By Greg Lutz*
One of the simplest means of improving the efficiency of selection is to improve the degree to which individual genetic values are estimated. Remember that in most cases only a small portion of the phenotypic (observable) variation we measure is attributable to directly inherited genetic influences. This portion is referred to as the heritability of the trait, and the relative magnitude of this portion of variation is an important factor in the design of an improvement program. When the genetic potential of individual fish is obscured by environmental conditions and/or competition, selection can be difficult to pursue.
One method of increasing accuracy when assessing genetic values involves compiling repeated measurements of an individual’s performance in certain traits – number of eggs per gram of body weight (evaluated over a number of spawns) would be one example in tilapia. Another approach is to select individuals based on the performance of their offspring. This practice increases the generation interval for selection, but it can be useful when selecting for traits with low heritability (relatively high levels of environmental influence or comparatively small amounts of additive genetic variation to work with). It’s also useful for traits that are expressed in only one sex (again fecundity characteristics), or traits that describe slaughter-related characters such as fillet yield.
A third approach to selection is to estimate an individual’s genetic worth, wholly or partially, based on the overall performance of its immediate family. As pointed out above, when heritability is low, a large portion of the superiority or inferiority of any individual we observe will tend to result from influences that are not heritable – and under these circumstances, the mean of an entire family may tell us more about a fish’s genetic merit than its individual phenotype. In reality, this is the only set of circumstances in which family selection is actually more efficient than mass selection, and worth the extra labor, infrastructure and operational costs required to conduct it. The progress per generation, however, can be much less than what might be attainable using offspring performance as a basis for selection.
The positive trade-off with this type of approach lies in the reduction of the generation interval when using information from overall family performance. Family selection relies solely on family averages. Any given individual within the family, be it the largest or the smallest, is assigned the same “value” to reflect the ranking of the family against other families. In this practice, all (or a random sample of) the members of those families with the highest averages are retained for breeding purposes. For traits with low heritabilities, family selection is generally more efficient than mass selection. But questions come into play in terms of costs and the accumulation of inbreeding.
In species like tilapia, significant differences will be apparent between male and female fish, so families should be ranked by sex. The family score can then be the average of the male and female averages, or families can be ranked separately for each sex and females from the families with the highest female rankings are mated with males from the families with the highest male rankings. The first approach is probably more appropriate for a tilapia breeding program. Family selection requires discreet families, produced by specific, identifiable males and females. Unfortunately, tilapia are usually spawned in large groups and are notoriously uncooperative when single pair matings are attempted. Since male tilapia are often inclined to harass unreceptive females to the point of death when housed as single pairs, one simple trick to improve the chances of success for production of multiple families is to surgically remove the upper lip (maxillary and pre-maxillary bones) of the males. An illustrated protocol can be found here: http://pubs.iclarm.net/Pubs/GIFTmanual/pdf/GIFTmanual-04.pdf
Once you manage to create the families, another inherent problem with family selection is that it requires an accurate way to keep track of them, preferably while raising them all together in the same environment. While there are certainly ways to do this, an inability to identify fish to particular families in many commercial settings often requires growers to rely on mass selection even for traits with low heritabilities. When all fish of interest cannot be raised together in a common environment, using individuals’ deviations from group averages has sometimes been proposed as a means to avoid bias from separate environments and trends over time, but thought must be given as to which, if not all, of its contemporaries an individual will be evaluated against (its full-siblings, its half-siblings, all contemporaneous individuals within the same pond, or cage, or recirculating system, etc.) Common environments, or competition and social hierarchies within common environments (something tilapia are notorious for), become important considerations in these evaluations. When all these problems are taken together, in some circumstances the optimal selection approach for tilapia involves ranking individuals within each family. This approach, known as “within-family selection,” may be of particular value if there is no alternative but to segregate families in order to track their performance.
Within-family selection is most efficient when differences between families are largely due to differences in their rearing environments. Even under the most similar conditions attainable, subtle differences in culture environments can still exist among family groups, but with this approach they do not directly bias the evaluation criterion. An added benefit of within-family selection is the ease involved in selecting members of every family to replace the preceding generation. This is a worthwhile practice, because when every family contributes equally to the breeding population the effective population size theoretically becomes two times the actual population size. In this way, the accumulation of inbreeding depression is slowed considerably, and in spite of the claims of some genetic consultants inbreeding can accumulate more rapidly in a family selection program than under mass selection conditions.
In either case, if some sort of physical marker is used to identify fish by family, be it branding, clipping, or PIT tagging, another question arises… how many individuals to mark per family? Depending on the size of the breeding stock being used, a family of tilapia fingerlings can easily exceed 700 or even 1000 individuals. No one wants to tag that many fingerlings, especially if some 100 – 200 families are being compared (which is often considered a necessary scale of production in family selection in order to minimize the accumulation of inbreeding). In 2014 Chao Song and his colleagues demonstrated that when channel catfish fingerlings exhibited family differences of 0.2, approximately 55 animals per family were required for reliable detection of these differences. However, when the true differences dropped to 0.15, 90 individuals per family would be required to detect them.
Molecular markers can provide an alternative with some significant advantages – every individual is born with the required information to trace it back to its family of origin, if you have the right set of markers available. Under some conditions, such as when broodstock are spawned en masse in groups, family selection and within-family selection are generally not possible without molecular marking techniques. Even when molecular markers are available, they are usually not cheap so a trade-off must be considered between the number of fish being genotyped and the amount of useful information to be gained.
Occasionally, an individual’s performance can be combined with that of its family to provide a composite score. This approach is referred to as ‘combined selection.’ Modeling and quantitative theory indicate that gains from combined selection will consistently surpass those from mass selection or family selection. However, as in other situations the requirement to identify and track each individual may not justify the additional efficiency of this type of selection, especially in an aquaculture setting.
