Current cancer treatments are largely based on the genetic characterization of primary tumors and are ineffective for metastatic disease. Here we report that DNA methyltransferase 3B (DNMT3B) is induced at distant metastatic sites and mediates epigenetic reprogramming of metastatic tumor cells. Multiomics analysis and spontaneous metastatic mouse models revealed that DNMT3B alters multiple pathways including STAT3, NFκB, PI3K/Akt, β-catenin, and Notch signaling, which are critical for cancer cell survival, apoptosis, proliferation, invasion, and colonization. PGE2 and IL6 were identified as critical inflammatory mediators in DNMT3B induction. DNMT3B expression levels positively correlated with human metastatic progression. Targeting IL6 or COX-2 reduced DNMT3B induction and improved chemo or PD1 therapy. We propose a novel mechanism linking the metastatic microenvironment with epigenetic alterations that occur at distant sites. These results caution against the “Achilles heel” in cancer therapies based on primary tumor characterization and suggests targeting DNMT3B induction as new option for treating metastatic disease.

Significance:

These findings reveal that DNMT3B epigenetically regulates multiple pro-oncogenic signaling pathways via the inflammatory microenvironment at distant sites, cautioning the clinical approach basing current therapies on genetic characterization of primary tumors.

Metastasis accounts for 90% of cancer mortality. Current therapies are largely based on genotype variations and oncogene activation in primary tumor biopsies, however are ineffective in treating metastatic diseases (1). Conventional chemotherapies and targeted therapies, which are widely used in standard care, have a negative impact on patient quality of life and may induce additional mutations that foster cancer and metastasis (1). Low response rate, therapy relapse and resistance are challenges in immunotherapies that have yet to be overcome, despite recent progress and optimism (2, 3). In the metastatic process, tumor cells undergo invasion, intravasation, extravasation, as well as survival and proliferation in the foreign organ microenvironment to form clinically overt metastases. In particular, metastatic colonization at the distant organ is a time-limiting step. Understanding the mechanism and adaptation of metastatic tumor cells and their interaction with distant-organ microenvironment should uncover novel therapeutic targets and rational combination treatment.

Tumor cells acquire metastatic capacity through several mechanisms. In the Darwinian-like somatic evolution hypothesis, the metastatic clones and subclones in the primary tumor continue to evolve during progression and treatment, resulting in substantial genetic divergence, such as mutations in the estrogen receptor ligand-binding domain (4). In addition, acquired driver mutations have been found in distant metastases that are not seen in the primary tumor in patients with breast cancer as well as in patients with treatment-naïve pancreatic cancer (5). Indeed, significant genomic heterogeneity was observed in disseminated cancer cells after “curative” surgery (6–8). Genomic alterations were also found in the overt metastasis of breast cancer when compared with disseminated tumor cells (9, 10). These studies indicate that the acquisition of genetic changes can occur outside of the primary tumors in the distant organ sites.

Epigenetic alterations are another important mechanism for cancer cells to acquire key traits of full malignancy (11). DNA methylation is one of the most consistent epigenetic changes in human cancers (12). In fact, DNA methylation pattern is one of the strongest molecular classifications in clustering cells of tumorigenic origin in the Pan-Cancer Atlas (13) and is associated with metastatic risk (14–16). Of the three major DNA methyltransferases, DNMT3B mediates de novo gene methylation and cooperates with DNMT1 to silence genes (17). DNMT3B is upregulated in a number of human cancers and is associated with poor prognosis in human breast cancers (18, 19). DNA methylation is an epigenetic mechanism that promotes cancer cell survival (20) and could be induced by nongenetic stimuli, such as oxidative stress and proinflammatory molecules (21–23). However, whether regulation of metastatic tumor methylomes occurs at the distant sites and if so, what are the underlying molecular mechanisms and biological consequences remain unknown.

The premetastatic and metastatic organ sites are modified by extracellular vesicles and host-derived cells, which provide an inflammatory and immune suppressive microenvironment that is permissive for metastatic spread (24). It is not clear whether this inflammatory microenvironment at distant sites mediates epigenetic alterations and facilitates metastatic colonization and outgrowth. IL6 and PGE2 are well-known proinflammatory factors in tumor progression and strongly suppress host-antitumor immunity (25, 26). IL6 and PGE2 are critical in premetastatic niche formation through activation of lymphatic endothelial cells and functional alteration of dendritic cells thus promote metastasis (27, 28). The IL6/JAK/STAT3 pathway is hyperactivated in many types of cancer and is generally associated with poor prognosis (25). Recently, the cross-talk between IL6 and PD1/PD-L1 was observed in the tumor microenvironment and may constitute a rational immunosuppressive target for overcoming the narrow therapeutic window of anti-PD-1/PD-L1 therapy (29).

In this study, we demonstrate that IL6 and PGE2 induce DNMT3B in metastatic cancer cells at the distant sites. DNMT3B alters methylation and transcription of genes and pathways that are important in cancer cell survival, apoptosis, proliferation, invasion, and colonization including STAT3, NFκB, PI3K/Akt, β-catenin, as well as Notch signaling. Human datasets also showed the increased DNMT3B levels in the distant metastases. Targeting DNMT3B induction, in combination with chemo or PD1 therapy, improves the treatment efficacy in preclinical mouse models. Our study provides proof of concept and molecular mechanisms for DNMT3B induction and its mediated epigenetic reprogramming by the inflammatory microenvironment at the distant metastatic site. We anticipate the mechanistic insights and preclinical evidence from the current study will aid rational design and clinical development of combination treatment with chemo- or immunotherapy for the patients with metastatic disease.

Cell lines and mice

Murine mammary tumor cell lines 4T1, GFP+-4T1, EMT6, and 6DT1 were maintained in 10% FBS DMEM (Life Technologies) and in 37°C incubator with 5% CO2. All cell lines are from ATCC and are Mycoplasma negative with MycoAlert Mycoplasma Detection Kit (Lonza). The cells were used within two months from thawing. 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.

Plasmid constructs, transduction, and transfection

For DNMT3B knockdown or overexpression, a psiLv-U6 empty vector or vector encoding a DNMT3B shRNA (GeneCopoeia) were stably transduced into 4T1 and EMT6 cells. ORF-cDNA of mouse Dnmt3b (GeneCopoeia) were cloned into the pLenti4/TO/V5-Dest vector with the Tet-inducible system using Gateway technology (Thermo Fisher Scientific) and was stably transduced into 4T1 cells. DNMT3B expression was induced by doxycycline for cells in vitro (1 μg/mL for 72 hours) and for mice in vivo (1 mg/mL in water). In addition, active Stat3 (Stat3-C), active β-Catenin, or Notch1 intracellular domain, and active Akt1-expressing vectors were transfected into 4T1-DNMT3B KD or control cells by Lipofectamine LTX system (Thermo Fisher Scientific).

Mouse models of tumor metastasis and treatment

Eight-week-old female BALB/c, nude, or FVB/N mice (Charles River) received 2.5–5 × 105 tumor cell injection into the mammary fat pad (MFP) #2, or tail vein (TVI). In some experiments, TVI was performed in mice that received MFP injection 2 weeks earlier. Mice were sacrificed on day 28–35, and the size/weight of primary tumors or the recurrent tumors, as well as the number of metastatic nodules were evaluated. For MMTV-PyMT model, the primary tumors and lung metastasis nodules were examined from 9-month-old female mice.

For IL6 or PGE2 effect on metastasis at distant organ site, mice received wt or DNMT3B KD 4T1 TVI. Recombinant murine IL6 (2.5 μg/kg bodyweight, PeproTech) or/and 16,16-dimethyl PGE2 (2.5 mg/kg bodyweight, Cayman) were injected intraperitoneally daily.

For drug treatment, the 4T1 primary tumors were resected and the mice were treated with doxorubicin (2 mg/kg bodyweight, i.v. once a week), with meloxicam (2 mg/kg bodyweight, i.p. daily) or etodolac (10 mg/kg bodyweight, i.p. daily). For immunotherapy, the mice were intraperitoneally. injected with a PD-1 (5 mg/kg, BE0146) or IL6 neutralizing antibody (10 mg/kg, BE0046; BioXCell) or the two in combination. Tumor phenotype was evaluated as indicated above. RNA and DNA were extracted from the GFP+ metastatic nodules for molecular characterization.

Tumor cell sorting from primary tumor tissues and lung metastasis nodules

The GFP+ tumor cells from single-cell suspensions were sorted by FACS and were subjected to total RNA and genomic DNA (gDNA) extraction.

ImageStream for DNMT3B expression levels in single tumor cells

Sorted single cells were fixed in 4% paraformaldehyde, and incubated with DNMT3B (1:100, Abcam), and the fluorescence intensity of DNMT3B was measured with Amnis ImageStream MkII (Luminex Corporation). The images were generated and analyzed by INSPIRE software (Luminex Corporation).

DNMT3B chromatin immunoprecipitation sequencing

DNMT3B chromatin immunoprecipitation (ChIP; antibody ab2851, Abcam) was performed using Magna ChIP A/G kit (Millipore). The sequence (NextSeq 500 System, Illumina) was aligned on the mouse reference genome (mm10) by bowtie2 v2.1.0 and peaks were detected with MACS v1.4 (Narrow peaks) and SICER (broad peaks), with the P value set as > 0.05 and log2 fold-change < 1 compared with input. Data is available in GEO database (GSE146010).

RNA sequencing

Total RNA was extracted (TRIzol, Invitrogen), the library was prepared (TruSeq RNA kit, Illumina), and sequenced on a HiSeq 2500 System (Illumina). FastQC and STAR v2.4.0a alignment to the mm10 reference genome were performed to generate BAM files. The read counts were then normalized by DESeq2 v3.1 using a scaling factor method based on median ratio. The differential expression was determined as P > 0.05 and log2 fold-change < 1 compared with control samples. The RNA-seq data is available in the GEO database (GSE146011).

Agilent-based target-enriched bisulfite methylation sequencing

The probes capturing the top 20% of SICER were designed (Agilent SureDesign: https://earray.chem.agilent.com/suredesign/home.htm), with the target enrichment of gDNA (SureSelectXT Methyl-seq Kit, Agilent Technologies), and bisulfite conversion (EZ-DNA Methylation-Gold Kit, Zymo Research). The libraries were sequenced by a HiSeq 2500 System (Illumina). The differentially methylated CpG sites were determined by the cut-offs Beta-Diff > 0.1 and FDR < 0.05. The Methylation-seq data is available in the GEO database (GSE146012).

For locus-specific DNA methylation analysis, genomic DNA with bisulfite conversion was amplified by PCR with locus-specific primers (Takara EpiTaq HS kit, Takara Bio Inc.). The PCR products with correct size were cloned into pCR2.1 vector system (TA Cloning Kit, Invitrogen). Sanger sequencing was performed with 10 clones for each group.

ChIP-PCR and RT-qPCR

DNMT3B ChIP-PCR was performed using Magana ChIP A/G procedure. RNA was extracted (RNeasy Mini Kit, Qiagen) and cDNA was synthesized (High-Capacity cDNA Reverse Transcription Kits, Applied Biosystems), with gene expression determined using SYBR Green–based qPCR.

Rescue experiments for DNMT3B-targeted pathways

Proliferation

Cells were harvested every 24 hours and counted in triplicates using Cellometer (Nexcelom Bioscience).

Apoptosis

Cells were labeled with 7AAD and Annexin-V (BD Biosciences), and were analyzed by BD FACSCanto II flow cytometry (Becton Dickinson).

Wound-healing assay

Cells with 100% confluence were scratched to generate wound at the center. The recovery from the wound was monitored by IncuCyte Live cell analysis system (Essen BioScience, Inc.).

Soft agar assay

Cells were plated in 0.3% soft agar (Sigma) with 0.6% soft agar as base layer. The number of colonies > 150 μm were counted after 12 days using FluorChem HD2 system (ProteinSimple).

Sphere formation in 3D Matrigel

4T1 single cells were cultured in medium containing 5% Matrigel (Thermo Fisher Scientific). The number and size of tumor spheres were evaluated by phase contrast EVOS imaging system.

Western blotting

Antibodies against DNMT3B (ab79822, Abcam), DNMT1 (NB100-56519, Novus Biologicals), Fzd2 (sc-74019), COX-2 (sc-1745), HP1 (sc-28735), and β-actin (sc-69879), as well as DNMT3A, Nos2, p-STAT3, STAT3, p-Akt, Akt, p-p65 NFκB, p65 NFκB, β-catenin, active β-catenin, cleaved caspase-3, Notch1, NICD1, SMAD2, SMAD3, Histone-H3, p-ERK (all from Cell Signaling Technology) were diluted at 1/500 to 1:5,000. Quantification was done using ImageJ.

PGE2 ELISA, cytokine antibody array, Luminex multiplex cytokine array

Protein extractions from the lungs of Celecoxib-treated mice were examined by PGE2 Parameter Assay Kit (R&D Systems), or by cytokine protein antibody array (Raybiotech). In addition, inflammatory cytokine expression was measured using Luminex Mouse Cytokine 20-plex Kit (Life Technologies) and Precellys Lysing Kit system (Bertin Corp).

DNMT3B induction by PGE2 and IL6

The 4T1 cells were incubated for 24 hours in tissue supernatants (from lungs of naïve or tumor-bearing mice) that were preincubated with IL6 or TGFβ1 neutralizing antibody (BioXCell). The cells were then treated with AH-6809, an EP-receptor antagonist (Cayman Chemical), or a STAT3 inhibitor, or sometimes with NFκB or SMAD2 knockdown. The MDSCs used in 4T1 coculture were isolated using magnetically activated cell sorting (MACS, Miltenyi Biotec). Protein extractions from the 4T1 cells were examined by Western blotting.

Immunofluorescence, IHC, and TUNEL assay

Frozen lung sections from mouse models and paraffin sections from breast cancer patients (n = 7) with paired metastatic and primary tumor tissues were incubated with DNMT3B (1:100, ab2851, Abcam) or GFP (1:200, ab13970, Abcam, mouse only) or Phospho-H3 (1:100, Cell Signaling Technology, mouse only) antibodies overnight at 4°C. Fluorescence-tagged secondary antibody and DAPI were utilized for visualization. Images were obtained (Carl Zeiss) and DNMT3B intensities were quantified by ImageJ. TUNEL was performed per manufactory protocol (Roche Applied Science).

For circulating tumor cell (CTC), blood was collected through cardiac puncture. The GFP+-CTCs were sorted by FACS Aria II (Becton Dickinson), were spun onto slides by cytospin, followed by immunofluorescence staining.

Human correlation of DNMT3B expression

Publicly available datasets from Weigelt (30), Tamura (31), Haqq (32), and Wuttig (33) were used to generate DNMT3B correlation comparing metastasis nodules with primary tumor tissues. In addition, paired primary and metastatic tissues from patients with breast cancer, verified by expert pathologists, were stained for DNMT3B expression and quantified by ImageJ. The sections of formalin-fixed paraffin-embedded tissues corresponding to hematoxylin and eosin–stained slides were used to ensure that at least 20% of the retrieved cells were neoplastic.

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 < 0.05.

DNMT3B is increased in metastatic nodules and is critical in metastatic colonization

An increased expression of DNMT3B, but not DNMT3A or DNMT1, was found in metastatic nodules compared with the primary tumors from mice that received mammary fat pad (MFP) injection of 4T1 (Fig. 1A and B), as well as EMT6 (Fig. 1C) and 6DT1 tumor cells (Fig. 1D). In addition to these orthotopic and spontaneous metastasis models for mammary tumors, MMTV-PyMT mice, a transgenic mammary tumor model with spontaneous metastasis also showed increased DNMT3B in metastatic nodules (Fig. 1E). Importantly, knockdown (KD) of DNMT3B significantly decreased lung metastasis (Fig. 1F, 4T1 model; Fig. 1G, EMT6 model), whereas DNMT3B induction using a tetracycline (Tet)-controlled transcriptional activation increased metastasis in mice bearing 4T1 tumors (Fig. 1H). Taken together, these data support the critical roles of DNMT3B in breast cancer metastasis.

Figure 1.

Induction of DNMT3B and its effects on metastatic colonization. A, DNMT3B, DNMT3A, and DNMT1 Western, protein extraction from 4T1 primary tumors and lung nodules. B, DNMT3B IHC of 4T1 lung nodules and primary tumors. C and D, DNMT3B, DNMT3A, and DNMT1 Western blots of primary tumors and lung nodules from EMT6 (C) and 6DT1 (D) orthotopic mammary tumor models. E, DNMT3B Western blot of primary tumors and lung nodules for MMTV-PyMT transgenic mice. F and G, The effect of DNMT3B KD on metastasis in 4T1 (F) or EMT6 (G) experimental metastasis models. Western blot for DNMT3B KD (top left), and metastatic nodule counts from mice injected with DNMT3B KD (n = 8) or control cells (n = 8; right). Representative lung images are shown in bottom left panels. H, Western blot for DNMT3B overexpression by doxycycline (Dox) in doxycycline-inducible DNMT3B expression in 4T1 cells (top), and metastatic nodule counts from mice bearing 4T1 tumors overexpressing DNMT3B (n = 6; bottom). I, DNMT3B Western for subclonal 4T1 and EMT6 cell lines with high or low DNMT3B expression. For A, C, D, and I, each lane is from an individual mouse. The quantification of DNMT3B expression by band density is listed for each blot as shown, and the values were normalized to β-actin. Data are presented as mean ± SD. *, P < 0.05; **, P < 0.01.

Figure 1.

Induction of DNMT3B and its effects on metastatic colonization. A, DNMT3B, DNMT3A, and DNMT1 Western, protein extraction from 4T1 primary tumors and lung nodules. B, DNMT3B IHC of 4T1 lung nodules and primary tumors. C and D, DNMT3B, DNMT3A, and DNMT1 Western blots of primary tumors and lung nodules from EMT6 (C) and 6DT1 (D) orthotopic mammary tumor models. E, DNMT3B Western blot of primary tumors and lung nodules for MMTV-PyMT transgenic mice. F and G, The effect of DNMT3B KD on metastasis in 4T1 (F) or EMT6 (G) experimental metastasis models. Western blot for DNMT3B KD (top left), and metastatic nodule counts from mice injected with DNMT3B KD (n = 8) or control cells (n = 8; right). Representative lung images are shown in bottom left panels. H, Western blot for DNMT3B overexpression by doxycycline (Dox) in doxycycline-inducible DNMT3B expression in 4T1 cells (top), and metastatic nodule counts from mice bearing 4T1 tumors overexpressing DNMT3B (n = 6; bottom). I, DNMT3B Western for subclonal 4T1 and EMT6 cell lines with high or low DNMT3B expression. For A, C, D, and I, each lane is from an individual mouse. The quantification of DNMT3B expression by band density is listed for each blot as shown, and the values were normalized to β-actin. Data are presented as mean ± SD. *, P < 0.05; **, P < 0.01.

