In this study, we collected the tumor tissues and adjacent non-tumor liver tissues of 41 patients with ICC and 20 patients with CHC, and compiled the clinical and pathological data information of these samples (Table S1, Fig. 1e). To identify HBV integration sites in the liver cancer genome and study the virus–host interaction, we conducted HIVID analysis on these 61 samples. With HBV capture sequencing, HBV integration sites in the host genome were effectively detected at 1 base-pair resolution. We detected 493 HBV integration sites in ICC patients, of which 417 were from tumor samples and 76 were from non-tumor samples. And 246 HBV integration sites were detected in CHC patients, of which 156 were located in the genome of tumor samples and 90 were in non-tumor samples (Table S2–S5). Next, we analyzed the characteristics of HBV integration in tumor and non-tumor tissues. The results show that the occurrence ratio of HBV integration is higher in tumor (29/41) than non-tumor tissue (23/41) for ICC patients (70.7% vs 56.1%). While for CHC patients, the HBV integration ratio is a bit lower in tumor (12/18, 66.7%) than non-tumor tissue (15/20, 75%) (Fig. 1a, b). Furthermore, in ICC patients, the average number of integration sites in tumor is significantly higher than in non-tumor tissue (13.45 vs 3.3, p value = 0.00018, Student’s T test). Similar results were found in CHC, where the average number of integration sites in tumor is higher than that in non-tumor tissue (13.7 vs 6.4, p value = 0.091, Student’s T test), although the difference is not statistically significant (Fig. 1c, d).

HBV integrations’ distribution in our samples. a HBV integration rate in tumor and non-tumor tissues of ICC samples; b HBV integration rate in tumor and non-tumor tissues of CHC samples; c the comparison of HBV breakpoint numbers between tumor and non-tumor tissues in ICC; d the comparison of HBV breakpoint numbers between tumor and non-tumor tissues in CHC; e the overview of clinical indicators, number of HBV integrations in each sample, and HBV integrations profiles in the genes of recurrent genes of high frequency

To investigate the function of HBV integrations in our samples, we analyzed the distribution of HBV integration sites across the genome. As a result, we found that HBV integration might prefer to affect the CDS (coding sequence) region of the human genome in CHC tumor samples. And the frequency of HBV integrations in tumor tissue of CHC patients was also significantly higher in intergenic region than in non-tumor tissue (Fig. S1), suggesting that HBV is likely to affect the gene function through integrating into coding region (CDS) and intergenic regulating region. To study the function of the HBV integrations in tumor tissues and adjacent non-tumor tissues, we made an annotation for the HBV integration sites. There are 44 genes with HBV integrations in tumor tissues and 30 genes in non-tumor tissues. Further analysis indicated that HBV tended to integrate into a few hot spots in tumor samples in both ICC and CHC, while in non-tumor tissues, the HBV integrations tend to be scattered across the genome (Fig. 2). This result indicated that HBV integration in some target genes might play etiological role in tumorigenesis. Notably, the data of both ICC and CHC indicated a recurrently integrated genes FN1 in non-tumor tissue, which was consistent with previous research on HCC [24, 25], suggesting that the adjacent non-tumor tissues of all three PLC types had the analogous profile of HBV integration. In ICC patients, the genes with HBV integrations in more than 1 tumor samples were TERT (4), ZMAT4 (2), MET (2), ANKFN1 (2), and PLXNB2 (2); while in non-tumor tissues, only FN1 contained HBV integrations in 2 samples, which is consistent with the results of our previous study of HCC [24]. For CHC patients, TERT and ALKBH5 are found to be integrated by HBV fragments in 4 tumors and 2 tumors respectively; interestingly, FN1 is also the only gene with HBV integration event occurred in more than 2 non-tumor samples, which is consistent with the results of ICC samples (Fig. 2a, b). When calculated by number of integration events, we found that TERT still had the high number of integration sites in tumors of both ICC (27) and CHC (22). While the genes with second highest number of HBV integration were HAUS5 (8) and ANKFN1 (14), respectively, in ICC and CHC, we observed that the landscape of HBV integration in the three subtypes of liver cancer showed significant differences (Fig. S2). Only a small number of insertion sites were shared by ICC, CHC, and HCC, underscoring that integration patterns are distinct in the three different types of primary liver cancer (Fig. S3) [24]. Besides, we also found that many breakpoints in human genome could be integrated by more than one HBV fragments; and in the same way, each HBV breakpoints may integrate into more than one location in human genome. Furthermore, HBV breakpoints occurred frequently at the end of X protein and the head of preC/C protein in all three types of tumor tissues. Besides, breakpoints in CHC also distributed in Pre-S1/S and Pre-S2/S region (Fig. 2c).

