The m⁶A Reader YTHDF2 Promotes Bladder Cancer Progression by Suppressing RIG-I–Mediated Immune Response

YTHDF2 regulates RIG-I–mediated innate immune signaling to support bladder cancer progression, highlighting the functional importance of m⁶A modifications in bladder cancer and uncovering therapeutic opportunities to improve patient outcomes.

Abstract

N⁶-Methyladenosine (m⁶A) is the most prevalent internal modification of mammalian mRNAs. Recent studies have shown that m⁶A methyltransferases METTL3 and METTL14 play important roles in urothelial bladder carcinoma (BLCA). To provide a more comprehensive understanding of the m⁶A regulatory landscape in bladder cancer, we investigated the role of YTHDF2, a crucial m⁶A reader, in BLCA. YTHDF2 was frequently upregulated at both the RNA and protein level in BLCA. Functionally, YTHDF2 promoted the proliferation and tumor growth of BLCA cells in vitro and in vivo, respectively. Integrative RNA sequencing and m⁶A sequencing analyses identified RIG-I as a downstream target of YTHDF2. Mechanistically, YTHDF2 bound to the coding sequence of DDX58 mRNA, which encodes RIG-I, and mediated its degradation in an m⁶A-dependent manner. Knockdown of RIG-I inhibited apoptosis and promoted the proliferation of BLCA cells. Depleting RIG-I was also able to reverse the effects of YTHDF2 deficiency. YTHDF2-deficient BLCA cells implanted orthotopically in recipient mice activated an innate immune response and promoted recruitment of CD8⁺ T lymphocytes into the tumor bed and the urothelium. Moreover, YTHDF2 deficiency enhanced the efficacy of Bacillus Calmette-Guérin immunotherapy treatment. This study reveals that YTHDF2 acts as an oncogene in BLCA. YTHDF2 inhibits RIG-I to facilitate immune evasion, supporting testing YTHDF2 inhibition in combination with immunotherapy.

Significance:

YTHDF2 regulates RIG-I–mediated innate immune signaling to support bladder cancer progression, highlighting the functional importance of m⁶A modifications in bladder cancer and uncovering therapeutic opportunities to improve patient outcomes.

Graphical Abstract

Introduction

Urothelial bladder carcinoma is the most common malignant disease of the urinary system, with the eleventh-highest incidence and ninth-highest mortality of all cancers worldwide (1, 2). Bladder cancer exhibits a rather unique mode of tumor evolution: Seventy-five percent of patients are diagnosed with non–muscle-invasive disease but with tumor heterogeneity is likely to have a role in the management of these patients, as it might affect the selection of non–muscle-invasive bladder cancer (NMIBC)–risk groups, surveillance monitoring strategies, intravesical therapies, and early application of radical therapy (3). High-recurrence frequency and the risk of progression to muscle-invasive disease necessitate frequent surveillance, trans-urethral resection, and therapeutic intervention, making this disease the most expensive lifetime treatment as compared with all other cancers (2, 4). The efficacy of targeted therapies for bladder carcinoma has been recently reviewed (5), and highlights a clear deficit for a deeper understanding of the molecular mechanisms underlying the pathogenesis of bladder carcinoma recurrence and progression.

N⁶-Methyladenosine (m⁶A) is the most abundant intrinsic mRNA modification in eukaryotes occurring at the consensus motif, DRACH (where D denotes A, G or U, and H denotes A, C or U; ref. 6). m⁶A modification is a dynamic and reversible process that requires three types of proteins: An m⁶A “writer,” an “eraser,” and a “reader” (7, 8). The writers install the m⁶A modification onto the target RNA as part of a core methyltransferase complex, comprising two catalytic subunits (METTL3 and METTL14) and a regulatory subunit (WTAP). The erasers function to demethylate m⁶A on RNA, with two main m⁶A erasers (FTO and ALKBH5) discovered to date. Finally, the m⁶A-modified RNAs are recognized by specific “reader” proteins that bridge methyl-selective RNA binding with diverse cellular processes and modulate RNA metabolism (e.g., mRNA splicing, mRNA export, translation initiation and mRNA decay) via m⁶A-dependent regulation of gene expression. Reader proteins can be broadly divided into two categories: “Nuclear readers,” which include heterogeneous nuclear ribonucleoproteins (HNRNPC, HNRNPG, and HNRNPA2B1) and YTHDCs (YTHDC1 and YTHDC2), and “cytoplasmic readers,” such as eIF3, YTH domain-containing family proteins (YTHDF1, YTHDF2, and YTHDF3), and insulin-like growth factor 2 mRNA-binding proteins (IGF2BP1, IGF2BP2, and IGF2BP3; ref. 6). m⁶A modification is critical to numerous biological processes, including stemness, differentiation, and responses to stress and DNA damage (6, 8). Indeed, aberrant RNA methylation levels contribute to the pathogenesis of various malignancies, including melanoma, glioblastoma, and pancreatic cancers (9–14).

Several recent studies have shown m⁶A levels to be significantly elevated in bladder tumors as compared with normal bladder tissues, with elevated expression caused by an aberrant expression of m⁶A methyltransferases METTL3 and METTL14 (15–21). Upregulation of METTL3 leads to increased m⁶A levels in several genes, including AFF4, MYC, ITGA6, CDCP1, ITGA6, and DGCR8 (15–18, 20, 21). METTL14, on the other hand, has been shown to inhibit bladder tumor-initiating cell self-renewal by targeting Notch1 mRNA stability (19). These studies suggest a potential role for METTL3 and METTL14 in bladder carcinoma tumorigenesis and establish the importance of m⁶A modification in bladder carcinoma. However, whereas these studies focus on the function of m⁶A methyltransferases in bladder carcinoma, the role of m⁶A readers in bladder carcinoma tumorigenesis remains elusive. YTHDF2 is a well-studied m⁶A reader that primarily regulates mRNA stability. It recognizes and binds to m⁶A-modified RNAs through its YTH domain and recruits the CCR4-NOT deadenylase complex via its N-terminal domain to accelerate the degradation of target RNAs (22). Although previous studies have established its significance in the pathogenesis of several cancer types (11, 12, 23–28), the function of YTHDF2 in bladder carcinoma remains unexplored.

Here, we report that YTHDF2 promotes bladder cancer progression from NMIBC to MIBC. YTHDF2 is overexpressed in bladder carcinoma tissues, promoting bladder carcinoma proliferation and inhibiting bladder carcinoma cell death. We show that, mechanistically, YTHDF2 targets the coding sequence (CDS) region of DDX58, which encodes for RIG-I, an RNA helicase, and thereby accelerates DDX58 mRNA degradation in an m⁶A-dependent manner. Elevated YTHDF2 in patients with bladder carcinoma results in RIG-I tumor-suppressor degradation. In addition, deficiency in DDX58 signaling abrogates cancer cell death and promotes the progression of bladder carcinoma. Furthermore, we show that restoring RIG-I signaling by YTHDF2 depletion is sufficient to reactivate immune cell function and leads to cancer cell eradication. Our findings highlight a role for RIG-I signaling in bladder carcinoma tumorigenesis and identify YTHDF2 as a potential therapeutic target for bladder carcinoma.

Materials and Methods

Clinical specimens

Human bladder carcinoma tissues and paired adjacent normal tissues were obtained from the Department of Urology, Luohu People's Hospital (Shenzhen, China). After surgical resection, tumor and tumor-adjacent mucosal tissue within 3 to 5 cm from the edge of the tumor were cut into species, snap-frozen in liquid nitrogen and stored at −80°C. A 500 μL of ice-cold lysis buffer was added to approximately 5-mg piece of tissue that was homogenized with an electrical homogenizer, and then maintained under constant agitation for 2 hours at 4°C. Samples were then centrifuged for 20 minutes at 12,000 rpm at 4°C in a microcentrifuge. The tubes were gently removed from the centrifuge and placed on ice, the supernatant was aspirated, and placed in a fresh tube kept on ice for western blot analysis. This study was approved by the Ethics committee of Luohu People's Hospital.

Cell culture and transfection

All cells were incubated at 37°C in 5% CO₂. The human bladder carcinoma cell line MGH-U3 was purchased from Leibniz-Institut Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH (DSMZ). SW780 (Cat# CRL-2169), 5637 (Cat# HTB-9), T24 (Cat# HTB-4), and UMUC3 (Cat# CRL-1749) were purchased from the ATCC. MBT2 cells were obtained from the First Affiliated Hospital of Anhui Medical University. MB49 cells were purchased from Emd Millipore (Merck, Cat# SCC148). The identities of these cell lines were confirmed by STR analysis. Mycoplasma was tested every month. Cells were cryopreserved within the first 5 passages upon generating, and were used within 20 passages after thawing. Cells were cultured in MEM (MT10013CV; Corning) or RPMI-1640 (22400089, Gibco) supplemented with 10% FBS (GenStar), 1% l-glutamine (25030081; Gibco), and 1% penicillin/streptomycin antibiotic cocktail (10378016, Gibco). For transient KD of target genes, 100 pmol of each siRNA was transfected into 1 × 10⁵ cells using Lipofectamine RNAiMAX (13778150, Invitrogen) and Opti-MEM (31985088; Gibco), according to the manufacturer's instructions. Cells were harvested 24 to 36 hours post-transfection. All siRNAs were synthesized by GenePharma. siRNA sequences are listed in Supplementary Table S1. Poly(I:C) LMW (tlrl-picw, InvivoGen, France) was transfected into SW780 or MGHU3 cells at a dose of 10 μg/2×10⁶ cells using Lipofectamine RNAiMAX; cells were harvested at 6–12 hours post-transfection.

Cell viability and proliferation assay

Cell viability and proliferation were determined with a Cell Counting Kit-8 (E606335, Sangon Biotech) and a BeyoClick EdU-488 (C0071S, Beyotime). For the CCK-8 assay, SW780 or MGHU3 cells were transfected with the indicated siRNAs, then seeded into 96-well plates at a density of 1,000 cells/well. According to the manufacturer's recommendations, 10 μL of CCK-8 solution was added into the wells and the mixture was cultured for 3 hours. The absorbance was recorded at 450 nm.

For the EdU assay, cells were cultured in 24-well plates and transfected with the indicated siRNAs. At 24 hours post-transfection, cells were incubated with 50 μmol/L EdU for 2 hours at 37°C and then fixed in 4% formaldehyde for 30 minutes. Glycine was added to neutralize the formaldehyde. After permeabilizing with 0.3% TritonX-100 for 10 minutes at room temperature, 1 × Click reaction cocktail (100 μL) was used to react with the EdU for 30 minutes. Finally, 1 × Hoechst (100 μL) was used to stain the nuclei. Cells were visualized using a Leica DM750 microscope.

Colony formation assay

Stably transfected bladder carcinoma cells were plated into 6-well plates (800 cells/well) and cultured with DMEM for about 2 weeks. Proliferating colonies were stained with 1% crystal violet staining solution (E607309; Sangon Biotech). The colonies were counted and photographed.

Generation of stable knockin or knockout cell lines

For stable knockin, the full-length YTHDF2-CDS was cloned into a lentivector-based FG-EH-DEST for lentivirus production using the Gateway Cloning System (Invitrogen). For stable knockout (KO) of target genes, sgRNAs targeting the genes of interest were cloned into the lentiCRISPR v2 vector (sgRNA primer sequences are listed in Supplementary Table S1). Plasmids with packaging plasmids psPAX2 and pMD2.G were transfected into HEK293T cells using polyetherimide for lentivirus production. Lentivirus particles were harvested at 48 and 72 hours after transfection and concentrated by ultracentrifugation. Finally, concentrated lentivirus particles were added to SW780, MGHU3, or MB49 cells for 6 hours, and then removed with fresh culture medium. To isolate YTHDF2 knockin or YTHDF2 KO clones, transduced cells were selected by treatment with 5 μg/mL puromycin for 2–3 days, then diluted to one cell per 100-μL media and plated into the wells of a 96-well plate. Single cell–derived colonies were expanded and subjected to western blot validation.

Immunoblotting

The immunoblotting procedures were performed as previously described (29). Tissues were mechanically fragmented into small pieces. Tissue homogenates and cultured cells were lysed in RIPA buffer (25 mmol/L Tris-HCl pH 7.6, 150 mmol/L NaCl, 1% NP-40, 1% sodium deoxycholate, 0.1% SDS) containing phosphatase inhibitor cocktail (524629, Roche) and protease inhibitor cocktail (B14012, Bimake) on ice for 30 minutes, and then centrifuged at 12,000 × g at 4°C for 5 minutes. The supernatants were collected, and the protein concentration was quantified by the BCA method. Protein lysates (20 μg per lane) were separated with SDS-PAGE and transferred to polyvinylidene difluoride membranes. Membranes were blocked for 1 hour in 5% milk followed by overnight incubation with primary antibodies at 4°C. Membranes were washed for 3 × 10 minutes at room temperature in TBST. Mouse or rabbit horseradish peroxidase (HRP)–conjugated secondary antibodies were incubated for 1 hour in 5% milk at room temperature, followed by washing 3 × 10 minutes at room temperature in TBST. Proteins were detected using Immobilon Western HRP Substrate (WBKLS0500, Millipore).

The following primary antibodies were used in this study: YTHDF2 (1:2,000, 24744–1-AP, Proteintech; RRID: AB_2687435), RIG-I (1:2,000, D261405; Sangon Biotech, RRID: AB_2537772), p-IRF3 (1:1,000, 37829; Cell Signaling Technology, RRID: AB_2799121), p-TBK1 (1:1,000, 5483; Cell Signaling Technology, RRID: AB_10693472), Flag (1:4,000, 80010–1-RR, Proteintech, RRID: AB_2882940), β-actin (1:10,000, RM2001, RayBiotech, RRID: AB_2756462), and GAPDH (1:20,000, 60004–1-Ig, Proteintech, RRID: AB_2107436).

IHC and immunofluorescence

For IHC assays, tumors were resected and fixed in 4% paraformaldehyde overnight, and then dehydrated with a gradient alcohol series. Tissue microarrays (TMA) or formalin-fixed paraffin-embedded sections were dewaxed, rehydrated, and subjected to high-temperature antigen retrieval. Slides were quenched of endogenous peroxides using 3% H₂O₂ for 15 minutes, and then blocked in 5% goat serum in PBS/0.1% Tween-20 (PBS-T) for 1 hour. Slides were then stained overnight with primary antibodies in blocking buffer, rinsed, and then incubated with HRP-conjugated secondary antibodies at 1:200 dilution in PBS-T for 45 minutes at room temperature. Staining was visualized with DAB chromogen.

For immunofluorescence, tumor sections were washed three times with PBS, permeabilized with 0.1% Triton X-100, blocked in 1% BSA diluted in PBS for 45 minutes at room temperature and incubated with the primary antibodies over night at 4°C. Sections were then washed and incubated with FITC, Alexa Fluor-555 or Cy5-labeled secondary antibodies for 1 hour at room temperature. Tumor sections were washed thrice and counter-stained with DAPI. Immunofluorescence images were captured using a Zeiss LSM 800 confocal microscope at ×20 magnifications and analyzed using ImageJ software.

The following primary antibodies were used: YTHDF2 (1:200, 24744–1-AP, Proteintech, RRID: AB_2687435), RIG-I (1:100, 20566–1-AP, Proteintech, RRID: AB_10700006), EPCAM (1:200, EM1111, HuaBio, RRID: AB_2928115), CD4 (1:200, 25229, Cell Signaling Technology, RRID: AB_2798898), CD8 (1:200, A11856, ABclonal, RRID: AB_2758823), and CD11B (1:500, 66519–1-Ig, Proteintech, RRID: AB_2881882).

Human bladder cancer TMA (Cat# HBlaU108Su01) was purchased from Shanghai Outdo Biotech Co., Ltd. (http://www.superchip.com.cn/).

Relative cell death assays

The supernatants of treated cells were harvested and analyzed for LDH activities using an LDH cytotoxicity assay kit (88953, Pierce), according to the manufacturer's instructions. Relative cell death was determined as (experimental cell death – baseline cell death)/ (maximum cell death – baseline cell death) × 100%. Maximum cell death was indicated by a positive control for 100% cell death generated by lysis buffer from the kit, whereas baseline cell death was indicated by a negative control of 0% cell death because of no treatment.

