Maladaptive secondary sexual characteristics reduce the reproductive success of hybrids between native and non‐native salmonids

Abstract Human‐mediated hybridization between introduced and native species is one of the most serious threats to native taxa. Although field studies have attempted to quantify the relative fitness or reproductive success of parental species and their hybrids, only a few studies have unraveled the factors determining the fitness of hybrids. Here, we hypothesized that maladaptive secondary sexual characteristics may reduce fitness of hybrids between two fish species. To test this, we evaluated the reproductive success of introduced brook trout (BT: Salvelinus fontinalis), native white‐spotted charr (WSC: S. leucomaenis) and their hybrids in a natural stream in Hokkaido, Japan, where the two parental species show remarkably different male secondary sexual characteristics, such as elongated jaws and deeper bodies. We predicted that introgression from WSC is maladaptive for BT males because the BT male has more prominent secondary sexual characteristics. Our results suggest that both sexual selection and outbreeding depression in males and females significantly influence an individual's reproductive success. Our results also suggest that asymmetric introgression may increase the risks to persistence in the recipient species.

Modliszewski, Wright, Wu, & Willis, 2008), ecological factors affecting hybrid fitness have been much less well investigated in humanmediated hybridization (but see, While, Michaelides, & Heathcote, 2015). Since it is generally difficult to track hybrid individuals simultaneously with their parental species in the wild while measuring fitness component(s), several previous studies have illustrated the patterns or consequences of hybrid fitness in the wild (Muhlfeld et al., 2009;Taylor, Gerlinsky, Farrell, & Gow, 2012), but less frequently have the ecological mechanisms driving such patterns or consequences been studied. As an outstanding example, flycatcher birds from a natural hybrid zone have been well investigated for factors such as hybrid inviability (Alatalo, Gustafsson, & Lundberg, 1982), hybrid infertility (Alund, Immler, Rice, & Qvarnstrom, 2013), ecological inviability (Tegelstrom & Gelter, 1990;Veen et al., 2001), and sexual selection against hybrids (Svedin et al., 2008). Understanding the factors influencing hybrid fitness is critical for forecasting the impacts of human-mediated hybridization between native and nonnative species (Allendorf et al., 2001;Ellstrand, Prentice, & Hancock, 1999;Fitzpatrick et al., 2010).
Sexual selection can be a key factor influencing hybrid fitness (Lemmon & Lemmon, 2010;Svedin et al., 2008;While et al., 2015), but has been studied much less than other postzygotic factors, such as genetic incompatibilities (Brown & Eady, 2001;Howard, Gregory, Chu, & Cain, 1998;Noor, 1997), or hybrids performing poorly in parental niches (Grant & Grant, 1997;Hatfield & Schluter, 1999;Naisbit, Jigginns, & Mallet, 2001). A study by Svedin et al. (2008) estimated the fitness components of collared (Ficedula albicollis) and pied (Ficedula hypoleuca) flycatchers and their hybrids and found that sexual selection against the intermediate phenotype (i.e., intermediate plumage characters) of hybrid males accounted for approximately one-third of the total postzygotic isolation between the two species, and approximately 75% of this selection was attributable to sexual selection. Despite the potential importance of sexual selection, its role in postzygotic isolation has only been investigated in a few other cases, such as acoustic signals of the chorus frog (Pseudacris spp, Lemmon & Lemmon, 2010) and the body colors and head length of the wall lizard (Podarcis muralis, While et al., 2015).
In this study, we explored the effects of sexual selection against hybrids between native and non-native salmonid species (i.e., charr: Salvelinus genus). We focused on secondary sexual characteristics of hybrid males because male-male competition often results in remarkable sexual dimorphisms in salmonid fishes (Fleming, 1996;Fleming & Gross, 1994;Quinn & Foote, 1994).
During the breeding season, males develop elongated jaws, develop an angular tip or kype on the lower jaw, and large canine-like teeth emerge from the hooked snout which appear to function as specialized weapons in fighting competitors. In addition, the enlargement of the dorsal hump appears to act as a shield against attack from competitors and block access to ovipositing females by competing males. Males also show brilliant nuptial coloration, which may attract females. Such developed secondary sexual characteristics can contribute to reproductive success through male-male competition and female choice (Fleming & Gross, 1994;Quinn & Foote, 1994). Hybrids between introduced brook trout (BT), Salvelinus fontinalis, with native white-spotted charr (WSC), S. leucomaenis, occur in a boreal stream of Japan (Fukui, May-McNally, Katahira, Kitano, & Koizumi, 2016;Kitano, Ohdachi, Koizumi, & Hasegawa, 2014). Here, the male secondary sexual characteristics between these species express obviously distinct features; BT has more intense breeding coloration, deeper bodies, and larger jaws (Figure 1, see Results). These observations allowed us to test whether the male reproductive characters can be one of the factors/mechanisms determining the hybrid fitness in the wild. For a similar and well-investigated example, introgression between non-native rainbow trout (Oncorhynchus mykiss) and native westslope cutthroat trout (O. clarkii lewisi) has spread across watersheds in North America (Boyer, Muhlfeld, & Allendorf, 2008;Hitt, Frissell, Muhlfeld, & Allendorf, 2003;Yau & Taylor, 2014).
F I G U R E 1 Representative males of introduced brook trout (above photo), hybrid (middle photo), and native white-spotted charr (below photo). Photographs have been taken in the present study area. Bottom drawing picture represents locations for 16 landmarks identified on each male charr photograph:snout length (1-2), under-jaw length (15-16), pectoral fin (13-14), dorsal fin height (3-5), hump height 1 (3-4), hump height 2 (6-12), caudal peduncle (8-10), caudal length 1 (7-9), caudal length 2 (9-11). Theses nine measurements are used in the principal component analysis. For females, only caudal peduncle (8-10) was used for subsequent statistical analysis with the zero-inflated model  Muhlfeld et al. (2009) found that the reproductive success in post F1 hybrids declined rapidly with an increasing genetic proportion of rainbow trout ancestry, although the mechanisms reducing the reproductive success of hybrids remained unknown. Here, we conducted an intensive field survey with parentage analysis in a tributary of the upper Sorachi River, Hokkaido, Japan, to (a) quantify the reproductive success of parental species and their hybrids in a wild population, (b) evaluate whether introgression causes outbreeding depression in the life stages from maturation to the juveniles of the next generation, and (c) test whether male reproductive characters affects hybrid fitness. For (b) and (c) we focused on BT and its hybrid because sample size was limited for pure WSC. We specifically predicted that transfer of the less well developed secondary sexual characteristics of native WSC would reduce the male reproductive success of introduced BT.

