Cyclin A2/E1 activation defines a hepatocellular carcinoma subclass with a rearrangement signature of replication stress

Abstract

Cyclins A2 and E1 regulate the cell cycle by promoting S phase entry and progression. Here, we identify a hepatocellular carcinoma (HCC) subgroup exhibiting cyclin activation through various mechanisms including hepatitis B virus (HBV) and adeno-associated virus type 2 (AAV2) insertions, enhancer hijacking and recurrent CCNA2 fusions. Cyclin A2 or E1 alterations define a homogenous entity of aggressive HCC, mostly developed in non-cirrhotic patients, characterized by a transcriptional activation of E2F and ATR pathways and a high frequency of RB1 and PTEN inactivation. Cyclin-driven HCC display a unique signature of structural rearrangements with hundreds of tandem duplications and templated insertions frequently activating TERT promoter. These rearrangements, strongly enriched in early-replicated active chromatin regions, are consistent with a break-induced replication mechanism. Pan-cancer analysis reveals a similar signature in BRCA1-mutated breast and ovarian cancers. Together, this analysis reveals a new poor prognosis HCC entity and a rearrangement signature related to replication stress.

Fig. 1

Diverse mechanisms leading to CCNA2 activation in HCC. a Summary of structural rearrangements (top) and viral insertions (bottom) affecting CCNA2 gene identified in 751 HCC from the LICA-FR, TCGA and ICGC-JP cohorts. b Sorted CCNA2 expression (log scale) in the LICA-FR and TCGA cohorts. Gene expression was obtained from RNA-seq data and is given in fragments per kilobase of exons per million reads (FPKM). Samples harboring structural variants (SV) or viral insertions are indicated with a color code. c Functional consequences of AAV2 and HBV insertions in CCNA2. Viral insertions identified in the LICA-FR cohort were precisely mapped using WGS or viral capture data, and RNA-seq reads were aligned on the reconstructed chimeric DNA to identify the transcription start sites and predicted translation initiation sites of abnormal transcripts. d CCNA2 fusions identified in the LICA-FR, TCGA and ICGC-JP cohorts. The transcription start site of the fusion transcript is represented together with the predicted translation initiation site. Fusions with KIAA1109, LIPC and TDO2 involve 5′ exons not annotated in transcript databases but expressed in normal liver. e Schematic representation of cyclin A2 protein with functional domains. D-box Destruction box; Ub, Ubiquitination targeting sequences. f Western blot analysis of cyclin A2 using antibodies targeting the N-terminal (top) or C-terminal (middle) domains. Tumors with viral insertions or gene fusions are compared with tumors without CCNA2 alteration and non-tumoral liver controls

Fig. 2

Viral and non-viral mechanisms of CCNE1 activation in HCC. a Summary of structural rearrangements (top) and viral insertions (bottom) affecting CCNE1 gene identified in 751 HCC from the LICA-FR, TCGA and ICGC-JP cohorts. b Sorted CCNE1 expression (log scale) in the LICA-FR and TCGA cohorts. Gene expression was obtained from RNA-seq data and is given in fragments per kilobase of exons per million reads (FPKM). Samples harboring structural variants (SV), focal amplifications and viral insertions are indicated with a color code. c Functional consequences of structural rearrangements affecting CCNE1 regulatory region. RNA-seq read counts along CCNE1 locus are represented in normal liver (top) and in 4 tumors harboring structural rearrangements upstream CCNE1 trancription start site (TSS). H3K27Ac chromatin immunoprecipitation sequencing (ChIP-seq) signal and chromatin states in adult liver were obtained from the ROADMAP consortium and are depicted below each reconstruted DNA sequence. EnhA: active enhancer; EngG: genic enhancer; EnhWk: weak enhancer; TssA: active TSS; TssFlnk: flanking TSS; TxWk: weak transcription; Quies: quiescent chromatin

Fig. 3

Clinical and molecular features of cyclin-activated HCC. a t-SNE plots depicting the classification of HCC from the LICA-FR and TCGA cohorts based on their transcriptional profiles. Tumors harboring CCNA2 or CCNE1 activating alterations are indicated with a color code. b Clinical characteristics, driver genes and deregulated pathways associated with CCN-HCC in the LICA-FR (top) and TCGA (bottom) cohorts. c Overall survival in CCN-HCC as compared with other HCC in the LICA-FR cohort. Only HCC with curative resection (R0) were included

