Epigenetic inactivation of the CpG demethylase TET1 as a DNA methylation feedback loop in human cancers

Promoter CpG methylation is a fundamental regulatory process of gene expression. TET proteins are active CpG demethylases converting 5-methylcytosine to 5-hydroxymethylcytosine, with loss of 5 hmC as an epigenetic hallmark of cancers, indicating critical roles of TET proteins in epigenetic tumorigenesis. Through analysis of tumor methylomes, we discovered TET1 as a methylated target, and further confirmed its frequent downregulation/methylation in cell lines and primary tumors of multiple carcinomas and lymphomas, including nasopharyngeal, esophageal, gastric, colorectal, renal, breast and cervical carcinomas, as well as non-Hodgkin, Hodgkin and nasal natural killer/T-cell lymphomas, although all three TET family genes are ubiquitously expressed in normal tissues. Ectopic expression of TET1 catalytic domain suppressed colony formation and induced apoptosis of tumor cells of multiple tissue types, supporting its role as a broad bona fide tumor suppressor. Furthermore, TET1 catalytic domain possessed demethylase activity in cancer cells, being able to inhibit the CpG methylation of tumor suppressor gene (TSG) promoters and reactivate their expression, such as SLIT2, ZNF382 and HOXA9. As only infrequent mutations of TET1 have been reported, compared to TET2, epigenetic silencing therefore appears to be the dominant mechanism for TET1 inactivation in cancers, which also forms a feedback loop of CpG methylation during tumorigenesis.


Results and Discussion
Epigenomic identification of TET1 as a methylated target in multiple cancers. During our analysis of whole-genome CpG methylation profiles (methylomes) of multiple tumor cell lines and primary tumors 31 , the promoter of one of the CpG demethylases, TET1, turned out to be a target in multiple methylomes (Fig. 1A). Bioinformatics analysis of the methylome data showed significant positive enrichment of CpG methylation (Cut off = 2) at the TET1 promoter and exon 1 region in multiple tumors, including nasopharyngeal carcinoma (NPC) xenografts (C15, C18) and primary tumor (OCT83), esophageal squamous cell carcinoma (ESCC) cell lines (KYSE140, KYSE510), hepatocellular carcinoma (HCC) cell lines (HuH7, HepG2) and primary tumor (418T), as well as nasal NK/T-cell lymphoma (NKTCL) cell lines (SNK6, NK-YS) and primary tumor (NK1) (Fig. 1A). The TET1 promoter and exon 1 region contain a typical CpG island ( Fig. 2A), indicating that CpG methylation most likely regulates its expression in human cells.
We thus further examined the expression and methylation profiles of TET1 in multiple cancers. Results showed that, although all three TET genes (TET1, −2, −3) were ubiquitously expressed in a series of human normal adult and fetal tissues (Fig. 1B), only TET1 neither TET2 nor TET3, was frequently downregulated or totally silenced in a variety of tumor cell lines including multiple carcinomas (nasopharyngeal, esophageal, lung, gastric, colon, breast, cervical, renal) and lymphomas (Hodgkin, non-Hodgkin and NKTCL), while TET1 is readily expressed in all immortalized normal epithelial cell lines of different tissue origins ( Fig. 2 and Suppl. Fig. S1A).
Methylation-specific PCR (MSP) primers for TET1 was tested for not amplifying any not-bisulfited DNA, confirming the detection specificity of TET1 methylation in our study (Fig. 2B). Then by MSP, we detected TET1 promoter methylation in virtually all downregulated cell lines of nasopharyngeal, esophageal, lung, gastric, colon, breast, cervical and renal carcinomas, as well as Hodgkin (HL), non-Hodgkin (NHL) and NKTCL lymphomas, but not in immortalized normal epithelial cell lines (Fig. 2C,D; Table 1). Moreover, TET1 downregulation and methylation were infrequently detected in hepatocellular (HCC) and prostate cancer cell lines but not in the bladder and melanoma cell lines examined (Suppl. Fig. S1B).
We further investigated whether TET1 promoter methylation directly mediates its repression. DNA methyltransferase inhibitor 5-aza-dC (Aza) was used or in combination with histone deacetylase (HDAC) inhibitor to treat tumor cell lines of nasopharyngeal, esophageal, colon, breast and renal, all with methylated and downregulated TET1. After the treatment, restoration of TET1 expression was observed, along with increased unmethylated promoter alleles as detected by MSP (Fig. 3B). Demethylation of the TET1 promoter was confirmed by BGS analysis, which shows dramatically demethylated CpG sites (Fig. 3C), indicating that CpG methylation directly mediates TET1 silencing in tumor cells.
