ALS Cognitive Decline Traced to Distinct Brain Cell Patterns by Region
Spatial transcriptomics reveals that language vs. executive dysfunction in ALS stem from different cellular and molecular failures in the prefrontal cortex.
Summary
Many people with ALS develop thinking problems on top of muscle weakness, but the cause has been unclear. Researchers mapped gene activity across two prefrontal brain regions in donated tissue from people with ALS, with and without cognitive impairment, plus healthy controls. They found that executive problems, such as planning and decision-making, were linked to lower mitochondrial and synaptic activity in deep-layer neurons of the dorsolateral prefrontal cortex. Language problems instead tracked with a widespread reaction involving reactive glial cells and blood-vessel-associated immune changes. Neither the C9ORF72 mutation nor TDP-43 protein pathology predicted who had cognitive impairment. The work suggests ALS is not one uniform process in the brain. Different cognitive symptoms reflect different cellular disruptions, which could eventually guide more targeted treatments.
Detailed Summary
Amyotrophic lateral sclerosis (ALS) is best known for destroying motor neurons, yet about half of patients show extra-motor involvement. Roughly 35% fall in a mid-spectrum range with mild deficits in language, executive function and behavior. It has been unclear whether these varied clinical pictures share one root mechanism or arise from region-specific cellular disruptions. That distinction matters for anyone thinking about neurodegeneration as a family of age-related brain diseases, and for designing targeted therapies.
The team used brain tissue from the Edinburgh Brain Bank, stratified by the Edinburgh Cognitive and Behavioral ALS Screen (ECAS): 19 ALS donors with normal cognition, 17 with cognitive impairment (ALSci), and 17 age-matched neurotypical controls. They sampled BA46 in the dorsolateral prefrontal cortex (executive function) and BA44/45 (Broca's area, language). Methods included 246 Visium spatial transcriptomic arrays from 45 donors (953,073 spots), Multiome snRNA/snATAC-seq (144,260 nuclei) and multiplexed COMET imaging (about 3 million cells, 17 donors). Language was the most frequently impaired domain. Neither C9ORF72 expansion nor TDP-43 pathology predicted cognitive status.
Glial reactivity was common. Reactive microglia and stressed oligodendrocyte progenitor cells increased in both ALS and ALSci, while reactive astrocytes increased significantly only in ALSci. The response appeared across cortical layers L1-L6 and white matter, with AP-1 motif enrichment in open chromatin. Donors with reactive gliosis had significantly lower language scores, an association specific to BA44/45. Imaging confirmed it: astrocytes were less round (more ramified), and IBA1+ microglial area fell without a drop in cell number, consistent with a shift toward an amoeboid morphology.
Using the Splotch Bayesian model, the authors derived 23 spatial gene modules and analyzed 21 of them. Language deficits correlated with upregulation of gliosis-linked modules M2 and M20 in both regions. M2 contained endothelial, microglial and neuronal submodules whose spatial cross-correlation rose from control to ALS to ALSci, suggesting intermixing of normally segregated processes. Around aSMA-positive vessels, IBA1+ signal was reduced in all ALS donors, whereas elevated CD68 (phagocytic activation) appeared selectively in ALSci.
Executive dysfunction followed a different pattern. Only two mitochondria-associated modules, M8 and M9, correlated with executive scores, and only in BA46 layers L5-L6. L5 IT, L6 IT and L6 IT Car3 excitatory neurons, plus L5 PVALB+ basket and L5-L6 LAMP5+/LHX6+ interneurons, showed reduced expression in ALSci, including CALM2 and GAPDH. Healthy aging also lowered M8/M9, but across L1-L5 in both regions, so spatial profiling could separate age-related from disease-related change. Neuronal subtype proportions did not differ between groups.
The findings argue that ALS cognitive heterogeneity reflects distinct, region-specific cellular programs rather than a single process, and that TDP-43 pathology only partly predicts them. Caveats: the data are postmortem and correlative, endothelial cells were undersampled (1.1% of nuclei), and the imaging measures cannot resolve individual-cell morphology. Visium spots also capture multiple cells, and the text provided was truncated, so the discussion could not be reviewed.
Key Findings
- Cohort stratified by ECAS into 19 ALS (normal cognition), 17 ALSci (cognitive impairment) and 17 age-matched controls; language was the most frequently impaired domain
- Neither C9ORF72 repeat expansion nor TDP-43 pathology predicted cognitive status in the cohort
- Reactive microglia and stressed OPCs rose in both ALS and ALSci, but reactive astrocytes rose significantly only in ALSci
- Donors with reactive gliosis had significantly lower language scores, an association specific to BA44/45
- Language deficits correlated with upregulation of gliosis-linked spatial modules M2 and M20 across both regions, and M2 submodule cross-correlation rose from control to ALS to ALSci
- Executive impairment correlated with downregulation of mitochondrial modules M8/M9 only in BA46 deep layers (L5-L6), driven by L5/L6 IT excitatory neurons and PVALB+/LAMP5+ interneurons
- Perivascular IBA1+ area fell in all ALS donors, while elevated perivascular CD68 appeared selectively in ALSci; the dataset spans 246 Visium arrays, 144,260 nuclei and about 3 million imaged cells
Methodology
Postmortem prefrontal cortex (BA44/45 and BA46) from a single biorepository, the Edinburgh Brain Bank, was profiled with 10x Visium spatial transcriptomics (246 arrays, 45 donors), paired snRNA/snATAC-seq via Multiome (24 samples per region, 144,260 nuclei), and Lunaphore COMET multiplexed immunofluorescence (17 donors, 31 sections, about 3 million cells). Cell types were annotated with a classifier trained on 2.6 million public snRNA-seq profiles. Spatial modules were derived with the Splotch Bayesian hierarchical model and Leiden clustering, then correlated with ECAS domain scores and compared across control, ALS and ALSci groups. Quality control found no systematic confounding by RNA integrity or postmortem interval, and one outlier donor with high interferon gene expression was excluded.
Study Limitations
The data come from postmortem tissue with a wide postmortem interval (24-231 hours) and a modest donor number, and the results are correlative. Endothelial cells were undersampled in snRNA-seq (1.1% of nuclei), Visium spots capture multiple cells, and IBA1 area and astrocyte roundness are coarse proxies that cannot resolve per-cell morphology. Conflict-of-interest information was not included in the supplied text.
Enjoyed this summary?
Get the latest longevity research delivered to your inbox every week.
Enter your email to subscribe: