Molecular characterization of long-term survivors of hepatocellular carcinoma

Hepatocellular carcinoma is one of the most fatal cancers, and the majority of patients die within three years. However, a small proportion of patients overcome this fatal disease and survive for more than five years. To determine the molecular characteristics of long-term survivors (survival ≥ 5 years), we analyzed the genomic and clinical data of hepatocellular carcinoma patients from The Cancer Genome Atlas and the International Cancer Genome Consortium databases, and identified molecular features that were strongly associated with the patients’ prognosis. Genes involved in the cell cycle were expressed at lower levels in tumor tissues from long-term survivors than those from short-term survivors (survival ≤ 1 years). High levels of positive regulators of the G1/S cell cycle transition (cyclin-dependent kinase 2 [CDK2], CDK4, Cyclin E2 [CCNE2], E2F1, E2F2) were potential markers of poor prognosis. Hepatocellular carcinoma patients with TP53 mutations were mainly belonged to the short-term survivor group. Abemaciclib, an FDA-approved selective inhibitor of CDK4/6, inhibited the cell proliferation and tumor growth of hepatocellular carcinoma cells in vitro and in vivo. Thus, high G1/S transition-related gene levels and TP53 mutations are promising diagnostic biomarkers for short-term survivals, and abemaciclib may be a potential targeted drug for hepatocellular carcinoma.


INTRODUCTION
The incidence of liver cancer has recently been increasing rapidly, by about 3% in women and 4% in men each year [1]. Hepatocellular carcinoma (HCC) is the second leading cause of cancer-related mortality, accounting for the deaths of 21,600 men and 10,180 women annually [1,2]. Notably, the five-year survival rate of HCC patients is only 18% in the United States [1]. Although most HCC patients die within three years, some patients overcome this fatal disease and survive for more than 5 years. Understanding the mechanisms that promote long-term survival (LTSs, survival ≥ 5 years) would be very valuable for the diagnosis and treatment of HCC [3]; however, little is known about the molecular characteristics of LTSs.
Numerous previous studies have identified genes associated with HCC oncogenesis, including NOTCH2 and β-CATENIN [4][5][6]; however, most of these studies have focused on differentially expressed genes (DEGs) between normal and tumor tissues. For example, AGING NOTCH2 was reported to be highly expressed in HCC tumor tissues [4], and Yes-associated protein (a member of the Hippo signaling pathway) was found to induce the oncogenesis and excessive growth of HCC [7,8]. Though it is very important to identify the DEGs between normal and tumor tissues, it is also of great value to study the DEGs between tumor tissues from LTSs and short-term survivors (STSs, survival ≤ 1 years) of HCC [3], because these genes may be critical determinants for the survival time of patients or potential biomarkers for prognostic prediction and therapy. However, such genes have not yet been identified.
Uncontrolled cellular proliferation is a hallmark of tumorigenesis and cancer progression [9]. Cellular proliferation is dynamically and strictly controlled during different phases of the cell cycle, especially the G1/S and G2/M transitions [9]. The G1/S transition is the most important breaker in the cell cycle, and after that period, extracellular stimulants are no longer required [10,11]. This transition is tightly regulated by multiple genes and pathways, and its disruption is associated with cancer development [12,13]. For example, cyclin D1 (CCND1), a promoter of the G1/S transition, is overexpressed in estrogen receptor-positive breast cancer cells [14]. The G1/S transition is considered as an effective target for cancer therapies; for instance, palbociclib, a selective cyclin-dependent kinase 4/6 (CDK4/6) inhibitor that leads to G1/S arrest, was found to significantly prolong the progression-free survival of patients with hormone-receptor-positive metastatic breast cancer [15,16]. Although the G1/S transition is very important in tumor development and therapy, its roles on HCC have not been clearly elucidated.
In this study, we investigated novel molecular biomarkers of HCC to provide effective prognostic predictors and therapeutic targets. For this purpose, we performed a combinational analysis of whole-exome sequencing data, whole-transcriptome sequencing data and patients' clinical information from the Liver Hepatocellular Carcinoma (LIHC) project in The Cancer Genome Atlas (TCGA) database and the LIRI-JP project in the International Cancer Genome Consortium (ICGC) database. We then conducted in vitro and in vivo experiments to validate the reliability of these biomarkers in HCC.

The molecular differences between HCC tissues from LTSs and STSs
In this study, we first compared the global gene expression profiles of treatment-naive, surgically resected HCC samples from LTSs (n = 40; TCGA-LIHC cohort) and STSs (n = 68; TCGA-LIHC cohort) ( Figure  1A and Table 1). The median survival time was much shorter in STSs than that in LTSs ( Figure 1B). Based on our predefined cut-off criterion (false discovery rate [FDR] < 0.01), 1199 DEGs were identified, including 760 upregulated and 439 downregulated genes in STS tumor tissues versus LTS tumor tissues ( Figure 1C).
We then used the Database for Annotation, Visualization and Integrated Discovery (DAVID) to perform Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses of the DEGs. We found that the cell cycle was the most significant pathways enriched in STS tumor samples ( Figure 1D, 1E). Gene Set Enrichment Analysis (GSEA) using GO terms as the reference gene set revealed that GO terms associated with cell cycle regulation (e.g., mitotic cell cycle, mitotic nuclear division, DNA replication, cell cycle G1/S phase transition) were enriched in primary HCC samples from STSs compared with those from LTSs ( Figure 1F). We then performed GSEA using canonical pathways mainly from Reactome and the Pathway Interaction Database (PID) as the reference gene set. The results also demonstrated that cell cycle-related pathways such as the mitotic cell cycle, DNA strand elongation, G1/Sspecific transcription and E2F pathway were significantly enriched in tumor tissues from STSs ( Figure 1G).
Metastasis and the anticancer immune response are very important determinants of the treatment response and prognosis of HCC patients [17,18]. Therefore, we assessed the expression of genes involved in these processes in HCC samples from STSs and LTSs. Surprisingly, no obvious differences were observed between the two groups (Supplementary Figure 1), suggesting that these pathways do not critically influence the survival of HCC patients.
To confirm the correlation between cell cycle-related gene expression and the survival of HCC patients, we constructed a protein-protein interaction (PPI) network with the upregulated genes in STS HCC samples, and used the Molecular Complex Detection (MCODE) algorithm to determine the top significant modules in the network. Consistently, many nodes in module 1 were target genes of E2F transcription factors (e.g., BUB1B, CDCA8, CENPE, KIF2C, NUP107, PLK1 and RAD21) ( Figure 1H), which promote the G1/S transition in the cell cycle via the CyclinD-CDK4/6-Retinoblastoma-E2F pathway. In addition, many nodes in module 1 were cell cycle regulators (e.g., BUB1 and CCNB1) ( Figure 1H). Most of the nodes in module 2 were also target genes of E2Fs (e.g., MCM2-6, ORC6  and CHEK1) ( Figure 1I). These data suggested that genes involved in cell cycle regulation, but not in cancer metastasis or antitumor immunity, are essential determinants of the survival time of HCC patients.

