🔬 oncology

Longitudinal multi-omics characterization of the malignant evolution in multirelapsing glioblastoma

This study presents a longitudinal multi-omics analysis of a single patient with multirelapsing glioblastoma, linking specific genomic and cellular evolution patterns across 11 surgeries over 31 months to clinical progression, treatment resistance, and extracranial metastasis.

Original authors: Lackman, M. H., Wardell, C., Darrigues, E., De Loose, A., Lyle, G. A., Xue, Y., Learned, K., Cheney, A., Vaske, O. M., Karaman, S., Le Joncour, V. J., Rodriguez, A.

Published 2026-06-22

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Lackman, M. H., Wardell, C., Darrigues, E., De Loose, A., Lyle, G. A., Xue, Y., Learned, K., Cheney, A., Vaske, O. M., Karaman, S., Le Joncour, V. J., Rodriguez, A.

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

Imagine a patient's brain tumor not as a static lump, but as a living, breathing city that is constantly rebuilding itself, changing its laws, and adapting to every attack launched against it. This paper tells the story of one such city—a Glioblastoma (GBM)—observed over 31 months as it fought a war against various medical treatments.

Here is the story of that battle, broken down into simple terms.

The Patient and the Timeline

The researchers followed a young man with a very aggressive brain cancer. Over the course of nearly three years, he underwent 11 different surgeries to remove tumors that kept popping up in different parts of his brain, and eventually, even outside his skull (in his nose and neck lymph nodes).

Usually, scientists only get to look at a tumor once or twice. But here, they had a rare "front-row seat" to the entire movie of the disease, collecting fresh tissue samples every time the tumor returned. They treated these samples like time capsules, analyzing their DNA and how their genes were "speaking" (turning on and off) at every stage.

The Genetic Family Tree

Think of the tumor as a family. When the patient was first diagnosed, the tumor had a few "family secrets" (mutations) that started the whole thing, including a broken gene called TP53 (which the patient was born with) and a few others like RB1 and NF1.

As the tumor grew and moved to new locations, it didn't just copy itself perfectly. It was like a family tree branching out.

The Ancestors: All the new tumors shared the original "family secrets."
The New Branches: As the tumor evolved, it picked up new mutations (like TSC2, CD22, and FANCI) that helped it survive better.
The Survivors: One specific branch of the tumor family (called "Clone Epsilon") was incredibly tough. It appeared early and was the main survivor in the final years, even in the tumors that spread outside the brain.

The Shape-Shifting City (Cellular Evolution)

The most fascinating part of the story is how the tumor cells changed their "personality" or "uniform" to survive. The researchers identified four main types of "citizens" in this tumor city:

Mesenchymal (MES): The "warriors." They are tough, aggressive, and good at fighting inflammation.
Astrocyte-like (AC): The "builders." They look like normal brain support cells.
Oligodendrocyte-like (OPC): The "insulators."
Neural Progenitor-like (NPC): The "stem cells" or "seeds." They are young, flexible, and can turn into anything.

The Twist:
When the doctors attacked the tumor with radiation or drugs, the tumor didn't just die; it changed its uniform.

When hit with certain drugs, the "warrior" cells (MES) would shrink back, and the "builder" cells (AC) would take over.
However, as the tumor got worse and started spreading to the nose and neck, something strange happened. The tumor cells started acting more like neurons (brain cells). They began mimicking the brain's own electrical wiring. The paper suggests that by pretending to be neurons, the tumor cells could sneak through the brain's defenses and travel to new places, including outside the skull.

