MSK team finds a way to predict breast cancer drug resistance before treatment begins

Published in Nature. Memorial Sloan Kettering Cancer Center.

What if oncologists could read a tumor's escape plan before it executes? That is the question at the center of a new study from Memorial Sloan Kettering Cancer Center, where researchers have identified genetic signatures that predict, before treatment starts, exactly how certain breast cancers will develop resistance to one of the most widely used drug classes in metastatic disease.

The drugs in question are CDK4/6 inhibitors, a mainstay of treatment for estrogen receptor-positive metastatic breast cancer. They work by blocking proteins that drive cancer cell division. But about 10% of patients develop resistance through a specific mechanism: their tumors lose a protective gene called RB1, which normally acts as a brake on cell growth. Once RB1 is gone, CDK4/6 inhibitors lose their target.

Two warning flags in the genome

The MSK team, led by physician-scientists Pedram Razavi, MD, PhD, and Sarat Chandarlapaty, MD, PhD, with first author Anton Safonov, MD, analyzed genomic data from more than 5,800 breast cancer patients. They identified two pre-treatment warning signs that a tumor is likely to shed its RB1 gene during CDK4/6 inhibitor therapy.

The first is homologous recombination deficiency (HRD), a condition in which cancer cells cannot properly repair broken DNA. The second is the tumor's starting genetic makeup, particularly whether it already carries only a single copy of RB1. Tumors with one copy are far more likely to lose the remaining copy under the selective pressure of treatment.

Patients born with mutations in the BRCA2 gene were especially vulnerable. The inherited BRCA2 defect creates a DNA repair environment that makes additional mutations in RB1 more likely to accumulate. These patients consistently did poorly on standard CDK4/6 inhibitor therapy.

"To our knowledge, this is the first example showing that a complete genomic analysis of breast cancer, including both inherited and tumor-specific alterations, can predict the precise biological mechanism of resistance before therapy even begins," Razavi said.

From prediction to preemptive strike

The findings did not stay in the realm of theory. In laboratory experiments using patient-derived xenograft models from BRCA2-mutant breast cancers, co-first author Minna Lee, MD, demonstrated that CDK4/6 inhibitors performed poorly against these tumors, which were prone to losing RB1 during treatment. PARP inhibitors, drugs that exploit the DNA repair deficiency rather than fighting against it, consistently worked better in HRD-positive tumors.

The convergence of genomic data, lab evidence, and clinical observations was strong enough to bypass the usual incremental trial process. A global, randomized phase 3 clinical trial called EvoPAR-Breast01 is now enrolling patients with newly diagnosed ER-positive, HRD-positive metastatic breast cancer. The trial tests whether the combination of the PARP inhibitor saruparib and the hormonal therapy camizestrant outperforms standard CDK4/6 inhibitor combinations.

"There aren't many examples where translational data were compelling enough to move directly into a phase 3 study without developing earlier clinical evidence," Razavi said.

The reversion twist

One of the study's more intriguing findings involves what happens after resistance develops. Some tumors developed "reversion mutations" that restored their DNA repair function. Once HRD was reversed, these tumors appeared to regain sensitivity to CDK4/6 inhibitors. This raises an interesting possibility: using PARP inhibitors early might not only improve initial outcomes but could also restore the effectiveness of CDK4/6 inhibitors for later use.

"Cancers don't have endless ways to escape treatment," Razavi said. "They are one- or two-trick ponies, and those tricks are often determined by their inherited or tumor-specific genetic features. If we can predict what they're capable of, we can intercept it before the resistance happens."

What this study cannot tell us yet

The genomic analysis is retrospective. The predictions are based on patterns observed in existing patient data, and while the laboratory models confirmed the biological logic, the definitive test is whether the EvoPAR-Breast01 trial actually improves patient outcomes. Phase 3 results will take years. It also remains to be seen how broadly the HRD-based prediction applies across the full diversity of breast cancer subtypes and patient populations.

The study draws on MSK's unusually large genomic dataset and its integrated research infrastructure, advantages not every cancer center can replicate. Whether the approach generalizes to other institutions with smaller datasets or less comprehensive genomic profiling is an open question.

Still, the conceptual advance is significant. Rather than waiting for resistance to emerge and then reacting, the approach aims to read a tumor's vulnerabilities upfront and choose the treatment most likely to prevent its escape. That shift, from reactive to preemptive oncology, has been a long-stated goal. This study offers one of the clearest examples of how genomics might get us there.