Influence of genetic dissimilarity in the reproductive success and mate choice of brown trout – females fishing for optimal MHC dissimilarity



    1. Södertörn University College, School of Life Sciences, Huddinge, Sweden
    2. Population Biology and Conservation Biology, Department of Ecology and Evolution, Evolutionary Biology Centre, Uppsala University, Uppsala, Sweden
    Search for more papers by this author

    1. Population Biology and Conservation Biology, Department of Ecology and Evolution, Evolutionary Biology Centre, Uppsala University, Uppsala, Sweden
    2. Institute of Freshwater Research, Drottningholm, Sweden
    Search for more papers by this author

    1. Institute of Freshwater Research, Drottningholm, Sweden
    2. Animal Ecology, Department of Ecology and Evolution, Uppsala University, Uppsala, Sweden
    Search for more papers by this author
  • M. GRAHN

    1. Södertörn University College, School of Life Sciences, Huddinge, Sweden
    Search for more papers by this author

Lars Forsberg, Södertörn University College, School of Life Sciences, Box 4101, SE 141 04 Huddinge, Sweden.
Tel.: +46 8 608 4734; fax: +46 8 608 4510; e-mail address:


We examined the reproductive success of 48 adult brown trout (Salmo trutta L.) which were allowed to reproduce in a stream that was controlled for the absence of other trout. Parentage analyses based on 11 microsatellites permitted us to infer reproductive success and mate choice preferences in situ. We found that pairs with intermediate major histocompatibility complex (MHC) dissimilarity mated more often than expected by chance. It appears that female choice was the driving force behind this observation because, compared with other individuals, males with intermediate MHC dissimilarity produced a larger proportion of offspring, whereas female reproductive output did not show this pattern. Hence, rather than seeking mates with maximal MHC dissimilarity, as found in several species, brown trout seemed to prefer mates of intermediate MHC difference, thus supporting an optimality-based model for MHC-dependent mate choice.


Mate choice can include considerations of genetic compatibility (Trivers in Campbell, 1972; Tregenza & Wedell, 2000; Neff & Pitcher, 2005) and the most extensively studied genes in this context are the MHC class I and class II loci (Apanius et al., 1997; Penn & Potts, 1999; Bernatchez & Landry, 2003; Piertney & Oliver, 2006). These genes are primarily involved in vertebrate adaptive immune response regulation by encoding an essential part of the MHC molecule, presenting self- and nonself peptides to T cells at cell surfaces (Klein, 1986). Consequently, these loci are under pathogen-mediated selection (Apanius et al., 1997; Penn & Potts, 1999) and the polymorphism is the highest of all coding loci found among vertebrates (Apanius et al., 1997).

Several hypotheses have been suggested to describe the selection underlying the widespread MHC polymorphism in natural populations. The intimate role of MHC genes in the immunological framework focuses attention on the pathogen-driven selection. According to the overdominance hypothesis (Doherty & Zinkernagel, 1975), heterozygote advantage should sustain polymorphism, whereas the negative frequency-dependent selection hypothesis holds that polymorphism is preserved when rare alleles have a selective advantage (Takahata & Nei, 1990; Slade & Mccallum, 1992). When MHC polymorphism is maintained by pathogen-driven selection, reproductive behaviours resulting in matings with MHC-dissimilar mates are also favoured (Penn & Potts, 1999). The key mechanisms in such MHC-based sexual selection models are negative assortative (disassortative) mating, i.e. avoidance of MHC similar mates, and associated fitness benefits (Brown & Eklund, 1994; Penn & Potts, 1999). One benefit could be enhanced pathogen resistance in offspring, either by increasing MHC heterozygosity (Potts & Wakeland, 1990) or by presenting a ‘moving target’ for evolving pathogens (Potts & Slev, 1995; Penn & Potts, 1999). An alternative and nonexclusive hypothesis involves avoidance of inbreeding (Potts & Wakeland, 1993; Brown & Eklund, 1994), whereby MHC-based disassortative mating preferences function to prevent kin mating, thereby reducing the negative consequences of inbreeding, similar to other genetic incompatibility systems, e.g. the SI system in plants (Matton et al., 1994).

