PARP inhibitors are a class of targeted cancer drugs that exploit a weakness in tumors with defective DNA repair. They work by blocking the PARP enzyme, which cells use to repair single-strand DNA breaks. When PARP is inhibited in a tumor that already has a broken homologous recombination repair (HRR) pathway, the tumor cannot fix the resulting DNA damage and dies. This concept is called synthetic lethality and has been highly successful in ovarian and breast cancers with BRCA1/2 mutations.
However, in endometrial cancer, the use of PARP inhibitors remains investigational. A key question is how to identify which endometrial cancer patients are most likely to benefit, since not all patients with BRCA mutations necessarily have fully defective homologous recombination in practice, and some patients without BRCA mutations may still have HRR deficiency through other mechanisms.
This paper reports an exploratory biomarker analysis embedded within the MITO END 3 trial, a randomized phase II clinical trial conducted across multiple Italian centers that studied chemotherapy combinations in patients with advanced endometrial cancer (recurrent or metastatic disease).
The study assessed homologous recombination deficiency (HRD) using two different biomarker approaches applied to tumor tissue from trial participants.
The first is gLOH (genomic loss of heterozygosity), which measures large-scale chromosomal changes that accumulate when homologous recombination is defective. When cells cannot properly repair DNA breaks, entire chromosomal segments can be lost, a pattern detectable by genomic sequencing. The threshold used in this study was the ARIEL3 cutoff of gLOH greater than or equal to 16%, validated in ovarian cancer trials.
HRDsig is a machine learning algorithm that detects a characteristic mutational signature (the pattern of specific types of DNA mutations) associated with HRR-deficient tumors. It produces a score from 0 to 1, with a cutoff of 0.7 or higher indicating HRD positivity. Unlike gLOH, which measures structural genomic consequences of HRD, HRDsig detects the mutational footprint left by HRD on the tumor's DNA sequence, potentially capturing deficiency even when gLOH has not yet accumulated.
Of 102 evaluable patients with adequate tumor tissue for analysis, results were strikingly different depending on the measurement method used. By HRDsig, only 5 patients (4.9%) were positive. By gLOH, 10 out of 63 evaluable patients (15.9%) were positive using the threshold validated in ovarian cancer.
Importantly, the two methods showed limited overlap: being gLOH-high did not reliably predict being HRDsig-positive and vice versa. This discordance is biologically meaningful. gLOH reflects accumulated genomic damage from past HRD, while HRDsig reflects ongoing or recent HRD-associated mutagenesis. A tumor could show one without the other depending on when HRD arose during tumor evolution and how much time has elapsed.
These low prevalence rates contrast with ovarian cancer, where HRD positivity rates of 50% or higher are seen. This suggests that endometrial cancer is biologically distinct from ovarian cancer in its HRR pathway biology, and that the ovarian cancer biomarker thresholds and expected prevalences cannot be directly applied to endometrial cancer without recalibration.
The researchers examined the clinical and molecular characteristics of the HRD-positive patients to understand what kind of endometrial cancer tends to show HRR deficiency. Most HRD-positive tumors were endometrioid histotype, high grade, and carried TP53 mutations, which is associated with the copy-number high molecular subtype of endometrial cancer (also called TCGA-D or p53-abnormal). This subtype is generally associated with the worst prognosis.
Notably, none of the HRD-positive tumors were MSI-H (microsatellite instability high), the molecular subtype associated with mismatch repair deficiency and high immunotherapy response. This non-overlap makes biological sense: MSI-H and HRD deficiency are generally distinct mechanisms of genomic instability with different treatment implications.
Perhaps the most surprising finding was that 7 out of 8 patients with BRCA1 or BRCA2 mutations were HRDsig-negative. This is counterintuitive since BRCA1/2 are key HRR genes, but it is consistent with findings in other cancer types and may reflect that BRCA variants in these patients are variants of uncertain significance, or that compensatory mechanisms have restored functional HRR despite the mutation. This finding reinforces that BRCA mutation status alone is not a reliable surrogate for functional HRD in endometrial cancer.
The primary clinical message from this study is that HRD is uncommon in advanced endometrial cancer and that the small fraction of patients who are HRD-positive may represent a distinct subgroup that could benefit from PARP inhibitor therapy. This finding has direct relevance for trial design: future studies testing PARP inhibitors in endometrial cancer should use prospective HRD screening to enrich for the patients most likely to respond.
The discordance between gLOH and HRDsig, and the lack of correlation between BRCA mutation status and HRDsig positivity, highlights a broader challenge in the field: there is no single universally accepted test for HRD. Different assays measure different aspects of HRR deficiency and may not identify the same patients. Determining which test best predicts PARP inhibitor benefit in endometrial cancer specifically is an open research question.
The study also found that HRR pathway mutations were present in 24 of 102 patients (23.5%) by sequencing, yet showed no correlation with HRDsig positivity. This further confirms that having a mutation in an HRR gene does not guarantee functional HRD, and that functional assays measuring genomic consequences of defective repair are more informative than mutation catalogs alone. Overall, the results support a focused but cautious approach to PARP inhibitor development in endometrial cancer, targeting the small molecularly confirmed HRD-positive subgroup.