FGFR3△7–9 promotes tumor progression via the phosphorylation and destabilization of ten-eleven translocation-2 in human hepatocellular carcinoma

Overexpression of fibroblast growth factor receptor 3 (FGFR3) correlates with more severe clinical features of hepatocellular carcinoma (HCC). Our previous study has shown that FGFR3∆7–9, a novel splicing mutation of FGFR3, contributes significantly to HCC malignant character, but the epigenetic mechanism is still elusive. In this study, through mass spectrometry and co-immunoprecipitation studies, we discover a close association between FGFR3∆7–9 and the DNA demethylase Ten-Eleven Translocation-2 (TET2). Unlike other certain types of cancer, mutation of TET2 is rare in HCC. However, activation of FGFR3∆7–9 by FGF1 dramatically shortens TET2 half-life. FGFR3∆7–9, but not wild-type FGFR3, directly interacts with TET2 and phosphorylates TET2 at Y1902 site, leading to the ubiquitination and proteasome-mediated degradation of TET2. Overexpression of a phospho-deficient mutant TET2 (Y1902F) significantly reduces the oncogenic potential of FGFR3∆7–9 in vitro and in vivo. Furthermore, FGFR3∆7–9 significantly enhances HCC cell proliferation through the TET2-PTEN-AKT pathway. Specifically, TET2 offsets the elevation of p-AKT level induced by FGFR3∆7–9 through directly binding to PTEN promoter and increasing 5-hmC. Therefore, through phosphorylation and inhibition of TET2, FGFR3∆7–9 reduces PTEN expression and substantiates AKT activation to stimulate HCC proliferation. Together, this study identifies TET2 as a key regulator of the oncogenic role of FGFR3∆7–9 in HCC carcinogenesis and sheds light on new therapeutic strategies for HCC treatment.


Introduction
Liver cancer is one of the most common cancers, and is the fourth lethal malignancy worldwide 1 . Moreover, the incidence of liver cancer continues to increase by 2 to 3% annually during 2007-2016 in United States 2 . Both genetic and epigenetic alterations play critical roles in HCC. In our previous study, we demonstrated that overexpression of fibroblast growth factor receptor 3 (FGFR3) plays an important role in HCC development 3 . Moreover, we have shown that FGFR3 Δ7-9, a novel mutant transcript, links exon 6 to exon 10 directly and promotes the proliferation, migration, and metastasis of HCC cells both in vitro and in vivo 4 .
DNA methylation, which occurs mainly on the 5th carbon atom of cytosine (5-methylcytosine, 5-mC) in the context of CpG islands, has an important role in normal development and tumor development 5 . DNA methylation and demethylation are important for regulating chromosome structure, as well as gene expression. Ten-Eleven Translocation-2 (TET2) is a member of the dioxygenase family, which converts 5-mC to 5-hydroxymethylcytosine (5-hmC) and leads to demethylation at CpG islands 6 . TET2 is one of the most frequently mutated genes in hematopoietic malignancies, and its disruption is an early event in the onset of disease 7 . Patients carrying TET2 mutations often show significantly reduced global 5-hmC levels 8 . Decreased expression of TET proteins and lower 5-hmC levels are also general hallmarks of many solid cancer types, including liver, breast, lung, gastric, prostate, and breast cancer, as well as glioblastoma and melanoma [9][10][11][12] . Moreover, the decrease in 5-hmC levels can be attributed to the high proliferation rate of cancer cells 13 . However, the detailed regulation mechanisms of TET2 in HCC are unclear.
Searching for proteins that interact with and mediate the function of FGFR3 Δ7-9 , we found that that FGFR3 Δ7-9 can significantly enhance proliferation in HCC cells through the TET2-PTEN-AKT signaling pathway. We demonstrated that TET2 was significantly downregulated in HCC, and lower TET2 expression levels were associated with poor prognosis. Moreover, FGFR3 Δ7-9 could directly interact with and phosphorylate TET2 at Y1902, which led to the ubiquitination and proteosomal degradation of TET2. TET2 downregulation by FGFR3 Δ7-9 attenuated PTEN expression, which strengthened AKT phosphorylation and promoted the proliferation and survival of HCC cells. The functional link between FGFR3 Δ7-9 and tumor suppressor TET2 revealed in our study may help develop new therapeutic strategies for HCC treatment.

