The mechanisms of how cancer cells are selected and evolve to establish distant metastatic colonies remain unclear. Tumor heterogeneity and lack of biomarkers are some of the most difficult challenges in cancer biology and treatment. Here using mouse models for triple-negative breast cancer (TNBC) metastasis, we report heterogeneous expression of DNA methyltransferase 3B (DNMT3B) in both mouse and human primary tumors. High levels of DNMT3B were correlated with poor clinical outcomes in multiple human breast cancer datasets. Mechanistically, clonal cells with high DNMT3B (DNMT3BH) showed higher vimentin (VIM) expression and displayed enhanced epithelial-to-mesenchymal transition capacity. Deletion of VIM diminished the metastatic phenotype of DNMT3BH cells. Importantly, in preclinical mouse models in which the primary tumors were surgically removed, perioperative targeting of DNMT3B in combination with chemotherapy markedly suppressed tumor recurrence and metastasis. Our studies identify DNMT3B-mediated transcription regulation as an important mediator of tumor heterogeneity and show that DNMT3B is critical for tumor invasion and metastasis, reinforcing its potential as a target for treating metastatic disease.

Implications:

Our findings of transcriptome changes mediated by DNMT3B provide new mechanistic insight for intratumor heterogeneity and chemoresistance, and therapeutic targeting of DNMT3B in combination with chemotherapy offer additional treatment options for metastatic disease especially for patients with TNBC.

This article is featured in Highlights of This Issue, p. 1589

Advances in early detection and targeted therapies have significantly improved outcomes for patients with breast cancer. However, there are limited therapeutic options to treat metastatic disease, resulting in significant patient mortality (1). Triple-negative breast cancer (TNBC) is an aggressive subtype characterized by extensive intratumoral heterogeneity, and frequently develops resistance to chemotherapy (2). Tumor heterogeneity and a lack of biomarkers remain to be some of the most difficult challenges in achieving high therapeutic efficacy (3, 4). Aside from the distinct genomic mutations that promote emergence or expansion of therapy-resistant cancer cells (5, 6), epigenetic and transcriptomic changes have been recently identified as key factors in tumor heterogeneity and evolution (7–10). In addition, neoadjuvant therapy can promote the development of resistant subpopulations in human breast tumors through transcriptional reprogramming without genomic alterations (11, 12). These studies demonstrate that tumor cells are selected for and evolved from a heterogeneous population of primary tumors through acquired transcriptional alterations. Therefore, identification of transcriptional regulatory mechanisms that drive metastatic clonal emergence or expansion is critical to enhance our understanding of the metastatic process and treatment options.

DNA methylation pattern is one of the strongest signatures to classify cancer cell type with its origin in The Cancer Genome Atlas (TCGA) analysis (13). DNA methylation is also one of the key molecular mechanisms that regulates epithelial–mesenchymal transition (EMT), which provides cellular plasticity for metastatic advantage (14, 15). Of the three DNA methyltransferases (DNMT) that catalyze DNA methylation at the 5′ cytosine of CpG sites, DNA methyltransferase 3B (DNMT3B) mediates de novo DNA methylation and is required to silence genes through promoter hypermethylation during development (16). Aberrant overexpression of DNMT3B is reported in several human cancers, including breast cancer, while its expression levels are low in many adult tissues (17–19). Preclinical studies demonstrated that overexpression of DNMT3B promotes primary tumor progression of melanoma and colon cancer (20–22). Recently, we reported that DNMT3B is induced at metastatic sites and facilitates metastatic colonization (23). However, the intratumoral DNMT3B heterogeneity and its mechanism of function remain to be investigated.

Surgical resection of primary tumors or metastatic nodules has been reported to mobilize inflammatory immune cells and to increase inflammatory molecules in circulation (24, 25). In addition, neoadjuvant and adjuvant chemotherapy, which are the most common way to treat cancer, also increase the systemic secretion of inflammatory molecules (26, 27). It is not clear whether targeting the surgery- or chemotherapy-induced inflammation would be a valid approach to suppress metastatic spread and to minimize therapeutic adverse effects.

In this study, we report intratumoral heterogeneity of DNMT3B in breast cancer tissues, and that high DNMT3B (DNMT3BH) levels are correlated with poor patient survival and more aggressive subtypes of breast cancer. In mouse models, tumor cells with high DNMT3B level display enhanced migration and metastatic colonization. Importantly, targeting DNMT3B expression with perioperative application of anti-inflammatory drugs and targeting DNMT3B enzymatic activities potentiated the chemotherapeutic efficacy in suppressing breast cancer metastasis. Our studies reveal a novel mechanism of DNMT3B in metastatic spread and targeted therapies aimed at the specific clonal cells with high DNMT3B expression will likely provide additional treatment options.

Cell lines and mice

The murine mammary tumor cell lines 4T1 and 6DT1 were maintained in 10% FBS DMEM (Life Technologies) in a 37°C incubator with 5% CO2. All cell lines were obtained from ATCC and were verified to be Mycoplasma negative with the MycoAlert Mycoplasma Detection Kit (Lonza). The cells were used within 2 months after thawing. To establish clonal cell lines with high or low DNMT3B expression, clonal cells were derived from single-cell seeding, and the clonal cell lines maintained the high or low DNMT3B levels after three passages were selected. The experiments with the clonal cell lines were performed within 2 weeks after thawing. The generation of the DNMT3B knockdown and DNMT3B-inducible cell line was described in our previous study (23). Cell line authentication has not been performed. All mice were housed at the NCI animal facility, and experiments were approved by the NCI Animal Care and Use Committee.

IHC and immunofluorescence

Paraffin sections of mouse mammary tumors and a human breast tumor tissue array (BRM961, US Biomax Inc.) were used. DNMT3B expression was detected with a DNMT3B antibody (1:100, ab2851, Abcam), and hematoxylin was utilized as a nuclear counterstain. For immunofluorescence (IF) staining, the frozen lung and tumor sections were fixed with 4% formaldehyde (28906, Thermo Fisher Scientific), then incubated with vimentin (VIM; Proteintech), Ki67 (M7240, Agilent), or Annexin V (11060-1-AP, Proteintech) antibodies overnight at 4°C after blocking with 10% normal serum. Fluorescence-tagged secondary antibodies and DAPI were utilized for visualization. Images were obtained (Carl Zeiss LSM 880), and quantitative analyses of the IHC and IF images were performed by ImageJ.

Human data analysis

Publicly available human datasets, GEO-GSE6532 (28), Yau (2010) (29), TCGA-BRCA, and METABRIC (30) for patient survival, GEO-GSE12276 (31) and GEO-GSE25065 (32) for metastasis status of patients, TCGA-BRCA for tumor stages and subtypes, were used to investigate the correlation of DNMT3B expression with stages, cancer subtypes, and patient survival. The gene expression associated with promoter methylation and its correlation with breast cancer patient survival was analyzed using the methylation expression index (MEI) as reported previously (33). The MEI for 10 DNMT3B targeted genes from the previous study (23) were determined in the GSE59000 dataset and applied to GEO-GSE6532, Yau (2010), TCGA-BRCA, and METABRIC datasets for patient survival analysis.

Mouse tumor models and treatment

Tumor cells (4T1 or 6DT1, 2.5 × 105) were injected into the #2 mammary fat pad (MFP) or tail vein injection (TVI) of 8-week-old female BALB/c, nude, or FVB/N mice (Charles River). In some experiments, TVI was performed in mice that received MFP injection 2 weeks earlier. Mice were sacrificed at given timepoints as described in each experimental design. Weight of primary tumors and the number of metastatic nodules were evaluated at necropsy, and primary tumors, lungs, and circulating tumor cells (CTC) were collected for further analysis.

For drug treatment experiments, the 4T1 or 6DT1 primary tumors were surgically removed on day 14. The mice were treated with paclitaxel (6 mg/kg bodyweight, i.v. once a week), gemcitabine (2 mg/kg bodyweight, i.v. once a week), 5-azacytidine (5-Aza; 1 mg/kg body weight, i.p. three times per week), or celecoxib (1,500 ppm in diet). To target surgery-induced inflammation, celecoxib was given to the mice perioperatively or long term, as shown in the experimental design (Fig. 7C, top). For preclinical treatment, mice were sacrificed when the recurrent tumor size reached > 2 cm or mice showed labored breathing or bodyweight loss of more than 10%.

Western blotting

Protein extraction from cultured tumor cells, primary tumors, and metastatic nodules was performed using a RIPA protein extraction buffer containing protease and phosphatase inhibitors (Thermo Fisher Scientific). Antibodies against DNMT3B (ab79822, Abcam), VIM (10366-1-AP, Proteintech), β-Actin (sc-69879, Santa Cruz Biotechnology), p-STAT3 (9145, Cell Signaling Technology), STAT3 (12640, Cell Signaling Technology), and p-p65 NFκB, (3033, Cell Signaling Technology) were diluted at 1:500 to 1:1,000. Quantification of Western blot density was measured by ImageJ software, each band was normalized by corresponding β-actin band.

Flow cytometry

Mouse primary tumors and lungs were chopped into small pieces using sharp blades followed by incubation in tissue dissociation buffer (Collagenase, Dispase, and DNase in DMEM) to generate a single-cell suspension. CTCs were collected from the blood through cardiac puncture. The samples were stained with DAPI to determine cell viability and analyzed by flow cytometry (LSRFortessa, BD Biosciences). The data were analyzed by FlowJo (BD Biosciences).

Cytokine antibody array

Proteins were extracted from the lungs of control or celecoxib diet-treated mice and were examined using a cytokine protein antibody array (AAM-CYT-1000-2, Raybiotech) per the manufacturer's instructions.

Wound-healing migration assay

For Vim knockout in 4T1 cells, the cells were transduced with Pspcas9-2A-puro-px459 constructs containing Vim single-guide RNAs, followed by 2 μg/mL puromycin selection for 2 days. The knockout of VIM was verified by Western blots. Control or VIM-KO cells were incubated until they reached 100% confluency and were then treated with mitomycin c for 1 hour to inhibit proliferation. 700–800 μm wide wounds were made using an IncucyteWoundmaker (Essen BioScience, Inc.). Images to monitor cell migration were taken at 0, 6, 12, 18, and 24 hours by Incucyte S4 (Essen BioScience, Inc.), and the relative wound recovery at 12 hours after wounding was quantitated by Image J with MRI wound-healing tool plug in.

Transwell invasion assay

Transwell inserts of 24-well plate (Millipore Sigma) were coated with matrigel (Millipore Sigma). Cancer cells were seeded into the inserts and incubated for 18 hours. The invaded cells were fixed by 5% glutaraldehyde for 10 minutes then stained with 0.5% crystal violet solution for 20 minutes. The matrigel and cells in the inserts were removed by wet cotton swap. The numbers of invaded cells at the bottom of inserts were measured by EVOS cell imaging system.

Statistical analysis

GraphPad Prism v6.01 and R were used for graphs and statistics. Unless otherwise indicated, data were expressed as mean ± SD. All data were analyzed using the Student t test for comparison of two groups or one-way ANOVA for three groups or more. Differences were considered statistically significant when the P value was < 0.05.

DNMT3B expression is heterogeneous in primary tumors, with high DNMT3B expression correlating with poor prognosis for patients with breast cancer

Our previous study showed that DNMT3B is increased in tumor cells from metastatic nodules compared with those from primary tumors (23). In further examining the primary tumor tissues, we observed heterogeneous levels of DNMT3B; some tumor cells expressed high (DNMT3BH) while others expressed modest or low levels (DNMT3BL). The DNMT3BH and DNMT3BL cell heterogeneity was consistent in both mouse mammary tumors (from 4T1 and 6DT1 mouse models; Fig. 1A; Supplementary Fig. S1A and S1B) and primary tumor tissues from patients with breast cancer (Fig. 1B). High expression of DNMT3B was not detected in the normal mammary glands (Supplementary Fig. S1A). We hypothesize that DNMT3BH tumor cells may exhibit distinct biological behavior in the primary tumor microenvironment. Using multiple clinical datasets containing a large number of breast tumor tissues, we first found that the high DNMT3B levels correlated with decreased metastasis-free survival (GSE6532 and Yau), decreased relapse-free survival (TCGA), or decreased disease-specific survival (METABRIC; Fig. 1C), which was not observed for DNMT3A and DNMT1 in the same datasets (Supplementary Fig. S1C). In addition, significantly higher DNMT3B levels were found in the primary tumors from patients with metastases compared with those without metastases (Fig. 1D). High levels of DNMT3B but not DNMT3A or DNMT1 were also observed in advanced clinical stages as well as HER2-enriched and basal-like aggressive subtypes (Fig. 1E; Supplementary Fig. S1D). These results suggest that high expression of DNMT3B in primary tumors correlates with more aggressive tumor subtypes and poor prognosis for patients with breast cancer.

Figure 1.

Expression of DNMT3B in primary tumors and its clinical correlation with human breast cancers. A, DNMT3B IHC of 4T1 mammary tumors and percentage of DNMT3BH cells in all tumor cells. Intensity threshold above 150 in each nucleus was set for DNMT3BH cells by Image J. B, Representative DNMT3B IHC of human breast tumors and percentage of DNMT3BH cells in all tumor cells. Intensity threshold above 150 in each nucleus was set for DNMT3BH cells by Image J. The white arrows indicate tumor cells with high DNMT3B expression, and the black arrows indicate low DNMT3B expression (tissue array, 13 sections of primary tumors examined). C, Kaplan–Meier plots for correlation between DNTM3B expression and patient survival in four different human breast cancer datasets. D, DNMT3B expression levels in breast tumors from patients with or without metastatic diseases (GEO-GSE12276 and GEO-GSE25065). E, DNMT3B expression levels in breast tumors at different stages (left) and in different subtypes (right). Data are presented as mean ± SD. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Figure 1.

Expression of DNMT3B in primary tumors and its clinical correlation with human breast cancers. A, DNMT3B IHC of 4T1 mammary tumors and percentage of DNMT3BH cells in all tumor cells. Intensity threshold above 150 in each nucleus was set for DNMT3BH cells by Image J. B, Representative DNMT3B IHC of human breast tumors and percentage of DNMT3BH cells in all tumor cells. Intensity threshold above 150 in each nucleus was set for DNMT3BH cells by Image J. The white arrows indicate tumor cells with high DNMT3B expression, and the black arrows indicate low DNMT3B expression (tissue array, 13 sections of primary tumors examined). C, Kaplan–Meier plots for correlation between DNTM3B expression and patient survival in four different human breast cancer datasets. D, DNMT3B expression levels in breast tumors from patients with or without metastatic diseases (GEO-GSE12276 and GEO-GSE25065). E, DNMT3B expression levels in breast tumors at different stages (left) and in different subtypes (right). Data are presented as mean ± SD. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Close modal

We have previously reported several genes that are downregulated by DNMT3B-mediated promoter hypermethylation (23). The majority of these genes showed a negative correlation between promoter methylation and gene expression in the human dataset examined here (GSE59000), which has both gene expression and DNA methylation data from same primary tumor samples (Supplementary Table S1). We next selected the 10 top-ranked genes with a Spearman r of less than −0.2 and performed the MEI analysis as described previously (33). The lower MEI values indicate lower gene expression with increased promoter methylation of the 10 genes. A scatter plot between gene expression and DNA methylation of CLDN4 was shown as an example (Fig. 2A). The MEI was then applied to determine patient survival rates in multiple human breast cancer datasets. Patients with low MEI, indicating a decreased overall expression of the 10 DNMT3B target genes, showed significantly decreased survival in all datasets examined (Fig. 2B). Together our data support a strong correlation between high DNMT3B expression and aggressive cancer types as well as poor prognosis of patients with breast cancer mediated by epigenetic and transcriptomic changes.

