KMT2D/ZNF460-induced COL9A1-mediated extracellular matrix stiffness maintains the cancer stem cell pool to promote colorectal cancer progression

Abstract

Collagen is a central component of the extracellular matrix (ECM) in tissues, and ECM can promote tumor cell immune evasion. Our research aimed to expound the biological function of the collagen alpha-1(IX) chain (COL9A1) in colorectal cancer (CRC) and the upstream mechanism regarding KMT2D/ZNF460. COL9A1 contributed to maintaining colorectal cancer stem cells (CCSC) self-renewal and proliferative capacity, and COL9A1 knockdown attenuated CCSC stemness, which was activated by 20.0 kPa polyacrylamide gels. Silencing of COL9A1 hampered tumor growth and stemness in mice induced by AOM/DSS and improved the tumor microenvironment (TME) in xenograft-bearing mice. Mechanistically, KMT2D promoted COL9A1 expression by mediating H3K4me1 modification of the enhancer and recruiting ZNF460. In the presence of attenuated KMT2D signaling, its effect on CCSC stemness and CRC progression was similar to that of knockdown of COL9A1, both of which have therapeutic benefits for CRC tumors. Again, the reactivation of COL9A1 reversed this trend. In conclusion, KMT2D mediates H3K4me1 modification of enhancers and recruits ZNF460 to activate COL9A1, which enhances ECM stiffness and self-renewal of CCSC to remodel TME, contributing to CRC progression.

Supplementary Information

The online version contains supplementary material available at 10.1007/s10565-025-10053-3.

Introduction

As a relatively rare group of cancer cells in the tumor microenvironment (TME), cancer stem cells (CSCs) hold properties of immune privilege to escape from the attack of the immune system, regulated by the microenvironment of CSCs (An and Hu 2023). Colorectal cancer (CRC) ranks the third most prevalent form of tumor, with a significant body of research linking relapses to the failure of CSC subpopulations to be eradicated which have been shown to express various markers, including CD133 and leucine-rich-repeat-containing G-protein coupled receptor 5 (LGR5) (Morgan et al. 2023; Morgan and Mortensson 2018). Extracellular matrix (ECM) stiffness can modulate CSCs through different molecular signaling pathways (Safaei et al. 2023). Lysyl oxidase (LOX), a secreted, copper-dependent amine oxidase and collagen cross-linker, was enhanced in CRC tissues and correlated with the tumor stage, lymph node metastasis, and tumor grade in CRC patients (Wei et al. 2017). Therefore, a deeper understanding of the effect of matrix stiffness on CSC characteristics might be beneficial to the development of innovative cancer therapies in CRC.

Collagens are the major components of ECM, and various collagen proteins are found to be upregulated in tumors, which predicts an unsatisfactory prognosis in cancer patients (Jiang et al. 2022). For instance, the link between ECM-Collagen I and OSBPL2 has been recently implicated in CRC to accelerate the discovery of new treatment options for CRC (Lin et al. 2024). Combining several bioinformatics algorithms, we identified here that collagen alpha-1(IX) chain (COL9A1) is a novel ECM-related gene that is also enhanced in CRC patients with high LGR5 expression. It is interesting to note that collagen has a regulatory effect on the function and phenotype of tumor-infiltrating immune cells, including T cells, which suggests that tumor-associated collagen may possess immune-modulatory properties within the TME, consequently impacting cancer progression (Romer and Thorseth 2021). Two distinct types of lymphocytes, CD8⁺ T cells and natural killer (NK) cells, serve as a prognostic indicator in CRC (Fionda et al. 2022). NK cells are versatile players in immunity because of their powerful and immediate response against transformed cells and the ability to manipulate the subsequent adaptive immune response (Della Chiesa et al. 2022). The high density of mature dendritic cells (DCs) is considered a potential target in treating many malignancies, and DCs, accountable for the activation of T cells, which enhances the secretion of diverse cytokines, are rapidly evolving immunotherapeutic targets in eradicating residual tumor cells (Mosinska et al. 2017). Even though the correlation between COL9A1 expression and the infiltration of cytotoxic lymphocytes and DCs has been predicted in CRC, the role of COL9A1 in immune escape mediated by these cells is mostly undecided. We designed this study to examine the association of COL9A1 with colorectal CSC (CCSC) phenotypes and immune escape and to expound the mechanisms through which COL9A1 expression is regulated in CRC.

Materials and methods

Sample collection

Twenty CRC samples and matched 20 normal paracancerous tissues were collected from Harbin Medical University Affiliated Cancer Hospital from June 2019 to April 2022, and the diagnosis of CRC was verified by histological examination by pathologists. The study protocol was approved by the Institutional Review Board of Harbin Medical University Affiliated Cancer Hospital (approval No: KY2019-22) following the Declaration of Helsinki. Written informed consent was obtained from each patient.

AOM/DSS model

All animal experiments were approved by the Animal Ethical and Welfare Committee of Harbin Medical University Affiliated Cancer Hospital (approval No: KY2022-11). The animals were treated following the Guide for the Care and Use of Laboratory Animals, as established by the National Institutes of Health. All mice were kept in a suitable pathogen-free environment.

C57BL/6 mice (8 weeks, Vital River, Beijing, China) were administered intraperitoneally with a single dose of AOM (A5486, Merck KGaA, Darmstadt, Germany) at 10 mg/kg. After 7 days, the mice were given drinking water containing 2% DSS (42,867, Merck KGaA) for one week and then fed regular drinking water for two weeks (Ma et al. 2018). DSS treatments were repeated for two additional cycles. Control mice were not treated with AOM/DSS.

For lentiviral gene intervention in AOM/DSS-induced CRC mice, the mice were starved overnight before the start of each DSS cycle and injected intracolonically with short hairpin RNA (sh) COL9A1, sh KMT2D, and overexpression (oe) COL9A1-concentrated lentiviruses (designed and purchased by VectorBuilder, Guangzhou, Guangdong, China). On day 120, mice were euthanized with 150 mg/kg sodium pentobarbital (i.p.), and tumor load was assessed. Tumors larger than 1 mm were counted, measured, and weighed. Colon tissue was collected for subsequent studies.

Immunohistochemical staining (IHC)

Sections of human CRC tissues and mouse colon tissues were first deparaffinized, rehydrated, and incubated with 3% peroxide-methanol for 30 min. Nonspecific antigen binding was sealed with 10% BSA. The sections were probed with primary antibodies to KMT2D (1:200, PA5-57490, Invitrogen Inc., Carlsbad, CA, USA), ZNF460 (1:250, 25,299–1-AP, ProteinTech Group, Chicago, IL, USA), COL9A1 (1:500, 61–492, ProSci, Poway, CA, USA), Collagen I (1:500, PB9939, Wuhan Boster Biological Technology Co., Ltd., Wuhan, Hubei, China), LOX (1:100, ABT112, Merck KGaA), CD133 (1:250, NB120-16518, Novus Biological Inc., Littleton, CO, USA), LGR5 (1:200, NBP1-28,904, Novus Biological) overnight at 4 °C, After the incubation with secondary antibody (1:2000, ab97051, Abcam, Cambridge, MA, USA) at 37 °C for 50 min, the sections visualized with diaminobenzidine (D12384, Merck KGaA). Then, sections were counterstained with hematoxylin and visualized by microscopy. The stained tissue sections were evaluated by two pathologists. Staining results were assessed according to the rate of positive cell staining.

Separation of CCSC

The sgRNA targeting the last exon of LGR5 (exon 18) was constructed as LGR5 P2 A-GFP using a plasmid vector expressing GFP. The donor vector carrying the LGR5-GFP targeting construct was then electroporated into a human CRC cell line, HCT 116 (CL-0096, Procell, Wuhan, Hubei, China). Successfully transfected cells were seeded into ultra-low attachment 6-well plates and cultured in DMEM/F12 (51445 C, Merck KGaA) plus N2, B27, 20 ng/mL epidermal growth factor (EGF, PHG0313, Thermo Fisher Scientific Inc., Waltham, MA, USA) and 20 ng/mL bFGF (100-18B, Thermo Fisher). After one week, GFP⁺ clones were observed by fluorescence microscopy and individually hand-selected into new wells for propagation. After successive passages of sphere formation, LGR5⁺ CCSCs were screened from spheres by fluorescence‑activated cell sorting (FACS) for further experiments (Zhu et al. 2022).

