KRT17 Promotes T-lymphocyte Infiltration Through the YTHDF2–CXCL10 Axis in Colorectal Cancer

KRT17 is highly expressed in deficient mismatch repair colorectal cancer, where it promotes T-cell infiltration via the YTHDF2–CXCL10 axis. KRT17 expression correlates with improved immunotherapy responsiveness, highlighting a potential prospective biomarker and therapeutic target for immunotherapy.

Abstract

Poor infiltration of T lymphocytes has been regarded as a crucial mechanism of tumor immune escape. Here, we demonstrate a protective role of KRT17 in colorectal cancer, where KRT17 reversed the tumor immunosuppressive microenvironment by increasing T-lymphocyte infiltration. High-throughput RNA sequencing suggested that KRT17 was significantly upregulated in deficient mismatch repair (dMMR) tumors compared with proficient mismatch repair (pMMR) tumors. In a colorectal cancer cohort of 446 cases, KRT17 expression positively correlated with better clinical outcomes. Krt17 overexpression decreased xenograft tumor growth in immune-competent mice. T-cell depletion in a murine model showed that the presence of T lymphocytes was necessary for Krt17-mediated disruption of tumorigenesis. Mass spectrometry and coimmunoprecipitation assays suggested KRT17 caused YTHDF2 degradation through the ubiquitin-proteasome system. Through high-throughput RNA immunoprecipitation sequencing, we found that CXCL10 was the target gene of the N6-methyladenosine (m⁶A) “reader” YTHDF2. KRT17 synergized with anti–PD-1 for better tumor control in an immunotherapy-resistant murine model. In a cohort of patients with colorectal cancer receiving pembrolizumab, high KRT17 expression was found within the tumors of responders. Collectively, we elucidated a critical role of KRT17 in colorectal cancer to prevent immune escape. These findings present new insights into potential therapeutic strategies and effective markers of immunotherapy reactivity against pMMR tumors.

Introduction

The immune system safeguards the host from most pathogens and aberrant cells through multifactorial mechanisms, including innate and adaptive effector approaches. A malfunctioning immune system can cause a variety of illnesses, including autoimmune conditions, tumorigenesis, and pathogenic infections. Extensive internal and extrinsic factors, such as defective immunological molecules (1) and the application of immunosuppressants (2), can result in restricted antitumor immune responses and tumor progression. Emerging clinical evidence has demonstrated that immune checkpoint inhibitors (ICI) have durable antitumor activity in multiple cancer types (3, 4). Understanding the mechanisms by which tumors compromise immunity will be crucial for disease prevention and therapeutic interventions.

Immune escape refers to a process by which tumor cells evade immunologic detection and/or eradication through multiple pathways, including genetic or epigenetic alterations, cellular cross-talk, and recruitment of suppressive cells to the tumor microenvironment (TME; ref. 5). Reduced T-lymphocyte infiltration into tumors is one of the critical mechanisms behind immune escape (6). Rapidly accumulating data have begun to elucidate that tumors inhibit T-lymphocyte infiltration through multiple pathways, including immunoediting of tumor neoantigens (7), modified chemokine expression profiles (8), hijacking of the T-cell checkpoint pathways (9), and abnormal vasculature (10). Approximately 85% of patients with colorectal cancer have almost no cytotoxic T-lymphocyte (CTL) infiltration, and these patients usually have poor survival outcomes and high recurrence risk (11, 12). However, the mechanisms governing T-lymphocyte infiltration in colorectal cancer remain largely unknown, and elucidating the potential regulatory mechanisms behind CTL infiltration is essential for developing effective strategies for tumor immunotherapy.

In this study, we identified a role and molecular mechanism of KRT17 in T-lymphocyte infiltration in colorectal cancer. To explore key factors that regulated CTL infiltration, RNA sequencing (RNA-seq) was performed to compare transcriptomic differences between deficient mismatch repair (dMMR) and proficient mismatch repair (pMMR) tumor tissues. We identified that KRT17 was highly expressed in dMMR tumors. In subcutaneous MC38 and CT26 models, Krt17 overexpression increased T-lymphocyte infiltration to reverse immune escape. Mechanistically, KRT17 promoted ubiquitin-mediated degradation of YTHDF2, which decreased the decay of N6-methyladenosine (m⁶A)-modified CXCL10 transcripts. Importantly, high KRT17 expression was found within the tumors of responders in a cohort of patients with colorectal cancer who had received pembrolizumab. In summary, these findings indicate that KRT17 is a promising immunotherapeutic target and an effective marker of immunotherapy reactivity in colorectal cancer.

Materials and Methods

Colorectal cancer patient samples

Frozen, fresh colorectal cancer tissues were obtained from three dMMR and three pMMR cases to analyze transcriptome alterations by RNA-seq as described previously. Twenty paired colorectal cancer samples and normal adjacent tissues (including 10 dMMR and 10 pMMR cases) were used to analyze KRT17 mRNA expression by qRT-PCR (details below). Five paired colorectal cancer samples and normal adjacent tissue (including 5 dMMR and 5 pMMR cases) were used to analyzed KRT17 protein expression by western blot assay (see below). Tumor tissues were collected immediately after surgery. Some tissues were then stored at −80°C with RNAlater solution (Invitrogen). The remaining tissues were directly stored at −80°C. All patients did not undergo either chemotherapy or radiotherapy before surgery. 446 patients with colorectal cancer (224 dMMR and 222 pMMR patients) who received surgery at our hospital provided formalin-fixed, paraffin-embedded (FFPE) colorectal cancer tissue samples. Patients did not receive neoadjuvant chemoradiotherapy, except high-risk stage II and III patients. Clinical parameters were obtained from the patient electronic medical records located in our hospital. All human tissue samples were obtained with written informed consent from donors. Only paraffin sections from 446 patients with colorectal cancer were collected retrospectively. A total of 120 patients were randomly selected from 446 patients to analyze the difference of KRT17 expression in tumor tissue and adjacent normal tissue, including 60 dMMR and pMMR cases. All procedures were carried out in accordance with the declaration of Helsinki and were approved by the Institutional Review Committee of the Sixth Affiliated Hospital of Sun Yat-sen University.

Colorectal cancer patient cohort treated with anti–PD-1

Patient eligibility criteria included pathologic confirmation of colorectal cancer, at least four weeks of prior treatment with the anti–PD-1 drug pembrolizumab, adequate tumor biopsy samples, available baseline patient and disease information, and sufficient archival tissue available for analysis. We collected endoscopic biopsy samples from all included patients with colorectal cancer. All tissues were fixed in formalin and then embedded in paraffin. Clinical, radiographic, and treatment data for each patient were collected retrospectively. All human tissue samples were obtained with written informed consent from donors. All procedures were carried out in accordance with the declaration of Helsinki. A cohort of 30 patients with colorectal cancer receiving pembrolizumab was established in our study, and all patients were histologically diagnosed dMMR cancer, including 15 cases in each of the responder and non-responder groups. The objective clinical response was defined by RECIST version 1.1.

Cell lines and cultures

All human colon cancer cell lines, including DLD1, HCT8, WiDr, Lovo, SW48, HCT116, RKO, HCT15, and HEK293T, were all obtained from the ATCC. The mouse colon cancer cell line CT26 was obtained from the ATCC and the murine colon cancer cell line MC38 was obtained from the National Infrastructure of Cell Line Resource (Beijing, China). DMEM (Gibco, Thermo Fisher Scientific) mixed with 10% FBS (Gibco, Thermo Fisher Scientific) was used to culture all cells in a 5% CO₂ atmosphere. All cell lines were validated by STR DNA finger-printing and the reauthentication took place in 2021. All cell lines were routinely tested for Mycoplasma. Cells were generally passaged 1–2 times before being used for subsequent experiments. All cell lines were passaged within 15 times in our study.

Protocols for transfection and lentivirus infection are described previously in the indicated section. After cells were transfected 48 hours with the indicated plasmid, 25-μm MG132 (Sigma-Aldrich) or DMSO (Sigma-Aldrich) was added into medium and cultured 12 hours. For IFNγ stimulation, MC38 and CT26 cells were transfected with vector and Krt17 plasmids with 50 ng/mL IFNγ (NovoProtein) or BSA (Sigma-Aldrich) for 48 hours.

Animal models

BALB/C nude mice (4–5 week), BALB/c mice, and C57BL/6J mice (6–8 weeks) were obtained from VitalRiver Laboratory Animal Technology. All mice were fed in pathogen-free conditions at the Experimental Animal Center of the Sixth Affiliated Hospital of Sun Yat-sen University. For the xenograft tumor model using nude mice, a total of 2 × 10⁵ MC38 or CT26 cells transfected with vector or with Krt17 lentivirus were resuspended in 100 μL of a mixture with PBS and Matrigel (Corning; 1:1) and then subcutaneously inoculated into right flank of each nude mice. For the xenograft tumor model using immune-competent mice, a total of 2 × 10⁵ MC38 cells transfected with vector or with Krt17 lentivirus were subcutaneously inoculated into right flank of each C57BL/6J mouse; a total of 2 × 10⁵ CT26 cells transfected with vector or with Krt17 lentivirus were subcutaneously inoculated into right flank of each BALB/c mouse. For in vivo depletion of CD8⁺ T cells, either MC38 or CT26 tumor-bearing mice were injected intraperitoneally with 200 μg (10 mg/kg) of either anti-CD8 (Bio X Cell) or IgG (Bio X Cell) every four days starting on day 4 after tumor inoculation. For the in vivo anti–PD-1 treatment study, either C57BL/6J mice bearing MC38 tumors or BALB/c mice bearing CT26 tumors received intraperitoneal injection with either 200 μg (10 mg/kg) of anti–PD-1 (Bio X Cell) or IgG (Bio X Cell) every three days starting on day 6 after tumor inoculation. For the in vivo neutralizing anti-CXCL10 treatment study, BALB/c mice bearing CT26 tumors received intraperitoneal injection with either 40 μg (2 mg/kg) of anti-CXCL10 (RD system) or IgG (RD system) every three days starting on day 15. Tumor volumes were measured on the basis of the following formula: Volume (mm³) = (long diameter × short diameter²)/2. Tumor long and short diameters were monitored by caliper, and mice were checked every two days. For tumor growth assays, mice were sacrificed when tumor volume of any one mouse reached 2,000 mm³ or there was a clear difference in tumor volume between control and experimental groups. For mouse survival analysis, the mouse was sacrificed when its own tumor volume reached 2,000 mm³ and survival time was calculated from the time after tumor inoculation. Tumor weights were also measured. Tumor tissues were collected and then used in all following experiments. Animal experiments were authorized by the Institutional Animal Care and Use Committee of Sun Yat-sen University and complied with the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health (National Academies Press, 2011) in China.

Tissue dissociations

Tumor tissues obtained from patients with colorectal cancer and mouse models were dissociated to perform extract RNA or protein. For RNA extraction, tumor tissues were removed from the −80°C freezer and were cut into approximately 3-mm pieces on ice with a blade. The tissues were transferred to a new 2-mL Eppendorf tube, and then 4-mm steel balls (Servicebio) and 1-mL TRizol reagent (Invitrogen) were added. A tissue crusher (Qiagen Tissue Lyser LT) was used to dissociate the tissues with a frequency of 60 Hz, a time of 10 minutes. After tissue dissociations, RNA was extracted (described below). For protein extraction, 500-μL T-PER tissue protein extraction reagent (Thermo Fisher Scientific) was used instead of 1-mL TRizol reagent, and the remaining steps were as described above.

Cell isolations and activation

Spleens were isolated from BALB/c mice or C57BL/6J mice. A 70-μm cell filter (Corning) was placed in a 10-cm cell culture dish (Corning), 6 mL lymphocyte separation solution (BioLegend) was added, and mouse spleens were quickly minted and ground. The lymphocyte separation solution was transferred into 15-mL centrifuge tube (Corning), and 1-mL serum-free DMEM (Gibco) was slowly added into centrifugal tube. Samples were centrifuged at 800 x g for 30 minutes at 4°C. The tunica albugnean layer was collected and washed once by adding 5 mL precooled serum-free DMEM (Gibco), and centrifuged at 250 x g for 10 minutes at 4°C. Finally, the supernatant was discarded. For cell activation, the isolated cells were cultured in 8 mL DMEM (Gibco) containing 10% FBS (Gibco), 50 U/mL IL2 (MCE), and 200 μL mouse CD3/CD28 dynbeads (Invitrogen) for 48 hours.

m⁶A-RNA immunoprecipitation assay

Samples used for the m⁶A-RNA immunoprecipitation (MeRIP) assay included MC38 and CT26 cells, cells treated with 50 ng/mL IFNγ (NovoProtein) or BSA (Sigma-Aldrich), cells transfected with vector and Krt17 plasmids, and/or cells transfected with coexpressing plasmids. Each group contained three biological replicates, and m⁶A abundance of Cxcl10 mRNA was evaluated by qRT-PCR. Total RNA was obtained by TRizol Reagent (Invitrogen). 300-ng RNA was retained as an input sample for qRT-PCR (described below), and the residual sample was used for MeRIP. MeRIP was performed by adding residual RNA into 500 μL immunoprecipitation (IP) buffer [10 mmol/L Tris, 150 mmol/L NaCl, 0.1% NP-40 (APE×BIO), 100 U RNase inhibitor (Omega), pH 7.4] and incubated with 4 μg/mL m⁶A antibody (Abcam). The m⁶A-IP mixture was incubated for 2 hours at 4°C, and then Dynabeads Protein A (Life Technologies) coated with BSA was added into the m⁶A-IP mixture and incubated for an additional 2 hours at 4°C. The mixture was then washed in IP buffer and eluted by competition with 50 μg/mL N6-methyladenosine (Santa Cruz Biotechnology). 20-μL RLT buffer (Thermo Fisher Scientific) was used to elute the bound RNA from beads. The ReverTra Qpcr RT Master Mix kit (TOYOBO) was used to reverse-transcribe the extracted IP RNA and input RNA into cDNA library according to its instructions. m⁶A enrichment of Cxcl10 was evaluated by qRT-PCR.

qRT-PCR

qRT-PCR samples included human and mouse colon cancer cell lines, whole-tumor tissues of patients with colorectal cancer, and whole-tumor tissues of the xenograft tumor model. Tissues were dissociated before RNA extraction (described above). Total RNA was extracted by TRizol Reagent (Invitrogen). The ReverTra Qpcr RT Master Mix kit (TOYOBO) was used to reverse-transcribe the extracted RNA into cDNA library according to its instructions. The SYBR Green PCR Master Mix (Applied Biosystems) was used to perform qRT-PCR on LightCycler 96 (Roche). According to its instructions, a volume of 10-μL RT-qPCR reaction system was prepared as follows: 5 μL 2X SYBR Green Mix + 0.25 μL PCR Forward Primer (10 μmol/L) + 0.25 μL PCR Reverse Primer (10 μmol/L) + 1 μL cDNA template sample + 3.5 μL ddH2O. Each sample contained three biological replicates. We applied the 2^–ΔΔCT method and normalized to β-actin to determine gene expression. For MeRIP–qRT-PCR, equal amounts of either IP RNA or input RNA from each sample was extracted for construction of the cDNA library. Reverse transcription and PCR processes are similar to those described above. The relative m⁶A abundance of Cxcl10 was measured by its expression levels of MeRIP sample under normalization by its expression levels of input sample. All genes were assessed by qRT-PCR and their primer sequences were shown (Supplementary Tables S1 and S2).

RNA-seq analysis

Using TRizol reagent (Invitrogen), total RNA was isolated from mouse xenograft tumor tissues (MC38 cell line with vector and Krt17 lentivirus) or colorectal cancer tumor tissues (colorectal cancer cases with dMMR and pMMR). For each group, three biological replicates were used. Tissues were dissociated before RNA extraction (described above). RNA purity was examined on a Bioanalyzer 2200 device (Aligent), and the RNA integrity number of all samples ranged from 7 to 9. Ribosomal RNA was then removed using the RiboMinus Eukaryote Kit (Qiagen), and a cDNA library was created by a NEB Next Ultra RNA Library Prep Kit (New England Biolabs). cDNA fragments were purified and end repaired, and a poly(A) tail was added. Finally, using an Illumina HiSeq 3000 (Illumina), deep sequencing was carried out, aligning the clean reads to the reference genome (GRCH37.p13 NCBI). DESeq2 software was used to examine the differences in gene expression between two groups. The genes with the parameter of absolute fold change >2 and FDR < 0.01 were defined differentially expressed genes. qRT-PCR was used to confirm the expression of relevant genes (primer sequences shown in Supplementary Tables S1 and S2).

Plasmid construction, transfection, and lentivirus infection

For KRT17, Krt17, Ythdf2, Cxcl10-expressing plasmids, the full-length ORF sequences with tag of these genes (KRT17 GeneID: 3872, Krt17 GeneID: 16667, Ythdf2 GeneID: 213541, Cxcl10 GeneID: 15945) were, respectively, subcloned into the pcDNA3.1 vector (Thermo Fisher Scientific). The activity of the internal ribosome entry site (IRES) was examined with pCMV-IRES-Renilla Luciferase-IRES-Gateway-Firefly Luciferase (pIRIGF) vector (Hebei Yancheng Biotechnology). Colorectal cancer cells, including DLD1, HCT8, MC38, and CT26 cell lines, were transfected with above vectors using Lipofectamine 3000 (Invitrogen). For construction of shKRT17 plasmids, shKRT17 sequences were cloned into pLKO.1 vector (Hebei Yancheng Biotechnology). All constructs were verified by sequencing. For stable lentivirus infection, HEK293T cells were transfected with the indicated plasmids, psPAX2 and pMD2G (Addgene) according to the manufacturer's instructions. Colorectal cancer cells were then transfected with the above supernatant in the presence of 8 μg/mL polybrene (Beyotime) and selected with puromycin (MCE, China) after infection. For construction of coexpressing cells used in rescue experiments, equal amounts of Krt17 and Ythdf2 plasmids or lentivirus were added into the medium, the next steps of transfection or lentivirus infection were consistent with the above steps. The oligonucleotide sequences for vector construction are listed in Supplementary Table S3.

IP assay

The cells transfected with the relevant plasmids were lysed in Pierce IP lysis buffer (Thermo Fisher Scientific) in the presence of a protease inhibitor cocktail (including protease inhibitors and phosphatase inhibitors; Thermo Fisher Scientific) for 30 minutes at 4°C. Centrifugation was conducted on cell lysates at 13,000 x g for 20 minutes at 4°C. The antibodies used in IP assays included anti-KRT17 (1:30, Abcam) and anti-YTHDF2 (1:50, Abcam). Next, the antibodies were applied to the 200-μL Pierce Protein A/G magnetic beads (Thermo Fisher Scientific) for 4 hours at 4°C to crosslink the antibodies. The antibody-crosslinked beads were mixed with the 500-μL cell lysates overnight at 4°C. After being washed with washing buffer (Cayman) five times, the beads were incubated with 2× sample loading buffer (Beyotime) and then boiled for 10 minutes. The lysates were resolved on SDS-PAGE (described below).

Western blotting

Western blotting samples included human and mouse colon cancer cell lines, whole-tumor tissues of patients with colorectal cancer, and whole tumor from the xenograft tumor model. Tissues were dissociated before protein extraction (described above). Total protein was extracted by T-PER tissue protein extraction reagent (Thermo Fisher Scientific). The concentration of protein samples was analyzed with the BCA standard protein kit (Dingguo), and quantitative experiments were carried out on the basis of its instructions. 5× SDS loading buffer (Dingguo) was used. 1–10 μg protein per sample was loaded into gel based on the sample type and primary antibody. The blots were blocked with 5% nonfat milk (Biotopped), and 0.45-μm polyvinylidene difluoride membrane (Millipore) was used to transfer protein from the 10% running gel (Servicebio). Primary antibodies used for western blotting are listed in Supplementary Table S4. The dilution of primary antibodies was in accordance with their instructions. Secondary antibodies of HRP-label IgG included anti-mouse (1:1,000, Servicebio) and anti-rabbit (1:1,000, Servicebio). Mini-PROTEAN Tetra Cell, Mini Trans-Blot Module, PowerPac Basic Power, and ChemiDoc (Bio-Rad) were used. SDS-PAGE Running Buffer Powder, SDS-PAGE Transfer Buffer Powder, TBS powder, and Tween-20 (Servicebio) were used. Blotting for β-actin (1:1,000, Proteintech) was used as a loading control.

