Abstract Liver metastasis is the main lethal cause of colorectal cancer (CRC). The knowledge of role and mechanism of circular RNA (circRNA) in liver metastasis of CRC is still inadequate. In this study, whole-transcriptome analysis was performed using three datasets ([27]GSE147597, [28]GSE147602 and [29]GSE147603). A total of 14 potential circRNAs were identified, after which their structural patterns and binding miRNAs were obtained. Next, 45 differentially expressed miRNAs (DEmiRNAs) between CRC without and with liver metastasis were acquired, consisting 38 upregulated and 7 downregulated miRNAs. After conducting intersection analysis, expression validation and correlation analysis, miR-761 and miR-424-5p were selected as the most potential miRNAs linked to liver metastasis of CRC. Subsequently, the target genes of miR-761 or miR-424-5p were predicted and differentially expressed genes (DEGs) between CRC without and with liver metastasis were obtained. 257 genes that were commonly appeared in predicted genes and DEGs were significantly enriched in “epithelial-to-mesenchymal transition” and “signaling by Robo receptor”. Among these enriched genes, only TPM2, SRPX and SRGAP1 were significantly negatively correlated with miR-424-5p and were positively linked to hsa_circ_0000375 in CRC without or with liver metastasis. Collectively, the current findings elucidated a potential hsa_circ_0000375-miR-424-5p-TPM2/SRPX/SRGAP1 network contributing to liver metastasis of CRC. Keywords: Circular RNA (circRNA), hsa_circ_0000375, miR-424-5p, Colorectal cancer (CRC), Liver metastasis, Bioinformatic analysis 1. Introduction Colorectal cancer (CRC) is one of the most common malignancies and ranks the third cause of cancer-related deaths all over the worldwide [[30]1,[31]2]. Liver is the leading target organ of blood metastasis of CRC. As reported, nearly 20 % of patients with CRC present synchronous liver metastasis, and another 20 % CRC patients develop liver metastasis within 5 years after early diagnosis [[32]3]. Furthermore, liver metastasis is the main lethal reason of CRC, with the five-year overall survival less than 15 % [[33]4]. Despite the molecular mechanism of liver metastasis of CRC has been strenuously studied, its knowledge is still inadequate and need to be further explored. Circular RNAs (circRNAs), a group of novel, endogenous and non-coding RNAs, have covalently closed loops without 5′-cap and 3′-polyadenylated tail structure which makes them more stable than linear counterparts [[34][5], [35][6], [36][7]]. Recently, the critical roles of circRNAs in initiation and progression of human malignancies have been gradually unlocked, including CRC [[37]8,[38]9]. For example, Liu et al. confirmed that circIFT80 enhanced the progression of CRC by targeting beta-catenin [[39]10]; Fan et al. showed that circ_0000395 facilitated CRC development via upregulating MYH9 [[40]11]. However, the specific role and mechanism of circRNA in liver metastasis of CRC have not been fully elucidated and deserve to be further studied. In our previous work [[41]12], a potential hsa_circ_0001955/hsa_circ_0000977-mediated miRNA-mRNA regulatory network involved in the transition from normal tissues to CRC was established by whole-transcriptome analysis. The oncogenic role of hsa_circ_0001955 in CRC was following validated by Gao et al. [[42]13]. In this study, we aimed to identify the potential circRNA-miRNA-mRNA network contributing to liver metastasis of CRC. The findings from the current study might provide crucial clues for understanding the molecular mechanism of CRC liver metastasis and developing effective therapeutic targets in treating liver metastasis of CRC. 2. Results 2.1. Screen of 14 potential circRNAs associated with liver metastasis of CRC To study the role and mechanism of circRNA in liver metastasis of CRC, a circRNA dataset [43]GSE147597 was firstly employed. [44]GSE147597 contained 10 CRC patients without liver metastasis and 10 CRC patients with liver metastasis. No statistical difference of baseline of age between the two groups was observed (without liver metastasis: 64.9 ± 1.853; with liver metastasis: 61.9 ± 4.905; P = 0.574). Using GEO2R online tool, data normalization, UMAP analysis and expression density were successively performed ([45]Fig. 1A–C). A total of 959 significant differentially expressed circRNAs (DECs), consisting of 432 upregulated and 527 downregulated DECs in CRC without liver metastasis, were identified ([46]Fig. 1D–E). Furthermore, 14 DECs with |Fold change|>1.5 were chosen for subsequent analysis. As shown in [47]Table 1 and [48]Fig. 2A-N, 2 DECs were significantly upregulated whereas 12 DECs were markedly downregulated in CRC without liver metastasis compared with CRC with liver metastasis. The 14 circRNAs were considered as the potential circRNAs associated with liver metastasis of CRC. Fig. 1. [49]Fig. 1 [50]Open in a new tab Screening differentially expressed circRNAs (DECs) between CRC without liver metastasis and with liver metastasis. (A) The normalization of the 20 tissue samples in [51]GSE147597 by GEO2R. (B) The UMAP analytic result of [52]GSE147597 by GEO2R. (C) The expression density of [53]GSE147597. (D) The significant DECs between CRC without and with liver metastasis. (E) The volcano plot of the DECs between CRC without and with liver metastasis in [54]GSE147597. Table 1. The potential differentially expressed circRNAs (DECs) between CRC without liver metastasis and CRC with liver metastasis. circBase ID Adjusted P-value t B Log[2]FC[55]^a hsa_circ_0003270 0.006 4.469 0.599 1.709 hsa_circ_0082182 0.004 4.814 1.366 1.626 hsa_circ_0043278 0.003 −5.092 1.983 −4.587 hsa_circ_0006220 0.030 −3.123 −2.374 −2.510 hsa_circ_0001666 0.029 −3.156 −2.304 −2.284 hsa_circ_0006502 0.010 −3.963 −0.533 −1.968 hsa_circ_0088045 0.013 −3.770 −0.963 −1.925 hsa_circ_0092330 0.002 −5.540 2.960 −1.725 hsa_circ_0088046 0.014 −3.721 −1.073 −1.703 hsa_circ_0000375 0.008 −4.269 0.151 −1.698 hsa_circ_0080210 0.011 −3.854 −0.777 −1.671 hsa_circ_0092289 0.004 −4.918 1.598 −1.620 hsa_circ_0064644 0.016 −3.601 −1.337 −1.572 hsa_circ_0042435 0.003 −5.221 2.266 −1.556 [56]Open in a new tab ^a FC: CRC without liver metastasis vs. CRC with liver metastasis (without/with). Fig. 2. [57]Fig. 2 [58]Open in a new tab Expression determination of candidate circRNAs associated with liver metastasis of CRC. The expression levels of hsa_circ_0003270 (A), hsa_circ_0082182 (B), hsa_circ_0043278 (C), hsa_circ_0006220 (D), hsa_circ_0001666 (E), hsa_circ_0006502 (F), hsa_circ_0088045 (G), hsa_circ_0092330 (H), hsa_circ_0088046 (I), hsa_circ_0000375 (J), hsa_circ_0080210 (K), hsa_circ_0092289 (L), hsa_circ_0064644 (M) and hsa_circ_0042435 (N) in CRC without liver metastasis compared with CRC with liver metastasis. *P < 0.05. 2.2. Prediction of binding miRNAs of the circRNAs associated with liver metastasis of CRC The corresponding genome location and parental genes of 14 potential circRNAs were obtained from circBase database ([59]Table 2). It has been widely acknowledged that circRNAs exert their biological functions by competitively binding to miRNAs, at least partially [[60]14]. The structural patterns of potential circRNAs were drew by CSCD database. Consequently, the structural loops of 11 circRNAs were presented in [61]Fig. 3A-L. Intriguingly, all the 11 circRNAs had microRNA response elements (MREs). Subsequently, the possible miRNAs that could bind to these circRNAs were predicted. For better visualization, a circRNA-miRNA regulatory network was established using Cytoscape software as shown in [62]Fig. 4. Table 2. The locations and parental genes of candidate circRNAs. circBase ID Location Parental gene hsa_circ_0003270 chr9:128,099,296-128,099,870 