Introduction

The ecological processes that shape genetic divergence among and within species are a key area of research in biology. The Trinidadian guppy (Poecilia reticulata) remains an extensively studied system because of the rapid divergent evolution among populations. For example, guppy populations differ predictably in morphological and behavioural traits in response to selection from different predation regimes (Houde 1997). Additionally, there is a high degree of genetic divergence at neutral loci among populations from different drainages (Carvahlo et al. 1991) and even populations within rivers (Suk and Neff 2009). Paradoxically, despite the adaptive and neutral divergence among guppy populations, strong reproductive isolation or other signs of incipient speciation among guppy populations appear to be absent (reviewed in Magurran 1998).

Recently, it has been shown that the MHC class IIB, an important gene in the adaptive immune system, is not as diverged across guppy populations in Northern Trinidad as microsatellite loci (van Oosterhout et al. 2006; Fraser et al. 2009). Here we hypothesize that pathogens are acting as a homogenizing selective agent at this important fitness locus and preventing genetic divergence across populations. Gyrodactylus turnbulli and G. bullatarudis (hereafter referred to as gyrodactylus) are major pathogens of wild guppies and they are found in many populations across different ecological factors and drainages (Lyles 1990; Martin and Johnsen 2007). Thus, gyrodactylus are an excellent candidate as an agent of homogenizing selection in the guppy. Gyrodactylus are small (ca. 0.5 mm long) monogenean fluke parasites that attach to the integument of their host and feed on its protective mucus and skin. The exponential population growth of gyrodactylus on their host, due to their short generation time and direct life cycle, can lead to a damaging or even lethal infection in guppies (e.g. Scott and Anderson 1984). Indeed, van Oosterhout et al. (2007a), using a mark recapture experiment, have provided direct evidence of gyrodactylus mediated selection.

Both innate and adaptive resistance to gyrodactylus has been shown to be heritable in the guppy (Scott 1985; Cable and van Oosterhout 2007a), but a clear link to MHC variation has yet to be established. In one study, van Oosterhout et al. (2006) surveyed the upper and lower Aripo populations and compared variation of MHC and parasite intensity at the population level. They showed that the observed allele frequencies at the MHC could be maintained by overdominance selection, suggesting that gyrodactylus could be an agent of selection on the MHC; although, recently van Oosterhout (2009) has also argued that the observed diversity could be maintained by selection at closely linked loci. Therefore, it remains to be determined what effect the MHC has on gyrodactylus resistance.

Here, we examine the relationship between MHC variation and parasite resistance in guppies from 10 populations from five rivers and three drainages. We examine MHC allele and genotype effects on parasite infection from gyrodactylus and use these data to examine the potential causative factors of selection on the MHC in the wild.

Materials and methods

During May of 2006, we collected an average of 41.1 ± 2.6 (SD) guppies from 10 populations across Northern Trinidad (N = 412 total). The populations comprised upstream and downstream locales from the Aripo and Guanapo rivers in the Caroni drainage, the Quare and Turure rivers in the Oropouche drainage, and the Yarra river on the northern coast. Fish were individually packaged to avoid cross-contamination and transported back to our lab at the University of the West Indies, St. Augustine where they were housed individually before examination. All fish were analyzed within 24 h of being collected.

First, fish were lightly anesthetized with MS222 and an external examination was conducted and standard length and sex were recorded. The fish were then euthanized with an overdose of MS222 and a skin smear was taken, two gill arches were excised at random, followed by a whole body dissection as outlined in Roberts and Smail (2001). The number of Gyrodactylus turnbulli and G. bullatarudis flukes, without identifying species, on each individual was recorded. The two species have similar pathologies in guppies and infection with either species results in similar parasite load and parasite-related mortality (Cable and van Oostehout 2007b). No flukes were observed on individuals from the upper Guanapo, upper Quare and upper Turure sites and we therefore excluded these populations from analysis of parasite and genotype correlations. No other parasites, besides gyrodactylus, were found in any individual. Other parasites, particularly digeneans, have been reported to infect guppies in these populations but are also known to vary spatially and temporally and may account for the lack of other parasites found in this study (Cable and van Oosterhout 2007a).

Whole body tissue samples were preserved in 95% ethanol for genetic analysis. From our sample, a subset of fish was chosen randomly for genetic analysis. Briefly, an average of 14 fish from each population (range: 11–17) were genotyped at the MHC class IIB gene using a combination of single strand conformation polymorphism (SSCP), cloning and sequencing (see Fraser et al. 2009 for full details). The MHC class IIB reading frame was determined by aligning our sequences with the reading frame identified for the guppy in van Oosterhout et al. (2006) and three species of African cichlids (Haplochromis lividus, Oreochromis alcalicus, and Aulonocara hansbaenschi) (Figueroa et al. 2000; Ono et al. 1993). The amplified 230 bp were subsequently trimmed to 218 bp and 72 codons that constituted an open reading frame. The MHC class IIB gene is duplicated in the guppy and therefore individuals can have up to four different alleles (McConnell et al. 1998; van Oosterhout et al. 2006).

To determine whether or not individual genes or genotypes were related to gyrodactylus infection (number of flukes) we used an algorithm adapted from Pitcher and Neff (2006). First, alleles that differed by three amino acids or fewer and had unique peptide binding residues were grouped together into an allele “type” and we focused on the five most common allele types; all other types were grouped into a “rare” category. The total number of flukes was transformed using log10 +1. The algorithm relates infection to both additive and non-additive genetic effects by using the general linear model:

$$ {\text{Log}}_{10} \left( {{\text{number}}\;{\text{of}}\;{\text{flukes}} + 1} \right) = \mu + S\left( {{\text{a}}_{\text{i}} \times {\text{A}}_{\text{i}} } \right) + {\text{S}}\left( {{\text{n}}_{\text{ij}} \times {\text{N}}_{\text{ij}} } \right) $$

where μ is the mean number of flukes infecting individuals, ai is the additive coefficient for allele i, Ai is the number of copies of allele i, nij is the non-additive coefficient associated with alleles i and j, and Nij was coded as either 1 if individuals had at least one copy of both alleles i and j or 0 otherwise. Not all combinations of alleles were observed and therefore some non-additive effects were not assessed. We also did not consider any three-way non-additive effects. First, significance of the general linear model was determined using a bootstrap routine where we randomized parasite load with respect to genotype data and then determined the proportion of runs that the R2 was higher than the R2 of the non-randomized (observed) data. Significance of each coefficient was assessed using a bootstrap routine that was repeated for a total of 1,000 samples. Each bootstrap resampled the data with replacement and re-calculated the coefficients; P-values were then determined as the proportion of bootstrap runs in which a given coefficient was either above or below zero. We also considered a model that included a population effect to control for any potential spurious association between the frequency of the a-type alleles and intensity of the infection across populations.

Results

Gyrodactylus infection was found in seven of the 10 populations surveyed (Fig. 1). Across the seven populations that exhibited infection, we found that 25% of fish were infected with the parasite (range: 10–38%, N = 295). Infected individuals had an average of 1.96 ± 0.17 (SD; range: 1–8) flukes. Standard length of the fish was not related to the number of flukes present (Spearman’s ρ = 0.077, N = 412, P = 0.12). Females had more flukes (mean = 0.44 ± 0.07) than males (mean = 0.26 ± 0.07), but this difference was not significant (t410 = 1.87, P = 0.062). Furthermore, there was no difference in the number of males and females analyzed across populations (χ2 = 0.24 n = 412 P < 0.63).

Fig. 1
figure 1

Location and MHC class IIB allele frequencies of 10 guppy (Poecilia reticulata) sampling sites in Northern Trinidad. Populations comprise Lower Yarra (LY), Upper Yarra (UY), Lower Quare (LQ), Upper Quare (UQ), Lower Turure (LT), Upper Turure (UT), Lower Guanapo (LG), Upper Guanapo (UG), Lower Aripo (LA), and Upper Aripo (UA). Pie charts display the frequencies of the five most common allele types and the remaining rare alleles category for each population. Bar graphs show the proportion of individuals that were parasitized with Gyrodactylus turnbulli and G. bullatarudis flukes, black indicating proportion of individuals parasitized and white indicating the proportion that were not parasitized

Full size image

We found a total of 39 different amino acid sequences (see Fraser et al. 2009), which were binned into the five most common allele types and one rare type category. Across all populations, the a-type allele was the most common with a mean frequency of 69% (range: 33–100%), whereas, the other types of alleles were less common (range: 0–15%; Fig. 1). The general linear model significantly explained the variation in parasite load (Table 1). The a-type allele had a significant additive effect on the number of flukes (Table 1); individuals with a copy of an a-type allele had fewer flukes (0.51 ± 0.15) than individuals without an a-type allele (1.00 ± 0.33). There was no other significant additive or non-additive effect of any other allele type (Table 1). We also confirmed this additive effect of the a-type allele on the incidence of infection (infected = 1 non-infected = 0) using a logistical ordinal regression. We found that the a-type had a significant negative effect on incidence of infection (χ2 = 4.12, df = 1, 98, P = 0.042). Inclusion of the population effect in the general linear model was not significant (P = 0.59 for comparison of the R2 to the model without the population term) and did not change the significant negative effect of the a-type allele (P = 0.036 with the inclusion of the population effect). Across populations, there was also a negative relationship between the frequency of the a-type allele and the frequency of parasite infection, although the relationship was not significant (Spearman’s ρ = −0.30, n = 10, P = 0.40).