RNA extraction, cDNA synthesis, and qPCR

Total RNA samples were isolated with TRizol Reagent (15596026, Life Technologies) following the manufacturer's guidelines. For cDNA synthesis, 1 μg total RNA or immunoprecipitated RNA samples were used for reverse transcription in 20 μL reaction volume using the HiScript III RT SuperMix (R323–01, Vazyme). Quantitative PCR (qPCR) was performed with RealStar Green Fast Mixture (2X; A302; GenStar) in a Roche LightCycler 480 system. GAPDH, β-actin, and/or 18S rRNA were used as endogenous controls. Each reaction was run in triplicate. Sequence-specific primers for the genes detected are listed in Supplementary Table S2.

RNA stability assay

Cells were transfected with the indicated plasmid or siRNAs, and then treated with 5 μg/mL actinomycin D (SBR00013, Sigma-Aldrich) at 24-hours post-transfection. Cells were then harvested at different time points. qRT-PCR was used to detect the relative expression of mRNA at various time points.

RNA-binding immunoprecipitation and qPCR

RNA-binding immunoprecipitation and qPCR (RIP-qPCR) was used to validate the interactions between YTHDF2 and its target mRNAs. RIP assay was performed as described previously (30). Briefly, SW780 cells with a stable knockin of Flag-tagged YTHDF2 were grown in 150-mm culture plates at 80% confluence. Cells were washed once with ice-cold PBS and then lysed in RNA immunoprecipitation (RIP) buffer (150 mmol/L KCl, 25 mmol/L Tris pH 7.4, 5 mmol/L EDTA, 0.5 mmol/L DTT, and 0.5%NP40) comprising freshly added 100 U/mL RNAase inhibitor (B600478, Sangon Biotech) and protease inhibitors. Then, 100 μL of whole-cell extract was incubated with Protein A/G magnetic beads (B26101, Bimake) conjugated with YTHDF2 antibody or m⁶A-specific antibody and negative control IgG antibody for 6 hours at 4°C. The conjugated beads were then washed thrice with RIP buffer, and then incubated with RNase-free DNase I for 15 minutes at 37°C, followed by Proteinase K for 15 minutes at 37°C to remove protein. Finally, the coprecipitated RNAs were extracted and subjected to qRT-PCR analysis.

RNA-seq and MeRIP-seq

For RNA-seq, total RNA from SW780 cells with stable expression of YTHDF2 KO and corresponding control cells were extracted using TRizol Reagent and submitted to the Beijing Genome Institute (BGI) for library construction and sequencing. Total RNA quality and quantity were assessed using an Agilent 2100 Bioanalyzer and NanoDrop 2000. mRNAs were enriched using oligo(dT)-attached magnetic beads and subsequently fragmented into small pieces. First-strand cDNA was generated using random hexamer-primed reverse transcription, followed by a second-strand cDNA synthesis. The mRNA libraries were then sequenced on a BGISEQ-500 platform.

Methylated RNA immunoprecipitation sequencing (MeRIP-seq) was conducted by LC-Bio. Briefly, mRNA was purified using the NEBNext Poly(A) mRNA Magnetic Isolation Kit (E7490S, NEB). RNA fragmentation was performed by incubation with magnesium ions at 94°C using NEBNext Magnesium RNA Fragmentation Module (E6150S, NEB). The specific anti-m⁶A antibody was applied for m⁶A pull down. The enriched mRNA fragments were then used to construct libraries with VAHTS Total RNA-seq (H/M/R) Library Prep Kit for Illumina (NR603, Vazyme). Sequencing was carried out on an Illumina HiSeq 2500 with paired-end 150-bp read length.

Sequencing data analysis

For RNA-seq data, low-quality reads were removed by SOAPnuke and Trimmomatic (31, 32). The retained clean reads were mapped to the human genome hg38 by using HISAT2 v2.1.0. RSEM v1.2.8 was used to estimate gene and transcript expression levels. DEGseq2 was used for geometric normalization and FPKM calculation with the following parameters: Fold Change ≥2 and adjusted P value ≤ 0.001.

For m⁶A-seq data analysis, we applied the customized pipeline easym6A to process our m⁶A-seq data as previously described (33). Briefly, adapters of raw sequencing data were clipped away by Cutadapt v1.15. The retained clean reads were first mapped to ribosomal RNAs (rRNA) by HISAT2 v2.1.0. All unmapped reads were then aligned to genomes using HISAT2 v2.1.0 with default parameters. PCR duplicates were evaluated by MarkDuplicates from Picard v2.17.10. We then calculated the PCR duplicate proportion, which we defined as the number of PCR duplicate reads divided by the total number of mapped reads. Another three metrics, nonredundant fraction and PCR bottlenecking coefficients 1 (PBC1) and 2 (PBC2), were quantified using ENCODE standards. Input samples from m⁶A-seq data were used to estimate gene expression levels by StringTie v1.3.4d. Peaks were detected with exomePeak, MeTPeak, and MACS2 v2.1.1. Finally, HOMMER v4.9 was used for motif enrichment analysis based on the top 2000 peaks ranked by their fold enrichment levels. Metagene analysis of m⁶A distribution on transcripts was performed using the R package Guitar. Integrative Genomics Viewer was used for m⁶A peak visualization.

Xenograft mouse model, orthotopic implantation, and intravesical BCG treatment

All animal experiments were conducted in accordance with protocols approved by the Ethics committee of Luohu People's Hospital and followed the relevant ethical regulations regarding animal research. BALB/c nude mice were used for in vivo xenograft studies, and C57BL/6 mice for orthotopic implantation and Bacillus Calmette–Guérin (BCG) studies. All mice were purchased from Guangdong Medical Laboratory Animal Center and kept under specific pathogen-free conditions. For xenograft studies, female mice between 4 and 6 weeks of age were used. Monoclonal cell lines showing a nearly complete absence of Ythdf2 and Ddx58 were used for in vivo experimentation. Mice were implanted with 2 × 10⁶ MGHU3 cells resuspended in 100 μL of Matrigel Matrix High Concentration (354248) and PBS mixture subcutaneously into the right flanks. Subcutaneous tumor growth was measured weekly by calipers, and the volume was calculated using the formula: volume = (length × width²)/2.

The MB49 orthotopic bladder cancer model was generated as previously described (34). Briefly, 6- to 8-week-old female C57BL/6 mice were anesthetized with an intraperitoneal injection of pentobarbital sodium salt (Y0002194, Sigma-Aldrich). Then, a 1-cm incision was made through the skin and abdominopelvic wall to expose the internal organs. Urine was then aspirated from the bladder, and 50 μL (1×10⁶ cells) MB49 cells were slowly injected into the bladder lumen. After surgery, the muscle layers and the skin were sutured with 6.0 resorbable sutures.

For BCG instillation treatment, frozen stocks of BCG (R19019, REBio) were thawed and resuspended in PBS to a final concentration of 2 mg/mL; PBS alone was used as control. Mice were placed under anesthesia with an intraperitoneal injection of pentobarbital. A 24-gauge catheter was inserted into the bladder through the urethra, and 50 μL of BCG (1 mg per mouse) or PBS was injected into the bladder.

Quantification and statistical analysis

Statistical analysis was performed using GraphPad Prism 8.0. Error bars show the mean ± SEM or mean ± SD, as indicated. Significance was determined using a Student two-tailed unpaired t test with 95% confidence intervals, unless otherwise indicated; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant (P > 0.05).

Data Availability

Datasets generated in this study have been deposited to Sequence Read Archive (SRA; ID: PRJNA909345). The data analyzed in this study were obtained from Gene Expression Omnibus (GEO) at GSE77952, GSE128959, GSE3167, GSE13507, and GSE176178, and The Cancer Genome Atlas (TCGA) database. Source Data for Immunoblotting have been provided (Supplementary Fig. S9). All other raw data are available upon request from the corresponding author.

Results

YTHDF2 is associated with bladder carcinoma progression

To assess the potential role of YTHDF2 in bladder carcinoma, we first examined YTHDF2 expression in public GEO datasets deposited at the NCBI. We found that YTHDF2 mRNA expression was significantly elevated in bladder carcinoma tumors as compared with normal bladder tissues in the GSE3167 dataset (Supplementary Fig. S1A). Consistently, in TCGA, YTHDF2 was also found to be significantly upregulated in tumors as compared with adjacent normal tissues (Supplementary Fig. S1A). We next validated YTHDF2 upregulation in a bladder carcinoma cohort consisting of 56 pairs of bladder cancer tissues and adjacent nontumor tissues by qRT-PCR (Supplementary Fig. S1B, left). The TCGA dataset also showed that the YTHDF2 expression was upregulated in tumors, when compared with normal tissues (Supplementary Fig. S1B, right). We also determined that the protein level of YTHDF2 in bladder carcinoma specimens was remarkably increased in tumor tissues (Supplementary Fig. S1C). We also analyzed the YTHDF2 expression in different bladder carcinoma subtypes. In the TCGA dataset, we found that YTHDF2 was more expressed in the Luminal Papillary and Luminal Unstable subtypes (Supplementary Fig. S1D).