Figure 2.

Correlation between MEI of DNMT3B targeted genes and patient survival. A, Scatter plot of correlation for CLDN4 gene expression and promoter methylation (left), and Spearman r values for correlation between gene expression and promoter methylation of 10 DNMT3B-targeted genes (right) in a human breast cancer dataset (GSE59000). B, Kaplan–Meier plots for correlation between MEI of the 10 DNMT3B-targeted genes and patient survival from four different human breast cancer datasets (P < 0.05, P < 0.01).

Figure 2.

Correlation between MEI of DNMT3B targeted genes and patient survival. A, Scatter plot of correlation for CLDN4 gene expression and promoter methylation (left), and Spearman r values for correlation between gene expression and promoter methylation of 10 DNMT3B-targeted genes (right) in a human breast cancer dataset (GSE59000). B, Kaplan–Meier plots for correlation between MEI of the 10 DNMT3B-targeted genes and patient survival from four different human breast cancer datasets (P < 0.05, P < 0.01).

Close modal

DNMT3BH tumor cells have higher metastatic capacity than DNMT3BL cells

To investigate the biological behavior of DNMT3BH tumor cells in the primary tumor microenvironment, DNMT3BH and DNMT3BL 4T1 clonal cell lines were established from mouse 4T1 cells (Fig. 3A) and labeled with GFP and RFP, respectively. The expression levels of DNMT3B were quantified by measuring DNMT3B band density, with an average 1.26 for DNMT3BH, 0.70 for DNMT3BL cells and 0.68 for normal mammary gland (Fig. 3A; Supplementary Fig. S2A). The protein levels of DNMT1 and DNMT3a were not different between DNMT3BH and DNMT3BL cells (Fig. 3A), and were not affected by DNMT3B knockdown (Supplementary Fig. S2B). Three DNMT3BH or DNMT3BL clonal cell lines were pooled for in vivo experiments to minimize the clonal variation. The pooled cells were injected into mouse MFPs. As expected from our previous studies (23), there were more metastases from DNMT3BH cells compared with DNMT3BL cells (Fig. 3B). DNMT3B knockdown in DNMT3BH cells diminished this phenotype (Fig. 3B). To model the coexisting of DNMT3BH and DNMT3BL cells in primary tumors, a clonal competition experiment was performed in which an equal number of DNMT3BH and DNMT3BL cells were coinjected into the MFP (Fig. 3C). The lung nodules from DNMT3BH and DNMT3BL clonal cell lines were counted separately by GFP or RFP label (Fig. 3C). We observed significantly more lung nodules from DNMT3BH cells than DNMT3BL cells (Fig. 3D, left). DNMT3B knockdown in DNMT3BH cells markedly decreased the number of lung nodules (Fig. 3D, left). This observation was also well documented by IF imaging of the lungs as DNMT3BH and DNMT3BL cells were labeled with GFP and RFP, respectively (Fig. 3D, right). Additional flow cytometry analysis of single-cell suspensions from the lungs of mice that received coinjection showed an increased number of DNMT3BH cells compared with the DNMT3BL cells in the lung (Fig. 3E; Supplementary Fig. S2C).

Figure 3.

Metastatic capacity of DNMT3BH tumor cells compared with DNMT3BL cells. A, DNMT3B, DNMT1, and DNMT4A Western blot analysis for DNMT3BH and DNMT3BL clonal 4T1 cell lines. B, Surface lung metastases (met) counts from mice that received MFP injections of DNMT3BH, DNMT3BL, and DNMT3BH-KD 4T1 cells. C, Schematic experimental design for metastatic potential analysis of DNMT3BH, DNMT3BL, DNMT3BH-KD 4T1 cells. D, Quantitation of DNMT3BH-GFP or DNMT3BL-RFP lung metastases at week 4 from mice that received MFP injection of tumor cells (n = 5; left). Representative images are on the right. The white dotted lines indicate the lung boundary. E, Flow cytometry analysis of GFP+ or RFP+ tumor cells at week 4 from the lungs of mice that received MFP injection of tumor cells (left), with representative plots of GFP+ or RFP+ tumor cells (right). F, DNMT3BH and DNMT3BL metastases in the lungs in a time course experiment (2, 3, and 4 weeks, MFP injection) in which the DNMT3BH and DNMT3BL cells were injected together for clonal competition. Top left, schematic design; bottom left, quantitative data; Right, representative IF images. Quantifications of band density are presented in Western blot analysis. Data are presented as mean ± SD. **, P < 0.01; ***, P < 0.001.

Figure 3.

Metastatic capacity of DNMT3BH tumor cells compared with DNMT3BL cells. A, DNMT3B, DNMT1, and DNMT4A Western blot analysis for DNMT3BH and DNMT3BL clonal 4T1 cell lines. B, Surface lung metastases (met) counts from mice that received MFP injections of DNMT3BH, DNMT3BL, and DNMT3BH-KD 4T1 cells. C, Schematic experimental design for metastatic potential analysis of DNMT3BH, DNMT3BL, DNMT3BH-KD 4T1 cells. D, Quantitation of DNMT3BH-GFP or DNMT3BL-RFP lung metastases at week 4 from mice that received MFP injection of tumor cells (n = 5; left). Representative images are on the right. The white dotted lines indicate the lung boundary. E, Flow cytometry analysis of GFP+ or RFP+ tumor cells at week 4 from the lungs of mice that received MFP injection of tumor cells (left), with representative plots of GFP+ or RFP+ tumor cells (right). F, DNMT3BH and DNMT3BL metastases in the lungs in a time course experiment (2, 3, and 4 weeks, MFP injection) in which the DNMT3BH and DNMT3BL cells were injected together for clonal competition. Top left, schematic design; bottom left, quantitative data; Right, representative IF images. Quantifications of band density are presented in Western blot analysis. Data are presented as mean ± SD. **, P < 0.01; ***, P < 0.001.

Close modal

To further understand the metastatic behavior of DNMT3BH tumor cells at different stages of metastatic progression, we performed a time course experiment in which the DNMT3BH and DNMT3BL cells were coinjected into MFPs, and lungs were evaluated for GFP and RFP metastatic nodules at 2, 3, and 4 weeks after injection (Fig. 3F, top left). At week 2, no DNMT3BH or DNMT3BL cancer cells were detected in the lung (Fig. 3F, bottom left and IF). At week 3, there were DNMT3BH-GFP and mixed DNMT3BH-GFP/DNMT3BL-RFP lung nodules (Fig. 3F, bottom left and IF). However, no lung nodules contained only DNMT3BL-RFP cells (Fig. 3F, bottom left). At week 4, some lung nodules comprised of only DNMT3BL-RFP cells were observed but at lower frequency than DNMT3BH-GFP nodules (Fig. 3F, bottom left and IF). In addition, the sizes of DNMT3BL lung nodules at week 4 were smaller than the DNMT3BH lung nodules, as shown in the representative images (Supplementary Fig. S2D). These results suggest that DNMT3BH cells colonize earlier and faster than the DNMT3BL cells.

