Somatic mutations and single-nucleus transcriptomics reveal uniquely human properties of neuronal aging

Problem Statement

Human brain aging is frequently modeled in mammals that rarely exhibit natural age-related neurodegenerative conditions. While somatic mutations (single-nucleotide variants [SNVs] and short insertions/deletions [Indels]) accumulate in all cells, the relationship between genomic stability, mutation accumulation rates, and lifespan across species remains poorly understood. Previous studies in intestinal crypts suggested that longer-lived species accumulate somatic mutations at slower rates, resulting in a relatively consistent end-of-life mutation burden across mammals. However, it is unknown if this evolutionary "ceiling" applies to post-mitotic neurons, which cannot be replenished by stem cells and thus face irreversible accumulation of DNA damage. Given the human susceptibility to neurodegenerative diseases and the potential for decades-long lifespans compared to close relatives, a systematic comparison of neuronal genomic changes across species is critical to determine if human aging involves unique molecular vulnerabilities.

Methodology

The authors employed a multi-species, single-cell approach to characterize somatic mutational landscapes and transcriptomic changes in the dorsolateral prefrontal cortex (PFC).

1. Development of SCAN2-Zoo: To analyze somatic mutations in non-human species, the authors modified the existing human-specific algorithm SCAN2 into SCAN2-Zoo. This new pipeline replaces the requirement for phased heterozygous SNPs (which are scarce in non-human datasets) with a maximum variant allele frequency (maxVAF) approach to construct allelic balance curves, maintaining high fidelity for somatic mutation calling across diverse diploid species.
2. Sample Collection and Sorting: The study utilized postmortem brain tissue from six mammalian species: human, chimpanzee, rhesus macaque, marmoset, mouse, and ferret. To minimize cell-type heterogeneity, the authors developed a fluorescence-activated nuclei sorting (FANS) strategy using antibodies against NeuN, SOX6, and SOX10 to isolate pure populations of glutamatergic neurons and MGE-GABAergic neurons (parvalbumin/somatostatin-expressing).
3. Single-Cell Whole Genome Sequencing (scWGS): Sorted single nuclei underwent whole genome amplification via Primary Template Amplification (PTA) and were sequenced at 25–30X coverage. Mutations were called using SCAN2-Zoo.
4. Validation with META-CS: To validate findings, a subset of samples (human and rhesus macaque) was analyzed using Multiplexed End-tagging Amplification of Complementary Strands (META-CS), a duplex sequencing technology that relies on strand consensus to distinguish true somatic variants from artifacts.
5. Mutational Signature Analysis: The authors performed de novo mutational signature extraction using Non-Negative Matrix Factorization (NMF) on trinucleotide spectra and compared them against the COSMIC database. They also analyzed existing single-neuron data from neurodegenerative diseases (Alzheimer's, CTE, FTD, ALS).
6. Single-Nucleus Transcriptomics (snRNA-seq): To assess functional consequences, the authors performed snRNA-seq on sorted nuclei from adult and aged humans, chimpanzees, and rhesus macaques. Cells from different species were processed in the same reaction to mitigate ambient RNA artifacts, with species assignment performed bioinformatically based on unique molecular identifier (UMI) mismatches. Differential gene expression (DGE) analysis was conducted on pseudobulk counts.

Key Results

1. Conservation of Annual Mutation Rates, Divergence of End-of-Life Burden
Contrary to the pattern observed in intestinal crypts, the annual rate of somatic SNV and Indel accumulation in cortical neurons was remarkably similar across all six species (~18–24 SNVs/year).

• Human: ~18 SNVs/year.
• Mouse/Ferret: ~18–24 SNVs/year.
• Consequence: Because humans live significantly longer, the end-of-life mutational burden in human neurons is vastly higher (>12-fold more SNVs and hundreds more Indels) than in mice, and significantly higher than in chimpanzees and macaques.

2. Species-Specific Mutational Signatures
Despite similar annual rates, the underlying mutational mechanisms differ:

• SBS-C (Human-Specific): A signature dominated by C>T transitions (specifically at CpG and non-CpG sites) accumulates significantly faster in humans (~15% of total mutations) compared to non-human primates (near zero). This signature resembles COSMIC SBS30, associated with base excision repair (BER) deficiency and oxidative stress.
• SBS-A (NHP-Specific): A signature dominated by T>A and T>C transitions accumulates at high rates in rhesus macaques and other non-human primates but is lower in humans.
• Disease Correlation: The human-specific SBS-C signature is significantly elevated in neurons from patients with Alzheimer's disease, CTE, FTD, and ALS compared to neurotypical controls.

3. Transcriptomic Dysregulation in Human Aging
Single-nucleus transcriptomics revealed that human aging neurons undergo pervasive transcriptional changes not observed to the same extent in chimpanzees or macaques.

• Gene Downregulation: Aged human neurons show extensive downregulation of hundreds of genes, whereas aged non-human primates show significantly fewer downregulated genes.
• Functional Pathways: The downregulated genes in humans are highly enriched for proteostasis (chaperones, ribosomal proteins, proteasome components) and mitochondrial function (respiratory complexes).
• Specific Markers: The gene CIRBP, associated with long-lived species like bowhead whales, was uniquely downregulated (>2-fold) in aged humans.

Significance and Claims

The paper claims that human neuronal aging is uniquely severe at the molecular level compared to close evolutionary relatives and standard animal models.

• Evolutionary Constraint: The authors propose that while there may be an evolutionary constraint on the annual rate of mutation accumulation in neurons (perhaps to maintain synaptic plasticity or avoid immediate toxicity), there is no such constraint on the end-of-life burden. This allows the burden to accumulate to levels that are potentially damaging in long-lived humans but are not reached in shorter-lived species.
• Limitations of Animal Models: The findings suggest that animal models (mice, macaques) fail to recapitulate the "uniquely human" late-life condition characterized by high somatic mutation burdens, specific mutational signatures (SBS-C), and widespread proteostasis/mitochondrial collapse.
• Neurodegeneration Link: The accumulation of the human-specific SBS-C signature, linked to oxidative stress repair pathways, and its elevation in neurodegenerative diseases, suggests that the genomic instability inherent in human aging may predispose the brain to neurodegeneration.
• Mechanism of Decline: The extensive gene downregulation in humans is not solely driven by the higher mutational burden in coding regions (as no excess burden was found in downregulated genes), suggesting that other forms of genetic instability or stress pathways triggered by the aging genome drive the transcriptional decline.

In summary, the study establishes that human neurons accumulate somatic mutations at a rate similar to other mammals but over a much longer timeframe, resulting in a uniquely high mutational load and a specific mutational signature linked to oxidative stress. This genomic instability coincides with a profound, species-specific collapse in proteostasis and mitochondrial function, defining a molecular landscape of aging that is not fully captured by current non-human models.