Skip to main content
NCBI home page
Cellular Oncology logoLink to Cellular Oncology
. 2024 Mar 23;47(4):1375–1389. doi: 10.1007/s13402-024-00933-x

DNA hypomethylation patterns and their impact on the tumor microenvironment in colorectal cancer

He Huang 1,#, Qian Li 2,#, Xusheng Tu 3,#, Dongyue Yu 4,#, Yundong Zhou 5,#, Lifei Ma 6,7,#, Kongyuan Wei 8, Yuzhen Gao 9, Guodong Zhao 10, Ruiqin Han 11,✉, Fangdie Ye 12,✉, Chunlian Ke 1,✉
PMCID: PMC12974014  PMID: 38520647

Abstract

Background

Recent research underscores the pivotal role of immune checkpoints as biomarkers in colorectal cancer (CRC) therapy, highlighting the dynamics of resistance and response to immune checkpoint inhibitors. The impact of epigenetic alterations in CRC, particularly in relation to immune therapy resistance, is not fully understood.

Methods

We integrated a comprehensive dataset encompassing TCGA-COAD, TCGA-READ, and multiple GEO series (GSE14333, GSE37892, GSE41258), along with key epigenetic datasets (TCGA-COAD, TCGA-READ, GSE77718). Hierarchical clustering, based on Euclidean distance and Ward's method, was applied to 330 primary tumor samples to identify distinct clusters. The immune microenvironment was assessed using MCPcounter. Machine learning algorithms were employed to predict DNA methylation patterns and their functional enrichment, in addition to transcriptome expression analysis. Genomic mutation profiles and treatment response assessments were also conducted.

Results

Our analysis delineated a specific tumor cluster with CpG Island (CGI) methylation, termed the Demethylated Phenotype (DMP). DMP was associated with metabolic pathways such as oxidative phosphorylation, implicating increased ATP production efficiency in mitochondria, which contributes to tumor aggressiveness. Furthermore, DMP showed activation of the Myc target pathway, known for tumor immune suppression, and exhibited downregulation in key immune-related pathways, suggesting a tumor microenvironment characterized by diminished immunity and increased fibroblast infiltration. Six potential therapeutic agents—lapatinib, RDEA119, WH.4.023, MG.132, PD.0325901, and AZ628—were identified as effective for the DMP subtype.

Conclusion

This study unveils a novel epigenetic phenotype in CRC linked to resistance against immune checkpoint inhibitors, presenting a significant step toward personalized medicine by suggesting epigenetic classifications as a means to identify ideal candidates for immunotherapy in CRC. Our findings also highlight potential therapeutic agents for the DMP subtype, offering new avenues for tailored CRC treatment strategies.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13402-024-00933-x.

Keywords: Colorectal cancer, Epigenetic alterations, Immune checkpoint therapy, DNA methylation, Tumor microenvironment

Introduction

Colorectal cancer (CRC) remains one of the most prevalent and lethal malignancies worldwide [1–3]. Despite advances in understanding its molecular underpinnings, CRC continues to pose significant challenges in terms of early detection, treatment, and management [4, 5]. One of the critical aspects of CRC that has garnered increasing attention in recent years is the role of epigenetic modifications, particularly DNA methylation, and its impact on tumor behavior and patient outcomes [6, 7].

DNA methylation, a key epigenetic mechanism, involves the addition of a methyl group to the DNA, typically at the 5th carbon of cytosine rings in CpG dinucleotides [8, 9]. This process plays a crucial role in regulating gene expression without altering the DNA sequence [10]. Aberrations in DNA methylation patterns have been implicated in various cancers, including CRC [11, 12]. These aberrations can lead to the silencing of tumor suppressor genes or the activation of oncogenes, contributing to cancer initiation and progression [13, 14].

In CRC, the phenomenon of DNA hypomethylation, particularly in the context of demethylated promoters and enhancers, has gained prominence [15, 16]. Hypomethylation in these regions can lead to the overexpression of oncogenes and contribute to genomic instability, thereby influencing tumor behavior and patient prognosis [17, 18]. Understanding these epigenetic alterations is crucial for developing targeted therapies and improving patient outcomes.

However, despite the recognized importance of DNA methylation changes in CRC, a comprehensive analysis of the tumor microenvironment landscape in the context of these changes remains limited [19]. The tumor microenvironment, consisting of various cell types, including immune and stromal cells, plays a pivotal role in cancer progression and response to therapy [20, 21]. The crosstalk between tumor cells and their microenvironment, mediated through genetic and epigenetic mechanisms, is critical for understanding tumor behavior and devising effective treatments.

Moreover, the heterogeneity of CRC at the molecular level poses additional challenges [22, 23]. CRC is not a uniform disease but consists of various subtypes with distinct molecular characteristics, each responding differently to therapies [24, 25]. This heterogeneity necessitates a more nuanced understanding of the disease at the molecular level to tailor treatments to individual patient needs effectively [26].

Given this background, our study aims to delve deep into the epigenetic landscape of CRC, with a specific focus on DNA hypomethylation and its association with the tumor microenvironment. We seek to unravel the complex interplay between epigenetic modifications and the microenvironment in CRC and how these interactions influence tumor behavior and patient outcomes. Our research is grounded in the hypothesis that DNA hypomethylation, particularly in enhancer regions, contributes significantly to tumor heterogeneity and the modulation of the tumor microenvironment in CRC. By exploring this aspect, we aim to contribute to a more nuanced understanding of CRC pathogenesis and open new avenues for targeted therapies.

Thanks to the advancements in bioinformatics technologies, researchers are now capable of unraveling the mechanisms of diseases through high-throughput data analysis [27, 28]. Consequently, we adopt a comprehensive multi-omics approach, integrating data from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) to achieve these goals. This method allows for an in-depth analysis of DNA methylation patterns and their correlation with the tumor microenvironment, alongside examining the regulatory networks of transcription factors affected by these epigenetic changes. Our study’s significance extends beyond expanding the understanding of CRC biology; it has substantial translational implications. By identifying specific epigenetic patterns linked to tumor behavior and the microenvironment, we aim to pave the way for novel diagnostic and therapeutic strategies in CRC. These could include leveraging DNA methylation patterns as biomarkers for early detection and prognosis, and developing epigenetic therapies tailored to modulate the tumor microenvironment, ultimately enhancing treatment efficacy and patient care in CRC.

