3.1 Transcriptional levels of the LILRB family in patients with liver cancer
The LILRB family contains five subtypes: LILRB1, LILRB2, LILRB3, LILRB4 and LILRB5. We compared the mRNA expression of the LILRB family in different cancers with that in corresponding normal tissues based on the Oncomine database. As shown in Fig. 1, compared with that of normal tissues, in addition to LILRB4, other LILRB family members were downregulated in liver tumor tissues. All the data came from four datasets, and all the tumor data came from HCC[28–31]. Three of the datasets had the same results even if two datasets did not achieve the condition of fold change >1.5, while the Mas`s dataset had the opposite results (Table.2).
The table shows that (1) the transcriptional levels of LILRB1 were significantly downregulated in HCC patients in three datasets with fold changes of -1.602, -1.251, and -1.182; (2) compared with normal tumors, LILRB2 was also downregulated in liver tumors in three datasets with fold changes of -2.098, -1.441, -1.281, and -1.708; (3) compared to normal tissues, LILRB3 was found to be decreased in HCC with fold changes of -2.078, -1.391, -1.341, and -1.322; (4) while the results from the other three datasets showed that there was no difference in LILRB4 expression between liver tumors and normal tissues, LILRB4 was upregulated in liver tumors in Mas`s datasets with fold changes of 1.7; and (5) compared with normal tissues, LILRB5 expression was significantly decreased in HCC patients in three datasets with fold changes of -1.656, -2.826, and -1.321, respectively.
To verify the above results further, GEPIA 2.0 was used to compare the mRNA expression of the LILRB family between liver tumor tissues and normal tissues. The LIMMA method was used to compare liver tumor tissues and their paired normal samples regarding mRNA expression of the LILRB family. The results showed that LILRB1, LILRB2, LILRB3, and LILRB5 were downregulated, and compared with liver normal tissues, LILRB4 expression was upregulated in liver tumor tissues, which was consistent with the results in the Oncomine database (Fig. 2).
To assess LILRB expression at the protein level, we detected LILRB family expression in liver tumors and their corresponding TFLs by using immunohistochemistry. We found that the expression of LILRB1, LILRB2, LILRB3 and LILRB5 was lower in liver tumors than in TFL, while the expression of LILRB4 was high in liver tumors than in TFL (Fig. 3).
3.2 The relationship between the LILRB family and tumor stage in liver cancer patients
We compared the relationship between LILRB family expression and four tumor stages in liver cancer patients by using GEPIA 2.0. One-way ANOVA was used to analyze these results. In agreement with the literature[16], although there was no significant correlation between LILRB family expression and tumor stage in liver cancer patients, compared with that in other stages of liver cancer patients, the expression of the LILRB family in stage IV liver cancer patients decreased (Fig. 4).
3.3 Prognostic value of LILRB family mRNA expression in all liver cancers
Kaplan-Meier Plotter was used to examine the prognostic value of LILRB family mRNA expression levels in all liver cancers. We compared the correlation between mRNA expression of LILRB family members and OS, RFS, PFS and DSS of liver cancer patients. The results revealed that patients in LILRB2 and LILRB5 low group had shortened OS and DSS and low expression of all LILRB family members predicted poorer patients RFS and PFS (p<0.05) (Fig. 5). Median survival times of LILRB high groups in OS, PFS, PFS and DSS were longer than these of LILRB low groups, when there was statistical difference between these two groups (Table.3).
Next, we further investigated the prognostic value of LILRB family mRNA expression in liver cancer with different histologic stages by using Kaplan-Meier Plotter. The results were as follows: (1) The high mRNA expression of LILRB1 was associated with better RFS and PFS in stages Ⅰ and Ⅱ. (2) The high mRNA expression of LILRB2 was associated with longer OS, RFS and PFS in stages Ⅰ, Ⅱ and Ⅲ. High mRNA expression of LILRB2 was predicted to have better DSS in stages Ⅱ and Ⅲ. (3) High expression of LILRB3 mRNA was correlated with better RFS and PFS in stage Ⅱ, while low expression of LILRB3 mRNA was associated with better DSS in stage Ⅰ. (4) The low mRNA expression of LILRB4 was associated with better DSS in stage Ⅰ. However, the high expression of LILRB4 mRNA was correlated with better RFS and PFS in stage Ⅱ as well as better PFS in stage Ⅲ. (5) High mRNA expression of LILRB5 was predicted to be associated with better OS in stage Ⅱ and Ⅲ patients and with better PFS in stage Ⅱ patients, while low mRNA expression of LILRB5 was associated with better OS in stage Ⅰ patients (Supplementary Table S1).
3.4 Prognostic value of LILRB mRNA expression with risk factors in liver cancer
The two main risk factors, alcohol consumption and hepatitis virus, could induce liver cancer. Therefore, we compared the relationship between LILRB mRNA expression and these two risk factors by using Kaplan-Meier Plotter. Regarding alcohol consumption, we summarized the following: (1) The high mRNA expression of LILRB1 was associated with better OS and DSS in liver cancer patients in alcohol-consuming cohorts and better RFS and PFS in liver cancer patients in nonalcohol-consuming cohorts. (2) In liver cancer with alcohol consumption cohorts, high LILRB2 mRNA expression was correlated with better OS, RFS, PFS and DSS. The liver cancer patients with no alcohol consumption habit who expressed high LILRB mRNA had better RFS, PFS and DSS. (3) The high mRNA expression of LILRB3 was associated with better RFS in liver patients in nonalcohol-consuming cohorts. (4) The high mRNA expression of LILRB4 was only correlated with better RFS and PFS in liver cancer patients in alcohol consumption cohorts and better RFS in liver cancer patients in nonalcohol-consuming cohorts. (5) Regardless of whether liver cancer patients drink, the high mRNA expression of LILRB5 in these two kinds of patients was predicted to have better OS, RFS, PFS and DSS (Supplementary Table S2).
