The benefits of interpopulation hybridization diminish with increasing divergence of small populations

Authors


Correspondence: Nina Pekkala, Department of Biological and Environmental Science, PO Box 35, FIN-40014, University of Jyväskylä, Jyväskylä, Finland.

Tel.: +358 40 770 5056; fax +358 14 617 239; e-mail: pekkala.nina@gmail.com

Abstract

Interpopulation hybridization can increase the viability of small populations suffering from inbreeding and genetic drift, but it can also result in outbreeding depression. The outcome of hybridization can depend on various factors, including the level of genetic divergence between the populations, and the number of source populations. Furthermore, the effects of hybridization can change between generations following the hybridization. We studied the effects of population divergence (low vs. high level of divergence) and the number of source populations (two vs. four source populations) on the viability of hybrid populations using experimental Drosophila littoralis populations. Population viability was measured for seven generations after hybridization as proportion of populations facing extinction and as per capita offspring production. Hybrid populations established at the low level of population divergence were more viable than the inbred source populations and had higher offspring production than the large control population. The positive effects of hybridization lasted for the seven generations. In contrast, at the high level of divergence, the viability of the hybrid populations was not significantly different from the inbred source populations, and offspring production in the hybrid populations was lower than in the large control population. The number of source populations did not have a significant effect at either low or high level of population divergence. The study shows that the benefits of interpopulation hybridization may decrease with increasing divergence of the populations, even when the populations share identical environmental conditions. We discuss the possible genetic mechanisms explaining the results and address the implications for conservation of populations.

Introduction

Anthropogenic destruction and fragmentation of natural habitats, as well as other mainly anthropogenic factors, cause reductions in population sizes of many species and increase isolation between populations (Ewers & Didham, 2006). Small and isolated populations are prone to extinction not just for demographic but also for genetic reasons (e.g. Saccheri et al., 1998; Bijlsma et al., 2000; Keller & Waller, 2002; Spielman et al., 2004; Frankham, 2005; O'Grady et al., 2006). In the short term, the most important genetic threat faced by small populations is inbreeding depression, that is, the reduced fitness of offspring produced by mating between related individuals. Inbreeding depression results mainly from the unmasking of recessive deleterious alleles in homozygous genotypes, but overdominant loci and epistatic relationships involving dominance may also contribute (Lynch, 1991; Charlesworth & Charlesworth, 1999; Carr & Dudash, 2003; Charlesworth & Willis, 2009). In the long term, accumulation and fixation of deleterious alleles and loss of adaptive potential through genetic drift further decrease the viability of small populations (Lande, 1994, 1998; Lynch et al., 1995a,b; Whitlock, 2000; Bijlsma & Loeschcke, 2012). Alleviating the genetic problems of small and isolated populations has consequently become an important topic in conservation of endangered species and populations.

Introduction of genetic material from other populations has been suggested as a strategy to improve the viability of inbred populations (Tallmon et al., 2004; Hedrick, 2005; Edmands, 2007; Frankham et al., 2011; Hedrick et al., 2011). Interbreeding by individuals from genetically differentiated populations can lead to heterosis, an increase in fitness of the hybrid offspring, because the effects of recessive deleterious alleles are masked in heterozygous individuals and because heterozygosity in overdominant loci is restored (Lynch, 1991; Lynch & Walsh, 1998; Whitlock et al., 2000; Lippman & Zamir, 2007). Positive fitness effects can also follow from new favourable interactions between loci and from disruption of negative interactions that may have been fixed in small populations through genetic drift (Lynch, 1991; Erickson & Fenster, 2006; Edmands et al., 2009). Well-known examples of a ‘rescue’ of the threatened population by augmentation of gene flow are those of an adder (Vipera berus) population in Sweden (Madsen et al., 1999), the Florida panther (Puma concolor coryi) (Johnson et al., 2010) and the greater prairie chicken (Tympanuchus cupido pinnatus) in Illinois (Westemeier et al., 1998). In many studies that involve hybridization between natural populations, it is not clear how much of the success of the management effort can be attributed to demographic reasons and how much can be attributed to genetic reasons. Nevertheless, convincing evidence for the genetic benefits of interpopulation hybridization comes from many experimental studies (reviewed in Tallmon et al., 2004).

Mixing of genetic material from differentiated populations can, however, also result in outbreeding depression, that is, in a decrease in fitness of the hybrid offspring below that of the nonhybrid parents (e.g. Parker, 1992; Gharrett et al., 1999; Fenster & Galloway, 2000; Edmands & Deimler, 2004; Galloway & Etterson, 2005; Burton et al., 2006). An obvious mechanism for outbreeding depression is the disruption of local adaptations (Templeton, 1986; Lynch & Walsh, 1998), but outbreeding depression can also occur in crosses among populations that have adapted to identical environmental conditions. In isolated populations, the combined effects of drift and selection can lead to the evolution of different multilocus genotypes that work well together, the so-called coadapted gene complexes (Templeton, 1986; Lynch, 1991; Fenster et al., 1997; Lynch & Walsh, 1998). Interpopulation hybridization can break up the coadapted gene complexes, resulting in outbreeding depression. Also, bringing together alleles that are neutral or beneficial individually but have deleterious effects when combined can result in outbreeding depression (Phillips & Johnson, 1998; Orr & Turelli, 2001; Edmands, 2007; Presgraves, 2010). There are numerous examples of outbreeding depression from experimental studies of interpopulation hybridization (reviewed in Edmands, 2007). Concern about the possibility of outbreeding depression is one reason for the infrequent use of interpopulation hybridization in the management of endangered natural populations (Frankham et al., 2011).

Whether interpopulation hybridization has a positive or a negative impact on population viability depends on the relative magnitudes of the positive and negative effects of hybridization. One of the key factors predicted to influence the magnitudes of heterosis and outbreeding depression is the level of genetic divergence between the populations (Lynch, 1991; Falconer & Mackay, 1996; Lynch & Walsh, 1998; Whitlock et al., 2000). Without selection, heterosis from masking of recessive deleterious alleles should increase linearly with the divergence of the populations. In contrast, outbreeding depression due to epistatic relationships is expected to develop slowly in the first stages of population divergence, but then develop at accelerating speed as populations become increasingly diverged (Orr, 1995; Orr & Turelli, 2001). Thus, assuming that heterosis results from divergence in single-locus genotypes whereas outbreeding depression involves divergence at two or more interacting loci, the positive effects of hybridization should predominate at low to intermediate population divergence, whereas at higher levels of divergence there is an increasing risk of outbreeding depression. Consistent with this expectation, an optimal outcrossing distance has been found at intermediate (most often geographic) distances in several plant species (reviewed in Edmands, 2002). However, other types of relationships (including no clear relationship) between parental divergence and offspring fitness have also been reported (Edmands, 2002; Galloway & Etterson, 2005; Willi et al., 2007).

Another factor that can influence the outcome of interpopulation hybridization is the amount of genetic variation introduced. As hybridization can have both positive and negative effects on population viability, determining the optimal amount of introduced genetic variation is difficult. In general, it appears that rather low levels of immigration are enough to cause an increase in fitness of a small population (reviewed in Mills & Allendorf, 1996; Tallmon et al., 2004). However, in some cases, a long-term fitness increase may be achieved only if higher amounts of genetic variation are introduced (Bijlsma et al., 2010; Hedrick & Fredrickson, 2010). On the other hand, even low levels of immigration from a genetically incompatible population can potentially cause considerable damage (Mills & Allendorf, 1996; Edmands & Timmerman, 2003). In the management of threatened populations, the amount of genetic variation introduced can be manipulated with the choice of the source population (i.e. by using a source population that has either low or high genetic variability), with the number of introduced individuals or with the number of source populations that are genetically different from each other.

The effects of interpopulation hybridization can also vary between generations following the hybridization. Heterozygosity peaks in the first hybrid generation and is diluted thereafter, predicting that heterosis should also peak in the first generation and decrease in subsequent generations. Fitness may be reduced in generations following the hybridization also because recombination further breaks up parental gene combinations and exposes harmful epistatic interactions involving recessive alleles (Lynch, 1991; Lynch & Walsh, 1998). Indeed, even though outbreeding depression is often already observed in the first hybrid generation, the negative effects of hybridization are generally more severe in the second generation after hybridization (reviewed by Tallmon et al., 2004; Edmands, 2007). Then again, outbreeding depression may also be a relatively short-lived phenomenon, as it is possible that natural selection removes the unfit genotypes from a population (Templeton, 1986). Among the few studies that have been continued beyond the first or second generation after hybridization, some have found evidence that outbreeding depression is not long lasting (Edmands et al., 2005; Erickson & Fenster, 2006; Hwang et al., 2011).

The various factors that can alter the outcome of interpopulation hybridization make it difficult to predict the consequences of hybridization in any particular occasion. Disentangling the influence of various factors on the long-term effects of interpopulation hybridization is important for informed management of endangered species and populations. Despite the urgent need for this kind of knowledge, previous studies on the effects of interpopulation hybridization that go beyond the second or third generation after hybridization are scarce (but see Edmands et al., 2005; Erickson & Fenster, 2006; Bijlsma et al., 2010; Hwang et al., 2011).

Our main objectives in this study were to experimentally test the effects of (i) the level of population divergence and (ii) the number of source populations on the long-term consequences of hybridization between isolated populations. Using Drosophila littoralis as a model species, we first created inbred populations that were maintained in size of 10 individuals (sex ratio 1:1) throughout the experiment. We then established hybrid populations with offspring from two or four inbred populations at two levels of population divergence (F = 0.30 and F = 0.63). We measured viability of the inbred and hybrid populations for seven generations after the hybridization events as proportion of populations facing extinction and as offspring production measured relative to a large control population. We found the outcome of interpopulation hybridization to be influenced by the level of population divergence, but not by the number of source populations.

Materials and methods

Establishment of the inbred populations and the control population

A laboratory population of the boreal drosophilid Drosophila littoralis was established in May 2006 from 157 males and 99 females collected from a natural population by the Tourujoki River in Jyväskylä, central Finland. A total of 34 of the 99 females had been inseminated in the wild and produced fertile eggs after transfer to the laboratory. The rest of the females were mated randomly with the wild-caught males. Population size was increased to approximately 400 breeding pairs at the second laboratory generation (F2). For the first four laboratory generations (F1–F4), the population was maintained by breeding individual pairs that were chosen randomly among nonsiblings, thus reducing inbreeding in the population. For the next two generations (F5–F6), the flies were allowed to mate randomly with nonoverlapping generations as a population of approximately 500 breeding pairs.

