Adaptation can occur over a very short time scale (Thompson 1998), suggesting that wild populations could potentially adapt to the increasing number of novel pathogens emerging in the environment. In vertebrates, genes of the major histocompatibility complex (MHC) are involved in pathogen resistance by encoding cell-surface proteins that bind peptide fragments derived from pathogens and present them to T cells that activate a specific immune response (Potts & Wakeland 1990). These genes are among the most polymorphic in vertebrates and most of the variability is concentrated at the peptide-binding region (PBR) involved in pathogen binding (Hughes & Yeager 1998). Theory predicts that, at the population level, high MHC diversity should confer resistance to a diverse array of pathogens (reviewed in Bernatchez & Landry 2003; Sommer 2005; Piertney & Oliver 2006). Recent studies conducted at a large spatial scale supported a positive association between MHC diversity and pathogen diversity in the wild, independent of historical colonization processes, in humans (Prugnolle et al. 2005) and Atlantic salmon (Dionne et al. 2007). Yet, the exact underlying mechanism involved in pathogen resistance at the individual level and how this translates into the maintenance of MHC diversity at the population level is not well understood (Sommer 2005; Piertney & Oliver 2006).

At the population level, MHC diversity is thought to be maintained by balancing selection through (i) heterozygote advantage, (ii) frequency-dependent selection, or (iii) variable selection in time and space (Nei & Hughes 1991; Hedrick 2002). The heterozygote advantage (or overdominance) hypothesis suggests that the fitness of heterozygote individuals is greater than that of both homozygotes because of their ability to present a broader spectrum of antigens to T cells (Hughes & Nei 1988). MHC heterozygote advantage was observed in humans for resistance to hepatitis B and HIV infections (Thursz et al. 1997; Carrington et al. 1999) and in mice for resistance to Salmonella infections (Penn et al. 2002; McClelland et al. 2003). On the other hand, frequency-dependent selection occurs when the advantage of possessing an MHC allele depends on its frequency in the population and is usually referred to as negative frequency-dependent selection or rare allele advantage (Takahata & Nei 1990). The variable selection in time and space hypothesis reasons that MHC polymorphism is maintained by shifting pathogen compositions that occur through time and space (Hedrick 2002). Support for the latter two hypotheses is provided by the association of certain MHC alleles with resistance to specific diseases, such as those found for malaria and HIV in humans (Hill et al. 1991; Carrington et al. 1999), Marek's disease in chicken (Briles et al. 1977) and furunculosis, infectious salmon anaemia and infectious haemopoietic necrosis virus in salmonids (Langefors et al. 2001; Grimholt et al. 2003; Miller et al. 2004). Further support for the latter hypothesis was also found in a temporal study conducted on the great reed warbler, Acrocephalus arundinaceus, where change in frequency of two MHC class I alleles contrasted with the stochastic variability of microsatellite markers (Westerdahl et al. 2004). When contrasting the heterozygote advantage and the frequency-dependent selection hypotheses in pathogen resistance, the latter has received more support in the literature (Sommer 2005). However, the exact pathogen resistance mechanism occurring at the individual level in the wild is yet to be clarified. Moreover, although the importance of pre-existing genetic variation, referred to as ‘standing genetic variation’, has been emphasized for adaptation and rapid evolution (Barrett & Schulter 2008), its importance for pathogen resistance in the wild is yet to be evaluated. Finally, although associations between pathogenic infection and population abundance have been observed in the wild (Krkošek et al. 2007; Miller & Vincent 2008), investigating demographic consequences of immune maladaptation is still in its infancy.

Anadromous populations of Atlantic salmon, Salmo salar, reproduce and spend the first years of their life in rivers along the North American and European Atlantic coasts, where a major decline in abundance has been observed during the last decades (Caron et al. 2005). After a feeding period of 1–3 years at sea, the majority of Atlantic salmon returns to spawn in natal rivers (Stabell 1984) and evidence suggests that they adapt to local conditions (reviewed in Taylor 1991; García de Leániz et al. 2007). Recently, adaptation to local temperature regime and bacterial community prevailing in the natal river has been supported through large-scale variations in MHC class IIβ diversity specifically at the PBR, suggesting the influence of pathogen-driven balancing selection on MHC diversity in Atlantic salmon (Dionne et al. 2007). Also, indirect evidence was found for lower fitness of immigrants coming from genetically and thermally distinct regions, suggesting a potential influence of local adaptation in limiting gene flow (Dionne et al. 2008). Evidence of adaptation through disassortative mating at the PBR in MHC class II was also found, possibly providing better pathogen resistance for the progeny (Landry et al. 2001). Evidence of selection at MHC class II was also found at the juvenile stage. Indeed, observed genotype frequencies of hatchery juveniles introduced in a river were significantly different after six months than the expected frequencies from parental crosses (de Eyto et al. 2007). However, uncertainty still persists whether early juvenile stages are able to mount specific immune responses (Bakke & Harris 1998). More importantly, information about selective agents, namely pathogens infecting Atlantic salmon in the wild along with the associated infection rates, is lacking. This lack of knowledge is even more pronounced for juvenile stages, as most pathogen resistance studies have been conducted on adults in laboratory conditions (Langefors et al. 2001; Grimholt et al. 2003; Miller et al. 2004).