HBV integration breakpoints across human genome. a Sample frequency of integration breakpoints in ICC patients; b sample frequency of integration breakpoints in CHC patients; each bar represents the sample frequency of HBV integration breakpoints at a particular locus in the human genome (hg38). Tumor (red) and non-tumor (blue) samples with HBV integrations are shown on the inner and outer circles, respectively. Histogram axis units represent number of samples. Some loci with a high frequency of integration are marked. c The distribution of breakpoints on human genome and their corresponding source location on HBV genome

To further investigate the interaction between host and virus, we also analyzed the distribution of the breakpoints on HBV genome. We found an enrichment of HBV breakpoints in the region of 1400–1900 bps containing the whole HBx gene and 5′end of the Precore/Core genes in both tumor and non-tumor samples of ICC and CHC tissues (Fig. 3a, b), which was similar to that of HCC tissue we reported before [24], suggesting that HBV virus might produce HBx protein through fully integrated HBx gene and maintain regulation promoter sequence of precore/core gene to facilitate its survival and replication in human cell. Previous studies also showed that the expression of HBx was a factor overcoming silencing of cccDNA or HBsAg expression as a possible modulator of the adaptive immune system. Furthermore, core and HBx proteins were potential candidates for direct involvement in the integration process due to their DNA-binding activities [32, 33]. In China, HBV genotypes B and C account for the majority of HBV carrier patients, with genotype B predominating in the central region and genotype C predominating southern and northern region [34, 35]. The main genotypes are genotypes B and C in ICC and CHC patients of our study, and genotype C accounted for more than 50% (Fig. 3c).

Integration breakpoints of HBV genome. a The breakpoints distribution in ICC patients. Histogram axis units represent number of breakpoints in 100-bp intervals. HBV genes and the orientation are shown in innermost circles. b The breakpoints distribution in CHC. c The HBV genotypes of integration breakpoints in tumor and non-tumor of ICC and CHC patients

HBV affects the stability of the genome by integrating into the human genome, thereby causing cell carcinogenesis. CpG islands, which are closely related to genome stability, are mainly located in the promoter and exon regions near gene transcription regulatory regions [36]. Genomic and epigenetic regulations of CpG sites across the chromosomes play an important role in liver cancerization [37, 38]. In this study, we found that for both ICC and CHC patients, the HBV integrations in the tumor samples are significantly enriched in CpG island region compared to random distribution, indicating that HBV integration is not a random event, but could provide certain selection advantages for cancer cell. Furthermore, the occurrence of HBV integrations in CpG regions is significantly more frequent in tumor than in non-tumor tissues in CHC (Fig. 4), suggesting that HBV preferentially integrates into the specific regions which may affect the gene expression and regulation in cells and induce cell cancerization.

Comparison of the breakpoints in the CpG island region of 47 ICC and 20 CHC patients. P values were calculated by Chi-square test. The expected ratio is calculated based on random distribution

Since HBV DNA could integrate into the human genome in the early stages of liver cancer, and it is reported that HBV may lead to chromosomal instability after integrated into HCC genome [24, 39, 40], we next analyzed the distribution of HBV breakpoints on chromosome level in our data. As a result, in ICC patients, the HBV integration events were significantly enriched in many chromosomes including chr 4, 5, 7, 8, 17 and non-enriched in chr 1, 2, 3, 10, 12, 13, 14, 20 in tumors; while in non-tumor tissues, only chromosomes 17 and 19 were enriched with HBV integrations (Fig. 5a). As for CHC patients, only chromosomes 2, 5, 17 and 19 are enriched for HBV integrations in tumors and chromosomes 12 and 21 are enriched for HBV integrations in non-tumor tissues (Fig. 5b). We observed that no common overrepresented chromosomes were found between ICC and CHC adjacent non-tumor tissues. In addition, this result is also distinct from the situation of HCC in our previous study where chromosomes 5, 16, 17 and 19 were enriched with HBV integrations [24]. Interestingly, as overrepresented chromosomes in HCC, two (chromosomes 5 and 17) were also enriched with HBV integrations in both ICC and CHC tumors.

Whole chromosome and chromosomal ends enrichment or non-enrichment of HBV integration in ICC (a) and CHC (b). Each bar of whole chromosome represents the expected (assuming uniform, random distribution, expected in yellow) and the observed (actual numbers in tumor tissue: tumor in blue; non-tumor tissue: non-tumor in red) ratio of HBV integration breakpoints at a particular chromosome in human genome. Ratios are numbered. Each bar of chromosomal ends represents the expected (assuming uniform, random distribution, Expected in yellow) and the observed (actual numbers in tumor tissue: tumor in blue; non-tumor tissue: non-tumor in red) ratio of HBV integration breakpoints at the 2 M region of chromosomal ends in human genome. Ratios are numbered. Red star represents statistically significant difference between non-tumor liver samples and random distribution. Blue star represents statistically significant difference between tumor samples and random distribution. (P values < 0.05) P values were calculated by Chi-square test

As telomeres at chromosome ends play a critical role in genome stability, we then checked the distribution of integration sites in chromosome ends. As a result, there was a significant increase in the frequency of HBV integration at the telomeres in chromosomes 1, 2, 4, 5, 8, 10, 15, 16, 18 and 22 in ICC tumors, while only chromosomes 17 and 19 showed enrichment of HBV integrations in adjacent non-tumor tissues (Fig. 5a). While, in CHC tumor, HBV preferred to integrate into the ends of chromosomes 1, 4, 5 and 12, and only the ends of chromosome 12 were enriched for HBV integrations in adjacent non-tumor tissues (Fig. 5b). Most of the overrepresented chromosomes in CHC tumors were also overrepresented in ICC tumor (chr1, 4, 5), while no common overrepresented chromosomes existed for adjacent non-tumor tissue of ICC and CHC.

