A 16-gene signature distinguishes anaplastic astrocytoma from glioblastoma

Abstract

Anaplastic astrocytoma (AA; Grade III) and glioblastoma (GBM; Grade IV) are diffusely infiltrating tumors and are called malignant astrocytomas. The treatment regimen and prognosis are distinctly different between anaplastic astrocytoma and glioblastoma patients. Although histopathology based current grading system is well accepted and largely reproducible, intratumoral histologic variations often lead to difficulties in classification of malignant astrocytoma samples. In order to obtain a more robust molecular classifier, we analysed RT-qPCR expression data of 175 differentially regulated genes across astrocytoma using Prediction Analysis of Microarrays (PAM) and found the most discriminatory 16-gene expression signature for the classification of anaplastic astrocytoma and glioblastoma. The 16-gene signature obtained in the training set was validated in the test set with diagnostic accuracy of 89%. Additionally, validation of the 16-gene signature in multiple independent cohorts revealed that the signature predicted anaplastic astrocytoma and glioblastoma samples with accuracy rates of 99%, 88%, and 92% in TCGA, GSE1993 and GSE4422 datasets, respectively. The protein-protein interaction network and pathway analysis suggested that the 16-genes of the signature identified epithelial-mesenchymal transition (EMT) pathway as the most differentially regulated pathway in glioblastoma compared to anaplastic astrocytoma. In addition to identifying 16 gene classification signature, we also demonstrated that genes involved in epithelial-mesenchymal transition may play an important role in distinguishing glioblastoma from anaplastic astrocytoma.

Figure 1. The schematic representation of the work flow of statistical analysis.

The expression of 175 genes was subjected to prediction analysis of microarray (PAM) in the training set. PAM analysis identified the 16 discriminatory genes between AA and GBM which was further validated by principal component analysis (PCA) and cross validated probability by PAM. The 16-gene signature was further validated in test set and three independent cohorts of patient samples (GSE1993, GSE4422, TCGA and GSE4271).

Figure 2. Identification of 16-gene signature in training set.

A. Plot showing classification error for the 175 input genes from PAM analysis in the training set. The threshold value of 2.5 corresponded to 16 genes which classified AA (n = 30) and GBM (n = 78) samples with classification error of 0.12. B. Heat map of one-way hierarchical clustering of 16 PAM-identified genes in AA and GBM patient samples of the training set. A dual-color code was used, with red and green indicating up- and down regulation, respectively.

Figure 3. PCA and cross validated probabilities of AA and GBM samples of training set.

A. PCA was performed using expression values of 16- PAM identified genes between AA (n = 30) and GBM (n = 78) samples in training set. A scatter plot is generated using the first two principal components for each sample. The color code of the samples is as indicated. B. The detailed probabilities of 10-fold cross-validation for the samples of training set based on the expression values of 16 genes are shown. For each sample, its probability as AA (orange color) and GBM (blue color) are shown and it was predicted by the PAM program as either AA or GBM based on which grade's probability is higher. The original histological grade of the samples is shown on the top.

Figure 4. PCA and cross validated probabilities of AA and GBM samples of test set.

A. PCA was performed using expression values of 16- PAM identified genes between AA (n = 20) and GBM (n = 54) samples in test set. A scatter plot is generated using the first two principal components for each sample. The color of the samples is as indicated. B. The detailed probabilities of 10-fold cross-validation for the samples of test set based on the expression values of 16 genes are shown. For each sample, its probability as AA (orange color) and GBM (blue color) are shown and it was predicted by the PAM program as either AA or GBM based on which grade's probability is higher. The original histological grade of the samples is shown on the top.

Figure 5. PCA and cross validated probabilities of AA and GBM samples of TCGA dataset.

A. PCA was performed using expression values of 16- PAM identified genes between grade III glioma (n = 27) and GBM (n = 152) samples in TCGA dataset. A scatter plot is generated using the first two principal components for each sample. The color code of the samples is as indicated. B. The detailed probabilities of 10-fold cross-validation for the samples of TCGA dataset based on the expression values of 16 genes are shown. For each sample, its probability as grade III glioma (orange color) and GBM (blue color) are shown and it was predicted by the PAM program as either grade III glioma or GBM based on which grade's probability is higher. The original histological grade of the samples is shown on the top.

Figure 6. Discordant samples do not exhibit clinical features and molecular markers of its histologic grade.

A. The average Age at Diagnosis along with standard deviation is plotted for Authentic AAs (n = 21), Authentic GBMs (n = 37), Discordant AAs (n = 8) and Discordant GBMs (n = 20). *** denotes that P<0.001, ** denotes that P<0.01 and * denotes that P<0.05. B. The Kaplan Meier survival analysis of Authentic AAs (n = 13), Authentic GBMs (n = 165) and Discordant GBMs (n = 13). The median survival was significantly different across the groups with P<0.001. C. The percentage of samples showing CDKN2A/2B loss, EGFR amplification and p53 mutation is plotted for Authentic AAs (n = 14), Authentic GBMs (n = 37), Discordant AAs (n = 5) and Discordant GBMs (n = 2).

Figure 7. KEGG pathway analysis showed the upregulation of Focal adhesion pathway in GBM.

67 differentially expressed genes between AA and GBM from GSE1993 was subjected to network analysis in Pathway Express which identified the focal adhesion as the significantly differentially regulated pathway between AA and GBM. The input genes present in the network are represented in red and blue indicating the upregulation or downregulation respectively in GBM as compared to AA.