CD96 was enriched in glioblastoma, IDH-wildtype, and mesenchymal glioma
To clarify differences of CD96 expression pattern in four grades of glioma malignancy, the mRNA level of CD96 was examined in CGGA and TCGA databases separately. The results revealed that glioblastoma (WHO IV) showed a higher CD96 expression in CGGA (Fig. 1A) and TCGA (Fig. 1B) cohorts than WHO grade II and grade III glioma. Compared to other pathological subsets (namely oligodendroglioma, oligoastrocytoma, and astrocytoma), CD96 expression was up-regulated relatively in glioblastoma (Figs. 1C and 1D). Isocitrate dehydrogenase (IDH) mutation distributes in almost 40% glioma which has an outsize impact on glioma development and progression. To this end, the expression profile of CD96 in different IDH states was also explored. We found that CD96 was consistently enriched in IDH-wildtype (IDHWT) in CGGA as well as in TCGA database (Figs. 1E and 1F). This result suggested that CD96 expression was more prevalent without IDH mutation (IDHWT) than with IDH mutation (IDHMUT) in glioma. In order to investigate the expression pattern of CD96, the distribution of CD96 in different molecular subtypes was analyzed. Compared with the other three subtypes (classical, neural, and proneural), CD96 was highly enriched in mesenchymal subtype in both CGGA and TCGA cohorts (Figs. 2A and 2B). To further validate our findings, we carried out ROC curves analysis of CD96 expression and mesenchymal subtype in all grade glioma. Intriguingly, we observed that area under the curve (AUC) was up to 78.6.% and 92.8% in CGGA and TCGA datasets separately (Figs. 2C and 2D). These findings indicated that CD96 acted as a potential biomarker for mesenchymal subtype glioma.
CD96 was significantly associated with immune functions in glioma
Considering CD96 expression was closely tied to malignancy, we inferred that CD96 became an integral part of glioma progression and then performed GO analysis to unfold its role. 54 and 176 genes positively correlated with CD96 were listed by Pearson correlation analysis (Pearson |R| > 0.6) in CGGA and TCGA datasets. According to GO enrichment analysis, we found that genes most relevant to CD96 were more involved in T cell activation and regulation of lymphocyte proliferation in CGGA and TCGA databases, respectively (Figs. 3A and B). To further elucidate the immune function of CD96 in glioma, 1,540 genes from the AmiGO 2 website (http://amigo. geneontology.org/amigo) were reported to be associated with the immune response and we selected 362 and 354 genes most relevant to CD96 (Pearson |R|> 0.3) in CCGA and TCGA cohorts (Table S1) for heatmap analysis. Thereinto, 357 genes were strongly positively correlated with CD96 expression, while 5 genes had a significantly negative relationship with CD96 in CGGA dataset. While 347 and 7 genes displayed directly and inversely proportional in TCGA dataset (Fig. 4). To sum up, CD96 was directly correlated with most immune responses and negatively correlated with few immune responses in glioma.
The correlation between CD96 and T cell mediated immunity in glioma
To fully understand the relationship between CD96 and T cell immune in glioma, we performed gene set variation analysis (GSVA) to assess differential activities of pathways between sets of genes. As delineated in Figs. 5A and 5B, these relationships were similar in both CGGA and TCGA databases. Specifically, CD96 showed a positive correlation with T-helper 1/2 type immune response (GO:0042088 and GO:0042092), T-helper 1/2 cell cytokine production (GO:2000556 and GO:2000553), and natural killer cell mediated cytotoxicity directed against tumor cell target (GO:0002860). Conversely, CD96 was correlated negatively with T cell mediated immune response to tumor cell (GO:0002842) and T cell mediated cytotoxicity directed against tumor cell target (GO:0002852). This result re-validated that the special immune function of CD96 is to act an inhibitory role in T cell immune to tumor cells in glioma.
Correlation Between CD96 And Other Immune Checkpoints
As therapeutic targets, a growing number of immune checkpoints have been emerged and examined in clinical trials or clinical situations [25, 26]. Thereby, we analyzed the relationship between CD96 and other immune checkpoints, such as PD-1, CTLA-4, TIGIT, TIM-3, NR2F6, and GITR. Pearson correlation analysis revealed that CD96 was tightly associated with PD-1, CTLA-4, TIGIT, and TIM-3. Co-expression of PD-1 with CD96 was consistent with the previous research . These results were validated in CGGA and TCGA datasets mutually (Figs. 5C and 5D), implying the possible synergistic effects of CD96 with these checkpoint members. Accordingly, we postulated CD96 may contribute significantly to the inflammatory response in glioma and used the previous method to test. As shown in Figure S1, CD96 was positively associated with HCK, MHC-I, MHC-II, STAT1, STAT2, and FCGR2A, especially with LCK. This finding additionally evidenced the vital immune function of CD96 in glioma.
Correlation between CD96 and immune infiltration level in GBM and LGG
Tumor-infiltrating lymphocytes are an independent predictor of sentinel lymph node status and survival in cancers. Hence, we investigated whether CD96 expression was connected with immune infiltration levels in GBM (glioblastoma multiforme) and LGG (Low Grade Glioma) by TIMER website tools. Our findings manifested that CD96 expression related positively with infiltrating levels of dendritic cell (r = 0.504), neutrophil (r = 0.487), macrophage (r = 0.431), and CD8 + T cell (r = 0.406) in LGG. Simultaneously, CD96 had a marginal positively association with B cell (r = 0.329) and CD4 + T cell (r = 0.37) infiltration level in LGG. On the other hand, CD96 expression has no significant relationships with tumor purity (r = -0.117) infiltrating levels of B cell (r = 0.101), CD8 + T cell (r = -0.157), CD4 + T cell (r = -0.1), and so on in GBM (Fig. 6). These differences implied that CD96 would make more contribution to immune infiltration in LGG than GBM, especially in dendritic cells.
CD96 predicted worse survival in glioma
Owing to the high relevance between CD96 and immune suppressor in glioma, the prognostic impact of CD96 was verified via Kaplan–Meier method. Overall survival analysis in glioma and GBM patients demonstrated that high expression of CD96 predicted relatively poor survival in CGGA and in TCGA cohorts (Fig. 7). Eventually, we evaluated the independence of the clinicopathological significance of CD96 in glioma by univariate and multivariate Cox regression analyses. The results indicated that CD96, age, gender, WHO grade, IDH mutation, and h1p19q codeletion status were closely associated with overall survival and revealed CD96 is an independent prognosticator for glioma patients (Table S2).