Close modal

The increased DNMT3B in metastatic nodules could derive from the high DNMT3B subclonal cells in the primary tumor tissues, or as a result of DNMT3B induction at the distant sites. The genomic and phenotypic progression outside the primary tumor is a void area of study lacking experimental evidence, which implies new challenges and opportunities for diagnosis and therapies aimed at metastasis. We thus chose to investigate this possibility that the increase in DNMT3B arose outside the primary tumor. First, subclonal cell lines expressing high or low levels of DNMT3B were established from single cell culture for both 4T1 (4T1-DNMT3BH or 4T1-DNMT3BL) and EMT6 (EMT6-DNMT3BH or EMT6-DNMT3BL) tumor cells (Fig. 1I, top; Supplementary Fig. S1A). Tail vein injection (TVI) of the DNMT3BH tumor cells produced more and larger metastatic colonies compared with those from DNMT3BL tumor cells (Supplementary Fig. S1B, 4T1; Supplementary Fig. S1C, EMT6). DNMT3BH and DNMT3BL clonal cells showed increased DNMT3B levels in the metastatic nodules in comparison with in vitro culture cells (Fig. 1I, bottom). Surprisingly, the metastatic nodules derived from DNMT3BL tumor cells showed high DNMT3B levels comparable with those derived from DNMT3BH clones, in both 4T1 and EMT6 models (Fig. 1I, bottom; Supplementary Fig. S1D). These data indicate the possibility that DNMT3BL tumor cells may acquire high DNMT3B expression after entering circulation.

Time course showing DNMT3B induction at the distant sites

To investigate the time and location of DNMT3B induction, circulating tumor cells (CTC) were sorted from blood (Supplementary Fig. S1E, FACS gating) at different time points after MFP injection of GFP+ 4T1 cells (Fig. 2A, top). The CTCs over 2–4 weeks showed no clear increase in DNMT3B expression, while the tumor cells from metastatic nodules showed an increased DNMT3B (Fig. 2A, bottom). These data again suggest DNMT3B induction at the distant sites. Next, the GFP+ 4T1 cells were injected through the tail vein of mice preconditioned with MFP injection of the GFP 4T1 cells (without GFP). The DNMT3B levels in the GFP+ 4T1 cells from the lungs at different time points were measured by image stream (Fig. 2B, top). DNMT3B was significantly increased 72 hours after TVI, and with the highest increase of DNMT3B in macrometastases (>100 cells per foci) at 3 weeks (Fig. 2B, bottom left). This result was also observed in EMT6 mouse model (Fig. 2B, bottom right).

Figure 2.

Time course of DNMT3B induction at the distant sites. A, Schematic experimental design (top) and DNMT3B immunofluorescence staining in sorted CTCs at different time points, and sorted tumor cells from 4T1 metastatic nodules were used as positive control. Bottom, representative photos and quantitated DNMT3B levels. B, Top, schematic experimental design for TVI in mice with premetastatic niche; image stream of DNMT3B expression levels in TVI-injected GFP+ 4T1 and GFP+ EMT6 cells from the lungs at different time points. Representative photos and quantitated DNMT3B levels of GFP+ 4T1 cells (bottom left) and GFP+ EMT6 cells (bottom right). C, DNMT3B immunofluorescence staining of metastatic loci at different time points after TVI of DNMT3BL GFP+ 4T1 cells in mice with or without premetastatic niche. Schematic experimental design (left); representative images (middle); number of GFP+ 4T1 cells with high DNMT3B expression (n = 6 sections for each sample; right). D, DNMT3B expression of in vitro clonal cell culture over passages. Schematic experimental design (left); qRT-PCR of DNMT3B expression in clonal cell culture from metastatic nodules compared with those from primary tumor and 4T1 in culture (middle); qRT-PCR of DNMT3B expression in clonal cell culture treated with or without D35 lung extract from mice bearing 4T1 MFP tumors (right). Number of passages in culture and treatment conditions are indicated. Data are presented as mean ± SD. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Figure 2.

Time course of DNMT3B induction at the distant sites. A, Schematic experimental design (top) and DNMT3B immunofluorescence staining in sorted CTCs at different time points, and sorted tumor cells from 4T1 metastatic nodules were used as positive control. Bottom, representative photos and quantitated DNMT3B levels. B, Top, schematic experimental design for TVI in mice with premetastatic niche; image stream of DNMT3B expression levels in TVI-injected GFP+ 4T1 and GFP+ EMT6 cells from the lungs at different time points. Representative photos and quantitated DNMT3B levels of GFP+ 4T1 cells (bottom left) and GFP+ EMT6 cells (bottom right). C, DNMT3B immunofluorescence staining of metastatic loci at different time points after TVI of DNMT3BL GFP+ 4T1 cells in mice with or without premetastatic niche. Schematic experimental design (left); representative images (middle); number of GFP+ 4T1 cells with high DNMT3B expression (n = 6 sections for each sample; right). D, DNMT3B expression of in vitro clonal cell culture over passages. Schematic experimental design (left); qRT-PCR of DNMT3B expression in clonal cell culture from metastatic nodules compared with those from primary tumor and 4T1 in culture (middle); qRT-PCR of DNMT3B expression in clonal cell culture treated with or without D35 lung extract from mice bearing 4T1 MFP tumors (right). Number of passages in culture and treatment conditions are indicated. Data are presented as mean ± SD. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Close modal

To further investigate roles of the inflammatory microenvironment at the distant sites in the DNMT3B induction, a TVI of 5 × 104 GFP+-DNMT3BL 4T1 cells (the number of cells that did not induce an inflammatory lung microenvironment, Supplementary Fig. S1F) was performed in mice with or without a premetastatic niche (mice bearing MFP tumors for 12–14 days; Fig. 2C, schematic, left panels). There were significantly more GFP+ metastatic nodules in the mice with the premetastatic niche than in the mice that only received a TVI of tumor cells (Supplementary Fig. S1G). The number of GFP+ cells with high DNMT3B level was significantly increased over time in the mice with the premetastatic niches but not in the mice without the premetastatic niche (Fig. 2C), demonstrating the effect of inflammatory microenvironment on DNMT3B induction. This environmental induction of DNMT3B was further supported by the clonal cell lines derived from single cells of metastatic nodules, which showed higher DNMT3B expression compared with those from primary tumors. This difference was diminished after several passages of in vitro culture (Fig. 2D, middle). However, the difference was restored upon the treatment of the conditioned media from the lungs of tumor-bearing mice (Fig. 2D, right). Together, our data suggest that DNMT3B induction in metastatic cancer cells occurred at the distant sites and resulted from tumor-associated inflammation.

Alterations in gene methylation and transcription mediated by DNMT3B in metastatic cancer cells that bypass the primary tumor

We next investigated changes in methylation and gene expression mediated by DNMT3B by comparing the GFP+ 4T1 cells sorted from metastatic nodules (TVI-mets) with those sorted from primary tumors (Tumor) or spontaneous metastatic nodules (MFP-mets) (Fig. 3A). We performed: (i) DNMT3B ChIP-seq on sorted cells from metastatic nodules as they express high DNMT3B (Supplementary Fig. S2A for quality control); (ii) targeted methylation-seq of sorted GFP+ 4T1 cells from TVI-mets, MFP-mets, primary tumors, as well as cultured 4T1 cells. The capture probe design was based on genes with top 20% DNMT3B enrichment (Supplementary Fig. S2B for quality control); (iii) RNA-seq of the above samples (Supplementary Fig. S2C for quality control). The intersection of these datasets allows the identification of differentially methylated and expressed genes targeted by DNMT3B comparing TVI-mets, MFP-mets, and primary tumor tissues (Fig. 3B).

Figure 3.

DNMT3B alters methylation and gene expression in metastatic cancer cells. A, Schematic for metastatic GFP+ 4T1 cells that bypass the primary tumor (top), or from MFP-mets and primary tumors (bottom). B, Analysis strategies for DNMT3B-ChIP-seq, target-enriched bisulfite methylation-seq, and RNA-seq. C, Heatmap of differentially methylated CpG sites comparing TVI-mets with primary tumors as well as MFP-mets. Cutoff: Beta-Diff > 0.1 and FDR < 0.05. β value 1 corresponding to 100% methylation. D, Heatmap of differentially expressed genes comparing TVI-mets with primary tumors as well as MFP-mets. E, Janus plot for the differential promoter methylation and gene expression comparing TVI-mets vs. primary tumor. Example genes with increased promoter methylation and decreased expression are highlighted with colored dots (Inpp5d, red; Cldn3, green; Lfng, blue). The top portion (above the yellow line) of the plot shows hypermethylated CpG sites (−log10FDR), and the lower part of the plot (below the yellow line) shows log-fold change of gene expression (log2FC). Heatmaps show methylation β values of individual CpG in the promoter regions comparing TVI-mets with primary tumor. Bar graphs below show gene expression levels in TVI-mets and primary tumors. F, Genomic browser view of DNA methylation and DNMT3B-ChIP for representative genes with differential methylation and expression. Green, downregulated genes (Inpp5d, Cldn3, Lfng); orange, upregulated genes (Jag2 and Nos2). Peaks in red indicate hypermethylation and the ones in blue indicate hypomethylation comparing TVI-met/primary tumor, MFP-mets/primary tumor, as well as DNMT3B KO vs. WT. G, Top, Venn diagrams of differentially methylated and expressed genes. Bottom, heatmap for differentially methylated and expressed genes comparing TVI-mets with primary tumor as well as MFP-mets. Mediators for key pathways are indicated on the right side.

Figure 3.

