Cell-Free DNA from Ascites and Pleural Effusions: Molecular Insights into Genomic Aberrations and Disease Biology

This article is featured in Highlights of This Issue, p. 773

The expanding application of targeted treatments for malignancies highlights the importance of monitoring the genetic aberrations of a tumor over time (1). Many treatment paradigms have relied on the analysis of a single biopsy to provide genomic profiling of an individual's tumor. Nonetheless, inter-tumoral as well as intra-tumoral heterogeneity can lead to an incomplete genomic picture when analyzing a sample from a single tissue biopsy. Smaller subclonal populations of mutations may be undetectable if not present within the specific anatomic location within a biopsied tumor or at distant unbiopsied metastatic sites (2).

It is clear that assays that are able to represent global mutational burden across multiple metastatic sites will be a cornerstone of effective targeted therapy. Liquid biopsies, specifically the analysis of plasma cell-free DNA (cfDNA), may help overcome the spatial bias of single tissue samples by capturing the global heterogeneous tumor genome of both primary and metastatic lesions from a single fluid sample (3–5). In this investigation, we consider the potential of analyzing the genetic alterations present in the cfDNA of ascites or pleural effusion samples from individuals with metastatic cancer across histologies.

The advantages of monitoring ctDNA from ascites or pleural effusions includes the fact that abundant DNA can be obtained through (i) minimally invasive procedures, (ii) are collected for clinically therapeutic and diagnostic purposes, (iii) can be collected from advanced stage malignancy, (iv) can have larger DNA quantities for analyses compared with plasma, (v) can be collected serially to understand clonal dynamic change within the cells, and (vi) can be paired with viable cells for patient-derived xenograft (PDX) analyses. The detection of relevant mutations will be presented in the context of a larger cohort of individuals who had fluid removed during therapeutic paracenteses and thoracenteses.

Samples were collected from eleven subjects receiving therapeutic removal of ascites or pleural effusions and prepared for analysis. Of these, samples from seven individuals passed quality control (QC) thresholds of the median of the absolute values of all pairwise differences (MAPD; ≤0.3) and SNP quality control of normal diploid markers (ndSNPQC; ≥26). Four of these subjects had genomic data from previous analysis of solid tumor tissue. All individuals gave informed consent in accordance with the University of California San Diego Internal Review Board guidelines and consented for review of medical records.

Of the seven subjects whose effusate samples passed QC thresholds, six individuals had metastatic solid tumor including of the lung cancer, breast cancer, pancreas cancer, colon cancer, rectal cancer, and renal cell carcinoma. One human sample from an individual who had cirrhosis served as a negative control. One subject who had colon cancer was cytology negative within the pleural effusion and likely had the effusion from a cause other than malignancy.

Samples were centrifuged at 1,500 rpm for 5 minutes and approximately 30 mL of the clear supernatant aspirated and frozen at −80°C. cfDNA was isolated using commercially available DNA purification kits (Plasma/Serum Circulating DNA Purification Maxi Kit, Norgen Bioteck Corp) and quantified using the NanoDrop 2000 UV-Vis Spectrophotometer (Thermo Scientific) and PicoGreen dsDNA quantitation kit (Life Technologies) following the manufacturer's protocols.

Genomic microarray hybridization was performed using the OncoScan FFPE Assay Kit (Affymetrix, Inc.) using an optimal input of 80 ng per sample. As ctDNA analyses with this assay have not been verified in large studies, this was a research use only application of the assay. The assay may detect clear gains and losses at input as low as 40 ng (3, 4). The Oncoscan platform contains more than 220,000 Molecular Inversion Probes interrogating approximately 900 cancer genes and relevant somatic mutations, and the assay can detect aberrations in DNA fragments of 50–125 kb (3, 4). The somatic mutation panel contains hotspot 74 mutations over 9 genes (BRAF, EGFR, IDH1, IDH2, KRAS, NRAS, PIK3CA, PTEN, TP53). The complete probe-set information is available in Supplementary Table S1 including copy number variations (CNV) and somatic mutations (SMV).