At the seventh laboratory generation (F7, referred to as generation 0 from here on), 16 small populations of 10 individuals each (five males and five females) were established from the laboratory population (hereafter inbred populations) (Fig. 1). Also, a large control population consisting of 500 individuals (250 males and 250 females) was established, in order to control for possible environmental variation in time and thus to make the comparison between different time points easier (Lynch & Walsh, 1998 p. 263). All populations were maintained at the same density of breeding adults (five males and five females) per bottle, with nonoverlapping generations. Thus, the inbred populations consisted of one bottle each, and the control population consisted of 50 bottles. In addition, one to three extra bottles were set up for each inbred population at each generation (when enough adult flies were available) to protect against the loss of populations and to get enough adult flies for use in other experiments. However, each new generation of the inbred populations was always continued with randomly chosen flies from only one of the bottles. The control population was kept panmictic by mixing the offspring from the 50 bottles prior to randomly picking the flies for the subsequent generation (see 'Maintenance of the populations' for details).

Figure 1.

Schematic representation of establishment of the study populations. The control population and the inbred populations were maintained throughout the experiment. The hybrid populations were maintained for seven generations. Low divergence (LD) hybrid populations were established at the low level of divergence between the inbred populations (F = 0.30). High divergence (HD) hybrid populations were established at the high level of divergence between the inbred populations (F = 0.63). N = number of individuals in a population, rep. = number of replicate populations.

Establishment of the hybrid populations

To manipulate the level of population divergence, the hybrid populations were established using offspring collected from the inbred populations at two points in time: after 7 and after 15 generations from the establishment of the inbred populations (Fig. 1). The level of divergence between the inbred populations was estimated from genetic variation detected at eight nuclear microsatellite loci (see 'Estimation of population divergence'). The hybrid populations established at the low level of population divergence (generation 7, F = 0.30) will be referred to as the LD hybrid populations (LD as in ‘Low Divergence’). The hybrid populations established at the high level of population divergence (generation 15, F = 0.63) will be referred to as the HD hybrid populations (HD as in ‘High Divergence’).

To manipulate the number of source populations, the hybrid populations were established with offspring collected from either two or four inbred populations. At the low level of population divergence, we established 16 populations with individuals from two inbred populations and 16 populations with individuals from four inbred populations. The combinations of inbred source populations were assigned randomly, while ensuring that each combination was different from the others. Only males or females were taken from any inbred population when constructing a hybrid population, so that all first generation offspring produced in the hybrid populations would be hybrids (Table S1, provided as online supplementary, identifies the source population of each male and female fly used in establishment of the hybrid populations). At the high level of population divergence, we established ten populations with individuals from two inbred populations and nine populations with individuals from four inbred populations. The lower number of hybrid populations at the high level of population divergence was due to the loss of four of the 16 inbred populations and decreased offspring production in the extant inbred populations (only nine of the 16 inbred populations were continued at generation 15). When establishing the hybrid populations at the high level of population divergence, when possible, we used the same combinations of source populations that were used at the low level of population divergence. In addition, some new combinations of source populations were randomly assigned to replace previously used combinations that were no longer available because of extinctions and low productivity of the inbred populations (see Table S1). Each hybrid population was established with five males and five females and maintained at the same density of breeding adults, with nonoverlapping generations, by starting each new generation with randomly chosen flies from the previous generation (see 'Maintenance of the populations' for details).

Maintenance of the populations

The populations were maintained in plastic bottles containing 50 mL of malt medium (Lakovaara, 1969) at 19 °C with relative humidity of 60% and with constant light. Under the rearing conditions used, the egg-to-adult development time of the flies was approximately 3 to 4 weeks. For generations 0–9 (from the establishment of the inbred populations and the control population), we maintained the populations as follows: At the start of each generation, five mature, randomly chosen males and females (age, 16–23 days from eclosion) from the previous generation were placed in a bottle to mate and lay eggs. After 5 days in the bottle, the parental flies were removed. To avoid selection for early reproduction and fast egg-to-adult development and to time the collection of the offspring to the peak emergence time, the first eclosed offspring from each population were counted and discarded 21 days after the removal of parental flies. Seven days later (28 days after the removal of parental flies), all newly eclosed offspring were collected, counted and separated according to sex under CO2 anaesthesia. Based on a preliminary experiment (results not shown), under the rearing conditions used, D. littoralis males mature at the earliest 10 days after eclosion. Thus, when collected 0–7 days after eclosion, we could assume that the offspring were unmated. The collected offspring were kept in plastic vials (diameter, 23.5 mm; height, 75.0 mm; 8 mL of malt medium) at maximum density of ten flies per vial and changed to fresh vials every 7 days. Sixteen days from the collection of the offspring, the next generation of each population was started at the defined population size by randomly picking the parental flies from amongst the collected offspring. All the bottles (including the extra bottles) of each inbred population were established with flies collected from a single randomly chosen bottle that was designated a priori to serve as the source of flies for the next generation (we call this bottle the primary bottle). Only in the case that the primary bottle did not produce at least five males and five females (enough flies to start the next generation), a replacement bottle was randomly drawn from amongst the extra bottles, but also then the next generation was started with flies collected from only one bottle (see also 'Extinction').

