Twenty-Two Tumors Reveal the Mutational Anatomy of One of the Deadliest Uterine Cancers

Whole-exome sequencing of uterine carcinosarcoma — a rare, aggressive malignancy — maps the genes most frequently disrupted and exposes how tumor purity and stage shape what researchers can see.

Uterine carcinosarcoma kills more than half the women it touches, often within two years of diagnosis. It accounts for a thin sliver of all uterine cancers — roughly 5 percent — yet is responsible for a wildly disproportionate share of uterine cancer deaths. The tumors are biological chimeras: part carcinoma, part sarcoma, two malignant tissue types tangled inside a single mass. That duality has frustrated pathologists and oncologists for decades, leaving treatment plans borrowed from better-understood cancers and prognoses that remain stubbornly poor.

At Johns Hopkins, a team sequenced the exomes of 22 such tumors alongside matched normal tissue from the same patients, building a mutation-by-mutation portrait of the disease. The cohort is small — 22 women — but for a cancer this rare, it represents one of the most detailed genomic inventories assembled. What emerges is a landscape dominated by familiar cancer genes behaving in specific, sometimes surprising, patterns, set against a backdrop of quality-control checks that determine how far the findings can be trusted.

A Cohort of Postmenopausal Women

Uterine carcinosarcoma — sometimes called malignant mixed Müllerian tumor — overwhelmingly strikes women after menopause. The Johns Hopkins cohort reflects that reality. The age-at-diagnosis distribution clusters where clinicians expect it, in the sixth and seventh decades of life. Any patient diagnosed before 50 is clinically unusual and could harbor germline alterations in genes like TP53 or the BRCA DNA-repair pathway that predispose to early cancer. In a 22-patient study, even one or two younger cases can shift the distribution meaningfully, so establishing the age profile is the first step before interpreting any mutation data.

Treatment Arms and Stage: Checking for Structural Bias

Before examining a single mutation, it matters how the 22 patients were distributed across treatment regimens and disease stages. An imbalance in either dimension can confound any attempt to link genomic findings to clinical outcomes.

The treatment-arm chart reveals whether enrollment was roughly balanced or whether one regimen dominates. In a cohort this size, any arm with fewer than five patients is effectively underpowered for subgroup analysis — those cases describe individual trajectories, not population-level trends.

Stage composition carries similar weight. Uterine carcinosarcoma is notoriously diagnosed late; many women present with stage III or IV disease because the tumor spreads aggressively through the myometrium and into the peritoneum. A cohort dominated by advanced-stage tumors would be typical of real-world referral patterns. A cohort with substantial early-stage representation, by contrast, offers a rarer window into the genomic landscape before metastatic evolution has layered on additional mutations.

How Many Mutations, and How Varied?

The mutation burden of a tumor — how many DNA-altering changes it carries per megabase of sequenced genome — is one of the most informative single numbers in cancer genomics. It sets a baseline: a tumor with very high burden may respond to immunotherapy, while one with few mutations may be driven by a small number of potent alterations.

Across the 22 uterine carcinosarcomas, the nonsynonymous tumor mutational burden (TMB) — counting only mutations that change protein sequence — reveals the cohort's baseline mutation rate and, critically, flags any hypermutated outliers. Specimens with TMB above roughly 10 mutations per megabase deserve special attention: they may be mismatch-repair deficient or carry mutations in the proofreading domain of DNA polymerase epsilon (POLE), both mechanisms that unleash a flood of mutations and, paradoxically, can make tumors more visible to the immune system.

Specimens with TMB above roughly 10 mutations per megabase may be mismatch-repair deficient — a paradox that can make tumors more visible to the immune system.

The Genes That Keep Breaking

Gene-level recurrence is the backbone of any cancer exome study. A mutation in a single patient could be a passenger — biologically inert noise. A mutation in the same gene across five, ten, or fifteen patients is a signal.

The oncoprint of the 25 most frequently altered genes across the 22 tumors is where the biology crystallizes. TP53 — the genome's most famous tumor suppressor — should appear altered in the vast majority of samples, a pattern consistent with near-universal TP53 disruption in uterine carcinosarcoma. Beyond TP53, genes like KRAS, PIK3CA, and FBXW7 define therapeutically relevant subgroups. A patient whose tumor carries both TP53 loss and PIK3CA activation, for example, sits at the intersection of two druggable pathways.

