Coexpression analysis of CD133 and CD44 identifies proneural and mesenchymal subtypes of glioblastoma multiforme

Abstract

Accumulating evidence suggests that the stem cell markers CD133 and CD44 indicate molecular subtype in Glioblastoma Multiforme (GBM). Gene coexpression analysis of The Cancer Genome Atlas GBM dataset was undertaken to compare markers of the Glioblastoma Stem-Progenitor Cell (GSPC) phenotype. Pearson correlation identified genes coexpressed with stem cell markers, which were then used to build a gene signature that classifies patients based on a CD133 coexpression module signature (CD133-M) or CD44-M subtype. CD133-M tumors were enriched for the Proneural (PN) GBM subtype compared to Mesenchymal (MES) subtype for CD44-M tumors. Gene set enrichment identified DNA replication/cell cycle genes in the CD133-M and invasion/migration in CD44-M, while functional experiments showed enhanced cellular growth in CD133 expressing cells and enhanced invasion in cells expressing CD44. As with the 4 major molecular subtypes of GBM, there was no long-term survival difference between CD44-M and CD133-M patients, although CD44-M patients responded better to temozolomide while CD133-M patients benefited from radiotherapy. The use of a targeted coexpression approach to predict functional properties of surface marker expressing cells is novel, and in the context of GBM, supports accumulating evidence that CD133 and CD44 protein marker expression correlates with molecular subtype.

Keywords: cancer genome atlas; cancer stem cells; coexpression; glioblastoma; molecular subtype.

Figure 1. Cancer stem cell marker coexpression modules are associated with molecular subtype

(A) The coexpression module signatures of a range of stem cell markers was used to assign a patient specific similarity score. This score was then related to the assigned molecular subtype (top horizontal bar). (B) Enrichment of PN subtype in CD133-M and enrichment of CD44-M in MES. All patients in the 2013 data freeze were assigned to either a CD133 or CD44 coexpression module subtype (CD133-M or CD44-M) and the percentage of each molecular subtype within each coexpression subtype calculated. Coexpression module signature analysis of the Rembrandt (C) and (D) and the TCGA Agilent microarray dataset.

Figure 2. The coexpression modules identified in GBM tumors are reflected in GSPCs

(A) Coexpression module analysis of Bhat el al 2013 gene expression data. Hierarchical clustering of GSPC cultures by CD133-M and CD44-M partitions samples into 2 groups, largely overlapping with the assignment of PN (purple horizontal bar) and MES (red horizontal bar). (B) Coexpression module analysis of Gunther et al 2008 reveals 2 groups by clustering that largely overlap with the suspension or semi-adherent growth pattern described in the study. (C) Coexpression module analysis of Shats et al shows that CD133+ (blue) sorted cells are assigned to the CD133 coexpression signature and CD133− cells (red) are assigned to the CD44 coexpression signature. (D) Coexpression module analysis of Rath et al shows both normal neural and GBM cells that are CD133+ (blue) are enriched for CD133-M and CD133− cells are enriched for CD44-M (red).

Figure 3. Enrichment of proliferative pathways for CD133-M and enrichment of invasive pathways for CD44-M and CD15-M

The Pearson correlation score of all genes with CD133, CD44 and CD15 mRNA was used as input to pre-ranked GSEA. (A) Enrichment of proliferative pathways in CD133 correlated genes. (B) and (C) Enrichment of invasive pathways for CD44 and CD15 correlated genes.

Figure 4. Functional validation of the coexpression module and pathway analysis

(A) FACS analysis of the expression of surface protein markers using a panel of GSPCs. Cells were labeled with CD15, CD44 and CD133 antibodies. Expression of CD44 and CD133 is represented in the center panel. The expression of CD15 is shown separately for CD44+ cells (top) and CD133+ cells (bottom). (B) Repopulation ability is highest for CD44/CD133 double positive GSPCs. Limiting dilutions of cells ranging from 50 to 1 cell per well were sorted into plates and scored for sphere formation after 7 days. Sphere forming efficiency was estimated by ELDA. Error bars represent sem of three independent GSPC lines. (C) CD44 expressing GSPCs are a more invasive subpopulation. The surface area of the sorted GSPCs was quantified 7 days after the administration of extracellular matrix. Mean invasion is represented relative to the CD44−/CD133− subpopulation. Error bars represent sem of five independent GSPC lines. (D) Temozolomide sensitivity of sorted GSPCs was measured 7 days after sorting. 50μM of temozolomide was applied daily. Mean temozolomide sensitivity is presented relative to the CD44−/CD133− subpopulation. Error bars represent sem of five independent GSPC lines.

Figure 5. Differential response to therapy for CD44-M and CD133-M classified patients in the TCGA Agilent cohort

(A) Survival advantage for CD133-M classified patients in the TCGA RNAseq cohort. (B) Survival analysis of the TCGA Agilent microarray GBM dataset. CD133-M patients receive greater benefit from radiotherapy. CD133-M classified patients in the Agilent cohort were stratified on the basis of receiving temozolomide (C) or radiotherapy (D). All other covariates are set to the mean of the dataset. CD44-M patients receive greater benefit from temozolomide. CD44-M classified patients were stratified on the basis of receiving temozolomide (E) or radiotherapy (F).