Methods

DNA methylation and gene expression data acquisition

We utilized the Illumina Human Methylation (450K) platform to quantify DNA methylation profiles. These data, along with the K-array, were sourced from the UCSC Xena platform (https://xenabrowser.net), encompassing projects TCGA-READ (n = 106) and TCGA-COAD (n = 347). Additionally, gene expression data for TCGA-COAD (n = 512) and TCGA-READ (n = 177) were obtained for transcriptome analysis. This data was quantified using fragments per kilobase million (FPKM), which were subsequently converted to transcripts per kilobase (TPM) to enhance compatibility across samples. Moreover, Fire Browse (https://firebrowse.org) provided essential data on copy number segments.

Survival, mutation, and clinicopathological data collection

Comprehensive data on overall survival, somatic mutations, and clinicopathological features were collated from cBioPortal (https://www.cbioportal.org). We conducted a meticulous matching process with survival information, resulting in a cohesive dataset of 330 primary CRC tumor cases and 21 standard samples, integrating both gene expression profiles and DNA demethylation data.

External epigenetic and transcriptomic data sets

Four additional CRC datasets were gathered from the Gene Expression Omnibus (GEO), comprising one epigenetic dataset (GSE77718, n = 192 [96 tumors and 96 adjacent normal tissues]) [29] and three transcriptomic datasets (GSE14333 [30], n = 196; GSE37892 [31], n = 130; GSE41258 [32], n = 198). Recurrence-free survival (RFS) data were extracted from associated literature for these transcriptome datasets. For microarray expression data involving multiple probe IDs for the same gene symbol, median values were employed.

Pre-processing of gene expression profiles and DNA methylation

The DNA methylation profile of the TCGA cohort required a synthesis of β matrices from the two projects. We employed the ComBat algorithm from the R package “sva” to mitigate potential batch effects [33]. This algorithm, known for its ability to correct technical variations, utilizes an empirical Bayes method. Given the nature of the combined β matrix, we first log-transformed the β-values before applying a reverse logit transformation and subsequent ComBat adjustment, as per the recommended pipeline [34]. Following this, a rigorous filtering procedure was implemented using the “ChAMP” R package. Our criteria for filtering included: probes with less than 3 beads per probe in at least 5% of samples, all multi-hit probes, probes with detection P-values over 0.01, non-CpG probes, SNP-related probes, and probes on the X and Y chromosomes [34, 35]. Similarly, for gene expression profiles, the ComBat algorithm was utilized to address batch effects in the combined TCGA-COAD and TCGA-READ transcriptomic data.

Unsupervised hierarchical clustering for cancer-specific loss of DNA methylation

Adopting the strategy previously outlined in studies [36], we applied specific filtering criteria to identify probes exhibiting high methylation (average β-value > 0.8) in normal tissue, coupled with variable methylation in CRC (standard deviation of β > 0.2 in CRC tumors). This approach was intended to enrich for sites that lose methylation during cancer progression. In addition to β-values, we utilized M-values for their enhanced capability in quantifying methylation strength [37]. We performed unsupervised hierarchical clustering using Euclidean distance and Ward’s method to categorize 330 primary tumor samples based on their β-values. By cutting the dendrogram at k = 3, we identified three distinct clusters.

Calculation of microenvironment cell abundance and pathway enrichment

To detect the presence of infiltrating stromal or immune cells in tumor tissues, we employed the R package “ESTIMATE” [38]. In the TCGA CRC cohort, the DNA methylation score of infiltrated lymphocytes was computed following the methodology detailed in a literature [39]. Additionally, the “MCPcounter” R package was utilized for quantifying two stromal and eight immune cell populations in the heterogeneous tissues, derived from transcriptomic data.

Differential analysis and functional enrichment

The “ChAMP” package’s champ.MP function facilitated the identification of differentially methylated probes (DMPs). A probe k was classified as significantly methylated if its β-value exceeded 0.3 in a specific subtype, yet remained below 0.2 in a group with an FDR < 0.25 and p < 0.05. Conversely, this criterion was inverted for probes that significantly lost methylation. DNA methylation at the CpG level was analyzed through generalized gene-set testing using the “missMethyl” R package, with the hallmark genes background sourced from the MSigDB [40, 41]. The “limma” R package [42] was employed for gene set enrichment analysis (GSEA), creating a pre-ranked gene list based on descending log2Fold change values from differential expression analysis. Functional enrichment was determined based on hallmark pathways using the “clusterProfiler” R package [43]. Additionally, functional enrichment analysis based on the gene list was conducted using Enrichr (https://maayanlab.cloud/Enrichr) [44].

Cancer Subtype Characterization:

The “movics” R package was instrumental in characterizing identified tumor phenotypes. We considered mutational frequency, clinical characteristics, the fraction of the copy number-altered genome, and other parameters as default [45]. Mutation landscapes were further analyzed using the “maftools” R package [46]. The focal somatic Copy Number Alterations (CNAs) were identified using GISTIC2.0 on Gene Pattern (https://www.genepattern.org) with defined parameters including amplification/deletion thresholds at 0.1, a q-value threshold of 0.05, and a confidence level of 0.95 [47].