DNMT3BH cells display an enhanced EMT phenotype

The weights of primary tumors derived from DNMT3BH cells or DNMT3BL cells were not significantly different. DNMT3B knockdown in DNTM3BH cells showed only a modest reduction in tumor weight (Supplementary Fig. S2E). However, when we further examined the clonal composition of the primary tumors from the coinjection of DNMT3BH and DNMT3BL cells, there was higher percentage of DNMT3BH cells than DNTM3BL cells (Supplementary Fig. S2F). This result indicates that DNMT3BH cells have higher capacity for cell proliferation and clonal expansion over DNMT3BL cells in the primary tumors where different clonal cells coexist and compete. We next determined whether DNMT3BH cells have a better capacity to escape from the primary tumors. We first measured the number of CTCs at the early timepoints of tumor cell dissemination (Fig. 4A). We observed a higher number of CTCs in the animals injected with DNMT3BH cells than those injected with DNMT3BL cells, which was even more pronounced at day 17 and day 20 than day 14 (Fig. 4B; Supplementary Fig. S3A). Consistently, in the coinjection of DNMT3BH and DNMT3BL cells, the number of DNMT3BH CTCs was markedly higher than the DNMT3BL CTCs (Fig. 4B). The number of lung metastases from DNMT3BH cells was significantly higher than those from DNMT3BL cells while there was no difference in primary tumor weights between those composed of DNMT3BH or DNMT3BL cells at any of the early timepoints examined (Supplementary Fig. S3B).

Figure 4.

DNMT3BH tumor cells displayed enhanced dissemination. A, Schematic experimental design for a time course of clonal competition of DNMT3BH and DNMT3BL 4T1 tumor cells in early stages of cancer cell dissemination. B, Quantitated data of CTCs from the clonal competition and time course experiment (left). Data were collected 11, 14, 17, and 20 days after tumor cell injection. Representative flow cytometry results of CTCs with indication of cell frequencies (right). C, Quantitated data of VIM-positive cells among DNMT3BH and DNMT3BL cells in the primary tumors (left) from the clonal competition and time course experiment, the same as in B. Representative image of primary tumors (right). Data are presented as mean ± SD. **, P < 0.01; ***, P < 0.001.

Figure 4.

DNMT3BH tumor cells displayed enhanced dissemination. A, Schematic experimental design for a time course of clonal competition of DNMT3BH and DNMT3BL 4T1 tumor cells in early stages of cancer cell dissemination. B, Quantitated data of CTCs from the clonal competition and time course experiment (left). Data were collected 11, 14, 17, and 20 days after tumor cell injection. Representative flow cytometry results of CTCs with indication of cell frequencies (right). C, Quantitated data of VIM-positive cells among DNMT3BH and DNMT3BL cells in the primary tumors (left) from the clonal competition and time course experiment, the same as in B. Representative image of primary tumors (right). Data are presented as mean ± SD. **, P < 0.01; ***, P < 0.001.

Close modal

To further determine the biological properties that lead to the increased number of the DNMT3BH tumor cells in circulation, we measured proliferation (Ki67), apoptosis (annexin V), and EMT state (VIM) of cancer cells in the primary tumors. There was an increase in the number of Ki67-positive DNMT3BH cells at day 20 but not at earlier timepoints, and no difference in the number of annexin V-positive cells between DNMT3BH and DNMT3BL cells (Supplementary Fig. S3C and S3D). Interestingly, there were clearly more VIM-positive tumor cells in the DNMT3BH population than that in the DNMT3BL population (Fig. 4C). In DNMT3BH cells displaying more mesenchymal-like morphology, almost all were VIM positive, as shown in IF images (Fig. 4C, right; Supplementary Fig. S3E). These data revealed that DNMT3BH cells have enhanced EMT phenotype.

High DNMT3B increased VIM expression and the EMT phenotype

We next investigated DNMT3B regulation of VIM and EMT. We observed that DNMT3BH cells expressed higher levels of VIM than DNMT3BL cells (Fig. 5A). DNMT3B knockdown decreased Vim mRNA (Fig. 5B). 5-Aza, an inhibitor targeting DNMT enzymatic activity, also reduced Vim mRNA expression (Fig. 5C). In addition, there were DNMT3B enrichment and high DNA methylation in gene body regions of Vim, as well as Snai1, another well-known EMT marker (Fig. 5D; Supplementary Fig. S4A and S4B). These data suggest a possible regulatory mechanism for Vim and Snai1 expression. Consistently, DNMT3BH cells showed higher levels of Twist2 and Snai1, compared with DNMT3BL cells (Fig. 5E). To further investigate the functional significance of high VIM in DNMT3BH cells, VIM was knocked out in pooled DNMT3BH and DNMT3BL cells (Fig. 5F). VIM knockout diminished the migration of DNMT3BH tumor cells in a wound-healing assay, whereas it had a minimal effect on the DNMT3BL cells (Fig. 5G). VIM knockout also significantly suppressed the invasion of DNMT3BH tumor cells in a matrigel-coated transwell assay, while it did not have effect on DNMT3BL cells (Fig. 5H). Together, these results suggest that the DNMT3BH cells display enhanced EMT through increased expression of Vim, thus intravasate and disseminate from the primary tumors to initiate metastatic colonization.

Figure 5.

DNMT3B increased VIM expression. A, VIM Western blot analysis for DNMT3BH and DNMT3BL clonal cell lines. B, mRNA levels of Dnmt3b and Vim in DNMT3BH, DNMT3BL, and DNMT3BH-KD cells. C, mRNA levels of Dnmt3b and Vim in DNMT3BH cells with 5-Aza treatment. D, DNMT3B enrichment on Vim from DNMT3B chromatin immunoprecipitation (ChIP) sequencing data (GSE146010). E, mRNA levels of Twist1, Twist2, Zeb1, Zeb2, Snai1, Snai2, and Cdh1 in DNMT3BH and DNMT3BL cells. F, DNMT3B and VIM Western blot analysis for pooled DNMT3BH and DNMT3BL clonal cells with and without VIM knockout. G, Wound-healing migration assay for pooled DNMT3BH and DNMT3BL clonal cells with and without VIM knockout. Relative wound recovery in time course (left), representative images at 0 hour and 12 hours after scratch (middle), and relative wound recovery at 12 hours (right). The edge of migrating cells was indicated in yellow. H, Transwell invasion assay for pooled DNMT3BH and DNMT3BL clonal cells with or without VIM knockout. Representative images of invaded cells (left) and number of invaded cells per field (right). Quantifications of band density are presented under each Western blot analysis. Data are presented as mean ± SD. **, P < 0.01; ***, P < 0.001.

Figure 5.

DNMT3B increased VIM expression. A, VIM Western blot analysis for DNMT3BH and DNMT3BL clonal cell lines. B, mRNA levels of Dnmt3b and Vim in DNMT3BH, DNMT3BL, and DNMT3BH-KD cells. C, mRNA levels of Dnmt3b and Vim in DNMT3BH cells with 5-Aza treatment. D, DNMT3B enrichment on Vim from DNMT3B chromatin immunoprecipitation (ChIP) sequencing data (GSE146010). E, mRNA levels of Twist1, Twist2, Zeb1, Zeb2, Snai1, Snai2, and Cdh1 in DNMT3BH and DNMT3BL cells. F, DNMT3B and VIM Western blot analysis for pooled DNMT3BH and DNMT3BL clonal cells with and without VIM knockout. G, Wound-healing migration assay for pooled DNMT3BH and DNMT3BL clonal cells with and without VIM knockout. Relative wound recovery in time course (left), representative images at 0 hour and 12 hours after scratch (middle), and relative wound recovery at 12 hours (right). The edge of migrating cells was indicated in yellow. H, Transwell invasion assay for pooled DNMT3BH and DNMT3BL clonal cells with or without VIM knockout. Representative images of invaded cells (left) and number of invaded cells per field (right). Quantifications of band density are presented under each Western blot analysis. Data are presented as mean ± SD. **, P < 0.01; ***, P < 0.001.

