Nature Genetics
20, 315 - 316 (1998)
doi:10.1038/3776
Peromysci, promiscuity and imprintingLaurence D HurstDepartment of Biology and Biochemistry, University
of Bath, Claverton Down, Bath, BA2 4SD,
UK. l.d.hurst@bath.ac.uk I don't like flying—especially in planes with only one engine. What
if it fails? If you share my fear, you will understand why evolutionists are
fascinated by genomic imprinting. In a diploid organism, selection will typically
favour two functional alleles, in case one 'fails' as a consequence of mutation.
Just what sort of selection gives rise to imprinting, where one copy of a
gene is inactivated, leaving the organism vulnerable to mutation? In a study
presented on page x (ref.1), Paul Vrana and colleagues
attempt to test an appealing and popular hypothesis, one that supposes that
imprinting is a consequence of evolutionary conflict between maternally and
paternally derived genes2.
In genomic imprinting, the choice of which allele to inactivate is dependent
upon the sex of the parent from which the allele was derived. For example,
a fetus inherits a copy of the insulin-like growth factor 2 gene (Igf2
) from both its parents, but only the father's gene is expressed. The
conflict model2 proposes that selection acting on paternally
and maternally derived genes in the same fetus is different. Each maternally
derived allele in a fetus has a 50% likelihood of also being present in any
other given fetus from the same mother. Consequently, any detrimental effect
of this allele on other progeny, or on the mother, could reduce the probability
of that allele spreading. In contrast, a rare paternally derived allele in
a fetus sired by a father who has no other offspring with the same mother
will not be present in any other progeny of that mother—any detriment
to these infants, or to the mother's future reproductive prospects, might
therefore increase the probability of the allele's spread. Generally, when
there is multiple paternity, selection will favour paternally derived alleles
that obtain more resources from the mother than is optimal for maternally
derived alleles. The maternal genome will therefore tend to silence growth
promoters, while the paternal genome will tend to silence growth suppressors2.
Testing the conflict hypothesis
 |  | Peromyscus maniculatus Photograph kindly provided by Wallace Dawson |
| |
How would one know if this hypothesis is correct? One test would be to
determine whether monogamous species maintain imprinting. Under monogamy,
the conflict hypothesis would predict that the probability of two progeny
containing a particular allele is the same, regardless of parentage; one would
expect to see an absence of imprinting. Vrana et al. have tested this
prediction by characterizing two closely related species of rodent (Peromyscus
)—one monogamous and one polyandrous—and the progeny obtained
from crossing them.
The hybrid progeny show parent-of-origin effects, suggesting that the pattern
of imprinting is different in each of the parents, with the possibility that
it is lost in the monogamous one. If the father is from the promiscuous species,
the progeny are large; if he is from the monogamous species, they are small.
This is consistent with the conflict hypothesis, as the paternal genome of
the monogamous species should be 'less demanding'. Against expectations, however,
Vrana et al. find that imprinting is maintained in the monogamous species
and conclude that the difference in size between the two reciprocal crosses
are due to a disruption of imprinting. How do these findings reflect on the
conflict hypothesis? Vrana et al. comment that, although an absence
of imprinting would have been strong supportive evidence, its maintenance
does not warrant rejection of the hypothesis. One possible explanation is
that monogamy has only recently evolved. Alternatively, the monogamous species
may not really be monogamous—there is evidence for a very low rate of
partner exchange3. It is, however, questionable whether this
rate is high enough to maintain genomic imprinting when selection against
it (due to exposure of deleterious recessives) may be acting. The necessary
parameter estimates—such as rate and effect of deleterious mutations
and degree of between-brood competition—needed to test this theory are
not sufficiently rigorous to provide an unequivocal answer. The vagaries of
interpretation also apply to suggestive evidence of imprinting in Arabidopsis
—as near an obligate 'selfer' as one can find4; like
a monogamous mammal, it is not expected to have imprinting, yet shows parent-of-origin
effects5. This pattern of prediction, followed by apparent rejection,
followed by post-hoc qualifiers typifies attempts to test the conflict hypothesis
if not science in general. Genetic systems in conflict often show rapid sequence
evolution, but imprinted genes do not6. This is contrary to
the prediction that maternally and paternally imprinted genes should antagonistically
evolve, although, once again, there have been suggestions for why imprinted
genes do not adhere to expectation7.
