Kdm6a Deficiency Activates Inflammatory Pathways, Promotes M2 Macrophage Polarization, and Causes Bladder Cancer in Cooperation with p53 Dysfunction

Diagnosed in more than 430,000 patients worldwide every year, bladder cancer is the fourth most common cancer in men and the 11th most common cancer in women (1). Ninety percent of bladder cancer is urothelial carcinoma, with localized bladder cancer broadly categorized into non-muscle–invasive bladder cancer (NMIBC) and muscle-invasive BC (MIBC; ref. 2). NMIBC accounts for approximately 75% of newly diagnosed bladder cancer cases, and recommended treatment strategies include transurethral resection and intravesical instillation of bacillus Calmette–Guérin (BCG), or chemotherapy (1). While NMIBC is generally not life-threatening, it is associated with a high rate of recurrence (50%–70%; refs. 3, 4). However, MIBC, which accounts for the remaining 25% of bladder cancer cases, has a poor prognosis despite the use of radical cystectomy or radiotherapy (5). In addition, the overall cost of bladder cancer management is high, owing primarily to the necessities of multiple interventions and long-term surveillance of patients with NMIBC, as well as perioperative costs for patients with MIBC undergoing cystectomy (6).

Histologically, bladder cancer can be categorized into papillary and nonpapillary types that correlate with distinct genetic alterations, although these are believed to arise from common progenitor cells. Papillary NMIBC (Ta and T1), which is characterized by frequent FGRF3 mutations, develops via epithelial hyperplasia, while nonpapillary NMIBC (carcinoma in situ; CIS), which is characterized by p53 mutations, develops via flat dysplasia. MIBC (T2 and above) is thought to develop mainly from non-papillary NMIBC (CIS; refs. 1, 7–9).

High-throughput DNA sequencing of bladder cancer samples revealed the widespread occurrence of mutations in epigenomic regulators, such as histone modifiers (7, 10, 11). Of these, KDM6A [lysine (K)-specific demethylase 6A], also known as UTX (ubiquitously transcribed tetratricopeptide repeat X chromosome), functions as a demethylase for lysine 27 on histone H3 (H3K27; ref. 12) and also contributes to the methylation of lysine 4 on histone H3 (H3K4) and acetylation of H3K27 as a component of COMPASS (complex of proteins associated with Set 1)-like complexes (13, 14). KDM6A is frequently mutated in patients with both NMIBC and MIBC (7, 10, 11). Indeed, studies demonstrated that KDM6A deficiency results in increased proliferation in bladder cancer cells (15, 16) and in enhanced susceptibility to bladder cancer induced by BBN (N-butyl-N-(4-hydroxybutyl) nitrosamine), a cigarette smoke–like mutagen, in mice (17). Recently, the contribution of KDM6A mutations to papillary NMIBC pathogenesis has been reported (11); however, the pathogenic mechanisms underlying KDM6A-mutated MIBC via nonpapillary NMIBC remain to be fully clarified.

To address this issue, we generated mice with a bladder urothelium-specific deletion of KDM6A. In addition, to analyze the multistep process involved in the carcinogenesis of MIBC via nonpapillary NMIBC, Kdm6a-deficient mice were crossed with mice heterozygous for p53, whose mutation/deletion frequently overlaps with the KDM6A mutation in MIBC specimens (10, 18), and offspring were subsequently exposed to BBN. Phenotypic, pathologic, molecular, cellular, and therapeutic findings with regard to the mice and their relevance to patients with bladder cancer are presented.

To generate urothelium-specific Kdm6a-deficient mice, Kdm6a^flox/flox mice, in which exons 11 and 12 in the Kdm6a gene were flanked by two loxP sites, were mated to UpkIICre transgenic mice that expressed Cre recombinase specifically in the urothelium (19). The resultant Kdm6a^flox/flox UpkIICre⁻ and Kdm6a^flox/flox UpkIICre⁺ mice (hereafter referred to as control [Ctrl] and Kdm6a-deficient [Kdm6a^Δ/Δ] mice, respectively) were further crossed with p53^+/− mice, which were provided by RIKEN BioResource Research Center (Tsukuba, Japan, RBRC01361). Four different genotypes of mice: Ctrl p53^+/+, Ctrl p53^+/−, Kdm6a^Δ/Δ p53^+/+, and Kdm6a^Δ/Δ p53^+/− were used in this study. All animal experiments were carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the Hiroshima University Animal Research Committee (permission no. 29–58).

Mouse bladder cancer cell line, MBT2, was purchased from JCRB Cell Bank (Japanese Collection of Research Bioresources Cell Bank, Osaka, Japan). Human bladder cancer cell line, RT4 and 5637, were provided by Dr. Mayuko Kanayama (Juntendo University, Tokyo, Japan) and the Pathology Core of the Bladder Cancer SPORE at MD Anderson Cancer Center (Houston, TX), respectively. Cell lines were authenticated by DNA fingerprinting using the AmpFISTR Amplification or AmpFISTR Profiler PCR Amplification protocols (Life Technologies) and were proven to be negative for Mycoplasma infection using CycleavePCR mycoplasma Detection Kit (Takara Bio Inc.).

For carcinogen-induced tumorigenesis, mice at 8–10 weeks of age were fed drinking water containing 0.05% BBN (Tokyo Chemical Industry) in light-shielded bottles ad libitum. Dissected bladders were sectioned sagittally, fixed in 10% buffered formalin, and subjected to pathologic analysis.

