Transcript
Outbreeding effect on embryo development in hybrids of allopatric pink salmon ( Oncorhynchus gorbuscha ) populations, a potentialconsequence of stock translocation Ivan A. Wang 1 , Sara E. Gilk 2 , William W. Smoker, Anthony J. Gharrett ⁎ Fisheries Division, School of Fisheries and Ocean Sciences, University of Alaska Fairbanks, 11120 Glacier Highway, Juneau, AK 99801, USA Abstract Previously, we demonstrated that F 2 pink salmon ( Oncorhynchus gorbuscha ) hybrids between spatially separated even-broodline populations had significantly lower survivals than their controls and that both F 1 and F 2 hybrids between spatially separated odd- broodline pink salmon populations had reduced survivals. Those observations confirmed that outbreeding depression occurs, but didnot provide insight into the mechanism. Development time, which often has a genetic basis in pink salmon, is probably locallyadaptive. It is a trait that may contribute to outbreeding depression in intercrosses of genetically divergent fish populations. Weanalyzed development times (accumulation of temperature days; ATUs) at hatching of hybrids from the same even- and odd-year broodline crosses in which reduced survival had been observed. First-generation hybrids between Auke Creek (Juneau, eastern Gulf of Alaska) females and Pillar Creek (Kodiak, western Gulf of Alaska) males and Auke Creek controls were made in 1996 and 1997,incubated at Auke Creek, and released to the ocean. Second-generation hybrids, controls, and both backcrosses were made in 1998and 1999. In 2001, F 1 hybrids were made between Pillar Creek females and both Auke and Pillar creek males and incubated at Pillar Creek. All crosses of Pillar Creek fish (hybrids at Auke Creek and controls at Pillar Creek) developed more slowly (hatched after accumulating more ATUs) in both the first and second generations. Trajectories of development times of back-crossed salmon lay between those of F 2 controls and hybrids. Comparison of development times of F 2 crosses and backcrosses suggested that neither anecological (locus-by-locus) nor a genetic (epistatic) model of outbreeding alone adequately explained observed differences. Prudent conservation of biodiversity requires consideration of biogeographic history as well as the genetic structure and local environment of the populations involved when actions are contemplated in which there is potential for intercrosses.© 2007 Elsevier B.V. All rights reserved. Keywords: Outbreeding depression; Embryonic development; Allopatric hybridization; Pink salmon; Oncorhynchus gorbuscha ; Local adaptation 1. Introduction Translocation of propagated stocks has been a com-mon “ solution ” to management and conservationconcerns, particularly in fisheries, for more than acentury. Even though growing evidence predicts that outbreeding depression may occur if individuals fromgenetically different populations intercross (e.g. for animals, Gharrett and Smoker, 1991; Phillip andClaussen, 1995; Edmands, 1999; Gharrett et al., 1999;Marshall and Spalton, 2000; Marr et al., 2002; Gilk et al., 2004; Miller et al., 2004), the practice is commonand continues to generate rancorous discussion (e.g.,Mudrack and Carmichael, 2005). Understanding both Available online at www.sciencedirect.com www.elsevier.com/locate/aqua-online ⁎ Corresponding author. Tel.: +1 907 796 6445; fax:+1 907 796 6447. E-mail address:
[email protected] (A.J. Gharrett). 1 Current address: University of New Mexico, Biology Department,MSC03 2020, Albuquerque, NM 87131, USA. 2 Current address: Alaska Department of Fish and Game, 333Raspberry Road, Anchorage, AK 99518, USA.0044-8486/$ - see front matter © 2007 Elsevier B.V. All rights reserved.doi:10.1016/j.aquaculture.2007.08.002Aquaculture 272S1 (2007) S152 – S160 the potential effects and mechanisms of outbreedingdepression is important to prudent conservation.Two models explain how outbreeding depression canoccur: the ‘ ecological ’ or locus-by-locus model resultsfrom accumulated effects, and the ‘ genetic ’ or epistaticmodel results from interactions between loci (Shields,1982; Templeton, 1986; Emlen, 1991; Lynch, 1991;Edmands,2007).Elementsofbothmodelscanbeatwork simultaneously. Both models assume that genetic com- positions of populations diverge from a combination of random drift and local adaptation. The locus-by-locusmodel predicts that hybrids of populations adapted todifferent environments would