By Jiuyi Li et al.  

doi: 10.1158/1078-0432.CCR-19-0299

Glioma is one of the most-deadly types of human malignancies. Despite intensive research over the past decades, effective therapies are still lacking. The current diagnosis guideline is limited in its capacity to assign gliomas into molecular subtypes with distinct pathogenic mechanisms, thereby generating artificial heterogeneities and complexities, hindering the elucidation of pathogenic mechanisms and the development of subtype-specific treatment.

In 2014, the team led by Professor Xiaolong Fan has reported an EM/PM classification scheme for glioma (PNAS 111:3538-3543, 2014). The EM and PM genes are consistently co-expressed with EGFR or PDGFRA, they are involved in the neural stem cell differentiation to early astrocytes and the differentiation of oligodendrocyte progenitor cells, respectively. The co-expression of the EM and PM genes are evolutionally conserved during the brain development, and subsets of the EM and PM genes are frequently altered in glioma genome. Based on the reciprocal expression pattern between the EM and PM gene module, adult gliomas can be defined as the EM subtype (with high EM expression but low PM expression), and the PM subtype (with high PM expression but low EM expression). The EM/PM subtyping is independent of morphological subtyping, as both the EM and PM subtypes contain all different morphological subtypes. These EM and PM subtypes are distinct in their prognosis, resemblance to brain development cell lineage, and pattern of genomic alterations. 

To translate the EM/PM classification scheme into a diagnostic tool, the team led by professor Fan has designed a TaqMan low-density RT-PCR array and trained a support vector machine-based prediction model, and validated the efficacy of the array and the prediction model in about 200 individual glioma samples from Beijing Neurosurgical Institute/Beijing Tiantan Hospital and Gothenburg University in Sweden. Complemented with low-coverage whole genome sequencing and detection of IDH mutations, adult gliomas are individually diagnosed as being of the EM or PM subtypes, resembling distinct neural lineages, and carrying a relatively quiescent or actively evolving genome. Among many transcriptome-based molecular classification schemes proposed for glioma, this diagnostic platform is the first in translating the principle of big data analysis into a technique platform of individual diagnosis. An application of this diagnostic platform may improve risk stratification and treatment decision for glioma.