Introduction
Despite recent advances in diagnostic and therapeutic tools, including surgical techniques, pancreatic ductal adenocarcinoma (PDAC) remains one of the most challenging diseases, with a 5-year survival rate of < 10%1. Eighty percent of patients are diagnosed with advanced-stage carcinomas, often beyond the reach of surgical intervention and typically categorized as American Joint Committee on Cancer stage III or more severe2. Even with aggressive treatments employing chemotherapy agents, such as gemcitabine or fluorouracil, the median survival is less than 1 year2.
Deoxyribonucleic acid (DNA) methylation is a key epigenetic mechanism in the development and progression of PDAC3. This modification involves the covalent addition of methyl groups to DNA, predominantly at cytosine–guanine dinucleotide sequences (CpG sites)4. Epigenetic alterations can modulate gene transcription and play pivotal roles in fundamental biological processes, including cellular differentiation, tumorigenesis, and immune response regulation, with particularly profound implications in cancer4. Among various methylation patterns, promoter methylation represents a functionally critical subset that targets gene promoter regions and typically results in transcriptional silencing5. This aberrant hypermethylation can inactivate tumor suppressor genes, thereby contributing to malignant transformation and tumor progression. In PDAC, promoter hypermethylation of key regulatory genes leads to their transcriptional downregulation, playing a significant role in tumorigenesis and disease progression6.
DNA methylation patterns of PDAC may serve as early detection7,8 and prognostic biomarkers9. Furthermore, understanding promoter methylation in PDAC is key to guiding therapeutic interventions10. Specifically, demethylating agents show promise in reversing some of these epigenetic alterations. As research on PDAC advances across various fields, the potential for personalized therapy becomes promising. However, a significant problem with current treatments is that they treat cancer as a homogeneous disease. PDAC exhibits significant intratumoral heterogeneity (ITH) in morphology11,12, genetics13,14, and gene expression15,16, contributing to treatment resistance.
Although research on intertumoral methylation profiles in PDAC has advanced17,18, few studies have focused on intratumoral promoter methylation profiles in PDAC19. Recent findings have suggested that DNA methylation patterns may be linked to specific molecular subtypes and clinical outcomes in PDAC20,21. However, the extent of intratumoral DNA methylation heterogeneity and its relationship to tumor evolution, histological features, and gene expression patterns remain largely unexplored. This study aimed to investigate promoter methylation heterogeneity within PDAC and elucidate its relationship with clinicopathological features. Focus was placed on potential functional promoters identified using the Illumina methylation platform22 based on the Encyclopedia of DNA Elements data23. By integrating methylation data from multiple tumor regions with histopathological analysis, as well as copy number variation (CNV) and gene expression profiling from The Cancer Genome Atlas (TCGA), this study aimed to provide a comprehensive characterization of the epigenetic landscape in PDAC and contribute to developing more accurate diagnostic tools and treatment strategies, paving the way for personalized epigenetic-based therapy and improved outcomes for patients with PDAC.
Materials and methods
Patient selection and clinicopathological review
All histological slides from 116 consecutive surgical cases of PDAC diagnosed at Kyorin University Hospital between 2018 and 2022 were reviewed. Intraductal papillary mucinous neoplasm-derived carcinomas were excluded from the cohort. The degrees of differentiation within these ductal adenocarcinomas were identified, and each area was classified as well-differentiated, moderately differentiated, poorly differentiated, or showing squamous cell differentiation according to the World Health Organization Classification of Tumors of the Digestive System11. Squamous differentiation was evaluated as dense, opaque eosinophilic polygonal tumor cells regardless of whether keratinization (keratohyalin granules and/or pearls) and/or intercellular bridges were observed.
Cases that met all three of the following criteria were selected: (1) histological invasion diameter ≥ 20 mm, (2) histological tumor cellularity ≥ 30%, and (3) adenocarcinoma with multiple differentiation in the tumor, each occupying at least 10 mm2. In six cases, six to nine lesions representing different histological regions were collected to reflect the overall histological characteristics of the whole tumor. Forty-four tumor and five normal pancreatic tissue samples were selected. Histology (hematoxylin-and-eosin slides) was reviewed by two gastrointestinal pathologists (K.K. and A.H.) who were independently blinded to clinical data. Clinicopathological information, including age, sex, drinking and smoking history, medical history, clinical stage, operative procedure, preoperative and postoperative treatment, and follow-up, was retrieved from medical records (Supplementary Table 1).
DNA extraction and array-based DNA methylation analysis
Fifteen consecutive 10 μm sections were cut from formalin-fixed, paraffin-embedded (FFPE) blocks containing each sample. After deparaffinization, xylene treatment, and immersion in water, the target areas of each sample were quickly collected using a trimming knife and a toothpick. The macrodissected tissue sections were collected in a microcentrifuge tube, and DNA was extracted using the QIAamp DNA FFPE Tissue Kit (Qiagen). DNA quality and quantity were examined using Nanodrop and Qubit. Genomic DNA was quantified at 500 ng, and bisulfite conversion was performed using the EZ DNA Methylation Kit (Zymo Research). Subsequent methylation analysis was performed using the Infinium Methylation EPIC BeadChip (Illumina) in strict accordance with the manufacturer’s recommended protocol. Raw signal intensities were obtained from IDAT files using R version 4.4.2 (http://www.r-project.org).
Preprocessing of DNA methylation data
The beta values (β values) for each sample were individually normalized using background correction and dye bias correction for both color channels. Probes associated with non-CpG sites, sex chromosomes, and single nucleotide polymorphisms (SNPs) and those with a mean β value of < 0.1 were excluded from the dataset. To analyze methylation in PDAC promoter regions, probe sites from cancer samples belonging to promoter regions were selected by identifying functional promoter regions labeled as either “Promoter Associated” or “Promoter_Associated_Cell_type_specific” in the “Regulatory Feature Group” from the official Illumina annotation list (https://support.illumina.com/array/array_kits/infinium-methylationepic-beadchip-kit/downloads.html).
Data filtering and dimensionality reduction with promoter probes
Following normalization, probes were filtered to retain only those located in promoter-associated regions. Additional filters were applied to remove probes on sex chromosomes and those affected by SNPs or cross-reactivity. Methylation values (β values) were transformed to M-values for statistical analysis. Dimensionality reduction was performed using the plotMDS function from limma package version 3.64.1 by using the top 1000 variable probes to produce a multidimensional scaling (MDS) plot, which is a distance-preserving method that aims to represent data points in a lower-dimensional space while preserving the pairwise distances.
