Sex‐specific effects of outbreeding on offspring quality in pike (Esox lucius)

Abstract Intraspecific genetic admixture occurs when previously separated populations within a species start interbreeding, and it can have either positive, negative, or neutral effects on reproductive performance. As there currently is no reliable predictor for the outcome of admixture, an increased knowledge about admixture effects in different species and populations is important to increase the understanding about what determines the response to admixture. We tested for effects of admixture on F1 offspring quality in three subpopulations of pike (Esox lucius). Gametes were collected in the field, and eggs from each female were experimentally fertilized with milt from a male from each population (one “pure” and two “admixed” treatments). Three offspring quality measures (hatching success, fry survival, and fry length) were determined and compared between (a) pure and admixed population combinations and (b) the sex‐specific treatments within each admixed population combination (based on the origin of the male and female, respectively). The results suggested that although there were no overall effects of admixture on offspring quality, the consequences for a given population combination could be sex‐specific and thus differ depending on which of the parents originated from one or the other population. All offspring quality traits were influenced by both maternal ID and paternal ID. Sex‐ and individual‐specific effects can have implications for dispersal behavior and gene flow between natural populations, and are important to consider in conservation efforts.

The consequences of genetic admixture, how response patterns might vary among populations within species and according to sex of immigrants within populations must be further investigated to increase the understanding of what determines the outcome of admixture, to search for any general response patterns, and to facilitate successful management (Rius & Darling, 2014).
The present study investigates the effects of intraspecific genetic admixture on different aspects of F1 offspring quality in anadromous subpopulations of pike (Esox lucius) (Figure 1). Pike is a long-lived, iteroparous, top-predatory fish, with a circumpolar distribution, and an important model species in studies of ecology and evolution . Moreover, pike is a highly valued species for recreational and commercial fishing (Lehtonen, Leskinen, Selen, & Reinikainen, 2009;Pierce, Tomcko, & Schupp, 1995) and subjected to large-scale stocking programs throughout its distribution (Craig, 1996;Larsen, Hansen, Nielsen, Jensen, & Loeschcke, 2005;Skov & Nilsson, 2007).
In this study, pike from three subpopulations (for sampling locations see Figure 2a) that display different local adaptations (e.g., larval traits , body size and growth rate , vertebral number , and salinity tolerance (Sunde, Tamario, Tibblin, Larsson, & Forsman, 2018)) were used in a common garden admixture experiment. Three offspring quality traits (hatching success, fry survival, and fry length) were measured and compared between (a) population combinations ("population combination comparisons") and (b) the two treatments within each population combination, based on the origin of the male and female, respectively ("sex-separated comparisons"). The population combination comparisons, where both treatments from a population combination were pooled, can be considered to primarily mimic the movement of a group of individuals of mixed sexes (e.g., management efforts such as supplementations and translocations of a group of individuals of mixed sexes). The sex-separated admixture comparisons on the other hand, primarily mimic the effects that would be expected following immigration of a single individual (e.g., natural dispersal), or of a group of individuals of the same sex.
The main objectives were to investigate whether (a) there were any effects of genetic admixture on F1 offspring quality, (b) the magnitude and/or direction of admixture effects varied between F I G U R E 1 An adult pike (Esox lucius) in the water by the inlet to one of the study localities (Harfjärden). Photograph: Anders Forsman populations and population combinations, and (c) the consequences of interbreeding were sex-specific and varied depending on the source population of the male and female, respectively.
However, previous studies have also found evidence of local adaptations in larval traits , body size and growth rate , vertebral number , and salinity tolerance   This, in combination with the fact that pike has external fertilization and can be reared in common garden environments makes it amenable for experimental studies of interpopulation genetic admixture.
Adult pikes from each subpopulation were caught during 4 days (March 31, April 1, April 4, and April 5, 2016) using fyke nets that were placed in the inlet of the streams leading to the spawning grounds Tibblin, Berggren, et al., 2016), and ripe individuals (Harfjärden: 11 females, 12 males; Lervik: 10 females, 12 males; Okne: 11 females, 10 males) were stripped of gametes. Stripping was synchronized between the localities to avoid systematic error due to different storing times of gametes before artificial fertilization was conducted. To ensure high-quality eggs, only females with eggs that had no visual trace of blood were selected, and the first batch of eggs from each selected female was discarded to avoid premature fertilization activation of the eggs (opening of the micropyle) (Craig, 1996;Raat, 1988). The gametes were collected in separate sterile plastic tubes (eggs: 50 ml Falcon tubes, milt: 2 ml Eppendorf tubes) that were immediately put on ice in cooler boxes. The tubes were continuously kept on ice (up to 4 hr) during the transportation to the Kalmarsound Laboratory of Linnaeus University in Kalmar, Sweden, until fertilization.