Results
FGFR3 Δ7-9 interacts with TET2 and phosphorylates TET2 at Y1902 Our previous study found that FGFR3 Δ7-9 has a stronger ability to promote the progression of HCC cells than wild type FGFR3 4 . TET2 was one of the potential downstream substrates of FGFR3 Δ7-9 by tandem mass spectrometry (data not shown). In a co-immunoprecipitation assay with ectopically expressed FGFR3 Δ7-9 and TET2, we found that HA-tagged FGFR3 Δ7-9 could precipitate myc-TET2 from transfected 293T cells (Fig. 1a). Conversely, myc-TET2 could also pull down His-FGFR3 (Fig. 1b). In addition, the physical interaction between TET2 and FGFR3 Δ7-9 was supported by experiments conducted in an HCC cell line (SMMC-7721) stably expressing FGFR3 (WT) or FGFR3 Δ7-9 . Our results demonstrated that FGFR3 Δ7-9 , but not wild type FGFR3, was able to coimmunoprecipitate with endogenous TET2 in a reciprocal manner (Fig. 1c, d).
To further identify the binding site of TET2 to FGFR3 Δ7-9 , we first truncated TET2 into fragments T1, T2, T3, and T4 out of 2002 amino acids (Fig. 1e). All four fragments could not interact with FGFR3 Δ7-9 , and only full length TET2 may mediate binding (Fig. 1f). Hence, the key binding site of TET2 to FGFR3 Δ7-9 was narrow-downed to1887-2002 fragment. Considering the distinctive role of FGFR3 Δ7-9 as a tyrosine kinase, TET2 may be tyrosine phosphorylated by FGFR3 Δ7-9 and we verified this hypothesis using a multistage mass spectrometry method. As expected, in this fragment, one solo tyrosine phosphorylation site of TET2 was identified at Y1902 (Fig. 1g). Then, we mutated TET2 Y1902 to Y1902F to pinpoint the speculated binding site to FGFR3 Δ7-9 of TET2. As expected, the Y1902F mutation of TET2 abolished the interaction between TET2 and FGFR3 Δ7-9 (Fig. 1h). More importantly, FGFR3 Δ7-9 could significantly increase the tyrosine phosphorylation of wild-type TET2 but not the Y1902F mutant (Fig. 1i). Hence, we confirmed Y1902 was the phosphorylated site of TET2 by FGFR3 Δ7-9 . In addition, when FGFR3 Δ7-9 was silenced by siRNA, the tyrosine phosphorylation of TET2 by FGFR3 Δ7-9 could be apparently abolished (Fig. 1j). When FGF1, ligand of FGFR3, was further included in system, the tyrosine phosphorylation of TET2 could be increased consequently (Fig. 1k). Altogether, our findings indicated that Y1902 is the phosphorylation site of TET2 by FGFR3 Δ7-9 .
FGFR3 Δ7-9 negatively regulates TET2 stability and promotes its ubiquitination The above results indicated the interaction between FGFR3 Δ7-9 and TET2. To further study the effect of FGFR3 Δ7-9 on TET2, we first determined whether FGFR3 Δ7-9 affected the expression of TET2. We examined the expression of TET2 at both the mRNA and protein levels by qPCR and immunoblotting, respectively, upon overexpression of FGFR3 Δ7-9 in SMMC-7721 and HepG2 cell lines. We found that both the mRNA and protein levels of TET2 were not altered by FGFR3 (Fig. 2a,  b). However, although TET2 mRNA level was barely affected by FGFR3 Δ7-9 , TET2 protein level was significantly decreased upon overexpression of FGFR3 Δ7-9 in both HCC cells (Fig. 2c, d). Meanwhile, downregulation of TET2 protein by FGFR3 Δ7-9 overexpression could be reversed by FGFR3 Δ7-9 knockdown (siFGFR3 Δ7-9 ) in Fig.  2e. On the other hand, activated by FGF1, FGFR3 Δ7-9 could further inhibit TET2 protein expression (Fig. 2f).
Based on above results, we used inhibitors of lysosome (NH 4 Cl), calpain (Calpeptin), and caspase (Z-VAD-FAK) to investigate potential mechanism. Results in Fig. 2K showed that none of these treatments could affect the accelerated TET2 degradation by FGFR3 Δ7-9 . Instead, treatment with MG132, an inhibitor of the proteasome, effectively blocked TET2 half-life shortening by FGFR3 Δ7-9 activation, indicating that FGFR3 Δ7-9 -mediated degradation of TET2 protein might be involved in the ubiquitin-proteasome pathway. Based on above data, we found that FGFR3 Δ7-9 promoted TET2 ubiquitylation (Fig. 2L). Given the possible phosphorylation site of TET2 by FGFR3 Δ7-9 in Y1902, we then found TET2 Y1902F mutation could abolish the ability of FGFR3 Δ7-9 to promote the ubiquitylation of TET2 (Fig. 2M). Together, these data strongly suggested that FGFR3 Δ7-9 could promote TET2 ubiquitin-proteasome degradation in a Y1902 phosphorylation-dependent manner.