Cell lines

Mouse colonic mucosal epithelial cells (iCell-0108a, Cellverse Bioscience Technology Co., Ltd., Shanghai, China) were cultured in mouse colonic mucosal epithelial cell-specialized medium (iCell-0108a-001b, Cellverse). Fetal human colon (FHC) cells (KL-C2079H, Kang Lang, Shanghai, China) and mouse colon tumor 26 (CT26, CL-0071, Procell) were maintained in RPMI-1640 (PM150110, Procell) medium plus 10% FBS + 1% P/S and placed in a culture environment of 37 °C, 5% CO₂.

All lentiviral vectors were packaged using the procedures specified by VectorBuilder. LGR5⁺ CCSCs were infected with sh COL9A1, sh KMT2D, oe ZNF460, oe COL9A1 lentiviruses, and their control lentiviruses in the presence of polybrene infection/transfection reagent (TR-1003, Merck KGaA). The cells were screened with 1 μg/mL puromycin for 14 d to harvest stably infected cells. Target gene expression in stable clones was verified by RT-qPCR.

CT26 cells were infected with concentrated lentiviruses containing sh COL9A1, sh KMT2D, oe COL9A1, and their control lentiviruses using polybrene infection/transfection reagent (TR-1003, Merck KGaA) for 24 h. The screened clones (as mentioned above) were collected for further culture (Zhang et al. 2020). After verification of target gene expression in infected cells by RT-qPCR, these cells were used in tumor transplantation experiments.

RNA isolation and qPCR

Total RNA was isolated from LGR5⁻ or LGR5⁺ HCT 116 cells as well as CT26 using TRIzol (15596026 CN, Invitrogen). The cDNA was then synthesized by iScript Select cDNA Synthesis Kit (1,708,896, Bio-Rad Laboratories, Hercules, CA, USA). RT-qPCR was performed using Power SYBR Green PCR premix (4,367,659, Thermo Fisher). The 2^−ΔΔCt method was used to calculate relative gene expression by normalizing to GAPDH. Table 1 details the primer sequences used.

Table 1.

PCR primers used in this study

Gene name	Forward sequence (5’−3’)	Reverse sequence (5’−3’)
KMT2D (Human)	GGAATGGGTAGCTCTTTGGCGA	TGCCGAATCAGCAGCTCTCGTA
KMT2D (Mouse)	ACACACTATAAACGGCCCCA	TACGCGAACGAGCCAGATAC
ZNF460 (Human)	GACATTTACCCAGGAGGAGTGG	GCCAGATGAGAAGGGATAGGCT
COL9A1 (Human)	TGGAGTGGAAGGACCAAGAGGA	GTGCTGATCTGTCGGTGCTCTA
COL9A1 (Mouse)	AGAAATCGGCGAGCCAGGAAGA	CCAACTATGCCAGTGATGCCTC
GAPDH (Human)	GTCTCCTCTGACTTCAACAGCG	ACCACCCTGTTGCTGTAGCCAA
GAPDH (Mouse)	CATCACTGCCACCCAGAAGACTG	ATGCCAGTGAGCTTCCCGTTCAG

Note: COL9A1, collagen alpha-1(IX) chain; KMT2D, histone-lysine N-methyltransferase 2D; ZNF460, zinc finger protein 460; GAPDH, glyceraldehyde-3-phosphate dehydrogenase

Western blot

LGR5⁻ or LGR5⁺ HCT 116 cells, mouse colonic mucosal epithelial cells, FHC, or CT26 cells were lysed using RIPA Buffer (R0278, Merck KGaA) plus protease and phosphatase inhibitor mixture (P1045, Beyotime, Shanghai, China). The lysate supernatant was concentrated 20-fold in an ultrafiltration tube and diluted 1:1 with RIPA buffer. An equal amount of total protein lysate was subjected to 10% SDS-PAGE and transferred to a PVDF membrane. After sealing with 5% goat serum in TBS/Tween-20 for 1 h, the membranes were probed overnight at 4 °C with primary antibodies. The antibodies are as follows: KMT2D (1:500, ABE1867, Merck KGaA), ZNF460 (1:2000, 26–545, ProSci), COL9A1 (1:1000, 61–492, ProSci), Collagen I (1:2000, ab34710, Abcam), LOX (1:1000, ABT112, Merck KGaA), CD133 (1:1000, NB120-16518, Novus Biological), LGR5 (1:100, NBP1-28,904, Novus Biologicals), β-actin (1:2000, PA1-16,889, Invitrogen). Visualization of the proteins was achieved using SuperSignal West Pico PLUS Chemiluminescent Substrate (34,580, Thermo Fisher) following incubation with secondary antibody (1:10,000, ab97051, Abcam).

Sphere formation

LGR5⁺ CCSCs (5,000) were plated into ultra-low attachment 6-well plates and cultured in DMEM/Ham Nutrient Mix F12 (51445 C, Merck KGaA) plus N2 (17,502,001, Gibco, Carlsbad, CA, USA), B27 (A3582801, Gibco), 20 ng/mL EGF, and 20 ng/mL bFGF. Spheres were observed and counted after one week, sphere formation rate = (number of spheres)/5000 × 100% (Zhu et al. 2022).

Polyacrylamide gels

A large coverslip was immersed in 0.1 M NaOH and air-dried. After that, 3‐aminopropyltrimethoxysilane (281,778, Merck KGaA) was spread across each coverslip and soaked in 0.5% glutaraldehyde (G5882, Merck KGaA) in PBS for 30 min. Coverslips were covered with nitrocellulose to increase polyacrylamide binding. Then, 25 μL of acrylamide/bisacrylamide mixture (containing 3% or 10% acrylamide and bisacrylamide at a concentration of 0.01% or 0.3%) was placed on the coverslips to prepare gels with a stiffness of 2.0 and 20.0 kPa, respectively, and covered with a small circular coverslip. The rest steps were conducted as previously reported (Pelham 1997).

Colony formation assays

Cells were plated at 500 cells/well into 6-well plates pre-laid with 2.0 kPa or 20.0 kPa polyacrylamide gels and incubated for 2 weeks. The medium was renewed every 3 days during colony growth. Colonies with more than 50 cells were counted under a light microscope.

Transwell assay

Cell migration and invasion assays were performed using Corning Transwell 24-well plates (CLS3422, Merck KGaA). LGR5⁺ CSCs were first pre-cultured on polyacrylamide gels of 2.0 kPa or 20.0 kPa for 48 h to form mechanical memory in response to physical and mechanical stimuli. Then, 200 μL of LGR5⁺ CSCs in serum-free medium was seeded in the upper chamber, and 500 μL of medium containing 10% FBS was loaded into the lower chamber. After 12 h, non-migrating/non-invasive cells attached to the surface of the upper chamber were removed. The cells were fixed, treated with 0.5% crystal violet for 4 h at 37 °C, and imaged. Transwell filter membranes were pre-coated with Matrix-Gel (C0376, Beyotime) for the invasion assay.

Immunofluorescence staining

LGR5⁺ CCSCs (1 × 10⁵), which inherently contained GFP green fluorescence, were fixed in 4% paraformaldehyde for 20 min and permeabilized with 0.5% Triton X-100 for 5 min. The cells were blocked in 10% goat serum for 1 h at room temperature and probed with the anti-Ki67 (1:100, PB9026, Boster) at 4 °C overnight and with IgG (1:500, ab150083, Abcam) at room temperature for 1.5 h, followed by counter-staining with DAPI. Fluorescence signals were imaged.

