Integrative phenotypic and gene expression data identify myostatin as a muscle growth inhibitor in Chinese shrimp Fenneropenaeus chinensis

Growth traits, largely determined by muscle growth, are the most critical economic traits in shrimp breeding. Myostatin (Mstn) is a conserved inhibitor of muscle growth in vertebrates, but until now solid evidence supporting a similar function of Mstn in invertebrates has been lacking. In the present study, we examined the Mstn expression along with growth trait data in a Fenneropenaeus chinensis population, to establish a potential correlation between Mstn and growth. The heritabilities of FcMstn expression, body weight at 190 days of culture, body weight and length at 230 days of culture, and average daily gain were estimated using 773 individuals and a thirteen-generation pedigree. The results showed FcMstn expression was negatively correlated with the growth traits, and the mean FcMstn expression in females was significantly lower than that of males, indicating Mstn negatively regulates muscle growth in shrimp, and its lower expression may underscore the faster growth of females. Low heritabilities were detected for FcMstn expression, suggesting that the expression of Mstn might be heritable in shrimp. These results provide strong support for a growth inhibitory function of Mstn in F. chinensis, and suggest a potential method for selective breeding of this species without substantial experimental resources and labor force.

Shrimp is an important aquaculture species with high single output values, which takes a great role in the economic development of the world. To improve the production has always been the important goal in shrimp farming. Faster growth is essential for increasing production, which could make for more economic benefits and sustainable development for shrimp breeding. Selective breeding is an important propelling force for the development of efficient and sustainable production 1 . In the beginning, mass selection was the most commonly used method for selection breeding in aquaculture, because it was easy to manipulate 2 . Afterwards, family-based selection became the alternative methods in aquatic species due to its effect for all types of traits 3 . The family-based selective breeding programs have improved the target traits a lot and made great contributions to the development of shrimp industry [4][5][6] . However, there were still some difficulties in family-based selective breeding, such as the high investment and operational costs 7 , and long and complex test periods. Therefore, new methods and technologies are urgently required for accelerating the genetic improvement of important traits in shrimp.
Along with the development of molecular biology technology, marker assisted selection (MAS) and genome-wide selection (GWS) were suggested as effective methods for selecting for target traits 8,9 . However, the markers used for MAS need to be closely linkaged with the larger effective quantitative trait locus (QTL) of the target trait, and GWS needs large number of markers to be genotyped in large scale samples. Although the costs of MAS and GWS technologies are gradually decreasing as new breakthroughs in methodology, they are still

Results
Descriptive statistics. Finally, a total of 773 individual records from 44 full-sib families, including 383 females and 390 males were analyzed in the present study. The number of observations, mean value, maximum and minimum, standard deviation and coefficient of variation for BW1, BW2 and BL2, ADG, and FcMstn expression at individual (all individuals, females, and males, respectively) and family (family mean with all individuals, females, and males, respectively) levels are shown in Table 1. The 25th percentiles, median percentiles, 75th percentiles, minima and maxima of the five traits at individual and family levels are displayed in Fig. 1. The results revealed the BW1, BW2, ADG, and FcMstn expression values both varied substantially among the individuals and families, but the BL2 has lower variance at the two levels. However, the variance was higher when analyzed at the individual level than the family level as indicated by the higher standard deviation (SD) and coefficient of variation (CV) for all the traits (Table 1). For the growth traits (BW1, BW2, BL2, and ADG), the means of females were significantly higher than those of males at both of individual and family levels (P < 0.001). However, for the FcMstn expression, the mean of females was significantly lower than that of males at the two levels (P < 0.001).

Variance components and heritability estimations.