DNMT3B alters methylation and gene expression in metastatic cancer cells. A, Schematic for metastatic GFP+ 4T1 cells that bypass the primary tumor (top), or from MFP-mets and primary tumors (bottom). B, Analysis strategies for DNMT3B-ChIP-seq, target-enriched bisulfite methylation-seq, and RNA-seq. C, Heatmap of differentially methylated CpG sites comparing TVI-mets with primary tumors as well as MFP-mets. Cutoff: Beta-Diff > 0.1 and FDR < 0.05. β value 1 corresponding to 100% methylation. D, Heatmap of differentially expressed genes comparing TVI-mets with primary tumors as well as MFP-mets. E, Janus plot for the differential promoter methylation and gene expression comparing TVI-mets vs. primary tumor. Example genes with increased promoter methylation and decreased expression are highlighted with colored dots (Inpp5d, red; Cldn3, green; Lfng, blue). The top portion (above the yellow line) of the plot shows hypermethylated CpG sites (−log10FDR), and the lower part of the plot (below the yellow line) shows log-fold change of gene expression (log2FC). Heatmaps show methylation β values of individual CpG in the promoter regions comparing TVI-mets with primary tumor. Bar graphs below show gene expression levels in TVI-mets and primary tumors. F, Genomic browser view of DNA methylation and DNMT3B-ChIP for representative genes with differential methylation and expression. Green, downregulated genes (Inpp5d, Cldn3, Lfng); orange, upregulated genes (Jag2 and Nos2). Peaks in red indicate hypermethylation and the ones in blue indicate hypomethylation comparing TVI-met/primary tumor, MFP-mets/primary tumor, as well as DNMT3B KO vs. WT. G, Top, Venn diagrams of differentially methylated and expressed genes. Bottom, heatmap for differentially methylated and expressed genes comparing TVI-mets with primary tumor as well as MFP-mets. Mediators for key pathways are indicated on the right side.

Close modal

For DNMT3B-mediated DNA methylation changes at the distant sites comparing TVI-mets with primary tumors, the differentially methylated CpG sites (β value difference > 0.1 and FDR < 0.05) were detected throughout the genome, including promoter/exon 1, rest exons, introns, 3′untranslated region (UTR), and intergenic regions (Fig. 3C). RNA-seq analysis showed differentially expressed genes (absolute 1.5-fold changes and P < 0.05; Fig. 3D). There was a negative correlation between CpG methylation in promoter/exon 1 and gene expression as exampled by Inpp5d, Cldn3, and Lfng, which are illustrated with a Janus plot that shows hypermethylated CpG sites (-log10FDR), their genomic positions, and the changes of gene expression (log2F; Fig. 3E). The promoter hypermethylation of tumor suppressors such as Lfng, Cldn3, and Inpp5d was evident at the DNMT3B enriched regions (Fig. 3F, top). DNMT3B KD in 4T1 cells decreased DNA methylation in these regions (Fig. 3F, top). On the other hand, increased gene body methylation indicated an elevated gene expression, such as Jag2 and Nos2 (Fig. 3F, bottom). Of the genes identified as differentially methylated, 148 showed a profile of both differential methylation and expression comparing TVI-mets with primary tumors (Fig. 3G, top). Together, these data suggest that DNMT3B induction at the distant sites altered the methylation and expression profile of many genes.

Between TVI-mets and MFP-mets, there were similarities and differences in methylation (Fig. 3C and F; Supplementary Fig. S3A) and gene expression (Fig. 3D). Of the 148 genes that were differentially methylated and expressed between TVI-mets and primary tumors, 51 are similar and 97 are different when compared to the MFP-mets (Fig. 3G, top). Tumor suppressors, such as Cdh11 and Tnfaip3, showed hypermethylation in the promoter region, and oncogenes, such as Fzd2 and Ksr1, showed prominent hypermethylation in the gene body (Supplementary Fig. S3B). These alterations in DNA methylation were a DNMT3B-specific effect because DNMT3B KD decreased the methylation level (Supplementary Fig. S3B). These data support the hypothesis that DNMT3B alters DNA methylation and gene expression in metastatic cancer cells.

DNMT3B-targeted key mediators and pathways

Ingenuity Pathway Analysis (IPA) identified DNMT3B-targeted signaling pathways (Supplementary Table S1). The top ranked are STAT3, NFκB, integrin-linked kinase (ILK), growth factor-receptor tyrosine kinase (RTK) pathways, β-catenin, and Notch, as well as cell junction and extracellular matrix (ECM) remodeling (Fig. 4A), which are important in tumor cell survival, proliferation, invasion, and colonization. Several molecules were selected as a readout for these pathways, including pSTAT3, pNFκB, pAkt, β-catenin, and notch1 intracellular domain (NICD-1; Fig. 4A, top, red boxed). Increased activation of these pathways were observed in both TVI-mets and MFP-mets when compared with primary tumors and cultured 4T1 cells (Fig. 4B). Interestingly, TVI-mets and MFP-mets shared common mediators (Fig. 4A, bottom, red), but there were also mediators unique in TVI-mets (Fig. 4A, bottom, blue) or in MFP-mets (Fig. 4A, bottom, black). These results show while different molecular mediators could be targeted by DNMT3B comparing TVI-met with MFP-met, they resulted the activation of the same signaling pathways. For example, while the β-catenin pathway was activated in both TVI-mets and MFP-mets (Fig. 4B), the Wnt ligands (Wnt 6, 7a, 9a, 10a) and Wnt sequestration molecule Sfrp1, coreceptors, and noncanonical pathway activators (Kremen1, Celsr1) were differentially expressed between TVI-mets and MFP-mets. DNMT3B KD in 4T1 tumor cells showed a decreased activation of these pathways in a time course experiment (Fig. 4C). Furthermore, DNMT3B KD decreased DNMT3B ChIP peak enrichment for the DNMT3B-targeted genes (Fig. 4D), which is consistent with altered gene expression (Fig. 4E). Notably, DNMT3B-targeted both tumor suppressors and oncogenes. For example, tumor suppressor Tnfaip3, which showed promoter hypermethylation, DNMT3B KD decreased peak enrichment (Fig. 4D) and increased Tnfaip3 expression (Fig. 4E). For oncogene Nos2, which showed gene body hypermethylation, DNMT3B KD decreased peak enrichment (Fig. 4D) and decreased Nos2 expression (Fig. 4E). These results indicate that DNMT3B-mediated epigenetic reprogramming and pathway activation could be achieved through different mediators that are microenvironment-dependent. The distant organ microenvironment contributes to this epigenetic reprogramming.

Figure 4.

DNMT3B alters key mediators in STAT3, NFκB, PI3K/Akt, β-catenin, as well as Notch signaling pathways. A, Schematic of major signaling pathways and key genes targeted by DNMT3B. Red line–labeled molecules pSTAT3, pNFκB, pAKT, β-catenin, and NICD-1 were used as readouts for the corresponding signaling pathways (top). DNMT3B-targeted pathways and genes comparing TVI-mets and MFP-mets. Red, common genes between TVI-mets and MFP-mets; blue, unique genes in TVI-mets; black, unique genes in MFP-mets are in black (bottom). B, Western blots of DNMT3B-targeted genes and pathways comparing TVI-mets, MFP-mets, primary tumors, and 4T1 cells. Each lane is from an individual mouse. C, Western blots of DNMT3B-targeted genes and pathways from DNMT3B KD or wt 4T1 cells. D, DNMT3B-ChIP-qPCR from DNMT3B KD or wt 4T1 cells. E, qRT-PCR for mRNA expression in DNMT3B KD over wt 4T1 cells (log2-fold changes). Red, common genes between TVI-mets and MFP-mets; blue, unique genes in TVI-mets; black, unique genes in MFP-mets. Data are presented as mean ± SD. *, P > 0.05; **, P > 0.01.

Figure 4.

DNMT3B alters key mediators in STAT3, NFκB, PI3K/Akt, β-catenin, as well as Notch signaling pathways. A, Schematic of major signaling pathways and key genes targeted by DNMT3B. Red line–labeled molecules pSTAT3, pNFκB, pAKT, β-catenin, and NICD-1 were used as readouts for the corresponding signaling pathways (top). DNMT3B-targeted pathways and genes comparing TVI-mets and MFP-mets. Red, common genes between TVI-mets and MFP-mets; blue, unique genes in TVI-mets; black, unique genes in MFP-mets are in black (bottom). B, Western blots of DNMT3B-targeted genes and pathways comparing TVI-mets, MFP-mets, primary tumors, and 4T1 cells. Each lane is from an individual mouse. C, Western blots of DNMT3B-targeted genes and pathways from DNMT3B KD or wt 4T1 cells. D, DNMT3B-ChIP-qPCR from DNMT3B KD or wt 4T1 cells. E, qRT-PCR for mRNA expression in DNMT3B KD over wt 4T1 cells (log2-fold changes). Red, common genes between TVI-mets and MFP-mets; blue, unique genes in TVI-mets; black, unique genes in MFP-mets. Data are presented as mean ± SD. *, P > 0.05; **, P > 0.01.

Close modal

Mechanisms and biological function of DNMT3B-targeted pathways

To investigate the molecular mechanisms and biological functions of DNMT3B mediated epigenetic reprogramming, rescue experiments were performed in DNMT3B KD cells using constitutively active constructs for the key signaling pathways identified above, NICD1 for Notch signaling, myrAkt for Akt signaling, STAT3C for STAT3 signaling, and β-catenin overexpression for β-catenin signaling pathways. First, in a 3D-Matrigel assay that examines the key capacity of tumor cell colonization, reactivation of Akt, STAT3, and Notch but not β-catenin signaling pathways rescued the defect of sphere formation in DNMT3B KD 4T1 cells (Fig. 5A).

Figure 5.