Intuitively, consideration of MHC compatibility between mates is an appealing ingredient of mate choice, given the role of MHC genes in the discrimination of self from nonself at the cellular level. It has recently been found that MHC peptide ligands are used as olfactory cues and that MHC-linked olfactory receptor genes are involved in the process of discrimination in mice and fish (Boehm & Zufall, 2006), e.g. in three-spined sticklebacks (Gasterosteus aculeatus) (Milinski et al., 2005). These findings indicate a molecular mechanism by which individuals sense compatibility through the body odour of conspecifics. The high variability of MHC loci in combination with the ability of MHC alleles to influence body odour provides the necessary ingredients for a genetically based recognition system at the organismal level (Grafen, 1990).

The first indication of MHC-based mating preference was reported 30 years ago when congenic female mice were found to prefer MHC-dissimilar mates (Yamazaki et al., 1976). Since then, MHC-based mate preferences have been further described in mice (Yamazaki et al., 1988; Potts et al., 1991; Penn & Potts, 1998) and several other vertebrates, e.g. humans (Ober et al., 1997; Jacob et al., 2002), reptiles (Olsson et al., 2003), birds (Freeman-Gallant et al., 2003; Bonneaud et al., 2006) and fish (Landry et al., 2001; Reusch et al., 2001). In some mating systems, however, there may be no room for consideration of MHC compatibility, and several studies have found no correlations between MHC characteristics and mate choice (Paterson & Pemberton, 1997; Ekblom et al., 2004; Westerdahl, 2004; Richardson et al., 2005). Moreover, some features of the adaptive immune system predict that, under certain circumstances, optimally dissimilar mates are preferable to mates of even greater dissimilarity. Although disassortative mate preferences increase pathogenic scope or reduce the risk of inbreeding, at some point the positive effects of reproducing with disparate mates will be exceeded by negative effects such as self-reactive autoimmune responses and a net loss of T cells (Nowak et al., 1992; Deboer & Perelson, 1993) or by other factors such as outbreeding depression (Bateson, 1983; Thornhill, 1993) and fitness loss caused by the disruption of co-adapted gene complexes (see Hendry et al., 2000; Neff, 2004; Bonneaud et al., 2006). This would result in a trade-off, and the optimal and preferred choice would be mates of intermediate MHC dissimilarity (Penn & Potts, 1999). The predominant finding in studies examining MHC-based mate choice, however, has been disassortative preferences, whereas conclusive empirical support for the theoretically predicted optimally based MHC-dependent mating preference has been scarce in the literature (but see Reusch et al., 2001; Bonneaud et al., 2006).

Genetically based mate choice can function to increase genetic quality in offspring by means of ‘good genes’ or ‘compatible genes’ (reviewed by Neff & Pitcher, 2005). Good genes are characterized by alleles capable of increasing fitness irrespective of the architecture of the rest of the genome, and the intrinsic additive genetic variation responds to directional selection. In such mating systems, individuals of the choosing sex will have similar mate preferences for individuals of the other sex possessing the ‘good gene’, e.g. a favourable allele. In mating systems based on ‘compatible genes’, on the other hand, the interactions of male and female genotypes are more important, and mate preference varies among individuals searching for genotypes to complement their own. Neff & Pitcher (2005) propose that mating systems are simultaneously influenced by the effects of ‘good genes’ and ‘compatible genes’ and that they alternate between states of high and low levels of additive and nonadditive genetic variation over evolutionary time scales. This means that the underlying selective mechanisms for MHC-based mate choice vary over time, and previous MHC studies in salmonid species have found features consistent with both the ‘good’ (Lohm et al., 2002) and the ‘compatible’ genes’ (Arkush et al., 2002) mating systems.

Salmonids have previously been found to discriminate conspecifics based on MHC class IIβ characteristics, for example arctic charr (Salvelinus alpinus) (Olsen et al., 1998) and Atlantic salmon (Salmo salar) (Landry et al., 2001). MHC-based mate preferences are most likely to evolve in species at risk of inbreeding (Penn & Potts, 1999) and as brown trout reproduce in their natal stream as a result of strong homing behaviour, the probability of meeting close kin at spawning sites is obvious. Brown trout is a polygamous species with a mating system mainly driven by intrasexual competition; females hold and defend territories whereas males compete for access to females (Petersson & Jarvi, 1997; Petersson et al., 1999). During reproduction females bury their eggs in gravel and the chosen male sheds sperm over them. The fertilized eggs hatch and the fry (or alevins) emerge from the gravel on their own without parental care. Hence, the mating system diminishes the scope for some confounding factors common in studies of mate choice, e.g. cryptic female choice, parental care and biased investment in zygotes sired by specific males.