Clustering number optimization
The Shi-Tomasi criterion and average silhouette width methods were employed to determine the optimal number of clusters for methylation data analysis. By using the NbClust package in R, we applied both methods to the top 1000 most variable probes and to all promoter probes. The Elbow method was used to plot the within-cluster sum of squares against different numbers of clusters (k = 1–5) and to identify the point wherein the addition of more clusters would yield diminishing returns. The Silhouette method calculated the average silhouette width for different numbers of clusters, with higher values indicating better clustering quality. The optimal number of clusters was determined by consensus between these approaches, with consideration to both statistical significance and biological relevance.
Hierarchical clustering analysis and heatmap visualization
Hierarchical clustering analysis was conducted using the same top 1,000 most variable probes as those employed in the MDS analysis. The pheatmap package in R was utilized to generate both dendrograms and heatmaps. Clustering was performed using the ward.D2 method, and β values were represented with a color-coded scale bar in the heatmaps. On the basis of the clustering optimization results, samples were classified into two groups: T1 and T2 (see the “Results” section for further details).
Intratumoral methylation heterogeneity
ITH was quantified using methylation β values (bVals) from the top 1,000 most variable and all promoter probes. For each patient, ITH was calculated as the mean pairwise distance: dist(1 − cor(bVals)), where cor(bVals) is the Pearson correlation coefficient of β values between sample pairs. This distance measures methylation profile dissimilarity, with higher values indicating greater heterogeneity.
Differential methylation analysis
Differential methylation analysis was conducted to identify differentially methylated probes (DMPs) between the two primary sample clusters by using all probes. A linear model was fitted to the M-values, and contrasts were specified to compare methylation differences between clusters. Significant DMPs were identified on the basis of log fold change and adjusted p-values (log2 fold change ≥ 1, adjusted p-value < 0.05). For differentially methylated region (DMR) analysis, the demarcate package24 was used to annotate and detect DMRs across the genome. Gene Ontology (GO) analyses were performed on the top 200, 500, and 1000 differentially methylated genes to explore their functional significance by using DAVID Bioinformatics (https://david.ncifcrf.gov/tools.jsp).
TCGA data acquisition
Methylation data from TCGA were obtained from the Genomic Data Commons Data Portal (https://portal.gdc.cancer.gov/) in the form of raw IDAT files generated using the Illumina Human Methylation 450 platform. Ribonucleic acid sequencing (RNA-seq) data were downloaded from the same portal to analyze the gene expression profiles. The subtypes of each case were analyzed according to Table S1 in the TCGA paper17. Clinical data, including patient survival information and prognostic variables, were downloaded from the Broad Institute’s Firehose portal (https://gdac.broadinstitute.org/). The clinical, methylation, and RNA-seq datasets were integrated to investigate the relationship between methylation patterns, gene expression, patient outcomes, and cancer subtypes.
TCGA methylation data analysis merged with the Kyorin dataset
The methylation β values from the TCGA and Kyorin datasets were merged to comprehensively compare methylation patterns. Both datasets were preprocessed by filtering for high-quality samples with purity levels > 0.2 and excluding probes with missing or low-quality measurements. Common CpG probes between the datasets were identified, thus resulting in 642 probes retained for subsequent analysis. MDS visualization was performed using these probes on the merged dataset.
Downstream methylation analysis with TCGA dataset
Hierarchical clustering analysis and heatmap visualization for the TCGA dataset were performed using the merged 657 probes, following the same approach as described above. The classification into two groups (T1 and T2) was also applied to the TCGA dataset. Additionally, the same method used for the Kyorin dataset was employed to identify DMRs in the TCGA dataset.
Integration of methylation and expression analyses with TCGA dataset
Based on the TCGA methylation profiles (T1 and T2 groups) identified as described above, Gene Set Enrichment Analysis (GSEA) was performed in conjunction with gene expression data. Additionally, differentially expressed genes between the two groups were identified using the Wilcoxon test. Furthermore, a comparative analysis of methylation profiles and gene expression profiles (Collisson classification, Moffitt classification, Bailey classification), as well as copy number alterations (CNAs), was conducted based on Table S1 from the TCGA paper17. An oncoprint was then generated to visualize the results.
MYC target gene data download
MYC target genes were identified using chromatin immunoprecipitation sequencing data obtained from ChIP-Atlas (https://chip-atlas.org)25. The top 1,000 MYC target genes, based on average scores within 1000 bp of the transcription start site, were included in this study.
CNV analysis
CNV analysis was performed using methylation data from EPIC arrays (Kyorin dataset) by using the Conumee 2.01 package26. The analysis pipeline involved creating CNV annotations on the basis of genomic bins and CpG probe locations, ensuring reliable coverage across the genome. Bins were iteratively merged with neighboring bins until the required minimum number of probes was reached; those exceeding the maximum size were filtered out. Genome annotations, including chromosomal sizes and gap regions, were sourced from the University of California, Santa Cruz, and only CpG probes located on autosomal chromosomes were retained. Exclusion regions were subtracted, and optional detailed regions were annotated for further investigation. The final set of bins was used to define genomic regions for CNV analysis.
Genomic segments were classified as gained or lost by using a cutoff of 0.1. For each patient, the fraction of altered regions was computed as the proportion of gained or lost segments relative to total regions to summarize the extent of CNAs in the tumor genome.
Phylogenetic analysis
The phylogenetic tree was constructed in R using the “ape” package27. First, a distance matrix was calculated to represent dissimilarities in DNA methylation data and to assess the correlation of values between samples. Then, an algorithm was applied to generate a balanced evolutionary tree by minimizing the total branch length, thus ensuring an efficient representation of the relationships among the samples. This tree provides a visual depiction of the similarities and differences between the samples. Finally, a graphical representation of the tree was created to facilitate interpretation and analysis.
Statistical analysis
Data analysis was performed using R software (version 4.4.2). Categorical variables were compared using Fisher’s exact test or the chi-square test, as appropriate. Overall survival was estimated from the date of initial diagnosis by using the Kaplan–Meier method, and survival differences were assessed using the log-rank test. All statistical tests were two sided. Statistical significance was defined as an adjusted p-value and/or false discovery rate (FDR) q-value < 0.05, unless otherwise specified in the Methods section. For all analyses, including both univariate and multivariate analyses, 95% confidence intervals were applied.
Ethics approval and informed consent
This study protocol was approved by the Ethics Committee of Kyorin University School of Medicine (Approval Code: R03-187). The study was conducted in accordance with the Ethical Guidelines for Life Sciences and Medical Research Involving Human Subjects (revised on March 27, 2023), the 1964 Declaration of Helsinki, and its subsequent amendments or equivalent ethical standards. Participants were provided with an opportunity to opt out of the study if they were deemed unable to provide informed consent, and detailed procedures for opting out were made available on a publicly accessible website. All participant data were anonymized and processed to ensure non-identifiability, with strict measures taken to protect privacy.
Results
Methylation profiling in pancreatic cancer
A comprehensive methylation analysis was performed on a cohort of six pancreatic cancer cases from Kyorin University. Guided by a detailed pathological evaluation of tumor regions, samples were collected from 44 cancerous and 5 normal sites by considering histological features across entire tumors (Fig. 1a, Supplementary Fig. 1, Supplementary Table 1). Following functional annotation and probe selection, 27,454 probes were identified as promoter probes for downstream analysis. Both the Elbow and Silhouette methods, which were applied to the top 1,000 most variable probes and to all promoter probes, indicated that 2 was the optimal number of clusters (Supplementary Fig. 2).
Fig. 1 [Images not available. See PDF.]
Distinct methylation profiles in the Kyorin dataset. (a) Schematic of multiregional sampling in PDAC cases (Case 3). (b) Clustering analysis and heat maps of TCGA dataset using the top 1,000 promoter probes. (c) MDS visualization based on the methylation patterns of the Kyorin samples. Normal samples and T1 group (T1 profile) samples are circled with dots. (d) Intratumoral methylation heterogeneity based on the mean pairwise distance using the top 1,000 promoter probes. The p values for each case represent the results of the Wilcoxon test compared with Case 5. (e) GO analysis of the hypomethylated regions in the T2 group.
Hierarchical clustering analysis revealed that all normal samples were grouped into one cluster, whereas the cancer samples were divided into two distinct clusters (Fig. 1b). One cancer cluster displayed methylation profiles that were relatively similar to normal pancreatic tissue (T1 profile), whereas the other exhibited markedly different methylation patterns (T2 profile). MDS analysis using the top 1,000 most variable promoter methylation probes further supported this clustering result (Fig. 1c).
Histological characteristics and heterogeneity of methylation profile in pancreatic cancer
From a histological differentiation perspective, samples exhibiting the T1 profile frequently included well-differentiated pancreatic ductal carcinomas (Fig. 1b; p < 0.001, well-differentiated vs. other histology, two-sided Fisher’s exact test). There were varying degrees of methylation profile heterogeneity across the cases. All six samples from Case 5 displayed T1 profiles, whereas four of eight samples from Case 1 exhibited T1 profiles. In Cases 3 and 4, only one sample showed a T1 profile, with the remaining samples displaying T2 profiles. Cases 2 and 6 consisted of samples with T2 profiles.
ITH analysis using the top 1,000 promoter probes supported these findings (Fig. 1d). Case 5 exhibited the least heterogeneity, and Case 6 showed a similar trend (p = 0.512 compared to Case 5). By contrast, Cases 1, 2, 3, and 4 demonstrated greater heterogeneity compared with Case 5 (p < 0.001 for Cases 1, 2, and 3; p = 0.037 for Case 4; all p-values were compared to Case 5).
Biological characteristics of methylation profiles
This grouping identified 1,877 DMRs between the T1 and T2 groups (Supplementary Table 2). GO analysis of the top 200 regions with statistically significant differences in methylation β values between the T1 and T2 groups revealed that genes associated with “regulation of DNA-templated transcription” (FDR q = 2.99 × 10−10) and “regulation of transcription by ribonucleic acid polymerase II” (FDR q = 1.30 × 10−5) exhibited decreased methylation in the T2 group than in the T1 group (Fig. 1e, Supplementary Table 4). These characteristics were consistently observed when the top 100, top 500, and all 660 regions were used as inputs (Supplementary Fig. 4). By contrast, the biological features associated with the genes with increased methylation in the T2 group were dependent on the number of regions analyzed (Supplementary Fig. 5). In the top 200 regions, genes related to “negative regulation of transcription by RNA polymerase II” and “negative regulation of miRNA transcription” showed increased methylation in the T2 group relative to the T1 group, but they did not reach significance (both FDR q > 0.05) (Supplementary Table 5).
Reproducibility of methylation profiles using TCGA data
After confirming no significant difference in variability between the Kyorin and TCGA datasets via MDS analysis (Fig. 2a), we reanalyzed the TCGA data by using 642 probes common to both the 1000 probes selected from the Kyorin cohort (EPIC array) and those available on the 450 K array used in the TCGA methylation analysis (Fig. 2b). On the basis of the optimal cluster number determined by the Elbow and Silhouette methods (Supplementary Fig. 6), the TCGA samples were also classified into two groups (Fig. 2b): T1 showed methylation profiles similar to those of normal pancreatic tissue, and T2 displayed distinct methylation patterns.
Fig. 2 [Images not available. See PDF.]
Distinct methylation profiles in TCGA dataset. (a) MDS visualization based on the methylation patterns of all Kyorin and TCGA samples. (b) Clustering analysis and heat maps of TCGA dataset using overlapping probes between Kyorin and TCGA datasets. (c) GO analysis of the hypomethylated regions in the T2 group. (d) Kaplan–Meier analysis of the overall and progression-free survival comparing the T1 and T2 groups.
Biological and clinical characteristics of methylation profiles in TCGA cohort
DMR analysis between the T1 and T2 groups in the TCGA dataset identified 7,722 regions with variable methylation levels (Supplementary Table 6). GO analysis of the top 200 regions with the most statistically significant differences in methylation β values revealed that genes involved in “regulation of transcription by RNA polymerase II” and “positive regulation of transcription by RNA polymerase II” exhibited lower methylation levels in the T2 group than in the T1 group (FDR q = 3.29 × 10−21 and q = 1.23 × 10−14, respectively) (Fig. 2c, Supplementary Table 7). These trends remained consistent when the top 500, 1000, or 2000 regions were used as input (Supplementary Fig. 6). By contrast, genes with a tendency toward hypermethylation in the T2 group relative to the T1 group did not show enrichment for any distinct biological processes (FDR q = 1.0) (Supplementary Table 8). Various terms were identified when analyzing the top 500, 1000, and 2000 regions, but none reached statistical significance (Supplementary Fig. 7).
