The length of the expanded CAG repeat accounts for something like 70% of variance in age of onset in Huntington’s Disease.  This is due in part to the very late onset observed in people with fewer than 40 repeats and the very early onset seen with really long repeats, say, more than 50.  In the range where most HD carriers fall– the 40 to 50 CAG range– the correlation is less tight, and CAG length only explains perhaps 40% of variance.

By subtracting log actual age of onset (and we use age of motor onset, i.e. onset of motor symptoms for this) from log predicted age of onset (based on CAG length) we can get a residual age of onset for each person which is independent of CAG length.

It’s believed that a lot of this residual age of onset can be explained by other genetic factors besides CAG length.  Djousse 2003 provided the earliest study; the best evidence comes from U.S.–Venezuela Collaborative Research Project 2004.  The latter study looked at the large Venezuelan HD pedigree including a few hundred sibling pairs and found that siblings explained about 40% of variance in one another’s age of onset.  Since siblings only share half their genetic information, it could be extrapolated that as much as 80% of the variance is heritable.  They also looked at parent-child, avuncular and cousin relationships and put together an overall estimate that 38% of age of onset was genetic.  The paper does not explain the magic by which they separated out the effect of shared environmental factors (i.e. nurture, not nature), which they claim explain another 22% of variance, but I suspect it had to do with an assumption that only siblings share their environment and so you can triangulate from the lesser degrees of similarity among cousins and so forth.

In any event, it looks like about 40% of residual age of onset is genetic.  We’d like very much to know what those genetic factors are– maybe there are protective or harmful variants that lie in druggable genes.  For a review of why we care and what’s been tried so far, see Gusella & MacDonald 2009.

But frustratingly, no study has yet uncovered convincing age-of-onset modifiers for HD.  To my knowledge, the two most recent studies on the subject are still those mentioned in my GWAS post.  Andresen 2007 reviews 12 variants identified as significant in earlier studies and, for 11 of them, fails to replicate their results, indicating that earlier findings were either bogus or specific to one population.  The only SNP that Andresen found might really be associated is in GRIN2A, which codes for the NR2A subunit of the NMDA glutamate receptor, but it was marginal with p = .04.    Gayan 2008 did a linkage scan using a ~6,000-SNP genotyping array on 443 related individuals from the Venezuelan pedigree and found a couple of “potentially” associated loci.  As far as I know, no later study has tried to validate this result or determine exactly what variant or gene at those loci might have been causal.

The GRIN2A finding is interesting, especially because of the apparent interaction between coffee, GRIN2A and Parkinson’s.  But it appears to me it’s not yet universally accepted that GRIN2A definitely has a modifier role in HD age of onset.  To be on the safe side, as of today you could still say that the HD research community has not conclusively identified any genes or variants that definitely contribute to the apparent 40% heritability of residual age of onset.

update 2013-01-17 & 2013-02-18: I learned of a few other papers relevant to this topic:

  • Li 2006 reports on the results of the sib pair-based HD MAPS study, finding that the 6q23-24 locus is associated with residual age of onset, reaching genome-wide significance.
  • Metzger 2008 (ft), a candidate gene study, claims that a SNP in huntingtin associated protein 1 is associated with age of onset.
  • Lee 2009 attempts to replicate a purported residual age of onset association with GRIK2 (both a SNP and a TAA repeat length polymorphism) in 2,911 subjects and finds no association.
  • Lee 2012a reports complete dominance of the longer HTT allele, meaning there is no association between the other allele’s CAG length and residual age of onset.  The analysis includes people with 1 expanded and 1 normal allele as well as 10 people with 2 expanded alleles.
  • Lee 2012b reports on an HTT haplotype analysis, and finds no evidence that either the haplotype of the expanded allele or of the normal allele are associated with age of onset.