Comparative mapping reveals quantitative trait loci that affect spawning time in coho salmon (Oncorhynchus kisutch)

Spawning time in salmonids is a sex-limited quantitative trait that can be modified by selection. In rainbow trout (Oncorhynchus mykiss), various quantitative trait loci (QTL) that affect the expression of this trait have been discovered. In this study, we describe four microsatellite loci associated with two possible spawning time QTL regions in coho salmon (Oncorhynchus kisutch). The four loci were identified in females from two populations (early and late spawners) produced by divergent selection from the same base population. Three of the loci (OmyFGT34TUF, One2ASC and One19ASC) that were strongly associated with spawning time in coho salmon (p < 0.0002) were previously associated with QTL for the same trait in rainbow trout; a fourth loci (Oki10) with a suggestive association (p = 0.00035) mapped 10 cM from locus OmyFGT34TUF in rainbow trout. The changes in allelic frequency observed after three generations of selection were greater than expected because of genetic drift. This work shows that comparing information from closely-related species is a valid strategy for identifying QTLs for marker-assisted selection in species whose genomes are poorly characterized or lack a saturated genetic map.


Introduction
Spawning time in salmonids is an important sexlimited life-history trait that determines fertilization and progeny emergence dates and also affects the probability of survival and growth rate of small fry (Quinn et al., 2002).An increase in the reproductive period in salmon farming allows better management of fish production (to account for seasonal variations) and increases the period during which eggs are available on the market (Gall and Neira, 2004).
In coho salmon (Oncorhynchus kisutch), it is possible to shift the spawning time of cultivated populations (Quinn et al., 2002).Estimates of heritability in cultivated popula-tions from Chile range from 0.24 ± 0.07 (Gall and Neira, 2004) to 0.40 ± 0.06 (Neira et al., 2006) and this trait responds to early spawning selection.Traditional selection works well, with the phenotypic response to selection for early spawning fluctuating between -2.74 ± 0.7 and -3.23 ± 1.3 days per generation (Neira et al., 2006).However, phenotypic selection is still inefficient as it is impossible to impose selection directly on males.Furthermore, this phenotype is expressed during the reproductive age, at the end of the salmon's life.In this context, marker-assisted selection could increase the response to early or late spawning time selection in the short term.
Despite its importance as a Chilean farmed species, coho salmon remains poorly characterized from a genetic standpoint.In contrast to rainbow trout and Atlantic salmon (Salmo salar) (Davidson et al., 2010;Palti et al., 2011), the lack of a dense genetic map for this species has delayed the search for QTLs related to commercial and life-history traits (Araneda, 2005).Although a map has been published for coho salmon, it is based on an analysis of 48 F 2 individuals and has low resolution, with only 133 co-dominant markers spanning 429.7 cM in the female map (McClelland and Naish, 2008).This map has allowed the QTL mapping of minor effects for growth rate, length, weight and hatch-ing time (McClelland and Naish, 2010;O'Malley et al., 2010).Nevertheless, it is possible to use a comparative approach to discover new QTL by using genetic markers linked to QTL in closely-related species.This approach has been used to identify QTL associated with temperature tolerance in Arctic char (Salvelinus alpinus) (Somorjai et al., 2003a) based on previously-identified QTL from rainbow trout (Jackson et al., 1998;Perry et al., 2001).
The main aim of this study was to use a comparative approach to identify microsatellite loci associated with potential QTL that affect spawning time (SPT-QTL) in coho salmon; the microsatellites used are reportedly linked to this trait in rainbow trout.If the microsatellite markers linked to SPT-QTL in rainbow trout are conserved in coho salmon then we would expect to find strong allelic heterogeneity between populations under divergent selection for spawning time.Such heterogeneity would indicate an association between these microsatellite loci and SPT-QTL in coho salmon.The identification of loci linked to SPT-QTL should allow marker-assisted selection for spawning time in coho salmon.

Experimental population and phenotypic evaluation
The fish used in this study were reared in the Coho Breed Improvement Program facilities (Centro de Mejoramiento Genético) located in Coyhaique in southern Chile (S 45°34.422'W 72°04.436' W).The program started with two-year classes in 1992 and 1993, both of which were closed populations managed under a two-year reproductive cycle.The populations consisted of 30-35 males that were mated with 100-120 females in each cycle followed by selection for harvest weight and early spawning using an animal model (Winkler et al., 1999).In 1995, a divergent selection experiment was initiated using two sets of fish as breeders: those that spawned during the first third of the spawning season (40 females and 13 males, N e = 39.2;early spawning population) and those that spawned during the last third of the spawning season (40 females and 12 males, N e = 36.9;late spawning population).The effective size was held essentially constant for the next three generations (N e » 40) by mating 12-14 males with 40 females (Araneda et al., 2009).Both populations were selected for three generations and spawning time was recorded as the number of days starting from December 31 st to the date of spawning for every season (Gall and Neira, 2004).In 2001, blood samples for DNA extraction were obtained from 20 females from the early spawning population and 20 females from the late spawning population.Additionally, DNA samples from 40 base population females were obtained from our sample bank.The average difference in spawning time between early and late populations in 2001 was 85 days (Araneda et al., 2009).

Microsatellite loci and PCR conditions
Nine microsatellite loci were used to screen for associations with spawning time (Table 1).Six and three microsatellite loci were previously identified as linked and unlinked with spawning time QTL (SPT-QTL) in rainbow 516 Araneda et al.  (Guyomard et al., 2006) and 14.2 cM from a SPT-QTL closely linked to OmyFGT34TUF (Sakamoto et al., 1999;O'Malley et al., 2003).For all descriptions of rainbow trout linkage groups we used the nomenclature proposed by Guyomard et al. (2006).
The forward primers used to amplify each locus were dye labeled and PCR amplicons were run on an automated sequencer (Model ABI377, Applied Biosystems) with GeneScan-500 ROX as the size standard.The thermal profile was 94 °C for 2 min, followed by 30 cycles at 94 °C for 30 s, 57 °C to 70 °C for 1 min (see Table 1 for the specific annealing temperature of each primer pair), 72 °C for 1 min, and a final 5 min extension step at 72 °C.For some primer sets, we used a touchdown protocol to improve the PCR fragment resolution (Table 1).PCR was done in a total volume of 15 mL containing 1.5 mL of 10x PCR buffer, 4.0 mM of each dNTP, 0.4 mM of primer, 1.8 mM MgCl 2 , 0.5 units of Taq DNA polymerase (Invitrogen) and 40 ng of DNA from each individual.DNA was extracted from blood samples using a phenol/chloroform protocol (Medrano et al., 1990) and quantified spectrophotometrically (Hewlett Packard model 8452A spectrophotometer).

Association analysis
Marker-trait associations were assessed using three statistical methods: (1) First, we applied the L D statistic, a multiple comparison approach based on contingency tables between microsatellite alleles and populations (Araneda et al., 2009;Colihueque et al., 2010).This procedure tests the null hypothesis that two populations are homogeneous with respect to the probability distribution of microsatellite al-leles; the alternative hypothesis is that at least one allele is excessively associated with a particular population (Choulakian and Mahdi, 2000).For every locus, the highest value of L D across alleles was compared to the chi-squared value, with one degree of freedom of 13.8 being equivalent to an LOD score of 3.0 [Z » c 2 /2 log(10)], which corresponded to an a level of approximately 0.0002.(2) Second, we used an c 2 Monte-Carlo bootstrapping algorithm with 10,000 iterations to test allelic heterogeneity between populations (Zaykin and Pudovkin, 1993).( 3) Finally, to assess genetic drift, we used a 99% confidence interval (CI) for allelic frequency variance for a locus with two alleles in which: 2 as an estimate of genetic drift, where p is the frequency of the most frequent allele, q is the pooled frequency of all other alleles, a = 0.01 and N e = 37 (the lowest value in our populations), so that d.f. was 2N e -1 = 73.We also estimated the prediction of change by drift for the allele frequency with the highest L D value for every locus after three generations as: p p S

