Large mosaic copy number variations confer autism risk

Maxwell A. Sherman, Rachel E. Rodin, Giulio Genovese, Caroline Dias, Alison R. Barton, Ronen E. Mukamel, Bonnie Berger, Peter J. Park, Christopher A. Walsh, Po‐Ru Loh

TL;DR

This study reveals that large "mosaic" copy number variants (mCNVs)—mutations that occur after fertilization and affect only a subset of the body's cells—are a significant and previously overlooked risk factor for Autism Spectrum Disorder (ASD). By analyzing over 17,000 individuals, researchers found that large mCNVs (>4 Mb) are more than 11 times more frequent in children with autism than in their unaffected siblings, often involving massive chromosomal rearrangements that would likely be lethal if present in every cell.

What Problem Does This Paper Address?

While genetics plays a massive role in autism, a large portion of ASD cases remain unexplained by known inherited or de novo germline mutations (mutations present in the sperm or egg). Scientists have long suspected that "mosaicism"—mutations that arise during early embryonic development—might fill this gap.

While previous studies have linked mosaic single-letter DNA changes (SNVs) to autism, the role of mosaic Copy Number Variants (mCNVs)—large-scale deletions or duplications of DNA chunks—remained largely unexplored. The challenge lies in detection: because these mutations only exist in a fraction of cells, they produce very faint signals in standard genetic tests, making them difficult to distinguish from experimental noise.

Methods

The authors employed a sophisticated computational pipeline to "sift" through genotype array data:

  • MoChA (Mosaic Chromosomal Alterations): They used this specialized caller to analyze two key signals: Log-R Ratio (LRR), which measures total DNA intensity (detecting gains or losses), and B-allele frequency (BAF), which detects imbalances between maternal and paternal DNA contributions.
  • Phased Haplotypes: To increase sensitivity, they used "phasing"—determining which genetic markers sit on the same physical chromosome. By using parental data, they achieved "near-perfect" long-range phasing, allowing them to detect mutations even when they were present in as few as 3% of the sampled cells.
  • Experimental Validation: To prove these mutations weren't just artifacts of blood or saliva samples, they performed whole-genome sequencing and digital droplet PCR (ddPCR) on post-mortem brain tissue to confirm the presence of mCNVs in actual neurons.

Datasets

The study utilized two massive, well-characterized autism cohorts:

  1. Simons Simplex Collection (SSC): 2,591 families.
  2. Simons Powering Autism Research for Knowledge (SPARK): 8,866 families.
  3. Total Sample: 12,077 ASD probands (affected individuals) and 5,500 unaffected siblings.
  4. Comparison Group: Data from 454,993 individuals in the UK Biobank was used to compare the effects of mosaic vs. germline mutations in the general population.
  5. Validation Set: Post-mortem brain tissue from 60 additional individuals.

Key Insights and Ideas

  • The "Lethality" Hypothesis: Mosaic CNVs in autism are significantly larger (median 7.8 Mb) than typical germline CNVs (median 0.92 Mb). The authors hypothesize that these massive mutations are only "survivable" because they are mosaic. If they were present in every cell (germline), they would likely result in miscarriage.
  • Size Over Cell Fraction: Interestingly, the size of the mutation correlated with the severity of autism symptoms, but the percentage of cells affected did not. This suggests that even a small fraction of cells carrying a massive genetic error can disrupt brain development.
  • Different Genomic "Hotspots": Surprisingly, the researchers did not find mosaic versions of the "classic" autism mutations (like the 16p11.2 deletion). Instead, mosaic risk comes from unique, massive events across the genome, suggesting mosaicism and germline mutations contribute to ASD through different biological pathways.
  • Compound Heterozygosity: In one case, a child had a mosaic deletion on one chromosome and an inherited rare variant on the other for the same gene (NRXN1), essentially "knocking out" the gene's function in a subset of cells.

Key Results

  • Significant Burden: Large mCNVs (>4 Mb) were found in 25 probands but only 1 sibling, resulting in an Odds Ratio of 11.4.
  • Clinical Correlation: Increased mCNV size was significantly associated with higher scores on the Social Communication Questionnaire (SCQ), a measure of ASD severity (P = 0.016).
  • Prevalence: The authors estimate that these large mosaic events explain approximately 0.2% of all ASD cases.
  • Brain Confirmation: The team successfully validated a complex 10.3 Mb duplication in the post-mortem brain of an individual with ASD, proving these mutations propagate into the central nervous system.

Limitations and Future Work

  • Rarity: Despite the large dataset, mCNVs are still rare. This limited the authors' ability to find "recurrent" mutations (the same spot hitting multiple people) or to fully parse how different cell fractions affect the phenotype.
  • Tissue Proxy: Most of the study relied on DNA from blood or saliva. While the authors argue high-cell-fraction mutations are likely systemic, future work will need to look more directly at brain tissue across larger cohorts.
  • Mechanism of Origin: The study identifies that these mutations happen, but not why. Future research into the environmental or genetic factors that predispose an embryo to mosaic chromosomal instability is needed.

Summary updated

Abstract

Abstract Although germline de novo copy number variants are a known cause of autism spectrum disorder (ASD), the contribution of mosaic (early-developmental) copy number variants (mCNVs) has not been explored. Here, we assessed the contribution of mCNVs to ASD by ascertaining mCNVs in genotype array intensity data from 12,077 ASD probands and 5,500 unaffected siblings in the Simons Simplex Collection (SSC) and Simons Powering Autism Research for Knowledge (SPARK) cohorts. We detected 46 mCNVs in probands and 19 mCNVs in siblings ranging from 49 kb to 249 Mb and affecting 2.8-73.8% of cells. In both cohorts, probands carried a significant burden of large (>4 Mb) mCNVs (P = 0.043 and P = 6.6 × 10 −3 in SSC and SPARK, respectively), which were present in a total of 25 probands but only 1 sibling (OR=11.4, 95% CI=1.5-84.2). Surprisingly, we did not observe mosaic analogues of the short de novo CNVs recurrently observed in ASD. Event size positively correlated with severity of ASD symptoms (P = 0.016), and four probands exhibited clinical symptoms consistent with syndromes previously associated with genes or regions disrupted by their respective mosaic mutations. In analyses of post-mortem brain tissue from 60 additional probands, we further detected and experimentally validated two mCNVs including a complex 10.3 Mb duplication on chromosome 2. These results indicate that mosaic CNVs contribute a previously unexplained component of ASD risk.
BibTeX
@inproceedings{sherman2021large,
  author = {Maxwell A. Sherman and Rachel E. Rodin and Giulio Genovese and Caroline Dias and Alison R. Barton and Ronen E. Mukamel and Bonnie Berger and Peter J. Park and Christopher A. Walsh and Po‐Ru Loh},
  title = {{Large mosaic copy number variations confer autism risk}},
  booktitle = {Nature Neuroscience},
  year = {2021},
  doi = {10.1038/s41593-020-00766-5},
}

Loading related content...

Figure