A 1988 article by Horstgen-Schwark and Langholz on the prospects of selecting for late maturity in Nile tilapia (Oreochromis niloticus) illustrates a number of the concepts involved in the various approaches to selection discussed above. The goal of their study was to evaluate a breeding program designed to delay maturation. The initial breeding population consisted of 35 full-sib families of fish (a full sib family has both the mother and father in common). Half of the fish from each family were slaughtered at 136 days of age and gonadosomatic index (GSI) and a visual assessment of gonadal development (VAGD- my term, not the authors’) were recorded for each slaughtered fish. These values were then used as selection criteria for males and females, respectively.
Families were ranked separately for each sex, based on average GSI for males and average VAGD for females. Selection was directional in each trait: for the lowest values, to reflect late maturation of both males and females. In this sense, family selection was being practiced, but with a twist: based on separate rankings for male maturity within families and female maturity within families. At this point, another twist was added: families were ranked by mean weight, and only families equal to or above the overall average were considered for selection of late-maturing animals. This was considered necessary to avoid the selection of families with high proportions of underdeveloped offspring. So, families were selected based not only on late maturation of males and/or females, but also on growth.
Females from those families with the lowest average VAGD values were spawned with males from families with lowest average GSI values, but full-sib matings were not allowed to avoid any complications from inbreeding depression. Additionally, only the heaviest fish within each selected family were used to produce the next generation, in order to reduce the time interval between generations. As a result, only about 25% of the late-maturing males and 50% of the late-maturing females were spawned. This equated to within-family selection for growth, combined with the family selection for maturation and growth that had already been carried out.
Two males and four females were selected randomly from each family to serve as control lines for each generation. When compared to unselected controls, selected lines exhibited a reduction of 1.25 standard deviations for both GSI in males and VAGD in females after two generations of selection. These responses were statistically significant.
Bolivar reported in 1999 on within-family selection trials for growth in Nile tilapia. The selection criteria in this study was fairly easy to track: growth during 16 weeks post-hatching. The rationale for this work was the development of a selection strategy that could be applied with relatively limited facilities. Over 12 generations, within-family selection for growth was applied to tilapia grown in tanks. A continuous linear response for weight at 16 weeks was apparent over the entire course of the study.
Calculations based on observed selection response indicated a heritability for weight at 16 weeks of 0.38 in the base population, with a potential genetic gain of roughly 12% per generation. While the selection program was carried out in tanks, selected lines were evaluated for their performance in hapas and ponds to determine if gains would be of value under commercial conditions. Substantial selection response was apparent in both environments, suggesting that relatively modest facilities can be utilized to produce significant genetic improvements in extensive tilapia production settings, without the need for complex or expensive techniques like molecular markers.
In many cases, complex family-selection based improvement programs have been promoted within the tilapia industry as a means to generate gains in growth while minimizing the accumulation of inbreeding (remember, inbreeding only accumulates… kind of like enemies). However, this is a costly trade-off because family selection is much less efficient due to the fact that only 50% of the additive genetic variance in a population is expressed ‘between’ families (that is the unfortunate reality, based on statistical equations). Consider a large operation that requires 7 million larvae per month (not uncommon among some large tilapia operations). If we are generous, in terms of a fecundity estimate, and assume each female will produce roughly 1,000 larvae per month, a minimum of 7000 females will be required to produce this amount, and probably some 2,000 to 3,000 males. This equates to an effective population number (Ne) of some 8,500 animals. Now, further assume that some portion of the breeding stock on hand is resting during any given month, so let’s say the farm needs some 15,000 animals on hand to meet the year-round demand for fry.
If we opt for mass selection, simply choosing superior animals with no regard to their pedigrees, out of an annual production of some 84 million fry, some 15,000 will be destined to replace their parents as breeding stock. This is a mere 0.018%. Sounds bad in terms of inbreeding. Really bad. Now, if we used the fry from any given month for our selection activities and if 100% of the superiority observed in those 15,000 fish was directly heritable, they might be represented by as few as 15 females and perhaps 5 males – an Ne of around 15. If a particular group of breeding individuals WERE so superior as to contribute such a disproportional amount of offspring in the selected portion of the next generation, then by all means this superiority should be retained within the population in spite of the potential for increased inbreeding levels. Here’s an inside tip: in simple terms this has historically been referred to as SELECTION.
But of course, such high heritabilities are never the case, since most of the variation in performance we observe is attributable to non-heritable factors, as has already been established. Let’s assume a fairly reasonable heritability (say, one that might justify family selection) of 0.20. Now to get those gains we are talking about a projected effective population number of 375 (based on 375 females and 125 males). Inbreeding would be expected to accumulate at roughly 1/2Ne per generation, or 0.0013.
If the same operation were using family selection, with, say, 180 families per generation, the effective population size would be 180 females and 180 males, with an Ne of 360. Inbreeding would be expected to accumulate at 0.0014. Actually, not even a superior number, in spite of the costs, labor and infrastructure involved. Now, if there were a way to guarantee that every family contributed equally to the next generation of breeding stock, this accumulation is effectively cut in half, and some benefit is gained. However, the accumulation of inbreeding in all of these hypothetical cases is negligible, and easily offset by selection gains, largely as a result of the high fecundity of tilapia.
C. Greg Lutz, has a PhD in Wildlife and Fisheries Science from the Louisiana State University. His interests include recirculating system technology and population dynamics, quantitative genetics and multivariate analyses and the use of web based technology for result-demonstration methods.