Estimates of the variance component and heritability for BW1, BW2, ADG, BL2, and FcMstn expression at all-individual, female-individual, and male-individual levels are provided in Table 2. For BW1, BW2, and BL2, the females have the highest heritabilities, followed by all individuals, but the males have the lowest heritabilities. Excepting for BL2 in males, the heritability estimates of the five traits (BW1, BW2, BL2, ADG, and FcMstn expression) were all significantly different from zero at the three levels (P < 0.05). According to the classification reported by Cardellino and Rovira 30 (i.e., low, 0.05-0.19; medium, 0.20-0.44; high, 0.45-0.64; and very high, >0.65), the FcMstn expression have low heritabilities (0.05 ± 0.03, 0.06 ± 0.06, and 0.08 ± 0.06, respectively) at all-individual, female-individual, and male-individual levels; the heritability estimates of BW1 (0.75 ± 0.14) and BW2 (0.66 ± 0.13) in females were very high; the heritability of BL2 (0.59 ± 0.13) in females was high; the heritability estimates of BW1 (0.35 ± 0.09), BW2 (0.34 ± 0.09), BL2 (0.22 ± 0.07), and ADG (0.22 ± 0.08) in all individuals, and BW1 (0.25 ± 0.09), BW2 (0.26 ± 0.10), and ADG (0.44 ± 0.11) in males were medium. The heritabilities of BW1 were not significantly different from those of BW2 at all the three levels (P > 0.05). The heritability estimates of BW1, BW2, and BL2 in females were significantly greater than those in all individuals and males (P < 0.05), but there was no significant difference between all individuals and males (P > 0.05). For the heritabilities of ADG and FcMstn expression, there was no significant Genetic and phenotypic correlation analysis. The correlation analysis based on phenotypic and breeding values between traits at all-individual, female-individual, and male-individual levels are given in Table 3. Importantly, there were significant negative genetic correlations between FcMstn expression and other growth traits (BW1, BW2, BL2, and ADG) at the three levels, excepting for the non-significant negative genetic correlation with ADG in males. Especially, there were extremely significant genetic correlation between FcMstn expression and BW1 (−0.91), BW2 (−0.97), BL2 (−0.98), and ADG (−0.84) in females, and there were also significant genetic correlation between FcMstn expression and BW1 (−0.86), BW2 (−0.85), and BL2 (−0.64) in males. In generally, there are significantly positive correlations between the growth traits, and r g values were higher than r p values; females have the highest r g and r p values, followed in all individuals, but the males have the lowest correlation values, and almost all of the correlations were significantly different from zero (Table 3). BW2 is significantly correlated with BL2 (r g was 0.99, 1.00, and 0.91, respectively), BW1 (r g was 0.97, 1.00, and 0.81, respectively) and ADG (r g was 0.71, 0.94, and 0.63, respectively) in all individuals, females, and males. In addition, BW1 is significantly correlated with BL2 (r g was 0.98, 1.00, and 0.84, respectively) in all individuals, females, and males, and BL2 is also significantly correlated with ADG (r g was 0.76, 0.95, and 0.57, respectively) at the three levels.

Discussion
Growth traits, largely determined by muscle growth, are the most concerned economic traits in the shrimp breeding 31 . Sexual dimorphism for growth of shrimp is very distinct; specifically, females tend to grow faster and bigger than males, such as in F. chinens 6 , L. vannamei 4,32 , and P. monodon 33,34 . Mstn is a conserved inhibitor of muscle growth in vertebrates, but until now solid evidence supporting a similar function of Mstn in invertebrates has been lacking. In the present study, we firstly to establish a potential correlation between the expression of Mstn and growth using a large sample size in shrimp, containing the difference of its expression level between females www.nature.com/scientificreports www.nature.com/scientificreports/ and males. In addition, we firstly estimated the heritability of the FcMstn expression with the large sample size and a deep pedigree.
According to the classification reported by Cardellino and Rovira 30 , the growth trait has medium to high heritability (0.24-0.52) in shrimp 5,6,35,36 , which was consistent with our results, illustrating that the genetic factor played important role during their growing process. Some scholars pointed out that variation in gene expression may play more important roles than differences between variant forms of proteins for evolution 37 . The DNA region that affects gene expression is highly variable, containing more than 0.6% polymorphism 38 . The naturally occurring polymorphisms of nucleotides could affect the gene expression in vivo [39][40][41] . In the present study, the growth rates varied substantially among the individuals and among the families even their growth were selected for thirteen generations, and females generally grew faster than males (Table 1). This phenomenon may be related to different expression levels of some genes among the individuals and between females and males.