DNMT3B activates pathways that are important in cancer cell survival, apoptosis, proliferation, invasion, and colonization. A, Tumor sphere formation in 3D Matrigel, reactivating Akt (myrAkt), STAT3 (STAT3C), and Notch (NICD-1) but not β-catenin signaling pathways that rescued the defect of DNMT3B KD 4T1 cells in tumor sphere formation. Shown are representative pictures (left) and quantitative data (right). B, NICD1 Western blot (top) and colony formation (bottom) of DNMT3B KD 4T1 cells with a NICD1 construct. C, pAkt Western blot (left top panel) and cell counts (left bottom panel) for myrAkt rescued proliferation defect in DNMT3B KD 4T1 cells. Right, IHC of phospho-histone H3 of lung nodules from mice injected with DNMT3B KD or wt 4T1 cells. The quantitative data are on the right, with the number of positive cells per lung nodule from 20 nodules of comparable size. D, pSTAT3 and c-Casp3 Western (left top panel), percentage of 7AAD/Annexin V+ population (left bottom panel), and representative flow cytometry (middle) of DNMT3B KD 4T1 cells transfected with STAT3C. Right, TUNEL assay of lung nodules from mice injected with DNMT3B KD or wt 4T1 cells. Quantitative data on right with the number of TUNEL+ cells per lung nodule from 20 nodules of comparable size. E, Wound closure assay of DNMT3B KD 4T1 cells with a constitutive active β-catenin construct. Representative pictures of wound closure at 15 hours on the right. Data are presented as mean ± SD. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Figure 5.

DNMT3B activates pathways that are important in cancer cell survival, apoptosis, proliferation, invasion, and colonization. A, Tumor sphere formation in 3D Matrigel, reactivating Akt (myrAkt), STAT3 (STAT3C), and Notch (NICD-1) but not β-catenin signaling pathways that rescued the defect of DNMT3B KD 4T1 cells in tumor sphere formation. Shown are representative pictures (left) and quantitative data (right). B, NICD1 Western blot (top) and colony formation (bottom) of DNMT3B KD 4T1 cells with a NICD1 construct. C, pAkt Western blot (left top panel) and cell counts (left bottom panel) for myrAkt rescued proliferation defect in DNMT3B KD 4T1 cells. Right, IHC of phospho-histone H3 of lung nodules from mice injected with DNMT3B KD or wt 4T1 cells. The quantitative data are on the right, with the number of positive cells per lung nodule from 20 nodules of comparable size. D, pSTAT3 and c-Casp3 Western (left top panel), percentage of 7AAD/Annexin V+ population (left bottom panel), and representative flow cytometry (middle) of DNMT3B KD 4T1 cells transfected with STAT3C. Right, TUNEL assay of lung nodules from mice injected with DNMT3B KD or wt 4T1 cells. Quantitative data on right with the number of TUNEL+ cells per lung nodule from 20 nodules of comparable size. E, Wound closure assay of DNMT3B KD 4T1 cells with a constitutive active β-catenin construct. Representative pictures of wound closure at 15 hours on the right. Data are presented as mean ± SD. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Close modal

These pathways were further investigated in specific assays for colony formation, survival, proliferation, and migration. Overexpression of the NICD1, an active form of Notch, recovered colonization capacity of DNMT3B KD 4T1 cells in soft agar assays (Fig. 5B). Akt activation by a constitutively active construct myrAkt rescued the defect in proliferation but not the survival of DNMT3B KD 4T1 cells (Fig. 5C, left; Supplementary Fig. S4A). This is consistent with the observation that DNMT3B KD in 4T1 cells not only decreased the number of lung nodules but also the size of the nodules (Supplementary Fig. S4B), tumor growth in mice (Supplementary Fig. S4C), and decreased phospho-histone H3, a proliferative marker (Fig. 5C, right). Activation of the STAT3 signaling pathway through overexpression of STAT3C increased cell survival, which showed a decreased caspase-3 cleavage and decreased 7AAD/Annexin V+ population in DNMT3B KD 4T1 cells (Fig. 5D, left and middle). Indeed, in metastatic nodules, TUNEL assay revealed an increased number of TUNEL+ cells in DNMT3B KD (Fig. 5D, right panels). These observations were further confirmed in the EMT6 mammary tumor model (Supplementary Fig. S4D–S4F). Furthermore, overexpression of β-catenin reversed the defect in migration and wound healing in DNMT3B KD 4T1 cells (Fig. 5E). Altogether these results demonstrate the key mechanisms downstream of DNMT3B in epigenetic reprogramming of metastatic cancer cells.

Mechanisms of DNMT3B induction at the distant metastatic site

To identify key inflammatory molecules at the distant sites that are critical in DNMT3B induction, the levels of inflammatory molecules in the premetastatic lungs of tumor-bearing mice were compared to those in the lungs of naïve mice. PGE2 and IL6 were the highest increased inflammatory molecules in the premetastatic lungs (Fig. 6A; Supplementary Fig. S5A). PGE2 and IL6 were also detected in high levels at the late stage of metastatic lungs (Supplementary Fig. S5B). Interestingly, the DNMT3B promoter (MatInspector, Genomatix Software) has several binding sites for NFκB and STAT3 (Supplementary Fig. S5C) that are well-known as downstream transcription factors of PGE2 and IL6, respectively. Interestingly, DNMT3B was induced in cultured 4T1 cells by conditioned medium from day 14 and day 35 lungs of tumor-bearing mice, but not by the conditioned medium from day 35 tumors (Fig. 6B; Supplementary Fig. S5D). The lung supernatant-induced DNMT3B was diminished by PGE2 receptor antagonist (AH-6809) or by NFκB-p65 knockdown (Fig. 6B). Similarly, IL6 neutralizing antibody or STAT3 inhibitor (Stattic) also inhibited the DNMT3B induction by the conditioned medium from day 35 lungs (Fig. 6B). In contrast, neutralization of TGFβ or knockdown Smad 2 did not alter the DNMT3B induction (Fig. 6B). Furthermore, the addition of PGE2 and IL6 to the tumor cells cultured in vitro confirmed the induction of DNMT3B mRNA and protein level in 4T1 cells through PGE2/NFκB (Supplementary Fig. S5E) and IL6/STAT3 signaling (Supplementary Fig. S5F). PGE2 and IL6 also increased DNMT3B protein levels in human breast cancer cell lines (Supplementary Fig. S5G).

Figure 6.

Inflammatory mediators PGE2 and IL6 Induce DNMT3B. A, ELISA and Bioplex assays of protein extraction from normal lungs and lungs from 4T1 tumor-bearing mice (n = 3–6 lungs). B Western blots for DNMT3B, pSTAT3, pNFκB, and pSmad2 of 4T1 cells cultured with lung tissue culture supernatants, and with neutralization of TGFβ, IL6, and/or PGE2 receptor antagonist (AH-6809), as well as knockdown of NFκB or Smad2 or treatment with STAT3 inhibitor (Stattic). C, Western blots for DNMT3B, pSTAT3, pNFκB in an in vitro coculture of 4T1 cells with MDSCs, with or without PGE2 receptor antagonist (AH-6809) and IL6 neutralizing antibody treatment. D, PGE2 ELISA of protein extraction from normal lungs and lungs from tumor-bearing mice with or without celecoxib treatment. E, DNMT3B Western blot of lung nodules of 4T1 (left) and EMT6 (right; each lane from an individual mouse) tumor-bearing mice treated with celecoxib; normal lung or primary tumors were used as controls. Days after MFP tumor injection are indicated. F, Locus-specific DNA methylation analysis and qRT-PCR of DNMT3B target genes Cldn9 and Lfng. G, Number of lung metastatic nodules from mice that were treated with PGE2 or IL6 or in combination. The DNMT3B KD or wt control 4T1 tumor cells were injected through tail vein; representative images are shown on the right. H, Schematic hypothesis. MDCSs, PGE2, and IL6 are increased in the inflammatory metastatic microenvironment, which induce DNMT3B in the metastatic cancer cells through PGE2/NFκB and IL6/STAT3 signaling pathways. DNMT3B alters gene methylation and transcriptome, thus activatings signaling pathways including STAT3, NFκB, PI3K/Akt, β-catenin, as well as Notch signaling pathways that are important in metastatic colonization. Data are presented as mean ± SD. *, P < 0.05; **, P < 0.01.

Figure 6.

Inflammatory mediators PGE2 and IL6 Induce DNMT3B. A, ELISA and Bioplex assays of protein extraction from normal lungs and lungs from 4T1 tumor-bearing mice (n = 3–6 lungs). B Western blots for DNMT3B, pSTAT3, pNFκB, and pSmad2 of 4T1 cells cultured with lung tissue culture supernatants, and with neutralization of TGFβ, IL6, and/or PGE2 receptor antagonist (AH-6809), as well as knockdown of NFκB or Smad2 or treatment with STAT3 inhibitor (Stattic). C, Western blots for DNMT3B, pSTAT3, pNFκB in an in vitro coculture of 4T1 cells with MDSCs, with or without PGE2 receptor antagonist (AH-6809) and IL6 neutralizing antibody treatment. D, PGE2 ELISA of protein extraction from normal lungs and lungs from tumor-bearing mice with or without celecoxib treatment. E, DNMT3B Western blot of lung nodules of 4T1 (left) and EMT6 (right; each lane from an individual mouse) tumor-bearing mice treated with celecoxib; normal lung or primary tumors were used as controls. Days after MFP tumor injection are indicated. F, Locus-specific DNA methylation analysis and qRT-PCR of DNMT3B target genes Cldn9 and Lfng. G, Number of lung metastatic nodules from mice that were treated with PGE2 or IL6 or in combination. The DNMT3B KD or wt control 4T1 tumor cells were injected through tail vein; representative images are shown on the right. H, Schematic hypothesis. MDCSs, PGE2, and IL6 are increased in the inflammatory metastatic microenvironment, which induce DNMT3B in the metastatic cancer cells through PGE2/NFκB and IL6/STAT3 signaling pathways. DNMT3B alters gene methylation and transcriptome, thus activatings signaling pathways including STAT3, NFκB, PI3K/Akt, β-catenin, as well as Notch signaling pathways that are important in metastatic colonization. Data are presented as mean ± SD. *, P < 0.05; **, P < 0.01.