The aim of this study was to investigate mate choice in relation to genetic similarity at MHC and microsatellite loci in brown trout (S. trutta). Inferences of mate choice preferences were made after examining mating patterns as an effect of dissimilarity of pairs and by investigating the effects of individual genetic properties on the reproductive success of individuals.

Materials and methods

Mate choice experiments

The reproductive success of 48 adult anadromous brown trout (S. trutta L.) was examined in two consecutive replicates in an experimental stream in Dalälven river, Älvkarleby, central Sweden (60°38′N and 17°26′E) (see Dannewitz et al., 2004). This 110-m-long semi-natural habitat is equipped with traps at the inlet and outlet and can be drained of water. All fish used in the experiment were native to the Dalalven River and were caught in a permanent trap in Alvkarleby. Some fish had an entirely wild origin whereas others were reared in a hatchery for the first 2 years of their life before being released into the river as smolts. A previous study found that wild or reared origin did not affect the reproductive success of fish (Dannewitz et al., 2004). The first replicate commenced in October 2000 when 12 female and 12 male breeders were released into the experimental stream, which was free from other breeders (including jacks), and allowed to spawn at will. Ovulating females and males with running milt were selected for the experiments to minimize the risk of bias derived from variation in maturation status. In the following spring, 384 offspring were randomly sampled in the outlet trap. The second replicate started in October 2001 and was performed in the same way as the previous one, with 12 new breeders of each sex, 384 randomly collected offspring in the spring and so forth. More details about the experimental design and set-up can be found elsewhere (Dannewitz et al., 2004). The work was approved by the Ethics Committee for Animal Research and followed standards set by the Swedish Ministry of Agriculture (licence 34 3632-92).

Microsatellite and parental analysis

Breeders and sampled offspring were screened for genetic variation at 11 microsatellite loci (see Dannewitz et al., 2003, 2004). Parentage of the sampled offspring was established by comparing the alleles of offspring with those in each of the potential parental crosses (O'Reilly et al., 1998). The number of offspring produced by each individual parent fish was estimated by multiplying the fraction of offspring from a parent fish in the trap sample with the total number of offspring in the trap. The total number of offspring in the trap was estimated by dividing the mass of all caught offspring with the mean mass of 300 randomly selected offspring from the trap. Based on the parental analysis we were able to trace mating events and reconstruct the families produced (for details see Dannewitz et al., 2004).

MHC typing and analysis

To characterize the MHC class IIβ locus of the breeders we examined a 257-bp-long sequence of exon 2, encoding the peptide-binding sites in the β-chain of the MHC molecule. Following DNA extraction (Laird et al., 1991) and quantification using a Helios spectrophotometer (Unicam Instruments, Cambridge, UK), amplification with polymerase chain reaction (PCR) was performed in a Mastercycler gradient (Eppendorf, Hamburg, Germany) with a 25-μL mixture containing 100 ng of DNA template, 20 pmol of primer, 0.9 μg of BSA, 1x PCR reaction buffer, 25 nmol of MgCl2, 125 μm dNTP and 0.2 units of AmpliTaq DNA Polymerase (Applied Biosystems, Foster City, CA, USA). The forward primer was TVS 4501 (Langefors et al., 2000) labelled with fluoresceine and the reverse primer was MG 14 with a 40-bp GC-clamp at the 3′-end (Langefors et al., 2000). The PCR temperature profile started with a 3-min denaturation at 94 °C, followed by 40 cycles at 95 °C for 30 s, 30 s annealing at 55 °C and 30 s extension at 72 °C. Cycling was followed by a 10-min extension at 72 °C and the temperature profile ended with a 4 °C hold. PCR products were mixed in 10X BlueJuiceTM Gel Loading Buffer (Invitrogen, Carlsbad, NM, USA) in accordance with manufacturer's instructions and separated using denaturating gradient gel electrophoresis (DGGE) with equipment supplied by the CBS Scientific Company (Delmar, NY, USA). DGGE gels with 5% 19 : 1 acrylamide–bisacrylamide were run at 1900 Vh and scanned in a FLA-3000 (Fuji Photo Film Co., Tokyo, Japan). All individuals were typed in both 0–80% and 20–60% urea gradient gel (Langefors et al., 2000). At this stage either one or two distinct bands were found in all individuals and those with one band were further investigated by varying the annealing temperature in the PCR, DGGE runtime, gel gradient and analysis of alleles in produced offspring. Each visualized allele of all 48 individuals was collected from the gels (Langefors et al., 2001) and amplified using the same PCR conditions as before but with unlabelled and non-GC-clamped primers. PCR products were precipitated in NH4Ac and forward and reverse sequenced (BigDye® Terminator v1.1 Cycle Sequencing Kit; Applied Biosystems, Foster City, CA, USA). After NaAc precipitation and suspension in a loading dye of formamide and dextran, followed by 3 min of denaturation, the fragments were sequenced with an ABI 377 (Applied Biosystems, Foster City, CA, USA) in accordance with manufacturer's instructions. The sequences were aligned and translated into amino acid sequences using BioEdit 5 (Hall, 1999), giving a total of 34 unique DNA sequences representing 31 translated amino acid sequences. DNA sequences were submitted to Genbank (url: and can be retrieved using the following accession numbers: DQ491027DQ491060.

