Haig (2004) — Genomic Imprinting and Kinship: How Good is the Evidence?

Bibliographic Reference (APA 7.0)

Haig, D. (2004). Genomic imprinting and kinship: How good is the evidence? Annual Review of Genetics, 38, 553–585. https://doi.org/10.1146/annurev.genet.37.110801.142741

Core Argument

The kinship theory of genomic imprinting proposes that parent-specific gene expression evolves at a locus when (1) changes in the aggregate expression level at that locus influence a fitness trade-off between related individuals, and (2) at least one of the affected individuals has different probabilities of carrying the maternal versus paternal alleles of the individual in which the gene is expressed (asymmetric kin). The strongest asymmetry occurs in mother-offspring interactions: a fetus’s maternally derived alleles are necessarily present in its mother (m = 1), while its paternally derived alleles are necessarily absent (p = 0, in an outbred species). The theory predicts that paternally expressed genes will be selected to favor higher levels of maternal investment (e.g., enhancing prenatal growth), while maternally expressed genes will favor lower levels.

Haig distinguishes a “weak version” (a logical statement about how selection must act if parent-specific expression exists and affects asymmetric kin) from a “strong version” (the empirical claim that fitness effects on asymmetric kin have been the principal drivers of imprinting at most imprinted loci). This review assesses the strong version by examining four imprinted gene clusters: the Beckwith-Wiedemann cluster (IGF2/H19), the Tme cluster (Igf2r/Air), the callipyge cluster, and the GNAS complex locus.

Methods

This is a qualitative narrative review. Haig does not conduct a meta-analysis or systematic literature search. His approach is:

1. Cluster selection: Two clusters were chosen because they contain the flagship loci IGF2 and IGF2R (Beckwith-Wiedemann and Tme clusters). The remaining two (callipyge and GNAS) were “picked blind, before I had given much thought to their evolutionary interpretation,” to avoid stacking the deck in favor of the theory.
2. Focus on phenotypes, not mechanisms: The review emphasizes the physiological and phenotypic effects of imprinted genes, rather than molecular mechanisms of imprinting establishment or maintenance, because phenotypes determine selective outcomes.
3. Omission: The Prader-Willi/Angelman cluster at 15q11–q13 is deliberately excluded because Haig had recently published a separate evolutionary interpretation of Prader-Willi syndrome.
4. Evidence appraisal: For each cluster, Haig reviews uniparental disomy (UPD) phenotypes, individual gene knockout phenotypes, and phylogenetic distribution of imprinting, then evaluates whether the collective pattern supports or contradicts kinship theory predictions.

Key Findings

1. The Beckwith-Wiedemann cluster (IGF2/H19/KCNQ1OT1/CDKN1C) broadly supports the theory. Paternal UPD at 11p15.5 causes prenatal overgrowth and enlarged placentas. The cluster contains at least two paternally expressed enhancers of intrauterine growth (IGF2 and KCNQ1OT1, the latter acting indirectly by silencing maternally expressed alleles). Maternally expressed genes in this cluster generally act as growth inhibitors (CDKN1C, IPL), with Mash2 being a notable exception — its maternal-specific expression is explicable under the theory if partial reallocation of cells from spongiotrophoblast to giant cells enhances fetal growth via placental hormones. The noncoding RNA H19 shows evidence of growth-inhibitory effects, but its precise role remains unclear.

2. The Tme cluster (Igf2r/Air) provides strong support for the kinship theory. Igf2r is a maternally expressed inhibitor of prenatal growth: inactivation of the maternal copy causes a roughly 30% increase in fetal size in mice, mediated by elevated circulating IGF-II. The paternally expressed noncoding RNA Air is required for silencing the paternal Igf2r allele; deletion of the paternal copy of the Air promoter results in biallelic Igf2r expression and reduced birth weight. IGF2R is imprinted in marsupials, rodents, and artiodactyls but not in primates — a pattern consistent with a single origin of imprinting followed by loss in the Euarchonta lineage. The theory predicts that IGF2R’s growth-inhibitory effects should be attenuated in primates, where the gene has been subject to selection on both parental alleles for at least 75 million years.

3. The GNAS complex locus provides provisional support for the kinship theory, particularly via the “huddling hypothesis.” Mice with a maternally-inherited Gnas knockout (mKO) have increased birth weight, develop obesity, and show reduced nonshivering thermogenesis (NST) in brown adipose tissue (BAT). Mice with a paternally-inherited knockout (pKO) have decreased birth weight, lean phenotype, and increased NST. Haig argues that this pattern is consistent with the kinship theory: within huddles of littermates, matrilineal relatedness is typically higher than patrilineal relatedness (due to multiple paternity), so maternally derived alleles should favor greater contribution to the “public good” of communal heat production, while paternally derived alleles should favor free-riding. The complex tissue-specific imprinting of Gs-alpha — biallelic in most tissues but monoallelic (maternal) in renal proximal tubules, thyroid, and pituitary — suggests that imprinting can be exquisitely context-dependent rather than gene-wide.

