Investigators isolated nuclei from frozen parietal-cortex tissue and used fluorescence-activated cell sorting to enrich NeuN-positive neuronal populations. Two sorting rounds produced samples in which more than 70% of nuclei were tetraploid, although one Alzheimer’s sample was discarded because it did not meet that threshold. Researchers then extracted genomic DNA from 430 to 550 neuronal nuclei per sample and amplified it with PicoPLEX before preparing exome libraries for sequencing. The resulting reads were aligned to the human reference genome, and SNVs and INDELs were called with VarScan 2 and LoFreq. Variants were classified as germline or somatic primarily by allele frequency: a threshold of 0.25 separated the low-frequency population interpreted as somatic from the predominantly germline variants at higher frequencies.

The rise in mutations was highly specific to single-base changes. Overall, Alzheimer’s-derived tetraploid neurons contained 16% more mutations than diploid neurons from the same patients, with the difference attributable to a 17% increase in sSNVs. Somatic INDEL counts were similar in tetraploid Alzheimer’s neurons, diploid Alzheimer’s neurons and control neurons. A second analysis with LoFreq, which can identify variants occurring at frequencies as low as roughly 0.05% to 1%, again found significantly more sSNVs in tetraploid cells. Most were concentrated at allele frequencies of 0.005 to 0.03, with a pronounced peak at 0.01, whereas diploid neurons had a flatter distribution. In the broader analysis, the proportion of somatic variants rose by 45% in tetraploid neurons relative to control neurons.

Mutation type provided another striking clue. Tetraploid neurons were strongly enriched for thymine-to-cytosine transitions, written T>C, which are equivalent to adenine-to-guanine changes on the complementary DNA strand. This class accounted for roughly 64% of variants unique to tetraploid neurons, compared with about 37% of variants shared by diploid and tetraploid populations and 34% of variants unique to diploid neurons. Such transitions have been associated with oxidative DNA damage, including damage linked to lipid peroxidation. The T>C enrichment was also concentrated among variants below an allele frequency of 0.05, indicating that it was especially pronounced in the ultra-low-frequency population enriched in Alzheimer’s tetraploid cells.

The authors also compared the variant profiles with established single-base substitution signatures derived from large cancer-sequencing datasets. SBS123 made the largest contribution in most samples, but its relative share was higher in the tetraploid Alzheimer’s group, reaching 64% to 66% of fitted mutations, compared with 49% to 57% in controls and 52% to 59% in diploid Alzheimer’s neurons. The two tetraploid samples also shared a secondary combination of SBS116 and SBS5 that was not consistently present in the diploid groups. Cosine similarities between observed and reconstructed spectra ranged from 0.938 to 0.958. These analyses support a shift in the mixture of mutational processes associated with tetraploidy, rather than the appearance of a wholly tetraploid-specific signature.

Gene-level analysis showed that Alzheimer’s tetraploid neurons accumulated especially many sSNVs in pathways governing oxidative stress and DNA repair. Total SNVs increased by 38% in genes associated with antioxidant activity, 31% in genes involved in the response to reactive oxygen species, 29% in genes linked to hydrogen peroxide activity and 26% in genes associated with oxidative stress. Somatic variants in antioxidant and reactive-oxygen-response genes increased by 64% and 61%, respectively, both larger than the 45% genome-wide rise. Among DNA-repair genes, total SNVs increased by 31% and sSNVs by 64% in tetraploid cells. Base-excision, mismatch and nucleotide-excision repair genes showed significant sSNV increases, whereas double-strand-break-repair genes did not.

Not every disease-related category behaved as expected. Somatic variants also accumulated in genes associated with Parkinson’s disease, including a 64% increase over the genome-wide value, while sSNVs in Alzheimer’s risk genes rose by 59%. The pattern varied among individual genes, with significant increases in ABCA7, APOE, BIN1, CD33, EPHA1, MAPT, PSEN2, SMARCA4 and PLD3, but decreases in CD2AP and PSEN1. The study therefore found no simple, exclusive association between mutations in Alzheimer’s genes and the tetraploid state. A computational molecular pathogenicity analysis instead suggested that the mutations affecting genes in cancer and senescence pathways were more likely than random variants to be predicted to damage protein function.

That pathogenicity analysis used a score ranging from zero for the least pathogenic prediction to five for the most pathogenic and combined outputs from 12 computational tools. Cancer-related and senescence-related sSNVs had significantly steeper score distributions than generic genes, as did variants in several oxidative-stress and DNA-repair categories. Alzheimer’s and Parkinson’s risk genes did not show higher scores than control genes. These are computational predictions, not measurements of protein activity. The study did not perform functional assays, matched RNA sequencing or targeted validation of individual variants, so the data cannot establish that the mutations impaired DNA repair, altered gene expression or drove cell-cycle re-entry.

A central technical concern is that the neuronal DNA underwent whole-genome amplification before sequencing. PicoPLEX can introduce allelic imbalance, allelic or locus dropout and polymerase-derived errors, all of which may alter the frequencies used to distinguish somatic from germline variants. The distinct T>C spectrum in tetraploid-specific variants and its resemblance to independently established mutational signatures argue against a uniform amplification bias fully explaining the population-level pattern. They do not, however, rule out stochastic errors affecting individual low-frequency calls. The investigators also used a non-parametric bootstrap with 5,000 resamples to assess uncertainty in group means and pairwise differences; its confidence intervals were consistent with the original parametric analyses, but it did not increase the study’s effective sample size or statistical power.

The results are compatible with a self-reinforcing model of Alzheimer’s-associated neuronal tetraploidy. Oxidative stress could damage DNA, including genes involved in antioxidant responses and DNA repair; mutations in those genes could then impair genome maintenance and favor aberrant cell-cycle activity and genome duplication. Once tetraploidy was established, increased genomic stress could further elevate somatic mutation burden. Conversely, severe Alzheimer’s pathology could cause cellular stress and genome instability in already-tetraploid neurons, producing the same pattern. Because all three Alzheimer’s brains were at Braak stage V/VI and no early or mild cases were examined, the cross-sectional data cannot determine temporal order or causality. The clearest conclusion is therefore that Alzheimer’s-associated tetraploid neurons carry a specific enrichment of sSNVs with an oxidative-associated T>C pattern, particularly in oxidative-stress and DNA-repair genes, that merits validation in larger cohorts and with sequencing methods capable of confirming rare variants without amplification.

Subject of Research: Somatic single nucleotide variants in Alzheimer’s-associated tetraploid neurons

Article Title: Specific enrichment of somatic single nucleotide variants in Alzheimer-associated tetraploid neurons

Article References: López-Sánchez, N., Rábano, A., & Frade, J. M. (2026). Specific enrichment of somatic single nucleotide variants in Alzheimer-associated tetraploid neurons. Frontiers in Molecular Neuroscience, 19, Article 1929505. https://doi.org/10.3389/fnmol.2026.1929505

Image Credits: AI Generated

DOI: 10.3389/fnmol.2026.1929505

Keywords: Alzheimer’s disease, neuronal tetraploidy, somatic mutations, single nucleotide variants, whole-exome sequencing, oxidative stress, DNA repair, mutational signatures, post-mortem brain, genome instability, Specific, enrichment