Table 1 Additive and non-additive genetic effects of five MHC class IIB allele types and a rare alleles category on the number of Gyrodactylus turnbulli and G. bullatarudis flukes found on guppies (Poecilia reticulata)
Full size table

Discussion

Our results show an advantage of the most common MHC class IIB allele type found in the guppy populations from Northern Trinidad. The number of a-type alleles was associated with a decrease in the number of gyrodactylus flukes infecting individual guppies, which indicates an additive effect of the allele type on resistance to gyrodactylus. We found no additive effect of any other type of allele and no non-additive effects of any pair of alleles. Gyrodactylus is a pervasive pathogen and is known to cause serious harm and to be lethal to guppies (e.g. Scott and Anderson 1984; Cable and van Oosterhout 2007a, b; van Oosterhout et al. 2007a, b, 2008). Therefore the a-type allele should confer a strong fitness advantage, which most likely accounts for its high frequency in populations. The hypothesis that the a-type is being maintained by homogenizing selection is further supported by lower pairwise population genetic differentiation at the MHC than at neutral microsatellite loci driven mainly be the high frequency of the a-type alleles (Fraser et al. 2009). Our results do not, however, support those of van Oosterhout et al. (2006), who argued that homogenizing selection in the upper and lower Aripo populations was due to balancing selection resulting from a heterozygote advantage. We found no evidence for non-additive effects on parasite load at least not across the seven populations that had gyrodactylus. Similarly, van Oosterhout (2009), using selection coefficient of gyrodactylid-induced mortality rate estimated from van Oosterhout et al. (2007a), concluded that parasite selection was high but insufficient to explain the level of inferred overdominance selection on the MHC. We conclude that directional selection exerted by gyrodactylus in each population is responsible for the high frequency of the a-type alleles and homogenizing allele frequencies at the MHC in these otherwise genetically disparate guppy populations.

Although the results presented here are correlational, the hypothesis that the a-type allele confers resistance to gyrodactylus is further supported by experimental results from infectivity trials (Fraser and Neff in review). We experimentally infected guppies from a stock population by transferring two flukes and then monitored the spread of the parasite infection on individual fish over the course of 10 days. We found that an a-type allele had a negative additive effect on parasite load. Taken together, these data provide strong evidence that the a-type allele provides resistance to gyrodactylus infection in the guppy.

The established view in fishes has been that response to gyrodactylus infection is provided by the innate immune system. However, evidence is mounting that the adaptive immune system is also important in gyrodactylus resistance. While the innate immune system reacts to infections in a non-specific manner, the adaptive immune system recognizes foreign antigens, through the MHC, and reacts to infections with specificity and memory (Klein 1990). Kania et al. (2007) showed that after infection with Gyrodactylus salaris, Baltic Atlantic salmon (Salmo salar) showed increased expression in MHC class I genes (also see Lindenstrøm et al. 2004). In guppies, Scott (1985) found that fish previously infected with gyrodactylus were better able to combat infection, even if the previous infection had been 6 weeks prior (also see Cable and van Oosterhout 2007a). Similarly other studies have found resistance conferred by the MHC to other monogenean parasites (Ottová et al. 2007), parasites in the phylum platyhelminths (Wegner et al. 2003), and other parasitic worms known as helminths (Schad et al. 2005). Therefore it is likely that the MHC can recognize antigens from gyrodactylus and mount a reaction using the adaptive immune system. Our results lead us to propose that specific amino acid variants of the MHC contribute to the recognition of gyrodactylus infection and link it to the adaptive immune system and subsequent resistance and clearance of the infection.

A recent model by Eizaguirre et al. (2009) shows how the pleiotropic effect of the MHC in local adaptation and mate choice contexts can accelerate and stabilize speciation among populations. The premise of the authors’ model is that local selection on the MHC by the pathogen community differs among populations and mate choice for optimal MHC genotypes then leads to reproductive barriers across the populations. We have instead found that despite the high level of population divergence at neutral loci, the guppy populations studied here are more similar at the MHC than expected by neutral evolution (Fraser et al. 2009; also see Miller et al. 2001). Therefore, it is also possible that natural selection at the MHC and subsequent mate choice for MHC genotype can actually counteract speciation. Indeed, we have provided some of the first data implicating the widespread gyrodactylus pathogen as an agent of homogenizing selection across otherwise genetically disparate populations. Future work implicating adaptive mate choice for the MHC in the guppy could provide an important explanation for the apparent lack of reproductive barriers among the populations in Northern Trinidad.