Results
Table 2 shows a reduction in the number of alleles from 1995 to 2001 in nearly all of the loci sampled.This table also shows the range of allele sizes and the frequency of the most frequent allele across the three populations that were used to assess drift.The complete allele distributions and frequencies are shown in Tables S1 to S3 and the as-Comparative QTL in coho salmon 517 sessment of genetic drift is shown in Table S4 (all in Supplementary Material).
Six loci showed allelic heterogeneity among fish belonging to early and late spawning populations, which suggested that these loci could be associated with spawning time.Subsequent association analyses indicated that three loci (One2ASC, One19ASC and OmyFGT34TUF) were strongly associated with spawning time (p < 0.0002) and a fourth locus, Oki10, was close to the limit of significance (Table 3).All four microsatellite loci that were possibly associated with spawning time showed differences in the allelic distribution of early and late spawning females compared to females from the base population (Figure 1).In particular, for OmyFGT34TUF, alleles 139 and 143 occurred at a high frequency in late spawning females (32.5% and 40%, respectively), but were infrequent in early spawning females (5% and 2.5%, respectively).On the same locus, allele 185 also occurred at a high frequency (25%) and was found exclusively in early spawning females (Figure 1).Locus One2ASC had significantly higher allele 214 and 242 frequencies (30% and 37.5%, respectively) in late spawning compared to early spawning females (0% and 5%, respectively), and locus One19ASC had a high frequency (50%) of allele 232 in late spawning females compared to a frequency of only 10% in early spawning females (Figure 1).In early spawners, the latter locus also showed a high proportion (35%) of an exclusive allele (222).Finally, locus Oki10 contained two alleles (223 and 231) exclusive to the late spawning group that both had high frequencies (20% and 27.5%, respectively), while allele 129 (frequency of 20%) was observed exclusively in the early spawning group (Figure 1).
The genetic drift effect estimated by using the most frequent allele in 1995 showed an average change in allele frequency of 5% due to drift per generation, with an upper confidence interval (99%CI) limit of 7.3% (Table 4).The estimate of change due to drift, based on the allele frequency with the highest L D value, showed a drift effect that was always inferior to the change in gene frequency observed in 2001 for the three loci associated with spawning time (One2ASC, One19ASC and OmyFGT34TUF) and for Oki10.Thus, the frequency change expected due to drift was always inferior to the change observed after three generations of selection.For the other five loci, the change observed after selection was in the range of drift prediction (Table 4).