Integrative analysis of transcriptome expression and enhancer DNA methylation

Our integrated analysis combined gene expression and DNA methylation, using the R package to correlate the two datasets. The approach involved identifying probes from the Infinium Human Methylation 450k BeadChip annotation, followed by a Mann-Whitney U test to validate the null hypothesis of equal or greater gene expression in a specific group compared to the reference group. This method helped identify putative genes overexpressed due to hypomethylation of enhancer probes. Enriched motifs for enhancers were identified using ELMER, focusing on hypomethylated enhancers linked to target genes. We used Hypergeometric Optimization of Motif Enrichment (HOMER) to detect motif occurrence within ± 250 base pair regions of each probe, referencing HOCOMOCO (Human Comprehensive Model Collection version 11). Fisher’s exact test quantified enrichment for each probe set, with the Benjamini-Hochberg procedure correcting for multiple testing. A probe set was significantly enriched for a particular motif if it had a confidence interval of 95%, an odds ratio above 1.1, a motif occurrence frequency 10 times in the probe set, and an FDR < 0.05. Regulatory transcription factors were identified using ELMER, revealing a negative association between expression and DNA methylation at enriched motifs. Candidate motif-transcription factor pairs were evaluated using the Mann-Whitney U test, hypothesizing greater or equal gene expression in a group of interest compared to the reference group. Top-ranking transcription factors, based on the -log10 P-value, were considered candidate upstream regulators, particularly those in the top 5%.

Regulon analysis

Employing the “RTN” R package, as described [48, 49], we reconstructed the transcriptional regulatory network for enhancer-associated transcription factors. Spearman rank-order correlation and mutual information analysis identified potential associations between regulators and targets in the transcriptome expression profile. Permutation analysis with an FDR threshold of  > 0.00001 and resampling (1000 times using bootstrapping) eliminated unstable associations, maintaining a consensus bootstrap rate of  > 95%. Weaker associations were filtered out using inequality filtering, and previously calculated regulons were extended to external datasets for new sample evaluation.

Inference of therapeutic response and statistical analyses

The “pRRophetic” R package predicted chemotherapeutic sensitivity in CRC samples using expression profiles from 727 cancer cell lines [50], with drug sensitivity and phenotype data sourced from GDSC 2016 (https://www.cancerrxgene.org). IC50 values, indicative of treatment sensitivity, were estimated for each sample, and prediction accuracy was assessed using 10-fold cross-validation [50]. For immunotherapy, subclass mapping inferred clinical responses to Immune Checkpoint Inhibitors (ICIs) [51], leveraging a publicly available dataset of 47 melanoma patients responsive to immunotherapy [52]. SubMap assessed transcriptomic similarities across different phenotypes and datasets.

Statistical analyses

R version 4.0.2 was employed for statistical analyses, using Fisher’s exact test for categorical data, the Kaplan-Meier estimator for survival analysis, and the Mann-Whitney U test for continuous data. A two-sided p-value < 0.05 was deemed statistically significant for all unadjusted comparisons.

Results

Identification of demethylators in CRC patients

To elucidate the role of DNA hypomethylation in CRC biology, we conducted an extensive analysis using Illumina 450k arrays on 330 TCGA CRC samples. Our initial step involved identifying 107,204 probes highly methylated in normal tissues. From these, we selected 5471 probes that exhibited variable methylation across tumor samples to focus on CpG sites that lose methylation in CRC. Through unsupervised hierarchical clustering, we distinguished three distinct tumor phenotypes based on their methylation profiles: DMP-High (HDMP; n = 61), DMP-Low (LDMP; n = 122), and DMP-Non (NDMP; n = 147). This classification revealed profound hypomethylation patterns in these groups (Fig. 1). Interestingly, there was no significant correlation between these demethylated phenotypes and major clinicopathological features (all, P > 0.05; Supplementary Table S1). However, the survival analysis indicated that the HDMP phenotype was associated with poorer clinical outcomes (P = 0.01; Fig. 1), identifying it as an aggressive form of DMP in CRC.

Fig. 1.

Fig. 1

Open in a new tab

Identification of demethylators in CRC patients from the TCGA cohort. A Heatmap showing the DNA methylation landscape using probes with high methylation in normal tissues and variable methylation across CRC tumor cases in the TCGA cohort. B Kaplan-Meier curves comparing overall survival (OS) among three groups of demethylators. C Density plot displaying the differential methylated probes between demethylator positive (DMP) cases and others. Probes with hypermethylation and hypomethylation are marked in red and blue, respectively. D Heatmap illustrating the DNA methylation landscape of epithelial-mesenchymal transition (EMT) using pathway representative genes, based on methylation M-value (Color figure online)

Further, we analyzed differentially methylated probes between DMP and other cases. Our findings revealed 1520 probes undergoing DNA methylation loss, while only 16 probes showed methylation gain in the DMP group compared to other cases (Fig. 1; Supplementary Table S2), suggesting a predominant loss of DNA methylation in CRC’s DMP. Generalized Gene-Set Testing (GGST) indicated significant hypomethylation in the epithelial-mesenchymal transition (EMT) pathway in the DMP group (P = 0.0001, FDR = 0.006; Fig. 1, Supplementary Table S3).

To validate these findings, we employed the same analytical approach on an external dataset (GSE77718). We identified 100,765 probes with high DNA methylation in normal tissues, noting a significant overlap with the TCGA cohort (73.5% vs. 78.2% in GSE77718; P < 0.001; Fig. 2). Subsequent clustering of 449 highly variable probes in tumor tissues revealed three demethylation factors similar to those observed in the TCGA cohort (Fig. 2). An analysis of differentially methylated probes in this dataset identified 2649 probes losing DNA methylation and only 15 gaining methylation in the DMP group compared to other cases (Fig. 2) [53]. Additionally, GGST demonstrated a significantly lower methylation of the EMT marker pathway in the DMP group compared to other cases (Fig. 2, Supplementary Table S4), reinforcing the association of hypomethylation with aggressive CRC phenotypes.

Fig. 2.