Kaplan–Meier survival analysis (Fig. 2d,e) demonstrated that patients in the T2 group (median = 12.8 months) had significantly shorter disease-free survival (DFS) than those in the T1 group (median not reached; survival rate > 50% at the longest follow-up; p = 0.04) (Fig. 2e). A similar trend was observed in multivariate analysis, although it did not reach statistical significance (p = 0.051) (Supplementary Fig. 8).
Association between methylation and expression profiles.
GSEA was employed to investigate the relationship between methylation profiles (T1 and T2) and expression profiles using TCGA data. Results revealed significant upregulation of DNA repair-related and MYC target genes in the T2 group (FDR q = 0 and 0.009, Fig. 3a, Supplementary Table 9). In contrast, the T1 group showed increased pancreatic β-cell-related gene expression, although this difference did not reach statistical significance (FDR q = 0.77, Fig. 3b, Supplementary Table 10). Differential expression analysis between the T1 and T2 groups identified 2,228 significantly differentially expressed genes (DEGs), with 192 of the top 1,000 MYC target genes showing upregulation in the T2 group (Supplementary Table 11, Fig. 3c) though MYC expression was not different between the two groups (Supplementary Fig. 9). GO analysis of these DEGs demonstrated the upregulation of RNA splicing-related genes in T2 compared to T1 (Fig. 3d), whereas genes associated with “the disruption of the cell wall of another organism” were downregulated (Fig. 3e). Comparison to previously established expression subtypes revealed that Collison’s quasi-mesenchymal and Bailey’s squamous types were exclusively observed in the T2 group, showing a significant correlation with methylation profiles (p < 0.01). Moffitt’s classification showed a higher proportion of basal-like tumors in T2, although this difference was not statistically significant (p = 0.293). T2 tumors exhibited more pronounced CNAs, but this trend also lacked statistical significance (p = 0.376). Results suggested that the distinct methylation profiles (T1 and T2) are associated with specific gene expression patterns and molecular subtypes in pancreatic cancer.
Fig. 3 [Images not available. See PDF.]
Expressional pattern of distinct methylation profiles in TCGA dataset. (a,b) GSEA analysis showing enriched pathways in the T2 (a) and T1 (b) groups. (c) Differential gene expression of MYC target genes (Top 1,000) between T1 and T2 groups. (d) GO analysis of upregulated genes in the T2 group. (e) GO analysis of downregulated genes in the T2 group. (f) Association between methylation profiles, expression subtypes, and copy number variants.
Intratumoral evolutionary diversity of methylation profiles
The characteristics of PDAC methylation were investigated from an evolutionary perspective, with a focus on all promoter regions. An analysis of methylation heterogeneity based on inter-sample relationships across all promoters revealed that Cases 1, 2, 3, and 4 exhibited greater heterogeneity than Cases 5 and 6; this finding is consistent with the results obtained using both the top 1,000 and all promoter regions as input (Supplementary Fig. 10a and Fig. 1d). From the perspective of CNAs, Cases 1, 3, 4, and 6 exhibited greater genomic changes, whereas Cases 2 and 5 showed relatively fewer alterations (Supplementary Fig. 10b).
In the phylogenetic analysis, Case 3 illustrates a case where CNA events and morphological phenotypes align closely with the methylation-based evolutionary tree (Fig. 4a). The earliest sample, Sample 1, displays a T1 profile characterized by minimal CNA events and a moderately differentiated morphology; therefore, it serves as the baseline for the tumor’s development (Fig. 4b). As the tumor progresses, Sample 2 shifts to a T2 profile, which is marked by the simultaneous loss of the CDKN2A tumor suppressor gene and the amplification of the 9p chromosomal region, accompanied by a transition to poorly differentiated morphology (Fig. 4b–d). Thereafter, Samples 4 and 6, which also exhibit T2 profiles, show further evolution with the additional gain of the KRAS oncogene, coinciding with a very poorly differentiated squamous morphology.
Fig. 4 [Images not available. See PDF.]
Intratumoral evolutionary diversity of methylation profiles. (a) Phylogenetic trees based on methylation patterns in a pancreatic adenosquamous carcinoma (Case 3). (b) Representative histological images of the tubular (samples 1 and 2) and squamous (samples 4 and 6) components. (c) CNAs of the entire tumor. (d) CNAs in individual samples (samples 1, 2, 4, and 6). Chromosome 9p gain (red arrow), CDKN2A/B copy number loss (blue arrow), and KRAS amplification (yellow arrow) are indicated.
Similarly, Cases 1 and 4 exhibit evolutionary patterns akin to Case 3 (Supplementary Figs. 11 and 12). In Case 4, only Sample 1, which is positioned near the normal tissue on the phylogenetic tree, displays a T1 profile. The remaining seven samples exhibit T2 profiles and consistently harbor genomic alterations such as chrosome 8q gain with MYC amplification and chromosome 9p loss involving CDKN2A/B deletion. Although Case 1 does not include normal tissue, Samples 1, 2, 4, and 6, which are characterized by well-differentiated histology and T1 methylation profiles, cluster closely together on the phylogenetic tree and show minimal CNAs. By contrast, the remaining samples exhibit T2 profiles and are positioned distantly from the T1 group on the tree. Notably, chromosome 7p gain including EGFR amplification is observed in Samples 3, 5, and 8.
Case 2 provides a different perspective. All tumor samples displayed T2 profiles, associated with moderately to poorly differentiated tumors (Supplementary Fig. 13a). Although the CNA events were less prominent in number and amplitude than Case 6 (see the next part), the phylogenetic tree revealed longer branch lengths and greater diversity among the samples, thus indicating increased evolutionary divergence (Supplementary Fig. 13b,c).
Intratumoral limited heterogeneity of methylation profiles
Compared with the previous cases (Case 1, 3, and 4), Case 5 displayed a distinct profile. All tumor samples in this case exhibited T1 profiles, which are associated with well-differentiated tumors, and showed remarkably similar histological appearances (Fig. 5a,b). In the phylogenetic tree, these samples clustered tightly together, separated from normal samples, and had very short internal and external branch lengths. This tight clustering reflects a high degree of similarity among the tumor samples, consistent with the presence of very few CNA events. The limited genetic alterations likely contribute to the stable morphology observed in these well-differentiated tumors, as demonstrated by fewer CNAs overall and in individual samples (Fig. 5b–d).
Fig. 5 [Images not available. See PDF.]
