Present address: Cardiff School of Biosciences, Cardiff University, PO Box 915, Cardiff CF10 3TL, Wales, UK.
Present address: Institute for Evolution and Biodiversity, University of Münster, Hüfferstrasse 1, 48149 Münster, Germany.
Present address: Konrad–Lorenz Institute for Ethology, Savoyenstrasse 1a, 1160 Wien, Austria.
Manfred Milinski, Department of Evolutionary Ecology, Max-Planck Institute of Limnology, August-Thienemann Strasse 2, 24306 Plön, Germany. Tel.: +49 4522 763254; fax: +49 4522 763310; e-mail: email@example.com
Genes of the major histocompatibility complex (MHC) are indispensable for pathogen defence in vertebrates. With wild-caught three-spined sticklebacks (Gasterosteus aculeatus) we conducted the first study to relate individual reproductive parameters to both MHC class I and II diversities. An optimal MHC class IIB diversity was found for male nest quality. However, male breeding colouration was most intense at a maximal MHC class I diversity. One MHC class I allele was associated with male redness. Similarly, one MHC class IIB allele was associated with continuous rather than early female reproduction, possibly extending the reproductive period. Both alleles occurred more frequently with increasing individual allele diversity. We suggest that if an allele is currently not part of the optimum, it had not been propagated by choosy females. The parasite against which this allele provides resistance is therefore unlikely to have been predominant the previous year – a step to negative frequency-dependent selection.
Teleost fish are ideal for the study of MHC class I and II diversities, as, unlike other vertebrates, the MHC class I and class II regions are not linked, therefore allowing the effects of each class to be tested independently (Sato et al., 2000; Stet et al., 2003). Three-spined sticklebacks (Gasterosteus aculeatus, Linnaeus, 1758) have probably up to 12 MHC class I and six MHC class IIB loci (Sato et al., 1998), that are extremely similar to one another due to interlocus gene conversion and/or recent duplication (Reusch et al., 2004; Reusch & Langefors, 2005; Schaschl & Wegner, 2007). The different number of allelic variants can be achieved through heterozygosity, or possibly, through different haplotype copy number of MHC genes. In the stickleback, evidence for the existence of an optimal (about six alleles) rather than a maximal (12 alleles) individual MHC class IIB diversity is mounting: the mean individual diversity in a natural population was found to be 5.8 alleles (Reusch et al., 2001) and wild-caught sticklebacks carried a minimal parasite burden when they had 5.2 MHC class IIB alleles (Wegner et al., 2003b). Moreover, individuals with an intermediate number of MHC class IIB alleles were infected least after experimental exposure to a defined parasite dosage (Wegner et al., 2003a; Kurtz et al., 2004). In this study, we investigate both MHC class I and II genes. Previous MHC-related studies in the stickleback have been restricted to MHC class IIB.
In the present study, we related individual diversity of both MHC class IIB and MHC class I to relevant reproductive traits of the three-spined stickleback, namely male nest quality, the intensity of male breeding colouration and the number of egg clutches spawned by females. MHC-based pathogen resistance might be particularly important under poor food conditions as trade-offs are often weaker or sometimes not even visible at all under good environmental conditions (Stearns, 1992). If food is limited, a potentially costly suboptimal immune response might not be compensated for by a higher food intake. This is the first study to investigate MHC class I with respect to reproductive parameters of the three-spined stickleback.
Material and methods
Three-spined sticklebacks were caught in the Große Plöner See (Northern Germany) in December 2003. In the laboratory, they were kept first under spring temperature and light conditions and then under summer conditions (18 °C, 16 : 8 h light : dark cycle) until they achieved reproductive condition. Each fish's length (to the nearest mm) and weight (to the nearest mg) were measured 6–7 weeks after being caught (when the fish were transferred from spring to summer conditions). A body condition factor was calculated as 100W/Lb where W is the fish weight in g, L the fish length in cm and b the regression coefficient calculated from the logarithm-transformed values of length and weight (Frischknecht, 1993). In the following, this body condition index is referred to as initial body condition. Under summer conditions, fish were fed ad libitum with frozen chironomid larvae and additional live food every day (well-fed group) or only frozen chironomid larvae once every second day (poorly fed group) without knowledge of the fishes’ MHC identity. Individual fish were randomly assigned to the two food groups.