| Study area and data collection from the field
Our field study was conducted in the Higashi-Furebetsu stream, a tributary of the upper Nunobe stream, in central Hokkaido, Japan (Supporting Information Figure S1). Based on species distribution data from a previous survey (Fukui et al., 2016), we considered that the tributary was suitable to investigate the interactions between Photographs were taken for all captured age-1+ and older fishes to measure the external characteristics of all candidate parental fish using an acrylic water tank (40 cm × 15 cm × 10 cm) and digital camera (Panasonic; DMC-FT3-S). Nine morphological characters ( Figure 1) were measured from the photographic data using the software Image J (version 1.48v; US National Institutes of Health, Bethesda, Maryland) corrected by the fork length measured during the field sampling (i.e., calculated by relative lengths). Morphological differences between the three male groups (i.e., BT, HYB, and WSC) were examined by principal component analysis (PCA). All the nine characters measured were ln-transformed and size-adjusted using their FL prior to the PCA. Since the first principal component (PC1) included secondary sexual characters such as snout length or hump height, which are known as sexually selected variables (Fleming & Gross, 1994;Quinn & Foote, 1994), and showed a significant difference between WSC and BT (see Results), PC1 was used as a value of secondary sexual characteristics in the subsequent statistical analysis.

| Data analysis
Parental species and hybrids were genetically determined using the software STRUCTURE (Pritchard et al., 2000). When all samples (i.e., BT, WSC, and HYB) were analyzed, the most likely number of putative populations (K) was two, representing WSC and BT. Based on K = 2, we calculated a hybrid index as the proportion WSC ancestry: this ranges from 0.001 for pure BT (no WSC allele) to 0.999 for pure WSC (no BT allele). First generation hybrids between BT and WSC have a hybrid index of 0.5 with the 13 microsatellite loci and heterozygous for alleles from each of the parental taxa at all the seven diagnostic loci (Supporting Information Table S1). We assumed that fish with a hybrid index of 0.5 but not heterozygous at all the diagnostic loci were post-F 1 hybrids, although it was difficult to distinguish backcrossing hybrids from other post-F1 hybrids.
We performed parentage analysis using CERVUS v3.0.7 (Kalinowski, Taper, & Marshall, 2007) to identify parentage among candidate parents (the adults captured in October and December of 2013) and offspring (the 0+ fry sampled in May, June, August, October of 2014, and 1+ parr sampled in September 2015). The CERVUS analysis uses a maximum-likelihood approach to predict parent-offspring relationships. For each offspring, a metric of the likelihood that each potential father is the actual father is first calculated as a log-likelihood (LOD) score. Then, CERVUS uses the range of LOD score distributions to calculate a critical LOD score against which LOD scores from the actual population can be evaluated. The LOD scores are then computed for each potential father using their actual genotype for each individual offspring. The critical LOD score calculated from the simulation step is used to assign fathers to each offspring (Kalinowski et al., 2007). The same procedure was also applied to mother-offspring relationships. Allele frequencies using all genotyped individuals, the expected frequency of heterozygotes (H e ), and the observed frequency of heterozygotes (H o ) were also calculated using CERVUS.
We defined reproductive success as the number of offspring produced by each male or female. To investigate how body size, proportion WSC ancestry, and secondary sexual characteristics influence reproductive success, we used zero-inflated models with negative binomial distribution (ZIB) to model counts of offspring (Zuur, Ieno, Walker, Saveliev, & Smith, 2009), including zeros for males that may have mated, but their offspring was not sampled, because the reproductive success, is a count variable with many zeros (Supporting Information Figure S2). The statistical analysis with ZIB was only conducted for BT and HYB because the