Triple-negative breast cancer (TNBC) is a subtype of breast cancer that lacks expression of estrogen receptor, progesterone receptor, and the HER2 but is enriched with cancer stem cell–like cells (CSC). CSCs are the fraction of cancer cells recognized as the source of primary malignant tumors that also give rise to metastatic recurrence. 5-Hydroxymethylcytosine (5hmC) is a DNA epigenetic feature derived from 5-methylcytosine by action of tet methylcytosine dioxygenase enzymes (e.g., TET1); and although TET1 and 5hmC are required to maintain embryonic stem cells, the mechanism and role in CSCs remain unknown. Data presented in this report support the conclusion that TET1 and TET1-dependent 5hmC mediate hydrogen peroxide (H2O2)–dependent activation of a novel gene expression cascade driving self-renewal and expansion of CSCs in TNBC. Evidence presented also supports that the H2O2 affecting this pathway arises due to endogenous mechanisms—including downregulation of antioxidant enzyme catalase in TNBC cells—and by exogenous routes, such as systemic inflammation and oxidative stress coupled with obesity, a known risk factor for TNBC incidence and recurrence.

Implications:

This study elucidates a pathway dependent on H2O2 and linked to obesity-driven TNBC tumor-initiating CSCs; thus, it provides new understanding that may advance TNBC prevention and treatment strategies.

Triple-negative breast cancer (TNBC) tumors are defined by the lack of estrogen receptor (ER), progesterone receptor, and the HER2 oncogene. Poor survival rates and limited treatment options for TNBC further underscore that the molecular drivers of this aggressive subtype of breast cancer remain undefined (1). TNBCs are, however, characterized as having an intrinsic basal-like gene expression signature and enrichment of cancer stem cell–like cells (CSC; refs. 2, 3). CSCs are the minor fraction of cancer cells recognized as the source of malignant tumor initiation (4, 5), and they have limited proliferative potential and give rise to therapy resistance and metastatic recurrence (6–8). Therefore, to understand what drives TNBC, research is needed to better define the mechanisms driving CSCs.

Toward this end, we recently discovered that methyl-CpG–binding domain protein 2, variant 2 (MBD2_v2) is required for expression of core pluripotency gene NANOG and essential for CSC self-renewal and expansion (9–11). In addition, we reported that hydrogen peroxide (H2O2)—a reactive oxygen species (ROS) and redox signaling factor (12)—regulates the expression of serine- and arginine-rich splicing factor 2 (SRSF2) that is necessary for expression of MBD2_v2 (9). We considered the epidemiologic evidence that obesity is a risk factor for TNBC incidence and poor outcomes (13–18) and that the mechanism remains unknown; and also in a follow-up study, we reported experimental results supporting that because obesity induces a systemic increase in proinflammatory signaling factors (e.g., cytokines) resulting in higher levels of H2O2, obesity can fuel an increase in SRSF2 expression to promote TNBC cell tumorigenicity in vivo (9).

To better understand this mechanism driving CSCs in TNBC and how obesity is a risk factor for TNBC requires identifying signaling factors upstream of SRSF2 that are similarly regulated by obesity-induced ROS, specifically H2O2. For new leads, in a recent article, we posed the hypothesis that the epigenetic feature 5-hydroxymethylcytosine (5hmC) mediates the effect of H2O2 to regulate CSC-related TNBC gene expression based on evidence that by an unknown mechanism tet methylcytosine dioxygenase 1 (TET1) and TET1-dependent 5hmC are required to maintain embryonic stem cells (ESC; refs. 19, 20). We developed an approach to perform genome-wide sequencing analysis of 5hmC and mRNA, and among other genes identified a positive correlation between TAR DNA-binding protein (TARDBP) gene-localized 5hmC enrichment and TARDBP mRNA expression (21). Moreover, both were downregulated in TNBC cells in response to treatment with catalase (CAT; ref. 21), an antioxidant enzyme expressed in all mammalian cells that specifically targets and neutralizes H2O2 (22). Because the biochemistry surrounding how TARDBP regulates SRSF2 mRNA is known (23), these observations collectively provided direction for our continued research of a novel obesity-linked pathway regulating SRSF2 and TNBC CSCs presented herein.

Existing reports are conflicting as to the potential role of TET1 in TNBC (24, 25). Furthermore, although the role of 5-methylcytosine (5mC) to repress expression of tumor-suppressor genes in cancer is well documented (26–28), the role of 5hmC to upregulate gene expression (29, 30) remains understudied in cancer. A report from Tsai and colleagues also raises the possibility that TET1 upregulates cancer cell genes independent of its function to convert 5mC to 5hmC (31). The current study establishes the role of TARDBP to upregulate SRSF2 and promote CSCs and is the first to report evidence that TET1 via 5hmC upregulates TARDBP. This study also breaks new ground by uncovering that dysregulation of CAT in TNBC cells allows for H2O2-dependent signaling that increases TET1 levels in TNBC and more so for obese patients, thus further elucidating a mechanism underlying why obesity is a risk factor for TNBC.

Cell culture and treatments

Previously established, patient tumor-derived TNBC cell lines (32, 33) MDA-MB-468, MDA-MB-231, and HCC70 were originally acquired from the ATCC. MDA-MB-468 and MDA-MB-231 cell lines were routinely cultured in 10% FBS DMEM media (Thermo Fisher Scientific) at 37°C with 5% CO2. HCC70 cells were cultured in 10% FBS RPMI-1640 media (Thermo Fisher) at 37°C with 5% CO2. The SUM149 cells were developed and acquired from Dr. Stephen Ethier (34), and cultured in 10% FBS Ham's F-12 (Thermo Fisher) media, at 37°C with 5% CO2. Human breast epithelial cell line MCF10A was acquired from the Karmanos Cancer Institute, formerly the Michigan Cancer Foundation. MCF10A cells were cultured in 5% horse serum DMEM/F-12 (Thermo Fisher) media containing 20 ng/mL of EGF, 0.5 mg/mL of hydrocortisone, 100 ng/mL of cholera toxin, and 10 μg/mL of insulin at 37°C with 5% CO2. Mycoplasma testing was performed on all cell lines with the MycoAlertTM Mycoplasma Detection Kit (product no. LT0–118, Lonza Group Ltd.) and the MycoAlertTM Assay Control Set (product no. LT07–518) following the manufacturer's directions. All cell lines were authenticated by short tandem repeat analysis using the PowerPlex(r) 16 System (Promega). For each frozen aliquot of verified and confirmed mycoplasma-free cells, experiments were performed after at least 2 and fewer than 12 passages. CAT protein, L-2-hydroxyglutarate (2HG), and cobalt chloride (CoCl2) were from Sigma-Aldrich.

Semiquantitative real-time RT-PCR

qRT-PCR analysis was performed as described previously (9). Briefly, RNA was isolated from pulverized snap-frozen tumor tissue, or from cultured cell lines using the Qiagen RNeasy Kit. RNA was converted to cDNA using the High-Capacity RNA-to-cDNA Kit (Thermo Fisher). Note that 100 ng cDNA was combined with SYBR Green PCR Master Mix (Thermo Fisher) in 20 μL reactions in 96-well plates. Reactions were run in triplicate using the StepOnePlus Real-Time PCR System (Thermo Fisher). mRNA-targeted primer pairs are listed in Supplementary Materials and Methods. PUM1 and/or β-Actin served as controls. Relative levels of expression were calculated using the delta-delta Ct method (35).

Mouse tumor samples originated from our previous study of MDA-MB-468 cell line–derived tumors harvested from lean control mice and diet-induced obesity (DIO) mice and flash frozen in liquid nitrogen (9). Treatment-naïve, surgically resected flash-frozen patient tumor samples were acquired from the Karmanos Cancer Institute and BioIVT. Our analysis of deidentified patient tumor specimens was prereviewed and approved by the Wayne State University Institutional Review Board. Written informed consent was obtained from all tissue donors, and methods conformed to the standards set by the Declaration of Helsinki. Obesity was defined by body mass index (BMI) threshold (BMI > 30 kg/m2; ref. 36).

