DNA methylation regulator-based modification patterns define distinct immune phenotypes and immunotherapy responses in colorectal cancer

doi:10.21203/rs.3.rs-1819659/v1

Download PDF

Research Article

DNA methylation regulator-based modification patterns define distinct immune phenotypes and immunotherapy responses in colorectal cancer

https://doi.org/10.21203/rs.3.rs-1819659/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background: A major challenge in immunotherapy is identifying populations that can benefit from treatment. DNA methylation is one of the most extensively studied epigenetic regulation mechanisms and reportedly involved in immune regulation and anticancer immune responses. How the integrated DNA methylation modification pattern affects the tumor microenvironment (TME) and the response to immunotherapy is poorly understood, however.

Methods: In this study, samples from 1,691 colorectal cancer (CRC) patients recruited from five independent cohorts were used to comprehensively evaluate DNA methylation regulator patterns and TME characteristics. DNA methylation regulator patterns were determined using unsupervised hierarchical clustering analysis based on 20 DNA methylation regulators. The CIBERSORT and single-sample gene set enrichment analysis algorithms were used to respectively assess tumor-infiltrating immune cell relative abundance and immune-related pathways in each sample. A principal component analysis algorithm–based DNA methylation scoring scheme (DM_Score) was devised to quantify DNA methylation levels in candidates. The predictive value of the DM_Score for immunotherapy was evaluated in two cohorts. The association between the DM_Score and TME was assessed in an external cohort from our center.

Results: Three DNA methylation regulator patterns associated with various clinical outcomes and TME characteristics were identified, consistent with classical immune phenotypes (immune-desert, immune-inflamed, and immune-excluded). The DM_Score established based on 270 DNA methylation-related signature genes strongly predicted unfavorable CRC prognoses associated with various mutations and TME landscapes. In addition, the DM_Score was positively correlated with many recognized biomarkers of immunotherapy response, including tumor mutational burden, tumor immune dysfunction and exclusion, immunophenoscore, PD-L1 mRNA, and microsatellite instability. In the two immunotherapy cohorts, a significant survival advantage was observed in patients with a high DM_Score after anti-PD-1/PD-L1 therapy.

Conclusions: We present the first comprehensive analysis of cross-talk between DNA methylation modification patterns and TME characteristics in CRC. Quantitative evaluation of DNA methylation levels in individual tumors can aid in guiding more-precise immunotherapy strategies.

Colorectal cancer

DNA methylation

Immunotherapy

Tumor microenvironment

Immune phenotype

DNA methylation is one of the first discovered and most-studied mechanisms of epigenetic regulation (1). 5-Methylcytosine is generated by the addition of a methyl group to the carbon-5 position of the cytosine-phosphate-guanine dinucleotide in a reaction catalyzed by DNA methyltransferase (DNMT), and this is the primary form of DNA methylation in mammals (2). DNA methylation is essential for maintaining the stability of genetic material and regulating gene expression patterns. Previous studies confirmed that aberrant DNA methylation is involved in aging, developmental disorders, and various cancers (3, 4).

Colorectal cancer (CRC) is the third most common carcinoma worldwide and the second leading cause of cancer deaths (6). Highly efficacious individualized therapeutic strategies for CRC are still lacking, despite recent advancements in multimodal treatments (5). The development of CRC involves the stepwise accumulation of a series of genetic and epigenetic malignant transformations, such as hyper-methylation of promoters of specific genes, including CDKN2A, MLH1, RUNX3, MGMT, and APC (7, 8). Promoter hyper-methylation of these genes leads to dysregulation of DNA repair, cell cycle progression, and signaling pathway checkpoints (9). Furthermore, DNA methylation can defined the function of both tumor-reactive and bystander CD8⁺ tumor-infiltrating lymphocytes, which are the main antitumor immune cells (10).

Although immunotherapy has successfully shifted the paradigm for CRC treatment, not all patients benefit from immunotherapies such as anti-PD-L1 antibody treatment, which is of greatest benefit to CRC patients with high levels of microsatellite instability (MSI) and high tumor mutational burden (TMB) (11). The effectiveness of immunotherapy is also impacted by signal crosstalk between cellular components of the tumor microenvironment (TME) (12, 13). Sub-dividing TME of CRC into three classical immune phenotypes (immune-inflamed, excluded, and desert) by TME composition reveals distinct immune features and treatment responses (14, 15). Mounting evidence has confirmed that aberrant DNA methylation in the promoter region of specific genes leads to immune evasion and increased immunotherapy resistance as the TME changes (16–19). Goltz et al. found that PD-L1 methylation is negatively associated with PD-L1 mRNA expression and worse clinical outcomes in CRC patients (20). Jung et al. demonstrated that altered DNA methylation serves as an independent predictive marker of increased immunotherapeutic resistance (19). A recent study found that patients undergoing treatment with Durvalumab in combination with DNA methylation inhibitors experienced significantly enhanced treatment efficacy (17, 21). However, the global effect of DNA methylation on the TME as well as the impact on immunotherapy response remains unclear. Hence, a comprehensive investigation of how multiple DNA methylation regulators affect the TME and pharmacogenomic landscape could aid the development of precision immunotherapies.

In this study, we integrated 1,691 CRC patients from five independent cohorts to comprehensively evaluate their DNA methylation regulator patterns. Our analysis revealed three DNA methylation regulator patterns with various TME characteristics, consistent with classical immune phenotypes. Furthermore, a DNA methylation scoring system (DM_Score) based on DNA methylation-related signature genes was devised to quantify the level of DNA methylation in individual patients. Importantly, our results demonstrate that the DM_Score is a reliable biomarker for predicting the response to immunotherapy in CRC patients.

Data Collection and Processing

Cohort 1: TCGA-CRC cohort

A flowchart of the study is shown in Figure S1A. Data regarding the RNA transcriptome, somatic mutations, copy number variation (CNV), and corresponding clinical information were obtained from The Cancer Genome Atlas (TCGA) database (https://portal.gdc.cancer.gov/). The RNA sequencing data from TCGA (fragments per kilobase of exon model per Million (FPKM) format) were transformed into transcripts per kilobase million (TPM) format.