In this study, we demonstrated that epigenetic silencing is a common regulatory mechanism for TET1 inactivation at the transcriptional level in multiple human cancers. Additional alternative mechanisms regulating expression and activities of TET family members have been reported 32 . For examples, high mobility group AT-hook 2 (HMGA2), a chromatin remodeling factor, suppresses TET1 expression by directly binding to its promoter or indirectly through other components in breast cancer cells 24 . Polycomb repressive complex 2 (PRC2) mediates Tet1 downregulation through H3K27me3 histone mark deposition 33 . PARP activity increases TET1 expression levels through maintaining a permissive chromatin state 34 . miR-22 suppresses TET expression levels in breast cancer cells through directly targeting the 3′ -untranslated regions (UTRs) of TET mRNAs 27 . As direct substrates of calpains (calcium-activated cysteine proteases), TET proteins also undergo calpain-mediated degradation 28 . Nuclear exclusion of TET1 and TET2 is significantly correlated with loss of 5mC in glioma and colon Scientific RepoRts | 6:26591 | DOI: 10.1038/srep26591 cancer 29,30 . Thus, TET expression could be regulated at multiple levels of transcription, post-transcription or post-translation in different cell context, although TET1 silencing through promoter CpG methylation appears to be more common and predominant in multiple tumors.      Table 1), but infrequently in primary ESCC, lung, prostate tumors and other non-Hodgkin lymphomas (Suppl. Fig. S2, Table 1). TET1 methylation could even be detected in 50% of 16 nose swab samples from suspected NPC patients (Fig. 4B). In contrast, TET1 methylation was not detected in a panel of human normal adult and fetal tissues except for being barely seen in normal small intestine and colon (Fig. 4C). Further detailed BGS methylation analysis confirmed the presence of methylated promoter alleles in primary tumors but not normal tissues (Fig. 4D). TET1 downregulation was also detected in paired primary tumors of several tissue types (lung, stomach, colon, rectum, breast and kidney) and primary NPC tumors (Fig. 4E). Furthermore, through online GENT and Oncomine database analysis, we found that TET1 mRNA levels were significantly reduced in multiple solid tumors and leukemia, compared with their corresponding normal tissues (Suppl. Fig. S3). These results clearly demonstrate that TET1 silencing by promoter CpG methylation is a common event for multiple tumors of epithelial and lymphoid origins. Several studies have shown that TET genes are readily expressed in normal esophageal, gastric, colon, liver and breast tissues by PCR or immunohistochemistry 22,23,25 , but decreased in tumor cell lines and primary tumors to varied grades, with TET1 as the most significantly downregulated member. A previous report through analyzing Cancer Genome Atlas TCGA database found that TET1 is downregulated in primary tumors of colorectal, breast and lung since early stage, and associated with patient poor survival 23 . TET1 is significantly decreased at mRNA and protein levels in gastric primary tumors compared to surgical margins and associated with tumor localization and TNM grades 35 . DNA methylation and bivalent histone marks at the CpG island 3′ -shore mediate TET1 silencing in gastric cancer 36 . Reduced TET1 expression or 5 hmC level in breast cancer tissues could be biomarkers for breast cancer progression 37 . TET1 methylation in colorectal cancer tissues, not TET2 and TET3 38 , has been found as an early event in CRC tumorigenesis, thus as a valuable biomarker for metastasis prediction 39 . Our results are consistent with these previous studies. TET1 methylation appears to be tumor-specific and thus could serve as a potential epigenetic biomarker for cancer detection. Genetic alteration of TET1 is uncommon in human cancers. As alterations of cancer gene are through either genetic or epigenetic mechanisms, we further investigated possible genetic alterations of TET1 in cancers. Somatically acquired mutations of TET1 in human cancers were analyzed using the COSMIC database. Only < 1% of tumor cases (most cases with ≤ 0.25%) had detectable TET1 mutations (Fig. 5A), consisting of 80% of missense mutations, 10% of nonsense and 10% of synonymous mutations (Fig. 5B), with most of the mutations located in coding regions (Fig. 5C). We also detected hemizygous deletion of TET1 in some tumor cell lines with TET1 silencing and methylation, but not in TET1-expresssing cells (Suppl. Fig. S4A,B). Consistently, TET1 gene deletion was also observed in solid tumors by analyzing DNA copy number alterations using the Oncomine database (Suppl. Fig. S4C). These results demonstrate that TET1 mutation is uncommon in human cancers, although TET1 deletion is indeed present in some tumor samples.

TET1 functions as a tumor suppressor which requires its catalytic activity.