Stable gene expression and transient knockdown

TARDBP knockdown employed DharmaFECT Reagent (Dharmacon) and Silencer Select negative control siRNA and siRNA targeting TARDBP mRNA (Thermo Fisher; product nos. 4309846, s23829, and s530935). TET1 knockdown employed ON-TARGET plus SMART pool siRNA chemically modified to disrupt interaction with transcripts containing only partial complementarity, thereby disrupting potential off-target effects (product no. L-014635–03–0005, Dharmacon). Prepared lentiviral particles from Vector Builder were used according to supplier recommendations to stably express human CAT (NM_001752.3), human TET1 (NM_030625.3), and mCherry under control of a cytomegalovirus (CMV) promoter along with a gene for puromycin resistance.

Mammosphere assay

The presence and self-renewal capacity of CSCs were examined in TNBC cell lines using the mammosphere formation assay, as performed by us previously (10). Briefly, 1,000 single cells were seeded in 1.5 mL of the FBS-free sphere formation media (1:1 DMEM:F-12 media plus with B-27 and N-2 supplements; Thermo Fisher) in 6-well ultralow attachment plates (Corning Inc.), adding 0.5 mL/well fresh FBS-free sphere formation media after 96-hour incubation at 37°C with 5% CO2. After 7 days of incubation, the mammospheres formed were counted and reported as a fraction of the total number of cells seeded. Images were taken using a Nikon Eclipse TE2000-U microscope.

5-HmC DNA immunoprecipitation combined with real-time PCR analysis (hMeDIP-PCR)

The Qiagen DNeasy Kit was used to isolate cell DNA in preparation for 5hmC DNA immunoprecipitation (hMeDIP) using reagents from Abcam according to supplier instructions. Briefly, DNA samples in 1.5 mL tubes were sonicated by 4 to 12 second pulses at 20% intensity, with 40 seconds on wet ice between pulses. hMeDIP was performed using 3,500 or 4,000 ng sonicated DNA per sample. The output elution was 40 μL of immunoprecipitated DNA. From this 18 μL was used in a 40 μL SYBR Green PCR reaction. The primer pair used in this analysis targets 76 base pairs in the CpG island (CGI) overlapping the transcription start sight of the TARDBP gene (sense AATGAGGAACAGAGGGAAAC and anti-sense GCTGAACCTGAAGACTGAATA). Primers were designed using OligoArchitect (Sigma-Aldrich) and assessed to solely target TARDBP using the BLAST tool https://blast.ncbi.nlm.nih.gov/Blast.cgi and PCR Melt curve analysis (Supplementary Fig. S2). The equation 100 × 2[Ct(adjusted input) – Ct(IP)] was used to calculate percentage of input DNA recovered by hMeDIP.

5HmC oligonucleotide competition assay

The oligonucleotide used was previously shown to capture methyl binding domain proteins that bind 5mC- and 5hmC-modified CpGs (37). Complementary strands of 5hmC-modified oligonucleotide sequence (5′-TGAGTACAGCTATCA TC/iHydMe-dC/GGTGACGCACTACATCGA-3′ and 5′-TCGATGTAGTGCGTCAC/i5HydMe-dC/GGATGATAGCTGTACTCA-3′) were synthesized by Integrated DNA Technologies and annealed by heating at 95°C for 5 minutes, cooling down to 65°C over 30 minutes, and then incubating for 60 minutes at room temperature. Oligos of the same sequence without 5hmC modification were similarly treated and used as the control. Cell lines in 6-well plates at 70% confluence were pretreated with 25 μmol/L chloroquine diphosphate (Sigma-Aldrich) for 5 hours and then transfected with the double-stranded oligos (6–10 μL of 10 μmol/L oligo stock in 2 mL media) using transient transfection reagent DharmaFECT (Dharmacon), following the manufacturer's instruction. After 48 hours, total RNA was isolated and qRT-PCR was performed to compare expression of TARDBP in cells treated with the 5hmC-modified oligo versus cells treated with the unmodified oligo. Using the delta–delta Ct method (35), for loading control, the TARDBP Ct values were normalized to the combined average Ct values of breast cancer housekeeping genes PUM1 and β-Actin (38).

Western blot analysis

Details on these methods are in Supplementary Materials and Methods. Scanned films of full-length blots are in Supplementary Fig. S1.

Statistical analysis and bioinformatics

Statistical analysis was performed using Bioconductor R 3.3.2. Graphs were generated with GraphPad Prism. P values ≤ 0.05 are reported as significant, and *P ≤ 0.05, **P < 0.01, and ***P < 0.001 indicate level of significance. Linear regression or Student t test (two-sided) was applied when two conditions were compared, including qRT-PCR analysis. Pearson test was used for analysis of correlation. For analysis of three or more experiment conditions, a mixed-model approach was applied. To identify significant differences among gene expression data from The Cancer Genome Atlas (TCGA), ANOVA analysis with Tukey Honestly significant difference (HSD) test was performed. For analysis of gene expression according to intrinsic subtyping, microarray data and PAM50 subtype annotation from TCGA were downloaded via cBioPortal. For analysis of clinically relevant subtypes, RNA-sequencing data from TCGA were downloaded from The National Cancer Institute Genomic Data Commons data portal.

TARDBP regulates SRSF2 expression and CSCs, and TARDBP expression is inhibited by CAT

The objective of this study was to map factors upstream of SRSF2 driving TNBC CSCs, and as described above, we predicted that TARDBP is immediately upstream of SRSF2. We validated this prediction using TNBC cell lines MDA-MB-468 and SUM149, which express similarly abundant endogenous levels of SRSF2. In both cell lines, siRNA knockdown of TARDBP caused a marked decrease in SRSF2 mRNA and protein levels (Fig. 1AC). Next, we used the mammosphere formation assay to examine the effect of TARDBP knockdown on the presence of viable, self-renewing CSCs in TNBC cell cultures. TARDBP knockdown reduced the numbers of mammosphere-forming CSCs and inhibited the proliferative outgrowth of the mammospheres that did form (Fig. 1D and E). These results are consistent with effects of SRSF2 knockdown (9) and thus consistent with SRSF2 serving as the intermediary factor.

Figure 1.

TARDBP regulates SRSF2 expression and CSCs, and TARDBP expression is inhibited by CAT. A and B, qRT-PCR measurement of transient TARDBP knockdown and resulting change to SRSF2 mRNA in 2 TNBC cell lines, MDA-MB-468 and SUM149. Bars, SD 3 replicates, representative of 2 experiments using 2 independent siRNA against TARDBP relative to nonsilencing control siRNA. C, Western blot analysis of transient TARDBP knockdown and resulting change to SRSF2 protein. Representative of 2 experiments for each of 2 TNBC cell lines. β-Actin is the loading control. D, To measure effects on CSCs, TNBC cell lines in mammosphere formation assays were treated with nonsilencing control siRNA or either of 2 siRNAs independently targeting TARDBP. Bars, SEM 3 independent experiments. E, Representative mammosphere images, Bar = 50 μm. (F) SRSF2 and (G) TARDBP mRNA data from TCGA noncancer breast tissue (normal) and breast cancer samples. Other subtypes include ER-positive and HER2-positive breast cancers combined. Line equals the median, and means were compared. (H) SRSF2 and (I) TARDBP mRNA levels in TCGA breast cancer samples plotted according to intrinsic subtypes. J, qRT-PCR analysis of TARDBP in each of four TNBC cells treated with recombinant CAT protein (1,000 units/mL, 48 hours) relative to corresponding vehicle-treated control cells. Bars, SEM 3 independent experiments. K, qRT-PCR analysis of CoCl2-induced (100 μmol/L, 48 hours) TARDBP mRNA relative to vehicle-treated controls. Bars, SEM 3 experiments. L, qRT-PCR analysis of TARDBP mRNA in transduced HCC70 and MDA-MB-468 TNBC cells, comparing CAT and mCherry transgenes. (M) HCC70 and (N) MDA-MB-468 stably transduced to overexpress CAT or mCherry were treated with CoCl2, and qRT-PCR analysis measured the effects of treatment on TARDBP levels relative to vehicle-treated controls. Bars, SEM 3 independent biological replicates.