MHC class I and class II genotyping
Genomic DNA was extracted from part of a single dorsal spine from each individual fish using a DNA extraction kit (Invitek, Berlin, Germany). The MHC class IIB exon 2 (β1 domain) diversity was determined by using capillary electrophoresis (CE) single-strand conformation polymorphism (SSCP) as described in Binz et al. (2001), with modifications as described in Reusch et al. (2001). MHC class II loci are extremely similar to one another (< 2% nucleotide divergence, Reusch et al., 2004) due to interlocus gene conversion and/or recent duplication (Reusch & Langefors, 2005). Recently, CE-SSCP genotyping has been extended to determine individual MHC class I exon 2 (α1 domain) diversity in the three-spined stickleback (Schaschl & Wegner, 2006). Like for MHC class II, MHC class I genes are generated probably by repeated and recent gene duplication (Schaschl & Wegner, 2007). Thus, it was not possible for either of the two classes to devise a PCR strategy that is able to amplify a single locus only.
Nest quality, male breeding colouration and number of egg clutches
Nests were classified according to five categories: 0 = no nest; 0.5 = nest messy, no visible entrance; 1 = nest messy, entrance vaguely visible; 1.5 = nest neatly glued, but entrance only vaguely visible; 2 = nest neatly glued and entrance clearly visible. The rationale behind this classification was to assess whether and how well the nest was maintained, i.e. if the filaments of the nest were glued regularly and if the entrance was maintained by the male performing the ‘creeping-through’ behaviour (Wootton, 1976). Male sticklebacks use a secretion produced in their kidneys to glue their nest and thus keep it in shape (Wooton, 1976). The only vaguely visible entrance represents an opening that is not regularly used and glued. All nests were assessed by the same person without knowledge about the males’ MHC genotype.
The red breeding colouration of males was analysed 1 week after the assessing period. Males were photographed ventrally (without knowledge of the males’ MHC genotype) after having been stimulated with a ripe female for 15 min the previous day. The males were held in place using a sponge in a water-filled plastic box which had a window of filter glass (high-resolution skylight filter; Hama, Monheim, Germany). The pictures were taken within a dark box using a digital camera (Olympus E20-p; Olympus Corporation, Tokyo, Japan) with a 36-mm macrolens. Shutter speed was 1/60 s with aperture 7. For illumination, we used four cold lights (KL 1500 LCD; Leica; Solms, Germany) with 3300 K colour temperature. An intensity analysis of the red colouration was performed with IP Lab 3.6.2 for Mac OS 9.2.2 (Scanalytics, Inc., Rockville, MD, USA). On each picture, we selected a defined area of the throat covering the majority of the red parts. For this area, a mean red intensity was determined using the R/RGB model [8-bit red–green–blue colour model, i.e. values 0 (lowest intensity)–255 (highest intensity) for each of the colour channels], i.e. each pixel's red colour intensity was divided by the pixel's total colour intensity to be independent of brightness (Frischknecht, 1993). All pictures were measured twice in random order, on two consecutive days by the same person. The marking of the area and the resulting red values were highly repeatable: 95.9% for the marked area as well as 99.98% for the mean colouration.
During a 25-day period, starting 4 weeks after fish had been placed under summer conditions, each female's gravidity was assessed (without knowledge of the females’ MHC genotype) every day except days 15, 20 and 24. A female was assumed to have spawned overnight if she had been assessed as being very ripe on one day but not on the following day. It is known that female sticklebacks spawn their eggs when gravid even in the absence of males.
Males were dissected and screened for parasites without knowledge of the males’ MHC genotype. As not all dissections could be performed on the same day, only parasite species that could not be eliminated by the host were counted. These included the three digenean trematodes Diplostomum sp. (Nordmann, 1832), Cyathocotyle prussica (Mühling, 1896) and Echinochasmus sp. (Dietz 1909) and the nematode Contracaecum sp. (Railliet & Henry, 1915). Individuals of the respective parasite species were counted and an overall parasite load was calculated for each fish using the following approach (Rauch et al., 2006). For each parasite species, a relative number of individuals per fish was determined (number of individuals per fish divided by the maximum number of individuals of that respective species found on a single fish). The relative numbers of the four parasite species were summed up and averaged over the four species. This index allows parasite burden to be quantified and thus, comparisons to be made between individuals.
