Blood Tests Can Detect the Genetic Fingerprints of Endometrial Cancer — but the Tumor's Biology Determines How Loud the Signal

A 44-patient study at Memorial Sloan Kettering used circulating tumor DNA to profile uterine cancers, revealing that mutational burden and molecular subtype — not stage alone — shape what a liquid biopsy can find.

Every year, roughly 66,000 Americans are diagnosed with endometrial cancer — cancer of the uterine lining. Most are caught early, treated with surgery, and cured. But a meaningful fraction recur, sometimes years later, and the tools for spotting a recurrence early remain blunt: imaging, symptoms, rising protein markers in the blood. None of them can say, with molecular specificity, whether a tumor's DNA is still circulating in the body.

Circulating tumor DNA — fragments of cancer genomes that leak into the bloodstream — offers a tantalizing alternative. A simple blood draw, processed through targeted sequencing, might detect the same driver mutations found in the original tumor. The question is whether the signal is strong enough, and consistent enough, across the full spectrum of endometrial cancers. At Memorial Sloan Kettering, researchers sequenced cell-free DNA from 44 endometrial cancer patients to find out. The dashboard that follows maps what they found: which mutations appear in the blood, how mutational burden varies by tumor type and grade, and where the biological signal fades into noise.

A Cohort Anchored in Endometrioid Disease

The 44 patients in the MSK cohort reflect the epidemiology of endometrial cancer — but imperfectly, as any single-center study must. Endometrioid carcinoma, the most common histologic subtype, dominates the enrollment. That is expected: endometrioid tumors account for roughly 80 percent of all uterine corpus cancers. But serous carcinomas and carcinosarcomas — rarer, more aggressive subtypes that may shed ctDNA differently — are present in much smaller numbers. That imbalance matters. Any subtype-specific conclusion drawn from a handful of serous or carcinosarcoma specimens should be treated as hypothesis-generating, not definitive.

Grade, Stage, and the Risk Factors That Shape ctDNA Shedding

Beyond histology, two clinical features are expected to influence whether tumor DNA appears in the blood: tumor grade and surgical stage. Higher-grade tumors — grade 3, in the three-tier system — tend to be more biologically aggressive, more genetically unstable, and more likely to release DNA into circulation. The cohort's grade distribution shows how heavily high-grade disease is represented, a critical calibration point for interpreting detection rates.

FIGO stage tells a complementary story. Most endometrial cancers present at stage I, confined to the uterus. Advanced-stage disease (stages III and IV) is less common but more likely to produce a detectable ctDNA signal simply because there is more tumor in more places. A cohort skewed toward late-stage disease would overestimate the sensitivity of a liquid biopsy in unselected clinical practice.

Lymphovascular space invasion — the infiltration of tumor cells into blood or lymphatic vessels — adds another dimension. LVSI-positive tumors carry higher recurrence risk and may, in principle, shed more DNA into circulation. The proportion of LVSI-positive cases, and the size of the indeterminate or unknown category, determines whether this risk factor can be meaningfully tested.

A cohort skewed toward late-stage disease would overestimate the sensitivity of a liquid biopsy in unselected clinical practice.

The Mutational Landscape: Hotspots, Hypermutation, and the Biology Beneath

With the cohort's composition established, the data turn to what the blood actually reveals. The protein-position map of detected variants is the first clue. In endometrial cancer, somatic mutations are not scattered randomly across the genome; they cluster at well-known hotspots — PIK3CA H1047, PTEN R130, TP53 R175 — that represent driver events fueling tumor growth. Tall bars at these positions confirm that the ctDNA panel is capturing biologically meaningful, actionable mutations, not just background noise.

Tumor mutational burden, measured as non-synonymous mutations per megabase of sequenced DNA, provides a broader view. Endometrial cancer is unusual among solid tumors in its molecular heterogeneity: microsatellite-stable tumors cluster at low TMB, microsatellite-instable (MSI-high) tumors occupy a middle band, and POLE-ultramutated tumors push into territory above 100 mutations per megabase. The TMB distribution across these 44 specimens should reflect that tiered biology — and the density plots stratified by histologic subtype and tumor grade test whether it does.

