Blood samples were obtained from 138 patients with ovarian cancer between January 2017 and January 2020, and ctDNA sequencing was performed on Illumina HiSeq sequencer (Illumina, San Diego, CA) with a median unique exon coverage depth of 1,500× in a College of American Pathologists (CAP) and Clinical Laboratory Improvement Amendments (CLIA) certified laboratory (3D Medicine Inc: Shanghai, China) [5]. Hybrid capture-based NGS genomic profiling of a well-designed 150 cancer gene panel was performed on ctDNA. Somatic alterations were identified and clinical information including age, gender, and tumor histology was collected. A waiver of informed consent form was signed by each patient, and the study was approved by the ethics committee of the hospital.

Briefly, 20 ml of peripheral whole blood was collected for genomic profiling of ctDNA. The blood was centrifuged in Streck tubes at 1,600 g for 20 min at room temperature to separate the plasma. Then, the plasma layer was carefully transferred to a new 1.5 ml Eppendorf tube, followed by room-temperature centrifugation at 16,000 g for 10 min to remove residual cells and debris. The buffy coat was then transferred to a new tube for genomic DNA (gDNA) extraction. The QIAamp Circulating Nucleic Acid Kit (Qiagen) was used to extract ctDNA from the plasma, respectively, following the standard protocols, and then fragmented to a size ranging from 200 bp to 400 bp using Covaris S2 SonoLAB (Covaris). DNA concentrations were determined by the Qubit dsDNA HS Assay Kit (LifeTechnologies).

The assay methodology and procedure were as previously described [5]. Briefly, 20–100 ng of cfDNA was extracted to create adapted sequencing libraries. The ctDNA libraries were prepared by the Accel-NGS 2 S Plus DNA Library Kit (SWIFT) with unique identifiers (UIDs, also called barcodes) to tag individual DNA molecules. The captured DNAs were then amplified by PCR, and the final DNA concentrations and sizes were respectively measured by Qubit and Caliper. The captured libraries for FFPE gDNA were loaded into the Illumina HiSeq sequencer. The fraction of ctDNA was estimated using the maximum somatic allele frequency (MSAF). GAs including single nucleotide variation (SNV), insertions/deletions, copy number variations (CNV) and gene fusions were assessed and the corresponding criteria are the same as our previous study [6]. Germline alterations were excluded. We included in this study only samples with a tumor cell percentage >20.

The paired-end reads were mapped by BWA [7] MEM algorithm. SNVs were called by MuTect [8] with default parameters. Small insertions and deletions were called from the union of Varscan 2 [9] and Pindel [10] with default parameters. Fusions were called by selfdeveloped scripts with at least 5 pairs of reads spanned over the breakpoints between two partner genes. The CNVs of tumor tissues were calculated by BIC-seq2 [11] with default parameters, and the CNVs of ctDNA samples were called by a method reported by Jacob J. Chabon et al. [12]. All mutations were manually reviewed using integrative genomics viewer (IGV) [13] to further eliminate falsepositive results. The probability density distributions of mutant and wild-type fragments were calculated by Gaussian kernel smoothing using StatsModels 0.8.0.

Categorical variables were described as number and proportions. Categorical relationships were examined by using Pearson’s chi-square test with the Yates continuity correction when applicable and p value <0.05 was considered statistically significant. The SPSS22.0 software (SPSS, Inc., Chicago, IL, USA) was carried out for statistical analysis.

Hybrid capture-based genomic profiling was performed on plasma samples collected from 138 patients with ovarian cancer. The baseline characteristics for the patients were described in Table 1. In brief, the disease histology of ovarian cancer was ovarian serous cystadenocarcinoma (100%). The median age was 57 years.

The MSAF was determined as the maximum allele frequency for all the mutations detected per sample in blood. The MSAF greater than zero was used as the evidence of ctDNA in the blood and ctDNA was detected in 115 (83%) samples. The median MSAF across all cases was 0.026 (range, 0.0003–0.2822), and among cases with evidence of ctDNA present, the average of reportable GA was 3.83 GA/case (Table 1).

Among 115 cases with evidence of ctDNA, the most frequently altered genes were TP53 (72%), KRAS (11%), LRP1B (10%), ZNF703 (9%) and NF1 (8%) (Figure 1A). We further compared common GAs detected in ctDNA with those detected in tissue samples from patients with ovarian cancer and with those from TCGA [14]. The frequencies of common GAs were largely similar between ctDNA and tissue samples in our tissue database for patients with ovarian cancer, while different from TCGA (Figure 1B). These included TP53 (72% vs. 81% vs. 63.8%), KRAS (11% vs. 9% vs. 0.41%), LRP1B (10% vs. 13% vs. 2.9%), ZNF703 (9% vs. 1% vs. 0%), NF1 (8% vs. 13% vs. 4%) and so on. The gene amplification detected in ctDNA samples was inconsistent with our tissue samples. The most commonly detected gene amplification in of blood was PTK2. The detected amplification rates of PDGFRA (3.5% vs. 0%), DDR2 (2.6% vs. 0.2%) and KIT (2.6% vs. 0.2%) were higher in blood than tissue (p < 0.05), while the amplification rate of MYC was lower in blood than tissue (4.3% vs. 11.3%, p = 0.02, Figure 1C). Gene rearrangement was more frequently detected in ctDNA samples from this study than in tissue samples (Figure 1D).

Potentially actionable GAs informing selection of matched targeted therapies and clinical trials or predicting lack of response to antibody therapies were identified. GAs of BRCA1/2 (HRR associated-genes) were observed in 6.1% of ctDNA cases and 14.1% of tissue cases. Somatic SNV alterations in other HRR associated-genes including ATM, CHEK1/2 and RAD50 were detected in 6.1%, 5.2% and 1.7% of patients in ctDNA compared to 4.9%, 2.0% and 1.2% of patients in the tissue database (Figure 2A). In addition, a total of 26.1% of cases harboring at least one alteration in PI3K/AKT/mTOR pathways in ctDNA sample (vs. 34.5% in tissue sample, Figure 2B). And 13.0% of ctDNA samples showed at least one VEGF signal/angiogenesis pathways associated-genes mutations, compared with 20.0% in the tissue samples. However, the mutation frequency of AKT, PIK3CA, PTEN and STK11 in PI3K/AKT/mTOR pathway was significantly lower than that in tissue samples (p < 0.05, Figure 2C). The mutation frequencies of RAS pathways (12.2% vs. 12.7%) and NOTCH pathways (13.9% vs. 14.1%) were similar between ctDNA and tissue. (Figures 2D,E). In addition, we found the ratio of MSI-H was not different in ctDNA and tissue samples (1.3% vs. 1.4%, Figure 2F).