Trio exome with two candidate variants: which one is causal?
A 4-year-old boy has infantile-onset seizures, burst-suppression on EEG and global developmental delay. His parents are unaffected and unrelated. Trio exome sequencing reports two heterozygous variants in the same gene, which encodes a neuronal voltage-gated potassium channel subunit, and the laboratory cannot decide which to pursue.
Variant A: c.1102C>T, p.Arg368Ter, in the final exon, 14 codons from the stop codon. Inherited from the unaffected mother. gnomAD allele frequency 0.6% overall, 0.9% in East Asians. Three unaffected adult relatives carry it.
Variant B: c.845G>A, p.Gly282Asp, in the S4 transmembrane segment. Absent from both parents' blood DNA by Sanger sequencing and absent from gnomAD (0 of 118,000 alleles). Gly282 is identical from zebrafish to human. Four in silico predictors call it damaging. No functional data exist.
The referring clinician suggests the two variants could be "two hits" in a recessive disorder. The family asks about future pregnancies; they have an unaffected 2-year-old daughter.
1. Which variant should be pursued as the cause, and why? Address the two-hit hypothesis directly.
Read the worked answer
Variant B is the candidate; variant A is very unlikely to be causal.
Variant A fails on frequency first. An allele frequency of 0.6% means roughly 1 in 167 alleles in the general population. A severe, infantile-onset, fully penetrant dominant disorder cannot be carried at that frequency, and the variant also sits in three unaffected adult relatives. Its position argues the same way: a stop 14 codons from the natural terminus, in the final exon, removes almost nothing and will not trigger nonsense-mediated decay, because a premature termination codon must lie well upstream of the last exon-exon junction to be recognised. It looks like a rare benign or hypomorphic variant.
Variant B carries the strongest single piece of evidence available in rare disease: de novo occurrence in a child with a severe, early phenotype and unaffected parents. Add absence from 118,000 population alleles, a conserved residue in the S4 voltage sensor - a position where substitutions plausibly alter gating - and concordant in silico prediction. In silico tools are supporting evidence only; they cannot make a variant pathogenic on their own.
The two-hit hypothesis does not fit. Biallelic recessive disease requires the parents to be carriers, whereas here one variant is de novo. Also, the two variants must be shown to be in trans before any recessive model is even arithmetically possible; if they sit in cis, one allele carries both lesions and the other is normal. Phasing should be done regardless, because it is cheap and it closes off the question.
2. What confirmatory work would you require before signing this out as likely pathogenic, and what pitfalls would you actively check for?
Read the worked answer
First, confirm the de novo call independently. Sanger sequencing of a trio detects mosaicism only down to roughly 15–20%. Amplicon deep sequencing at high depth should be run on both parents to look for low-level mosaicism, because parental mosaicism changes both the interpretation and the counselling. If the first result is borderline, test a second tissue - buccal, skin fibroblasts, or paternal sperm - since blood alone can miss tissue-restricted mosaicism.
Second, check identity and maternity/paternity formally. A false de novo call is one of the classic errors in trio analysis and is usually caused by sample swap, mislabelled tubes or non-paternity rather than biology. The exome data already contain the markers to do this cheaply.
Third, exclude a second hit or a missed structural event: deletion/duplication analysis of the same gene, and a check for a second de novo variant elsewhere that better fits the phenotype. Exomes miss CNVs and deep intronic or repeat variants.
Fourth, pursue function. Patch-clamp recording of the mutant channel in heterologous cells answers the question the clinic actually needs: loss of function versus gain of function. That distinction changes treatment logic, since some channelopathies are treated by blocking the channel and others by activating it. Until then the variant is a strong candidate, not a proven cause, and the report should say so.
3. What recurrence risk should the family be given, and how does the answer change if mosaicism is found?
Read the worked answer
If variant B is confirmed de novo in the child's blood and no mosaicism is detected in either parent's blood or gonads, the empiric recurrence risk is about 1%. That figure exists because gonadal mosaicism is documented in a minority of families with apparently de novo variants of this kind - a mutation present in a fraction of germ cells is invisible in blood and can recur. It is an empiric estimate, not a mechanistic certainty, and the family should hear it described that way.
If deep sequencing detects, say, 3% mutant reads in the father's sperm, the recurrence risk rises to approximately that fraction per pregnancy, and prenatal testing or preimplantation genetic testing can be offered. The wording matters: a variant is de novo in the tissue tested, not de novo in the person.
The unaffected 2-year-old daughter is at population risk unless a parent is mosaic, but the practical way to address her is to test her for variant B if the parents want certainty - one targeted test is cheaper and more reassuring than a risk calculation. All counselling should be explicit about what remains unknown: long-term outcome is inferred from the phenotype, not from the variant alone, and no functional data yet exist.