| Admixture experiment
The collected gametes were used in a common garden admixture experiment. To that end, each female was paired with one male from each subpopulation, and each male was also paired with one female from each subpopulation. In general, each individual contributed gametes to only one family (female/male pair) in each population combination. However, occasional mismatches occurred due to unequal sex ratios of the captured individuals on different days. The eggs from each female were divided into six separate batches (25 eggs in each) which each was experimentally fertilized with milt from a male from one of the subpopulations (making two replicates each of one "pure" and two different "admixed" treatments per female).
All families were thus duplicated, resulting in a total of approximately 4,800 eggs divided between 192 experimental units. The artificial fertilization and the entire experiment was carried out in a constant room with a temperature of 12°C, using tap water that had been aerated for a minimum of four hours. The design was chosen such that the temperature was within the natural range (Frost & Kipling, 1967;Nilsson, Engstedt, & Larsson, 2014), and because other studies have successfully used similar approaches Tibblin, Koch-Schmidt, Larsson, & Stenroth, 2012;Tibblin et al., 2015).
The artificial fertilizations were conducted following the protocol used by Berggren et al. (2016) and Sunde et al. (2018). In short, eggs from one female were placed in a small glass bowl (50 mm diameter), and an excess of milt (approximately 200 μl) from one male was poured onto the eggs. Water (approximately an equal volume to the eggs) was subsequently added to the gametes to initiate fertilization, and the bowl was gently swirled in circular motions to mix the gametes. The gametes were let to rest for two minutes for fertilization to occur, and excess milt was then removed by rinsing the eggs with water three times. The fertilized eggs were then gently placed into separate plastic bins. Eggs and hatched fry were continuously kept in their natal bins throughout the experiment. Each bin consisted of two 800 ml plastic cups, placed in each other, with the bottom of the inner cup replaced with a plastic net (mesh size approximately 1.5 × 1.5 mm). This design of the bins enabled complete daily water exchanges throughout the experiment (including the time of egg incubation, when eggs are sensitive to movements) by addition of water to the cups.
The 192 experimental units were placed randomly with regards to treatment and source population in the constant room, to avoid systematic effects associated with differential air flow, temperature or light. To ensure that there was food available when fry reached the end of the yolk-sac phase, an excess of live brine shrimp (Artemia salina) were provided three times per day starting 5 days post hatch.
At 8 days post hatch, fry were also provided with an excess of

| Data collection
Data on different aspects of offspring quality were obtained from the photograph series. Offspring quality was estimated as (a) hatching success of eggs, (b) fry survival during 15 days following hatching, and (c) fry body length at 15 days post hatch (Figure 2b-d). We know from a previous study  that there is a time period during which unfertilized eggs are not visibly discernible as dead, such that unsuccessful fertilization cannot be reliably distinguished from postfertilization embryonic mortality. We therefore estimate hatching success as a composite measure from eggs to successful hatchlings.