Intratumoral limited heterogeneity in methylation profiles. (a) Phylogenetic trees based on methylation patterns in a PDAC (Case 5). (b) Representative histologic images of samples 1 and 6. (c) CNAs of the entire tumor. (d) CNAs in individual samples (samples 1 and 6). Chromosome 9p gain (red arrow) is indicated40.
Case 6 also exhibited tight clustering of tumor samples in the phylogenetic tree, with short branch lengths indicating epigenetic similarity (Supplementary Fig. 14). However, unlike Case 5, all tumor samples in Case 6 displayed T2 profiles, which are associated with moderately to poorly differentiated tumors.
Discussion
This comprehensive analysis of DNA methylation in PDAC revealed significant ITH, thus providing novel insights into the epigenetic landscape of this aggressive malignancy. The identification of two distinct methylation profiles (T1 and T2) with varying degrees of heterogeneity across cases underscores the complex nature of PDAC epigenetics and its potential implications for tumor biology and clinical outcomes.
The correlation between methylation profiles and gene expression patterns, particularly the upregulation of DNA repair-related genes and MYC target genes in the T2 group, provides insight into the functional effect of epigenetic alterations in PDAC. In both the Kyorin and TCGA cohorts, the T2 group showed methylation profiles that are distinct from normal pancreatic tissue. GO analysis of DMRs revealed reduced methylation in the genes involved in transcriptional regulation, which is consistent with a previous study28. GSEA further demonstrated the increased expression of DNA repair genes and MYC targets, consistent with previous findings on the role of MYC signaling in PDAC progression and therapeutic resistance29. Although MYC expression itself remained unchanged, the elevated expression of its downstream targets suggests regulation through post-translational modifications or chromatin-level mechanisms. MYC activity is known to be influenced by phosphorylation and isomerization, which affect its nuclear localization30, and by the surrounding chromatin environment, including histone modifications such as H3K4me331,32. These mechanisms may allow MYC to exert transcriptional effects even without changes in its expression level, thus potentially explaining the observed upregulation of its target genes in the T2 group.
The T2 group also showed significant overlap with aggressive molecular subtypes of PDAC, including Collisson’s quasi-mesenchymal and Bailey’s squamous types, and was associated with shorter DFS, consistent with findings reported in Bailey’s study32. These results suggest that promoter-based methylation profiling can effectively stratify PDAC into biologically and clinically meaningful subgroups. The reanalysis of TCGA data in this study adds further support to the molecular relevance of methylation profiles, which was not fully addressed in earlier studies17.
This study highlights the value of integrating methylation-derived CNA profiles with phylogenetic reconstruction to understand PDAC evolution. In Cases 1, 3, and 4, CNA patterns, histological features, and phylogenetic tree structures were well aligned, thus suggesting that both epigenetic and genetic alterations contribute to tumor progression. T1 profile tumors with well-differentiated morphology tended to appear early, whereas T2 profile tumors with poor differentiation emerged later and showed greater genomic instability. The observed shift from T1 to T2 methylation profiles during tumor evolution, which was accompanied by poorly differentiated morphology and squamous features, is consistent with the findings of recent studies demonstrating the role of epigenetic changes in promoting PDAC progression and ITH33. However, similar phylogenetic clustering did not always reflect similar biological characteristics. For example, Case 5 showed minimal CNAs and stable morphology, whereas Case 6 exhibited extensive CNAs and poorly differentiated features despite tight clustering. Case 2 displayed relatively few CNAs but greater phylogenetic divergence; this phenomenon is potentially driven by specific events such as ATM deletion and MYC/KRAS amplification. These findings underscore the importance of considering both epigenetic and genetic factors when evaluating tumor heterogeneity, as indicated in a previous paper34.
As shown in this study, ITH in DNA methylation is increasingly recognized as a defining feature of many cancers35, 36–37. The variation in methylation heterogeneity observed among PDAC cases, from clearly separated clusters to very limited differences, indicates that numerous factors shape the epigenetic profile of this disease. Previous studies have connected microbial communities with DNA methylation in gastrointestinal cancers38, but the link between the microbiota and methylation patterns in PDAC remains unclear and requires further investigation. Although this study did not find evidence supporting the association between the gut microbiota and PDAC methylation patterns, earlier research has shown that gut microbes influence host epigenetics primarily through the production of metabolic compounds39. These observations suggest that shifts in the tumor microenvironment, particularly those caused by obstructive jaundice, may influence methylation profiles, a hypothesis we plan to explore in future studies. Additionally, the effect of chemotherapy on methylation heterogeneity, particularly in cases treated with neoadjuvant therapy, supports the concept that treatment can trigger epigenetic changes40, thus highlighting the dynamic nature of the epigenetic landscape. These findings suggest that methylation profiling has potential not only for understanding tumor development but also as a valuable tool for classifying tumors and predicting clinical outcomes in PDAC41,42. The methylation profiles identified in this study may offer promising opportunities for predicting and evaluating the efficacy of epigenetic-modifying drugs such as DNA methyltransferase (DNMT) inhibitors, potentially enabling more personalized therapeutic approaches, as has been suggested in recent literature regarding other cancer types43,44. The remarkable similarity between phylogenetic trees based on methylation and those based on CNAs suggests that epigenetic and genomic changes may follow parallel evolutionary trajectories in certain contexts; this underscores the value of integrating these data types to obtain comprehensive insights into the molecular mechanisms driving tumor progression and informing future therapeutic strategies.
Although this study yielded valuable insights, a few limitations must be acknowledged. First, the small number of eligible cases reflects the difficulty of delineating regions with clear histological differentiation in PDAC, i.e., the cohort may not fully represent the broader patient population. Additionally, the retrospective nature and short survival in PDAC limit comprehensive survival analysis, as evidenced by variable outcomes within methylation subgroups (e.g., Case 2: more than 50 months survival vs. Case 6: short disease-free survival, both T2 profiles). Future investigations will include smaller tissue regions to increase sample numbers and allow more robust prognostic analysis. Second, gene expression was assessed only in the TCGA cohort; we are now preparing transcriptomic profiling of representative Kyorin cases that display either homogeneous or heterogeneous methylation patterns. Third, the use of different methylation platforms across cohorts complicates direct comparison. To address this issue, we plan to regenerate or reprocess methylation data for every cohort on a single platform such as the Infinium MethylationEPIC array to create a unified dataset that will strengthen cross cohort validation.