Discussion
Association analyses are always suspect because of the higher rate of false positives produced by spurious associations between phenotypes and non-causative marker loci.Such spurious associations can be produced by population subdivisions or genetic drift (Pritchard and Rosenberg, 1999).The reduction in the number of alleles from 1995 to 2001 was possibly a by-product of divergent selection for spawning time instead of a consequence of genetic drift.Our results indicated the association of three microsatellite loci (One2ASC, One19ASC and OmyFGT34TUF) with spawning time in coho salmon while a fourth locus (Oki10) had a suggestive association.In four loci the changes in allele frequencies were higher than expected by drift, which is consistent with a marker locus under co-selection with the QTL region.A similar pattern of co-selected markers linked to QTL has been shown for ethanol drinking in mice (Belknap et al., 1997) and such co-selection is proof of a true QTL (Abiola et al., 2003).As additional evidence, it should be noted that three of these loci were previously linked with QTL for the same trait in rainbow trout linkage groups RT24 and RT19 (Sakamoto et al., 1999;O'Malley et al., 2003).We have thus identified 518 Araneda et al. four SSR loci that are potentially useful in marker-assisted selection for early or late spawning time in coho salmon.
Our findings, along with previous evidence from QTL mapping in rainbow trout, support the presence of two QTL regions that affect spawning time in coho salmon.The proposed position of both QTL is based on assumed synteny between the chromosomes of coho salmon and rainbow trout; this assumption suggests that these QTL Comparative QTL in coho salmon 519  were present in ancestral genomes from which these species originated.
We hypothesize that one of these QTL is located close to the region bearing the loci One19ASC and One2ASC in a coho salmon linkage group syntenic with RT24 of rainbow trout.The RT24 linkage group of rainbow trout contains Ots4BLM and OmyPuPuPyDU, but these loci are located 23.5-24.5 cM from the pair One19ASC/One2ASC and, in agreement with our association analysis, they have never been linked with SPT-QTL (Sakamoto et al., 1999;O'Malley et al., 2003).The second QTL must be located in a linkage group syntenic to the rainbow trout linkage group RT19, in a region between Oki10 and OmyFGT34TUF, possibly near the latter locus.We expect that these putative SPT-QTL positions will be confirmed by formal linkage studies using these marker loci when coho salmon have a saturated genetic map.
Chromosome segment conservation among salmon species is being increasingly documented through the construction of genetic maps for salmonids and comparative genomic studies (Danzmann et al., 2005(Danzmann et al., , 2008;;Timusk et al., 2011).In addition, comparative QTL mapping is actively being undertaken for salmon species belonging to different genera.This approximation has been used to identify QTL for upper temperature tolerance among rainbow trout and Arctic char (Somorjai et al., 2003b), as well as for body weight and Fulton's condition factor among Oncorhynchus, Salvelinus and Salmo (Reid et al., 2005).Further evidence of synteny and conservation of the different priming sites for these microsatellites markers lies in the feasibility of using heterologous primers to amplify microsatellite loci across all salmon species (Araneda et al., 2008;Danzmann et al., 2008).Currently, all evidence obtained from comparative QTL mapping indicates that chromosome regions that affect the quantitative variation of several fitness-related traits in salmon, e.g., body weight, growth rate, spawning time and temperature tolerance, must have been present before the separation of lineages that gave rise to the modern salmonid species (O' Malley et al., 2003;Somorjai et al., 2003b;Reid et al., 2005).Table S4 -Assessment of the effect of genetic drift.
The approximate genetic drift was estimated in a simple way by using the formula for a locus with two alleles S = pq/sqrt(2Ne), where Ne = 37 was applied because this was the lowest value in our populations and the harmonic means across three generations must be close to this value.The confidence intervals (Lower lim and Upper lim) for S (drift) were estimated from the CI of the variance.
Drift was estimated using as p the frequency from the most frequent allele in 1995 and q = 1-p.

Locus
where p 0 is the allele frequency in the base population (1995) and p 3 is the frequency of the same allele in 2001 (after three generations of selection).

Figure 1 -
Figure 1 -Distribution of alleles in the four microsatellite loci of coho salmon showing the association with spawning date.White bars correspond to allelic frequencies for the early spawning population, black bars indicate allelic frequencies for the late spawning population and grey bars indicate allelic frequencies for the base population.

Table 1 -
Primer sequences, annealing temperatures and SPT-QTL linkage evidence for nine microsatellite loci used in the QTL screening of female coho salmon selected from early and late spawning populations.

Table 2 -
Allelic characteristics of nine microsatellite loci in coho salmon females from base, early and late populations.
p: frequency of the most common allele.

Table 3 -
Association analysis between eight microsatellite loci and spawning time in coho salmon females selected for early and late spawning time.

Table 4 -
Change due to genetic drift predicted in alleles with the highest L D value relative to the initial frequency (base population) across all nine loci.
E: Maximum frequency observed in the early population.L: Maximum frequency observed in the late population.
Drift was estimated using as p the frequency in 1995 from the allele with the highest LD value of the two 2001 populations.Maximum differences between observed and expected (by drift) alleles frequencies.Frequencies observed after 3 generations Change by drift after 3 generations