4. The callipyge cluster does not yet provide compelling support for the kinship theory. The known functions of imprinted genes in this cluster (DLK1 inhibiting adipogenesis, DIO3 inactivating thyroid hormones) are not easily interpreted in terms of the theory’s predictions. DLK1 is paternally expressed and inhibits adipocyte differentiation — consistent with the idea that paternally derived alleles may favor reduced allocation to fat storage — but the opposite pattern (maternal knockout of Gnas reduces adiposity) complicates any simple interpretation. DIO3, which inactivates thyroid hormone, could theoretically reduce fetal metabolic costs that are borne by the mother, but this remains conjectural.

5. The kinship theory has succeeded for a subset of loci but has not provided a satisfying explanation for the majority of imprinted genes. Haig characterizes the glass as “half full or half empty” and notes that many imprinted genes may be “innocent bystanders” swept up in the establishment of imprinting at neighboring loci, or may have phenotypic effects too subtle to have been detected in knockout studies. For example, multiple imprinted genes affect white adipose tissue (WAT) but show no consistent pattern across genes or species. Imprinted effects on BAT may have been systematically missed because BAT is recruited postnatally in cold conditions, and most knockout studies are conducted at warm ambient temperatures.

Concepts Introduced or Used

• Kinship theory of genomic imprinting (weak and strong versions): The hypothesis that parent-specific expression evolves because of fitness effects on asymmetric kin — individuals with different probabilities of carrying the maternal versus paternal alleles of the expressing individual.
• Asymmetric kin: Relatives whose probability of identity-by-descent differs for maternal and paternal alleles (e.g., an offspring’s mother has m = 1, p = 0; a maternal half-sib has m = 0.25, p = 0).
• Maternal investment: Any investment by a mother in an offspring that increases the offspring’s survival at the cost of the mother’s ability to invest in other offspring (Trivers’ definition). The kinship theory predicts paternally expressed genes favor higher maternal investment than maternally expressed genes.
• Huddling hypothesis for BAT imprinting: In species that produce litters, littermates huddle for warmth, making heat generation in BAT a public good. Because matrilineal relatedness within huddles exceeds patrilineal relatedness (due to multiple paternity), maternally derived alleles favor higher NST contribution, while paternally derived alleles favor lower contribution (free-riding).
• Polar overdominance: The unusual inheritance pattern of the callipyge mutation, where only paternal heterozygotes express the muscle hypertrophy phenotype; maternal heterozygotes and homozygotes of either parental origin are normal.
• Functional haploidy of imprinted genes: Because only one parental allele is expressed at an imprinted locus, the gene is effectively hemizygous. A single inactivating mutation or epigenetic silencing event at the active allele eliminates gene function entirely.
• Innocent bystander hypothesis: Some genes in imprinted clusters may show parent-specific expression not because imprinting was selected at those loci, but as a side effect of selection for imprinting at a neighboring gene.

Entities Referenced

• IGF2 (Insulin-like growth factor 2): Paternally expressed, enhancer of prenatal growth. Imprinted in therian mammals (marsupials and eutherians) but biallelic in monotremes and birds.
• H19: Maternally expressed noncoding RNA, reciprocally imprinted with IGF2. Possible growth-inhibitory effects; possible tumor-suppressor and tumorigenic roles reported in different cell lines.
• IGF2R (Insulin-like growth factor 2 receptor, also CI-MPR): Maternally expressed inhibitor of fetal growth. Binds IGF-II and targets it to lysosomes for degradation. Imprinted in marsupials, rodents, and artiodactyls; biallelic in primates (Euarchonta).
• CDKN1C (p57^KIP2^): Maternally expressed cyclin-dependent kinase inhibitor. Mutations cause Beckwith-Wiedemann syndrome. Antagonistic to IGF2 in cell proliferation.
• KCNQ1OT1 (LIT1): Paternally expressed noncoding RNA from the KvDMR1 imprinting control region. Silences nearby maternally expressed genes in cis.
• Mash2 (ASCL2): Maternally expressed transcription factor required for spongiotrophoblast development. Not imprinted in humans.
• Air (Antisense Igf2r RNA): Paternally expressed noncoding RNA required for silencing the paternal Igf2r allele and the maternal Slc22a2/Slc22a3 alleles.
• Gs-alpha (GNAS): Alpha subunit of the heterotrimeric G-protein Gs. Biallelic in most tissues, but monoallelic maternal expression in renal proximal tubules, thyroid, and pituitary. Mutations cause Albright’s hereditary osteodystrophy (AHO) and pseudohypoparathyroidism type Ia (PHP-Ia).
• DLK1 (Pref-1): Paternally expressed inhibitor of adipocyte differentiation. Knockout causes increased adiposity.
• DIO3 (Type 3 deiodinase): Paternally expressed enzyme that inactivates thyroid hormones. Highly expressed in placenta and fetal tissues.
• UCP-1 (Uncoupling protein-1): Mitochondrial protein responsible for nonshivering thermogenesis in brown adipose tissue. Expression levels differ in Gnas mKO vs. pKO mice.
• Beckwith-Wiedemann syndrome (BWS): Human overgrowth syndrome associated with paternal UPD 11p15.5, loss of imprinting at IGF2, or inactivating mutations of maternal CDKN1C.
• Pseudohypoparathyroidism type Ia (PHP-Ia): AHO with PTH resistance, caused by maternal inheritance of inactivating Gs-alpha mutations.
• Callipyge: Sheep mutation causing muscle hypertrophy in paternal heterozygotes only (polar overdominance), caused by a single-base substitution dysregulating multiple imprinted genes on chromosome 18.

