GENETICS AND BROODSTOCK MANAGEMENT OF COHO SALMON

 

James M. Myers1, Per O. Heggelund2, Greg Hudson3, and Robert N. Iwamoto1

 

1Northwest Fisheries Science Center

2725 Montlake Boulevard East
Seattle, Washington 98112

 

2AquaSeed Corporation

4530 Union Bay Place NE

Seattle, Washington 98105

 

3Domsea Broodstock, Inc.

10420-A 173rd Avenue SW

Rochester, Washington 98579

 

. 

ABSTRACT

 

The success of finfish broodstock operations for both aquaculture and captive restoration purposes will depend on accurate prediction and management of the influences of a myriad of genetic effects.  Broodstock development programs for aquacultural species have historically relied on terrestrial models.  As with terrestrial species, much of the early genetic research on finfish focused on estimates of inbreeding depression (through sib mating), estimates of heritabilities from covariance analysis, and estimates of genetic effects through interspecific or interstrain hybridization.  Recent innovations in biotechnology, including transgenic manipulations and isolation of DNA markers, have shown some promise to assist traditional broodstock improvement programs.  This report will first provide a general overview of some basic genetic principles that have proven useful for developing classical broodstock programs and may be equally important for the production of genetically modified (clonal, polyploid, or transgenic) broodstocks.   The remainder of the report will focus on two aspects of broodstock development programs, genotype-environment interactions and inbreeding, and present research data from a multi-generational coho salmon (Oncorhynchus kisutch) selection program.

 Keywords:  coho salmon, selection, genotype-environment interactions, inbreeding   

1.  INTRODUCTION

Broodstock management generally involves the control of two seemingly antagonistic processes:

1)      The accumulation of novel genotypes as a direct outcome of the modification of certain traits (classical selection) or 2) through genetic manipulation and the maintenance of genetic variability through avoiding the loss of alleles or the accumulation of deleterious alleles.  The amount and kind of genetic intervention will depend on whether genetic improvement (e.g. commercial aquaculture crop) or genetic homeostasis (e.g. a captive broodstock program for maintenance of an endangered species) is the principal goal of the program.  In either case, well-defined program goals as well as the initial choice of the foundation stock(s) are critical.  In addition, any broodstock genetics program will have a central breeding plan that is comprised of two principal components: selecting the parents for the next generation (selection) and determining how the selected parents will be mated (mating systems) (Turner and Young 1969).

The design of most broodstock programs is based on several assumptions.  First, a number of studies involving the effects of inbreeding through sib matings predict significant declines in reproduction and growth traits with inbreeding coefficients (DFs, the probability of two  alleles being identical by descent) of 10% or more.  Inbreeding can be estimated through pedigree analysis or through direct experimental measures of changes in genetic variability (allozymes and DNA).  Second, improvements in phenotypic traits through selection are dependent on selection intensity and the heritability (h2, the proportion of  phenotypic variation due to additive genetic variation) of the trait.  Furthermore, changes in the phenotypic characteristic of one trait (both desired and unintended) can occur indirectly through selection on another trait.  The magnitude of these changes is dependent on the genetic correlation between the two traits, the heritability of both traits, and the selection intensity (Falconer 1989).

In most cases, broodstock programs for fish and shellfish have had to rely on models derived from terrestrial organisms; however, as our understanding of the response of aquatic organisms to long-term selection programs accumulates, it may be necessary to reevaluate many of the underlying assumptions used in developing commercial broodstock programs.  Furthermore, the success of captive broodstock programs to preserve the genetic characteristics of a natural population may be limited by the magnitude of inadvertent genetic changes due to artificial propagation (e.g. domestication) (Hard and Hershberger 1995).  Finally, the development of broodstock programs for genetically manipulated or transgenic aquatic organisms does not benefit from the same volume of existing scientific literature as is found for classical selection programs; however, many aspects of the management of genetically modified organisms will be the same as those for classical selection programs.

This report will focus on two of the issues involved with broodstock genetics:  genotype-environment interactions (GxE) and inbreeding.  It will draw principally from the growing quantitative genetics information from a long-term (13 generations) coho salmon (Oncorhynchus kisutch) breeding program.  This report proposes that GxE interactions and inbreeding are universal issues of concern in the development of a broodstock regardless of the program goal or type of genetic manipulation or management employed.

1.1  BACKGROUND

1.1.1.  GENOTYPE-ENVIRONMENT INTERACTIONS

Many traits have continuous rather than discrete distributions and are under the control of many genes with small, independent effects.  Ultimate expression of the trait will depend on the form and nature of that genetic control as well as a composite of influencing environmental factors.  Estimating the relative contributions of genotype and environment is largely accomplished through the analysis of variance and subsequent variance component analyses (Falconer 1989).  Certain combinations of genotype and environment may not yield a phenotypic value that is equal to the sum of the two factors.  In those cases, a third function must be added, which changes the equation for phenotypic expression to

            P = G + E + f(GE).                                                                

GxE interactions may be manifested either as changes in magnitude of the differences among the genotypes or as changes in ranking of the different genotypes.  It is the latter type of interactions that are biologically important.  Furthermore, significant interactions are more likely to be detected when environmental differences are large rather than small.  Genotype differences, on the other hand, may be large or small (Hohenboken 1985).