The results of the relationship between mRNA expression of the LILRB family and different hepatitis virus infection statuses are shown in Supplementary Table S3. Regarding hepatitis virus-infected liver cancer patients, the results showed that (1) the high mRNA expression of LILRB2 was associated with better RFS and PFS; (2) the high mRNA expression of LILRB3 and LILRB5 was correlated with better PFS; and (3) the low mRNA expression of LILRB4 was predicted to have better OS and DSS. In contrast, the high mRNA expression of LILRB4 was associated with better RFS and PFS. Regarding nonhepatitis virus-infected liver cancer patients, we concluded that high mRNA expression of LILRB1, LILRB2 and LILRB5 was associated with better OS, RFS, PFS and DSS, and that high LILRB4 mRNA expression was correlated with better RFS.
3.5 The relationship between LILRB family expression and immune cell infiltration
The LILRB family is considered to be an immune inhibitory receptor. TILs are associated with prognostic indicators for liver cancer[32]. Therefore, we speculated that there was a correlation between LILRB family expression and TIL infiltration in liver cancer patients. Timer database was used to analyze the suspected association. The results showed that all LILRB family members were significantly associated with tumor purity, B cells, CD8+ T cells, CD4+ T cells, macrophages, neutrophils and dendritic cells (Fig. 6).
The TISIDB database was used to assess the relationship between LILRB family expression and TIL subsets. Although there was no significant correlation of LILRB5 expression with activated CD4+ T cells and CD56dim natural killer cells, the expression of other LILRB family members in liver cancer was strongly associated with various TIL subsets (Supplementary Tables S4-S8). Together these ଁndings suggested that the LILRB family plays an important role in recruitment and regulation of immune inଁltrating cells in liver cancer.
3.6 The mutations of LILRB family genes
The cBioPortal online tool was used to analyze the mutations of LILRB family genes. The main mutation types of LILRB family genes in liver cancers were missense mutations and amplifications (Fig. 7A). The mutations were concentrated in hepatocellular carcinoma and hepatobiliary cancer. A total of 5.42% of LILRB genes in 369 cases of hepatocellular carcinoma were altered, and 3.38% of LILRB genes in 916 cases of hepatobiliary cancer were altered (Fig. 7B).
Then, we analyzed the correlation between each LILRB member and their mRNA expression by using the cBioPortal online tool. The statistical method was Spearman. The results showed that there was a significant relationship between each LILRB member (Table.4) (p<0.05).
3.7 Predicted functions and pathways of the LILRB family in liver cancer
To understand the biological significance and consequences of the protein profiling of the LILRB family, we constructed a PPI network by using STRING. The Cytoscape software was used to process it. As shown in Fig. 8, PPI analysis showed known and predicted interactions between LILRB family members and 50 proteins.
Next, GO and KEGG analyses for the LILRB family and these 50 genes were performed using the DAVID online tool. The GO analysis showed that there were 38 items of biological process (BP), 25 items of cellular component (CC) and 23 items of molecular function (MF) with FDR less than 0.05. The first 10 items were used to plot. The GO analysis for BP showed that LILRB family members and their related genes were mostly enriched in the immune response and antigen processing and presentation of antigens (Fig. 9A). The GO analysis for CC revealed that LILRB family members and their related genes were significantly enriched in the plasma membrane and cell surface (Fig. 9B). The GO analysis for MF showed that LILRB family members and their related genes were associated with beta-2-microglobulin binding, peptide antigen binding and MHC class I protein binding (Fig. 9C). In KEGG analysis, 27 pathways were related to the function of LILRB family members (Fig. 9D). Among them, antigen processing and presentation, cell adhesion molecules (CAMs), phagosomes, natural killer cell-mediated cytotoxicity, and leukocyte transendothelial migration were involved in antitumor immunity in liver cancer.
Based on the KEGG results and previous researches[33–35], we hypothesized that LILRB family members and their related genes are involved in the pathways of antigen processing and presentation and natural killer cell-mediated cytotoxicity (Fig. 10A and 10B).
Our previous study verified that compared with their corresponding tumor-free liver tissues (TFLs), the percentage of infiltrating CD1c+ myeloid DCs (mDCs) was significantly decreased in liver tumor tissues [7]. Considering that infiltrating NK cells in liver cancer might be involved in the signaling of the LILRB family, we further compared the percentage of NK cells in liver tumor tissues and the corresponding TFL. Human NK cells can be segregated into three major subsets: CD56brightCD16dim/−, CD56dimCD16+, CD56dimCD16dim/− NK cells. The strategy of gating these three NK cells subsets were according to the literature[36]. The results shown that the percentage of CD56brightCDdim/− NK cell in liver tumor tissues was lower than that from TFL tissues (Tumor vs TFL: 30.84 %±7.77 % vs 43.85 %±4.12 %, p<0.05) (Fig. 11). In contrast, the percentage of CD56dimCD16+ NK cell in liver tumor tissues was higher than that from TFL tissues (Tumor vs TFL: 49.98 %±8.66 % vs 29.73 %±4.78 %, p<0.05).