Figure 1.

TARDBP regulates SRSF2 expression and CSCs, and TARDBP expression is inhibited by CAT. A and B, qRT-PCR measurement of transient TARDBP knockdown and resulting change to SRSF2 mRNA in 2 TNBC cell lines, MDA-MB-468 and SUM149. Bars, SD 3 replicates, representative of 2 experiments using 2 independent siRNA against TARDBP relative to nonsilencing control siRNA. C, Western blot analysis of transient TARDBP knockdown and resulting change to SRSF2 protein. Representative of 2 experiments for each of 2 TNBC cell lines. β-Actin is the loading control. D, To measure effects on CSCs, TNBC cell lines in mammosphere formation assays were treated with nonsilencing control siRNA or either of 2 siRNAs independently targeting TARDBP. Bars, SEM 3 independent experiments. E, Representative mammosphere images, Bar = 50 μm. (F) SRSF2 and (G) TARDBP mRNA data from TCGA noncancer breast tissue (normal) and breast cancer samples. Other subtypes include ER-positive and HER2-positive breast cancers combined. Line equals the median, and means were compared. (H) SRSF2 and (I) TARDBP mRNA levels in TCGA breast cancer samples plotted according to intrinsic subtypes. J, qRT-PCR analysis of TARDBP in each of four TNBC cells treated with recombinant CAT protein (1,000 units/mL, 48 hours) relative to corresponding vehicle-treated control cells. Bars, SEM 3 independent experiments. K, qRT-PCR analysis of CoCl2-induced (100 μmol/L, 48 hours) TARDBP mRNA relative to vehicle-treated controls. Bars, SEM 3 experiments. L, qRT-PCR analysis of TARDBP mRNA in transduced HCC70 and MDA-MB-468 TNBC cells, comparing CAT and mCherry transgenes. (M) HCC70 and (N) MDA-MB-468 stably transduced to overexpress CAT or mCherry were treated with CoCl2, and qRT-PCR analysis measured the effects of treatment on TARDBP levels relative to vehicle-treated controls. Bars, SEM 3 independent biological replicates.

Close modal

Profiles of expression in breast cancer patient tumors are also consistent with a TARDBP–SRSF2 regulatory link. Analysis of TCGA data revealed that SRSF2 and TARDBP mRNA expression levels are concordantly higher in TNBC relative to noncancer breast tissue (Fig. 1F and G). Across breast cancer subtypes defined by intrinsic gene expression patterns, SRSF2 and TARDBP display corresponding profiles, including high expression in basal-like/TNBC cancers and low expression in luminal A cancers, which are ER-positive and the subtype least likely to recur (ref. 2; Fig. 1H and I).

In addition to regulating SRSF2, if TARDBP is directly involved in the novel pathway we are mapping, then TARDBP will be inhibited by the antioxidant enzyme CAT. Using four TNBC cell lines—MDA-MB-468, MDA-MB-231, HCC70, and SUM149—data demonstrate that for each cell line TARDBP mRNA levels were decreased in cultures treated with recombinant CAT protein relative to matched vehicle-treated controls (Fig. 1J). We proceeded to measure levels of TARDBP after cells were perturbed by CoCl2 treatment, which is used to induce a biologically relevant and relatively sustained increase in cellular H2O2 not achieved by adding H2O2 directly to cultures (39). Consistent with the effect of CAT treatment to inhibit TARDBP mRNA expression, CoCl2 treatment upregulated TARDBP mRNA expression (Fig. 1K).

To reinforce these results, HCC70 and MDA-MB-468 cells were selected to stably overexpress CAT and measure the effect on TARDBP levels relative to mCherry-expressing control cells. In cells transduced with the CAT expression vector, CAT mRNA and protein levels were markedly increased relative to low levels in mCherry-expressing control cells (averaging 56-fold according to qRT-PCR, Supplementary Fig. S3). However, unlike the effect of exogenous CAT treatment (Fig. 1J), TARDBP expression was not decreased in CAT-overexpressing cells relative to mCherry-expressing control cells (Fig. 1L). At this point, we speculated that an unknown compensatory mechanism re-established TARDBP levels in the stable CAT-overexpressing cells, and we proceeded to test the response to CoCl2 treatment. The effect of CoCl2 treatment to increase TARDBP expression in mCherry-expressing control cells was inhibited in CAT-overexpressing cells (Fig. 1M and N).

TET1 and TARDBP-localized 5hmC regulate TARDBP expression

The next goal was to address our hypothesis that TET1-dependent 5hmC upregulates TARDBP mRNA expression. We used a lentiviral approach to stably overexpress TET1 in HCC70 cells with the expectation that we would be able to measure the effect of TET1 overexpression to increase TARDBP mRNA and TARDBP-localized 5hmC levels relative to mCherry-transduced control cells. Unexpectedly, results indicated that TET1 levels were increased in both mCherry control and TET1-transduced cells; this was established by comparing each with parent (nontransduced) cells in qRT-PCR and Western blot analysis (Fig. 2A and B). TARDBP mRNA levels were also increased in both mCherry and TET1-transduced cells (Fig. 2B and C) relative to parent cells. We performed 5-hydroxymethylated DNA immunoprecipitation combined with real-time PCR analysis (hMeDIP-PCR). Concordant with the effect of lentiviral transduction technique to upregulate TET1 and TARDBP, 5hmC levels localized to the CGI encompassing the TARDBP transcription start site (TSS) were increased in mCherry and TET1-transduced HCC70 cells relative to parent cells according to hMeDIP-PCR analysis (Fig. 2D). Similar results were observed when SUM-149 cells were transduced (Fig. 2E–G). It appears that the lentiviral approach induces TET1 and TARDBP expression irrespective of the recombinant gene being introduced to cells, and this would also explain why CAT treatment decreased TARDBP levels (Fig. 1J), but stable CAT overexpression did not (Fig. 1L).

Figure 2.

TET1 and 5hmC regulate TARDBP expression. A, qRT-PCR analysis of TET1 mRNA in HCC70 cells stably expressing transgenes mCherry or TET1 relative to parent cells. Bars, SD 3 replicates, representative of 2 experiments. B, Western blot analysis of TET1 and TARDBP in nuclear lysates from HCC70 cells—parent, mCherry- and TET1-transduced—representative of 2 experiments. Nucleoporin p62 (NUP62) is the loading control. C, qRT-PCR analysis of TARDBP mRNA in HCC70 cells expressing transgenes mCherry or TET1 relative to parent cells. D, TARDBP-localized 5hmC levels according to hMeDIP-PCR analysis of parent, mCherry- and TET1-transduced HCC70 cells. Bars, SEM 3 experiments. E–G, As in A–B above, qRT-PCR and Western blot analysis of TET1 and TARDBP levels in SUM149 cells stably expressing transgenes mCherry or TET1 relative to parent cells. H and I, Decreased TET1 and TARDBP mRNA levels in TET1-transduced SUM149 cells due to transient TET1 knockdown (TARGETplus SMARTpool siRNA) relative to nonsilencing control siRNA measured by qRT-PCR. Bars, SD 3 replicates, representative of 2 experiments. J, Decreased TARDBP mRNA in SUM149 cells due to 2HG treatment (10 mmol/L, 48 hours), measured by qRT-PCR. Bars, SEM 3 experiments. K, Levels of 5hmC localized to the CGI surrounding the TARDBP TSS in parent and TET1-transduced SUM149 cells, and effect of 2HG treatment of TET1-transduced SUM149 cells according to hMeDIP-PCR. L and M, Effect of 2HG treatment on TARDBP mRNA and TARDBP-localized 5hmC levels in TET1-transduced HCC70 cells measured by qRT-PCR and hMeDIP-PCR. Bars, SEM of at least 3 independent experiments. N–P, Effect of TARGETplus SMARTpool siRNA targeting TET1 on TET1 and TARDBP levels in parent (nontransduced) HCC70 cells, by Western blot and qRT-PCR analysis. Representative of 2 independent experiments performed in triplicate. Bars, SD. Q, Mammosphere formation assays using HCC70 cells, comparing transient TET1 knockdown and nonsilenced control cultures. Bars, SEM 3 independent experiments. R, Representative mammosphere images. Bar = 50 μm.

