Development and validation of an interferon signature predicting prognosis and treatment response for glioblastoma

Abstract

Background: Interferon treatment, as an important approach of anti-tumor immunotherapy, has been implemented in multiple clinical trials of glioma. However, only a small number of gliomas benefit from it. Therefore, it is necessary to investigate the clinical role of interferons and to establish robust biomarkers to facilitate its application. Materials and methods: This study reviewed 1,241 glioblastoma (GBM) and 1,068 lower grade glioma (LGG) patients from six glioma cohorts. The transcription matrix and clinical information were analyzed using R software, GraphPad Prism 7 and Medcalc, etc. Immunohistochemical (IHC) staining were performed for validation in protein level. Results: Interferon signaling was significantly enhanced in GBM. An interferon signature was developed based on five interferon genes with prognostic significance, which could reflect various interferon statuses. Survival analysis showed the signature could serve as an unfavorable prognostic factor independently. We also established a nomogram model integrating the risk signature into traditional prognostic factors, which increased the validity of survival prediction. Moreover, high-risk group conferred resistance to chemotherapy and high IFNB1 expression levels. Functional analysis showed that the high-risk group was associated with overloaded immune response. Microenvironment analysis and IHC staining found that high-risk group occupied a disorganized microenvironment which was characterized by an enrichment of M0 macrophages and neutrophils, but less infiltration of activated nature killing (NK) cells and M1 type macrophages. Conclusion: This interferon signature was an independent indicator for unfavorable prognosis and showed great potential for screening out patients who will benefit from chemotherapy and interferon treatment.

Keywords: Glioblastoma; immune response; interferon; microenvironment; prognosis.

Figure 1.

The distribution pattern of the interferon signature in the CGGA microarray cohort. A, The expression pattern of the signature’s five genes with other clinical characteristics. B, The risk score was elevated significantly in higher grade gliomas. C, The higher risk patients were specifically enriched in the IDH1 wild type and mesenchymal subtype in GBM. (* means P < 0.05, ** means P < 0.01, *** means P < 0.001, **** means P < .0001).

Figure 2.

Prognostic value of the interferon signature in GBM. A, The high-risk group exhibited an unfavorable prognosis in GBM of CGGA microarray cohort. In the other five validation cohorts, there was also a significantly shorter survival times in the high-risk group compared with those patients in the low-risk group (B–F). High-risk group patients occupied a significantly reduced overall survival compared with low-risk group patients when stratified by age (G and H), KPS (I and J), IDH1 mutation status (K and L) and MGMT promoter status (M and N) in CGGA microarray cohort.

Figure 3.

The risk signature mediated chemotherapy resistance in GBM of CGGA microarray cohort. A, For low-risk group, the patients with radio-chemotherapy survived significantly longer than those patients who received radiation alone. B, High-risk group patients did not benefit well from adjuvant chemotherapy. For all GBM patients who received chemotherapy, only the MGMT promoter methylated patients with a low-risk score had a survival advantage over the unmethylated ones; survival time of MGMT promoter-methylated patients with a high-risk score was similar to that of unmethylated patients (C and D). In GBMs with methylated MGMT promoters, only patients with a low-risk score benefited from chemotherapy (E and F). （* means P < 0.05, ** means P < 0.01, *** means P < 0.001, **** means P < 0.0001）.

Figure 4.

The low and high-risk groups of CGGA microarray cohort exhibited different immune status and biological process. A, GO analysis based on up-regulated genes correlated with the high risk group. B, GSVA results showed that immune relevant processes and some oncogenic pathways were most significantly enriched in the high risk group. GSEA results showed that there was a positive enrichment of two immune related terms in the high risk group (C and D). PCA results exhibited that the low risk and high risk groups were generally distributed in different directions based on immune related genes (E).

Figure 5.

The relationship between the interferon signature with the glioma microenvironment. A, There was a significantly negative correlation between the tumor purity and the risk score. The high-risk group occupied a higher immune score and stromal score (B and C). MCP analysis showed there was an extreme enrichment of immune and stromal cell in the high-risk group (D). CIBERSORT results showed that the high-risk group was associated with more macrophages and neutrophils; while activated NK cells and M1 type macrophages were enriched in low-risk group (E). Positive correlations exist between risk score and M0 macrophages as well as neutrophils (F and G), while the risk score showed a negative correlation with M1 macrophages and activated NK cells (H and I). J, IHC staining confirmed the CIBERSORT analysis results. （A-I were performed in CGGA microarray cohort; J was conducted in the First Hospital of China Medical University; * means P < 0.05, ** means P < 0.01, *** means P < 0.001, **** means P < 0.0001）.

Figure 6.

Survival analyses stratified by interferon showed heterogeneity between low and high-risk group of CGGA microarray cohort. There were no significant survival differences in the low or high risk groups when stratified by IFNG (A and B) or IFNA2(C and D). E, In the low-risk group, the high IFNB1 group showed a benefit survival compared with low IFNB1 expression patients. F, In the high-risk group, there was a shorter survival of high IFNB1 expressing patients.