Ten generations after the establishment of the inbred populations and the control population, we noticed that the emergence of flies took place earlier than before. We implemented no direct selection on timing of reproduction or on egg-to-adult development time of the flies, as the first emerging flies were always discarded. However, the random collection of parental flies for each generation may have caused positive selection on fecundity of the flies. If a genetic correlation exists between fecundity and development time, selection on fecundity may have caused correlated evolution of faster development. To maintain the collection of flies at the peak emergence time (and to avoid causing false increase in population extinction risk), from generation 10 onwards (the 4th generation of the LD hybrid populations), we changed the procedure so that the offspring were collected between 17 and 24 days after the removal of the parental flies, that is, 4 days earlier than before. We kept the generation length constant by starting the next generation 20 days from the collection of the flies. Thus, the age of the flies at introduction to the bottle was now 20–27 days from eclosion. The difference between the two procedures is minor, as the age of the flies at collection is the same (0–7 days from eclosion), and the age at introduction to the bottle is overlapping (16–23 and 20–27 days from eclosion). The change in the maintenance procedure did not affect the measure of offspring production as we continued to count the emerging flies for 28 days from the removal of the parental flies (see 'Offspring production'). Note also that we always measured offspring production relative to the large control population that was changed to the new maintenance procedure at the same time as the inbred and hybrid populations.

Extinction

Offspring production (see below) and extinctions in all established populations (the inbred populations, the control population and the hybrid populations) were followed for seven generations after both hybridization events. The inbred and hybrid populations were considered extinct if they produced less than five female and/or male offspring during the 7 days before the offspring were collected (see 'Maintenance of the populations'). This definition of extinction was chosen because of the need to maintain the breeding population size constant from generation to generation in order to compare offspring production between the inbred and hybrid populations. Defined as such, the extinction risk of the study populations can be considered an overestimate of the true extinction risk. On the other hand, extinction risk is likely to be higher in nature than it is in the favourable conditions of the laboratory. In any case, defined as such, extinction of a population does indicate very low offspring production in the population.

Because the hybrid populations consisted of only one bottle per population, in statistical analyses, also the inbred populations were scored as extinct if not enough offspring were produced from the primary bottle (see 'Maintenance of the populations'). However, the inbred populations were in reality continued from one of the extra bottles when possible. Thus, the inbred populations considered extinct in the statistical analyses of offspring production and extinctions at the low level of population divergence were still used for the establishment of the hybrid populations at the high level of population divergence, if they had been successfully continued from the extra bottles.

Offspring production

Offspring production of the populations was counted for 28 days after removing the parental flies from the bottles. Offspring production of the inbred populations was counted from all available bottles, including the extra bottles, to improve the accuracy of the measure. Thus, the number of bottles used for counting population offspring production was one to four for the inbred populations (most often two), one for the hybrid populations and 50 for the control population. The per capita offspring production was obtained by dividing the total number of offspring by the number of bottles used for counting the offspring (in other words, we thus measured per bottle offspring production, but since each bottle had five pairs of parental flies, our measure corresponds to per capita offspring production). For the inbred and hybrid populations, we calculated the per capita offspring production relative to the per capita offspring production of the control population, measured at the same generation, in order to control for possible environmental variation in time (Lynch & Walsh, 1998 p. 263). When a population was considered extinct (see above), offspring production of the population was recorded as zero from the extinction onwards.

Estimation of population divergence

The level of divergence between the inbred populations at generations 7 and 15 was estimated from genetic variation detected at eight nuclear microsatellite loci. The eight loci chosen for the study (Vir4, Vir11, Vir32, Vir38, Vir99, Mon6, Mon17, Mon26) were polymorphic in the original large population and showed no evidence of linkage disequilibrium (for details, see Routtu et al., 2007). Genomic DNA was extracted from flies preserved in 70–95% ethanol. After air-drying to remove traces of ethanol, the individuals were crushed in a microcentrifuge tube with a hand-held pestle. Qiagen DNeasy Tissue Kit reagents were used for extraction following the manufacturer's protocol modified for use with the Kingfisher magnetic particle processor (Thermo Scientific, Waltham, MA, USA). PCRs were carried out in a volume of 10.5 μL. The reaction mix contained 1× Mg-Free Buffer (Biotools, Madrid, Spain), 200 μm dNTPs (Fermentas, Helsinki, Finland), 1 μm R-primer, 1 μm F-primer, 1.5 mm MgCl2 (Biotools), 1 unit of Taq DNA Polymerase (Biotools), and 1 μL template DNA. The thermocycling conditions included: initial denaturation at 94 °C for 3 min followed by 30 cycles of denaturation at 94 °C, annealing at 52 °C (Mon26) or at 55°C and extension at 72 °C, and a final extension at 72°C for 10 min; using Bio-Rad C1000 or S1000 thermal cyclers (Bio-Rad Laboratories, Hercules, CA, USA). The PCR products were denaturated with formamide together with GeneScan™ 500 LIZ™ Size Standard, separated with an ABI Prism 3130xl Genetic Analyser, and visualized with GeneMapper v.4.0 software (all Applied Biosystems, Carlsbad, California, USA).

For randomly mating isolated populations of equal size derived from the same source population, increase in neutral marker homozygosity (F) equals the increase in among-population variance in allele frequencies, that is, population divergence (Hartl & Clark, 1997). We estimated population divergence by estimating the increase in homozygosity (F) in the inbred populations. The observed (Ho) and expected heterozygosities (He) were calculated with GenAlEx v.6.41 software (Peakall & Smouse, 2006), pooling the data from the inbred populations (Table 1). We then calculated the increase in homozygosity due to finite population size for the inbred populations at generations 7 and 15 as Ft = 1 − Ho,t/He,1(Control), where Ho,t is the observed heterozygosity in the inbred populations at generation t, and He,1(Control) is the expected heterozygosity in the control population at generation 1. At generation 7, the estimated F was 0.30, and at generation 15, the estimated F was 0.63. The control population sustained a high level of heterozygosity throughout the experiment and conformed to Hardy–Weinberg expectations at all sampled generations (Table 1).