Cell cycle pathways were greatly enriched in matched tumor tissues from STSs of HCC
Matched primary tumor tissues and normal liver tissues were available for nine of the STSs included in this study. Therefore, we analyzed the global gene expression profiles of these tissues. After filtering the data based on a predefined cut-off criterion (FDR < 0.01), we identified 1356 DEGs between the tumor and normal tissues of STSs, including 668 upregulated and 688 downregulated genes in the tumor tissues versus normal liver tissues ( Figure 2A). GSEA revealed that gene sets involved in cell cycle regulation (e.g., the mitotic cell cycle, DNA replication and E2F-mediated regulation of DNA replication) were enriched in the primary tumor tissues ( Figure 2B).
Next, we used DAVID for GO and BioCarta pathway analyses of the DEGs between tumor and normal tissues from STSs. The BioCarta pathway analysis indicated that CDKs and cyclins, which regulate DNA replication and the cell cycle, were overrepresented in STS tumor samples ( Figure 2C). In the GO analysis, cell cycle-related terms such as mitotic nuclear division AGING tissues from STSs. Gene sets were annotated with canonical pathways mainly from Reactome. NES = normalized enrichment score; P = nominal P value; FDR = false discovery rate. (C) and (D) BioCarta pathway analysis (C) and GO analysis (D) of DEGs via DAVID. Blue bars that cross the threshold line (P < 0.01) represent pathways that differed significantly between primary tumor tissues and matched normal liver tissues from nine STSs. (E) The MCODE algorithm was used to identify the most significant module in the PPI network constructed from the upregulated genes in STS HCC samples. The MCODE score of this module was 37. and the G1/S transition of the mitotic cell cycle were also significantly upregulated in STS tumor samples ( Figure 2D).
We then constructed a PPI network with the upregulated genes in STS HCC samples, and used MCODE algorithm to identify the most significant module ( Figure 2E). Many nodes in this module were target genes of E2Fs (e.g., CDK1, BUB1B, CDCA8, CDC20, CCNB2, CENPE, PLK1 and RAD21) ( Figure 2E). These findings demonstrated that genes involved in cell cycle pathways were upregulated in tumor tissues from STSs.

The upregulation of cell cycle pathways was associated with a poor prognosis in HCC patients
To further investigate the relationship between the expression of cell cycle-related genes and the prognosis of HCC patients, we analyzed the expression levels of 300 mitotic cell cycle genes from the Reactome in 40 LTS and 68 STS tumor samples. After stratifying the tumor samples based on their gene expression patterns and examining the proportion of LTSs, we observed two clusters. Cluster I expressed higher levels of mitotic cell cycle genes and captured a higher proportion of STSs (77.3%), while cluster II expressed lower levels of mitotic cell cycle genes and captured a lower proportion of STSs (40.5%) ( Figure 3A). Then, we examined the prognoses of the patients in these clusters, and found that cluster I exhibited a poorer survival probability, whereas cluster II exhibited a better prognosis ( Figure 3B).
To confirm these results, we stratified the tumor samples from LTSs and STSs into two clusters based on their expression of cell cycle G1/S phase transition genes from the GO resource. Cluster I expressed higher levels of cell cycle G1/S phase transition genes and captured a higher proportion of STSs (78.0%), while cluster II expressed lower levels of cell cycle G1/S phase transition genes and captured a lower proportion of STSs (44.9%) ( Figure 3C). Moreover, cluster I exhibited a poorer prognosis ( Figure 3D). These data suggested that the upregulation of cell cycle pathways is associated with a poorer prognosis in HCC patients.

