KMT9 Controls Stemness and Growth of Colorectal Cancer

Colorectal cancer, which includes hereditary, sporadic, and colitis-associated forms, is one of the leading causes of cancer-associated deaths worldwide (1). Four distinct consensus molecular colorectal cancer subtypes (CMS1–4) have been defined based on gene expression signatures, DNA methylation status, somatic copy number alterations, miRNA regulation changes, and presence of genetic aberrations in tumor suppressor genes [e.g., tumor protein p53 (TP53), adenomatous polyposis coli protein (APC)] or oncogenes [e.g., kristen rat sarcoma viral oncogenes (KRAS); refs. 2–6]. To date, systemic therapeutic options for colorectal cancer include chemotherapy (adjuvant and neoadjuvant) and to a lesser extent, therapeutic antibodies directed against growth factor receptors, for example, vascular endothelial growth factor receptor (VEGFR; ref. 7). Despite treatment, 30% to 40% of human patients relapse and suffer from tumor recurrence (8). This has been attributed to the acquirement of genetic aberrations during therapy and survival of cancer stem/initiating cells (CSC; ref. 9). CSCs and adult intestinal stem cells in the healthy gut have similar characteristics with respect to their self-renewal and differentiation capacity (10). For example, leucine-rich repeat containing g-protein–coupled receptor 5 (LGR5), a well-established target of the WNT signalling pathway, is expressed in benign intestinal stem cells and also defined as a CSC marker because LGR5-expressing (LGR5⁺) tumor cells have a high clonogenic capacity (11–13). Currently, resistant CSC populations are poorly characterized and therapeutic strategies for targeting CSCs remain to be identified (14, 15). One important feature of CSCs is their dynamic ability to switch between proliferative or differentiated states by modulating gene expression, which suggests the existence of epigenetic regulation (16).

Histone methyltransferases (HMT) catalyze the transfer of a methyl group from S-adenosyl-methionine (SAM) to lysine or arginine residues of histones. Histone methylation regulates various biological processes including proliferation, cell cycle, and stemness (17). Aberrant expression of histone methytransferases contributes to global changes of the histone methylation landscape, which has been associated with colorectal cancer development, progression, and patient survival (18). Therefore, targeting epigenetic regulators such as HMTs has been proposed as therapeutic strategy for colorectal cancer (19–21).

Recently, we identified the novel histone lysine methyltransferase KMT9 (22). KMT9 functions as an obligatory heterodimer composed of KMT9α (also named N6AMT1) and KMT9β (also named TRMT112), and their interaction is required for SAM binding and methyltransferase activity (22). KMT9 monomethylates lysine 12 of histone H4 (H4K12me1), thereby controlling genes that regulate proliferation of prostate and lung cancer cells (22, 23). Of note, high levels of KMT9 have been associated with poor patient survival in prostate and lung cancer (22, 23). Here, we investigated the function of KMT9 in colorectal cancer in vitro in human and murine organoid systems, as well as in vivo in murine models of colorectal cancer. Our data demonstrate that KMT9 is an essential regulator of colorectal cancer cell proliferation and stemness, which establishes KMT9 as a potential therapeutic target for colorectal cancer.

pLenti6-miKMT9α was constructed by inserting the DNA sequence corresponding to a miRNA against human KMT9α into pLenti6/V5-DEST according to the manufacturer's instructions (Life Technologies). Cloning details can be obtained upon request. Details regarding the miRNA sequences used for the cloning can be found in the Supplementary Table S1.

Normalized The Cancer Genomes Atlas (TCGA) gene expression data were downloaded with TCGA-Assembler Version 2.0 (https://github.com/compgenome365/TCGA-Assembler-2;.rsem.genes.normalized_results) and CMS classification of human colorectal cancer samples was calculated using CMScaller, a package in R, as described previously (24).

Apc^fl/fl p53^fl/fl Kras^G12D/+ Kmt9α^fl/fl, Apc^fl/fl p53^fl/fl Kras^G12D/+ Kmt9α^+/+, Rosa26-CreERT2xKmt9α^wt/wt (Kmt9α^wt/wt), Rosa26-CreERT2xKmt9α^fl/fl (Kmt9α^ind-fl/fl), Villin1 (Vil1)-CreERT2xKmt9α^wt/wt (Kmt9α^IEC-wt/wt), Villin1 (Vil1)-CreERT2 xKmt9α^fl/fl (Kmt9α^IEC-fl/fl) mice were used for organoids generation and in vivo experiments. The mice were maintained in a temperature- and humidity-controlled animal facility with a 12-hour light/dark cycle and free access to water. Animals were sacrificed using cervical dislocation and tissues were immediately collected for further experiments.

