Comprehensive profiling of extrachromosomal circular DNAs in colorectal cancer progression Authors: Lu Lu, Mingjie Chen, Guicheng Zhang, Yujing Liu, Xiangyuan Xu, Zenghua Jiang, Yangxian Xu, Tao Liu, Fan Yang, Guang Ji, Hanchen Xu Published: 2024 ABSTRACT AbstractColorectal cancer (CRC) development involves a series of molecular events that drive the progression from normal colorectal epithelium to adenoma and eventually to invasive carcinoma. While the involvement of extrachromosomal circular DNAs (ecDNAs) in cancer genome remodeling has been established, their specific roles in CRC formation remain unclear. Using Circle-Sequencing and whole transcriptomic sequencing, we comprehensively profile circular DNAs and transcriptomes in healthy individuals, colorectal adenoma, and CRC patients. Our delineate analyses characterize the key circular DNAs involved in oncogene expression through the normal-adenoma–carcinoma continuum and highlight that immune response-related pathways and cell cycle pathways, are the dominat events in CRC progression. Notably, chr8 ecDNA 64950741–114379093 exhibits robust up-regulation during CRC progression. Further validation in a new cohort of 50 CRC patients confirms the higher expression of chr8 ecDNA 64950741–114379093 and its strong correlation with poor prognosis. Thus, these findings provide unprecedented insights into the landscape of circular DNAs in CRC and highlights the potential of chr8 ecDNA 64950741–114379093 as a promising biomarker and therapeutic target for CRC management.Keywords: Circular DNAs, chr8 ecDNA 64950741–114379093, CRC progression, BiomarkerSubject terms: Cancer, Genetics, Biomarkers, Oncology Introduction Colorectal cancer (CRC) is a complex disease that manifests as a result of a multi-step progression model, evolving from normal colorectal epithelium, to adenoma, and finally, to adenocarcinoma1,2. The progression initiates with early alterations in the APC genes, followed by further genetic and epigenetic changes, including the amplification and mutations of various oncogenes3. These alterations considerably impact cellular behavior, paving the way for invasive growth and development. Considering the complexity of these transitions, an in-depth understanding from various perspectives is important to unravel the pathogenic mechanisms that drive CRC progression and to develop novel target therapies. Moreover, with the continuous evolution of CRC, early detection with promising biomarkers is increasingly crucial for improving patient outcomes, and ultimately, refine prevention, diagnosis and prognosis for CRC treatment. Extrachromosomal circular DNA (referred to as circular DNA throughout the text) has recently gained significant attention in oncological studies due to its profound role in genomic restructuring and cancer evolution4. Circular DNA can be classified into two primary categories based on its size: eccDNA (small circular DNAs) and ecDNA (large copy number amplified circular DNAs). Characterized by a distinct structure originating from chromatin-related activities, circular DNAs participate the formation of super-enhancers, exhibit increased chromatin accessibility, establish long-range chromatin contacts, and contribute to DNA repair mechanisms5. This unique profile facilitates somatic rearrangements, a process instrumental in tumorigenesis. For instance, computational analysis of whole-genome sequencing data for 3212 cancer patients have revealed that circular DNA amplification is a common event occurred across various cancer types6. As an intrinsic factor, circular DNAs are implicated in a range of vital biological processes, including immune surveillance evasion7, enhancement of metastasis and invasiveness8, stimulation of cell proliferation4,8 and modulation of drug sensitivity9–11. Additional studies indicate that circular DNA amplification promotes intratumoral genetic heterogeneity, thereby accelerates tumor evolution12,13. Notably, circular DNAs are often characterized by their enrichment in amplified oncogenes, resulting in increased levels of oncogene transcription14, an fundamental activity commonly altered in caner. Recent research has illuminated the role of circular DNAs in amplifying oncogenes such as EGFR, MDM2, CDK4, and BRAF, a common alteration in multiple cancer types including CRC6. This process involves the induction of chromosomal rearrangement by circular DNAs, positioning oncogenes in close proximity to regulatory elements such as enhancers or promoters, and consequently leading to the creation of chimeric genes15–17. However, the precise role of particular circular DNAs in the dynamic progression of cancers like CRC remains elusive. Intriguingly, patients with cancers harboring circular DNAs typically experience worse survival outcomes compared to those whose tumors lack circular DNA-driven oncogene amplification6,18,19. This correlation, coupled with the unique structure and inherent stability of eccDNAs, render it an ideal candidate for biomarker identification in cancer research. Thus, studying eccDNA could lead to the identification of new regulators governing the transition from normal epithelium to colorectal adenoma (CRA) to carcinoma in CRC development, as well as yield promising biomarkers for early detection of CRC. Here, by implementing Circle-seq, we comprehensively characterize the molecular properties including the length distribution, genomic positions and regulatory components of circular DNAs in healthy individuals, colorectal adenoma, and CRC samples. Integrating with whole transcriptomic sequencing, we analyze the circular DNA density in conjunction with transcript abundance, revealing their significance in oncogene expression. Importantly, our study highlights immune responses pathways such as antigen presentation that are coupled with corresponding immune cell infiltrations become more prominent during the progression of CRC. One particular ecDNA named chr8 ecDNA 64950741–114379093 involves in this event. Notably, ecDNA chr8 ecDNA 64950741–114379093 shows a gradual increase during CRC development and exhibits a strongly correlation with unfavorable prognosis in an independent cohort of 50 newly identified CRC patients. These findings provide invaluable insights into the potentials of colorectal cancer prevention, diagnosis, and the development of therapeutic interventions. Materials and methods Circle-Seq and data analysis Circle-sequencing was employed to identify circular DNA in CRC, CRA and health control (HC). The process followed Circle-Seq methods previously reported5, as summarized as follows. The samples were first subjected to overnight incubation at 50 °C with the addition of Proteinase K (Thermo Fisher, Waltham, MA, USA). Circular DNAs were then extracted using a Plasmid Mini AX kit (A & A Biotechnology) according to the manufacturer’s instructions. Subsequently, mitochondrial circular DNA was removed using FastDigest MssI (Thermo Scientific), and residual linear DNA was eliminated with Plasmid-Safe ATP-dependent DNase (Epicenter, Madison, WI, USA). The purified samples were then used as templates to amplify circular DNAs by φ29 polymerase amplification (REPLI-g Midi Kit, QIAGEN, Germany). The amplification reactions were carried out at 30 °C for 46–48 h. The library was constructed using NEBNext® Ultra II DNA Library Prep Kit for Illumina (New England Biolabs) and sequenced using an Illumina NovaSeq 6000 sequencer in 150 bp paired-end mode. The raw soft-clipped read counts of the breakpoint were obtained using Samtools software (version 0.2), and the detection of circular DNAs was performed with Circle-map software (v1.1.4). The differentially expressed circular DNAs in each comparison were filtered using edgeR software (v0.6.9). Annotation of the circular DNAs was carried out using the Bedtools software (v2.27.1). Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were performed using clusterProfiler (v3.16) based on the differentially expressed circular DNA-associated genes. Annotation of oncogenes was obtained from the ONGene database20. Transcriptome sequencing and data analysis Total RNA was extracted from 5 replicates for the CRC, CRA, and HC groups. RNA sequencing was performed using an Illumina NovaSeq 6000 sequencer. High-quality clean reads were aligned to the reference genome (UCSC hg19) using HISAT2 software (v2.0.4). Differentially expressed mRNA was identified by calculating the fold change and p value based on the gene level FPKM obtained from the cuffdiff software. GO and pathway enrichment analyses were further carried out using clusterProfiler software (v3.16) on the filtered differentially expressed mRNAs. Statistical analysis All statistical plots and analyses were executed using the R (v4.04) programming package in this study. Significance was determined using unpaired Student’s t test and the χ2 test (for clinicopathological analysis) where appropriate. The R packages survminer (v0.4.9), survival (v3.3-1) and ggplot2 were applied for survival analysis and data visualization. Overall survival (OS) was calculated from the date of CRC diagnosis to the date of death or last follow-up. Survival rates were estimated using the Kaplan‒Meier (KM) product limit method and were compared with the log-rank test to determine significance. KM estimators were adjusted by inverse probability weighting (IPW) because potential confounders existed. Multivariate and univariate analyses were used to assess differences in outcome on the basis of circular DNA level and other baseline features. Differences were statistically significant at *p < 0.05, **p < 0.01, ***p < 0.001. Additional methods The sample collection and verification are described as the Supplementary Methods. Ethics approval and consent to participate The human cancer tissues used in this study were approved by the institute ethical committee of Longhua Hospital affiliated to Shanghai University of Traditional Chinese Medicine. Patient consent for publication We informed consent was obtained from all subjects. Experimental method declaration We confirmed that all experiments were performed in accordance with relevant guidelines and regulations. Results Detection of circular DNAs in HC, CRA and CRC samples by circle-sequencing analysis In an effort to discern the critical factors that occur during the transition from normal epithelium to adenoma and finally to adenocarcinoma, we focused our study on the potential role of circular DNAs. Utilizing the Circle-Seq method21,22, we detected circular DNAs in samples obtained from CRC, CRA, and HC groups. The study comprised five biological replicates for each group, surgically retrieved from Longhua Hospital from 2017 to 2020.The inclusion criteria for CRA patients consisted of a polyp biopsy pathologically diagnosed as tubular adenoma, villous adenoma, or tubular choriocapillaris adenoma. CRC patients included in the study had preoperative colonoscopic biopsy pathology-confirmed colorectal adenocarcinoma and underwent radical surgery. Histological examination revealed neatly arranged glands in the HC group. In contrast, pathological staining of CRA patients exhibited elongated glands with crowded nuclei and deep staining. Furthermore, the histopathological features of CRC samples demonstrated infiltrative tumor growth and marked gland dysplasia, with cells arranged irregularly (Fig. 1A). These distinctions in pathological features provide insight into the different stages of disease progression within the samples studied. Through the application of Circle-Seq (Fig. 1B), we isolated nucleic acids, removed linear genomic DNA, and constructed a sequencing library following rolling-circle amplification. The sequenced reads were mapped and annotated, leading to the identification of various types of circular DNAs in our study, including exonic, intergenic, intronic, and sense-overlapping ones. Circular DNA detection in HC, CRA and CRC samples by circle sequencing. (A) Sample collection and clinical diagnosis. Entroscopy (upper panel) and H&E (lower panel) staining of HC, CRA and CRC. (B) Illustration of Circle-Seq procedures. DNA was extracted from clinical samples, and circular DNAs were purified by removing mitochondrial circular DNAs and residual linear DNA. Comprehensive analysis and description of circular DNA landscape in CRC The landscape of circular DNA in CRC has been seldom thoroughly explored. In our study, we underscored biological variation and characterized two types of circular DNA: ecDNA (length > 25 kb) and dominant eccDNA (length < 25 kb). Circle-seq identified over 80,000 distinct circular DNAs in 15 tissue samples, with varying numbers commonly detected in all replicates within each group (HC, CRA, and CRC) (Supplementary Table 1, Fig. 2A). EccDNA accounted for > 99% of all circular DNAs detected, with 52,509, 75,974 and 60,126 ecDNAs detected in HC, CRA and CRC samples, respectively (Fig. 2A). Our data consistently identified a total of 1401 confident circular DNAs in at least 3 of the 5 biological replicates (Fig. 2B), which were annotated to 47 oncogenes, including well-identified MYCN and EGFR (Fig. 2C, Supplementary Table 2). Among the CRC-specific confident circular DNAs, 2% (119 out of 3377) were derived from oncogenes. In addition, the analysis of length distribution revealed consistency across the three groups, with the eccDNA curve peaking around 400 bp and ecDNA primarily showing a median distribution of 33.04 Mb (Fig. 2D). All chromosomes contributed to circular DNAs, with exceptions noted in chromosomes 22 and Y (Fig. 2E,F). A higher circularization density was observed in CRA (Fig. 2G), with eccDNA relatively evenly spread across all chromosomes and ecDNA mainly found on chromosome 2. Our findings also aligned with prior studies regarding the occurrence level in repeat sequences, including SINEs, LINEs, satellites, centromeres, and other repeat regions in the human genome (Fig. 2H). Moreover, we detected disproportionately high levels of 5ʹ UTR and 3ʹ UTR and relatively low levels in the Alu region in the mapping results, which supports the nonrandom generation of circular DNAs (Fig. 2I). These findings not only expand our understanding of the role of circular DNAs in CRC but also highlight the potential importance of these structures in oncogenesis. Characterization of circular DNAs in HC, CRA and CRC. (A) Histograms showing the number statistics of eccDNA (red) and ecDNA (blue) in each group. (B) Venn diagram showing the overlap of confidential (at least 3 out 5 in each group) circular DNAs among the three groups. (C) Venn diagram demonstrating 1401 overlapping confidential circles annotated to oncogenes. (D) Length distribution of ecDNA and eccDNA (in right side window). (E) Chromosome distribution of ecDNA normalized to chromosome length, and ideogram representing the density of ecDNA origins. (F) Chromosome distribution of eccDNA normalized to chromosome length and an idogram displaying the position of ecDNA origins. (G) Groupwise comparison of the chromosome distribution of circular DNAs. (H) Circular DNAs annotated to functional regions normalized to region size. (I) Circular DNA percentage annotated to repeat regions. Features of eccDNAs and ecDNAs detected in a comparative study between CRC and HC samples Previous study has established that circular DNA can stimulate accessible chromatin, leading to elevated oncogene expression17. In our study, we initially correlated the frequencies of 119 consistently identified oncogene-derived circular DNAs with their corresponding host gene mRNA transcript levels within each group. Examples of high-density circular DNAs leading to increased mRNA transcripts include EPCAM, ERBB3, CDH17, CCND1 and KLF5 and CDX2 in CRC samples (Fig. 3A), which is consistent with recent findings23–28. Investigating the relationship between DNA circularization and transcription levels, we compared the CRC and HC groups using a cutoff standard of an absolute fold change of 2 and a P-value of 0.05. We identified 34 eccDNA candidates (24 upregulated, 10 downregulated) and 38 ecDNA candidates (29 upregulated, 9 downregulated) that overlapped (Fig. 3B,C, Supplementary Fig. S1B, Supplementary Tables 3, 4). To further elucidate the changes in circular DNA and its corresponding elements, we analyzed the fold change of each transcript and corresponding circularized DNA. We observed a diminished count in counterparts for shorter-length eccDNAs. Importantly, our study unveiled new oncogene-linked circularities and transcriptions, particularly ALK, CCNE1, NCOA3, and TLE1, with a predominant localization in the first and second quadrants (Fig. 3D). Considering that ecDNA originates from genomic regions containing multiple full-length oncogenes, this observation suggests that DNA circularization could lead to the simultaneous upregulation of mRNA and offers potentially valuable insights for understanding the development of CRC. Comparative analysis of eccDNAs and ecDNAs in CRC and HC samples. (A) Circular DNA counts relative to the average transcription level of CRC and HC samples. (B,C) Venn diagrams showing differentially expressed mRNAs and eccDNA (B) and ecDNA (C) with the same host genes (|fold change| > 2 and p value < 0.05) in the CRC vs. HC comparison. (D) Scatter plot showing differentially expressed eccDNA (left) and ecDNA (right) with the same host genes. Red dots indicate oncogenes, with arrows marking the newly-identified key oncogenes. Identification and analysis of circular DNAs in colorectal cancer progression Next, we aimed to investigate the main contributors of circular DNAs in the progression of CRC development. Using a cluster-based bioinformatics approach to analyze circular DNA expression (Supplementary Fig. S2), we discerned two distinct trends involving 5281 circular DNAs, showing a progression from healthy individuals to CRA and culminating in CRC (Fig. 4A). A heatmap visually represented these unique expressions of circular DNA across the Health-CRA-CRC groups (Fig. 4B). To refine the data, we initially introduced a filtering strategy that targeted circular DNAs detectable in at least 20% of all samples. And then collect the circular DNAs showing a progressive, yet significant increase from HC, to CRA, and finally CRC, aligned with the two observable trends identified earlier. Last, following validation through a public eccDNA atlas database, we found 699 circular DNAs of varying lengths and expressions that satisfied all the established criteria (Fig. 4C). To discover the upstream drivers in potential circulation of these DNAs, we explore the patterns within the sequence of flanking circular DNA junction sites, particularly recurrent nucleotide motifs that might hint at a preference for DNA circularization21,29–31. Our of binding motifs based on these 699 circular DNAs showed significant enrichment of ZNF770, ZEB1, TEAD1 and HIF1A (Fig. 4D). These transcription factors, equipped with splicing functions, could potentially facilitate the circularization of certain circular DNAs in an innovative manner. Notably, 699 circular DNAs, encompassing over 2000 genes, are involved in various functions during the transformation from normal epithelium to adenoma to adenocarcinoma. Leveraging an unbiased clustering methodology, we discerned these transcriptomic variations into five distinct clusters. A pronounced upward trend was evident in clusters C4 and C5, illustrating the intricate dynamics of CRC progression and highlighting the key signaling pathways (Fig. 4E). Prominent oncogenes, notably PIK3CA and GAB2, were consistent markers across all analyzed groups. Additionally, we observed regulators fundamental to cell cycle and DNA replication dynamics, such as CCND1, FOSL1, and E2F5, becoming progressively more prevalent as CRC advanced (Fig. 4E). Of note, pathways such as Toll-like receptor signaling, cell cycle, DNA replication, antigen processing and presentation, and Wnt signaling ranked among the most enriched (Fig. 4E). From our findings, 13 circular DNAs, encompassing crucial genes pivotal to these pathways, underscoring their potential significance in CRC evolution (Fig. 4F). Collectively, these insights shed light on the molecular intricacies underlying CRC pathogenesis. Unraveling circular DNAs in the evolution of colorectal cancer. (A) Cluster analysis via the Mfuzz package identified two clusters with increasing trends based on Circle-Seq. (B) A heatmap illustrating differential circular DNA expressions in three groups. (C) A comprehensive Circos map: The outer ring marks chromosome positions, with red bars highlighting circular DNAs from our cluster analysis. Blue bars represent gene distributions related to these circular DNAs, while green indicates their lengths. Corroboration with the eccDNA atlas database is presented in the innermost blue rings, revealing 699 pivotal circular DNAs. (D) Upper, examination of trinucleotide sequences adjacent to the initiation and termination loci of circular DNA. Lower: Enrichment assessment of motifs recognized by transcription factors. (E) Heatmap indicates enriched signaling trajectories and oncogenes in five clusters, highlighting specific oncogenes and representative pathways via KEGG enrichment. (F) A Sankey diagram displays circular DNAs-genes-pathways connections, emphasizing the significance of clusters 4 and 5. Identification and analysis of transcriptomes in the progression of colorectal cancer In order to thoroughly delineate the normal epithelium-adenoma-adenocarcinoma sequence on multiple levels, we utilized RNA-seq to profile gene expression along with the expressions of circular DNA across identical samples. Intriguingly, our transcriptome analysis unveiled specific gene profiling and identified two distinct trending clusters as well (Fig. 5A). Of note, more than 3000 genes were identified from these clusters and were found to be characterized in pathways associated with the cell cycle, NF-kB signaling, and immune response-related pathways including T cell receptor signaling pathway and antigen processing and presentation (Fig. 5B,C). Within these pathways, a marked amplification of 353 genes, consistent with those pinpointed as crucial circular DNAs-associated entities, was manifested with CRC progression (Fig. 5D). To further corroborate these findings, we examined immune cell infiltration by employing a de-convolution strategy that used average expressions of key genes to represent the corresponding immune cell types. Our analysis of immune cell infiltration revealed that dendritic cells, macrophages, dendritic cells, naïve B cells, and memory CD4+ T cells (Fig. 5E), all of which are engaged in antigen presentation, are predominant in CRC groups, as opposed to their lesser presence in healthy groups. The congruence of these data points toward immune response, cell cycle, as primarily altered routes in the transition, which is consistent with the findings from circular DNA sequencing. Exploring the transcriptomic progression in colorectal cancer. (A) Mfuzz cluster analysis captures two notable trend-ascending clusters based on RNA-seq. (B) Heatmap represents differential gene expression across the three groups. (C) Pathway enrichments trace the transcriptomic shifts in CRC development, with dot sizes indicating gene percentage in pathways. (D) Line chart depicts hallmark gene expression within the top enriched pathways.353 genes, identified as central entities in relation to circular DNAs, are accentuated during CRC progression. Representative genes in the cell cycle and antigen presentation domains are distinguished by color. (E) Estimated immune landscapes across groups. Immune cell infiltration profiles were established through the CIBERSORT method, revealing 20 distinct immune cell types. Characterize the increase of ecDNA chr8 ecDNA 64950741–114379093 is associated with poor prognosis in CRC patients We further investigated the prognostic role of circular DNAs in early CRC detection which is based on the 699 circular DNAs identified along the normal epithelium–adenoma–adenocarcinoma transition. Considering the size of circular DNAs displays characteristics specific to both tissue and disease, we refined our analysis by examining the length distribution, specifically focusing on feature peaks at roughly 1 kb and 400 bp, with top-rated fold changes. Eventually, we collected one ecDNA (chr8 ecDNA 64950741–114379093) and four eccDNAs (including chr1 eccDNA 165876369–165876745, chr6 eccDNA 3616397–3616760, chr13 eccDNA 20602532–20603506, chr1eccDNA 216066357–216066753) for additional validation, all of which demonstrated a progressive increase during CRC progression (Fig. 6A,B). We designed divergent primers to magnify sites specific to circular DNA in HC samples and two CRC specimens. The ensuing results demonstrated successful amplification of the targeted sequences, and the products were verified by Sanger sequencing (Supplementary Fig. S3). Our analysis confirmed that chr8 ecDNA 64950741–114379093 is an ecDNA encompassing various coding and noncoding genes, whereas the other four eccDNAs seem to be partially derived from genomic areas (Supplementary Table 5). Gel electrophoresis revealed that these circular DNAs were amplified in CRC samples, yet the sequence of one particular product, eccDNA (chr1eccDNA 216066357–216066753), did not match the anticipated sequence composition, leading us to exclude it from further investigation (Supplementary Fig. S3). Prognostic implications of amplified ecDNA chr8 64,950,741–114,379,093 in CRC. (A) Flowchart delineating a selection process, spotlighting 1 ecDNA and 4 eccDNAs. (B) Expression trajectory of the identified five circular DNAs throughout CRC progression. (C) IPW-adjusted KM curve analysis of the high- and low-abundance groups, p value as indicated. (D) Univariate Cox regression and multivariate Cox regression analysis (*p < 0.05; **p < 0.01; ***p < 0.001). In our study, we included another independent cohort of 50 CRC patients, quantifying the four candidates using real-time PCR. The detailed information of each patient is shown in Supplementary Table 6. To eliminate the bias introduced by possible factors including age, sex, N stage, T stage, and M stage, we applied the adjusted IPW methods to adjust the KM curve (Fig. 6C). This adjustment revealed a poor survival with high levels of chr1 eccDNA 165876369–165876745 (p = 0.0432) and chr8 ecDNA 64950741–114379093 (p = 0.0221) (Fig. 6C). Additionally, univariate analysis confirmed chr8 ecDNA 64950741–114379093 (median hazard ratio is 2.193, 1.367–3.518) as a detrimental survival indicator, independent of tumor stages and age (Fig. 6D). Most importantly, our findings identified chr8 ecDNA 64950741–114379093 as one of the prominent circular DNAs significantly associated with antigen-related immune cell infiltration (Supplementary Fig. S4), aligning with a crucial event in CRC progression. Together, this discovery suggests the potential of chr8 ecDNA 64950741–114379093 as an influential factor in these immune responses, further underscoring its significance in the context of CRC oncogenesis. Discussion The complex relationship between circular DNAs and colorectal cancer progression reveals new avenues for understanding the disease at a molecular level. In an in-depth exploration of CRC, CRA, and HC samples, the present study provides significant insights into the types, distribution, functions, and potential clinical relevance of circular DNAs. Utilizing Circle-Seq, a method far more sensitive for detecting eccDNA than whole-genome or ATAC-seq32, our study robustly characterized abundant eccDNAs with a length distribution around 400 bp, while ecDNA distribute with a median of 33.04 Mb. The study further revealed that circular DNAs are produced by all chromosomes except for chromosomes 22 and Y, an insight that extends our understanding of chromosomal behavior. Contrasting with the findings in Moller’s paper, where chromosomes 17 and 19 were more prominent (Fig. 2E), our data pointed to a more even distribution of the genic source across chromosomes. This divergence could be explained by an averaging effect, particularly as the healthy control samples within our study showed elevated levels of chromosomes 19 and 17. Further enriching the characterization, our study detected substantial reads mapping to a variety of repetitive elements in the human genome, reinforcing the ubiquity of circular DNAs across all examined types. Moreover, the generation of these circular DNAs seemed far from random, a pattern underlined by an overrepresentation in the 5ʹ UTR and 3ʹ UTR and a converse paucity in the Alu region—a phenomenon consistent with previous research33,34. This specific distribution may hold significant implications for understanding the underlying genomic processes and warrants further exploration. The transition from normal colorectal epithelium to adenoma and ultimately to adenocarcinoma is a multistep process characterized by genetic and epigenetic changes3. Our findings demonstrate that eccDNAs are instrumental in this progression by enhancing oncogene expression. Our study identified 119 CRC-specific circular DNAs, including well-known genes like EGFR, MYCN, EPCAM, previously found in glioblastoma and gastric cardia adenocarcinoma23,35–37. We also uncovered potential oncogenes such as PIK3CA, FOXO1, CCND1, KLF5, which were rarely linked to circular DNA occurrence. Unlike the shorter lengths of eccDNA, which rarely encompass full-length gene, ecDNA can contain more than one gene body. We notably discovered that upregulated ecDNAs simultaneously host several heightened oncogenes, like chr8 ecDNA 64950741–114379093 and chr1 ecDNA 73884064–236889168 (Supplementary Fig. S5). However, this observation doesn't necessarily prove that a large DNA loop can amplify the expression of several oncogenes, and additional evidence is required to understand the underlying regulatory mechanisms. In addition, exploring circular DNA’s biogenesis and mechanisms of formation, our research investigated the sequence patterns at circular DNA junction sites, finding similarities to Sarah T. K. Sin’s work and also identified a “spacer” in the downstream edge but reduced at the start38. This dual-direction-repeat pattern, previously suggested as a generation mechanism, may induce DNA circularization through homologous recombination5. Additionally, inspired by Matthew’s work on architectural proteins facilitating DNA loop formation39, we explored potential binding in these sequences through motif enrichment analysis, which could unravel the transcriptional regulatory mechanism in DNA circularization for further study. Research shows that circular DNAs in specific inflammatory diseases and cancers often play a role in regulating cell growth and immune-related signaling functions, thereby contributing to tumorigenesis, particularly through pathways like WNT and RAS40,41 In our study, we pinpointed immune response-related pathways, cell cycle pathways as the dominant forces in CRC formation, as evidenced by both circular DNA mapped genes and clustered transcriptomes enrichment analysis. Specifically, we confirmed the utilization of 699 circular DNA mapping genes in immune responses, including TCR signaling, CD28 family co-stimulation, MHC presentation pathways, and Wnt signaling, among the top 20 enriched pathways in Reactome and GO analysis (Supplementary Fig. S6). This suggests the immune system as a primary responder to tumor evolution and these circular DNAs may serve dual roles in immunomodulation. They could either encompass full gene transcripts pivotal to certain pathways or act as potent immunostimulants, which is supported by recent studies elucidate that the immunostimulatory potential of eccDNAs hinges not on their specific sequences, but rather on their inherent circularity through STING signals42. Amid the complexity of CRC screening via colonoscopy43, our research discovered a specific indicator, chr8 ecDNA 64950741–114379093, that shows a consistent increase during CRC development, enhancing our understanding of how eccDNAs may be linked to adverse prognoses. This correlation between eccDNA accumulation and cancer phenotype emphasizes the potential utility of eccDNAs as promising biomarkers. The role of eccDNAs in tumorigenesis is emerging as a significant area of interest, particularly their function in the amplification of key oncogenes and contribution to DNA repair mechanisms. Despite these advances, the precise role of specific eccDNAs in the dynamic progression of cancers remains an area ripe for further exploration. Future research focusing on the mechanistic understanding of eccDNA formation, function, and its impact on cellular behavior will undoubtedly advance the development of innovative strategies for early detection, prevention, diagnosis, and treatment of CRC. In conclusion, the characterization of eccDNAs not only enhances the understanding of CRC at the molecular level but also provides a foundation for future research endeavors that may lead to innovative diagnostic and targeted therapeutic interventions for this complex disease. Supplementary Information Supplementary Figures. Supplementary Information. Supplementary Table 1. Supplementary Table 2. Supplementary Table 3. Supplementary Table 4. Supplementary Table 5. Supplementary Table 6. Acknowledgements This work is supported by National Nature Science Foundation of China (82322076 to H.C.X., 82004004 to F.Y.), Shanghai Frontiers Science Center of Disease and Syndrome Biology of Inflammatory Cancer Transformation (2021KJ03-12 to G.J.), Shanghai Rising-Star Program (20QA1409300 to H.C.X.), the Program for Young Eastern Scholar at Shanghai Institutions of Higher Learning (QD2019034 to H.C.X.), Young Elite Scientists Sponsorship Program by CACM (2021-QNRC2-B26 to H.C.X.). Author contributions H.C.X., F.Y., L.L., and G.J. conceived and designed this study. Y.J.L., X.Y.X., Z.H.J., Y.X.X., and T.L. contributed to the collection of the patient materials and clinical information. L.L., M.J.C., and G.C.Z., prepared the sequencing of tissues, and performed the sequencing data mining. The manuscript was drafted by F.Y., L.L. and H.C.X, reviewed by all authors. H.C.X., F.Y., and G.J. supervised all aspects of this work. Data availability All high-throughput sequencing data were deposited in the public GEO database (GSE226741). Data relevant to the study are included in the article or uploaded as supplementary information. The datasets used during the current study were available from the corresponding authors on reasonable request. Competing interests The authors declare no competing interests. Footnotes Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Fan Yang, Email: fanyangforever@gmail.com, Email: fan.yang@shsmu.edu.cn. Guang Ji, Email: jg@shutcm.edu.cn, Email: jiliver@vip.sina.com. Hanchen Xu, Email: hcxu@shutcm.edu.cn, Email: hanson0702@126.com. Supplementary Information The online version contains supplementary material available at 10.1038/s41598-024-70455-x. References Associated Data Supplementary Materials Supplementary Figures. Supplementary Information. Supplementary Table 1. Supplementary Table 2. Supplementary Table 3. Supplementary Table 4. Supplementary Table 5. Supplementary Table 6. Data Availability Statement All high-throughput sequencing data were deposited in the public GEO database (GSE226741). Data relevant to the study are included in the article or uploaded as supplementary information. The datasets used during the current study were available from the corresponding authors on reasonable request.