Family variation in surface feeding behavior of steelhead fry predicts growth rate under hatchery conditions

Hatchery reared salmonid fish often have lower fitness than their natural-origin counterparts when spawning in the wild. Although this difference appears to result from rapid adaptation to captivity, it is not known what traits are under selection. We hypothesize that variation in traits that confer a growth rate advantage to some individuals in the novel hatchery environment are under strong selection because survival at sea is correlated with size at release. Here we show that full sibling families of steelhead trout (Oncorhynchus mykiss) show substantial variation in propensity to feed at the surface as fry, and that the more surface-oriented families grew faster under hatchery conditions. We hypothesize that surface-oriented fry gain an initial growth advantage that persists through size at release. Because surface orientation is a correlate of generalized boldness, hatcheries may inadvertently select for that phenotype, which could explain the fitness differences observed between hatchery and natural-origin fish in the wild.


33
Hatchery-reared salmonid fish generally have lower fitness than natural-origin fish when 34 spawning in the wild, and this effect is true even for early-generation hatchery fish (1-7).
35 Common garden experiments in steelhead show that the fitness reduction appears to be genetic, 36 rather than simply an environmental effect of the hatchery (1,6,8). That first-generation hatchery 37 steelhead perform worse than natural-origin fish in the wild, but much better as broodstock in the 38 hatchery, suggests that the genetic difference results from rapid adaptation to captivity (9).

44
The size of hatchery salmon at release is positively correlated with their probability of 45 survival at sea (10-15). Therefore, a plausible hypothesis to explain rapid domestication is that 46 hatcheries select for physiological or behavioral traits that allow some fish to grow quickly in the 47 unnatural conditions in a hatchery (11,16,17). Such selection might be especially intense on 48 steelhead, which are typically raised to smolting in one year in hatcheries versus the normal two 49 years that they take in the wild (18,19). If those hatchery-favored traits are maladaptive in the 50 wild, then that could explain why hatchery fish quickly evolve to have lower reproductive 51 success than natural-origin fish in the wild environment.

52
Reduced anxiety/fearfulness is an almost universal outcome of domestication in 53 vertebrates (20). Traits such as quickness to feed, and propensity to be near the surface in fish 77 We hypothesize that this variation reflects heritable variation along the shyness-boldness axis, 78 and that the bolder families get an initial growth advantage that they maintain through the rest of 79 the rearing period. If these bolder families wind up larger at release, then subsequent size-80 selective mortality at sea would result in selection for boldness. If it is indeed true that bolder 81 fish gain an initial growth advantage during the first few days of learning to feed on artificial 82 food, then it may be possible to reduce their initial advantage by altering the feeding method. In 83 particular, if shy fish are startled by humans, then using automated feeders might reduce the 84 performance difference between bold and shy families. Here we tested whether families that 85 scored higher on a correlate of boldness (propensity to feed at the surface as fry) also grew faster 86 in mixed-family hatchery tanks, and whether that relationship was stronger when fish were hand-87 fed versus fed by automatic feeders.

88
We used a group of 11 full-sibling families of steelhead to test the following predictions: 142 missing from some tanks, although all families were represented in all treatments (S1 Table).

143
Fry were ponded directly into the 400 L tanks on the same day that their siblings were set 144 up in 25-L buckets to quantify surface feeding (May 30 th , 2018). Tanks were randomly assigned 145 to one of three feeding treatments, with three replicate tanks per treatment. The first treatment 146 was fed by hand, as this style of broadcast feeding is typical in hatchery settings. The second 147 two treatments utilized the same amount of feed, but using belt autofeeders to allow fish to feed 148 without interacting with humans. The first autofeeder treatment ("auto-bolus") mimicked the 149 hand feeding schedule with parcels of food entering the tank with each round of hand feeding.  One possible caveat to the result is that three of the families (CC, DD and EE) were 206 spawned 9 days earlier than the other eight. CC and DD were among the largest, and also had 207 the highest average nose-pokes values, although EE was unremarkable (Fig 2). As is standard 208 hatchery practice, the three earlier families were kept in cooled water for a week to delay their 209 development until the rest were spawned in order to ensure all families were ponded at similar 210 sizes. Thompson et al. (39) showed, using a different steelhead hatchery program, that families 211 from earlier spawn dates wind up larger at release, presumably owing to the cooling effect. This 212 effect was significant in a year when spawning was spread evenly over a seven-week period, 213 although not in a year when it was spread mostly over a two-week period. Thus, although it is 214 conceivable that the sizes of families CC and DD were slightly inflated by the cooling, the effect 215 of a 9-day difference is likely to be very small. We cannot statistically adjust for spawn date 216 because there are so few families, and families CC and DD make spawn date highly confounded 217 with nose-pokes. We therefore estimated the possible effect using the size vs. spawn-date

234
If hatcheries are indeed selecting for highly bold behavior in fry and juveniles, then it 235 might be possible to alter the way fry are trained to feed in order to reduce the initial growth 236 disparity between shy and bold fish. For example, perhaps one could separate out the initially 237 shy fish and feed them separately until they catch up in size to the others. In this experiment, 238 using automated feeders instead of hand-feeding did not reduce the correlation between behavior 239 and final body size. So that change in husbandry seems unlikely to substantially reduce the 240 intensity of selection. Of course, interventions such as this presume that a goal of the hatchery is 241 to produce adult fish having fitness as similar as possible to that of natural-origin fish, rather than 242 a goal of maximizing production. Regardless, these data support the hypothesis that inadvertent