Liquid Biopsy: It's the Bloody Truth!

In this issue of Clinical Cancer Research, Tukachinsky and colleagues from Foundation Medicine evaluated the clinical utility of comprehensive genomic profiling (CGP) of cell-free circulating tumor DNA (ctDNA) for use as predictive biomarkers in patients with metastatic castration-resistant prostate cancer (mCRPC; ref. 1). The goal of predictive oncology is to develop analytically and clinically validated biomarkers that can identify patients suitable for matched targeted biologic or immunologic therapies that would not otherwise be offered, leading to better treatment outcomes. Such approaches can optimize care delivery through improved survival or reduced costs and toxicities related to the use of ineffective agents. In advanced prostate cancer, predictive medicine has become standard of care in 2020 with the demonstration of improved survival with olaparib in men with metastatic mCRPC and certain homologous repair deficiencies and the durable efficacy observed with pembrolizumab in microsatellite instability high (MSI-H) mCRPC. In addition, the identification of germline variants in DNA repair enzymes can have a major impact on familial cancer risk and screening, and thus both somatic and germline testing has been recommended in National Comprehensive Cancer Network and many other global prostate cancer guidelines.

Tukachinsky and colleagues (1) assessed the genomic landscape of plasma ctDNA from a heterogenous group of patients from the TRITON clinical trials of rucparib as well as a deidentified clinical cohort of men with advanced disease. The TRITON cohorts included patients who had progressed on one prior androgen receptor signaling inhibitor (ARSi) and those who had progressed on—one to two lines of ARSi followed by taxane-based chemotherapy. The Foundation Liquid assay evaluated 62–70 genes and was designed to detect common and potentially actionable substitutions/indels, rearrangements, and amplifications in ctDNA, including MSI status.

Overall, 94% of patients were found to have detectable ctDNA with a median ctDNA fraction of 7.5% with an interquartile range of 0.8% to 33.7%. This heterogeneity clearly represents the heterogeneity of clinical disease burden based on prognostic factors known to relate to ctDNA fraction such as the pattern of spread to bone or visceral sites, PSA levels, symptoms, lactate dehydrogenase (LDH), and other factors (2). Similar to this previous work, Tukachinsky and colleagues found that the ctDNA fraction was consistently higher in patients who had progressed on more lines of therapy, likely representing greater disease burden. The authors found oncogenic and deleterious genomic alterations in 79.5% of all patients, the most common of which were alterations in TP53 and AR, consistent with previous work on ctDNA profiling of patients with mCRPC by Sonpavde and colleagues (3). Only 3%–5% of samples were uninformative due to lack of detectable tumor DNA, and 20% of patient lacked any detectable alterations, but this proportion is likely to be much higher in clinical practice without specific criteria for when to perform testing.

In this analysis, the authors identified at least one DNA damage repair (DDR) gene alteration in 30% of samples. This included 8.9% with BRCA1/BRCA2 alterations, 5.9% with CDK12 alterations, and 1.4% with MSI-H status, thus this assay may identify nearly a third of patients with presently actionable alterations that could benefit from a novel predictive oncology agent with FDA approval such as olaparib, rucaparib, or pembrolizumab. While this study has major limitations due to the lack of clinical outcomes, this kind of landscape research remains important to understand the appropriate disease settings and potential implications of testing.

The authors were also able to detect androgen receptor (AR) alterations in 42% of patients. These alterations included mutations known to predict ARSi resistance and rearrangements previously reported in patient with prior ARSi exposure. However, this assay is not optimized for the detection of AR genomic structural rearrangements nor upstream AR enhancer gains which may be clinically relevant. AR amplifications were detected in just 13% of patients but this number increased to 41% in a subset of patients with ctDNA fraction ≥ 20%. Finally, ctDNA was compared with patient-matched tissue samples to evaluate for concordance. Tissue samples collected a median 758 days before plasma and showed 75% positive percent agreement (PPA) whereas samples collected within 30 days of ctDNA had a much higher PPA. This difference in concordance between remote and contemporary tissue biopsies highlights the ability of ctDNA to capture genomic alterations that have evolved over time with therapy. For example, 2.4% of patients with BRCA1/BRCA2 alterations were seen exclusively in liquid biopsy with BRCA1 mutations being more frequent in patients with prior taxane chemotherapy, and MSI-H disease is more frequently somatic than germline in men with mCRPC. Furthermore, alterations captured exclusively in ctDNA may represent the clonal heterogeneity of metastatic prostate cancer not identified with a tissue biopsy sample. In a study by Wyatt and colleagues on the concordance of ctDNA alterations compared with time-matched tissue biopsy, 33% of somatic mutations were detected exclusively in liquid biopsy samples (4).

