Liquid Biopsy in Ovarian Cancer May Help Guide Clinical Decisions

Jaana Oikkonen, PhD

Jaana Oikkonen, PhD

A new single-center report published inJCO Precision Oncologyoffers a proof of concept for using circulating tumor DNA (ctDNA) to guide clinical decisions in high-grade serous ovarian cancer.

“There is an urgent need to find effective therapy targets for high-grade serous ovarian cancer in general and for relapsed, chemotherapy-resistant disease in particular. [Our] approach allows the selection of treatment options that target subclones that persist during therapy,” the study authors, led by Jaana Oikkonen, PhD, of the University of Helsinki in Finland, wrote. “This is a substantial improvement in the management of recurrent solid cancers, where tumors are not usually sampled either because of the risk of potential intervention complications or simply because the sample could be insufficient or not representative of the disease.”

The investigators noted that they believe their research to be the first comprehensive, open-source ctDNA workflow for detecting clinically actionable alterations in solid cancers.

They collected 78 longitudinal ctDNA samples and 21 tumor tissue samples from 12 patients with high-grade serous ovarian cancer at Turku University Hospital, Finland. They applied a sequencing panel of more than 500 genes, followed by variant and copy number alterations (CNA) calling, filtering, and prioritization of clinically relevant aberrations.

“The motivation for ctDNA analysis was to perform a minimally invasive genomic survey of clinically actionable tumor aberrations in the patient,” they wrote. “To establish this connection, we used tumor tissue samples as a ground truth for patient mutational burden. Altogether, 265 mutations in 185 genes and CNA aberrations in 113 genes passed our calling and filtering criteria.”

Patients received either primary debulking surgery followed by 5 to 6 cycles of platinum-taxane chemotherapy, or neoadjuvant chemotherapy. Oikkonen et al stratified patients on the basis of their platinum-free interval (PFI). Half of the patients were poor responders, defined as having either platinum-resistant disease (PFI <6 months; n = 3) or partially platinum-sensitive disease (PFI, 6 to 12 months; n = 3). Patients with platinum-sensitive disease (PFI >12 months; n = 6) were considered good responders.

They extracted cell-free DNA (cfDNA) from plasma samples and subjected them to 1000&times; targeted Illumina Hi-Seq sequencing using the Oseq Solid Cancer Panel, which tests for more than 500 clinically actionable genes.

The investigators used the ctDNA data analysis pipeline of the Anduril computational ecosystem to detect mutations. They tracked mutations detected in at least 1 plasma or tumor tissue sample with at least 1% variant allele frequency (VAF) but not in the corresponding germline blood control. They filtered the samples by identifying clusters of variants shared between patients but not known in cancer. They used detected somatic alterations to confirm the presence and quantity of ctDNA.

When they found potentially clinically actionable alterations, the results were validated through immunohistochemistry. In situ hybridization was used for prominent alterations that have been shown to exist in patients&rsquo; tumor tissue.

Oikkonen et al found that the median concordance of mutations detected from plasma to tumor tissue samples from the same patient was 79%. They found similar median concordance (74%) with CNAs for plasma samples with tumor content greater than 30%. The number of detected CNAs highly correlated with the proportion of ctDNA from cfDNA (n = 101; correlation coefficient, 0.88).

Mutations inTP53were reported in all of the patients in the cohort. Additionally, there was also a correlation found between ctDNATP53VAF and CA-125 levels (median, 0.67; range, 0.16-0.97).

&ldquo;These results show that the mutations and CNAs detected from plasma are in good concordance with those detected from tissue samples, which is the prerequisite for the use of detected mutations and CNAs from ctDNA in clinical decision making,&rdquo; the investigators wrote.

They also hypothesized that genes whose mutation VAFs remained stable or increased during treatment in poor responders could reveal the pathways involved in chemotherapy-resistant disease. Pathways enriched in the poor-responding patients versus the good-responding patients included transcription, p53, and chromatin regulatory pathways and DNA double-strand break repair pathways. The most selectively enriched pathways were all related to chromatic regulation in the poor-responding patients.

Oikkonen et al identified potential clinically actionable mutations or CNAs in 7 patients (58%). They also identified 4 major targetable processes: mammalian target of rapamycin (mTOR; 2 patients), DNA repair (3 patients), epidermal growth factor receptor (EGFR; 3 patients), and cyclin dependent kinases (CDKs; 1 patient).

Both patients with mTOR pathway mutations responded well to platinum-based therapy. In the DNA-repair group, 1 patient had aBRCA2mutation, which indicated possible sensitivity to PARP inhibitors, although the frequency was low in the tumor cells. Another DNA-repair patient had a germline deletion inRAD51C; this patient had an extended response to platinum-based chemotherapy. The third DNA-repair patient had a somatic deletion ofFANCAplus concurrent deletions inCDKN1BandCDKN2B. This patient progressed 8.9 months after the last platinum cycle, despite having no residual tumor following surgery. The authors suggested these mutations would suggest sensitivity to CDK4/6 inhibitors.

Of the 3 patients who had EGFR-sensitive mutations, 1 had anERBB4mutation, while another had aMAPKmutation and a simultaneousMAPK1amplification. The findings for the thirdEGFR-mutated patient, who harbored anERBB2amplification in her ctDNA, led the clinical team to adjust her treatment asERBB2is a predictive biomarker for trastuzumab (Herceptin). Although this patient experienced clinical and pathologic complete response to initial therapy of neoadjuvant chemotherapy,TP53VAF remained detectable throughout treatment, indicating a poor response to platinum taxane, and she did indeed progress 5 months after completing initial treatment.

On the basis of theERBB2amplification detected, this patient&rsquo;s treatment was changed to trastuzumab 600 mg subcutaneously once every 3 weeks, combined with reduced doses of carboplatin (AUC 4) and dose-dense paclitaxel (80 mg/m²on days 1 and 8). The patient responded to this combination well, and a reduction in CA-125 value from 840 IU/mL to 19 IU/mL was achieved after only 3 treatment cycles, Oikkonen et al reported.

The investigators believed that longitudinal ctDNA sampling can be used to monitor response to first-line therapy and to identify patients with ovarian cancer who are likely to respond poorly to treatment. &ldquo;Rapid discovery of resistant cell population expansion provides an early opportunity to interfere with the development of recurrence,&rdquo; they wrote. &ldquo;This is particularly important in chemo-resistant relapse in which larger tumor burden is associated with low therapy responses, with both conventional and targeted therapies.&rdquo;

Reference:

Oikkonen J, Zhang K, Salminen L, et al. Prospective Longitudinal ctDNA WorkflowReveals Clinically Actionable Alterations in Ovarian Cancer [published online May 3, 2019].JCO Precision Oncology.doi: 10.1200/PO.18.00343.