In summary, a relatively random distribution of integration sites is observed in non-tumor samples compared to tumor samples. The results also suggested that affected chromosomes/telomeres in tumors of CHC were a subset in that of ICC, which might be explained by that fact that CHC has some common pathological features of ICC and HCC. And the different profiles of HBV integrations in the non-tumor tissues of ICC and CHC suggested that the HBV integration events were randomly occurred in initiation stage as non-tumor liver tissue may be early status of tumor. These results underscoring that the preferential integrations might be closely related to maintain the stability of the genome.

To explore the effect of HBV integration on gene expression, we selected tumors and adjacent non-tumor liver tissues from three ICC and four CHC patients for RNA-seq and WGS. For RNA-seq data, we measured the level of gene expression and calculated the difference of the expression levels for each gene between tumor and non-tumor tissue (Table 1). As a result, in the samples of ICC patients, we found 11 integrated genes with differentially expressed between tumor and non-tumor tissue. Out of the 11 genes, 8 were up-regulated and 3 were down-regulated in tumors. Out of the 8 up-regulated genes, 2 were integrated in non-tumor samples, and 6 were integrated in tumor samples. 7 out of the 8 genes contained breakpoints in intron, one is in intergenic region and another one is in promoter region. On the other hand, there are 3 genes with breakpoints in non-tumor samples showing down-regulated expression level in tumor samples. All the 3 integrations occurred in intron (Fig. 6a). It is notable that TERT with integrations in both intergenic and promoter regions were up-regulated.

Change of expression level of integrated genes in a ICC and b CHC. Upregulated: tumor > non-tumor; down-regulated: tumor < non-tumor

In the samples of CHC patients, we found 15 differentially expressed genes with HBV integration between tumor and non-tumor tissues. Out of the 15 genes, 12 were up-regulated and 3 were down-regulated in tumor. Of the 12 genes, 8 contained HBV breakpoints in non-tumor tissues and 4 were in tumor tissues. When further digging deeper into the genome structure, we found that 10 out of the 12 breakpoints were in intron and 2 were in promoter. The 4 integration sites in 3 genes co-occurring with down-regulation of gene expressions were all located in non-tumor samples. Out of the 3 integrated genes, 2 contained HBV breakpoints in intron and 1 contained breakpoint in intergenic region (Fig. 6b).

In summary, most of the HBV integration were in intron. And in ICC, the genes with HBV integrations in tumor tend to be up-regulated and the gene down-regulated in non-tumor tissue has the correlation with HBV integration in non-tumor tissue, which means that no matter HBV integration occurs in tumor or non-tumor tissue, the integrations have the tendency to co-occur with the up-regulation of the corresponding tissue. And in both ICC and CHC, all genes with HBV integrations in tumor showed up-regulations. It is worth noting that only TERT contained HBV integration breakpoint in both ICC and CHC samples, and in both intron and promoter region.

Then, we focused on the up-regulated genes with HBV integration in tumor tissue. In ICC samples, the expression value of HBV-integrated genes, including MET, WNT2, BRD9, and TERT, in tumor seems to be higher than the paired non-tumor tissues. Consistently, the expression values of these genes in integrated tumor tissues are also likely higher than that of non-integrated tumor tissues (Figs. S4–S7). The results suggest that the expression of some viral integrated genes might be up-regulated due to the HBV integrations.

To further explore the relationship between HBV integrations, gene expressions and genomic variations, we also analyzed the WGS data for the same batch of samples of RNA-seq. We found that the HBV integration in the oncogenic-related gene TERT in ICC could increase its expression level (Fig. 7c) and reduce downstream DNA copy numbers (Fig. 7d). This suggests that HBV fragments could integrate near the oncogene, causing large-scale genome variations on nearby genomic sequences, and, at the same time, changing the expression level of the oncogene genes. In CHC tissues, we also found that HBV integration could reduce DNA copy numbers and up-regulate the gene expression levels of TERT (Fig. 7a, b).

The effect of HBV integration, the genome variation, and transcription near TERT. a Differences in TERT gene expression between tumor tissues with HBV integration in CHC (tumor, upper panel) and adjacent non-tumor tissues without HBV integration (lower panel). b The difference in DNA copy number between tumors with HBV integration, tumors without HBV integration, and adjacent non-tumor tissues in CHC; c differences in TERT gene expression between tumor tissues with HBV integration in ICC (tumor, upper panel) and adjacent non-tumor tissues without HBV integration (lower panel). d The difference in DNA copy number between tumors with HBV integration and adjacent non-tumor tissues in ICC