Breaking Up Isn't Hard to Do: Isolating Cell-free DNA Fragments in Osteosarcoma

In this issue of Clinical Cancer Research, Udomruk and colleagues present data from pretreatment blood specimens from a retrospective cohort of 50 patients with osteosarcoma (1), the most common malignant bone tumor in children and adolescents. They profiled cell-free DNA (cfDNA) at diagnosis and determined healthy controls had the largest median fragment size of 177 bp while those with metastatic osteosarcoma had the shortest, with a median size of 151 bp. They then showed that patients with osteosarcoma harboring cfDNA with a short-length peak of ≤150 bp on fragment analysis had inferior overall survival compared with patients without short fragments. Importantly, the presence of short cfDNA fragments remained significant in multivariable modeling accounting for risk factors such as tumor size and metastatic status. The authors next performed whole-genome sequencing from both the short (≤150 bp) and long fragments (>150 bp) of DNA and found that the short fragments were enriched for mutations compared with the longer fragments which were likely enriched with healthy, germline DNA. Finally, an in silico analysis showed that tumor-derived cfDNA content as measured by copy-number variation was 2.3-fold higher in the short fraction compared with the long fraction.

For patients with osteosarcoma, there have been no recent remarkable therapeutic advances and survival rates have been stagnant. In addition to clinical stage, several biomarkers have been demonstrated to be prognostic, such as primary tumor necrosis following neoadjuvant chemotherapy, which is a strong predictor of outcome (2). Recently, specific somatic copy-number alteration patterns such as MYC amplification have been recognized as predictive of outcome; however, this approach may suffer from the lack of tumor material in this disease (3). Thus, a unified goal of those involved in osteosarcoma research is to have reliable, real-time monitoring of disease to assess tumor burden and response to treatment.

Liquid biopsies represent a potentially inexpensive and noninvasive method to achieve real-time disease monitoring goals. The enhanced detection of circulating tumor DNA by fragment size analysis as applied here for the first time exclusively in osteosarcoma builds on the few published articles describing the use of blood-based biopsies for this disease (4–9), all of which require molecular assays that look beyond point mutations, which are generally nonrecurrent and infrequent (10). Specifically, Shulman and colleagues showed that circulating osteosarcoma-derived cfDNA was detectable at diagnosis in 57% of patients with localized osteosarcoma using low passage whole-genome sequencing (5). While the presence or absence of osteosarcoma-derived cfDNA was not significantly associated with outcome in this cohort, increased tumor fraction at diagnosis was associated with increased rates of mortality. This approach has also been used in cohorts of patients in Europe (7). More recently, detection of aberrant methylation in osteosarcoma-derived cfDNA was also shown to be prognostic in samples taken from patients both before and after surgical resection (6). The current research by Udomruk and colleagues expands on these works by showing that detection of aberrant copy number may be enhanced by evaluating just the small fragment size cfDNA, perhaps allowing for detection below the theoretical limitation of 3% circulating tumor fraction based on the ichorCNA algorithm (11). Thus, it may be that fragment analysis has the potential to improve the sensitivity and specificity of cfDNA assays.

The use of fragment analysis for detecting circulating tumor DNA is intriguing for its simplicity, as typical next generation sequencing–based or PCR-based approaches are often costly, time-intensive, and require significant technical expertise for both sample preparation and downstream analysis (12). For routine tracking of cfDNA to become clinical standard, a rapid turnaround time with easy interpretation is necessary. As this approach appears to have remarkable sensitivity of 100% for distinguishing patients from healthy controls, it may be able to provide an easy to interpret present or absent conclusion. Furthermore, fragment analysis has a rapid turnaround time and requires no costly reagents or expertise for analysis. This is an ideal feature as a clinically applicable test, considering the rarity of osteosarcomas, and the challenge of access to care for many patients. In addition to the clinical utility of fragment length analysis, the ability to enrich for short circulating tumor–derived DNA fragments may also enable tracking of genetic evolution over time through deeper sequencing as also suggested in a recent report from Christodoulou and colleagues (13). This may enable analyses without the need for germline and/or repeated acquisition of tumor material, which is important given the often-limited availability of samples and the invasiveness of biopsies in this disease.

While this study successfully demonstrates the feasibility of cfDNA fragment length analysis as a method for risk stratifying patients with osteosarcoma, several questions remain. First, before this approach could be used in the clinic, further prospective validation in an independent sample set which appropriately accounts for potential differences in pediatric and adult patients, variability in sample processing, and analysis are required. Second, the current analysis demonstrates the ability to stratify patients according to outcome, suggesting that the assay is likely identifying patients with clinically localized disease who may have subclinical metastases and more aggressive biology. Conversely, it is possible that this type of approach may identify patients with metastatic disease that would be more likely to be cured with standard therapy. However, larger patient cohorts are needed to confirm this. Another major opportunity would be to evaluate how this assay may vary over the course of therapy as the current study only evaluated samples at diagnosis. Understanding how the cfDNA fragment length changes over time in multiple subgroups may illuminate how this assay could be utilized for monitoring and help the field tie this information together with emerging and current biomarkers such as deep panel sequencing or percent tumor necrosis following neoadjuvant therapy (4, 14).

The potential of circulating tumor DNA analysis in osteosarcoma is clear, whether as a prognostic biomarker, disease burden marker, or a method to explore mechanisms of treatment resistance. The research by Udomruk and colleagues contributes to the body of evolving literature focused on the application of liquid biopsies in this difficult to cure cancer, where outcomes have not improved for decades. The research team makes a compelling case for integrating this approach to analyzing cfDNA in osteosarcoma (Fig. 1). A major accomplishment of the research team was their foresight to bank diagnostic blood samples from patients, which highlights the importance of a coordinated international effort to enroll all children with cancer in clinical studies that integrate storage of biospecimens.

M.A. Applebaum reports personal fees from Innervate Radiopharmaceuticals outside the submitted work. No disclosures were reported by the other authors.

Relevant funding support has been provided by The New York Community Trust (D.A. Weiser) and National Pediatric Cancer Foundation (D.A. Weiser and M. Hayashi).