Isolated urothelium or bladder cancer cells were homogenized using an ultrasonic processer with RIPA buffer (Nacalai Tesque Inc.). Western blot and IHC analyses were performed as described previously (20, 21). For immunofluorescence staining, cultured cells were treated with 4% paraformaldehyde, followed by 0.5% Triton X-100. Samples on slides were imaged by Olympus confocal laser microscopy FV1000-D using five different fields of view per sample. ImageJ software (NIH; http://imagej.nih.gov/ij/) was used according to a standard procedure for the quantification of images. The primary antibodies used in these assays are as follows; anti-KDM6A (Cell Signaling Technology, #33510), anti-H3K27me3 (Cell Signaling Technology, #9733), anti-H3K4me1 (Active Motif #39297), anti-H3K27ac (Abcam, ab4729), anti-F4/80 (Bio-Rad, MCA497GA), anti-Ki-67 (Abcam, ab15580), anti-Stat3α (Cell Signaling Technology, #8768), anti-phospho-Stat3 (Cell Signaling Technology, #9145), and β-Actin (Sigma-Aldrich, A1978). For immunofluorescence staining, Alexa Fluor 488–conjugated anti-rabbit IgG (Invitrogen) antibody was used as the second antibody. The nuclei were counterstained with Hoechst 33342 solution (Dojindo).

The urothelium specimens or tumor tissues were minced and trypsinized. Flow cytometry was performed on a FACSCanto II flow cytometry system (Becton Dickinson). The antibodies used in this assay are as follows; FITC-conjugated anti CD44 (BD Biosciences 56859), APC-conjugated anti-BrdU (BD Biosciences 51-23619L), FITC-conjugated anti-F4/80 (BioLegend 123107), APC-conjugated anti-CD206 (BioLegend 14707), and PE/Cy7-conjugated anti-CD14 Antibody (BioLegend 123309). FlowJo software was used to analyze the data.

Total RNA was extracted with TRIzol reagents (Life Technologies) from isolated urothelium or bladder cancer cells and subjected to cDNA synthesis using SuperScript VILO (Invitrogen) according to the manufacturer's protocol. Copy DNA was run on a Step One Plus Real-Time PCR system (Applied Biosystems) with SYBR Green Real-Time PCR Master Mixes (Thermo Fisher Scientific) for quantitative real-time PCR. The results were normalized using hypoxanthine phosphoribosyltransferase (hprt) as an endogenous control. The primers used in these assays are as follows: Cxcl1 (5′-GCTGGGATTCACCTCAAGAA-3′ and 5′-AAGGGAGCTTCAGGGTCAAG-3′), Has1 (5′-GACGGAGAAGAGAGAATCCAGG-3′ and 5′-GATGATCGTGAGTGCTCGCC-3′), Arc (5′-AAGCAGCAGACCTGACATCC-3′ and 5′-TTCTCCTGGCTCTGTAGGCT-3′), Kdm6a (5′-CAGATCCAAATTCTGGCCAGTCC-3′ and 5′-CAGAAGCAAGTGCAGATACATGG-3′), Ccl2 (5′-CTGCTGTTCACAGTTGCCG-3′ and 5′-GCACAGACCTCTCTCTTGAGC-3′), Il6 (5′-TGCCTTCTTGGGACTGATGC-3′ and 5′-TGAAGTCTCCTCTCCGGACT-3′), human Il6 (5′-AGACAGCCACTCACCTCTTCAG-3′ and 5′-TTCTGCCAGTGCCTCTTTGCTG-3′), and human Ccl2 (5′-AGAATCACCAGCAGCAAGTGTCC-3′ and 5′-TCCTGAACCCACTTCTGCTTGG-3′).

Total RNA was extracted as described above, and then converted into libraries using a SureSelect Strand-Specific RNA Library Prep kit (Agilent Technologies). Transcriptome analysis was performed using a next-generation sequencer (HiSeq 2500; Illumina). The generated sequence tags were mapped onto the mouse genomic sequence (Mouse Genome Browser GRCm38/mm10). GSEA software was downloaded from Broad Institute (http://software.broadinstitute.org/gsea/downloads.jsp) and analyses were performed as described previously (22). Positively and negatively enriched gene sets are shown in Supplementary Tables S1–S4.

The plasmid, pX330-U6-Chimeric BB-CBh-hSpCas9, was used to express Kdm6a-targeted single guide (sg)RNAs in MBT2 cells. sgRNA sequences were designed as described using CRISPR direct (https://crispr.dbcls.jp). A px330 vector, with or without Kdm6a-targeted sgRNA sequences, was transfected into MBT2 cells, then the cells were single-cell sorted on a FACSAriaI II flow cytometry system (Becton Dickinson), sequentially passaged and subjected to sequencing to confirm alterations in the Kdm6a gene. Cells transfected with the empty vector were used as a control.

To determine cell proliferation and the cell cycle, mice were fed drinking water containing BrdU (3 mg/mL; Nacalai Tesque Inc) for 4 days. The isolated urothelial cells were stained with FITC-conjugated anti-CD44 and an APC BrdU Flow Kit (BD Pharmingen) according to the manufacturer's protocol. Kdm6a-knockout and control MBT2 cells were also treated with BrdU for 6 hours. These cells were analyzed using flow cytometry as described above.

Two different siRNAs against human Kdm6a (siKdm6a-1 and siKdm6a-2, Stealth RNAi, Thermo Fisher Scientific) or control siRNA (siCtrl, Silencer, Thermo Fisher Scientific) were transfected using Lipofectamine RNAiMAX reagent (Thermo Fisher Scientific).

Single-cell suspensions of Kdm6a-knockout and control MBT2 cells were seeded at a density of 1.0 × 10⁵ cells/well in serum-free DMEM/F12 (Gibco/Thermo Fisher Scientific) supplemented with 20 ng/mL EGF (Invitrogen), 20 ng/mL basic fibroblast growth factor (Invitrogen), 1% N-2 Supplement (Invitrogen), 2% B-27 Supplement (Invitrogen), and 1% penicillin and streptomycin. Cells were cultured for 2 weeks and then the number of tumorspheres was counted.