TET2 expression is downregulated in HCC tissues and predicts poor prognosis in HCC patients
To explore the effects of downregulation of TET2 induced by FGFR3 Δ7-9 , we firstly investigated the expression of TET2 and the relationship of FGFR3 in HCC. Based on the TCGA transcriptomic data, a number of cancer types, including HCC, we observed that lower TET2 expression levels in cancer tissues when compared with matched normal tissues (Fig. 3A, B). In our data panel, mRNA expression of TET2 was measured in 32 pairs of HCC and patient-matched normal tissues by quantitative RT-PCR (qRT-PCR). Consistent with the TCGA dataset, as shown in Fig. 3C, TET2 mRNA levels were significantly lower in HCC tissues than in normal tissues. Lower expression of TET2 in tumors was further confirmed at the protein level by western blot analyses (Fig. 3D). To investigate the role of TET2 in HCC progression, we determined the expression levels of TET2 in HCC clinical samples by immunohistochemistry (IHC) staining of tissue microarrays (TMAs), which contain 78 pairs of tumor and adjacent normal tissues. In our cases, positive expression of TET2 was detected in 20 (25.6%) of the tumor tissues, while only 50 (43.1%) of the adjacent normal specimens showed a positive signal (P = 0.009) ( Table 1). From this, we confirmed that TET2 did be downregulated in HCC tissues in clinical scenario. Clinicopathological features analyses showed that decreased IHC signal of TET2 was correlated with larger tumor size (P = 0.036, Table 1). However, no significant correlation was observed between TET2 expression and other parameters, such as gender, age, TNM stage, distant metastasis, or relapse (P > 0.05, Table 1). Regarding the clinical prognosis, lower TET2 expression levels were associated with a worse 3-year overall survival (P = 0.035, Fig. 3E). IHC staining also indicated that the level of TET2 is lower in HCC samples with FGFR3 Δ7-9 ( Fig. 3F, P = 0.035). Overall, our clinical data suggested that downregulation of TET2 is correlated with poor prognosis and the protein level of TET2 was downregulation in HCC with FGFR3 Δ7-9 .
FGFR3 Δ7-9 decreases the level of PTEN and TET2 restores PTEN expression PTEN is critical for inhibiting cancer cell migration, invasion and proliferation. We observed lower PTEN expression levels in HCC tumor tissues versus corresponding normal tissues in the TCGA data, and low PTEN mRNA levels were associated with a worse overall (see figure on previous page) Fig. 2 FGFR3 Δ7-9 negatively regulates TET2 stability. A qPCR analysis of TET2 expression in SMMC-7721 and HepG2 cells infected with empty vector or FGFR3-expressing plasmid. B Western blot analysis of TET2 expression in SMMC-7721 and HepG2 cells infected with empty vector or FGFR3expressing plasmid. C qPCR analysis of TET2 expression in SMMC-7721 and HepG2 cells infected with empty vector or FGFR3 Δ7-9 -expressing plasmid. D Western blot analysis of TET2 expression in SMMC-7721 and HepG2 cells infected with empty vector or HA-FGFR3 Δ7-9 -expressing plasmid. E Western blot analysis of TET2 expression in SMMC-7721/FGFR3 Δ7-9 and HepG2/FGFR3 Δ7-9 after indicated treatment; F Western blot analysis of TET2 expression in SMMC-7721/FGFR3 Δ7-9 and HepG2/FGFR3 Δ7-9 in the presence and absence of FGF1 G SMMC-7721 cells were treated with 20 μg/ml CHX, and whole-cell lysates (WCL) were collected at the indicated time points for immunoblot analysis. H Semiquantification with GAPDH as a loading control. Relative TET2 levels at time 0 were set as 100%. I SMMC-7721/FGFR3 Δ7-9 cells were treated with 20 μg/ml CHX in the presence or absence of FGF1, and whole-cell lysates (WCL) were collected at the indicated time points for immunoblot analysis. J Semiquantification with GAPDH as a loading control. Relative TET2 levels at time 0 were set as 100%. K FGFR3 Δ7-9 -mediated TET2 degradation is blocked by the proteasome inhibitor, MG132. SMMC-7721 cells were transfected with HA-FGFR3 Δ7-9 and then treated with inhibitors of lysosome (NH 4 Cl, 20 mM), calpain (Calpeptin, 50 μM), caspase (Z-VAD-FMK, 100 μM), or 26S proteasome (MG132, 10 μM) for 24 h, followed by immunoblotting analyses. L SMMC-7721/FGFR3 Δ7-9 and control cells were treated with MG-132 (10 μM for 24 h). Cell lysates were immunoprecipitated with antibodies against TET2 or IgG. Ubiquitination levels were analyzed by western blot. M Cells were transfected with indicated plasmids. TET2 ubiquitylation was examined by coupled IP-western. F Statistical analysis of the correlation between the levels of TET2 and FGFR3 Δ7-9 (P = 0.035; P value was obtained using a Pearson χ2 test). Fig. S1A and S1B). Moreover, lower PTEN protein levels also showed a worse disease free survival and overall survival (analyzed through the TRGAted: https://nborcherding.shinyapps.io/TRGAted/) (Supplementary Fig. S1C and S1D). PTEN loss is inversely correlated with constitutive activation of the PI3K/AKT signaling pathway 14 , indicating PTEN inhibits the activation of the PI3K/AKT pathway.