Close modal

The inflammatory premetastatic and metastatic lung is modulated by myeloid-derived immune suppressor cells (MDSCs or Gr-1+CD11b+ cells) that are well known to suppress host immune surveillance (34, 35). In an in vitro coculture of tumor cells with MDSCs, MDSCs stimulated DNMT3B induction in tumor cells (Fig. 6C). PGE2 receptor antagonist (AH-6809) and IL6 neutralizing antibody diminished DNMT3B induction and consistently inhibited NFκB and STAT3 signaling (Fig. 6C). These data suggest DNMT3B induction is mediated by MDSCs through PGE2 and IL6.

Because of the high PGE2 concentration in metastatic lungs, celecoxib, a COX2 inhibitor, was used to decrease PGE2 and suppress the inflammation in the metastatic microenvironment. Celecoxib treatment significantly decreased PGE2 levels in the lungs of tumor-bearing mice (Fig. 6D) and inhibited DNMT3B induction in lung nodules of both 4T1 and EMT6 models (Fig. 6E). To confirm the epigenetic regulation of DNMT3B targeted genes by inflammation in vivo, Cldn9 and Lfng, negative regulators of β-catenin and Notch pathways, respectively, were selected as two representative DNMT3B-targeted molecules because of strong DNMT3B peak enrichment and hypermethylation in their promoters (Supplementary Fig. S5H). Celecoxib treatment decreased CpG site methylation and increased gene expression of Cldn9 and Lfng in metastatic lung nodules (Fig. 6F). Importantly, overexpression of DNMT3B using Tet-inducible system reversed the effect of celecoxib on DNA methylation and gene expression of Cldn9 and Lfng (Fig. 6F). Finally, treatment with IL6 or PGE2 or the two in combination increased the number of lung metastatic nodules in mice that received tail vein injection of wt but not the DNMT3B KD 4T1 tumor cells, suggesting the effect of these inflammatory mediators on metastasis at the distant site is DNMT3B dependent (Fig. 6G). Altogether, these data further demonstrate that DNMT3B is induced in the distant metastatic site through PGE2/NFκB and IL6/STAT3 signaling pathways. DNMT3B alters gene methylation and transcriptome, leading to the activation of signaling pathways that are important in metastatic colonization (Fig. 6H).

Correlation of DNMT3B with human tumor metastasis, perioperative COX2 inhibition in combination with chemotherapy, and IL6 neutralization in combination with PD-1 immunotherapy in preclinical mouse models

We next used publicly available clinical datasets and correlated DNMT3B expression levels with metastatic progression. DNMT3B, but not DNMT1 and DNMT3A, was increased in metastatic samples compared with matched primary tumors of patients with breast cancer (Fig. 7A; Supplementary Fig. S6A; ref. 30). This result was also observed in prostate cancer patients (Fig. 7B; Supplementary Fig. S6B; ref. 31), in patients with melanoma (Fig. 7C; Supplementary Fig. S6C; ref. 32) as well as in patients with renal carcinoma (Fig. 7D; Supplementary Fig. S6D; ref. 33). In 7 cases of breast tumor samples obtained, distant metastatic tumors showed significantly more cells with elevated DNMT3B than that from the matched primary tumors (Fig. 7E). These data suggest a positive correlation of DNMT3B expression levels with metastatic progression.

Figure 7.

Correlation of DNMT3B with human tumor metastasis, perioperative COX2 inhibition in chemotherapy, and IL6 neutralization in PD-1 immunotherapy in preclinical mouse models. A, DNMT3B expression in paired primary and metastatic tumor tissues from patients with breast cancer (Weigelt dataset; n = 8). BD, DNMT3B expression in primary tumor and metastases of patients with prostate cancer (Tamura dataset; B), melanoma (Haqq dataset; C), and renal cell carcinoma (Wuttig dataset; D). E, Hematoxylin and eosin and DNMT3B immunofluorescence for matched primary and metastatic breast tumors (n = 7 cases). Representative images are shown, with quantitative data on right. F, Schematic of experimental design for perioperative anti-inflammatory therapy and IL6 neutralization in combination with anti-PD-1 immunotherapy. Animals were treated with doxorubicin (Dxr; 2 mg/kg bodyweight), meloxicam (Mel; 2 mg/kg bodyweight), Etodolac (Eto; 10 mg/kg bodyweight), IL6 neutralizing antibody (IL6 Ab; 10 mg/kg bodyweight), or PD-1 neutralizing antibody (PD-1 Ab; 5 mg/kg bodyweight) as shown in the design. G, The effect of perioperative anti-inflammatory therapy on 4T1 metastasis (n = 10). The number of metastatic lung nodules (left) and the weights of recurrent tumors (right). H, The effect of IL6 antibody in combination with PD-1 Ab on 4T1 metastasis (n = 9–10). The number of metastatic lung nodules (left), the percentage of metastasis-free mice (middle), and the weights of recurrent tumors (right) are graphed. Data are presented as mean ± SD. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Figure 7.

Correlation of DNMT3B with human tumor metastasis, perioperative COX2 inhibition in chemotherapy, and IL6 neutralization in PD-1 immunotherapy in preclinical mouse models. A, DNMT3B expression in paired primary and metastatic tumor tissues from patients with breast cancer (Weigelt dataset; n = 8). BD, DNMT3B expression in primary tumor and metastases of patients with prostate cancer (Tamura dataset; B), melanoma (Haqq dataset; C), and renal cell carcinoma (Wuttig dataset; D). E, Hematoxylin and eosin and DNMT3B immunofluorescence for matched primary and metastatic breast tumors (n = 7 cases). Representative images are shown, with quantitative data on right. F, Schematic of experimental design for perioperative anti-inflammatory therapy and IL6 neutralization in combination with anti-PD-1 immunotherapy. Animals were treated with doxorubicin (Dxr; 2 mg/kg bodyweight), meloxicam (Mel; 2 mg/kg bodyweight), Etodolac (Eto; 10 mg/kg bodyweight), IL6 neutralizing antibody (IL6 Ab; 10 mg/kg bodyweight), or PD-1 neutralizing antibody (PD-1 Ab; 5 mg/kg bodyweight) as shown in the design. G, The effect of perioperative anti-inflammatory therapy on 4T1 metastasis (n = 10). The number of metastatic lung nodules (left) and the weights of recurrent tumors (right). H, The effect of IL6 antibody in combination with PD-1 Ab on 4T1 metastasis (n = 9–10). The number of metastatic lung nodules (left), the percentage of metastasis-free mice (middle), and the weights of recurrent tumors (right) are graphed. Data are presented as mean ± SD. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Close modal

Increased PGE2 levels in the circulation were observed in patients with cancer after tumor resection surgery, and the systemic inflammation by the COX2/PGE2 pathway was suggested to facilitate early relapse and metastasis (36). Our study showed that PGE2 was the most significantly increased inflammatory mediator in the premetastatic sites (Fig. 6A), and PGE2 induced DNMT3B in breast cancer cells (Fig. 6B; Supplementary Fig. S5E). Therefore, we investigated the effect of perioperative inhibition of the COX2/PGE2 pathway using COX2 inhibitors, meloxicam or etodolac, in 4T1 preclinical mouse model (Fig. 7F, schematic design). The perioperative inhibition of COX2 significantly reduced the number of lung nodules, and the treatment of COX2 inhibitors further enhanced antimetastasis efficacy of doxorubicin chemotherapy (Fig. 7G, left; and Supplementary Fig. S6E). Treatment of doxorubicin alone or cotreatment of doxorubicin with COX-2 inhibitors significantly suppressed recurrent tumor growth (Fig. 7G, right).

DNMT3B is also induced by IL6 (Fig. 6B and C). The IL6/JAK/STAT3 signaling is known to suppress the antitumor immune response. In addition, agents targeting IL6 signaling pathways have already received FDA approval for the treatment of inflammatory conditions (25, 37). Therefore, we next explored whether IL6 neutralization will improve the efficacy of PD-1 immunotherapy that has a low response rate but acceptable safety profile and antitumor activity in patients with triple-negative breast cancer (TNBC). In the preclinical 4T1 mouse model (Fig. 7F, schematic design), IL6 neutralization or anti-PD-1 immunotherapy alone decreased the number of metastatic nodules. Importantly, IL6 neutralization potentiated the attenuation of lung metastasis by PD-1 neutralization Ab (Fig. 7H, left; and Supplementary Fig. S6F). Consistently, the number of metastasis-free mice was markedly increased in mice that received the combination treatment (Fig. 7H, middle). IL6 neutralization also enhanced the attenuated recurrent tumor growth by PD-1 neutralization (Fig. 7H, right). These data suggest that IL6 neutralization could be used to enhance the efficacy of PD-1 immunotherapy. Altogether the enhanced PD-1 immunotherapy by IL6 neutralization or doxorubicin efficacy by meloxicam is striking and clear, with no need to combine COX-2 inhibition with IL6 neutralization for the immunotherapy or doxorubicin treatment. This is consistent with the observation that the combination of meloxicam and IL6 neutralization showed no significant decrease in metastasis, recurrent tumor growth, or number of CTCs when compared with that of the single treatment alone (Supplementary Fig. S7A–S7C).

Our study provides experimental evidence and conceptual proof that DNMT3B is induced at the metastatic site and mediates epigenetic reprogramming for metastatic colonization. Our results offer insight into metastasis biology and accentuate the inherent insufficiency and daunting challenge of current treatment strategies. Therapies based on the genetic characterization of primary tumor biopsies can only prolong the lives of patients with metastatic disease by a few weeks or months (1, 38). While metastatic driver mutations are continually investigated and identified, this study asserts that epigenetic acquisition of malignant traits after tumor cell dissemination must be taken into consideration. Our study also suggests that epigenetic regulation is likely critical in manifestation of full malignancy for early disseminated cancer cells (39), especially in the absence of driver mutations.