Spatial variation in infectivity suggested a trend for higher infection rates and pathogen diversity in juvenile salmon from south shore compared with north shore rivers in the St Lawrence system. This result was concordant with the evidence of a higher pathogen selection pressure in southern warm than northern cold habitats found in a large-scale study conducted on wild Atlantic salmon populations (Dionne et al. 2007). Indeed, the bacterial community in each of the six rivers was analysed during the same year and bacterial diversity was higher in south shore than in north shore rivers (38.9±12.8 and 2.5±2.4×10³IODmm⁻², respectively; see Dionne et al. (2007) for bacterial community analyses). Temporal variation in infectivity also suggested higher infection rates in June when water temperature rise rapidly from approximately 6°C to over 18°C (Ministère des Ressources Naturelles et de la Faune du Québec). Pathogenic bacteria and other parasites may increase in abundance and virulence once a temperature threshold is reached, as is the case for myxozoa infections that mostly occur at more than 15°C (Uhland et al. 2000), and as previously observed in the laboratory for other fish pathogens (e.g. Larsen et al. 2004). The observed decrease in infection rate thereafter suggested that either juveniles lost their infection with time, following the pathogen natural life cycle or through immune resistance, or mortality occurred following infection. Although these hypotheses are not mutually exclusive, our data are more consistent with the latter hypothesis, as a significant decrease in frequency of a MHC susceptibility allele towards myxozoa infection, as well as a trend towards the increase in a MHC resistance allele, were observed during the summer while no significant change was detected for other MHC and microsatellite alleles. Overall, we found evidence for spatio-temporal variation in infection rate concordant with variation in pathogen selection pressure in the environment, and also with the immune competence associated with specific MHC alleles (see below).

Twelve species of potentially pathogenic bacteria were found in kidneys of wild juvenile salmon, all of which had a low prevalence in the populations. Some of them were known pathogens, namely Mycoplasma infecting farmed cod and other vertebrates (Messick et al. 2002; Nylund et al. 2006), the pathogenic bacteria Rickettsiella mainly infecting arthropods (Cordeaux et al. 2007), the opportunistic fish pathogen Pseudomonas causing septicaemia in stressful conditions (Thune et al. 1993), Shewanella responsible for ulcer disease in some molluscs and fish (Cai et al. 2006) and Piscirickettsia causing epizootic diseases in salmonids (Fryer & Hedrick 2003). Other bacteria, namely Sphingomonas, Methylobacterium, Bosea, Lactobacillus and Burkholderia, have been previously found in soil and freshwater environments and some such as Burkholderia can become opportunistic pathogens (Berriatua et al. 2001). Although prevalence was low for these bacteria, some of them such as Piscirickettsia are qualified as epizootic, affecting a small number of individuals but, under favourable environmental conditions, possessing the capacity to spread rapidly in the population (Fryer & Hedrick 2003). As such, these could constitute latent pathogens representing a potential health risk for salmon populations in the future. The most prevalent pathogen infecting juvenile salmon was a myxozoa of the Myxobolus clade, a group of fish parasites first discovered in farmed salmonids in Europe (Hoffman 1990). Myxobolus infections, as well as other documented myxozoa infections, have been detected in the USA since the 1950s and epidemiological and molecular evidence suggests that these parasites were initially introduced through fish transported from Europe (Hedrick et al. 1993; Andree et al. 1999; Kent et al. 2001; Miller & Vincent 2008). Myxobolus infections have been recently reported in eastern Canada in some fish species such as the log-perch, Percina caprodes, and the banded killifish, Fundulus diaphanus (Cone & Marcogliese 2004; Cone et al. 2006). However, to our knowledge, this is the first published report of a myxozoa infection in wild Atlantic salmon in eastern Canada. As such, this most probably represents an emerging salmon parasite in this system. Myxozoa parasites are responsible for multiple diseases in salmonids, including the well-documented whirling disease caused by Myxobolus cerebralis, which can severely compromise survival in some species (Kent et al. 2001). Even though MHC allele 7 covaried with regional variability, MHC alleles 7 and 8 were associated with resistance and susceptibility to myxozoa infection, respectively, after controlling for spatio-temporal variability in infection rate. Previous studies conducted in the laboratory also found an association between an MHC allele and incidence of diseases, namely furunculosis, infectious salmon anaemia and infectious haemopoietic necrosis virus in Atlantic salmon (Langefors et al. 2001; Grimholt et al. 2003; Miller et al. 2004). However, very few studies have attempted to test for an association between MHC alleles and parasite resistance in wild populations (but see Harf & Sommer (2005) and references therein on small mammals and Madsen & Ujvari (2006) on pythons). In addition to identifying an association between MHC alleles and susceptibility, we have also attempted to disentangle the influence of spatio-temporal variability in pathogen selection pressure from immune resistance processes in determining infection rates in the wild, which has always been a challenge in the past (Wegner et al. 2003). The significant shift in the frequency of the susceptibility allele 8 over the summer supports the hypothesis of ‘real-time’ myxozoa selection pressure on juvenile Atlantic salmon causing mortality over the summer. An increasing number of studies have demonstrated the effect of natural selection on MHC genes over an evolutionary time scale in the wild (Miller et al. 2001; Prugnolle et al. 2005; van Oosterhout et al. 2006; Dionne et al. 2007), but few have identified differential mortality associated with MHC allelic composition during a single generation (Piertney & Oliver 2006; but see de Eyto et al. 2007). Assuming that this susceptibility allele frequency change is strictly associated with mortality, we can speculate that the deleterious consequences of maladaptation in these populations resulted roughly in a 5 per cent reduction in population abundance from June to August (allele frequency changes from 0.088 to 0.040, figure 3). In another system, the introduction of M. cerebralis was associated with a drastic decline in rainbow trout (Oncorhynchus mykiss) recruitment over 18 years (Miller & Vincent 2008). Altogether, these results then suggest that Myxobolus parasites could have demographic consequences on infected populations, and underline the importance of immune adaptation for population persistence. Admittedly, more in-depth studies will be necessary to rigorously assess the deleterious consequences of immune maladaptation in salmon populations. Such studies will become increasingly important in the face of climate change, which may cause rapid variations in pathogen communities and selection pressures.