The accumulating evidence from salmonid research indicates that, because of the difficulty in predicting the presence and magnitude of interactions a priori, GxE interactions should be considered when developing selection programs.  Iwamoto et al. (1984) and Heath et al. (1994) detected significant interactions among families and two rearing temperatures in the incidence of precocial maturation in coho salmon and chinook salmon (O. tshawytscha), respectively.  Similarly, Wild et al. (1994) reported significant interactions for early sexual maturity among Atlantic salmon (Salmo salar) families and different net-cage sites in Norway.  Balfry et al. (1997) found that while resistance of chinook salmon to vibriosis has a significant genetic component, there were no significant GxE interactions for families grown in two net-pen environments.  Significant GxE interactions (but without changes in rank) were reported by Hanke et al. (1989) for growth of underyearling Atlantic salmon grown under different photoperiod treatments.  Wangila and Dick (1988) found that specific growth rate of two rainbow trout (O. mykiss) strains and their hybrids were differentially affected by differences in the rearing temperature.  Beacham (1987) determined that families of chum salmon (O. keta) fry reared in a higher temperature rearing environment and a lower temperature freshwater rearing environment showed changes in rank in body weight after 196 days of rearing.  Iwamoto et al. (1986) determined that GxE interactions, while significant, were not biologically important for growth performance of three rainbow trout strains and their hybrids under several density and ration treatment levels.  Finally, significant GxE interactions including rank changes were reported by Iwamoto (1982) for growth and seawater survival of two strains of coho salmon and their hybrids reared under four different rearing temperatures and three photoperiod treatments.

In natural salmonid populations, local adaptation is viewed as an evolutionarily important form of GxE (Taylor 1991).  For captively reared populations, the magnitude of GxE interactions and the resulting potential for selection effects, such as domestication, may be a major concern.  For the commercial fish culturist, strong GxE effects may determine the general utility of a broodstock under a variety of culture conditions.  Furthermore, the accuracy of heritability estimates and selection programs derived under certain conditions may be minimized under different culture conditions if GxE effects are significant. 

Under culture conditions, if the performance of a given genotype is fairly uniform under a variety of environmental conditions (small GxE interaction), such a strain could be used as a general-purpose broodstock and be expected to perform moderately well under a range of conditions.  On the other hand, large GxE interactions suggest the need for special purpose broodstocks for each environmental condition or anticipated change in the environment.  Consequently, measurement of the magnitude of these responses will be particularly important in defining the range of the utilization of specific broodstocks in the industry.

Conservation programs that utilize captive broodstocks need to assess the risk that captive culture may directly select for traits that lower the fitness of fish under natural conditions or simply relax selection for genotypes that are adapted to the native ecosystem for that population.

1.1.2.  INBREEDING

Inbreeding is defined as the probability of two alleles in an individual being identical by descent, and is normally the result of mating related individuals.  The rate of inbreeding is a function of the characteristics of the foundation stock as well as limited population sizes in subsequent generations (Falconer 1989).  The deleterious effects of inbreeding have been documented in aquatic species, but there has been particular emphasis on salmonids (Aulstad and Kittelsen 1971; Kincaid 1976, 1983, and 1995; Gjerde et al. 1983; Su et al. 1996). The majority of these studies produced relatively high inbreeding levels (∆F=10-25%) through sib matings.  Hershberger et al. (1990a) analyzed the growth performance of coho salmon under selection and increasing levels of inbreeding. Despite accumulated inbreeding levels after four generations approaching those of full-sib mating, there was no apparent decrease in growth performance.  Whether selection gains masked deleterious effects or the accumulation of inbreeding levels over several generations does not result in the same deleterious effects as has been reported for closely-related (sib) matings was not determined.

2.  MATERIALS AND METHODS

2.1  A CASE STUDY: THE DOMSEA COHO SALMON BROODSTOCK SELECTION PROGRAM

A selection and breeding program began in 1977 by the University of Washington, the Washington Sea Grant Program, and Domsea Farms, Inc. to develop coho salmon broodstock for the marine net-pen industry in the state of Washington.  The founding population had been derived from the Washington Department of Fish and Wildlife’s Skykomish Hatchery, Skykomish, WA in 1971 and 1972, and had been subjected to three generations of mass selection for growth (by mating several hundred adults) before the initiation of the Domsea Farms broodstock program (Novotny 1975).  A selection scheme, raising 40 families of 600 individuals, was designed to yield maximum response and to be useful under commercial culture conditions.  This scheme involved several different types of concurrent selection (e.g., family and individual) and used a selection index that incorporated estimates for heritability, relative economic values, genetic correlations, and mean values on all the traits of interest (survival and growth rate) (Iwamoto et al. 1982, Saxton et al. 1984, Hershberger et al. 1990b).  Breeding was conducted by a circular mating procedure to minimize the possibility of crossing closely related families while assortatively mating the top performing families.  On a theoretical basis, these steps should limit the change in inbreeding to about 2% per generation (Hershberger and Iwamoto 1984).  Because of the 2-year life cycle of coho salmon under captive culture, odd- and even-year spawning broodstock were developed as independent broodlines.  Initially, odd- and even-year lines were founded independently; however, in 1991 a catastrophic accident at the rearing facilities eliminated the even-year line.  A new even-year line was produced from three-year old odd-year line spawners.  As of 1997, the odd- and even-year lines have been separated for three generations.

As part of the selection program, a number of reproductive traits (spawner weight and length, fecundity, egg size (at the eyed stage), survival to the eyed stage, and survival to ponding) and growth traits (length and weight at 1800 temperature units (T.U.s) post ponding, and weight after 3.5 and 8 months of saltwater rearing) were monitored.  In 1986, the program was modified to rear broodstock in freshwater throughout their life cycle.  This change in rearing strategy substantially reduced the mortality normally experienced following transfer to salt water and during the summer months prior to maturation.  The timing of the measurements previously made in salt water was adjusted to reflect the different temperature regime experienced by the all-freshwater reared broodstock. 

