Mednotes

L2 Genetics Genetics of autism

Properties2
tagsgenetics, lecture
aliasesL2 Genetics Genetics of autism

15 min read

Overview

This lecture works through the genetic evaluation of a family with a child with Autism Spectrum Disorder (ASD), moving through a fixed framework of approaches: recurrence risk, pedigree analysis, monogenic syndromic disorders, chromosomal microarray, exome/genome sequencing, and common variation. The throughline is that ASD is causally heterogeneous: no single test currently gives families a definitive causal answer, and most of the genetic liability is polygenic rather than attributable to one identifiable factor.

What is ASD? Definition and clinical spectrum

  • Autism Spectrum Disorder (ASD) has broad and narrow definitions, characterised by three domains:
    • Social communication deficits (gaze, play, affection, connection)
    • Restrictive repetitive behaviours (routine, play, domestic rituals, roles)
    • Language deficits (modulation in social contexts, reciprocity)
  • Prevalence: 1-2% of the population.
  • The spectrum (broad to narrow) nests: Asperger’s disorder, Autistic disorder, Childhood disintegrative disorder, Pervasive developmental disorder not otherwise specified (Devlin and Scherer, 2012).
  • Comorbidity ranges across the spectrum: anxiety/ADHD 30-50%, cognition deficits 40-60%, medical comorbidities 25-40%.
  • Severe (classical) autism: frequent association with intellectual disability, medical comorbidities frequent.
  • Heritability of broad-definition autism: ~90% in twin concordance studies (twin studies as an instrument for heritability carry caveats); more modern genome array-based estimates ~60%.

Genetic work-up framework

The lecture repeats one framework for systematically evaluating a family with a child with ASD:

  • Empirical recurrence risks
  • Pedigree analysis
  • Detection of an underlying single gene disorder
  • Chromosomal analysis / comparative genomic hybridisation (microarray)
  • Exome/genome sequencing
  • Predictive risk scores / common variation

The two questions families ask are: what caused it, and will it happen again? The key to diagnosis is recognising that autism (a) carries substantial genetic underpinnings and (b) is causally heterogeneous.

Recurrence risk and pedigree analysis

  • Recurrence risk for all types of ASD: c. 1/6-1/4 (15-25%), not as neat as a Mendelian disease.
  • Heritability for strict autism (without intellectual disability or other comorbidities) is lower, ~40%.
  • Male:female ratio is 4:1 (low female:male ratio).
  • Bottom line: recurrence risks are high, and families need clear answers to guide reproductive choices.
  • A pedigree of two unaffected parents with one affected child illustrates a sporadic (de novo) presentation of an autosomal dominant disorder: the mutation arose de novo in the child rather than being inherited.

Monogenic syndromic causes: Fragile X and tuberous sclerosis

  • Monogenic syndromic associations are sometimes said to explain up to 10% of ASD, but this is considered an unlikely, over-stated figure; a more realistic estimate is 3-4%.
  • Two commonly cited examples: Fragile X syndrome and Tuberous sclerosis.

Fragile X syndrome

  • X-linked: males affected; females can be non-penetrant or show variable expressivity.
  • Caused by a triplet (CGG) repeat expansion in FMR1 (Xq27); routine clinical test available.
  • CGG repeat classification: 5-44 = normal, 45-54 = intermediate, 55-200 = premutation, >200 = pathologic.
  • Full expansions cause moderate-severe intellectual disability in males; some ASD-like behaviours are noted.
  • Pedigrees show variable penetrance/expressivity across generations.

