Bibliographic Reference
Wilkins, J. F., & Haig, D. (2003). What good is genomic imprinting: The function of parent-specific gene expression. Nature Reviews Genetics, 4(5), 359–368. https://doi.org/10.1038/nrg1062
Core Argument
Genomic imprinting (parent-specific gene expression) is an evolutionary puzzle because monoallelic expression forgoes the principal advantage of diploidy — protection against deleterious recessive mutations. The paper systematically evaluates three competing hypotheses for what countervailing selective advantage could outweigh this cost: (1) evolvability models, which propose that imprinting enhances population-level adaptability; (2) the ovarian-time-bomb (OTB) hypothesis, which proposes that imprinting protects females from invasive trophoblastic disease; and (3) the kinship theory, which proposes that imprinting arises from evolutionary conflict between maternally and paternally derived alleles over resource allocation to kin. The authors argue that the kinship theory has received the most extensive theoretical development and best explains the known properties of imprinted genes, but acknowledge that the strong version of the theory — which claims kinship effects were the predominant selective force — has yet to explain the evolution of imprinting at most imprinted loci.
Methods
Theoretical review. The paper synthesizes three classes of evolutionary model — game-theoretic, population-genetic, and quantitative-genetic — and evaluates empirical evidence for each hypothesis. It draws on comparative data across mammals (marsupials, monotremes, eutherians), plants, and model organisms; UPD and knockout phenotypes in mice; and the directionality of imprinting (which parent’s allele is silenced) as a test of predictions.
Key Findings
-
The kinship theory predicts directionality correctly where other hypotheses cannot. Madumnal (maternally derived) alleles of growth enhancers should be silent, and padumnal (paternally derived) alleles of growth inhibitors should be silent — a pattern explained by asymmetric coefficients of matrilineal relatedness (r_m) and patrilineal relatedness (r_p) when females sometimes bear offspring by more than one male. The OTB can also predict this directionality for trophoblast genes but requires bystander effects for non-trophoblast genes. Evolvability models make no directional predictions at all.
-
The “loudest-voice-prevails” principle resolves intragenomic conflict. Game-theoretic and quantitative-genetic models consistently find that whichever allele (madumnal or padumnal) favors the larger amount of a given gene product produces that amount at evolutionary equilibrium, and the other allele is silenced. This is a simple form of conflict resolution: the allele that favors higher expression presents the other with a fait accompli.
-
Imprinting of growth inhibitors is evolutionarily less stable than imprinting of growth enhancers. The conflict between cis-acting imprinted loci (expressed in offspring, evolving according to madumnal/padumnal interests) and trans-acting imprinting machinery (expressed in parents, evolving according to maternal/paternal interests) creates an asymmetry. Maternal genes favor lesser demands than either madumnal or padumnal genes, so there is no maternal incentive to reactivate a silenced growth enhancer. But paternal genes sometimes have incentive to reactivate silenced growth inhibitors — making padumnal silencing less stable. This explains why most madumnally silent loci are silenced by direct promoter methylation, while most padumnally silent loci are silenced indirectly via antisense transcripts.
-
Population-genetic and game-theoretic models yield different predictions because they use different equilibrium concepts. Population-genetic models track the frequency dynamics of a small number of alleles over time; game-theoretic models use the criterion of non-invasibility (an allele near fixation that cannot be displaced by any rare alternative). The authors argue the population-genetic results can be misleading because some allele pairs considered by those models are unlikely to compete in natural populations.
-
No hypothesis fully explains the empirical landscape. Evolvability models lack specificity (their logic should apply at most loci in most organisms, but imprinting is restricted). The OTB is over-specific (applies only to trophoblast growth in mammals) and cannot explain imprinting in plants or in non-invasive tissues. The kinship theory has the best fit to the phylogenetic pattern — imprinting is associated with viviparity in both mammals and plants, absent in oviparous model organisms (Caenorhabditis, Drosophila, Danio) — but the strong version “is yet to meet the challenge of explaining the evolution of imprinting at most imprinted loci.”
