Mapping recurrent mosaic copy number variation in human neurons

Abstract

When somatic cells acquire complex karyotypes, they often are removed by the immune system. Mutant somatic cells that evade immune surveillance can lead to cancer. Neurons with complex karyotypes arise during neurotypical brain development, but neurons are almost never the origin of brain cancers. Instead, somatic mutations in neurons can bring about neurodevelopmental disorders, and contribute to the polygenic landscape of neuropsychiatric and neurodegenerative disease. A subset of human neurons harbors idiosyncratic copy number variants (CNVs, "CNV neurons"), but previous analyses of CNV neurons are limited by relatively small sample sizes. Here, we develop an allele-based validation approach, SCOVAL, to corroborate or reject read-depth based CNV calls in single human neurons. We apply this approach to 2,125 frontal cortical neurons from a neurotypical human brain. SCOVAL identifies 226 CNV neurons, which include a subclass of 65 CNV neurons with highly aberrant karyotypes containing whole or substantial losses on multiple chromosomes. Moreover, we find that CNV location appears to be nonrandom. Recurrent regions of neuronal genome rearrangement contain fewer, but longer, genes.

Fig. 1. SCOVAL: identification of copy number variation using read-depth and allele imbalance.

Overview of SCOVAL. A Single nuclei and bulk dural fibroblast DNA were analyzed using 10X platforms. (Images from vecteezy.com) B Single nuclei library quality is assessed based on median absolute deviation (MAD) and copy number thresholds are established using population statistics. Graphs depict schematized data; vertical red lines illustrate threshold strategy. C Candidate CNVs are identified based on altered read depth across consecutive genomic bins. D Heterozygous SNPs are phased using bulk linked-reads in chromosomal segments (“hap 1” or “hap 2”). E Absolute log2 ratios derived from “hap 1”/“hap 2” are calculated across ~100 SNP windows (see text). A deletion with concordant loss of heterozygosity (log2 ratio <> 0) is illustrated. F A highly aberrant CNV neuron (#5) shows representative Gingko calls (blue bars), duplications (e.g., green arrow), heterozygous deletions (e.g., black arrow), and homozygous deletions (e.g., orange arrow) and qualitatively concordant increases in absolute log2 ratio (white<purple). The genome is plotted from left to right on the x-axis, read-depth is in the upper panel (CN state on the Y-axis), and absolute log2 ratios are reported in the lower panel.

Fig. 2. CNV neurons can have highly aberrant karyotypes.

A The observed CNV per neuron [(purple bars, counts (y-axis),CNVs/neuron (x-axis)] distribution deviates (P < 0.0001) from Poisson expectations (dashed blue line). Arrows indicate neurons with monosomic chromosomes. B, C Deletions cluster in a subset of CNV neurons. B Counts (y-axis) of the cumulative percent of each chromosome deleted (n = 2097 neurons * 22 autosomes) in CNV neurons. C Neuronal genomes (n = 2097) are arranged in a cells-by-chromosome matrix, ranked by the total percentage of their genome containing deletions. Cell #226 is the first CNV neuron among 2097 total neurons with the smallest observed single deletion (blue = unaffected chromosome, yellow <50%, orange = 50−99%, red 100%). D–F Among 65 neurons with the most aberrant genomes, some have similar karyotypes. D Hierarchical clustering identifies two groups (yellow, red) with the least divergence from similarity (y-axis). E Red cluster neurons [cells #32, 33, and 47 in (C)] have similar CNV profiles. Read-depth is plotted as in Fig. 1F. The yellow cluster (cells #17 and #19) is shown in Fig. S6B. F Concordant read-depth is observed on opposite haplotypes in the most similar pair [#32(red) and #33(blue)]. When overlapping, events on cell #47 (green) match the #32 haplotype, but never the #33 haplotype. Chromosome 3 is plotted from left to right. Haplotype log2 ratio (upper panel) and corresponding read-depth (lower panel, blue = diploid) plots show overlapping deletions and LOH for each haplotype.

Fig. 3. Analysis of CNV distribution relative to random null model.

A Empirical read-depth plots of two CNV neurons (left panels) and representative permutations (right two panels) are displayed as in Fig. 1F. B Relative to 10,000 permutations of real data (represented by blue dotted line and error bars), high and low CNV burden are enriched at the extremities of the Gaussian distribution (green bars). C Circos plot shows that hotspots (red, outer tier) and cold spots (blue, outer tier) cluster on distinct chromosomes. Thirty-three pathogenic CNVs (blue, purple, inner tier) never overlap hotspots. Eleven (blue) overlap cold spots. D Violin plot (mean +/- SD) showing gene enrichment in cold spots (left, N = 56, mean gene hits = 57.71 ± 58.93) and depletion in hotspots (right, N = 83, mean gene hits = 15.65 ± 32.40) relative to other 5 Mb regions (N = 404 controls, mean gene hits 37.79 ± 43.71) with chi-square P < 0.001 for cold spots vs. controls and hotspots vs. controls.

Fig. 4. Recurrent CNV breakpoints across multiple neurons.

A UCSC Genome Browser view of all CNVs detected on Chromosome 1 (47 neurons, rows). Seven neurons (red) contain CNVs that share a breakpoint region (CNVB). B Representative CNVB (red) on Chromosome 1 overlaps (±250 kb) two genes (lower panel). C Number of breakpoints identified in each Ginkgo bin (y-axis) relative to bin size (x-axis), shown for bins containing two or more CNVs (red) and averaged across all permutations in control set (blue line). D–F Violin plots show real and permuted datasets, normalized by bin size, when examined for D number of breakpoints, E number of long (>100k) genes (****p < 0.0001 for one-tailed t-test), and F transcripts per million bp (TPM) values of the longest gene in each bin.