be most fit in a (hypothet-ical) intermediate environment and would probably havediminished fitness in either parental environment. Thelocus-by-locus model also predicts that outbreedingdepression would occur in the first and later generationsand persist until selection reduces deleterious alleles. Theepistatic model predicts outbreeding depression that results from disruption of coadapted complexes of epis-tatically acting genes, which result from local adaptation(Shields, 1982; Templeton, 1986; Emlen, 1991; Lynch,1991).BecauseF 1 individualspossesscompletecopiesof each parent's coadapted genome, outbreeding depressionmay not occur in the first generation, and heterosis mayoccur. However, as a consequence of independent assortment and recombination of genes during meiosis,fitness losseswould beexpected inthe F 2 and subsequent generations. The number, type, and degree of differenceand extent of linkage between exchanged genes woulddetermine the magnitude of fitness reduction (Shields,1982).We are using pink salmon as a model for studies of local adaptation and outbreeding depression in wildvertebrate populations. In our experiments, we conduct crosses in late summer from wild fish and raise embryosin incubators before marking and releasing them thefollowing spring. After release, they emigrate from their natal stream to sea. They mature and return to spawnabout 15 months later. Therefore, most of their life cycleis in the wild and prerelease rearing is at ambient streamtemperatures. Previously, we observed reduced marinesurvival in F 2 hybrids between spatially separated allo- patric populations of the same pink salmon broodyear (Gilk et al., 2004), which are consistent with geneticoutbreeding depression. Reduced marine survival wasalso observed in F 1 hybrids of the odd broodline of spatially separated populations (Gilk et al., 2004).Studies of outbreeding depression in vertebrate pop-ulations are not abundant, and little effort has beendevoted to determine the biological mechanisms that may lead to reduced fitness in vertebrates. One of the potential mechanisms may relate to timing. Timings of adult spawning runs, development, and emigration aregenerally important components in local adaptation of a population. Alterations in timing have the potential toreduce survivals substantially. For example, develop-menttime – thetimebetweenfertilizationandhatching – is important in local adaptation of a population andis presumably under selection by seasonally variablestream flows and incubation temperatures (Hebert et al.,1998). Early- and late-run pink salmon from the samestreamusuallyexperiencedifferenttemperatureregimes.However, development times of early- and late-run fishdiffer even when their embryos are incubated under thesame thermal regime (Goddard, 1995; Hebert et al.,1998). Genotype-by-environment effects were alsodetected in experiments in which both early- and late-runtemperatureregimeswereusedtoincubateearly-andlate-run-embryos, which suggest that epistasis may playa role in outbreeding depression of development time inPacific salmon (Goddard, 1995; Hebert et al., 1998).Our objective in this researchwas to evaluate the effect of hybridization on development timing in pink salmon populations. We compared development time of thehybrids between two spatially separated populations withdevelopment time of the controls (1) to test hypotheticaldifferences betweenthe populations of pink salmon,(2) toevaluatethepossibilitythatoutbreedingdepressionmayberelatedtodevelopmentratesofhybrids,and(3)toexaminethe genetic mechanisms that control development timingin hybrids. To do this, we examined the accumulation of temperature units between fertilization and hatching(ATUstohatch)inthesamecrossesforwhichoutbreedingdepression of survival during the marine phase of the lifecycle has been documented (Gilk et al., 2004).Our focus population was Auke Creek (near Juneau,Alaska), and the source of semen for hybrids produced at Auke Creek was Pillar Creek (Kodiak Island, Alaska).The streams are at nearly the same latitude on the coast of the Gulf of Alaska, but about 1000 km apart; there islittledirect genetic exchange and the populations probablyevolved independently since the streams were colonizedfollowing the recent glaciation. The spawning seasons of the two populations coincide from late August to earlySeptember.Wereplicatedtheobservationsintheindependentodd-and even-year pink salmon broodlines. We compared theATUs to hatch in F 1 hybrid and control crosses, in F 2 hybrid and control crosses, in the two types of back-crosses, and between the backcrosses and F 2 hybrid andcontrol crosses. We analyzed the variation of ATUs tohatch with a nonparametric resampling program, whichallowed us to explore possible genetic mechanisms for S153 I.A. Wang et al. / Aquaculture 272S1 (2007) S152 – S160 observeddifferences.WealsocomparedATUstohatchof first-generationhybridstothoseofcontrolsatPillarCreek. 