distributions of reproductive success versus proportion WSC ancestry did not fit to the ZIB for WSC, likely attributable to the low sample size of pure WSC. The number of offspring was used as a response variable, whereas FL (i.e., body size: both males and females), PC1 (i.e., secondary sexual characters: males only), and the height of the caudal peduncle (females only) were used as independent variables. Because body size in salmonid fishes is important not only for male-male interactions during mating but also for the establishment of territories in non-breeding season (e.g., Fleming, 1996), we used both FL and PC1 as the predictor variables to separate these effects. The height of the caudal peduncle in females is associated with the depth of eggs buried and, thus, has a significant effect on reproductive success (Fleming & Gross, 1994). Since a significant correlation was observed between FL and PC1 (Spearman's r = 0.646, p < 0.001), PC1 was expressed as the residuals from its liner regression against FL. The transformation of PC1 value individuals with remarkable secondary sexual characteristics, such as higher hump height than average for their size, had a large positive residual value. All statistical analyses were performed using the software R version 3.2.4 (R Development Core Team, 2011).

| RE SULTS
In total, 666 offspring and 268 potential parents (149 females Table S1). There were a number of loci that were not in Hardy-Weinberg equilibrium. This is likely because random mating did not occur among BT, WSC, and HYB.
We analyzed morphological variation in males and we confirmed differences of secondary sexual characters between BT and WSC males. A total of 109 of the 112 males were successfully measured the morphological characters. Analysis of variance (ANOVA) indicated that these groups differed significantly from each other along the first principal component (PC1, F 2,107 = 27.951, p < 0.0001).
There was no significant difference among male types in PC2. The BT and WSC males were clearly distinguished by the secondary sexual characteristics (Figures 1 and 2, Table 1). Tukey's pairwise comparisons indicated that BT and WSC males significantly differed from one another in PC1 (p < 0.0001), and that HYB and WSC males significantly differed from one another in PC1 (p < 0.0001), but that BT and HYB groups were not significant distinct from one another (p = 0.4322). The PC1 is based primarily on variation in secondary sexual characters such as hump height and lower jaw length ( Figure 1, Table 1) and explained 40.4% of the variation and PC2 explained an additional 19.6% of the variation.
Of the 637 offspring genotyped, 326 individuals (51%) were assigned to a sampled father, and 367 (58%) were assigned to a sampled mother by CERVUS. Of the 112 mature males, 59 (53%) appeared to produce at least one offspring (maximum 82 offspring) with the median of 1. Of the 124 mature females, 89 (72%) appeared to produce at least one offspring (maximum 84 offspring) with the median of 2.
The reproductive success was a count variable with many zeros as expected (Supporting Information Figure S2).
The ZIB showed that the proportion of WSC ancestry among mature males and females had a strong negative effect on reproductive success (Table 2). Hybrids had lower reproductive success than that of both parental species in males and females ( Figure 3a). Quadratic equations fitted to the median number of offspring per parent among males and females (males, γ 2 = 0.8673, y = 3.3482x 2 -3.7054x + 0.875; females, γ 2 = 0.9496, y = 12.5x 2 -11.071x + 2.2857) estimated that reproductive success decreased by 30.6% and 25.0% for males and females, respectively, in BT with 20 percent proportion WSC ancestry (Figure 3b).
For male BT and HYB, FL had only a small, but statistically significant positive effect on reproductive success (Table 2, Figure 4), but variation along PC1 (i.e., prominence of secondary sexual characters) had a 20x stronger positive effect on reproductive success (Table 2, Figure 4). For female BT and HYB, FL had a significant and positive effect on reproductive success, but caudal peduncle depth had no detectable effect on reproductive success (Table 2, Figure 4).