The expression signature of G1/S phase transition inducers was an independent prognostic factor in HCC patients
CDK4, CDK6 and CDK2 are major inducers of the G1/S phase transition, and function in complexes with their cyclin partners [19]. In total, 13 positive regulators of the CyclinD-CDK4/6-Retinoblastoma-E2F pathway and the CyclinE-CDK2 pathway directly promote the cell cycle transition from G1 into S phase [19]. Therefore, we compared the expression of these genes between STSs and LTSs. The levels of eight positive regulators (CDK2, CDK4, CCNE1, CCNE2, E2F1, E2F2, E2F3 and TFDP1) were significantly increased in STS than in LTS tumor samples ( Figure 4A, 4B), whereas the levels of the other five positive regulators (CDK6, CCND1, CCND2, CCND3 and TFDP2) did not differ between the two groups (Supplementary Figure 2). Then, we performed an unsupervised hierarchical clustering analysis to determine the expression patterns of these eight positive regulators in 355 HCC samples from TCGA. Two groups were observed: cluster I expressed higher levels of these genes, had a higher proportion of STSs (30.1%) and exhibited significantly poorer survival ( Figure 4C, 4D), whereas cluster II expressed lower levels of these genes, had a lower proportion of STSs (11.8%) and exhibited a better prognosis ( Figure  4C, 4D).
We then performed Cox proportional hazards regression analyses to investigate whether the expression pattern of these eight positive regulators or patients' clinical characteristics were associated with overall survival. In a univariate analysis, five factors (residual tumor, pathologic stage, T stage, treatment procedures and the expression pattern of the eight positive regulators) were associated with patients' overall survival with P values < 0.05 (Table 2). However, in a multivariate analysis, only R2 residual tumors (hazard ratio = 9.5270, P = 0.03652) and high levels of the eight positive regulators (hazard ratio = 1.8178, P = 0.00478) were found to be independent prognostic factors in HCC patients (Table 2). Subsequently, we examined the effects of each of the eight G1/S phase transition inducers on patients' prognoses. Notably, higher levels of seven of these genes (CDK2, CDK4, CCNE1, CCNE2, E2F1, E2F2 and E2F3) were associated with poorer survival ( Figure 4E and Supplementary Figure 3). Thus, these cell cycle-related genes that promote the G1/S transition are promising prognostic markers in HCC patients.

HCC patients with TP53 mutations tended to be STSs
Next, we investigated whether there were differences in somatic mutations between LTSs and STSs from TCGA. TP53 is a well-known negative regulator of the AGING  G1/S phase transition of the cell cycle [20][21][22]. We found that TP53 mutations accounted for more than 28% of the mutations in HCC patients, and the majority of TP53 mutations were missense mutations ( Figure 5A, 5B). Interestingly, HCC patients with TP53 mutations mainly belonged to the STS group (Table 3) and exhibited a poor overall survival rate ( Figure 5C). HCC patients with TP53 mutations and high CDK4 (or E2F1) levels had the shortest survival expectancy of all the groups ( Figure 5D). These data suggested that TP53 mutations contribute to and are promising diagnostic markers of STSs.

The expression of G1/S phase transition inducers and TP53 mutations predicted survival in the ICGC dataset
We then investigated whether the eight positive regulators of the G1/S phase transition and TP53 mutations exhibited similar prognostic trends in other HCC datasets. We chose the LIRI-JP project in the ICGC database, which contained data from 203 HCC patients. Consistent with our previous results ( Figure 4E and Supplementary Figure 3), higher levels of seven inducers of the G1/S phase transition (CDK2, CDK4, CCNE2, E2F1, E2F2, E2F3 and TFDP1) were associated with poorer survival ( Figure 6A). Moreover, HCC patients with wild-type TP53 and low CDK4 (or E2F1) expression had the longest survival expectancy ( Figure 6B). These data confirmed that the levels of seven G1/S phase transition inducers and TP53 mutations predicted survival in HCC patients.

Abemaciclib significantly inhibited the proliferation of HCC cells in vitro
Because higher levels of CDK4 and its downstream effector E2F1 were observed in STSs ( Figure 4E and Figure 6A), we speculated that inhibiting CDK4 might prolong the survival of HCC patients. Therefore, we assessed the effects of abemaciclib (the first FDAapproved CDK4 inhibitor for breast cancer treatment) on the HCC cell lines Hep3B and Huh7 [23]. Retinoblastoma (RB) is inactivated upon its phosphorylation by CDK4, so we detected the protein levels of phosphorylated RB and total RB in these two cell lines. While total RB protein levels remained relatively constant in Hep3B and Huh7 cells, phospho-S780-RB levels decreased significantly following abemaciclib treatment ( Figure 7A). Abemaciclib also AGING The patients were stratified into mutant TP53 group and wild type TP53 group according to TP53 mutational status.  AGING notably reduced the clone numbers of these two cell lines ( Figure 7B, 7C). A Cell Counting Kit 8 (CCK-8) assay ( Figure 7D) and a 5-ethynyl-2´-deoxyuridine (Edu) assay ( Figure 7E) confirmed that abemaciclib repressed HCC cell proliferation. These data demonstrated that abemaciclib substantially inhibited the proliferation of HCC cells in vitro.