Fig. 4

Cyclin-activated HCC display a specific signature or structural rearrangements. a Six rearrangement signatures identified across 350 HCC genomes by non-negative matrix factorization. Structural rearrangements were classified in 38 categories considering their type (del: deletion, dup: tandem duplication, inv: inversion, trans: inter-chromosomal translocation) and size, and distinguishing clustered from non-clustered events. The probability of each rearrangement category in each signature is represented, with rearrangement types indicated above and rearrangement sizes below. b Unsupervised classification of 45 HCC from the LICA-FR cohort based on the contribution of rearrangement signatures in each tumor. Significant molecular alterations associated with the cluster of tumors having a high contribution of signature RS1 are represented below. P-values were obtained using Fisher’s exact tests. c Validation of the association between the RS1 signature and CCN-HCC in the TCGA and ICGC-JP series. The middle bar, median; box, interquartile range; bars extend to 1.5 times the interquartile range. d CIRCOS plot representing the structural rearrangement profile of a representative CCN-HCC (FR-961T, harboring a GSTCD-CCNA2 fusion). e Copy-number profile showing the accumulation of focal gains along chromosome 14 in tumor FR-961T. Structural rearrangements are overlaid on the copy-number profile with a color code indicating the type of event. trans: inter-chromosomal translocation; dup: tandem duplication; inv: inversion. f Three types of rearrangements leading to focal chromosome gains in CCN-HCC. A representative example of each type of event is shown with a copy-number plot above and a schematic representation of the rearranged chromosome below. Structural rearrangements are represented with the same color code as in e. Dashed lines on schematic chromosome reconstructions represent the abnormal junctions detected in WGS data

Fig. 5

Hotspot analysis of rearrangement signature 1 (RS1) breakpoints. a The density of RS1 breakpoints along chromosome 4 is displayed above replication timing, RNA-seq expression, H3K27Ac ChIP-seq profile and chromatin state. Replication timing was determined using Repli-Seq data from the liver cancer cell line HepG2. RNA-seq profile was generated from a normal liver sample. H3K27Ac and chromatin states in normal adult liver were obtained from the ROADMAP consortium. The legend for chromatin state color codes is displayed in c. Hotspots corresponding to highly expressed liver enzymes are annotated (UGDH, UDP-glucose 6-dehydrogenase; UGT, UDP glucuronosyltransferase family cluster; ALB, albumin; HSD17B, hydroxysteroid 17-Beta dehydrogenases 11 and 13, ADH, alcohol dehydrogenase cluster; ACSL1, acyl-CoA synthetase long chain family member 1). b RS1 breakpoint density in topologically associated domains (TADs). TADs were defined in human embryonic stem cells (H1) and classified based on gene expression in normal liver, replication timing and chromatin state. For each comparison, breakpoint density was normalized to be 1 in the group with the lowest density. Error bars indicate the 95% confidence interval. c Enrichment of rearrangement breakpoints in ROADMAP chromatin states for the 6 rearrangement signatures identified in HCC. For each signature, the fold-change between the observed and expected number of breakpoints falling within each chromatin state is represented, and chromatin states with a >2-fold enrichment are annotated. d Quantile-quantile plot of RS1 breakpoint enrichment p-values across 500 kb windows. e Proportion of TERT promoter alterations in CCN-HCC and other HCC analyzed by WGS. f Contribution of the 6 rearrangement signatures to TERT promoter rearrangements in CCN-HCC and other HCC

Fig. 6

Pan-cancer analysis of the RS1 signature a Violin plots representing the number of rearrangements attributed to signature RS1 across patients within each cancer type in the ICGC PCAWG data set. For each cancer type, we assessed the association between tumors with ≥ 50 RS1 events and tumors with a rearrangement breakpoint < 80 kb from CCNA2 or CCNE1 gene using Fisher’s exact tests. ns: not significant. The definition of cancer codes and number of samples per cancer type are available in Supplementary Data 9. b Number of RS1 events across 524 breast cancer genomes and association with BRCA1 alterations and CCNE1 amplifications. PD13296a, the only tumor with both BRCA1 mutation and CCNE1 amplification, has the highest number of RS1 events in the series. c Number of RS1 events across 80 ovarian cancer genomes and association with BRCA1 alterations and CCNE1 amplifications. P-values were obtained using one-sided Wilcoxon rank-sum tests. d Number of RS1 events in liver, breast and ovarian cancers with or without CCNA2, CCNE1 and BRCA1 alterations. The middle bar, median; box, interquartile range; bars extend to 1.5 times the interquartile range. e, Violin plots representing the distribution of tandem duplication sizes across liver, breast and ovarian cancers with or without CCNA2, CCNE1 and BRCA1 alterations. f Violin plots representing the replication timing of duplication and inter-chromosomal translocation breakpoint loci in liver and breast cancers with or without CCNA2, CCNE1 and BRCA1 alterations. Replication timing was determined using Repli-Seq data from the HepG2 cell line for liver cancer and from the MCF-7 cell line for breast cancer. g Proposed connexion beween rearrangement signatures in CCN-HCC and in BRCA1-inactivated breast and ovarian cancers