survival (Supplementary
Overexpression of FGFR3 significantly downregulated the expression of PTEN and increased the level of p-AKT, and these effects were more remarkable in FGFR3 Δ7-9 than in wild type FGFR3 (Fig. 5a). Meanwhile knockdown TET2 further increased the level of p-AKT and decrease the level of PTEN in present of FGFR3 Δ7-9 (Supplementary Fig. S2). In Fig. 5b, we found that TET2 (Y1902F) could significantly increase PTEN levels and decrease p-AKT levels. It revealed that it was TET2 (Y1902F) that could affect the downregulated of PTEN and upregulation of p-AKT induced by FGFR3 Δ7-9 .
TET2 upregulates PTEN expression via increasing 5-hmC level in PTEN promoter area From above, we found that TET2 (Y1902F) dramatically increased PTEN levels, and we then explored the possible underlining mechanisms. Based on the TCGA database, we found that PTEN methylation level is higher in tumor than normal tissues ( Supplementary Fig. S3A), and PTEN methylation level is negative related to PTEN expression ( Supplementary Fig. S3B). Moreover, the methylation level of PTEN is associated with poor prognosis (Supplementary Fig. S3C). Therefore, regulation of methylation level of PTEN is one of mechanisms for altering the expression of PTEN in HCC. Furthermore, TCGA data analyses demonstrated that PTEN was positively correlated with TET2 (Fig. 5e), and TET2 (Y1902F) could significantly increase PTEN mRNA levels (Fig. 5F). TET2's function was made possible by regulating its targets' 5-hmC level in promoter area, and PTEN is canonically modulated by promoter methylation status. Consequently, we analyzed the 5-hmC content in the promoter region of PTEN. Evidenced by quantification of 5-hmC levels in genomic DNA by methylation-sensitive qPCR, we found that knockdown of TET2 decrease the level of 5-hmc in PTEN promoter region ( Supplementary  Fig. S4A). Furthermore, we found that TET2 (Y1902F) induced much higher levels of 5-hmC in PTEN promoter region compared with its mock and TET2 (WT) counterparts (Fig. 5g). Meanwhile, the PTEN promotor methylation level was lowest in the TET2 (Y1902F) group ( Supplementary Fig. S4B). Furthermore, the results from chromatin immunoprecipitation (ChIP) assay definitely confirmed the direct binding of TET2 to the PTEN promoter using three sets of primers (Fig. 5h). Moreover, we found cell proliferation increased by FGFR3 Δ7-9 could be inhibited by LY294002, an inhibition of PI3K/AKT pathway (Fig. 5i). We also used Wortmannin, a specific and irreversible inhibitor for PI3K/AKT, and found the consist result with LY294002 ( Supplementary Fig. S5). In all, our results indicated that TET2 could directly bind to the PTEN promoter, then increase its 5-hmC level and boost PTEN transcription consequently.