Eight-week-old C3H/HeNCrl mice (Charles River Japan Inc., Yokohama, Japan) were intraperitoneally injected with 3% Brewer thioglycollate medium and intraperitoneal macrophages were harvested 3 days after treatment.

Cell migration and M2 polarization assays were performed using a transwell coculture system in 24-well plates (Corning). For the cell migration assay, primary macrophages were seeded at a density of 1.0 × 10⁶ cells/well to top wells with an 8.0-μm pore size, cocultured with EV or KO cell–derived conditioned media (EV CM or KO CM) in bottom wells for 24 hours, and migrated cells were stained. For the M2 polarization assay, primary macrophages were seeded at the same density to the top wells with a 0.4-μm pore size, co-cultured with EV CM or KO CM in the bottom wells for 72 hours, which was exchanged every 24 hours, and then collected and subjected to flow cytometric analysis to examine M2 polarization.

For IL6- and CCL2-induced assays, recombinant mouse IL6 and CCL2 (MCP-1; PeproTech), were added in the serum-free media at the concentrations of 0.1, 1.0, or 10.0 ng/mL. Primary macrophages treated with IL6 and CCL2 were subjected to cell migration and M2 polarization assays as described above.

MBT2 cells were adjusted to 1.0 × 10⁴ cells/50 μL of Hank Balanced Salt Solution (Gibco) and mixed on ice with 50 μL of Matrigel (Corning) just prior to implantation. Cells were injected subcutaneously into 8-week-old C3H/HeNCrl mice (Charles River Japan Inc). Tumor volumes were measured with calipers and calculated as (w² × l)/2, where w = width and l = length. At the endpoint of the experiment, tumor weight was measured and tumor cells were subjected to flow cytometric and Western blot analyses.

MR16-1, an anti-mouse IL6 receptor antibody (Chugai Pharmaceutical Co., Ltd., Tokyo, Japan) and propagermanium, a CCR2 inhibitor (Sanwa Kagaku Kenkyusho Co., Nagoya, Japan), were used in this study. Mice were intraperitoneally injected with 100 μg MR16-1 or IgG isotype (three times a week) and fed normal chow or chow containing 0.005% propagermanium (Oriental Yeast Co., Ltd.) for 3 weeks.

The Cancer Genome Atlas datasets for patients with bladder cancer (18) were downloaded from the cBioPortal for Cancer Genomics (www.cbioportal.org) and Broad GDAC Firehose (http://gdac.broadinstitute.org). RNA-sequencing z-scores of gene expression were normalized, and patients were categorized as follows: KDM6A^high (patients with high order 25% of KDM6A z-scores; 101 patients) and KDM6A^low (patients with low order 25% of KDM6A z-scores; 101 patients). The correlation between two targeted gene expressions using normalized RNA-seq z-scores were analyzed using Spearman rank correlation coefficient. The significance of mutual exclusivity between the targeted genes was calculated on the cBioPortal website. A tissue composition analysis of total macrophages and M2 macrophages, assumed to reside in the tumor and tissue microenvironment, was analyzed using xCell (23, 24) and precalculated TCGA data (http://xCell.ucsf.edu/). The disease-free survival (DFS) of patients with bladder cancer was classified according to the absence or presence of targeted gene mutations and analyzed using log-rank tests.

All statistical analyses in our study were performed using GraphPad Prism 7. Fisher exact test, an unpaired t test, a χ² test, and a log-rank test were used to make comparisons between two groups. Multiple group comparisons were performed using a one-way ANOVA test. In vivo data are expressed as mean ± SEM. Significant differences are noted by asterisks as follows: *, P < 0.05; **, P < 0.01 and ***, P < 0.001.

RNA-seq and ChIP-seq data obtained in this study have been deposited under accession numbers DRA008016 and DRA008091, respectively in the DNA Data Bank of Japan (DDBJ) and BioSample.

To generate mice with a urothelium-specific deletion of Kdm6a, Kdm6a^flox/flox mice, in which exons 11 and 12 of the Kdm6a gene were flanked by two loxP sites, were crossed with UpkIICre mice in which Cre was specifically expressed in the urothelium (ref. 19; Fig. 1A). The deletion of the Kdm6a gene product in the urothelium was confirmed by quantitative (q)PCR (Fig. 1B) and Western blot analysis using an anti-KDM6A antibody (Fig. 1C). After a 1-year observation period, the bladders of Kdm6a^flox/flox UpkIICre⁻ (Ctrl) and Kdm6a^flox/flox UpkIICre⁺ (Kdm6a^Δ/Δ) mice were subjected to pathologic examination. No obvious changes were observed (left two panels of Fig. 1D and left two columns of Fig. 1E), indicating that Kdm6a deficiency per se is not sufficient to induce bladder cancer.

Previous reports and our analysis using TCGA have demonstrated that a mutation and/or deletion of the p53 gene frequently and significantly coexists with the KDM6A mutation in patients with MIBC (refs. 10, 18; Supplementary Fig. S1A). To analyze the cooperative role of Kdm6a dysfunction and p53 insufficiency, we crossed Ctrl and Kdm6a^Δ/Δ mice with mice heterozygous for p53 (p53^+/−). Analysis of Ctrl p53^+/− and Kdm6a^Δ/Δ p53^+/− mice at 1 year of age revealed that although no pathologic abnormalities were found in Ctrl p53^+/− mice (third panel of Fig. 1D and third column of Fig. 1E), more than half of Kdm6a^Δ/Δ p53^+/− mice developed dysplasia-CIS (fourth panel of Fig. 1D and fourth column of Fig. 1E), which was frequently associated with the peeling of dysplastic epithelial cells, a characteristic of bladder cancer (indicated by arrows in the bottom right panel of Fig. 1D), and thicker lamina propria (Supplementary Fig. S1B, middle). In addition, bladders of Kdm6a^Δ/Δ p53^+/− mice with dysplasia-CIS were macroscopically more vascularized than those with normal urothelium (Supplementary Fig. S1C).