In conclusion, this study comprehensively characterizes DNA methylation heterogeneity in PDAC, thus revealing distinct methylation profiles with prognostic and biological significance. The observed associations between methylation patterns, gene expression, and molecular subtypes underscore the importance of epigenetic mechanisms in shaping PDAC biology. Future research should focus on elucidating the functional consequences of these methylation changes and exploring their potential as biomarkers for patient stratification and therapeutic targeting. A deeper understanding of the epigenetic landscape in PDAC may pave the way for more effective diagnostic, prognostic, and therapeutic approaches to managing this aggressive malignancy.
Acknowledgements
We thank Kazunari Tanabe, Ayumi Sumiishi, Namiko Kondo, and Kaoruko Kojima for their excellent technical assistance. We are also grateful to Saori Funakoshi for her support with data collection.
Author contributions
Study concept and design: A.H. Collection of the clinical samples and data: K.K. K.O., H.O. and Y.S. Collection of the pathologic data: K.K., K.N., T.S., J.S. and A.H. Analysis of methylation and sequencing data: K.K, Y-J.H, K.S. and A.H. Analysis of clinicopathological data and interpretation: K.K. and A.H. Manuscript writing: K.K., J.S. and A.H. All authors have read and approved the final paper. All other authors declare that they have no conflicts of interest in relation to this work.
Funding
This work was supported in part by a Grant-in-Aid for Research Activity Start-up (20K22821), the Yasuda Medical Foundation, and the Kyorin University Collaborative Research Project Grant and the Mochida Memorial Foundation for Medical and Pharmaceutical Research.
Data availability
The datasets used and/or analyzed during this study are available from the corresponding author on reasonable request.
Declarations
Competing interests
The authors declare no competing interests.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
1. Siegel, R. L., Miller, K. D., Fuchs, H. E., Jemal, A. & Cancer Statistics CA Cancer J. Clin.71, 7–33. https://doi.org/10.3322/caac.21654 (2021).
2. Park, W., Chawla, A. & O’Reilly, E. M. Pancreat. Cancer JAMA326, https://doi.org/10.1001/jama.2021.13027 (2021).
3. Roy, S., Singh, A. P. & Gupta, D. Unsupervised subtyping and methylation landscape of pancreatic ductal adenocarcinoma. Heliyon. 7https://doi.org/10.1016/j.heliyon.2021.e06000 (2021).
4. Bird, A. DNA methylation patterns and epigenetic memory. Genes Dev.; 2002; 16, pp. 6-21.1:CAS:528:DC%2BD38Xos1ehsQ%3D%3D [DOI: https://dx.doi.org/10.1101/gad.947102] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/11782440]
5. Jones, PA; Baylin, SB. The fundamental role of epigenetic events in cancer. Nat. Rev. Genet.; 2002; 3, pp. 415-428.1:CAS:528:DC%2BD38XksFOmsrw%3D [DOI: https://dx.doi.org/10.1038/nrg816] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/12042769]
6. Vincent, A et al. Genome-wide analysis of promoter methylation associated with gene expression profile in pancreatic adenocarcinoma. Clin. Cancer Res.; 2011; 17, pp. 4341-4354.1:CAS:528:DC%2BC3MXos1Smt7g%3D [DOI: https://dx.doi.org/10.1158/1078-0432.Ccr-10-3431] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/21610144][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3131423]
7. Wu, H. et al. Noninvasive detection of pancreatic ductal adenocarcinoma using the methylation signature of Circulating tumour DNA. BMC Med.20https://doi.org/10.1186/s12916-022-02647-z (2022).
8. Majumder, S et al. Methylated DNA in pancreatic juice distinguishes patients with pancreatic cancer from controls. Clin. Gastroenterol. Hepatol.; 2020; 18, pp. 676-683e673.1:CAS:528:DC%2BC1MXhsFyqtb7M [DOI: https://dx.doi.org/10.1016/j.cgh.2019.07.017] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/31323382]
9. Gu, M et al. A novel methylation signature predicts inferior outcome of patients with PDAC. Aging (Albany NY); 2021; 13, pp. 2851-2863.1:CAS:528:DC%2BB3MXhsFWhsL7M [DOI: https://dx.doi.org/10.18632/aging.202347] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/33550277]
10. Chiappinelli, KB et al. Inhibiting DNA methylation causes an interferon response in cancer via DsRNA including endogenous retroviruses. Cell; 2015; 162, pp. 974-986.1:CAS:528:DC%2BC2MXhsVensLnJ [DOI: https://dx.doi.org/10.1016/j.cell.2015.07.011] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/26317466][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4556003]
11. WHO Classification of Tumours Editorial Board 5th Ed. WHO Classification of Tumours of the Digestive System., vol. 1 (IARC, 2019).
12. Taherian, M., Wang, H. & Wang, H. Pancreatic ductal adenocarcinoma: Molecular pathology and predictive biomarkers. Cells. 11. https://doi.org/10.3390/cells11193068 (2022).
13. Hayashi, A et al. A unifying paradigm for transcriptional heterogeneity and squamous features in pancreatic ductal adenocarcinoma. Nat. Cancer; 2020; 1, pp. 59-74.1:CAS:528:DC%2BB38XjsFeitrg%3D [DOI: https://dx.doi.org/10.1038/s43018-019-0010-1] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/35118421][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8809486]
14. Chan-Seng-Yue, M et al. Transcription phenotypes of pancreatic cancer are driven by genomic events during tumor evolution. Nat. Genet.; 2020; 52, pp. 231-240.1:CAS:528:DC%2BB3cXotFGksg%3D%3D [DOI: https://dx.doi.org/10.1038/s41588-019-0566-9] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/31932696]
15. Campbell, PJ et al. The patterns and dynamics of genomic instability in metastatic pancreatic cancer. Nature; 2010; 467, pp. 1109-1113.2010Natur.467.1109C1:CAS:528:DC%2BC3cXhtlCjt7%2FP [DOI: https://dx.doi.org/10.1038/nature09460] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/20981101][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3137369]
16. Yachida, S et al. Distant metastasis occurs late during the genetic evolution of pancreatic cancer. Nature; 2010; 467, pp. 1114-1117.2010Natur.467.1114Y1:CAS:528:DC%2BC3cXhtlCjtrvI [DOI: https://dx.doi.org/10.1038/nature09515] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/20981102][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3148940]
17. Raphael, BJ et al. Integrated genomic characterization of pancreatic ductal adenocarcinoma. Cancer Cell.; 2017; 32, pp. 185-203e113.1:CAS:528:DC%2BC2sXhtlClsbbN [DOI: https://dx.doi.org/10.1016/j.ccell.2017.07.007]
18. Espinet, E et al. Aggressive PDACs show hypomethylation of repetitive elements and the execution of an intrinsic IFN program linked to a ductal cell of origin. Cancer Discov; 2021; 11, pp. 638-659.1:CAS:528:DC%2BB3MXosV2hsLg%3D [DOI: https://dx.doi.org/10.1158/2159-8290.CD-20-1202] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/33060108]
19. Liu, X. et al. Multi-omics analysis of intra‐tumoural and inter‐tumoural heterogeneity in pancreatic ductal adenocarcinoma. Clin. Transl. Med.12https://doi.org/10.1002/ctm2.670 (2022).