Cohort 2: meta-GEO cohort

Normalized matrix files and paired clinical data were directly downloaded from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo). Four GEO cohorts with available survival and tumor stage data (GSE103479, GSE17538, GSE39582, and GSE72970) were merged into the meta-GEO cohort. Finally, a total of 1,619 patients from four GEO CRC cohorts (the meta-GEO cohort) and TCGA-CRC cohort were enrolled for further analysis. Basic information regarding these cohorts is summarized in Table S1. Batch effects in different RNA sequencing array data were corrected using the ComBat algorithm of the R package sva (Figures S1B and C) (22).

Cohort 3: Immunotherapey cohort

To evaluate the potential therapeutic value of the DM_Score, two immunotherapy cohorts (GSE78220 and IMvigor210) were recruited for the study (15, 23). The raw count data of pre-therapy biopsy samples were normalized and converted to the TPM format for further analysis.

Cohort 4: Wuhan cohort

Eight fresh CRC tissues were collected from the Gastrointestinal Surgery Department of Renmin Hospital of Wuhan University during March and May 2021, followed by a high-throughput transcriptome sequencing using an Illumina HiSeq sequencing platform. This study was approved by the Medical Ethics Committee of the Renmin Hospital of Wuhan University (WDRY2021-K126), and all patients have signed a prior informed consent.

RNA-Sequencing

Total RNA was extracted from all samples using Takara RNAiso Plus (TaKaRa BIO, Dalian, China) and quality inspected using a Nano-300 micro-dlight photometer (ALLSHENG, Hangzhou, China) and Agilent 2100 Bioanalyzer (Applied Biosystems, Carlsbad, CA, USA) according to the manufacturers' instructions. The library was constructed using KAPA RNA Hyper Prep Kit (Kapa Biosystems, Inc., Woburn, MA, USA), and sequencing was performed using a Novaseq 6000 platform (Illumina, San Diego., CA, USA) with a sequencing data volume of 10G. The reads were first mapped to the latest UCSC transcript set using Bowtie 2 ( Version 2.1.0), and gene expression levels were estimated using RSEM (http://deweylab.github.io/RSEM/, version 1.2.15) (24). Trimmed mean M-values were used to normalize gene expression data. Eight CRC patients were divided into high-DM_Score and low-DM_Score groups according to the median DM_Score.

Immunohistochemistry (IHC)

Formalin-fixed, paraffin-embedded specimens of eight CRC tissues from our cohort were prepared. Immunohistochemical staining and scoring of the CRC tissues was carried out using antibodies against PD-L1 (1:1000; Servicebio, GB11339A), CD8A (1:200; Servicebio, GB11068-1), and TGFB1 (1:500; Servicebio, GB111876) were performed, as previously described procedures (25). The expression levels of PD-L1, CD8A, and TGFB1 protein were evaluated according to the total histochemical score, which was based on both the staining intensity and proportion of positive cells.

Unsupervised Clustering of 20 DNA Methylation Regulators

The K-means clustering algorithm was used to identify different DNA methylation modification patterns using the R package ConsensusClusterPlus (26). The clustering was run 1,000 times to ensure classification stability. DNA methylation regulators were identified from published literature (27). In total, 20 regulators were identified as participating in the DNA methylation process, including 3 writers, 3 erasers, and 14 readers (Table S2).

Gene Set Variation Analysis (GSVA)

GSVA enrichment analysis was conducted to characterize the biological processes involved in different DNA methylation patterns using the R package GSVA (28). The gene set “c2.cp.kegg.v7.1” for GSVA analysis was obtained from the Molecular Signature Database (https://www.gsea-msigdb.org/gsea/msigdb/index.jsp).

Calculation of the Abundance of Tumor-Infiltrating Immune Cells (TIICs)

The relative abundance of tumor-infiltrating immune cells (TIICs) in each sample was quantified by the single-sample gene set enrichment analysis algorithm. Immune gene sets defining each TIIC were obtained from a previous study (29).

Immune-Related Gene Signatures

The immune-related gene signatures used in this study, including those associated with angiogenesis, antigen processing machinery, cytolytic activity, epithelial-mesenchymal transition (EMT), immune checkpoints, the major histocompatibility complex or human leukocyte antigen, the pan-fibroblast TGF- response signature, the interferon response, and WNT targets, were obtained from previous research (Table S3). The TMB score was generated by dividing the total number of mutations by the number of exons in each sample. MSI and immunophenoscore (IPS) data were downloaded from The Cancer Immunome Database (https://tcia.at/home). The Tumour Immune Dysfunction and Exclusion (TIDE) score, a computational method to model tumor immune evasion mechanisms, was obtained using the online tool (http://tide.dfci.harvard.edu/) (30).

Identification of Differentially Expressed Genes (DEGs)

Based on the expression of the 20 DNA methylation regulators, three distinct DNA methylation modification patterns were identified. DEGs among distinct phenotypes were calculated using the R package limma (31). Overlapping DEGs were identified based on a Venn graph drawn using the R package VennDiagram (32). The significance threshold for DEGs was defined as an adjusted P-value of less than 0.001.

Generation of the DM_Score

To quantify the DNA methylation modification levels in individual CRC patients, a DNA methylation scoring system based on the specialized genes was constructed, namely the DM_Score, which was calculated as follows. First, prognosis-related overlapping DEGs were extracted using univariate Cox regression analysis. Second, principal component analysis (PCA) was employed for dimension reduction of the prognostic DEGs, and both principal component 1 (PC1) and 2 (PC2) were chosen as the signature scores. Finally, the DM_Score was defined as the sum of PC1 and PC2. The formula for the DM_Score calculation is shown below, where i denotes the expression of DNA methylation phenotype–related genes.

Statistical Analysis

All statistical analyses and drawings in this study were carried out using R software (version 4.0.2, https://cran.r-project.org/src/base/R-4/). All statistical analyses were two-sided, and P<0.05 was set as the significance threshold unless otherwise indicated. Data for three or more groups were compared using the Kruskal-Wallis or chi-square test. Survival curves were drawn using Kaplan-Meier survival analysis, and the Log-Rank test was used to evaluate the significance of differences. The optimal cutoff values of DNMT3B and the DM_Score were determined using the surv_cutpoint function of the R package survminer (https://CRAN.R-project.org/package=survminer) in the meta-GEO and TCGA-CRC cohorts.