KDM6A demethylases trimethylated H3K27 (H3K27me3) and also promotes monomethylated H3K4 (H3K4me1) and acetylated H3K27 (H3K27ac; refs. 13, 14). Thus, we analyzed H3K27me3, H3K4me1, and H3K27ac levels in the urothelia of mice with four different genotypes, Ctrl p53^+/+, Kdm6a^Δ/Δ p53^+/+, Ctrl p53^+/−, and Kdm6a^Δ/Δ p53^+/−. As shown in Fig. 1F and G, H3K27me3 levels were significantly higher in the Kdm6a^Δ/Δ urothelium than the Ctrl urothelium on a p53^+/− background and H3K4me1 levels were significantly lower in the Kdm6a^Δ/Δ urothelium than the Ctrl urothelium on both of p53^+/+ and p53^+/− backgrounds, whereas H3K27ac levels were comparable between the Kdm6a^Δ/Δ and Ctrl urothelia. In addition, IHC staining with an antibody to a macrophage marker, F4/80, exhibited a significant accumulation of macrophages in Kdm6a^Δ/Δ bladder specimen on a p53^+/− background (Fig. 1H and I). These results indicate that Kdm6a deficiency cooperates with p53 haploinsufficiency to induce dysplasia-CIS with increased H3K27me3 and decreased H3K4me1 levels, associated with an accumulation of inflammatory cells.

Kdm6a^Δ/Δ p53^+/− mice developed dysplasia-CIS but invasive carcinoma was not observed, suggesting that additional factors are required for disease progression. To address this possibility, mice with four different genotypes, Ctrl p53^+/+, Kdm6a^Δ/Δ p53^+/+, Ctrl p53^+/−, and Kdm6a^Δ/Δ p53^+/−, were treated with BBN and their bladders were histologically analyzed at 10 weeks after treatment (Fig. 2A). Interestingly, all Kdm6a^Δ/Δ p53^+/− mice developed dysplasia-CIS, of which about 30% progressed to MIBC (Fig. 2B, fourth panel and Fig. 2C, black bar in the fourth column). Staining with Ki-67, a proliferation marker, revealed that Ki-67-positive area was significantly increased in Kdm6a^Δ/Δ samples compared with Ctrl samples on a p53^+/− background (right two panels of Fig. 2D and right two columns of Fig. 2E). We also examined proliferation and stem cell properties of bladder cells by incorporation of bromodeoxyuridine (BrdU) and staining with anti-CD44, a bladder cancer stem cell marker, respectively, by flow cytometric analysis at 8 weeks after treatment (Fig. 2A; ref. 25). Interestingly, compared with Ctrl cells, Kdm6a^Δ/Δ cells showed a significant increase only in CD44-positive cells on a p53^+/+ background (Fig. 2F, middle), whereas they exhibited significant increases in BrdU-positive and BrdU/CD44-double positive cells in addition to CD44-positive cells on a p53^+/− background (Fig. 2G). These results indicate that Kdm6a deficiency expands the number of bladder cells with a cancer stem cell phenotype and that p53 haploinsufficiency confers the ability to proliferate, which cooperatively progresses to MIBC in combination with BBN.

We then investigated molecular changes underlying the development and progression of bladder cancer. To this end, RNA samples extracted from the urothelia of BBN-treated Ctrl p53^+/+, Kdm6a^Δ/Δ p53^+/+, Ctrl p53^+/−, and Kdm6a^Δ/Δ p53^+/− mice were subjected to sequencing. To identify early gene expression changes, we used samples at 4 weeks after BBN treatment (Fig. 3A) when pathologic changes were not observed.

To identify gene(s) targeted by Kdm6a deficiency, we first compared Ctrl and Kdm6a^Δ/Δ mice on a p53^+/+ background. Of the upregulated genes, the top five most abundant transcripts in Kdm6a^Δ/Δ mice were Has1, Ccl2, Arc, Il6, and Cxcl1 (Fig. 3B, left); enhanced expression levels of these genes in the Kdm6a^Δ/Δ urothelium were confirmed by qPCR (Fig. 3B, right). To analyze pathway changes, the results of the transcriptome were subjected to GSEA (26). Because 3 of the top 5 upregulated genes, Ccl2, Il6, and Cxcl1, belong to the cytokine/chemokine family, we focused on pathways related to cytokine/chemokine signaling. Interestingly, in the analysis using Kyoto Encyclopedia of Genes and Genomes (KEGG) gene sets, we found significant positive enrichment for “KEGG CYTOKINE-CYTOKINE RECEPTOR INTERACTION” and “KEGG CHEMOKINE SIGNALING PATHWAY” in the Kdm6a^Δ/Δ urothelium (Fig. 3C, arrows in the left panel). Of note, in the “KEGG CYTOKINE-CYTOKINE RECEPTOR INTERACTION” gene set, we found that Ccl2, Il6, and Cxcl1, to be the top 3 upregulated genes (Fig. 3C, underlined in the right panel), indicating that enhanced expression of these genes plays a fundamental role in disease pathogenesis.

To investigate whether these changes were a direct effect of KDM6A deficiency, we performed RNA-seq analysis using the BBN-untreated urothelia from Ctrl and Kdm6a^Δ/Δ mice and the results were subjected to GSEA. In accordance with the results shown in Fig. 3C, the “KEGG CYTOKINE-CYTOKINE RECEPTOR INTERACTION” and “KEGG CHEMOKINE SIGNALING PATHWAY” were also significantly enriched in the Kdm6a^Δ/Δ urothelium compared with the Ctrl urothelium (Supplementary Fig. S1D), indicating that the activation of these pathways was essentially induced by Kdm6a deficiency itself, irrespective of BBN. In addition, we analyzed the role of BBN in the expression profiles of top two upregulated proinflammatory genes, Ccl2 and Il6 (Fig. 3C, right). The expression of these genes in Ctrl and Kdm6a^Δ/Δ urothelia without or with BBN treatment were assessed by FPKM values of RNA-seq. As shown in Supplementary Fig. S1E, expression levels of both genes were very low without BBN [BBN(−)] but induced by BBN treatment [BBN(+)], and interestingly, the increase rate of the expression of both genes by BBN was much greater in the Kdm6a^Δ/Δ urothelium than in the Ctrl urothelium.

