Thanks for reading and we would be delighted to hear your comments/suggestions.

I’m also incredibly proud of our multi-institutional, collaborative team—this work reflects the efforts of researchers across disciplines and career stages, including three undergraduate students who are co-authors on this study!

We are deeply grateful to all the patients and their families whose generosity made this study possible.

This supports the conclusion that joint Hi-C/ONT-based reconstructions yield more accurate structural predictions than those based solely on short-read WGS.

We compared ecDNA structure and transcription in two tumors: in P59, EGFR and CDK4 were predicted to be on separate ecDNAs; in P68, on the same ecDNA. Xenium data showed stronger EGFR-CDK4 expression correlation in P68 compared to p59.

We found that ecDNA+ tumors form nuclear EGFR transcriptional hubs.
EGFR FISH on post-Xenium slides showed >70% of these hubs localized to the nucleus and overlapped with EGFR copy number - unlike control gene FABP7 with similar expression but no hub formation.

We found that EGFR ecDNA tumors are hypoxic, mesenchymal-rich, and pericyte-enriched.

Given our comprehensive whole-genome sequnecing datasets, we compared the spatial organization of ecDNA vs Linear Amplification GBMs

We found that IDH-mutant gliomas frequently harbored inflammatory microglia expressing CX3CR1 specifically within AC-like malignant neighborhoods, rather than OPC-like niches—despite the OPC-like malignant cell state being more prevalent in IDH-mutant gliomas.

To investigate the spatial tumor microenvironment of our glioma samples, we applied a single-cell spatial transcriptomics approach (10x Xenium), generating 4.6 million cells and over 775 million transcripts.
We identified distinct cell states in IDH-wildtype vs mutant gliomas

We next asked whether treatment contributed to ecDNA loss in recurrences. Recurrent tumors arose from residual regions of the primary site, suggesting ecDNAs were spatially restricted subclones eliminated during initial surgery.

We also examined EGFR expression levels, which showed a marked reduction in the recurrent tumors. Similarly, single-nucleus whole-genome analysis revealed the disappearance of the focal amplification on chromosome 7, while clonal alterations were retained.

Hi-C and WGS data revealed the loss of ecDNA signals in recurrent tumor (P-12r1) while preserving other structural variations, such as chromothripsis on chr 3

We also identified two GBM patients (P12 and P29) in whom ecDNA molecules were not implicated in tumor recurrence and were absent in both subsequent surgical specimens.

Hi-C and long-read data were essential to improve ecDNA reconstructions.
For example, for p59, the short-read reconstruction suggested EGFR and CDK4 genes are on the same ecDNA circle.
However, Hi-C and ONT WGS-based reconstructions suggested these genes reside on two distinct circles.

🔹 Resolving ecDNA Structure and Function
We used short- and long-read whole-genome sequencing plus Hi-C to reconstruct complex extrachromosomal DNA (ecDNA) structures. This revealed:

Tumors can harbor multiple ecDNA circles with distinct oncogenes.

We performed an integrative multi-omic and spatial transcriptomic analysis of 93 gliomas, leveraging 449 unique specimens to uncover how oncogenic drivers reshape tumor architecture and the surrounding microenvironment in human gliomas.

Omg, this is super funny but also very painful!!!