90 ,--------------------, 80 ,·�---------------;----' 70 60 +--=,,.,-= �------------------! 50 ,-�:--...:�_ �----------------l Fig. 2. An example of the relation- 40 ship between the expected loss of microsatellite alleles (percent allele loss) and the actual number of broodstock contributing gametes to the next generation (effective breeding number) after one, 1 0 ,-�--r---------=:::::::�--,=------l 0 -t------,---=�==--,--�---,----""Ll�-� 0 50 five and 10 generations. So What Have We Done? We examined variation at a number of microsatellite loci in random samples from a farm and compared this to variation in a sample from the wild population from which the hatchery broodstock had been collected. The variation was measured in terms of number of alleles observed in each sample, and the frequency that each allele occurs in the sample. We did this work looking at two species from farms in two countries. In one study we examined a random sample from each of two different farms and compared each with a sample from the different wild population from which broodstock had been collected. We examined three independent microsatellite loci, in 50 individuals in each sample. At one farm there was a loss of 37 percent (12 of 32) of the alleles compared with the wild sample, and for the other farm a 40 percent allele loss (8 of 20) was observed compared with the sample from the wild population. In our second study we took four random samples of 64 individuals from a farm stock and compared the variation at five microsatellite loci with that in a sample of 61 individuals from a wild population from which the broodstock were obtained. In this instance we observed allele losses of 39 percent to 67 percent in the farm samples (wild population sample had 83 alleles). What Can We Say of This Loss? First, most of the 'lost' alleles (those not observed in the farm) were observed at low frequencies ( <0. 10) in the wild sample, so the loss was not unexpected. However, we did observe some major shifts in the frequency of some alleles. For instance, a change in frequency from 0.01 in the wild to 0.30 in the farm, and from 0.30 to 0.60 in either direction occurred. Importantly, we also observed alleles that had become fixed at 100 percent in a farm sample, resulting in no variation at that DNA region. We, therefore, observed loss of alleles and changes in frequency of alleles between the wild and hatchery samples. Should this be of concern to a farmer and is the observed change statistically and biologically significant? Statistical Significance To measure the statistical significance of the change we compared the observed loss with an expected loss. The expected loss is the likelihood of observing the allele in the farm sample based on the frequency that we observed the allele in the wild sample, the number of generations of isolation between the samples, (in this case only one) and the number of parents actually contributing to the farm stock (this is the effective breeding number). Figure 2 shows an example of the relationship between the effective breeding number and expected loss of alleles, based on allele frequencies observed in the wild or founder population and the numb�r of genera� tions of separation. The relationship is based on the observed frequency of the alleles in the stock from which the broodstock were collected. Basi100 1 50 200 250 300 Effective breeding number cally and intuitively, the lower the effective breeding number, the higher the expected loss, and the greater the number of generations of isolation the greater the loss. In the example derived from one of our study samples, a loss of 40 percent of the alleles after just one generation would be expected if the effective number had been only 10 individuals. The actual number of broodstock spawned to create the cohort suggested by the hatchery records was much greater than this. The significance of the observed change can be affected by a difference in the number of individuals examined in each sample, the larger a sample size the greater the chance of seeing more alleles, particularly the rarer ones, and the choice of those individuals. Sample size was not an issue for us inasmuch as our wild and farm samples were of similar size. Despite similar sample sizes, however, in our second study we observed a large difference among the four farm cohorts with losses varying from 39 percent to 67 percent. All four represented statistically significant losses compared with the wild sample. If we combined all four samples into a single farm sample, despite the greater number of individuals, 256 compared with 64, we still observed a 23 percent loss compared with the wild sample, but that change was not statistically significant. This highlights the need to keep sample sizes similar, but also how the choice of the individuals in a sample can influence the result. We made an assumption in these calculations that all hatchery broodstock contributed to the next generaWORLD AQUACULTURE 7
RkJQdWJsaXNoZXIy MjExNDY=