Tuberous sclerosis

  • Autosomal dominant; 1 in 6,000 newborns.
  • Caused by loss of both alleles at either of two genes, TS1 and TS2 (regulate tissue growth); large genes with many variants, detected by sequencing.
  • Intracerebral space-occupying lesions (“tubers”); other manifestations in heart, skin, kidneys; very variable neurological presentation.
  • Readily diagnosed clinically by an experienced practitioner; ASD reasonably common.
  • Organ-system manifestations: brain 90% epilepsy, 80-90% subependymal nodules (SEN), 10-15% subependymal giant cell astrocytoma (SEGA), 90% TAND, 50% intellectual disability, 40% ASD; other 50% oral fibromas, 50% retinal astrocytic hamartomas; heart 90% cardiac rhabdomyoma in infants (20% in adults); lung (women) 80% asymptomatic LAM, 5-10% symptomatic LAM (can cause respiratory failure), 10% MMPH in men and women; kidney 70% angiomyolipoma, 35% simple multiple cysts, 5% polycystic kidney disease, 2-3% renal cell carcinoma; skin 75% angiofibroma, 20-80% ungual fibroma, 25% fibrous cephalic plaques, >50% shagreen patches, 90% focal hypopigmentation.
  • Diagnostic criteria: major features include hypomelanotic macules (>3, ≥5mm), angiofibromas (>3) or fibrous cephalic plaque, ungual fibromas (>2), shagreen patch, multiple retinal hamartomas, cortical dysplasias, subependymal nodules, subependymal giant cell astrocytoma, cardiac rhabdomyoma, lymphangioleiomyomatosis (LAM), angiomyolipomas (>2). Minor features include confetti skin lesions, dental enamel pits (>3), intraoral fibromas (>2), retinal achromatic patch, multiple renal cysts, nonrenal hamartomas. Definite diagnosis = 2 major features, or 1 major + ≥2 minor. Possible diagnosis = 1 major, or ≥2 minor. LAM and angiomyolipomas together, without other features, do not meet criteria for a definite diagnosis.
  • Mutations found via sequencing may be de novo or inherited (from penetrant or non-penetrant relatives); both loci and all mutation forms must be considered (deletions - whole gene or intragenic - and point mutations).
  • Key principle: beware exclusionary thinking - a negative genetic sequencing test very often is NOT exclusionary, because large genes with diverse variant types can be missed.

Chromosomal anomalies and CNVs

  • Chromosomal microarray can detect large chromosomal anomalies and microdeletions/duplications (example: a deletion at the tip of chromosome 1).
  • Chromosomal anomalies account for 1-2% of autism; they are frequently private, produce unique patterns of anomalies (neurological sequelae very frequent), and other medical comorbidities are common.
  • Can be recurrent, e.g. maternal duplication of 15q11-13, which spans genes including TUBGCP5, CYFIP1, NIPA2, NIPA1, MAGEL2, SNURF/SNRPN, UBE3A, ATP10A, GABRB3, GABRA5, GABRG3, OCA2.
  • 15q11-q13 is an imprinted locus: gene function differs between maternally and paternally derived alleles, due to parent-specific methylation silencing critical genes. Maternal derivation of a duplication of this region leads to autism; which specific genes drive this is not fully resolved.
  • CNV burden data (Scherer et al 2010): deletions and duplications are found with increased frequency in ASD cases, and cases are more likely to have multiple and larger rearrangements than controls. Overall CNV presence case:control ratio 1.19 (p=0.012); deletions-only ratio 1.26 (p=0.008); duplications-only ratio 1.16 (not significant). Larger CNVs (≥500kb) are especially enriched: all ≥500kb ratio 1.69 (p=0.005); duplications-only ≥500kb ratio 1.82 (p=0.007).
  • Interpretation problems: CNVs with apparent causal relevance are often inherited from healthy (non-penetrant) parents, and some are variable in expression, e.g. 16p12.2 deletion/duplication is “frequent” (0.6% of ASD cases) with variable comorbidities (developmental delay ± obesity, non-ASD psychiatric disorders, or normality). Duplications are less penetrant than deletions in general.
  • Recurring CNV regions remaining significant after correction include 16p11.2 deletion and duplication, 15q13.2-3 deletion, 15q11.2-q13 duplication, 7q11.23 duplication, and 17q12 deletion; recurrent de novo events were also strongly implicated at 16p11.2 deletion (0.37% of probands), 15q11.2 duplication (0.18%), 15q11.2-13.1 duplication (0.16%), 15q13.2-13.3 duplication (0.13%), 16p11.2 duplication (0.13%), and 7q11.23 duplication (0.09%).
  • Evolutionary aside: some loci linked to increased brain volume across primate evolution are also ASD loci, often driven by gene duplication (e.g. ARHGAP11B at 15q13, linked to neocortex expansion), raising the question of whether ASD reflects an evolutionary trade-off.
  • Take-home for chromosomal microarray: small imbalances are resolvable; background CNVs are present in healthy people (genuinely benign); non-penetrance for potentially pathogenic CNVs is common; a single unitary anomaly is seldom the cause. The “mutational load” concept: many independent genetic factors must be co-inherited before the ASD phenotype manifests.