2.1.1  GENOTYPE-ENVIRONMENT INTERACTIONS

To evaluate the possible consequences of maintaining broodstocks in freshwater relative to salt water, randomly selected portions of each full-sib family were reared in each environment after the initial 7-month freshwater common rearing period.  In 1985, the full-sib families were apportioned to a saltwater net-pen and to land-based circular tanks with a well-water source of approximately 100C constant temperature.  They were reared under those conditions for the subsequent 18 months to maturation and spawning.  In 1986, in addition to the two environments mentioned above, the full-sib families from that brood-year were also reared at an additional freshwater location in 13.50C spring water.  In both cases, individuals from each of the full-sib families were measured for length at approximately 8 months post-transfer adjusted for the differences in rearing temperatures.   Weight was estimated using the length data, and a length-weight regression was generated for each site by subsampling the population at each location/environment.  Because the same regression equation was applied to all families within a location, genetically influenced differences for weight among families as well as GxE interactions were biased to some unknown degree.

Four statistical methods were used for measuring GxE interactions, each with a basis in analysis of variance and variance component analysis: 1) statistical significance of the interaction term; 2) percentage of total variation contributed by each interaction term; 3) the ratio of the interaction term to the sum of its component parts; and 4) the genetic correlation approach (Robertson 1959).

The analysis of variance model used in tests for significance and variance component partitioning was:

            Yijk = u + Gi + Ej  + GEij  + 0ijk

            where

u is the overall mean;

                        Gi  is the effect of the ith family;

                        Ej   is the effect of the jth environment;

                                GEij  is the effect of the ith family and the jth environment;

0ijk is the deviation of individual k from the mean of the family-environment subgroup

Statistical significance (P < .05) was determined for each main effect and the interaction term (mixed effects model where families were random and environment was fixed).  Analysis of variance tests were run using Statview 5.1 (SAS Institute, Cary, NC, USA).  Once variance components were partitioned, the percentage of each component part to the total variance was determined.  Two ratio estimates (1:  VGE/VTotal; 2:  VGE/VG +VE+VGE) were examined as arbitrary determinants of the importance of GxE interactions in each case (VGE/VTotal  > 5%; VGE/(VG +VE+VGE ) > 10%).   Variance components were further used to calculate genetic correlations (rG; where rG = VG / (VG + VGE).  The genetic correlation is a measure of whether genetic control of performance  in one environment is similar to that in a second environment.  Generally, if rG is < 0.8, GxE interactions are considered biologically and practically important (Robertson 1959).

The significance of family rank change in different rearing environments for BY1985(F6BR73) and 1986(F6BR74) was estimated using the Mann-Whitney Test (Zar 1974).

2.1.2  INBREEDING

Estimates of inbreeding (Wright’s inbreeding coefficient) for each of the families sampled were computed from the family pedigree using the CompuPed v4.0 program (RCI Software, Loveland, CO, USA).  The inbreeding level of the founding generation, F3BR73 (Fig 3), was assumed to be 0.00 for the purposes of the calculation. Estimates of inbreeding levels under a random mating scenario (excluding sib matings) were derived using the equation  ∆F=1/(2N+4) for each generation (Falconer 1989).

In 1986, to test the effects of intensive inbreeding (brother-sister mating) fifth generation females (F= 0.059) from one family were crossed with full-sib males or outcrossed with males from unrelated families.  Individuals from inbred and outbred families were weighed after 2,000 degree-days of rearing.

2.1.2.1  Allozyme Analysis

For the current study, 159 and 165 juveniles representing the 40 full-sib families from BY1997 and BY1998, respectively, were examined for allozyme variation using methods outlined by Aebersold et al. (1987).  The following 66 loci were resolved (locus nomenclature follows Shaklee et al. 1990):  AAT12**, sAAT3**, sAAT4**, ADA1**, ADA2**, mAH1**, mAH2**, mAH3**, sAH**, mAAT1**, AK**, FBLD3**, FBLD4**, ALAT**, CKA1**, CKA2**, CKC1**, CKC2**, CKB**, EST1**, EST4**, FH**, bGLUA**, bGALA**, GAPD3**, GAPD2**, GAPD4**, GAPD5**, PEPA**, PEPC**, GPIB1**, GPIB2**, GPIA**, GR**, HAGH**, mIDH1**, mIDH2**, sIDH1**, sIDH2**, LDHA1**, LDHA2**, LDHB1**, LDHB2**, LDHC**, PEPB1**, PEPLT**, aMAN**, MDA12**, MDB12**, mMDH2**, mMDH3**, MPI**, PNP1**, PNP2**, PGDH**, PGK1**,PGK2**, PGM1**, PGM2**, PEPD2**, PK2**, IDDH1**, sSOD1**, TPI1**, TPI2**, TPI4**, and TPI3.

Since no baseline exists for the genetic content of the founding population, the results of this analysis were compared to existing data from five coho salmon populations (Little Pilchuck (N=120), Harris Creek (N=120), Grizzly Creek BY91 (N=100), Grizzly Creek BY93 (N=100), and Lewis Creek (N=67)) which are found in the same watershed (Snohomish and Skykomish River Basins) as the founding population for the Domsea Farms broodstock.  Allozyme data from the populations was analyzed with the BIOSYS computer program (Swofford and Selander 1981).