Figure 2.

TET1 and 5hmC regulate TARDBP expression. A, qRT-PCR analysis of TET1 mRNA in HCC70 cells stably expressing transgenes mCherry or TET1 relative to parent cells. Bars, SD 3 replicates, representative of 2 experiments. B, Western blot analysis of TET1 and TARDBP in nuclear lysates from HCC70 cells—parent, mCherry- and TET1-transduced—representative of 2 experiments. Nucleoporin p62 (NUP62) is the loading control. C, qRT-PCR analysis of TARDBP mRNA in HCC70 cells expressing transgenes mCherry or TET1 relative to parent cells. D, TARDBP-localized 5hmC levels according to hMeDIP-PCR analysis of parent, mCherry- and TET1-transduced HCC70 cells. Bars, SEM 3 experiments. E–G, As in A–B above, qRT-PCR and Western blot analysis of TET1 and TARDBP levels in SUM149 cells stably expressing transgenes mCherry or TET1 relative to parent cells. H and I, Decreased TET1 and TARDBP mRNA levels in TET1-transduced SUM149 cells due to transient TET1 knockdown (TARGETplus SMARTpool siRNA) relative to nonsilencing control siRNA measured by qRT-PCR. Bars, SD 3 replicates, representative of 2 experiments. J, Decreased TARDBP mRNA in SUM149 cells due to 2HG treatment (10 mmol/L, 48 hours), measured by qRT-PCR. Bars, SEM 3 experiments. K, Levels of 5hmC localized to the CGI surrounding the TARDBP TSS in parent and TET1-transduced SUM149 cells, and effect of 2HG treatment of TET1-transduced SUM149 cells according to hMeDIP-PCR. L and M, Effect of 2HG treatment on TARDBP mRNA and TARDBP-localized 5hmC levels in TET1-transduced HCC70 cells measured by qRT-PCR and hMeDIP-PCR. Bars, SEM of at least 3 independent experiments. N–P, Effect of TARGETplus SMARTpool siRNA targeting TET1 on TET1 and TARDBP levels in parent (nontransduced) HCC70 cells, by Western blot and qRT-PCR analysis. Representative of 2 independent experiments performed in triplicate. Bars, SD. Q, Mammosphere formation assays using HCC70 cells, comparing transient TET1 knockdown and nonsilenced control cultures. Bars, SEM 3 independent experiments. R, Representative mammosphere images. Bar = 50 μm.

Close modal

To more thoroughly implicate TET1-dependent 5hmC in the regulation of TARDBP, we employed the TET1-transduced TNBC cells and the hMeDIP-PCR method to measure the effects of reversing TET1 upregulation on TARDBP-localized 5hmC levels. Attempts to knockdown-overexpressed TET1 levels in TET1-transduced SUM149 cells by transient approach using ON-TARGETplus SMARTpool siRNA elicited a 50% reduction of TET1 (Fig. 2H) and a reduction in TARDBP expression that was significant but not pronounced (Fig. 2I). We proceeded to treat TET1-transduced SUM149 cells with 2HG, which inhibits TET enzymatic activity (40), and observed a more pronounced reduction of TARDBP expression levels (Fig. 2J). hMeDIP-PCR analysis of TET1-transduced SUM149 cells demonstrated the effect of 2HG treatment to decrease levels of 5hmC localized at the CGI surrounding the TARDBP TSS (Fig. 2K). We performed the same experiment using HCC70 cells, also lentivirally transduced with the TET1 transgene. Although the trend was similar, 2HG treatment of TET1-transduced HCC70 cells was less effective at reducing TARDBP expression and TARDBP-localized 5hmC (Fig. 2L and M). This appears to be consistent with there being higher levels of TET1 and TARDBP-localized 5hmC in the TET1-transduced HCC70 cells relative to TET1-transduced SUM149 cells (compare relationship of TET1 band to control gene in Fig. 2B and F, also y axes in Fig. 2K and M).

We performed the reciprocal experiment using ON-TARGETplus SMARTpool siRNA to transiently knockdown TET1 in parent cells and thereby observed again that when TET1 levels were manipulated, TARDBP levels were affected in a manner consistent with the conclusion that TET1 plays a role in upregulating TARDBP expression (Fig. 2NP). Furthermore, we observed a decrease in the fraction of mammosphere-forming CSCs that paralleled the decrease of TARDBP mRNA (Fig. 2Q) and the proliferative outgrowth potential of mammospheres that did form was also decreased (Fig. 2R) as in Fig. 1D and E.

We developed a competitive oligonucleotide assay to further confirm the role of 5hmC to regulate TARDBP expression. Parent (nontransduced) TNBC cell line cultures were transiently transfected with double-stranded DNA oligonucleotide that was either 5hmC-modified or unmodified, and after an incubation period, mRNA was harvested and TARDBP mRNA levels were analyzed by qRT-PCR. We theorized that the 5hmC-modified oligonucleotide could sequester 5hmC-binding epigenetic readers that would otherwise bind to and transactivate the endogenous 5hmC-modified TARDBP promoter. Therefore, if 5hmC has a role in regulating TARDBP expression, then relative to the unmodified control oligonucleotide, the 5hmC-modified oligonucleotide should decrease TARDBP expression measured by qRT-PCR analysis. As theorized, in each of four TNBC cell lines investigated, transfection of the 5hmC-modified oligonucleotide caused a decrease in TARDBP expression relative to the unmodified control oligonucleotide (Fig. 3, first column). Moreover, examination of the raw Ct-value RT-PCR data confirmed that the 5hmC-modified oligonucleotide decreased TARDBP (Fig. 3, second column) and did not affect the expression levels of breast cancer housekeeping genes (38), PUM1 and β-Actin (Fig. 3, third and fourth columns).

Figure 3.

Oligonucleotide competition assay and control gene data. A–D, In each of four TNBC cell lines investigated—SUM149, HCC70, MDA-MB-468, and MDA-MB-231—qRT-PCR analysis measured a decrease in TARDBP mRNA in cells transfected with the 5hmC-modified oligonucleotide compared with cells transfected with the unmodified oligonucleotide (first column). Examination of the raw Ct-value RT-PCR data corroborated that the 5hmC-modified oligonucleotide decreased TARDBP mRNA (second column) and did not affect the mRNA levels of breast cancer housekeeping genes PUM1 and β-actin (third and fourth columns, respectively). Each data point represents an independent biological replicate, the mean of 3 technical replicates. Lines, mean and SEM. Ct values (log2) are inverse to the amount of mRNA in the sample.

Figure 3.

Oligonucleotide competition assay and control gene data. A–D, In each of four TNBC cell lines investigated—SUM149, HCC70, MDA-MB-468, and MDA-MB-231—qRT-PCR analysis measured a decrease in TARDBP mRNA in cells transfected with the 5hmC-modified oligonucleotide compared with cells transfected with the unmodified oligonucleotide (first column). Examination of the raw Ct-value RT-PCR data corroborated that the 5hmC-modified oligonucleotide decreased TARDBP mRNA (second column) and did not affect the mRNA levels of breast cancer housekeeping genes PUM1 and β-actin (third and fourth columns, respectively). Each data point represents an independent biological replicate, the mean of 3 technical replicates. Lines, mean and SEM. Ct values (log2) are inverse to the amount of mRNA in the sample.