Kmt9α^wt/wt, Kmt9α^ind-fl/fl, Kmt9α^IEC-wt/wt, and Kmt9α^IEC-fl/fl mice between 10 and 12 weeks of age were given an intraperitoneal injection of 10 mg/kg body weight of azoxymethane (AOM; Sigma). Each experiment was conducted using 24 mice from different litters [n = 12 control mice (Kmt9α^wt/wt/Kmt9α^IEC-wt/wt) and n = 12 Kmt9α^ind-fl/fl or Kmt9α^IEC-fl/fl mice]. One week after the first intraperitoneal AOM injection, animals were given ab libitum access to drinking water with 1.25% dextran sodium sulfate (DSS; MP Biomedicals) for seven days followed by another seven days of normal drinking water, for a total of 5 cycles. A second intraperitoneal AOM injection was given after the first cycle. For in vivo investigation of KMT9α, mice were injected with 1 mg of tamoxifen for five days and fed with tamoxifen containing food during the entire procedure. Body weight was measured once a week. Following the last cycle of normal drinking water, mice were sacrificed using cervical dislocation. Tumors were measured with a caliper and tumor volume was calculated by the formula V = 4/3 × 3.142 × ((width+length)/4)³. Tumors were fixed in 10% formalin for subsequent embedding or alternatively frozen in liquid nitrogen for subsequent analyses.

For establishment of healthy colon organoids, colonic crypts of C57BL/6 mice were isolated as described previously (25). For generation of APKK (Apc^KO/Kras^G12D/p53^KO/Kmt9α^KO) and APK (Apc^KO/Kras^G12D/p53^KO) organoids, colonic crypt of Apc^fl/fl p53^fl/fl Kras^G12D/+ Kmt9α^fl/fl, and Apc^fl/fl p53^fl/fl Kras^G12D/+ Kmt9α^+/+ mice were isolated as described previously (25). The deletion of the floxed sequences was mediated by infection with Cre-expressing adenovirus (BioCat GmbH). For generation of AOM/DSS tumor organoids, colonic tumors were excised from Kmt9α^wt/wt or Kmt9α^ind-fl/fl mice after AOM/DSS treatment. Tumor tissue was manually dissected and a single-cell suspension containing tumor stem cells was generated using the Tumour Dissociation Kit, mouse (Miltenyi Biotec) according to the manufacturer's protocol. Patient-derived organoids (PDO) were isolated from human colorectal cancer tissue using Tumor Dissociation Kit, human (Miltenyi Biotec) according to the manufacturer's protocol. The single cells obtained were resuspended in a solution containing growth factor-reduced Matrigel (Corning) and Advanced DMEM-F12 medium (Thermo Fisher Scientific) in a 1:1 ratio. For each dome, approximately 1,000 cells were seeded in a 50 μL drop of Matrigel/Advanced DMEM-F12 in 24-well plates. The Matrigel was allowed to polymerize at 37°C for 20 minutes and then covered with 600 μL of culture medium.

Healthy colon organoids and colorectal cancer PDOs were maintained in IntestiCult Organoid Growth Medium [Stemcell Technologies, catalog #06005 (for mouse organoids) and catalog #06010 (for human organoids)] supplemented with penicillin/streptomycin. Mouse tumor organoids were maintained in basal medium [Advanced DMEM-F12 supplemented with penicillin/streptomycin, HEPES 10 mmol/L (Invitrogen), Glutamax 1× (Invitrogen), N2 1× (Gibco), B27 1× (Gibco), and N-acetylcysteine 1 mmol/L (Sigma)]. For AOM/DSS tumor organoids, the basal medium was supplemented with 50 ng/mL EGF (Peprotech). After 3 to 5 passages, the organoids were frozen and cryopreserved as stocks for future experiments. In general, the organoids were used between passage numbers 7 and 15. For in vitro deletion of Kmt9α, AOM/DSS tumors organoids were treated with 1 μmol/L 4-hydroxytamoxifen or EtOH (vehicle) as a control. Mouse and human organoids were subcultured in Matrigel every 5–7 days or every 14 days, respectively.

Chromatin immunoprecipitation (ChIP) experiments were performed as previously described (26). Two days after seeding, AOM/DSS tumor organoids were incubated with 1 μmol/L tamoxifen or EtOH (vehicle) as a control. Five days after tamoxifen incubation, organoids were dissociated into single-cell suspension using TrypLE (Gibco). Cell pellets were washed twice with cold PBS, cross-linked with 1% PFA for 15 minutes at 4°C, and then rinsed twice with ice-cold PBS. The pellets were resuspended in TSE I buffer (20 mmol/L Tris-HCL pH 8, 2 mmol/L EDTA, 150 mmol/L NaCL, 0.1% SDS, and 1% Triton X-100) and sonicated for 1 hour at 4°C (Bioruptor, Diagenode). Immunoprecipitation was performed with GammaBind G-Sepharose beads (GE-Healthcare) and specific antibodies for anti-KMT9α (#27630, lot 20062017, Schüle Lab); anti-H4K12me1 (#27429, lot 27062017, Schüle Lab); anti-KMT9β (#28358, lot 03042018, Schüle Lab). Libraries were prepared from immunoprecipitated DNA according to standard methods. ChIP sequencing (ChIP-seq) libraries were sequenced using a HiSeq 2000 (Illumina) at the sequencing core facility of the MPI-IE, Freiburg. Reads were aligned to the mm10 build of the mouse genome using Bowtie 2 (RRID:SCR_016368; ref. 27). Data were further analyzed using the peak finding algorithm MACS 1.42 (28) using input as control. All peaks with FDR greater than 2.0% were excluded from further analysis. The reads were used to generate the genome-wide intensity profiles, which were visualized using the IGV genome browser (29). HOMER (RRID:SCR_010881; ref. 30) was used to annotate peaks (annotatePeaks.pl) and to calculate overlaps between different peak files (mergePeaks). The genomic features (promoter, exon, intron, 3′UTR, and intergenic regions) were defined using Refseq (RRID:SCR_003496). Seqplots (https://github.com/Przemol/seqplots) was used to visualize the signals in heatmaps. Data are deposited under GSE150506.

Two days after seeding, Kmt9α^ind-fl/fl AOM/DSS tumor organoids were treated with 1 μmol/L tamoxifen or EtOH (vehicle) as a control. Five days later, the organoids were dissociated into single-cell suspensions using TrypLE for 20 minutes at 37°C. After dissociation, single-cell suspensions were washed twice in PBS, centrifugated for 5 minutes at 300 × g, and counted using a LUNA automated cell counter (Logos Biosystems). Single cell capture, reverse transcription, and library preparation were carried out on the Chromium platform (10× Genomics) with the single-cell 3′ reagent v2 protocol according to the manufacturer's recommendations using 1 × 10⁴ cells as input per reaction well. The two final libraries [tamoxifen/EtOH (vehicle)] were pooled and sequenced on two Illumina NovaSeq SP lanes (paired-end 26 bp + 96 bp). Raw sequencing data were processed and aligned to the mouse genome (mm10) using the CellRanger pipeline (10x Genomics version 3.1, SCR_017344). Data are deposited under GSE150506. Previously published single-cell mRNA sequencing (scRNA-seq) data from 23 Korean patients with colorectal cancer (GSE132435) were analyzed using Seurat v3 as described above (31). Details regarding the analysis of scRNA-seq data can be found in the Supplementary Materials and Methods.

RNA from AOM/DSS organoids treated with tamoxifen or EtOH (vehicle), APK and APKK tumor organoids, miCtrl or miKMT9α transduced-PDO organoids, and normal colonic epithelial cells was isolated using RNeasy Mini columns (Qiagen). Total RNA from KMT9α-proficient and -deficient AOM/DSS tumors (n = 5 for each group) was isolated using TRIzol (Invitrogen). RNA samples were sequenced by the standard Illumina protocol to create raw sequence files (.fastq files) at Novogene, London. Reads were aligned to the mm10 build of the mouse genome using STAR version 2.7 (RRID:SCR_004463; ref. 32). The aligned reads were counted with HOMER software (RRID:SCR_010881; analyzeRepeats) and differentially expressed genes were identified using EdgeR (RRID:SCR_012802; ref. 33). RNA sequencing (RNA-seq) experiments from tumor organoids were performed in biological triplicates. P values <10^–6 were considered as statistically significant. For transcriptome analyses from whole-tumor tissue of AOM/DSS tumors, the 3,000 most significantly deregulated transcripts (2,306 genes) were used to perform gene set enrichment analyses (GSEA; RRID:SCR_003199). Data are deposited under GSE150506.

Fresh-frozen human tumor and adjacent healthy colonic tissue were provided by the Ontario Tumour Bank, which is supported by the Ontario Institute for Cancer Research through funding provided by the Government of Ontario. Tissue was kept on ice and manually dissected and minced for subsequent RIPA lysis. For total protein isolation from organoids, Matrigel-cultured organoids were harvested using TrypLE to break up the domes and were washed with PBS. Organoids and human tissue were lysed in ice-cold RIPA buffer (1 mmol/L EDTA, 50 mmol/L Tris-HCl pH7.5, 0.1% SDS, 150 mmol/L NaCl, 1% NP-40, 1% sodium deoxycholate) containing complete EDTA-free Protease Inhibitor Cocktail (Roche) for 10 minutes on ice. After centrifugation for 10 minutes at 13,000 rpm at 4°C, supernatant was collected and protein concentration was determined using Bradford assay. The list of the antibodies used for Western blot analysis can be found in the Supplementary Materials and Methods.

