Data from: Spatiotemporal modeling reveals high-resolution invasion states in glioblastoma
Data files
Dec 27, 2023 version files 1.40 GB
-
cNMF_outputs.zip
-
GBM_Xenograft_Visium.zip
-
README.md
-
Spot_metadata.csv
Jul 20, 2024 version files 1.40 GB
-
cNMF_outputs.zip
-
GBM_Xenograft_Visium.zip
-
README.md
-
Spot_metadata.csv
Abstract
Diffuse invasion of glioblastoma cells through normal brain tissue is a key contributor to tumor aggressiveness, resistance to conventional therapies, and dismal prognosis in patients. A deeper understanding of how components of the tumor microenvironment (TME) contribute to overall tumor organization and to programs of invasion may reveal opportunities for improved therapeutic strategies. Towards this goal, we applied a novel computational workflow to a spatiotemporally profiled GBM xenograft cohort, leveraging the ability to distinguish human tumor from mouse TME to overcome previous limitations in analysis of diffuse invasion. Our analytic approach, based on unsupervised deconvolution, performs reference-free discovery of cell types and cell activities within the complete GBM ecosystem. We present a comprehensive catalogue of 15 tumor cell programs set within the spatiotemporal context of 90 mouse brain and TME cell types, cell activities, and anatomic structures. Distinct tumor programs related to invasion were aligned with routes of perivascular, white matter, and parenchymal invasion. Furthermore, sub-modules of genes serving as program hubs were highly prognostic in GBM patients. The compendium of programs presented here provides a basis for rational targeting of tumor and/or TME components. We anticipate that our approach will facilitate an ecosystem-level understanding of immediate and long-term consequences of such perturbations, including identification of compensatory programs that will inform improved combinatorial therapies.
Methods
The dataset includes 10x Genomics Visium spatial transcriptomic profiles of 23 GBM xenograft samples established from six brain tumor initiating cell lines (BTIC) derived from four patients (BT143, BT238, BT134 and BT161). BTIC lines from two patients (BT143 and BT238) were established from samples obtained from the tumor core (x), the contrast-enhancing highly vascularized tumor margin (y), and the highly diffuse leading edge of the tumor (z) through pre-operative MRI-guided resections. Upon implantation, early, mid, and late timepoints of growth are sampled based on known time to endpoint. Biological replicates of these mouse brain sections are obtained 30-40um apart. Raw spatial transcriptomic data were processed using the SpaceRanger software (v1.3.1) to generate FASTQ files. Sequences were aligned to the hybrid genome reference sequence GRCh38-mm10-2020-A and barcode/UMI counting was performed using Space Ranger pipeline’s default settings and all samples were aggregated using the SpaceRanger Aggr pipeline. The aggregated gene expression data was factorized into gene expression programs using the Consensus Non-negative Matrix Factorization algorithm to generate two matrices – one specifying the contribution of each spot to the identified programs (usage matrix) and the other specifying contribution of each gene to the programs (gene-score matrix). The factorization was performed separately for human (Tumor) and mouse (Brain & TME) across all spots. Rank 15 and Rank 90 were chosen for the human and mouse data, respectively. The usage matrix from each factorization result was normalized such that the usage values per spot sum to 1. Next to quantify the signal of programs relative to the total signal from the two genomes, we further scaled the cNMF-usage matrices whereby human program usages per spot were multiplied by genome admixture ratios and mouse program usages per cell by 1-admixture. The admix-scaled usage matrices were used for further downstream analyses.