| D ISCUSS I ON
While past studies have shown the reduction in fitness or reproductive success in hybrids (e.g., Muhlfeld et al., 2009;Lancaster, Bradshaw, Goldsworthy, & Sunnucks, 2007), underling mechanisms have been rarely clarified (but see Svedin et al., 2008;Lemmon & Lemmon, 2010;While et al., 2015). We hypothesized that intermediate morphology is maladaptive especially the one related to sexual selection. Our results suggested that PC1 of morphological traits, which was associated with male secondary sexual characteristics, significantly affect the reproductive success of non-native salmonid and its hybrids with native species. Importantly, hybrid index was also selected as a significant factor affecting male reproductive success, suggesting that not only maladaptive morphology but also some other undetermined factor(s) reduce the fitness of hybrids.
Hybrid index was also negatively associated with female reproductive success. Although the sample size and number of examined loci were limited, our results were largely consistent with the predictions derived from the general patterns of salmonids' mating systems. In addition, more prominent secondary sexual characteristics of nonnative species might result in asymmetric introgression from non-native to native species, which is a serious concern from conservation perspective.

| Roles of morphology on reproductive success of salmonids
Because of remarkable morphological variations during breeding seasons, many studies have attempted to clarify the factors affecting reproductive success in salmonids (e.g., Fleming & Gross, 1994;Tentelier et al., 2016). Among them, body size appeared to be the most consistent determinant as the variation could range from <10 cm to>100 cm, even within populations (L'Abbe-Lund, 1989).
Our population was less variable but still had more than twofold difference in fork length (129-323 mm). In males, large individuals become dominant and outcompete small males for access to females (Blanchfield, Ridgway, & Wilson, 2003;Dickerson, Brinck, Willson, Bentzen, & Quinn, 2005;Fleming, 1996Fleming, , 1998Quinn & Foote, 1994). In females, fecundity increases exponentially with body size and, thus, large females have much greater advantages (van den Berghe and Gross, 1989). While we found a significant effect of body size in both sexes, the size advantage was not very strong especially for males (Table 2; Figure 4). It is likely that small males use alternative tactics, such as sneaking, in the presence of bigger males to increase their reproductive success (Thériault et al., 2007). Alternatively, we might have missed some potential par- The statistical analysis was conducted separately for males and females.
TA B L E 2 Zero-inflated count model with negative binomial distribution of male and female reproductive success F I G U R E 3 (a) Number of offspring per male (upper left) and female (upper right) versus the proportion native whitespotted charr ancestry (i.e., "0" side is pure BT and "1" is pure WSC). The left plot includes 112 fathers and 327 juvenile assignments, and the right plot includes 124 mothers and 367 juvenile assignments using parentage analysis. Each point represents an estimate for an individual fish from a spawn year. Circles represent brook trout, white-spotted charr, and their hybrids. such as elongation of the jaw, development of an angular tip or kype, or dorsal hump, which is referred even in the Descent of Man (Darwin, 1871). Despite the wide acceptance of salmonids' researchers, surprisingly few studies have actually demonstrated the roles of the sexual traits on reproductive success (Fleming & Gross, 1994). These exaggerated traits are usually observed in large sea-run species or individuals, but even relatively small stream-resident brook trout show such traits (Kazyak, Hilderbrand, & Holloway, 2013), providing a good opportunity to assess their roles. In fact, PC1 clearly distinguished between brook trout and white-spotted charr based on secondary sexual characteristics. Thus, the significant positive effect of PC1 represents one of the rare instances of evidence for the role of male secondary sexual characteristics on reproductive success in the field.
The caudal peduncle area is also reported as an important factor for females in digging during nest construction and appears to influence the depth of egg burial (Fleming & Gross, 1994). By contrast, we did not find a relationship between caudal peduncle depth and reproductive success. This may be because our study occurred in an area where the substrate had been modified by humans and