Discussion
FGFR family of proteins have an extracellular ligandbinding domain, a transmembrane domain, and a split intracellular kinase domain 15 . Members of the FGFR family can bind to FGFs, resulting in receptor dimerization, autophosphorylation, and downstream signal transduction 16 . Investigations demonstrated that FGFRs might play important roles in a variety of processes, including cellular proliferation, differentiation, and angiogenesis [17][18][19] . In our previous study, we found that, among FGFR family members, only FGFR3 was elevated in HCC specimens, which is positively correlated with poor clinicopathologic parameters 3 .
Genome Instability and Mutation is one of the hallmarks of Cancer 20 . Activating mutations of FGFR3 play important roles in the progression of chondrosarcoma, superficial bladder tumors and urothelial cell carcinomas 21 . Several novel mutant transcripts caused by aberrant splicing and activation of cryptic splice sequences have been reported in digestive tract tumors 22 . In our previous study, we identified a novel transmembrane mutation of FGFR3, FGFR3 Δ7-9 , which lacks exons encoding the immunoglobulin-like III domain related to bind FGFs. FGFR3 Δ7-9 showed a much stronger affinity to FGFs than wild-type FGFR3, and promoted the proliferation, migration, and metastasis of HCC cells both in vitro and in vivo 4 . FGFR3 Δ7-9 performs aberrant ligand binding ability and self-activity independent FGF1 stimulation. Besides, FGFR3 Δ7-9 also shows aberrant binding and phosphorylating ability. We further identified an FGFR3 Δ7-9 's aberrant binding protein, TET2. Emerging evidence suggests that TET families may play unique roles in many biological processes, such as gene control mechanisms, DNA methylation, and involvement in many diseases, especially cancer. As a tumor suppressor, loss of function of TET2 caused by mutation or deletion has previously been reported in several types of cancer 7,23-25 . Although loss of 5-hmC has been observed in various cancers 9,26,27 , TET2 mutations are only frequently detected in leukemia. Using cBio Cancer Genomics Portal (cBioPortal) databases, we also found that mutation of TET2 was frequent in cancers, such as uterine cancer, acute myeloid leukemia, and melanoma ( Supplementary Fig. S6A). However, the data from 373 individual HCC patients in the TCGA dataset showed that <1% of specimens had TET2 mutations (3 out 373) (Supplementary Fig. S6B). Therefore, unlike other certain type of cancer, mutation of TET2 was a rare event in HCC. IDH1/2 mutations resulting in production of oncometabolite D-2-HG inhibits TETs and many other epigenetic enzymes, but are also highly restricted to gliomas and leukemia 28,29 . Our data now suggest altered post-translational modification as another mechanism for deregulating TET2 functions in HCC. Several pathways regulating TETs protein stability have been reported, such as the caspase 30 and calpains 31 (see figure on previous page) Fig. 4 Phosphorylation at the Y1902 site of TET2 increases HCC cell proliferation in vitro. A-B Representative images of clone formation on plastic plates with SMMC-7721 FGFR3Δ7-9 and HepG2 FGFR3Δ7-9 cell lines expressing TET2 (wild type and Y1902F mutant) or empty vector. Quantification of cell growth is shown in B (n = 3, independent t-test, *P < 0.05, **P < 0.01). C CCK-8 cell proliferation assay for SMMC-7721 FGFR3Δ7-9 /HepG2 FGFR3Δ7-9 cells that stably express TET2 (WT) or TET2 (Y1902F). Cells transfected with empty vector were used as a control. D Cell proliferation of indicated sublines, measured by EdU incorporation (40×). E The ratio of apoptotic cells in indicated sublines. *P < 0.05, **P < 0.01. F-G Indicated sublines were serum-deprived (starved) for 24 hours and then maintained in culture medium with 10% FBS (released) for 0, 6 or 12 h. **P < 0.01.
pathways. Our data show that TET2 stability could also be regulated by the ubiquitin-proteasome pathway in HCC. FGFR3 Δ7-9 significantly reduced the protein level of TET2, rather than affected the mRNA level of TET2, and we found it was the ubiquitin-proteasome pathway that got involved in FGFR3 Δ7-9 -mediated TET2 degradation. Furthermore, we found that FGFR3 Δ7-9 , but not wild-type FGFR3, could directly interact with TET2 and phosphorylate the Y1902 site to destabilize TET2.
Aberrant promoter hypermethylation leading to inappropriate transcriptional silencing of tumor suppressor genes is often found in various human neoplasms, including HCC, colorectal and gastric cancers [32][33][34] . TET2 is one member of the TET family that enzymatically converts 5-methylcytosine to 5-hmC, which is an intermediate of DNA demethylation 35 . The results presented here demonstrate that TET2 is an inhibitor of FGFR3 Δ7-9 . Meanwhile, TET2-Y1902F mutation exhibited a more remarkable inhibition of FGFR3 Δ7-9 . Oncogenic FGFR3 signaling occurs through the PI3K/AKT, Ras/Raf/ MEK/ERK, and JAK/STAT pathways [36][37][38] . FGFR3 Δ7-9 more potently induced phosphorylation of the ERK and AKT kinases, leading to abnormal downstream signaling through the ERK and PI3K/AKT/mTOR pathways. In this study, we found that FGFR3 Δ7-9 could significantly activate the PI3K/AKT and ERK pathways, and inhibited the expression of PTEN. However, TET2 could promote the re-expression of PTEN. Moreover, TET2 could reduce the elevated p-AKT induced by FGFR3 Δ7-9 , which depended on the upregulation of PTEN. We further confirmed that TET2 could directly bind to the PTEN promoter and increase 5-hmC. Therefore, through the phosphorylation and inhibition of TET2, FGFR3 Δ7-9 reduced PTEN expression, and then rescued the level of p-AKT to trigger the cell growth advantage.
In summary, by exploring the signaling pathways unique to HCC and absent from other types of cancer, we demonstrated, for the first time, that FGFR3 Δ7-9 phosphorylates the tumor suppressor TET2 at Y1902, which promotes TET2 ubiquitination and destruction. Thus, the dysregulation of TET2 targets, such as PTEN, promotes HCC progression. These findings have revealed important aspects of the functional roles of oncogenic FGFR3 Δ7-9 signaling in HCC and may contribute to designing new therapeutic strategies to treat hepatic cancer.

