The flow chart of our research is shown in Figure 1. There were a total of 521 samples in our present study, including 480 colon cancer samples and 41 normal samples (both from the TCGA database). 2843 immune genes were downloaded from the ImmPort database (Supplementary data 1).
DEirlncRNAs in Colon Cancer Database Samples
A total of 1,342 immune-related gene mRNA expression matrices (Supplementary data 2), and 694 immune-related lncRNA expression matrices (Supplementary data 3) were extracted from colon cancer samples. Immune-related lncRNA expression differentiation analysis found that the expression of 71 immune-related lncRNAs had significant differences (|logFC|>1, adj.P<0.05). The DEirlncRNAs in colon cancer are illustrated in the corresponding heat map (Figure 2A) and volcano plot (Figure 2B).
Establishment and Validation of the RiskScore Model
After filtering the DEirlncRNA 1-or-0 matrix, we finally identified 1550 pairs of lncRNA. The univariate Cox regression and LASSO analysis identified 8 lncRNA pairs to construct the risk model (Figures 3A and 3B), where hazard ratios (HRs) of AC103740.1|LEF1−AS1, LINC02391|AC053503.5 and WWC2−AS2|AL355916.2 were less than 1, while hazard ratios (HRs) of AC104090.1|NEURL1−AS1, AC099524.1|AL161908.1, AC074011.1|AL078601.2, AL355916.2|LINC01723, and AP003392.4|LINC00598 were greater than 1 (Figure 3C). The Cox multivariate regression analysis revealed that AC103740.1|LEF1−AS1 (P=0.002; HR, 95%CI = 0.490 [0.311-0.730]), AC104090.1|NEURL1−AS1 (P=0.002; HR, 95%CI = 2.010[1.279−3.159]), and AC074011.1|AL078601.2 (P=0.010; HR, 95%CI = 1.797[1.151−2.808]) were considered to be independent significant prognostic factors for colon cancer (Figure 3D). The predicted AUC values were 0.776 at 1 year, 0.703 at 2 years, and 0.686 at 3 years, respectively (Figure 4A), and the highest AUC value was 0.776 at 1 year (Figure 4B). The ROC analysis revealed that the riskscore model with 0.776 of AUC value was better fitted than the clinical risk models including Age with 0.560 of AUC value, Gender with 0.474 of AUC value and Stage with 0.727 of AUC value (Figure 4C). Figure 4C showed that the clinical stage also had a higher AUC value, which indicated that the clinical stage might be an important prognostic factor. AIC analysis revealed that the cutoff value of the risk score was 1.283 (Figure 4D). Based on this cutoff value, patients with the risk score higher than 1.283 were included into the high-risk group, and patients with the risk score lower than 1.283 were included into the low-risk group.
Survival analysis of the model and the correlations between the risk score and clinical characteristics
The risk score and survival status of the scatter plot shown that patients in the low-risk score group have longer survival time (Figures 5A and 5B). Kaplan–Meier survival curve analysis showed that patients in the low-risk group had better survival status than those in the high-risk group (Figure 5C). The analysis of the correlations between the risk score and clinical characteristics (Figure 6) showed that the risk score was significantly correlated with the N stage (Figure 6E), which implied that patients in the advanced stage were more likely to be at risk than those in the early stage. The results of univariate (Figure 6G) and multivariate (Figure 6H) Cox regression analysis showed that Stage (P<0.001, HR=2.547,95%CI=[1.966-3.299]) and riskScore (P<0.001, HR=1.820, 95%CI=[1.542-2.147]) were independent prognostic factors.
In our present study, we used the CIBERSORT algorithm to calculate the infiltration of immune cells of colon cancer in the high- and low-risk groups. In order to reveal the differences of the tumor microenvironment (TME) in colon cancer between the high- and low-risk groups, ssGSEA was used to analyze the level of immune cell infiltration and immune cell function in each group, which found that there were statistical differences in the immune cells infiltration rate of B cells memory and Neutrophils between the high- and low-risk groups, and the immune infiltration rate of the low-risk group was higher than that of the high-risk group (Figure 7A). Immune-related biological processes were significantly different between the high- and low-risk groups except for Th2_cells and TIL (Figure 7B). The survival analysis of immune cell infiltration subtypes showed there was significant survival difference between the high- and low-risk groups for B cells navie (P=0.022; Figure 7C), Macrophages M0 (P=0.012; Figure 7D), NK cells resting (P=0.034; Figure 7E), Plasma cells (P=0.021; Figure 7F), and T cells CD8 (P=0.018; Figure 7G). The survival analysis of immune-related biological processes showed there was significant survival difference between the high- and low-risk groups for APC_co_inhibition (P=0.002; Figure 7H), B_cells (P=0.031; Figure 7I), CCR (P=0.014; Figure 7J), Check-point (P=0.025; Figure 7K), DCs (P=0.010; Figure 7L), HLA (P=0.031; Figure 7M), iDCs (P=0.009; Figure 7N), Mast_cells(P<0.001; Figure 7O), MHC_class_I(P=0.037; Figure 7P), Neutrophils (P=0.023; Figure 7Q), Parainflammation(P=0.009; Figure 7R), pDCs (P=0.002; Figure 7S), T_cell_co-inhibition (P=0.023; Figure 7T), T_helper_cells (P=0.023; Figure 7U), Tfh (P=0.032, Figure 7V), Th1_cells (P=0.006; Figure 7W), and TIL(P=0.009; Figure 7X). There was a significantly negative correlation between the immune cell infiltration of neutrophils and the risk score (R=-0.39, p=0.0097, Figure 8A), while the immune cell infiltration of T cells CD8 was positively correlated with the risk score (R=0.35, p=0.021; Figure 8B) . The high expression of immune checkpoint inhibitor related genes PDCD1 (Figure 8C) and LAG3(Figure 8D )were significantly related to the high risk score, while there was no statistically significant association between the expression level of CTAL-4 and the risk score (Figure 8E).
Response to Drug Sensitivity
Our present study assessed the correlations between IC50 level of 8 drugs including Carboplatin, Cisplaein, Dasatinib, Erlotinib, Gefitinib, Imatinib, Oxaliplatin and AZD6244 and the risk score. The higher IC50 of Carboplatin, Cisplaein, Gefitinib and Imatinib was significant correlated with the high risk score (Figures 8F-8I). The lower IC50 of Dasatinib, Erlotinib,Oxaliplatin and AZD6244 was significant correlated with the high risk score (Figures 8J-8M).