The Landscape of Genetic and Transcriptional Alterations of DNA Methylation Regulators in CRC

As shown in Figure 1A, a total of 20 DNA methylation regulators, including 3 writers, 3 erasers, and 14 readers, were identified in this study. First, we summarized the incidence of somatic mutations and CNV in these 20 regulators in CRC. Overall, maftool analysis showed a low somatic mutation rate among the DNA methylation regulators. TET methylcytosine dioxygenases 1 and 3 exhibited the highest mutation rates (both 6%), whereas MBD3 exhibited the lowest mutation rate (0%) among the 20 regulators (Figure 1B). CNV was a ubiquitous phenomenon among the DNA methylation regulators in CRC, with most exhibiting CNV deletions, whereas DNMT3B exhibited an increased frequency of CNV amplifications. The CNV profiles of the 20 DNA methylation regulators are presented in Figure 1C and D. Significant synergy between mRNA expression and CNV alterations was observed among the DNA methylation regulators between normal and CRC samples. Compared with normal tissues, most of the DNA methylation regulators (except ZBTB38) with amplified CNV were expressed at markedly higher levels in CRC tissues (Figure 1E). Furthermore, the expression of all but one of the DNA methylation regulators (except MBD3) differed significantly among the distinct consensus molecular subgroups (CMSs) (Figure 1F). This analysis revealed that the expression of and epigenetic imbalance among DNA methylation regulators plays a pivotal role in the tumorigenesis and progression of CRC.

Prognosis and Immune Landscape of DNA Methylation Regulators in CRC

Kaplan-Meier survival curves demonstrated that 11 of the 20 identified regulators were associated with overall survival (OS) in the meta-GEO cohort (Figure S2A-K). Figure 2A summarizes the interactive relationship and prognostic significance of these 20 DNA methylation regulators in CRC. Our analyses revealed a significant cross-talk network existing not only between regulators of the same functional subtype but also between writers, erasers, and readers. In particular, the strong correlation observed between UHRF1 (reader) and DNMT1 (writer) indicated functional cooperation between these regulators.

We also investigated the relationship between DNA methylation regulators and the immune TME in CRC. Correlation analyses showed that the expression of many regulators (particularly DNMT3B) was significantly negatively correlated with the relative abundance of TIICs (Figure 2B). We therefore conducted a more detailed examination of the role of DNMT3B in immune-related processes, MSI status, and immunotherapy response.

Gene Set Enrichment Analysis results revealed that immune-related processes, including the T-cell receptor signaling pathways, were significantly enriched in the high-DNMT3B expression group (Figure 2C). A significant difference in OS between the high- and low-DNMT3B expression groups was observed in the meta-GEO cohort (P = 0.031, Figure 2D). Interestingly, we found that DNMT3B mRNA expression correlated with the mean methylation of promoter sites positively, not negatively in the traditional sense (Figure 2E). Furthermore, DNMT3B mRNA expression was significantly downregulated in the MSI-high (MSI-H) group (Figure 2F). DNMT3B mRNA expression also differed significantly among distinct CMSs, suggesting a role in promoting CRC heterogeneity (Figure 2G). In the IMvigor210 cohort, a greater proportion of CRC patients with high DNMT3B expression exhibited complete response (CR) or partial response (PR) than patients with low DNMT3B expression (31% vs. 17%, respectively, Figure 2H). Patients in the high-DNMT3B expression group exhibited a significant survival advantage after treatment with the PD-L1 inhibitor Atezolizumab (P = 0.034, Figure 2I). Overall, these results indicate a close relationship exists between the DNA methylation regulators, the TME, and immunotherapy response.

DNA Methylation Modification Patterns Mediated by the 20 Regulators in CRC

The above studies confirmed that DNA methylation is closely related to the prognosis and immune microenvironment of CRC patients. To build on these findings, we attempted to identify distinct DNA methylation patterns using consensus clustering based on the expression of those 20 regulators. Accordingly, three distinct modification pattern clusters, namely Cluster A (n = 481), Cluster B (n = 848), and Cluster C (n = 362), were identified in the meta-GEO and TCGA cohorts. The PCA plot of the 20 DNA methylation regulators revealed significant separation between the three pattern clusters (Figure 3A). Kaplan-Meier survival curves demonstrated a significant survival advantage for patients in Cluster B, whereas those in Cluster C exhibited a clearly poorer outcome in the meta-GEO cohort (P = 0.027, Figure 3B). Figure 3C shows a heatmap of the three DNA methylation modification pattern clusters annotated according to clinical and pathologic parameters.

GSVA was performed to explore how the three DNA methylation modification patterns affect biological behavior. Cluster B was markedly enriched in immune-related biological processes, such as the complement and coagulation cascades and cytokine-cytokine receptor interactions, whereas carcinogenic activation and stromal-related pathways, such as adherens junction and the MAPK signaling pathway, were significantly enriched in Cluster C (Figures 3D-F). We therefore further evaluated the relationship between DNA methylation modification patterns and immune phenotypes. Interestingly, despite increased infiltration of immune cells, particularly innate immune cells (e.g., eosinophils, myeloid-derived suppressor cells, macrophages, mast cells, monocytes, natural killer cells, and neutrophils), CRC patients in Cluster C did not exhibit a significant survival advantage (Figure 3G). A previous study showed that stromal activation in the TME can inhibit antitumor immune effectors (33). Single-sample gene set enrichment analysis confirmed that stromal-related processes, such as EMT1, EMT2, and the pan-fibroblast TGF- response signature, were significantly activated in Cluster C (Figure 3H). We also found that CD8⁺ T-effector cells and cytolytic activity were significantly enhanced in Cluster B, which was associated with higher antitumor activity and better prognosis. Collectively, these findings indicate that the three DNA methylation modification patterns are significantly associated with distinct clinical outcomes and immune landscapes. Cluster A represents an immune-desert phenotype characterized by low TIICs, whereas Cluster B represents an immune-inflamed phenotype marked by higher antitumor activity, and Cluster C can be considered an immune-excluded phenotype featuring stromal activation.

Construction of the DNA Methylation-Related Gene Clusters

Given the heterogeneity of DNA methylation modification, DEGs within the three clusters were identified to investigate the molecular behavior of each modification pattern in depth. A total of 1,183 DNA methylation modification pattern–related genes were identified within the three modification patterns (Figure 4A). Enrichment analysis demonstrated that these genes were primarily associated with RNA modification- and DNA methylation-related events (Figure 4B and C).