| Analytical approach
One of the advantages with using pike as study species is that it has external fertilization, which enables utilization of eggs and milt from the same individuals in all treatments. This design allows controlling for variation in the response variables owing to any parental effects, and thus allows for improved power of the statistical evaluation of experimental treatments and fixed effects. Including all possible pairwise population combinations in the experimental design also enables comparisons between pooled population combinations (regardless of which sex in a specific combination of populations originates from which of the populations; henceforth "population combination comparison") as well as comparisons between all pairwise population treatments separated by origin of the two sexes, respectively (henceforth "sex-separated admixture comparison"). To evaluate the different ways by which admixture, source populations, and parental identity might affect offspring quality, data from the experiment were analyzed in two different ways. This was necessary to address different hypotheses and to analyze effects at different levels of resolution.

| Population combination comparisons
First, we addressed the overall questions whether and how admixture influenced offspring quality in different combinations of populations, and whether the effect of admixture differed among the populations. To that end, we used separate analyses on the three offspring quality traits and compared the performance of the six population combinations (the three populations pure, "HH," "OO," and "LL," and all three hybrid population combinations, "OH," "LO," and "HL"). Population combination was treated as a fixed factor, and family (female/male pair) and replicate (nested in family) were with a binomial fit and a logit-link function using the same R package.

| Sex-specific comparisons
Next, we set out to answer the question whether offspring quality measures within population combinations varied depending on the source of the male and female, respectively. For that, we used simi- replicate (nested within family) were treated as random factors.
Differences between sex-separated pairwise admixed population combinations were based on comparisons of least squares means.

| Tests of statistical significance
Null hypothesis significance testing and statistical inference based on p-values is controversial and can be complicated, particularly when analyses include complex models, multiple statistical tests, and when p is near the typical threshold 0.05 (Wasserstein & Lazar, 2016). A p-value by itself does not measure the size of a difference or the importance of an effect, nor does it provide a good measure of evidence in favor or against the null hypothesis, and results must, therefore, be interpreted with caution (Wasserstein & Lazar, 2016). Yet, statistical significance is still considered important by some and may be used for meta-analyses.
We therefore adhere to the tradition yet treat p-values carefully.
For both the population combination comparisons and the sexspecific comparisons, statistical significance of fixed factors and of interactions was assessed using the Type III test of fixed effect.
Models were fitted to the data using the restricted maximum likelihood method (Laplace approximation). The effects and statistical significance of parental identity were assessed using covariance parameter estimates. We do not report results associated with the replicate factor.

| Ethics note
All applicable national guidelines for the care and use of animals were followed. The study conforms to Directive 2010/63/EU. Ethical approval for the study was granted by the Ethical Committee on Animal F I G U R E 3 Effects of admixture and population identity on viability and performance of eggs and offspring in Esox lucius. Data for three populations (H = Harfjärden, L = Lervik, O = Oknebäck) that were purebred and mixed. Figure shows means ± SE based on raw data. Results from statistical analysis of variation among groups are presented in Table 1 Research in Linköping, Sweden (approval Dnr ID 83). Permission for field studies was granted by the County Administrative Board in Kalmar (approval 623-1681-13).

| RE SULTS
The analyses revealed the following overall results:

| Hatching success
Average hatching success was approximately 55% (range 43%-74%). This is comparable with estimates obtained in a previous experiment conducted in the wild . The population combination comparison suggested that hatching success differed between population combinations (p = 0.04, Figure 3, Table 1). However, no differences were present between the pure populations and their two respective hybrid groups. Further, the results from the sex-separated comparison showed that hatching success was influenced by maternal populations (p < 0.0001), but was not influenced by paternal population (p = 0.48) or the interaction between maternal and paternal population (p = 0.92; Table 2).
Additionally, the least squares means comparisons showed that there were sex-specific effects of admixture on hatching success that depended on the source populations of the male and female parent within two out of the three population combinations (HL vs. LH, p = <0.0001, and HO vs. OH, p = 0.0004, Figure 4, Table 3). The probability associated with the third population combination (OL vs. LO) was also relatively low (p = 0.095), potentially reinforcing the idea that there is a sex-specific effect.

| Fry survival
Survival of fry was high (>85%) in all groups. The population combination comparison showed that fry survival did not differ between the population combinations (p = 0.90, Figure 3, Table 1). The sexspecific comparisons further showed that neither paternal population (p = 0.84), maternal population (p = 0.36), nor the interaction between maternal and paternal population (p = 0.59) influenced the response (Table 2). Additionally, the least squares means comparisons indicated no strong sex-specific responses, that is the response did not seem to differ within population combinations depending on the source population of the male and female parent (p > 0.05, Figure 4, Table 3).

| Fry length
The population combination comparison showed that fry body length did not differ between population combinations (p = 0.52, Figure 3, Table 1). The sex-separated comparisons suggested that fry length varied between maternal source populations (p = 0.04), but was not influenced by paternal population (p = 0.33) or the interaction between maternal and paternal population (p = 0.41; Table 2). The least squares means comparisons suggested that there was a sex-specific effect of admixture on fry length within the Harfjärden-Okne population combination (HO vs. OH, p = 0.02, Figure 4, Table 3).

| Effects of maternal and paternal identity
There were moderate to strong effects of maternal identity and paternal identity on hatching success and fry survival (as  (Ppop), and the interaction between maternal population and paternal population on hatching success, fry survival, and fry length are presented. Covariance parameter estimates associated with random effects of maternal and paternal identity are presented in Table 4. evidenced by the covariance parameter estimates reported in Table 4).  to investigate the effects of genetic admixture on different aspects of offspring quality.

| D ISCUSS I ON
The analytical approach in the present study included two different analyses, to do comparisons at different levels of resolution ("population combination comparison" and "sex-separated comparison"). Results, therefore, inform about the effects of genetic admixture that might result from different types of gene flow.

| Effects of admixture
The population combination comparison suggested that the effect of genetic admixture on survival and fry body length did not differ between any of the population combinations. Further, hatching success differed between population combinations, but no differences were found between any of the purebred populations and its two respective hybrid population combinations. This suggests that genetic admixture does not have any effect on offspring quality in the pike populations included in this study. That admixture does not substantially reduce or increase aspects of F1 offspring performance is interesting and relevant to population management. However, results from admixture studies must be interpreted with caution, especially if they will be used as guidelines for management. Several studies have found that different populations within species can respond in different ways to genetic admixture (Escobar et al., 2008;Hufford et al., 2012;Sunde & Forsman, 2016;Tinnert et al., 2016;Tortajada et al., 2010). Thus, even though the results showed that admixture did not affect offspring quality of pike in the included populations and population combinations, this outcome is not necessarily transferable to other pike populations. In addition, previous studies have found that the response to genetic admixture can differ between generations (Huff, Miller, Chizinski, & Vondracek, 2011;Tinnert et al., 2016), and between different environments (see McClelland & Naish, 2007). Thus, that no overall effects of admixture were found in this study, does not ensure that there will not be any effects under natural conditions (with different selective regimes) or in later generations (e.g., due to break-up of coadapted gene complexes). Due to the complex nature of the interacting mechanisms determining the response to genetic admixture, it is important that the role of admixture is further studied in different species and populations to search for general patterns that might be used to predict the direction of the response (Rius & Darling, 2014).
We observed a trend that fry in admixed population combinations were smaller than fry in purebred treatments (Figure 3). Body size as a fitness estimate might be an especially important proxy in pike where cannibalism in fry is common (Craig, 1996). A large body could enable fry to prey on smaller conspecifics, and protect them from being victims of cannibalism. Previous investigations based on Q ST /F ST comparisons and common garden breeding experiments have also indicated that different subpopulations of pike have evolved genetically based adaptive differences in growth rates and body size . This indicates that there might be genetic incompatibilities between the locally adapted subpopulations that resulted in the somewhat smaller size of admixed fry in the present study.
Perhaps the most novel conclusion emerging from this study was that the consequences of genetic admixture for a given population combination can vary depending on the origin of the male and female, respectively. Such sex-specific differences within population combinations were found for both hatching success and fry body length (Figure 4). A possible explanation for this is that antagonistic selection can result in differences in genetic architecture of life history traits between the sexes (Forsman, 1995;Lande, 1980;Rice, 1984;Roff, 1992;Trivers, 1972;Zwoinska et al., 2016), which might contribute to the sex-specific responses to genetic admixture.
Admittedly, our experimental design does not enable us to discard the possibility that the observed differences and effects were mediated to some extent by environmentally induced plasticity and parental effects. Regardless the underlying mechanisms, sex-specific effects have the potential to affect the ecological and evolutionary dynamics of the populations.
In addition, we found that both maternal and paternal identity influenced all offspring quality measures, indicative of genetic or epigenetic parental effects, or a combination of the two. This conforms with the findings of Siddique et al. (2017) who report that parental identity had a strong influence on embryonic development in pike.

| Implications for management and natural dispersal
The pike populations included in this study, despite being located in close proximity to each other in an open system, are genetically separated . This fine-scaled genetic structuring is caused by the homing behavior observed in pike. The results suggested that admixture did not materially impair offspring quality TA B L E 4 Effects of maternal identity (MID) and paternal identity (PID) on different aspects of offspring quality in experimental crossings of three natural populations of Esox lucius Note. Random effects are indicated by covariance parameter estimates obtained using generalized linear mixed models (glmer) and general linear mixed models (lmer), respectively. The fitted statistical models included fixed explanatory variables (maternal source population, paternal source population, and the interaction between maternal and paternal source population) reported in Table 2. in any of the population combinations, and conversely, none of the populations performed substantially better when admixed. In addition, previous studies have found that the pike populations harbor different local adaptations (e.g., larval traits ; body size and growth rate ; vertebral number ; and salinity tolerance ). Because of differences in local adaptations, the benefits of such population-specific adaptations are likely higher in the natural environments than in the laboratory (Forsman, 2014), and dispersal from the environments that individuals are adapted to is likely associated with additional disadvantages to the immigrants in the wild.  Willi & Van Buskirk, 2005), and the suitability of these metrics as proxies to predict the outcome of admixture thus likely differ among species (Edmands, 2002). Our study populations differ in regards to geographic separation and neutral genetic differentiation. However, we did not find any overall effects of genetic admixture on offspring quality in any of the populations combinations. Thus, based on the results from this study, it is not possible to determine if there is any correlation between the degree of geographic distance or genetic differentiation and the response to admixture in pike. Whether and how geographic distance and genetic differentiation influence the direction of the outcome of admixture in pike therefore needs to be further investigated.
It should be emphasized that estimates of neutral genetic diversity and differentiation do not necessarily reflect adaptive diversity (Leinonen et al., 2008;Reed & Frankham, 2001;Tinnert, 2017;Willi, Van Buskirk, & Hoffmann, 2006), and that the outcome of admixture is determined by functional genetic archi-

| CON CLUS IONS
The results of the population combination comparisons indicate that genetic admixture does not have any overall effects on F1 offspring quality in pike. However, these result should be interpreted with caution because previous studies have shown that the direction of the response can differ between different populations (Escobar et al., 2008;Hufford et al., 2012;Sunde & Forsman, 2016;Tortajada et al., 2010), between generations (Huff et al., 2011;Tinnert et al., 2016), and between different environments (see McClelland & Naish, 2007). The lack of an overall effect in the present study is thus not universally transferable to all pike populations.
An interesting finding in our study was that the responses to genetic admixture for some population combinations were sexspecific, and thus could differ depending on the origin of the male and female respectively. Future studies are necessary to assess the generality of such sex-specific effects of admixture, and attempts to explore whether sex-specific responses to admixture are generally associated with sex-biased dispersal may expand our knowledge and understanding about dispersal behavior and evolution of biodiversity, and aid development of management actions aimed at preservation and revitalization of endangered and declining populations and species.