All statistical analyses were conducted in R statistical package version 2.3.0. Post hoc tests are reported directly in the respective results section. To test for significant correlations between the fishes’ (males and females) initial body condition and their individual MHC class I and MHC class IIB diversities, we conducted two independent regressions incorporating a linear as well as a quadratic term in each analysis (function LM).
First, a multiple correlation was performed to remove potential co-linearities in the models. As initial body condition was correlated with food group, this variable is expressed as the residuals from this regression in all following analyses.
Males were tested for significant effects on nest quality with an ordinal logistic regression incorporating food group, number of MHC class I alleles, number of MHC class IIB alleles, parasite load and initial body condition [function polr, Ananth & Kleinbaum, 1997 (choice of model); Fox & Andersen, 2004 (methods and results report)]. We also included all possible two-way interactions with food group in the model. We then performed independent ordinal logistic regressions to test for an MHC class I and class IIβ optimum.
An analysis of covariance (ancova) was performed to test the effect of the above-mentioned factors on male breeding colouration. Independently, the relationship between the number of MHC variants (class I and IIB) and redness was also considered as quadratic function. To test whether the difference in colour intensity between the reddest and the palest quarters of the test fish is linked to a difference in their MHC class I allele composition, we performed an analysis of similarity (anosim, Clarke & Gorley, 2006). We based the similarity matrix on the Bray–Curtis index (Bray & Curtis, 1957, for the choice of the index see Clarke et al., 2006). Afterwards, a SIMPER test was conducted to estimate whether specific MHC variants contributed to the difference between the two groups. Finally, an anova was performed to test for significant effects of the identified MHC variants on male redness.
Number of egg clutches
An ancova was conducted to analyse the influence of food group, individual allele diversity of MHC class I and class IIB as well as all possible interactions with food group on the number of egg clutches a female spawned. We also tested for quadratic correlations between individual diversity of MHC class I and IIB and the number of egg clutches spawned. To see whether the possession of a specific MHC class IIB allele variant accounted for differences in spawnings at the end of the reproductive period, we conducted an anova. For an analysis of the course of reproductive investment across the 25-day breeding period, repeated measurement manovas were performed with time as the repeated factor and the possession of the identified allele and food group as between-subject variables. To obtain the respective significance levels, repeated measurement manovas were conducted on the respective subsamples with time as the repeated factor and the possession of the identified allele as between-subject variable or just with time as the repeated factor.
The data set consisted of 191 fish (95 males and 96 females). Not all parameters considered were available for all fish resulting in variable sample sizes. For MHC class IIB a mean individual allele diversity of 5.6 ± 0.1 SE was calculated with a median number at six alleles. The mean individual allele diversity for MHC class I was 7.1 ± 0.1 SE with seven being the most frequent individual allele number.
In both, males and females, initial body condition was significantly correlated with a quadratic function of individual MHC class IIB diversity (Regression, F1,173 = 4.131, P = 0.044, y = 1.31 + 0.11NMHC II − 0.009NMHC II2) with a maximum calculated for individual diversity of 6.23 alleles.
Nest quality was significantly associated with a quadratic function of individual MHC class IIB diversity [Ordinal logistic regression, N = 72, value = −0.269, SE = 0.119, = 5.066, P = 0.0244, Table 1(a1)]. Most individuals maintaining a high-quality nest carried an intermediate number of MHC class IIB alleles, whereas most individuals with a low or high allelic diversity had nests of lower quality (Fig. 1). Males possessing six alleles had the highest proportion of nests of the highest quality, depicting an optimum at six alleles (Fig. 1).
Table 1. Significant variables correlated with (a) nest quality, (b) redness, (c) parasite load, (d) number of egg clutches.