The TET1 catalytic domain (CD) (containing the Cys-rich and DSBH regions) remains intact hydroxylase activity in embryonic development and reprogramming 6,13 , displaying ability to induce 5 hmC formation, demethylation and gene transcription in differentiated cells 33 . We test whether the catalytic activity of TET1 was required for its possible tumor suppression functions, using TET1-CD and its enzymatic dead mutant (TET1-CD-mut) (Fig. 6A). Ectopic expression of TET1-CD significantly suppressed tumor cell clonogenicity (to ~40-50% of control cells) in colony formation assays of NPC, ESCC, gastric, colon and breast tumor cells, while the TET1-CD-mut lost this ability (Fig. 6B). TUNEL assay showed significantly increased numbers of apoptotic cells in TET1-CD expressing-tumor cells, compared with vector or TET1-CD-mut controls (Fig. 6C). These results demonstrate that TET1 possesses bona fide tumor suppressive functions in tumor cells of multiple types.
Consistent with our results, several recent studies have shown similar tumor-suppressive functions of TET1 in cancer cells. TET1 inhibits proliferation and invasion of colon 23 , breast 24,25 , renal 40 and prostate 25 cancer cells in vivo and in vitro. TET1 deficiency promotes B-lineage differentiation, leading eventually to B-cell lymphoma 41 . TET1 suppression as a key event of the RAS programming is required for KRAS-induced cellular transformation 26 . Thus, loss of function of TET1 is a common event during multiple tumorigenesis of solid tumors or hematologic malignancies.

TET1 induces TSG promoter demethylation in tumor cells. Several studies identified TET1 target
genes in mouse ES cells and some tumor cells, using RNA-or ChIP-sequencing or hydroxymethylated DNA immunoprecipitation sequencing (hMeDIP-seq) 12,24,26,27,33,[42][43][44][45] . A series of TET1-targeted genes including TSGs have been identified, such as TIMP 25 , HOXA9 and HOXA7 24 , and Wnt signaling antagonists DKK3 and DKK4 23 . To further explore the molecular mechanism of TET1 in tumor suppression, we examined some known and potential target TSGs to assess the demethylase activity of TET1 in tumor cells. Mild upregulation of HOXA9, HOXA5, PCDH7, TCF4, MEIS1, SLIT2 and ZNF382 at mRNA levels was observed in TET1-CD-expressing carcinoma cells by semi-quantitative RT-PCR (Fig. 6D) and qRT-PCR (Fig. 6E). Meanwhile, we also detected decreased methylated alleles of HOXA9, SLIT2 and ZNF382 promoters in TET1-CD-expressing tumor cells, but not in TET1-CD-mut-expressing cells, with increased unmethylated promoter alleles observed concurrently, suggesting that TET1 indeed functions as a CpG demethylase to demethylate and reactivate multiple TSGs in tumor cells (Fig. 6F). In addition to HOXA9, we also found that TSGs like SLIT2, ZNF382, PCDH7, TCF4, MEIS1 and HOXA5 as TET1 target genes which could be demethylated and reactivated by TET1 in tumor cells. Other mechanisms besides demethylase activity could also be involved in regulating target genes by TET1, such as recruiting PRC2 42 , PRDM14 43 , Sin3A co-repressor complex 44 and MBD3/NURD complex 45 . Further studies on TET1-targeted gene regulation in human cancers would help us to understand more of its role in cancer development.
The discovery of TET enzymes, in addition to DNMTs, establishes a fundamental etiologic role of CpG methylation in human cancers. In response to environment carcinogens 46-48 like chemical carcinogens and tumor viruses, DNMT activities and expression levels are induced and increased in cells, displaying stronger maintenance and de novo methylation capacity, leading to specific gene CpG island hypermethylation. The epigenetic alterations, especially promoter CpG methylation of TSGs, facilitate genome instability, disrupted cellular signaling and even further genetic mutations, thus are crucial to tumor initiation and progression 1,49 . Remarkably, promoter CpG methylation-mediated silencing of the CpG demethylase TET1 in human cancers, which in turn, further leads to increased 5 mC levels in tumor cells, thus forming a DNA methylation feedback loop mediated by DNMT/CpG methylation and TET1 (Fig. 7).
In summary, our study comprehensively examined TET1 expression and methylation status in multiple tumors, and demonstrated that promoter CpG methylation is a predominant mechanism for TET1 inactivation in human cancers. The tumor-specific methylation of TET1 could serve as a valuable, epigenetic non-invasive biomarker. TET1 as a tumor suppressor and CpG demethylase in tumor cells requires its intact catalytic domain, which provides new insight into the epigenetic master role of TET1 in tumor pathogenesis. Our findings enlighten us on the mechanistic elucidation of the importance of CpG methylation in human cancers.