Experimental mice were housed in the pathogen-free barrier facility of the University Medical Center Freiburg in accordance with institutional guidelines and all experiments were approved by the regional board.

Human tumor organoids were established at Georg-Speyer-403 Haus from fresh human tumor tissue according to regional regulations and the experiments were approved by the regional ethics committee (Ethikkommission Universitätsklinikum Frankfurt/Main 274/18). Informed consent was obtained from all donors of tissue.

Data are represented as mean ± SEM. Significance was calculated by two-tailed Student t test, by one-way ANOVA, and Tukey multiple comparisons test as indicated in the figure legends. Statistical significance was set to P < 0.05 and is represented as following: ****, P < 0.0001; ***, P < 0.001; **, P < 0.01; *, P < 0.05; ns, not significant. Sample sizes are indicated where appropriate.

Additional methods including virus production and organoid transduction, organoid size assessment, colonic epithelial cells isolation, core histone isolation, list of the antibodies used for Western blot analysis, cell proliferation assay, flow cytometry, quantitative RT-PCR analysis, scRNA-seq, hematoxylin and eosin, and immunohistochemical staining and TUNEL assay can be found in the Supplementary Materials and Methods.

To ensure data availability, all RNA-seq, scRNA-seq, and ChIP-seq data have been deposited at GEO under GSE150506. All data that support the findings of this study are available from the corresponding authors upon reasonable request without any restrictions.

All unique materials are readily available from the authors without any restrictions.

To investigate whether KMT9 plays a functional role in colorectal cancer, we profiled KMT9α and KMT9β mRNA expression in healthy human colon and primary colon adenocarcinoma tissues. A large cohort TCGA (34) of 256 patients with colorectal cancer revealed a significant increase in both KMT9α and KMT9β mRNA in colorectal cancer tissue compared with healthy colon (Fig. 1A and B). CMS stratification revealed significant overexpression of KMT9α in CMS2-, CMS3-, and CMS4- but not in CMS1-tumors, whereas KMT9β mRNA was significantly increased in all subtypes (Supplementary Fig. S1A–S1C). Moreover, Western blot analyses showed that both KMT9α and KMT9β protein levels were strongly increased in human colorectal cancer tissue compared with patient-matched healthy colon (Fig. 1C). We therefore hypothesized that KMT9 might play a functional role in colorectal tumorigenesis.

Because KMT9α is indispensable for KMT9′s histone methyltransferase activity (22), we engineered mice with conditional Kmt9α alleles by flanking exon 2 and 3 with loxP sites (Kmt9α^fl/fl) to unravel potential functions of KMT9 in colorectal tumorigenesis. For initial experiments, Kmt9α^fl/fl mice were crossed to the Rosa26-CreERT2 deleter strain (35) to produce Kmt9α^ind-fl/fl mice for tamoxifen-inducible deletion of Kmt9α. Crosses of mice with Kmt9α wild-type (wt) alleles to Rosa26-CreERT2 mice, referred to as Kmt9α^wt/wt, served as controls. We treated Kmt9α^ind-fl/fl and Kmt9α^wt/wt mice with AOM and DSS to induce inflammatory colorectal tumor growth (36, 37) and established three-dimensional (3D) epithelial organoid cultures from AOM/DSS–associated tumors (Supplementary Fig. S1D). In the absence of tamoxifen treatment, KMT9α and KMT9β protein levels were increased in four independent AOM/DSS tumor organoids from Kmt9α^wt/wt (Ctrl) and Kmt9α^ind-fl/fl (#1, #2, #3) mice compared with healthy colon organoids (Supplementary Fig. S1E). This increase is in accordance with observations we made in human patient samples (Fig. 1A–C). Following tamoxifen treatment, KMT9α protein was efficiently depleted in the three independent Kmt9α^ind-fl/fl AOM/DSS tumor organoids compared with control Kmt9α^wt/wt AOM/DSS tumor organoids (Fig. 1D). Importantly, upon loss of KMT9α, AOM/DSS tumor organoids displayed a shrinkage in size (Fig. 1E), and the proliferation rate was strongly decreased (Supplementary Fig. S1F).