Little and large
When the conflict hypothesis was first presented, the gigantism of human
infants with paternal uniparental disomy (UPD) associated with Beckwith-Wiedeman
syndrome, was cited as supportive evidence2. Individuals with
a UPD have two copies of one chromosome from one parent. A paternal UPD should
be associated with over-expression of growth promoters and under-expression
of growth suppressors and so individuals with a paternal UPD should be large.
Of the six paternal UPDs associated with growth perturbations, however, only
Beckwith-Wiedeman syndrome displays overgrowth8. Again, one
might imagine post-hoc explanations for this unexpected pattern. Perhaps UPDs
have too great a change in dosage? Indeed, a moderate overdose of paternal
genes in Arabidopsis results in large endosperm, but higher levels
can lead to abortion and small size9.
The data obtained by Vrana et al. also shed some light on this problem.
If the conflict hypothesis is correct, only the paternal genes in large hybrids
should be bi-allelically expressed. This is observed, however, for two of
three maternally expressed genes and three of four paternally expressed genes.
Evidence from small hybrids is closer to the expected pattern. All paternally
expressed genes have monoallelic expression, while two of three maternally
expressed genes have biallelic expression, although only one is expressed
in the fetus (the other is biallelically expressed in the placenta). These
data do not provide compelling evidence for the predicted link between the
direction of an imprint and the direction of the growth effect. Neither do
analyses of mice depleted of imprinted genes10; five mutants
(null for Igf2, H19, Igf2r, Peg3, Grf1)
appear to support the relationship, another two (null for Gtl2 and
Gnas) probably do but need clarification, and a further seven do not.
Some of these seven mutants have no growth phenotype (for example, mice that
are deficient in both Ins1 and Ins2, and mice that lack the
Smn region of Snrpn)—although it could be that the change
is too small to detect.
The conflict hypothesis highlights a difference between fetal expression
of maternal and paternal genes with respect to the demand on nutrients from
mother, either in utero or until the termination of weaning. In adulthood,
there is no conflict between genes in the same individual over the optimal
rate of growth or of provisioning to offspring and therefore imprinting is
not expected to persist in the adult. This is supported by the observation
that the Igf2 imprint is abolished shortly after birth and that the
growth effects of Grf1 persist only until termination of weaning. This
is not, however, the case for Mest, a paternally expressed gene. Mice
with mutant Mest show fetal growth retardation, as expected from the
conflict hypothesis, but the adult females neglect their young11,
a behaviour that cannot obviously be explained by the conflict hypothesis.
While it may fit in some cases, the accumulating evidence, especially that
pertaining to adult behaviour, suggests that the conflict hypothesis is not
going to explain everything. Given, too, that the existence of allelic exclusion
implies another reason for haploid expression, it is time we thought some
more about the advantages of flying with only one engine.
REFERENCES
- Vrana, P.B., Guan, X.-J., Ingram, R.S. & Tilghman, S.M. Nature Genet. 20, 362-365 (1998). | Article | PubMed | ISI | ChemPort |
- Moore, T. & Haig, D. Trends Genet. 7, 45-49 (1991). | Article | PubMed | ISI | ChemPort |
- Foltz, D.W. Am. Nat. 117, 665-675 (1981). | Article | ISI |
- Abbott, R.J. & Gomes, M.F. Heredity 62, 411-418 (1989). | ISI |
- Scott, R.J., Spielman, M., Bailey, J. & Dickinson, H.G. Development 125, 3329-3341 (1998). | PubMed | ISI | ChemPort |
- McVean, G.T. & Hurst, L.D. Proc. R. Soc. Lond. B Biol. Sci. 264, 739-746 (1997). | Article | PubMed | ISI | ChemPort |
- Haig, D. Proc. R. Soc. Lond. B Biol. Sci. B 264, 1657-1662 (1997). | PubMed | ChemPort |
- Hurst, L.D. & McVean, G.T. Trends Genet. 13, 436-443 (1997). | Article | PubMed | ISI | ChemPort |
- Scott, R.J., Spielman, M., Bailey, J. & Dickinson, M.G. Development 125, 3329-3341 (1998). | PubMed | ISI | ChemPort |
- Hurst, L.D. & McVean, G.T. Curr. Opin. Genet. Devel. (in press).
- Lefebvre, L. et al. Nature Genet. 20, 163-169 (1998). | Article | PubMed | ISI | ChemPort |
- Miller, N. et al. Genomics 46, 509-512 (1997). | Article | PubMed | ISI | ChemPort |
|