To determine which population of patients with bladder carcinoma is mostly relevant to YTHDF2, we classified 402 TCGA patients to YTHDF2-low and -high, according to the transcriptome data. Interestingly, we found that higher expression of YTHDF2 is associated with FGFR3 mutation, suggesting that it may have a role in luminal papillary MIBC and NMIBC (Fig. 1A). Although high frequency of recurrence as superficial tumors is the major problem of low-grade Ta tumors, these tumors are still able to progress to T1 and T2 invading the muscle (35, 36). These findings suggested that YTHDF2 may play a vital role in patients harboring mutated FGFR3. Progression from NMIBC to MIBC is the major problem among these patients. Because most of the bladder cancer samples in TCGA have already progressed (MIBC, ≥T2), to evaluate the effects of YTHDF2 in bladder carcinoma progression, we analyzed the expression of YTHDF2 in another two datasets, GSE77952 and GSE128959. As shown in Fig. 1B, the expression of YTHDF2 was statistically higher in MIBC when compared with NMIBC tumors, and in recurrent and progressed tumors compared with primary tumors. MIHC staining for YTHDF2 expression of NMIBC and MIBC patients also indicated that YTHDF2 is upregulated in MIBC (Fig. 1C), suggesting that YTHDF2 may promote bladder carcinoma progression. To further examine the association between YTHDF2 expression and tumor progression in bladder carcinoma, we performed IHC on two bladder carcinoma TMA containing 128 cases of bladder carcinoma and 40 adjacent normal tissues (HBlaU060CS02 and HBlaU108Su01). As shown in Fig. 1D and E, YTHDF2 expression was significantly upregulated in MIBC samples as compared with NMIBC samples. Analysis of overall survival (OS) using TMA-based IHC staining of YTHDF2 in bladder carcinoma samples illustrated that those with lower expression of YTHDF2 had better outcome, manifesting as a longer OS (Fig. 1F). To validate whether YTHDF2 promotes NMIBC recurrence, we also compared the expression of YTHDF2 in 3 pairs of primary and recurrent bladder carcinoma samples. As shown in Supplementary Fig. S1E, YTHDF2 expression was higher in recurrent tumors than in primary tumors. Collectively, these data suggest that YTHDF2 may contribute to bladder carcinoma progression.

Figure 1.

YTHDF2 is highly upregulated in patients with bladder carcinoma. A, Association of YTHDF2 expression and mutations in TCGA bladder carcinoma. B, Relative YTHDF2 expression level in NMIBC and MIBC from the GEO dataset GSE77952 (left) and in recurrence samples compared with primary tumors from GEO dataset GSE128959 (right). C, The fluorescence intensity of YTHDF2, EpCAM (epithelial marker), and DAPI by multiplex immunofluorescence (MIHC) staining assays. D, Representative images of TMA-based IHC staining of YTHDF2 in bladder carcinoma samples. E, The IHC expression differences between MIBC and NMIBC. Statistical analysis was performed with the nonparametric Mann–Whitney test. F, The overall survival using TMA-based IHC staining of YTHDF2 in bladder carcinoma samples. P values in A were calculated using the χ² test. P values in D and E were calculated with the nonparametric Mann–Whitney test. Survival rate (F) was compared by log-rank (Mantel–Cox) test. *, P < 0.05; **, P < 0.01; ****, P < 0.0001.

YTHDF2 promotes proliferation and suppresses apoptosis in bladder carcinoma cells

To explore the potential role of YTHDF2 in bladder carcinoma, we used two siRNAs (Supplementary Table S1) to transiently knockdown (KD) YTHDF2 in two bladder carcinoma cell lines, SW780 and MGHU3, which harbor mutated FGFR3, express higher levels of YTHDF2 than other Luminal Papillary bladder carcinoma cell lines, and could progress to muscle-invasive in vivo. The KD efficiencies were confirmed by qPCR and immunoblotting (Supplementary Fig. S2A and S2B). KD of YTHDF2 significantly suppressed bladder carcinoma cell proliferation, as determined using CCK-8 (Fig. 2A) and colony-forming (Fig. 2B) assays. Furthermore, through flow cytometry, we found that the percentage of apoptotic cells was higher in YTHDF2 KD SW780 cells compared with WT cells (Fig. 2C; Supplementary Fig. S2C), indicating that KD of YTHDF2 promoted apoptosis. We also detected an increase in cell death using an LDH assay. Similarly, YTHDF2 silencing resulted in a significant increase in TNF/zVAD/cycloheximide–induced necroptosis (Fig. 2D), and dramatically suppressed the migratory ability of bladder carcinoma cells in wound-healing and Transwell cell invasion assays consistent with a recent report (Supplementary Fig. S2D–S2E; ref. 18). From these findings, we reasoned that YTHDF2 inhibition may be able to enhance chemosensitivity. As shown in Fig. 2E, YTHDF2 KD caused a 3- to 5-fold decrease in the IC₅₀ values for fluorouracil, gemcitabine, and mitomycin C as compared with control cells, indicating that YTHDF2 deficiency diminishes drug resistance. In summary, these results indicate that YTHDF2 deficiency impairs bladder carcinoma cell proliferation, enhances cell death, and potentiates drug sensitivity.

Figure 2.

Inhibition of YTHDF2 affects the proliferation, cell death, and drug sensitivity of bladder carcinoma cells. A, Effects of knockdown of YTHDF2 expression on cell growth/proliferation in SW780 and MGHU3 cells analyzed by the CCK-8 assay. B, Representative images from the colony-forming assay (left) and colony number analysis (right). C, Effects of knockdown of YTHDF2 expression on cell apoptosis in SW780 and MGHU3 cells. Apoptosis was measured by analysis of double Annexin V/PI-stained cells. D, Effects of the knockdown of YTHDF2 on relative cell death determined by lactate dehydrogenase release of SW780 cells transfected either with control siRNA or siRNA against YTHDF2, then treated with PBS or TNFα (10 ng/mL) plus cycloheximide (10 μg/mL) and zVAD (20 μmol/L) for 12 hours. E, Effects of YTHDF2 knockdown on drug sensitivity in SW780 cells analyzed by the CCK-8 assay. F, Effects of knockout of YTHDF2 on tumor growth in a subcutaneous xenograft model. The size of tumors was measured at the indicated time points, tumors were extracted, and weighed after mice were sacrificed. Data in A–F are presented as the means ± SEM. P values in A–F were calculated using Student t test. *, P < 0.05; **, P < 0.01; ****, P < 0.0001.

To further verify the oncogenic function of YTHDF2 in bladder carcinoma in vivo, we generated YTHDF2 KO MGHU3 cells using CRISPR/Cas9 (sgRNA primer sequences are listed in Supplementary Table S1). YTHDF2 KO and control MGHU3 cells were subcutaneously implanted into the right flanks of BALB/c nude mice. As shown in Fig. 2F, the inactivation of YTHDF2 in these cell lines effectively suppressed tumor growth in nude mice, as reflected by the significant reduction in tumor size and weight when compared with that of nontargeting sgRNA-treated (control) cells (Fig. 2F; Supplementary Fig. S2F). Taken together, our results suggest that YTHDF2 plays a pivotal role in promoting bladder carcinoma growth and suppressing bladder carcinoma cell death.