HCT 116 cells were cultured at a density of 5 × 10⁵ cells per well on sterile glass coverslips in 24-well plates, fixed using 4% paraformaldehyde, permeabilized with 0.5% Triton X-100, and sealed using goat serum. The cells were stained overnight with anti-KMT2D (1:50, 701,869, Invitrogen) and anti-H3 K4 me1 (1:100, 49–1003, Invitrogen) and with goat anti-rabbit IgG (A-11008, Invitrogen) and DAPI. The mean fluorescence intensity was quantified using Image J.

Hematoxylin–eosin (HE) staining

HE staining kits (C0105S, Beyotime) were used to assess inflammatory infiltration of mouse colon tissues. The paraffin-embedded sections of mouse colon tissue were dewaxed and hydrated, then stained with hematoxylin for 10 min, treated with differentiation solution for 30 s, and then stained with eosin for 1 min, washed with 70% ethanol, and then subjected to dehydration, clearing, sealing, and observation. The scoring system was based on a previous study (Johansson et al. 2014).

Masson's staining

The extent of collagen deposition in mouse colon tissues was assessed using a Masson’s trichrome staining kit (HT15, Merck KGaA). After dewaxing and hydration, the paraffin-embedded sections were treated with hematoxylin, Biebrich scarlet-acid fuchsin, phosphotungstic phosphomolybdic acid, and aniline blue (all for 5 min). Finally, the slides were incubated in 1% acetic acid for 2 min, dehydrated, cleared, and then visualized on the microscope. For quantification of collagen fiber deposition, images were analyzed under the microscope using Image-Pro Plus 6.0 image analysis software (Media Cybernetics, Bethesda, MD, USA).

Subcutaneous xenografts

CT26 cells (5 × 10⁵) infected with sh Ctrl, sh COL9A1, sh KMT2D, sh KMT2D + oe Ctrl, or sh KMT2D + oe COL9A1 were suspended in 100 μL PBS and administered subcutaneously into the dorsal right ventral side of BALB/c mice (211, Vital River) (n = 6). After measuring subcutaneous tumor size in mice for the first time on the fifth day, the monitoring was performed every 4 days thereafter for 25 days. Tumor volume (V) was assessed with this formula: V(mm³) = 0.5 × length × width². On day 25, the mice were euthanized by 150 mg/kg sodium pentobarbital (i.p.).

Isolation of tumor-infiltrating lymphocytes (TIL)

To isolate TIL, BALB/c mouse subcutaneous tumors derived from mouse CT26 cells were minced using scissors and a scalpel, and detached in Hanks'buffered salt solution (HBSS, 88,284, Thermo Scientific) containing 1 mg/ml collagenase (C5138, Merck KGaA), 0.1 mg/mL DNAse I (DN25, Merck KGaA), and 2.5 U/mL hyaluronidase (H1115000, Merck KGaA) for 2 h to prepare tumor cell suspensions. The resulting suspension was passed through a 70-μm cell filter and resuspended in PBS + 3% BSA to 1 × 10⁶ cells/mL. Mouse TILs were purified by Mouse CD90.2 MicroBeads (130–121–278, Miltenyi Biotec, Bergisch Gladbach, Germany) and incubated under 3000 IU/mL of IL-2 (HY-P7077, MedChemExpress, Monmouth Junction, NJ, USA) for 5 days, after which the TILs were harvested for the following experiments.

Flow cytometry

Excised BALB/c mouse subcutaneous tumors derived from CT26 cells were sectioned into small pieces and dissociated using the Tumor Dissociation Kit (130–096–730, Miltenyi Biotec). For the isolation of mouse CD8⁺ T cells from tumors, Dynabeads FlowComp Mouse CD8 Kit (11462D, Invitrogen) was used. Antibodies against APC-CD3 (17–0031–82, Invitrogen), FITC-CD8 (11–0081–82, Invitrogen), FITC-CD19 (11–0193-82, Invitrogen), PE-Ki67 (12–5698-82, Invitrogen), APC-granzyme b (Gzmb, 17–8898-82, Invitrogen), PE-Perforin (Prf1, 12–9392-82, Invitrogen), APC-interferon-gamma (IFN-γ, 17–7311-82, Invitrogen) were used.

NK cells were labeled using CD3 and CD49b-FITC (ab272259, Abcam). CD11c⁺ cells were isolated from mouse tumors using CD11c MicroBeads UltraPure (130–125–835, Miltenyi) and stained with MHCII-FITC (11–5321-82, Invitrogen), CD103-PE-Cy7 (121,425, BioLegend), XCR1-PE (148,203, BioLegend), and IRF8-APC (17–9852-82, Invitrogen).

Co-culture and functional analysis of CT26 and CD8⁺T/NK cells

Spleen CD8⁺ T lymphocytes from BALB/c mice were isolated using the Dynabeads FlowComp Mouse CD8 Kit. CD8⁺ T cells were pre-stimulated with 200 U/mL IL-2, Dynabeads Mouse T Activator CD3/CD28 (11452D, Gibco) for 48 h before the experiment. Afterwards, CD8⁺ T cells were co-cultured with CT26 cells in a 5:1 ratio for 48 h.

Primary NK cells were isolated from mouse spleens using the EasySep Mouse NK Cell Isolation Kit (19,855, NovoBiotechnology Co., Ltd., Beijing, China) and co-cultured with CT26 cells at a 5:1 ratio for 24 h. Cells were subsequently collected for analysis of NK cell functionality (IFN-γ⁺ NK cells) and toxicity to tumor cells.

Cell viability of CT26 cells was analyzed at 0, 24, 48, and 72 h using the CCK-8 (C0005, TargetMol, Boston, MA, USA).

ChIP assays

ChIP assays were conducted using a ChIP kit (17–295, Merck KGaA). HCT 116 cells were cross-linked with 1% formaldehyde, lysed, sonicated to generate DNA fragments (an average length of 500 bp), and subjected to IP using specific antibodies. ZNF460 antibody, KMT2D antibody (1:10, ABE1867, Merck KGaA), H3 K4 me1 antibody (1:10, NB21-1021, Novus Biological), and isotype control (ab171870, Abcam) were used in this assay. The magnetic beads were washed after precipitation, and the DNA was eluted, purified, and subjected to qPCR using promoter- and enhancer-specific primers for COL9A1.

Luciferase reporter assays

Luciferase reporter gene vectors were purchased from Promega (Madison, WI, USA). The pGL3-basic (firefly luciferase, Fluc) pRL-TK (Renilla luciferase, Rluc) dual-luciferase reporter gene kit (YanjiangBio, Shanghai, China) was used, and the COL9A1 promoter sequence was subcloned into the dual luciferase reporter gene vector. Next, 500 ng of reporter plasmid was co-transfected into HCT 116 cells for 24 h. The relative Fluc/Rluc activity was calculated.

Plasmid construction and cell transfection

Two single-stranded sgRNA primers with complementary sequences were annealed for 20 min at 90 °C in an annealing buffer consisting of 50 mM Tris–Cl (pH = 7.5–8.0), 250 mM NaCl, and 5 mM EDTA and cloned into the epiCRISPR plasmid (135,960, Addgene, Watertown, MA, USA). HCT 116 cells were cultured in 6-well plates at 5 × 10⁵ cells/well, and sgRNA plasmids were transfected using Lipofectamine 3000 (L3000001, Invitrogen) to construct KMT2D-LCD1^delete (sgRNA A-B) versus KMT2D-LCD2^delete (sgRNA C-D) cells, and cells transfected with sgVector plasmid were used as controls. The sgRNA primers for KMT2D are listed in Table 2.

Table 2.

sgRNA primer

Primer name	Forward sequence (5’−3’)	Reverse sequence (5’−3’)
hKMT2D-A	CCGGGGACAATAGGGCAGAATCA	AACTGATTCTGCCCTATTGTCCC
hKMT2D-B	CCGCAGGACTGTACCTCTGACAG	AACCTGTCAGAGGTACAGTCCTG
hKMT2D-C	CCGGGGTGTAGCAGTCCTCTAGT	AACACTAGAGGACTGCTACACCC
hKMT2D-D	CCGGAACTTGTGTCTTATGCCAC	AACGTGGCATAAGACACAAGTTC

Statistical analysis

For statistical analyses in the study, GraphPad Prism 8.0.2 (GraphPad, San Diego, CA, USA) was utilized. The data is shown as mean ± SEM. For comparisons between any two groups, paired or unpaired t-tests were utilized. For multiple group comparisons, ANOVA (one-way or two-way when appropriate) was performed, with Tukey’s post hoc test. p < 0.05 indicates a statistically significant difference between the data.