Close modal

TET1 expression is high in TNBC and inhibited by CAT that is suppressed in TNBC

Good and colleagues reported that TET1 expression is higher in TNBC tumors relative to noncancer breast tissue, HER2-positive, and hormone receptor–positive clinical subtypes (41). We analyzed expression data from TCGA according to intrinsic subtyping to clarify that the relatively high levels in the TNBC cohort are more specifically attributable to enrichment of basal-like cancers (Fig. 4A). In addition to regulating TARDBP (Figs. 2 and 3), if TET1 is directly involved in the novel pathway we are mapping, then TET1 will be inhibited by CAT. We reviewed our genome-wide dataset from a previous study (21) and identified that treatment of TNBC cell line MDA-MB-468 with recombinant CAT reduced TET1 mRNA expression by 60% (P = 0.05, Fig. 4B, inset). We reproduced this result using MDA-MB-468 cells and three additional TNBC cell lines (MDA-MB-231, HCC70, and SUM149), and for each observed that treatment with CAT protein significantly decreased TET1 mRNA levels, translating to decreased protein levels (Fig. 4B and C). We next employed the transduced HCC70 and MDA-MB-468 cell lines introduced above and noted that TET1 expression, like TARDBP (Fig. 1L), was not decreased in stable CAT-overexpressing cells relative to mCherry control cells (Fig. 4D). However, for both cell lines, CoCl2 treatment of mCherry-expressing control cells significantly induced TET1 levels and CoCl2 treatment of CAT-overexpressing cells did not (Fig. 4E and F). In a final set of experiments using these cells, we performed mammosphere assays. Relative to vehicle treatment, CoCl2 treatment of mCherry control cell cultures increased the fraction of mammosphere-forming CSCs, but CoCl2 treatment of CAT-overexpressing cell cultures did not (Fig. 4G and H).

Figure 4.

TET1 expression is high in TNBC and inhibited by CAT. A, TET1 mRNA data from TCGA according to breast cancer intrinsic subtype. Line equals median, and means were compared. B and C, qRT-PCR and Western blot analysis of TET1 in each of four TNBC cells treated with recombinant CAT protein (1,000 units/mL, 48 hours) relative to corresponding vehicle-treated control cells. The blot is representative of 2 independent experiments. NUP62 is the loading control. Bars, SEM of 3 independent replicates. Inset, TET1 levels from RNA-sequencing analysis of CAT-treated MDA-MB-468 cells relative to vehicle-treated controls (mean of 3 sample sets). D, qRT-PCR analysis of TET1 mRNA in transduced HCC70 and MDA-MB-468 TNBC cells, comparing CAT and mCherry transgenes. (E) HCC70 and (F) MDA-MB-468 stably transduced to overexpress CAT or mCherry were treated with CoCl2, and qRT-PCR analysis measured the effects of treatment on TET1 levels relative to vehicle-treated controls. Bars, SEM 3 experiments. (G) HCC70 and (H) MDA-MB-468 stably transduced to overexpress CAT or mCherry were treated with CoCl2, and mammosphere assays measured the effects of treatment on CSCs relative to vehicle-treated controls. Bars, SEM 3 independent biological replicates.

Figure 4.

TET1 expression is high in TNBC and inhibited by CAT. A, TET1 mRNA data from TCGA according to breast cancer intrinsic subtype. Line equals median, and means were compared. B and C, qRT-PCR and Western blot analysis of TET1 in each of four TNBC cells treated with recombinant CAT protein (1,000 units/mL, 48 hours) relative to corresponding vehicle-treated control cells. The blot is representative of 2 independent experiments. NUP62 is the loading control. Bars, SEM of 3 independent replicates. Inset, TET1 levels from RNA-sequencing analysis of CAT-treated MDA-MB-468 cells relative to vehicle-treated controls (mean of 3 sample sets). D, qRT-PCR analysis of TET1 mRNA in transduced HCC70 and MDA-MB-468 TNBC cells, comparing CAT and mCherry transgenes. (E) HCC70 and (F) MDA-MB-468 stably transduced to overexpress CAT or mCherry were treated with CoCl2, and qRT-PCR analysis measured the effects of treatment on TET1 levels relative to vehicle-treated controls. Bars, SEM 3 experiments. (G) HCC70 and (H) MDA-MB-468 stably transduced to overexpress CAT or mCherry were treated with CoCl2, and mammosphere assays measured the effects of treatment on CSCs relative to vehicle-treated controls. Bars, SEM 3 independent biological replicates.

Close modal

As described above, recombinant CAT protein served as a tool to reveal effects of H2O2 on gene expression. The current study also led us to realize that CAT is inherently relevant to TNBC. Based on our past experiences stably overexpressing a variety of genes, it seemed to take an inordinate amount of time to establish routinely passaged cell lines overexpressing CAT. This led us to first consider that CAT expression may be endogenously suppressed in TNBC. Follow-up analysis of TCGA data revealed that CAT expression is downregulated in breast cancer tumors relative to noncancer breast tissue. Among normal breast tissue and clinical breast cancer subtypes, CAT expression displays the greatest downregulation in TNBC cancer (Fig. 5A); and among intrinsic subtypes, CAT expression is significantly lower in HER2-enriched and basal-like tumors (Fig. 5B).

Figure 5.

CAT is suppressed in TNBC. A, CAT mRNA data from TCGA noncancer breast tissue (normal) and breast cancer samples. Other subtypes include ER-positive and HER2-positive breast cancers combined. Line equals the median. The means were compared. B, CAT mRNA data from TCGA according to breast cancer intrinsic subtype. C–G, qRT-PCR analysis of CAT and TET1 levels in MCF10A breast epithelial cells and TNBC cell lines, CoCl2-treated (100 μmol/L, 48 hours) cultures relative to vehicle-treated controls. Bars, SEM 3 experiments.

Figure 5.

CAT is suppressed in TNBC. A, CAT mRNA data from TCGA noncancer breast tissue (normal) and breast cancer samples. Other subtypes include ER-positive and HER2-positive breast cancers combined. Line equals the median. The means were compared. B, CAT mRNA data from TCGA according to breast cancer intrinsic subtype. C–G, qRT-PCR analysis of CAT and TET1 levels in MCF10A breast epithelial cells and TNBC cell lines, CoCl2-treated (100 μmol/L, 48 hours) cultures relative to vehicle-treated controls. Bars, SEM 3 experiments.

Close modal

We further considered the possibility that CAT expression is suppressed in TNBC cells and reasoned that in cellular systems operating normally to maintain low-level redox homeostasis, an increase in H2O2 levels will cause an induction of CAT. We hypothesized that this response mechanism is dysregulated in basal TNBC cells, thereby maintaining CAT at low levels while H2O2 levels are elevated. We used the noncancer, basal epithelial cell line MCF10A to test the concept. In line with the expectation for a normally operating cellular system, CoCl2 treatment of MCF10A caused an increase in CAT expression and TET1 expression remained unchanged (Fig. 5C). In three of four TNBC cell lines tested, CoCl2 treatment did not increase CAT expression (Fig. 5D–F). HCC70 appears to be unique in that CoCl2 treatment caused CAT to increase, although the levels of induction appear to be attenuated relative to MCF10A cells (Fig. 5G). Consistent across the four TNBC cell lines, TET1 expression was increased by CoCl2 treatment (Fig. 5D–G). Collectively, these results support that H2O2 upregulates TET1 expression, and high TET1 in basal TNBC is coupled with suppression of antioxidant CAT.

The pathway from TET1 to NANOG is linked to obesity

The current study builds on our recent discoveries that H2O2 and SRSF2 are necessary for MBD2_v2 expression and that MBD2_v2 is in turn essential for NANOG expression and self-renewal and expansion of TNBC tumor-initiating CSCs (9–11). We previously hypothesized that because obesity induces a system-wide increase in proinflammatory signaling factors (e.g., cytokines) resulting in higher levels of H2O2, obesity can fuel an increase in MBD2_v2 expression to promote TNBC tumorigenicity in vivo (9). In support of this hypothesis, we reported that MBD2_v2 expression is higher in TNBC tumors from obese women (BMI > 30 kg/m2) relative to nonobese women (BMI < 30 kg/m2; ref. 9). We now present similar evidence from qRT-PCR analysis of TNBC tumor samples to support that each factor identified as being upstream of MBD2_v2—i.e., SRSF2, TARDBP, and TET1 as well as downstream NANOG—is upregulated in tumors from obese women (Fig. 6A–D). Reinforcing these results, analysis of tumors derived from inoculated MDA-MB-468 cells revealed that TET1 and TARDBP expression levels are higher in TNBC tumors that formed more frequently in a DIO mouse model relative to lean controls (Fig. 6E) as we previously reported for SRSF2, MBD2_v2, and NANOG (9). Figure 7 integrates our previously published findings (9–11) and results presented herein, and summarizes current understanding of this novel pathway driving TNBC CSCs.

