The genome scan for QTL affecting IPN resistance confirmed a major QTL explaining nearly all of the genetic variation for this trait. This implies that, in this population, the genetic component of resistance to IPN is largely under the control of one QTL, presumably corresponding to one gene, a surprising find given that disease resistance traits are generally presumed to be complex. The results agree remarkably well with the other study identifying a major QTL for IPN-resistance in Atlantic salmon [25
]. In both studies the QTL explained ~25% of the phenotypic variation, it segregated in a large fraction of mapping parents (7 out of 20 parents in the Scottish study; 72 out of 176 'fry' mapping parents in the present study), and the estimated QTL positions were highly similar. Only a single marker (Alu333) was shared between the two studies in the vicinity of the QTL, but the position of this marker relative to the QTL peak was very consistent. Thus, it seems reasonable to assume that the two studies have in fact identified one and the same QTL.
In terms of the minor QTL identified, on the other hand, the overlap between the two studies was less striking. In the Scottish study, one other genome-wide significant QTL was detected, located on linkage group 26, in addition to a chromosome-wide significant QTL in linkage group 19. Of these, only the QTL on linkage group 26 was found to be harbour a (suggestive) QTL in the present work (linkage group 26 in Houston et al. [25
] corresponds to linkage group 28 from the present work). The one other genome-significant QTL identified in the present study was identified on linkage group 4. These differences could be due to differences between the populations studied, with some polymorphisms segregating in one populnation but not in others, or they could be related to sampling since each population was represented by only a limited number of mapping parents in the genome scans. Also, one cannot rule out that some of the QTL could be false positives.
The major QTL on linkage group 21 was found to explain a large fraction of the phenotypic variance in both the post-smolt and fry data sets. In general, the fraction of variance explained by a QTL detected in a genome scan tends to be overestimated due to the so-called Beavis effect [26
]. It should be noted, however, that the Beavis effect gets less severe as the experimental power increases and as the observed fraction of variance increases. Thus, our own calculations based on the formulas presented by Xu [28
] showed that, for a QTL explaining 28% of the genetic variation, the bias would be almost negligible even if only 100 offspring had been genotyped. Similarly, given the sample sizes of this study, the upward bias would be insignificant even for QTL of very minor effects. We would argue that the fraction of variance explained by the major QTL on linkage group 21 is more likely to have been underestimated, since the correction for selective genotyping assumes that the 9% least and 9% most resistant post-smolts were genotyped, while in reality the set of "most resistant animals" was drawn randomly from the ~25% of animals that survived within each family. The fraction of genetic variance explained by the QTL on the post-smolt stage was calculated on the basis of the heritability estimated in fry, since a good estimate of the heritability at the post-smolt stage was not available. The heritabilities at the two life stages are not necessarily identical.
Body weight has been found to be correlated with immune functions in some studies (e.g. [29
]), an argument for the use of body weight as a covariate in analysis for disease resistance QTL. In this study, individual fish could only be weighed during or after the challenge test, and we were thus not able to include weight as a covariate in the test for QTL (since surviving fish lived longer and thus could have put on extra weight during the challenge). We did, however, analyse the data for body weight QTL, using the 'affected/resistant' status of the fish as a fixed effect in that analysis (Moen et al., in prep.). Linkage group 21 was found not to harbour experiment-wide significant QTL for body weight, meaning that the observed QTL for IPN resistance is not a correlated effect of a QTL for body weight.
For ease of presentation, segregation data from male and female mapping parents were combined in QTL analysis, but only female map distances were used. In large parts of the Atlantic salmon genome, recombination does not occur in males [31
], meaning that males do not contribute to the local positioning of QTLs in such regions. In other regions, male recombination rates are very high relative to female recombination rates. The F curves for linkage group 21 exemplify these particularities well; the sharp increase in the F statistic between OMM1197 and Omi27TUF is caused by these two markers being genetically unlinked in males, even though the female genetic distance between them is only 2 cM. For other parts of the linkage group, no male recombination events were found in the whole data set.
The major QTL for IPN-resistance on linkage group 21 had a strong effect in both post-smolts and in fry. These are very distinct life stages, given that fry live in fresh water and post-smolts in salt water. The overlapping confidence intervals for QTL position and the concurrence of QTL genotypes and linkage phases between fry and post-smolt stages strongly indicates that the same gene controls IPN resistance at both life stages, although one cannot rule out that two different, but linked genes determine resistance at the two life stages. A functioning adaptive immune system is most probably developed at a later stage than that of the experimental fish in the fry test [32
], indicating that the gene underlying the QTL is likely to pertain to the innate immune system. Controlled challenge tests represent somewhat of an 'artificial' environment compared to natural farming conditions, and may lack certain natural stressors that could influence the pattern of resistance to a pathogen. The major QTL on linkage group 21, however, does not appear to be affected by such environmental effects, since it was found to have a large effect both in a field trial [25
] and in an experimental trial.