Univariate Cox regression analyses identified 270 genes with prognostic value. Based on the expression levels of these 270 genes, an unsupervised clustering analysis was conducted and identified three stable genomic subtypes in the meta-GEO and TCGA cohorts: gene.Cluster A (n = 591), gene.Cluster B (n = 845), and gene.Cluster C (n = 255). Survival analysis of the meta-GEO cohort revealed that gene.Cluster A and gene.Cluster B were associated with better prognosis, whereas gene.Cluster C was associated with the worst clinical outcome (Figure 4D). This clustering result was consistent with the DNA methylation modification patterns (Figure 4E).

In addition to MBD4, significant differences in expression between the three clusters were observed among all DNA methylation regulators (Figure 4F). Similar to the Cluster C modification pattern (immune-excluded phenotype), gene.Cluster C was associated with a wide range of both immune and stromal activation phenotypes (Figure 4G and H). These results further demonstrated that CRC is associated with three distinct DNA methylation modification patterns, which are represented by distinct clinical outcomes and immune landscapes.

Development of the DM_Score

Considering the individual complexity and heterogeneity of the DNA methylation modifications, we developed a scoring system to quantify the DNA methylation level of each CRC patient (i.e., the DM_Score). The CRC patients were divided into high- and low-score subgroups with optimal cutoff values of 0.915 and 1.985 in the meta-GEO and TCGA-CRC cohorts, respectively. Patients with a low DM_Score exhibited longer survival times than those in the high-DM_Score group, consistent with the meta-GEO (Figure 5A, P < 0.001) and TCGA-CRC (Figure 5B, P = 0.019) cohorts.

A Sankey diagram was generated to visualize the attribute changes in the DNA methylation modification patterns, gene clusters, DM_Score, and survival status for individual CRC patients (Figure 5C). Patients in Cluster C exhibited the highest median DM_Score, whereas those in Cluster A exhibited the lowest median DM_Score (Figure 5D, P < 2.2 × 10^-16). Similarly, gene.Cluster C was associated with a significantly higher DM_Score than gene.Cluster A and gene.Cluster B (Figure 5E, P < 2.2 × 10^-16). DM_Score was significantly associated with AJCC stage and survival status in the meta-GEO cohort (P < 0.01, Figure 5F and G). Becht et al. suggested that CMS2 represents immunologically “cold” tumors lacking significant lymphoid/myeloid infiltration and major histocompatibility complex expression (34). Among the four CMSs in CRC, CMS1 exhibited the highest DM_Score, whereas CMS2 exhibited the lowest DM_Score in TCGA-CRC cohort (Figure 5H).

To characterize the impact of DM_Score on the TME, we also investigated the correlation between DM_Score and TIICs. A heatmap was generated and showed that DM_Score was positively associated with vast majority of TIICs and stroma-related signatures in the meta-GEO cohort (Figure 5I). It is thus reasonable to conclude that a higher DM_Score represents a higher immune-activating phenotype, whereas a low DM_Score can be hypothesized to indicate an immune-desert phenotype.

Prognostic Value of the DM_Score

Univariate and multivariate Cox regression analyses confirmed that the DM_Score provided robust and independent prognostic value for predicting clinical outcomes in both the meta-GEO (Figure 5J and K) and TCGA-CRC (Figure 5L and M) cohorts. We then assessed the prognostic value of the DM_Score in terms of OS in CRC. Subgroup analyses revealed that the risk signature remained robust as an independent prognostic factor in terms of age (age ≥ 65 and < 65 years, Figure S3A and 4B), gender (male and female, Figure S3C and 3D), and AJCC stage (III + IV, Figure S3E). The OS of low-DM_Score patients was markedly superior to that of high-DM_Score patients in all clinicopathologic subgroups. These results further confirmed the utility of the DM_Score as a robust prognostic predictor in CRC.

Mutational Profile of TCGA-CRC Cohort DM_Score Groups

Accumulating evidence suggests that a close relationship exists between immunotherapeutic effect and somatic mutations (35, 36). We therefore contrasted the mutational landscapes between TCGA-CRC cohort high- and low-DM_Score groups using the R package maftools (Figure 6A and B) (37). We then analyzed the differences in mutation frequencies in common cancer driver genes between the high- and low-DM_Score groups. BRAF (20% vs. 2% [high vs. low DM_Score, respectively], P < 0.001, Figure 6C) and SMAD4 (18% vs. 8%, P = 0.003, Figure 6D) exhibited higher somatic mutation rates in the high-DM_Score group, whereas TP53 (50% vs. 77%, P < 0.001, Figure 6E) and APC (66% vs. 94%, P < 0.001, Figure 6F) exhibited a higher frequency of mutations in the low-DM_Score group. Pearson’s correlation analysis revealed that DM_Score was positively correlated with TMB (R=0.22, P = 8.7 ×10^-7, Figure 6G), thus indirectly suggesting that patients with a high DM_Score would experience a durable clinical response to immunotherapy. These results provide more details about the relationship between DM_Score classification and genome somatic mutations and reveal the complexity of interactions between DNA methylation modification and individual somatic mutations.

Relationship between DM_Score grouping and classical immune and molecular subtypes

To further validate the relationship between DM_Score and the TME, we examined the protein expression levels of PD-L1, CD8A, and TGFB1 between high- and low-DM_Score patients in the Wuhan cohort. As shown in Figure 7A, compared with the low-DM_Score group, the high-DM_Score group exhibited higher PD-L1 and CD8A expression but lower TGFB1 expression.

The CMS classification describes the four molecular subgroups of CRC according to genetic and epigenetic perspectives: CMS1 (MSI Immune), CMS2 (Canonical), CMS3 (Metabolic), and CMS4 (Mesenchymal) (38). Figure 7B and C shows that the CMS2 subgroup was predominant in the low-DM_Score group and CMS4 in the high-DM_Score group in both the GSE39582 and TCGA-CRC cohorts (P < 0.001, χ² test). CMS2 and CMS3 are reportedly correlated with reduced immune infiltration and immunoreactivity in CRC (39). The sum of the percentage of CMS2 and CMS3 was >80% in the low-DM_Score group, which again confirmed that CRC patients with a low-DM_Score can be classified as bearing immunologically “non-inflamed” tumors.