Figure 6.

Genes in the TNBC gene expression cascade from TET1 to NANOG are linked to obesity. A–D, TET1, TARDBP, SRSF2, and NANOG mRNA expression levels in tumors from obese patients with TNBC with BMI > 30 relative to levels in tumors from patients with BMI < 30, measured by qRT-PCR. Data are mean-normalized and log2 transformed. E, TET1 and TARDBP levels in TNBC cell line MDA-MB-468–derived tumors harvested from DIO (n = 3) and lean control mice (n = 3). TET1 and TARDBP levels, increased in DIO samples (P = 0.04 and P = 0.06 respectively), were correlated (P = 0.02). qRT-PCR analysis of each sample was performed in triplicate and mean normalized.

Figure 6.

Genes in the TNBC gene expression cascade from TET1 to NANOG are linked to obesity. A–D, TET1, TARDBP, SRSF2, and NANOG mRNA expression levels in tumors from obese patients with TNBC with BMI > 30 relative to levels in tumors from patients with BMI < 30, measured by qRT-PCR. Data are mean-normalized and log2 transformed. E, TET1 and TARDBP levels in TNBC cell line MDA-MB-468–derived tumors harvested from DIO (n = 3) and lean control mice (n = 3). TET1 and TARDBP levels, increased in DIO samples (P = 0.04 and P = 0.06 respectively), were correlated (P = 0.02). qRT-PCR analysis of each sample was performed in triplicate and mean normalized.

Close modal
Figure 7.

Graphical depiction of the pathway under study regulating TNBC tumor-initiating CSCs. The figure represents a convergence of findings from our previous reports (9–11) and results from the current study. The antioxidant CAT, which specifically targets and neutralizes H2O2, is downregulated in TNBC. This allows for upregulation of redox signaling by H2O2, driving a gene expression cascade from TET1 through TARDBP and SRSF2 to support expression of MBD2_v2 and ultimately maintain CSC self-renewal. Moreover, increasing H2O2 leads to an increase in gene expression and expansion of CSCs. Obesity raises levels of proinflammatory signaling factors (e.g., cytokines) that raise levels of H2O2 systemically and in breast cancer cells as further evidenced and discussed in ref. (9). This may explain the link between obesity and increased expression of the genes in this pathway.

Figure 7.

Graphical depiction of the pathway under study regulating TNBC tumor-initiating CSCs. The figure represents a convergence of findings from our previous reports (9–11) and results from the current study. The antioxidant CAT, which specifically targets and neutralizes H2O2, is downregulated in TNBC. This allows for upregulation of redox signaling by H2O2, driving a gene expression cascade from TET1 through TARDBP and SRSF2 to support expression of MBD2_v2 and ultimately maintain CSC self-renewal. Moreover, increasing H2O2 leads to an increase in gene expression and expansion of CSCs. Obesity raises levels of proinflammatory signaling factors (e.g., cytokines) that raise levels of H2O2 systemically and in breast cancer cells as further evidenced and discussed in ref. (9). This may explain the link between obesity and increased expression of the genes in this pathway.

Close modal

This report contributes new insights to cancer research and breaks new ground in mapping a molecular signaling pathway driving CSCs in TNBC that is also contributing to why obesity is a risk factor for TNBC incidence and recurrence. Each gene in the novel pathway presented herein has recently been implicated in TNBC tumorigenicity in vivo: MBD2_v2 and SRSF2 by our laboratory (9), NANOG by Thiagarajan and colleagues (42), TARDBP by Key and colleagues (43), and TET1 by Wu and colleagues (25). Moreover, studies including ours have shown that for MBD2_v2, SRSF2, TARDBP, and TET1, high tumor expression associates with worse outcomes for patients with TNBC (9, 41, 43). However, prior to the current study the mechanism driving upregulation of these genes in TNBC had not been elucidated. Moreover, for TET1 and TARDBP the downstream effectors linking them to TNBC cell tumorigenicity were also unknown. Now we have addressed these unknowns and together with our previously published work provide evidence for the regulatory links depicted in Fig. 7. Furthermore, data in the current report support that the H2O2 regulating this pathway can arise from endogenous mechanisms, including downregulation of CAT, and by exogenous perturbations such as due to systemic inflammation coupled with obesity.

According to a recent literature search, this is the first report documenting that CAT expression is downregulated in TNBC tumors and that suppression of CAT expression in TNBC is key to activation of a CSC driver pathway. Now, it remains to be understood how redox signaling by H2O2 regulates TET1 and how CAT expression is dysregulated in cancer cells. These are lines of investigation we plan to pursue. We have also considered that the gene expression data from breast cancer patient tumors may be indicating that the same or a similar CSC-driver pathway functions in other breast cancer subtypes, particularly the HER2-enriched subtype.

The work we undertook leads us to appreciate that the lentiviral-based gene transfer method may be causing upregulation of TET1 and TARDBP in TNBC cells irrespective of the transgene being introduced, and there is reason to think this is directly related to the nature of the transduced lentivirus vector. Although the experimental system is replication-incompetent, the transgene vector does include long terminal repeat and TAR DNA element sequence from human immunodeficiency virus type 1 (HIV-1). TARDBP was discovered and so named in the course of HIV-1 research because it binds the TAR DNA element and inhibits HIV-1 gene expression (44). We speculate that stably introducing this element into the DNA of target cells changes the dynamics of TARDBP signaling and activates a response mechanism to upregulate TARDBP, via upregulation of TET1 according to our results. We observed upregulation of TET1 in transduced cells after 3 weeks of passaging and stable selection. Perhaps at an earlier time point, we would have measured lower TET1 levels in mCherry control cells and been misled in our research. More work should be done to validate this phenomenon, but based on our observation whether researchers used certain stable or transient gene modifying approaches may explain some of the discrepancies among results from previous studies investigating the role of TET1 in TNBC as well as ESCs (19, 24, 25).

5hmC enrichment in CpG islands at or near the TSS, attributed to the conversion of 5mC to 5hmC by TET1, upregulates gene expression (29, 30). Moreover, across ESCs and tumor tissues, the majority of 5hmC modifications are stable (29, 45, 46), which is consistent with 5hmC serving a direct gene regulatory role. However, possibly due to experimental challenges, the role of 5hmC in cancer remains understudied. Herein, we are the first to present that TET1-dependent 5hmC localized to the CGI encompassing the TARDBP TSS regulates TARDBP gene expression and thereby affects a pathway regulating CSCs. This finding is based on the convergence of multiple methods of analysis including an oligonucleotide competition assay we developed that provided additional data to confirm the role of 5hmC. Transfecting cells with a 5hmC-modified oligonucleotide caused a decrease in TARDBP expression relative to transfection with the unmodified control oligonucleotide (Fig. 3). The results are consistent with the rationale that the 5hmC-modified oligonucleotide specifically can sequester a 5hmC-binding protein that would otherwise bind to and transactivate the 5hmC-modified endogenous TARDBP promoter. The oligonucleotide sequence used was previously confirmed to capture methyl binding domain proteins (37), and knowing which specific factors are sequestered by the 5hmC-modified oligonucleotide is not necessary for the purpose of demonstrating the role of 5hmC. However, evidence points to potential involvement of the epigenetic reader methyl binding domain protein 3 (MBD3). Although TET1 preferentially binds 5mC (45, 47), MBD3 preferentially binds 5hmC and actually activates gene expression (30). Moreover, chromatin immunoprecipitation and sequencing analysis (ChIP-seq) data from Shimbo and colleagues (48), downloaded and displayed in Supplementary Fig. S4, show that MBD3 binding maps to the CGI surrounding the TARDBP TSS in TNBC cells. Further work needs to be done to confirm exactly which epigenetic regulator or transcription factor binds 5hmC to upregulate TARDBP, but we predict that coordinated upregulation of that factor(s) increases the potency of TET1 upregulation in vivo.