Materials and methods
Patients, tissues specimens, and cell culture Seventy-eight sets of HCC tissues and adjacent normal liver tissues (confirmed by pathology) were collected from patients who underwent curative surgery in Ruijin Hospital Shanghai Jiao Tong University School of Medicine from Jan. 2014 to Dec. 2015. Moreover, another two cohorts collected between 2016 and 2017 in accordance with the same protocol were also used in this study for qRT-PCR or western blot. Written informed consent was obtained and the research was approved by the local Ethics Committee and the Institutional Review Board. All patients were fully informed of the experimental procedures and were made aware of the potential risks and complications of the proposed treatment scheme. Patient data including demo-graphics, operative procedures, pathology, and complications, were prospectively collected. The human HCC cell lines SMMC-7721 was obtained from Shanghai Institute of Biochemistry and Cell Biology, Chinese Academy of Sciences. HepG2 and human embryonic kidney cell 293T were purchased from ATCC. HepG2 cells were grown in Eagle's minimum essential medium, SMMC-7721 cells were cultured in RPMI 1640 medium and 293T cells were maintained in DMEM. Culture medium was added 10% FBS, 100 IU/ml penicillin and 100 IU/ml streptomycin.

RNA isolation and quantitative real-time PCR
Total RNA was extracted using Trizol reagent (Ther-moFisher, 15596026) and complementary DNA was synthesized with the Reverse Transcription system (Toyobo, FSQ-101) according to the manufacturer's instructions. Real-time PCR was performed using SYBR Green PCR Master Mix (ThermoFisher, 4367659). The primers used for amplification are listed in Supplementary Table S1.

Enzymatic chromatin immunoprecipitation assay
Enzymatic chromatin immunoprecipitation (ChIP) assays were performed using the Enzymatic ChIP Kit (Cell Signaling Technology, #9005) according to the manual. Briefly, HCC cells were cross-linked in 1% formaldehyde solution for 10 min at room temperature, followed by the addition of 125 mM of glycine for 5 min. Antibodies including anti-TET2 (Cell Signaling Technology, #18950) and normal IgG were used for each immunoprecipitation. Immunoprecipitated and input DNAs were subjected to qRT-PCR analyses. The primers used for amplification are listed in Supplementary Table S1.

