Clonal evolution in liver cancer at single-cell and single-variant resolution

The genetic heterogeneity in hepatocellular carcinoma (HCC) has been extensively studied by bulk or multi-region sequencing [1, 2], and more recently at single-cell sub-chromosomal scale [3, 4]. Analysis at single-variant resolution, however, is still lacking. To address this issue, here we employed a single-cell strategy to dissect the single-variant clonal structure of HCC, and investigate the relationship between genetic and phenotypic heterogeneity (Fig. 1a). A total of 5 HCC patients (HCC1, HCC2, HCC5, HCC8 and HCC9) were analyzed, including one (HCC8) with both primary tumor and the intrahepatic portal vein tumor thrombus (PVTT) (Additional file 1: Supplementary Methods, and Additional file 2: Fig. S1).

Great inter-tumor genetic heterogeneity of HCC was revealed by pseudo-bulk whole exome sequencing (WES), with different somatic mutations and mutational signatures observed among patients (Additional file 2: Fig. S2 and Additional file 3: Table S1). To further explore the intra-tumor heterogeneity, ~ 60 mutation sites (Additional file 4: Table S2) were then selected from each patient for target sequencing on single cells. High quality single-cell mutation data were obtained with good correlations between mutated cell fractions and WES-derived variant allele frequency values, as well as with low allele drop-out rates in most samples (Additional file 2: Fig. S3).

The clonal structures of liver tumor tissues were then uncovered by single-cell mutational profiles. Both HCC1 and HCC2 exhibited a single-clone structure with limited heterogeneity, while a multi-clone structure was observed in HCC9 (Additional file 2: Fig. S4). The evolutionary history of tumor cells in HCC9 was reconstructed by employing mutation combination analysis in single cells. The initiated cell was malignantly transformed to the founder Clone-1 as 9 genes were mutated, where 4 of them are COSMIC Census drivers [5]. Other clones were then derived from Clone-1 by acquisition of extra subclonal mutations.

A common origin but independent evolution pattern was observed in primary tumor and metastatic PVTT. Single cells from both tumor tissues in HCC8 shared mutations on 10 genes, where 7 of them are Census drivers: CCR4, KMT2A, ZEB1, AKT3, MUC16, BRD4 and ETV6 (Fig. 1b, c). Clone-1 with the 10 shared mutations represented a common origin, and other cells in both tumor tissues had divergent extra mutations (Fig. 1c). This implied an early stage metastasis followed by independent evolution, consistent with recent observation of early metastatic seeding in other types of solid tumor [6]. Two clones within PVTT shared PVTT-private clonal mutations on 3 genes, where KIAA1549 and NCOR1 are Census drivers related to tumorigenesis [7, 8]. Mutations on these two genes thus might be metastasis-related early genetic events. Interestingly, nine out of ten mutations in Clone-1 were T > A substitutions (Fig. 1d) related to carcinogen aristolochic acids [9], consistent with previous suggestion of early rather than late exposure for HCC development [10]. This suggested that early genetic event during liver tumorigenesis may be related to specific etiology. The phylogenetic tree of single cells also supported the multi-clone structure in HCC8 (Fig. 1e), which represented genuine tumor phylogeny different from that derived from bulk or multi-region sequencing [11].

The inter- and intra-tumor genetic heterogeneity in HCC were found consistent with phenotypic heterogeneity by parallel single-cell RNA-Seq (Additional file 2: Fig. S5). Tumor cells from 4 patients formed separate clusters, illustrating patient-specific transcriptomic profiles (Fig. 2a–c and Additional file 2: Fig. S6a, b). For intra-tumor heterogeneity, the sub-chromosomal scale copy number inference from global transcriptomic profiles in HCC1 or HCC2 were quite similar within each patient, consistent with their single-clone structures (Additional file 2: Fig. S6c, d). For HCC9, both copy number inference and differentially expressed gene analysis identified tumor sub-populations, echoing its relatively higher genetic heterogeneity (Fig. 2d, e). Interestingly, the genes mutated specifically in each tumor showed similar expression patterns among single cells from different patients, and clustering of those genes exhibited a mixture of patient origin (Fig. 2f). This suggested that the tumor-specific mutations in HCC might cause phenotypic heterogeneity by altering the expressions of other genes rather than their own. The direct link between genetic and phenotypic heterogeneity in HCC, however, still await further clarification with new single-cell multi-omics tool that could co-detect point mutation and gene expression [12].

In summary, here we reconstructed single-cell clonal evolution in human liver cancer at single-variant resolution. The common origin but independent evolutionary fate for primary and metastatic liver tumors observed here may help understanding liver cancer progression, and single-cell dissection of both genetic and phenotypic heterogeneity will provide information for their functional linkage.