Prob > χ2
Models were chosen on the basis of AIC criteria. Significant at *P = 0.05, **P = 0.01, ***P = 0.001.
†Residuals from the correlation between parasite load and food group.
‡Residuals from the correlation between initial body condition and food group.
There was a significant interaction effect between food level and parasite load on nest quality [ordinal logistic regression, N = 72, value = −0.516, SE = 0.253, = 4.143, P = 0.042, Table 1(a1)]. This resulted in decreasing nest quality with increasing parasite load when fish were under restricted food conditions, whereas parasite load did not have a detectable impact when fish were under good food conditions.
Male breeding colouration was not significantly correlated with the quadratic term of either MHC class I or MHC class IIB diversity (regression, F1,70 = 0.749, P = 0.390, and F1,70 = 0.030, P = 0.862). However, males were more intensely coloured with an increasing number of their MHC class I alleles [ancova, F1,68 = 8.528, P = 0.005, Fig. 2, Table 1(b)]. This can be expected if specific alleles that are not part of the optimum allow for high redness. To search for such alleles we compared the reddest quarter of males with the least coloured quarter and found that these two groups differed significantly in their MHC class I allele composition (anosim, N = 40, global R = 0.073, P = 0.033). One specific MHC class I allele contributed most to this difference (SIMPER test, 6.85%). Therefore, we tested whether the presence of this allele was related to a more intense colouration performing an anova on redness as a function of this allele. Males were more brightly coloured when they carried the specific allele (F1,44 = 4.5166, P = 0.039). To test whether the likelihood of possessing this allele indeed increased with the number of alleles per individual, we performed a binomial logistic regression on the whole population for each of this allele. The presence of this MHC class I variant increased with an increasing individual MHC class I diversity [GLM (binomial, logit link), N = 176, = 32.963, P < 0.0001].
Number of egg clutches
The quadratic terms of neither individual MHC class I nor MHC class IIB diversity correlated significantly with the number of egg clutches a female spawned (ancova, F1,72 = 0.085, P = 0.771 and F1,72 = 0.319, P = 0.574 respectively). However, females under good food conditions spawned significantly more egg clutches than poorly fed females [ancova, F1,69 = 44.031, P < 0.0001, Figs 3 and 4, Table 1(c)]. It seems that it is not current food availability alone that determines the number of egg clutches a female spawns. We found a significant interaction between food level and individual MHC class IIB diversity on the number of egg clutches (ancova, F1,69 = 9.921, P = 0.002, Fig. 3). Under good food conditions females spawned fewer clutches the more MHC class IIB alleles they carried, whereas the number of clutches spawned by poorly fed females was independent of allelic diversity. We tested whether this result can be explained by a conditional strategy to spawn either early or continuously. We divided the observation time into five subperiods consisting of 5 days each to compare the MHC class IIB composition of those females that spawned in the last subperiod (N = 20) with that of females that did not (N = 19). The analysis of similarity did not show any significant differences in the MHC class IIB allele composition of these two groups (anosim, R global = −0.001, P = 0.474, average dissimilarity = 54.45%). However, the SIMPER test showed that there was one allele that accounted for 8.61% of the allelic difference between the two groups. The likelihood of possessing this allele increased with increasing allele numbers per individual [(GLM (binomial, logit link), = 21.293, P < 0.0001].
For the entire reproductive period, the two food groups differed significantly in the reproductive investment per subperiod of individuals with and without that specific allele (repeated measurements manova, time × food group × allele: N = 75, F4,68 = 2.894, P = 0.028, Fig. 4). Whereas there was an overall effect of time under low food conditions showing that the number of spawnings decreased rapidly (repeated measurements manova, N = 36, F4,31 = 21.912, P < 0.0001, Fig. 4), we found no difference between individuals carrying the specific allele and those not carrying it (repeated measurements manova, time × allele: N = 36, F4,31 = 0.785, P = 0.544, Fig. 4). However, individuals of the well-fed group differed significantly in their reproductive investment across the breeding period depending on the presence of the specific allele (repeated measurements manova, time × allele: N = 39, F4,34 = 4.057, P = 0.009, Fig. 4). Individuals not carrying the allele reproduced early in the reproductive period followed by a decrease of spawning rate (repeated measurements manova, N = 25, F4,21 = 11.741, P < 0.0001, Fig. 3b), whereas females that possessed the allele spawned continuously throughout the whole observation period (repeated measurements manova, N = 14, F4,10 = 0.625, P = 0.655, Fig. 4).