Thorsson et al. developed an immune subtype classification of 33 non-hematological cancer types using an integrated immunogenomics method (40). They identified and characterized six immune subtypes with potential guiding value for cancer management, particularly immunotherapy. Only four immune subtypes were identified in TCGA-CRC cohort, predominated by C1 (wound healing), which was almost evenly distributed among the two risk groups (Figure 7D). It was found that C2 was more highly infiltrated with activated CD8⁺, follicular helper T cells, regulatory T cells, and neutrophils, representing a “hot” TME immunological status (41). The proportions of C2 in the GSE39582 cohort were 28% and 11% in the high- and low-DM_Score subgroups, respectively (Figure 7E).

Role of the DM_Score in Predicting the Immunotherapy Response

Next, we investigated the suitability of the DM_Score as a biomarker to predict the efficacy of immunotherapies that have recently shown breakthroughs in cancer treatment, particularly anti-PD-1/PD-L1 therapy (42, 43). To date, TMB, MSI, and PD-L1/PD-1 mRNA have been the most widely used predictors of immunotherapy outcomes (44). Patients in the high-DM_Score group exhibited higher TMB in TCGA-CRC cohort, suggesting that they might benefit more from immunotherapy (P = 2 × 10^-5, Figure 8A). IPS, a new machine learning-based marker, reportedly offers superior predictive capability for response to immune checkpoint blockade therapy in solid tumors (45). As we expected, the IPS was significantly higher in the high-DM_Score group, suggesting these patients might obtain greater benefit from anti-CTLA-4 and anti-PD-1 combination therapy (P = 0.011, Figure 8B).

The TIDE score was higher in the high-DM_Score group compared with the low-DM_Score group (P = 1 × 10^-8, Figure 8C). Similarly, patients in the high-DM_Score group exhibited higher levels of PD-L1 mRNA both in TCGA-CRC (P = 8.3 × 10^-16, Figure 8D) and meta-GEO (P = 6.8 × 10^-9, Figure 8E) cohorts. In addition, CRC patients with MSI-H/deficient mismatch repair (dMMR) always exhibited a significantly higher DM_Score compared with CRC patients with microsatellite stable (MSS), MSI-low (MSI-L), or proficient mismatch repair (pMMR) status in both TCGA-CRC (P < 2.2 × 10^-16, Figure 8F) and GEO GSE39582 (P = 2 × 10^-8, Figure 8G) cohorts.

In the GSE78220 cohort, patients with a high DM_Score exhibited markedly prolonged survival after anti-PD-1 therapy (P < 0.001, Figure 8H). Meanwhile, all patients in the low-DM_Score group exhibited sustained disease progression after Pembrolizumab therapy, whereas 68% of patients in the high-DM_Score group achieved CR or PR (Figure 8I). In the IMvigor210 cohort, we calculated the DM_Score of each patient treated with the PD-L1 inhibitor Atezolizumab. Similar to the performance in the GSE78220 cohort, we verified longer survival expectation in the high-DM_Score subgroup (P = 0.018, Figure 8J). CRC patients achieving CR or PR exhibited a higher DM_Score than patients with stable disease or progressive disease (P = 0.0014, Figure 8K). As illustrated in a stacked bar plot, a higher percentage of CR or PR was observed in the high-DM_Score group than the low-DM_Score group (35% vs. 20%, Figure 8L). Taken together, the results from these analyses strongly suggest that DNA methylation modification pattern is correlated with immunotherapy response and that the DM_Score is useful for selecting patients who might benefit from anti-PD-L1/PD-1 immunotherapy.

Increasing evidence indicates that DNA methylation plays a prominent role in immunological processes by regulating immune recognition, immune responses, and antitumor activity (46–48). Studies of DNA methylation regulator patterns conducted to date have primarily focused on single regulators or TME cell type. However, more comprehensive profiling of DNA methylation modification patterns could enhance understanding of the interaction between DNA methylation and the TME and contribute to the development of more effective immunotherapy strategies.

The present study examined the expression patterns of 20 DNA methylation regulators and revealed three distinct patterns associated with diverse TME characteristics representing three classical immune subtypes. Patients in Cluster B exhibited an immune-inflamed phenotype and had the best prognosis, whereas patients in Cluster A and Cluster C exhibited a non-inflamed phenotype and the poorest prognosis. Cluster A exhibited the lowest levels of antigen presentation, stimulatory immunomodulator expression, and infiltrating immune cells. In addition, Cluster A patients exhibited the highest level of WNT signaling, which is indicative of a “cold” immune microenvironment. Cluster B was characterized by the activation of antigen processing machinery and a high level of CD8⁺ T-effector cell infiltration, indicating a “hot” immune microenvironment.

Consistent with the high level of immune cell infiltration and activation of immune-related pathways in Cluster C, we also observed significant activation of angiogenesis, EMT, and TGF-β pathways, all of which are indicative of stromal activation (49, 50). Previous studies demonstrated that cancer patients with inflammatory phenotypes exhibit a better prognosis, but we found that patients in Cluster C and gene.Cluster C consistently had the worst prognosis, despite high levels of immune cell infiltration (51, 52). This contradiction might be explained by the presence of desmoplastic stromal barriers in the TME, which act to suppress antitumor immune responses (51, 52). In the immune-excluded phenotype, CD8⁺ T cells densely colonize the fibroblast- and collagen-rich peritumoral stromal zones instead of the tumor parenchyma; this stromal barrier limits the response to antitumor immunity (54, 55). Considering the important regulatory role of stromal activation in tumor immunosuppression, traditional treatment strategies combined with specific inhibitors targeting TGF-β represent a promising approach for reshaping the TME and restoring antitumor efficacy (56, 57). Galunisertib, a specific molecular inhibitor of TGF-β receptor I kinase, was approved for use in combination with Nivolumab (a PD-1 inhibitor) in clinical trials to treat multiple malignancies (NCT02734160, NCT02423343) (53, 58). In brief, the results of these trials confirmed that DNA methylation modification patterns play an important role in shaping and modifying both the immune and stromal TME phenotypes, suggesting that DNA methylation affects the response to immunotherapy. Based on the results of the above analyses, we speculate that Cluster C or gene.Cluster C CRC patients would benefit more from combination treatment consisting of TGF-β inhibitors and immunotherapy.