Cell proliferation, migration, and invasion assays

DLD1 cells transfected with overexpressed plasmids (vector or KRT17 plasmids) and HCT8 cells transfected with knockdown plasmids (shNC, shKRT17–1, and shKRT17–2 plasmids) were used to determine the effect of KRT17 on colorectal cancer cell phenotypes. Cell proliferation was examined using the proliferation curve and plate colony formation. For the proliferation curve, 6,000 cells were plated into 96-well plates (Corning) with serum-free DMEM (Thermo Fisher Scientific), and cell proliferation was monitored every 10 hours using an IncuCyte Essens Bioscience incubator (Essens Bioscience). For plate colony formation, 600 cells were seeded in 6-well plates (Corning) with serum-free DMEM (Thermo Fisher Scientific) and cultured for 10 days. The colonies were then fixed with 4% paraformaldehyde (Servicebio) and stained with 0.1% crystal violet (Servicebio) for 15 minutes at room temperature. Cell colonies were imaged and counted with inverted microscope (Leica). Cell migration was determined by a wound healing assay. A total of 4×10⁶ cells were plated into 6-well plates (Corning) and incubated until confluency was reached. A rectilinear scratch was performed using a 100-μL pipette (Corning) tip. After 24 hours, cells were fixed with 4% paraformaldehyde (Servicebio) for 15 minutes, and then stained with 0.1% crystal violet (Servicebio) for 15 minutes at room temperature. Wound closure was imaged using an inverted microscope (Leica) and ImageJ software (NIH) was used for quantitative analysis. For invasion assays, 24-well plates (Corning) with 8-μm pore size chamber inserts (Corning) were used. A total of 4×10⁴ cells were seeded in the upper chamber well with 200 μL serum-free DMEM and 800 μL of DMEM (Thermo Fisher Scientific) with 20% FBS (Thermo Fisher Scientific) was added into the lower chamber. After 48 hours, cells migrating through the membrane were fixed with 4% paraformaldehyde (Servicebio) for 15 minutes, and then stained with 0.1% crystal violet (Servicebio) for 15 minutes. The migrating cells were imaged using an inverted microscope (Leica) and ImageJ software (NIH) was used for quantitative analysis.

Cycloheximide-chase experiments

MC38 or CT26 cells transfected with vector or Krt17 plasmids were used to perform cycloheximide-chase experiments. Using DMEM (Thermo Fisher Scientific) with 10% FBS (Thermo Fisher Scientific), a total of 2×10⁵ cells were plated into 6-well plates (Corning). 100 μg/mL cycloheximide (Thermo Fisher Scientific) was added to the medium, and cells were harvested at 0, 2, 4, 6, 8, 10 hours for protein extraction. Western blotting was used to analyze the expression of YTHDF2 protein (described above).

ELISA

For analysis of CXCL10 production, 100 μL supernatant was collected after MC38 and CT26 cells were transfected with indicated plasmid (vector and Krt17 plasmids, coexpressed plasmids, or vector and Cxcl10 plasmids) in presence of 50 ng/mL IFNγ for 48 hours. A RayBio Human or Mouse Cytokine Antibody Array (Raybiotech, Inc.) was used, and a standard curve was performed according to the manufacturer's protocol. The ELISA plate was assessed using an enzyme-label instrument (Thermo Fisher Scientific) with wavelength set at 450 nm. The sample concentration was calculated by substituting the instrument readings into the formula obtained from the standard curve.

IHC and scoring

All samples were FFPE tissues from a cohort of 446 patients with colorectal cancer, a cohort of colorectal cancer treated with pembrolizumab, and animal models described above. Mouse/Rabbit Polymer Test System Universal kit (ZS-bio) was used to perform IHC according to its instructions. Briefly, paraffin sections were routinely dewaxed and hydrated. Citrate solution (Servicebio) was selected and placed in microwave (Midea) on medium-high heat for 8 minutes for antigen repair. The citrate solution was cooled to room temperature naturally. Endogenous peroxidase blocking agents (ZS-bio) were added and incubated for 10 minutes at room temperature. Primary antibodies used for IHC are listed in Supplementary Table S4. After incubation overnight at 4°C, enzyme-labeled anti-goat IgG polymer (ZS-bio) was incubated for 20 minutes at room temperature. Diaminobenzidine solution (ZS-bio) was used to stain at room temperature. Microscopy (Leica) was used to observe staining results and to collect images. Tumor region of human and mouse tumor tissues was assessed for the expression of KRT17, YTHDF2, and CXCL10. The staining scores were quantified on the basis of the staining intensity and proportion of positively stained cells. The proportion of positively stained cells was determined by the percentage of positive stained area using the following four groups: grade 0 = 0, grade 1 = 1% to 25%, grade 2 = 26% to 50%, grade 3 = 51% to 75%, and grade 4 = >75%. The staining intensity was divided into the following four groups: grade 0 with no staining, grade 1 with weak staining, grade 2 with medium staining, and grade 3 with strong staining. The staining scores were calculated using the following formula: Staining scores = staining intensity × proportion of positively stained cells. The expressions of KRT17, YTHDF2, and CXCL10 were determined by the staining scores on a scale from 0 to 12 and divided into low and high groups based on the median. The CD3⁺ and CD8⁺ T-cell density per patient was assessed on tumor parenchyma (TP) and invasive margin (IM), which determined by counting the number of positive cells per tissue sample using tumor tissue sections, as described previously (13). Using similar method, the infiltration level of T lymphocyte on tumor region of mouse tumor tissue was assessed.

Immunofluorescent staining

Mouse MC38 and CT26 cell lines were chosen to evaluate cellular localization of KRT17 and YTHDF2 proteins. A cell slide (Thermo Fisher Scientific) was spread into a well of 12-well plate (Corning), and each well was inoculated with 2×10⁴ MC38 or CT26 cells. 4% paraformaldehyde (Servicebio) was used to fix cells for 30 minutes at room temperature. 0.5% Triton X-100 (Servicebio) was added to break cell membrane for 30 minutes at room temperature and 1% BSA (Thermo Fisher Scientific) was used to block for 1 hour at room temperature. Specific primary antibodies, including anti-KRT17 (1:200, Abcam) and anti-YTHDF2 (1:200, Abcam) were incubated with the cells overnight at 4°C. Samples were added with Alexa 488- or 594-conjugated goat antibodies (1:1,000, Thermo Fisher Scientific) against rabbit or mouse IgG. The cells assessed using a confocal laser scanning microscope (Leica TCS-SP8) after being counterstained with DAPI (Merck Millipore).

Actinomycin D assay

For the half-life of mRNA assessment, after MC38 and CT26 cells were transfected with indicated plasmid (vector and Krt17 plasmids, or coexpressed plasmids) with 50 ng/mL IFNγ for 48 hours, gene transcription was blocked by adding 5 mg/mL Actinomycin D (Sigma-Aldrich) to the cells. DMSO was used as a negative control. Cells were harvested at 0, 6, 12, 24 hours and the stability of Cxcl10 was analyzed by qRT-PCR as described above.

In vitro T-cell migration assays

In a Transwell device with a polycarbonate membrane with an 8-μm pore size (Corning), an in vitro migration experiment was performed. The top compartment contained resuspended 2×10⁴ activated CD8⁺ T cells after two washes using sterile PBS (BOSTER), and the bottom chamber contained 1 mL nondiluted conditioned media from MC38 or CT26 cells transfected with indicated plasmids (vector and Krt17 plasmids, co-expressed plasmids, or vector and Cxcl10 plasmids) with 50 ng/mL IFNγ (NovoProtein) for 48 hours. After the coculture for 24 hours, the media of the bottom chamber were collected; it was centrifuged at 13,000 x g for 20 minutes at 4°C. After centrifugation, the supernatant was discarded and cells at the bottom of the eppendorf tube were resuspended in 20 μL of PBS. Using Cell counting analyzer (Nexcelom), cell numbers in the bottom chamber were then counted.

RNA immunoprecipitation assay and RNA IP sequencing

The Magna RIP RNA-Binding Protein IP Kit (Merck Millipore) was used to carry out the RNA immunoprecipitation (RIP) assay according to its instructions. For RIP, anti-YTHDF2 (1:1,000, Abcam) was used and included mouse MC38 and CT26 cells, and those transfected with vector and Krt17 plasmids and those transfected with coexpressing plasmids. Equal amounts of either IP RNA or input RNA from each sample was extracted for construction of the cDNA library. Reverse transcription and PCR processes are similar to those described above. The relative expressions of indicated genes (Cxcl10, Ccl2, and Ccl7) were measured by its expression in the RIP sample after normalization of its expression in the input sample. For RIP sequencing (RIP-seq), the mouse MC38 cell line was chosen to extract RNA. For next-generation sequencing (NGS), input RNA and RIP RNA were generated by Geneseed Biotech Company. The TruSeq Stranded mRNA Sample Prep Kit (Illumina) was used for NGS library preparation, a BioAnalyzer High Sensitivity DNA chip (Geneseed Biotech) was used for quantification, and a 19Illumina HiSeq 2500 was used for deep sequencing. Using Bowtie2 version 2.1.0, the reads were first mapped to the most recent UCSC transcript set (14), and RSEM version 1.2.15 was used to assess gene expression. The RSEM approach was used to quantify the RNA-seq reads and RIP-seq reads at the gene level, and they were standardized to fragments per kilobase of exon model per million mapped fragments.