Table 1. The results of the genetic analysis. Gen. = sampled generation, npopulations = number of populations sampled, nsamples/population = number of samples per population (mean and minimum), nsamples/locus = number of samples per locus (mean and minimum), He = expected heterozygosity (all inbred populations pooled), Ho = observed heterozygosity (all inbred populations pooled), F is calculated as Ft = 1 − Ho,t/He,1(Control), where Ho,t is the observed heterozygosity in the inbred populations at generation t, and He,1(Control) is the expected heterozygosity in the control population at generation 1
PopulationGen. n populations n samples/population n samples/locus He (SE)Ho (SE) F
Mean (min)Mean (min)
Inbred71611.8 (9)128.1 (27)0.507 (0.078)0.364 (0.058)0.30
151211.6 (1)94.8 (39)0.500 (0.098)0.195 (0.051)0.63
Control1110592.1 (77)0.523 (0.083)0.512 (0.087)0.02
614238.5 (32)0.542 (0.089)0.562 (0.084)−0.07
2413833.5 (17)0.535 (0.085)0.505 (0.073)0.03

Statistical analyses

The effects of hybridization (inbred vs. hybrid populations), the level of population divergence (low vs. high level of divergence) and the number of source populations (two vs. four source populations) on proportion of extinct populations were analysed with Fisher's exact test (2-sided) using PASW Statistics 18 software. The effects of hybridization, the level of population divergence, the number of source populations and generation (generations from the establishment of the LD and HD hybrid populations) on offspring production were examined with mixed model analysis using SAS v9 software (SAS Institute Inc., Cary, NC, USA), accounting for covariance between the observations from the same populations at different generations. To be sure that our analysis was not biased by unequal contributions of populations to the different hybridization treatments (inbred populations vs. hybrid populations), in analyses that involved comparisons between inbred and hybrid populations, we excluded data from those HD hybrid populations that were established with individuals from inbred populations that could not be continued at generation 15 (see Table S1). The number of HD hybrid populations in these comparisons was 13. In contrast, when comparing the two hybrid population treatments (LD and HD hybrid populations), all 19 HD hybrid populations were included. The 95% confidence intervals for estimated marginal means of offspring production in different treatment combinations were obtained by multiplying the standard errors of the estimates by 1.96.

Results

Extinction

The proportion of populations that faced extinction during the seven generations of observation was not different between hybrid populations established from two and four source populations (0 of 16 populations, i.e. 0.0% vs. 1 of 16 populations, i.e. 6.3% among the LD hybrid populations, respectively, = 1.00; 3 of 10 populations, i.e. 30.0% vs. 2 of 9 populations, i.e. 22.2% among the HD hybrid populations, respectively, = 1.00). Therefore, the treatments of two and four source populations were pooled for further analyses.

Estimated with the pooled data, the proportion of extinct populations was significantly lower among the LD hybrid populations than among the concurrent inbred populations (Table 2). In contrast, the proportion of extinct populations was not significantly different between the HD hybrid populations and the concurrent inbred populations (Table 2). Comparing the LD and HD hybrid populations, the proportion of extinct populations was significantly lower among the LD than the HD hybrid populations (Table 2). The proportion of extinct populations did not differ between inbred populations measured at generations 7–13 (i.e. concurrent to LD hybrid populations) and at generations 15–21 (i.e. concurrent to HD hybrid populations) (Table 2, P = 0.434).

Table 2. Numbers of populations that went extinct and survived in different treatments during the seven-generation observation, together with tests of statistical significance. For both LD (low divergence) and HD (high divergence) hybrid populations, the populations established from two and four source populations are pooled together. Generation of extinction denotes generations in the inbred populations. Note that B. excludes data from hybrid populations that were established with individuals from inbred populations that could not be continued at generation 15
 ExtinctSurvivedPer cent extinct (%)Fisher's exact testGeneration of extinction
A.LD hybrid populations1313.1P < 0.00113
Concurrent inbred pop.10662.58,9,9,9,9,10,12,12,12,13
B.HD hybrid populations21115.4P = 0.17819,20
Concurrent inbred pop.4544.418,18,20,20
C.LD hybrid populations1313.1P = 0.02213
HD hybrid populations51426.318,18,19,20,21

Offspring production

The hybrid populations established from four source populations tended to produce less offspring than the hybrid populations established from only two source populations (Fig. 2), but the effect was not statistically significant (F1,49.2 = 2.91, = 0.094). Also, the interactions of the number of source populations with the level of divergence and with generation were nonsignificant (F1,54.8 = 0.02, = 0.890 and F6,79.1 = 1.77, = 0.117, respectively). Therefore, the treatments of two and four source populations were pooled for further analyses.

Figure 2.

Mean per capita offspring production relative to the control population for the low divergence (LD) and high divergence (HD) hybrid populations, established from either two (2SP) or four source populations (4SP) (error bars indicate 95% confidence intervals). Note that data from all established HD hybrid populations are included (n = 19).

We analysed the pooled data with a mixed model assuming heterogenous autoregressive covariance structure and nonequal covariance matrices in the inbred and hybrid populations. The analysis shows that, overall, hybridization significantly increased offspring production, but the effects of hybridization differed between the two levels of population divergence (the interaction between hybridization and level of population divergence was significant; see Table 3 and Fig. 3). Generation also had a significant effect on offspring production, and the effect of generation differed between the inbred and hybrid populations (Table 3). More specifically, there was a highly significant negative linear relationship between offspring production and generation in the inbred populations (F1,108 = 14.78, < 0.001), but no linear trend in the hybrid populations (F1,125 = 0.01, = 0.939).