The heterozygote advantage hypothesis was not supported by our data, which might not be surprising in the context of a predominant pathogen, as suggested by Wegner et al. (2004). Admittedly, not all possible pathogens were identified in this study, which might have influenced the relationship observed between heterozygosity and infection rate. However, what was interesting was the detection of the reverse pattern, with significantly more MHC heterozygotes and higher MHC amino acid genetic distance in infected than in uninfected fish. A similar trend was also observed in Atlantic salmon with furunculosis in the laboratory (Langefors et al. 2001). One possible explanation could be the low frequency of the MHC susceptibility allele 8 (overall frequency less than 0.05), which would result in this allele being mostly restricted to heterozygous individuals, with the result that heterozygotes appear more infected than expected. In such a case, MHC amino acid genetic distance should follow this pattern only if the susceptibility allele is more differentiated in terms of amino acids compared with other alleles, which was indeed found at PBR but not at overall polymorphic sites (t-test comparing p-distances associated with allele 8 with p-distances associated with all other allele comparisons: p=0.029 and 0.39, respectively), in concordance with the results of the MANOVA. Additionally, higher MHC heterozygosity in infected fish could also be partially explained by spatial variability, as a tendency for higher MHC heterozygosity was observed in south shore rivers where infections were more frequent (H_O south shore, 0.90±0.05; north shore, 0.81±0.05; t-test: p=0.085). This suggested that the heterozygote advantage hypothesis should be tested with care in the wild, as frequency and identity of MHC alleles conferring susceptibility as well as spatial variability in MHC heterozygosity could influence the relationship between MHC heterozygosity (or other MHC individual genetic distance) and infection rate.

Our results, showing an association between MHC alleles and infection level, are more compatible with the negative frequency-dependent selection and the variable selection in time and space hypotheses (Takahata & Nei 1990; Hedrick 2002), but possibly better support the latter. Although rare alleles were not found to be more advantageous than the others, this study suggested that at the individual level, when facing an emerging pathogen, possessing the right alleles to initiate an appropriate immune response may be important for contemporary adaptation. As the pathogen community is changing, different alleles could be selected for through time, maintaining MHC polymorphism at the population level over an evolutionary time frame, as suggested by the variable selection in time and space hypothesis (Hedrick 2002) and as evidenced in the temporal study of Westerdahl et al. (2004). Indeed, at the population level, balancing selection maintaining MHC diversity could represent the combined influence of multiple directional selection pressures associated with each pathogen over time. However, allele frequency and infection rates would need to be followed across generations to find a stronger support for these hypotheses and evaluate whether a resistance process will develop over an evolutionary time frame in these myxozoa-exposed populations.