Fig. 2

Open in a new tab

Validation of demethylation factors in CRC patients using GEO dataset. A Heatmap presenting DNA methylation in normal tissues and variable methylation in tumor cases within the GEO dataset, using probes with high methylation in normal tissues. B Density plot showing differential methylation probes between DMP and other cases, with hypermethylation and hypomethylation indicated in red and blue, respectively. C Venn diagram demonstrating overlapping probes with high DNA methylation in normal tissues between the TCGA cohort and GEO dataset. D Heatmap depicting the DNA methylation landscape of EMT, based on methylation M-value, using pathway representative genes (Color figure online)

Enhancer-associated demethylation and transcription factors

Leveraging the comprehensive clinicopathological, survival, and multi-omics data from the TCGA cohort, our study primarily focused on unraveling the complexities of tumor heterogeneity in colorectal cancer (CRC). We honed in on genomic regions particularly susceptible to DNA methylation changes in Demethylated Phenotypes (DMP). In these regions, a significant enrichment of demethylated enhancers was observed in comparison to the Illumina 450k array background, particularly among the 1520 probes showing DNA methylation loss (46% vs. 22%; P < 0.001; Fig. 3). We utilized DNA methylation data to identify functional alterations at transcriptional enhancers in tumors and integrated these with matched expression profiles from the TCGA cohort. This integration facilitated the reconstruction of transcription factor (TF) networks using the ELMER pipelines. Our analysis revealed that approximately 2924 enhancer probes underwent DNA methylation loss in the DMP groups, identified using a β-value > 0.2 and FDR < 0.05 (Supplementary Table S5).

Fig. 3.

Fig. 3

Open in a new tab

Identification of enhancer-associated transcription factors. A Bar graph showing the specific regional distribution of probes losing DNA methylation in DMP versus other cases, against the background of the Illumina 450k array. B Heatmap correlating DNA methylation with gene expression, showing epigenetic activation patterns due to low methylation at enhancer probes. C Dot plot illustrating enrichr functional enrichment for 780 epigenetically activated genes. D, E Display the distribution of enrichment scores for Myc targets and oxidative phosphorylation in DMP and other cases. F Box plot presenting the distribution of regulatory activities of 14 transcription factors (TFs) in DMP versus other cases

Subsequently, each of these probes was linked to a nearby gene, resulting in the identification of 8781 nearby genes (Supplementary Table S6) and 4575 putative probe-gene pairs (Supplementary Table S7), with overall gene expression in the DMP group being equal to or greater than that in other cases (Fig. 3). Notably, 780 genes within these probe-gene pairs demonstrated biological significance in the Myc targets and oxidative phosphorylation pathways (Odds Ratio [OR]: 3.6 and 3.1, respectively; P < 0.001, FDR < 0.001) (Fig. 3, Supplementary Table S8). Gene Set Variation Analysis (GSVA) further indicated an overexpression of these pathways in the DMP group compared to other cases (Myc targets: P = 0.041; oxidative phosphorylation: P = 0.005; Fig. 3E).

In our motif analysis, approximately 280 enriched probe sets were considered for specific motifs (Supplementary Table S9). This led to the identification of 17 regulatory TFs, including GBX2, MSX2, PITX2, and others, based on motif and expression association analyses (Supplementary Table S9). The activity of these TFs, forming regulons, was assessed through their transcriptome expression profiles. Among these TFs, 12 showed significantly higher regulon activity in the DMP group compared to other cases (Fig. 3).

Given the known association between activated oxidative phosphorylation and poor clinical outcomes in CRC, we explored the potential role of oxidative phosphorylation as an energy source driving tumor aggressiveness. Unsupervised hierarchical clustering was performed on three external datasets with transcriptome profiles based on genes involved in oxidative phosphorylation (Fig. 4C). In each dataset, two subtypes were identified, with the subtype exhibiting activated oxidative phosphorylation generally correlating with poorer Recurrence-Free Survival (RFS) rates (P = 0.044 for GSE14333, P = 0.071 for GSE37892, P = 0.061 for GSE41258; Fig. 4F), further emphasizing the significance of metabolic pathways in CRC prognosis.

Fig. 4.

Fig. 4

Open in a new tab

Prognostic value of oxidative phosphorylation in CRC patients. Unsupervised clustering of genes representing oxidative phosphorylation pathway identifies two subtypes in (A) GSE14333, (B) GSE37892, and (C) GSE41258 datasets. Kaplan-Meier survival curves for relapse-free survival (RFS) rates in (D) GSE14333, (E) GSE37892, and (F) GSE41258

Tumor microenvironment landscape of DMP in CRC patients

In our study, Demethylated Phenotypes (DMP) in colorectal cancer (CRC) were characterized by the activation of the Myc targets pathway, which has been previously associated with the suppression of tumor immunity. To investigate this further, we conducted differential expression analysis (Fig. 5, Supplementary Table S10) and Gene Set Enrichment Analysis (GSEA) based on log2FoldChange values derived from these results. This analysis revealed that DMP significantly downregulated several immune-related pathways, including inflammatory responses (Normalized Enrichment Score [NES]: −1.94, P = 0.001, FDR = 0.003), IFN-α responses (NES: −1.79, P = 0.001, FDR = 0.003), and IFN-γ responses (NES: −1.93, P = 0.001, FDR = 0.003) (Fig. 5).

Fig. 5.

Fig. 5

Open in a new tab

Tumor microenvironment landscape in CRC patients with DMP. A Heatmap showing tumor microenvironment status, including gene expression of various immune checkpoint targets (top panel), estimated abundance of 10 microenvironment cell types (middle panel), and MeTIL scores (bottom panel). B Box plot illustrating the expression level distribution of immune checkpoint targets. C Box plot showing the distribution of tumor microenvironment cell abundance. D–F Display the distribution of immune escape score (IES), stromal escape score (SES), and MeTIL scores between DMP and other cases. G Volcano plot indicating differentially expressed genes between DMP and other cases. H GSEA chart showing downregulated hallmark pathways in DMP compared to other cases, including inflammatory response, IFN-α response, and IFN-γ response

Furthermore, we quantified the infiltration levels of ten types of cells within the tumor microenvironment and assessed the expression of immune checkpoints in CRC samples (Fig. 5). Our findings showed a significant increase in the expression levels of several immune checkpoints in DMP, including CD274 (PDL1), CD247 (CD3), PDCD1 (PD1), PDCD1LG2 (PDL2), TNFRSF9 (CD137), CTLA4 (CD152), and TNFRSF4 (CD134), which are potential targets for immunotherapy (Fig. 5). We noted that DMP was typically characterized by an immune-depleted tumor microenvironment (Fig. 5), further evidenced by significantly lower immune enrichment scores (IES) (Fig. 5), stromal enrichment scores (SES) (Fig. 5), and MeTIL (Fig. 5) compared to other CRC cases (all P < 0.001).