2.1.2.2  Transferrin

Electrophoretic analyses were conducted on serum samples from 100-120 adult fish in each of four years (1977, 1978, 1985, and 1986).  The electrophoretic procedures employed were those reported in Utter et al. (1970) for analysis of serum transferrins in coho salmon. For the BY1997 and BY1998, 159 and 165 juveniles representing the 40 full-sib families were analyzed according to Van Doornik et al. (1995).

3.  RESULTS

3.1  HERITABILITY ESTIMATES AND RESPONSE TO SELECTION

Hershberger et al. (1990b) summarized the genetic estimates and performance of the broodstock after four generations of selection.  They reported significant gains in body weight for both the odd-year and even-year broodstocks for the 8-month saltwater monitoring period (239.0g to 432.5g for the odd-year and 296.2 to 666.7g for the even-year line) reflecting an average gain of 10.1% per generation (Figure 1).   Heritability estimates derived from sib analysis for length and weight ranged between 0.18 to 0.33 and showed minor change over the four generations of intensive selection (Table 1).  An internal control line also showed increases in length and weight gain over the time period suggesting that domestication selection (Gjedrem 1979, Doyle 1983) had occurred. 

3.2  GENOTYPE-ENVIRONMENT INTERACTIONS

Family (G) and Environment (E) effects were significant for length in BY1985 (Table 2). The GxE interaction was statistically significant (P<0.01), indicating that the relative growth of families in saltwater and freshwater rearing affected growth of each family differently.  All variance component ratio estimates (VGE/Vtotal = 5.5% and VGE/(VG + VE + VGE = 25.6%) as well as the genetic correlation coefficient (rg = .532) indicated that GxE interactions were important.

For the 1986 broodstock, the splitting and subsequent rearing of full-sib families in the freshwater and saltwater environments mentioned above plus an additional freshwater environment permitted several additional comparisons of the data (Table 2).  When length data from all three environments were combined, the two main effects (families and environment) and the interaction term were all statistically significant and comprised significant portions of the total variation (.097, .136, and .051, respectively).   Moreover, the variance component ratio estimates as well as the genetic correlation coefficient all exceeded the criteria indicating the biological and practical importance of GxE interactions.  The average growth performance at each of the sites varied between years (Table 2).  In 1985 fish reared at the freshwater site were heavier, while during 1986, fish at the saltwater site were larger . For the 1985 weight data, the environment main effect was not significant and comprised just 0.1% of the total variation compared with 9.8% of the total variation for the corresponding length data.   All variance component ratio estimates exceeded the arbitrary significance criteria.  In 1986, the same two-environment comparison indicated a significant environment effect, comprising 15.3% of the total variation.   The three-environment analysis (comparing one saltwater and two freshwater sites) yielded an even stronger environmental effect , accounting for 58.2% of the total variation.

Comparisons of the relative ranking of families at each of the rearing sites during each of the broodyears also revealed significant GxE interactions (Table 3).  There were considerable changes in rank among families reared at different sites.  On average, only about half of the top 15 families from any one site were among the top 15 families at the other sites.  Furthermore, the correlation in family ranks between the two freshwater sites used in rearing the BY1986 families was substantially greater (r = 0.714) than that between either of the freshwater sites and the saltwater site (r = 0.187 and 0.070).  This would suggest (although only one year of data is available) that differences in freshwater and saltwater rearing had a substantial effect on relative family performance.

Finally, genotype x environment interactions were also observed when heritabilities were computed separately for each of the environments in each of the broodyears.  In BY 1985, under normal rearing conditions, similar estimates of heritability, around 0.20, for length and weight were derived for both fresh and salt water.  However, in BY 1986, heritability estimates for length and weight ranged from 0.35 to 0.55 for both freshwater rearing sites, but only from 0.05 to 0.06 for length and weight at the saltwater site.

3.3  INBREEDING

Genetic improvement has been rapid and significant due in part to the intense selection imposed upon the population every generation.   Program design created significant genetic bottlenecks in every generation, with six males and six females from each of 10 full-sib families serving as the nucleus for the succeeding generation.  Because of the limited numbers of spawners per generation, the mating of close relatives became inevitable.  By the fourth generation of the selection program (BY1983), all individuals were related to some degree.

Analysis of the pedigree records for the even- and odd-year broodlines (Figs. 2 and 3) determined that inbreeding coefficients of 16.30% and 14.95% had accumulated in the odd- and even-year broodlines, respectively after 9 and 10 generations of selection (Table 4).  Based solely on population size and excluding self- and sib-mating, the inbreeding coefficients after 9 (even-year) and 10 (odd-year) generations were estimated to be 21.88% and 22.80%, respectively.  Although the mating design was successful in reducing the rate of inbreeding accumulation, both estimates are well within the levels expected to result in inbreeding depression (Gall 1987).  

Furthermore, allozyme analysis indicated that although overall heterozygosity levels were equivalent to natural populations from the watershed where the Domsea stock originated, there was a loss in variability at more than 70% of the loci found to be variable in the natural stocks (Table 5).   At 6 of the 30 variable loci, the frequency of rare alleles was actually higher in the two Domsea lines than in the natural populations.  Given that the two Domsea lines diverged from a common stock in 1991 (with 3 generations of subsequent isolation) the similarity in allozyme variability between the two Domsea lines may indicate that gene frequencies have stabilized in recent years and that much of the divergence from the founding natural populations occurred earlier in the selection program.