We have reasoned that the novel H2O2-regulated pathway under study serves as a link between obesity and TNBC because obesity induces an increase in circulating proinflammatory signaling factors (e.g., cytokines) resulting in higher levels of H2O2 systemically and localized to cancer cells (9). Moreover, a molecular mechanism linking obesity to CSCs would be consistent with the epidemiologic evidence that obesity is a risk factor for TNBC tumor incidence and recurrence (13–18). We previously used high-throughput microarray data to uncover a significant, positive association between high MBD2_v2 expression in TNBC tumors and obesity (9). The current study used a more specific, targeted approach to analyze TET1, TARDBP, SRSF2, and NANOG gene expression levels and links them to obesity. Finding that all of these genes are upregulated in TNBC and apparently more so in TNBC tumors from obese women and DIO mice adds to the evidence supporting the conclusion that these are factors in a pathway driving TNBC tumor-initiating CSCs depicted in Fig. 7. Obesity is a risk factor for TNBC (13–18), and roughly 50% of patients with TNBC are obese (49). Mapping this novel pathway driving TNBC tumor-initiating CSCs contributes to understanding the public health risk for cancer as obesity rates continue to rise (50).

G. Dyson reports grants from NIH during the conduct of the study. A. Bollig-Fischer reports grants from NCI and Believe Foundation during the conduct of the study. No potential conflicts of interest were disclosed by the other authors.

B. Bao: Methodology, data acquisition, writing original draft, writing-review and editing. E.A. Teslow: Methodology, data acquisition, writing-review and editing. C. Mitrea: Data analysis, writing-review and editing. J.L. Boerner: Resources, writing-review and editing. G. Dyson: Data analysis, interpretation, writing-review and editing, A. Bollig-Fischer: Conceptualization, funding acquisition, methodology, data analysis and interpretation, writing original draft, writing-review and editing.