Figure 3.

Mean per capita offspring production relative to the control population for the low divergence (LD) and high divergence (HD) hybrid populations, and the inbred populations (INBRED) (error bars indicate 95% confidence intervals). For both LD and HD hybrid populations, the populations established from two and four source populations are pooled together. For HD hybrid populations, data are included only for the populations established from inbred populations that were continued at generation 15 (n = 13).

Table 3. The effects of hybridization, the level of population divergence, and generation (from the establishment of the hybrid populations) on offspring production. For both LD (low divergence) and HD (high divergence) hybrid populations, the populations established from two and four source populations are pooled together. For HD hybrid populations, data are included only for the populations established from inbred populations that were continued at generation 15 (n = 13)
SourceNum. DFDen. DF F Sig.
Hybridization (Inbred vs. Hybrid)141.820.73< 0.001
Divergence (Low vs. High)158.62.130.150
Generation61548.53< 0.001
Hybridization × Divergence1547.170.010
Hybridization × Generation61503.840.001
Divergence × Generation61650.800.575

To analyse the interaction between hybridization and the level of population divergence in more detail, we performed the tests of simple effects (Winer, 1971) for these variables, pooling data over generations. The analysis reveals that hybridization significantly increased offspring production at the low level of population divergence, but this effect was not significant at the high level of population divergence (Table 4, Fig. 3). The analysis also shows that offspring production was significantly higher in the LD hybrid populations than in the HD hybrid populations (Table 4, Fig. 3). Offspring production did not differ between inbred populations measured at generations 7–13 and at generations 15–21 (Table 4, Fig. 3).

Table 4. The effect of hybridization, separately for the low and high levels of population divergence, and the effect of the level of population divergence, separately for the inbred and hybrid populations, on offspring production. The seven generations of observation are pooled together. For both LD (low divergence) and HD (high divergence) hybrid populations, the populations established from two and four source populations are pooled together. For HD hybrid populations, data are included only for the populations established from inbred populations that were continued at generation 15 (n = 13)
SourceHybridizationDivergenceNum. DFDen. DF F Sig.
Hybridization (Inbred vs. Hybrid)Low143.236.92< 0.001
Hybridization (Inbred vs. Hybrid)High153.12.240.140
Divergence (Low vs. High)Inbred134.80.390.534
Divergence (Low vs. High)Hybrid173.216.54< 0.001

To compare the LD and HD hybrid populations in more detail, we performed further analysis with a data set that included all established hybrid populations, that is, also the HD hybrid populations that were established with individuals from inbred populations that could not be continued at generation 15 (Table 5, Fig. 4). This analysis also revealed that the LD hybrid populations produced significantly more offspring than the HD hybrid populations. Notably, judged from the confidence intervals, the LD hybrid populations (mean, 1.18; 95% CI, 1.073–1.293) produced significantly more offspring relative to the large control population, and the HD hybrid populations (mean, 0.73; 95% CI, 0.60–0.87) produced significantly less offspring relative to the control population. Generation had a highly significant effect on offspring production, and the effect of generation was similar for both LD and HD hybrid populations (the interaction between divergence and generation was nonsignificant, Table 5). Therefore, the effect of generation on offspring production was studied with data pooled from the LD and HD hybrid populations. The effect of generation was best explained by a cubic (third-degree) trend (F1,92.5 = 19.55, < 0.0001; Fig. 4). Linear and quadratic trends were nonsignificant (= 0.668 and = 0.124, respectively).

Figure 4.

Mean per capita offspring production relative to the control population for the low divergence (LD) and high divergence (HD) hybrid populations (error bars indicate 95% confidence intervals). For both LD and HD hybrid populations, the populations established from two and four inbred source populations are pooled together. Note that data from all established HD hybrid populations are included (n = 19).

Table 5. The effects of the level of population divergence and generation (from the establishment of the hybrid populations) on offspring production of the hybrid populations. For both LD (low divergence) and HD (high divergence) hybrid populations, the populations established from two and four source populations are pooled together. The data are included from all established HD hybrid populations (n = 19)
SourceNum. DFDen. DF F Sig.
Divergence (Low vs. High)150.426.19< 0.001
Generation683.56.71< 0.001
Divergence × Generation683.50.650.691

Discussion

We used experimental D. littoralis populations to study the effects of population divergence and the number of source populations on the long-term consequences of interpopulation hybridization. We found that hybridization between small, inbred populations increased population viability and that the effect was persistent for at least seven generations following the hybridization. However, the increase in viability was significant only when hybridization was done at the low level of divergence (F = 0.30) between the inbred populations. Furthermore, when established at the low level of population divergence, offspring production of the hybrid populations was higher than in the large control population. In contrast, when established at the high level of divergence (F = 0.63), offspring production of the hybrid populations was lower than in the large control population. The number of source populations used to establish the hybrid populations had no significant effect on population viability. We will now discuss these results in more detail and conclude with the implications the results have on conservation of endangered populations.

