Diverse somatic genomic alterations in single neurons in chronic traumatic encephalopathy

Abstract

Chronic traumatic encephalopathy (CTE) is a neurodegenerative disease linked to exposure to repetitive head impacts (RHI), yet little is known about its pathogenesis. Applying two single-cell whole-genome sequencing methods to hundreds of neurons from prefrontal cortex of 15 individuals with CTE and 4 with RHI without CTE, we revealed increased somatic single-nucleotide variants in CTE, exhibiting a pattern previously reported in Alzheimer's disease (AD). Furthermore, we discovered high burdens of somatic small insertions and deletions in a subset of CTE individuals, resembling a known pattern, ID4, also found in AD. Our results suggest that neurons in CTE experience stereotyped mutational processes shared with AD; the absence of similar changes in RHI neurons without CTE suggests that CTE involves mechanisms beyond RHI alone.

Fig. 1.. Study design and sSNV burden in CTE neurons.

(A) Cohort design: neurotypical control, repetitive head impacts (RHI), chronic traumatic encephalopathy (CTE), and Alzheimer’s disease (AD). Illustrations of CTE and AD show characteristic tau pathology patterns (in brown). The number of cases included in this study are shown for each clinical condition. (B) Single-cell whole-genome sequencing (WGS) experimental approach. Nuclei are isolated from postmortem human brain tissue and subjected to fluorescence-activated nuclear sorting for the neuronal nuclear marker NeuN. Nuclei are sorted, lysed, and subjected to primary template amplification (PTA) and multiplexed end-tagging amplification of complementary strands (META-CS). Amplified genomic DNA is then assayed using WGS to identify somatic mutations. PTA data is used to determine somatic mutation burden, while META-CS data is used to identify strand-related signatures. (C) sSNV burden in CTE (red), RHI (light blue), and neurotypical control brains (dark blue) with a significant increase in CTE but not in RHI when compared to controls. sSNV burden (from each neuron as a point) estimated by SCAN2 is fitted against age by clinical conditions using LME models (neurotypical control: dark blue; RHI: light blue, P = 0.725; CTE: red, P = 0.003). P-values compare each clinical condition against controls. (D) Similar to (C) with added neurons from AD (green) brains (AD: green, P = 0.002 using the LME model). P-values compare each clinical condition against controls. (E) Excess sSNV burden in RHI, CTE, and AD compared to neurotypical control after adjusting for age. Data are mean ± standard error. The dashed blue line shows sSNVs attributable to age (zero excess). P-values are from two-tailed Wilcoxon tests. (F) Excess sSNVs in each CTE case ordered by increasing age (indicated in parentheses). The dashed blue line shows sSNVs attributable to age (zero excess). (G) Circos plot showing the PTA sSNV density distribution of CTE cases across the whole genome. Each CTE case is depicted by color in a circular track.

Fig. 2.. sSNV mutational signatures in CTE neurons.

(A, B) Mutation spectra of Signatures A and C. (C-F) sSNV burden separated by contributions from Signature A (C, D) and Signature C (E, F) in CTE (red), AD (green), and neurotypical control (dark blue) neurons. These signatures decompose sSNV burden into age- and disease-specific effects. Their contributions are fitted against age by clinical conditions using LME models (C, Signature A, CTE: P = 0.088; D, Signature A, CTE: P = 0.088, AD: P = 0.037; E, Signature C, CTE: P = 7.7 × 10⁻⁴; F, Signature C, CTE: P = 7.7 × 10⁻⁴, AD: P = 0.006). P-values compare each clinical condition against controls. (G) Ratios of each signature’s total contribution to age-related contribution in neurotypical control, CTE, and AD neurons. Age-contributed sSNVs of each signature are obtained from the LME model in neurotypical controls. A ratio > 1 indicates a higher contribution from the signature compared to age-matched controls. Bars in each box plot from top to bottom show the first, second (median), and third quartile; whiskers extend 1.5 interquartile range (IQR). (H) Relative contribution of Signatures A and C in CTE and AD after adjusting for age. Based on the ratios shown in (G), the relative contribution of each signature is calculated by removing the median ratio of neurotypical controls (representing age effect) from the median ratio of each disease. (I) Mutational spectra of neurotypical control, CTE, and AD neurons are fitted to the COSMIC SBS database of cancer mutational signatures. Residual mutational patterns for CTE and AD are obtained by subtracting mutation profiles of age-matched controls from those of CTE and AD to show disease-specific contributions.

