By shedding light on the cellular origins of circulating DNA (cirDNA), this research provides important insights into the mechanisms of cirDNA production in cancer. Contrary to expectations, the increased cirDNA in patients with cancer was not derived predominantly from neoplastic cells or surrounding nonneoplastic epithelial cells; rather, the excess cirDNA originated primarily from leukocytes, implying a systemic impact of cancer on cell turnover or DNA clearance.
Bert Vogelstein's previous research has provided groundbreaking paradigms of the molecular basis of cancer and the genetic mechanisms governing tumor progression. Now, the leading team he has built over the past two decades offers a striking contribution to the analysis of novel, tumor-derived genetic material present in patients’ blood: circulating DNA (cirDNA or cell-free DNA as noted in Mattox and colleagues). Reporting work done in collaboration with the eminent teams of Drs. Jeanne Tie and Dennis Lo, they describe, within this issue of Cancer Discovery (1), a new paradigm of the cells of origin of cancer patients’ cirDNA.
Numerous studies have observed elevated total cirDNA concentrations in the blood of patients with cancer as compared with cancer-free controls (2). The precise origins of this excess cirDNA, however, have remained elusive, hampering any clinical exploitation of the phenomenon. Within this context, and with the ultimate goal of identifying the major tissue source of elevated cirDNA levels, Mattox and colleagues (1) have conducted an in-depth analysis of cirDNA methylation patterns in both patients with cancer and individuals without cancer. Focusing specifically on colorectal, lung, ovarian, and pancreatic cancers, their study involved a retrospective examination of clinical cohorts, comprising 178 patients with cancer and 64 controls.
Mattox and colleagues’ study focused on patients with cancer with a high level of cirDNA concentration (>15 ng/mL). CirDNA was extracted from standard EDTA blood plasma before tumor resection. Additional samples collected 24 hours after pancreatic tumor resection were also analyzed. The total concentration of cirDNA was determined by qPCR analysis, with any possible contamination by genomic DNA from leukocytes being excluded by an assessment of the fraction of high-molecular-weight DNA using the RealSeqS assay or Agilent BioAnalyzer. To determine the contribution of each cell type, more than 5,000 differentially methylated regions were analyzed using whole-genome sequencing of bisulfite-treated cirDNA. The researchers then applied various matrices and algorithms of deconvolution analysis, which is the most advanced and comprehensive method of determining the cellular origin of cirDNA (3). Colorectal and pancreatic cancer cirDNA samples were also examined using SafeSeqS mutation analysis and copy-number variation detection by methylation data analysis to estimate the part of DNA derived from neoplastic cells (1).
Building on Yuval Dor's contributions to the deconvolution analysis of cirDNA assay (3), the study by Mattox and colleagues offers further intriguing insights into the source of elevated cirDNA levels. Their results show that leukocytes, particularly neutrophils, are the major contributors of cirDNA in both noncancer individuals and patients with cancer, regardless of their cirDNA levels (1). For both patients with cancer and controls, more than half of the total cirDNA detected originated from leukocytes, with two thirds of this deriving from neutrophils. As for lymphocyte-derived cirDNA, B cells and T cells contributed significantly to the cirDNA pool to a degree contrary to what would be expected from their proportions in the circulation. Minor tissue contributors included the liver, colon, heart, brain, and lungs. As compared with controls, patients with colorectal cancer showed higher cirDNA contributions from colonic epithelium, and it was determined that this cirDNA was derived from the neoplastic cells themselves. Thus, contrary to expectations, the study found no evidence that neoplastic cells or surrounding nonneoplastic epithelial cells were major sources of elevated cirDNA in patients with cancer.
This study also explored cirDNA dynamics following surgery, which is typically associated with increased cirDNA levels. Interestingly, the majority of postsurgery cirDNA in patients with pancreatic cancer was also derived from leukocytes, as it is before surgery. Furthermore, there was a striking increase (46-fold on average) in cirDNA derived from hepatocytes after surgical intervention, which is probably due to postsurgery liver damage (1).
In view of these findings, the researchers speculate that the excess cirDNA seen in patients with cancer both before and after surgery is due to a systemic effect rather than localized tissue damage. Having identified an increase in DNA release from various tissue sources, including leukocytes, hepatocytes, and colon and lung epithelial cells, they conclude that such effects point to a systemic cause (1). This finding also suggests that cancers may affect cell turnover or DNA clearance, leading to increased cirDNA contributions from all major tissue sources. They further hypothesize that this may be caused by the secretion of certain molecules by neoplastic or endothelial cells in tumors. Another possibility they entertain involves a potentially impaired clearance of cirDNA in patients with cancer.