Immunoprecipitation and immunoblot analyses
Immunoprecipitation analyses were performed using the Direct Magnetic IP/Co-IP Kit (ThermoFisher, 88828) according to the manual. Briefly, indicated antibodies were bound to the beads for 60 min. Cell lysates were incubated with antibody-bound beads overnight at 4°C. Washed the beads twice with Wash Buffer and once with ultrapure water. Eluted bound antigen.

Cell viability and colony formation assay
The Cell Counting Kit-8 (CCK-8, Dojindo, Kumamoto, Japan) was used for the cell proliferation assay. In brief, cells were plated in 96-well plates at 2000 cells per well (four biological replicates) and incubated at 37°C with 5% CO 2 . Cell viability was quantified by measuring OD450 every 24 h using a microplate reader (Epoch; BioTek, Winooski, VT). For the colony formation assay, 1000 cells were plated per well in 6-well plates and cultured at 37°C with 5% CO 2 for 14 days. Colony formation was detected by staining with 0.1% crystal violet in methanol for 30 min. Data were obtained from three independent experiments.

Apoptosis detection and cell cycle analyses
Flow cytometric assays of apoptosis and cell cycle were performed as previously described 39 . Briefly, both attached and floating cells were harvested, washed twice with ice-cold PBS and suspended in 100 µl binding buffer. Cells were incubated with 3 µl FITC-Annexin V and 5 µl PI at room temperature for 15 min in the dark. Next, an additional 300 μl 1× Binding Buffer was added to each sample, and then apoptosis was analyzed by flow cytometry (FACSCalibur; Becton Dickinson, Sparks, MD) according to the manufacturer's instructions. For cell cycle analysis, singlecell suspensions were fixed with 70% cold ethanol at 4°C overnight. Afterwards, samples were washed twice with cold PBS, incubated with PI (50 µg/ml) at 37°C for 30 min in the dark, and then analyzed by flow cytometry (FACSCalibur).

EdU staining
An EdU dye assay was performed using the Cell-Light EdU Apollo 567 In Vitro Imaging Kit (RiboBio Technology, Guangzhou, China) according to the manufacturer's instructions. The labeled cells were counted under a fluorescence microscope. The experiments were independently repeated in triplicates.
Quantification of 5-hmC levels in genomic DNA by methylation-sensitive qPCR Genomic DNA was incubated with T4 Phage β-glucosyltransferase (New England Biolabs, Ipswich, MA) by following the manufacturer's protocol. First, 100 ng of glucosylated genomic DNA was digested with MspI, or without enzyme (mock) at 37°C overnight and then incubated for 20 min at 80°C for enzyme deactivation. HpaII-resistant or MspI-resistant DNA fraction was quantified by qPCR and normalizing to the mock control. MspI-resistant DNA represents the 5hmC DNA fraction, whereas the fraction of 5-mC DNA was calculated by subtracting the 5-hmC fraction from the resistance to HpaII. Primers were listed in Supplementary Table S1.

Xenograft tumor model
HCC cell xenograft models in mice were established as previously described 4 . To establish the xenograft tumor model, HCC cells were injected into right frank of nude mice for in vivo tumor growth assay. SMMC-7721 FGFR3Δ7-9 /Vector, SMMC-7721 FGFR3Δ7-9 /TET2, and SMMC-7721 FGFR3Δ7-9 /TET2 Y1902F cells (1 × 10 6 cells) were subcutaneously injected into 4-week-old male BALB/ c nude mice (Institute of Zoology, Chinese Academy of Sciences). Each group included 4 mice. Tumor nodules were measured and were calculated using the following formula: tumor volume = (width 2 × length)/2. Mice were euthanized 6 weeks after injection. All animal studies were approved by the Ethics Committee of Ruijin Hospital, Shanghai Jiao Tong University School of Medicine.

Statistical analyses
An ANOVA and Student's t test were used for comparison among groups. The Mann-Whitney U test was used for comparison of tumor volume. Categorical data was evaluated with a c2 test or Fisher exact test. The survival curve was plotted using the Kaplan-Meier method and compared by the log-rank test. A P value less than 0.05 was considered to be significant.