Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2012 May;123(5):727-38.
doi: 10.1007/s00401-012-0941-4. Epub 2012 Feb 10.

A Prognostic Gene Expression Signature in Infratentorial Ependymoma

Affiliations
Free PMC article

A Prognostic Gene Expression Signature in Infratentorial Ependymoma

Khalida Wani et al. Acta Neuropathol. .
Free PMC article

Abstract

Patients with ependymoma exhibit a wide range of clinical outcomes that are currently unexplained by clinical or histological factors. Little is known regarding molecular biomarkers that could predict clinical behavior. Since recent data suggest that these tumors display biological characteristics according to their location (cerebral vs. infratentorial vs. spinal cord), rather than explore a broad spectrum of ependymoma, we focused on molecular alterations in ependymomas arising in the infratentorial compartment. Unsupervised clustering of available gene expression microarray data revealed two major subgroups of infratentorial ependymoma. Group 1 tumors over expressed genes that were associated with mesenchyme, Group 2 tumors showed no distinct gene ontologies. To assess the prognostic significance of these gene expression subgroups, real-time reverse transcriptase polymerase chain reaction assays were performed on genes defining the subgroups in a training set. This resulted in a 10-gene prognostic signature. Multivariate analysis showed that the 10-gene signature was an independent predictor of recurrence-free survival after adjusting for clinical factors. Evaluation of an external dataset describing subgroups of infratentorial ependymomas showed concordance of subgroup definition, including validation of the mesenchymal subclass. Importantly, the 10-gene signature was validated as a predictor of recurrence-free survival in this dataset. Taken together, the results indicate a link between clinical outcome and biologically identified subsets of infratentorial ependymoma and offer the potential for prognostic testing to estimate clinical aggressiveness in these tumors.

Figures

Fig. 1
Fig. 1
Expression profiling reveals two major subgroups of gene expression in infratentorial ependymomas. a) Unsupervised clustering of 67 infratentorial ependymomas using 250 highly variable probe sets reveals two major sample clusters. Age: patients <10 years (orange), >10 years (blue) Grade: II (red), III (green) Unknown (white). b) Heat map of the top 40 genes differentially expressed between group 1 and group 2.
Fig. 2
Fig. 2
Distribution of age, grade and gender among the microarray defined subgroups. Group 1 ependymomas were predominantly composed of younger patients.
Fig. 3
Fig. 3
Kaplan-Meier analysis of survival association with 10-gene signature. Formalin-fixed, paraffin embedded ependymoma samples were subjected to QRT-PCR as described in Materials and Methods. The samples were ranked by metagene score and dichotomized into 2 groups using the median metagene score as the cutoff. Survival is shown for the lower metagene scores (blue) vs. the higher metagene score (red). a) Recurrence-free survival according to the 10-gene set. The median recurrence-free survival for the metagene-defined unfavorable group was 80 weeks while for the favorable group it was not reached. b) Overall survival according to the 10-gene set. The median overall survival for the unfavorable group was 345 weeks, and was not reached for the favorable group. The log rank test was used to determine statistical significance.
Fig. 4
Fig. 4
The unfavorable 10-gene signature is significantly associated with younger age at diagnosis. Statistical significance of age was determined using the Fisher’s exact test, two tailed.
Fig. 5
Fig. 5
Subgroups A and B described in the Witt et al report are similar to transcriptomal subgroups 1 and 2. Gene expression microarray data from the Toronto and Heidelberg datasets was used for clustering using the gene list described in Fig. 2B. Heat map of the top 40 genes differentially expressed between Group 1 and Group 2 in the a) Heidelberg dataset and b) the Toronto dataset. 38 of the 40 most differentially expressed genes in Groups 1 and 2 were evaluable in the Heidelberg data set while 38 of the 40 differential genes were present in the Toronto dataset. Color coding is as follows, Group A tumors: orange; Group B tumors: blue.
Fig. 6
Fig. 6
Expression of LAMA2 and NELL2 in the transcriptomal subgroups. LAMA2 was over expressed in Group 1 ependymomas while NELL2 expression was significantly higher in Group 2 ependymomas. The Student’s t-test was used to determine statistical significance.
Fig. 7
Fig. 7
Kaplan-Meier analysis of survival association with 10-gene signature in the Heidelberg dataset. Gene expression data from the Heidelberg dataset was used to calculate a metagene score for the 10-gene signature as described in materials and methods. The samples were ranked by metagene score and dichotomized into 2 groups using the median metagene score as the cutoff. Survival is shown for the higher metagene scores (blue) vs. the lower metagene score (red). a) Recurrence-free survival according to the 10-gene set. The median recurrence-free survival for the metagene-defined unfavorable group was 110 weeks while for the favorable group it was not reached. b) Overall survival according to the 10-gene set. The log rank test was used to determine statistical significance

Similar articles

See all similar articles

Cited by 48 articles

See all "Cited by" articles

Publication types

MeSH terms

Feedback