QTs IV: “miraculous” and “missing” heritability
(1) Selection should “use up” VA, by fixing the favorable alleles.
But it doesn’t (at least in many cases).
The Illinois Long-term Selection Experiment (1896-2015, and continuing*),
has artificially selected corn for high and low oil and protein content.
… but the “up” lines
are still increasing at
a nearly constant rate!
Both are now up by
more than 10 s.d.!
How is this possible?
The “down” lines were
abandoned in the late
20th century…
*http://mooselab.cropsci.illinois.edu/longterm.html
Anth/Biol 5221, 9 December 2015
(2) For many traits with high heritabilities, only small fractions of VA
can be accounted for by the mappable quantitative trait loci (QTL).
Human height is one of many examples.
More than 200 QTL can be identified by
linkage disequilibrium with SNP marker loci.
h2 = 0.98
Logic of
QTL
mapping in
a study of
hybrids
between
Mimulus
cardinalis
and
Mimulus
lewisii.
But these 200 QT loci, together, explain only
around 5% of the observed heritability!
Where’s the rest of it?
h2 = 0.82
Utah Genetic
Reference Project
Why is schizophrenia so common?
It afflicts more than 1 in 200 people.
It is highly heritable.
And it dramatically lowers fitness.
“There is little doubt about the existence of a
fecundity deficit in schizophrenia. Affected
individuals have fewer children than the
population as a whole. This reduction is of the
order of 70% in males and 30% in females. The
central genetic paradox of schizophrenia is why,
if the disease is associated with a biological
disadvantage, is this variation not selected out?
To balance such a significant disadvantage, a
substantial and universal advantage must exist.
Thus far, all theories of a putative advantage
have been disproved or remain unsubstantiated.”
– From the Wikipedia article “Schizophrenia”
“Data from a PET [positron
emission tomography] study
suggest that the less the
frontal lobes are activated
(red) during a working memory
task, the greater the increase
in abnormal dopamine activity
in the striatum (green),
thought to be related to the
neurocognitive deficits in
schizophrenia.”
And the heritabilities are amazing! (Thank you, twins!)
Note that several conditions
considered distinct “diseases” by
mental-health professionals are
strongly correlated genetically
(MZ twins are diagnosed with
different ones of them much
more often than are DZ twins).
This has fueled debate about the
reality of the distinctions.
What is the genetic “architecture” of these susceptibilities?
Mutations in a few genes with large effects?
If so, which genes are they?
Or is it polymorphisms at many loci, each making a small contribution?
And in either case, what kinds of mutations are involved?
Unconditionally deleterious?
Or good-news/bad-news tradeoffs?
Amino-acid and nucleotide substitutions?
Or duplications and deletions (copy-number variation)?
QTL mapping suggests that there are modest numbers
of genes with major effects, on many chromosomes.
Ripke et al. Nat. Genet. 45, 1150-1159 (2013)
And some of these are plausible “candidate” genes
Many play well established
roles in neurotransmission,
in or near synapses.
But again there’s this large and familiar problem:
The major-effect genes explain just a fraction of the heritability!
Where’s the rest of it?
Could susceptibility be a highly polygenic quantitative trait?
This was proposed
48 years ago by
Irving Gottesman.
(Gottesman & Shields, PNAS 58,
199-205 (1967)
But if that’s the case,
where is all this heritable
variation hiding?
And why is there so much
of it, given that selection
against schizophrenia is
very strong?
American Journal of
Human Genetics 35,
1161-1178 (1983)
One possibility:
frequent copynumber mutations
New deletion
of ~2 Mb
But there are other possibilities, for example:
(1) very many genes with alleles of very small effect
(2) very many genes with alleles of infrequent effect (low penetrance)
(3) complex genetic interactions (epistasis, dominance, etc.)
(4) epigenetics (imprinting or other environmental “pseudo-heritability”)
However, not all of these can easily explain the very high levels of
heritability seen in the twin studies and other direct pedigree analyses.
Nor can they all easily explain why selection experiments do not run out
of additive variance, even when mean values have been changed by many
standard deviations.
But possibility #1 (very many loci, each with very small effects) can
explain all of the observations, at least in principle.
(See Rockman [2012] on the course web site for an entertaining recent
review of this idea, emphasizing evolution more than medicine.)
Recent evidence for many alleles of small effect, I
“SNP heritability” studies
estimate the total variance
in “liability” explained by
SNPs, which are used to
estimate the pairwise
genetic relationships of
cases and controls.
Genetic variation is
indicated to the degree
that “case-case and
control-control pairs are,
on average, more similar
across the genome than
case-control pairs.”
Cross-Disorder Group of the Psychiatric Genomics Consortium
Nat. Genet. 45, 894-994 (2013)
Recent evidence for many alleles of small effect, II
Combined
analysis of two
large studies
Type 2 diabetes
Myocardial infarction
Rheumatoid Arthritis
Ripke et al.,Nat. Genet. 45, 1150-1159 (2013)
“We estimate that
8,300 independent,
mostly common SNPs
(95% credible
interval of 6,300–
10,200 SNPs)
contribute to risk
for schizophrenia
and that these
collectively account
for at least 32% of
the variance in
liability.”
If more loci contribute to VA, there is more scope for evolution.
x min
p = 0.5
6
loci
x max
p = 0.75
p = 0.95
x min
x max
p = 0.5
Mean up
3.1 s.d.
100
loci
p = 0.75
p = 0.5
25
loci
p = 0.75
p = 0.95
p = 0.95
Mean up
6.3 s.d.
Mean up
12.9 s.d.
Summary
Large amounts of heritable variation could be caused by very small
contributions to VA from many different loci.
Such “small-effect” alleles may help to explain why VA is not quickly
depleted by strong, continued selection (the “miraculous” heritability).
They may also account for the variation that cannot be associated with
markers, even in enormous QTL studies (the “missing” heritability).
Deleterious small-effect mutations could drift upward at many loci, or
hitchhike with closely linked adaptive mutations, because selection
against the disfavored allele at each of the many loci would be weak.
Thus our surprising susceptibility to schizophrenia and related disorders
might be, in part, a consequence of our species’ history of small
population size, or a side effect of recent rapid mental evolution.
If so, is Wikipedia’s assertion necessarily correct?
“The central genetic paradox of schizophrenia is why, if the disease
is associated with a biological disadvantage, is this variation not
selected out? To balance such a significant disadvantage, a
substantial and universal advantage must exist.”