To establish a correlation between the epigenetic phenotype and tumor immunity, we explored the relationship between regulon activity, derived from the epigenetically regulated transcriptome, and the tumor microenvironment. We extended the regulon networks, originally constructed from the TCGA cohort, to include data from three external datasets with available gene expression data (GSE14333, GSE37892, and GSE41258). In these datasets, regulon activities for 14 transcription factors (TFs) were calculated, followed by unsupervised hierarchical clustering (Fig. 6C). Consistently, tumors with regulon-activated landscapes typically exhibited an immune-depleted phenotype.

Fig. 6.

Fig. 6

Open in a new tab

Validation of epigenetic regulation in the tumor microenvironment. Circular heatmap showing tumors with activated regulatory factors of 14 TFs in external GEO datasets (including GSE14333, GSE37892, and GSE41258), typically characterized by an immune exhaustion phenotype

Chromosomal instability of DMP and genomic heterogeneity

In our examination of colorectal cancer (CRC), we delved into the genomic heterogeneity and chromosomal instability associated with Demethylated Phenotypes (DMP). By analyzing the mutational landscape, we identified 87 genes exhibiting differential mutational frequencies between DMP and other CRC cases (P < 0.05), each with an overall mutation frequency greater than 5% across the cohort (Fig. 7). Notably, these genes manifested more mutations in non-DMP cases than in DMP, as detailed in Supplementary Table S11. Further, our analysis of broad-level Copy Number Alterations (CNAs) (Fig. 7) led to an estimation of chromosomal instability through Fractional Genome Altered (FGA) scores. Intriguingly, we found DMP to harbor a higher degree of chromosomal instability, characterized by significant copy number gains or losses compared to other cases (both P < 0.01; Fig. 7, Supplementary Table S12).

Fig. 7.

Fig. 7

Open in a new tab

Genomic heterogeneity, chromosomal instability, and potential therapeutic approaches for CRC patients with DMP. A OncoPrint illustrating the distribution of differentially mutated genes between DMP and other cases. B Bar graph showing the distribution and scores of fraction of genome altered (FGA) and fraction of genome gained/lost (FGG). C Comparison of generalized copy number alterations (CNA) across the entire cohort and between different epigenetic phenotypes. D Heatmap of subclass analysis results, indicating the potential limited benefit from immune checkpoint inhibitors in DMP. E Box plot displaying the distribution of estimated IC50 based on the GDSC database between two epigenetic phenotypes

Potential therapeutic strategy for DMP

In light of the discovery of an immune-depleted microenvironment in DMP, we probed the differences in response to immune checkpoint blockade between DMP and other CRC cases. Surprisingly, DMP did not exhibit transcriptomic similarity to melanoma subtypes responsive to anti-PD1 blockade, in contrast to other CRC cases that might benefit from such immunotherapy (Fig. 7). This finding suggests the potential of epigenetic classifications in identifying suitable candidates for immunotherapy among CRC patients. Given the poor clinical outcomes associated with DMP, we pursued potential anti-CRC drugs that might be effective for DMP. Utilizing an in silico approach for drug screening, we applied a predictive ridge regression model correlating drug sensitivity with cell line data for each CRC case in the TCGA cohort (Supplementary Table S13). This analysis identified six drugs with potential efficacy against DMP compared to other cases, namely Lapatinib, RDEA119, WH.4.023, MG.132, PD.0325901, and AZ628, all displaying a significant ΔIC50 greater than 0.2 (P < 0.05; Fig. 7). This insight opens new avenues for targeted therapeutic strategies in treating DMP in CRC, potentially improving outcomes for patients with this challenging phenotype.

Discussion

This study represents a significant contribution to understanding CRC pathogenesis, particularly focusing on the role of DNA hypomethylation in shaping the tumor microenvironment. Our findings reveal new insights into the epigenetic landscape of CRC, highlighting the importance of DNA hypomethylation, especially in enhancer regions, and its interplay with the tumor microenvironment.

The association between DNA hypomethylation and tumor heterogeneity in CRC, as observed in our study, aligns with and expands upon previous research. For instance, studies like Timp et al. and Hinoue et al. have demonstrated the role of DNA methylation in CRC, but our research provides a more detailed exploration of hypomethylation in enhancer regions [6, 54] . This focus is crucial, as enhancers play a significant role in regulating gene expression, and their alteration through hypomethylation could lead to the activation of oncogenic pathways and suppression of tumor suppressor genes.

Our multi-omics approach, utilizing data from TCGA and GEO, allowed us to deeply analyze the relationship between DNA methylation patterns and the tumor microenvironment. This comprehensive analysis is a step forward from studies that have primarily focused on individual aspects of CRC pathology. By integrating various data types, we could uncover complex interactions that would not be apparent when examining genomic or epigenomic data in isolation. This approach aligns with the current trend in cancer research, which emphasizes the importance of multi-dimensional data analysis for a holistic understanding of cancer biology, as suggested by the Pan-Cancer Atlas initiative.

One of the key findings from our study is the significant downregulation of immune-related pathways in CRC with DNA hypomethylation, particularly in the Myc targets pathway. This observation adds to the growing body of literature on the immunosuppressive nature of certain CRC subtypes. Studies like Puccini et al. have discussed the role of the tumor microenvironment in immunotherapy response, and our findings provide a molecular basis for this phenomenon, showing how epigenetic changes can modulate the tumor microenvironment to create an immune-depleted landscape [55] . This insight has crucial implications for the development of immunotherapies in CRC, as it suggests that targeting epigenetic modifications could enhance the efficacy of these treatments.

Furthermore, our research highlights the potential of DNA methylation patterns as biomarkers for CRC. The ability to identify specific epigenetic signatures associated with tumor behavior and patient outcomes could revolutionize CRC diagnostics and prognostics. This aspect of our study correlates with the findings of Li et al., who emphasized the potential of epigenetic biomarkers in cancer. However, our study extends this concept by linking these biomarkers to the tumor microenvironment, providing a more comprehensive understanding of their significance [56].