Unexpectedly, two pathways concerning cell proliferation, “KEGG DNA REPLICATION” and “KEGG CELL CYCLE”, were negatively enriched in the Kdm6a^Δ/Δ urothelium (Fig. 3C, dotted arrows in the left panel). Thus, to analyze the effect(s) of Kdm6a deficiency and p53 haploinsufficiency on cell proliferation, we compared the results of “KEGG DNA REPLICATION” and “KEGG CELL CYCLE” pathways as follows: (a) Ctrl p53^+/+ versus Kdm6a^Δ/Δ p53^+/+, (b) Ctrl p53^+/− versus Kdm6a^Δ/Δ p53^+/−, (c) Ctrl p53^+/+ versus Ctrl p53^+/−, and (d) Kdm6a^Δ/Δ p53^+/+ versus Kdm6a^Δ/Δp53^+/−. Of these, (a) and (b) compare Ctrl and Kdm6a^Δ/Δ on p53^+/+ and p53^+/− backgrounds, respectively, and (c) and (d) compare p53^+/+ and p53^+/− on Ctrl and Kdm6a^Δ/Δ backgrounds, respectively (Fig. 3D, top). As shown in the lower panels of Fig. 3D, both pathways were negatively enriched by Kdm6a deficiency on both p53^+/+ and p53^+/− backgrounds (gray bars, also see the top four panels of Supplementary Fig. S1F), but conversely were positively enriched by p53 haploinsufficiency on both Ctrl and Kdm6a^Δ/Δ backgrounds (black bars, also see bottom four panels of Supplementary Fig. S1F). The overall upregulated and downregulated pathways in (a)–(d) are listed in Supplementary Tables S1–S4. These results collectively indicate that Kdm6a deficiency activates cytokine and chemokine pathways and BBN enhances the expression of proinflammatory genes, whereas p53 haploinsufficiency promotes cell proliferation.

We next analyzed the effects of Kdm6a deficiency on proliferation and gene expression changes using breast cancer cells. MBT2, a mouse breast cancer cell line (27), was used in this study. MBT2 is deficient in p53 (28) and our sequencing data for Kdm6a revealed no mutations or deletions in the coding region (not shown); thus, we attempted to generate Kdm6a-deficient MBT2 cells by the CRISPR/Cas9 system (Fig. 4A, left). We established four independent clones in which both Kdm6a alleles were homozygously mutated (KO clone-1–4 in the right panel of Fig. 4A), and the absence of KDM6A protein in these clones was confirmed by Western blot (Fig. 4B). Empty vector (EV)-introduced clones, which showed no mutation and expressed KDM6A, were used as a control (EV clones, Fig. 4A and B).

KDM6A is known to directly demethylase H3K27me3 through the Jmjc domain (12) and also contribute to H3K4me1 and H3K27ac (13, 14). Thus, we examined differences in H3K27me3, H3K4me1, and H3K27ac levels between EV and KO cells. As shown in Supplementary Fig. S2A, KO cells exhibited significantly increased H3K27me3 and significantly decreased H3K4me1 levels compared with EV cells, whereas H3K27ac levels were comparable. Because the most prominent gene expression changes in the Kdm6a-deficient urothelium were the enhanced expression of proinflammatory cytokines/chemokines, we examined the expression levels of Ccl2 and Il6, the top two most upregulated genes in the Kdm6a^Δ/Δ urothelium (Fig. 3C). In agreement with the results shown in Fig. 3C, upregulation of Ccl2 and Il6 were detected in KO clones (Fig. 4C). The upregulation of both genes was also detected by silencing KDM6A using siRNA in human bladder cancer cell lines, RT4 and 5637 (Supplementary Fig. S3A), both of which have wild-type KDM6A (http://cancer.sanger.ac.uk/cell_lines).

We then analyzed EV and KO cells in terms of stemness and proliferative activity. We first performed a sphere-forming assay that identifies stem cell potential. As shown in Fig. 4D, KO cells generated a significantly higher number of colonies compared with EV cells in both small-sized (100 μm∼500 μm) and large-sized (>500 μm) groups. In agreement with this result, KO cells expressed significantly more CD14, a bladder cancer stem cell marker (ref. 29; Supplementary Fig. S2B), and the enhanced expression of CD14 and other basal stem cell markers, Krt5 and Krt14 (30, 31), were also detected in the Kdm6a^Δ/Δ urothelium (Supplementary Fig. S2C). In addition, KO cells exhibited a reduced proliferation rate and were more commonly found in the G₀–G₁ phase and less commonly in the S-phase in the cell cycle compared with EV cells (Supplementary Fig. S2D and S2E). These results indicate that Kdm6a deficiency promotes stem cell activity in MBT2 cells, as was observed in the Kdm6a^Δ/Δ urothelium.

We next examined the effect of EV and KO cells on the behaviors and characteristics of macrophages, which play important roles in tumorigenesis (32, 33). Primary macrophages isolated from wild-type mice were cocultured with conditioned medium (CM) from EV and KO cells (EV CM and KO CM). In a Boyden chamber migration assay (34), macrophages cocultured with KO CM exhibited significantly enhanced migration compared with those with EV CM (Fig. 4E). Similar results were obtained using THP-1, a human monocyte/macrophage cell line, cocultured with the CM of siKDM6A-treated RT4 and 5637 cells (Supplementary Fig. S3B). In addition, staining cells for F4/80 and CD206, an M2 macrophage marker (35), revealed that the percentage of F4/80⁺, CD206⁺ cells among macrophages cocultured with KO CM was significantly higher those with EV CM (Fig. 4F).

To examine whether IL6 and CCL2 were the main source to drive macrophage migration and M2 macrophage polarization, we analyzed primary macrophages treated with various concentrations of recombinant IL6 and CCL2. As shown in Fig. 4G and H, we found that the migrated area and the percentages of F4/80⁺, CD206⁺ cells increased in a dose-dependent manner, and IL6 and CCL2 at 10 ng/mL significantly accelerated the migration and M2 polarization of primary macrophages. These results indicate that the increased migration ability and promoted M2 polarization induced by the CM from KO clones (Fig. 4E and F) were mainly due to the direct effect of enhanced expression of Il6 and Ccl2 (Fig. 4C).

We finally proceeded to in vivo tumorigenic assays. Taking advantage that MBT2 was established from inbred C3H/HeNCrl (C3H) mice, EV and KO cells were subcutaneously transplanted into CH3 mice and tumor sizes were measured. We found that KO tumors grew significantly faster than EV tumors (Fig. 4I, left), and at the endpoint, the mean weight of KO tumors was significantly greater than that of EV tumors (Fig. 4I, middle and right). In addition, as shown in Fig. 4J, the percentage of F4/80⁺, CD206⁺ M2 macrophages in KO tumors was significantly higher than in EV tumors. We then examined the activation of the STAT3, a major target of inflammatory responses (36). We found that the phosphorylation of STAT3 was significantly increased in KO tumors compared with EV tumors (Fig. 4K). In addition, using GSEA, the “KEGG JAK STAT SIGNALING PATHWAY” was found to be significantly enriched in the Kdm6a^Δ/Δ urothelium on both p53^+/+ and p53^+/− backgrounds (Supplementary Fig. S2F). Moreover, “HALLMARK IL6 JAK STAT3 SIGNALING” and “HALLMARK INFLAMMATORY RESPONSE” were significantly enriched in the Kdm6a^Δ/Δ urothelium without BBN treatment (Supplementary Fig. S2G). These results collectively indicate that Kdm6a deficiency conferred cancer stem cell characteristics and a growth advantage on bladder cancer cells by promoting M2 macrophage polarization and activating proinflammatory STAT3 signaling, possibly due to increased H3K27me3 and decreased H3K4me1.

We then investigated the therapeutic potential of the inhibition of IL6 and/or CCL2 on the growth of bladder cancer, with or without a Kdm6a deficiency. Propagermanium, an inhibitor of CCR2 that is a receptor for CCL2, and MR16-1, a neutralizing antibody against the mouse IL6 receptor, were used in this study.

Mice transplanted with EV or KO cells were treated with vehicle, propagermanium, MR16-1, or a combination of propagermanium and MR16-1. The therapy was started 7 days after transplantation when the cells were engrafted, and tumor volumes were measured every 7 days until day 28 (Fig. 5A). In mice transplanted with EV cells, no significant changes were observed in tumor growth and in the mean tumor weight at the endpoint (Fig. 5B, top). In contrast, in mice transplanted with KO cells, tumors treated with the combination therapy exhibited markedly retarded growth compared with those treated with vehicle, propagermanium, or MR16-1 (Fig. 5B, bottom left). At the endpoint, the mean weight of tumors treated with combination therapy was significantly lower than that of tumors treated with vehicle (Fig. 5B, bottom right). In addition, the analysis of F4/80⁺, CD206⁺ M2 macrophages in EV tumors revealed that the mean percentages of positive cells were comparable among all groups (Fig. 5C, top). In contrast, KO tumors treated with MR16-1 and combination therapy showed significantly reduced numbers of F4/80⁺, CD206⁺ M2 macrophages compared with those treated with vehicle (Fig. 5C, bottom). These results collectively demonstrate that the dual inhibition of IL6 and CCL2 effectively suppresses the growth of bladder cancer cells with a Kdm6a deficiency, possibly by preventing the inflammatory response in the tumor microenvironment.

We finally explored the clinical relevance of our model using genomic and clinical data of patients with MIBC available from TCGA (18). Patients were categorized on the basis of a normalized RNA-seq z-score for KDM6A expression as either KDM6A^high (patients with the highest 25% of z-scores; 101 patients) or KDM6A^low (patients with the lowest 25% of z-scores; 101 patients), which would correspond to our Ctrl and Kdm6a^Δ/Δ mice, respectively. The normalized RNA-seq z-scores for Ccl2 and Il6 expression, the top two genes upregulated by Kdm6a deficiency (Fig. 3C), in KDM6A^low were significantly higher compared with KDM6A^high (Fig. 6A). In addition, GSEA using KEGG gene sets revealed that “KEGG CYTOKINE–CYTOKINE RECEPTOR INTERACTION” pathway was significantly positively enriched in KDM6A^low (Fig. 6B). We then estimated the enrichment of stromal and immune cells in the tumor microenvironment using the xCell webtool (23, 24), which employs a compendium of gene signatures with the removal of dependencies between closely related cell types. The estimated xCell scores of total and M2 macrophages were significantly higher in KDM6A^low compared with patients with KDM6A^high MIBC (Fig. 6C). These results show that the phenotypes in human MIBC samples with KDM6A^low closely correspond to those in Kdm6a^Δ/Δ mice and Kdm6a KO clones (Figs. 1–4). In addition, we found that the KDM6A mutation was significantly and positively correlated with KDM6A^low compared with KDM6A^high (Fig. 6D). Thus, the findings in our mice would be applicable not only for human bladder cancer with KDM6A^low but also with a KDM6A mutation.