Single-gene / exome sequencing and the mutational burden

  • Single gene mutations are “the difficult end of the spectrum”: variants in some genes are identified recurrently and are risk-associated, and can be tracked in single genes across a few families.
  • Whole exome/genome sequencing studies now find an excess of nonsense mutations in cases vs controls (2-4x more) - the mutational load concept again.
  • The most rigorous causal criteria are de novo, non-synonymous changes predicted to alter protein function, but these criteria are almost certainly overly stringent.
  • Identified genes are likely the tip of the iceberg. Genes implicated cluster in synaptic function and chromatin remodelling and confer independent risk for ASD. Rare variants confer significant personal liability but are not common enough to explain most of the heritability.
  • Protein-protein interaction network of genes disrupted by de novo mutations in ASD clusters into three functional groups: synaptic function (e.g. GRIN2B, SYNGAP1, DLG4, SHANK2, CASK, NRXN1, RIMS1, RBFOX1), Wnt signalling (e.g. PSEN1, CTNNB1, CHD8), and chromatin remodelling (e.g. ADNP, SMARCC4, ARID1B, MLL3, MLL5). CHD8 is the most recurrently implicated gene (largest node, truncating mutations).
  • Mutational burden study (O’Roak et al, Nature 2012): trio (parents + child) exome analysis of 209 families found 126 severe, disruptive de novo mutations (33 truncating), about a third localising to a single protein-protein interaction network. Recurrent mutations were found in only two genes (two instances each); targeted sequencing of a further 1,700 cases found mutations in three of the genes identified.

    The lecture's quoted estimate from this paper is cut off mid-sentence on the slide: "...if 20-30% of our de novo point mutations are considered to be pathogenic, we can estimate between 384 and 821 loci..." The conclusion of this estimate is not given in the source material.

    • The findings point to extreme locus heterogeneity, mutational heterogeneity, and oligogenic inheritance.
    • Example case: a single individual with five separate gene disruptions (de novo truncating SNVs in CHD8 and CUBN, plus deletions of PITRM1, SPNS3, and RCAN1), illustrating how multiple independent hits can combine (“oligogenic”) in one person.