Data files containing the intensity values for individual probes were converted to probe set analysis results using the OncoScan Console v1.2 Software using the FFPE reference. All seven samples met the recommended quality control thresholds of MAPD (≤0.3) and ndSNPQC (≥26). MAPD is a global measure of the variation of all microarray probes across the genome representing the median of the distribution of changes in log2 ratio between adjacent probes and is a measure of short-range noise in the microarray data. Lower MAPD values are representative of higher quality control. ndSNPQC is a measure of how well genotype alleles are resolved in the microarray data. Larger ndSNPQC values are representative of higher quality control.

Copy number calls were calculated and binned using the TuScan Algorithm and data visualized in Nexus Express OncoScan 3.1 (BioDiscovery). Somatic mutations were scored for each individual, where the MutScores (a measure of the signal response of the marker relative to the expected signal distribution of this marker in the absence of the mutation) exceeded a predetermined threshold per marker (marker thresholds vary from 5–7). CNVs were scored against a global panel of approximately 900 cancer-related genes and individual amplifications, duplications, and regions exhibiting loss of heterozygosity were scored per sample (Supplementary Table S2 and S3). CNVs were then evaluated in the COSMIC database to select for genes implicated in carcinogenesis, and this is reported for all seven cases in Supplementary Table S3. Both somatic mutations and selected CNVs detected by the Oncoscan FFPE Assay were confirmed using targeted NGS (Illumina MiSeq). Specific genomic aberrations identified in Case 1 were validated by quantitative PCR with remaining sample. Variants were selected for Sanger Sequencing validation if they were suspected to be oncogenic from the COSMIC database and had a CNV of 0 or of 3 or higher. Where applicable, samples were compared with tumor molecular profile and plasma cell free DNA molecular profile.

Four individuals had tumor tissue from metastatic sites sent for targeted NGS (Foundation Medicine) before their ascites cfDNA analysis. The assay sequenced the entire exons of 236 cancer-related genes, and select introns from 19 genes often rearranged in cancer (full list available at http://www.foundationone.com/genelist.php#). This method of sequencing allows for detection of copy number alterations, gene rearrangements, and somatic mutations with 99% specificity and >99% sensitivity for base substitutions at ≥5 mutant allele frequency and >95% sensitivity for copy number alterations. A threshold of ≥8 copies was used for gene amplification.

For the cfDNA analyses, barcoded sequencing libraries were generated from 5 to 30 ng of plasma cfDNA. The exons of 54 cancer genes, including all coding exons of 18 genes and the recurrently mutated exons in an additional 36 genes, were captured using biotinylated custom bait oligonucleotides (Agilent), resulting in a 78,000 base-pair (78 kb) capture footprint. Samples were paired-end sequenced on an Illumina Hi-Seq 2500, followed by algorithmic reconstruction of the digitized sequencing signals. The coverage depth across all coding sequences in all samples averaged approximately 10,000X. Illumina sequencing reads were mapped to the hg19/GRCh37 human reference sequence, and genomic alterations in cfDNA were identified from Illumina sequencing data by Guardant Health's proprietary bioinformatics algorithms. These algorithms quantify the absolute number of unique DNA fragments at a given nucleotide position, thereby enabling ctDNA to be quantitatively measured as a fraction of total cfDNA. The limit of detection for single-nucleotide variants in cfDNA by the Guardant360 assay is 0.1% (http://www.guardanthealth.com/).

Cell-free DNA yields ranged from 3.9 ng to 38.4 ng/mL of effusate in the seven subjects evaluated, indicating variability in DNA content. In each case (Tables 1 and 2), analysis of the ascites cfDNA found a range of detectable aberrations, and one individual with cirrhosis (patient 7, Table 2) that served as a negative control had minimal copy number variations detected. All evaluable subjects are listed in Tables 1 and 2 wherein copy number variations (CNV) of ≥3 are listed. All CNVs and CNVs filtered by COSMIC are listed in Supplementary Tables S2 and S3, respectively.

A 63-year-old Asian man (nonsmoker) had a right upper lobe mass detected after work-up for persistent cough. While attempting curative surgery for perceived localized disease, intraoperative pleural deposits consistent with extensive tumor progression were observed and the surgery was aborted. Biopsy of the tumor was consistent with adenocarcinoma. Hot spot testing by PCR-based gene sequencing was negative for ALK, EGFR, or ROS1 aberrations (Response Genetics). The patient completed five cycles of carboplatin and paclitaxel with bevacizumab, and his tumor burden remained stable for 17 months.