Transcriptome sequencing and m6a sequencing identifies RIG-I as a downstream target of YTHDF2

The m⁶A reader protein YTHDF2 exerts its function through the direct recruitment of the CCR4–NOT complex and promotes the degradation of m⁶A-modified transcripts (22). We performed RNA sequencing in SW780 cells to further delineate the functional implications of YTHDF2 and to identify its downstream targets in bladder carcinoma. We identified a total of 516 differentially expressed genes (DEG; 333 upregulated genes, 183 downregulated genes) in YTHDF2 KO SW780 cells (Fig. 3A and B; Supplementary Table S3). Gene ontology (GO) analysis of the DEGs revealed enriched pathways mainly associated with type I IFN signaling, defense response, and immune system processing (Supplementary Fig. S3A). DEGs were then assigned to pathway visualization in an unbiased fashion according to the Kyoto Encyclopedia of Genes and Genomes (KEGG) database. We found a significant enrichment of genes involved in RIG-I–like receptor signaling pathway, glycosaminoglycan biosynthesis, and TGFβ signaling pathway, as well as genes that encode for cell adhesion molecules (Fig. 3C). By searching the molecular signature database of gene set enrichment analysis, we found that the upregulated transcripts were significantly enriched for genes associated with IFNγ response, IFN signaling, and response to viruses, whereas the downregulated genes were significantly enriched for pathways associated with seleno aminoacid metabolism, translational initiation, and eukaryotic translation elongation (Fig. 3D; Supplementary Fig. S3B).

Figure 3.

Identification of genes and pathways in YTHDF2 knockout cells. A, Volcano plot showing DEGs in YTHDF2 KO cells compared with non-targeting sgRNA controls, significantly downregulated (blue) and upregulated (red). B, Heatmap showing the most significantly downregulated or upregulated genes in control versus YTHDF2 KO SW780 cells. C, KEGG pathways enrichment in the set of DEGs between YTHDF2 KO and control SW780 cells. D, Gene set enrichment analysis of upregulated genes in YTHDF2 KO cells for an “interferon signaling” signature. E, qPCR analysis of representative upregulated genes in SW780 and MGHU3 cells. F, Top consensus m⁶A motifs identified by HOMMER in YTHDF2 KO and control cells. G, Metagene profile showing the distribution of m⁶A peaks across the length of transcripts composed of three rescaled nonoverlapping segments, 5′-UTR, CDS, and 3′-UTR. H, Pie chart depicting the distribution of m⁶A peaks in different transcript segments in control and YTHDF2 KO cells. I, MA plot of m⁶A peaks detected by meRIP-seq in control or YTHDF2 KO cells. J, GO enrichment analysis of the upregulated m⁶A peaks (G) in different peaks between YTHDF2 KO and control cells. Data in E are presented as the means ± SEM. P values in E were calculated using the Student t test. **, P < 0.01; ***, P < 0.001.

To validate our RNA-seq data, we transfected SW780 and MGHU3 cells with two siRNAs targeting YTHDF2 and performed qPCR analysis. As shown in Fig. 3E, depletion of YTHDF2 was accompanied by an increased expression of genes associated with cell death and IFN-response, such as RIPK1, FADD, IFI44, RSAD2, and CXCL10. In addition, most of the upregulated genes were enriched in antiviral signaling pathways, implying a pivotal role for YTHDF2 in bladder carcinoma cell immune response. To confirm this hypothesis, we infected YTHDF2 KO and control SW780 cells with eGFP-tagged vesicular stomatitis virus. We observed lower levels of virus particle production, as detected by GFP signal, in YTHDF2-depleted cells at 16 hours post-infection (Supplementary Fig. S2G). Furthermore, we detected more IFNβ production in the culture medium of YTHDF2 KO SW780 cells (Supplementary Fig. S2H). These results suggest an enhanced immune response in YTHDF2-depleted SW780 cells. Overall, DEGs resulting from YTHDF2 depletion were principally associated with IFNβ response.

To identify potential downstream targets of YTHDF2 in bladder carcinoma cells, we carried out immunoprecipitation of m⁶A methylated RNA followed by sequencing (m⁶A-seq). After combining the peaks in replicates, a total of 8,974 joint m⁶A peaks were identified by MACS, exomePeak, and MeTPeak methods in SW780 cells (Supplementary Fig. S3C). To identify which biological processes were involved in m⁶A modification in bladder carcinoma cells, m⁶A peaks were subjected to enrichment analysis using the genomic region enrichment of annotations tool (GREAT, http://great.stanford.edu/). We found that m⁶A peaks were significantly enriched in genes involved in ncRNA metabolic processes, macromolecular complex disassembly, regulation of type I IFN production, and regulation of intrinsic apoptotic signaling pathways (Supplementary Fig. S3D). Consistent with previous studies, the most common m⁶A motif “GGAC” was significantly enriched. These m⁶A peaks were predominantly located in exons and were highly enriched near stop codons and in 3′UTRs in both the control and YTHDF2 KO cells (Fig. 3F–H).

We next compared the abundance of the m⁶A peaks between the control cells and YTHDF2 KO cells, with a total of 438 and 2,300 m⁶A peaks, respectively, shown to be significantly decreased and increased (P < 0.005; fold-change >1.2; Fig. 3I). We performed GO enrichment analysis to understand the biological pathways involved in these differential m⁶A changes. We found that the upregulated m⁶A peaks were enriched in pathways involved in the regulation of type I IFN production and DNA repair (Fig. 3J), whereas the downregulated m⁶A peaks were mainly involved in mitochondria organization, ncRNA processing, and translation elongation (Supplementary Fig. S3E). As a cytoplasmic reader of m⁶A, YTHDF2 mainly mediates the turnover of its RNA targets by recruiting the CCR4–NOT deadenylase complex to initiate deadenylation and decay of transcripts containing m⁶A for committed degradation. Therefore, we assumed that when reducing or abrogating YTHDF2 (KD or KO), the stability of its target RNAs should be enhanced. Given that YTHDF2 mainly promotes the decay of its target mRNAs, mRNA levels and m⁶A abundance of target genes were expected to be increased with YTHDF2 depletion. Thus, we filtered 383 upregulated genes that were found in the RNA-seq with 2,300 m⁶A peaks that were shown to be elevated in YTHDF2 KO cells. From this analysis, we identified 33 candidate genes, including DDX58 (Fig. 4A). Considering that RIG-I has a critical function in type I IFN signaling and also regulates cancer cell survival and death, we further analyzed our m⁶A-seq data for DDX58. We found that m⁶A modifications accumulated in DDX58 transcripts. In addition, the m⁶A peak in the last exon of DDX58 mRNA was increased remarkably upon YTHDF2 KO (Fig. 4B). These results indicate that DDX58 might be a direct downstream target of YTHDF2.

Figure 4.

RIG-I mRNA is identified as a target of YTHDF2. A, Bioinformatic analysis filtered RIG-I as a downstream target of YTHDF2. B, Integrative Genomics Viewer tracks displaying the results of m⁶A-seq (red) and RNA-seq (blue) read distributions in DDX58 mRNA in control and YTHDF2 KO cells. C and D, qPCR analysis of DDX58 mRNA (C) and immunoblotting (D) for RIG-I protein level in bladder carcinoma cells transiently transfected with control siRNAs (siNC) and YTHDF2 siRNAs (siYTHDF2). E, Validation of m⁶A modification in DDX58 mRNA in SW780 cells using YTHDF2-specific or m⁶A antibody by RIP-qPCR. F,DDX58 mRNA decay assay using actinomycin D treatment in SW780 cells transfected with siNC or YTHDF2 siRNAs. G,DDX58 mRNA decay assay using actinomycin D treatment in SW780 cells transfected with siMETTL3, siFTO, or siNC. H, Schematic illustration of the m⁶A site and its mutants in the CDS of DDX58 mRNA (near the stop codon). I, qPCR and immunoblotting analysis of RIG-I and its mutants in HEK293T cells. J, qPCR shows the level of DDX58 and its mutant mRNA in HEK293T cells transfected with YTHDF2 or empty vector after actinomycin D at the indicated time points. The relative levels of DDX58 transcripts in F, G, and J were normalized to the housekeeping gene β-actin, and DDX58 mRNA level was analyzed by qPCR. Data in C, E–G, I, and J are presented as the means ± SEM. P values in C, E–G, I, and J were calculated using the Student t test. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant (P > 0.05).