While this study makes significant strides in elucidating the epigenetic landscape of CRC, particularly regarding DNA hypomethylation and its impact on the tumor microenvironment, it is not without limitations. Firstly, our reliance on retrospective data from repositories such as TCGA and GEO poses inherent constraints related to sample heterogeneity and historical data collection methods. Additionally, the observational nature of this study limits our ability to establish causality between epigenetic changes and tumor behavior. Our findings, primarily associative, require validation through prospective, experimental studies to confirm the functional impact of identified epigenetic alterations. Furthermore, the complexity of epigenetic regulation in CRC necessitates more extensive, multi-layered investigations, integrating proteomic and metabolic data, to fully comprehend the intricate interplay of genetic and epigenetic factors in CRC pathogenesis and progression.

In conclusion, our study unveils pivotal insights into the epigenetic underpinnings of CRC, particularly emphasizing the role of DNA hypomethylation in shaping the tumor microenvironment. This understanding opens up new possibilities for diagnosis, prognosis, and therapeutic intervention in CRC, especially for immune-depleted subtypes that currently have limited options. By pinpointing potential therapeutic targets within the epigenetic landscape, we highlight the promising avenue of targeting DNA hypomethylation to modulate the tumor environment and potentially amplify the effectiveness of immunotherapies. Our findings not only contribute to a deeper understanding of CRC biology but also open new avenues for diagnosis, prognosis, and treatment. The potential of epigenetic therapies to improve patient outcomes in CRC is an exciting prospect, and further research in this area could lead to significant advances in the field.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Supplementary Material 1 (5.6MB, pdf)
Supplementary Material 2 (4MB, xlsx)

Acknowledgements

Not applicable.

Author contributions

C.K., F.Y., and R.H. contributed to the study design and critical revision of the manuscript. H.H., Q.L., X.T., Y.Z., D.Y., and L.M. carried out the study and drafted the manuscript. H.H., Q.L., X.T., Y.Z., D.Y., L.M., Y.G., K.W., and G.Z. analyzed the data. All authors read and approved the final manuscript.

Funding

This research was supported by Medical Products Administration of Guangdong Province (2021YDZ03), the Science and Technology Research Project of Hebei Higher Education Institutions (QN2021012), the National Natural Science Foundation of China (81902498, H2022405002), Hubei Provincial Natural Science Foundation (2019CFB177), Natural Science Foundation of Hubei Provincial Department of Education (Q20182105), Chen Xiao-ping Foundation for the development of science and technology of Hubei Provincial (CXPJJH11800001-2018333), The Foundation of Health and Family planning Commission of Hubei Province (WJ2021Q007), Innovation and entrepreneurship training program (201810929005, 201810929009, 201810929068, 201813249010, S201910929009, S201910929045, S202013249005, S202013249008 and 202010929009) and The Scientific and Technological Project of Taihe hospital (2021JJXM009).

Data availability

The original contributions presented in the study are included in the article, further inquiries can be directed to the corresponding author.

Declarations

Ethical approval

Not applicable.

Competing interests

None declared.

Footnotes

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

He Huang, Qian Li, Xusheng Tu, Dongyue Yu, Yundong Zhou and Lifei Ma have contributed equally to this work.

Contributor Information

Ruiqin Han, Email: hrq@ibms.pumc.edu.cn.

Fangdie Ye, Email: yefangdie@163.com.

Chunlian Ke, Email: kechunlian@163.com.