2. Materials and methods 2.1. Source populations Because of their two-year semelparous life cycle, twogenetically isolated broodlines of pink salmon (even and oddyears) occur in many streams. Consequently, we conducted two parallel, independent experiments at Auke Creek, AK (58.38°N,134.64°W) each of which included two generations; one usedeven-broodline and the other used odd-broodline fish. A thirdexperiment was conducted at Pillar Creek, AK Kodiak Island(57.81°N, 152.42°W) in which we observed only the first generation. Pillar Creek is 1048 km (great circle distance) fromAuke Creek.Population sizes at Auke Creek range between 2000 and30,000 (data on file, US NOAA Fisheries, Auke Bay Laboratory,Juneau, AK 99801) and at Pillar Creek between 1000 and 40,000(Vining and Brennan, 2001). Auke Creek is a high gradient, lake-fed stream that is 350 m long; Pillar Creek is a low gradient,reservoir-fedstreamthatisabout1800mlong.Occasionally,Pillar (Vining and Brennan, 2001) and Auke (Fukushima and Smoker, 1997) creeks are dry in July or August. Between August and November — the typical spawning times of spawning andhatching (Fig. 1), Pillar Creek is, on average, colder, than AukeCreekby1to2°(Celsius).AresearchhatcheryisoperatedatAukeCreekbyUS NOAAFisheries; afisheryenhancementhatchery isoperated at Pillar Creek by Kodiak Regional AquacultureAssociation. 2.2. F 1 matings and incubation At Auke Creek Hatchery on both 29 August 1996 and 2September 1997, we used 40 randomly chosen Auke Creek females to make F 1 crosses (Gilk et al., 2004). The F 1 controlswere produced from 40 Auke Creek males, and F 1 hybridswere produced from 40 Pillar Creek males. Semen from Pillar Creek males was taken the day before crosses were made,whereas semen from Auke Creek males was taken on the dayof mating. We used a 2×2 replicated-incomplete-factorialmating design. Each of two Pillar Creek males and two AukeCreek males were crossed with two Auke Creek females to produce a block of eight families; twenty such blocks wereformed from other breeders to produce 80 control and 80hybrid families. The design was intended for quantitativegenetic analyses not reported here. Control and hybrid fishwere incubated in separate FAL ™ (Marisource, Milton, WA)vertical incubator cabinets, 16 trays per cabinet; each tray wasdivided into 10 compartments. Zygotes from each family weredivided into two approximately equal replicate groups, whichwere assigned randomly to compartments in the incubators.Both hybrid and control incubators received Auke Creek Hatcherywaterfromthesamepipe.Waterflowsforeachcabinet were approximately 8 l/min before the eyed stage (whenunfertilizedeggswereremoved)and23l/minthereafter.Loadingdensities were approximately 2000 eggs/tray; most compart-ments were sparsely populated — at most two layers of eggs.Watertemperatureswererecordeddailytothenearest0.1°C.Wetreated the incubating eggs once a week with formaldehyde(1:6000 in static water) for one-hour periods before hatching toreduce fungus and hydra infestations. There were no obviousdifferences in pre-hatching mortality rates between crosseswithin an experiment (data not shown).Hybrid and control fry were marked with different pelvic finexcisions and released to Auke Creek at high tide after dark near the peak of natural pink salmon emigration from Auke Creek between mid-April and early May. Pelvic fin clips assigned tohybrid and control crosses were reversed in the odd- and even-year experiments.At Pillar Creek Hatchery on 6 September 2001, we crossed20 Pillar Creek females with 20 Pillar Creek males to make F 1 control crosses; and we fertilized the eggs from the same 20Pillar Creek females with semen from 20 Auke Creek males tomake F 1 hybrid crosses. Crosses were single-pair matings.Semen from Auke Creek males was collected the day beforeuse, and semen from Pillar Creek males was taken on the dayof mating. The embryos were incubated in one FAL ™ verticalincubator cabinet with four trays, each of which was dividedinto 10 compartments. Two trays held hybrids and two heldcontrols; each family had its own compartment. All familieswere incubated with Pillar Creek Hatchery water, and water temperatures were recorded daily to the nearest 0.1 °C. Wetreated the incubating eggs twice a week with formaldehyde prior to hatching. One of the males used in the Pillar Creek Hatchery experiment was infertile. Fish from the Pillar Creek Hatchery experiment were not released. 