Multiple parasites are driving major histocompatibility complex polymorphism in the wild

Authors


 K. Mathias Wegner, Max-Planck-Institute of Limnology, Department of Evolutionary Ecology, August – Thienemann-Str. 2, 24306 Plön, Germany. Tel.: +494 522 763 252; fax: +494 522 763 310; e-mail: wegner@mpil-ploen.mpg.de

Abstract

Abstract Parasite mediated selection may result in arms races between host defence and parasite virulence. In particular, simultaneous infections from multiple parasite species should cause diversification (i.e. balancing selection) in resistance genes both at the population and the individual level. Here, we tested these ideas in highly polymorphic major histocompatibility complex (MHC) genes from three-spined sticklebacks (Gasterosteus aculeatus L.). In eight natural populations, parasite diversity (15 different species), and MHC class IIB diversity varied strongly between habitat types (lakes vs. rivers vs. estuaries) with lowest values in rivers. Partial correlation analysis revealed an influence of parasite diversity on MHC class IIB variation whereas general genetic diversity assessed at seven microsatellite loci was not significantly correlated with parasite diversity. Within individual fish, intermediate, rather than maximal allele numbers were associated with minimal parasite load, supporting theoretical models of self-reactive T-cell elimination. The optimal individual diversity matched those values female fish try to achieve in their offspring by mate choice. We thus present correlative evidence supporting the ‘allele counting’ strategy for optimizing the immunocompetence in stickleback offspring.

Introduction

Hosts and parasites often engage in arms races (Van Valen, 1973). These associations should result in close correlations between virulence traits of pathogen strains (or species) and the host's resistance genes, finally leading to diversification and balancing selection on both, virulence and resistance genes. The major histocompatibility complex (MHC) of vertebrates is the prime example for resistance genes with extensive genetic polymorphism attributed to balancing selection (Klein, 1986). MHC-gene encoded proteins present antigens to T lymphocytes and initiate specific immune responses. An excess of replacement over silent nucleotide substitutions in antigen binding regions (ABR) is strong evidence for balancing selection (Hughes & Nei, 1988). The crucial role of MHC molecules in antigen presentation also suggests pathogens as selective agents, but support for this hypothesis is limited and indirect (Hughes & Nei, 1988). A series of studies found associations between certain MHC alleles and infection with single pathogens (Hill et al., 1991; Thursz et al., 1995; Paterson et al., 1998; Godot et al., 2000; Langefors et al., 2001; Meyer & Thomson, 2001), supporting the notion of direct selection on MHC alleles by pathogens. Although in line with hypotheses on explaining MHC polymorphism, these findings are not sufficient to explain the extreme diversity of these genes. This is because any mechanism of balancing selection explaining MHC polymorphism with a single infectious agent would require extreme dynamics of host–parasite coevolution (Peters & Lively, 1999), which seems to be unlikely considering the stable distributions of macroparasites (Shaw et al., 1998). Rather, numerous pairs of single parasite-resistance allele associations will result in diversifying selection on MHC alleles within host populations (Apanius et al., 1997). This is particularly relevant when taking into account that the majority of animal species are parasites (Windsor, 1998), and that most animals suffer from multiple infections simultaneously. Surprisingly, we are not aware of any study examining the correlation between diversity of local parasite communities and MHC polymorphism.

Consequently, the goal of this study is to cover the full range of naturally occurring macroparasites in wild populations of three-spined sticklebacks (Gasterosteus aculeatus L.) for explaining MHC diversity. We predict that local populations exposed to a more diverse array of parasites will be more diverse in terms of their MHC genes. We focus here on MHC class II genes because they are most critical for triggering the immune response against nonviral, extra cellular antigens (Hughes & Yeager, 1998). Moreover, some MHC class II haplotypes have been associated with clinical severity of cestode infections in humans (Godot et al., 2000) and infection with intestinal nematodes in a ruminant population (Paterson et al., 1998) indicating functional significance of MHC class II genes in defence against macroparasites. Within class II genes, the second exon of the β chain was selected because it composes parts of the functional important peptide binding groove, and has been shown to be the most polymorphic part in many class II genes (Ohta, 1998; Hughes, 1999).