We developed the DM_Score system to quantify DNA methylation modification profiles for individual patients as a means of aiding the development of approaches to diminish the DNA methylation heterogeneity in CRC. DNA methylation modification patterns associated with the immune-excluded phenotype exhibited higher DM_Scores, whereas the immune-desert phenotype was associated with lower DM_Scores. This inference was further validated in an external cohort from our center. Further analyses indicated that the DM_Score is an independent biomarker of unfavorable prognosis in CRC. As the DM_Score was positively correlated with recognized biomarkers of the response to immunotherapy, including TMB, PD-L1 mRNA, IPS, and MSI, it could be utilized to identify immunotherapy responders. By examining two immunotherapeutic cohorts, we verified that patients with high DM_Score are more likely to benefit from immunotherapy than conventional postoperative adjuvant therapy, even though they have worse survival. However, this study was limited due to the application of traditional transcriptome sequencing data, which prevented comprehensive assessment of the intratumoral heterogeneity in patients treated with a variety of modalities.

Conclusively, our study is the first to comprehensively characterize the DNA methylation modification patterns of 20 DNA methylation regulators and systematically evaluate their corresponding TME characteristics based on analyses of CRC samples. We also developed an individual scoring system (DM_Score) to quantify the DNA methylation level in CRC patients, which can facilitate improvements in personalized immunotherapeutic strategies. Our analyses identified three different CRC immune subtypes based on DNA methylation regulator patterns.

Ethics approval and consent to participate

This study was approved by the Medical Ethics Committee of the Renmin Hospital of Wuhan University, and all patients have signed a prior informed consent (WDRY2021-K126).

All methods were performed in accordance with the relevant guidelines and regulations in the ethics approval and consent to participate in declarations

Consent for publication

Not applicable.

Availability of Data and Material

All data can be obtained from TCGA-CRC cohort (https://portal.gdc.cancer. gov/repository) and GEO cohort (https://www.ncbi.nlm.nih.gov/geo/). The RNA transcriptome data and corresponding clinical information of the IMvigor210 cohort was downloaded from the website http://research-pub.gene.com/ IMvigor210CoreBiologies/.

The GEO accession numbers for the high-throughput sequencing reported from our cohort is GSE174519.

Competing interests

The authors declare that they have no competing interests.

Funding

The study was supported by the grants of the Wu Jieping Medical Foundation (320.2710.1855) and Chenxiao-ping foundation for the development of science and technology of Hubei province (CXPJJH121003-2103).

Authors' contributions

YBZ and CY designed this study; CY, HC, CHW and FYC performed the data analyses and drafted the manuscript. CY HC and CHW carried out data management and constructive discussions. All authors approved the final manuscript.

Acknowledgment

We acknowledge TCGA and GEO database and contributors for uploading

their meaningful datasets.

Author details

¹Department of Gastrointestinal Surgery, Renmin Hospital of Wuhan University, No. 99 ZhangZhiDong Street Wuchang District, Wuhan, Hubei, P. R. China.

²Key Laboratory of Hubei Province for Digestive System Disease, Renmin Hospital of Wuhan University, Wuhan, China.