Mass spectrometry

The gel from the IP assay using anti-KRT17 (Abcam) was cut into tiny pieces. Peptides were dispersed in 0.1% formic acid (FA; Thermo Fisher Scientific) and 2% acetonitrile (ACN; Thermo Fisher Scientific) following trypsin digestion (Promega) and then were immediately placed onto a reversed-phase analytical column (Thermo Fisher Scientific). The gradient consists of increasing solvent B (0.1% FA in 80% ACN) from 5% to 50% in 40 minutes, while maintaining a consistent flow rate of 300 nL/min, as recommended by the manufacturer. Using Q Exactive hybrid quadrupole-Orbitrap mass spectrometer (Thermo Fisher Scientific), MS analysis was conducted. Detection model included reflection mode and positive ion spectroscopy, the ion source accelerated voltage was 20 KV, and each peptide map was accumulated by 200 bombardments. Thermo Fisher Scientific's Q Exactive was used to perform tandem mass spectrometry (MS/MS) on the peptides, which was then connected online to perform ultra-performance liquid chromatography. The m/z scan range for MS scans was 350 to 1800 m/z. A fixed initial mass of 100 m/z was chosen. By using Uniprot Aedis Aegypti as a search engine, MASCOT software (Matrix Science) was used to identify proteins.

Statistical analysis

All data are presented as mean ± standard deviation (SD) unless otherwise noted. Data analysis was conducted using SPSS16.0 statistical package (IBM). For continuous variables with normal distributions, a two-tailed Student t test or one-way ANOVA was used to establish statistical significance, whereas either Mann–Whitney or Kruskal–Wallis test was applied for skewed distributions. For survival analysis, Kaplan–Meier plots and log-rank tests were used. GraphPad Prism 8 (GraphPad) was used to make ROC curve and analyze AUC. Statistical significance was defined as a P value of 0.05.

Availability of data and materials

The RNA-seq data of colorectal cancer cases with different MMR status used in the study (HRA002906) are available in the GSA-Human public repository. The RNA-seq data of xenograft tumor tissues in mice (CRA007974) are available in GSA. The RIP-seq data in this study were deposited at GSA database with an accession number (CRA007975). The MS data have been stored to the ProteomeXchange Consortium via the PRIDE partner repository (15) with the dataset identifier PXD036433.

Results

High KRT17 expression indicates favorable clinical outcomes in dMMR colorectal cancer

dMMR tumors tend to have massive CTL infiltration in colorectal cancer (11). To uncover specific transcriptomic differences that might influence CTL infiltration, we performed a transcriptomic analysis using three dMMR cases and three pMMR cases (Fig. 1A). By analyzing RNA-seq data, we found that the mRNA transcript expression of 159 genes was upregulated and 62 genes were downregulated in dMMR tumors compared with pMMR tumors (Fig. 1B). The expression of the top 5 upregulated genes was then detected in each group of 10 colorectal cancer tissues using qRT-PCR. These results suggested only KRT17 (keratin, type Ι cytoskeletal 17) was differentially expressed between dMMR and pMMR tumor tissues (Fig. 1C and D). Moreover, mRNA expression was also detected in 20 paired tissues and showed KRT17 expression was higher in tumor tissues compared with normal tissues (Supplementary Fig. S1A). A similar association was found in stratified analyses of patients with dMMR or pMMR status (Supplementary Fig. S1A). Western blots confirmed that protein expression of KRT17 was higher in dMMR colorectal cancer tissues than pMMR colorectal cancer tissues (Fig. 1E). Similarly, Western blot analysis of 10 paired colorectal cancer samples and normal adjacent tissues showed KRT17 was highly expressed in tumor tissues (Supplementary Fig. S1B).

Figure 1.

High KRT17 expression indicates favorable clinical outcomes in dMMR colorectal cancer. A, Schematic showing the screening schedule to find out the key role of KRT17 on transcriptome analysis using three dMMR and pMMR cases. B, Volcano plot of different expressed genes obtained by DEseq2 analysis in dMMR tumor tissues compared with pMMR tumor tissues. 159 significantly upregulated or 62 downregulated genes are plotted in red and blue points. C, Heat map of the top five up- or downregulated genes screened in dMMR and pMMR tumor tissues. D, qRT-PCR analysis of top five upregulated genes in 10 (each) dMMR and pMMR tumor tissues. E, Western blot analysis of KRT17 protein expression in 10 dMMR and pMMR tumor tissues. F, Immunostaining of KRT17 in the tumor in a representative human colorectal cancer case of different clinical grade with dMMR compared with a pMMR case; scale bars, 100 μm (top), 20 μm (bottom). G, Quantification of KRT17 expression by IHC score in all 446 human colorectal cancer samples. H and I, Kaplan–Meier survival curve of patients with colorectal cancer layered by the KRT17 expression in tumor tissue sections. Values are represented as mean ± SD. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns (no significance), by two-tailed Student t test (D), one-way ANOVA (G) or log-rank test (H and I).

To explore clinical characteristics of KRT17 expression, IHC was used to examine KRT17 expression in 446 colorectal cancer cases, including 224 dMMR and 222 pMMR cases (Fig. 1F). In all colorectal cancer samples, KRT17 expression gradually decreased in more advanced TNM stages (Fig. 1G). Similar results were found in our subsequent stratified analyses of patients with dMMR or pMMR status (Fig. 1G). In addition, we analyzed 120 paired tumor tissues, including 60 (each) dMMR and pMMR cases, which suggested that KRT17 expression in tumor tissues was higher than in corresponding normal adjacent tissues (Supplementary Fig. S1C and S1D). Kaplan–Meier curves showed that higher KRT17 expression associated with better overall survival [OS; log-rank test, P = 0.0004; hazard ratio (HR), 0.3345; 95% confidence interval (CI), 0.1921–0.5824] and disease-free survival (DFS) in all cases (log-rank test, P < 0.0001; HR, 0.3287; 95% CI, 0.2214–0.4878; Fig. 1H and I). Stratification of the cohort into patients with dMMR (n = 224) or pMMR (n = 222) found that high KRT17 expression associated with good OS in dMMR cases (log-rank test, P < 0.0001; HR, 0.1536; 95% CI, 0.0735–0.3210), but not in pMMR cases (log-rank test, P > 0.05; Fig. 1H). However, patients with high KRT17 expression had less recurrence at 3 years, regardless of MMR status (dMMR: log-rank test, P < 0.0001; HR, 0.2018; 95% CI, 0.1143–0.3562; pMMR: log-rank test, P = 0.0471; HR, 0.5292; 95% CI, 0.2965–0.9446; Fig. 1I). These data indicate a potential interplay between KRT17 expression and MMR status in tumor cells, which affects the prognosis of patients with colorectal cancer.

KRT17 suppresses tumor growth by increasing CTL infiltration into tumors

On the basis of the better prognosis of patients with colorectal cancer with high KRT17 expression, we considered whether KRT17 expression affected tumor growth. In all colorectal cancer samples, there was no significant correlation between KRT17 expression and proliferating tumor cells. Similar results were also found in dMMR and pMMR cases (Supplementary Fig. S1E). To further determine whether KRT17 affected colon cancer phenotypes, we tested KRT17 expression in human colorectal cancer cell lines by qRT-PCR (Supplementary Fig. S2A). The DLD1 cell line had the lowest KRT17 expression, thus we chose these cells for overexpression. Because of the weak migratory ability of the HT29 and WiDr cell lines (16), we chose the cells with the third highest KRT17 expression for knockdown. The efficacy of overexpression and knockdown in the two colorectal cancer cell lines was validated using qRT-PCR (Supplementary Fig. S2B) and Western blot (Supplementary Fig. S2C) assays. Next, the proliferation curve showed that neither overexpression nor knockdown of KRT17 changed cell growth (Supplementary Fig. S2D). Similarly, the colony formation assay indicated that either overexpression or knockdown of KRT17 did not affect the number and size colonies following a 10-day incubation (Supplementary Fig. S2E). In addition, neither the migratory nor invasive activities were not changed by either KRT17 overexpression or knockdown (Supplementary Fig. S2F and S2G). These results suggested that KRT17 had no effect on colon cancer phenotypes in vitro.

On the basis of the distinct difference of TME between dMMR and pMMR colorectal cancers, we assumed that KRT17 expression in tumor cells would affect tumor progression by changing the immune TME. To explore the effect of KRT17 on immune TME in tumors of different MMR status, we chose the MC38 and CT26 mouse colorectal cancer cell lines. Next, we constructed MC38 and CT26 with stable overexpression of Krt17 (Supplementary Fig. S2H and S2I) and subcutaneously inoculated into immune-deficient nude mice. The results indicated that Krt17 had no effect on the tumor growth in immune-deficient nude mice in vivo (Supplementary Fig. S2J and S2M). We next subcutaneously inoculated the two cell lines with or without stable Krt17 overexpression into immune-competent mice. Both MC38 and CT26 tumors with Krt17 overexpression were substantially smaller than tumors from vector cells (Fig. 2A). Consistently, the tumors from the Krt17 overexpression group showed decreased weight compared with tumors from the vector group (Fig. 2B and C). These results indicated that KRT17 inhibited tumor growth in immune-competent mice.

Figure 2.

KRT17 suppresses tumor growth by increasing CTL infiltration into tumors. A, Average growth curves of subcutaneous xenograft tumors in immune-competent mice after inoculation of MC38 or CT26 cells with vector or with Krt17 lentivirus. MC38, n = 6. CT26, n = 7. B and C, Representative pictures and tumors weight of xenograft tumors in immune-competent mice after inoculation of MC38 or CT26 cells with vector or with Krt17 lentivirus. B, Scale bars, 1 cm. D, Volcano plot of different expressed genes obtained by DEseq2 analysis in xenograft tumor tissues of MC38 with Krt17 lentivirus compared with the tumor tissues with vector lentivirus. 136 significantly upregulated or 156 downregulated genes are plotted in red and blue points. E, Meta-enrichment analysis summary for 292 significantly coregulated genes. F and G, qPCR analysis of IFNγ or Gzmb expression in the indicated tumors of xenograft tumor models; n = 5. H, Representative IHC staining of CD3 and CD8 in sections of mouse tumor tissues from the vector and Krt17 overexpression groups; scale bars, 100 μm (left), 20 μm (right). I and J, Quantification of CD3⁺ and CD8⁺ T lymphocytes in mouse tumor tissues; n = 5. K, Mice were given IgG or anti-CD8 antibody starting on the fourth day after tumor inoculation and treated on the indicated days for a total of 4 treatment. L, Tumor growth in mice bearing vector and Krt17 overexpression tumors treated IgG or anti-CD8 antibody. MC38, n = 7 (top). CT26, n = 6 (bottom). M and N, Representative pictures and tumors weight of xenograft tumors in mice after inoculation of MC38 or CT26 cells with vector or with Krt17 lentivirus along with IgG or anti-CD8 antibody treatment. M, Scale bars, 1 cm. O, Immunostaining of KRT17, CD3, and CD8 in the tumor in a representative human colorectal cancer case of different KRT17 expression with dMMR compared with a pMMR case; scale bars, 100 μm (left), 20 μm (right). P and Q, Quantification of CD3⁺ and CD8⁺ T lymphocytes in TP in all 446 human colorectal cancer samples, including 224 dMMR colorectal cancer samples and 222 pMMR samples. Values are represented as mean ± SD except for (A); l, mean ± SEM. TP, tumor parenchyma. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, no significance, by two-tailed Student t test (C, F, G, I, J, P, and Q), one-way ANOVA (A and N) or two-way ANOVA (L).

To fully understand the gene changes occurring in TME with Krt17 overexpression, we next conducted RNA-seq to investigate specific transcriptomic alterations of tumor tissues from tumor models. Transcriptomic analysis revealed that Krt17 overexpression in xenograft tumors resulted in significant alterations, including the upregulation of 136 genes and downregulation of 156 genes (Fig. 2D). Among the differentially expressed genes in tumors with Krt17 overexpression, we found increased expression of cytokines and cytotoxic molecules (e.g., IFNg, Cxcl11, Gzma, Cd2, and H2Aa). Gene oncology analysis was subsequently performed on these differentially expressed genes, and enriched pathways were found to be primarily associated with the immune system process, T-cell activation, chemotaxis, cytokine production, response to cytokine, and lymphocyte homeostasis (Fig. 2E). The expression of both IFNγ and granzyme B, two key markers of T lymphocytes, was consistently higher in the Krt17 overexpression group than in the vector group in the subcutaneous MC38 and CT26 models (Fig. 2F and G). Taken together, these data suggested that KRT17 played a key role in T-lymphocyte responses.

To further verify these results, we then examined the presence of CTLs in tumor tissue of xenograft tumor models. IHC was used to determine CTL infiltration as indicated by CD3⁺ and CD8⁺ T cells (Fig. 2H). These results indicated that there were significantly more CD3⁺ and CD8⁺ T cells in tumor tissues from the Krt17 overexpression group than in those from the vector group (Fig. 2I and J). Given that KRT17 appeared to have an inhibitory effect on tumorigenesis, we also investigated the potential contribution of CD8⁺ T cells to the observed decline in tumor growth using anti-CD8 in the xenograft tumor models (Fig. 2K). Treatment with anti-CD8 led to significantly faster tumor growth compared with the IgG control group. Moreover, there was no difference in either tumor volume or tumor weight between vector and Krt17 overexpression tumors in the anti-CD8 group (Fig. 2L–N). This demonstrated the indispensable role of CTL responses on KRT17-mediated effects on tumorigenesis.

KRT17 positively correlates with T-lymphocyte infiltration

To verify the potential association between KRT17 expression and T-lymphocyte infiltration in colorectal cancer, we assessed T-lymphocyte infiltration in 446 colorectal cancer samples, including 224 dMMR and 222 pMMR cases. We examined CD3⁺ and CD8⁺ cells in two tumor regions: The TP and IM as stated in previous research (11). Regardless of MMR status, both CD3⁺ and CD8⁺ cells in the TP were more abundant in high KRT17 expression tumors than those with low KRT17 expression (Fig. 2O–Q).

To further demonstrate the effect of KRT17 expression and CTL infiltration in the TP on prognosis, we performed stratified analyses of the cohort. We divided the cohort into two groups based on the median number of CD3⁺ and CD8⁺ cells in TP. Among patients with low number of CD3⁺ cells in TP, high KRT17 expression associated with good OS and DFS in dMMR cases, but not in all cases and pMMR cases (Supplementary Fig. S3A). However, among patients with high CD3⁺ cells in the TP, those with higher KRT17 expression had better OS and DFS in all cases and dMMR cases, but not in pMMR cases (Supplementary Fig. S3B). Survival and recurrence in all cases was stratified by KRT17 expression regardless of the number of CD8⁺ cells in the TP, and similar results were shown in dMMR cases but not in pMMR cases (Supplementary Fig. S3C and S3D). In addition, tumors with high KRT17 expression had more CD3⁺ and CD8⁺ cells in the IM than those with low KRT17 expression (Supplementary Fig. S4A and S4C). Stratification of the cohort into patients with different number of CD3⁺ and CD8⁺ cells in the IM found results similar to those in the TP (Supplementary Fig. S4D–S4G). Collectively, these observations suggested that KRT17 had a protective role in tumor progression by upregulating CTL infiltration in colorectal cancer.

KRT17 facilitates proteolytic degradation of YTHDF2 via ubiquitination

To elucidate the molecular mechanism behind KRT17-mediated regulation of the immune system process, we next performed a co-IP experiment using anti-KRT17 in conjunction with MS to uncover KRT17-interacting proteins. 10 immune-related proteins were found to bind to the KRT17 protein, among which YTHDF2 ranked first according to their respective protein scores (Supplementary Fig. S5A). The two peptide sequences for the YTHDF2 protein were identified by MS (Fig. 3A). We then performed reciprocal co-IP experiments and confirmed that KRT17 bound to YTHDF2 (Fig. 3B). Subsequent immunofluorescence assays indicated that this interaction may take place in the cytoplasm (Fig. 3C).

Figure 3.

KRT17 facilitates proteolytic degradation of YTHDF2 via ubiquitination. A, The identified YTHDF2 amino acids in immunoprecipitation (IP) products by mass spectrometry. B, Western blot analysis of IP using anti-KRT17 or anti-YTHDF2 antibody in MC38 and CT26 cells with indicated antibodies. C, Immunofluorescent staining of KRT17 and YTHDF2 in MC38 and CT26 cells transfected with vector and Krt17. D, qRT-PCR analysis of Krt17 and Ythdf2 expressions in MC38 and CT26 cells transfected with vector and Krt17. E, qRT-PCR analysis of Krt17 and Ythdf2 expressions in xenograft tumor tissue from the vector and Krt17 overexpression groups. F, Western blot analysis of KRT17 and YTHDF2 expressions in MC38 and CT26 cells transfected with vector and Krt17. G, Western blot analysis of KRT17 and YTHDF2 expressions in xenograft tumor tissue from the vector and Krt17 overexpression groups. H, Immunostaining of KRT17 and YTHDF2 in the tumor in a representative human colorectal cancer case of different KRT17 expression with dMMR compared with a pMMR case; scale bars, 100 μm (left), 20 μm (right). I, Quantification of KRT17 and YTHDF2 expressions in all 446 human colorectal cancer samples, including 224 dMMR colorectal cancer samples and 222 pMMR samples. J, Western blot analysis of KRT17 and YTHDF2 protein levels in MC38 or CT26 cells transfected with vector or Krt17 with cycloheximide (CHX, 100 μg/mL) treatment at the indicated time. The quantification of YTHDF2 expression is shown in the bottom. K, Western blot analysis of KRT17 and YTHDF2 protein levels in MC38 and CT26 cells transfected with vector or Krt17 with MG132 (25 μm) for 12 hours. L, Western blot analysis of YTHDF2 ubiquitin levels after IP using anti-YTHDF2 antibody in MC38 and CT26 cells transfected with the indicated plasmids. M, Western blot analysis of IP using anti-YTHDF2 antibody in MC38 and CT26 cells with indicated antibodies. N, Western blot analysis of IP using anti-YTHDF2 antibody in MC38 and CT26 cells transfected with vector or Krt17 plasmids. Values are represented as mean ± SD. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns (no significance), by two-tailed Student t test (D, E, and I), one-way ANOVA (J).

To further explore function of KRT17 binding, we tested the expression of YTHDF2 in Krt17-overexpressing cells. We found that the abundance of Ythdf2 mRNA level did not change with Krt17 overexpression in vitro and in vivo (Fig. 3D and E). However, YTHDF2 protein levels were decreased under Krt17 overexpression in vitro and in vivo (Fig. 3F and G). Because PSMB8 and SMAD3 have been reported to function in T-lymphocyte homeostasis (17, 18), we tested the effect of Krt17 overexpression on their protein expression. The results indicated that Krt17 overexpression did not change the protein level of PSMB8 and SMAD3 (Supplementary Fig. S5B). Thus, these results indicated that Krt17 overexpression inhibited YTHDF2 protein levels. To further verify the clinical correlation between KRT17 and YTHDF2 protein expressions in colorectal cancer cases, we assessed 446 colorectal cancer cases using IHC to analyze the expression of these two proteins (Fig. 3H). YTHDF2 protein expression negatively correlated with KRT17 protein expression in all cases, and a subgroup analysis showed similar results in dMMR and pMMR cases (Fig. 3I).

To confirm that KRT17 decreased the protein stability of YTHDF2, we designed a cycloheximide-chase experiment using tumor cells transfected with vector or Krt17. The YTHDF2 protein level was tested every two hours after cycloheximide treatment by Western blot assays. These results indicated that Krt17 overexpression lessened the half-life of the YTHDF2 protein (Fig. 3J). We also found that the effect of KRT17 on YTHDF2 protein levels could be inhibited by MG132, a proteasome inhibitor (Fig. 3K). Collectively, these results suggested that KRT17 induced the proteolytic degradation of YTHDF2. Consistent with these results, we also revealed that KRT17 fully enhanced YTHDF2 ubiquitination in both MC38 and CT26 cells (Fig. 3L). It has been reported that S-phase kinase–associated protein 2 (SKP2), which acts as the substrate recognition component of E3 ubiquitin ligase complexes, binds to YTHDF2 and prompts YTHDF2 proteolysis by the CRL–NEDD8 pathway (19). Therefore, we investigated whether KRT17 promoted YTHDF2 ubiquitination through the CRL–NEDD8 pathway. Co-IP assays verified that YTHDF2 bound to SKP2 in both MC38 and CT26 cells (Fig. 3M). Moreover, the interaction between YTHDF2 and SKP2 was stronger in Krt17-overexpressing cells than in vector cells (Fig. 3N). In summary, these results suggested that KRT17 prompted ubiquitination-mediated degradation of the YTHDF2 protein in colorectal cancer.