In host–parasite associations, the target of selection is the individual. Evidently, the diversity among single genotypes is necessary to explain population-wide polymorphism (Doherty & Zinkernagel, 1975; McClelland et al., 2000; Penn et al., 2002). This is particularly relevant when the genomic architecture of the MHC region is considered. There are only few species expressing only one MHC class IIB-locus (such as salmon Salmo salar L., Langefors et al., 1998), or having very compact regions (chicken Gallus gallus L.; Kaufman & Salomonsen, 1997). In most other species, however, the MHC covers an extensive region of up to 4 Mbp and most loci of class I and class II are duplicated (Klein, 1986). In particular the functionally important antigen-binding region (ABR) of MHC class II β chain has multiple expressed gene products in humans and other species (Swarbrick & Crawford, 1997; Málaga-Trillo et al., 1998; McConnell et al., 1998; Hughes, 1999). Assuming functional equivalency of these duplicated genes, individuals may thus possess between one (all loci homozygous with functionally similar copies at all loci) and two times the number of loci different alleles. Given that three-spined sticklebacks possess up to six MHC class IIB-loci (Sato et al., 1998; Reusch et al., 2001a) high levels of diversity can be created within individuals.

However, maximum individual diversity in MHC genes may not be beneficial. On theoretical grounds, Nowak et al. (1992) suggested an optimal, intermediate number of MHC alleles. A high diversity at MHC class IIB genes not only allows presentation of more antigens. As a negative consequence too many MHC variants will also result in presentation of more self-peptides, with subsequent elimination of corresponding self-reactive T-cell receptor lines (Mason, 2001). This process is exacerbated in class II molecules because they are heterodimers. Consequently, a combination of α and β chain will result in quadratic increase of functional class II molecules and T-cell elimination. Balancing selection would then outweigh diversifying selection and favour an optimal number of MHC alleles. Additionally, recent findings of Reusch et al. (2001a) and Aeschlimann et al. (P. B. Aeschlimann, M. A. Häberli, T. B. H. Reusch, T. Boehm and M. Milinski, unpublished data) suggest that ‘allele counting’, a strategy of sexual selection, is also favouring intermediate MHC diversity in offspring. Female fish select males with more alleles when they have less alleles than the population average themselves, while they tend to select males with lower individual MHC diversity when they have more MHC class IIB alleles than the population average (Reusch et al., 2001a). Assuming an adaptive value of such a sophisticated choice mechanism, one possible cause maximizing the offspring's fitness may be an optimal MHC diversity with lowest parasite burden.

Accordingly, we predict, in line with Nowak et al.'s (1992) theoretical expectation, that individual parasite load is lowest at an intermediate individual diversity of MHC class IIB genes. We test these predictions in a survey of parasite infections in wild caught sticklebacks from eight populations, and correlated their parasite load with MHC class IIB diversity in individuals and populations.

Materials and methods

Sampling design and parasite screening

Three-spined sticklebacks (G. aculeatus) were caught during three periods (September, November, March) from eight water bodies (lakes and rivers) belonging to three catchments in northern Germany (Table 1). Twenty-five fish of each catch were screened for parasite infections under a dissecting microscope. These fish harboured a total of 15 parasite species: Ciliata: Trichodina sp., Apiosoma sp.; Microsporidia: Glugea anomala; Crustacea: Argulus foliaceus; Monogenea: Gyrodactylus sp.; Digenea: Diplostomum spathaceum, Echinochiasmus sp.(?), Cyathocotyle prussica; Cestoda: Proteocephalus filicollis, Valipora campylancristrota; Acanthocephala: Acanthocephalus lucii, A. clavula; Nematoda: Camallanus lacustris, Contracaecum sp., Anguillicola crassus. The diversity of the local parasite community was expressed as Simpson's diversity index D using following formula:

