Underdominance

Underdominance, also termed heterozygote disadvantage, is a genetic and evolutionary condition in which diploid heterozygotes (A₁A₂) at a locus or chromosomal structural rearrangement exhibit strictly lower Darwinian fitness than both corresponding homozygotes (A₁A₁ and A₂A₂), such that w₁₂ < w₁₁ and w₁₂ < w₂₂. It is the exact mathematical converse of overdominance (heterozygote advantage): whereas overdominance acts as a form of balancing selection that maintains stable genetic polymorphism at an intermediate allele frequency, underdominance generates disruptive positive frequency-dependent selection that drives a population toward fixation of whichever allele exceeds a critical threshold frequency (p̂).

In deterministic population genetics, an underdominant locus possesses an unstable internal equilibrium at allele frequency p̂ ∈ (0, 1). If the initial frequency p₀ of allele A₁ falls below p̂, natural selection eliminates A₁ (p → 0); if p₀ exceeds p̂, natural selection drives A₁ to complete fixation (p → 1). Because rare newly mutated alleles arise at an initial frequency of p₀ = 1/(2*N*) ≪ p̂, nearly every novel underdominant mutation appears initially in heterozygous form and is purged by selection before it can reach the threshold required for fixation.

Despite this deterministic barrier against the invasion of new mutations—often referred to as Wright's Paradox of Chromosomal Evolution—underdominant structural rearrangements such as reciprocal translocations, Robertsonian fusions, and pericentric inversions frequently distinguish closely related eukaryotic species and anchor parapatric hybrid zones. As analyzed by Sewall Wright (1941) and Russell Lande (1979), fixation of strongly underdominant rearrangements requires stochastic genetic drift in small, semi-isolated demes, inbreeding, or coupling to meiotic drive to push a novel arrangement across the unstable adaptive valley (p > p̂), after which it acts as a postzygotic barrier contributing to reproductive isolation. In contemporary synthetic biology, this bistable threshold property is engineered into underdominance gene drives to achieve spatially confined replacement of disease-vector mosquito populations.

Allele-Frequency Dynamics and the Unstable Threshold p̂

Consider a single bi-allelic autosomal locus in a randomly mating diploid population with alleles A₁ at frequency p and A₂ at frequency q = 1 − p. Under Hardy–Weinberg proportions prior to viability or fertility selection, zygotes occur at frequencies p² (A₁A₁), 2*pq* (A₁A₂), and q² (A₂A₂). Assign relative fitnesses to the three genotypes using selection coefficients s₁ > 0 and s₂ > 0 measured relative to the heterozygote, or equivalently by normalizing homozygote fitnesses:

w₁₁ = 1 − s₁, w₁₂ = 1 − s_h, w₂₂ = 1 − s₂ (with s_h > s₁, s₂)

For algebraic convenience, setting the heterozygote fitness reduction relative to each homozygote as S₁ = w₁₁ − w₁₂ > 0 and S₂ = w₂₂ − w₁₂ > 0 (or in the standard normalized form w₁₁ = 1, w₁₂ = 1 − s, w₂₂ = 1 + t, where S₁ = s > 0 and S₂ = s + t > 0), the mean fitness w̄ of the population is:

w̄(p) = p²w₁₁ + 2*pqw*₁₂ + q²w₂₂ = w₂₂ − 2*q*(w₂₂ − w₁₂)*p* + (w₁₁ − 2*w*₁₂ + w₂₂)*p*²

By Wright's fundamental equation of allele-frequency change, the single-generation change Δ*p* = *p*' − p is:

Δ*p* = [ p q / (2*w̄*) ] (dw̄ / dp) = [ p (1 − p) / w̄ ] [ (S₁ + S₂)*p* − S₂ ]

Setting Δ*p* = 0 reveals three equilibrium states:

1. Fixation of A₂ (p = 0): Locally stable because Δ*p* < 0 for 0 < p < p̂.
2. Fixation of A₁ (p = 1): Locally stable because Δ*p* > 0 for p̂ < p < 1.
3. Internal polymorphic equilibrium (p = p̂): Obtained by setting (S₁ + S₂)*p̂* − S₂ = 0:

p̂ = S₂ / (S₁ + S₂) = (w₂₂ − w₁₂) / (w₁₁ − 2*w*₁₂ + w₂₂)

In the standard parameterization where w₁₁ = 1 − s₁, w₁₂ = 1 − (s₁ + s₂), or where w₁₂ = 1 and homozygotes have fitness 1 + s₁ and 1 + s₂, this simplifies directly to p̂ = s₂ / (s₁ + s₂). When underdominance is symmetric (w₁₁ = w₂₂ = 1 and w₁₂ = 1 − s), S₁ = S₂ = s, and the unstable threshold sits exactly at p̂ = ½ (50%).