YTHDF2-mediated regulation of RIG-I is dependent on m⁶A modifications

KD of YTHDF2 in bladder carcinoma cells led to a significant upregulation in both DDX58 mRNA and protein levels in low invasive MGHU3 and SW780 cell lines harboring mutated FGFR3 (Fig. 4C and D). Given to the finding that YTHDF2 regulates DDX58 in FGFR3-mutated cell lines, we investigated whether the regulation also operates in UMUC3 and T24, FGFR3 WT and P53-mutated invasive cell lines. Our data reveal that alteration of RIG-I expression in these two cell lines after YTHDF2-KD was not significant, as compared with MGHU3 and SW780 cell lines (Supplementary Fig. S4A). To verify that YTHDF2 regulation of DDX58 is dependent on its m⁶A modifications, we performed RIP using either YTHDF2 antibodies or m⁶A-specific antibodies followed by RT-qPCR in Flag-tagged YTHDF2 stable knockin SW780 cells. As shown in Fig. 4E, when compared with IgG controls, we saw an enrichment of DDX58 mRNA in the fraction bound to both the YTHDF2 antibodies and m⁶A-specific antibodies, implying that DDX58 transcripts are indeed methylated and bound to YTHDF2. As YTHDF2 silencing potently affects target gene expression, we performed mRNA decay assays in YTHDF2 KD cells. Fig. 4F shows that DDX58 half-life was significantly more prolonged in YTHDF2-depleted cells than in control cells. Moreover, we observed accelerated decay of DDX58 mRNA upon KD of the FTO eraser in SW780 cells, whereas KD of the methyltransferase gene METTL3 prolonged DDX58 mRNA half-life (Fig. 4G). These results suggest that m⁶A modification is involved in regulating the mRNA stability of DDX58.

SRAMP server (an m⁶A modification site predictor; http://www.cuilab.cn/sramp) was used to predict the potential m⁶A modification sites on DDX58 transcripts. Consistent with our m⁶A-seq data, a conserved DRACH motif was found in the last exon near the stop codon of DDX58 (Fig. 4B and H). To confirm this finding, we mutated the GGACT motif to GGCCT and GGTCT in DDX58 mRNA (Fig. 4H). As expected, mutations in the m⁶A site of DDX58 significantly increased the mRNA stability and the protein level of RIG-I in HEK293T cells (Fig. 4I). In addition, ectopic YTHDF2 induced a significant decrease in the half-life of wild-type DDX58; this decrease was largely diminished by mutations in the m⁶A consensus sites (Fig. 4J). We also overexpressed WT and mutant RIG-I in YTHDF2 KO and WT SW780 cells, respectively. The results showed that mutations in the m⁶A site of DDX58 significantly increased the mRNA stability and the protein level of RIG-I in SW780 cells, whereas no significant difference was observed in YTHDF2 KO SW780 (Supplementary Fig. S4B–S4D). Taken together, our data show that YTHDF2 mediates the m⁶A-dependent decay of DDX58 mRNA.

RIG-I acts as a tumor suppressor in bladder carcinoma

To illustrate the role of RIG-I in bladder carcinoma, we first detected its mRNA expression in tumor samples. We found a slight decrease in DDX58 mRNA levels in tumors as compared with adjacent normal tissues (Supplementary Fig. S5A). Consistently, we also observed a decrease in RIG-I protein levels in bladder carcinoma tumors (Supplementary Fig. S5B). Furthermore, TMA-based IHC analysis of 40 pairs of bladder carcinoma tumors and adjacent normal tissues revealed that RIG-I is downregulated in bladder tumors (Fig. 5A). In the TCGA dataset, we found that RIG-I showed reduced expression in the Luminal Papillary subtype (Supplementary Fig. S5C). Moreover, we found a negative correlation between RIG-I and YTHDF2 protein levels in 68 bladder carcinoma tumors (Fig. 5A and B), and found DDX58 expression to be negatively correlated with YTHDF2 mRNA levels in the TCGA bladder carcinoma dataset (Fig. 5C). Collectively, these data reveal RIG-I as downregulated and negatively correlated with YTHDF2 in bladder carcinoma.

Figure 5.

Knockdown of RIG-I promotes bladder cancer cell proliferation. A, Representative IHC images for RIG-I and YTHDF2 in bladder cancer tumor and adjacent bladder tissue. B, Correlation of protein level of YTHDF2 and RIG-I in 68 tumor tissues of bladder carcinoma TMA (HBlaU108Su01) analyzed by IHC assay. C, Correlation of mRNA level of YTHDF2 and DDX58 in TCGA bladder carcinoma. D, Immunoblotting showing RIG-I signaling by poly(I:C) stimulation in SW780 and MGHU3 cells. E, qRT-PCR showing relative mRNA expression of IFNβ and its target genes in SW780 and MGHU3 cells with poly(I:C) stimulation. F, Effects of knockdown of DDX58 on cell proliferation measured by CCK-8 assays. Data are presented as the means ± SEM. P values were calculated using Student t test. *, P < 0.05, **, P < 0.01, ***, P < 0.001. (C, Plot was generated with DrBioRight, drbioright.org.)

Given that RIG-I is an important molecule in promoting type I IFN (IFNβ) signaling, we next tested whether RIG-I signaling operates in bladder carcinoma cells. As shown in Fig. 5D, SW780 and MGHU3 cells transfected with the polyinosinic: polycytidylic acid immunostimulant [poly(I:C)] had significantly elevated levels of RIG-I along with IRF3 and TBK1 phosphorylation, which represent the primary activation of IFN-I signaling. Moreover, poly(I:C) transfection also upregulated the expression of IFNβ as well as several target genes, including ISG15, CXCL10, and CCL5 (Fig. 5E). This phosphorylation of IRF3 and TBK1 caused by poly(I:C) stimulation could be impaired by DDX58 KD (Supplementary Fig. S5D). These data suggest that RIG-I signaling is still functional in bladder carcinoma cells. Moreover, we sought to identify correlations with tumor-infiltrating immune cells based on the RNA-seq using TCGA datasets. We found significant positive correlations between DDX58 mRNA level and CD4⁺ T cells, CD8⁺ T cells, neutrophils, and myeloid dendritic cells (DC), with YTHDF2 mRNA levels negatively correlated with the presence of these immune cells (Supplementary Fig. S5E and S5F).

CCK-8 and colony forming assays indicated that DDX58 silencing promoted cell proliferation and viability (Fig. 5F; Supplementary Fig. S5G). Moreover, DDX58 depletion led to marked enhancement of cell migration and invasion, as shown by wound-healing (Supplementary Fig. S5H) and Transwell (Supplementary Fig. S5I) assays. As expected, DDX58 KD impaired the expression of apoptosis-associated IFN-stimulated genes, such as TRAIL, XAF1, and OAS1 (Supplementary Fig. S5J).

Previous studies reported that RIG-I signaling mediates several types of tumor cell–intrinsic tumor cell death, such as apoptosis, pyroptosis, autophagic cell death, and immunogenic cell death (37, 38). To test whether RIG-I exerts the same effects in bladder carcinoma cells, we transfected SW780 cells with poly(I:C) to trigger RIG-I signaling, and measured changes in cell death based on the LDH assay. As shown in Supplementary Fig. S5K, without poly(I:C), there was no difference in the degree of cell death between the DDX58-deficient and control cells. Yet, following poly(I:C) stimulation, DDX58 KD cells showed fewer dead cells as compared with the control counterpart; this suggests that residual RIG-I signaling was involved in cell death in bladder carcinoma cells (Supplementary Fig. S5K). Overall, these results implicate RIG-I as a tumor suppressor in bladder carcinoma via the activation of IFNβ signaling and that RIG-I induces cell death in bladder carcinoma cells.

Depleting RIG-I partially restores the phenotype induced by YTHDF2 deficiency in bladder carcinoma

To confirm that the observed phenotypes were mediated by dysregulation of the YTHDF2–RIG-I axis, we conducted several functional restoration assays. We found that YTHDF2 deficiency led to impaired cell growth and viability in SW780 cells, and that this could be reversed by DDX58 depletion (Fig. 6A–C). In addition, we noted that cell death induced by YTHDF2 deficiency could be partially recovered by KD of DDX58 (Fig. 6D).

Figure 6.