Table 1.  Sampled three-spined stickleback ( Gasterosteus aculeatus ) populations, their parasite diversity as Simpson's D , MHC class II B and microsatellite diversity for pooled populations and single time points. Trammer see samples from November 2000 and March 2001 were omitted because of the low number of fish caught.
Catchment*
Habitat
 Population
Dieksee
SC
Lake
Kl. Plöner see
SC
Lake
Trammer see
SC
Lake
Vierer see
SC
Lake
Schwale
ST
River
Rönnau
TR
River
Söhren
SC
River
Wedeler Au
EL
Estuary
  • *

    SC=Schwentine, ST=Stör, TR=Trave, EL=Elbe, nd=not determined.

Parasite diversity
Simpson's DTotal 3.761 1.3663.929 3.421 1.1011.0001.346 2.368
September 2000 3.907 2.0543.929 4.449 1.0201.0001.346 2.191
November 2000 2.276 1.359nd 3.338 1.100ndnd 4.294
March 2001 2.479 2.217nd 2.375 1.003ndnd 1.924
MHC
Total allelesMean23242523.66713131125
September 200023252525131311nd
November 20002324nd2413ndnd28
March 20012323nd2213ndnd22
Microsatellites
Alleles locus−1Total 6.86 7.148.5710.29 7.14nd5.0013.29
September 2000 5.86 5.577.29 6.43 6.00nd5.0010.57
November 2000 5.00 7.00nd 6.43 6.29ndnd10.71
March 2001 5.29 5.14nd 6.29 4.86ndnd 8.57
image

where S is the number of species entering the index and Pi the relative abundance of species i to the total abundance of parasites.

Further information on populations and parasites can be obtained from Kalbe et al. (2002).

Characterization of MHC class IIB variation

Immunogenetic variation was assessed in a larger sample of 1017 fish, which comprised the random subsample of 434 individuals subjected to dissection and parasite screening. Characterization of MHC class IIB genes involved an amplification of a 124-bp sequence internal to the ABR of the exon 2 of MHC class II heavy β-chain genes (Binz et al., 2001) with polymerase chain reaction (PCR) reactions using two different combinations of primer pairs (Reusch et al., 2001a). As primers differed in two bases, each pair amplifies a unique subset of MHC class IIB sequences. To identify different sequences we separated the fluorescently labelled PCR products using single-stranded conformation polymorphism (SSCP) on an ABI 310 Capillary Sequencer (Applied Biosystems, Weiterstadt, Germany). For simplicity, we refer to different class IIB sequences as being alleles although they stem from different loci. Only those of the 434 dissected fish with unambiguous PCR products in both reactions were included in further analyses, finally leading to a sample size of n=299. Both primer pairs amplify ≈80% of all present alleles (Reusch et al., 2001a). Additionally, we might have missed a certain proportion of the variation within the remaining base pairs of the whole exon. However, Reusch & Wegner (T. B. H. Reusch & K. M. Wegner, unpublished data) found that in 56 more completely sequenced exons (211 bp) only three sequences were identical in the part of the exon included into the genotyping method used here, meaning that 95% of variation is covered by our typing method. Furthermore, as our null hypotheses can only be rejected when finding variation, any lack in resolution of our genotyping technique is always conservative with respect to our hypotheses. Due to presumably up to six duplications of the β chain loci (Sato et al., 1998) each primer amplified on average more than two alleles. In our data set the first primer pair amplified 2.78 (±0.96), whereas the second primer pair amplified 3.01 (±0.99) alleles per individual, resulting in 5.83 (±1.55) alleles/individual ranging from a minimum of two to a maximum of nine in total. Such a range in identifiable allele numbers can either be explained by interindividual variance in the number of loci present, a phenomenon previously described for teleost fish (Málaga-Trillo et al., 1998), or by recent duplications bearing functionally equivalent sequences. Work on the development of single locus primers is in progress but difficult because of the recency of locus duplications (T. B. H. Reusch & K. M. Wegner, unpublished data). Alternatively, the amplification of PCR products from duplicated genes may overestimate the functional variability, as not all copies of the gene necessarily get expressed. However, preliminary expression studies (K. M. Wegner, T. B. H. Reusch & T. Boehm, unpublished data) revealed that most of the sequences covered by our genotyping are expressed.