Concepts Introduced or Used
genomic-imprinting, parent-specific gene expression, madumnal-allele, padumnal-allele, kinship-theory, conflict-theory, inclusive-fitness, matrilineal-relatedness, patrilineal-relatedness, loudest-voice-prevails-principle, demand-enhancer, demand-inhibitor, ovarian-time-bomb-hypothesis, evolvability, functional-haploidy, uniparental-disomy, loss-of-imprinting, cis-trans-conflict, non-invasibility, game-theoretic-model, population-genetic-model, quantitative-genetic-model, viviparity
Entities Referenced
- Genes: IGF2, IGF2R (M6P/IGF2R), Mest (mesoderm-specific transcript), Peg3 (paternally expressed 3), GNAS1
- Species: Mus musculus, Peromyscus maniculatus (deer mouse, high multiple paternity), Peromyscus polionotus (oldfield mouse, low multiple paternity), marsupials (opossum), monotremes, sheep, Arabidopsis (and other plants), Danio rerio (zebrafish), Drosophila melanogaster, Caenorhabditis elegans
- Methods: game-theoretic modelling (non-invasibility criterion), formal population genetics, quantitative genetics
- Key researchers: David Haig, Robert Trivers, Laurence Hurst, Hamish Spencer, Yoh Iwasa, Susannah Varmuza, Mellissa Mann
- Databases: Harwell imprinting website (>60 imprinted transcripts in mice as of 2003)
Limitations (as stated by authors)
- The kinship theory’s strong version has not yet explained the evolution of imprinting at most imprinted loci. The authors acknowledge this as an open challenge that future empirical work must address.
- UPD data are ambiguous. Paternal UPDs are sometimes associated with growth retardation and maternal UPDs with growth enhancement, opposite to the predictions of the simplest version of the kinship theory. The theory can accommodate this through models of resource allocation between placental and embryonic compartments (Iwasa et al., 1999), but the authors caution that UPDs “might not provide strong evidence against the theory, but they definitely cannot be interpreted as providing strong evidence in its favour.”
- Post-weaning effects of imprinted genes (e.g., Mest and Peg3 effects on maternal behavior) pose a challenge. The madumnal and padumnal alleles of a mother are equally likely to be transmitted to her ova, so they would seem to benefit equally from her level of maternal care. Explanations invoking hidden asymmetries of relatedness (e.g., costs to maternal half-sisters, inbreeding effects) have been proposed but their underlying assumptions remain untested.
- The persistence of imprinting in presumably monogamous species (P. polionotus) is not a strong refutation of the kinship theory because selection to restore biallelic expression is weak — of the same order as the mutation rate. However, this conclusion considers only germline mutations; somatic mutation costs of functional haploidy have not been formally modelled.
- The game-theoretic vs. population-genetic model dispute is unresolved in part because the two approaches use different concepts of evolutionary equilibrium.
- The authors frame the risk of prematurely rejecting vs. prematurely accepting the strong version of the kinship theory as a judgment call the field must weigh.
Relevance to Clonal Evolution
Wilkins and Haig do not directly address cancer, but their framework has direct implications for understanding somatic evolution in tumors:
(a) Imprinted genes as developmental vulnerabilities exploited by cancer. If imprinted genes serve adaptive functions in development — particularly growth regulation, resource allocation, and placental invasiveness — then their disruption in cancer represents the somatic exploitation of a developmental architecture that was shaped by kinship conflict. The same genes that mediate fetal-maternal resource transfer (IGF2, IGF2R) are recurrently dysregulated in cancer.
(b) One-hit kinetics of imprinted tumor suppressors. Because one allele at an imprinted locus is already epigenetically silenced, the remaining active allele is functionally hemizygous. A single genetic or epigenetic hit to the active allele suffices for complete loss of function, halving the waiting time compared to a biallelically expressed tumor suppressor that requires two inactivating events. This accelerates the tempo of clonal-evolution.
(c) Loss of imprinting (LOI) as faster-than-mutation path to altered gene dosage. Imprinting is an epigenetic state, not a DNA sequence change. LOI — the reactivation of the silenced allele — can double gene dosage without any mutation, occurring on a potentially faster timescale than mutational activation. For growth-promoting imprinted genes (e.g., IGF2), LOI is a recurrent event in multiple cancer types and represents an epigenetic mechanism for clonal expansion that operates in parallel with and potentially faster than genetic driver mutations. This connects imprinting to the broader question of the tempo of clonal-evolution: epigenetic changes may provide faster routes to fitness gains than waiting for mutations, particularly at loci where the chromatin infrastructure for silencing already exists and can be eroded.