Close modal

DNMT3B is a key target of anti-inflammatory drugs to inhibit metastasis

Our previous studies identified prostaglandin E2 (PGE2) as one of the most significantly increased inflammatory molecules in the metastatic microenvironment that is critical for DNMT3B induction through the STAT3 and NFκB signaling pathways (23). Celecoxib is the most widely prescribed FDA-approved selective inhibitor of COX2, a prostaglandin-endoperoxide synthase also known as cyclooxygenase 2, and it significantly decreases PGE2 production. We thus first investigated the effects of celecoxib on the levels of various cytokines in the lungs of 4T1 tumor-bearing mice by a cytokine array experiment. Treatment with celecoxib decreased the levels of protumor cytokines (GCSF, VEGF, IL1a, eotaxin, SCF) and increased antitumor cytokines (XCL1, IGFBP6, PF4; Fig. 6A). Treatment with celecoxib decreased DNMT3B and inhibited the activation of STAT3 and NFκB in the lung nodules (Fig. 6B). To further examine the cause-effect role of DNMT3B for metastasis suppression by celecoxib, GFP-labeled 4T1 cells with a doxycycline-inducible overexpression of DNMT3B vector were injected through the tail vein into the mice bearing primary tumors (Fig. 6C). As expected, celecoxib decreased the number of metastatic nodules (Fig. 6D), while doxycycline-induced overexpression of DNMT3B significantly increased the number of metastatic nodules (Fig. 6D). Importantly, doxycycline induction of DNMT3B in tumor cells rescued the metastasis suppression by celecoxib treatment (Fig. 6D, left and middle), with no difference in primary tumor weight (Fig. 6D, right). Thus, DNMT3B is a key target in metastasis suppression by celecoxib.

Figure 6.

Anti-inflammatory drug targeting of PGE2 in DNMT3B-mediated metastasis. A, Heatmap from cytokine array experiments on protein extracts of lungs from 4T1 tumor-bearing mice treated with celecoxib or control (left), with the quantitative data for the fold changes of the protumor and antitumor cytokines (right). B, DNMT3B, pSTAT3, and pNFκB Western blots of primary tumor, normal lung, and metastatic nodules from mice treated with celecoxib or control. C, Schematic experimental design for celecoxib treatment and Dox-DNMT3B induction (left). Nude mice were injected (TVI) with 4T1-GFP cells (with inducible DNMT3B) 14 days after the initial injection of 4T1 tumor cells in the MFP. The mice were treated with doxycycline to induce DNMT3B expression. D, The number of GFP+ lung nodules from mice that received TVI of 4T1-GFP cells and representative images (middle), and the weights of primary tumors (right). Quantifications of band density are presented in Western blot analysis. Data are presented as mean ± SD. *, P < 0.05; **, P < 0.01.

Figure 6.

Anti-inflammatory drug targeting of PGE2 in DNMT3B-mediated metastasis. A, Heatmap from cytokine array experiments on protein extracts of lungs from 4T1 tumor-bearing mice treated with celecoxib or control (left), with the quantitative data for the fold changes of the protumor and antitumor cytokines (right). B, DNMT3B, pSTAT3, and pNFκB Western blots of primary tumor, normal lung, and metastatic nodules from mice treated with celecoxib or control. C, Schematic experimental design for celecoxib treatment and Dox-DNMT3B induction (left). Nude mice were injected (TVI) with 4T1-GFP cells (with inducible DNMT3B) 14 days after the initial injection of 4T1 tumor cells in the MFP. The mice were treated with doxycycline to induce DNMT3B expression. D, The number of GFP+ lung nodules from mice that received TVI of 4T1-GFP cells and representative images (middle), and the weights of primary tumors (right). Quantifications of band density are presented in Western blot analysis. Data are presented as mean ± SD. *, P < 0.05; **, P < 0.01.

Close modal

Perioperative application of celecoxib in combination with 5-Aza or chemotherapeutic drugs inhibits metastasis

Cancer surgery is fundamentally important in the treatment of solid tumors. However, recent studies suggest that surgical removal of primary tumors or metastatic lesions induces systemic inflammatory responses that stimulate metastatic spread and dormant tumor reactivation (24, 25). As PGE2 is a key inflammatory mediator (25), we tested whether celecoxib would decrease cancer metastasis in the 4T1 preclinical mouse model following surgical resection of the primary tumor (Fig. 7A). In addition, considering the adverse cardiovascular events observed for COX-2 inhibitors in previous clinical trials (34), we directly compared the effect of perioperative short-term celecoxib treatment (7 days, 1,500 ppm in chow) and those from long-term treatment (23 days, 1,500 ppm in chow; Fig. 7A). Both short- and long-term celecoxib significantly decreased the number of metastases, with no difference between the short-term versus the long-term treatment (Fig. 7B, left). Of note, long-term but not perioperative celecoxib treatment decreased recurrent primary tumor weight (Fig. 7B, right).

Figure 7.

Perioperative application of celecoxib in combination with 5-Aza or chemotherapies in metastasis treatment. A, Schematic experimental design for surgical removal of primary tumors and perioperative versus long-term celecoxib treatment. B, Number of metastases in the lungs (middle) and the weights of recurrent tumors in mice treated with perioperative versus long-term celecoxib. C, Top: Schematic experimental design for combination of chemotherapeutics with perioperative and long-term celecoxib treatment; bottom: number of metastases in the lungs (bottom left) and the weights of recurrent tumors (bottom right) from mice that received paclitaxel (Pac), gemcitabine (Gem), or 5-Aza in combination with perioperative (S, short) or long-term (L, long) celecoxib (Celx). D, Percent survival of 4T1 (left) or 6DT1 (right) tumor-bearing mice with single or combined drug treatment with gemcitabine or 5-Aza alone, the two in combination, or each of the two drugs in combination with perioperative celecoxib. Data are presented as mean ± SD. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Figure 7.

