Landscape of immune cell gene expression is unique in predominantly WHO grade 1 skull base meningiomas when compared to convexity

Abstract

Modulation of tumor microenvironment is an emerging frontier for new therapeutics. However in meningiomas, the most frequent adult brain tumor, the correlation of microenvironment with tumor phenotype is scarcely studied. We applied a variety of systems biology approaches to bulk tumor transcriptomics to explore the immune environments of both skull base and convexity (hemispheric) meningiomas. We hypothesized that the more benign biology of skull base meningiomas parallels the relative composition and activity of immune cells that oppose tumor growth and/or survival. We firstly applied gene co-expression networks to tumor bulk transcriptomics from 107 meningiomas (derived from 3 independent studies) and found immune processes to be the sole biological mechanism correlated with anatomical location while correcting for tumour grade. We then derived tumor immune cell fractions from bulk transcriptomics data and examined the immune cell-cytokine interactions using a network-based approach. We demonstrate that oncolytic Gamma-Delta T cells dominate skull base meningiomas while mast cells and neutrophils, known to play a role in oncogenesis, show greater activity in convexity tumors. Our results are the first to suggest the importance of tumor microenvironment in meningioma biology in the context of anatomic location and immune landscape. These findings may help better inform surgical decision making and yield location-specific therapies through modulation of immune microenvironment.

Figure 1

Gene co-expression network reveals immune function to correlate strongly with meningioma location. (A) Gene dendrogram illustrating modules. The grey denotes genes which are not implicated with any modules. Labels indicate cytokines which are positively correlated with the module meta-gene expression (Pearson correlation>0.6, p < 0.05). (B) Boxplot depicting the correlation between location and meta-gene expression level of the only significant module with DAVID annotations (Mann Whitney p = 0.005). As indicated in (C), this module maps to diverse immune processes, and is therefore labeled as “immune response (universal)”. C: Gene ontology terms ranked by Bonferroni p-value.

Figure 2

Network demonstrating the connectivity of immune cell fractions and cytokine transcriptomics in convexity (A) and skull base (B) meningiomas. Cytokines are represented in blue whereas immune cells are represented in red. Edge thickness is proportional to Pearson correlation and node size is proportional to eigenvector centrality, a measure of the influence of a particular node in the network. Notably, the eigenvector centrality ranges from 0 to 0.39 (IL-6) in the convexity model and from 0 to 0.50 (CCL3) in the skull base model. Neutrophils and activated mast cells are significantly correlated with cytokines in the convexity model using our thresholding criteria, while the skull base model contains only cytokines.

Figure 3

Immune cell connectivity of meningiomas by location. A-B: Histogram of Person’s correlations for the top 3 cell fractions, ranked by connectivity, of convexity (A) and skull base (B) meningiomas. C: Difference between cell connectivity values, comparing convexity to skull base (“SB”), where positive values (blue) indicate greater connectivity in convexity while negative values (red) indicate greater connectivity in skull base meningiomas. D and E: Ranking of eigenvector centrality of each cell type for convexity (D) and skull base (E) tumours. F: Difference between eigenvector centrality, with the same conventions as (C). Note the highly connected mast cells and neutrophils in convexity meningiomas and T gamma-delta cells, monocytes, and plasma cells in skull base (“SB”) meningiomas. A = activated, M = mature, N = naïve, fh = follicular helper, gd = gamma-delta, R = resting.