To investigate whether KMT9α loss affects tumor formation and growth in vivo, Kmt9α^fl/fl mice were crossed to the Villin1 (Vil1)-CreERT2 deleter strain (38), which allows tamoxifen-inducible deletion of Kmt9α specifically in intestinal epithelial cells (IEC) of Kmt9α^IEC-fl/fl mice (Fig. 1F). Tamoxifen-treated Kmt9α^IEC-fl/fl mice were compared with tamoxifen-treated Kmt9α^IEC-wt/wt mice (hereafter termed Kmt9α^IEC-KO and Kmt9α^IEC-WT, respectively). Upon AOM/DSS treatment, Kmt9α^IEC-WT mice developed colorectal tumors as expected (36, 37). Immunohistochemical and qRT-PCR analyses of the tumors observed in Kmt9α^IEC-WT mice revealed a significant increase in KMT9α mRNA and protein levels in AOM/DSS tumors compared with adjacent healthy tissue (Fig. 1G; Supplementary Fig. S1G). In contrast to Kmt9α^IEC-WT mice, Kmt9α^IEC-KO animals displayed a dramatic decrease in tumor burden characterized by a reduced number of microscopic and macroscopic colon tumors per mouse as well as a significantly smaller tumor size and mass (Fig. 1H). The few small tumors found in Kmt9α^IEC-KO mice displayed reduced Kmt9α expression in comparison with tumors in Kmt9α^IEC-WT mice (Fig. 1I). Analysis of cell proliferation and apoptosis showed that KMT9α-depleted tumors had reduced levels of proliferation marker KI67 (Fig. 1J) and increased apoptotic activity (Fig. 1K). In the absence of AOM/DSS treatment, colon tissue from Kmt9α^IEC-KO and Kmt9α^IEC-WT mice did not show any apparent morphologic differences (Supplementary Fig. S1H and S1I). Moreover, proliferation analysis of healthy colon crypts by KI67 staining did not reveal any significant differences between Kmt9α^IEC-WT and Kmt9α^IEC-KO mice without AOM/DSS treatment (Supplementary Fig. S1J). More importantly, transcriptome analysis performed on purified epithelial colon cells from AOM/DSS treatment-naïve Kmt9α^IEC-WT and Kmt9α^IEC-KO mice revealed only 12 significantly differentially expressed genes in KMT9α-deficient compared with KMT9α-proficient colon, thereby emphasizing the specific role of KMT9α in colorectal tumor tissue (Supplementary Table S2). Together, these data demonstrate that inflammation-associated colorectal tumorigenesis in mice is controlled by KMT9α.

To gain mechanistic insight into KMT9α-mediated gene regulation, we determined the transcriptomes of vehicle- and tamoxifen-treated Kmt9α^ind-fl/fl AOM/DSS tumor organoids by RNA-seq. The intersection of the differentially expressed gene sets for the three Kmt9α^ind-fl/fl tumor organoids revealed a common pool of 1,183 KMT9α-dependent genes (Fig. 2A). GSEA for these 1,183 genes uncovered terms associated with “cell cycle” and “apoptosis” as significantly deregulated biological processes (Fig. 2B). Accordingly, we found a significant downregulation of numerous genes involved in cell-cycle control in KMT9α-depleted AOM/DSS tumor organoids (Fig. 2C). qRT-PCR analysis validated reduced expression of cell-cycle regulators such as aurora kinase b (Aurkb), e2f transcription factor 1 (E2f1), establishment of sister chromatid cohesion N-acetyltransferase 2 (Esco2), minichromosome maintenance complex component 6 (Mcm6), pcna clamp associated factor (Pclaf), proline rich 11 (Prr11), rad51 paralog c (Rad51c), replication protein a2 (Rpa2), and dna topoisomerase II alpha (Top2a) upon KMT9α loss (Supplementary Fig. S2A). To validate these findings, we analyzed whether KMT9α depletion in AOM/DSS tumor organoids resulted in changes to the cell-cycle phase distribution by flow cytometry. KMT9 depletion was associated with an increase in G₀–G₁ cells and a reduction of the S-phase population (Fig. 2D), which suggests that loss of KMT9 decreased cell proliferation. Importantly, these analyses also revealed an early apoptotic sub-G₀ population in the KMT9α-depleted tumor organoids, which is consistent with our in vivo findings (Fig. 1K). Analysis of the representative AOM/DSS tumor organoids #3 uncovered significant upregulation of proapoptotic genes upon KMT9α depletion (Supplementary Fig. S2B). Upregulation of genes such as phorbol-12-myristate-13-acetate-induced protein 1 (Pmaip1) and transforming growth factor beta 2 (Tgfb2) was verified by qRT-PCR (Supplementary Fig. S2C). Further supporting the notion that KMT9α depletion promotes apoptosis of AOM/DSS tumor organoids, we detected an increase in cleaved caspase-3 levels (Supplementary Fig. S2D) and in Annexin V–positive (Annexin V⁺) and DAPI-negative (DAPI^–) apoptotic cells (Supplementary Fig. S2E) upon loss of KMT9α. Because KMT9 writes the H4K12me1 histone mark, we asked whether ablation of KMT9α resulted in decreased H4K12me1 levels. Western blot analyses showed a strong decrease in H4K12me1 in KMT9α-depleted AOM/DSS tumor organoids #3 (Supplementary Fig. S2F). To investigate whether the differentially regulated cell-cycle genes identified were direct KMT9 target genes, we analyzed the genomic localization of KMT9α, KMT9β, and H4K12me1 by ChIP-seq in AOM/DSS tumor organoids #3. We uncovered 5,652 KMT9α, KMT9β, and H4K12me1 colocalizations (Fig. 2E) that were enriched around the transcription start site (TSS) of target genes (Fig. 2F). In total, we observed the presence of KMT9α, KMT9β, and H4K12me1 at the promoter of 3,239 genes (Fig. 2E). Intersection of these 3,239 targets with the 5,882 differentially expressed genes observed for AOM/DSS tumor organoids #3 uncovered 1,168 differentially expressed, direct KMT9 target genes (Fig. 2G). GSEA of the direct KMT9 targets revealed a significant enrichment of terms related to cell cycle (Fig. 2H). For instance, direct target genes with promoter presence of KMT9α, KMT9β, and H4K12me1 included cell-cycle regulators described above such as Aurkb, Mcm6, Prr11, Rad51c, and Rpa2 (Fig. 2I and C; Supplementary Fig. S2A). Importantly, global transcriptome analysis performed from whole colorectal tumor tissue obtained from AOM/DSS–treated Kmt9α^IEC-WT and Kmt9α^IEC-KO mice also revealed cell-cycle regulation by KMT9, which is consistent with our data obtained from AOM/DSS organoids (Supplementary Fig. S2G and S2H). Together, our data show that KMT9 directly controls cell-cycle progression with a concomitant control of apoptotic state.