20. Fraunhoffer, N. A. et al. Multi-omics data integration and modeling unravels new mechanisms for pancreatic cancer and improves prognostic prediction. NPJ Precis. Oncol.6https://doi.org/10.1038/s41698-022-00299-z (2022).
21. Li, X. et al. Classification and prognosis analysis of pancreatic cancer based on DNA methylation profile and clinical information. Genes (Basel). https://doi.org/10.3390/genes13101913 (2022).
22. Infinium Methylation EPIC Manifest Column Headings. https://support.illumina.com/content/dam/illumina-support/documents/downloads/productfiles/methylationepic/infinium-methylationepic-manifest-column-headings.pdf
23. Consortium, EP et al. Expanded encyclopaedias of DNA elements in the human and mouse genomes. Nature; 2020; 583, pp. 699-710.2020Natur.583.699E1:CAS:528:DC%2BB3cXhsVygtLnK [DOI: https://dx.doi.org/10.1038/s41586-020-2493-4]
24. Peters, TJ et al. De Novo identification of differentially methylated regions in the human genome. Epigenet. Chromatin; 2015; 8, 6.1:CAS:528:DC%2BC2MXjt1Cit7k%3D [DOI: https://dx.doi.org/10.1186/1756-8935-8-6]
25. Oki, S. et al. ChIP-Atlas: a data-mining suite powered by full integration of public ChIP-seq data. EMBO Rep.19https://doi.org/10.15252/embr.201846255 (2018).
26. Daenekas, B. et al. Conumee 2.0: enhanced copy-number variation analysis from DNA methylation arrays for humans and mice. Bioinformatics. 40https://doi.org/10.1093/bioinformatics/btae029 (2024).
27. Janzen, T; Etienne, RS. Phylogenetic tree statistics: A systematic overview using the new R package ‘treestats’. Mol. Phylogenet. Evol.; 2024; 200, 108168.1:CAS:528:DC%2BB2cXhslGitrzM [DOI: https://dx.doi.org/10.1016/j.ympev.2024.108168] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/39117295]
28. Wang, SS et al. Whole-genome bisulfite sequencing identifies stage- and subtype-specific DNA methylation signatures in pancreatic cancer. iScience; 2024; 27, 109414.2024iSci..27j9414W1:CAS:528:DC%2BB2cXnslKjsb8%3D [DOI: https://dx.doi.org/10.1016/j.isci.2024.109414] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/38532888][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10963232]
29. Doha, Z. O. & Sears, R. C. Unraveling MYC’s role in orchestrating tumor intrinsic and tumor microenvironment interactions driving tumorigenesis and drug resistance. Pathophysiology. 30, 400–419. https://doi.org/10.3390/pathophysiology30030031 (2023).
30. Su, Y et al. Post-translational modification localizes MYC to the nuclear pore basket to regulate a subset of target genes involved in cellular responses to environmental signals. Genes Dev.; 2018; 32, pp. 1398-1419.1:CAS:528:DC%2BC1cXitlOrtbjJ [DOI: https://dx.doi.org/10.1101/gad.314377.118] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/30366908][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6217735]
31. Yang, Z et al. Hijacking a key chromatin modulator creates epigenetic vulnerability for MYC-driven cancer. J. Clin. Investig.; 2018; 128, pp. 3605-3618. [DOI: https://dx.doi.org/10.1172/JCI97072] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/29870403][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6063490]
32. Bailey, P et al. Genomic analyses identify molecular subtypes of pancreatic cancer. Nature; 2016; 531, pp. 47-52.2016Natur.531..53B1:CAS:528:DC%2BC28XjtFOisr0%3D [DOI: https://dx.doi.org/10.1038/nature16965] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/26909576]
33. Szymonski, K et al. Spatial recognition and semi-quantification of epigenetic events in pancreatic cancer subtypes with multiplexed molecular imaging and machine learning. Sci. Rep.; 2025; 15, 6518.1:CAS:528:DC%2BB2MXkvVKmtbk%3D [DOI: https://dx.doi.org/10.1038/s41598-025-90087-z] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/39987295][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11846859]
34. Guo, M; Peng, Y; Gao, A; Du, C; Herman, JG. Epigenetic heterogeneity in cancer. Biomark. Res.; 2019; 7, 23. [DOI: https://dx.doi.org/10.1186/s40364-019-0174-y] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/31695915][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6824025]
35. Ferreyra Vega, S et al. Spatial heterogeneity in DNA methylation and chromosomal alterations in diffuse gliomas and meningiomas. Mod. Pathol.; 2022; 35, pp. 1551-1561.1:CAS:528:DC%2BB38XhslGgs73L [DOI: https://dx.doi.org/10.1038/s41379-022-01113-8] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/35701666][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9596370]
36. Lv, X., Mao, Z., Sun, X. & Liu, B. Intratumoral heterogeneity in lung cancer. Cancers (Basel). 15https://doi.org/10.3390/cancers15102709 (2023).
37. Gallon, J et al. DNA methylation landscapes of prostate cancer brain metastasis are shaped by early driver genetic alterations. Cancer Res.; 2023; 83, pp. 1203-1213.1:CAS:528:DC%2BB3sXpt1Cjur0%3D [DOI: https://dx.doi.org/10.1158/0008-5472.CAN-22-2236] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/36749655][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10102845]
38. Wang, Q et al. Multi-omic profiling reveals associations between the gut mucosal microbiome, the metabolome, and host DNA methylation associated gene expression in patients with colorectal cancer. BMC Microbiol.; 2020; 20, 83.1:CAS:528:DC%2BB3cXnslGku7s%3D [DOI: https://dx.doi.org/10.1186/s12866-020-01762-2] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/32321427][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7178946]
39. Krautkramer, KA et al. Diet-microbiota interactions mediate global epigenetic programming in multiple host tissues. Mol. Cell.; 2016; 64, pp. 982-992.1:CAS:528:DC%2BC28XhvFGgtL7P [DOI: https://dx.doi.org/10.1016/j.molcel.2016.10.025] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/27889451][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5227652]
40. Menghi, F. et al. Genomic and epigenomic BRCA alterations predict adaptive resistance and response to platinum-based therapy in patients with triple-negative breast and ovarian carcinomas. Sci. Transl. Med.14, eabn1926. https://doi.org/10.1126/scitranslmed.abn1926 (2022).