Mean Fitness Landscape and Adaptive Valleys

Evaluating the second derivative of mean population fitness w̄(p) with respect to p demonstrates why p̂ is unstable:

d²w̄ / dp² = 2(S₁ + S₂) > 0

Because d²w̄ / dp² is strictly positive, p = p̂ represents a global minimum (an adaptive valley) on Wright's mean fitness surface w̄(p), whereas the two homozygote fixation boundaries p = 0 (w̄ = w₂₂) and p = 1 (w̄ = w₁₁) represent two alternative local adaptive peaks. Because natural selection under random mating increases mean population fitness (Δ*w̄* ≥ 0), a population initialized anywhere to the left of the valley floor (p < p̂) climbs the left slope to p = 0, while a population initialized to the right (p > p̂) climbs the right slope to p = 1.

Chromosomal Rearrangements and Wright's Shifting Balance

The most pervasive natural manifestation of underdominance occurs in individuals heterozygous for major chromosomal structural rearrangements:

• Reciprocal translocations: When non-homologous chromosomes exchange segments, structural homozygotes (either standard/standard or translocation/translocation) undergo normal bivalent pairing at meiosis I and produce 100% euploid gametes. In a translocation heterozygote, however, homologous pairing requires forming a quadrivalent cross configuration at pachytene. During anaphase I disjunction, alternate segregation yields balanced (euploid) gametes, whereas adjacent-1 and adjacent-2 segregation yield unbalanced gametes carrying duplications and deficiencies that cause embryonic lethality or gametophytic abortion. Because alternate and adjacent segregation occur at roughly equal frequencies, translocation heterozygotes suffer approximately 50% semisterility (s ≈ 0.50).
• Pericentric and paracentric inversions: If a single crossover occurs within the inverted loop of an inversion heterozygote during meiosis I, half of the resulting chromatids carry large duplications and deletions (forming dicentric bridges and acentric fragments in paracentric inversions, or unbalanced recombinant chromosomes in pericentric inversions), reducing heterozygote fertility in proportion to the genetic map length of the inverted segment unless crossover suppression or achiasmate meiosis intervenes.

Russell Lande (1979) calculated the fixation rate U of a novel chromosomal rearrangement with heterozygote disadvantage s arising at mutation rate μ in a finite diploid population of effective size N_e using Kimura's diffusion approximation:

U ≈ 2 N_e μ √(s / [ π N_e ]) exp(−*N*_e s)

Because the fixation probability decays exponentially with rate constant N_es (proportional to exp[−*N*_es]), a translocation with s = 0.50 has a virtually zero probability of fixing in a population of N_e = 1,000 (e⁻⁵⁰⁰ ≈ 10⁻²¹⁷), but fixes readily if the species is subdivided into very small local demes (N_e ≈ 5–20) experiencing frequent local extinction and recolonization, partial self-fertilization, or female meiotic drive.

Maternal–Fetal Rh(D) Blood Group Incompatibility

Before the introduction of prophylactic anti-D immunoglobulin (RhoGAM) in the late 1960s, the human Rhesus (Rh) blood group locus (RHD on chromosome 1p36.11) operated under asymmetric underdominance. Individuals homozygous for the deletion of the RHD gene (dd) are Rh-negative, whereas DD and Dd individuals express the Rh(D) transmembrane antigen on erythrocytes.