Estimating the genetic dissimilarity of individuals

Different approaches can be used to estimate the genetic dissimilarity of individuals and between pairs of individuals. Estimators such as levels of heterozygosity, number of shared alleles and allelic distances have previously been applied to describe and quantify genetic differences. For the analyses of individual reproductive success in this study, the genetic dissimilarity of individuals was defined in scores with reference to the other individuals, so that genetically unique and dissimilar individuals were assigned higher scores compared with more genetically common and similar individuals.

The individual microsatellite scores were based on a pairwise identity index calculated for all breeders using IDENTIX 1.1 (Belkhir et al., 2002). The coefficient of relatedness computed by IDENTIX 1.1 has low sampling variance compared with other coefficients of relatedness (Belkhir et al., 2002). The program estimates the degree of relatedness of individuals using a permutation method that compares the observed distribution of the moments of pairwise relatedness coefficients to that expected in unstructured populations. In this computation, individual genotypic microsatellite data were translated into a numeric difference between each pair of individuals in the replicate. Individual microsatellite scores were calculated as the average individual numeric difference to the other individuals in the replicate.

The individual MHC scores were based on the MHC alleles sequences found in individuals. These scores were computed using two different approaches, a summation method (after Landry et al., 2001) and a maximal distance method. First, Poisson-corrected pairwise amino acid distances were calculated in MEGA3 (Kumar et al., 2004) for each pair of individuals and their two alleles. The qualitative measure of amino acid distances may be the most relevant estimator of MHC divergence in the context of mate choice (Landry et al., 2001). This computation generates four numbers reflecting the sequence differences between the alleles found in the two compared individuals. Following the summation method, the four numbers were added up to provide an estimate of MHC similarity between two individuals (from Landry et al., 2001). In the maximum distance method, only the largest of the four distances was used to estimate dissimilarity between two individuals. As with the microsatellite scores, individual MHC scores were separately calculated as the average individual difference to other individuals in the replicate, both for sums and maximal distances. Henceforth, the summation approach will be referred to as the MHC score (sum) and the maximal distance approach as the MHC score (MD).

Statistical analyses

The mean and variance of genetic distances among confirmed pairs were tested against expectations under random mating conditions by means of a permutation test. Prior to the permutation test, the two data sets from the two replicates were pooled using the residuals from a Gaussian-distributed generalized linear model (GLM) analysis with the replicate as a factor, using R v. 2.4.1 statistical software (R Development Core Team, 2006). In the permutation test, the mean and variance were calculated for 61 randomly sampled pairs (the same number as the actual pairs) from the 288 potential mating constellations in each of 50 000 replicates using R v. 2.4.1 statistical software (R Development Core Team, 2006). The frequency distributions thus represent random mate choice in 50 000 mating events including the 288 potential pairs. The resampling strategy prevented individuals from different replicates being associated as mates in the test. The observed mean and variance in confirmed pairs were compared with the mean and variance of the resampled data.

The effect of individual MHC and microsatellite scores on the number of produced offspring was analysed in the two replicates, both separately and pooled together, as the total number of produced offspring varied considerably between the two replicates. Variables in the pooled analysis were transformed from numerical to proportional data to take account of the inter-year variation. Response variables in these analyses were achieved matings and number or proportion of offspring. Explanatory variables were individual weight and individual MHC score (sum), MHC score (MD) or microsatellite score. Body mass was included in the model due to the fact that a previous study performed in a stream aquarium with fish from the same population showed that male body mass explained about 18% of the variation in the male position in the dominance hierarchy (Petersson & Jarvi, 1997). The statistical associations between response and explanatory variables were estimated by means of Poisson-distributed GLMs, where the effects of explanatory variables were evaluated with sequential anova using the R v. 2.3.1 statistical package (R Development Core Team, 2006). Sequential models are dependent on the order of explanatory variables, while variation explained by a term fitted is removed before testing subsequent terms (Venables & Ripley, 2002). Hence, the effect of individual body mass was controlled for before testing the impact of individual genetic scores on the response variables. The square of explanatory variables tested nonlinear associations.


