Cortical organoids anchor a conserved heat-stress signature and cerebellar vulnerability map

What the work claims

Pagni et al. claim that the molecular response to heat stress is conserved across very different human experimental systems and that this conservation can be used to understand why the cerebellum, and particularly Purkinje neurons, are selectively damaged in heatstroke. They integrate transcriptomes from a controlled sauna study in 15 healthy volunteers, a clinical heatstroke cohort of 19 patients and 19 controls, heat-shocked primary cortical neurons, and heat-shocked cortical brain organoids. They report that a programme of downregulated ribosome biogenesis, RNA processing, translation, and metabolism is shared across all four systems, and that the cerebellum consistently shows the lowest enrichment of these downregulated genes. In clinical heatstroke, the upregulated signature is instead enriched in cerebellar regions and in granule and Purkinje neurons. Connectivity Map analysis then nominates the mTOR inhibitor KU-0063794 as a compound predicted to reverse the heatstroke signature.¹

How it works

The study pulls together four independent heat-stress datasets. Sauna-exposed peripheral blood mononuclear cells were collected at baseline, after 15 minutes at 75-78 degrees Celsius, and after one hour of recovery; the one-hour recovery timepoint provided the strongest signature and was used for cross-system comparison. Clinical heatstroke transcriptomes came from blood sampled before cooling and again after cooling. Primary cortical neurons were heat-shocked at 45 degrees Celsius for 30 minutes and sampled at multiple recovery timepoints. Cortical organoid transcriptomes came from a published dataset of control and heat-shocked organoids.

Differential expression analysis followed a common pipeline. Across the four datasets, 790 genes were differentially expressed in common, and 436 of them were concordantly downregulated. Gene Ontology enrichment identified 72 biological-process terms shared among the downregulated gene sets, including ribosome and ribonucleoprotein-complex biogenesis, translation, and energy metabolism. No upregulated Gene Ontology term was shared across all four systems, suggesting that the induced stress response is more context-dependent than the conserved suppressive programme.

To test regional vulnerability, the authors mapped heat-responsive gene sets onto a human brain atlas of 13 regions. For downregulated genes, the cerebellum ranked lowest in every dataset, meaning its baseline expression profile is least aligned with the conserved suppression programme. For upregulated genes, the pattern inverted in the clinical heatstroke cohort: the cerebellum and cerebellar hemisphere were the two highest-ranked regions. A similar inversion appeared at the cellular level using a published single-nucleus RNA-sequencing atlas of the adult human cerebellum comprising 5,601 nuclei from four neurologically normal donors. Across all datasets, granule and Purkinje neurons showed the strongest enrichment for downregulated heat-response genes; in the clinical heatstroke upregulated signature, they again ranked highest among cerebellar cell types.

The Purkinje-specific vulnerability argument rests on genes that are both heat-responsive and differentially expressed between granule and Purkinje neurons. After stripping out cell-type identity genes, the Purkinje-higher set was reproducibly enriched for synapse organisation, chemical synaptic transmission, regulation of membrane potential, and ion transport. The authors interpret this to mean that Purkinje vulnerability reflects their high baseline investment in excitability and synaptic function rather than a failure of canonical heat-shock pathways.

For drug repurposing, the heatstroke signature was queried against the LINCS L1000 Connectivity Map using the top 150 upregulated and top 150 downregulated genes. Thirty-nine compounds reversed the signature at the conventional threshold, but most were profiled in only one cell line. Restricting to compounds tested in at least two cell lines left eight candidates, of which four had annotated mechanisms. KU-0063794, an mTOR inhibitor, was the named compound. A separate granule-versus-Purkinje query identified 17 high-confidence reversers, including trequinsin.¹

Where a skeptic should push

The central weakness is that none of the experimental systems are cerebellar. The heat-response signatures are derived from blood, cortical neurons, and cortical organoids; the cerebellar vulnerability inference is made by projecting those signatures onto baseline adult cerebellar transcriptomes. That design cannot distinguish whether Purkinje neurons react differently to heat or simply express different genes at rest. A direct heat-stress experiment in cerebellar organoids or tissue would be needed to test the causal claim.

The clinical heatstroke signature is also confounded. Heatstroke is not pure hyperthermia; it involves systemic inflammation, coagulopathy, endothelial injury, and circulatory failure. The blood transcriptome therefore reflects whole-body physiology, not just brain stress. The finding that the cerebellum becomes enriched for upregulated genes in this cohort could be driven by peripheral processes that happen to correlate with cerebellar baseline expression, rather than by a cerebellar-specific pathogenic programme.

The organoid result requires cautious interpretation. Cortical organoids showed the largest and most strongly downregulated transcriptional response, but the authors themselves note that this may be an artefact of their reductionist environment. Organoids lack vascular perfusion, immune input, thermoregulation, and systemic metabolic buffering. Their exaggerated stress response could make them oversensitive screens that flag pathways not relevant to intact organisms.

Finally, the Connectivity Map predictions are in silico only. KU-0063794 and trequinsin were not tested experimentally in neurons or organoids in this study. The positive-control analysis, which correctly recovered approved mTOR inhibitors for a tuberous-sclerosis signature, validates the pipeline in principle, but it does not validate any specific heatstroke prediction.

Implication for brain organoid stress and repurposing

The clearest contribution for organoid-based drug discovery is methodological. The study shows how to combine organoid transcriptomics with public human exposure and clinical datasets to generate a cross-system stress signature. That signature can then be queried against repurposing libraries, giving researchers a rational starting point rather than an open-ended phenotypic screen. For teams building brain organoid models of environmental or hypoxic injury, the work provides a list of conserved pathways to use as positive controls and a warning that organoids may amplify stress responses relative to intact tissue.

The opportunity is to use cortical organoids as a high-sensitivity detector for human neuronal stress programmes. Because they are human, three-dimensional, and multicellular, they capture processes that monocultures miss. If a compound such as KU-0063794 is tested in heat-stressed organoids and reproducibly normalises the conserved suppression signature, that would be a stronger repurposing hypothesis than the in silico prediction alone.

The threat is mistaking the organoid response for the patient response. The exaggerated downregulation in organoids could produce false-positive drug targets: a compound that rescues the organoid signature might simply be counteracting culture stress rather than heatstroke pathophysiology. The authors are appropriately cautious, but readers looking for a quick repurposing win may ignore the caveat. Another threat is the mismatch between cortical organoids and cerebellar disease. Even if the organoid signature is validated, it does not prove that the same mechanisms operate in Purkinje neurons. Cerebellar organoids or slice models would be a necessary next step.

Overall, the work is best viewed as a hypothesis-generating platform. It links organoid biology to clinical heatstroke through a conserved transcriptional programme and points to mTOR-related pathways as candidates for experimental protection. It does not yet deliver a therapy, but it does show how organoids can sit at the centre of a translational pipeline that connects environmental exposure, human genetics, and computational repurposing.

The bottom line

This is a thoughtful integration study that uses cortical organoids as one anchor for a conserved human heat-stress signature. The evidence for a conserved suppressive programme is strong, but the cerebellar vulnerability claim is inferential and needs direct cerebellar experiments. The repurposing predictions are plausible starting points, not validated therapies. What would strengthen the work is experimental testing of KU-0063794 or trequinsin in heat-stressed neurons or organoids, followed by validation in a cerebellar model. What would weaken it is evidence that the cerebellar enrichment disappears when systemic confounders are controlled.