| Reproductive success of hybrids
Our results clearly showed that both male and female hybrids have lower reproductive success compared to the parent species ( Figure 3). This should be partly due to the maladaptive morphology as discussed above, but other undermined factors also reduce the reproductive success of hybrids because hybrid index had significant effects in both sexes. Unfortunately, the processes underlying the reduced hybrid reproductive success in our study are unknown, because our estimate of reproductive success represents a function of selection at different life stages, including spawning, eyed egg, egg-to-fry emergence and juvenile survival: selection could have acted at any or all of these stages. Previous laboratory studies have documented lower fitness of hybrids between BT and WSC at multiple stages (Suzuki & Fukuda, 1971, 1973, 1974. First, gonadal development of F1 hybrid males was reduced compared to the parental species (with lower quantity of milt) in both reciprocal crosses.
Second, survival and growth rates at the early life stages (i.e., hatching and eyed eggs) of both F2 and some backcrossed hybrids were lower than those of parental species, although some hybrid vigor was also observed in F1 hybrids. Future studies should determine the relative reduction in hybrid fitness at each of the stages.

| Conservation implications
We showed that reproductive success of male BT (and its hybrid) was reduced by the maladaptive phenotypes derived from WSC.
This, in turn, suggests that native WSC may suffer from asymmetric introgression from non-native BT, which have more remarkable secondary sexual characteristics. That is, hybrid WSC with more pronounced secondary sexual characteristics (via introgression with BT) might enjoy higher reproductive success compared to pure WSC. If this is the case, some genes of the non-native species will persist and spread even when pure BT disappear (Fukui et al., 2016), which potentially reduces the likelihood of persistence of pure WSC.
Although we could not directly assess the effects of introgression on the reproductive success of WSC due to insufficient sample size, our previous broader-scale survey showed that most hybrids were derived from backcross with white-spotted charr (Fukui et al., 2016).
Future study should focus on the direction of hybridization and its mechanisms.
Reduced hybrid reproductive success may inhibit the spread of non-native alleles and the development of hybrid swarms. We used, however, only 13 microsatellite markers to assess the levels F I G U R E 4 Reproductive success (number of offspring assigned per individual) as a function of male body size (a) and PC1 (b), and female body size (c) for brook trout and hybrid spawners. The p-values of above predictor variables which were calculated in the zero-inflated models (see Table 2

| Limitations of the study
While genomic data from next-generation sequencing (NGS) has recently been accumulating, we used only 13 microsatellite loci in the present study. Consequently, the hybrid index should be less accurate compared to NGS data. Also, we cannot know how much introgression has been occurring in the native species, as well as the loci under selection. In addition, parents were not determined for 40%-50% of juveniles, indicating that we could have missed some mature adults.
We tried to collect most of the adults in the study reaches by 2-4 passes of electrofishing until few fish were captured in the last pass.
However, since breeding dispersal is common in stream salmonids (Northcote, 1997;Hutchings and Gerber, 2002), some individuals might have moved to and reproduced in the study tributary between our sampling periods. Importantly, mainstem migratory fish often grow bigger (Koizumi, Yamamoto, & Maekawa, 2006) and, therefore, may attain higher reproductive success compared to smaller residents. We might have failed to collect such large migratory adults, which could explain the low parentage assignment. However, this is the nature of field studies in salmonids: similar attempts for estimating reproductive success showed the equivalent levels of parentage assignments (32%-61%: Thériault et al., 2007;Muhlfeld et al., 2009;Serbezov, Bernatchez, Olsen, & Vøllestad, 2010). Most importantly, even with the limited sampling, the results showed consistent patterns predicted from previous studies. Another important aspect, we missed is annual variation. A more detailed picture will be clarified with an NGS approach and additional surveys.

| CON CLUS ION
This study showed the relative reproductive success of hybrids and their parental species in the wild, despite that there are few field studies on directly quantifying relative fitness of hybrids and their parental species in human-mediated hybrid zones. The reproductive success of hybrids was lower than that of the two parental species.
Our results supported that both natural selection derived from outbreeding depression and sexual selection can affect the decision of hybrid fitness, and suggest that the more adaptive male traits derived from non-native species may facilitate the genetic extinction of native population via asymmetric introgression. As this study could not completely separate effects of sexual selection from other factors which can influence hybrid fitness, to quantify other fitness components, such as survival or growth, would appear to be an effective approach for unraveling further causes reducing the hybrid fitness.

ACK N OWLED G M ENTS
We appreciate the staff of the University of Tokyo Hokkaido Forest for supporting our field sampling. We thank T.

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R CO NTR I B UTI O N S
SF and IK designed the study. SF and SLM conducted the field survey. SF performed DNA experiments and statistical analysis, and wrote the original manuscript. All authors contributed to the revisions of the manuscript.