During the first replicate of the experiment, an estimated 18 478 offspring were produced in 30 different full-sib families and an estimated 3965 detected in 31 families in the second replicate. Analyses of reproduction and mate choice were performed using two different approaches. First, the reproduction between pairs of individuals, which is the most intuitive unit for studying mate choice preferences, was analysed. Analyses were also performed on the level of individual fish, examining individual reproductive output in relation to individual genetic characteristics when compared with the population means.

Dissimilarity between mates

The permutation test showed that the variances of genetic distances among the randomized pairs exceeded the observed variance in the confirmed pairs when estimating MHC amino acid distance between individuals using both the summation method and the maximal distance approach (one-tailed permutation test, P < 0.05, Fig. 1a). By contrast, the observed variation of microsatellite distances among confirmed pairs did not differ significantly from random expectations (Fig. 1b).

Figure 1.

 Observed variance (arrows) in genetic difference among confirmed pairs compared with distributions of expected results from random mating. The variances of the resampled pairs are plotted on the X-axis and their frequencies on the Y-axis. The dotted line shows the 95% lower percentile of the distribution of variances in genetic distance for 50 000 random resamples of all potential mating combinations. (a) The observed variance of MHC distances (sum) among confirmed pairs (0.0406) was less than the lower 95% confidence limit (0.0413). (b) The observed variation of microsatellite distances among confirmed pairs did not differ significantly from random mating.

Variance was used as a statistic rather than the mean in the permutation test because most of the pairs had an intermediate distance. Based on a hypothesis of preferences for intermediate mates, the observed mean of confirmed pairs would be expected to be similar to the mean of all potential pairs. In fact, the observed mean of MHC differences among confirmed pairs (0.8774) was similar to the resampled mean (0.855) and well within the 95% confidence intervals (two-tailed 95% CI 0.7996 and 0.9092). The observed mean for the microsatellite differences among the confirmed pairs (0.598) was also similar to the resampled mean (0.595) and within the 95% confidence limits (two-tailed 95% CI 0.5691 and 0.6205).

Reproduction of individuals

The number of successful matings varied between individuals: one male reproduced with seven females while four individuals did not reproduce at all. In the first replicate, males had mated with, on average, 2.7 females (SE ±0.62) and females had mated with, on average, 2.8 males (SE ±0.31). In the second replicate both males and females had reproduced with, on average, 2.6 mates (SE ±0.42 and 0.53 for males and females respectively). There was no apparent reproductive skew that would have indicated a strong ‘good gene’ effect. The large number of MHC alleles in relation to the sample size precluded a thorough analysis of the ‘good gene’ effect. We found that males with high microsatellite scores (i.e. greater dissimilarity to other individuals) had mated with more partners than males with low microsatellite scores (Table 1e) and produced a significantly higher proportion of offspring when compared with their counterparts with low microsatellite scores (Table 2e; Fig. 2e). We found no association between individual MHC scores and the number of matings (Table 1a–d). However, when estimating the individual MHC scores by means of the maximal distance method, we found a significant nonlinear effect indicating that males with intermediate MHC scores had produced a higher proportion of offspring compared with other males (Table 2c; Fig. 2c).

Table 1.   Results from generalized linear models testing the effects of individual MHC and microsatellite scores on the number of mates in males and females.
Proportion of matings dependent ond.f.Deviance P
  1. * denotes significant P-value with an α-value of 0.05.