Zhang J, Yang C, Wu C, Cui W, Wang L. DNA Methyltransferases in Cancer: Biology, Paradox, Aberrations, and Targeted Therapy. Cancers (Basel) (2020) 12(8):2123. doi:10.3390/cancers12082123
Skvortsova K, Stirzaker C, Taberlay P. The DNA methylation landscape in cancer. ESSAYS BIOCHEM (2019) 63(6): 797-811. doi:10.1042/EBC20190037
Michalak EM, Burr ML, Bannister AJ, Dawson MA. The roles of DNA, RNA and histone methylation in ageing and cancer. Nat Rev Mol Cell Biol (2019) 20(10): 573-589. doi:10.1038/s41580-019-0143-1
Dor Y, Cedar H. Principles of DNA methylation and their implications for biology and medicine. LANCET (2018) 392(10149): 777-786. doi:10.1016/S0140-6736(18)31268-6
Bray F, Ferlay J, Soerjomataram I, Siegel RL, Torre LA, Jemal A. Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin (2018) 68(6): 394-424. doi:10.3322/caac.21492
van Rooijen S, Carli F, Dalton S, Thomas G, Bojesen R, Le Guen M et al. Multimodal prehabilitation in colorectal cancer patients to improve functional capacity and reduce postoperative complications: the first international randomized controlled trial for multimodal prehabilitation. BMC CANCER (2019) 19(1): 98. doi:10.1186/s12885-018-5232-6
Tse J, Jenkins LJ, Chionh F, Mariadason JM. Aberrant DNA Methylation in Colorectal Cancer: What Should We Target? Trends Cancer (2017) 3(10): 698-712. doi:10.1016/j.trecan.2017.08.003
Jung G, Hernandez-Illan E, Moreira L, Balaguer F, Goel A. Epigenetics of colorectal cancer: biomarker and therapeutic potential. Nat Rev Gastroenterol Hepatol (2020) 17(2): 111-130. doi:10.1038/s41575-019-0230-y
Weisenberger DJ, Liang G, Lenz HJ. DNA methylation aberrancies delineate clinically distinct subsets of colorectal cancer and provide novel targets for epigenetic therapies. ONCOGENE (2018) 37(5): 566-577. doi:10.1038/onc.2017.374
Yang R, Cheng S, Luo N, Gao R, Yu K, Kang B et al. Distinct epigenetic features of tumor-reactive CD8+ T cells in colorectal cancer patients revealed by genome-wide DNA methylation analysis. GENOME BIOL (2019) 21(1): 2. doi:10.1186/s13059-019-1921-y
Ganesh K, Stadler ZK, Cercek A, Mendelsohn RB, Shia J, Segal NH et al. Immunotherapy in colorectal cancer: rationale, challenges and potential. Nat Rev Gastroenterol Hepatol (2019) 16(6): 361-375. doi:10.1038/s41575-019-0126-x
Fu Y, Liu S, Zeng S, Shen H. From bench to bed: the tumor immune microenvironment and current immunotherapeutic strategies for hepatocellular carcinoma. J EXP CLIN CANC RES (2019) 38(1):396. doi:10.1186/s13046-019-1396-4
Pitt JM, Marabelle A, Eggermont A, Soria JC, Kroemer G, Zitvogel L. Targeting the tumor microenvironment: removing obstruction to anticancer immune responses and immunotherapy. ANN ONCOL (2016) 27(8): 1482-1492. doi:10.1093/annonc/mdw168
Picard E, Verschoor CP, Ma GW, Pawelec G. Relationships Between Immune Landscapes, Genetic Subtypes and Responses to Immunotherapy in Colorectal Cancer. FRONT IMMUNOL (2020) 11369. doi:10.3389/fimmu.2020.00369
Mariathasan S, Turley SJ, Nickles D, Castiglioni A, Yuen K, Wang Y et al. TGFbeta attenuates tumour response to PD-L1 blockade by contributing to exclusion of T cells. NATURE (2018) 554(7693): 544-548. doi:10.1038/nature25501
Emran AA, Chatterjee A, Rodger EJ, Tiffen JC, Gallagher SJ, Eccles MR et al. Targeting DNA Methylation and EZH2 Activity to Overcome Melanoma Resistance to Immunotherapy. TRENDS IMMUNOL (2019) 40(4): 328-344. doi:10.1016/j.it.2019.02.004
Taylor K, Loo YH, Chakravarthy A, Wang B, Shen SY, Ettayebi I et al. An open-label, phase II multicohort study of an oral hypomethylating agent CC-486 and durvalumab in advanced solid tumors. J IMMUNOTHER CANCER (2020) 8(2):e000883. doi:10.1136/jitc-2020-000883
Sasidharan NV, Toor SM, Taha RZ, Shaath H, Elkord E. DNA methylation and repressive histones in the promoters of PD-1, CTLA-4, TIM-3, LAG-3, TIGIT, PD-L1, and galectin-9 genes in human colorectal cancer. CLIN EPIGENETICS (2018) 10(1): 104. doi:10.1186/s13148-018-0539-3
Jung H, Kim HS, Kim JY, Sun JM, Ahn JS, Ahn MJ et al. DNA methylation loss promotes immune evasion of tumours with high mutation and copy number load. NAT COMMUN (2019) 10(1): 4278. doi:10.1038/s41467-019-12159-9
Goltz D, Gevensleben H, Dietrich J, Dietrich D. PD-L1 (CD274) promoter methylation predicts survival in colorectal cancer patients. ONCOIMMUNOLOGY (2017) 6(1): e1257454. doi:10.1080/2162402X.2016.1257454
Jones PA, Ohtani H, Chakravarthy A, De Carvalho DD. Epigenetic therapy in immune-oncology. NAT REV CANCER (2019) 19(3): 151-161. doi:10.1038/s41568-019-0109-9
Leek JT, Johnson WE, Parker HS, Jaffe AE, Storey JD. The sva package for removing batch effects and other unwanted variation in high-throughput experiments. BIOINFORMATICS (2012) 28(6): 882-883. doi:10.1093/bioinformatics/bts034
Hugo W, Zaretsky JM, Sun L, Song C, Moreno BH, Hu-Lieskovan S et al. Genomic and Transcriptomic Features of Response to Anti-PD-1 Therapy in Metastatic Melanoma. CELL (2016) 165(1): 35-44. doi:10.1016/j.cell.2016.02.065
Langmead B, Salzberg SL. Fast gapped-read alignment with Bowtie 2. NAT METHODS (2012) 9(4): 357-359. doi:10.1038/nmeth.1923
Ye L, Zhang T, Kang Z, Guo G, Sun Y, Lin K et al. Tumor-Infiltrating Immune Cells Act as a Marker for Prognosis in Colorectal Cancer. FRONT IMMUNOL (2019) 102368. doi:10.3389/fimmu.2019.02368
Wilkerson MD, Hayes DN. ConsensusClusterPlus: a class discovery tool with confidence assessments and item tracking. BIOINFORMATICS (2010) 26(12): 1572-1573. doi:10.1093/bioinformatics/btq170
Meng Q, Lu Y, Ruan D, Yu K, Chen Y, Xiao M et al. DNA methylation regulator-mediated modification patterns and tumor microenvironment characterization in gastric cancer. Molecular therapy. Nucleic acids (2021) 24695-710. doi:10.1016/j.omtn.2021.03.023
Hanzelmann S, Castelo R, Guinney J. GSVA: gene set variation analysis for microarray and RNA-seq data. BMC BIOINFORMATICS (2013) 147. doi:10.1186/1471-2105-14-7
Zhang B, Wu Q, Li B, Wang D, Wang L, Zhou YL. m(6)A regulator-mediated methylation modification patterns and tumor microenvironment infiltration characterization in gastric cancer. MOL CANCER (2020) 19(1): 53. doi:10.1186/s12943-020-01170-0
Jiang P, Gu S, Pan D, Fu J, Sahu A, Hu X et al. Signatures of T cell dysfunction and exclusion predict cancer immunotherapy response. NAT MED (2018) 24(10): 1550-1558. doi:10.1038/s41591-018-0136-1
Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. NUCLEIC ACIDS RES (2015) 43(7): e47. doi:10.1093/nar/gkv007