Limitations (as stated by authors)

• The review deliberately focuses on four clusters and does not attempt comprehensive coverage of all known imprinted genes. The choice of the Beckwith-Wiedemann and Tme clusters may be perceived as “stacking the deck” in favor of the theory.
• The Prader-Willi/Angelman cluster at 15q11–q13 is excluded; its exclusion limits the completeness of the assessment.
• Functional data are scarce for many imprinted genes, and knockout phenotypes may miss subtle effects that are nonetheless evolutionarily significant.
• The kinship theory’s predictions for BAT function depend on ambient temperature during rearing — effects may have been systematically missed because most knockout studies are conducted at warm temperatures.
• The review is qualitative and interpretive; evolutionary explanations for imprinted loci are inherently post hoc and difficult to test experimentally.
• For the Euarchonta loss of IGF2R imprinting, testing the prediction that growth-inhibitory effects are attenuated in primates requires experimental manipulation that is not feasible in humans.

Relevance to Clonal Evolution

Haig’s kinship theory addresses organismal (germline-level) evolution, not somatic evolution. However, the paper’s subject matter — genomic imprinting — has direct consequences for cancer evolution that make it highly relevant to clonal evolution research:

1. Functional haploidy and single-hit TSG inactivation. At an imprinted tumor suppressor gene, only one allele is transcriptionally active. A single mutational or epigenetic event — affecting the sole active allele — can abrogate gene function, whereas a biallelically expressed TSG requires two hits (Knudson). This means imprinted loci are “pre-sensitized” for loss-of-function in somatic cells.

2. Loss of imprinting (LOI) as an epigenetic driver mechanism. Haig discusses LOI of IGF2 in Beckwith-Wiedemann syndrome (a developmental overgrowth disorder) and in children with isolated somatic overgrowth. LOI at IGF2 — whereby the normally silent maternal allele becomes activated — is also a well-established finding in many sporadic cancers (colorectal, Wilms tumor, hepatoblastoma). LOI represents an epigenetic driver event that increases the dosage of a growth factor without any coding sequence mutation, and it can occur as an early step in clonal expansion.

3. Imprinted growth-regulatory genes are enriched at the key nodes of cancer pathways. Several of the imprinted genes Haig reviews — IGF2 (growth factor), CDKN1C (cell cycle inhibitor), H19 (noncoding RNA with context-dependent tumor-suppressive and oncogenic roles), DLK1 (adipogenesis/differentiation regulator) — are directly implicated in proliferation control. The kinship theory provides an evolutionary explanation for why imprinted genes are disproportionately involved in growth regulation: the mother-offspring conflict over resource allocation creates selective pressure for parent-specific expression specifically at loci that influence growth. This evolutionary constraint logic means that the set of imprinted genes is enriched for exactly the functional categories that, when dysregulated, provide a fitness advantage to somatic clones.

4. LOI-driven clonal expansion. When LOI occurs at a growth-promoting imprinted locus (e.g., IGF2) or at a growth-inhibitory locus (e.g., silencing of H19 or CDKN1C), the affected cell gains a proliferation advantage. Haig’s review documents the antagonistic relationship between IGF2 and CDKN1C — both in the same imprinted cluster — meaning that epigenetic dysregulation of the entire cluster (as occurs with KvDMR1 methylation defects) can simultaneously activate a growth enhancer and silence growth inhibitors. This creates conditions for a clone to undergo clonal-expansion driven entirely by epigenetic rather than genetic changes.

5. Tissue-specific imprinting and tumor specificity. Haig emphasizes that imprinting can be tissue-specific (e.g., Gs-alpha is biallelic in most tissues but monoallelic in renal proximal tubules, thyroid, and pituitary). This implies that the vulnerability to single-hit inactivation is also tissue-specific, which may contribute to tissue tropism of certain cancer types associated with imprinted gene dysregulation.

6. BAT vs. WAT allocation as a metabolic trade-off parallel to cancer metabolism. Haig’s huddling hypothesis frames the BAT/WAT allocation as a physiological trade-off driven by kin selection. In somatic evolution, metabolic trade-offs between biosynthesis (anabolism for proliferation) and energy dissipation may be analogously relevant — cancer cells rewire metabolism (Warburg effect), and understanding how imprinted genes regulate metabolic allocation may shed light on metabolic vulnerabilities that clonal-expansion exploits.