Current understanding of extrachromosomal circular DNA in cancer pathogenesis and therapeutic resistance

The underlying molecular mechanisms that lead to the biogenesis of extrachromosomal circular DNA have yet to be fully clarified. Both DNA damage and the corresponding DNA repair strategies, the important events of genomic homeostasis observed particularly in tumorigenesis [23‐26], are processes that have been proven to produce extrachromosomal circular DNA species [27‐29] (Fig. 3). The microDNA is often observed from GC rich regions, especially within the 5′ and 3′ UTRs [30], which largely share with the regions susceptible to the formation of R-loops [31, 32]. These data suggested that R-loop-associated DNA damage response may contribute to eccDNA generation. Similarly, DNA-damaging agents, such as hydroxyurea (HU) [33], have helped provide evidence for the role of DNA damage repair (DDR) in extrachromosomal circular DNA production. Specific proteins in DDR, for example, DNA-PKcs (a central player in nonhomologous end joining (NHEJ)) [34] and MSH3 (involved in mismatch repair (MMR)), are necessary for the formation of extrachromosomal circular DNA [35]. In support of these findings, Dillon et al. [31] noted that deletion of the DNA repair protein MSH3 resulted in a dramatic decrease in the amount of circular DNA in human ovarian and prostate cancer cell lines. In addition, immunoblot analysis has shown that depletion of DNA-PKcs by shRNA or inhibitors causes the decreased amplification of dihydrofolate reductase (DHFR) on ecDNA and the elimination of ecDNA in the methotrexate (MTX)-resistant colon cancer cell line HT-29 [36]. This finding supports the notion that NHEJ is important for the formation of ecDNA. Moreover, the depletion of the Ku70/80 heterodimer significantly decreases the level of circular extrachromosomal DNA and inhibits the proliferation of human osteosarcoma SAOS2 cells [37]. Another DDR pathway, homologous recombination (HR), is also involved in extrachromosomal gene amplification. The attenuation of HR activity by shRNA-mediated BRCA1 depletion reduces the amount of extrachromosomal circular DNA and circular DNA-amplified DHFR in MTX-resistant colon cancer cells [38]. However, in contrast to these findings [38], defects in the HR protein RAD54 were shown to induce a markedly increased number of subpopulations containing extrachromosomal circular DNA in MTX-resistant subclones from HeLa cervical cancer cells [39]. This contradiction might be due to the lack of sufficient repetitive elements in the HeLa cell genome for the recombination pathway to loop out extrachromosomal circular DNA. To some extent, additional compensatory mechanisms, such as NHEJ, could contribute to the formation of extrachromosomal circular DNA after RAD54 inhibition, but this requires further exploration.

Previous reports support the general idea that extrachromosomal circular DNA might originate from either repeat or nonrepeat sequences within a genome. A survey of tandem repeat sequences in eccDNA found that chromosomal tandem repeats were excessively represented in eccDNA from the HeLa and SW620 cancer cell lines [40]. Moreover, the transcription of tandem repeat sequences can potentially lead to the accumulation of eccDNA via intrachromosomal HR between tandem repeats, such as extrachromosomal rDNA circles (ERCs), ultimately leading to genome instability [41, 42]. However, in HeLa S3 cells, nearly half of the eccDNA fragments were entirely composed of nonrepetitive DNA sequences [43]. In human lymphoblastoid cell lines, an analysis of microDNA features showed that abundant microDNA is mainly derived from nonrepetitive genomic regions [30], supporting the contribution of nonhomologous recombination in circular DNA production. Thus, these unique sequences could serve as valuable resources for the generation of circular DNA. Moreover, as dysfunction of DNA repair pathways causes the accumulation of DNA damage involved in tumorigenesis [44], this underlying knowledge provides novel insights into the formation of circular DNA by genomic sequences and might present promising genomic strategies to restrict tumor progression by influencing circular DNA abundance.

