https://www.biorxiv.org/content/10.64898/2026.06.12.730900v1
This tweet drew the paper to my attention:
I have seen previous work about the exchange of mitochondria via nanotubes. Hence I don’t know how much of this paper (preprint) is that new.
chatGPT(5.5):
Summary
The uploaded paper, “Tunneling nanotubes mediate mitochondrial homeostasis in cancer,” argues that cancer cells use tunneling nanotubes-like connections (TNTs) to exchange mitochondria with each other and with non-tumoral cells in the tumour microenvironment. The main model is that glioblastoma cells can both acquire healthy mitochondria from astrocytes and export damaged mitochondria to astrocytes, thereby improving mitochondrial fitness and resilience.
The authors study glioblastoma models including U251 cells and patient-derived glioblastoma stem-like cells, plus primary astrocytes. They show by live imaging that mitochondria move through TNTs in a saltatory, active-transport-like manner, with mitochondria sometimes travelling through membrane bulges or “gondolas.” They distinguish actin-based TNTs from thicker, more stable microtubule-containing TNTs, and find that both can mediate mitochondrial transfer, although the dominant TNT type varies by cell model.
A key result is that mitochondrial transfer between astrocytes and tumour cells is bidirectional but asymmetric. Astrocytes transfer mitochondria to tumour cells much more efficiently than tumour cells transfer mitochondria to astrocytes. In the reported FACS experiments, astrocytes transferred mitochondria to about 50.7% of U251 cells, 5.4% of GSC(-) cells and 18.0% of GSC(+) cells, whereas transfer from tumour cells to astrocytes was lower, around 8.0% from U251 and about 1–2% from GSCs.
The fate of the transferred mitochondria differs by direction. Tumour-derived mitochondria entering astrocytes appear damaged or depolarised, remain globular, fail to integrate into the astrocytic mitochondrial network, and colocalise with lysosomal markers, suggesting degradation by mitophagy. Conversely, astrocyte-derived mitochondria entering tumour cells can integrate into the tumour mitochondrial network, retain membrane-potential staining, and appear functionally active, especially when many mitochondria are transferred.
The RNA-seq experiment strengthens the functional claim. GSC(+) cells that had received astrocyte-derived mitochondria clustered separately from non-recipient GSC(+) cells and showed broad transcriptional activation, including upregulation of genes linked to mitochondrial metabolism, electron transport, membrane potential regulation, transmembrane transport, and motor proteins involved in organelle movement. The authors interpret this as evidence that astrocyte-derived mitochondria enhance mitochondrial metabolic activity in tumour cells.
The paper also attempts to address the difficult question of whether TNT-like mitochondrial transfer occurs in vivo. Because intravital subcellular microscopy was not suitable for deep-brain glioblastoma imaging, the authors used orthotopic tongue cancer models. They report TNT-like actin-containing structures in live tumours with mitochondria inside them, plus mitochondrial puncta in recipient tumour cells, and also evidence that mitochondria from the tumour microenvironment can enter tumour cells in vivo.
Claimed novelty
The strongest novelty is the proposed two-way mitochondrial homeostasis model: cancer cells may use astrocytes both as a source of healthy mitochondria and as a sink for damaged mitochondria. Previous work has shown mitochondrial donation to cancer cells, but this paper combines donation, disposal, mitophagy in recipient astrocytes, and transcriptional activation in recipient glioblastoma stem-like cells into one integrated model.
A second novel element is the dynamic live-cell characterisation of TNT-mediated mitochondrial transfer. The paper gives measurements of TNT diameter, length, lifetime, cytoskeletal composition and mitochondrial movement, rather than relying only on fixed-cell imaging. This is important because thin actin-based TNTs can be disrupted by fixation.
A third novel contribution is the use of intravital subcellular microscopy to visualise TNT-like structures in live tumour-bearing animals. The in vivo data are not as definitive as the in vitro data, but they are technically significant because TNTs are thin, dynamic and lack specific markers, making them hard to prove in living tissue.
A fourth novelty is the direction-specific interpretation of mitochondrial fate: tumour-to-astrocyte transfer looks like clearance of damaged mitochondria, while astrocyte-to-tumour transfer looks like functional metabolic support. That asymmetry is biologically interesting and potentially therapeutically relevant.
Critique
The paper is conceptually strong, but the central claim that TNTs mediate mitochondrial homeostasis in cancer is stronger for the in vitro data than for the in vivo data. In vitro, the authors directly visualise TNTs, mitochondrial movement and recipient-cell outcomes. In vivo, they observe TNT-like structures and mitochondrial puncta in recipient cells, but they acknowledge that they could not always capture complete transfer events from donor to recipient. Thus, the in vivo evidence supports plausibility rather than fully proving functional TNT-mediated transfer in live tumours.
A second limitation is that the in vivo model is head and neck squamous cell carcinoma in the tongue, not glioblastoma in the brain. This is understandable technically, because deep-brain intravital imaging is difficult, but it weakens the direct translation of the in vivo findings to GBM. The paper’s glioblastoma conclusions rely mainly on culture and organoid-style systems, while the strongest live-animal imaging is from a different cancer context.
A third issue is the specificity of TNT-mediated transfer. The authors attempt to control for secreted mitochondrial transfer using a filter system, which is useful, but mitochondrial material can move by multiple routes: extracellular vesicles, cell fragments, phagocytosis-like uptake, direct cell contact, cytonemes, tumour microtubes, or other membrane bridges. Because TNTs lack a unique molecular marker, the attribution to TNTs remains partly morphological and operational.
A fourth concern is the interpretation of the RNA-seq result. The finding of 1,982 upregulated genes and no downregulated genes is striking, but it raises questions. It may reflect genuine activation after mitochondrial acquisition, but it could also reflect sorting effects, different cell states, stress responses, donor-cell material contamination, differences in cell cycle, or selection of a more metabolically active subpopulation. The RNA-seq shows association between mitochondrial receipt and transcriptional activation; it does not by itself prove that transferred mitochondria are the causal driver of all those changes.
A fifth limitation is that functional metabolic assays seem less central than imaging and transcriptomics. The paper infers enhanced tumour-cell mitochondrial activity from membrane-potential dyes, mitochondrial integration and gene-expression changes. Stronger support would come from direct measurements such as Seahorse respiration, ATP production, mitochondrial DNA tracking, clonal fitness, survival under stress, radiation/temozolomide resistance, or tumour growth after blocking transfer.
A sixth critique is therapeutic. The authors suggest TNTs as potential therapeutic targets, but TNT biology overlaps with normal repair and intercellular rescue mechanisms. Blocking TNTs or mitochondrial transfer could impair beneficial mitochondrial donation in neurons, immune cells or damaged tissues. The paper would be stronger if it identified tumour-selective mechanisms, such as specific motor proteins, adhesion molecules, actin regulators or stress pathways that are required in tumour TNTs but less essential in normal tissue.
Overall assessment
This is a technically ambitious and biologically interesting preprint. Its strongest contribution is the model that tumour cells can maintain mitochondrial quality by importing healthy mitochondria from astrocytes while exporting damaged mitochondria for astrocytic degradation. The live-cell imaging and directional transfer experiments are persuasive, and the RNA-seq data suggest a real metabolic shift in recipient tumour cells.
The main weakness is that the most direct mechanistic evidence is in vitro, while the in vivo evidence is suggestive rather than definitive, and partly comes from a non-GBM tumour model. The next decisive experiments would be to block TNT formation or mitochondrial transport genetically or pharmacologically, then test whether this reduces mitochondrial rescue, tumour-cell metabolism, therapy resistance and tumour growth in relevant glioblastoma models.