8/ Second, we describe the diversity of cellular states and their pathway-based functional activities, with an expanded set of malignant cell states, including glial progenitor cell-like, neuronal-like, and cilia-like states that were previously depleted by tumor dissociation

9/ Third, after controlling for the frequencies of cellular states, we find that the remaining variation between GBM samples highlights three baseline gene expression programs which we labeled Neuronal, Glial, and Extracellular Matrix.

10/ These three layers of heterogeneity are inter-related and partially associated with specific genetic aberrations, thereby defining three stereotypic ecosystems in GBM. This work provides an unparalleled view of the multi-layered transcriptional architecture of GBM.

12/ Before that, I would like to say thank wonderful co-first authors, @AvishaySptizer, Kevin Johnson and Luciano Garofano. I thank my great mentor Mario Suva and outstanding co-PIs Tirosh Lab, Roel Verhaak, Antonio Iavarone and Anna Lasorella.

13/ Also appreciate all collaborators, patients and their families who generously provided the samples.

Heterogeneity only now coming in view through large scale efforts - this clearly demonstrates how in #gbm and other heterogeneous cancers if we don’t study 100+ patients then we are not going to solve the problem. Thx to all for opp to contribute to this effort.

1/ Truly excited to share our new study that I had to privilege to co-lead during my PhD alongside great friends and collaborators Masashi Nomura, Kevin C. Johnson and Luciano Garofano, which was published Nature Genetics! nature.com/articles/s4158…

2/ In this study we profiled the longitudinal evolution of glioblastoma at single-cell resolution - altogether 121 treatment-naïve and exposed tumors from 59 patients, 430K nuclei, full clinical annotation and whole exome/genome sequencing.

3/ We first leveraged this large cohort to revisit the intra- and inter-tumor heterogeneity in GBM and characterized three multi-layered transcriptional ecosystems. Read more about this study in this great thread by @Masashi → x.com/M_Nomura/statu…

4/ And now for the results! Across the cohort the most consistent change at recurrence was reduced malignant cell fraction, with a reciprocal increase in glial and neuronal TME cells. This was observed in ~66% of patients, suggesting increased tumor integration into brain tissue.

5/ Contrary to previous studies, we did not observe a significant enrichment of mesenchymal-like (MES-like) states at recurrence. Instead, recurrence trajectories were diverse and patient-specific, with no single state dominating across the cohort.

6/ Despite global conservation, individual matched tumor pairs showed frequent divergence, with the majority switching at least one transcriptional layer. Overall, 72% of all possible transitions were observed across the cohort.

7/ However, in certain sub-groups, specific trajectories do tend to recur more often. First, MGMT-Low tumors (likely TMZ responders) tend to lose MES-like states whereas MGMT-High tumors (likely non-responders) tend to gain MES-like at recurrence.

8/ Interestingly, quantifying MGMT activity from the single-cell expression data outperformed promoter methylation as a prognostic marker in this cohort.

9/ Second, small deletion phenotype (linked to radiotherapy) was associated with hypoxia-related states at recurrence, likely reflecting selection of radioresistant subpopulations. SBS21 (MMR-deficiency signature) also increased post-treatment in both CARE and GLASS cohorts.

10/ Overall genetic divergence (SNVs + CNAs) correlated with transcriptional evolution, suggesting that GBM cell state shifts during recurrence are at least partially genomically driven.

11/ To sum-up, our multi-center study offers a high-resolution atlas of GBM recurrence dynamics, shaped by treatment response and TME context and underscores the tremendous heterogeneity of this devastating disease.

12/ Big thanks to the PIs that supervised the project Tirosh Lab, Mario Suva, Roel Verhaak, Antonio Iavarone and Anna Lasorella, all collaborators and institutions involved in the CARE consortium Weizmann Institute, MGH Pathology, Broad Institute, Yale School of Medicine and Miami University.

13/ Big thanks also to the institutions that provided tumor samples Duke University, MD Anderson Cancer Center, Tokyo University Hospital, Pitié-Salpêtrière Hospital, St. Michael's Hospital, Seoul National University and NORLUX and funders NIH, National Cancer Institute, Mark Foundation and Sontag Foundation.

14/ This of course wouldn't have been possible without the generous contribution of the patients and their families, for which we are ever thankful.