Microsatellite genotyping

Genome-wide variability has previously been shown to explain most of MHC population structure and diversity (Boyce et al., 1997; Hedrick et al., 2001), at least in endangered species with comparatively small population sizes. Hence, we tried to estimate the role of selectively neutral processes (genetic drift) for explaining MHC polymorphism using microsatellite markers. Accordingly, we typed the majority of individuals comprised in the MHC genotyping (i.e. n=920) at seven polymorphic microsatellite loci (Largiader et al., 1999) under conditions for PCR amplification described by Reusch et al. (2001b).

Statistical methods

We compared the total allele diversity of MHC with the genome-wide variability (i.e. all alleles present for MHC and alleles/locus for microsatellites, respectively) and with parasite diversity at the population level. In two of the three linear regressions test of normality or homogeneity of variance failed. Therefore, we used nonparametric Spearman's rank correlation coefficients ρ. To determine the relative importance of either pairwise correlation the other variable was statistically kept constant by using partial correlations. As the individual data did not violate assumptions of normality and homogeneity of variance, the correlation among individual MHC diversity and parasite load was estimated by ordinary least square regression. Individual parasite load varied strongly between habitats. To reduce variance introduced by this factor, we used residuals around population mean as an indicator of individual parasite load. Confidence limits for the minimum of the quadratic polynomial were obtained by bootstrapping the individual data set 1000 times. Quadratic polynomials were fitted to each bootstrap replicate and minima were calculated.

Results

Parasite vs. MHC diversity in populations

In the study populations, we found high levels of diversity ranging from 13 to 28 different MHC class IIB alleles (Table 1). This estimate was rather conservative because only 80% of alleles will be detected by our genotyping method and some sequences cannot be resolved by SSCP (Reusch et al., 2001a). Populations being exposed to a wider array of macroparasites tended to be more diverse in terms of their MHC class IIB genes (partial Spearman's rank correlation=0.808, P < 0.001, Fig. 1a). As a null model, any genetic diversity should be largely governed by population size and structure (Nei et al., 1975; Kimura, 1983). Accordingly, we found a marginally significant correlation in the number of MHC class IIB alleles and microsatellite alleles per locus at the population level (partial Spearman's rank correlation: 0.495, P=0.052; Fig. 1b). On the other hand, we found no correlation of parasite diversity with genome-wide variability (partial Spearman's rank correlation= −0.248, n.s.; Fig. 1c).

Figure 1.

Populations of three-spined stickleback ( Gasterosteus aculeatus ): relationships between neutral, genome-wide variability (measured as microsatellite alleles/locus), parasite diversity (measured as Simpson's diversity D ) and immunogenetic diversity (measured as number of distinct alleles present within populations). As we found genetic differentiation between sampling dates at several occasions (Dieksee September–November: FST =0.010*, Kl. Plöner See September–November: FST =0.027***, Vierer See September–March: FST =0.011*, Wedeler Au September–March: FST =0.016** November–March: FST =0.015** with *** P  < 0.001, ** P  < 0.01 and * P  < 0.05) each sampling date of all populations were considered separately in the analysis. However, patterns are consistent when dates are pooled (MHC alleles vs. Simpson's D : ρ =0.745, P =0.009; MHC vs. microsatellite alleles/locus: ρ =0.780, P =0.009; microsatellite alleles/locus vs. Simpon's D : ρ =−0.409, P =n.s.). Panel (a) shows the relationship of MHC diversity to parasite diversity, panel (b) shows the relationship of MHC to microsatellite alleles/locus and panel (c) shows the relationship of parasite diversity to genome-wide variability.