Ythdf2 overexpression blocks the immunomodulatory effect of KRT17

YTHDF2 has reported to be an oncogene in several cancers, including lung cancer (20), ovarian cancer (21), and acute myelogenous leukemia (22). However, the role of YTHDF2 in colorectal cancer still remains poorly understood. Therefore, IHC was conducted to analyze YTHDF2 expression levels in 446 colorectal cancer cases, including 224 dMMR and 222 pMMR cases (Supplementary Fig. S5C). YTHDF2 expression gradually increased with more advanced TNM stages in all cases, and subgroup analyses of patients with either dMMR or pMMR status also had similar results (Supplementary Fig. S5D). Notably, patients with high YTHDF2 expression had worse OS (log-rank test, P < 0.0004; HR, 3.540; 95% CI, 2.019–6.206) and DFS in all cases (log-rank test, P < 0.0001; HR, 3.244; 95% CI, 2.171–4.847; Supplementary Fig. S5E and S5F). Stratification of the cohort showed that higher YTHDF2 expression associated with worse OS in dMMR cases (log-rank test, P < 0.0001; HR, 7.440; 95% CI, 3.513–15.76), but not in pMMR cases (log-rank test, P > 0.05; Supplementary Fig. S5D). However, patients with higher YTHDF2 expression had higher recurrence at 3 years, regardless of MMR status (dMMR: log-rank test, P < 0.0001; HR, 4.024; 95% CI, 2.265–7.146; pMMR: log-rank test, P = 0.0013; HR, 2.679; 95% CI, 1.510–4.751; Supplementary Fig. S5E). These data suggested that YTHDF2 likely function as an oncogene in colorectal cancer.

To explore whether KRT17 affected the immune environment in colorectal cancer by prompting degradation of YTHDF2, we first generated mouse colorectal cancer cell lines with stable overexpression of Ythdf2 (Fig. 4A and B). Next, we also confirmed that KRT17-mediated suppression of tumors growth was dependent on YTHDF2, as Ythdf2 overexpression partially impaired the effect of Krt17 overexpression on tumor growth (Fig. 4C and E). Moreover, overexpression of both Ythdf2 and Krt17 decreased CTL infiltration compared with only Krt17 overexpression (Fig. 4F and G). Consistently, Ythdf2 overexpression decreased both IFNγ and granzyme B expressions (Fig. 4H). These results confirmed that the immunomodulatory effect of Krt17 overexpression was dependent on YTHDF2 degradation.

Figure 4.

Ythdf2 overexpression blocks the immunomodulatory effect of KRT17. A, qRT-PCR analysis of Ythdf2 expression in Ythdf2 overexpression MC38 or CT26 cells. B, Western blot analysis of YTHDF2 expression in Ythdf2 overexpression MC38 or CT26 cells. C, Average growth curves of subcutaneous xenograft tumors in immune-competent mice after inoculation of MC38 or CT26 cells with indicated lentivirus (n = 7). D and E, Representative pictures and tumors weight of xenograft tumors in immune-competent mice after inoculation of MC38 or CT26 cells with indicated lentivirus. D, Scale bars, 1 cm. F, Representative IHC staining of CD3 and CD8 in sections of mouse tumor tissues from the four group; scale bars, 100 μm (left), 20 μm (right). G, Quantification of CD3⁺ and CD8⁺ T lymphocytes in mouse tumor tissues (n = 5). H, qPCR analysis of IFNγ or Gzmb expression in the indicated tumors of xenograft tumor models (n = 5). I, Immunostaining of YTHDF2, CD3, and CD8 in the tumor in a representative human colorectal cancer case of different YTHDF2 expression with dMMR compared with a pMMR case; scale bars, 100 μm (left), 20 μm (right). J, Quantification of YTHDF2 expression, CD3⁺ and CD8⁺ T lymphocytes in the tumor in all 446 human colorectal cancer samples, including 224 dMMR colorectal cancer samples and 222 pMMR samples. Values are represented as mean ± SD except for C (mean ± SEM). TP, tumor parenchyma. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns (no significance); by two-tailed Student t test (A, H, and J), one-way ANOVA (E and G) or two-way ANOVA (C).

We further verified the interaction between YTHDF2 expression and T-lymphocyte infiltration in our colorectal cancer cohort of 446 patients (Fig. 4I). Patients with higher YTHDF2 expression had less CTL infiltration in the TP in all cases; similar results were found after stratification of the cohort by MMR status (Fig. 4J). Survival and recurrence in all cases was stratified by YTHDF2 expression, regardless of different number of CD3⁺ and CD8⁺ T cells in the TP; similar results were shown in dMMR cases but not in pMMR cases (Supplementary Fig. S6A–S6D). In addition, tumors with higher YTHDF2 expression had less CD3⁺ and CD8⁺ T cells in the IM than those with low KRT17 expression (Supplementary Fig. S7A–S7C). Stratification of the cohort into patients with different number of CD3⁺ and CD8⁺ T cells in the IM found results similar to those in the TP (Supplementary Fig. S7D and S7G). These findings suggested that YTHDF2 had an apparent impact in tumorigenesis by inhibiting adaptive immune responses. In summary, Ythdf2 overexpression abolished the effects of KRT17 on adaptive immune responses.

CXCL10 is a target of YTHDF2

YTHDF2 has been shown to facilitate the decay of the m⁶A-modified transcripts (23). To explore which genes YTHDF2 protein could bind to, we performed high-throughput RIP-seq using anti-YTHDF2. We found that 151 genes bound by YTHDF2 overlapped with upregulated genes in Krt17 overexpression tumors and 14 immune-related genes between these overlapped (Fig. 5A). Among these genes, Ccl2, Ccl7, and Cxcl10 have previously been reported to promote lymphocyte migration (24). We performed RIP assays to validate our RIP-seq results and found that CXCL10 was bound to YTHDF2 (Fig. 5B).

Figure 5.

CXCL10 is one of the target genes of YTHDF2. A, Venn diagram showing the screening schema to find out the target gene of YTHDF2 by RNA-seq of xenograft tumor tissues and RIP-seq using anti-YTHDF2 antibody. B, qRT-PCR analysis of RIP using anti-YTHDF2 antibody in MC38 and CT26 cells. C, qRT-PCR analysis of MeRIP using m⁶A antibody in MC38 and CT26 cells. D, qPCR analysis of Cxcl10 expression in the indicated tumors of xenograft tumor models (n = 5). E, qPCR analysis of Cxcl10 expression in MC38 and CT26 transfected with vector or Krt17 plasmids. F, qRT-PCR analysis of Cxcl10 expression in MC38 and CT26 cells transfected with vector and Krt17 plasmids with IFNγ (50 ng/mL) for 48 hours. G, Gene set enrichment analysis to assess specific enrichment of response of IFNγ in xenograft tumor tissues of MC38 with Krt17 lentivirus or with vector lentivirus. H, qRT-PCR analysis of MeRIP using m⁶A antibody in MC38 and CT26 with IFNγ (50 ng/mL) for 48 hours. I, qRT-PCR analysis of MeRIP using m⁶A antibody in MC38 and CT26 transfected with vector or Krt17 plasmids with BSA or IFNγ (50 ng/mL) for 48 hours. J, mRNA stability of Cxcl10 by qRT-PCR in MC38 and CT26 cells transfected with vector or Krt17 plasmids with stimulation of IFNγ. K, Schematic diagram of CD8⁺ T-cell migration assays. L, Quantification of CD8⁺ T cells passing through the membrane of a Transwell system. M, Immunostaining of YTHDF2 and CXCL10 in the tumor in a representative human colorectal cancer case of different KRT17 expression with dMMR compared with a pMMR case; scale bars, 100 μm (left), 20 μm (right). N, Quantification of YTHDF2 and CXCL10 expressions in 120 human colorectal cancer samples, including 60 dMMR colorectal cancer samples and 60 pMMR samples. Values are represented as mean ± SD. MeRIP, m⁶A-RNA IP. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns (no significance), by two-tailed Student t test (B–I, L, and N), one-way ANOVA (J).

Next, to investigate whether CXCL10 was modified with m⁶A methylation, we performed MeRIP assays using an m⁶A antibody. These results showed that Cxcl10 was methylated in both MC38 and CT26 cells (Fig. 5C). In addition, we found that Cxcl10 was upregulated in Krt17-overexpressing tumors compared with control tumors from two xenograft tumor models (Fig. 5D). However, Krt17 overexpression did not change the expression of Cxcl10 in vitro (Fig. 5E). CXCL10 is one of the IFNγ-stimulated genes (ISG), and previous studies have reported that the presence of IFNγ is necessary for immune regulation of m⁶A modification (25). Given this, we reasoned that CXCL10 was regulated by KRT17 when IFNγ was present. Transcriptional analysis of Krt17-overexpressing and vector cells in the presence of IFNγ suggested enhanced expression of Cxcl10 (Fig. 5G). CXCL10 promoted T-lymphocyte migration, which increased IFNγ expression (Fig. 2F and G). This finding was consistent with gene set enrichment analysis for responses to IFNγ in Krt17-overexpressing tumors showed increased IFNγ-related changes compared with control tumors (Fig. 5F).

Previous work has reported that IFNγ increases m⁶A modification (26). Thus, we assumed that IFNγ also increased the m⁶A levels of CXCL10. We performed MeRIP assays, which showed that IFNγ indeed increased the m⁶A levels of CXCL10 (Fig. 5H). We also showed that Krt17 overexpression increased m⁶A levels of CXCL10 with IFNγ stimulation (Fig. 5I). To ascertain whether the elevated mRNA of Cxcl10 in Krt17-overexpressing tumors was the result of increased mRNA stability, we next evaluated the half-life of Cxcl10 mRNAs. Vector and Krt17-overexpressing cells in the presence of IFNγ were administered actinomycin D for 0, 6, 12, and 24 hours. Results demonstrated that Krt17-overexpressing cells resulted in more stabilized Cxcl10 mRNAs than vector cells (Fig. 5J).