Abemaciclib inhibited the tumor growth of HCC cells in vivo
Finally, we assessed the effects of abemaciclib on HCC development in vivo in mice subcutaneously injected with Hep3B cells. Abemaciclib treatment significantly reduced the size ( Figure 8A) and weight ( Figure 8B) of the tumors in these mice. Dynamic analyses of tumor development also revealed that the tumor volume was significantly reduced after abemaciclib treatment ( Figure  8C). These data demonstrated that abemaciclib can significantly inhibit the proliferation of HCC cells in vivo.

DISCUSSION
Prognostic prediction for cancer patients is vital, not only for identifying potential STSs who will need special treatments, but also for designing appropriate strategies to prolong survival [24]. For example, B-Raf proto-oncogene (BRAF) mutation, especially V600E, is an ideal marker of poor prognosis and new treatment target in colon cancer patients [25,26]. However, the ability to predict HCC patients' prognoses is still limited [1]. Here, we identified molecular markers that could effectively distinguish STSs from LTSs of HCC, allowing HCC patients to be stratified into good and poor prognosis groups easily and efficiently. Patients with low levels of eight G1/S phase transition inducers survived for nearly 7 years, whereas patients with high levels of these eight genes had a survival expectancy of only 2.07 years. Besides G1/S phase transition inducers, TP53 mutations are another critical factors to STSs, and promising biomarker for HCC prognosis. Therefore, we speculated that HCC patients with heightened expression of G1/S phase transition inducers and TP53 mutations tent to be STSs; and these predicted poor prognosis patients desperately need to receive more timely and effective treatments.
Many proteins involved in the G1/S transition contribute substantially to the development and progression of HCC [27][28][29][30]. Aberrant expression and activation of AGING these genes can stimulate the proliferation of HCC cells [31,32], and uncontrolled cell proliferation promotes cancer initiation [33]. Here, we demonstrated that genes that induce cell proliferation were highly expressed in STSs, suggesting that they are critical for cancer development. Sonntag et al. reported that CCNE1 and CDK2 were crucial for the initiation but not the progression of HCC [23]. These contradictory results may largely because of their synergistic effects; thus, it is important to treat these proteins as a unit. Intriguingly, though immune pathway activation and metastasis are key determinants of cancer development and patient survival [34,35], we observed no significant differences in these two pathways between STSs and LTSs (Supplementary Figure 1). These interesting results may be contributed to the characteristics of HCC, including the short survival time and shortage of immunotherapies [18,36]. Further clinical studies will be valuable to elucidate the involvement of these pathways in the development and treatment of HCC.
Aberrant activation of CDK4, one of the eight positive regulators of the G1/S transition, is frequently seen in various cancers, including HCC [37,38]. As one of the most fatal cancers, HCC is an aggressive cancer with dismal prognosis, largely contributing to lack of efficient treatment [39]. In this study, we demonstrated that CDK4 expression was significantly lower in LTS HCC tissues than in STS HCC tissues, suggesting that inhibiting CDK4 might prolong the survival of HCC patients. In recent years, abemaciclib, a highly selective and FDA-approved oral inhibitor of CDK4, has been clinically applied with promising therapeutic effects [40][41][42]. However, its effects on HCC have not been clarified [43]. We found that abemaciclib suppressed HCC cell proliferation in vitro and tumor growth in vivo, suggesting that it was an attractive drug for HCC patients, especially those with high CDK4 expression. Thus, it is of value to investigate the effects of abemaciclib on HCC through further experimentations and clinical explorations. Additionally, Goel demonstrated that CDK4 inhibition activates anti-tumor immunity [44]; therefore it is of interest to treat STS patients with combined therapy including CDK4/6 inhibitor and immune-checkpoint inhibitor. Besides CDK4, CDK2 is highly expressed in the tumor tissues of STSs. Accumulating evidence suggest that CDK2 inhibition are particularly useful for several cancers including lung cancer, prostate cancer, and breast cancer [45]. Herein, HCC patients with high CDK2 might have particular susceptibility to CDK2 inhibition.
Our study had several limitations. First, the numbers of LTSs and STSs in the TCGA-LIHC cohort were relatively small (40 LTSs and 68 STSs). Moreover, only nine matched HCC and normal liver samples from the STSs were available. Second, the clinical applicability of the CDK4/6 inhibitor abemaciclib for HCC treatment is still unclear, and clinical trials are needed to investigate its efficacy. Further studies are also needed to determine whether positive regulators of the G1/S transition can be used as prognostic predictors and treatment targets in HCC patients.
In summary, we performed a comprehensive comparison of the transcriptome between STSs and LTSs of HCC, and demonstrated that genes that induce the G1/S transition were significantly enriched in STSs. TP53 mutations, the most common mutations in HCC, were also associated with STSs. Higher expression of eight G1/S transition inducers strongly predicted a poorer prognosis in HCC patients. In terms of HCC treatment, abemaciclib substantially inhibited the proliferation of H CC cells. This study has suggested new targets for the personalized diagnosis and treatment of STSs of HCC.

TCGA data
RNA-seq raw counts (HTSeq -Counts) data, Mutation Annotation Format (MAF) data (generated by the mutect2 algorithm from whole-exome sequencing data), clinical data and survival data were downloaded from the LIHC project of TCGA (https://cancergenome. nih.gov/). LTSs were defined as patients whose overall survival extended 5 years from surgery, while STSs were defined as patients who survived > 3 months and < 1 year from surgery, to exclude perioperative mortalities. In total, 361 HCC patients with survival data were included in our analyses, including 40 LTSs and 68 STSs. RNA-seq data from primary tumor tissues were available for 355 of these HCC patients. RNA-seq data from primary tumor tissues and matched normal tissues were available for nine STSs. Mutation data were available for 348 HCC patients, including 342 patients with RNA-seq data from primary tumor tissues.

ICGC data
Normalized RNA-seq data (fragments per kilobase million, FPKM), annotated mutation data, clinical data and survival data were downloaded from the LIRI-JP project of the ICGC database (https://www.icgc.org). In total, 203 HCC patients were included in our analyses. RNA-seq data from primary tumor tissues, along with mutation data and survival data, were available for all patients.

Prognostic analyses
Survival curves were constructed using the Kaplan-Meier method and compared using the log-rank test via AGING the survival R package. Univariate and multivariate Cox proportional hazards regression analyses were conducted using the survival R package.

DEG analysis
The DEGs between primary tumor tissues from LTSs and STSs or between primary tumor tissues and matched normal liver tissues from STSs were analyzed with the edgeR R/Bioconductor package [46]. Genes with FDRs < 0.01 were considered to be significantly DEGs. GO, KEGG and BioCarta pathway enrichment analyses of the DEGs were performed using DAVID (https://david.ncifcrf.gov/) [47].

PPI network analysis
We used the plug-in StringApp [48] in Cytoscape software [49,50] to construct PPI network of the DEGs with default parameters. Subsequently, Molecular Complex Detection (MCODE) [51] in Cytoscape was applied to screen top significant modules within PPI network with default parameters.

GSEA
GSEA was implemented through GSEAPreranked tool in GSEA software (http://www.broadinstitute.org/gsea/) [52]. The value of log2 (FC) calculated by edgeR package was used as ranking metric. We chose C5 collection that contains genes sets annotated by GO terms in the Molecular Signatures Database (MSigDB) (https://www.gsea-msigdb.org/gsea/msigdb) as the reference gene set in GSEA software. We also used the canonical pathways sub-collection of C2 collection in the MSigDB as the reference gene set in GSEA software. The annotated pathways in this sub-collection are mainly from BIOCARTA, KEGG, REACTOME and PID databases.

