Transcriptional signatures in histologic structures within glioblastoma tumors may predict personalized drug sensitivity and survival

Cymon N Kersch; Cheryl J Claunch; Prakash Ambady; Elmar Bucher; Daniel L Schwartz; Ramon F Barajas Jr; Jeffrey J Iliff; Tyler Risom; Laura Heiser; Leslie L Muldoon; James E Korkola; Joe W Gray; Edward A Neuwelt

doi:10.1093/noajnl/vdaa093

Transcriptional signatures in histologic structures within glioblastoma tumors may predict personalized drug sensitivity and survival

Neurooncol Adv. 2020 Aug 3;2(1):vdaa093. doi: 10.1093/noajnl/vdaa093. eCollection 2020 Jan-Dec.

Authors

Cymon N Kersch¹, Cheryl J Claunch^{2

3}, Prakash Ambady¹, Elmar Bucher^{2

3}, Daniel L Schwartz^{4

5}, Ramon F Barajas Jr^{3

4

6}, Jeffrey J Iliff⁷, Tyler Risom⁸, Laura Heiser^{2

3}, Leslie L Muldoon¹, James E Korkola^{2

3}, Joe W Gray^{2

3}, Edward A Neuwelt^{1

9

10}

Affiliations

¹ Department of Neurology, Blood-Brain Barrier Program, Oregon Health and Science University, Portland, Oregon, USA.
² Department of Biomedical Engineering, OHSU Center for Spatial Systems Biomedicine, Oregon Health and Science University, Portland, Oregon, USA.
³ Knight Cancer Institute, Oregon Health and Science University, Portland, Oregon, USA.
⁴ Advanced Imaging Research Center, Oregon Health and Science University, Portland, Oregon, USA.
⁵ Department of Neurology, Layton Aging and Alzheimer's Disease Center, Oregon Health and Science University, Portland, Oregon, USA.
⁶ Department of Radiology, Oregon Health and Science University, Portland, Oregon, USA.
⁷ Department of Neurology and Department of Psychiatry and Behavioral Sciences, University of Washington, Seattle, Washington, USA.
⁸ Department of Pathology, Stanford University, Stanford, California, USA.
⁹ Department of Neurosurgery, Oregon Health and Science University, Portland, Oregon, USA.
¹⁰ Office of Research and Development, Department of Veterans Affairs Medical Center, Portland, Oregon, USA.

Abstract

Background: Glioblastoma is a rapidly fatal brain cancer that exhibits extensive intra- and intertumoral heterogeneity. Improving survival will require the development of personalized treatment strategies that can stratify tumors into subtypes that differ in therapeutic vulnerability and outcomes. Glioblastoma stratification has been hampered by intratumoral heterogeneity, limiting our ability to compare tumors in a consistent manner. Here, we develop methods that mitigate the impact of intratumoral heterogeneity on transcriptomic-based patient stratification.

Methods: We accessed open-source transcriptional profiles of histological structures from 34 human glioblastomas from the Ivy Glioblastoma Atlas Project. Principal component and correlation network analyses were performed to assess sample inter-relationships. Gene set enrichment analysis was used to identify enriched biological processes and classify glioblastoma subtype. For survival models, Cox proportional hazards regression was utilized. Transcriptional profiles from 156 human glioblastomas were accessed from The Cancer Genome Atlas to externally validate the survival model.

Results: We showed that intratumoral histologic architecture influences tumor classification when assessing established subtyping and prognostic gene signatures, and that indiscriminate sampling can produce misleading results. We identified the cellular tumor as a glioblastoma structure that can be targeted for transcriptional analysis to more accurately stratify patients by subtype and prognosis. Based on expression from cellular tumor, we created an improved risk stratification gene signature.

Conclusions: Our results highlight that biomarker performance for diagnostics, prognostics, and prediction of therapeutic response can be improved by analyzing transcriptional profiles in pure cellular tumor, which is a critical step toward developing personalized treatment for glioblastoma.

Keywords: gene signature; glioblastoma; heterogeneity; transcriptomics.

Abstract

Grants and funding