3.1 Performance of CIBERSORT for TIICs evaluation in renal cancer
The TIICs composition was analyzed by CIBERSORT. As shown in Figure 1A, the samples from GEO and TCGA presented distinct proportions of 22 TIICs subsets. As the renal cancer samples in GEO were all WT patients, we speculated that the TIICs in WT samples were significantly different from those in other renal cancer samples. Then we analyzed the correlation of immune cell proportions in renal cancer samples from TCGA and GEO. It was revealed that the renal cancer samples from the two datasets had highly consistent immune cell proportions (Figure 1B), illustrating that CIBERSORT could evaluate the fraction of TIICs independently of data sources and platforms.
- Landscape of TIICs in renal cancer
The compositions of TIICs subsets in 122 paired paracancerous and renal cancer samples from the TCGA and 224 WT samples from GEO were calculated by CIBERSORT, respectively. As shown in Figure 2A-C and Table S1 the intragroup and intergroup differences in TIICs fractions were manifest. Thus, we inferred that the proportion of TIICs subsets was an inherent characteristic which varied significantly among different individuals. In addition, we analyzed the difference in TIICs composition between renal cancer and paired paracancerous samples. As shown in Figure 2D, M0 macrophages and CD8 T cells were significantly elevated, while naive B cells were significantly decreased in renal cancer samples compared with those in paired paracancerous samples.
- P-value of CIBERSORT represents the overall proportion of TIICs
It should be noted that instead of determining the actual values, CIBERSORT only calculates the relative ratios of TIICs subsets, which contributes to the dependency of results on each other. Therefore, we further analyzed the association between the P-value provided by CIBERSORT and TIICs composition. CIBERSORT P-value < 0.05 correlate with higher immune cell infiltrates, while no significantly different of the P-value ≥ 0.05 (Raza ea al. 2016; Yongfu et al. 2018). Figure 3A showed that the proportions of samples with P-value < 0.05 and P-value ≥ 0.05 in TCGA and GEO were obviously different.
It has been well proved that the mean expression values of GZMA and PRF1 represent the immune cytolytic activity and are positively associated with TIICs proportions (Constantinos et al. 2018; Narayanan et al. 2018; Gu et al. 2019; Tian et al. 2018; Gong et al. 2019; Rooney et al. 2015). Accordingly, we futher evaluated the mean expression values, which represented the immune cytolytic activity in the system. After analyzing their mean expression values, we found that the samples with P-value < 0.05 had higher immune cytolytic activity in both TCGA and GEO cohorts (p < 2.22e-16 and p = 0.043, respectively, Figure 3B and 3C). These results indicated that the proportion of TIICs in samples with P-value < 0.05 was higher in comparison to samples with P-value ≥ 0.05.
- Overall TIICs proportion, resting mast cells and activated memory CD4 T cells are associated with prognosis of renal cancer
To investigate the effects of overall TIICs proportion or 22 individual TIICs subset on renal cancer prognosis, the survival analyses of 1,010 renal cancer samples with survival information were performed. The survival curve of renal cancer samples stratified by P-value of 0.05 showed that samples with P-value < 0.05 presented inferior survival outcome in comparison to samples with P-value ≥ 0.05 (p < 0.0001, Figure 3D). It was suggested that high TIICs proportions might be associated with poor prognosis of renal cancer patients.
Subsequently, to further explore the effects of 22 individual TIICs subset on renal cancer prognosis, the univariate Cox regression analysis was conducted with 22 TIICs subsets as continuous variables. The relevant 95% confidence intervals and Hazard Ratios (HRs) were shown in Figure 4A. It was found that activated memory CD4 T cells and resting mast cells were significantly associated with the prognosis of renal cancer patients (HR = 9.4e + 05, p = 0.004 and HR = 3.7e – 03, p = 0.034, respectively). The survival curves showed a low proportion of resting mast cells and a high proportion of activated memory CD4 T cells were related to poor prognosis of renal cancer patients (both p < 0.0001).
- Different immune clusters are associated with the prognosis of renal cancer
The above results indicated TIICs alteration might affect the prognosis of renal cancer patients. Therefore, we speculated whether different immune clusters could be identified with the TIICs data. Firstly, the optimal number of clusters was determined as 3 using the within cluster sum of square errors (WSS) method (Figure S1). Subsequently, the hierarchical clustering of the samples was conducted by the Euclidean distance model (Figure 5A). After analyzed the association of different immune patterns with prognosis, cluster 1 exhibited superior survival outcomes, while cluster 2 exhibited inferior survival outcomes (p = 0.0057, Figure 5B). Moreover, the compositions of 22 TIICs subsets is significantly different among the cluster 1, cluster 2, and cluster 3 (Figure 5C). In additional, the immunes cell type abundances differ between clusters was revealed in violin plot Figure S2 and clinical differences of the three clusters was showed in the Figure S3.