Three Variants, One Patient: Deciding Which Change Is Pathogenic
A 12-year-old boy is referred with LDL-cholesterol of 9.8 mmol dm⁻³ and xanthomas on the Achilles tendons; his father had a myocardial infarction at 42. The mother's LDL is 3.1 mmol dm⁻³; the father's is 8.4 mmol dm⁻³ and he takes a statin. Sequencing of the LDLR gene (18 exons) in the boy finds three variants: (1) a four-base deletion in exon 7 removing bases 1054–1057; (2) a G>A change in exon 9 that replaces cysteine 396 with tyrosine, in a cysteine-rich repeat whose sequence is identical in mouse, rat and zebrafish; (3) a T>C change 34 bases into intron 14, far from either splice junction. The boy is heterozygous for (1) and (2); the father carries (1) and the mother carries (2). The laboratory classifies variant (2) as a 'variant of uncertain significance'. The family asks which variant is causing the disease and whether the boy's 9-year-old sister should be tested.
1. Rank the three variants by how confidently they can be called pathogenic, justifying each ranking from the molecular consequences.
Read the worked answer
(1) Frameshift deletion - pathogenic, high confidence. Removing four bases from an exon is not a multiple of three, so every codon downstream is read in the wrong frame. The result is a run of incorrect amino acids followed almost always by a premature stop codon, giving a truncated receptor that cannot bind LDL, plus likely nonsense-mediated decay of the mRNA. Loss of one functional copy is sufficient to raise LDL substantially on a statin-free background.
(2) Cys396Tyr - probably pathogenic, but genuinely uncertain. Cysteine is uniquely able to form disulphide bonds, and the LDL receptor's ligand-binding domain is built from cysteine-rich repeats stabilised by disulphides. Replacing one cysteine with tyrosine removes a partner for a disulphide bond, so the repeat is likely to misfold and be retained in the endoplasmic reticulum. Conservation of that residue across mouse, rat and zebrafish argues that any change is not tolerated. Neither argument proves this particular substitution damages this particular protein, hence 'uncertain'.
(3) Intron 14 +34T>C - least likely. It sits far from the splice junctions and from the branch point, so it is unlikely to disturb splicing, and it is not in a coding region.
2. The laboratory also reports a synonymous (third-base) change in a different patient and calls it benign. Explain why that default is usually correct but not always.
Read the worked answer
The genetic code is degenerate: several codons specify the same amino acid, and third-base changes often leave the protein sequence unchanged, so the default classification of 'benign' is reasonable.
There are well-recognised exceptions, and the reasoning behind them is worth stating:
- Splicing. Exonic sequence contributes to splice-site recognition and to exonic splicing enhancers. A synonymous change can abolish an enhancer and cause exon skipping, producing a protein missing a whole segment.
- mRNA stability and translation rate. Rare codons can slow translation; changing codon usage can alter folding or message half-life.
- Cryptic sites. A change can create a new splice donor that competes with the real one.
For the same reasons, a deep intronic variant like (3) is not automatically harmless: it could create a cryptic splice site or disrupt a regulatory element. Classification therefore requires RNA analysis (RT-PCR of the patient's transcript) rather than sequence inspection alone; if the transcript is spliced and sized normally, the case for benignity is much stronger.
3. Explain why identifying a variant in this boy does not by itself prove causation, and describe the evidence you would gather before offering predictive testing to his sister.
Read the worked answer
Finding a rare variant shows only that the sequence differs from the reference. Causation requires that the variant (a) segregates with disease in the family, (b) is absent or extremely rare in controls, (c) changes the protein in a way that plausibly destroys function, and (d) is shown to do so experimentally.
Here (a) is unusually informative because the parents are each heterozygous for a different variant and both have raised LDL (8.4 and 3.1 mmol dm⁻³). That is consistent with the boy being a compound heterozygote - one defective allele from each parent - which is the expected basis of severe familial hypercholesterolaemia. It does not prove that variant (2) is the defective one; the mother's milder lipid profile could reflect the nature of her allele or other modifiers.
Evidence to gather: measure the parents' LDL off treatment where safe; test whether variant (2) has ever been seen in unaffected elderly controls; run a minigene or patient RNA assay to see whether splicing is affected; and, for Cys396Tyr, an expression study checking whether the protein reaches the cell surface or is retained intracellularly.
For the sister, targeted testing for the two family variants is appropriate and inexpensive: it can definitively identify a second affected child who needs early treatment, which matters because untreated familial hypercholesterolaemia causes infarction in the fourth or fifth decade. Predictive testing in a child should be limited to variants that are actionable.