In clinical practice, CGP of ctDNA can inform management in patients that have progressed on standard therapy. Patients have often been exposed to hormonal therapy for up to several years by the time they develop mCRPC which can lead to new and consequential somatic alterations since the time of initial tissue biopsy. A metastatic tissue biopsy can provide helpful information but a majority of metastatic lesions exist in bone and biopsy yield is highly variable. An important clinical advantage of ctDNA is its minimal invasiveness and its ability to capture the clonal heterogeneity and tumor evolution over time in metastatic progression that may not be detected in a single tissue sample. As the authors show here, liquid biopsy can detect clinically relevant AR ligand-binding domain alterations, such as W742L/c, T878A, and L702H, which suggest resistance to ARSi therapy and glucocorticoid agonism that could guide steroid or treatment switching. Somatic and germline alterations in DDR detected by ctDNA as well as specific MSI and tumor mutation burden assays can potentially predict response to immune checkpoint and PARP inhibition, which can provide substantial clinical benefit in this patient population.

As discussed by the authors, one identified limitation in CGP of cell-free DNA (cfDNA) is that it can detect confounding genomic alterations and lead to false interpretation of the results. For example, a study by Jensen and colleagues showed that genomic alterations in DDR related to clonal hematopoiesis of indeterminate potential (CHIP) can be detected in cfDNA testing and result in the incorrect conclusion that a patient is a PARP inhibitor (PARPi) candidate, such as for ATM mutations (5). CHIP is relatively common in older men, and may explain the lack of responses to PARPis in some men with monoallelic DDR alterations. In addition to the potential for these false-positive findings, cfDNA testing can miss important actionable alterations such as deletions of BRCA1/BRCA2 and PALB2 (which can lead to patients with undetected DDR alterations) and TP53/RB1 loss (which results in neuroendocrine phenotype and sensitivity to platinum chemotherapy).

Another identified limitation is that with cfDNA alone, we are also left without clinically useful disease phenotypic information at the RNA or protein level. For example, detection of AR splice variant 7 (AR-V7) in circulating tumor cells (CTC) in patients with mCRPC is associated with significantly shorter survival with abiraterone or enzalutamide but not taxane chemotherapy and could lead a clinician to alternative therapy (6). Additional CTC biomarkers are emerging related to neuroendocrine prostate cancer, prostate-specific membrane antigen (PSMA) expression, and other aspects of mCRPC disease biology that could complement ctDNA assays in treatment selection.

In conclusion, Tukachinsky and colleagues have provided comprehensive evidence to support the routine clinical use of ctDNA in patients with mCRPC for the identification of patients with actionable genomic alterations in DDR and with AR alterations that predict ARSi resistance. Further study is clearly needed linking these findings to patient outcomes and benefit; however, this work highlights the wide variability in ctDNA fraction, which directly impacts the sensitivity of the test to detect important gene alterations. In this study, ctDNA fraction is higher in patients who have progressed on more lines of therapy consistent with prior work that show ctDNA increases with greater disease burden. Thus, clinicians should be mindful of the patient context for ordering these now FDA-approved and College of American Pathologists/CLIA (Clinical Laboratory Improvement Amendments)-certified assays. Factors such as disease risk/volume, progression, and prior therapy exposure and available treatment options when selecting patients for liquid biopsy (Fig. 1). In addition, clinicians should recognize the limitations of ctDNA testing that include low sensitivity for deletions, which can lead to false negatives for DDR alterations, and inability to test for AR splice variants that predict poor outcomes for ARSi therapy. To address these limitations, future work should consider the integration of phenotypic biomarkers such as in CTCs, exosomes, or other blood-based assays, or through molecular imaging, which can provide clinically relevant information to guide treatments and improve outcomes for patients.

A.J. Armstrong reports grants and personal fees from Merck, BMS, Astellas, Dendreon, Bayer, AstraZeneva, Janssen, and Pfizer; personal fees from Clovis; grants from Genentech, Constellation, and Beigene outside the submitted work. No disclosures were reported by the other author.