a. MHC score (sum) in males
 Body mass10.470.131
 MHC score10.100.472
 Squared MHC score10.030.706
 Model residual204.93 
b. MHC score (sum) in females
 Body mass1< 0.010.922
 MHC score10.030.724
 Squared MHC score10.040.679
 Model residual204.77 
c. MHC score (MD) in males
 Body mass10.480.143
 MHC score1< 0.010.948
 Squared MHC score1< 0.010.938
 Model residual205.06 
d. MHC score (MD) in females
 Body mass1< 0.010.922
 MHC score1< 0.010.915
 Squared MHC score10.010.871
 Model residual204.81 
e. Microsatellite score in males
 Body mass10.480.115
 Microsatellite score11.050.024*
 Squared microsatellite score10.060.578
 Model residual deviance203.95 
f. Microsatellite score in females
 Body mass1< 0.010.921
 Microsatellite score10.030.691
 Squared microsatellite score10.300.230
 Model residual deviance204.49 
Table 2.   Results from generalized linear models testing the effects of individual MHC and microsatellite scores on the production of offspring in males and females.
Proportion of offspring dependent ond.f.Deviance P
  1. * denotes significant P-value with an α-value of 0.05.

a. MHC score (sum) in males
 Body mass10.220.114
 MHC score10.240.100
 Squared MHC score10.010.778
 Model residual201.61 
b. MHC score (sum) in females
 Body mass10.010.815
 MHC score10.460.080
 Squared MHC score10.040.588
 Model residual202.33 
c. MHC score (MD) in males
 Body mass10.220.094
 MHC score10.070.337
 Squared MHC score10.350.036*
 Model residual201.43 
d. MHC score (MD) in females
 Body mass10.010.840
 MHC score10.010.806
 Squared MHC score10.030.679
 Model residual202.78 
e. Microsatellite score in males
 Body mass10.220.085
 Microsatellite score10.320.039*
 Squared microsatellite score10.100.229
 Model residual deviance201.43 
f. Microsatellite score in females
 Body mass10.010.840
 Microsatellite score10.010.781
 Squared microsatellite score10.020.759
 Model residual deviance202.80 
Figure 2.

 The association of individual reproductive success and genetic dissimilarity in males and females. MHC and microsatellite scores estimated as described in the text. Symbols represent individual fish in the first (○) and second (bsl00084) replicate. Dashed lines indicate 95% confidence intervals in panels with significant relationships. The Y-axis in all sections represent individual residual reproductive success estimated as the proportion of offspring after taking into account the effect of individual weight. The X-axis indicate male and female genetic scores. (a) Effect of male MHC score (sum). (b) Effect of female MHC score (sum). (c) Significant nonlinear effect of male MHC score (MD). (d) Effect of female MHC score (MD). (e) Significant linear effect of male microsatellite score. (f) Effect of female microsatellite score.

Most of the breeders had intermediate MHC scores (Fig. 3a,c). The distribution was similar in males and females (two sample Kolmogorov–Smirnov test, P =0.93). Most of the breeders also had an intermediate level of heterozygosity at the MHC locus, estimated as the pairwise amino acid distance of the two alleles found in individuals, a pattern similar to that found in natural populations of the three-spined stickleback (G. aculeatus) (Wegner et al., 2003). By contrast, most parental fish had low microsatellite scores (Fig. 3b,d). The frequency distributions of microsatellite scores were also independent of sex (two-sample Kolmogorov–Smirnov test, P = 0.93).

Figure 3.

 Frequency distributions of genetic scores in males and females in the experiment. Individuals were categorized based on the genetic score in six equivalent groups. (a) Distribution of MHC scores (MD) in males (n = 24). (b) Distribution of microsatellite scores in males (n = 24). (c) Distribution of MHC scores (MD) in females (n = 24). (d) Distribution of microsatellite scores in females (n = 24).

The MHC regions in several species have previously been found to contain nonexpressed pseudogenes (Klein, 1986). With the DNA-based typing protocol used here, amplification of an expressed locus cannot be guaranteed. Nevertheless, no signs of pseudogenes such as stopcodons or frameshift mutations were found in the DNA sequences retrieved separately from the 48 breeders. Furthermore, the pattern of the MHC allele sequence variation was consistent with expressed loci [Tajima's test in MEGA3 (Kumar et al., 2004); D = 2.21, P < 0.05: dN/dS in DnaSP (Rozas et al., 2003); Zbinding sites = 2.83, P =0.003, Znonbinding sites = 1.37, P = 0.086].

The protocol used to investigate the MHC repeatedly amplified either one or two alleles in individuals, supporting previous findings that salmonid genomes contain a single expressed MHC class IIβ locus (Langefors et al., 1998; Hansen et al., 1999; Grimholt et al., 2000; Langefors et al., 2000; Landry & Bernatchez, 2001; Shum et al., 2001).