GAPVD1 hsa_circ_0082182 chr7:128,317,617-128,323,309 FAM71F2 hsa_circ_0043278 chr17:35,797,838–35,800,763 TADA2A hsa_circ_0006220 chr17:35,800,605–35,800,763 TADA2A hsa_circ_0001666 chr6:170,626,457-170,639,638 FAM120B hsa_circ_0006502 chr9:140,458,885-140,459,606 WDR85 hsa_circ_0088045 chr9:114,860,749-114,864,565 SUSD1 hsa_circ_0092330 chr22:19,965,129–19,965,469 ARVCF hsa_circ_0088046 chr9:114,860,749-114,875,148 SUSD1 hsa_circ_0000375 chr12:6,657,590–6657991 IFFO1 hsa_circ_0080210 chr7:50,737,418-50773020 GRB10 hsa_circ_0092289 chr20:17,595,565–17,595,865 RRBP1 hsa_circ_0064644 chr3:29,910,348-29,941,246 RBMS3 hsa_circ_0042435 chr17:20,149,238–20209395 SPECC1 [63]Open in a new tab Fig. 3. [64]Fig. 3 [65]Open in a new tab The structural patterns of candidate circRNAs acquired from CSCD database. The structural pattern of hsa_circ_0003270 (A), hsa_circ_0043278 (B), hsa_circ_0006220 (C), hsa_circ_0001666 (D), hsa_circ_0006502 (E), hsa_circ_0088045 (F), hsa_circ_0088046 (G), hsa_circ_0000375 (H), hsa_circ_0080210 (I), hsa_circ_0064644 (J) and hsa_circ_0042435 (K). (L) The representation of MRE, RBP and ORF. Fig. 4. [66]Fig. 4 [67]Open in a new tab Establishment of a potential circRNA-miRNA regulatory network by Cytoscape software. 2.3. Identification of potential miRNAs associated with liver metastasis of CRC To further identify the potential miRNAs linked to liver metastasis of CRC, a miRNA dataset [68]GSE147603 possessed identical CRC patients with [69]GSE147597 was introduced. Differential expression analysis for [70]GSE147603 was performed. As presented in [71]Fig. 5A–B, a total of 45 differentially expressed miRNAs (DEmiRNAs) between CRC without and with liver metastasis were acquired, including 38 upregulated and 7 downregulated DEmiRNAs. Next, intersection analysis for the predicted miRNAs of DECs and DEmiRNAs was conducted. As suggested in [72]Fig. 5C, none of miRNAs that were commonly appeared in “Predicted miRNA” and “DEmiRNA” sets were found. Intriguingly, a total of 7 miRNAs were commonly appeared in both the two miRNA sets ([73]Fig. 5D). As presented in [74]Fig. 6A–G, all the 7 miRNAs were significantly downregulated in CRC with liver metastasis when compared with CRC without liver metastasis, which was identical with the differential expression analysis by GEO2R online tool. Correlation analysis for circRNA-miRNA pairs was performed ([75]Fig. 7A-L). The results revealed that miR-761 was significantly negatively correlated with hsa_circ_0043278 ([76]Fig. 7A), hsa_circ_0006220 ([77]Fig. 7B), hsa_circ_0088045 ([78]Fig. 7C), hsa_circ_0088046 ([79]Fig. 7D), hsa_circ_0000375 ([80]Fig. 7E), as well as miR-424-5p was markedly inversely associated with hsa_circ_0000375 ([81]Fig. 7G) in CRC without or with liver metastasis. Taken together, these findings suggested that miR-761 and miR-424-5p might act as key tumor suppressors in negatively mediating liver metastasis of CRC. Fig. 5. [82]Fig. 5 [83]Open in a new tab Identification of potential miRNAs related to liver metastasis of CRC. (A) The volcano plot of the differentially expressed miRNAs (DEmiRNAs) between CRC without and with liver metastasis in [84]GSE147603. (B) The expressed landscape of significant DEmiRNAs between CRC without and with liver metastasis in [85]GSE147603. “Red” and “Green” indicated that miRNAs were significantly upregulated and downregulated in CRC without and with liver metastasis compared with CRC with and with liver metastasis, respectively. (C) The intersection of predicted miRNAs of circRNAs (upregulated in CRC without liver metastasis) and DEmiRNAs (downregulated in CRC without liver metastasis). (D) The intersection of predicted miRNAs of circRNAs (downregulated in CRC without liver metastasis) and DEmiRNAs (upregulated in CRC without liver metastasis). (For interpretation of the references to color in this figure legend,