Chen H, Boutros PC. VennDiagram: a package for the generation of highly-customizable Venn and Euler diagrams in R. BMC BIOINFORMATICS (2011) 1235. doi:10.1186/1471-2105-12-35
Grauel AL, Nguyen B, Ruddy D, Laszewski T, Schwartz S, Chang J et al. TGFbeta-blockade uncovers stromal plasticity in tumors by revealing the existence of a subset of interferon-licensed fibroblasts. NAT COMMUN (2020) 11(1): 6315. doi:10.1038/s41467-020-19920-5
Becht E, de Reynies A, Giraldo NA, Pilati C, Buttard B, Lacroix L et al. Immune and Stromal Classification of Colorectal Cancer Is Associated with Molecular Subtypes and Relevant for Precision Immunotherapy. CLIN CANCER RES (2016) 22(16): 4057-4066. doi:10.1158/1078-0432.CCR-15-2879
Chan TA, Yarchoan M, Jaffee E, Swanton C, Quezada SA, Stenzinger A et al. Development of tumor mutation burden as an immunotherapy biomarker: utility for the oncology clinic. ANN ONCOL (2019) 30(1): 44-56. doi:10.1093/annonc/mdy495
Samstein RM, Lee CH, Shoushtari AN, Hellmann MD, Shen R, Janjigian YY et al. Tumor mutational load predicts survival after immunotherapy across multiple cancer types. NAT GENET (2019) 51(2): 202-206. doi:10.1038/s41588-018-0312-8
Mayakonda A, Lin DC, Assenov Y, Plass C, Koeffler HP. Maftools: efficient and comprehensive analysis of somatic variants in cancer. GENOME RES (2018) 28(11): 1747-1756. doi:10.1101/gr.239244.118
Guinney J, Dienstmann R, Wang X, de Reynies A, Schlicker A, Soneson C et al. The consensus molecular subtypes of colorectal cancer. NAT MED (2015) 21(11): 1350-1356. doi:10.1038/nm.3967
Lal N, White BS, Goussous G, Pickles O, Mason MJ, Beggs AD et al. KRAS Mutation and Consensus Molecular Subtypes 2 and 3 Are Independently Associated with Reduced Immune Infiltration and Reactivity in Colorectal Cancer. CLIN CANCER RES (2018) 24(1): 224-233. doi:10.1158/1078-0432.CCR-17-1090
Thorsson V, Gibbs DL, Brown SD, Wolf D, Bortone DS, Ou YT et al. The Immune Landscape of Cancer. IMMUNITY (2018) 48(4): 812-830. doi:10.1016/j.immuni.2018.03.023
Soldevilla B, Carretero-Puche C, Gomez-Lopez G, Al-Shahrour F, Riesco MC, Gil-Calderon B et al. The correlation between immune subtypes and consensus molecular subtypes in colorectal cancer identifies novel tumour microenvironment profiles, with prognostic and therapeutic implications. EUR J CANCER (2019) 123118-129. doi:10.1016/j.ejca.2019.09.008
van Dijk N, Gil-Jimenez A, Silina K, Hendricksen K, Smit LA, de Feijter JM et al. Preoperative ipilimumab plus nivolumab in locoregionally advanced urothelial cancer: the NABUCCO trial. NAT MED (2020) 26(12): 1839-1844. doi:10.1038/s41591-020-1085-z
Italiano A, Bellera C, D'Angelo S. PD1/PD-L1 targeting in advanced soft-tissue sarcomas: a pooled analysis of phase II trials. J HEMATOL ONCOL (2020) 13(1): 55. doi:10.1186/s13045-020-00891-5
Zhao P, Li L, Jiang X, Li Q. Mismatch repair deficiency/microsatellite instability-high as a predictor for anti-PD-1/PD-L1 immunotherapy efficacy. J HEMATOL ONCOL (2019) 12(1): 54. doi:10.1186/s13045-019-0738-1
Charoentong P, Finotello F, Angelova M, Mayer C, Efremova M, Rieder D et al. Pan-cancer Immunogenomic Analyses Reveal Genotype-Immunophenotype Relationships and Predictors of Response to Checkpoint Blockade. CELL REP (2017) 18(1): 248-262. doi:10.1016/j.celrep.2016.12.019
Morgan AE, Davies TJ, Mc AM. The role of DNA methylation in ageing and cancer. Proc Nutr Soc (2018) 77(4): 412-422. doi:10.1017/S0029665118000150
Henning AN, Roychoudhuri R, Restifo NP. Epigenetic control of CD8(+) T cell differentiation. NAT REV IMMUNOL (2018) 18(5): 340-356. doi:10.1038/nri.2017.146
Klumper N, Ralser DJ, Bawden EG, Landsberg J, Zarbl R, Kristiansen G et al. LAG3 (LAG-3, CD223) DNA methylation correlates with LAG3 expression by tumor and immune cells, immune cell infiltration, and overall survival in clear cell renal cell carcinoma. J IMMUNOTHER CANCER (2020) 8(1):e000552. doi:10.1136/jitc-2020-000552
Calon A, Espinet E, Palomo-Ponce S, Tauriello DV, Iglesias M, Cespedes MV et al. Dependency of colorectal cancer on a TGF-beta-driven program in stromal cells for metastasis initiation. CANCER CELL (2012) 22(5): 571-584. doi:10.1016/j.ccr.2012.08.013
Derynck R, Turley SJ, Akhurst RJ. TGFbeta biology in cancer progression and immunotherapy. NAT REV CLIN ONCOL (2021) 18(1): 9-34. doi:10.1038/s41571-020-0403-1
Xiao Y, Ma D, Zhao S, Suo C, Shi J, Xue MZ et al. Multi-Omics Profiling Reveals Distinct Microenvironment Characterization and Suggests Immune Escape Mechanisms of Triple-Negative Breast Cancer. CLIN CANCER RES (2019) 25(16): 5002-5014. doi:10.1158/1078-0432.CCR-18-3524
Bakouny Z, Braun DA, Shukla SA, Pan W, Gao X, Hou Y et al. Integrative molecular characterization of sarcomatoid and rhabdoid renal cell carcinoma. NAT COMMUN (2021) 12(1): 808. doi:10.1038/s41467-021-21068-9
Jiang Y, Zhan H. Communication between EMT and PD-L1 signaling: New insights into tumor immune evasion. CANCER LETT (2020) 46872-81. doi:10.1016/j.canlet.2019.10.013
Kather JN, Suarez-Carmona M, Charoentong P, Weis CA, Hirsch D, Bankhead P et al. Topography of cancer-associated immune cells in human solid tumors. ELIFE (2018) 7:e36967. doi:10.7554/eLife.36967
Echarti A, Hecht M, Buttner-Herold M, Haderlein M, Hartmann A, Fietkau R et al. CD8+ and Regulatory T cells Differentiate Tumor Immune Phenotypes and Predict Survival in Locally Advanced Head and Neck Cancer. Cancers (Basel) (2019) 11(9):1398. doi:10.3390/cancers11091398
Li S, Liu M, Do MH, Chou C, Stamatiades EG, Nixon BG et al. Cancer immunotherapy via targeted TGF-beta signalling blockade in TH cells. NATURE (2020) 587(7832): 121-125. doi:10.1038/s41586-020-2850-3
Lan Y, Zhang D, Xu C, Hance KW, Marelli B, Qi J et al. Enhanced preclinical antitumor activity of M7824, a bifunctional fusion protein simultaneously targeting PD-L1 and TGF-beta. SCI TRANSL MED (2018) 10(424):eaan5488. doi:10.1126/scitranslmed.aan5488
Melisi D, Oh DY, Hollebecque A, Calvo E, Varghese A, Borazanci E et al. Safety and activity of the TGFbeta receptor I kinase inhibitor galunisertib plus the anti-PD-L1 antibody durvalumab in metastatic pancreatic cancer. J IMMUNOTHER CANCER (2021) 9(3):e002068. doi:10.1136/jitc-2020-002068

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

DNA methylation regulator-based modification patterns define distinct immune phenotypes and immunotherapy responses in colorectal cancer

Status:

Version 1

Abstract

Figures

Introduction

Materials And Methods

Results

Discussion

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1