According to inter-marker levels of r2
, the extent of linkage disequilibrium increase with decreasing inter-marker distance from 5 cM. The increase in LD with decreasing distance is more pronounced in this study than in an earlier study, where the post-smolt mapping families were used to calculate levels of LD across all linkage groups [31
]. This is probably due to a higher density of markers being used in the present study relative to the previous study. In spite of significant amounts of LD being present between markers in the QTL region, combined linkage disequilibrium/linkage analysis (LDLA) could not increase the QTL mapping resolution (data not shown). Possibly, the marker density in the QTL region needs to be increased before the QTL can be positioned with significantly more accuracy. At present, additional markers in the QTL region are not available.
Based on four-marker haplotypes found in QTL-heterozygous parents, putative genotypes at the underlying polymorphism were deduced in parents that had been found negative in the test for QTL. As expected, a highly significant surplus of homozygous genotypes was found, although there were 13 discrepancies in the form of parents that were negative in the test for QTL while being subsequently deduced to have the heterozygous genotype. Of these 13 parents, four carried copies of the three haplotypes alleles that were ambiguous in terms of being linked to Q or q. These four, therefore, were more likely to be true homozygotes representing ambiguous haplotypes. For the remaining nine parents, no apparent reason for their discrepancy could be found.
The QTL for IPN-resistance has now been detected in two different populations, Norwegian and Scottish, with high heterozygosity observed in both. This raises the interesting question of how an allele with a highly negative effect on a trait can be retained in populations at such large frequencies (especially since the mode of inheritance appears to be purely additive). One possible explanation is that the low-resistance allele has a positive effect on another trait under natural or artificial selection in these populations. This hypothesis, however, is opposed by an almost complete lack of observed negative genetic correlations between IPN-resistance and other traits recorded by Aqua Gen and other breeding companies. More specifically, among 14 recorded traits IPN resistance was found to be negatively and significantly correlated with only one trait. The trait in question was not among those that have been under strong selection, and the (genetic) correlation was also not strong (r = 0.17) (Aqua Gen Ltd., unpublished data). It is quite possible, therefore, that this QTL is neutral under wild conditions, but non-neutral within an aquaculture setting. Disease pressures are highly different within these two settings, and IPN-resistance has in Norway been subject to artificial selection for at most two salmon generations, with only family selection being employed and IPN being only one of several traits in the breeding goals; probably not strong enough selection for large changes in allele frequency to occur.
Given that a single QTL explains such a large proportion of the genetic variance for this trait, it is reasonable to assume that the high-resistance allele will move towards fixation relatively rapidly even without MAS, although MAS can accelerate the process significantly, or permit more focus to be put on other traits without compromising the genetic gain for IPN resistance. Some researchers might argue that loss of phenotype-affecting genetic variation should be avoided, even if the allele going towards extinction has not been found to have positive effects on other traits. In such a case, the marker tools described in this paper could be used to deliberately retain the low-resistance allele at low frequency in the population.
The QTL presented here has been in use in the Aqua Gen breeding programme since 2007, when within-family MAS was done on 30 half-sib groups in order to select the most IPN-resistant fish as parents of the elite fish (i.e. the fish used for egg production). At the time, MAS could only be done within family, and only within offspring of parents that had already been screened for segregation of the QTL. Presently, MAS could be done on the bulk of the population using linkage information, since QTL genotypes and marker-QTL linkage phases were established for 72% of the parents of the current generation (a number that can be increased by adding genotypes). There is also a potential for MAS based on linkage disequilibrium, since haplotypes were found that were associated with the underlying gene at the population level. However, the markers contributing to the haplotype cover a 10 cM interval, and are not likely to be stable over generations. Thus, there is a need to further characterise the QTL in order to obtain the most efficient tool for MAS, and to uncover the gene that is underlying this trait. Recent development in genomics, including an Atlantic salmon SNP-chip containing 16,000 SNPs developed at the Centre of Integrative Genetics (CIGENE; Ås, Norway), could prove highly useful in this respect. Also, samples of Aqua Gen broodstock dating as far back as the early 1980s will be investigated in order to test the hypothesis of the high-resistance allele having increased in frequency due to aquaculture breeding.