Transferrin samples from Domsea coho salmon analyzed in 1977 and 1985 Hershberger et al. 1990a) provide a baseline for analyzing genetic changes.  Samples taken in 1998 and 1999 indicate that there had been a substantial shift in transferrin allelic frequencies, with the complete loss of the “A” allele in both odd- and even-year lines (Table 6).  Furthermore, the decrease in the frequency of the “A” allele was similar in the period from 1977-1985 and from 1985 to the present.  It is unknown whether the loss of the “A” allele occurred before or after the divergence of the two lines in 1991. 

While the general population failed to show any significant inbreeding effects after five generations of selection, the results of experimental intensive inbreeding in 1986, on the other hand, indicated significant growth depression in the resultant progeny.  Progeny from the highly inbred full-sib matings (F = 0.265) weighed a significant 25% less than their out-bred half-sibs (Table 7).

4.  DISCUSSION

Historically, strain selection has followed the principle that strains have presumably adapted to local environmental conditions and will demonstrate optimal fitness within those same environmental parameters (Namkoong 1979).  However, as Levins (1968) has suggested, the overall fitness of a population may be enhanced, not by specific adaptation to a restricted range of environmental circumstances, but by general adaptability to a wider array of conditions.  Under this argument, then, the long-term survival of a population may well depend on average performance under many environments.  This point may be especially relevant where environmental conditions are variable, and changes are largely unpredictable.

The homing fidelity exhibited by salmonids provides an important mechanism, the creation of discrete localized breeding populations, for GxE interactions to occur.  Differences in the size and degree of isolation (geographic separation of spawning populations, ecological distinctiveness, and homing fidelity) of each population will determine the potential for local adaptation.  A further consequence of the establishment of discrete populations is the potential for the accumulation of inbreeding.  However, the success of the salmonid life history strategy would suggest that the deleterious effects of inbreeding are not significant under natural conditions.  There may be sufficient gene flow between populations, the effective size of most populations may be sufficiently large to minimize inbreeding effects, or salmonids may be able to tolerate inbreeding levels if accumulated at a slow rate (perhaps due to their tetraploid genome). 

Although local adaptation as a form of GxE appears to be a significant factor in natural populations, it is less well understood how captive culture (as one type of environment) influences the selection or expression of commercially important traits or traits that are important in the overall fitness of the organism under natural conditions.  Domestication is widely recognized in terrestrial organisms; however, fish being poikilothermic may be much more sensitive to the modifications in their environment created by captive culture (water temperature, flow, and chemical composition).  Indirect evidence of domestication effects has been observed in the relatively poorer survival of hatchery fish relative to their naturally produced counterparts (Reisenbichler and McIntyre 1977, Fleming and Gross 1992, Hard and Hershberger 1995), although there are a number of non-genetic effects which can lower the fitness of hatchery fish (behavior, disease, rearing conditions).  Furthermore, the effect of selection for commercially-important traits on fitness-related traits under natural conditions is not well understood.  Kincaid (1993) cautioned against the use of domestic broodstocks for supplementary stocking of natural fisheries.

For the commercial aquaculturist, the detection of significant GxE effects as in this study suggests that consideration be given to the environment used to rear broodstock (Gjedrem 1992).  Ideally, broodstock-rearing conditions should resemble those used for normal production stock.  For a variety of reasons (disease transmission, high prespawning mortality, slow growth and delayed maturation, etc.) it may not be desirable to subject broodstock to normal rearing conditions.  An alternate strategy would be to rear a portion of each family under production conditions and select broodstock families on the basis of the production performance of their sibs (sib selection).  It should be emphasized that the above mentioned actions are designed to optimize selection gains.   In those years where controls were present, the Domsea select lines consistently outperformed the control lines, regardless of rearing environment.  The existence of significant GxE effects in this study does not necessarily imply that selection gains derived in one environment will not also be expressed, in part or in whole, in a different environment (this may be more true for the range of captive culture conditions than for the range of natural rearing conditions).  Gjerde and Gjedrem  (1984) did not observe significant GxE effects among strains of Atlantic salmon at different net-pen rearing sites; however, the variability of the environments studied may not have been substantial enough to elicit a strain specific response.  Similarly, in rainbow trout (O. mykiss) ,G x E interactions ( as expressed by changes in rank) occurred at the extremes of rearing conditions (very slow or very fast growth), while rank changes were relatively rare under normal commercial culture conditions (Iwamoto et al. 1986).  In this study, growth conditions during BY 1986, especially at the marine net-pen site where growth was exceptional, may have resulted in large GxE effects that would not be expressed under the normal range of rearing conditions.

Differences in the heritabilities that were calculated for the Domsea broodstock reared in different environments further illustrates the importance of considering environmental and GxE effects when interpreting genetic parameters.  A lack of understanding of the magnitude of environmental effects can lead to erroneous or improper genetic interpretations (Gall 1985).

If GxE effects are significant, then it is possible that there has been additional selection pressure on the broodstock.  Alternatively, the culture conditions experienced by the Domsea broodstock (especially during freshwater rearing) may result in the removal of selective forces that operate in natural populations.  Under natural rearing conditions and, even more so, during marine net-pen rearing, most stocks experience high mortality levels at saltwater entry (smoltification).  This potentially provides the opportunity for considerable selection pressure on a number of life history traits (osmoregulation transition, growth, disease resistance, predator avoidance, etc).  Alternatively, there is no specific life history stage during freshwater captive culture that provides a similar opportunity for selection pressure to be exerted by the environment (as exhibited by marked mortalities during any one life history phase).  It still remains to be seen how the changes observed have affected the fitnesses of these fish under their native environmental conditions.