About 24 months after diagnosis, the patient was noted to have progressive disease with an increase in the size of the pleural-based nodules. A second biopsy sample was taken from a pleural deposit, and targeted next-generation sequencing (NGS; Foundation Medicine) showed CDK4 and MDM2 amplification, but no somatic mutations (Table 1; Case 1). The individual began a clinical trial with an anti-PD1 checkpoint antibody. About two months later, CT scan showed progression, and he began a subsequent trial with a molecularly matched CDK4 inhibitor. He was taken off study because of toxicity with fatigue, nausea, and anorexia. His condition deteriorated, and metastases to the peritoneal cavity manifested with increased abdominal girth and ascites. He needed biweekly paracenteses for symptomatic relief. cfDNA from the ascites was analyzed for CNV using the OncoScan FFPE Assay Kit (ThermoFisher, Inc.) for research use. The results demonstrated clear high copy EGFR amplification (copy number = 15, Table 1, Case 1). EGFR amplification was verified by quantitative PCR (Supplementary Fig. S1). The patient died without receiving anti-EGFR–directed therapy.

An 84-year-old woman was found to have a locally advanced breast cancer. Core biopsy of the primary tumor was estrogen receptor (ER) 90% positive, progesterone (PR) negative, Her2Neu negative. The patient was started on an aromatase inhibitor and had a marginal response, and then underwent a mastectomy for local control. About three years after the initial diagnosis, the patient was noted to have a liver metastasis on imaging which was confirmed as metastatic breast carcinoma on biopsy. She was treated with a vaccine-based immunotherapy, but had progression with peritoneal carcinomatosis within 7 months of the trial initiation. The patient was treated with capecitabine at that time, but had significant gastrointestinal toxicity. NGS of her archival tissue was performed. She was noted to have an activating PIK3CA mutation (PIK3CA E542K) in the tissue, and the patient was transitioned to exemestane and everolimus therapy. During this treatment regimen, the patient showed tumor regression in the omental lesions and peritoneal thickening on imaging (Figure 1), and showed a decrease in tumor marker CA15.3 from 585 to 341 U/mL. The patient eventually required hospitalization for pneumonitis thought to be drug induced by the everolimus, and the exemestane and everolimus were discontinued. The patient was then given additional chemotherapy with capecitabine and navelbine but, upon progression, opted for a palliative treatment plan. Of note, the ascites cfDNA from a palliative paracentesis showed the original PIK3CA E542K mutation found in the tumor tissue NGS. Plasma cfDNA also showed this alteration (Guardant360 assay, 54 gene assay; http://www.guardanthealth.com/guardant360/).

Estimation of ascitic cell-free DNA size based on fluorescence-based capillary electrophoresis demonstrates an effusate cfDNA fragment size of approximately 150–160 bp and a smaller peak between 300–400 bp. A representative capillary electrophoresis profile is shown in Fig. 2. These fragment sizes correlate with those fragment sizes seen in plasma which are typically centered around 166 bp, approximately the length of DNA wrapped around a nucleosome plus linker, and a minor peak at approximately 340 bp has also been reported (5, 6). Within 30 mL of effusate material obtained from the seven subjects, DNA yield ranged from 119.3 to 1,153.2 ng (Tables 1 and 2). The MIP probes have 2 arms which are each about 20 bp long and 1 bp between the two arms. This makes the smallest fragment of intact DNA that may be recognized approximately 45–50 bases long.

Subject number 1 with metastatic lung cancer showed high copy gain of EGFR (CNV = 15) in ascites cfDNA that was not found in the analysis of solid tumor tissue (Table 1; Fig. 3A and B, Case 1). Primary tumor tissue showed only CDK4 and MDM2 amplification. Altogether, the ascites cfDNA sample from Subject 1 displayed 16 copy number alterations (Supplementary Table S2) and those that are cited within the COSMIC database are listed in Table 1 and Supplementary Table S3. The HFN1B gene on chromosome 17 (copy number = 11; Fig. 3A and B) and the EGFR aberration was further verified by quantitative PCR assay (Supplementary Fig. S1). These alterations were distinct from the CDK4 and MDM2 amplification identified in the solid tumor tissue. Detailed characterization of the genomic data revealed that 0.31% of the cfDNA genome detected in the ascites was impacted by chromosomal changes.