Perioperative application of celecoxib in combination with 5-Aza or chemotherapies in metastasis treatment. A, Schematic experimental design for surgical removal of primary tumors and perioperative versus long-term celecoxib treatment. B, Number of metastases in the lungs (middle) and the weights of recurrent tumors in mice treated with perioperative versus long-term celecoxib. C, Top: Schematic experimental design for combination of chemotherapeutics with perioperative and long-term celecoxib treatment; bottom: number of metastases in the lungs (bottom left) and the weights of recurrent tumors (bottom right) from mice that received paclitaxel (Pac), gemcitabine (Gem), or 5-Aza in combination with perioperative (S, short) or long-term (L, long) celecoxib (Celx). D, Percent survival of 4T1 (left) or 6DT1 (right) tumor-bearing mice with single or combined drug treatment with gemcitabine or 5-Aza alone, the two in combination, or each of the two drugs in combination with perioperative celecoxib. Data are presented as mean ± SD. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Close modal

The efficacy of perioperative celecoxib was further investigated in combined treatment with the chemotherapeutic drugs paclitaxel or gemcitabine as well as the DNA methyltransferase inhibitor 5-Aza, a drug targeting the enzymatic activity of DNMTs (Fig. 7C, top left). The combinations of perioperative celecoxib, low-dose 5-Aza, or chemotherapeutic agents decreased the number of metastasis nodules (Fig. 7C, bottom left) and tumor recurrence in 4T1 tumor-bearing mice (Fig. 7C, right). Particularly, the perioperative treatment of celecoxib combined with gemcitabine or 5-Aza showed similar efficacy against metastasis that were shown as long-term celecoxib in combination with these two drugs (Fig. 7C). We further evaluated the effect of perioperative celecoxib with 5-Aza or gemcitabine on the survival of tumor-bearing mice (Fig. 7D). The perioperative celecoxib in combination with gemcitabine significantly increased survival, and 40% of the mice survived 200 days after tumor injection (Fig. 7D, left), and stayed tumor free for almost additional 1 year. In mice bearing 6DT1 tumors, an even more aggressive breast cancer metastasis model than 4T1, the drug combinations of celecoxib, 5-Aza, and gemcitabine also significantly increased survival (Fig. 7D, right). Our data suggest that perioperative application of celecoxib, together with low-dose 5-Aza, will likely enhance the efficacy of chemotherapy in metastasis treatment.

Our studies show that DNMT3B expression is heterogeneous in primary tumors from both mouse models and human patients with breast cancer. In addition, high DNMT3B expression correlated with poor prognosis and was associated with later stages and more aggressive basal-like breast cancers. Mechanistically, we found that the DNMT3BH cells are more mesenchymal-like and have higher migratory capacity to escape from primary tumors when compared with DNMT3BL cells. In therapeutic targeting using preclinical mouse models, we found that perioperative application of celecoxib in combination with 5-Aza or chemotherapies significantly inhibits metastatic spread. Together with our previous report that DNMT3B is elevated in metastatic cancer cells and facilitates metastatic outgrowth (23), we have identified DNMT3B as a valuable therapeutic target against metastatic disease because of its critical roles in tumor dissemination and metastatic outgrowth, two critical stages of metastatic progression.

Our work supports the key role of transcriptional regulation in tumor heterogeneity. The genetic heterogeneity, such as mutations and copy-number variations, cannot fully explain the substantial subclonal complexity of tumors (3, 35, 36). The mutational landscapes of the metastases are also largely similar to that of the primary tumors (37, 38). Rather, histopathology-guided multi-region sampling demonstrates that the heterogeneity in primary tumors is not due to genomic alterations, but due to transcriptional regulation (8, 10). Importantly, preexisting intratumor heterogeneity is known to be responsible for acquired resistance during therapy (3, 39). Overcoming acquired resistance remains an obstacle to treatment success. Our studies identify DNMT3B as a key regulator in this transcription-mediated tumor heterogeneity and as a nongenetic predictive marker for high metastatic potential.

We propose that high expression of DNMT3B may confer epigenetic plasticity and allow the cells to readily undergo EMT and metastatic colonization. This notion is supported by our observation that DNMT3BH cells expressed more VIM and Snai1, and knockout of VIM diminished the metastatic advantage of DNMT3BH cells. Our findings suggest that DNMT3BH cells are more mesenchymal-like and have higher capacity to disseminate and initiate the metastatic cascade. While gene silencing by promoter hypermethylation is well established, DNMT3B can also methylate gene bodies to increase gene expression (40–42). We found DNMT3B enrichment and high DNA methylation in gene body regions of Vim as well as Snai1. In contrast, for methylation-sensitive genes in breast cancer such as Rassf1a, Grhl2, and Sfrp1, DNMT3B enrichment was found in their promoter regions, and DNMT3B knockdown increased expression of these genes (Supplementary Fig. S4C and S4D). We previously reported that DNMT3B is elevated in metastatic cancer cells and facilitates metastatic outgrowth through multiple signaling pathways (23). We now report that in the primary tumors, DNMT3BH cells show an enhanced migratory capacity through elevated VIM expression. Therefore, targeting DNMT3B could inhibit multiple steps of the metastatic cascade.

We explored the therapeutic efficacy of DNMT3B inhibition by using celecoxib and the DNA methyltransferase inhibitor 5-Aza. Celecoxib, a NSAID that once attracted huge attention as a potential drug to prevent colorectal adenoma, failed in long-term clinical trials because of an increased risk of severe cardiovascular harm (43). However, a clinical study with arthritis patients showed that moderate doses of celecoxib are comparable with other nonselective NSAIDs, such as ibuprofen and naproxen, in cardiovascular safety (34). We found that celecoxib, when applied perioperatively, showed similar effectiveness as long-term treatment in suppressing metastasis. 5-Aza, a cytosine analog that induces degradation of DNMTs, is an FDA-approved drug for myelodysplastic syndromes and chronic monomyelocytic leukemia. A study with patient-derived xenograft organoids showed anticancer activity of 5-Aza against TNBC, which was also highly correlated with DNMT3A and DNMT3B levels (44). However, the clinical benefit of 5-Aza has been inconsistent in trials for the treatment of solid tumors (45–47). In the current study, we showed markedly increased antimetastatic activity of 5-Aza when combined with celecoxib or chemotherapies, while treatment with 5-Aza as a single agent was not effective. Our data propose a novel combination of readily available and affordable drugs to target breast cancer metastasis. Drug repurposing is a fast-progressing area that holds the promise of rapid clinical impact at a lower cost than de novo drug development (48), which is particularly useful and urgently needed for TNBC, which currently lacks effective treatment options.

Perioperative celecoxib in combination with chemotherapy significantly prolonged the survival of mice after tumor removal surgery, and some mice even showed complete response. In a long-term follow-up of patients with estrogen receptor–positive breast cancer, tumor recurrence and metastasis peaked at 1 to 2 years after tumor resection, suggesting the relapse risk increased by surgery (49). Surgery has also been found to induce systemic expansion of inflammatory immune cells and caused outgrowth of dormant cells at distant sites (25). Consistent with our observations, perioperative usage of anti-inflammatory drugs improved disease-free survival in colon and breast cancers (50, 51). In an animal model of metastatic lung cancer, the preoperative inhibition of PGE2 together with the perioperative resolution of inflammation by resolvins, a family of omega-3 fatty acid derivatives, eliminated micrometastases, with 40% of mice achieving long-term survival without micrometastases (52). Our findings, together with other published evidence, provide a proof of principle for the perioperative usage of anti-inflammatory drugs to treat and to prevent metastatic progression. Altogether, our studies highlight the importance of DNMT3B in tumor dissemination and metastatic outgrowth. Because of its diverse functions at different stages of the metastatic cascade, targeting DNMT3B with celecoxib or 5-Aza in conjunction with chemotherapy could significantly suppress metastatic progression.

No disclosures were reported.

J.Y. So: Conceptualization, data curation, formal analysis, visualization, writing–original draft. H.H. Yang: Data curation. W.Y. Park: Investigation. N. Skrypek: Investigation. H. Ishii: Investigation. J.M. Chen: Investigation. M.P. Lee: Data curation, writing–review and editing. L. Yang: Conceptualization, resources, data curation, supervision, writing–original draft, writing–review and editing.

We thank Drs. Meera Murgai and Brandi Carofino for critical reading of the article and editorial suggestions. We are grateful to all expert technical supports provided by Animal Facility, Flow Cytometry Core and Confocal Microscopy Core, Center for Cancer Research, NCI. This work is supported by NCI intramural funding, to L. Yang

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.

Note: Supplementary data for this article are available at Molecular Cancer Research Online (http://mcr.aacrjournals.org/).