Analysis of theoretical (based on pedigree data) and actual genetic changes (based on allozyme data and changes in selected traits) would suggest that there has been a considerable alteration in the genetic composition of the Domsea coho salmon broodstock.  At roughly three-quarters of the variable loci, rare alleles were completely lost; however, at the remaining variable loci the frequency of rate alleles increased. Average heterozygosity levels  were similar in the Domsea broodstocks relative to the naturally spawning populations that contributed to the Domsea founding population.  Similarly, Winkler et al. (1999) in their electrophoretic examination of two Chilean coho salmon broodstocks noted founder effects as the reason for a reduction in number of polymorphic loci but not in overall heterozygosity relative to other coho salmon stocks.

In the absence of continuous monitoring it is not known whether the changes in allozyme variation noted in the current study are directional (under direct or indirect selection), or if they represent random changes.  Theoretically, allozyme variation is selectively neutral.  Furthermore, the method and physical location of culture has changed considerably over the course of the broodstock program: 1) marine net-pen culture (1973-1985), 2) freshwater culture with warmed groundwater (1986-1990), 3) freshwater culture with unheated groundwater (1991-1999).  In addition, several investigators (Suzumoto et al. 1977, Pratschner 1978, Winter et al. 1980) have suggested that specific transferrin genotypes may provide improved resistance to specific bacterial pathogens including those associated with bacterial kidney disease, vibriosis, coldwater disease, and furunculosis.  The observed shifts in transferrin frequencies may be a product of selection and inbreeding rather than inbreeding alone (Hershberger 1990a).

Phenotypically, the Domsea coho salmon broodstock have demonstrated considerable improvements in the commercially important traits for which they were selected, even in the presence of moderately high inbreeding.  Although there is substantial evidence for the loss of, or changes in, genetic variability, there has been little phenotypic expression of inbreeding depression (Myers et al. 1999).  Measurements of reproductive traits, female spawner weight, egg weight, and survival to ponding, which have been shown to be the most sensitive to increases in the inbreeding coefficients (Gall 1987), have not demonstrated any reduction since the inception of the broodstock.  Female weight has varied considerably, with year-to-year variations most probably due to rearing condition differences.  Survival to ponding, with the one exception of BY1987, has not indicated any significant decrease.  The traits under direct selection, pre-smolt and post-smolt growth and survival continue to show positive response to selection (data not shown).  It is possible that indirect selection on reproductive traits may have more than compensated for any inbreeding depression effects.

The apparent lack of a deleterious response, relative to other intensive inbreeding studies, may indicate that the accumulation of inbreeding levels (as calculated by pedigree analysis) over a number of generations does not result in the same expression of deleterious traits as was observed in sib crosses (Pirchener 1969).  Given the high fecundity of salmonids, there is considerable opportunity for deleterious alleles to be culled from the population without a major reduction in fecundity.  It would be possible to test this hypothesis by creating a number of full-sib crosses, as was done in 1986, from the existing Domsea broodstock and monitoring for changes in traits that normally exhibit inbreeding depression.  If deleterious recessive alleles have continued to be removed from the broodstock, the relative growth rate of the full-sib crosses should show less of an inbreeding depression than was observed in 1986.  Alternatively, naturally occurring coho salmon breeding populations are relatively small in number and coho may have developed physiological mechanisms to deal with low levels of inbreeding.  Relative to other salmon species, coho salmon have relatively low levels of genetic variation (as detected by allozyme analysis).  Furthermore, salmonids have evolved through a least one tetraploid event, and the duplication of genes provides a further buffer against inbreeding effects.

The cumulative past research with the Domsea stock, and more recently, the results in this report, have provided a growing understanding of several fundamental genetic processes that influence growth and survival performance of coho salmon under captive culture.  Similar programs such as with Atlantic salmon, rainbow trout , catfish, tilapia, marine shrimp, and oysters, have expanded the knowledge base for other aquatic species.  Despite species‑related differences, selecting a founding population(s), developing mating schemes, and testing performance are activities that are universal to broodstock programs. It is likely that this generalist approach will suffice even with the growing application of molecular technology innovations.  Biotechnology approaches such as hypervariable genetic markers or mitochondrial DNA for pedigree determination have considerable potential as tools to assist with developing such programs; however, associating specific DNA markers with phenotypic traits (marker-assisted selection), especially with traits under polygenic control, may be considerably more difficult (Ferguson and Danzmann 1998).  While the specific aspects of molecular biotechnology applications to investigate broodstock development and characterization are relatively new, the underlying effects of related genetic manipulation and of assessment methods are fundamentally the same as those found in classical selection programs.

 

5.  REFERENCES

 

Aebersold, P. B., G. A. Winans, D. J. Teel, G. B. Milner, and F. M. Utter. 1987. Manual for starch gel electrophoresis: a method for the detection of genetic variation. NOAA (National Oceanic and Atmospheric Administration) Tech. Rep. NMFS (National Marine Fisheries Service), Seattle, WA, 61pp.

 

Aulstad, D.  and A. Kittelsen.  1971. Abnormal body curvatures of rainbow trout (Salmo gairdneri ) inbred fry. J. Fish. Res. Board. Can., 28: 1918‑1920.

 

Balfry, S.K., D.D. Heath, and G.K. Iwama. 1997. Genetic analysis of lysozyme activity and resistance to vibriosis in farmed chinook salmon, Oncorhynchus tshawytscha (Walbaum). Aquacult. Res. 28(11): 893-899.