In this area, oncology has previously focused almost exclusively on the cirDNA of malignant cells [or mutant cirDNA (cir-mutDNA)]. Consequently, measurement of total cirDNA concentration has to date been poorly investigated, in particular as a single marker. Where it has been investigated, this has tended to be only in the context of (i) the postsurgery presence of residual or recurrent cancer, (ii) the association of cirDNA concentration in postoperative samples with overall survival, and (iii) the evaluation of treatment response (2). Despite the intriguing observation of its high variation among all patients with cancer (with or without the same malignancy), in particular the demonstration of its value as a prognostic factor, cirDNA concentration has been generally overlooked by the oncology community. Rectifying or at least addressing this oversight, Mattox and colleagues’ in-depth examination of cancer patient cirDNA cells of origin reveals the principal, systemic nature of cirDNA release. Their work is supported by previous studies of the cirDNA methylome, which indicates that cirDNA mainly originates from leukocytes and to a far lesser extent in various other tissues (3–6).
Although the significance of this study must therefore be acknowledged, several limitations do exist: (i) to date, the qPCR method for quantifying cirDNA has not been validated; (ii) analysis of plasma from heterogeneous cancer staging (I–IV) and malignancies was combined; (iii) the low number of patients hinders the possibility of comparing tumor types; (iv) it would have been interesting to know the evolution of cirDNA levels in serial postsurgery analysis; and (v) the authors offer a number of different hypothesis in support of their observation, but no conclusive explanation is provided.
One explanation not proposed by Mattox and colleagues is the possibility of a systemic release of cirDNA deriving from neutrophil extracellular traps (NET). Involved in the immune innate response that occurs within the first hours of infection, the release of NETs physically and chemically eliminates microbes. NET production has also been shown in sterile diseases, especially inflammatory diseases. In 2015, our team was among the very first to postulate that cirDNA found in cancer patient plasma may originate from NETs (2). Subsequently, we demonstrated that NET degradation leads to the release of mononucleosomes (7), which constitute the vast majority of cirDNA-associated structures. The historical observation of mononucleosomes (and to a lesser extent dinucleosomes) as the principal structures associated with cirDNA led to apoptosis being considered the main source of cirDNA (2). Our direct observation had previously revealed the importance of NET formation as a significant mechanism of cirDNA release. cirDNA concentration in patients with newly diagnosed metastatic colorectal cancer associated with NET protein markers (8). As is the case with other mechanisms such as apoptosis, necrosis, and active cellular or microvesicular release, the specifics of the contribution of NET formation to cirDNA release have yet to be elucidated. Nonetheless, its variation according to individual physiopathology is a definite possibility. Note that extracellular traps are generated from other blood cell types, but these mechanisms too are still poorly known. The significance of NETs as a principal source of cirDNA would appear to be supported by the study of the cirDNA fragmentome and methylome given the revelation that cirDNA originates mainly from leukocytes, especially neutrophils (1, 3). Further supporting this assertion is the fact that elevated levels are associated with diseases in which uncontrolled NET formation is observed, such as in lupus, sepsis, and COVID-19 (7).
Irrespective of the various systemic effectors of mechanisms of cirDNA release, as proposed by Mattox and colleagues (including molecules secreted by neoplastic, endothelial, and inflammatory cells, and abnormal cirDNA clearance in patients with cancer), or as we ourselves propose (NETs), the fact of their systemic nature would appear to be a clear indication that they depend mainly on the specific individual rather than on the tumor. Consequently, the value of the proportion of cir-mutDNA among total cirDNA fragments and, as a result, the mutation allele frequency (MAF), would depend on the patient-specific rate of background nontumor cirDNA released by different cells/tissues rather than sensu stricto on the rate of cir-mutDNA release by the tumor. This would explain the high variation of MAF (0.003% to 90%) observed in patients showing similar RECIST scores (9). In addition, this would cast serious doubt on the value of cirDNA-derived MAF as a surrogate biomarker for cancer in clinical settings, in particular when used for selecting mutation for the detection of minimal residual disease to limit adjuvant therapy. Furthermore, given the systemic nature of the cirDNA release mechanism, higher scrutiny should be applied to the specificity and sensitivity of mutation detection scoring. Similarly, the direct relationship previously thought to exist between cirDNA and cancer stage or tumor volume (2, 10) should also be reconsidered.
The implications of Mattox and colleagues’ findings are far reaching and highlight the further research needed to elucidate the complex mechanisms underlying cirDNA release and clearance in various clinical scenarios, including cancer and postsurgery conditions. Such research would further our understanding of the relationship between cancer progression, immune responses, and cirDNA, as well as the mechanisms that lead to elevated concentrations of cirDNA. In so doing, they would provide insights that would prove invaluable in the development of more effective diagnostic and monitoring tools in oncology.
Authors’ Disclosures
A.R. Thierry reports the following patents: WO 2012/028746 A1, WO 2016/063122 A1, WO2019110750 (A1), EP 18 305788.4, EP 19 30 6003, and 11194720 PCT application number PCT/EP2022/072147. E. Pisareva reports the following patent: 11194720 PCT application number PCT/EP2022/072147.
Acknowledgments
We are grateful to M. Ychou and T. Mazard for fruitful discussion. We also thank Cormac Mc Carthy (Mc Carthy Consultant, Montpellier) for English editing (financial compensation). The study was supported by a SIRIC Montpellier Cancer Grant INCa_Inserm_DGOS_12553 and by a REVEAL RHU ANR grant.