Genes, Variants, and the Recurrent Drivers in Blood

Gene-level tallies rank which somatic alterations appear most frequently across the cohort's circulating DNA. The usual suspects of endometrial cancer genetics — PTEN, PIK3CA, TP53, ARID1A, CTNNB1 — are expected to top the list. But the ranking also reveals something practical: which driver mutations are reliably detectable in a blood draw, as opposed to requiring direct tumor tissue.

The variant classification breakdown adds texture. Missense mutations typically dominate any cancer sequencing dataset, but endometrial cancer — particularly MSI-high tumors — produces an elevated rate of frameshift insertions and deletions caused by defective mismatch repair. A high frameshift or stop-gained fraction signals that the cohort includes a meaningful number of these mismatch-repair-deficient tumors. The variant type breakdown (SNPs versus insertions versus deletions) provides an independent check: if indels are nearly absent, it could reflect a limitation of the cell-free DNA platform rather than the tumor's actual biology.

Chromosomal distribution of variants offers yet another lens. Chromosomes harboring frequently mutated genes — chromosome 10 for PTEN, chromosome 3 for PIK3CA, chromosome 17 for TP53 — should show elevated variant counts relative to their size. An unexpectedly tall bar on a chromosome not enriched for panel genes could indicate a focal hypermutation event or a technical artifact worth investigating.

The Co-Mutation Landscape: Patterns That Point to Tumor Identity

The most information-dense view in the dashboard is the co-mutation matrix — a grid of the 25 most recurrently altered genes across all specimens. Each column is a patient; each row is a gene. Color indicates the type of alteration.

The patterns matter. PTEN and PIK3CA co-mutations are a hallmark of endometrioid biology, reflecting simultaneous activation of the PI3-kinase pathway through both loss of a brake (PTEN) and acceleration of a driver (PIK3CA). TP53 mutations, by contrast, tend to appear in specimens that lack PTEN loss — the molecular signature of serous-like endometrial cancer, a distinct and more aggressive subtype. These mutually exclusive patterns are not just academic taxonomy; they map directly to treatment strategies. A tumor with co-occurring PTEN and PIK3CA mutations might respond to PI3K inhibitors, while a TP53-driven tumor calls for a different approach.

Specimens with very few alterations across all 25 genes may simply have low tumor fraction in their circulating DNA, reducing the sensitivity of the liquid biopsy. These cases represent the practical floor of ctDNA detection — and the challenge that must be overcome before blood-based monitoring can replace tissue-based genotyping for endometrial cancer.

Quality Control: Separating Biology From Artifact

Before any of these findings are taken as final, every specimen must pass a basic quality check: how many variants did the sequencing call in each sample? A bar chart of per-sample variant yield is the primary technical sentinel.

Most specimens should fall within a consistent range. Bars dramatically taller than the cohort median almost certainly represent MSI-high or POLE-ultramutated tumors — biologically real outliers whose extreme mutational burden will dominate gene-frequency statistics if not handled explicitly. At the other end, specimens with zero or near-zero variants may indicate failed DNA extraction, very low circulating tumor fraction, or an early-stage tumor that simply does not shed enough DNA to detect. These samples risk being false negatives in any ctDNA positivity analysis and must be flagged.

A smooth gradient from low to high variant counts, without abrupt jumps, suggests uniform assay performance across the cohort — a reassuring sign that the patterns in the primary results reflect biology rather than batch effects or processing failures.

The promise of circulating tumor DNA in endometrial cancer is real but conditional. In 44 patients at MSK, a blood-based sequencing panel captured the same driver mutations — PTEN, PIK3CA, TP53, ARID1A — that define the molecular subtypes of the disease in tissue. Hypermutated tumors, driven by mismatch-repair deficiency or POLE mutations, broadcast the loudest signal into the bloodstream. But low-grade, early-stage, microsatellite-stable tumors — the majority of endometrial cancers in the population — remain the hardest to detect, their circulating DNA faint or absent. The next step is not simply enlarging the cohort, though that is necessary. It is asking whether serial blood draws over time, tracking the rise and fall of known driver mutations after surgery, can catch recurrence earlier than imaging or symptoms. That is where ctDNA moves from research curiosity to clinical tool — and where the sparse cells in this dashboard's cross-tabulations become the most important ones to fill.