As mentioned above, although DNA damage responses are now available to generate ecDNA, the specific mechanisms that control DM generation in cancers are still not fully understood (Fig. 3). Previously, chromothripsis was proven to explain the origin of DMs in some cancer cells [45]. Chromothripsis is a single-step catastrophic event that drives local chromosome breakage and end-to-end chromosomal fusion and serves as the underlying driving force linked to DMs. For example, the genesis of complex DMs in several cancer types, including glioblastoma (GBM) and small cell lung cancer, has been explained by the telomeric fusion of chromosomes broken by chromothripsis [46, 47]. Recently, another model for DM genesis was proposed, named the episome model [48]. In support of this idea, a growing body of research evaluates the fine structure of c-Myc-containing DMs in different types of cancer cells [49, 50]. These findings consistently identify a stepwise recombination process, starting from single-chromosome amplified episomes, for the generation of c-Myc-containing DMs. Moreover, different sizes and architectures of c-Myc amplicons can coexist within the same cell population, providing evidence of DM heterogeneity. Intriguingly, several fusion genes or circular RNA of plasmacytoma variant translocation 1 (PVT1) can occur frequently in cells with c-Myc amplicons, further highlighting the role of PVT1 as a breakpoint hotspot for chromosomal breaks during c-Myc amplification and synergistically contributing to tumorigenesis [51]. Other mechanisms of DM generation have also been reported, such as breakage-fusion-bridge (BFB) cycles [52, 53], although their detailed roles have yet to be determined. In brief, during mitosis, the BFB cycle involves anaphase bridge formation due to the presence of dicentric chromosomes, followed by bridge breakage, which generates a variety of chromosomal aberrations, including DMs.

Owing to different genetic backgrounds and tissue types, these circular DNA molecules have broad heterogeneity in size [54], including small circularized DNA entities, later referred to as eccDNA, and larger ecDNA (Table 1). Subsequent studies revealed that the quantity of circular DNA has been widely reported for different types of cancer and might have the potential to amplify and drive intratumor genomic heterogeneity. Because of the advanced molecular and functional profiles, eccDNA/ecDNA elements are now understood to be crucial to oncogenetic plasticity [47, 55] and may be associated with poor prognosis [56]. More importantly, the biological roles of DNA circularization in the regulation of cancer-associated genes facilitate the more effective adaptation of cancer cells to variable environmental stresses, propagating tumor pathologies and drug resistance [57].

The chimeric circularization and amplification of circular DNA is very common and typically has a profound impact on the enhanced expression of oncogenes. However, it is unclear whether circularization itself or subsequent copy number amplification induces upregulation of the various oncogenes. To address this question, a recent investigation conducted by Koche and colleagues [17] demonstrated that the majority of oncogene amplification as determined by WGS seemed to coincide with circular DNA amplification in neuroblastoma. Likewise, by applying haplotype phasing analysis, the authors found that circular DNA is mainly derived from amplified alleles. Together, these findings confirmed that extrachromosomal DNA circularization served as an important driver of the high level of oncogenic amplification. These observations further suggest the use of circular DNA as a re-emerging target for the development of novel anticancer therapeutic strategies.

Unexpectedly, however, some oncogenes with variable copy numbers contained in circular DNA were not altered in neuroblastoma cells [17], indicating that extrachromosomal circular DNA might be required but insufficient alone to increase gene amplification. Based on this striking phenomenon, we hypothesized that circular DNA might have additional important functions in tumor pathogenesis. To clarify this issue theoretically, subsequent quantitative assessment of the chromatin state found that ecDNA could also strongly promote the transcription of full-length or truncated oncogenes by integrating back into the active chromatin with their intact domain structure [4, 112].

To date, progress in genetic research has revealed additional roles of circular DNA beyond its function in promoting oncogene amplification or transcription. This circular DNA may also drive oncogenic remodeling in human cancers, resulting in significantly adverse clinical prognoses [113]. Owing to a lack of integrative and highly sensitive strategies to characterize circular DNA in cells [114], such genomic rearrangements have often remained relatively underestimated or undetected by previous WGS methods. By applying a sensitive analysis algorithm that combines WGS datasets with the complementary sequencing of circular DNA (Circle-seq), Koche et al. [17] generated a landscape of tumor-specific circular DNA molecules in human neuroblastoma cells. The authors concluded circle-derived somatic rearrangements followed by reintegration of the circular DNA molecules into the linear genome. The authors also reported that most of the somatic chromosomal rearrangements were from the regions of DNA circularization, providing a scientific basis for the proposed hypotheses that circular DNA plays a profound role in genome remodeling. One can imagine that such a widespread reintegration phenomenon might have important effects on chromosomal gene aberrations, especially in the overexpression of oncogenes and the activation of oncogenic kinase signaling [115].

Today, multiple therapeutic methods have been widely used for cancer patients in clinical practice, including standard chemotherapy and targeted therapy. However, drug resistance is an almost universal challenge in cancer therapy, ultimately leading to tumor recurrence and treatment-related morbidity [116]. The cellular and molecular mechanisms of therapy resistance are quite complicated and involve multiple factors and signaling pathways [117]. For example, extrachromosomal amplified DNA and its regulatory capacity have been determined to play key roles in driving tumor heterogeneity and regulating therapy resistance in various cancers [118, 119], including colon cancer and GBM (Table 2). Defining the extrachromosomal DNA that drives the amplification of drug-related genes would aid in providing novel therapeutic perspectives to reverse resistance and ameliorate outcomes. Extrachromosomal circular DNA could promote the amplification of drug resistance-associated genes, and it is a powerful driver of intercellular genetic heterogeneity [120]. The classical MTX resistance gene DHFR was proven to amplify primarily in the form of DMs, contributing to tumor progression and the development of MTX resistance in human colon cancer cells. The dramatic impairment of HR [38] and NHEJ [36] repair activity could suppress the level of DHFR-amplified DMs and considerably revert the sensitivity of HT-29 cells to MTX but not affect intrachromosomal DHFR amplification. These findings support previous conclusions that targeting DHFR in cancer cells harboring amplified DHFR-containing circular DNA has attractive therapeutic potential [121]. However, substantial work is urgently needed to investigate the factors that determine the sensitivity of DHFR-amplified cancers in anti-DHFR treatment.

In addition, there is evidence that regions of genomic rearrangement, such as genomic focal amplification sites in human cancers, might be closely related to strong gain-of-function mutations in oncogenes, contributing to the development of drug resistance in cancer therapy [122]. Circular DNA elements act as the cytogenetic hallmarks of genomic focal amplification in cancer cells [123], and the copy number of these elements with oncogenic mutations is altered in response to environmental changes [107]. An example of this is in nervous system neoplasms, including GBM and low-grade glioma, where the receptor tyrosine kinases (RTKs) are frequently mutated and commonly give rise to the driver variant [124]. In particular, the activating mutations of EGFR [125] and PDGF receptor a (PDGFRa) [126], two members of the RTK family, have usually been thought to reside primarily on DM structures. These extrachromosomal mutations are the driving factors that contribute to therapy resistance by increasing tumor heterogeneity. Studies have also indicated that although the constitutively active mutant of EGFR, EGFR variant III (EGFRvIII), plays critical pro-survival roles in GBM pathogenesis, and it also makes cancer cells more sensitive to EGFR inhibitors [127]. Therefore, resistance to EGFR-targeted therapy might develop if extrachromosomal mutations are eliminated. In recent years, several empirical studies have been conducted to address this issue. After treatment with the EGFR inhibitor erlotinib, EGFRvIII-bearing extrachromosomal DNA elements within GBM cells were markedly reduced, resulting in resistance to anti-EGFR therapeutics. However, after withdrawal of the drug, the reemergence of clonal EGFRvIII mutations on DMs effectively resensitized cancer cells to EGFR inhibitors, which could then induce cell death [128]. Unexpectedly, in some cancer cell subclones, resistance to anti-EGFR therapy might not be entirely mediated by EGFRvIII-positive DMs; rather, resistance might occur through other compensatory mechanisms, such as the potential oncogenic roles of extrachromosomal MDM2 amplification [128]. In addition, Vicario et al. [129] analyzed the amplification patterns of HER2 (another member of the RTK family) in response to different anti-HER2 therapies. Although the acquisition of resistance in HER2-positive breast cancer cells is often concomitant with HER2 protein loss, the copy number of HER2-positive DMs did not obviously change after treatment with the HER2 inhibitor trastuzumab. These data indicate that HER2 elimination-mediated resistance might not be due to the loss of HER2-containing circular DNA. Thus, future work is needed to evaluate the compensatory mechanisms that enable cancer cells with or without RTK-positive extrachromosomal DNA to continue to proliferate during anti-RTK treatment. By elucidating these dynamic regulatory mechanisms in detail, we will gain more comprehensive insight into therapy resistance, and this understanding could support the future development of novel therapeutic approaches.

Other extrachromosomal driver mutations could also occur during the amplification of DMs, and this might be critical for tumor heterogeneity and therapy resistance. By applying exome sequencing to seven GBM tissue samples, Nikolaev et al. [130] identified a novel class of extrachromosomal gain-of-function mutations mediated by focal amplification. These amplification-linked extrachromosomal mutations (ALEMs) were found in multiple members of the RTK family, such as EGFR and c-KIT. Because of the unequal segregation of DMs between daughter cells during cell division, some daughter clones may inherit a higher copy number of DMs carrying driver ALEMs and thus obtain a growth advantage (Fig. 4). This unequal segregation effect, also known as “hitchhiking,” might be widespread across different types of cancers [131]. Interestingly, based on the “hitchhiking” mechanisms, DM-originating ALEMs could enable cancer cells to undergo rapid adaptation to environmental stresses, thereby acquiring resistance to anticancer therapies [130]. Overall, these studies emphasize that extrachromosomal DNA contributes to therapy resistance in cancer cells in multiple ways. These mechanisms could be exploited in further research and future clinical trials.

Tumor-specific features of extrachromosomal circular DNA elements could be important for the development and progression of malignancies and for therapy resistance [8]. It is therefore important to explore the extrachromosomal DNA profiles of patients with advanced cancers in clinic (Table 3). It has been proposed that extrachromosomal circular DNA is the dominant manifestation of amplified drug-resistance genes and oncogenes in patient-derived tumor tissues and can be identified in multiple types of cancers [132, 133]. Therapeutic strategies that could reduce extrachromosomal DNA elements in cancer cells would likely downregulate the gene expression that contributes to aggressive phenotypes and drug resistance in advanced cancers. Indeed, increasing in vitro and in vivo preclinical evidence has demonstrated that the loss of DMs, accompanied by reduced levels of DM-form oncogenes (c-Myc, MYCN, MDR1, etc.), can be achieved upon drug treatment and further enhance the therapeutic sensitivity [134‐136]. Likewise, the results from independent studies have found that elimination of the amplified drug resistance genes on extrachromosomal circular DNA would be feasible therapeutic strategies to make the cells more sensitive to chemotherapy drugs, such as HU and MTX [137, 138]. Based on these preclinical findings, Raymond et al. recently performed a single-arm study of 16 patients with advanced ovarian carcinomas and found that orally administered HU at a noncytotoxic dose decreased the DM copy number in cancer cells and improved the prognosis dramatically [139]. Thus, noncytotoxic strategies that eliminate DMs could be helpful in personalizing treatment decisions in cancer patients. Although these findings are interesting, future clinical trials to verify the efficacy of noncytotoxic strategies are now warranted for ovarian cancer and other types of cancers with frequently amplified DMs.

In addition, it has recently been proven that some extracellular DNA is enriched in cancer-derived extracellular vesicles (EVs), further contributing to their stability [140]. In human body fluids, these cancer-derived EVs carrying circular extrachromosomal DNA has been suggested to trigger a multiplicity of physiological and pathological responses [141]. These EVs can also functionally transfer fragments of some genomic oncogenes, such as H-ras [142] and K-ras [143], and result in the oncogenic transformation of epithelial cells. It was further observed that EVs containing extrachromosomal DNA can activate the paracrine signaling with neighboring cells, ultimately contributing to malignant biological behaviors and modulation of anti-cancer therapy [144, 145]. A recent work conducted by Read et al. showed that EVs derived from prostate cancer cell lines transferred EGFR and androgen receptor (AR) into extracellular space, which enhances the malignant phenotypes of neighboring cells [146]. Of course, this comprehensive knowledge provides important and new insights into the utilization of extrachromosomal circular DNA in EVs as biomarkers for cancer research and treatment.

In light of their important roles in oncogene plasticity and drug resistance, circular DNA elements can be detected in the blood-based screening of plasma from cancer patients as a component of liquid biopsy [147, 148]. Especially, several studies have indicated the potential application of extrachromosomal DNA elements in body fluids as candidate biomarkers for the diagnosis and monitoring of several disorders, such as cancers [149]. Because of their extraordinary structural stability than linear DNA, circular DNA in the blood might serve as an ideal source of potential biomarker exploration [150]. In support of their value as clinical monitoring tools, several recent reviews highlighted that the length and abundance of circular DNA elements have been more prominent in the cancer patients compared with the healthy donors [9, 133]. The analysis of pre- and post-surgery plasma from cancer patients revealed significantly decreased levels of eccDNA in blood circulation following tumor resection [35]. Accordingly, Kumar’s group recently identified the microDNA released from cancer cells in serum and plasma from cancer patients and characterized these unique molecular profiles after anti-cancer treatment [94]. Likewise, the abundance of DMs in peripheral blood lymphocytes has been demonstrated to function as an independent risk factor for patients with lung cancer [151]. Sequencing analysis of eccDNA profiles revealed thousands of unique circular DNAs in the circulatory system, with a highly heterogeneous molecular mass profile; the size distribution ranged from several base pairs to several kilobase pairs [150]. Thus, the discovery of cancer-associated circular DNA in the circulation provides a scientific basis for further investigations of the use of these DNA elements as biomarkers for precision cancer therapies. The real-time monitoring of extrachromosomal circular DNA in circulating blood has provided novel insights into cancer pathology. The clinical application of detecting circular DNA in liquid biopsies to identify tumorigenesis and progression is attractive; however, some problematic issues remain. The most important is a lack of reliable methods for easily quantifying the abundance of circular DNA elements in the blood [152].

While circular extrachromosomal DNA has been able to be determined for some time, higher-resolution imaging and cytogenetic methods have only recently been developed to evaluate these molecules. Advances in sequencing technologies have allowed for the genome-scale screening and mapping of circular DNA from human cancer cells [153]. Based on short-read sequencing and connecting amplified DNA segments, some computational analysis tools have been developed to explore the profiles of circular DNA signals in tumor samples in an unsupervised manner [154], such as AmpliconArchitect [155]. Defining eccDNAs using AmpliconArchitect showed that oncogenes mainly amplify eccDNA elements and are surprisingly preserved during cancer progression, which has been shown in previous reports [11]. However, these short-read maps could not unambiguously align the long-range duplications crossing the circular structure. As a consequence, the characterization and functional profiles of circular DNA elements might have been drastically underestimated in cancer patients. To effectively solve these problems, the detection systems are expected to be improved in subsequent studies. Recently, several improved tools with high precision, such as AmpliconReconstructor [131] and Circle-Map [156], were developed to better map the physical structure of long contiguous reads. Based on the BioNano technology platform, AmpliconReconstructor was used to provide more conclusive evidence that oncogenes carrying ecDNAs produced substantially more transcripts in multiple cancer cells and tissue types. Today, other higher-resolution technologies are being developed to improve the resolution of circular DNA from high-throughput DNA sequencing data. For example, a higher-resolution image analysis algorithm, named ECdetect, has been developed and is used to effectively quantify ecDNAs from DAPI-stained metaphases in an unbiased and highly accurate fashion [11, 55]. Taken together, these effective tools will permit a deeper understanding of the characteristics of extrachromosomal DNA particles in cancer pathology and their association with clinical features.

It is also essential to expand the cytogenetic toolbox to explore the function of key components with respect to extrachromosomal DNA particles. For a long time, fluorescence in situ hybridization (FISH)-based approaches have been used to resolve amplified oncogenes linked with cancer development. However, FISH probes have been unable to discriminate the location of specific genes on chromosomal and extrachromosomal DNA molecules [157]. A more advanced cell-imaging technology, ecSeg [158], has been introduced to accurately quantify oncogene amplification from ecDNA at the single-cell level. Moreover, the application of genome editing techniques, such as CRISPR-Cas [159, 160], could be leveraged to advance our understanding of the basic biologic functions of these molecules, especially in circular DNA-driven oncogene amplification and drug resistance. Moreover, as delineated above, such methods could help researchers discover agents that target extrachromosomal circular DNA elements.