2.3. F 2 matings and incubation Maturing F 1 adults were collected as they returned to theAuke Creek Hatchery weir and sorted into four enclosures inthe stream according to fin clip (experimental group) and sex,where they were held for several days. Matings were made on Fig. 1. Average water temperature in Auke Creek Hatchery fromfertilizationtohatchinginfouryearsofexperiments(dashedline)andthehistorical averages from Auke (solid line and Pillar (dotted line) creeks.S154 I.A. Wang et al. / Aquaculture 272S1 (2007) S152 – S160 two occasions 4 or 5 days apart. Because the mating protocolused parents with similar return date and maturation dates, parent choice may have been biased with respect to return andmaturation dates because some early returning fish were not mature on the first spawn date and died before the secondspawn date. Consequently, the broodstock used to produce theF 2 crosses on a given spawn date may not have been acompletely random sample of the F 1 adults with respect toreturn timing.Onboth28August1998and1September1998fortheeven- broodline experiment and on both 4 September 1999 and8September1999fortheodd-broodlineexperiment,wemadeF 2 crosses from mature F 1 adults. On each day, we used a 2×2replicated-incomplete-factorialmatingdesigntomakeF 2 controlcrosses [control females (C ♀ )×control males (C ♂ )] and F 2 hybrid crosses [hybrid females (H ♀ )×hybrid males (H ♂ )] withtwenty each of control females, control males, hybrid females,and hybrid males. From gametes of those same individuals, wealso made both types of backcrosses: hybrid females×controlmales (H ♀ ×C ♂ ) and control females×hybrid males (C ♀ ×H ♂ )on both spawning dates in each year.The same procedures were used to incubate the F 2 crosses aswereusedfortheF 1 crosses,exceptthatfourcabinetsof16dividedtrayswererequired.Asbefore,eachtypeofF 2 crossandbackcrosswas assigned to a separate incubation cabinet. Embryos producedonthefirstspawningdateswereassignedtothelowerhalfoftheir cabinets and embryos made on the second spawning dates wereassigned to the upper half of their cabinets (an assignment that reduced any possible bias toward early hatching in younger embryosexposedtotheproductsofhatchingfromolderembryos.)Temperatures did not differ substantially within or betweenincubation cabinets throughout the experiment. For example, in broodyear 1998 temperatures in the tops and the bottoms of thecabinets on any day differed by 0.1 °C at most and cumulativetemperature units were 168.1±0.1 °C in all four cabinets for the2 months prior to hatching. There were no obvious differences in pre-hatching mortality between crosses within an experiment. 2.4. F 1 and F 2 observations We determined the midhatch date, defined as the first dayon which 50% or more of the eggs in a cell had hatched, for Fig. 2. Expected accumulated temperature units (ATUs) required for backcrosses, relative to F 2 controls and F 2 hybrids, to reach midhatch for twoalternative models of inheritance: locus-by-locus and epistatic outbreeding depressive effects.Table 1Bootstrap analyses (20,000 iterations) of differences in accumulated temperature units (ATUs) to reach various stages of hatching between F 1 hybrid(H) and F 1 control (C) crosses and between F 2 H ♀ ×H ♂ and C ♀ ×C ♂ crossesBroodyear Generation a Spawn date Beginningof midhatchdifference b Median of midhatchdifference c Ending of midhatchdifference d Average temperaturefertilization to end of hatching (°C)ATUs to midhatchfor median controlor C×C familyATUs to midhatchfor median hybridor H×H familyEven F 1 29 Aug 96 9 21 18 6.9 607 628Odd F 1 3 Sep 97 15 19 24 7.6 650 669Even F 2 28 Aug 98 5 15 19 6.7 561 576Even F 2 1 Sep 98 19 11 6 6.4 552 563Odd F 2 4 Sep 99 15 10 14 7.1 630 640Odd F 2 9 Sep 99 19 13 14 6.8 571 584 a For all spawn dates F 1 hybrids in the Auke Creek Hatchery experiment required more ATUs to reach median midhatch than F 