Rescue assays were conducted to confirm whether YTHDF2 was indispensable for KRT17-induced immune regulation. In vitro, Ythdf2 overexpression in Krt17-overexpressing MC38 and CT26 cells significantly decreased Cxcl10 expression in the presence of IFNγ (Supplementary Fig. S8A). To further verify the immune regulation of Krt17 overexpression dependence on m⁶A manner, we performed MeRIP assays. Results showed that Ythdf2 overexpression in Krt17-overexpressing cells decreased m⁶A-modified transcripts for Cxcl10 (Supplementary Fig. S8B). In addition, we assessed Cxcl10 expression in tumor tissues obtained from xenograft tumor models. Results showed that Cxcl10 was downregulated in Ythdf2-overexpressing tumors compared with control tumors (Supplementary Fig. S8C). Consistently, the analysis of mRNA stability revealed that Ythdf2-overexpressing cells significantly decreased stabilized Cxcl10 mRNAs relative to control cells (Supplementary Fig. S8D).

We analyzed the Cxcl10 in the supernatant from tumor cells transfected with the indicated plasmid by ELISA. The results showed that Krt17 overexpression increased CXCL10 in the supernatant, and Ythdf2 overexpression lessened the enhanced CXCL10 from Krt17 overexpression (Supplementary Fig. S8E and S8F). To directly verify T-lymphocyte regulation of KRT17, we next performed an in vitro migration experiment (Fig. 5K), which showed that the migratory ability of T cells was significantly improved following Krt17 overexpression. Moreover, Ythdf2 overexpression abolished this ability (Fig. 5L). Thus, we concluded that KRT17 regulated CXCL10 expression dependence on YTHDF2. In addition, we found that Cxcl10 expression had same effect as Krt17 overexpression (Supplementary Fig. S8G–S8M). There was more T-lymphocyte infiltration in tumor tissues from the Cxcl10 expression group than in those from the vector group (Supplementary Fig. S8N and S8O). In addition, Cxcl10 overexpression promoted IFNγ and Gzmb expressions (Supplementary Fig. S8P). Therefore, Cxcl10 overexpression was sufficient to inhibit tumor progression by increasing T-lymphocyte infiltration. To explore the correlation between YTHDF2 and CXCL10 expressions in colorectal cancer tissues, IHC was used to analyze YTHDF2 and CXCL10 protein expressions in 120 colorectal cancer samples, including 60 (each) dMMR and pMMR cases (Fig. 5M). These results suggested that CXCL10 expression negatively correlated with YTHDF2 expression in all cases and that a similar correlation was found in dMMR and pMMR cases (Fig. 5N). A correlation between CXCL10 and KRT17 expressions was also analyzed and suggested that CXCL10 expression positively correlated with KRT17 expression (Supplementary Fig. S8Q). Together, these results identified the KRT17–YTHDF2–CXCL10 axis as a regulator of T-lymphocyte migration in colorectal cancer tumors and suggested that Krt17 overexpression could be a viable new strategy for immunotherapy treatments.

KRT17 sensitizes colorectal tumors to immunotherapy

Less CTL infiltration is the most recognized cause of resistance to ICIs (27). Therefore, we hypothesized that KRT17 may be involved in the sensitivity of colorectal cancer to immunotherapy. Thus, we designed a treatment schedule to determine whether Krt17 overexpression influenced immunotherapy responses (Fig. 6A). We first treated C57BL/6 bearing MC38 colorectal tumors with anti–PD-1 or isotype control IgG. Anti–PD-1 had a modest effect on tumor growth compared with control IgG antibody treatment (Fig. 6B). In addition, mice bearing Krt17-overexpressing tumors had slower tumor growth and prolonged survival (Fig. 6B and C).

Figure 6.

KRT17 sensitizes colorectal cancer to immunotherapy. A, Schematic diagram of anti–PD-1 treatment models given IgG or anti–PD-1 antibody starting on the sixth day after tumor inoculation and treated on the indicated days for a total of 5 treatment. B and D, Tumor growth in mice bearing vector and Krt17 overexpression tumors treated IgG or anti–PD-1 antibody. n = 8. C and E, Survival of mice in each group. F, Immunostaining of KRT17, CD3, and CD8 in sections of tumor tissues by endoscopic biopsy and representative picture of MRI from the colorectal cancer responders and nonresponders with anti–PD-1 antibody treatment; scale bars, 100 μm (left), 20 μm (right). G–I, Quantification of KRT17 expression, CD3 and CD8 in all 30 human colorectal cancer samples, including 15 (each) responders and non-responders. J, ROC curve for the prognostic prediction model based on KRT17 expression. K, Schematic summarizing our proposed model for the CTL infiltration under KRT17 expression in colorectal cancer. KRT17 inhibits colorectal cancer immune evasion by ubiquitin-mediated degradation of YTHDF2 to decrease m⁶A-modified CXCL10 mRNA decay, which promotes CTL trafficking into tumor. Values are represented as mean ± SD. ROC, receiver operating characteristic; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns (no significance), by two-tailed Student t test (G–I); two-way ANOVA (B–E).

Because of pMMR tumors having resistance to ICIs, we further used BALB/c mice bearing CT26 colorectal tumors with anti–PD-1 or isotype control IgG. Inconsistent with our previous MC38 results, mice bearing CT26 showed a more limited effect of anti–PD-1 treatment compared with control IgG (Fig. 6D). CT26 tumors with Krt17 overexpression had apparently shrunk, and BALB/c mice bearing tumors had prolonged survival (Fig. 6D and E). These findings were similar with C57BL/6 mice bearing Krt17-overexpressing MC38 colorectal tumors with anti–PD-1 treatment. To confirm that CXCL10 was responsible for the enhanced immunotherapy response with KRT17, neutralizing CXCL10 antibody was used in the Krt17 overexpression model with anti–PD-1 treatment. Mice treated with neutralizing CXCL10 antibody had significantly increased tumor growth compared with those with IgG treatment (Supplementary Fig. S9A–S9C). In summary, these findings showed a generalizable role of Krt17 in immunotherapy responses of colorectal tumors.

Colorectal cancer tumors with high KRT17 expression exhibit a favorable response to anti–PD-1 therapy

To determine whether the above findings were translatable to human patients with patients with colorectal cancer, we retrospectively collected 30 colorectal cancer samples obtained from endoscopic biopsy. These patients had all received anti–PD-1 treatment at our hospital. Before PD-1 treatment, all patients were histologically diagnosed with dMMR colorectal cancer. We first examined KRT17 expression via IHC analysis in tumor tissue. Results showed higher KRT17 expression in tumors from responders compared with those who did not respond to anti–PD-1 therapy (Fig. 6F and G). Similarly, patients with high CXCL10 expression had a better response to pembrolizumab (Supplementary Fig. S9D). Moreover, there was a positive correlation between KRT17 and CXCL10 expressions (Supplementary Fig. S9E).

We next identified whether preexisting CD3⁺ and CD8⁺ T cells represented critical factors in the clinical response to PD-1 therapy. Because TP and invasive tumor margin were difficult to define in our endoscopic biopsy samples, we only counted CD3⁺ and CD8⁺ cells in the entire field of vision. The response group showed more CTL infiltration than the nonresponse group (Fig. 6H and I). We further explored the correlation between KRT17 expression and T-lymphocyte density in the endoscopic biopsy samples, and results showed a positive correlation (Supplementary Fig. S9F–S9H). ROC and AUC values suggested that KRT17 expression had better predictive ability than either CD3⁺ or CD8⁺ density in tumor tissues of endoscopic biopsy (Fig. 6J; Supplementary Fig. S9I and S9J), which might be a promising clinical marker for immunotherapy in colorectal cancer. In summary, KRT17 was highly expressed in immunotherapy responders and had a better predictive performance than CTL density when determining the potential sensitivity to immunotherapy based on endoscopic biopsy samples.

Discussion

Managing the immunosuppressive TME has been identified as a promising therapeutic approach to treat a variety of malignancies. Urgent solutions are needed address limited immunotherapy efficacy in pMMR and partly in dMMR colorectal cancer owing to reduced CTL infiltration (27, 28). The major novel findings of our study provide novel insights on enhancing T-lymphocyte infiltration to reverse immune escape in colorectal cancer (Fig. 6K). We found that dMMR tumors had higher expression of KRT17 relative to pMMR tumors, resulting in increased CTL infiltration. This in turn inhibited malignant progression and sensitized colorectal tumors to immunotherapy. Mechanistically, KRT17 mediated the proteolytic degradation of YTHDF2, which increased Cxcl10 mRNA expression by inhibiting degradation of m⁶A-modified transcripts. CTL infiltration increased with CXCL10 expression. On the basis of our combined in vivo and in vitro observations, the role of the KRT17–YTHDF2–CXCL10 axis was identified in patients with colorectal cancer with pMMR. This finding reinforces the clinical significance of our discovery. The observed connections between KRT17 and CTL infiltrations will also contribute to a better understanding of how KRT17 influences colorectal cancer etiology and may provide additional insights into prospective treatment options for treating pMMR tumors.

The mechanisms underlying the relationship between MMR status and a high CTL microenvironment in colorectal cancer remain largely unknown. The MMR system functions to remove DNA base mismatches, which mainly arise during either DNA replication or DNA damage. Status of dMMR specifically refers to mutations in four genes MSH2, MLH1, PMS2, and/or MSH6 that lose the ability to clear DNA base mismatches. As a result, these DNA base mismatches lead to the production of many mutation-related proteins. Therefore, one important factor may be the increased number of tumor neoantigens generated by high mutational load in dMMR tumors (29, 30). Mutated neoantigens are tumor-specific and may not induce immune tolerance (31). However, there were few studies exploring other mechanisms besides neo-antigenicity. As shown in the previous studies, the concordance rate between RNA-seq and qRT-PCR experiments varied from approximately 16% to 80% (32, 33). In this study, we found that 20% (1/5) genes identified by RNA-seq were validated in a greater number of samples by qRT-PCR. The results appeared to be consistent with previous reports (32, 33). In our studies, we demonstrated a regulatory mechanism for T-lymphocyte homeostasis in colorectal cancer. Specifically, Krt17 overexpression promoted T-lymphocyte migration into the TP and invasive tumor margin by increasing the production of the classical lymphocyte chemokine CXCL10. This may provide new theoretical guidance for understanding the immune regulation of different MMR states.

