Salk Scientists Map How Every Cell Type in the Mouse Brain Ages Differently
Salk Institute
Fifty-seven million people worldwide live with a neurodegenerative disease. That number is expected to double every 20 years. Aging is the single biggest risk factor, but the molecular machinery connecting aging to neurodegeneration remains poorly mapped. A new atlas published in Cell on March 11 by researchers at the Salk Institute begins to fill in that map with a level of detail that has not existed before.
What the atlas contains
The dataset is massive. Researchers profiled over 200,000 single cells across eight brain regions and 36 distinct cell types, using two complementary techniques: DNA methylation sequencing and chromatin conformation capture. They then added nearly 900,000 cells captured with spatial transcriptomics, which maps gene activity while preserving the physical location of each cell within the tissue.
The result is the most comprehensive single-cell, multi-omic atlas of brain aging to date. It is publicly available on Amazon Web Services and the Gene Expression Omnibus, where it will serve as a reference framework for human brain studies, including those generated by the NIH's BRAIN Initiative.
Co-corresponding author Joseph Ecker, a professor at Salk and Howard Hughes Medical Institute Investigator, described the atlas as a framework for understanding how aging reshapes the brain at the molecular level, cell type by cell type, region by region.
The epigenome does not age uniformly
The central finding is that aging is not a uniform process across the brain. Different cell types and different brain regions show distinct patterns of epigenetic change, the chemical modifications on top of DNA that alter gene expression without changing the underlying genetic code.
One key type of epigenetic change is DNA methylation, the addition or removal of small chemical tags that can silence or activate genes. The atlas revealed that age-related methylation changes were more pronounced in non-neuronal cells, such as the glia that support and protect neurons, than in neurons themselves. This suggests that the brain's support infrastructure may be more vulnerable to age-related epigenetic drift than the neurons it is supposed to maintain.
The researchers also found that transposable elements, sometimes called "jumping genes," lose DNA methylation as cells age. These repetitive DNA sequences make up roughly half of the genome and are normally kept silent by methylation. When they become active, they can disrupt normal gene function and contribute to cellular dysfunction. The finding is consistent with a growing body of evidence that epigenetic erosion at transposable elements may be a driver of age-related decline.
The same cell type, different aging trajectories
The spatial transcriptomics data added a dimension that bulk sequencing cannot provide. First author Qiurui Zeng, a graduate student in Ecker's lab, noted that the same cell type ages differently depending on its location in the brain. Non-neuronal cells in the posterior brain showed more inflammation-related changes than those in the anterior regions. This spatial heterogeneity means that studying aging in a single brain region and generalizing the results to the whole brain is likely to produce misleading conclusions.
The chromatin conformation data revealed additional age-related changes, including increased strength at boundaries between chromosomal neighborhoods called topologically associating domains (TADs). These boundaries, maintained by a protein called CTCF, help organize the genome into functional units. Stronger boundaries in aged cells suggest a reorganization of the genome's three-dimensional architecture that could alter which genes are accessible for activation.
Co-corresponding author Margarita Behrens, a research professor at Salk, emphasized the interconnected nature of brain aging. The death of one group of neurons can cascade into circuit-level malfunction, as seen in Parkinson's disease. Having cell-type-specific understanding of how aging proceeds will expand the range of therapeutic possibilities by identifying which cells change first and which changes matter most.
From atlas to prediction
The team has already begun using the dataset to build deep learning models that predict age-related gene expression changes from epigenetic data. This lays the groundwork for what could eventually become a virtual model of brain aging, one that could predict how specific interventions might slow or reverse age-related decline in targeted cell populations.
Hosting the dataset on AWS was a deliberate choice. Nearly 900,000 spatially resolved cells generate enormous amounts of data, and many research labs lack the computational infrastructure to process it locally. Cloud hosting removes that barrier, placing the atlas alongside other major neuroscience resources like the Allen Brain Atlas and the Seattle Alzheimer's Disease Brain Cell Atlas.
Mouse brains are not human brains
The atlas was built from mouse brains, which share broad organizational principles with human brains but differ in important ways. The specific cell types, their relative proportions, and the genes they express are not identical. Findings from this atlas will need validation in human tissue before they can inform clinical approaches to neurodegeneration.
The study captures aging at discrete time points rather than continuously, which means it cannot distinguish between age-related changes that accumulate gradually and those that occur in sudden shifts. And while the atlas is comprehensive for a mouse brain study, 36 cell types may still group together cells with distinct functional properties.
Still, as a reference framework, the atlas provides something the field has lacked: a systematic, high-resolution baseline against which future studies of neurodegeneration, therapeutic interventions, and protective factors can be compared. The aging brain is not a single thing going wrong in one way. It is many things going wrong differently in different places. Mapping that complexity is the prerequisite to addressing it.