Analysis and visualization of somatic mutations
We used the Maftools R/Bioconductor package [53] to extract detailed mutational information from the MAF file. The OncoPlot function [54] and the lollipopPlot function were used to create the OncoPlot of the top ten mutated genes and the lollipopPlot of TP53, respectively.

Hierarchical clustering
The gene expression data for two genes sets (the cell cycle G1/S phase transition in GO, and the mitotic cell cycle in Reactome) were extracted from normalized and log2-transformed RNA-seq data from 108 HCC samples in TCGA. Unsupervised hierarchical clustering was used to discover groups based on the expression patterns of the genes in these two genes sets. In total, 108 cell cycle G1/S phase transition genes from the GO resource and 300 mitotic cell cycle genes from the Reactome were used. The expression values of these genes in the 108 samples were hierarchically clustered with Euclidean distances and Ward linkages using the dendextend R package [55]

Western blotting
HCC cells were lysed with a strong radioimmunoprecipitation assay buffer containing Halt TM Protease Inhibitor Cocktail (Thermo, Waltham, MA, USA). The concentrations of the proteins in the lysate were determined with a bicinchoninic acid protein assay kit (Pierce, Rockford, IL, USA). Proteins were detected on a 12% sodium dodecyl sulfate-polyacrylamide gel electrophoresis and then transferred onto a polyvinylidene fluoride membrane. The membranes were incubated overnight at 4° C with primary antibodies for retinoblastoma (phospho S780) (ab47763, Abcam, AGING Cambridge, UK), retinoblastoma (ab181616) and GAPDH (BM1623, Boster, Wuhan, China), and then were incubated with the appropriate horseradish peroxidase-conjugated secondary antibodies for 2 hours. The proteins were detected using an Amersham Imager 600 (GE Healthcare Life Sciences, Boston, MA, USA).

Clone formation assay
For the clone formation assay, 1000 HCC cells were seeded in each well of a six-well plate and cultured in an incubator for more than 12 days. The medium was replaced every three days. Then, the cells were fixed with methanol for 20 min and stained with crystal violet. Pictures were taken with a digital camera (Nikon, Japan).

CCK-8 assay
HCC cells were cultured in 96-well plates at a density of 1000 cells/well. After the cells had been cultured for 12, 36 or 60 hours, each well was treated with 10 µL of the assay reagent (Dojindo Molecular Technologies, Kumamoto, Japan), and the plate was returned to the incubator for 2 hours. Then, the absorbance at 450 nm was recorded with a SpectraMax M5 microplate reader (Thermo Fisher Scientific, Waltham, MA, USA).

Animal model
Twelve male BALB/c nude mice (four weeks old) were obtained from SLAC Laboratory Animal Company (Shanghai, China). For tumor generation, the mice received a subcutaneous axillary injection of Hep3B cells (6×10 6 cells/mouse) in the right forelimb. The mice were gavaged with abemaciclib (50 ug/g) daily, beginning seven days after the cell injection. The mice were sacrificed after 28 days, and their tumor weights were evaluated.

Ethical approval and consent to participate
All the animal experiments were performed according to National Institutes of Health guide for the care and use of laboratory animals (NIH Publications No. 8023, revised 1978) and were approved by the Institutional Animal Care and Use Committee of Tongji University.

Statistical analysis
All the data represented three independent experiments, with the data expressed as the mean ± SD. Differences between the Control and Abemaciclib groups were analyzed using Student's t-test in SPSS 17.0. A p-value less than 0.05 was considered significant.

Availability of data and materials
All data analyzed during this study are included in this published article. Raw and processed data are stored in the laboratories are available upon reasonable request.  The patients were stratified into the high expression group and the low expression group according to the median of normalized RNA-seq data.