References

  • 1.H. Brenner, M. Kloor, C.P. Pox, Colorectal cancer. Lancet 383(9927), 1490–1502 (2014) [DOI] [PubMed] [Google Scholar]
  • 2.B. Duan, Y. Zhao, J. Bai, J. Wang, X. Duan, X. Luo, et al., Colorectal cancer: an overview, In: Gastrointestinal Cancers, ed. J.A. Morgado-Diaz (Exon Publications, Brisbane (AU), 2022). Copyright: The Authors.; The authors confirm that the materials included in this chapter do not violate copyright laws. Where relevant, appropriate permissions have been obtained from the original copyright holder(s), and all original sources have been appropriately acknowledged or referenced [Google Scholar]
  • 3.R.L. Siegel, K.D. Miller, N.S. Wagle, A. Jemal, Cancer statistics, 2023. CA Cancer J. Clin. 73(1), 17–48 (2023) [DOI] [PubMed] [Google Scholar]
  • 4.N.A. Johdi, N.F. Sukor, Colorectal Cancer Immunotherapy: options and Strategies. Front. Immunol. 11, 1624 (2020) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.M. Marcuello, V. Vymetalkova, R.P.L. Neves, S. Duran-Sanchon, H.M. Vedeld, E. Tham, et al, Circulating biomarkers for early detection and clinical management of colorectal cancer. Mol. Aspect. Med. 69, 107–122 (2019) [DOI] [PubMed] [Google Scholar]
  • 6.H. Hampel, M.F. Kalady, R. Pearlman, P.P. Stanich, Hereditary Colorectal Cancer. Hematol./Oncol. Clin. N. Am. 36(3), 429–447 (2022) [DOI] [PubMed] [Google Scholar]
  • 7.S. Sakata, D.W. Larson, Targeted Therapy for Colorectal Cancer. Surg. Oncol. Clin. N. Am. 31(2), 255–264 (2022) [DOI] [PubMed] [Google Scholar]
  • 8.A.L. Mattei, N. Bailly, A. Meissner, DNA methylation: a historical perspective. Trends Genet. 38(7), 676–707 (2022) [DOI] [PubMed] [Google Scholar]
  • 9.L.D. Moore, T. Le, G. Fan, DNA methylation and its basic function. Neuropsychopharmacol. 38(1), 23–38 (2013) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.M. Klutstein, D. Nejman, R. Greenfield, H. Cedar, DNA Methylation in Cancer and Aging. Cancer Res. 76(12), 3446–3450 (2016) [DOI] [PubMed] [Google Scholar]
  • 11.G. Jung, E. Hernández-Illán, L. Moreira, F. Balaguer, A. Goel, Epigenetics of colorectal cancer: biomarker and therapeutic potential. Nat. Rev. Gastroenterol. Hepatol. 17(2), 111–130 (2020) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.D. Müller, B. Győrffy, DNA methylation-based diagnostic, prognostic, and predictive biomarkers in colorectal cancer. Biochim. Biophys. Acta Rev. Cancer 1877(3), 188722 (2022) [DOI] [PubMed] [Google Scholar]
  • 13.B. Dariya, S. Aliya, N. Merchant, A. Alam, G.P. Nagaraju, Colorectal Cancer Biology, Diagnosis, and Therapeutic Approaches. Crit. Rev. Oncogenesis 25(2), 71–94 (2020) [DOI] [PubMed] [Google Scholar]
  • 14.V. Vymetalkova, P. Vodicka, S. Vodenkova, S. Alonso, R. Schneider-Stock, DNA methylation and chromatin modifiers in colorectal cancer. Mol. Aspect. Med. 69, 73–92 (2019) [DOI] [PubMed] [Google Scholar]
  • 15.K. Cervena, A. Siskova, T. Buchler, P. Vodicka, V. Vymetalkova, Methylation-Based Therapies for Colorectal Cancer. Cells 9(6) (2020) [DOI] [PMC free article] [PubMed]
  • 16.A.M. Jubb, S.M. Bell, P. Quirke, Methylation and colorectal cancer. J. Pathol. 195(1), 111–134 (2001) [DOI] [PubMed] [Google Scholar]
  • 17.A. Gutierrez, H. Demond, P. Brebi, C.G. Ili, Novel Methylation Biomarkers for Colorectal Cancer Prognosis. Biomolecules 11(11) (2021) [DOI] [PMC free article] [PubMed]
  • 18.V.A. Ionescu, G. Gheorghe, N. Bacalbasa, A.L. Chiotoroiu, C. Diaconu, Colorectal Cancer: from Risk Factors to Oncogenesis. Medicina (Kaunas, Lithuania) 59(9) (2023) [DOI] [PMC free article] [PubMed]
  • 19.V.V. Lao, W.M. Grady, Epigenetics and colorectal cancer. Nat. Rev. Gastroenterol. Hepatol. 8(12), 686–700 (2011) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Y. Chen, X. Zheng, C. Wu, The Role of the Tumor Microenvironment and Treatment Strategies in Colorectal Cancer. Front. Immunol. 12, 792691 (2021) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.C.R. Lichtenstern, R.K. Ngu, S. Shalapour, M. Karin, Immunotherapy, Inflammation and Colorectal Cancer. Cells 9(3) (2020) [DOI] [PMC free article] [PubMed]
  • 22.M.A. Senchukova, Genetic heterogeneity of colorectal cancer and the microbiome. World J. Gastrointestinal Oncol. 15(3), 443–463 (2023) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Y. Xi, P. Xu, Global colorectal cancer burden in 2020 and projections to 2040. Transl. Oncol. 14(10), 101174 (2021) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.M.W. Dougherty, C. Jobin, Intestinal bacteria and colorectal cancer: etiology and treatment. Gut. Microbes. 15(1), 2185028 (2023) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.G. Zhu, L. Pei, H. Xia, Q. Tang, F. Bi, Role of oncogenic KRAS in the prognosis, diagnosis and treatment of colorectal cancer. Mol. Cancer 20(1), 143 (2021) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.G. Rosati, G. Aprile, A. Colombo, S. Cordio, M. Giampaglia, A. Cappetta, et al., Colorectal Cancer Heterogeneity and the Impact on Precision Medicine and Therapy Efficacy. Biomedicines. 10(5) (2022) [DOI] [PMC free article] [PubMed]
  • 27.J. Liu, P. Chen, J. Zhou, H. Li, Z. Pan, Prognostic impact of lactylation-associated gene modifications in clear cell renal cell carcinoma: insights into molecular landscape and therapeutic opportunities. Environ. Toxicol. n/a(n/a) [DOI] [PubMed]
  • 28.H. Li, L. Zhou, W. Zhou, X. Zhang, J. Shang, X. Feng, et al., Decoding the mitochondrial connection: development and validation of biomarkers for classifying and treating systemic lupus erythematosus through bioinformatics and machine learning. BMC Rheumatol. 7(1), 44 (2023) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.T. McInnes, D. Zou, D.S. Rao, F.M. Munro, V.L. Phillips, J.L. McCall, et al., Genome-wide methylation analysis identifies a core set of hypermethylated genes in CIMP-H colorectal cancer. BMC Cancer 17(1), 228 (2017) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.R.N. Jorissen, P. Gibbs, M. Christie, S. Prakash, L. Lipton, J. Desai, et al., Metastasis-Associated Gene Expression Changes Predict Poor Outcomes in Patients with Dukes Stage B and C Colorectal Cancer. Clin Cancer Res 15(24), 7642–7651 (2009) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.S. Laibe, A. Lagarde, A. Ferrari, G. Monges, D. Birnbaum, S. Olschwang, A seven-gene signature aggregates a subgroup of stage II colon cancers with stage III. OMICS 16(10), 560–565 (2012) [DOI] [PubMed] [Google Scholar]