Finally, given that the histologic findings in Kdm6a^Δ/Δ p53^+/− mice closely resemble human non-papillary NMIBC and MIBC (Figs. 1 and 2), we examined whether a KDM6A mutation and/or a p53 mutation affects disease-free survival (DFS). We found that although the frequency of the KDM6A mutation did not differ between the two subtypes (Fig. 6E), that of the p53 mutation was significantly higher in the nonpapillary group (Fig. 6F). Interestingly, the KDM6A mutation was significantly correlated with reduced DFS in the nonpapillary group (Fig. 6G), whereas the p53 mutation did not affect DFS in either subtype (Fig. 6H). These findings suggest that although mutations in KDM6A and p53 frequently and significantly coexist in human MIBC samples (refs. 10, 18; Supplementary Fig. S1A), the presence of the KDM6A mutation predicts poor DFS in patients with nonpapillary MIBC, possibly by the increase of cancer stem cells (Figs. 2F, G, and 4D; Supplementary Fig. S2B and S2C) and resultant therapeutic resistance and tumor recurrence.

Comprehensive molecular characterization of MIBC has revealed that chromatin modifying genes are more frequently altered than in any other cancer (7, 10, 11, 18). This strongly suggests that epigenetic deregulation is particularly important in the pathogenesis of bladder cancer. However, the precise underlying mechanisms remain largely unknown.

In this report, we focused on KDM6A, which encodes a histone modifier and is frequently mutated genes in bladder cancer (7, 10, 11). By generating and analyzing urothelium-specific Kdm6a-deficient mice, we demonstrated that Kdm6a deficiency upregulated proinflammatory cytokines/chemokines (Figs. 3B and C and 4C; Supplementary Fig. S1D; Supplementary Fig. S2F and S2G), promoted M2 macrophage polarization (Figs. 1H and I and 4F, and H), and increased the proportion of cells harboring a cancer stem cell phenotype (Figs. 2F and G and 4D; Supplementary Fig. S2B and S2C). In addition, Kdm6a deficiency led to dysplasia-CIS in cooperation with p53 dysfunction (Fig. 1D and E), and this progressed to MIBC with BBN administration that markedly upregulated the expression of proinflammatory genes on a Kdm6a-deficient background (Fig. 2B and C; Supplementary Fig. S1E). It is noteworthy that pathways related to cell proliferation, such as “KEGG DNA REPLICATION” and “KEGG CELL CYCLE,” were rather suppressed by Kdm6a deficiency alone, but in turn activated when p53 dysfunction was introduced (Fig. 3D; Supplementary Fig. S1F). These results indicate that Kdm6a deficiency induces proinflammatory cytokine and chemokine production in the microenvironment and BBN aggravates inflammatory conditions, whereas p53 inactivation confers a proliferative advantage to tumor progenitor cells, which cooperatively contribute to the initiation and progression of bladder cancer. Our findings demonstrate the distinct roles of Kdm6a deficiency and p53 inactivation and also an additional role of BBN in bladder cancer development, which provides insights into the complex and multistep process of bladder cancer pathogenesis.

Inflammation plays a fundamental role in tumorigenesis (37, 38). Cytokines and chemokines produced by tumor cells induce remodeling of the microenvironment through the recruitment of various types of inflammatory and immune cells. Such diverse cells, in turn, secrete cytokines and chemokines and support the survival and proliferation of tumor cells, with this inflammatory circuit cooperatively promoting tumorigenesis (37, 38). The most important environmental cells are tumor-associated macrophages (TAM). TAMs with an M2 phenotype constitute a significant proportion of tumor-infiltrating cells and contribute to tumor growth, invasion, and metastasis by promoting angiogenesis and suppressing the antitumor immunoresponse (32, 33). In fact, the accumulation of TAMs in tumor tissues correlates with invasiveness, therapy resistance, and a poor outcome of bladder cancer (39–41). Interestingly, the combination of IL6 and CCL2, the most abundantly expressed and markedly upregulated gene products in Kdm6a deficiency with BBN (Figs. 3B and C and 4C; Supplementary Fig. S1E), efficiently converted myeloid cells into M2 macrophages (35). Thus, Kdm6a deficiency contributes to the development of bladder cancer, mainly through IL6- and CCL2-mediated M2 macrophage polarization and STAT3 activation (Figs. 1H and I and 4E, H, J, and K; Supplementary Fig. S2F and S2G).

It remains to be clarified how Kdm6a deficiency induces the upregulation of Il6 and Ccl2. KDM6A demethylases H3K27me3 per se (12) and also functions as a component of a COMPASS-like complex that promotes H3K4me1 and H3K27ac (13, 14). We found that Kdm6a deficiency resulted in increased H3K27me3 and reduced H3K4me1 levels in the urothelium and also in bladder cancer cells (Fig. 1F, and G; Supplementary Fig. S2A). Because H3K27me3 and H3K4me1 are regarded as a repressive and active histone marks, respectively, both of increased H3K27me3 and reduced H3K4me1 are considered to contribute to gene repression. Therefore, enhanced expression levels of Il6 and Ccl2 are assumed to be an indirect effect of Kdm6a deficiency. Indeed, we compared the H3K27me3 and H3K4me1 levels in the regulatory regions of both genes between EV and KO cells, but no obvious differences were observed (not shown). A global search of the changes in ChIP analysis for the changes of H3K27me3, H3K4me1, and/or H3K27ac levels will be necessary to identify genes responsible for the upregulation of Il6 and Ccl2 induced by Kdm6a deficiency.

