Comprehensive Genomic Profiling Helps Detect Patients With Ovarian Cancer Likely to Benefit from PARP Inhibition

Kathleen N. Moore, MD

Approximately 25% of patients withBRCAwild-type serous ovarian cancer may benefit from treatment with PARP inhibitors, along with 12.7% of patients with a non-serous histology, according to findings of genomic analyses of patients with ovarian cancer presented during the 2017 ASCO Annual Meeting.

In an oral presentation, Kathleen N. Moore, MD, assistant professor of gynecologic oncology, University of Oklahoma Health Sciences Center, Oklahoma City, Oklahoma, presented findings that demonstrate that comprehensive genomic profiling (CGP) is a valuable tool to integrate into routine ovarian cancer treatment decision making and clinical trial design.

It has been shown that defective homologous recombination DNA repair (HRD) is associated with high-grade serous ovarian cancer histology, longer survival, and platinum or PARP inhibitor sensitivity. HRD causes loss of heterozygosity (LOH), a pattern of allelic imbalance detectable by CGP.BRCAwild-type serous ovarian cancer can have LOH and respond to PARP inhibitors, whereas non-serous, or difficult to classify, ovarian cancers (NOS) are often less responsive to platinum-based therapy.

This study identified additional therapeutically relevant subsets of patients with ovarian cancer who might benefit from PARP inhibitors or targeted therapies. Homologous recombination (HR) is the primary high-fidelity DNA repair pathway for double strand breaks. Once HRD occurs, DNA repair fidelity in tumor cells is low, resulting in errors that can be detected on chromosomes as LOH.

BRCA1andBRCA2germline and somatic mutations are known to cause HRD. The ability to identify other causes of HRD can lead to the development of a predictive biomarker for therapeutics that target DNA damage response repair. Tumors with intact homologous recombination need alternative treatment strategies.

DNA (≥50 ng) extracted from formalin fixed, paraffin embedded tumor tissue obtained during clinical care from 4114 patients with advanced ovarian cancer was analyzed for all classes of clinically relevant genomic alterations by hybrid-capture, next-generation sequencing of up to 315 genes. Most samples were from the ovary (89%); 6% were from the Fallopian tube, and 5% from the peritoneum. Tumor subtypes included serous high grade and NOS (67%), 20% epithelial NOS, and 13% non-serous.

HRD was assessed using a combination ofBRCA1andBRCA2mutation status and LOH score across tumor histologies. In tumors with intact homologous recombination, genomic alterations were used to predict alterative targeted therapeutic options using potentially actionable mutations.

Proprietary algorithms simultaneously evaluated: 1) global genomic metrics, including LOH, with LOH-high defined as &ge;16% genomic LOH, LOH-low as <16% LOH; 2) microsatellite instability (MSI); and 3) tumor mutation burden (TMB), with TMB-high defined as 10 mutations/Mb.

For all patients, 17.2% had aBRCAmutation, 24.2% wereBRCAwild-type/LOH-high, and 58.7% wereBRCAwild-type/LOH-low. The median LOH score for tumors with aBRCAmutation was significantly higher than for those that wereBRCAwild-type (22.2 vs 9.8;P<1 x 10^-10).BRCA-mutated serous and non-serous LOH scores were similarly LOH-high (86% and 79%, respectively;P= .79). The median LOH score forBRCAwild-type serous disease was significantly higher than nonserous (12.8 vs 5.8;P<1 x 10^-10).

For serous (n = 2270) tumors, 56.2% wereBRCAwild-type/LOH-low, 25.1% wereBRCAwild-type/LOH-high, and 8.7% wereBRCA-mutated; for epithelial NOS tumors (n = 807) the results were 52.3%, 28.7%, and 19.0% respectively; non-serous tumors (n = 537) results were 81%, 12.7%, and 6.5%, respectively. Moore said these results indicate that epithelial NOS tumors belong in the serous category as their profiles were similar.

Among tumor subtypes, Moore noted that neuroendocrine tumors are 4.8%BRCA-mutated and 15.8% LOH-high. &ldquo;That&rsquo;s surprising and opens up some therapeutic opportunities. In mucinous tumors, the lack of aBRCAmutation is not surprising to any of us. But the presence of some LOH-high (8.1%) is surprising, and again, maybe it opens up some [therapeutic] opportunities.&rdquo;

The genomic alteration analysis showed molecularly distinct subsets and not a lot of commonalities between groups. The clinical significance of this is not yet known. Alterations in the PI3kinase/AKT pathway,EGFR,FGFR,HER2,MEK, and other mutations, as well as MSI were identified inBRCAwild-type, LOH-low tumors.

Moore said that translation of molecular information into clinical practice needs to be validated further in clinical trials. She said that patients with aBRCAmutation and/or LOH-high would benefit from PARP inhibitors or enrollment into a clinical trial. Those withBRCAwild-type/LOH-low, high TMB, and MSI might benefit from immune checkpoint inhibition or enrollment into a clinical trial; 66% of tumors in this group had other targetable genomic alterations, so these patients might benefit from targeted therapy or clinical trial enrollment. Moore said empiric therapies, including surgery, chemotherapy, radiation therapy, therapy resistance/recurrence, and palliative care would still be needed to treat the 18% or patients with untargetable genomic alterations.

Moore observed that patients with carcinosarcomas and mucinous carcinomas have routinely been excluded from clinical trials of agents; this CGP analysis shows they have genomic alterations that would indicate potential benefit from PARP inhibitors or similar agents, and in fact 82% of patients with ovarian cancer have genomic alterations, suggesting a mechanism of potential benefit for novel treatment approaches not captured by traditional histologic definitions.

Barbara Norquist, MD, assistant professor, Division of Gynecologic Oncology, University of Washington in Seattle, discussed the study presented by Moore. &ldquo;It&rsquo;s very common for people to think these changes only happen in high-grade serous tumors. I congratulate [the researchers] on this large study,&rdquo; she said.

Norquist&rsquo;s questions for future investigation included how to link this &ldquo;extensive and rich database&rdquo; to clinical outcomes, whether tumor resistance that develops over time might change the LOH score, which might affect the timing of testing, and how testing might be integrated into patient care.

Responding to a question about testing failure rate, Julia A. Elvin, MD, PhD, of Foundation Medicine, Inc., Cambridge, Massachusetts, the lead author of the study, said that at a tumor nuclear percentage of around 30% (ie, if approximately 1 in 3 cells in the tissue specimen comes from the tumor), there is enough signal to reliably detect LOH patterns in samples. This analysis is routinely performed on core biopsy samples, and only requires 50 ng of DNA, the equivalent of about 8,000 to 15,000 nucleated cells. &ldquo;Overall sample failure is about 7%, mainly because the tissue has been used up by prior molecular testing, so it is usually an amount issue, not a quality of sample issue,&rdquo; she said.

Reference:

Elvin JA, He Y, Sun J, et al. Comprehensive genomic profiling (CGP) with loss of heterozygosity (LOH) to identify therapeutically relevant subsets of ovarian cancer (OC).J Clin Oncol35, 2017 (suppl; abstr 5512).