Fig. 3.. sIndel burden and mutational signatures in CTE neurons.

sIndel burdens were identified from PTA scWGS data, with strand-related analysis inferred from mutational signatures extracted from META-CS scWGS data. (A) sIndel burden in CTE (red), RHI (light blue), and neurotypical control brains (dark blue) with a significant increase in CTE but not in RHI when compared to controls. sIndel burden (from each neuron as a triangle) estimated by SCAN2 is fitted against age by clinical conditions using LME models (neurotypical control: dark blue; RHI: light blue, P = 0.338; CTE: red, P = 0.004). P-values compare each clinical condition against controls. (B) Excess sIndel burden in RHI and CTE compared to neurotypical control after adjusting for age. Data are mean ± standard error. The dashed blue line shows sIndels attributable to age (zero excess). P-values are from two-tailed Wilcoxon tests. (C) Comparison of all types of sIndels across age-matched controls, RHI, and CTE. Data are mean burden per cell. Asterisk denotes significant changes in certain types of sIndels when compared to age-matched controls (P < 0.05, two-tailed Wilcoxon test). (D) Double-stranded (ds) and single-stranded (ss) Indel signatures in neurotypical controls and CTE extracted from META-CS data. (E) Absolute (top) and relative (bottom) contribution of dsIndel and ssIndel signatures in PTA-profiled neurotypical control and CTE neurons. The pair of dsIndel and ssIndel signatures used for decomposition is determined by the clinical condition of PTA neurons. CTE cases with a pronounced ssIndel contribution are indicated by brackets. Cells with < 15 Indels are not shown. (F) Decomposition of dsIndel and ssIndel signatures to COSMIC ID database.

Fig. 4.. Somatic ID4-like deletions in certain CTE individuals.

(A, B) Excess sIndels in each CTE (A) and AD (B) case ordered by increasing age (parentheses). CTE and AD cases are separated into either High-Indel group (High-Indel CTE: red, High-Indel AD: dark green) or Low-Indel group (Low-Indel CTE: yellow, Low-Indel AD: light green) based on whether they have an excess of sIndels (> 50). Data are mean ± standard error. The dashed blue line shows sIndels attributable to age (zero excess). (C) Principal component analysis (PCA) clustering of sIndels from each case across neurotypical controls, CTE, and AD. sIndels from each case were aggregated and stratified into 83 contexts defined by COSMIC. The first two dimensions are used for visualization with the percentage of variance explained shown for each dimension. Cases with < 15 sIndels are not shown. (D, E) Comparison of proportions of 2 to 4 bp deletions in dsIndels (D) and ssIndels (E) across age-matched controls, Low-Indel CTE, and High-Indel CTE. Each triangle represents an individual from META-CS data. High-Indel CTE vs. Age-matched controls: P = 1.0 × 10⁻⁴ and P = 7.5 × 10⁻⁴, two-tailed Wilcoxon test. Data are mean ± standard deviation. (F) Relative contribution of ID4 to dsIndels and ssIndels of each case from META-CS data shown as a heatmap. Case IDs are colored by their group assignment (age-matched control: dark blue, Low-Indel CTE: yellow, High-Indel CTE: red).

Fig. 5.. Enrichment analysis of sSNVs and sIndels in CTE neurons.

(A-D) sSNV (A, B) and sIndel (C, D) density as a function of gene expression and chromatin accessibility levels. Neuronal transcriptional profiles were characterized from snRNA-seq data sequenced in this study. Neuronal accessibility profiles were obtained from snATAC-seq data in Ganz et al. (23). Genes and open chromatin regions are separated into 10 (for sSNV) or 5 (for sIndel) equally sized groups with increasing levels indicating increasing expression or accessibility. Observed density of each expression or accessibility group is obtained by overlapping the original mutation call sets with regions of each group. Expected density is calculated from 1000 permutations of mutation call sets overlapped with regions of each group. Enrichment ratio is calculated by observed / expected density for each permutation, and mean ratio over 1000 permutations is used to construct the trend line by linear model (error bar indicates standard deviation). Pearson correlation coefficient (R) and two-tailed p-value (P) are shown. obs.: observed, exp.: expected. (E, F) Gene ontology (GO) analysis of genes where sSNVs (E) and sIndels (F) are located. GO terms with FDR-adjusted P < 0.01 in both CTE and control are reported.