Opportunities and challenges
Overall, our literature review indicates a strong interest in the consequences of population outcrossing across ecology and evolutionary biology. Yet, explicit and robust empirical estimates of heterosis in the context of natural dispersal within metapopulations are apparently remarkably lacking. The few broadly relevant studies are still mainly inspired by traditional research agendas of agriculture, conservation genetics and speciation fields. Consequently, they tend to focus on highly inbred and/or highly differentiated populations, or do not attempt to characterize the connectivity level across populations studied. Crosses between isolated populations provide valuable information on the genetic architecture of population differentiation and on possible ecological and evolutionary consequences within contexts such as genetic rescue or biological invasion. However, natural dispersal and interbreeding between these populations are unlikely, unless these populations come under secondary contact.
Although many study systems have been used to quantify demographic consequences of dispersal in natural populations (e.g. see Millon et al. 2019), most such study systems were notable for their absence from the list of retained studies that fulfilled our current criteria (Table 1). This may be because quantifying multi-generational fitness in wild populations is certainly challenging. For instance, 18 years of fieldwork still resulted in a small sample size for natural immigrants in song sparrows (Marr et al. 2002), constraining the strength of evidence for heterosis. However, the increasing number and length of available individual-based field studies should soon make such analyses feasible in at least some systems (e.g. reviewed in Clutton-Brock & Sheldon 2010). Our search found several studies contrasting fitness of immigrants and residents but that did not investigate the fitness of descendant generations, even if some fitness metrics include aspects of offspring fitness (e.g. number of fledglings in birds). The existence of such studies implies that failure to explicitly quantify and compare target individuals is not entirely due lack of data, but at least partly reflect that such ambitions are not currently on the radars of many population and evolutionary ecologists.
Logistical considerations are certainly prohibitive for some taxa. For example, invertebrates and other small sized taxonomic phyla are traditionally studied under artificial conditions, often using laboratory stock populations, due to the difficulties marking and identifying individuals in their natural environments. As a consequence, future estimates of heterosis in natural populations cannot rely on strategies such as long-term individuals-based studies, which typically involve vertebrates (but see “new methods” below). For other taxa, however, traditional approaches may simply require reconsideration. Ease of manipulation may present such a cost-effective approach in comparison to field-based parentage analyses in plants, that experimental crosses are often the method of choice. Alternatively, these choices may represent a historical oversight that is only now being corrected (see e.g. Ellstrand 2014).
The preponderance of experimental crossing approaches, however, may be another symptom of the disconnect between research silos. Indeed, we found that studies often presented methodological priorities that greatly limit inferences about demographic and evolutionary outcomes of outbreeding within the context of dispersal and resulting immigration in natural populations. For example, lack of knowledge regarding the origin of parental individuals may result in categories of parental and filial generations that include a mixed set of ancestries. Consequently, both means and variances within the categories compared are affected, rendering results difficult to interpret. Additionally, by not investigating the natural occurrence of interbreeding between populations, we lack information on the frequency with which different types of individuals are produced within a population. It then becomes difficult to predict the ultimate eco-evolutionary consequences of genetic effects manifested through outcrosses. For instance, even if F2 descendants of immigrants have very low fitness, any impact would be trivial if F2s are rarely conceived in the first place. F2s could be rare, even in the case of high fitness of F1s, if F1s rarely mate due to non-random mating within and among immigrant lineages, or if F1s are themselves rare and/or temporally segregated.
Note, however, that even our proposed research agenda may lead to variable recommendations regarding experimental approaches. In principle, the ability to manipulate breeding presents several advantages, such as a larger sample size across categories. In addition, artificial or semi-artificial experimental setting provide easier means to rear offspring under several environmental conditions and, for some taxa, the ability to rear different generations under the same environmental conditions. As maternal effects might be significant, experimental crossing also gives the opportunity for the systematic incorporation of such effects in the experimental design. Moreover, combining crosses between residents and immigrants from different sources into a global effect of outcrossing may be preferable in certain contexts. Logistically, it may be impossible to categorize immigrants into their exact population of origin and, even when possible, low source-specific immigration rates likely would prevent disentangling the heterotic effects across pairwise population combinations or filial generations. More importantly, when attempting to understand the eco-evolutionary consequences of immigration within the perspective of a focus population, the average effect of different population crosses may better represent the relevant outcome to the eco-evolutionary dynamics of that population. Therefore, methods and statistical considerations will depend on several logistic opportunities and limitations of the individual taxa/populations under study. In any case, methods applied must involve proper characterization and reporting of statistical expectations and errors.