Genetic Evaluation And A Preliminary Genomic Selection Research Of Economic Traits In Cynoglossus Semilaevis

Cynoglossus semilaevis, as commercially valuable and left-eyed flatfish, is an important marine fish species, widely distributed throughout coastal areas of China. However, the culture of C. semilaevis at present encountered some serious problems due to serious inbreeding that has occurred during artificial reproduction, such as unsatisfactory fecundity and quality of female, low hatchability of fertilized eggs, as well as low survival during larval and juvenile rearing, which seriously hindered the healthy development of C. semilaevis aquaculture. Therefore, advanced selective breeding technology is in urgent needed to breed excellent fish strains with high growth speed and resistant to diseases and adverse environment. In the present study, we evaluated the economical traits, including growth-related traits and disease resistance traits. Besides that, a preliminary research on genomic selection used in C. semilaevis breeding was also carried out. The results as follows:1. In order to know the relationship between body weight and morphometric traits in C. semilaevis, body weight（BW/g） and three morphometric traits, e.g., total length（TL/cm）, body height（BH/cm） and body depth（BD/cm） of 831 individuals of 12 months old C. Semilaevis were measured in this study. The best models for female and male individuals reflecting the relationship between each morphometric trait and body weight were obtained through model fitting screening, respectively. We established the regression equation with morphological traits as variables and body weight as the dependent variable and calculated the direct and indirect path coefficient and the coefficient of determination. The results showed that the best model fitted for female individuals was 090.3?004.0 xy? for TL – BW, 983.2?382.0 xy? for BH – BW and 461.2?912.61 xy? for BD – BW. And for male animals, they were 003.3?005.0 xy?for TL – BW, 147.3?300.0 xy?for BH – BW and 156.2?681.57 xy?for BD – BW. Path analysis showed that BH was the most predominant variable to affect BW（0.535） in females, while TL was the largest effect to BW（0.407） in males. Three morphological traits of female and male individuals were all extremely remarkable effects on the body mass. The multiple linear regression model equation were established as BW =-88.530 + 1.297*TL+ 15.862*BH+ 39.021*BW for females, and BW=-71.579 + 2.587*TL+ 7.351*BH + 35.595*BW for males.2. Phenotypic and genetic parameters for growth-related traits in C. semilaevis, were estimated in 22 full-sib families produced by normal and neo-male breeding stocks. As phenotypic males with female genotypes, neo-males are harmful in C. semilaevis aquaculture because they reduce overall production. The present study evaluated the difference in the growth-related traits: total length（TL）, body weight（BW） and square root of body weight（SQ_BW） at the age of 570 days between normal and neo-male offspring（neo-males used as male parents）. The difference in the proportion of females between normal and neo-male offspring was also assessed. Based on the linear mixed model, restricted maximum likelihood（REML） and best linear unbiased prediction（BLUP） were used to estimate various（co）variance components and estimated breeding values（EBVs） of growth-related traits. As a result, all the mean values of the three studied traits were significantly larger in normal offspring than in neo-male offspring. Additionally, the female proportion was significantly larger in normal offspring than in neo-male offspring. Heritability was 0.128 ±0.0662 for TL, 0.128 ±0.0655 for BW and 0.132 ±0.0629 for SQ_BW, all of which were low level heritabilities. The correlation coefficients of EBVs and phenotypic values of the target traits were 0.516 for TL, 0.524 for BW and 0.506 for SQ_BW, all of which were highly significant（P<0.01）. Genetic correlations among TL, BW and SQ_BW were positive high（0.921–0.969） and higher than those of phenotype（0.711– 0.748）, both of which had low standard errors（0.063–0.123 for genotype, and 0.010–0.018 for phenotype）. Compared with normal offspring, neo-male offspring have lower breeding values for each studied trait through EBVs comparison. Therefore, neo-male offspring should not be used as broodstock in a C semilaevis breeding programs.3. The main goal of this part was to estimate the heritabilities and genetic correlations for the early growth traits（total length, body depth and body weight） of C. semilaevis based on data observed from 80 families produced using male parents collected from four locations, i.e. location₁, location₂, location₃（location₃₁ and location₃₂）, and location₄. Since significantly different for growth traits between the group of fish with pure white bottom color and the rest（P < 0.05）, the trait of fish with pure white or not（named as trait of bottom color） was analyzed together with the growth traits. Heritability estimates for the four traits varied from 0.210 to 0.362, all of which were medium heritabilities. The relatively high heritability of bottom color（0.362） indicates that the available genetic variation in this trait should result in worthwhile response to selection. Genetic correlations among growth traits were positive and high（0.913 – 0.959）, and higher than those of the phenotypic（0.866 – 0.899）. Therefore, selection of one of the three traits will result in correlated increase in the other two as positive and high correlations. In addition, genetic correlations between bottom color and growth traits were positive and ranging from 0.241 to 0.353, higher than those of phenotypic（0.127 – 0.151）, which suggested selection of the bottom color can be