Common variation and the overall genetic architecture

  • Common variation accounts for most of the genetic signal in ASD. Overall: heritability 52% (inherited variants), environment 48%. Within genetics (62% of the outer bracket): inherited common variants 49.8% (each of >1000 alleles, individually very low risk); inherited rare variants 2.6% (rare inherited CNVs e.g. 1q21.1 dup, 15q11.2 del, 16p11.2/16p12.1 del, 16p13.11, 17q12 dup; rare inherited SNVs e.g. CNTN6, SHANK1, SHANK2, NRXN1); de novo variants 9.5% (de novo CNVs 2.9%, e.g. Williams-Beuren syndrome, Potocki-Lupski syndrome, Smith-Magenis syndrome, 22q13 del, 9q34 del, 16p11.2, 17q21.31 del; de novo SNVs/LGD 6.6%, e.g. SHANK3, CHD8, DYRK1A, GRIN2B, KATNAL2, RIMS1, SCN2A, POGZ, ADNP, ARID1B, TBR1). Environment contributes 38.1%.
  • The “green zone” (common variants and more) is hard to measure clinically because: a polygenic model with thousands of variants explains most complex traits like autism but is difficult to measure; most variants are of tiny individual effect; epigenetic variation is hard to capture clinically; mosaicism for de novo CNVs and other variants is poorly detected in blood.
  • Torre Ubieta et al (Nat Med, 2016) breakdown of ASD liability: strongly implicated de novo SNVs together account for 1.34% (individual genes each 0.08-0.21%, e.g. CHD8 0.21%); CNVs together 1.28% (e.g. NRXN1 del 0.32%, 16p11.2 del 0.31%); syndromic causes together 3.40% (Fragile X 1.94%, tuberous sclerosis 0.90%, Phelan-McDermid 0.28%, neurofibromatosis type 1 0.28%). Overall liability breakdown: unaccounted 41%, predicted common inherited 49%, non-additive 4%, rare inherited 3%, de novo 3%.
  • Autism is a disease with a strong genetic basis; common inherited factors account for most of this liability; only a small minority of ASD patients have an identifiable genetic factor of major effect. Much of the “gap” in explaining ASD’s genetics is attributable to common variation, though a large proportion remains unaccounted for.

Clinical work-up and limits of testing

  • Working up autism in practice: clinical assessment, plus chromosomal microarray for those with intellectual disability or other comorbidities.
  • Families often want a test with high positive predictive value, particularly for prenatal diagnosis.
  • The heterogeneity and complex (polygenic/oligogenic) architecture of ASD may preclude such a definitive test in the future, even once technology is optimised.

Self-test

  1. Define Autism Spectrum Disorder using its three core domains, and state its population prevalence.
  2. What is the recurrence risk range quoted for a sibling of a child with ASD, and how does the male:female ratio feature in the discussion of recurrence?
  3. Give the heritability estimates for broad-definition autism from twin studies vs modern genome array studies, and explain why “strict” autism has a lower heritability estimate.
  4. List the framework of approaches used to systematically evaluate a family with a child with ASD.
  5. Describe what the pedigree of two unaffected parents with one affected child illustrates about autosomal dominant conditions.
  6. Distinguish Fragile X syndrome from tuberous sclerosis in terms of inheritance pattern and underlying mutation type.
  7. Describe the FMR1 CGG repeat classification and its four category thresholds.
  8. What proportion of ASD is estimated to be explained by monogenic syndromic causes, and why is the commonly cited “10%” figure considered an overestimate?
  9. State the criteria for a definite clinical diagnosis of tuberous sclerosis.
  10. Explain the “beware exclusionary thinking” principle in genetic testing for tuberous sclerosis.
  11. What proportion of autism is attributable to chromosomal anomalies detectable by microarray, and what key limitation applies to interpreting a CNV found in an affected child?
  12. Explain how parent-of-origin (imprinting) affects the pathogenicity of a 15q11-13 duplication.
  13. Define the “mutational load” concept as it applies to chromosomal microarray findings in ASD.
  14. Describe the mutational burden findings from the O’Roak et al. trio exome sequencing study (number of families, de novo mutations found, and recurrence across genes).
  15. What is meant by “oligogenic inheritance”, and how does the five-gene-disruption case example illustrate it?
  16. List the three functional gene clusters into which ASD-associated de novo mutations group, and name the gene most recurrently implicated.
  17. According to the Torre Ubieta et al (2016) liability breakdown, what proportion of ASD liability is “unaccounted”, and what proportion is attributed to predicted common inherited variation?
  18. Why do families often want a highly predictive genetic test for ASD, and why might ASD’s genetic architecture make this difficult to achieve even with optimised technology?
  19. A family whose child has classical (severe) autism with intellectual disability asks for a single genetic test that will tell them the cause and the recurrence risk. Using what you know of the genetic architecture of ASD, explain why no single test can currently give a definitive answer.

Answers

Graph View

Backlinks