CSCs are essential for tumor initiation and maintenance, and thereby responsible for tumor relapse and treatment failure in patients with colorectal cancer (7, 39). To investigate whether KMT9α depletion would impact CSC function, we analyzed the expression of intestinal stem cell markers in AOM/DSS tumors from Kmt9α^IEC-WT and Kmt9α^IEC-KO mice using a previously described Lgr5 intestinal stem cell (ISC) signature (40). GSEA revealed significant negative enrichment of the intratumoral ISC signature upon loss of KMT9α (Fig. 3A). Indeed, a total of 30 ISC-related genes were significantly downregulated in KMT9α-depleted compared with KMT9α-proficient tumors, notably Apcdd1, Dach1, and Rhobtb3, which are previously described regulators of stemness in colorectal cancer (Fig. 3B; refs. 41–43). Importantly, GSEA performed on vehicle- and tamoxifen-treated Kmt9α^ind-fl/fl AOM/DSS tumor organoids also uncovered a negative enrichment of the Lgr5 signature with 117 stem cell–related genes significantly downregulated upon KMT9α depletion (Fig. 3C and D). To elucidate the transcriptomic underpinnings of KMT9 function in the CSC population, we performed scRNA-seq on Kmt9α^ind-fl/fl AOM/DSS tumor organoids cultured in the presence of EtOH (vehicle) or tamoxifen. Using a droplet-based microfluidics platform, we obtained 2,340 KMT9α-proficient and 1,303 KMT9α-deficient cells after quality filtering (Supplementary Fig. S3A–S3D). By performing unsupervised clustering and two-dimensional embedding using uniform manifold approximation and projection (UMAP), we identified four tumor cell subpopulations based on the expression of established marker genes (11, 40, 44, 45) and the differentiation states by pseudotime analysis: a “stem cell” (STEM) population (cluster 1), “cycling progenitor” (CP) population (cluster 2), a “transit-amplifying” (TA) cell population (cluster 3), and a “differentiated” (DIFF) cell population (cluster 4; Fig. 3E; Supplementary Fig. S3E–S3L). Population-dependent gene expression analyses showed ubiquitous Kmt9α mRNA expression in all four cell populations and marked decrease in Kmt9α mRNA in KMT9α-deficient cells relative to KMT9α-proficient cells (Fig. 3F). Analysis of the cycle phase distribution based on the expression of cell-cycle marker genes (Supplementary Fig. S3M) was in agreement with the role of KMT9α in cell-cycle regulation described above (Fig. 2D). Furthermore, in accordance with the bulk RNA-seq data, we observed downregulation of essential S-phase genes such as Mcm6 and Rpa2 by scRNA-seq upon KMT9α depletion (Supplementary Fig. S3N). Interestingly, numerous stem cell–related genes listed in the previously described Lgr5+ ISC signature (40) were downregulated upon loss of KMT9α in both precursor populations (Fig. 3G and H). In addition, we observed increased expression of proapoptotic genes in KMT9α-deficient STEM and CP populations, emphasizing the relevant role of KMT9 in tumoral stem cell maintenance (Fig. 3I; Supplementary Table S3). These data suggest that loss of KMT9α might affect the self-renewal potential and colony formation capacity of tumor stem/initiating cells. To test this hypothesis, single cells isolated from KMT9α-proficient and deficient AOM/DSS tumor organoids were evaluated for anchorage-independent sphere formation in a secondary replating assay. Importantly, the secondary sphere formation capacity of cells derived from KMT9α-deficient organoids was blocked (Fig. 3J).

