Every cell in the body carries a kind of biological clock encoded in its DNA methylation patterns - the addition of chemical tags to specific DNA sites. In healthy tissues, these patterns change predictably with age. Scientists have used this regularity to create DNA methylation age clocks - mathematical models that estimate a person's chronological age from a tissue sample with remarkable precision.
Two leading clocks are the Horvath clock, which uses 353 specific DNA sites to estimate age, and the Phenoage clock, which uses 513 sites and is better at predicting health-related aging outcomes. When these clocks are applied to cancer tumors, they frequently show a mismatch between the tumor's biological age and the patient's actual chronological age - a phenomenon called DNA methylation age drift.
In most cancer research, tumors that appear biologically older than expected (age acceleration, or DNAmaa) have been associated with poor outcomes. But what about tumors that appear biologically younger? This study investigated DNA methylation age deceleration (DNAmad) - where the tumor's biological clock appears to run slower than the patient's real age - in endometrial cancer, a phenomenon that had not been systematically studied before.
When the researchers applied the Horvath clock to 429 endometrial cancer tumors from The Cancer Genome Atlas (TCGA), the finding was striking: more than 90% of tumors displayed DNAmad - their DNA methylation age was younger than the patient's chronological age. This was unexpected, since age acceleration is the more commonly reported finding in cancer research.
The Phenoage clock gave somewhat different numbers (classifying more tumors as age-accelerated), but both clocks agreed in their rankings - tumors with the strongest deceleration on one clock tended to show the same on the other. To define a meaningful high-risk group, the researchers selected the top one-third of Horvath-based DNAmad tumors and filtered for those also confirmed by the Phenoage clock. This yielded 82 tumors (about 19% of the cohort) defined as high DNAmad positive (hDNAmad+).
Clinically, hDNAmad+ tumors were significantly more likely to occur in older patients, to be of serous or mixed (non-endometrioid) histological types, and to present at advanced stages and grades. Interestingly, hDNAmad+ was less common in patients with diabetes or hypertension - conditions known to be associated with age acceleration, suggesting the two phenomena may oppose each other.
Patients whose tumors fell into the hDNAmad+ group had significantly shorter overall survival and shorter progression-free survival compared to the hDNAmad- group - a finding confirmed by Kaplan-Meier survival curves with statistically significant log-rank tests. The difference was clinically meaningful, not just statistically significant.
To determine whether hDNAmad was an independent predictor or just a proxy for other known bad prognostic features, the researchers ran multivariate Cox regression analyses that also included age, tumor stage, grade, and histological type. The result: hDNAmad remained a significant independent predictor of shorter overall survival in multivariate analysis, along with advanced stage and grade. For progression-free survival, hDNAmad and advanced stage were the only independent factors.
This independence is important. It suggests hDNAmad captures biological information about the tumor that conventional staging and pathological grading do not fully reflect - adding genuinely new prognostic information that could one day help identify high-risk patients who need more aggressive surveillance or treatment, even when their clinical stage appears favorable.
At the genomic level, hDNAmad+ tumors were characterized by significantly more chromosomal instability. They had higher rates of copy number alterations (CNAs) across the genome and higher aneuploidy scores - meaning large portions of chromosomes were gained or lost compared to the normal two copies. Most hDNAmad+ tumors fell into the CN-high molecular subtype, the most genomically chaotic subtype of endometrial cancer with the worst prognosis.
The MYC gene amplification on chromosome 8q24.1 - a driver of cell proliferation in many cancers - was found in 61% of hDNAmad+ tumors compared to 35% of hDNAmad- tumors. The tumor suppressors TP53 (mutated in 70% vs. 30%), RB1, and CDKN2A - all key brakes on cell division and aging pathways - were far more frequently inactivated in hDNAmad+ tumors.
Strikingly, hDNAmad+ tumors showed strong enhancement of telomere maintenance - the cellular machinery that prevents chromosomes from shortening with each cell division. Amplification or gain of both the telomerase catalytic component (TERT, in 45% vs. 12%) and the telomerase RNA template (TERC, in 63% vs. 26%) were dramatically more frequent in hDNAmad+ tumors. This suggests these tumors have overcome one of the key molecular barriers to unlimited cell growth.
One of the most clinically significant findings concerns the tumor immune microenvironment. Using CIBERSORT analysis to estimate immune cell proportions from gene expression data, the researchers found dramatically fewer CD8 T cells (the primary cancer-killing immune cells) and myeloid cells in hDNAmad+ tumors compared to hDNAmad- ones. These tumors exhibit what is called an immunoexclusion phenotype - immune cells simply do not enter the tumor.
Analysis of immune checkpoint gene expression revealed a paradoxical but clinically important pattern: hDNAmad+ tumors had lower PD-L1 and CTLA4 expression but higher VTCN1 (B7-H4). PD-L1 and CTLA4 are the targets of the most commonly used immune checkpoint inhibitor drugs (pembrolizumab, nivolumab, ipilimumab). High VTCN1 is associated with immune suppression through a different pathway. This means hDNAmad+ tumors suppress immunity differently - and may not respond to the standard immunotherapy drugs that target PD-L1 and CTLA4.
Cancer immune cycle analysis confirmed that the primary defect was in step 4 of 7: the trafficking of immune cells from blood into tumor tissue. Out of 17 immune cell types analyzed, 14 showed significantly reduced recruitment to hDNAmad+ tumors. The consequence was reduced tumor cell killing - these tumors are effectively hiding from the immune system through a mechanism that standard immune checkpoint inhibitors are not designed to overcome.
The biological paradox - tumors that look young but behave aggressively - is explained by the following model: Normally, aging-related global DNA hypomethylation acts as a tumor suppressor, disrupting gene expression programs that promote cell proliferation and stemness. Cancer cells must overcome this barrier to grow aggressively. In hDNAmad+ tumors, significantly elevated levels of DNMT3A and DNMT3B (DNA methyltransferase enzymes) appear to re-methylate and rescue regions that should lose methylation with age, effectively overriding the aging-related tumor suppression mechanism.
Supporting this, hDNAmad+ tumors showed much higher stemness scores (a measure of stem cell-like properties that promote invasion and metastasis) and higher cell cycle scores, along with more frequent PIK3CA mutations and amplifications. The PI3K-AKT pathway, a key cell growth signaling network, was hyperactivated through two cooperating mechanisms: more frequent PIK3CA gain-of-function alterations plus extremely low expression of SCGB2A1, a newly identified PI3K inhibitor. Together, these changes drive runaway cell proliferation.
The practical implications are significant. hDNAmad could serve as a new biomarker to identify the most aggressive endometrial cancers - particularly those that, despite appearing early-stage, carry molecular features predicting early recurrence. Furthermore, the immunoexclusion profile and VTCN1 upregulation suggest that novel immunotherapy strategies targeting B7-H4 or other non-PD-L1 pathways, rather than standard checkpoint inhibitors, may be needed for this tumor subtype.