For Subject number 2, the ascites cfDNA detected a PIK3CA E542 point mutation that was also present in the original tissue biopsy and in blood based cfDNA analysis (Table 1; Fig. 4A and B, Case 2). This mutation was confirmed in the ascites by directed NGS through Illumina MiSeq v2 chemistry. Furthermore, the patient received PIK3CA-directed therapy with everolimus and had response with decrease in the size of her tumor (Fig. 1) and declining tumor marker CA15.3. At the time of the individual's paracentesis, which was 6 months after treatment with everolimus, a total of 318 copy number alterations were observed in the ascites (Supplementary Table S2). High copy gains (copy number = 6) were observed in chromosome 11 (CCND1), and copy loss of NF1 gene was detected in chromosome 22 (Fig. 4A and B). The alterations were consistent with tissue NGS results from the tumor biopsy. Additional alterations were seen, and a significant number of copy number alterations were observed to cluster on chromosome 12.

The remaining four individuals with cancer had genetic alterations identified that were subsequently filtered by correlating those with the COSMIC database (Table 2; Supplementary Table S3). One individual with cancer (Table 2, subject 4) with cytology-negative disease did not have high expression of CNVs in their pleural effusion, and the subject likely had an effusion from an etiology other than malignancy. An individual with cirrhosis showed only minor alterations in their ascites (Table 2, subject 7; Supplementary Table S3).

There is emerging evidence that cfDNA isolated from blood or plasma can be used to detect DNA shed from metastatic solid tumors (7–9). To the best of our knowledge, our data demonstrate, for the first time, that analysis of cfDNA isolated from ascites may provide important genomic information regarding an individual's cancer that complemented and expanded data obtained from tissue biopsies. Our index subject, a 63-year-old man (nonsmoker) with metastatic lung adenocarcinoma displayed a 15-fold amplification of the EGFR gene in ascites cfDNA analysis, but not in two separate lung biopsies. EGFR amplification was confirmed by quantitative PCR of DNA from ascites. We speculate that this EGFR amplification existed as a small subclonal population and/or was not detectable at the time of the tissue biopsy because of sampling bias. Importantly, this individual was never treated with anti-EGFR therapy because tissue analysis did not show EGFR alterations, and the subject died of disease shortly after ascites was analyzed. This case provides a rationale for the sampling of multiple fluid spaces in the future for important decision-making.

Data suggest that EGFR amplification in tumor tissue, in addition to activating EGFR tyrosine kinase mutations, correlate with sensitivity to anti-EGFR mAb-directed therapies (10–13). Necitumumab (an approved treatment for squamous lung cancer) is a second-generation, recombinant, human immunoglobulin G1 EGFR mAb that binds to EGFR with high affinity, competing with natural ligands and preventing receptor activation and downstream signaling. In murine non–small cell lung cancer xenograft models, the addition of necitumumab to gemcitabine and cisplatin resulted in a substantial increase in antitumor activity (14). In a recently published large multicenter clinical trial for the anti-EGFR antibody necitumumab in squamous cell lung cancer (SQUIRE), overall survival was more favorable in individuals with tumors harboring high EGFR expression assessed by IHC (13). Necitumumab has subsequently been approved by the FDA combined with a platinum-based doublet chemotherapy. An additional study assessed disease response to gefitinib in patients with EGFR amplification as detected by FISH, and there was a nonstatistically significant trend toward benefit. In this study, disease control was demonstrated in 63% of EGFR/FISH-positive patients versus 39% of FISH-negative patients (10). Further studies may consider the value of CNV data compared with IHC and FISH to direct therapy against EGFR amplification and characterize its association with response to anti-EGFR–based therapy.