To corroborate these findings in vivo, we crossed Kmt9α^fl/fl mice with the Lgr5-EGFP-IRES-CreERT2 deleter strain (11) for tamoxifen-inducible selective Kmt9α ablation in intestinal Lgr5⁺ stem cells. Tamoxifen-treated mice were challenged with AOM/DSS and Lgr5-expressing CSCs were intratumorally traced in vivo using enhanced green fluorescent protein (EGFP) reporter. In accordance with our hypothesis, we found that the percentage of viable Lgr5-EGFP–positive (Lgr5-EGFP⁺) CSCs was markedly decreased upon Kmt9α loss in the collected tumors (Fig. 3K). All together, our data demonstrate the importance of KMT9 in maintaining the self-renewal potential and viability of tumor stem/initiating cells.

Besides inflammation-induced colorectal tumorigenesis, sporadic colorectal cancer results from the progressive accumulation of genetic and epigenetic alterations (3, 46). To investigate whether KMT9 loss also affects sporadic colorectal carcinogenesis driven by key human colorectal cancer mutations, we generated epithelial tumor organoids from Apc^fl/fl p53^fl/fl Kras^G12D/+ Kmt9α^fl/fl and Apc^fl/fl p53^fl/fl Kras^G12D/+ Kmt9α^+/+ mice. Transduction of organoids with cre-expressing adenovirus resulted in APKK (Apc^KO/p53^KO/Kras^G12D/Kmt9α^KO) and APK (Apc^KO/p53^KO/Kras^G12D/Kmt9α^WT) colorectal cancer organoids. In line with our results presented above, we observed a severe impairment in growth for KMT9α-deficient APKK organoids in comparison with KMT9α-proficient APK colorectal cancer organoids (Fig. 4A), suggesting a critical role for KMT9 in sporadic tumor progression. To unravel the mechanistic role of KMT9α in sporadic colorectal cancer, we performed RNA-seq in APK and APKK organoids. In line with our results obtained with AOM/DSS tumor organoids, we observed a significant downregulation of genes involved in cell-cycle progression in APKK organoids (Fig. 4B). Furthermore, we identified 124 stem cell–related genes significantly downregulated by KMT9α depletion (Supplementary Fig. S4A). To corroborate our data in human colorectal tumors, we established patient-derived organoid (PDO) cultures from four patients with colorectal cancer (PDO #1, #2, #3, and #4) representing CMS2, CMS3, and CMS4 tumors (Fig. 4C). We did not consider CMS1 tumors due to the low expression levels of KMT9α in CMS1 compared with CMS2–4 tumor samples (Supplementary Fig. S1A). Human PDOs were transduced with lentivirus driving expression of either control miRNA (miCtrl) or miRNA directed against KMT9α (miKMT9α) and organoid growth was analyzed in vitro. Importantly, KMT9α knockdown efficiently reduced tumor organoid size in all investigated PDOs (Fig. 4D; Supplementary Fig. S4B and S4C). Consistent with this morphologic phenotype, transcriptomic analysis of the PDOs revealed a significant downregulation of genes involved in cell-cycle progression (Fig. 4E), as well as downregulation of multiple genes involved in CSC function of colorectal cancer (Supplementary Fig. S4D), thus corroborating the results obtained in murine inflammation-induced and sporadic colorectal tumorigenesis. We next investigated the function of KMT9α in the epithelial compartment of primary tumor tissue from patients with colorectal cancer. We analyzed single-cell transcriptome data from 23 Korean patients with colorectal cancer (GSE132435) and stratified epithelial tumor cells in KMT9α-proficient and KMT9α-deficient cell populations (Supplementary Fig. S4E; ref. 47). Interestingly, most of the genes involved in cell-cycle progression and genes downregulated upon depletion of KMT9α in murine AOM/DSS and in sporadic murine colorectal cancer organoids, were also downregulated in human epithelial KMT9α-deficient colorectal cancer cells (Fig. 4F). Furthermore, we analyzed the LGR5 ISC gene signature (40) and found numerous signature genes downregulated in epithelial KMT9α-deficient human colorectal cancer cells (Fig. 4G).