However, particular habitats were always associated with relatively high or low MHC diversity, with river sites having fewer alleles than lake habitats (Fig. 1). At the same time, the river samples showed significantly lower diversity in terms of parasite community and immunogenetics than lake and estuary populations, which were not different from each other. But when comparing lakes with the two other habitat types using microsatellites as evolutionary null-model, there was only a small difference between lakes and rivers (Fig. 2), whereas microsatellite allelic richness was highest in the estuarine population. Above population-level differences were still present when pooling time points to single populations, microsatellite alleles/locus: F1,5=4.636, P=0.084; number of MHC alleles: F1,5=223.908, P < 0.001; parasite diversity: F1,5=8.261, P=0.035.

Figure 2.

Three-spined stickleback ( Gasterosteus aculeatus ): Influence of habitat type (lake vs. river vs. estuary) sampled on parasite diversity (measured as Shannon's information index H ), numbers of MHC class II B alleles and microsatellite alleles/locus (±1 SE). Significant effects of habitat were found for all measurements (least square regressions with sampling date and habitat as factors, number of MHC alleles: F ratio 99.454, P  < 0.001; parasite diversity: F ratio 9.759, P =0.004; microsatellite alleles/locus: F ratio 21.338, P  < 0.001). Sampling date did not show a significant effect in the model ( F ratios ranged from 2.047, P =n.s. for parasite diversity to 2.66, P =n.s. for microsatellite alleles/locus). Pairwise differences were calculated using Tukey's post hoc test and letter codes (a and b) show group membership.

Parasite load and individual MHC diversity

We find support for our prediction of an optimal individual MHC diversity when correlating mean residual parasite load to the number of MHC alleles per fish (Fig. 3). Following this prediction we fitted a quadratic polynomial as the simplest function possessing a minimum and found the lowest parasite load at intermediate 5.17 (4.36–6.48 covering 95% of 1000 bootstrap replicates) alleles/fish. This function clearly explained more of the variation in the data than linear (R2=0.151) or increasing functions (exponential rise: R2=0.109). Furthermore individual genome-wide heterozygosity could not explain individual number of MHC alleles (R2=0.004, F1,253=1.006, P=n.s.) nor parasitation (R2 < 0.001, F1,253=0.352, P=n.s.), thereby excluding general heterozygote advantage. The polynomial regression remains significant when jack-knifing over allele numbers (Fig. 3). To our knowledge this is the first empirical evidence for Nowak et al.′s (1992) theoretical prediction for an optimal number of MHC alleles in individuals.

Figure 3.

Three-spined stickleback ( Gasterosteus aculeatus ) from eight populations in Schleswig-Holstein: relationship of residual parasite load (±1 SE) and number of MHC class II B alleles present in individual fish. Parasite diversity varies markedly between populations (see also Kalbe et al., 2002 ) around a global average of 3.86 parasite species/fish. To reduce variation introduced by effects of population affiliation residuals of an anova with population as factor ( anovaF5,251 =38.06, P  < 0.001) were used for subsequent analysis. The fitted quadratic polynomial has a minimum of 5.17 (95% C.I. 4.36–6.48, 1000 bootstrap replicates over individuals). Jack-knifing over allele-numbers resulted in six significant ( P  < 0.05) quadratic polynomials with minima lying within the range of the bootstrap replicates. The distribution of allele numbers in the data set is shown in the histogram below.

Discussion

Parasite vs. MHC diversity in populations

The extensive polymorphism in the eight studied populations (13–28 different MHC class IIB alleles; Table 1) indicates balancing selection working on the MHC loci. However, genetic drift which depends on parameters such as effective population size (Ne) also seems to have an influence on the amount of variability within the MHC. Accordingly, we found a positive correlation of microsatellite allele richness with MHC variability (Fig. 1b), which is in line with other recent findings (Gutierrez-Espeleta et al., 2001; Hedrick et al., 2001). Nevertheless, if parasites are the selective force behind the maintenance of MHC polymorphism parasite diversity should rather correlate with MHC variability than with genome-wide variability. Our data support this notion as populations being exposed to a wider array of macroparasites were also more diverse in terms of their MHC class IIB genes (Fig. 1a). The relative strength of the partial correlation outweighs that of MHC to microsatellite alleles/locus (see Fig. 1a, b). On the other hand, we find no correlation of parasite diversity with genome-wide variability (Fig. 1c). Accepting that MHC allele frequencies are not solely governed by neutral processes, but are also subject to parasite-induced selection, a low parasite diversity should be equivalent to relaxation of selection (Van Valen, 1965) making genetic variability at resistance genes less important for survival of the population.