41. Hatziapostolou, M et al. Promoter methylation leads to hepatocyte nuclear factor 4A loss and pancreatic cancer aggressiveness. Gastro Hep Adv.; 2024; 3, pp. 687-702. [DOI: https://dx.doi.org/10.1016/j.gastha.2024.04.005] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/39165427][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11330932]
42. Sato, N et al. Aberrant methylation of reprimo correlates with genetic instability and predicts poor prognosis in pancreatic ductal adenocarcinoma. Cancer; 2006; 107, pp. 251-257.1:CAS:528:DC%2BD28Xnsl2mtL4%3D [DOI: https://dx.doi.org/10.1002/cncr.21977] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/16752411]
43. Cabezón, M et al. Different methylation signatures at diagnosis in patients with high-risk myelodysplastic syndromes and secondary acute myeloid leukemia predict Azacitidine response and longer survival. Clin. Epigenet.; 2021; 13, 9.1:CAS:528:DC%2BB3MXmtFSlt7w%3D [DOI: https://dx.doi.org/10.1186/s13148-021-01002-y]
44. Harms, PW et al. Viral status predicts patterns of genome methylation and decitabine response in Merkel cell carcinoma. J. Investig. Dermatol.; 2022; 142, pp. 641-652.1:CAS:528:DC%2BB3MXitFSls7fF [DOI: https://dx.doi.org/10.1016/j.jid.2021.07.173] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/34474081]
You have requested "on-the-fly" machine translation of selected content from our databases. This functionality is provided solely for your convenience and is in no way intended to replace human translation. Show full disclaimer
Neither ProQuest nor its licensors make any representations or warranties with respect to the translations. The translations are automatically generated "AS IS" and "AS AVAILABLE" and are not retained in our systems. PROQUEST AND ITS LICENSORS SPECIFICALLY DISCLAIM ANY AND ALL EXPRESS OR IMPLIED WARRANTIES, INCLUDING WITHOUT LIMITATION, ANY WARRANTIES FOR AVAILABILITY, ACCURACY, TIMELINESS, COMPLETENESS, NON-INFRINGMENT, MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Your use of the translations is subject to all use restrictions contained in your Electronic Products License Agreement and by using the translation functionality you agree to forgo any and all claims against ProQuest or its licensors for your use of the translation functionality and any output derived there from. Hide full disclaimer
© The Author(s) 2025. This work is published under http://creativecommons.org/licenses/by-nc-nd/4.0/ (the "License"). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.
Abstract
Pancreatic ductal adenocarcinoma (PDAC) remains a lethal malignancy with poor prognosis. We investigated intratumoral deoxyribonucleic acid methylation heterogeneity by analyzing 44 tumor samples and 5 normal samples from 6 cases of PDAC by using high-resolution methylation arrays. Two distinct methylation profiles were identified: T1, which is similar to normal pancreatic tissue and is associated with well-differentiated histology, and T2, which is significantly different from normal tissue and is linked to poorly differentiated morphology and squamous features. Validation using The Cancer Genome Atlas (TCGA) confirmed these profiles and revealed the association of T2 with shorter disease-free survival (p = 0.04). Differentially methylated region analysis identified the substantial hypomethylation of transcription regulation genes in T2 profiles (false discovery rate [FDR] q < 0.001). Gene set enrichment analysis with TCGA gene expression data demonstrated the upregulation of DNA repair and MYC target genes in T2 samples (FDR q < 0.001). Phylogenetic analysis with our multi-sampling dataset suggested an evolutionary trajectory from T1 to T2 profiles coinciding with aggressive phenotypes and increased genomic instability. Cases exhibited varying degrees of intratumoral heterogeneity from distinctly separated clusters to minimal differences. This comprehensive characterization of the epigenetic landscape of PDAC provides insights into tumor evolution and heterogeneity with potential implications for patient stratification and the development of epigenetic-based diagnostic and therapeutic strategies.
You have requested "on-the-fly" machine translation of selected content from our databases. This functionality is provided solely for your convenience and is in no way intended to replace human translation. Show full disclaimer
Neither ProQuest nor its licensors make any representations or warranties with respect to the translations. The translations are automatically generated "AS IS" and "AS AVAILABLE" and are not retained in our systems. PROQUEST AND ITS LICENSORS SPECIFICALLY DISCLAIM ANY AND ALL EXPRESS OR IMPLIED WARRANTIES, INCLUDING WITHOUT LIMITATION, ANY WARRANTIES FOR AVAILABILITY, ACCURACY, TIMELINESS, COMPLETENESS, NON-INFRINGMENT, MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Your use of the translations is subject to all use restrictions contained in your Electronic Products License Agreement and by using the translation functionality you agree to forgo any and all claims against ProQuest or its licensors for your use of the translation functionality and any output derived there from. Hide full disclaimer
Details
1 Department of Pathology, Kyorin University School of Medicine, 6-20-2 Shinkawa, Mitaka, 181-8611, Tokyo, Japan (ROR: https://ror.org/0188yz413) (GRID: grid.411205.3) (ISNI: 0000 0000 9340 2869)
2 Cancer Biology and Genetics Program, Sloan Kettering Institute, Memorial Sloan Kettering Cancer Center, New York, NY, USA (ROR: https://ror.org/02yrq0923) (GRID: grid.51462.34) (ISNI: 0000 0001 2171 9952)
3 Department of Laboratory Medicine, Kyorin University School of Medicine, Tokyo, Japan (ROR: https://ror.org/0188yz413) (GRID: grid.411205.3) (ISNI: 0000 0000 9340 2869)
4 Department of Surgery, Kyorin University School of Medicine, Tokyo, Japan (ROR: https://ror.org/0188yz413) (GRID: grid.411205.3) (ISNI: 0000 0000 9340 2869)