1 controls( P ≪ 0.001, all 20,000 iterations) and F 2 H ♀ ×H ♂ crosses required more ATUs to each median midhatch than C ♀ ×C ♂ crosses. b Differences in ATUs between hybrids and controls when the first families reached midhatch. c Differences in ATUs between hybrids and controls when the median families reached midhatch. d Differences in ATUs between hybrids and controls when the last families reached midhatch.S155 I.A. Wang et al. / Aquaculture 272S1 (2007) S152 – S160 each replicate group of eggs. Beginning before hatching began(early November) and continuing until hatching ended (midDecember), we made daily observations of the number of eggsthat had hatched in each incubator cell and estimated the proportion to the nearest 10%. The midhatch dates of F 1 crosses made at Pillar Creek were similarly recorded. 2.5. F 1 analysis The mating design included correlated half-sib families inthe2×2factorialblockdesign.Toreducethedatatosinglefull-sib families for analysis, we conducted a simulation that ran-domly resampled midhatch data from each factorial block of eightfamilies.Wealsochoserandomlyonlyoneofthereplicatesfor each family. Specifically, we randomly reduced each block to two pairs of families, one control and one hybrid family that shared one of the females in the block and a second pair of families, one control and one hybrid that shared the secondfemale.Wecalculatedthedifferencesinaccumulatedtemperatureunits (ATUs, sum of daily water temperatures) between the daysoffertilizationandmidhatchbetweenhybridandcontrolfamiliesofeachpair.Foreachcross,weproduced20,000bootstrapmeansfromwhichwecalculatedtheproportionoftimes(outof20,000)thatthehatchingtimesofthehybridsexceededthehatchingtimesofthecontrols.Becauseeachiterationofthisanalysisusedonlyaquarter of the data, the analysis is conservative.For the few incubator cells in which all fish died beforehatching, midhatch data from families within a block weresubstituted in one of two ways. First, if one replicate of a familydied, the hatching data from the other replicate was substitutedfor the blank replicate; and second, if neither replicate from afamily survived to hatching, the data from the other male withinthecrosswassubstituted.Whenbothfamilieswithinacrossforagiven female died before hatching, the entire block of eight familieswasremovedfromtheanalysis.Inbroodyears1996and1997, we substituted hatching data for families incurring 100%mortality before hatching in fewer than 2% of the families.For the crosses evaluated at Pillar Creek Hatchery, we used aMann – Whitney Rank Sum U-Test (SigmaStat, SPSS Science,Chicago,IL)tocomparetheATUsatmidhatchforthetwogroups because assumptions of normality and equal variance were not met. We removed from the analysis one family from the Pillar Creek Hatchery experiment that experienced complete mortality before hatching. 2.6. F 2 analysis The F 2 crosses and backcrosses provided an opportunity toevaluate the extent to which locus-by-locus accumulation of effects (as opposed to epistatic interactions) expresses thetiming of midhatch as a phenotype. We used the sameresampling methods described for the F 1 crosses to analyzethe midhatch times of F 2 crosses, but extended them to includethe backcrosses. Control fish carried only Auke Creek genes,F 1 hybrid fish carried one Auke Creek and one Pillar Creek gene at each locus, an average of half of the genes in F 2 fishwere from Auke Creek (1/4 of loci would be homozygous and1/2 heterozygous), and 3/4 of the genes (at least one per locus)were in backcrosses. If development timing resulted only fromlocus-by-locuseffects,thephenotypeofthebackcrosses would be intermediate to the F 2 hybrid and control crosses. If de-velopment timing in the source populations (Auke Creek controls)weredueinparttoepistasis,onewouldexpecttoseealoss of that influence in the F 2 hybrids because both coadaptedAuke and Pillar creek coadapted genomes would be disrupted Fig. 3. Cumulative proportion of families reaching midhatch accordingto accumulated temperature units (ATUs) in F 1 hybrid (AukeCr. ♀ ×Pillar Cr. ♂ ) and control (Auke Cr. ♀ ×Auke Cr. ♂ ) crosses ineven- and odd-year pink salmon broodlines.Fig. 4. Cumulative proportion of families reaching midhatch accordingto accumulated temperature units (ATUs) in F 1 hybrid (Pillar Cr. ♀ ×Auke Cr. ♂ ) and control (Pillar Cr. ♀ ×Pillar Cr. ♂ ) crosses in 2001.S156 I.A. Wang et al. / Aquaculture 272S1 (2007) S152 – S160