Our second index subject was a woman with breast cancer who had a PIK3CA E542K mutation identified in her tumor tissue, cfDNA from blood plasma, and cfDNA from ascites. This subject received the approved regimen of everolimus with antiendocrine exemestane therapy, which correlated with radiologic tumor regression on CT scans (Fig. 1) and a serologic response in protein tumor markers. PIK3CA E542K is an activating mutation found on exon 9 of the PIK3CA gene (15). The BOLERO-2 trial which assessed the efficacy of everolimus plus exemestane for Her2Neu-negative breast cancer showed a greater progression-free survival benefit with everolimus compared with placebo for patients with a PIK3CA exon 9 mutation compared with those with other PIK3CA mutations (16, 17). A recent correlative analysis of genetic alterations and everolimus benefit in hormone-positive breast cancer demonstrated more efficacy in individuals with exon-specific mutations in PIK3CA exon 9 compared with exon 20 (17). Preclinically, in MCF7 breast cancer cells harboring the PI3KCA E542K mutation, the combination of everolimus with antiestrogen therapy can inhibit proliferation and trigger apoptotic cell death (18). Individuals who had higher chromosomal instability did less well than those who had lower chromosomal instability (17).

In the cases represented, analyses of ascites cfDNA demonstrated the presence of tumorigenic CNVs in cancer-associated genes (Supplementary Table S2 and S3). For Subject 2, PIK3CA E542K mutation, NF1 loss, and high copy CCND1 amplification was found in both tissue and ascites. For Subject 1, the CDK4 and MDM2 amplification present in the lung tumor was not detected in the ascites. Complete concordance between tumor tissue DNA and cfDNA may be confounded by timing of sample collection, therapy that a patient has received, and the use of different sequencing assays. It remains possible that some subset of tumor cells have a specific molecular profile that may be important in cell adhesion, migration, and metastasis to the peritoneal cavity, which would lead to these alterations to be disproportionally detected in ascites cfDNA.

Taken together, this analysis provides a potential rationale for a strategy of molecular analysis across fluid spaces including tissue, plasma, and effusions to provide a comprehensive analysis of relevant targets. Some of the limitations of this study include the fact that further analyses will need to be done to set appropriate thresholds geared toward therapeutic action-ability and address the hierarchy of subclonal populations. The significance of detecting discordant alterations in primary tissue versus blood versus effusions is unclear, and further studies will need to assess therapeutic predictive capacity with all versus any one of these modalities. This was a small sample size, and a larger study is underway to evaluate the predictive capacity of genetic alterations in effusions caused by cancer of various histologies, and an understanding of the predictive value of characterizing viable cells within various effusates versus effusate cfDNA. Furthermore, the Oncoscan was used for research use only to illustrate a biologic concept and, as the assay has not been verified for ctDNA evaluation, and further evaluation will be needed to characterize any indications beyond its intended use.

Our data demonstrate that cfDNA containing potentially actionable oncogenic alterations such as somatic mutations and CNVs can be discovered in ascites and pleural effusions. This information may be of significance when developing therapeutic strategies and anticipating different mechanisms of resistance. Through serial collection, one can obtain a longitudinal perspective of genomic change through time and under selective pressure by therapy. The ability to access viable cells in ascites and pleural effusions is a distinguishing factor which may have important implications in the creation of patient derived xenograft models and RNA and protein analyses. Ongoing studies are investigating concordance between tissue DNA, plasma cfDNA, and ascites cfDNA at the same time point and in a larger cohort of samples.

E. Fung is an employee at ThermoFisher Scientific. R. Kurzrock has ownership interest in Novena, Inc. and Curematch, Inc., reports receiving a commercial research grant from Genentech, Merck Serono, Pfizer, Sequenom, Foundation Medicine, and Guardant, and is a consultant/advisory board member for Sequenom, Actuate Therapeutics, and XBiotech. Hatim Husain has received research support from Trovagene, Inc. and Affymetrix and served on an advisory board for Foundation Medicine. No potential conflicts of interest were disclosed by the other authors.

Conception and design: H. Husain

Development of methodology: H. Husain, G. Gomez, E. Fung

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H. Husain, D. Quan, G. Gomez, B. Woodward, S. Venkatapathy, R. Duttagupta

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H. Husain, D. Nykin, N. Bui, S. Venkatapathy, R. Duttagupta, E. Fung, S.M. Lippman, R. Kurzrock

Writing, review, and/or revision of the manuscript: H. Husain, D. Nykin, N. Bui, R. Duttagupta, S.M. Lippman

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H. Husain, D. Nykin, B. Woodward, S.M. Lippman

Study supervision: H. Husain, B. Woodward

This project was funded by institutional funds at The University of California, San Diego. Funded in part by the National Cancer Institute grant P30 CA016672 (R. Kurzrock).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.