When an Rh-negative (dd) mother carries an Rh-positive heterozygous (Dd) fetus (fathered by a DD or Dd male), fetal erythrocytes crossing the placenta during parturition sensitize the maternal immune system to synthesize maternal anti-D IgG antibodies. In subsequent pregnancies with Dd fetuses, maternal IgG crosses the placenta and lyses fetal red blood cells, causing hemolytic disease of the newborn (erythroblastosis fetalis). Because every fetus eliminated by maternal–fetal incompatibility is strictly a heterozygote (Dd), selection acts exclusively against Dd offspring at rate s q²p (where q² is the frequency of dd mothers and p is the paternal D allele frequency). As J. B. S. Haldane (1942) pointed out, because the d allele was present at ~40% frequency in western Europe (q ≈ 0.40, p ≈ 0.60 > ½), deterministic underdominance should have slowly purged the rarer d allele unless countered by reproductive compensation in affected families or historical admixture between a pre-Neolithic European population nearly fixed for d (such as proto-Basque populations) and incoming Indo-European/Anatolian populations nearly fixed for D.

In 1968, entomologist Christopher F. Curtis proposed exploiting chromosomal translocations to drive disease-refractory genes into wild vector populations. Modern synthetic geneticists (Davis, Bax, & Greaves 2001; Akbari et al. 2013) construct engineered two-locus or single-locus underdominance gene drives (such as reciprocal toxin–antidote systems UD^MEL) in which construct A carries a maternal toxin and a zygotic antidote to construct B, while construct B carries a maternal toxin and a zygotic antidote to construct A.

Unlike homing CRISPR/Cas9 gene drives, which have an invasion threshold near zero (p̂ ≈ 0) and can spread globally from a tiny accidental release, an engineered underdominance drive with fitness advantage t in engineered homozygotes and lethality s = 1 in heterozygotes exhibits a high, controllable release threshold p̂ ≈ 0.27–0.65. Furthermore, Barton (1979) proved that in continuous space, an asymmetric underdominant allele propagates as a traveling Fisher–KPP wave with velocity c proportional to (s / 2 − s p̂) only if the local release exceeds the critical spatial nucleation radius R_crit ~ σ / √*s*. If the engineered mosquitoes disperse into neighboring non-target populations at low migrant fractions (m < p̂), local selection eliminates the transgene, making underdominance drives reversible via inundation with wild-type males.

Why does underdominance eliminate genetic variation whereas overdominance maintains it?

Under overdominance (heterozygote advantage), when an allele is rare, it resides almost exclusively in high-fitness heterozygotes (A₁A₂), causing selection to increase its frequency toward a stable internal equilibrium. Under underdominance (heterozygote disadvantage), a rare allele resides almost exclusively in low-fitness heterozygotes, causing selection to purge the rare allele and drive whichever allele is currently in the majority (> p̂) to complete fixation.

How can chromosomal rearrangements with 50% heterozygote sterility ever become fixed between species?

In a large panmictic population, a reciprocal translocation causing 50% heterozygote semisterility (s = 0.5) has a vanishingly small fixation probability scaling as exp(−*N*_es). However, as Sewall Wright and Russell Lande showed, if a species is structured into small, partially inbred demes (N_e ≈ 5–20) undergoing periodic bottlenecks and recolonization, random genetic drift can carry the rearrangement across the p̂ = 0.5 threshold in a local deme, from which it spreads by demic expansion.

Why are underdominance gene drives considered safer for localized vector control than homing CRISPR drives?

Homing CRISPR gene drives copy themselves in the germline, giving them an invasion threshold near zero so that a handful of escaped insects can convert an entire species across continents. Underdominance gene drives require the engineered allele to exceed a high threshold frequency (typically p̂ ≈ 30%–65%) to spread; accidental low-frequency migrants into neighboring regions are automatically eliminated by selection, and the drive can be reversed by releasing wild-type individuals.

1. Wright, S. (1941). "On the probability of fixation of reciprocal translocations." The American Naturalist, 75(761), 513–522. https://doi.org/10.1086/280996
2. Lande, R. (1979). "Effective deme sizes during long-term evolution estimated from rates of chromosomal rearrangement." Evolution, 33(1Part1), 234–251. https://doi.org/10.1111/j.1558-5646.1979.tb04678.x
3. Curtis, C. F. (1968). "Possible use of translocations to fix desirable genes in insect populations." Nature, 218(5139), 368–369. https://doi.org/10.1038/218368a0
4. Davis, S., Bax, N., & Greaves, C. (2001). "The engineering of spatial containment using 'underdominance' systems." Journal of Theoretical Biology, 212(1), 83–98. https://doi.org/10.1006/jtbi.2001.2357