Effect of hybridization at the low level of population divergence (F = 0.30)

The significantly improved viability of the hybrid populations established at the low level of population divergence compared to the inbred source populations can plausibly be explained by differential expression of deleterious recessive alleles in the populations (Lynch, 1991; Lynch & Walsh, 1998; Whitlock et al., 2000). In the inbred populations recessive, deleterious alleles are expressed in homozygous state, but in the hybrid populations the effects of recessive deleterious alleles are masked in heterozygous genotypes. However, we cannot exclude the possibility that some part of the superiority of the hybrid populations was due to loci with overdominant fitness effects (Lynch & Walsh, 1998; Lippman & Zamir, 2007), due to break-up of negative interactions among loci that may have become fixed in the inbred populations (Erickson & Fenster, 2006; Edmands et al., 2009) or due to formation of new beneficial interactions among loci in the hybrid populations (Lynch, 1991). Fitness increase following from interpopulation hybridization in populations suffering from inbreeding and drift is a common phenomenon and has been observed before in plants (e.g. Moll et al., 1965; Fenster & Galloway, 2000; Richards, 2000; Rhode & Cruzan, 2005; Willi et al., 2007), invertebrates (e.g. Edmands, 1999; Escobar et al., 2008; Coutellec & Caquet, 2011; Holleley et al., 2011) and vertebrates (e.g. Madsen et al., 1999; Marr et al., 2002; Hogg et al., 2006; Hedrick & Fredrickson, 2010).

Remarkably, offspring production of the hybrid populations established at the low level of population divergence also significantly exceeded offspring production in the large control population. This suggests that selection had removed (purged) at least some deleterious alleles from the inbred populations. Had there been no selection, average genotypes in the hybrid populations should be similar to genotypes in the control population, and the mean fitness of the hybrid populations should thus equal fitness in the control population (Falconer & Mackay, 1996; Crnokrak & Barrett, 2002). Notably, we also detected evidence for purging of genetic load in the inbred populations when we tracked the fitness of the inbred populations relative to the large control population generation by generation (Pekkala et al., 2012). Small population size can enhance selection against recessive highly deleterious alleles because in small populations, these alleles are expressed in homozygous state (Hedrick, 1994, 2002; Wang et al., 1999; Kirkpatrick & Jarne, 2000; Glemin, 2003). However, empirical studies of purging have provided inconsistent results (Ballou, 1997; Byers & Waller, 1999; Crnokrak & Barrett, 2002; Leberg & Firmin, 2008). Furthermore, purging by small population size, in comparison with purging by nonrandom mating, has been considered relatively inefficient on theoretical grounds (Glemin, 2003). The findings of the current study demonstrate that some purging of genetic load can occur also in populations of very small size.

Effect of hybridization at the high level of population divergence (F = 0.63)

In contrast to the hybrid populations established at the low level of population divergence, viability of the hybrid populations established at the high level of population divergence was not significantly better than viability of the concurrent inbred populations. A likely explanation for the reduced improvement in viability following hybridization at the high level of population divergence is that recessive, highly deleterious alleles had been to a greater extent purged from the inbred populations by the time the hybrid populations were established. Naturally, hybridization cannot mask the effects of recessive deleterious alleles if these alleles do not exist. Other possible explanations for the reduced benefits of hybridization at high level of population divergence are the break-up of coadapted gene complexes that may have developed in the inbred populations (Templeton, 1986; Lynch, 1991; Fenster et al., 1997; Lynch & Walsh, 1998) and the formation of new deleterious epistatic allele combinations in the hybrid populations (Phillips & Johnson, 1998; Orr & Turelli, 2001; Edmands, 2007; Presgraves, 2010). However, it is difficult to assess the role of these kinds of epistatic interactions in affecting the viability of the hybrid populations.

Notably, offspring production of the hybrid populations established at the high level of population divergence was significantly lower than in the large control population, even before any of the hybrid populations had gone extinct (Fig. 4, Table 2). A plausible explanation for the low offspring production in the hybrid populations relative to the control population is that the inbred populations had accumulated mildly deleterious mutations that are only partially recessive. Although in populations of small effective size selection against recessive highly deleterious alleles can be enhanced, selection against mildly deleterious alleles is relaxed, allowing them to accumulate in the populations (Lande, 1994; Lynch et al., 1995a; Wang et al., 1999; Whitlock, 2000; Glemin, 2003). As mildly deleterious alleles typically are only weakly recessive, their effects are expected to be masked only to a slight degree in the heterozygous hybrid genotypes (Whitlock et al., 2000). Results from our previous study support the conclusion that mildly deleterious alleles were accumulating in the inbred populations (Pekkala et al., 2012). However, it is also possible that negative epistatic interactions in the hybrid populations discussed in the previous paragraph contribute to the reduced fitness of the hybrid populations established at the high level of population divergence.

Effect of generation

We found the increase in viability of the hybrid populations established at the low level of population divergence to last for at least seven generations. Although in many previous studies heterosis observed in the first generation hybrid offspring has turned into outbreeding depression in the following generations (see Edmands, 2007), some of the previous studies have found positive effects of interpopulation hybridization to persist longer. For example, some studies carried out in plants (Moll et al., 1965; Willi et al., 2007) and Drosophila (Spielman & Frankham, 1992) have found the positive effects of hybridization to last at least for two or three generations. Very few studies have been continued beyond the second or third generation after hybridization. However, in a recent study using experimental D. melanogaster populations, Bijlsma et al. (2010) found that 10% immigration into highly inbred (but subsequently expanded) populations increased fitness of the populations for at least ten generations. Furthermore, some studies that have found outbreeding depression in early hybrid generations have found fitness to recover at later generations (Erickson & Fenster, 2006; Hwang et al., 2011). The long-lasting positive effects of hybridization might be partly due to selection against deleterious recessive alleles and allele combinations and positive selection favouring beneficial allele combinations (Templeton, 1986; Crnokrak & Barrett, 2002; Edmands et al., 2005; Erickson & Fenster, 2006; Hwang et al., 2011).