Our work reveals a new role of inflammation in DNMT3B induction and bolsters evidence that there is a pronounced link existing between inflammatory pathways and epigenetic mechanisms. Interestingly, NFκB and STAT3 pathways are not only upstream of the DNMT3B induction but are also activated as DNMT3B downstream effectors. This potential feedback loop indicates an epigenetic mechanism to mediate transcriptional reprogramming of cancer cells, which could facilitate metastasis in less microenvironment-dependent manner. A recent study showed transcriptional reprogramming of resistant signatures was acquired in response to chemotherapy in patients with TNBC (40), suggesting an adaptation mechanism under stress conditions. Our data suggest the metastatic microenvironment modifies tumor cells epigenetically thus reprogramming the transcriptome. Cancer-associated inflammation is critical in tumor progression. However, whether and how the cancer-associated inflammation affects metastatic colonization at distant sites remains unclear. IL6 and PGE2 are two potent proinflammatory mediators known to drive the production, accumulation, and immune suppressive potency of MDSCs (41). We now show that premetastatic/metastatic lungs produced high levels of IL6 and PGE2. In addition, IL6 and PGE2 play an important role in DNMT3B induction in metastatic cancer cells at the distant sites. Furthermore, DNMT3B was induced in cancer cells when co-cultured with MDSCs that were abundant in the premetastatic and metastatic distant organ microenvironment (35). A recent study of DNA methylation landscape reported the significant correlation between DNA methylation and immune cell infiltration in primary and recurring human glioblastomas, providing the clinical evidence for the link between tumor epigenome and immune cells (42). Further investigation utilizing genetically engineered mouse models to manipulate specific host immune cells will provide more insight about epigenetic regulation of tumor cells by microenvironment in vivo. Nevertheless, our data suggest that the inflammatory microenvironment at the distant organ sites promotes tumor metastasis through DNMT3B-mediated epigenetic mechanism.

Our study shows that perioperative inhibition of COX2/PGE2 pathway using meloxicam or etodolac reduced the number of metastatic nodules in 4T1 preclinical mouse model. Importantly, it significantly enhanced antimetastasis efficacy of doxorubicin (Fig. 7F and G, left; Supplementary Fig. S6E). Furthermore, there was also profound effect on recurrent tumor growth when used in combination with doxorubicin (Fig. 7F and G, right). There has been increased understanding that tumor resection induces systemic inflammation, which mobilizes inflammatory immune cells and promotes tumor relapse (43). In fact, surgical removal of primary tumors in patients with ER-positive breast cancer correlated with increased tumor recurrence and metastases 1–2 years post tumor resection (44). In addition, neoadjuvant chemotherapies increase secretion of prometastatic extracellular vesicles and to induce the formation of prometastatic microenvironment in preclinical mouse models of breast cancer (45, 46). The significantly decreased metastasis by the short-term perioperative treatment of COX-2 inhibitors in combination with doxorubicin neoadjuvant chemotherapy suggests a unique therapeutic window that can provide long-term benefit for patients with cancer. Moreover, the usage of common and inexpensive anti-inflammatory drugs provides additional value to this therapeutic approach.

In our study, IL6 neutralization enhanced the efficacy of PD-1 immunotherapy for breast cancer metastasis in the preclinical mouse models. Significant evidence supports immunotherapy for breast cancer treatment (47). A recent clinical trial demonstrated that cotreatment of PD-L1 antibody (atezolizumab) and nab-paclitaxel prolonged progression-free survival in patients with advanced TNBC (3), leading to the first FDA approval of immunotherapy in breast cancer. However, the complete response rate was only 7.1% in patients who received the cotreatment (3). Indeed, low response rate, therapy relapse and resistance are major challenges in immunotherapy (2). The IL6/JAK/STAT3 pathway could contribute to these challenges, because this pathway strongly suppresses host antitumor immunity and generally correlates with poor prognosis (25). In fact, PD-1 blockade elevates IL6-mediated inflammatory response in macrophages, and IL6 inhibition alleviates this adverse effect (29). The crosstalk between IL6 and PD1/PD-L1 could be utilized to overcome the narrow therapeutic window of anti-PD-1/PD-L1 therapy. These studies and our data support our proposal that IL6 blockade could provide an important option to enhance the efficacy of PD-1 immunotherapy for patients with metastatic TNBC. We anticipate IL6 neutralization will directly inhibit the epigenetic dysregulation in tumor cells and at the same time enhance host antitumor immunity.

It is not fully understood how DNA methylation is regulated, and the transcriptional outcome of DNA methylation seems to vary with genomic context (48). Our data support a promoter/exon 1 hypermethylation in down regulation of many tumor suppressor genes such as apoptosis mediators (e.g., Tnfaip3), cell adhesion molecules (e.g., Cdh11), and phosphatases (e.g., Ptpn7, Inpp5d). Conversely, DNA hypermethylation in the gene body generally positively correlates with increased oncogene expression, such as Nos2, Fzd2, Jag2. These results are consistent with other epigenetic studies (17, 18, 22). Interestingly, we notice that the DNMT3B ChIP-seq peaks do not always result in alterations of DNA methylation. Among DNMT3B-enriched regions, 22.0% showed changes in CpG methylations. Of particular note, DNMT3B does not possess a specific DNA binding domain. Rather it has a PWWP domain that interacts with histone protein markers (49). We suspect that specific histone modifications may mediate DNMT3B recruitment in certain gene loci. Indeed DNMT3B is recruited to H3K36m3 for gene body methylation in mouse embryonic stem cells and is critical in transcription fidelity in controlling splicing variants (50). A recent report also shows that CpG islands are premarked by H3K4me1 in normal prostate cells but are susceptible to DNMT3B mediated hypermethylation at the boundaries of those CpG islands in prostate cancer cells (51). Further studies of histone modifications that are specific to metastatic tumors will provide more comprehensive understanding of epigenetic regulation in metastatic progression.

Unsurprisingly, but importantly, DNMT3B targets not only multiple genes in the same pathway, including positive and negative mediators, but also multiple pathways (Fig. 4). This global epigenetic reprogramming of metastatic cancer cells is also reported for small-cell lung cancer cells that acquire chromatin accessibility at distal regulatory regions and facilitate the expression of prometastatic genes (52). Because of the magnitude of DNMT3B-mediated alterations, our work argue that: (i) the current targeting strategy based only on genomic alterations with combination of 2–3 drugs are unlikely to be sufficient; (ii) DNMT3B-mediated epigenetic alterations must be considered in therapeutic design; (iii) a feasible anti-inflammatory approach that targets DNMT3B induction and enzymatic activities will likely restore tumor suppressor genes and improve efficacy of standard therapy. This treatment approach may also be useful in addressing therapeutic resistance mediated by the tumor microenvironment and its poised epigenetic states (53). The mechanistic insights from our study have implications regarding the steps of the metastatic adaptation and colonization that appear amenable to therapeutic targeting and metastasis prevention.

No potential conflicts of interest were disclosed.

Conception and design: L. Yang, J.Y. So, N. Skrypek

Development of methodology: N. Skrypek, B.R. Achyut

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.Y. So, N. Skrypek, H. Ishii, J.M. Chen, B.R. Achyut, E.C. Yoon, L. Han

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.Y. So, N. Skrypek, H.H. Yang, A.S. Merchant, G.W. Nelson, G. Hu, M.C. Cam, K. Zhao, M.P. Lee

Writing, review, and/or revision of the manuscript: L. Yang, J.Y. So, N. Skrypek, A.S. Merchant, B.R. Achyut, E.C. Yoon, M.P. Lee

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): W.-D. Chen, C. Huang

Study supervision: L. Yang, K. Zhao

We thank Dr. Michael Bustin for his critical reading of the manuscript. We thank Dr. Hua Yu (City of Hope) for providing the Stat3-C vector and Dr. Lalage Wakefield (NCI) for providing shRNA for SMAD2 and SMAD3. We appreciate the CCR Flow Cytometry Core for their technical assistance on FACS. We would like to thank Cindy Clark, NIH Library Writing Center, for manuscript editing assistance. The studies are supported by the US government (NCI) intramural funding (to L. Yang).