1.
Steeg
PS
.
Targeting metastasis
.
Nat Rev Cancer
2016
;
16
:
201
18
.
2.
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
.
3.
Marusyk
A
,
Janiszewska
M
,
Polyak
K
.
Intratumor heterogeneity: the rosetta stone of therapy resistance
.
Cancer Cell
2020
;
37
:
471
84
.
4.
McDonald
KA
,
Kawaguchi
T
,
Qi
Q
,
Peng
X
,
Asaoka
M
,
Young
J
, et al
.
Tumor heterogeneity correlates with less immune response and worse survival in breast cancer patients
.
Ann Surg Oncol
2019
;
26
:
2191
9
.
5.
Angus
L
,
Smid
M
,
Wilting
SM
,
van Riet
J
,
Van Hoeck
A
,
Nguyen
L
, et al
.
The genomic landscape of metastatic breast cancer highlights changes in mutation and signature frequencies
.
Nat Genet
2019
;
51
:
1450
8
.
6.
Yates
LR
,
Knappskog
S
,
Wedge
D
,
Farmery
JHR
,
Gonzalez
S
,
Martincorena
I
, et al
.
Genomic evolution of breast cancer metastasis and relapse
.
Cancer Cell
2017
;
32
:
169
84
.
7.
Sheffield
NC
,
Pierron
G
,
Klughammer
J
,
Datlinger
P
,
Schonegger
A
,
Schuster
M
, et al
.
DNA methylation heterogeneity defines a disease spectrum in Ewing sarcoma
.
Nat Med
2017
;
23
:
386
95
.
8.
Tavernari
D
,
Battistello
E
,
Dheilly
E
,
Petruzzella
AS
,
Mina
M
,
Sordet-Dessimoz
J
, et al
.
Non-genetic evolution drives lung adenocarcinoma spatial heterogeneity and progression
.
Cancer Discov
2021
;
11
:
1490
507
.
9.
Hinohara
K
,
Polyak
K
.
Intratumoral heterogeneity: more than just mutations
.
Trends Cell Biol
2019
;
29
:
569
79
.
10.
Sharma
A
,
Merritt
E
,
Hu
X
,
Cruz
A
,
Jiang
C
,
Sarkodie
H
, et al
.
Non-genetic intra-tumor heterogeneity is a major predictor of phenotypic heterogeneity and ongoing evolutionary dynamics in lung tumors
.
Cell Rep
2019
;
29
:
2164
74
.
11.
Silwal-Pandit
L
,
Nord
S
,
von der Lippe Gythfeldt
H
,
Moller
EK
,
Fleischer
T
,
Rodland
E
, et al
.
The longitudinal transcriptional response to neoadjuvant chemotherapy with and without bevacizumab in breast cancer
.
Clin Cancer Res
2017
;
23
:
4662
70
.
12.
Kim
C
,
Gao
R
,
Sei
E
,
Brandt
R
,
Hartman
J
,
Hatschek
T
, et al
.
Chemoresistance evolution in triple-negative breast cancer delineated by single-cell sequencing
.
Cell
2018
;
173
:
879
93
.
13.
Hoadley
KA
,
Yau
C
,
Hinoue
T
,
Wolf
DM
,
Lazar
AJ
,
Drill
E
, et al
.
Cell-of-origin patterns dominate the molecular classification of 10,000 tumors from 33 types of cancer
.
Cell
2018
;
173
:
291
304
.
14.
De Craene
B
,
Berx
G
.
Regulatory networks defining EMT during cancer initiation and progression
.
Nat Rev Cancer
2013
;
13
:
97
110
.
15.
Carmona
FJ
,
Davalos
V
,
Vidal
E
,
Gomez
A
,
Heyn
H
,
Hashimoto
Y
, et al
.
A comprehensive DNA methylation profile of epithelial-to-mesenchymal transition
.
Cancer Res
2014
;
74
:
5608
19
.
16.
Lyko
F
.
The DNA methyltransferase family: a versatile toolkit for epigenetic regulation
.
Nat Rev Genet
2018
;
19
:
81
92
.
17.
Rhee
I
,
Bachman
KE
,
Park
BH
,
Jair
KW
,
Yen
RW
,
Schuebel
KE
, et al
.
DNMT1 and DNMT3b cooperate to silence genes in human cancer cells
.
Nature
2002
;
416
:
552
6
.
18.
Girault
I
,
Tozlu
S
,
Lidereau
R
,
Bieche
I
.
Expression analysis of DNA methyltransferases 1, 3A, and 3B in sporadic breast carcinomas
.
Clin Cancer Res
2003
;
9
:
4415
22
.
19.
Nosho
K
,
Shima
K
,
Irahara
N
,
Kure
S
,
Baba
Y
,
Kirkner
GJ
, et al
.
DNMT3B expression might contribute to CpG island methylator phenotype in colorectal cancer
.
Clin Cancer Res
2009
;
15
:
3663
71
.
20.
Linhart
HG
,
Lin
H
,
Yamada
Y
,
Moran
E
,
Steine
EJ
,
Gokhale
S
, et al
.
Dnmt3b promotes tumorigenesis in vivo by gene-specific de novo methylation and transcriptional silencing
.
Genes Dev
2007
;
21
:
3110
22
.
21.
Micevic
G
,
Muthusamy
V
,
Damsky
W
,
Theodosakis
N
,
Liu
X
,
Meeth
K
, et al
.
DNMT3b modulates melanoma growth by controlling levels of mTORC2 component RICTOR
.
Cell Rep
2016
;
14
:
2180
92
.
22.
Ibrahim
ML
,
Klement
JD
,
Lu
C
,
Redd
PS
,
Xiao
W
,
Yang
D
, et al
.
Myeloid-derived suppressor cells produce IL-10 to elicit DNMT3b-dependent IRF8 silencing to promote colitis-associated colon tumorigenesis
.
Cell Rep
2018
;
25
:
3036
46
.
23.
So
JY
,
Skrypek
N
,
Yang
HH
,
Merchant
AS
,
Nelson
GW
,
Chen
WD
, et al
.
Induction of DNMT3B by PGE2 and IL6 at distant metastatic sites promotes epigenetic modification and breast cancer colonization
.
Cancer Res
2020
;
80
:
2612
27
.
24.
Tohme
S
,
Simmons
RL
,
Tsung
A
.
Surgery for cancer: a trigger for metastases
.
Cancer Res
2017
;
77
:
1548
52
.
25.
Krall
JA
,
Reinhardt
F
,
Mercury
OA
,
Pattabiraman
DR
,
Brooks
MW
,
Dougan
M
, et al
.
The systemic response to surgery triggers the outgrowth of distant immune-controlled tumors in mouse models of dormancy
.
Sci Transl Med
2018
;
10
:
eaan3464
.
26.
Karagiannis
GS
,
Pastoriza
JM
,
Wang
Y
,
Harney
AS
,
Entenberg
D
,
Pignatelli
J
, et al
.
Neoadjuvant chemotherapy induces breast cancer metastasis through a TMEM-mediated mechanism
.
Sci Transl Med
2017
;
9
:
eaan0026
.
27.
Keklikoglou
I
,
Cianciaruso
C
,
Guc
E
,
Squadrito
ML
,
Spring
LM
,
Tazzyman
S
, et al
.
Chemotherapy elicits pro-metastatic extracellular vesicles in breast cancer models
.
Nat Cell Biol
2019
;
21
:
190
202
.
28.
Loi
S
,
Haibe-Kains
B
,
Desmedt
C
,
Lallemand
F
,
Tutt
AM
,
Gillet
C
, et al
.
Definition of clinically distinct molecular subtypes in estrogen receptor-positive breast carcinomas through genomic grade
.
