NEW YORK – Whole-genome sequencing could help to further categorize triple-negative breast cancers into groups that may respond to different treatment approaches, according to a new study.
An international team of researchers sequenced 254 triple-negative breast cancer tumors from patients from the Sweden Cancerome Analysis Network-Breast (SCAN-B) project, an ongoing population-based observational study of breast cancer patients in southern Sweden. About 9 percent of all breast cancers are triple-negative breast, meaning they lack the progesterone, estrogen, or HER2 receptors that other breast cancers often have and are associated with poor outcomes.
By applying the previously developed machine-learning algorithm HRDetect to the genome sequences, the researchers divided the tumors into three groups — low, intermediate, and high — based on how deficient they were in homologous recombination-based gene repair. These groups had varying prognoses and responses to treatments, as the researchers reported today in Nature Medicine.
"Using whole-genome sequencing, we can truly discriminate tumors that may or may not respond to current drugs among triple-negative breast cancer patients, a type of breast cancer that we still struggle to treat well," first author Johan Staaf, an associate professor of oncology and pathology at Lund University, said in a statement.
The SCAN-B study has been recruiting breast cancer patients from nine hospitals in the Skåne region of Sweden, which serves about 1.3 million people. During the timeframe of the study, 4,665 people in that population were diagnosed with invasive breast cancer, nine percent of whom had triple-negative disease. Of these, 340 patients were enrolled into the study and 237 patients had whole-genome sequencing data that was good enough for further analysis.
That 3 percent failure rate was largely driven by samples not reaching the required 30-fold sequencing depth, which the researchers said reflects a true estimate of sequencing success in a clinical scenario.
After determining likely somatic drivers and pathogenic germline mutations for this cohort, the researchers applied the mutational signature-based algorithm HRDetect to examine the "BRCA-ness" or homologous recombination repair deficiency of these samples. Nearly 60 percent of the triple-negative tumors were classified as HRDetect-high; about 35 percent were HRDetect-low; and about 5 percent fell in an intermediate group.
As the researchers expected, the HRDetect-high group was enriched for patients under the age of 50 years, with high-grade tumors and a basal-like expression subtype. But as they noted, this group also included tumors with ER-staining, from middle-aged patients, and with non-basal-like gene expression profiles, suggesting that the HRDetect-high phenotype is enriched for the typical basal-like tumors found among young patients with triple-negative disease, but also encompasses others. This, they argued, suggests HRDetect represents a novel means of stratifying tumors.
Patients whose tumors fell into these different HRDetect groups also responded differently to treatment. Patients in the HRDetect-high group had the best outcomes of the three. These tumors were both more likely to respond to adjuvant chemotherapy as well as to PARP inhibitors. The HRDetect-intermediate group, meanwhile, had the worst outcomes, though the HRDetect-low group also had poor outcomes.
But as the researchers noted, the molecular profiles of the HRDetect-intermediate and -low groups suggested potential actionable targets. The HRDetect-intermediate group, for instance, was enriched for tumors with CCNE1 amplifications and for hypermutators of a rearrangement signature of long tandem duplications. The HRDetect-low group, meanwhile, was enriched for alterations to the PIK3CA/AKT1 pathway.
"[I]mportantly, this approach also gives us clues to some of the mechanisms that are going wrong in the poor-outcome tumors, and hence how we might treat those tumors differently or how we might develop new drugs," Staaf added.