It is notable that in addition to KDM6A, loss-of-function mutations of several histone-modifying genes have been found in bladder cancer specimens (7, 10, 11). These include genes encoding methyltransferases such as KDM2D, KDM2C, and KDM2A (also known as MLL4/MLL2, MLL3, and MLL1, respectively). Intriguingly, the protein products of such genes (MLL4/MLL2, MLL3, and MLL1) are a core subunit of COMPASS-like complexes that incorporate KDM6A (13, 14). This leads us to the idea that mutations in KDM6A and MLL genes commonly contribute to bladder cancer pathogenesis through impairing the function of COMPASS-like complexes. The finding that mutations in KDM6A and MLL genes are almost mutually exclusive (7, 10) supports this idea. Indeed, we found significantly decreased H3K4me1 levels as well as increased H3K27me3 levels in the Kdm6a^Δ/Δ urothelium and Kdm6a-KO clones (Fig. 1F, and G; Supplementary Fig. S2A). Thus, it is strongly suggested that dysfunction of KDM6A contributes to bladder cancer development in a demethylase activity–independent (namely, COMPASS-like complex formation-dependent) manner in addition to a demethylase activity–dependent manner, as proposed in previous studies (16, 17). It is intriguing to know whether mice lacking Mll genes in the urothelium also exhibit inflammatory phenotypes with the activation of cytokine and chemokine pathways, as observed in our Kdm6a^Δ/Δ mice.

Until the recent introduction of immune checkpoint inhibitors (42, 43), standard therapies for bladder cancer remained unchanged for three decades and no molecular-based therapies have been developed (1, 44, 45). We examined the effect of blocking IL6 and/or CCL2 signaling on the proliferation of bladder cancer cells and demonstrated that a combination of an anti-IL6 receptor antibody and a CCR2 inhibitor efficiently suppressed the growth of Kdm6a-deficient bladder cancer cells in vivo, associated with a significant reduction of M2 macrophages in tumor tissues (Fig. 5). We also found that in patients with bladder cancer, the low expression of KDM6A was associated with activated cytokine and chemokine pathways. The expression levels of KDM6A negatively correlated with those of IL6 and CCL2, and were accompanied by increases in total and M2 macrophages, as was the case for mice (Fig. 6A–C). These results lead to the idea that this combination therapy may be effective in bladder cancer patients with KDM6A deficiency. Given that both drugs have already been approved for clinical use [e.g., anti-human IL6 receptor antibody (tocilizumab) for rheumatoid arthritis and a CCR2 inhibitor (propagermanium) for chronic hepatitis B], our therapeutic strategy will be readily applicable to clinical settings. Interestingly, a previous report described that the combined inhibition of IL6 and CCL2 retarded the growth of breast cancer cells (46). Thus, the administration of cytokine and/or chemokine antagonists may be a promising molecular-targeted therapy for various types of inflammation-based cancers.

Bladder cancer is about three to four times more prevalent in men than in women (17, 47). KDM6A is an X chromosome–specific gene and KDM6C (also known as ubiquitously transcribed tetratricopeptide repeat, Y chromosome [Uty]), a male counterpart of KDM6A, exists on the Y chromosome. Unlike KDM6A, mutations of KDM6C have rarely been identified (11), but loss of KDM6C is reported to be a common event in bladder cancer (16, 48). In fact, the simultaneous loss of both KDM6A and KDM6C was found in bladder cancer samples from men (16), and the introduction of KDM6A–KDM6C double mutations into male bladder cancer cell lines induced hyperproliferation (15). Interestingly, although KDM6A and KDM6C are highly homologous, KDM6C possesses very low demethylase activity for H3K27 (49). Thus, it is strongly suggested that KDM6A and KDM6C concomitantly suppress bladder cancer development in demethylase activity-independent and dosage-dependent manners. It is intriguing to note that KDM6C also functions as a subunit of a COMPASS-like complex and male mice deficient for both Kdm6a and Kdm6c develop bladder cancer associated with inflammation.

In summary, we generated and analyzed Kdm6a-deficient mice and investigated the mechanism(s) involved in how dysfunctional KDM6A contributes to bladder cancer development (Supplementary Fig. S4). Our findings provide novel insights into the pathogenesis of this heterogeneous and complex disease, and may present a novel therapeutic approach for bladder cancer with KDM6A deficiency.

No potential conflicts of interest were disclosed.

Conception and design: K. Kobatake, K. Ikeda, T. Ueda, T. Ichinohe, Z.-I. Honda, A. Matsubara, H. Honda

Development of methodology: K. Kobatake, K. Ikeda, Y. Nakata, N. Yamasaki, A. Kanai, K. Sentani, P.C. Black, H. Honda

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Kobatake, Y. Nakata, N. Yamasaki, A. Kanai, Y. Sera, T. Hayashi, M. Koizumi, Y. Miyakawa, Y. Sotomaru, P.C. Black, H. Honda

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Kobatake, K. Ikeda, Y. Nakata, N. Yamasaki, A. Kanai, T. Inaba, W. Yasui, P.C. Black, H. Honda

Writing, review, and/or revision of the manuscript: K. Kobatake, K. Ikeda, Y. Nakata, T. Inaba, O. Kaminuma, S. Horie, P.C. Black, H. Honda

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Kobatake, K. Ikeda, Y. Nakata, A. Kanai, K. Sentani

Study supervision: H. Honda

This work was in part supported by JSPS KAKENHI (grant number 17K11170, to Y. Nakata; 17K16795, to K.-I. Ikeda; and 19H03693, to H. Honda), a Research Grant from the Princess Takamatsu Cancer Research Fund (to H. Honda), the TERUMO LIFE SCIENCE FOUNDATION (to H. Honda), and The Kao Foundation for Arts and Sciences (to Y. Nakata). We thank Yuki Sakai and Sawako Ogata for animal care, genotyping, and molecular experiments. We also thank Dr. Junji Takeda (Osaka University, Osaka, Japan), Dr. Mayuko Kanayama (Juntendo University, Tokyo, Japan), the RIKEN BioResource Center, and Chugai Pharmaceutical Co., Ltd. for providing us with KY1.1 ES cells, RT4 cells, B6.Cg-Trp53<tm1Sia>/Rbrc mice (RBRC01361), and the anti-mouse IL6 receptor antibody, MR16-1, respectively.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.