RIG-I depletion can restore the phenotype of YTHDF2 deficiency in bladder carcinoma cells. A–D, Effects of knockdown (A, C, and D) or knockout (B) of either YTHDF2 or DDX58, or both YTHDF2 and RIG-I in SW780 cells on cell viability measured by CCK-8 assays (A), cell proliferation analyzed by colony formation assays (B), the percentage of cells in S-phase determined by EdU assays (C), and relative cell death calculated by LDH release (D). E–G, Tumor size and weight of subcutaneous tumors from the indicated groups. H, H&E staining and sizes of tumors generated by Ddx58 KO and control MB49 cells implanted in mouse bladder lumen. I, A schematic model of signaling mediated by YTHDF2 in tumorigenesis of bladder carcinoma cells. YTHDF2 high-expressing cells fail to activate the type I IFN response and to suppress cell death. RIG-I detects and binds cytosolic double-stranded RNA (dsRNA) by its DExD/H-box domain, then opens its auto-repressed conformation, and recruits to the CARD domain of the MAVS, which induces phosphorylation of TBK1, IRF3, and IRF7, leading to the production of type I IFNs and cell death. In YTHDF2 high-expressing bladder carcinoma cells, excessive YTHDF2 binds to m⁶A modified RIG-I mRNA, which promotes RIG-I mRNA degradation, therefore, causing a decrease in RIGI-I protein expression level in the cell. Cytosolic dsRNA cannot be effectively recognized by the low-level RIG-I protein, thus weakening the activation of IFNI signaling and promoting bladder carcinoma tumorigenesis. Data in A–D are presented as the means ± SEM. Data in F and G are presented as the means ± SD of the experiment with 5 mice per group. P values in A–D, F, and G were calculated using the Student t test. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

To further investigate the contribution of YTHDF2 in regulating RIG-I in the context of bladder carcinoma in vivo, we established xenografts in BALB/c nude mice by inoculating mice with YTHDF2 KO, DDX58 KO, and double KO (YTHDF2 KO and DDX58 KO) MGHU3 cells; control cells were treated with a scrambled sequence (Control). As shown in Fig. 6E–G; Supplementary Fig. S5L and S5M, KO of DDX58 significantly enhanced bladder carcinoma tumor growth, which mimics the effect of YTHDF2 overexpression, and was sufficient to restore the inhibitory effect of YTHDF2 KO on tumor growth.

We transplanted MB49 cells into the bladder lumen, with a CRISPR/Cas9 KO of Ddx58 or control. Ten days post-transplantation, three mice injected with Ddx58-KO cells showed MIBC stage whereas the tumors in control did not invade the muscle. Bladder volume of Ddx58-KO mice were about 3 times larger than in control (Fig. 6H). Taken together, these data demonstrate that DDX58 is a functionally important target of YTHDF2 (Fig. 6I).

YTHDF2 deficiency enhances the therapeutic efficacy of BCG immunotherapy

BCG immunotherapy is an intravesical immunotherapy based on the bacteria of Mycobacterium bovis (bovine TB) and has been used since 1976 for the treatment of bladder cancer (39). Considering that RIG-I expression shows a strong positive correlation with immune cells in TCGA bladder carcinoma dataset (Supplementary Fig. S5E), we anticipated that downregulating YTHDF2 could benefit BCG immunotherapy.

To this end, we first generated a bladder tumor model. We created and injected Ythdf2 KO MB49 cells (1×10⁶) and control MB49 cells (Control) into the bladder lumens of C57BL/6 mice with a surgical orthotopic approach. By 10 days’ post-inoculation, the Ythdf2 KO tumors were much smaller than their control counterparts (Fig. 7A; Supplementary Fig. S6A and S6B). Ythdf2-KO tumors exhibited superficial phenotype that did not invade the stroma or muscle, whereas control tumors showed a severe degree of muscle invasion (Fig. 7B). These results confirmed our finding that YTHDF2 promotes bladder carcinoma progression.

Figure 7.

Knockout of Ythdf2 enhances the efficiency of BCG therapy in orthotopic implantation bladder cancer model and shows therapeutic potential in bladder carcinoma. Ythdf2 KO and control MB49 cells were implanted in mouse bladders. A, Images of tumors showing effects of knockout of Ythdf2 on tumor sizes. B, Representative H&E, IHC, MIHC staining of bladder tissue section. C, Representative H&E staining of bladder tissue section from mouse survivors after BCG therapy and confocal micrographs of CD4⁺, CD8⁺, and CD11b⁺ immune cells infiltration. D, Quantification of CD8⁺ and CD11b⁺ cells (n = 12 FoV; FoV, field of view) in C. For each immune cell, we counted 12 fields of view. E, Experimental schematic. Mice were implanted with MB49 bladder carcinoma cells on day 0 and received intravesical BCG or PBS on days 3, 10, and 17. F, Survival curves of mice implanted with control or Ythdf2 KO MB49 cells and treated with BCG. G, Representive H&E staining of survivor mouse bladder after BCG therapy and confocal micrographs of CD8⁺ T lymphocyte infiltration. H, Quantification of CD8⁺ T lymphocyte infiltration (n = 6 FoV). I, The schematic view of the therapeutic programs and curative effects. YTHDF2 disables RIG-I, attenuating the BCG therapy response and perturbation of antitumor immune cell infiltration. Data in D and H are presented as the means ± SD. P values in D and H were calculated using the Student t test. Survival rate (F) was compared by log-rank (Mantel–Cox) test. *, P < 0.05; **, P < 0.01; ns, not significant (P > 0.05).

Given that YTHDF2-deficient cells produced more IFNβ and that IFNβ enhanced the recruitment of immune cells to the tumor microenvironment (TME), we next assessed whether YTHDF2 modulated the degree of immune infiltration within the TME. We observed no significant difference in IHC detection of CD4⁺ T lymphocytes between Ythdf2 KO and Control tumors. However, Ythdf2 KO tumors were significantly infiltrated by CD8⁺ T lymphocytes and CD11b-positive cells (Fig. 7C and D; Supplementary Fig. S6C), which suggest enhanced CD8⁺ cytotoxic T-cell activation. These data indicated that YTHDF2 promotes a pro-tumorigenic microenvironment by reducing tumor infiltration of critical mediators of the antitumor response.

Prior studies have suggested that T cells are required for BCG therapy. Thus, we next sought to characterize whether YTHDF2 deficiency could enhance the efficacy of BCG therapy. C57BL/6 mice were implanted with control or Ythdf2 KO MB49 cells, as above, and received three weekly intravesical treatments of BCG (Fig. 7E). We observed a prolonged survival rate among mice inoculated with Ythdf2 KO cells as compared with control mice: after 57 days’ post-inoculation, 7 control tumor-bearing mice had died whereas only 3 Ythdf2 KO tumor-bearing mice had died (Fig. 7F). The bladder of the only one surviving mouse transplanted with control MB49 was shown to be free of tumor. Similar situation was observed in 3 out of the 5 surviving mice with Ythdf2 KO cells, whereas the other 2 bladders still harbored tumors (Supplementary Fig. S6D and S6E). Furthermore, BCG treatment resulted in a robust infiltration of CD8⁺ cells (Fig. 7G). Remarkably, Ythdf2 KO enhanced CD8⁺ cytotoxic T-cell recruitment in combination with BCG (Fig. 7H). Most studies use MB49 or MBT-2 cells to investigate the therapeutic efficacy of BCG immunotherapy. Therefore, we repeated this experiment with MBT-2 cells, we also found that YTHDF2 deficiency improves the therapeutic efficacy of BCG immunotherapy and contributes to a prolonged survival rate, which is in line with the results of our experiments with MB49 cells (Supplementary Fig. S7A). Confocal and flow cytometry analysis also revealed that Ythdf2 KO tumors showed a higher percentage of CD8 and CD11b-positive cells (Supplementary Fig. S7B–S7D). Collectively, our data indicate that YTHDF2 deficiency benefits BCG immunotherapy by increasing CD8⁺ T-cell infiltration both in the TME and the urothelium (Fig. 7I).

Given that RIG-I is a functionally important target of YTHDF2, and YTHDF2 promotes bladder cancer progression by suppressing RIG-I–mediated immune response (Figs. 4–6). We speculated that RIG-I may also play a vital role in YTHDF2 mediating BCG immunotherapy. To further characterize the function of RIG-I to BCG response, we first analyzed the mRNA levels of DDX58 in WT and YTHDF2 KD MGHU3 cells treated by BCG. We found that BCG treatment considerably induced DDX58 expression in both WT MGHU3 and 5637 cells at 24 hours post infection. Moreover, in the YTHDF2 KD MGHU3 cells, BCG infection triggered remarkably strong expression of RIG-I (Supplementary Fig. S8A). We also performed this analysis using GEO data (GSE176178). We found that BCG response patients showed higher expression of RIG-I compared with nonresponder (Supplementary Fig. S8B). In Supplementary Fig. S5E, we showed that RIG-I was positively correlated with CD4⁺ T cells, CD8⁺ T cells, neutrophils, and myeloid DCs in TCGA bladder carcinoma dataset, which is similar to the immune response mediated by YTHDF2. Notably, the expressions of YTHDF2 and DDX58 were found to be significantly correlated (Supplementary Fig. S8C). In summary, the data suggest that YTHDF2 facilitates BCG immunotherapy by RIG-I–mediated immune response.