The Arms Race: Drugs vs. Adaptation

The patient was treated with a "precision medicine" arsenal, including immunotherapy, targeted drugs, and chemotherapy. The researchers watched how the tumor reacted in real-time:

The Immunotherapy (Nivolumab): The "warrior" cells (MES) initially lowered their shields (a protein called CD274) to hide from the immune system. But eventually, they raised their shields again, and the tumor grew back.
The Targeted Drug (Osimertinib): This drug was designed to stop a specific growth signal (EGFR). The tumor responded by killing off the "warrior" cells, but the "builder" cells (AC) survived and actually started overproducing the very signal the drug was trying to block, effectively outsmarting the medicine.
The Final Escape: Despite these changes, the tumor eventually found a way to thrive. The "neuron-like" cells became the dominant force in the final stages, driving the cancer to spread to the lymph nodes and the nasal cavity.

The Big Takeaway

This paper is like a detailed map of a battle that was lost, but it teaches us how the enemy fought.

Tumors are not static: They are dynamic cities that constantly rebuild and change their identity to survive attacks.
Adaptation is key: When you block one path, the tumor finds another. If you stop the "warriors," the "builders" take over.
The "Neuron" Trick: The most dangerous evolution was the tumor learning to act like brain cells (neurons). This allowed it to spread everywhere, even outside the brain.
The Need for Serial Sampling: The authors argue that to understand cancer, we can't just look at it once. We need to keep taking samples (like checking the weather every day, not just once a year) to see how the tumor is changing its strategy.

In short, this study shows that Glioblastoma is a master of disguise. It changes its genetic code, its cell types, and its behavior to dodge every treatment thrown at it, eventually finding a way to spread by pretending to be the very brain cells it is destroying.

Technical Summary: Longitudinal Multi-omics Characterization of Malignant Evolution in Multirelapsing Glioblastoma

Problem Statement
Glioblastoma (GBM) remains a formidable therapeutic challenge, with the standard of care unchanged since 2005. A primary barrier to understanding therapeutic failure and advancing drug development is the lack of longitudinal analyses of neoplastic material throughout the treatment course. Previous attempts at comprehensive molecular profiling of GBM progression have been limited by sampling only a single recurrence or failing to capture the dynamic interplay between genetic evolution, cellular plasticity, and clinical relapse. Furthermore, the short survival of GBM patients often precludes the collection of serial samples necessary to map the full trajectory of tumor evolution, particularly regarding extracranial metastasis and terminal progression.

Methodology
This study presents a unique, multidimensional characterization of a single adult male patient with Li-Fraumeni syndrome (germline TP53 mutation) and IDH-wildtype GBM. The patient underwent 11 neurosurgical resections over a 31-month period, spanning from initial diagnosis to terminal extracranial metastasis (lymph node).

The research employed a longitudinal multi-omics approach:

Genomic Analysis: Bulk DNA sequencing was performed on all 11 tumor samples (T1–T11) to identify somatic mutations, copy number variations (CNVs), and loss of heterozygosity (LOH). Phylogenetic trees were reconstructed using high-confidence variants to infer tumor evolution and clone distribution.
Transcriptomic Analysis: Bulk RNA sequencing was conducted on all 11 samples to assess gene expression dynamics. Single-cell RNA sequencing (scRNA-seq) was performed on samples T4 through T10 (covering 7 distinct anatomical sites), yielding over 41,000 high-quality cells.
Computational Integration:
- Deconvolution: Bulk RNA-seq data was analyzed using GBMDeconvolutR to infer cellular proportions (neoplastic, immune, stromal).
- Cell State Mapping: scRNA-seq data was integrated and clustered using Seurat to identify phenotypic cell states (Mesenchymal [MES], Astrocyte-like [AC], Oligodendrocyte Progenitor-like [OPC], and Neural Progenitor-like [NPC]).
- Trajectory Analysis: Monocle was used for pseudotime analysis to order cells from stem-like to differentiated states. RNA velocity (scVelo) was employed to predict future cellular states and transition directions.
- Therapy Response Modeling: "Quadriga plots" were developed to visualize the abundance and gene signature strength of the four major cell states alongside the expression of precision medicine targets (e.g., EGFR, JAK1/2, mTOR) in response to specific therapies.

