Gene expression signatures of ADHs across different HCC samples
In order to distinguish the expression levels of ADHs between normal and tumor liver tissues, the transcriptome of 50 normal liver and 269 HCC samples was identified. As shown in Fig. 1, the expression level of ADH1A, ADH1B, ADH1C, ADH4, and ADH6 was significantly downregulated in HCC samples compared to normal liver samples. Interestingly, the expression level of ADH5 was slightly but significantly upregulated in HCC samples (Fig. 1F). Furthermore, all the ADH genes showed a positive correlation with each other (Fig. 2).
Alcohol consumption increases the risk for liver cancer, and is also considered as a primary cause of HCC through the development of cirrhosis [26]. In the present study, our data showed that the expression level of ADH family members including ADH1A-ADH6 was significantly increased in alcohol consumption HCC patients compared to non-alcohol consumption ones (Fig. 3). Obviously, the expression of ADH1B and ADH4 was strongly upregulated (Fig. 3B and 3D).
In order to characterize the correlation between ADHs and tumor stages, the expression level of ADHs was identified in HCC samples with different T classifications. Our results showed that the expression level of ADHs including ADH1A, ADH1B, ADH1C, and ADH6 was obviously decreased with the progression of tumor malignancy (Fig. 4A-4C and 4F). And the expression level of ADH4 was higher in HCC patients with T2 classification than T1 and T3 classifications, but it was lowest in HCC samples with T3 classification (Fig. 4D). While the expression level of ADH5 was remarkably increased with the progression of tumor malignancy (Fig. 4E).
Association between ADHs expression and survival time of HCC patients
In order to explore the prognostic value of ADHs expression level in the patients with HCC, all the patients were categorized into high and low expression groups based on the expression level of ADHs as described above. Primarily, the association of ADHs expression and clinicopathological characteristics was determined with two-sided Fisher’s exact tests. Our results showed that the expression of ADH1B, ADH1C, ADH4, and ADH5 was significantly associated with age. Moreover, the expression of ADH1A, ADH1B, and ADH4 was significantly associated with histologic stage. In addition, ADH1A and ADH4 expressions were significantly associated with T classifications (Table 1).
Table 1
Association analysis between ADHs expression and clinicopathological variables in 269 HCC patients
Variables | ADH1A | ADH1B | ADH1C | ADH4 | ADH5 | ADH6 |
High | Low | p | High | Low | p | High | Low | p | High | Low | p | High | Low | p | High | Low | p |
All patients | 112 | 157 | | 174 | 95 | | 87 | 182 | | 186 | 83 | | 114 | 155 | | 143 | 126 | |
Gender | | | 0.685 | | | 0.261 | | | 0.391 | | | 0.029 | | | 0.105 | | | 0.003 |
Males | 78 | 113 | | 128 | 63 | | 65 | 126 | | 140 | 51 | | 87 | 104 | | 113 | 78 | |
Females | 34 | 44 | | 46 | 32 | | 22 | 56 | | 46 | 32 | | 27 | 51 | | 30 | 48 | |
Age (years) | | | 0.216 | | | 0.015 | | | 0.050 | | | 0.018 | | | 0.007 | | | 0.462 |
< 60 | 46 | 77 | | 70 | 53 | | 32 | 91 | | 76 | 47 | | 41 | 82 | | 62 | 61 | |
≥ 60 | 66 | 80 | | 104 | 42 | | 55 | 91 | | 110 | 36 | | 73 | 73 | | 81 | 65 | |
Histologic grade | | | 0.005 | | | 0.0009 | | | 0.060 | | | 0.001 | | | 0.800 | | | 0.165 |
Low (Grade 1 + 2) | 82 | 88 | | 123 | 47 | | 62 | 108 | | 130 | 40 | | 71 | 99 | | 96 | 74 | |
High (Grade 3 + 4) | 30 | 69 | | 52 | 48 | | 25 | 74 | | 56 | 43 | | 43 | 56 | | 47 | 52 | |
T classification | | | < 0.0001 | | | 0.253 | | | 0.466 | | | 0.039 | | | 0.680 | | | 0.274 |
Low (T 1 + 2) | 92 | 55 | | 103 | 61 | | 66 | 129 | | 142 | 53 | | 81 | 114 | | 108 | 87 | |
High (T 3 + 4) | 20 | 54 | | 40 | 34 | | 21 | 53 | | 44 | 30 | | 33 | 41 | | 35 | 39 | |
Alcohol consumption | | | > 0.999 | | | 0.790 | | | 0.785 | | | 0.054 | | | 0.439 | | | |
Yes | 39 | 55 | | 62 | 32 | | 29 | 65 | | 72 | 22 | | 43 | 51 | | 52 | 62 | 0.336 |
No | 73 | 102 | | 112 | 63 | | 58 | 117 | | 114 | 61 | | 71 | 104 | | 91 | 84 | |
Tumor recurrence | | | 0.269 | | | 0.899 | | | 0.118 | | | 0.088 | | | 0.805 | | | 0.903 |
Yes | 50 | 81 | | 84 | 47 | | 36 | 95 | | 84 | 47 | | 57 | 74 | | 69 | 62 | |
No | 62 | 76 | | 90 | 48 | | 51 | 87 | | 102 | 36 | | 57 | 81 | | 74 | 64 | |
Furthermore, univariate and multivariate Cox regression analyses were also performed to assess the prognostic value of ADHs expression and clinicopathological characteristics. As shown in Table 2, univariate Cox regression analysis results presented that high T classification, alcohol consumption, tumor recurrence were significantly associated with poor OS for HCC patients. Moreover, higher T classification and tumor recurrence were also significantly associated with poor RFS. Interestingly, high expression of ADH1A, ADH1C, ADH4, and ADH6 was considered as an independent factor with an improved prognosis for OS and RFS. While high expression of ADH1B was only was considered as an independent factor with an improved prognosis for RFS (Table 2). In addition, our multivariate Cox regression analysis results showed that higher T classification and tumor recurrence were also significantly associated with poor OS and RFS (Figs. 5 and 6). Interestingly, high expression of ADH1A, ADH1C, ADH4, and ADH6 was also considered as an independent factor with an improved prognosis for OS (Fig. 5). And high expression of ADHs without ADH5 was considered as an independent factor with an improved prognosis for RFS (Fig. 6).