The observed mean individual MHC class IIB allele diversity of 5.6 is close to both the mean allelic diversity found in a natural population (Reusch et al., 2001) and the optimal allele number found both here for body condition, i.e. 6.2, and in other studies (Aeschlimann et al., 2003; Milinski, 2003; Wegner et al., 2003a,b; Kurtz et al., 2004). Also in this study, we found that individuals owning an optimal MHC class IIB number of about six alleles were in better condition than individuals carrying a nonoptimal allele number. The hypothesis of the two opposing selective forces acting on the number of MHC alleles per individual – increased pathogen recognition versus increased autoreactivity and thus increased deselection of T-cell clones (Nowak et al., 1992; De Boer & Perelson, 1993; Borghans et al., 2003; review in Milinski, 2006) –should also apply to MHC class I. We predicted an optimal individual allele diversity also for MHC I. We would expect individuals carrying the median number of alleles to be the most frequent type. Indeed we found that the median was the most frequent type and had seven alleles.
Nest quality was influenced by an interaction between parasite load and food group. This means that poorly fed fish had nests of lower quality with increasing parasite load when compared with well-fed fish. Maintaining a high-quality nest involves regular glueing, which is likely to require a higher glue production. Glue is produced in the males’ kidneys, which were larger in males with high-quality nests (Barber et al., 2001b) and in well-fed males (Wootton, 1994). The pronounced difference that we found between the two food groups clearly emphasizes the importance of diet on individual performance and suggests that feeding regime should be carefully considered when conducting laboratory experiments (see also Barber et al., 2000; Candolin & Voigt, 2001). The effect of the interaction of parasite load and food level on nest quality fits our prediction from life-history theory (Stearns, 1992). Male sticklebacks might be able to compensate for parasitization under good food conditions, whereas this might not be the case when energy availability is low. Hence, as a trait used by females to evaluate males, nest quality seems to reflect condition in a complex way.
We did not find that males with an optimal MHC diversity had the brightest breeding colouration. Instead, we found a positive correlation between a male's redness and its MHC class I allele number. However, this is not necessarily in contradiction with optimal allele diversity if there is a specific allele whose possession outweighs the disadvantage of being above the optimal allele number by providing resistance to a new predominant pathogen. Indeed, we identified one MHC class I allele that was associated with a more intense breeding colouration. Various studies have found specific MHC alleles to be associated with resistance against certain pathogens (e.g. Langefors et al., 2001; Grimholt et al., 2003; Bonneaud et al., 2005; reviewed in Milinski, 2006) and in the stickleback, four MHC class IIB alleles were found that were each significantly related to the resistance against one specific macroparasite species (Wegner, 2004).
One reason for males to be redder when carrying an advantageous MHC class I allele might be that they can afford to reallocate the available antioxidants to the carotenoid-based breeding ornamentation (Brush & Reisman, 1965) instead of using them as radical scavengers. Male sticklebacks showing a higher MHC-based resistance could have a lower demand for innate immune activity, thus be less prone to oxidative stress, and therefore not so much in need of radical scavengers (von Schantz et al., 1999; Kurtz et al., 2004, 2006). After exposure to parasites, three-spined sticklebacks did reduce their carotenoid-based red colouration (Milinski & Bakker, 1990) and individuals with an optimal number of MHC class IIB alleles showed the lowest respiratory burst reaction, an innate immune mechanism that generates reactive oxygen molecules to kill pathogens (Kurtz et al., 2004). Moreover, sticklebacks with a nonoptimal MHC class IIB diversity have higher levels of MHC gene expression (Wegner et al., 2006), which correlates with increased oxidative stress (Kurtz et al., 2006).