Results

COL9A1 is involved in LGR5⁺CSC maintenance by affecting ECM stiffness

To analyze the transcriptome specificity of CCSC, we downloaded the gene expression data of 512 patients from the TCGA Colon Cancer (COAD) cohort from the GDC hub (https://gdc.xenahubs.net) [UNIT = log2(count + 1)] and categorized the patients into two groups, LGR5_high and LGR5_low according to their LGR5 (Ensembl_ID = ENSG00000139292.11) expression mean (10.40). Significantly differentially expressed genes (DEGs) were filtered using the R package limma (version 3.40.6) with the setting of |log₂FoldChange|> 1 and p < 0.01 (Fig S1A). The heatmap in Fig S1B shows the top 20 genes with the most significantly elevated or decreased expression in the LGR5_high sample (Fig S1B), where LGR5 (ENSG00000139292.11) is included among the genes with elevated expression, suggesting that the grouping was successful.

DEGs were then subjected to GO enrichment analysis. As shown in Fig S1C, DEGs were mainly localized to the cell projection membrane and collagen-containing ECM, involved in ECM organization and extracellular structure organization, and exerted molecular functions like ECM structural constituent, ECM structural constituent conferring, and tensile strength. DEG enrichment in ECM-receptor interaction was also observed in KEGG pathway enrichment results. Therefore, we selected genes enriched in the ECM-related GO or KEGG pathways for intersection, and there were four intersecting genes: COL9A3, FN1, COMP, and COL9A1 (Fig S1D).

The promotional effects of FN1 (Cai et al. 2018), COMP (Zhong et al. 2020), and COL9A3 (Mao et al. 2021) on CRC progression have been reported, whereas COL9A1 has not yet been studied in CRC. COL9A1 (ENSG00000112280.14) expression was significantly elevated in CRC samples with high LGR5 expression (log₂FoldChange = 1.210574855) (Fig S1E). Thereafter, analysis of the PROGgeneV2—Pan-Cancer Prognostics Database (http://www.progtools.net/gene/) database revealed that CRC patients with high COL9A1 expression tended to have worse OS (Fig S1F) and relapse-free survival (RFS) (Fig S1G).

Inhibition of COL9A1 attenuates the stemness and proliferative capacity of CCSCs

We then analyzed the expression of COL9A1 in human CRC and paracancerous tissues by IHC, and COL9A1 expression was higher in CRC tissues (Fig. 1A). Thereafter, we screened LGR5⁺ CCSCs from cell spheroids by FACS (Fig. 1B). COL9A1 expression in LGR5⁺ CCSCs and LGR5⁻ cells was examined. The levels of COL9A1 in LGR5⁺ CCSCs were significantly higher than those in LGR5⁻ cells (Fig. 1C, D).

Fig. 1.

Knockdown of COL9A1 significantly inhibits the stemness and proliferation of CCSCs. (A) IHC analysis of COL9A1 expression in human CRC tissues and matched paracancerous tissues (n = 20). (B) Screening of LGR5⁺ CCSCs from cell spheroids by FACS. (C) RT-qPCR detection of COL9A1 mRNA levels in LGR5⁺ CCSCs and LGR5⁻ cells. (D) Western blot analysis of COL9A1 protein expression in LGR5⁺ CCSCs and LGR5⁻ cells. LGR5⁺ CSCs were infected using three shRNAs targeting COL9A1. (E) Knockdown efficiency of sh COL9A1 by RT-qPCR. (F) The sphere-forming ability of LGR5⁺ CCSCs after the knockdown of COL9A1 was examined using sphere sphere-formation assay. (G) The protein expression of CD133 and LGR5 in LGR5⁺ CCSCs after the knockdown of COL9A1 was assessed using Western blot analysis. (H) The proportion of Ki67-positive LGR5⁺ CCSCs after the knockdown of COL9A1 was examined using immunofluorescence staining. The results are shown for three biological studies. The bars represent the mean ± SEM; ** p <.01, *** p <.001, and **** p <.0001, as determined by a t-test (A, C, D, F, G, H) or one-way ANOVA with Tukey’s post hoc test (E)

To elucidate the association of COL9A1 with stemness, we inhibited COL9A1 by sh COL9A1 lentiviruses in LGR5⁺ CCSCs and selected sh COL9A1 #3, which had the strongest knockdown effect, for the subsequent operation (Fig. 1E). It was found by sphere formation assay that LGR5⁺ CSC with COL9A1 knockdown had weaker tumor initiation ability than sh Ctrl-infected cells (Fig. 1F). Western blot assay showed that inhibition of COL9A1 also downregulated the levels of stemness markers (CD133 and LGR5) in LGR5⁺ CCSCs (Fig. 1G). In addition, we also detected a significant reduction in the percentage of Ki67⁺ CCSC after COL9A1 knockdown by immunofluorescence staining, demonstrating that COL9A1 knockdown reduced the proliferative capacity of CCSCs (Fig. 1H).

COL9A1-mediated ECM stiffness significantly enhances CCSC stemness and invasiveness

We analyzed the expression of ECM deposition protein Collagen I and ECM modulating enzyme LOX in LGR5⁺ CCSCs and LGR5⁻ cells by western blot analysis. Their expression was upregulated in LGR5⁺ CCSCs (Fig. 2A). To demonstrate that the stemness of CCSCs was related to COL9A1-mediated ECM stiffness in the TME, we cultured LGR5⁺ CSCs with sh COL9A1 in polyacrylamide gels at 2.0 kPa or 20.0 kPa. Upon colony formation assay, the number of colonies formed in the matrix gel with a stiffness of 20.0 kPa was higher than that in the matrix gel with a stiffness of 2.0 kPa, whereas silencing of COL9A1 expression reduced the number of colonies formed under both stiffnesses (Fig. 2B). Furthermore, we found that sh COL9A1 resulted in smaller and fewer spheres formed by cells at both 2.0 kPa and 20.0 kPa matrix gel, and that incubation in 20.0 kPa stiffness elicited greater sphere-forming ability (Fig. 2C). Finally, we found by western blot experiments that protein expression of CD133 and LGR5 was significantly elevated in cells cultured at 20.0 kPa stiffness compared to cells cultured on 2.0 kPa medium. However, the silencing of COL9A1 downregulated CD133 and LGR5 (Fig. 2D). We also analyzed the migration and invasion abilities of LGR5⁺ CSCs cultured on 2.0 kPa and 20.0 kPa matrix gel after COL9A1 knockdown. Transwell assay showed that cells cultured on 20.0 kPa medium exhibited significantly enhanced migration and invasion compared with those on 2.0 kPa, whereas knockdown of COL9A1 reduced the migration and invasion abilities under both conditions (Fig. 2E).

Fig. 2.