Unfortunately, parasite diversity is confounded by two major habitat types sampled in this study, lakes and rivers. The relationship among MHC and parasite diversity is weak within single habitats, whereas the data rather cluster according to habitat (Fig. 1a). In a previous study (Reusch et al., 2001b) we have already identified a phylogeographical signal indicating virtually no gene flow between even physically adjacent populations from different habitats. Similar rapid ecological speciation has been reported from species pairs on Vancouver Island, which exploit different food niches (Schluter, 1996). We cannot rule out other such confounding effects, albeit any ecological speciation depending on trophic niche is probably irrelevant for immunogenetic variation. One possible diverging process operating between habitat types may be different demographic histories or different effective population sizes Ne. Our and previous (Reusch et al., 2001b) microsatellite data, and general expectations from other fish species (DeWoody & Avise, 2000) suggest that Ne of the estuarine population is largest, whereas there is no significant difference between lakes and rivers based on our findings (Fig. 2). Therefore, the effect of genetic drift seems to be comparable in these two habitats. Variability at the MHC loci could be expected to show the same pattern, but instead we find elevated levels of immunogenetic diversity in lake populations relative to river populations. Such a pattern can more easily be explained by selective impact of parasite communities, which show the same pattern between habitats as MHC (Fig. 2). In conclusion, we have found supportive evidence for our initial hypothesis, but more phylogenetically independent populations need to be assessed to formally test the hypothesis and clarify the confounding effects induced by differences in habitat.

Parasite load and individual MHC diversity

If individuals with more MHC alleles are able to present more foreign peptides to T cells, we expected that immune responses could be mounted to resist a wider array of parasites. Additionally, following the theoretical prediction of an optimal number of MHC alleles within an individual (Nowak et al., 1992) we found higher parasite burdens at both extreme ends of the distribution (Fig. 3). We fitted a quadratic polynomial as the simplest function possessing a minimum on pooled data from all fish as the optimality function should be an intrinsic feature of the stickleback MHC. We found the lowest parasite load at 5.17 (4.36–6.48 covering 95% of 1000 bootstrap replicates) alleles/fish. The rise in parasite burden may indicate prior elimination of self-reactive T-cell lines in those individuals. This may only be important in populations with relatively high individual parasitic burden, e.g. lakes. In rivers, on the other hand, we could not find this pattern. Here, fish are frequently challenged with infections from only one parasite species (Kalbe et al., 2002). The chance of mounting a functional immune response against a single pathogen does not depend on high levels of T-cell diversity. Rather effective presentation of the antigen by an MHC allele and recognition by a single T-cell receptor line is necessary. Such an effective modulation can be achieved by a single specific MHC allele. Although the optimal number is an intrinsic property of MHC–T-cell complexes and should be global for sticklebacks, self elimination effects only appear in fish suffering from infestation by a wider array of parasite species. Note that the distribution of parasite load on the number of alleles is right-skewed which may indicate combinatoric interaction between the two antigen-presenting parts of the heterodimer (extracellular domains of α and β chain, Mason, 2001). This might also explain the rare occurrence of fish bearing maximal diversity with nine alleles as these fish will be strongly limited in their functional repertoire of T lymphocytes (Fig. 3).