Together, these findings demonstrate a decisive role of KMT9α in mouse and human sporadic colorectal carcinogenesis and as well as in inflammation-induced tumorigenesis, thereby identifying inhibition of KMT9 as a highly promising single-target approach for the treatment of colorectal cancer.

In this study, we used mouse models and tumor organoids derived from mice and human patients as well as human tumor tissue to uncover an essential role of KMT9, a novel H4K12me1 histone methyltransferase, in colorectal cancer. Our data show that growth of inflammation-induced tumors as well as sporadic colorectal cancer is dramatically reduced in the absence of KMT9α in vitro and in vivo. While we cannot exclude that additional KMT9 functions could potentially contribute to the observed phenotype, we demonstrate that a major consequence of KMT9 loss is the impairment of tumor growth via disrupting the regulation of genes involved in proliferation, cell-cycle progression, and apoptosis. So far, we cannot fully rule out the possibility that the loss of KMT9 is playing a role in the inflammatory or regenerative process after AOM/DSS treatment, which warrants further investigation. Moreover, we have also demonstrated that KMT9α ablation impairs tumor growth in CMS2-, CMS3-, and CMS4 PDOs. To date, it is well accepted that the interpatient heterogeneity of colorectal cancer has clinical implications on therapeutic responses (48). Thus, future investigations have to examine different colorectal cancer subtypes for their responsiveness to targeting KMT9 in vivo.

Besides reduction of tumor growth, elimination of colorectal CSCs to prevent tumor recurrence and metastasis remains one of the biggest clinical challenges in treating colorectal cancer. To date, most therapeutic strategies rely on combination therapy of a CSC-targeting agent with eradication of tumor mass (49), where targeting of CSCs is hindered due to its similarity with benign, noncancerous stem cells. In this study, we showed that in addition to a major impact on inhibiting tumor growth, targeting KMT9 also downregulated numerous stem cell–related genes and impaired CSC function and maintenance. Thus, inhibition of KMT9 in colorectal cancer may be a promising therapy to target colorectal CSCs.

Currently, the use of epigenetic drugs as monotherapy for colorectal cancer have largely failed to improve patient outcomes (50, 51). The main issues leading to the failure of these epigenetic agents include intratumor and interpatient heterogeneity and the essential function of epigenetic regulators in both malignant and benign cells. Importantly, we found that without AOM/DSS insult, mice with specific KMT9α ablation in intestinal epithelial cells survive without a noticeable phenotype and do not exhibit morphologic changes of the colon tissue. Thus, weak side effects of KMT9 targeting would be assumed in the clinical context. In consequence, the development of small-molecule inhibitors targeting KMT9 might be a promising therapeutic avenue for colorectal cancer treatment.

C. Berlin reports grants from Deutsche Forschungsgemeinschaft during the conduct of the study and outside the submitted work. F.R. Greten reports grants from DFG and HMWK during the conduct of the study. E. Metzger reports grants from Deutsche Forschungsgemeinschaft during the conduct of the study and outside the submitted work. R. Schüle reports grants from Deutsche Forschungsgemeinschaft (DFG) during the conduct of the study and outside the submitted work. No disclosures were reported by the other authors.

C. Berlin: Conceptualization, data curation, formal analysis, validation, investigation, methodology, writing–original draft. F. Cottard: Conceptualization, data curation, formal analysis, validation, investigation, methodology, writing–original draft. D. Willmann: Data curation, software, formal analysis. S. Urban: Investigation, methodology. S.M. Tirier: Data curation, software, formal analysis. L. Marx: Investigation. K. Rippe: Resources, methodology. M. Schmitt: Investigation, methodology. V. Petrocelli: Investigation, methodology. F.R. Greten: Conceptualization. S. Fichtner-Feigl: Conceptualization. R. Kesselring: Conceptualization, resources, investigation, methodology. E. Metzger: Conceptualization, writing–original draft. R. Schüle: Conceptualization, supervision, funding acquisition, writing–original draft.

This work was supported by grants of the Deutsche Forschungsgemeinschaft - Project ID 413517907 to C. Berlin, Project ID KE2164/2-1 - FOR2438 to R. Kesselring and Project ID 192904750 - SFB 992 Medical Epigenetics, Project-ID 403222702 - SFB 1381, Project-ID 89986987 - SFB 850, Schu688/15-1, and of the European Research Council (ERC AdGrant 322844) to R. Schüle.

The authors thank L. Walz, F. Pfefferle, M. Sum, and S. Schumacher for excellent technical assistance as well as B. Mauerer, H. Braumüller, J. Leung, and H. Greschik for creative input and insightful comments throughout the project. B. Ritter was of great help concerning human tumor organoid experiments. The authors would like to acknowledge the Lighthouse Core Facility for their assistance with FACS analysis.

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.