Columns in the oncoprint represent individual patients. Those with very few colored tiles may have low tumor purity — meaning the sequencing captured more normal tissue than tumor — or may genuinely belong to a genomically quieter subtype. Rows altered in at least five of 22 patients (roughly 23 percent) represent the most robust recurrence signals; genes altered in only one or two patients are contextual but statistically fragile in a cohort this size.

What Kinds of Mutations Dominate

Not all mutations are created equal. A missense mutation swaps one amino acid for another — sometimes devastating, sometimes harmless. A nonsense mutation truncates the protein entirely. A frameshift insertion or deletion scrambles the reading frame downstream, usually destroying function.

In most solid tumors, missense variants dominate the landscape. But uterine carcinosarcoma, with its frequent disruption of tumor suppressors like TP53 and FBXW7, is expected to carry an enriched fraction of truncating events — nonsense and frameshift mutations that knock genes out entirely rather than merely altering them. The variant-classification spectrum across all called mutations reveals whether loss-of-function events are indeed over-represented relative to typical endometrial cancers.

At the variant-type level, single-nucleotide changes (SNPs) should account for 85 to 95 percent of all calls in a standard exome-capture experiment. A disproportionately high fraction of insertions or deletions — above roughly 15 percent — could indicate microsatellite instability or, less interestingly, an alignment artifact worth investigating.

Does Stage Track with Mutation Burden?

One of the most intuitive questions in cancer genomics: do tumors that have spread farther carry more mutations? If mutation accumulation accompanies clinical progression, advanced-stage tumors should show a rightward shift in TMB compared with early-stage cases.

The stage-stratified TMB chart plots each specimen's nonsynonymous burden within its AJCC stage row. With 22 patients split across multiple stages, some rows will be thin — a stage with only one or two specimens cannot support statistical inference. But the visual pattern matters. If stage III–IV tumors consistently sit to the right of stage I–II tumors, the data support a mutation-accumulation model. If the spreads overlap entirely, the difference in aggressiveness between stages may be driven by qualitative changes — specific pathway activations, chromosomal rearrangements — rather than sheer mutational quantity.

A single high-TMB outlier in an early-stage row is particularly interesting. It would represent a biologically aggressive tumor caught before it spread — a case where molecular profiling might have flagged the danger that staging alone could not.

The Purity Problem: Can We Trust What We See?

Every tumor biopsy is a mixture. Tumor cells are tangled with normal stromal cells, immune infiltrates, and blood vessels. The fraction of the specimen that is actually tumor — its purity — directly affects whether somatic mutations can be detected. A specimen with 30 percent purity means 70 percent of the sequenced DNA is normal, and low-frequency tumor variants can drown in that noise.

The tumor-purity distribution across the 22 specimens is the single most important quality-control check. If most specimens cluster above 60 percent purity, the mutation calls are likely robust. If several fall below 40 percent, their TMB values and gene-level calls should be interpreted conservatively — some true mutations were probably missed.

The definitive test is whether purity correlates with TMB. A positive correlation — low purity mapping to low TMB — would mean that part of the spread in mutation burden is an artifact of specimen composition rather than genuine genomic diversity. Outliers that show high TMB despite low purity are especially noteworthy: they are likely genuinely hypermutated, and their true burden may be even higher than measured.

Twenty-two patients cannot settle the genomics of uterine carcinosarcoma. But they can — and do — establish its contours: a disease defined by near-universal TP53 loss, punctuated by actionable co-mutations in genes like PIK3CA and KRAS, with a mutation burden that varies enough across patients to suggest distinct molecular subtypes lurking beneath a single pathological diagnosis. The quality-control data serve as a standing caution: tumor purity and tissue handling shape what any exome study can detect, and the lowest-purity specimens in this cohort may be hiding drivers that sequencing simply could not reach. For the clinician, the immediate question is whether the hypermutated outliers — the tumors at the far right of the TMB distribution — would respond to immune checkpoint inhibitors the way mismatch-repair-deficient colorectal and endometrial cancers do. For the researcher, the next step is clear: validate these recurrence patterns in a larger, multi-institutional cohort, and pair exome data with transcriptomic profiling to understand how the carcinoma and sarcoma components of these tumors divide their mutational labor.