It has previously been shown that the male stickleback's red throat is an honest signal revealing a male's health (Milinski & Bakker, 1990) and resistance to parasite infection (Milinski & Bakker, 1991). Our findings support the good-genes hypothesis of Hamilton & Zuk (1982), suggesting that a male's condition might be shaped by the genetic value of its immune system, revealed by the intensity of the red ornamentation. Females choosing the reddest male would then gain fitness benefits for their offspring in two ways: through high-quality parental care offered by a healthy mating partner and also by ensuring that the offspring are endowed with a good immune system. This concurs with the finding that offspring of brighter stickleback males are more resistant to infection by the cestode Schistocephalus solidus (Barber et al., 2001a). Furthermore, when redness can be linked to resistance against presumably common pathogens, mate choice based on odour that reveals a male's MHC alleles may be improved by visual mate choice. As suggested by Aeschlimann et al. (2003) and Milinski (2003, 2006), this might guarantee that not only the optimal MHC allele number but also the relevant specific MHC alleles are included in the optimum for the offspring. Consequently, males with an intermediate number of MHC class I alleles should also be more likely to possess the specific allele. However, this was not the case in our study. Possibly, the pathogen against which this allele presumably provides resistance, was not one of the predominant parasite species in the stickleback population the year prior to our study, so that this allele could not be detected by a redder colouration and thus not be subjected to female choice. From this hypothesis, we would predict that such alleles would be found included in the optimum in the following year.
Number of egg clutches
Wootton (1994) and Fletcher & Wootton (1995) concluded that the number of clutches spawned per breeding season is dependent on the current food supply, whereas clutch and egg size are a function of female size at the start of the breeding season, thereby being only little affected by food availability. This would mean that food shortage occurring later would reduce the number of spawnings.
We found not only that females under good food conditions spawned significantly more egg clutches than poorly fed females, but also a significant interaction between food level and individual MHC class IIB diversity on the number of egg clutches. We found a conditional strategy to spawn either early or continuously under good food conditions depended on the possession of a specific MHC class IIB allele. Females that possessed the allele, probably providing resistance against a current parasite, spawned continuously throughout the whole observation period and would probably have continued beyond this time. The likelihood of possessing this allele increased with increasing allele numbers per individual. Hence, as suggested for the allele that correlated with intense breeding colouration in males, the parasite against which this allele presumably provides resistance, had not been predominant in the stickleback population in the previous year and was thus not included in the optimum.
As evidence from the field suggests (Wegner, 2004), MHC alleles are under negative frequency-dependent selection by parasites, and we expect that parasites abundant in one year will not be predominant in the following year. In conclusion, we predict the following scenario for negative frequency-dependent selection of MHC class I and IIB alleles by pathogens in sticklebacks. Females try to select a male that: (1) offers through olfactory cues the number of different alleles that, in combination with the females’ own alleles, achieve the immunogenetic optimum for the offspring; and (2) displays costly sexual signals which increase the probability that the complement from this male contains alleles that provide resistance against currently predominating parasites. We assume that the female assembles the optimal number of six alleles, which in the best case are composed of the alleles providing highest resistance against the currently predominating parasites. But how are advantageous MHC alleles of females selected for the offspring's optimum if the optimum is achieved by female choice? This study suggests that females with such alleles can achieve a higher reproductive output. Thus the ‘good genes’ of resistance of such successful females are probably ending up in the optimum that these females choose for their offspring. Thanks to the higher reproductive success of their mothers; these offspring will become numerous. Therefore, the previously dominating pathogens have only low prevalence in the following period of time. Now, other pathogens are predominating against which, e.g. any of the previous ‘top six’ MHC class IIB alleles no longer provide best resistance. Hence, other MHC alleles probably not contained in the previous optimum are providing highest resistance. The more alleles a fish carries, the more likely is this fish to carry the best resistance alleles just by chance, as suggested by this study. These alleles are then predicted to be part of the optimum of the next generation. This scenario awaits testing in a natural population during several generations. It opens the door to study the role of negative frequency-dependent selection in maintaining MHC polymorphism, and thus mate choice and sexual reproduction.
The authors would like to thank two reviewers for thoughtful suggestions, G. Augustin, A. Hasselmeyer and S. Liedtke for technical assistance.