CCSC stemness promoted by ECM stiffness and invasiveness in CRC is reversed by sh COL9A1. (A) Collagen I and LOX protein levels in LGR5⁺ CCSCs and LGR5⁻ cells were assessed using Western blot analysis. (B) Colonies formed by LGR5⁺ CCSCs after the knockdown of COL9A1 in gels with 2.0 kPa or 20.0 kPa stiffness. (C) Spheres formed by LGR5⁺ CCSCs after the knockdown of COL9A1 in gels with 2.0 kPa or 20.0 kPa stiffness. (D) The protein expression of CD133 and LGR5 in LGR5⁺ CCSCs after the knockdown of COL9A1 in gels with 2.0 kPa or 20.0 kPa stiffness was assessed using Western blot analysis. (E) The migration and invasion ability of LGR5.⁺ CCSCs after the knockdown of COL9A1 in gels with 2.0 kPa or 20.0 kPa stiffness was assessed using the Transwell assay. The results are shown for three biological studies. The bars represent the mean ± SEM; *p <.05, ** p <.01, *** p <.001, and **** p <.0001, as determined by a t-test (A) or two-way ANOVA with Tukey’s post hoc test (B, C, D, E)

Inhibition of COL9A1 hinders ECM deposition-mediated CSC stemness and tumor growth in vivo

The protein expression of COL9A1 was elevated in CT26 cells relative to mouse colonic mucosal epithelial cells (normal) by western blot analysis (Fig S2A). Therefore, we used three shRNAs targeting COL9A1 for CT26 cells and verified successful lentiviral infections by RT-qPCR. mCol9a1 [shRNA#1] lentivirus with the best knockdown efficiency was used for the following assays (Fig S2B). We induced CRC in mice by the AOM/DSS and knocked down COL9A1 by sh COL9A1 (Fig. 3A). After euthanizing mice after day 120, AOM/DSS significantly induced the development of colon tumors in mice compared to control mice, whereas subsequent knockdown of COL9A1 allowed mice to develop significantly fewer tumors in their colons than sh Ctrl (Fig. 3B). In addition, sh COL9A1-treated mice also had significantly longer colon lengths (Fig. 3C). Next, by IHC, COL9A1-positive staining was found to be significantly upregulated in the colon of AOM/DSS mice, while sh COL9A1 lentiviral infection successfully reduced COL9A1-positive staining (Fig. 3D). As analyzed by HE histological staining, the colon tissues of AOM/DSS-treated mice showed significant inflammatory infiltration with significantly elevated pathological scores; whereas the knockdown of COL9A1 significantly alleviated the pathology of the mouse colon (Fig. 3E).

Fig. 3.

Knockdown of COL9A1 attenuates CCSC stemness in AOM/DSS-induced primary CRC. (A) Schematic representation of the experimental procedure. (B) The number of tumors dissected from C57BL/6 mice induced with AOM/DSS. (C) Measurement of colon length in mice with different treatments. (D) IHC analysis of COL9A1 expression in colon tissues of mice with different treatments. (E) Analysis of inflammatory infiltration by HE staining. (F) The extent of collagen deposition was analyzed by Masson’s trichrome staining. IHC analysis of Collagen I and LOX (G) and CD133 and LGR5 (H) in primary tumors of mice. The bars represent the mean ± SEM (n = 6); *p <.05, *** p <.001, and **** p <.0001, as determined by one-way ANOVA with Tukey’s post hoc test

It was shown by Masson’s trichrome staining that the induction of AOM/DSS led to significant collagen fiber deposition in mouse colon tissues, but sh COL9A1 treatment downregulated the ECM accumulation in colon tissues (Fig. 3F). In addition, we also found by IHC analysis that the levels of Collagen I and LOX in mouse colon tissues were significantly upregulated by induction with AOM/DSS, which was reversed by COL9A1 knockdown (Fig. 3G). To further analyze CCSC in mouse colon tumors, we dissected the colorectal primary tumors and showed that positive staining for CD133 and LGR5 was significantly downregulated in tumors derived from mice injected with sh COL9A1 (Fig. 3H).

Knockdown of COL9A1 inhibits tumor formation and modulates the CRC immune microenvironment

We performed microenvironment cell populations counter (MCPcounter) in CRC in Sangerbox 3.0 (http://vip.sangerbox.com/) and found that COL9A1 expression in CRC was negatively correlated with infiltration of cytotoxic lymphocytes (Fig S2C) and myeloid dendritic cells (Fig S2D) in the tumor tissue.

CT26 cells with sh COL9A1 or sh Ctrl were administered into BALB/c mice to monitor tumor growth for 25 days. Knockdown of COL9A1 inhibited tumor growth, and tumor volume (Fig. 4A) and weight (Fig. 4B) were downregulated. Thereafter, tumors were collected for flow cytometry analysis showing that knockdown of COL9A1 upregulated the number of infiltrating CD8⁺ T cells and NK cells in mouse tumors (Fig. 4C). The expression of immune activation markers in TILs was evaluated by flow cytometry, and it was found that silencing of COL9A1 led to an elevation in the relative rate of IFN-γ, Gzmb, and Prf1 in CD8⁺ TILs (Fig. 4D). Notably, inhibition of COL9A1 expression resulted in a greater proliferative capacity of NK cells and an increased ratio of Ki67⁺ NK cells (Fig. 4E). Not only that, we also explored the effect of COL9A1 on CD8⁺ T cell and NK cell activity in vitro. We co-cultured COL9A1 knockdown-treated CT26 cells with CD8⁺ T cells, followed by flow cytometry. The level of IFN-γ secreted by CD8⁺ T cells in the co-culture system was significantly upregulated after knockdown of COL9A1 in CT26 cells (Fig S2E). CCK-8 analysis showed that COL9A1 knockdown significantly reduced CT26 cell survival, but the overall trend indicated that the cells were proliferating (Fig S2 F). In contrast, in the co-culture system, the viability of CT26 cells showed an overall decreasing trend, indicating negative cell number growth, which was stronger after knockdown of COL9A1 (Fig S2G). In addition to this, we also co-cultured CT26 cells with NK cells. Knockdown of COL9A1 in CT26 cells upregulated the level of IFN-γ secreted by NK cells (Fig S2H), and the viability of CT26 in the co-culture system was significantly downregulated (Fig S2I).

Fig. 4.

Silencing of COL9A1 deters tumor growth and activates anti-tumor immunity in vivo. CT26 cells knocking down COL9A1 were delivered subcutaneously into BALB/c mice, and the volume of the subcutaneous tumors was measured every 4 d (A). After 25 days, the tumors were removed for recording and weighed (B). The number of CD8⁺ T cells and NK cells infiltrated in the tumors (C), the relative rate of IFN-γ, Gzmb, and Prf1 was assessed in CD8⁺ TILs from mice (D), the relative rate of Ki67-positive NK cells (E), and number of cDC1/mDC expressing CD103, XCR1 and IRF8 infiltrated in mouse tumors (F) were analyzed using flow cytometry. (G) IHC analysis of Collagen I and LOX expression in subcutaneous tumor tissues of BALB/c mice. The bars represent the mean ± SEM (n = 6); *p <.05, ** p <.01, *** p <.001, and **** p <.0001, as determined by unpaired t-test (B, C, D, E, F, G) or two-way ANOVA with Tukey’s post hoc test (A)

Interestingly, we also identified by flow cytometry that the number of classical type 1 dendritic cells (cDC1/mDC) infiltrated in mouse tumors expressing CD11c⁺MHCII⁺CD103⁺ XCR1⁺IRF8⁺ was significantly upregulated by COL9A1 inhibition (Fig. 4F). It was also observed that the downregulation of COL9A1 expression significantly limited the proportion of Collagen I- and LOX-positive staining in mouse tumor tissues (Fig. 4G).

KMT2D mediates COL9A1 enhancer H3K4me1 modification and recruits ZNF460 to enhance COL9A1

Transcriptional regulatory signals around the COL9A1 promoter were predicted on the UCSC Genome Browser (https://genome.ucsc.edu/index.html) using Genome Track (ENCODE Regulation and JASPAR2022 TFBS hg38) (Fig. 5A). We observed that the transcription factor ZNF460 binds to the region around the COL9A1 promoter and that there is an H3 K4 me1 binding signal around this promoter region. H3 K4 me1 marks poised or active enhancers, and KMT2D (MLL4) catalyzes H3 K4 me1 (Kubo et al. 2024). In UALCAN (https://ualcan.path.uab.edu/index.html), we observed that protein expression of ZNF460 and KMT2D was elevated in CRC tissues (Fig. 5B). The specific binding site (Fig. 5C) of ZNF460 on the COL9A1 promoter fragment (chr6:70,278,762–70,279,761) was predicted in Jaspar (https://jaspar.genereg.net/), suggesting that ZNF460 has the potential to bind to the COL9A1 promoter. Whether ZNF460 binds to the COL9A1 promoter region to promote COL9 A1 transcription and whether this is dependent on KMT2D mediating the H3 K4 me1 modification of the enhancer fragment (chr6:70,279,762–70,283,236) is not yet known.