In the previous studies of F. merguiensis 29 and F. chinens (the results would be published in our other study named "Identification of myostatin and investigation of its inhibitory function on myogenesis and muscle growth in Chinese Shrimp, Fenneropenaeus chinensis"), significantly higher expression level of Mstn in muscle tissue was observed in the smaller shrimp than in the larger shrimp with about 10 individuals for each group. In the present study, we analyzed the expression of FcMstn in a large-scale sample (383 females and 390 males), and the results showed its expression varied substantially among individuals and among families (Table 1), which has provided important precondition and foundation for growth selection by analyzing its expression. Importantly, significantly negative genetic correlation was detected between the expression level of FcMstn and the growth traits (BW1, BW2, BL2, and ADG), especially in the females (r g ranged from −0.84 to −0.98) ( Table 3). In addition, www.nature.com/scientificreports www.nature.com/scientificreports/ the mean expression of Mstn in females was significantly lower than that of males at individual and family levels (P < 0.001). These results further demonstrate that Mstn negatively regulates muscle growth in shrimp, and its lower expression may underscore the faster growth of females. On this point, the results were found out some resemblance to the study on F. merguiensis Mstn that knock-down of FmMstn gene by RNAi can cause a stronger muscle development in F. merguiensis 29 . However, the study on P. monodon Mstn showed that shrimp with reduced expression level of Mstn mRNA displayed a slow growth rate compared with the control groups 26 , which might be caused by the effects of molt from the small sample scale and short experiment period. Because silencing of Mstn increased premolt duration, but the higher PmMstn expression in premolt stage led to slow growth 26 . Zhuo et al. 29 detected high polymorphism and identified several potential SNPs in Mstn of F. merguiensis. In addition, two alternative splicing isoforms of Mstn was found in F. merguiensis 29 and F. chinens (the results would be published in our other study named "Identification of myostatin and investigation of its inhibitory function on myogenesis and muscle growth in Chinese Shrimp, Fenneropenaeus chinensis"). Shin et al. 42   www.nature.com/scientificreports www.nature.com/scientificreports/ splicing isoforms of Mstn could negatively regulate pro-myostatin processing in muscle cells and prevent Mstn mediated inhibition of myogenesis in avian species. Consequently, we deduced that the variation in the expression of Mstn affected the growth rate of the shrimp, which might result from the difference of polymorphic nucleotides and alternative splicing isoforms among the individuals/families and between females and males. In the future study, we will carry out research to further verify the relationship between the polymorphic nucleotides or alternative splicing isoforms and the expression level of Mstn.
To investigate the potential use of the Mstn expression as quantitative trait for improving growth by selective breeding in shrimp, the heritability of its expression was estimated in the present study. In the previous heritability studies of gene expression, they all suggested a strong component of differential expression among genotypes [15][16][17][18][19] . However, because of the small sample sizes and lower genetic tie among individuals, most of these studies did not provide a direct estimate of heritability. The previous studies demonstrated that sample sizes of 100 individuals were too small to support robust estimates of heritability 13,43 . In the present study, we used a large scale sample (773 individuals from 44 full-sib families) and a thirteen generation pedigree for estimating the heritability of the expression of Mstn. Low heritability estimations were detected for the expression of Mstn at all-individual, female-individual, and male-individual levels, suggesting that the expression of Mstn might be heritable in shrimp. In addition, the significantly negative correlated between the expression of Mstn and the body weight and average daily gain indicated that there was selective potential for improving growth performance by analyzing the expression of Mstn at early stage before substantial experimental resources and labor force are invested. In future study, we will use more stages including the juvenile at rapid growth stage to verify the heritability of the expression of Mstn and its correlation with later stages.