J Clin Oncol
2007
;
25
:
1239
46
.
29.
Yau
C
,
Esserman
L
,
Moore
DH
,
Waldman
F
,
Sninsky
J
,
Benz
CC
.
A multigene predictor of metastatic outcome in early stage hormone receptor-negative and triple-negative breast cancer
.
Breast Cancer Res
2010
;
12
:
R85
.
30.
Curtis
C
,
Shah
SP
,
Chin
SF
,
Turashvili
G
,
Rueda
OM
,
Dunning
MJ
, et al
.
The genomic and transcriptomic architecture of 2,000 breast tumours reveals novel subgroups
.
Nature
2012
;
486
:
346
52
.
31.
Bos
PD
,
Zhang
XH
,
Nadal
C
,
Shu
W
,
Gomis
RR
,
Nguyen
DX
, et al
.
Genes that mediate breast cancer metastasis to the brain
.
Nature
2009
;
459
:
1005
9
.
32.
Hatzis
C
,
Pusztai
L
,
Valero
V
,
Booser
DJ
,
Esserman
L
,
Lluch
A
, et al
.
A genomic predictor of response and survival following taxane-anthracycline chemotherapy for invasive breast cancer
.
JAMA
2011
;
305
:
1873
81
.
33.
Figueroa
JD
,
Yang
H
,
Garcia-Closas
M
,
Davis
S
,
Meltzer
P
,
Lissowska
J
, et al
.
Integrated analysis of DNA methylation, immunohistochemistry and mRNA expression, data identifies a methylation expression index (MEI) robustly associated with survival of ER-positive breast cancer patients
.
Breast Cancer Res Treat
2015
;
150
:
457
66
.
34.
Nissen
SE
,
Yeomans
ND
,
Solomon
DH
,
Luscher
TF
,
Libby
P
,
Husni
ME
, et al
.
Cardiovascular safety of celecoxib, naproxen, or ibuprofen for arthritis
.
N Engl J Med
2016
;
375
:
2519
29
.
35.
Andor
N
,
Graham
TA
,
Jansen
M
,
Xia
LC
,
Aktipis
CA
,
Petritsch
C
, et al
.
Pan-cancer analysis of the extent and consequences of intratumor heterogeneity
.
Nat Med
2016
;
22
:
105
13
.
36.
Berglund
E
,
Maaskola
J
,
Schultz
N
,
Friedrich
S
,
Marklund
M
,
Bergenstrahle
J
, et al
.
Spatial maps of prostate cancer transcriptomes reveal an unexplored landscape of heterogeneity
.
Nat Commun
2018
;
9
:
2419
.
37.
Priestley
P
,
Baber
J
,
Lolkema
MP
,
Steeghs
N
,
de Bruijn
E
,
Shale
C
, et al
.
Pan-cancer whole-genome analyses of metastatic solid tumours
.
Nature
2019
;
575
:
210
6
.
38.
Li
C
,
Bonazzoli
E
,
Bellone
S
,
Choi
J
,
Dong
W
,
Menderes
G
, et al
.
Mutational landscape of primary, metastatic, and recurrent ovarian cancer reveals c-MYC gains as potential target for BET inhibitors
.
Proc Natl Acad Sci U S A
2019
;
116
:
619
24
.
39.
Easwaran
H
,
Tsai
HC
,
Baylin
SB
.
Cancer epigenetics: tumor heterogeneity, plasticity of stem-like states, and drug resistance
.
Mol Cell
2014
;
54
:
716
27
.
40.
Yang
X
,
Han
H
,
De Carvalho
DD
,
Lay
FD
,
Jones
PA
,
Liang
G
.
Gene body methylation can alter gene expression and is a therapeutic target in cancer
.
Cancer Cell
2014
;
26
:
577
90
.
41.
Duymich
CE
,
Charlet
J
,
Yang
X
,
Jones
PA
,
Liang
G
.
DNMT3B isoforms without catalytic activity stimulate gene body methylation as accessory proteins in somatic cells
.
Nat Commun
2016
;
7
:
11453
.
42.
Baubec
T
,
Colombo
DF
,
Wirbelauer
C
,
Schmidt
J
,
Burger
L
,
Krebs
AR
, et al
.
Genomic profiling of DNA methyltransferases reveals a role for DNMT3B in genic methylation
.
Nature
2015
;
520
:
243
7
.
43.
Solomon
SD
,
McMurray
JJ
,
Pfeffer
MA
,
Wittes
J
,
Fowler
R
,
Finn
P
, et al
.
Cardiovascular risk associated with celecoxib in a clinical trial for colorectal adenoma prevention
.
N Engl J Med
2005
;
352
:
1071
80
.
44.
Yu
J
,
Qin
B
,
Moyer
AM
,
Nowsheen
S
,
Liu
T
,
Qin
S
, et al
.
DNA methyltransferase expression in triple-negative breast cancer predicts sensitivity to decitabine
.
J Clin Invest
2018
;
128
:
2376
88
.
45.
Appleton
K
,
Mackay
HJ
,
Judson
I
,
Plumb
JA
,
McCormick
C
,
Strathdee
G
, et al
.
Phase I and pharmacodynamic trial of the DNA methyltransferase inhibitor decitabine and carboplatin in solid tumors
.
J Clin Oncol
2007
;
25
:
4603
9
.
46.
Jones
PA
,
Issa
JP
,
Baylin
S
.
Targeting the cancer epigenome for therapy
.
Nat Rev Genet
2016
;
17
:
630
41
.
47.
Oza
AM
,
Matulonis
UA
,
Alvarez Secord
A
,
Nemunaitis
J
,
Roman
LD
,
Blagden
SP
, et al
.
A randomized phase II trial of epigenetic priming with guadecitabine and carboplatin in platinum-resistant, recurrent ovarian cancer
.
Clin Cancer Res
2020
;
26
:
1009
16
.
48.
Corsello
SM
,
Bittker
JA
,
Liu
Z
,
Gould
J
,
McCarren
P
,
Hirschman
JE
, et al
.
The Drug Repurposing Hub: a next-generation drug library and information resource
.
Nat Med
2017
;
23
:
405
8
.
49.
Colleoni
M
,
Sun
Z
,
Price
KN
,
Karlsson
P
,
Forbes
JF
,
Thurlimann
B
, et al
.
Annual hazard rates of recurrence for breast cancer during 24 years of follow-up: results from the international breast cancer study group trials I to V
.
J Clin Oncol
2016
;
34
:
927
35
.
50.
Forget
P
,
Bentin
C
,
Machiels
JP
,
Berliere
M
,
Coulie
PG
,
De Kock
M
.
Intraoperative use of ketorolac or diclofenac is associated with improved disease-free survival and overall survival in conservative breast cancer surgery
.
Br J Anaesth
2014
;
113
:
i82
7
.
51.
Schack
A
,
Fransgaard
T
,
Klein
MF
,
Gogenur
I
.
Perioperative use of nonsteroidal anti-inflammatory drugs decreases the risk of recurrence of cancer after colorectal resection: a cohort study based on prospective data
.
Ann Surg Oncol
2019
;
26
:
3826
37
.
52.
Panigrahy
D
,
Gartung
A
,
Yang
J
,
Yang
H
,
Gilligan
MM
,
Sulciner
ML
, et al
.
Preoperative stimulation of resolution and inflammation blockade eradicates micrometastases
.
J Clin Invest
2019
;
129
:
2964
79
.

Supplementary data