Fig. 5.

KMT2D binds the COL9A1 enhancer via H3K4me1 modification and recruits ZNF460 to activate COL9A1. (A) Analysis of transcriptional regulatory signals around the COL9A1 promoter by Genome Track on the UCSC Genome Browser (https://genome.ucsc.edu/index.html). (B) The protein expression of ZNF460 and KMT2D in CRC tissues was predicted in UALCAN (https://ualcan.path.uab.edu/index.html). (C) Prediction of the binding site of ZNF460 on the COL9A1 promoter fragment (chr6:70,278,762–70,279,761) by Jaspar (https://jaspar.genereg.net/). (D) The enrichment level of COL9A1 by ZNF460 antibody after knockdown of ZNF460 in HCT 116 cells was examined using ChIP-qPCR. (E) The effect of the knockdown of ZNF460 on luciferase activity in HCT 116 cells was examined using a dual-luciferase assay. The enrichment level of COL9A1 by ZNF460 antibody after knockdown of KMT2D (F) or overexpression of KMT2D (G) by ChIP-qPCR analysis. The level of COL9A1 enhancer fragment enrichment by KMT2D antibody or H3K4me1 antibody after knockdown of KMT2D (H) or overexpression of KMT2D (I). KMT2D expression (J) and H3K4me1 levels (K) in HCT 116 cells after sgRNA transfection were examined using immunofluorescence staining. (L) The enrichment level of COL9A1 enhancer by anti-H3 K4 me1 after transfection with sgRNA by ChIP-qPCR analysis. (M) COL9A1 transcript levels in HCT 116 cells after sgRNA transfection by RT-qPCR analysis. The results are shown for three biological studies. The bars represent the mean ± SEM; *p <.05, ** p <.01, *** p <.001, and **** p <.0001, as determined by a t-test (E) or one-way ANOVA (J, K, M) and two-way ANOVA with Tukey’s post hoc test (D, F, G, H, I, L)

By IHC, we found that the positive staining of both ZNF460 and KMT2D was stronger in human CRC tissues than in paracancerous tissues (Fig S3A). Then, the expression of ZNF460 and KMT2D was higher in HCT 116 cells (Fig S3B). First, to verify the potential regulatory relationship between ZNF460 and COL9A1, we generated HCT 116 cells with stable knockdown of ZNF460 and applied sh ZNF460 #1 to subsequent experiments (Fig S3 C). ChIP-qPCR experiments showed that there was no enrichment of COL9A1 on IgG, but there was substantial COL9A1 enrichment by the ZNF460 antibody. The COL9A1 promoter fragment pulled down by the ZNF460 antibody was significantly reduced after the ZNF460 knockdown (Fig. 5D). The luciferase activity was reduced in HCT 116 cells after inhibition of ZNF460 (Fig. 5E). These data demonstrate that ZNF460 occupies the COL9A1 promoter region and activates its expression.

We subsequently selected sh KMT2D #2 with the best knockdown efficiency to construct sh KMT2D-treated HCT 116 cells (Fig S3D) and developed cells overexpressing KMT2D (Fig S3E). Interestingly, we found that neither knockdown nor overexpression of KMT2D affected the expression of ZNF460 in HCT 116 cells (Fig S3 F), but affected the mRNA expression of COL9A1, with knockdown of KMT2D downregulating COL9A1 expression, and overexpression of KMT2D enhancing its expression (Fig S3G). Thereafter, we performed ChIP-qPCR experiments on HCT 116 cells interfering with KMT2D expression and found that silencing of KMT2D downregulated the enrichment of COL9A1 by anti-ZNF460 (Fig. 5F); whereas enforced expression of KMT2D increased the enrichment of COL9A1 by anti-ZNF460 (Fig. 5G). Notably, we cloned primers for the enhancer of COL9A1 (chr6:70,279,762–70,283,236) and found that the enhancer fragments enriched by anti-KMT2D and anti-H3 K4 me1 were significantly downregulated after knockdown of KMT2D (Fig. 5H); the level of enrichment of COL9A1 enhancer fragments by anti-KMT2D and anti-H3 K4 me1 was elevated in cells after overexpression of KMT2D (Fig. 5I).

Recent findings suggest that KMT2D is equipped with two discrete low-complexity domains (LCDs) that play an important role in the assembly of membrane-less condensates (Li et al. 2021). We deleted the LCD1 and LCD2 structures of KMT2D by CRISPR-Cas9 and transfected them into HCT 116 cells. KMT2D protein levels were not decreased after LCD1 or LCD2 knockdown by immunofluorescence analysis, suggesting that deletion of KMT2D LCD1 or KMT2D LCD2 does not affect the stability of KMT2D protein expression (Fig. 5J). To explore the catalytic activity of KMT2D, the abundance of intracellular H3 K4 me1 was assessed using immunofluorescence. Knockout of either LCD1 or LCD2 significantly downregulated the level of H3 K4 me1 (Fig. 5K). Thereafter, ChIP-qPCR assay showed that LCD1 or LCD2 deletion significantly reduced the level of H3 K4 me1 enrichment at the COL9A1 enhancer (Fig. 5L), which led to downregulation of the COL9A1 mRNA level (Fig. 5M).

Activation of ZNF460 or COL9A1 induces CCSC stemness, inhibited by knockdown of KMT2D

To demonstrate the function of the KMT2D/ZNF460/COL9A1 axis in regulating CCSC stemness, we generated LRG5⁺ CCSCs with sh Ctrl, sh KMT2D, sh KMT2D + oe Ctrl, sh KMT2D + oe ZNF460, and sh KMT2D + oe COL9A1. The knockdown of KMT2D reduced the expression of KMT2D and COL9A1 but did not affect the ZNF460 expression. Subsequent reactivation of ZNF460 not only significantly elevated the level of ZNF460 but also reversed the repressive effect of knockdown of KMT2D on the expression of COL9A1 without affecting the expression of KMT2D. Sh KMT2D + oe COL9 A1 co-treatment only significantly elevated the expression of COL9 A1, but did not alter the expression of KMT2D and ZNF460 (Fig. 6A). The inhibitory effects of KMT2D knockdown on CCSC tumor initiation ability were all rescued by the combination of oe ZNF460 or oe COL9A1, and the tumor sphere-forming ability of CCSC was elevated (Fig. 6B). As expected, the downregulated protein levels of CD133 and LGR5 caused by KMT2D knockdown were reversed by the subsequent overexpression of ZNF460 or COL9A1 (Fig. 6C). Immunofluorescence staining also showed that reactivation of ZNF460 or COL9A1 rescued the inhibitory effect of the knockdown of KMT2D on the proliferative capacity of CCSC (Fig. 6D).

Fig. 6.

Activation of ZNF460 or COL9A1 reverses the inhibitory effect of sh KMT2D on CCSC stemness and proliferative capacity. (A) The mRNA expression of KMT2D, ZNF460, and COL9A1 in LRG5⁺ CCSCs infected with sh KMT2D combined with oe ZNF460 or oe COL9A1 was examined using RT-qPCR. (B) The sphere-forming ability of LGR5⁺ CCSCs after infection was examined using sphere-formation assays. (C) The protein expression of CD133 and LGR5 in LGR5⁺ CCSCs after infection was assessed using Western blot analysis. (D) The proportion of Ki67-positive LGR5⁺ CCSCs after infection was examined using immunofluorescence staining. The results are shown for three biological studies. The bars represent the mean ± SEM; *p <.05, ** p <.01, *** p <.001, and **** p <.0001, as determined by one-way ANOVA with Tukey’s post hoc test

KMT2D recruitment of ZNF460 mediates ECM stiffness to increase CCSC stemness via ZNF460

LRG5⁺ CSCs with sh KMT2D alone or in combination with ZNF460 or COL9A1 overexpression were subjected to colony formation assay and tumor sphere formation assay under polyacrylamide gels at different stiffness. The knockdown of KMT2D reduced the number of colonies and spheres formed, which were induced again by the overexpression of ZNF460 or COL9A1. The alterations were generally more pronounced with a stiffness of 20.0 kPa (Fig. 7A, B). The expression of stemness markers (CD133 and LGR5), Collagen I, and LOX was significantly higher in CCSC at 20.0 kPa gels than in cells of the 2.0 kPa group. Moreover, the expression of CSC markers and ECM markers was downregulated by KMT2D silencing and was significantly enhanced by the combined ZNF460 or COL9A1 overexpression (Fig. 7C). Finally, upon Transwell assay, it was found that KMT2D knockdown decreased the number of cells migrating and invading in both ECM stiffness, which was compromised by overexpression of ZNF460 and COL9A1. It was noteworthy that the promotion of cell migration and invasion in 20.0 kPa was highly significant by restoration of ZNF460 and COL9A1 (Fig. 7D).