utilized to enhance the selection of growth traits in a certain extent. Breeding performances of all the families were also analyzed, and the best 16 families was selected and treated as broodstock. This research obtained genetic parameters with high accuracy, because of a large number of family was used to analyze, which provides important basic information to implement selective breeding in C. semilaevis.4. Pathogen infection is an aquatic animal health issue that results in significant economic loss to the aquaculture industry. The objective of this part was to estimate genetic parameters for three disease-resistances traits, including trait₁（survival of challenge test with Edwardsiella tarda）, trait₂（survival of challenge test with Vibrio anguillarum） and trait₃（survival of natural infected）, in C. semilaevis. Heritabilities of the traits varied from 0.137 to 0.263, as estimated by the linear model（LM） and the threshold model（TM）. The trait₁ heritabilities（0.263 and 0.193 estimated by LM and TM） were higher than those of others（0.216 and 0.232 estimated by LM, 0.137 and 0.188 estimated by TM）. The estimates of heritabilities using LM were consistently higher than those of TM in this study. There were significant medium genetic correlations of 0.438 and 0.423 between trait₁ and trait₂ obtained from LM and TM（P < 0.05）. However, very low genetic correlations of trait₁ and trait₃（-0.100 for LM,-0.046 for TM）, as well as that of trait₂ and trait₃（0.047 for LM, 0.042 for TM） were obtained. Therefore, indirect selection for trait₁ and trait₂ was effective but almost ineffective for trait₁ and trait₃ as well as trait₂ and trait₃. Since there were not high genetic correlations among the three traits, it can be expected to obtain superior lines which can be resistant to two or more kinds of pathogens through cross mating. Otherwise, there was no significant difference in the predictive abilities of LM and TM, according to the correlation analysis of the two groups of family breeding values predicted by the two analysis models within each trait, respectively. The results provided important reference material for selecting C. semilaevis strains with high resistance to more than one kind of pathogen.5. Disease caused by Vibrio harveyi is one of the most serious health problems affecting the culturing of C. semilaevis in China. This part presents the first assessment of genetic variation associated with resistance to V. harveyi, and reports on a comparison of different measures of V. harveyi infection leading to a recommended breeding trait measure. C. semilaevis individuals from 54 full sibling families were communally challenged with V. harveyi. At termination of the challenge test, the families’ survival rates varied from 9.15% to 94.25% indicated there was a large difference of families resistant to V. harveyi. The variance components and heritabilities of disease resistance trait under different measures were estimated using the linear model, and gradually increased from 0.089 ± 0.021 to 0.449 ± 0.074. The total survival rate was 57.96%, when the heritability achive the highest, suggested a moderate proportion of genetic variance in V. harveyi resistance exists within the C. semilaevis population and provides good scope for selective breeding for this disease resistance. This research obtained genetic parameters with high accuracy, because of a large number of family was used to analyze, which provides important basic information to implement selective breeding in C. semilaevis. In addition, 12 families with high disease resistance were selected by family breeding values and breed as brood stocks, which were the basis of breeding resistant strains for C. semilaevis.6. Genomic selection（GS）, as reported, could improve the genetic progress mostly, decrease the inbreeding coefficient for the selective population and has attracted a lot of attentions in animal breeding. In this part, we detailedly summarized the progress of genomic selection in C. semilaevis from following eight parts: 1) construction of reference population; 2) collection of phenotypic information; 3) extraction of high quality genome DNA; 4) SNP genotyping by whole genome resequencing; 5) quality control and genotype imputation of SNP data obtained from genome sequencing; 6) several major statistical genetic algorithms were introduced to predict GEBVs; 7) evaluation of different statistical genetic algorithms; and 8) selection of candidates according to GEBVs. This strategy will provide scientific guidance for genomic selection used in C. semilaevis breeding.7. In order to obtain a appropriate density SNP data size and a excellent analysis model with high accuracy of predicting GEBVs, in this part, we analyzed the accuracy of estimating every SNP effect and predicting GEBV using different Bayes methods including Bayes A, Bayes B, Bayes Cπ, and Bayesian Lasso, based on the different number of SNP markers obtained through genomic data simulation. The results showed that the estimation accuracy dramatically increased, when the number of SNP markers increased from 1K to 6K. And the predicting accuracy increased slowly, when the SNP number added from 6K to 12 K, but a downward trend appeared when the SNP number was more than 12 K. The Bayes Cπ model has a higher superiority obtained from the DIC value comparison and the regression analysis of phenotypic value and GEBV. Comparison of the predicting accuracy of reference population progenies GEBV found that it is 10 K SNP genomic data and Bayes Cπ model used in genomic selection that will achieve the best effect. Those results provided an important reference to select and custom a suitable density SNP chip used in C. semilaevis genomic selection breeding.