Recent findings of Reusch et al. (2001a) and Aeschlimann et al. (P.B. Aeschlimann, M. A. Häberli, T. B. H. Reusch, T. Boehm and M. Milinski, unpublished data) suggest that ‘allele counting’, a strategy of sexual selection, is also influencing MHC diversity in offspring. Female sticklebacks with low individual diversity prefer the odour of males with more MHC alleles, whereas females with high individual diversity prefer males with low number of MHC alleles. Females ultimately aim for a total of 10 different alleles combined from their own and from the males. Our findings identify the adaptive value of such a sophisticated choice mechanism. By aiming at 10 different alleles (male + female) the fitness of progeny will be maximized as five alleles on average can be expected in the offspring after recombination, which is strikingly similar to the number of alleles with minimal parasite load from our data (i.e. 5.17).

As fish with an optimal level of diversity suffer least from parasitation, overdominance in this multilocus system can explain the extraordinary polymorphism of stickleback MHC class IIB genes. The effects of multilocus overdominance and the pattern found between habitats only appear when considering many naturally occurring parasite species simultaneously. Of the 434 examined sticklebacks none was parasite-free and only one specimen harboured a single parasite species (Kalbe et al., 2002). On average, 3.86 different parasite species were found per fish. It cannot be expected that sticklebacks suffer from an exceptionally high parasite burden compared with other host species given the sheer amount of parasites among all biota (Windsor, 1998). Many studies investigated only the effect of a single parasite or pathogen (Hill et al., 1991; Thursz et al., 1995; Paterson et al., 1998; Godot et al., 2000; Langefors et al., 2001) and identified associations between infections and certain MHC alleles. Such associations are prerequisite for balancing selection, but can only cause the striking diversification as combined effects from multiple parasite–host interactions.

Conclusions

In this paper we found a consistent relationship between parasite diversity among different habitats and MHC diversity. Although selectively neutral processes contributed to MHC correlated genetic structure, parasite diversity explained significantly more variance.

At the level of individuals, overdominance is the mechanism behind the balancing selection observed. Here, we understand overdominance as general diversity advantage, whereas the classical narrow sense definition is restricted to the synergistic interaction between two alleles of the same locus (Parsons & Bodmer, 1961). In the case of duplicated MHC loci, polymorphism can be maintained by a fitness advantage of individuals possessing different gene products at multiple, functional equivalent loci, even if they are homozygous at the individual gene locus. Furthermore, individuals cannot achieve an optimal MHC class IIB diversity in terms of parasite resistance with a single-copy gene (Fig. 3).

Individual parasite load cannot be reliably assessed by the choosing sex. As a solution for choosiness, Hamilton & Zuk (1982) interpreted sexually selected ornaments as honest signals for immunocompetence. In our data set the relationship between number of MHC alleles and parasite load suggests that immunogenetic variability can be used as a direct, reliable cue. This does not exclude that other sexually selected traits (e.g. nuptial colour, Milinski & Bakker, 1990) are advertising other resistance traits. As not only the number of alleles matters, but also their identity, nuptial colour might be an additional cue for immunocompetence (Milinski & Bakker, 1990). We also suggest that in this coevolutionary situation, the identity of beneficial alleles will vary both spatially and temporarily, precluding a fixation of individual haplotypes for an optimal level of diversity (i.e. five variants of MHC class IIB).

Surprisingly, in the case of MHC correlated sexual selection, individual MHC diversity will not become exaggerated as a handicap or run-away selection process but is instead optimized towards minimal parasite load (P.B. Aeschlimann, M. A. Häberli, T. B. H. Reusch, T. Boehm and M. Milinski, unpublished data). This shows that parasites can be regarded as a potential cause for natural and sexual selection. Both mechanisms of selection would interact to result in the striking MHC polymorphism found in the wild.

Acknowledgments

We would like to acknowledge S. Liedtke, C. Schmuck, G. Augustin and D. Lemcke for technical assistance, T. Boehm, D. Hasselquist, J. Kurtz, Å. Langefors, M. Milinski, R. Willmann and two anonymous referees for helpful comments and encouragement.

Received 10 October 2002;accepted 20 November 2002

Ancillary