Our discoveries also offer proof for a tumor-suppressive role of KRT17. However, preceding studies have revealed conflicting findings regarding the influence of KRT17 on tumor malignancy progression. For instance, KRT17 is upregulated in cervical, oral squamous cell carcinoma, and gastric cancer, and KRT17 is a poor prognostic biomarker in these cancers (34–36). In contrast, the protective role of KRT17 on tumor development has been supported by paired high-resolution lineage information with single-cell RNA-seq. These findings were further validated using CRISPR-based perturbations (37). Moreover, it has been reported that KRT17 affects skin tumorigenesis through an immunoregulatory mechanism involving many pathways, such as: (i) changed chemokine expression profile from Th2 polarization to Th1/Th17 signature (38); (ii) interaction with RNA-binding protein hnRNP K, thus regulating the expression of CXCR3 ligands CXCL9, CXCL10, CXCL11, and other chemokine genes (39); (iii) co-localization with a transcriptional regulator AIRE and bound to NF-κB consensus sequence at the promotor regions of pro-inflammatory genes (40). This suggests that the common immunoregulatory mechanism of KRT17 is to regulate the formation of a pro-inflammatory immune microenvironment in skin tumorigenesis. However, KRT17 has been reported to promote immune escape and ICIs resistance in head and neck cancer (41). Our findings demonstrated that KRT17 was highly expressed in dMMR tumors and prompted ubiquitination-mediated proteolytic degradation of YTHDF2. Ultimately, this increased CXCL10 expression and attracted T lymphocytes into the tumor. This disagreement may be caused by the functional differences of KRT17 in different tumor types, and more studies are needed to clarify the functional diversity of KRT17.

YTHDF2 is an m⁶A modification reader that promotes rapid decay of m⁶A-modified transcripts (23). Emerging evidence has demonstrated that YTHDF2 has significant effects on immune regulation and cytokine production (42). For instance, Wu and colleagues show that YTHDF2 increases the decay of KDM6B to promote H3K27me3 methylation of proinflammatory cytokines, which then decrease their expression (43). Depletion of Ythdf2 in tumor cells elevates the transcription of both Stat1 and Irf1 in the presence of IFNγ, thus increasing the induction of ISGs and sensitizing tumors to immunotherapy (25). Our research here also offers evidence for an immune escape role of YTHDF2. We showed that overexpression of Ythdf2 inhibited T-lymphocyte migration by decay of m⁶A-modified transcript of Cxcl10 with IFNγ stimulation, thereby promoting malignant phenotypes in colorectal cancer. The finding that overexpression of Ythdf2 partially attenuated the Krt17 overexpression phenotype suggests that other mechanisms may be in play. KRT17 has been reported to regulate target proteins through multiple pathways, including nucleus localization (40), nuclear export (39), and phosphorylation (44). Krt17 overexpression did not affect the protein level of PSMB8 and SMAD3, so other mechanisms or other molecules downstream require further exploration in the future. Various targets of YTHDF2 have been reported in cancer, including CEBPA, SOCS2, PD-1, CXCR4, and SOX10 (22, 26, 45). We identified CXCL10 as a key target of YTHDF2 that functions in an m⁶A-dependent manner to regulate the lymphocyte migration in colorectal cancer. Overall, these findings mechanistically provide the physiological role of the KRT17–YTHDF2 cascade in driving tumor-infiltrating lymphocyte migration and could provide a new immunotherapeutic strategy in colorectal cancer.

Recent research has revealed an important role for epigenomic reprogramming in tumor cells for tumor evasion. It has been shown that the histone modulator EZH2 modulates the transcription of the chemokines CXCL9 and CXCL10 by regulating DNMT1-mediated DNA methylation to prevent T-cell migration into tumors (46, 47). Depletion of the epigenetic regulator PBAF complex promotes chromatin accessibility to enhance tumor response to T-lymphocyte cytotoxicity (48). Collectively, these studies have shown that tumor cells enable immune escape at the transcriptional level by many epigenetic mechanisms. N6-methyladenosine (m⁶A) RNA methylation is the most common and conserved internal modification in RNA in eukaryotic cells (23, 49). Nevertheless, the RNA epitranscriptome's role in T-lymphocyte migration remains poorly understood. We uncovered a relationship between reversible m⁶A methylation with T-lymphocyte migration through the dynamic process of m⁶A-modified mRNA transcripts. We also revealed that KRT17 induced T-lymphocyte migration by preserving CXCL10 expression in an m⁶A-dependent manner. These data support the concept that CXCL10 is more abundant in dMMR colorectal cancer compared with pMMR colorectal cancer (11).

Despite the well-known breakthrough of immunotherapy, patients with colorectal cancer with pMMR status do not respond effectively to any immunotherapy alone. This is particularly concerning because patients with colorectal cancer with pMMR account for the largest proportion of colorectal cancer (27). The lack of T-lymphocyte infiltration into the tumor seems to be the main cause for this ineffectiveness. Therefore, there is interest in developing rational combinations that attract T lymphocytes into tumors of this type. Some preclinical and clinical studies have shown that combination of immunotherapy with radiotherapy, high-dose IL2, and IL10 have apparent responses in both melanoma and renal cell carcinoma through CD8⁺ T-cell expansion (50, 51). However, this response is dependent on continued high-dose drug therapy. The toxicity of such combinations has been a concern for physicians and has limited the development of combination tumor immunotherapy. Our results showed that KRT17 promoted ubiquitination-mediated degradation of YTHDF2, which decreased the decay of CXCL10 m⁶A transcript and increased T-lymphocyte infiltration in the presence of IFNγ. This suggests that increased T-lymphocyte infiltration leads to more IFNγ secretion into the TME, which in turn leads to increased CXCL10 expression and continues to attract more T lymphocytes. This positive feedback of T-lymphocyte infiltration may be initialized by low-dose and intermittent medication in tumors with high expression KRT17, which would avoid high-dose drug toxicity. Given this, our approach has great potential clinical value.

With the astounding achievement of ICIs therapies in recent years, there remains an urgent need to find effective markers that can distinguish patients who will significantly benefit from ICI therapy. Reduced CTL infiltration represents the most recognized reason. Herein, we found that KRT17 overexpression in tumor cells could enhance immunotherapy response in an immunotherapy-resistant model by the YTHDF2–CXCL10 axis. Therefore, gene-modifying techniques, such as CRISPR–Cas9 combined with targeting guide RNA (52), antibody-delivered short interfering RNA (53), or aptamer–short interfering RNA chimeras (54), can be used to modify KRT17–YTHDF2–CXCL10 axis in tumor cells, although this possibility requires further study. Moreover, although dMMR and microsatellite instability-high are the recognized predictors of ICI response in colorectal cancer (9), but some mutations in pMMR tumors may be also recognized by the immune system. There remain other unknown immune system regulators. Thus, the number of CD8⁺ cells is usually considered an indicator of response (55, 56). Specifically, KRT17 expression was a better predictor of ICI responsiveness than CD3⁺ or CD8⁺ density in endoscopic biopsy samples, which might be a promising clinical marker for immunotherapy in colorectal cancer. Our other major prospective clinical trial, which aims to investigate the effect of T-lymphocyte infiltration on the response to immunotherapy in rectal cancer, is ongoing (ClinicalTrials.gov, Number NCT 05450029). We will validate KRT17 as an immunotherapy biomarker based on this prospective clinical trial. A randomized controlled study targeting KRT17 might then be conducted to determine whether KRT17 could be used as an immunotherapy marker for clinical promotion in colorectal cancer.

Conclusions

We found that KRT17 was highly expressed in dMMR tumors, increased CTL infiltration into tumors, and sensitized responses to ICIs in colorectal cancer. This immune effect was mediated by elevated CXCL10 expression, whose mRNA transcript was stabilized owing to proteolytic degradation of the m⁶A reader YTHDF2. Our research demonstrates the critical role of KRT17 in reversing tumor escape and immunotherapy resistance. KRT17 also offers an opportunity to tackle the clinical problem of reduced immunotherapy response in pMMR tumors.

Supplementary Material

Supplementary Fig. 1

KRT17 is highly expressed in CRC tissues compared to tumor-adjacent tissue regardless of MMR status.

Supplementary Fig. 2

KRT17 has no effect on proliferation, migration and invasion of CRC cells.

Supplementary Fig. 3-4

KRT17 expression is positively correlated with CD3+ and CD8+ densities in tumor parenchyma and invasive margin in CRC.

Supplementary Fig. 5-7

YTHDF2 expression is negatively correlated with CD3+ and CD8+ densities in tumor parenchyma and invasive margin in CRC.

Supplementary Fig. 8

KRT17 exerts an immunomodulatory effect through YTHDF2-CXCL10 axis.

Supplementary Fig. 9

CXCL10 is indispensable for KRT7-mediated improved responsiveness to ICIs.

Supplementary Tables

Materials were used in the research.

Supplementary Legends

Detailed descriptions of supplementary figures.

Authors' Disclosures

No disclosures were reported.

Authors' Contributions

W. Liang: Validation, investigation, writing–original draft. H. Liu: Investigation, methodology. Z. Zeng: Investigation, visualization. Z. Liang: Formal analysis. H. Xie: Formal analysis. W. Li: Data curation. L. Xiong: Software. Z. Liu: Software. M. Chen: Data curation. H. Jie: Formal analysis. X. Zheng: Project administration, writing–review and editing. L. Huang: Resources, funding acquisition. L. Kang: Conceptualization, resources, funding acquisition, writing–review and editing.