Discussion

In the present study, we demonstrate that YTHDF2, a main cytoplasmic m⁶A reader protein mediating mRNA decay, plays a critical oncogenic role in bladder carcinoma tumorigenesis. YTHDF2 is aberrantly upregulated in recurrent tumors. We show that YTHDF2 KD significantly impairs proliferation and promotes apoptosis of human bladder carcinoma cells. We further show that depletion of YTHDF2 significantly inhibits tumorigenesis in vivo. Analysis by RNA-seq and m⁶A-seq, followed by validation and functional studies, suggests that DDX58 is a critical target gene of YTHDF2. As an RNA-binding protein, YTHDF2 recognizes and specifically binds to the CDS of DDX58 mRNA, which in turn leads to the downregulation of DDX58 mRNA and protein. Moreover, our RNA decay and mutagenesis assays indicate that the m⁶A site in the CDS of DDX58 is essential for YTHDF2 to post-transcriptionally regulate its expression. Our study demonstrates that the YTHDF2–RIG-I axis plays a critical role in the pathogenesis of bladder carcinoma (Fig. 7F). Furthermore, we provide evidence that this axis likely also plays an essential role in mediating the immune response of bladder carcinoma to BCG treatment.

The urothelium is the first-line of defense against pathogens. It exerts multiple functions in terms of responding to injury and infection, including generating an immune response (40). Increasing evidence supports a connection between cancer cell dedifferentiation and immune evasion. During tumor progression, cancer cells use both cell-autonomous and non–cell-autonomous mechanisms to change their susceptibility to immune recognition and destruction, including dysregulation of cancer cell intrinsic IFN signaling pathways (41, 42). Growing evidence supports an important link between such alterations to cancer cells and innate immune signaling during tumor development (43, 44). Thus, focusing on the activation of innate immunity could be a promising avenue in controlling tumor progression (33, 45, 46). Of note, we found that YTHDF2 expression is associated with mutated FGFR3, suggesting it mainly functions in NMIBC and luminal papillary MIBC. Although the prognosis of these tumors is known to be relatively more favorable (47), a portion of them still suffer from malignant progression upon recurrence (48). A pathological study also confirmed that the mutant FGFR3 promotes immune evasion contributing bladder cancer tumorigenesis (36). In this study, we found that the m⁶A reader protein YTHDF2 directly binds to the CDS of DDX58 mRNA to accelerate its degradation. This, in turn, causes the downregulation of RIG-I protein, a major transducer of RNA sensing in the activation of the immune response, thus promoting bladder carcinoma recurrence. Our study provides evidence that recurrence of bladder carcinoma impairs immune signaling through m⁶A readers, and this knowledge broadens our understanding of the interactions between innate immunity and tumorigenesis.

Bladder cancer remains a highly prevalent and significant health problem worldwide, requiring intrusive surveillance and resulting in high health economic costs (4). Surgery, alone or in combination with other treatments, is used in more than 90% of cases. Early-stage cancers may be treated by removing the tumor and then by administering immunotherapy (BCG) or chemotherapy drugs directly into the bladder (2). Notably, whereas BCG is more effective than chemotherapy in patients with high-risk NMIBC, approximately 40% of patients do not respond to BCG treatment (49). Moreover, the world remains in a global BCG shortage and will for the next 3–5 years (41). Thus, there is an urgent need to develop more effective therapies to strengthen the effect of BCG, and converting the initial response into a durable cure. Pattern recognition receptors (PRR) of the innate immune system, which recognize conserved pathogen-associated molecular patterns, are gaining interest as targets (50). RIG-I is the best characterized receptor of intracellular PRRs, which sense cytosolic RNA and trigger downstream immunomodulatory responses. Targeting regulatory components that indirectly activate RIG-I is thus a novel preclinical strategy to eliminate cancer cells (51–54). Interesting translational approaches have been suggested to harness retinoic acid-inducible gene-I–like receptor (RLR)–mediated apoptosis to eliminate cells that are persistently infected with viruses and to kill cancer cells. Many studies have reported the use of RIG-I ligands in preclinical studies (55). Poly (IC:LC) has been incorporated as an adjuvant to enhance the clinical efficacy of experimental vaccines in multiple malignancies, including glioma, ovarian cancer, and pancreatic carcinoma (56, 57). The RIG-I agonist, SLR20, induces RIG-I signaling and impairs tumor progression in breast carcinoma (38). In a recent trial in NMIBC patients, intravesical administration of coxsackievirus A21 (CVA21) caused induction of RIG-I and upregulation of IFN-inducible genes (58). These findings suggest the importance of the functional integrity of RIG-I signaling in bladder carcinoma tumor cells and its potential to serve as a target. Herein, we found that the RIG-I pathway can be suppressed by the aberrant expression of YTHDF2 in tumor cells of bladder carcinoma. The depletion of YTHDF2 triggers the activation of RIG-I–mediated IFN-I signaling and cell death. Analysis of tumor-infiltrating immune cells of the bladder carcinoma RNA-seq of the TCGA database revealed a negative correlation between YTHDF2 mRNA level and the presence of immune cells, such as CD4⁺ and CD8⁺ T cells, neutrophils, and myeloid DCs, suggesting immunosuppression in bladder carcinoma. YTHDF2 KO MB49 cells orthotopically transplanted into the bladder lumen showed the enhanced recruitment of CD8⁺ T cells to both the TME and to the urothelium. Two recent studies using transcriptomic tumor profiling on NMIBC clinical specimens identified the inflamed TME as associated with improved recurrence-free survival after BCG immunotherapy: BCG nonresponsive patients possessed an immunologically “cold” TME, and early recurrent patients showed repressed IFNα and IFNγ hallmarks and inflammation (59, 60); these findings suggested that tumor-intrinsic immunosuppression may be an obstacle for BCG immunotherapy. Our results may provide a rationale for understanding why BCG therapy is not effective in a significant fraction of patients with bladder carcinoma. Given the importance of RIG-I signaling in cancer immunology, the design of bifunctional siRNA combining YTHDF2 silencing with RIG-I activation by introducing a triphosphate group at the 5′ end of YTHDF2 siRNA (ppp-siRNA) may be exploited in bladder carcinoma treatment (61, 62). Therefore, our study may contribute to improving the efficiency of classical immunotherapy.

In summary, we provide compelling in vitro and in vivo evidence that YTHDF2 plays a critical oncogenic role in cell proliferation and tumorigenesis in bladder carcinoma through binding of its target gene DDX58, which, in turn, inhibits downstream signaling cascades. Our study highlights the functional importance of the innate immunity in bladder carcinoma tumorigenesis. In addition, given the functional importance of cancer cell-intrinsic RIG-I signaling, targeting YTHDF2 signaling by selective inhibitors may represent a promising therapeutic strategy to treat bladder carcinoma, especially in combination with BCG treatment. As YTHDF2 has also been implicated in other types of cancers, our discoveries may have a broader impact for cancer biology and cancer immunotherapy.

Supplementary Material

Supplementary Data

Supplementary table ST3

Supplementary Data

Supplementary Data clean version

Authors' Disclosures

J.P. Thiery reports personal fees from Biosyngen Pte Ltd. Singapore outside the submitted work. No disclosures were reported by the other authors.

Authors' Contributions

L. Zhang: Conceptualization, data curation, funding acquisition, methodology, writing–original draft. Y. Li: Conceptualization, funding acquisition, investigation, methodology, writing–review and editing. L. Zhou: Funding acquisition, methodology, writing–review and editing. H. Zhou: Validation, investigation. L. Ye: Resources. T. Ou: Supervision, writing–review and editing. H. Hong: Resources. S. Zheng: Resources. Z. Zhou: Resources. K. Wu: Formal analysis, methodology. Z. Yan: Software. J.P. Thiery: Supervision, writing–review and editing. J. Cui: Supervision, funding acquisition. S. Wu: Supervision, funding acquisition.