The temporal pattern of offspring production over the seven generations of observation was very similar in hybrid populations established both at the low and at the high levels of population divergence (see Fig. 4), and the pattern was highly significantly explained by a cubic (S-shaped) trend. This temporal pattern is intriguing, and although we cannot give a conclusive explanation for it, some speculation of the underlying mechanisms is warranted. Increasing fitness in the hybrid populations may be due to selection removing deleterious recessive alleles and possible deleterious multilocus allele combinations, as well as due to positive selection favouring beneficial allele combinations (Templeton, 1986; Crnokrak & Barrett, 2002; Edmands et al., 2005; Erickson & Fenster, 2006; Hwang et al., 2011). The negative fitness trend between generations 4 and 6 after hybridization could be due to increasing homozygosity as the hybrid populations approach Hardy–Weinberg equilibrium after maximal heterozygosity in the first generation hybrids and as genetic variability is lost from the hybrid populations by genetic drift (Falconer & Mackay, 1996; Lynch & Walsh, 1998). Decreasing fitness could also be expected from accumulation of mildly deleterious alleles because of relaxed natural selection (Lande, 1994; Lynch et al., 1995a; Wang et al., 1999; Whitlock, 2000) and from negative epistatic effects of hybridization (Edmands, 2007). Disentangling the importance of each mechanism is beyond the scope of the current study. Nevertheless, the results demonstrate that temporal patterns of viability in hybrid populations can be complex and that observing the fitness effects of hybridization only for one or a few generations may not predict viability over longer timescales.

In the inbred populations, a highly significant negative trend was observed between offspring production and generation. The negative trend can be explained by increasing homozygosity of recessive (highly) deleterious alleles and by the accumulation of mildly deleterious mutations. Apparently, purging of genetic load was not effective enough to overcome these effects.

Effect of number of source populations

The hybrid populations established from four inbred source populations tended to produce less offspring than the hybrid populations established from two source populations. Although the effect was not statistically significant, it was evident for hybrid populations established at both low and high levels of population divergence and might imply that with greater number of source populations there was more opportunity for negative effects of hybridization. To our knowledge, no previous studies have manipulated the number of source populations or the number of immigrants into small populations at a single time point to estimate the consequent fitness effects of hybridization. However, there are some studies in which the level of continuous immigration has been manipulated. For example, the level of immigration per generation did not affect fitness in small populations of a mustard species (Brassica campestris): 20% immigration per generation increased fitness as much as 50% immigration per generation (Newman & Tallmon, 2001). In contrast, in small populations of the house fly (Musca domestica), introduction of 50% immigration per generation increased fitness significantly more than introduction of 2.5% immigration per generation (Bryant et al., 1999). It is, however, difficult to compare the results of these studies to what was found in the current study, because of the differences in the continuity of the introduced genetic variation. Furthermore, since the effect of the number of source populations used was not statistically significant, the results in the current study must be interpreted cautiously.

Implications for conservation

Considering conservation practice, the design of the current study most closely reflects potential situations in reintroduction programmes of threatened species. When establishing a new population to natural habitat or to captivity, decisions need to be made on which, and on how many, source populations to use. Merging of different captive lineages has been successfully used, for example, in conservation of the endangered Mexican wolf (Canis lupus baileyi) (Hedrick & Fredrickson, 2010). Our results are of potential interest also to other similar applications, for example, when planning translocation of individuals between natural and/or captive populations of endangered species.

Scientists have been inconsistent with recommendations for using hybridization between isolated populations as a management tool, with some recommending a more cautious approach because of the risks of outbreeding depression (Edmands, 2007) whereas others call for more active approach (Frankham et al., 2011). Our results indicate that hybridization between isolated populations living in similar environments can yield substantial long-lasting fitness benefits for populations suffering from inbreeding and drift. However, we also found that the benefits of hybridization are reduced when the genetic divergence between populations is high. If highly inbred small populations indeed have purged recessive detrimental alleles and accumulated mildly detrimental partially recessive alleles, as might be expected based on the results of this study, interpopulation hybridization will be an efficient management tool only if immigrants are available that do not suffer from accumulation of mildly deleterious alleles. This would mean that in some cases the harmful effects of inbreeding and drift could be alleviated by interpopulation hybridization only if immigration was from a large population.

It must also be recognized that local adaptations, as well as the potential for adaptation to future environmental changes, have to be considered in applications of interpopulation hybridization (Templeton, 1986; Hedrick, 2005), although we did not address these questions in the current study. Furthermore, if time frames of several tens or hundreds of generations are considered, it is likely that hybridization can benefit the population only if a permanent increase in the effective population size is achieved (see e.g. Bijlsma et al., 2010; Hedrick & Fredrickson, 2010). If population size remains small, it can be expected that without constant gene flow, inbreeding and drift will continue to decrease the fitness of the population after the hybridization event.

Acknowledgments

We are grateful to all the persons involved in the maintenance of the fly populations and in acquiring the data. The Peerage of Science community and two anonymous reviewers are acknowledged for the valuable suggestions on the prior draft of the manuscript. The research was funded by the Academy of Finland (grant 7121616 to MP), the Centre of Excellence in Evolutionary Research, the Biological Interactions Graduate School and the Emil Aaltonen Foundation.

Ancillary