Next-generation sequencing (NGS) identified gender-dependent miRNA expression profiles in hepatocarcinogenesis
To detect miRNA expression profiling, the normal liver tissue samples (W) of wild-type non-transgenic mice (non-Tg) and hepatic tumors (T) and matched adjacent precancerous tissue samples (P) of Hras12V transgenic mice (Ras-Tg) from males and females were collected and named MW, MP, MT, FW, FP, and FT (M indicates male; F indicates female), respectively (Fig. S1). The miRNA expression profiles of these samples were assessed by NGS technology from isolated total RNAs. For every sample, an average of 27 million (ranged from 20-36 million) miRNA raw reads were obtained and the average clean ratio for the raw reads was 98.87% (ranged from 97.97-99.48%) (Table S1). Saturation analysis showed that the miRNAs sequenced and mapped by trimmed reads were saturated when the sequencing depth approached 10 million (Fig. S2), indicating that the sequencing depth used in our study was sufficient to achieved high transcriptome coverage. Pearson’s correlation and principal component analysis (PCA) indicated that the miRNA expression profiles of the six tissue groups vary significantly (Fig. 1A, B). The miRNAs read values were transformed to TPM (transcripts per million), and, by using the criteria of q-value ≤ 0.05 and fold-change ≥ 2, a total of 191 DEMIRs in males and 204 DEMIRs in females (Fig. 1C; Table S2) were identified at least in one paired comparison among W, P, and T. The identified miRNA expression profiling data revealed shared, but also unique expression patterns in hepatocarcinogenesis between sexes.
To assess the reliability of the quantitative miRNAomics analysis obtained by NGS, six of the identified differentially expressed miRNAs in both sexes were randomly selected and further evaluated in different males and females by RT-qPCR (Fig. 2). The expression changes of these miRNAs were consistent with the NGS data, both in males and in females (Table S2).
MiRNA expression profiling revealed different distribution of DEMIRs during hepatocarcinogenesis between sexes
To investigate the changes in miRNA expression profiles during hepatocarcinogenesis in each sex, we determined the numbers of miRNAs that their expression changed significantly, by pairwise comparison among W, P, and T of males and females. Interestingly, the changes in miRNA expression profiles differed between males and females (Fig. 3A, B). For both sexes, the most prominent changes were observed when T was compared with W and P, indicating that miRNAs may play crucial roles in hepatocarcinogenesis. In T versus W (T/W) and T versus P (T/P), the numbers of up-regulated miRNAs were approximately 3-4-fold higher than that of down-regulated miRNAs, implying that HCC-related miRNAs play a main role in the down-regulation of protein levels. Profoundly higher numbers of miRNAs whose expression was altered in T was observed in females, indicating that the aberrant regulation of a higher number of miRNAs is needed for hepatic transformation in females, reflecting the lower incidence of HCC observed in females. On the other hand, the number of miRNAs whose expression was altered in P versus W (P/W) was higher in males, indicating that males are more susceptible to Ras-induced carcinogenesis. The opposite changes in numbers of up- and down-regulated miRNAs in P/W suggests that different mechanisms are involved in oncogene stress in males and females.
Shared and unique miRNAs expression patterns during hepatocarcinogenesis between sexes
To identify shared and unique miRNAs involved in hepatic tumorigenesis in males and females, Venn analysis was firstly performed for T/P, T/W, and P/W in males and females (Fig. 3C, D; Table S2). And then, we classified the miRNAs into four categories (from (a) to (b)) describing particular variation trends of miRNAs during hepatic tumorigenesis. Further, shared and unique miRNAs were identified, depending on the categories (Table S3). To focus on the clear and definite variation tendency, we summarized at least two times significantly changed miRNAs in pairwise comparison among W, P, and T (Fig. 3E) (Fig. S3 summarized the expression patterns for all detected 264 DEMIRs).
In total, 144 miRNAs were found to show at least two times significant changes in a pairwise comparison between W, P, and T (Fig. 3E). Among them, 104 miRNAs were found in males, while 119 were found in females. There were 79 miRNAs shared to both sexes, while 25 and 40 miRNAs were unique in males and females, respectively. Among the shared miRNAs, 68 miRNAs were negatively associated with liver tumorigenesis. Among the unique miRNAs, 19 and 29 miRNAs were positively correlated to liver carcinogenesis in males and females, respectively, and 6 and 11 miRNAs were negatively correlated to hepatic tumors in males and females, respectively (Fig. 3E ; Table S3).
Additionally, the four categories of miRNA expression patterns can be further classified into several subtypes (Fig. 3E). The symbols used here indicate that “>” and “<”: miRNAs were significantly up- and down-regulated, respectively; and “=”: no significant difference in miRNA expression levels was observed between the samples. The category HCC-positive-related miRNAs (a) includes three subtypes: (1) T>P=W; (2) T=W>P; (3) T>W>P. The most miRNAs in this category were classified into subtype (1). The category Ras-positive-related miRNAs (b) includes two subtypes: (1) W<P=T and (2) W<P<T. These miRNAs were equally (or gradually) and significantly up-regulated in P and T compared with W. The category HCC-negative-related miRNAs (c) includes two subtypes: (1) T<P=W, (2) T=W<P. The most miRNAs in this category were classified into subtype (1). The category Ras-negative-related miRNAs (d) includes two types: (1) W>T=P and (2) W>P>T. These miRNAs were equally (or gradually) and significantly down-regulated in P and T compared with W.
In particular, 38 miRNAs in (a)-(1) type and 5 miRNAs in (c)-(1) type occupy the most proportions of miRNAs in variant trend types, which represent the shared miRNAs involved in hepatic tumor development (Table S3). Additionally, 4 (in b) common miRNAs in both sexes were found to be positively related to Ras oncogene expression. Moreover, the higher number of unique miRNAs in females compared to males (40 vs. 25) suggests that more changes in miRNAs expression are required in females to develop HCC (Table S3).
Identification of the regulatory networks that are involved in hepatocarcinogenesis both in males and females
To identify the regulatory networks that are involved in hepatocarcinogenesis both in males and females, the common HCC-related miRNAs (38 up-regulated (T>P=W); 5 down-regulated (T<P=W) in Fig. 3E) and mRNAs (646 up-regulated (T>P=W); 323 down-regulated (T<P=W) in Table S4) in males and females were selected for further analysis (Fig. 4I). Considering that mRNAs are negatively regulated by miRNAs, 323 down-expressed mRNAs may be regulated by 38 up-expressed miRNAs and 646 up-expressed mRNAs may results from 5 down-expressed miRNAs. Further, 35 TFs (27 in up-expressed and 8 in down-expressed mRNAs) were identified. Finally, four types of regulatory relationships among miRNAs, TFs, and genes were predicted (Fig. 4II; Table S5; Table S6).
To merge the four types of regulatory relationships predicted above (Fig. 4II), 3-node FFLs were formed (Fig. 4III; Table S7; Table S8). The network contained 1182 edges and 326 unique nodes in the 920 FFLs. Among the 1182 edges, 461 miRNA-gene, 4 miRNA-TF, 588 TF-gene, and 129 TF-miRNA pairs were predicted. Among the 326 nodes, 265 genes, 46 miRNAs, and 15 TFs were specific for HCC. Further, these FFLs were categorized into TF-FFLs (TF), miRNA-FFLs (miRNA), and composite-FFLs (composite) sub-networks (SNW) (Fig. 4IV). In these three SNWs, TF-SNW is the most redundant. By using the hub definition method proposed by Yu et al.[17], we determined the top 25% degree nodes of gene, miRNA and TF as the member of secondary TF-SNW to focus on the most significant components. To further analyze the potential function, all nodes in the three SNWs (secondary TF-SNW, miRNA-SNW, composite-SNW) were filtered based on the GEPIA and PUBMED database, identifying a group of significant components (Fig. 4V).
Differentially expressed miRNAs (DEMIRs) between males and females
To identify the DEMIRs between males and females involved in hepatic tumorigenesis, comparisons between the sexes (MW versus FW (MW/FW); MP versus FP (MP/FP); MT versus FT (MT/FT)) were performed. Interestingly, the highest number of DEMIRs was detected in MT/FT, and the lowest number of DEMIRs was detected in MP/FP (74 vs. 19), whereas 48 DEMIRs were identified in MW/FW (Fig. 5A). Venn diagram showed that the number of overlapped DEMIRs between MT/FT and MW/FW was higher than that between MP/FP and MT/FT (MW/FW) (Fig. 5B). Our findings suggest that RAS/ERK pathway activation reduces the differences in miRNA expression profiles between precancerous hepatocytes of the males and females. However, during hepatocarcinogenesis, the differences in miRNA expression profiles between hepatomas of the males and females are enhanced. The 29 overlapping miRNAs between MT/FT and MW/FW are possibly linked to the gender disparity of hepatocytes and hepatomas. Intriguingly, the opposite changes in numbers of up- and down-regulated miRNAs found between MW/FW and MT/FT suggest profound gender disparity of hepatocytes and hepatomas.
The miRNA clusters in the Dlk1-Dio3 genomic imprinting region (GIR) play important roles in hepatocarcinogenesis
Dlk1-Dio3 GIR is highly conserved in humans and mice, located on chromosome 14q32 and 12qF1, respectively. It contains three protein-coding genes and several non-coding RNA clusters (Fig. 5C). The miRNAs encoded in the GIR constitute the largest miRNAs cluster in both human and mouse and play crucial roles in most cancers, through the regulation of multiple pathways [18, 19]. In the present study, 53.5% of the detected DEMIRs (54 out of 101) between the two sexes were found to be located on Dlk1-Dio3 GIR (Fig. 5D; Table S9). The variation of DEMIRs in Dlk1-Dio3 GIR is consistent with that of total DEMIRs between sexes (Fig. 5A, D), suggesting that the DEMIRs in Dlk1-Dio3 GIR may play crucial roles in gender-disparity of hepatocytes and hepatocarcinogenesis.
Moreover, 54.2% (26 out of 48) DEMIRs in MW/FW were located on Dlk1-Dio3 GIR, and 25 of them were up-regulated in MW (Fig. 5D, E; Table S9). Intriguingly, in MP/FP, only 5.3% (1 out of 19) DEMIR, miR-341-3p, is located in Dlk1-Dio3 GIR (Fig. 5E; Table S9). This reflects the reduced gender disparity in miRNA expression in Dlk1-Dio3 GIR under the expression of Ras oncogene. However, in MT/FT, 71.6% (53 out of 74) DEMIRs were located in Dlk1-Dio3 GIR (Fig. 5E; Table S9), and in contrast to MW/FW, 53 miRNAs were down-regulated in MT. During tumorigenesis, the expression of miRNAs located in Dlk1-Dio3 GIR decreased in P and then increase again in T compared to W in males (Fig. 5F; Table S9). However, in females, the expression of miRNAs in Dlk1-Dio3 GIR tended to increase gradually in P and T compared to W (Fig. 5G; Table S9). These findings indicate that the overexpression of miRNAs of the Dlk1-Dio3 GIR plays a crucial role in hepatocarcinogenesis in both sexes. The significantly higher levels of these miRNAs in hepatoma in females compared to males imply their important role in the lower susceptibility of hepatocarcinogenesis observed in females.
Although the different expression patterns of miRNAs exist between males and females, the vast majority of DEMIRs located in Dlk1-Dio3 GIR were up-regulated in T comparing to P in both sexes (Fig. 5F, G). This finding indicates that the up-regulation of miRNAs located in Dlk1-Dio3 GIR may have similar and important roles in hepatocarcinogenesis in both sexes. In order to understand the functions of these DEMIRs on Dlk1-Dio3 GIR in hepatic tumorigenesis, their target genes were predicted using the TargetScan and miRDB database among the common 530 down-regulated mRNAs in T/P from both sexes (Table S10). Target enrichment analysis was performed using the Metascape database. Key metabolic pathways were identified. Among them, pathways related to retinol, steroid, lipid, bile acid, and bile salt were highly enriched (Table 1; Table S10). These results indicate that the miRNA encoded by Dlk1-Dio3 GIR play important roles in hepatocarcinogenesis by regulating multiple metabolic pathways.