Key Results

Genomic Evolution and Clonal Architecture:
- All 11 neoplastic lesions shared a common genomic ancestry defined by early mutations in NF1, RB1, GPS2, and APOB, alongside the germline TP53 mutation.
- The study identified 13 distinct tumor clones. While early mutations were ubiquitous, later-stage progression was driven by the acquisition of mutations in TSC2, PBRM1, CD22, and FANCI.
- Clones exhibited parallel evolution; for instance, RB1 and TSC2 mutations appeared in independent clones, suggesting a fitness advantage.
- Clone $\epsilon$ emerged early and dominated later samples, including the terminal extracranial metastases (nasal cavity and lymph node), indicating a persistent, long-lived subpopulation capable of widespread dissemination.
Transcriptomic Dynamics and Cell State Plasticity:
- Therapy-Induced Shifts: The tumor exhibited reversible shifts in phenotypic cell states in response to precision medicine. For example, treatment with the EGFR inhibitor Osimertinib (Tagrisso) led to a decrease in MES-like signatures and a compensatory enrichment of AC-like cells, which subsequently upregulated EGFR. Similarly, JAK/STAT inhibition (Jakafi) and mTOR inhibition (Everolimus) transiently reduced MES-like populations, which were replaced by AC-like cells.
- Immune Context: Early MES-like signatures correlated with high myeloid cell infiltration, suggesting the MES signature in bulk data often reflected immunoreactivity rather than intrinsic tumor state. Later stages showed immune exhaustion and a shift toward pro-neural signatures.
- Metastatic Trajectory: As the disease progressed toward terminal extracranial metastasis, there was a robust and consistent enrichment of pro-neural and neuron-like GBM cell signatures. RNA velocity analysis indicated a trajectory steering from MES/AC-like states toward pro-neural states with invasive features.
Extracranial Metastasis:
- The study confirmed that extracranial metastases (nasal cavity and neck lymph node) retained the same primordial genetic signature as the primary brain tumor but were characterized by a distinct upregulation of pro-neural and neuronal transcriptomic features. This suggests that neuronal mimicry is a core program driving both intracranial dissemination and extracranial spread.

Key Contributions

Unprecedented Longitudinal Scale: The study provides a rare, high-resolution view of GBM evolution across 11 surgical samples over 31 months, capturing the transition from primary tumor to terminal extracranial metastasis.
Multi-omic Correlation: It successfully links specific genomic events (e.g., TSC2, FANCI mutations) to transcriptomic shifts and clinical outcomes, demonstrating how genetic evolution drives phenotypic plasticity.
Therapy Response Mapping: By integrating single-cell data with clinical treatment history, the study delineates how specific targeted therapies (Nivolumab, Palbociclib, Osimertinib, Everolimus, Jakafi) induce specific, often reversible, cellular adaptations (e.g., MES-to-AC transitions).
Resource Generation: The authors developed "GBMaps," an open-access Shiny app for exploring the integrated single-cell data, and provided a comprehensive dataset linking genetic, molecular, cellular, and clinical progressions.

Significance
The paper claims to offer a "roadmap for understanding the intricacies of treatment response and failure" in GBM. Its primary significance lies in demonstrating that GBM cells exhibit remarkable adaptability, shifting between phenotypic states (MES, AC, pro-neural) in response to therapeutic pressure. Crucially, the study identifies that while therapies may transiently suppress specific clones, the tumor often adapts via alternative pathways or state transitions.

The authors conclude that the enrichment of neuronal and pro-neural features is a fundamental driver of both intracranial dissemination and extracranial metastasis. This finding prompts a re-evaluation of therapeutic paradigms, suggesting that targeting the plasticity of cell states and the neuronal mimicry of GBM cells may be necessary to prevent relapse. The study underscores the imperative of serial tissue sampling to decipher the mechanisms of resistance, arguing that static, single-timepoint analyses are insufficient to capture the dynamic nature of GBM evolution.