Fig. 7.

KMT2D/ZNF460 axis mediates COL9A1 expression to increase ECM stiffness to promote CCSC stemness and mobility. (A) Colonies formed by LGR5⁺ CCSCs in gels with 2.0 kPa or 20.0 kPa stiffness. (B) Spheres formed by LGR5⁺ CCSCs in gels with 2.0 kPa or 20.0 kPa stiffness. (C) The protein expression of Collagen I, LOX, CD133, and LGR5 in LGR5⁺ CCSCs in gels with 2.0 kPa or 20.0 kPa stiffness was assessed using Western blot analysis. (D) The migration and invasion ability of LGR5⁺ CCSCs in gels with 2.0 kPa or 20.0 kPa stiffness was assessed using the Transwell assay. The results are shown for three biological studies. The bars represent the mean ± SEM; *p <.05, ** p <.01, *** p <.001, and **** p <.0001, as determined by a two-way ANOVA with Tukey’s post hoc test

Overexpression of COL9A1 stimulates tumorigenesis in AOM/DSS-induced mice

We first infected CT26 cells using three shRNAs targeting KMT2D. The best knockdown efficiency among them was mKmt2 d [shRNA#3], which was therefore selected for subsequent experiments (Fig. 8A). Next, CT26 cells knocking down KMT2D combined with COL9A1 overexpression were generated by lentiviral infection, and RT-qPCR analysis verified the effect of infection. COL9 A1 levels downregulated by KMT2D knockdown were reversed by COL9A1 overexpression, and only KMT2D knockdown downregulated the expression of KMT2D (Fig. 8B). Primary CRC was induced in mice by the AOM/DSS, and then lentiviruses were delivered to the mice. After euthanizing the mice at the endpoint of the experiments (day 120), mouse colons were collected. Compared to AOM/DSS + sh Ctrl, sh KMT2D significantly reduced the number of colon tumors and increased colon length, but combined COL9A1 overexpression interventions resulted in more colon tumors and a shorter colon length (Fig. 8C, D). HE staining showed that the inflammatory infiltrate and downregulated pathological scores in the colon tissues of mice relieved by KMT2D knockdown were reversed by COL9A1 overexpression (Fig. 8E). Masson trichrome staining results also showed that inhibition of KMT2D significantly reduced collagen fiber deposition in colon tissues, whereas overexpression of COL9A1 exacerbated the collagen deposition pathology (Fig. 8F). Subsequently, we also found by IHC analysis that positive staining for Collagen I, CD133, and LGR5 was downregulated by sh KMT2D treatment in colon tissues; whereas reactivation of COL9A1 increased the positive staining for Collagen I, CD133, and LGR5 (Fig. 8G). The knockdown of KMT2D in colon tissues significantly downregulated the transcript levels of KMT2D and COL9A1. By contrast, the overexpression of COL9A1 only upregulated the levels of COL9A1, but did not affect the mRNA levels of KMT2D (Fig. 8H).

Fig. 8.

Overexpression of COL9A1 potentiates tumorigenesis in AOM/DSS-induced mice. Validation of sh KMT2D lentivirus efficiency (A) and sh KMT2D + oe COL9A1 (B) lentivirus efficiency in CT26 cells by RT-qPCR. After the euthanasia of C57BL/6 mice, the colon tissues of mice were dissected to analyze tumor load (C) and colon length (D). (E) Inflammatory infiltration of colon tissues was examined using HE staining. (F) The extent of collagen deposition in colon tissues was examined using Masson trichrome staining. (G) IHC analysis of CD133, LGR5, and Collagen I expression in primary tumors. (H) mRNA levels of KMT2D, and COL9A1 in primary tumors were examined using RT-qPCR. The bars represent the mean ± SEM (n = 6); *p <.05, ** p <.01, *** p <.001, and **** p <.0001, as determined by one-way ANOVA with Tukey’s post hoc test

KMT2D/COL9A1 axis mediates the immune escape of CRC cells in vivo

Consistently, mouse CT26 cells with the same infection were delivered subcutaneously into BALB/c mice, which were monitored for 25 days before being euthanized. Both the volume (Fig. 9A) and weight (Fig. 9B) of mouse tumors were downregulated by the knockdown of KMT2D, which were reversed by COL9A1 overexpression treatments. The CD8⁺ T cell and NK cell number were significantly elevated in mouse tumors with the suppression of KMT2D (Fig. 9C), and the relative rates of activation markers (IFN-γ, Gzmb, Prf1) of CD8⁺ T cells was also upregulated in mouse TILs (Fig. 9D). The NK cell proliferative capacity was also increased upon KMT2D knockdown (Fig. 9E). Combined overexpression of COL9A1 led to a decrease in CD8⁺ T cell and NK cell ratios in the tumors, downregulated IFN-γ, Gzmb, and Prf1 expression in the TILs, and reduced Ki67⁺ NK cell ratios.

Fig. 9.

Reactivation of COL9A1 reverses immune escape of CRC cells in vivo. (A) CT26 cells treated as described above were delivered into BALB/c mice via subcutaneous inoculation, and the subcutaneous tumor volume was measured every 4 days during the 25 d. (B) The weight of dissected tumors. The number of CD8⁺ T cells and NK cells infiltrated in the tumors (C), the relative rate of IFN-γ, Gzmb, and Prf1 was assessed in CD8⁺ TILs from mice (D), the relative rate of Ki67-positive NK cells (E), and number of cDC1/mDC expressing CD103, XCR1 and IRF8 infiltrated in mouse tumors (F) were analyzed using flow cytometry. (G) IHC analysis of Collagen I and LOX expression in subcutaneous tumor tissues of BALB/c mice. The bars represent the mean ± SEM (n = 6); *p <.05, ** p <.01, *** p <.001, and **** p <.0001, as determined by one-way (B, C, D, E, F, G) or two-way ANOVA with Tukey’s post hoc test (A)

Thereafter, in vitro co-culture experiments showed that KMT2D knockdown resulted in upregulation of the level of IFN-γ (Fig S4A) and reduction of CT26 cell viability (Fig S4B), which were overturned by reactivation of COL9A1. However, knockdown of KMT2D treatment inhibited the proliferation of CT26 cells. Additional overexpression of COL9A1 treatment promoted cell proliferation, which was contrary to the negative proliferation of CT26 in the co-culture system (Fig S4C). The same was true in the CT26 and NK cell co-culture system; knockdown of KMT2D resulted in upregulation of IFN-γ levels in NK cells (Fig S4D) and enhanced NK cell toxicity to CT26 cells (Fig S4E), which was rescued by reactivation of COL9A1.

We also identified by flow cytometry that the number of mDC infiltration in mouse tumors was also upregulated by knockdown of KMT2D, yet similarly reversed by overexpression of COL9A1 (Fig. 9F). Eventually, ECM-related protein expression in mouse tumors was analyzed by IHC, and the positive staining rate of Collagen I and LOX was downregulated by KMT2D knockdown, and upregulated by overexpression of ZNF460 or COL9A1 (Fig. 9G).