Traditionally, the heritability estimates for growth were mainly focused on harvest body weight in shrimp 6,35,44 . In the present study, the heritabilities of harvest body weight (BW1), body weight (BW2) and body length (BL2) of later stage, and average daily gain (ADG) were estimated for the G 13 of F. chinensis. In our previous study, the heritabilities of harvest body weight for the G 3 -G 5 of this population were 0.23-0.36, 0.22-0.36, 0.37-0.38 at all-individual, female-individual, and male-individual levels, respectively 6 , and the heritabilities in males were higher than those of females. In the present study, the heritabilities of BW1 for the G 13 were 0.35, 0.75, and 0.25 at the three levels. There was no significant difference between the heritabilities of G 3 -G 5 and that of G 13 both at all-individual and male-individual levels, but the heritability of BW1 for the G 13 was significantly higher than those of the G 3 -G 5 in females 6 ( Table 2). In the present study, the heritabilities of BW1, BW2, and BL2 in females were also significantly higher than those in males, which were consistent with the results of previous studies in M. rosenbergii 45,46 . They also reported heritabilities were higher in females than those in males 45,47,48 . When the females and males were treated as a separate trait, they have significant positive genetic correlations ( Table 2), indicating that the growth of females and males are controlled by the same genes, such as Mstn. There was no significant difference for heritability estimates between BW1 and BW2 at the three levels, and BW1 has significant genetic correlations with BW2, BL2, and ADG, indicating the later growth could be selected by the harvest body weight.
In summary, we analyzed the expression of Mstn along with growth trait data in 773 individuals (383 females and 390 males) from 44 full-sib families of a F. chinensis population, to establish a potential correlation between Mstn and growth in this study. In addition, the expression of Mstn was treated as a quantitative trait, and its heritability was estimated using these 773 samples and a thirteen generation pedigree. The results provide strong support for a growth inhibitory function of Mstn in shrimp, and its lower expression may underscore the faster growth of females. The result also demonstrated that the expression of Mstn might be heritable in shrimp, suggesting that there might be potential for improving growth performance by analyzing the expression of Mstn at early stage. The heritabilities of body weight and body length in females were significantly higher than those in males, indicating the females have higher selection potential. The high genetic correlations between females and males indicated that the growth of females and males are controlled by the same genes, such as Mstn.

Methods experimental materials. The experiment was conducted in the Mariculture Research Station of Yellow Sea
Fisheries Research Institute, Chinese Academy of Fishery Sciences, located in Qingdao City, Shandong Province, China in 2018. The experimental shrimp were selected from 105 full-sib families (including 28 half sib families) of the 13 th generation (G 13 ) of the selected population from our selective breeding program of F. chinensis. A nested mating design was used to produce the full-and half-sib families, and two females mated with one male to produce the half-sib families. The origin and the selection procedure of this selected population of F. chinensis, processes of family construction, hatching, and larvae rearing were described detailedly in our previous study 6 . The 105 families were reared separately in 200-L larvae-culture tanks after they were constructed. At the post-larvae stage in 35 days, 200 individuals of each family were moved to a larger tank (3 m 3 ), and the families were still separately reared. When the average body weight reached about 2.0 g, each family was tagged with a unique family code by injecting visible implant fluorescent elastomers (VIE) at three locations on the 6 th and 5 th abdominal segment. After tagging for each family, 60 healthy tagged individuals were randomly selected and equally divided into four concrete tanks (100 m 2 ) for rearing. During the rearing period, water was exchanged about 30% daily. The shrimp were fed with formulated diets four times every day, and a total daily supply accounted for 5% wet weight and adjusted daily. All the shrimp were individually tagged with numbered ring set on one ocular peduncle, and the family VIE tag, individual body weight (BW1) and sex were recorded at 190 days of culture (DOC).
In shrimp, the females grow faster than the males. In order to use enough females and males of each family for the next expression analysis, the families with lower survival rates (<50%) and with fewer females or males were eliminated. Finally, 44 full-sib families were selected, and about 20 individuals with a 1:1 sex ratio were selected from each family. All these selected individuals were moved to a cement tank (27 m 2 ) with close recirculation system for communal rearing. During the rearing period, the temperature was 22-24 °C. In the same way, the shrimp www.nature.com/scientificreports www.nature.com/scientificreports/ were fed with formulated diets according to their wet weight and fresh Ruditapes philippinarum. The family VIE, individual tag, sex, body weight (BW2), and body length (BL2) were recorded at 230 DOC. At the same time, the muscle tissue of the healthy survival individuals was separately dissected and frozen immediately in liquid nitrogen, and after that they were stored at −80 °C until RNA extraction. