MULTIGENERATIONAL OUTCOMES OF OUTCROSSING IN NATURAL POPULATIONS
The theoretical possibilities reviewed above highlight the potentially complex set of fitness outcomes from outcrossing due to natural immigration across spatially structured populations. We therefore conducted a literature review, aiming to evaluate the degree to which key effects have actually been estimated in wild populations where immigrants naturally outbreed with resident individuals, and thereby synthesise the overall body of empirical knowledge. We used search terms on Web of Science aimed to find ecological studies that estimate the fitness of the offspring of natural immigrants, including studies that do not specifically interpret results within the heterosis framework, as well as more traditional heterosis-related search terms (Fig. 5). Due to the high number of results initially found, we filtered for studies within the Web of Science category of “ecology or evolutionary biology”, aiming to exclude studies on domesticated or agricultural lines, resulting in 12,862 studies after duplicate removal. We validated our search by comparing the results to a list of previously known studies by our research group, and triaged the results based on the series of criteria presented in Table 1. The definition of criteria and the triage were done by DG in consultation with HJ and JMR.
We found 111 studies (0.86% of 12,862) on natural populations with at least one estimate of fitness or a fitness-related trait for both within and between population crosses for at least one generation (criteria 1-5; Supplementary Table S1). These studies exclusively focused on animals or plants (Fig. 6), encompassing 89 different species. Focal fitness metrics and traits varied across studies, but commonly included: germination, biomass, number of flowers, fruits and seeds produced and seed mass in plants; and number and size of offspring produced (e.g. clutch and egg size) and developmental time in animals. For both plants and animals, survival across different life stages and metrics of cumulative fitness were also common.
Over a third of the 111 studies (39.6%, N=44) presented crosses between populations for which the degree of connectivity was unclear (Fig. 7), whether due to a lack of knowledge regarding dispersal in the focal populations, lack of mention of it within the paper, or because the information presented was not sufficient for a categorization considering our specified criteria (Table 1, criteria 6). Another large proportion of studies (45%, N=50) included heterosis estimates for hybrid offspring resulting from introduction (N=1) or artificial crosses (N=49) between populations further than the distance of natural dispersal, such as crosses between populations separated by several hundreds of kilometers, and even populations from different continents. More than half of these studies (N=29) only estimated heterosis at this level of connectivity, while the others (N=21) combined estimates across different levels of connectivity (category “both” in Fig. 7). Among the remaining 17 studies (15%), which we could classify as only studying populations that are, or possibly are, interconnected by natural dispersal, only three reported fitness estimates for hybrid offspring resulting from interbreeding due to natural immigration (Fig. 7).
In two of the three studies conducted under the context of natural dispersal (Marr et al. 2002, Martinig et al. 2020), estimates were for the hybrid offspring between residents and natural immigrants, whereby long-term fitness and parentage data (18-29 years) allowed the categorization of individuals into resident or immigrant, and their offspring into the relevant filial generations. While Marr and collaborators (2002) explicitly aimed to estimate the fitness consequences of outcrossing following natural dispersal into an island population of song sparrows (Melospiza melodia ), Martinig and collaborators (2020) aimed to investigate sex-specific fitness consequences of dispersal in the North American red squirrels (Tamiasciurus hudsonicus ).
In the song sparrows, male immigrants showed lower reproductive performance than residents, but no difference in survival. Among comparisons of reproductive performance and survival including hybrid filial generations, no difference was found for hybrid F1s versus average immigrant-resident, hybrid F1s versus F2s, and hybrid F2s versus average F1-immigrant-resident (Marr et al. 2002). However, F1s produced between 27.1% (females) to 30.2% (males) more offspring than the average between immigrant and residents, and about 2 (females) to 3 times (males) more offspring than F2s. F2 males also showed decreased survival in comparison to F1s in both juvenile (46%) and adult (35.1%) stages. These heterotic effects are likely biologically significant, and the lack of strong statistical evidence for these differences possibly resulted from low power due to a small number of immigrants (14 females and 4 males vs. >100 residents). In turn, male immigrants of North American red squirrels had longer lifespans and produced more offspring than residents, while female immigrants had reduced lifetime reproductive success (Martinig et al. 2020). In addition, researchers found that both immigrant males and females produced daughters with lower lifetime reproductive success than resident males and females, respectively. However, no further comparisons or specifications of filial generations were presented, nor a distinction was made between offspring of immigrant-immigrant versus immigrant-resident crosses.
The third study (Bull & Sunnucks 2014) investigated the fitness of hybrid offspring along a contact zone between two races of a velvet worm (Euperipatoides rowelli ), aiming to investigate the maintenance of morphological differentiation between the races. Using body coloration patterns and haplotypes to classify individuals and embryos as either of the races or as hybrids, researchers found no evidence for a difference in embryo-to-adult survival across categories. However, hybrids represented a single category for which genetic admixture proportions deviated 15% in either direction from the pure parental races and, therefore, likely included different filial generations.
The remaining studies classified as using populations that were or likely to be connected by natural dispersal (N=14) conducted experimental crosses to obtain hybrid offspring. Most of these studies estimated fitness under artificial or semi-natural conditions (Fig. 7), with ten studies (71.4% of 14) investigating fitness within a single environmental condition, and only five estimating fitness for at least one component under at least one of the native parental environments. None of the studies using experimental crosses explicitly considered sex differences in F1 heterosis, likely due to focusing on hermaphroditic [plant] species or pre-sexual maturation life stages in animals. Moreover, only five of the 14 studies (35.7%) considered filial generations beyond the F1, including estimates for F2 (N=2), F3 (N=1) or backcrosses (N=3). These observations do not appreciably change when considering all of the 111 studies, with 70.3% (N=78) of the studies including only estimates for the F1 offspring, 70.3% including estimates for a single environmental condition, and only one study (Matsubayashi et al. 2011) that quantified male- and female-specific traits.
Among the studies using experimental crosses that included populations that are or likely are connected by natural dispersal (i.e. categories “Connected” and “Both” on Fig. 7), a mixture of evidence for positive, negative, or no heterosis was found. Heterosis often varied within studies, among traits or pairwise population combinations, and even among replicate crosses of the same populations. In some cases, such differences resulted from rearing environment (Ostevik et al. 2016) or maternal population of origin (Barnard-Kubow et al. 2016, Barnard-Kubow & Galloway 2017), demonstrating that outcomes may depend on environmental and on maternal effects, cytoplasmatic-nuclear genome interactions (Burton et al. 2006), or sex-chromosomal effects (Saavedra & Amat 2005). Some studies also found a relationship between heterosis and the genetic (Barker et al. 2019, Barnard-Kubow et al. 2016, Barnard-Kubow & Galloway 2017), geographical (Barnard-Kubow et al. 2016) or environmental distance (Pickup et al. 2013) between parental populations, or the size and the genetic diversity within populations (Pickup et al. 2013, Willi et al. 2007). Genetic or geographical distance (Barnard-Kubow et al. 2016, Barnard-Kubow & Galloway 2017, Willi & Van Buskirk 2005) and genetic diversity (Barker et al. 2019, Pickup et al. 2013, Willi & Van Buskirk 2005, Willi et al. 2007), however, were also reported to not affect heterosis in other studies or in different traits, environments, or filial generations within the same study. As an interesting example, fitness of hybrid offspring of the herb Campanulastrum americanum showed no relationship with genetic distance estimated from nuclear markers, but decreased with the genetic distance estimated from the predominantly maternally inherited chloroplast markers (Barnard-Kubow et al. 2016, Barnard-Kubow & Galloway 2017; but see Barnard-Kubow et al. 2017). Even further, maternal population of origin reversed the direction of heterosis within both F1 and backcrosses of reciprocal crosses for some traits and population combinations (Barnard-Kubow et al. 2016). Therefore, cytoplasmatic-nuclear interactions may not just affect heterosis levels per se, but also the correlation between heterosis and genetic distance.
Unfortunately, our review also revealed that explicitly quantifying and contrasting conditions under which heterosis is positive, negative, or absent across studies is constrained by forms of analyses and data reporting. Papers commonly presented a single statistical estimate for crosses between several populations, sometimes of different connectivity levels. Moreover, criteria for determination of heterosis varied enormously across studies, ranging from statistical comparisons between mean values of hybrid F1 and mean parental values, to more elaborate calculations involving the average offspring values of within and between population crosses (Supplementary Table S2). Finally, these studies also often implement methodological priorities that may introduce errors or biases within the context of our research agenda. For example, many study designs used a substantial number of offspring (e.g. >20) from a small number of parents (e.g. <5 parent pairs), rather than using offspring generated from more parental combinations. Also common was the use of multiple sires or pollen mixtures (sometimes from different source populations) for fertilization, with the aim of reducing the probability of unsuccessful crosses. Further, unsuccessful crosses were sometimes replaced, to ensure estimation of fruit or flower productivity in the offspring. Although these tactics are justified under alternative research contexts, key information regarding differences across families, pairwise population combinations, degrees of connectivity, or environmental characteristics is lost. Conclusions are also muddied by averaging of positive and negative outcomes within individual generations or pairwise population combinations. Furthermore, eliminating unsuccessful crosses from the data, or removing or confounding individual level variation via use of single or mixed sires, restricts proper variance and error estimations in the statistical analyses, and prevent documentation of cases in which outcrossing does not result in viable offspring. Therefore, estimates and conclusions about heterotic effects from these studies may not provide adequate evidence for heterotic effects within the context of eco-evolutionary consequences of dispersal in wild populations.