The major finding in this study was that pairs with an intermediate amino acid distance at the examined MHC class IIβ locus had mated more often than expected by chance. The permutation test showed that the high mating frequency of intermediate pairs was not merely an effect of random sampling in a population with a high proportion of intermediate pairs. We used a one-tailed test as we reasoned that the hypothesis of equal variances was one sided and that the expectation of greater variance among the observed matings compared with randomly generated matings was unrealistic.

We also found that males with intermediate MHC dissimilarity had produced a larger proportion of offspring compared with males with lower and higher MHC dissimilarity. This association was significant when MHC dissimilarity between pairs was estimated using the maximal distance method. The nonlinear effect of male MHC dissimilarity was significant in the first replicate but not in the second. As can be seen in Fig. 2a–d, offspring from parents with low MHC scores were absent in the second replicate. This pattern was not found in corresponding plots with microsatellite scores (Fig. 2e,f), indicating that offspring from parents with low MHC scores were not produced or were lost during the second replicate. The higher mortality is consistent with the lower total number of offspring in the second replicate.

Males with high microsatellite dissimilarity mated more often and produced a higher proportion of offspring than other males. If these patterns resulted from female preferences for genetically dissimilar males, it would decrease the risk of producing inbred offspring and indicate an active avoidance of inbreeding. Similar results were recently reported in the closely related Atlantic salmon (S. salar), as females that mated with more dissimilar males increased the chances of producing outbred offspring when estimated at five microsatellite loci (Garant et al., 2005).

A consequence of performing mate choice studies in situ is that preferences must be estimated subsequently. We used the mating of pairs and individual reproductive output as proxies for mate choice, similar to the design of Landry et al. (2001), where sampling of offspring was also conducted some time after the actual mating. Factors not taken into account in the natural settings could have affected individual reproductive success, although using the following rationale we argue that an effect of female mate choice can be detected in the reproductive output of males and that male mate choice can likewise be detected in the reproductive output of females.

If the observed patterns in male reproductive success (Fig. 2a,c,e) were caused by some factor other than female mate choice, a corresponding pattern would most likely have been found in the reproductive success of females (Fig. 2b,d,f). As illustrated in Table 3, no factor apart from female mate choice can explain why the observed pattern of reproductive success of intermediates was only found when the offspring were sorted based on paternal dissimilarity. For example, if there had been an unknown selective advantage in offspring from parents with intermediate MHC dissimilarity (such as a higher survival rate of the eggs, larvae or fry), the observed pattern in male reproductive success would most likely have been found in the reproductive success of females as well. Similarly, if mate choice had been a random process, independent of MHC in both sexes, the high proportion of MHC intermediate males (Fig. 3a) and the high proportion of MHC intermediate females (Fig. 3c), would have resulted in a high proportion of offspring from intermediate parents in both sexes. Using the same logic, the distribution of microsatellite scores of males (Fig. 3b) and females (Fig. 3d) was similar, with a high proportion of individuals with low scores. These distributions were not reflected in the patterns of male or female reproductive success (Fig. 2e,f). Instead, the positive correlation between male reproductive success and male microsatellite dissimilarity indicates a nonrandom female mate choice with respect to microsatellite dissimilarity and a female preference for males of a general genomic dissimilarity. We therefore propose that the observed patterns of male reproductive success were the results of female mate choice.

Table 3.   Potential causal factors related to the observation of higher reproductive output of MHC intermediate parents.
Factors with the potential to explain the higher reproductive output of MHC intermediate parentsIf true, pattern likely to be found in
Male reproductive successFemale reproductive success
1. MHC-dependent survival of eggs or fryxx
2. MHC-dependent behaviour of fryxx
3. Sampling biased on fry MHC propertiesxx
4. Large number of MHC intermediate malesx 
5. Large number of MHC intermediate females x
6. Male MHC-based mate choice of MHC-intermediate females x
7. Female MHC-based mate choice of MHC-intermediate malesx 

The summation and the maximal distance methods for estimating individual MHC dissimilarity differ from a biological perspective in the sense that the sum of distances between individuals reflects an approximation of the alleles in potential mates, whereas the maximal distance represents the largest distance of alleles in potential mates. The two methods are similar and comparable, as both assign higher scores to more divergent individuals and the individual scores generated by the methods are strongly positively correlated (F1,46 = 54.94, P < 0.001), and the coefficient of variance of the estimators was also similar, CV 0.1047 (sum) and 0.1042 (MD).

Although MHC-based mate choice preferences have been described in several vertebrate species, the findings have been dominated by disassortative preferences and only a few examples in the literature describe preferences for optimal MHC divergent mates (e.g. Reusch et al., 2001; Bonneaud et al., 2006). It has been argued that selection for intermediate number of alleles would not maintain high polymorphism over time in natural populations without unrealistically high mutation rates (Hedrick, 2004). On the other hand, mutations would not be the only source of MHC variation in natural populations, and processes such as recombination could occur (Wegner et al., 2004). MHC-dependent mate choice does not appear to be a universal phenomenon, as certain mating systems may have no room for genetic compatibility considerations, and several studies have found no correlations between MHC characteristics and reproductive success (e.g. Paterson & Pemberton, 1997; Ekblom et al., 2004; Westerdahl, 2004; Richardson et al., 2005). However, if the purpose of MHC-based preferences is optimal differences between mates, the outcomes of mate choice studies will be variable and highly dependent on the actual MHC diversity among potential mates. It is therefore to be expected that some studies will only find a slight deviation from a random mating pattern when comparing observed mean differences between mates and random expectation in populations where most mates are of intermediate dissimilarity. Hence, optimal MHC-dependent mate choice preferences may be more common than indicated in the literature.

Most species examined for MHC-based mate choice, e.g. mice, humans and birds, have a complex MHC with several loci coding for MHC molecules with MHC class I and class II genes arranged in tightly linked haplotypes (Klein, 1986). In teleostean fish, on the other hand, the classic MHC class I and II genes are typically found in different linkage groups (Hansen et al., 1999; Sato et al., 2000; Shum et al., 2001; Stet et al., 2003) and salmonid species usually express one MHC class IIβ locus (Langefors et al., 1998; Hansen et al., 1999; Grimholt et al., 2000; Landry & Bernatchez, 2001; Shum et al., 2001). In species possessing several MHC class IIβ loci, theory predicts selective mechanisms for an optimal number of expressed alleles (Penn & Potts, 1999) but this prediction is weakened in species expressing only one MHC class IIβ locus. It is tempting to interpret the preference for males with an optimal MHC dissimilarity on the basis of the optimal outbreeding theory (Bateson, 1983). However, the general genomic dissimilarity as reflected in the 11 microsatellite loci was not correlated with the MHC class IIβ dissimilarity in the 48 breeders (anova; effect of MHC scores on microsatellite scores; F3,44 =0.08, P = 0.78). Similar patterns of independence between MHC and neutral markers have previously been described in Atlantic salmon (S. salar) (Landry & Bernatchez, 2001) and sticklebacks (G. aculeatus) (Reusch et al., 2001). Under these conditions, MHC-based mate choice cannot function to reduce general genomic inbreeding, although it may prevent fitness costs associated with homozygosity in close-kin matings (Apanius et al., 1997).

Alternatively, female avoidance of the most MHC-dissimilar males may have evolved in response to preventing outbreeding depression at the MHC where individuals with elevated MHC dissimilarity are less well adapted to the local pathogenic environment compared with other individuals. Recently, McGinnity et al. (2003) reported a decrease in local adaptation and reduced fitness in a wild population of Atlantic salmon (S. salar) during interactions with escaped farm salmon without local adaptation, thus demonstrating the importance of local adaptation for fitness in a closely related species. Research on anadromous salmonids has found high levels of population differentiation at the MHC (Langefors et al., 1998; Landry & Bernatchez, 2001; Miller et al., 2001), indicating the local adaptation of these species’ MHC. It is likely that a similar MHC differentiation exists in the brown trout, with its closely related phylogenetical, ecological and behavioural patterns. If these assumptions hold true, it would suggest that female avoidance of highly MHC-dissimilar males could be a mechanism that has evolved in response to selection to ensure local adaptation of the MHC.

In conclusion, the permutation test showed that MHC intermediate pairs had mated more often than expected by chance. This process is most likely driven by female mate choice, as males with intermediate MHC dissimilarity had a higher reproductive output than other males, an observation not paralleled in the analysis of female reproductive output. The avoidance of males with highly dissimilar MHC would drive a process of local adaptation and an MHC splitting process between populations. A possible consequence of the preference for MHC intermediate males could therefore be reproductive isolation of populations over time. Such a process would be counteracted by the female preference for males with high general genomic dissimilarity.


We would like to thank P. Jacobsson, M. Lönn and K. Lehtila for their help. J. Höglund, J. Cunningham and two anonymous referees helped to prepare the manuscript. This work was supported by grants from the Swedish Energy Agency.