Discussion

It is imperative to consider the targeting of CSCs within the complex milieu of the TME to achieve the overarching objective of identifying novel avenues for the development of CSC-based cancer immunotherapy (Borlongan and Saha 2024). Based on our studies, we can propose a model in which KMT2D coordinates ECM stiffness and tumorigenesis by regulating CCSC stemness and immune escape in CRC (Fig. 10). As a methyltransferase, KMT2D catalyzes H3 K4 me1 modification in the enhancer region of COL9A1 and recruits ZNF460 to the promoter region of COL9A1 to enhance its expression. COL9A1 actively participates in the regulation of CCSC stemness and immune cell activity, thereby upholding CRC development and progression. Genetic repression of KMT2D triggers an immunomodulatory function, impeding ECM stiffness and the self-renewal capacity of CCSC, which was abated by the overexpression of ZNF460 or COL9A1.

Fig. 10.

Schematic diagram. KMT2D enhances ECM stiffness in CRC by mediating the H3 K4 me1 modification in the enhancer and recruiting the transcription factor ZNF460 to activate COL9A1 expression, which induces the self-renewal of CCSCs, impairs the immune response, and ultimately promotes the malignant progression of CRC

As summarized by Martino et al., collagens represent a superfamily of proteins with a high abundance in animals, and therapeutic strategies that target collagen and its interactions with the ECM may hold potential for the prevention of tumor progression and metastasis (De Martino 2023). Moreover, matrix stiffness in the invasion frontier is an imperative factor controlling the biological function of CSCs, including stemness maintenance, invasion, as well as metastasis (Liu et al. 2023a). Conti et al. have recently found that compared to LGR5⁻ CRC cells, LGR5⁺ CSCs are stiffer and adhere better to the ECM (Conti et al. 2024), highlighting the association between ECM stiffness and CCSC subpopulation maintenance. Interestingly, Zekri et al. showed that COL9A1 was significantly elevated in breast CSCs under different sorting standards (Zekri et al. 2022). Even though several members of the collagen family, including COL9A3 and COL2 A1, have been identified to be downregulated by curcumin to inhibit the ECM-receptor interaction pathway in LGR5⁺ CCSC (Mao et al. 2021), there is little information regarding the function of COL9A1 in regulating CCSC stemness. Our in vitro evidence that suppression of COL9A1 repressed the colony formation and sphere formation in LGR5⁺ CCSC under normal conditions and matrix gels with high stiffness substantiated our bioinformatics prediction results. These in vitro findings were reproduced in vivo in AOM/DSS-induced mice.

ECM is shaped by immune cells, which contribute to an immune suppressive microenvironment and hinder the efficacy of immunotherapies, making the ECM a barrier to protect cancer against treatments and supporting tumor progression (Yuan et al. 2023). The loss of COL9A1 has been reported by Probst et al. to destabilize the trabecular bone network, impair myeloid cell differentiation, and affect the innate immune response against Listeria (Probst et al. 2018). However, its immunoregulatory role in cancers has not been revealed. In the context of CRC, IFN-γ, released by cytotoxic T cells and NK cells, has been associated with prolonged patient survival, and the killing effects of CD8 + T cells are linked to the presence of cytotoxic molecules, such as granzymes and perforin (Zheng et al. 2023). Moreover, intratumoral NK and CD8⁺ T cells are activated with IFN-γ overproduction and proliferation (Ki67⁺), and intratumoral enrichment of XCR1⁺IRF8⁺CD103⁺ cDC1 (Bergamaschi et al. 2020). We consistently observed the activation of these markers upon the silencing of COL9A1 in BALB/c mice injected with CT26 cells.

As for the upstream mechanism responsible for the COL9A1 dysregulation, Imagawa et al. found that the CpG sites of the COL9A1 promoter were hypermethylated in osteoarthritic patients as compared with controls (Imagawa et al. 2014). Here, we identified the binding signals of transcription factor ZNF460 and H3 K4 me1 around the promoter region of COL9A1. High expression of ZNF460 was correlated with increased depth of invasion, lymph node metastasis, and distant metastasis, and predicted OS and recurrence-free survival in patients with colon cancer (Hao et al. 2021). In addition, ZNF460 could promote circRPPH1 expression, thereby increasing the migration and invasion of triple-negative breast cancer cells (Zhang et al. 2024). Intriguingly, H3 K4 methyltransferases mediate cell-autonomous regulation to sustain CSC, which provides critical insights into epigenetic regulatory mechanisms underlying breast cancer dormancy with KMT2D as a potential therapeutic target (Ferrer-Diaz et al. 2024). KMT2D knockout HCT116 cells provided strong evidence of the role of KMT2D in binding enhancer and maintaining global H3 K4 me1 levels (Froimchuk and Jang 2017). It has been revealed that tumors harboring the KMT2 mutation exhibited enhanced immunogenicity, infiltration of immune cells, and levels of immune cell cytotoxicity (Wang et al. 2024; Zhang 2021). Kim et al. employed KMT2D knockout mice to demonstrate that KMT2D regulates activation-induced naïve CD8 + T cell survival via the modulation of H3 K4 me1 levels in enhancer regions of genes associated with apoptosis and immune function (Kim et al. 2022). More relevantly, KMT2D loss-of-function variants were related to augmented levels of CD8⁺ T cells and activated NK cells in CRC (Liu et al. 2023b). Here, the fact that the activating effects of KMT2D silencing on CD8⁺ T cells, NK cells, and mDC were reversed by the overexpression of COL9A1 further indicated that KMT2D recruited ZNF460 to the promoter region and enabled H3 K4 me1 in the enhancer region to induce the expression of COL9A1.

Our current research has some limitations. First, direct evidence, such as CUT&Tag and ChIP-seqencing data, is required to further substantiate the mechanism by which KMT2D regulates COL9A1 expression via H3 K4 me1 modification. In addition, KMT2D has been identified as a tumor suppressor, and its mutational status significantly distinguished treatment-sensitive microsatellite instability cell lines from treatment-insensitive microsatellite instability cell lines (Takemon et al. 2024). Therefore, further fundamental experiments are needed to expand its clinical significance.

Our study establishes the critical role of KMT2D as a methyltransferase in CRC, coordinating ECM stiffness and immune escape through the maintenance of CCSC. For the first time, we demonstrate that blocking KMT2D inhibits CRC tumorigenesis and progression in vitro and in vivo and improves the immunosuppressive microenvironment in CRC. Moreover, our findings reveal that regulating KMT2D/ZNF460 on COL9A1 could expand our current concept of epigenetic activation in cancer therapy.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ECM

Extracellular matrix

CCSC

Colorectal cancer stem cells

CRC

Colorectal cancer

COL9A1

Collagen alpha-1(IX) chain

KMT2D

Histone-lysine N-methyltransferase 2D

ZNF460

Zinc finger protein 460

WT

Wild type

BSA

Bovine serum albumin

DMEM

Dulbecco's modified Eagle's medium

EGF

Epidermal growth factor

bFGF

Basic fibroblast growth factor

FACS

Fluorescence-activated cell sorting

TIL

Tumor-infiltrating lymphocytes

HBSS

Hanks' buffered salt solution

Author contributions

CXS conceived and designed the experiments and produced the initial draft of the manuscript. JMX, WJS, XYY and YXX collected the data and carried out the data analyses and statistical analysis. PH provided resources, supervision, and review of the manuscript. All the authors have read and approved the final submitted manuscript.

Funding

This work was supported by the Collaborative Innovation Cultivation Project for Universities in Heilongjiang Province (LJGXCG2023-087) and the Climbing Advantage Discipline Plan of Harbin Medical University Affiliated Cancer Hospital (PDYS-2024–14).

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

The study protocol was approved by the Institutional Review Board of Harbin Medical University Affiliated Cancer Hospital (approval no KY2019-22) and was conducted following the Declaration of Helsinki. Written informed consent was obtained from each patient. All animal experiments were approved by the Animal Ethical and Welfare Committee of Harbin Medical University Affiliated Cancer Hospital (approval no. KY2022-11) and carried out following the Guide for the Care and Use of Laboratory Animals of the NIH.

Consent for publication

Not applicable.

Conflicts of interest

The authors declare no competing interests.