We performed first survival analysis based on BCLC stages, E-S grades and recurrent status and then tried to correlate genomic profile of CNAs with clinicopathologic classifications. The BCLC staging classifies HCC based on liver functional status, physical status and cancer-related symptoms and linked the four stages with treatment algorithm [25, 26]. E-S grading is determined by tumor histologic and cytologic findings [27, 28]. The general genomic profile of CNAs from this case series showed a consistent pattern or shared similar CNAs from previous studies [3–18]. The most commonly seen CNAs in more than 30% of our case series, including gains of 1q, 5p and 8q and losses of 4q, 8p, 13q, 16q and 17p as shown in Fig. 2A, were all recurrent CNAs for HCC. Bioinformatic and gene expression analyses had been used to define candidate genes from these recurrent CNAs. Candidate genes as well as key genes and pathways from nine studies were listed in Supplemental Table 2. This list of genes showed large variations from different studies. Further gene functional analysis would be needed to validate the causal or modifying roles of these genes for HCC. Analysis of gene expression levels in CNAs noted that CNA gain or amplification of the MDM4 gene at 1q32 associated with poor metastasis-free survival [5, 16]. Mechanistically, gains of 1q resulted from hypomethylation of 1q heterochromatin and jumping translocations with two or more recipient chromosomes were seen in myelodysplastic syndrome acute myeloid leukemia [29]. Gains of 8q involving the MYC gene overexpression was observed in viral and alcohol-related HCCs [5]. Significantly upregulated genes ATAD2, SQLE, PVT1, ASAP1 and NDRG1 associated with gains of 8q24.13q24.3 was an unfavorable prognostic marker for HCCs [15]. Loss of 8p involving the DLC1, CCDC25, ELP3, PROSC, SH2D4A genes associated with poor outcomes for HCC; in vitro and in vivo analysis indicated that the PROCS, SHAD4A and SORBS3 genes have tumor-suppressive activities, along with the known tumor suppressor gene DLC1 [9]. Integrated analysis of somatic mutations and CNAs in HCCs identified 56 key genes and five pathways for HCC [8]. At least 38% (21/56) of these key genes could be mapped to the recurrent CNAs and affected four core pathways of p53/cell cycle control, chromatin remodeling, PI3K/Ras signaling, and oxidative and endoplasmic reticulum stress (Supplemental Table 2). Multiple genes from recurrent CNAs affected the p53/cell cycle control, including the CDKN2C gene at 1p32.3, the CDK11A/B genes at 1p36.33, the IRF2 gene at 4q35, the RB1 gene at 13q14.2 and the TP53 gene at 17p13.1. These recurrent CNAs and affected genes and pathways were considered an integral part of the comprehensive genetic landscape and genomic characterization of HCC [30, 31].
Correlating genomic CNAs with the HCC stages and grades could be helpful to interpret test results and guide clinical treatment. A dendrogram of the cluster analysis showed first the gain of 1q then gains of 8q and 5p followed by other CNAs; the gains of 1q and 8q were significantly associated with E-S grades II-IV [18]. Another study compared the recurrent CNAs between E-S grades I/II and III/IV and noted that gain of 8q was statistically more frequently seen in grade III/IV [14]. In general, cases in E-S grade III seemed to have more CNAs than those in grade II. More specifically, significant association of losses of 1p36.31p22.1, 4q13.2q35.2 and 10q22.3q26.13 and gains of 2q11.2q21.2 and 20p13p11.1 with E-S grade III were noted. After we stratified the cases according to BCLC stages, two regions significant in BCLC stage A were losses of 4q13.2q35.2 and 10q22.3q26.13 (Supplemental Fig. 3). The 4q13.2q35.2 included the IRF2 gene for p53/cell cycle control and SMARCAD1 gene for chromatin remodeling, while the 10q22.3q26.13 included the PTEN gene for PI3K/Ras signaling (Supplemental Table 2). These genes might be important for cancer cell differentiation and progression. Counterintuitively, our cases showed a better survival and less recurrence for E-S grade III than grade II at BCLC stage A. One hypothesis to explain this observation was that in BCLC stage A usually involved one tumor and the tumor in E-S grade III was cytologically more distinct from tumor in grade II, which might make the tumor in grade III more likely to be removed completely during the surgery than in tumors in grade II. This could also explain significantly less recurrence in E-S grade III than in grade II in BCLC stage A (P < 0.001). In BCLC stage C, the cancer cells have been spread to blood vessels, lymph nodes or other body organs. The recurrence rate after the surgery is similar between E-S grades II and III in this stage (P = 0.72). The loss of 1p36.31p22.1 and gains of 2q11.2q21.2 and 20p13p11.1 and additional gains of chromosomes 6, 7 and 20q were associated BCLC stage C, E-S grade III and recurrence for the worst survival outcome. The ARID1A, CDKN2C, CDK11A and CDK11B genes at 1p could affect the p53/cell cycle control and chromatin remodeling pathways (Supplemental Table 2). A study focused on gene expression from gains of 20q identify candidate genes contributing to unfavorable outcomes for HCC. Overexpression of the DDX27, B4GALT5, RNF114, ZFP64 and PFDN4 associated significantly with vascular invasion, and high RNF114 expression also associated with advanced tumor stage [12]. We didn’t find significant association between CNAs and BCLC stage or recurrence status. The association of CNAs with cytologic and histologic findings by E-S grades may reflect a link between molecular and cellular levels. Additionally, clustering analysis noted more CNAs in cases with poor survival outcomes but ROC analysis did not support the association of percentage of CNAs with clinicopathologic classifications.
This study provided preliminary results correlating genomic CNAs to clinicopathologic findings. However, two major limitations should be mentioned from this study. Firstly, the limited number of cases possibly introduced bias in case stratification. The results from this study need to be further validated from a large cohort of HCC cases. The second limitation was the technical challenge in tracking the clonal evolution and dissecting tumor heterogeneity [32, 33]. All tumor specimens were collected at the surgical procedures and thus made it difficult to look into initial event and accumulated aberrations from different tumor stages. HCC is a highly heterogeneous disease at the clinicopathologic level in association with comprehensive genomic heterogeneity from accumulated CNAs, somatic mutations and epigenetic alterations. A biopsy-based integrative diagnostic approach including morphology, immunohistochemistry, transcriptomic data, mutational profiles, CNA and methylome analysis have been proposed for future analysis of HCC [34].