Material and Methods
Cell lines and tissue samples. Human tumor cell lines of multiple tissue types were used [50][51][52][53][54][55]   Hodgkin (HL) and nasal natural killer (NK)/T-cell (NKTCL) lymphomas. Immortalized, non-transformed normal epithelial cell lines were used as "normal" controls. Cell lines were obtained from either American Type Culture Collection or collaborators. When needed, cell lines were treated with 10 μ mol/L 5-aza-2′ -deoxycytidine (Aza) (Sigma-aldrich, St Louis, MO) for 3 days, without or with further treatment with 100 nmol/l trichostatin A (TSA) (Cayman Chemical Co., Ann Arbor, MI) for additional ~16 h as previously 50,53 . Normal adult and fetal tissue RNA and DNA samples were purchased commercially (Stratagene, La Jolla, Ca; Millipore-Chemicon, Billerica, Ma). DNA samples of primary carcinomas, nose swab from suspected NPC patients, as well as surgical margin normal tissues, have been described previously 31,51,52 . Establishment of tumor methylomes by MeDIP-chip. Methylated DNA immunoprecipitation (MeDIP) coupled with promoter microarray hybridization was performed as previously 31 . Briefly, immunoprecipitation of methylated DNA was performed using monoclonal antibody against 5-methylcytidine (33D3, Diagenode, Seraing, Belgium) labeled with magnetic beads. Total input and immunoprecipitated DNA were labeled with Cy3 or Cy5, respectively, and hybridized to NimbleGen ™ HG18 Meth (385K CGI plus) promoter arrays or HG19 (2.1 M) Deluxe Promoter arrays (Array Star, Inc., MD). Normal epithelial cell lines and normal tissues were used as controls. Bioinformatics analysis of methylome data was performed as previously 31 .

Semi-quantitative RT-PCR and quantitative real-time PCR (qRT-PCR). Semi-quantitative RT-PCR
and quantitative real-time PCR were performed as described before 50,53 , with GAPDH as a control for all the samples shown in our previous publications 31,51,52 . qRT-PCR was carried out according to the manufacturer's protocol (HT7900 system; applied Biosystems), with SYBR Green master mix (applied Biosystems) used. Primers used are listed in Supplementary Table S1.
Bisulfite treatment of DNA samples and promoter methylation analysis. CpG island (CGI) analysis for TET1 promoter and exon 1 was performed using CpG island Searcher (http//ccnt.hsc.usc.edu/cpgis-lands2). Bisulfite modification of genomic DNA was carried out as described previously 56,57 . For MSP analysis, approximately 50 ng of bisulfited DNA for each sample was amplified with methylation-or unmethylation-specific primer set, according to our previous MSP protocol 58 . Bisulfite-treated DNA was also amplified using a set of BGS primers, then cloned into pCR4-TOPO vector (Invitrogen, Carlsbad, Ca), with 8-10 clones randomly picked and sequenced. MSP and BGS primers used are shown in Supplementary Table S1. Unmethylated gene alleles for these treated samples have been detected in our previous publications, which shows the good quality of these DNA samples 31,51,52 . Genetic deletion analysis for TET1. Homozygous deletion of TET1 coding exons 2 and 4 was examined using multiplex genomic DNA PCR, as previously described 51 . Primer sequences are shown in Supplementary  Table S1.
Colony formation assay of tumor cells. Human TET1 catalytic domain (TET1-CD) cDNA and its catalytic domain mutant (TET1-CD-mut) clones (Addgene, Cambridge, MA) were used as templates to generate TET1 constructs with an N-terminal Flag tag, and subcloned into pcDNA3.1 vector (Invitrogen, Carlsbad, Ca). Cells were cultured overnight in a 12-well plate and transfected with empty vector or TET1-CD, TET1-CD-mut-expressing plasmids using Lipofectamine 2000 (Invitrogen, Carlsbad, Ca). Forty-eight hours later, transfectants were replated in triplicate and cultured for 10-15 days in complete medium containing G418. Surviving colonies were stained with crystal violet (0.5% w/v) after methanol fixation, with visible colonies (≥ 50 cells) counted. TUNEL assay. Cells cultured on coverslips were fixed with 4% paraformaldehyde, and permeabilized with 0.1% triton X-100. TUNEL (Terminal deoxynucleotidyl transferase dUTP nick end labeling) staining was performed using the In Situ Cell Death Detection Kit (Roche, Mannheim, Germany).
Statistical analysis. Student's t-tests were performed. All reported p-values were two-sided, and p < 0.05 was considered statistically significant.