 

Beacham, T.D. 1987.  Genotype-environment interactions in growth of chum salmon.  Can.Tech. Rep. Fish. Aquat. Sci., no. 1517, 14pp.

 

Doyle, R.W. 1983.  An approach to the quantitative analysis of domestication selection in aquaculture.  Aquaculture, 33:  167-185.

 

Falconer, D.S. 1989. Introduction to Quantitative Genetics. 3rd Edition. Longman Group Ltd., Essex, U.K. 340pp.

 

Ferguson, M.M., and R.G. Danzmann.  1998.  Role of genetic markers in fisheries and aquaculture: useful tools or stamp collecting?  Can. J. Fish. Aquat. Sci.,  55: 1553‑1563.

 

Fleming, I.A. and  M.R. Gross.  1992. Reproductive behavior of hatchery and wild coho salmon (Oncorhynchus kisutch): does it differ?  Aquaculture, 103: 101-121.

 

Gall, G.A.E. 1985.  Quantitative genetic aspects of reproduction in salmonids.   Things are not always what they seem to be.  Pp 44-51.   In R.N. Iwamoto and S. Sower (eds)  Salmonid Reproduction.  An International Symposium.  Washington Sea Grant Program, Seattle, WA USA. 167 p.

 

Gall, G.A.E. 1987.  Inbreeding.  pp. 47-88.  In:  N. Ryman and F. Utter (eds).  Population Genetics & Fisheries Management.  University of Washington Press, Seattle, WA USA. 

 

Gjerde, B., K. Gunnes, and T. Gjedrem. 1983.  Effect of inbreeding on survival and growth in rainbow trout.  Aquaculture, 34(3-4): 327-332.

 

Gjerde, B., and T. Gjedrem.  1984.  Estimates of phenotypic and genetic parameters for carcass traits in Atlantic salmon and rainbow trout.  Aquaculture 36: 97-110.

 

Gjedrem, T. 1979.  Selection for growth rate and domestication in Atlantic salmon. Z. Teirz. Züechtugsbiol.  96:56-59.

 

Gjedrem, T.  1992.  Breeding plans for rainbow trout.  Aquaculture, 100: 73-83.

 

Hanke, A.R., G.W. Friars, R.L. Saunders, and J.M. Terhune.  1989.  Family x photoperiod interaction on growth in juvenile Atlantic salmon, Salmo salar.  Genome, 32(6):  1105-1112.

 

Hard, J.J. and W.K. Hershberger.  1995.  Quantitative genetic consequences of captive broodstock programs for anadromous Pacific salmon. in T.A. Flagg and C.V.W. Mahnken (eds.) An Assessment of the status of captive broodstock technology for Pacific salmon.  Bonneville Power Administration, Portland, OR, USA.   Final Report 93-56, pp. 2.1-2.75

 

Heath, D.D., R.H. Devlin, J.W. Heath, and G.K. Iwama.  1994.  Genetic, environmental, and interaction effects on the incidence of jacking in Oncorhynchus tshawytscha (chinook salmon).  Heredity, 72(2):  146-154.

 

Hershberger, W.K. and R.N. Iwamoto. 1984.  Systematic genetic selection and breeding in salmonid culture and enhancement programs.  In:  Carl J. Sinderman (Editor) Proceedings of the eleventh U.S.‑Japan meeting on aquaculture, salmon enhancement. Tokyo, Japan, 19‑20 October 1982, p. 29‑32, NOAA Tech. Rep. NMFS 27.

 

Hershberger, W.K., J. M. Myers, R. N. Iwamoto, and W.C. McAuley.  1990a.  Assessment of inbreeding and its implications for salmon broodstock development.  In:  R. S. Svrjcek (Editor) Genetics in aquaculture: Proceedings of the sixteenth U.S.‑Japan meeting on aquaculture; 20‑21 October 1987, Charleston, South Carolina (R.S. Svrjcek, ed.).  NOAA Tech. Rep. NMFS92, pp. 1-7.

 

Hershberger, W.K., J.M. Myers, R.N. Iwamoto, W.C. McAuley, and A.M. Saxton.  1990b.  Genetic changes in the growth of coho salmon (Oncorhynchus kisutch) in marine net‑pens, produced by ten years of selection.  Aquaculture, 85 (1990) 187‑197.

 

Hohenboken, W.D. 1985. Genotype x Environment Interaction. pp 151-165 In:  A.B. Chapman (Editor), General and Quantitative Genetics, Elsevier Science Publishers, Netherlands, 408p.

 

Iwamoto, R.N. 1982.  Strain-photoperiod-temperature interactions in coho salmon:  freshwater growth, smoltification, and seawater adaptation.  Ph.D. Dissertation, Univ. Washington, Seattle, 405pp.

 

Iwamoto, R.N., A.M. Saxton, and W.K. Hershberger. 1982.  Genetic estimates for length and weight of coho salmon (Oncorhynchus kisutch) during freshwater rearing.  J. Hered., 73: 187‑191.

 

Iwamoto, R.N., B.A. Alexander, and W.K. Hershberger.  1984.  Genotypic and environmental effects on the incidence of sexual precocity in coho salmon (Oncorhynchus kisutch).  Aquaculture, 43:  105-121.

 

Iwamoto, R.N., J.M. Myers, and W.K. Hershberger. 1986. Genotype-environment interactions for growth of rainbow trout, Salmo gairdneri.  Aquaculture, 57: 153-161.