Table 2
Univariate analysis of the expression of ADHs expression with OS and RFS in 269 patients with HCC
Variables | OS | | RFS |
HR (95%CI) | p value | | HR (95%CI) | p value |
Gender (males vs. females) | 0.904 (0.560–1.458) | 0.678 | | 0.798 (0.495–1.288) | 0.356 |
Age (≤ 60 vs. > 60 years) | 1.361 (0.856–2.162) | 0.193 | | 1.314 (0.827–2.089) | 0.247 |
Histologic grade Low (Grade 1 + 2) vs. High (Grade 3 + 4) | 1.051 (0.660–1.674) | 0.833 | | 1.144 (0.719–1.821) | 0.571 |
T classification Low (T1 + 2) vs. High (T3 + 4) | 3.819 (2.419–6.030) | 8.86e-09 | | 4.386 (2.766–6.952) | 3.21e-10 |
Alcohol consumption (yes vs. no) | 1.279 (0.795–2.058) | 3.10e-01 | | 1.036 (0.646–1.662) | 0.884 |
Tumor recurrence (yes vs. no) | 2.279 (1.383–3.755) | 0.001 | | 4.201 (2.507–7.039) | 5.06e-08 |
ADH1A expression (high vs. low) | 0.436 (0.261–0.727) | 0.001 | | 0.415 (0.248–0.693) | 0.0008 |
ADH1B expression (high vs. low) | 0.641 (0.405–1.015) | 0.058 | | 0.609 (0.385–0.964) | 0.034 |
ADH1C expression (high vs. low) | 0.516 (0.300-0.888) | 0.017 | | 0.518 (0.301–0.892) | 0.018 |
ADH4 expression (high vs. low) | 0.380 (0.240–0.600) | 3.41e-05 | | 0.386 (0.245–0.609) | 4.17e-05 |
ADH5 expression (high vs. low) | 0.708 (0.436–1.148) | 0.161 | | 0.677 (0.418–1.096) | 0.112 |
ADH6 expression (high vs. low) | 0.544 (0.342–0.864) | 0.009 | | 0.558 (0.351–0.886) | 0.013 |
Abbreviations: HR: hazard ratio, CI: confidence interval, p values were calculated with log-rank test. |
Next, the correlation between ADHs expression and OS/RFS was evaluated with Kaplan-Meier analysis and log-rank test. Our results showed that high expression of AHD1A, ADH1C, ADH4, and ADH6 was significantly associated with good OS and RFS in HCC patients (Figs. 7 and 8). Interestingly, the RFS rate of HCC patients with high ADH1B expression was significantly better than that of patients with low ADH1B expression (Fig. 8B).
Identification of involved pathways related to ADHs expression in HCC
In order to identify the potential biological functions of ADHs, GO terms and KEGG pathways enrichment analysis was performed with R project as above description. GO terms enrichment analysis revealed that plenty of pathways were well enriched, and 9 of them including ethanol oxidation, ethanol metabolic process, primary alcohol metabolic process, retinoid metabolic process, diterpenoid metabolic process, terpenoid metabolic process, retinol metabolic process, isoprenoid metabolic process, antibiotic metabolic process were enriched for all the members of ADHs family (Fig. 9A and Table S1). Moreover, KEGG pathways enrichment analysis results showed that 7 pathways were significantly enriched, including tyrosine metabolism, fatty acid degradation, retinol metabolism, glycolysis/gluconeogenesis, drug metabolism-cytochrome P450, metabolism of xenobiotics by cytochrome P450, chemical carcinogenesis (Fig. 9B and Table S2).
In order to investigate the role of ADHs in the pathogenesis of HCC, GSEA was performed between datasets with ADHs high expression and low expression. Our results unveiled that ADHs were enriched in serval signaling pathways (Table S3 and S4). Based on the enrichment with KEGG database, the high expression group of ADHs was positively associated with retinol metabolism, and fatty acid metabolism with the significant difference (Fig. 10A and 10B). And the low expression group of ADHs family members not including ADH5 was significantly but negatively associated with pathways in cancer (Fig. 10C). Furthermore, we utilized GSEA to enrich in cancer related pathways based on pathway interaction database (PID). As can be seen from Fig. 10D-10H, low expression group of ADHs family members without ADH5 was significantly enriched in several cancer related pathways, including ATR, FOXM1, MTOR, NOTCH, and P53 downstream pathway.