Heterosis via non-additive genetic effects
Understanding and predicting the fitness outcomes of outcrossing has been a core ambition in biology since at least the 19th century, motivated by agriculture and artificial breeding reports of high yield of hybrid offspring (reviewed in Crow 1987). Specifically, when highly inbred lines derived from individuals within single or weakly differentiated populations are crossed, F1 offspring commonly show “hybrid vigour” or “outbreeding enhancement”, where values of focal phenotypic traits exceed the mean across the parental lines, or even exceed the maximum parental value (Burke & Arnold 2001, Crow 1987, Lynch & Walsh 1998). These early findings fueled extensive study on “heterosis”, describing [positive or negative] deviations of offspring phenotypic value from the mid-parental phenotype expected under additive genetic effects (Burke & Arnold 2001, Lynch & Walsh 1998; Box 1, Fig. 3). This work has subsequently generated theoretical and empirical insights with consequences well beyond their applications to agriculture.
In wild populations, positive heterosis is commonly observed as the result of outbreeding with conspecifics that arrive or are artificially translocated to alleviate inbreeding depression in highly inbred populations (termed ‘genetic rescue’; Fig. 2; Derry et al. 2019, Frankham 2015, Hoffmann et al. 2021, Ingvarsson 2001, Whiteley et al. 2015). These beneficial effects have been shown to span multiple generations (Derry et al. 2019, Frankham 2016), leading to increased population growth (e.g. Åkesson et al 2016) and rapid fixation of immigrant’s genetic variants (e.g. Adams et al 2011). Here, heterosis primarily reflects genetic dominance (Box 1, Fig. 3A,C), via heterozygosity masking the consequences of deleterious recessive alleles or via overdominance (Charlesworth & Willi 2009, Crow 1948, Lamkey & Edwards 1999).
Meanwhile, at the other end of the spectrum of parental divergence, hybrid offspring of crosses between distantly related populations, including inter-(sub)species crosses, often show decreased fitness (termed negative heterosis or outbreeding depression; Lynch 1991, Rhymer & Simberloff 1996, Tallmon et al. 2004, Waser & Price 1989; Fig. 2). While hybrid inviability or sterility may imply reproductive isolation between lineages, these crosses often show imperfect isolation in the F1 offspring or fitness reductions that are only apparent in later generations (e.g. F2 and backcrosses). This may allow introgression of locally maladaptive variants that can lead to genetic or population extinction (Allendorf et al. 2004, Rhymer and Simberloff 1996). These consequences are often caused by negative effects of epistasis due to incompatibilities between highly diverged genomes (genetic, intrinsic or Dobzhansky-Muller incompatibility; Box 1, Fig. 3A,D) or by loss of within-lineage coadapted epistatic interactions (Demuth & Wade 2005, Dobzhansky 1950, Edmands & Deimler 2004, Orr 1995, Rundle & Whitlock 2001).
Despite such broad generalisations, it is now clear that fitness outcomes of interbreeding are far more complex. Crossing inbred lines can cause decreased fitness in hybrid offspring (Derry et al. 2019, Edmands 2007, Lynch 1991). Conversely, interspecific crosses may commonly generate beneficial effects, that even last to the F2 generation (Brice 2021). Furthermore, such crosses can result in novel or transgressive phenotypes (i.e. outside of the limits of parental phenotypes, e.g. best-parent heterosis) for fitness and fitness-related traits in F1 offspring (Atsumi et al. 2021, Grant & Grant 2010, Rieseberg et al. 1999). In turn, transgressive phenotypes have been linked to increased adaptive potential (Edelman & Mallet 2021), and increased ability of hybrids to colonize new environments (Pfennig et al. 2016, Rius & Darling 2014). Therefore, contrary to traditional expectations, crosses between either inbred lines or highly divergent lineages may result in either negative or positive effects.
This complexity in outcomes derives from two main facts. First, fitness consequences depend on more than just the degree of genetic divergence between parental lineages, but also on an axis of variation pertaining to the degree of ecological (i.e. adaptive) divergence between the parental lineages (Fig. 2). Here, fitness decreases with greater divergence between the parental environments as a consequence of genotype-environment mismatches, leading to extrinsic or ecologically dependent incompatibility even at the lower end of the spectrum of genetic divergence (Derry et al. 2019, Edmands & Deimler 2004, Rundle & Whitlock 2001, Thompson et al. 2022a,b). Second, both positive or negative heterosis can result from combinations of various genetic effects, including dominance, under/overdominance and epistasis, as outlined by traditional line-cross theory (Box 1; Boeven et al. 2020, Fu et al. 2015, Guo et al. 2014, Kusterer et al. 2007, Lynch 1991, Lynch and Walsh 1998, Mather & Jinks 1982, Shapira et al. 2014). Fitness outcomes, therefore, may depend not on one of the genetic effects alone, but on a balance between the relative magnitude of them, resulting in outcomes that can be unique to individual crosses and contexts (e.g. Armbruster et al. 1997).
In fact, fitness outcomes vary across traits that differ in underlying genetic architecture, and change across filial generations, because of changes in the magnitude of the different genetic effects (Burke & Arnold 2001, Brice et al. 2021, Fox et al. 2004, Lamkey & Edwards 1999, Lynch 1991, Roff & Emerson 2006, Rundle & Whitlock 2001, Rius & Darling 2014, Zhang et al. 2021). For instance, fitness components are often more affected by outcrossing than other traits (Whitlock et al. 2013), in line with findings that dominance and epistasis are more common in life history than in morphological traits (Roff & Emerson 2006). In turn, fitness gains in F1 offspring can be partially or entirely wiped out, or even reversed, in the F2 or subsequent generations, due to reduced dominance and increased epistatic interactions across filial generations (Box 1, Fig. 3D). Similarly, gene-by-environment interactions may change the relative magnitudes of genetic effects across rearing environments experienced by the hybrid offspring (e.g. Fox et al. 2004). Finally, loss of beneficial epistatic interactions that commonly arise from adaptive ecological differentiation (e.g. Carroll et al. 2003) can outweigh benefits of increased heterozygosity even for highly inbred populations in an environment-dependent manner (Edmands 2007, Frankham et al. 2011; Fig. 2). Outcomes of outbreeding, therefore, are affected by current environmental conditions experienced by the hybrid offspring, and also by how the genetic architecture of traits have been shaped by past evolutionary responses of the parental lineages.
A recent model of speciation (Dagilis et al. 2019) further emphasizes the complexity in fitness outcomes of outbreeding and how the balance between positive and negative effects connects the extremes of the genetic and ecological divergence continuum. Specifically, in addition to negative epistatic interactions arising in hybrid offspring (i.e. Dobzhansky-Muller incompatibilities), and loss of positive epistatic interaction within lineages (i.e. breakdown of coadapted gene complexes), Dagilis et al. (2019) allowed outcrosses to result in positive epistasis and loss of negative epistasis. Under these conditions, the same genetic effects can explain both positive and negative heterosis along a continuum of genetic and ecological divergence, as the balance between positive and negative epistatic effects (i.e. the fitness of the hybrid offspring) changes across the divergence continuum. Interestingly, the net balance between effects is extremely variable in the early stages of lineage divergence, due to the relative influence of drift and selection on focal loci (Dagilis et al. 2019). This is exactly the parameter space that demes within fragmented wild populations typically occupy, highlighting the potential for genetic effects to generate complex and dynamic forms of both positive and negative heterosis following natural dispersal and outcrossing in metapopulation systems.
Indeed, manifestation of heterosis does not require crosses between highly diverged or inbred lines or demes, but can arise from crosses among demes that show genetic structuring (i.e. population level inbreeding) due to adaptive (i.e. ecological) divergence or independent random fixation of alleles via genetic drift fostered by restricted dispersal (Barton 2001, Chevin et al. 2014, Crow 1948, Fenster et al. 1997, Lamkey & Edwards 1999, Simon et al. 2018, Whitlock et al. 2000). This scenario likely comprises most wild populations, since local adaptation is common across natural systems (Hereford 2009, Leimu & Fischer 2008) and genetic structure not only occurs in populations with geographical separation, but also along clines or center-edge continuum of distribution ranges (García-Ramos & Kirkpatrick 1997, Koski et al. 2019, Kottler et al. 2021). These empirical and theoretical developments suggest that heterosis could be widespread or even ubiquitous in wild systems, raising key questions of what fitness outcomes can - and generally do - arise following crosses between populations with intermediate degrees of genetic differentiation (Fig. 2).
Given that the same genetic effects shape the fitness of outcrossed offspring across the biological spectrum of genetic divergence, predictions regarding observed fitness outcomes of outcrossed offspring resulting from natural immigration can be formulated. For example, fitness consequences depend on the specific genetic architecture of a trait within a population and between the pair of populations, and can vary across life stages, sexes, and environments. Outcomes should consequently not be universal across traits, across life stages, or between sexes, and should vary among different pairwise combinations of populations of the same species, or even when the same pairwise combination experiences different environments. Yet, dominance effects should be more pronounced as non-adaptive genetic divergence between demes increases, generating a positive relationship between hybrid fitness in the F1 generation and pairwise genetic distance or divergence time between parental populations (Whitlock et al. 2000).
However, interactions with the environment may affect predicted outcomes. For instance, if dominance effects are environment-dependent, relative fitness for F1 and F2 offspring should vary across parental environments. If, instead, epistatic interactions are environment-dependent, F1 and F2 fitness should be similar across parental environments, but environment-dependent effects should emerge in reciprocal backcrosses (Dagilis et al. 2019, Rundle & Whitlock 2001). In the context of natural dispersal between spatially structured populations, backcrosses resulting from matings between F1 and residents (i.e. B1 or B2) will most likely live in the environment of the parental population that composes most of their genome (i.e. P1 or P2, respectively). Therefore, the fitness of these backcrosses should be higher than that of the F2 in these respective environments, due to effects of both dominance and epistasis (Edmands & Deimler 2004, Rundle & Whitlock 2001).
Finally, since coadapted gene complexes may play a big role in local adaptation, epistasis may be stronger than dominance in cases of ecologically driven divergence. For instance, populations may adaptively evolve within-lineage positive epistatic effects that, if broken during outbreeding, reduce the fitness of the hybrid offspring. Consequently a negative relationship between hybrid fitness and genetic distance between parental populations. Under such genetic architecture, fitness gains via dominance in F1 offspring may be outweighed by the loss of positive epistatic effects at larger genetic distances, leading to an optimal intermediate crossing distance between lineages (Lynch 1991, Price & Waser 1979; see also Fenster 1991). As beneficial dominance effects decrease, and negative between-lineage epistatic effects appear due to recombination, fitness declines for hybrids along the divergence continuum may become particularly severe for the F2 offspring and further recombinant generations. These conclusions mirror those of simulations using line-cross theory, whereby fitness of F1 and F2 offspring in respect to the mid-parent value decline as epistasis becomes stronger than dominance (Lynch 1991).
Overall, these theoretical developments reveal the degree to which complex multigenerational fitness effects resulting from natural immigration could arise, and provide tractable routes to predicting and rationalising observed variation.