  • 32.M.L. Martin, Z. Zeng, M. Adileh, A. Jacobo, C. Li, E. Vakiani, et al., Logarithmic expansion of LGR5(+) cells in human colorectal cancer. Cell. Signal. 42, 97–105 (2018) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.W.E. Johnson, C. Li, A. Rabinovic, Adjusting batch effects in microarray expression data using empirical Bayes methods. Biostatistics (Oxford, England) 8(1), 118–127 (2007) [DOI] [PubMed] [Google Scholar]
  • 34.Y. Tian, T.J. Morris, A.P. Webster, Z. Yang, S. Beck, A. Feber, et al., ChAMP: updated methylation analysis pipeline for Illumina BeadChips. Bioinformatics (Oxford, England) 33(24), 3982–3984 (2017) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.W. Zhou, P.W. Laird, H. Shen, Comprehensive characterization, annotation and innovative use of Infinium DNA methylation BeadChip probes. Nucleic Acids Res. 45(4), e22 (2017) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.A.D. Kelly, J. Madzo, P. Madireddi, P. Kropf, C.R. Good, J. Jelinek, et al., Demethylator phenotypes in acute myeloid leukemia. Leukemia 32(10), 2178–2188 (2018) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.P. Du, X. Zhang, C.C. Huang, N. Jafari, W.A. Kibbe, L. Hou, et al., Comparison of Beta-value and M-value methods for quantifying methylation levels by microarray analysis. BMC Bioinf. 11, 587 (2010) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.K. Yoshihara, M. Shahmoradgoli, E. Martínez, R. Vegesna, H. Kim, W. Torres-Garcia, et al., Inferring tumour purity and stromal and immune cell admixture from expression data. Nat. Commun. 4, 2612 (2013) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.J. Jeschke, M. Bizet, C. Desmedt, E. Calonne, S. Dedeurwaerder, S. Garaud, et al., DNA methylation-based immune response signature improves patient diagnosis in multiple cancers. J. Clin. Invest. 127(8), 3090–3102 (2017) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.B. Phipson, J. Maksimovic, A. Oshlack, missMethyl: an R package for analyzing data from Illumina’s HumanMethylation450 platform. Bioinformatics (Oxford, England) 32(2), 286–288 (2016) [DOI] [PubMed] [Google Scholar]
  • 41.A. Liberzon, C. Birger, H. Thorvaldsdóttir, M. Ghandi, J.P. Mesirov, P. Tamayo, The Molecular Signatures Database (MSigDB) hallmark gene set collection. Cell Syst. 1(6), 417–425 (2015) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.M.E. Ritchie, B. Phipson, D. Wu, Y. Hu, C.W. Law, W. Shi, et al., limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. 43(7), e47 (2015) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.T. Wu, E. Hu, S. Xu, M. Chen, P. Guo, Z. Dai, et al., clusterProfiler 4.0: a universal enrichment tool for interpreting omics data. Innovation (Cambridge (Mass)) 2(3), 100141 (2021) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.Z. Xie, A. Bailey, M.V. Kuleshov, D.J.B. Clarke, J.E. Evangelista, S.L. Jenkins, et al., Gene Set Knowledge Discovery with Enrichr. Curr. Protocols 1(3), e90 (2021) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45.X. Lu, J. Meng, Y. Zhou, L. Jiang, F. Yan, MOVICS: an R package for multi-omics integration and visualization in cancer subtyping. Bioinformatics (Oxford, England) (2020) [DOI] [PubMed]
  • 46.A. Mayakonda, D.C. Lin, Y. Assenov, C. Plass, H.P. Koeffler, Maftools: efficient and comprehensive analysis of somatic variants in cancer. Genome Res. 28(11), 1747–1756 (2018) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.C.H. Mermel, S.E. Schumacher, B. Hill, M.L. Meyerson, R. Beroukhim, G. Getz, GISTIC2.0 facilitates sensitive and confident localization of the targets of focal somatic copy-number alteration in human cancers. Genome Biol. 12(4), R41 (2011) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48.A.G. Robertson, J. Kim, H. Al-Ahmadie, J. Bellmunt, G. Guo, A.D. Cherniack, et al., Comprehensive Molecular Characterization of Muscle-Invasive Bladder Cancer. Cell 171(3), 540–56.e25 (2017) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 49.X. Lu, J. Meng, L. Su, L. Jiang, H. Wang, J. Zhu, et al., Multi-omics consensus ensemble refines the classification of muscle-invasive bladder cancer with stratified prognosis, tumour microenvironment and distinct sensitivity to frontline therapies. Clin. Transl. Med. 11(12), e601 (2021) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 50.P. Geeleher, N.J. Cox, R.S. Huang, Clinical drug response can be predicted using baseline gene expression levels and in vitro drug sensitivity in cell lines. Genome Biol. 15(3), R47 (2014) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 51.X. Lu, L. Jiang, L. Zhang, Y. Zhu, W. Hu, J. Wang, et al., Immune Signature-Based Subtypes of Cervical Squamous Cell Carcinoma Tightly Associated with Human Papillomavirus Type 16 Expression, Molecular Features, and Clinical Outcome. Neoplasia 21(6), 591–601 (2019) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 52.N. McGranahan, A.J. Furness, R. Rosenthal, S. Ramskov, R. Lyngaa, S.K. Saini, et al., Clonal neoantigens elicit T cell immunoreactivity and sensitivity to immune checkpoint blockade. Science 351(6280), 1463–1469 (2016) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 53.G.P. Wagner, K. Kin, V.J. Lynch, Measurement of mRNA abundance using RNA-seq data: RPKM measure is inconsistent among samples. Theory Biosci (Theorie in den Biowissenschaften) 131(4), 281–285 (2012) [DOI] [PubMed] [Google Scholar]
  • 54.W. Timp, H.C. Bravo, O.G. McDonald, M. Goggins, C. Umbricht, M. Zeiger, et al., Large hypomethylated blocks as a universal defining epigenetic alteration in human solid tumors. Genome Med. 6(8), 61 (2014) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 55.A. Puccini, H.J. Lenz, J.L. Marshall, D. Arguello, D. Raghavan, W.M. Korn, et al., Impact of Patient Age on Molecular Alterations of Left-Sided Colorectal Tumors. Oncologist 24(3), 319–326 (2019) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 56.Q. Li, Q. Lai, C. He, Y. Fang, Q. Yan, Y. Zhang, et al., RUNX1 promotes tumour metastasis by activating the Wnt/β-catenin signalling pathway and EMT in colorectal cancer. J. Exp. Clin. Cancer Res. 38(1), 334 (2019) [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Material 1 (5.6MB, pdf)
Supplementary Material 2 (4MB, xlsx)

Data Availability Statement

The original contributions presented in the study are included in the article, further inquiries can be directed to the corresponding author.

Articles from Cellular Oncology are provided here courtesy of Springer

RESOURCES