Multiparametric magnetic resonance imaging discerns glioblastoma immune microenvironmental heterogeneity

Cymon N Kersch,Leslie L Muldoon,Cheryl J Claunch,Rongwei Fu,Daniel Schwartz,Soonmee Cha,Jay Starkey,Edward A Neuwelt,Ramon F Barajas Jr

DOI: https://doi.org/10.1177/19714009231163560

2023-06-12

Abstract:Rationale and objective: Poor clinical outcomes for patients with glioblastoma are in part due to dysfunction of the tumor-immune microenvironment. An imaging approach able to characterize immune microenvironmental signatures could provide a framework for biologically based patient stratification and response assessment. We hypothesized spatially distinct gene expression networks can be distinguished by multiparametric Magnetic Resonance Imaging (MRI) phenotypes. Materials and methods: Patients with newly diagnosed glioblastoma underwent image-guided tissue sampling allowing for co-registration of MRI metrics with gene expression profiles. MRI phenotypes based on gadolinium contrast enhancing lesion (CEL) and non-enhancing lesion (NCEL) regions were subdivided based on imaging parameters (relative cerebral blood volume (rCBV) and apparent diffusion coefficient (ADC)). Gene set enrichment analysis and immune cell type abundance was estimated using CIBERSORT methodology. Significance thresholds were set at a p-value cutoff 0.005 and an FDR q-value cutoff of 0.1. Results: Thirteen patients (eight men, five women, mean age 58 ± 11 years) provided 30 tissue samples (16 CEL and 14 NCEL). Six non-neoplastic gliosis samples differentiated astrocyte repair from tumor associated gene expression. MRI phenotypes displayed extensive transcriptional variance reflecting biological networks, including multiple immune pathways. CEL regions demonstrated higher immunologic signature expression than NCEL, while NCEL regions demonstrated stronger immune signature expression levels than gliotic non-tumor brain. Incorporation of rCBV and ADC metrics identified sample clusters with differing immune microenvironmental signatures. Conclusion: Taken together, our study demonstrates that MRI phenotypes provide an approach for non-invasively characterizing tumoral and immune microenvironmental glioblastoma gene expression networks.