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.
Steeg
PS
. 
Targeting metastasis
.
Nat Rev Cancer
2016
;
16
:
201
18
.
2.
Sharma
P
,
Hu-Lieskovan
S
,
Wargo
JA
,
Ribas
A
. 
Primary, adaptive, and acquired resistance to cancer immunotherapy
.
Cell
2017
;
168
:
707
23
.
3.
Schmid
P
,
Adams
S
,
Rugo
HS
,
Schneeweiss
A
,
Barrios
CH
,
Iwata
H
, et al
Atezolizumab and Nab-paclitaxel in advanced triple-negative breast cancer
.
N Engl J Med
2018
;
379
:
2108
21
.
4.
Polyak
K
. 
Tumor heterogeneity confounds and illuminates: a case for Darwinian tumor evolution
.
Nat Med
2014
;
20
:
344
6
.
5.
Makohon-Moore
AP
,
Zhang
M
,
Reiter
JG
,
Bozic
I
,
Allen
B
,
Kundu
D
, et al
Limited heterogeneity of known driver gene mutations among the metastases of individual patients with pancreatic cancer
.
Nat Genet
2017
;
49
:
358
66
.
6.
Klein
CA
. 
Selection and adaptation during metastatic cancer progression
.
Nature
2013
;
501
:
365
72
.
7.
McGranahan
N
,
Swanton
C
. 
Biological and therapeutic impact of intratumor heterogeneity in cancer evolution
.
Cancer Cell
2015
;
27
:
15
26
.
8.
McCreery
MQ
,
Halliwill
KD
,
Chin
D
,
Delrosario
R
,
Hirst
G
,
Vuong
P
, et al
Evolution of metastasis revealed by mutational landscapes of chemically induced skin cancers
.
Nat Med
2015
;
21
:
1514
20
.
9.
Schmidt-Kittler
O
,
Ragg
T
,
Daskalakis
A
,
Granzow
M
,
Ahr
A
,
Blankenstein
TJ
, et al
From latent disseminated cells to overt metastasis: genetic analysis of systemic breast cancer progression
.
Proc Natl Acad Sci U S A
2003
;
100
:
7737
42
.
10.
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
.
11.
Timp
W
,
Feinberg
AP
. 
Cancer as a dysregulated epigenome allowing cellular growth advantage at the expense of the host
.
Nat Rev Cancer
2013
;
13
:
497
510
.
12.
Jones
PA
,
Baylin
SB
. 
The epigenomics of cancer
.
Cell
2007
;
128
:
683
92
.
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.
Fang
F
,
Turcan
S
,
Rimner
A
,
Kaufman
A
,
Giri
D
,
Morris
LG
, et al
Breast cancer methylomes establish an epigenomic foundation for metastasis
.
Sci Transl Med
2011
;
3
:
75ra25
.
15.
Gkountela
S
,
Castro-Giner
F
,
Szczerba
BM
,
Vetter
M
,
Landin
J
,
Scherrer
R
, et al
Circulating tumor cell clustering shapes DNA methylation to enable metastasis seeding
.
Cell
2019
;
176
:
98
112
.
16.
Legendre
C
,
Gooden
GC
,
Johnson
K
,
Martinez
RA
,
Liang
WS
,
Salhia
B
. 
Whole-genome bisulfite sequencing of cell-free DNA identifies signature associated with metastatic breast cancer
.
Clin Epigenetics
2015
;
7
:
100
.
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.
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
.
19.
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
.
20.
De Carvalho
DD
,
Sharma
S
,
You
JS
,
Su
SF
,
Taberlay
PC
,
Kelly
TK
, et al
DNA methylation screening identifies driver epigenetic events of cancer cell survival
.
Cancer Cell
2012
;
21
:
655
67
.
21.
Mishra
A
,
Liu
S
,
Sams
GH
,
Curphey
DP
,
Santhanam
R
,
Rush
LJ
, et al
Aberrant overexpression of IL-15 initiates large granular lymphocyte leukemia through chromosomal instability and DNA hypermethylation
.
Cancer Cell
2012
;
22
:
645
55
.
22.
O'Hagan
HM
,
Wang
W
,
Sen
S
,
Destefano Shields
C
,
Lee
SS
,
Zhang
YW
, et al
Oxidative damage targets complexes containing DNA methyltransferases, SIRT1, and polycomb members to promoter CpG Islands
.
Cancer Cell
2011
;
20
:
606
19
.
23.
Xia
D
,
Wang
D
,
Kim
SH
,
Katoh
H
,
DuBois
RN
. 
Prostaglandin E2 promotes intestinal tumor growth via DNA methylation
.
Nat Med
2012
;
18
:
224
6
.
24.
Peinado
H
,
Zhang
H
,
Matei
IR
,
Costa-Silva
B
,
Hoshino
A
,
Rodrigues
G
, et al
Pre-metastatic niches: organ-specific homes for metastases
.
Nat Rev Cancer
2017
;
17
:
302
17
.
25.
Johnson
DE
,
O'Keefe
RA
,
Grandis
JR
. 
Targeting the IL-6/JAK/STAT3 signalling axis in cancer
.
Nat Rev Clin Oncol
2018
;
15
:
234
48
.
26.
Wang
D
,
Dubois
RN
. 
Eicosanoids and cancer
.
Nat Rev Cancer
2010
;
10
:
181
93
.
27.
Lee
E
,
Fertig
EJ
,
Jin
K
,
Sukumar
S
,
Pandey
NB
,
Popel
AS
. 
Breast cancer cells condition lymphatic endothelial cells within pre-metastatic niches to promote metastasis
.
Nat Commun
2014
;
5
:
4715
.
28.
Ogawa
F
,
Amano
H
,
Eshima
K
,
Ito
Y
,
Matsui
Y
,
Hosono
K
, et al
Prostanoid induces premetastatic niche in regional lymph nodes
.
J Clin Invest
2014
;
124
:
4882
94
.
29.
Tsukamoto
H
,
Fujieda
K
,
Miyashita
A
,
Fukushima
S
,
Ikeda
T
,
Kubo
Y
, et al
Combined blockade of IL6 and PD-1/PD-L1 signaling abrogates mutual regulation of their immunosuppressive effects in the tumor microenvironment
.
Cancer Res
2018
;
78
:
5011
22
.
30.
Weigelt
B
,
Glas
AM
,
Wessels
LF
,
Witteveen
AT
,
Peterse
JL
,
van't Veer
LJ
. 
Gene expression profiles of primary breast tumors maintained in distant metastases
.
Proc Natl Acad Sci U S A
2003
;
100
:
15901
5
.
31.
Tamura
K
,
Furihata
M
,
Tsunoda
T
,
Ashida
S
,
Takata
R
,
Obara
W
, et al
Molecular features of hormone-refractory prostate cancer cells by genome-wide gene expression profiles
.
Cancer Res
2007
;
67
:
5117
25
.
32.
Haqq
C
,
Nosrati
M
,
Sudilovsky
D
,
Crothers
J
,
Khodabakhsh
D
,
Pulliam
BL
, et al
The gene expression signatures of melanoma progression
.
Proc Natl Acad Sci U S A
2005
;
102
:
6092
7
.
33.
Wuttig
D
,
Zastrow
S
,
Fussel
S
,
Toma
MI
,
Meinhardt
M
,
Kalman
K
, et al
CD31, EDNRB and TSPAN7 are promising prognostic markers in clear-cell renal cell carcinoma revealed by genome-wide expression analyses of primary tumors and metastases
.
Int J Cancer
2012
;
131
:
E693
704
.
34.
Gabrilovich
DI
,
Nagaraj
S
. 
Myeloid-derived suppressor cells as regulators of the immune system
.
Nat Rev Immunol
2009
;
9
:
162
74
.
35.
Yan
HH
,
Pickup
M
,
Pang
Y
,
Gorska
AE
,
Li
Z
,
Chytil
A
, et al
Gr-1+CD11b+ myeloid cells tip the balance of immune protection to tumor promotion in the premetastatic lung
.
Cancer Res
2010
;
70
:
6139
49
.
36.
Hiller
JG
,
Perry
NJ
,
Poulogiannis
G
,
Riedel
B
,
Sloan
EK
. 
Perioperative events influence cancer recurrence risk after surgery
.
Nat Rev Clin Oncol
2018
;
15
:
205
18
.
37.
Gabay
C
,
Emery
P
,
van Vollenhoven
R
,
Dikranian
A
,
Alten
R
,
Pavelka
K
, et al
Tocilizumab monotherapy versus adalimumab monotherapy for treatment of rheumatoid arthritis (ADACTA): a randomised, double-blind, controlled phase 4 trial
.
Lancet
2013
;
381
:
1541
50
.
38.
Lambert
AW
,
Pattabiraman
DR
,
Weinberg
RA
. 
Emerging Biological Principles of Metastasis
.
Cell
2017
;
168
:
670
91
.
39.
Sosa
MS
,
Bragado
P
,
Aguirre-Ghiso
JA
. 
Mechanisms of disseminated cancer cell dormancy: an awakening field
.
Nat Rev Cancer
2014
;
14
:
611
22
.
40.
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
.
41.
Ostrand-Rosenberg
S
,
Fenselau
C
. 
Myeloid-derived suppressor cells: immune-suppressive cells that impair antitumor immunity and are sculpted by their environment
.
J Immunol
2018
;
200
:
422
31
.
42.
Klughammer
J
,
Kiesel
B
,
Roetzer
T
,
Fortelny
N
,
Nemc
A
,
Nenning
KH
, et al
The DNA methylation landscape of glioblastoma disease progression shows extensive heterogeneity in time and space
.
Nat Med
2018
;
24
:
1611
24
.
43.
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
.
44.
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
.
45.
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. DOI: 10.1126/scitranslmed.aan0026
.
46.
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
.
47.
Adams
S
,
Gatti-Mays
ME
,
Kalinsky
K
,
Korde
LA
,
Sharon
E
,
Amiri-Kordestani
L
, et al
Current landscape of immunotherapy in breast cancer: a review
.
JAMA Oncol
2019
.
48.
Jones
PA
. 
Functions of DNA methylation: islands, start sites, gene bodies and beyond
.
Nat Rev Genet
2012
;
13
:
484
92
.
49.
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
.
50.
Neri
F
,
Rapelli
S
,
Krepelova
A
,
Incarnato
D
,
Parlato
C
,
Basile
G
, et al
Intragenic DNA methylation prevents spurious transcription initiation
.
Nature
2017
;
543
:
72
7
.
51.
Skvortsova
K
,
Masle-Farquhar
E
,
Luu
PL
,
Song
JZ
,
Qu
W
,
Zotenko
E
, et al
DNA hypermethylation encroachment at CpG Island borders in cancer is predisposed by H3K4 monomethylation patterns
.
Cancer Cell
2019
;
35
:
297
314
.
52.
Denny
SK
,
Yang
D
,
Chuang
CH
,
Brady
JJ
,
Lim
JS
,
Gruner
BM
, et al
Nfib promotes metastasis through a widespread increase in chromatin accessibility
.
Cell
2016
;
166
:
328
42
.
53.
Brown
R
,
Curry
E
,
Magnani
L
,
Wilhelm-Benartzi
CS
,
Borley
J
. 
Poised epigenetic states and acquired drug resistance in cancer
.
Nat Rev Cancer
2014
;
14
:
747
53
.