 

Kincaid, H.L.  1976.  Effects of inbreeding on rainbow trout populations.  Trans. Am. Fish. Soc., 105: 273‑280.

 

Kincaid, H.L. 1983.  Inbreeding in fish populations used for aquaculture.  Aquaculture, 33: 215‑227.

 

Kincaid, H.L. 1993.  Selective breeding and domestication.  Pp:307-309.  In: J.G. Cloud and G.H. Thorgaard (eds) Genetic Conservation of Salmonid Fishes.  Plemum Press, New York, NY USA.  314 p.

 

Kincaid, H.L. 1995. An evaluation of inbreeding and effective population size in salmonid broodstocks in federal and state hatcheries. Uses and effects of cultured fishes in aquatic ecosystems.  Am. Fish. Soc. 15, pp. 193-204.

 

Levins, R.  1968.  Evolution in changing environments.  Monogr. in Population Biology 2.  Princeton Univ. Press, Princeton, N.J., 120pp.

 

Myers, J.M., R.N. Iwamoto, D. Teel, D. Van Doornik, and W.K. Hershberger.  1999.  Coho salmon broodstock development‑‑‑1977 TO 1998.  Ten Generations of Systematic Selective Breeding. Bull. Natl. Res. Inst. Aquacult., Suppl 1: 63-70.

 

Namkoong, G. 1979.  Introduction to quantitative genetics in forestry. U.S. Dep. Agric. Bull., No. 1588, 342pp.

 

Novotny, A.J. 1975.  Net-pen culture of Pacific salmon in marine waters.  Marine Fisheries Review, 37: 36-47.

 

Pirchener, F.  1969.  Population genetics in animal breeding.  W.H. Freeman and Co., San Francisco, 274 p.

 

Pratschner, G.A.  1978.  The relative resistance of six transferrin phenotypes of coho salmon (Oncorhynchus kisutch) to cytophagosis, furunculosis, and vibriosis.  M.S. Thesis, Univ Washington, 71pp.

 

Reisinbichler, R.R.  and J.D.  McIntyre. 1977.  Genetic differences in growth and survival of juvenile hatchery and wild steelhead trout, Salmo gairdneri.  J. Fish. Res. Board Can., 34: 123-128.

 

Robertson, A. 1959.  The sampling variance of the genetic correlation coefficient.  Biometrics, 15: 469-485.

 

Saxton, A.M., W.K. Hershberger, and R.N. Iwamoto. 1984.  Smoltification in the net‑pen culture of coho salmon: quantitative genetic analysis.  Trans. Am. Fish. Soc., 113: 339‑347.

 

Shaklee, J.B., F.W. Allendorf, D.C. Morizot, and G.S. Whitt. 1990.  Gene nomenclature for protein-coding in fish.  Trans. Am. Fish. Soc., 119(1): 2-15.

 

Su,  G-S., L-E. Liliedahl, and G.A.E. Gall. 1996.  Effects of inbreeding on growth and reproductive traits in rainbow trout (Oncorhynchus mykiss).  Aquaculture, 142 (3-4): 139-148.

 

Suzumoto, B.K., C.B. Schreck, and J.D. McIntyre. 1977. Relative resistance of three transferrin genotypes of coho salmon (Oncorhynchus kisutch) and their hematological responses to bacterial kidney disease. J. Fish. Res. Board Can., 34(1):1-8.

 

Swofford, D. L., and R. B. Selander. 1981. BIOSYS-1: a FORTRAN program for the comprehensive analysis of electrophoretic data in population genetics and systematics. J. Heredity, 72:281-283.

 

Taylor, E. B.  1991.  A review of local adaptation in Salmonidae, with particular reference to Pacific and Atlantic Salmon.  Aquaculture, 98:185-207.

 

Turner, H.N. and S.S.Y. Young. 1969. Quantitative Genetics in Sheep Breeding. Cornell University Press, Ithaca, N.Y., 332pp.

 

Utter, F.M., W.E. Ames, and H.O. Hodgins.  1970.  Transferrin polymorphism in coho salmon (Oncorhynchus kisutch).  J. Fish. Res. Board Can. 27: 2371-2373.

 

Van Doornik, D.M., G.B. Milner, and G.A. Winans. 1995. An improved method of data collection for transferrin polymorphism in coho salmon (Oncorhynchus kisutch). Biochem. Genet. 33:257-260.

 

Wangila, B.C.C. and T.A. Dick. 1988.  Influence of genotype and temperature on the relationship between specific growth rate and size of rainbow trout.  Trans. Am. Fish. Soc., 117(6):  560-564.

 

Wild, V., H. Simianer, H-M. Gjoeen, B. Gjerde.  1994.  Genetic parameters and genotype x environment interaction for early sexual maturity in Atlantic salmon (Salmo salar).  Aquaculture, 128(1-2):  51-65.

 

Winkler, F.M., D. Bartley, and N.F. Diaz. 1999. Genetic differences among year classes in a hatchery population of coho salmon (Oncorhynchus kisutch) (Walbaum, 1792) in Chile.  Aquaculture 173: 423-431. 

 

Winter, G.W., C.B. Schreck, and J.D. McIntyre. 1980. Resistance of different stocks and transferrin genotypes of coho salmon, Oncorhynchus kisutch, and steelhead trout, Salmo gairdneri, to bacterial kidney disease and vibriosis.  Fish. Bull., U.S. 77:795-802.

 

Zar, J.H. 1974.  Biostatistical analysis.  Prentice-Hall, Englewood Cliffs, NJ.  620 pp.