This study was supported in part by funding from The Believe Foundation and Joann M. Deliz and James W. Deliz in memory of David Bergman (to A. Bollig-Fischer). The NIH:NCI Cancer Center Grant P30CA022453 to the Karmanos Cancer Institute supported contributions from the Biostatistics and Biobanking and Correlative Sciences Cores.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Bianchini
G
,
Balko
JM
,
Mayer
IA
,
Sanders
ME
,
Gianni
L
. 
Triple-negative breast cancer: challenges and opportunities of a heterogeneous disease
.
Nat Rev Clin Oncol
2016
;
13
:
674
90
.
2.
Prat
A
,
Parker
JS
,
Karginova
O
,
Fan
C
,
Livasy
C
,
Herschkowitz
JI
, et al
Phenotypic and molecular characterization of the claudin-low intrinsic subtype of breast cancer
.
Breast Cancer Res
2010
;
12
:
R68
.
3.
Ricardo
S
,
Vieira
AF
,
Gerhard
R
,
Leitao
D
,
Pinto
R
,
Cameselle-Teijeiro
JF
, et al
Breast cancer stem cell markers CD44, CD24 and ALDH1: expression distribution within intrinsic molecular subtype
.
J Clin Pathol
2011
;
64
:
937
46
.
4.
Reya
T
,
Morrison
SJ
,
Clarke
MF
,
Weissman
IL
. 
Stem cells, cancer, and cancer stem cells
.
Nature
2001
;
414
:
105
11
.
5.
Chaffer
CL
,
Weinberg
RA
. 
How does multistep tumorigenesis really proceed?
Cancer Discov
2015
;
5
:
22
4
.
6.
Liu
H
,
Patel
MR
,
Prescher
JA
,
Patsialou
A
,
Qian
D
,
Lin
J
, et al
Cancer stem cells from human breast tumors are involved in spontaneous metastases in orthotopic mouse models
.
Proc Natl Acad Sci U S A
2010
;
107
:
18115
20
.
7.
Fillmore
CM
,
Kuperwasser
C
. 
Human breast cancer cell lines contain stem-like cells that self-renew, give rise to phenotypically diverse progeny and survive chemotherapy
.
Breast Cancer Res
2008
;
10
:
R25
.
8.
Pattabiraman
DR
,
Weinberg
RA
. 
Tackling the cancer stem cells - what challenges do they pose?
Nat Rev Drug Discovery
2014
;
13
:
497
512
.
9.
Teslow
EA
,
Mitrea
C
,
Bao
B
,
Mohammad
RM
,
Polin
LA
,
Dyson
G
, et al
Obesity-induced MBD2_v2 expression promotes tumor-initiating triple-negative breast cancer stem cells
.
Molecular oncology
2019
;
13
:
894
908
.
10.
Bao
B
,
Mitrea
C
,
Wijesinghe
P
,
Marchetti
L
,
Girsch
E
,
Farr
RL
, et al
Treating triple negative breast cancer cells with erlotinib plus a select antioxidant overcomes drug resistance by targeting cancer cell heterogeneity
.
Sci Rep
2017
;
7
:
44125
.
11.
Teslow
EA
,
Bao
B
,
Dyson
G
,
Legendre
C
,
Mitrea
C
,
Sakr
W
, et al
Exogenous IL-6 induces mRNA splice variant MBD2_v2 to promote stemness in TP53 wild-type, African American PCa cells
.
Mol Oncol
2018
;
12
:
1138
52
.
12.
Bienert
GP
,
Schjoerring
JK
,
Jahn
TP
. 
Membrane transport of hydrogen peroxide
.
Biochim Biophys Acta
2006
;
1758
:
994
1003
.
13.
Pierobon
M
,
Frankenfeld
CL
. 
Obesity as a risk factor for triple-negative breast cancers: a systematic review and meta-analysis
.
Breast Cancer Res Treat
2013
;
137
:
307
14
.
14.
Gershuni
V
,
Li
YR
,
Williams
AD
,
So
A
,
Steel
L
,
Carrigan
E
, et al
Breast cancer subtype distribution is different in normal weight, overweight, and obese women
.
Breast Cancer Res Treat
2017
;
163
:
375
81
.
15.
Chen
L
,
Cook
LS
,
Tang
MT
,
Porter
PL
,
Hill
DA
,
Wiggins
CL
, et al
Body mass index and risk of luminal, HER2-overexpressing, and triple negative breast cancer
.
Breast Cancer Res Treat
2016
;
157
:
545
54
.
16.
Berclaz
G
,
Li
S
,
Price
KN
,
Coates
AS
,
Castiglione-Gertsch
M
,
Rudenstam
CM
, et al
Body mass index as a prognostic feature in operable breast cancer: the International Breast Cancer Study Group experience
.
Ann Oncol
2004
;
15
:
875
84
.
17.
Fontanella
C
,
Lederer
B
,
Gade
S
,
Vanoppen
M
,
Blohmer
JU
,
Costa
SD
, et al
Impact of body mass index on neoadjuvant treatment outcome: a pooled analysis of eight prospective neoadjuvant breast cancer trials
.
Breast Cancer Res Treat
2015
;
150
:
127
39
.
18.
Ewertz
M
,
Jensen
MB
,
Gunnarsdottir
KA
,
Hojris
I
,
Jakobsen
EH
,
Nielsen
D
, et al
Effect of obesity on prognosis after early-stage breast cancer
.
J Clin Oncol
2011
;
29
:
25
31
.
19.
Freudenberg
JM
,
Ghosh
S
,
Lackford
BL
,
Yellaboina
S
,
Zheng
X
,
Li
R
, et al
Acute depletion of Tet1-dependent 5-hydroxymethylcytosine levels impairs LIF/Stat3 signaling and results in loss of embryonic stem cell identity
.
Nucleic Acids Res
2012
;
40
:
3364
77
.
20.
Coulter
JB
,
O'Driscoll
CM
,
Bressler
JP
. 
Hydroquinone increases 5-hydroxymethylcytosine formation through ten eleven translocation 1 (TET1) 5-methylcytosine dioxygenase
.
J Biol Chem
2013
;
288
:
28792
800
.
21.
Mitrea
C
,
Wijesinghe
P
,
Dyson
G
,
Kruger
A
,
Ruden
DM
,
Draghici
S
, et al
Integrating 5hmC and gene expression data to infer regulatory mechanisms
.
Bioinformatics
2018
;
34
:
1441
7
.
22.
Giordano
CR
,
Terlecky
SR
. 
Peroxisomes, cell senescence, and rates of aging
.
Biochim Biophys Acta
2012
;
1822
:
1358
62
.
23.
Dreumont
N
,
Hardy
S
,
Behm-Ansmant
I
,
Kister
L
,
Branlant
C
,
Stevenin
J
, et al
Antagonistic factors control the unproductive splicing of SC35 terminal intron
.
Nucleic Acids Res
2010
;
38
:
1353
66
.
24.
Hsu
CH
,
Peng
KL
,
Kang
ML
,
Chen
YR
,
Yang
YC
,
Tsai
CH
, et al
TET1 suppresses cancer invasion by activating the tissue inhibitors of metalloproteinases
.
Cell Rep
2012
;
2
:
568
79
.
25.
Wu
MZ
,
Chen
SF
,
Nieh
S
,
Benner
C
,
Ger
LP
,
Jan
CI
, et al
Hypoxia drives breast tumor malignancy through a TET-TNFalpha-p38-MAPK signaling axis
.
Cancer Res
2015
;
75
:
3912
24
.
26.
Wild
L
,
Flanagan
JM
. 
Genome-wide hypomethylation in cancer may be a passive consequence of transformation
.
Biochim Biophys Acta
2010
;
1806
:
50
7
.
27.
Portela
A
,
Esteller
M
. 
Epigenetic modifications and human disease
.
Nat Biotechnol
2010
;
28
:
1057
68
.
28.
Das
PM
,
Singal
R
. 
DNA methylation and cancer
.
J Clin Oncol
2004
;
22
:
4632
42
.
29.
Huang
Y
,
Chavez
L
,
Chang
X
,
Wang
X
,
Pastor
WA
,
Kang
J
, et al
Distinct roles of the methylcytosine oxidases Tet1 and Tet2 in mouse embryonic stem cells
.
Proc Natl Acad Sci U S A
2014
;
111
:
1361
6
.
30.
Yildirim
O
,
Li
R
,
Hung
JH
,
Chen
PB
,
Dong
X
,
Ee
LS
, et al
Mbd3/NURD complex regulates expression of 5-hydroxymethylcytosine marked genes in embryonic stem cells
.
Cell
2011
;
147
:
1498
510
.
31.
Tsai
YP
,
Chen
HF
,
Chen
SY
,
Cheng
WC
,
Wang
HW
,
Shen
ZJ
, et al
TET1 regulates hypoxia-induced epithelial-mesenchymal transition by acting as a co-activator
.
Genome Biol
2014
;
15
:
513
.
32.
Prat
A
,
Perou
CM
. 
Deconstructing the molecular portraits of breast cancer
.
Mol Oncol
2011
;
5
:
5
23
.
33.
Dutil
J
,
Chen
Z
,
Monteiro
AN
,
Teer
JK
,
Eschrich
SA
. 
An interactive resource to probe genetic diversity and estimated ancestry in cancer cell lines
.
Cancer Res
2019
;
79
:
1263
73
.
34.
Forozan
F
,
Veldman
R
,
Ammerman
CA
,
Parsa
NZ
,
Kallioniemi
A
,
Kallioniemi
OP
, et al
Molecular cytogenetic analysis of 11 new breast cancer cell lines
.
Br J Cancer
1999
;
81
:
1328
34
.
35.
Bookout
AL
,
Cummins
CL
,
Mangelsdorf
DJ
,
Pesola
JM
,
Kramer
MF
. 
High-throughput real-time quantitative reverse transcription PCR
.
Curr Protoc Mol Biol
2006
;
Chapter 15:Unit 15 8
.
36.
WHO
. 
Technical report series 894: obesity: preventing and managing the global epidemic
.
Geneva
:
World Health Organization
; 
2000
.
37.
Song
CX
,
Diao
J
,
Brunger
AT
,
Quake
SR
. 
Simultaneous single-molecule epigenetic imaging of DNA methylation and hydroxymethylation
.
Proc Natl Acad Sci U S A
2016
;
113
:
4338
43
.
38.
Kilic
Y
,
Celebiler
AC
,
Sakizli
M
. 
Selecting housekeeping genes as references for the normalization of quantitative PCR data in breast cancer
.
Clin Transl Oncol
2014
;
16
:
184
90
.
39.
Chandel
NS
,
McClintock
DS
,
Feliciano
CE
,
Wood
TM
,
Melendez
JA
,
Rodriguez
AM
, et al
Reactive oxygen species generated at mitochondrial complex III stabilize hypoxia-inducible factor-1alpha during hypoxia: a mechanism of O2 sensing
.
J Biol Chem
2000
;
275
:
25130
8
.
40.
Evans
B
,
Griner
E
,
Reproducibility Project: Cancer Biology
. 
Registered report: oncometabolite 2-hydroxyglutarate is a competitive inhibitor of alpha-ketoglutarate-dependent dioxygenases
.
Elife
2015
;
4
:
e07420
.
41.
Good
CR
,
Panjarian
S
,
Kelly
AD
,
Madzo
J
,
Patel
B
,
Jelinek
J
, et al
TET1-mediated hypomethylation activates oncogenic signaling in triple-negative breast cancer
.
Cancer Res
2018
;
78
:
4126
37
.
42.
Thiagarajan
PS
,
Hitomi
M
,
Hale
JS
,
Alvarado
AG
,
Otvos
B
,
Sinyuk
M
, et al
Development of a fluorescent reporter system to delineate cancer stem cells in triple-negative breast cancer
.
Stem Cells
2015
;
33
:
2114
25
.
43.
Ke
H
,
Zhao
L
,
Zhang
H
,
Feng
X
,
Xu
H
,
Hao
J
, et al
Loss of TDP43 inhibits progression of triple-negative breast cancer in coordination with SRSF3
.
Proc Natl Acad Sci U S A
2018
;
115
:
E3426
E35
.
44.
Ou
SH
,
Wu
F
,
Harrich
D
,
Garcia-Martinez
LF
,
Gaynor
RB
. 
Cloning and characterization of a novel cellular protein, TDP-43, that binds to human immunodeficiency virus type 1 TAR DNA sequence motifs
.
J Virol
1995
;
69
:
3584
96
.
45.
Bachman
M
,
Uribe-Lewis
S
,
Yang
X
,
Williams
M
,
Murrell
A
,
Balasubramanian
S
. 
5-Hydroxymethylcytosine is a predominantly stable DNA modification
.
Nat Chem
2014
;
6
:
1049
55
.
46.
Ito
S
,
Shen
L
,
Dai
Q
,
Wu
SC
,
Collins
LB
,
Swenberg
JA
, et al
Tet proteins can convert 5-methylcytosine to 5-formylcytosine and 5-carboxylcytosine
.
Science
2011
;
333
:
1300
3
.
47.
Hu
L
,
Lu
J
,
Cheng
J
,
Rao
Q
,
Li
Z
,
Hou
H
, et al
Structural insight into substrate preference for TET-mediated oxidation
.
Nature
2015
;
527
:
118
22
.
48.
Shimbo
T
,
Takaku
M
,
Wade
PA
. 
High-quality ChIP-seq analysis of MBD3 in human breast cancer cells
.
Genom Data
2016
;
7
:
173
4
.
49.
Trivers
KF
,
Lund
MJ
,
Porter
PL
,
Liff
JM
,
Flagg
EW
,
Coates
RJ
, et al
The epidemiology of triple-negative breast cancer, including race
.
Cancer Causes Control
2009
;
20
:
1071
82
.
50.
Ward
ZJ
,
Bleich
SN
,
Cradock
AL
,
Barrett
JL
,
Giles
CM
,
Flax
C
, et al
Projected U.S. state-level prevalence of adult obesity and severe obesity
.
N Engl J Med
2019
;
381
:
2440
50
.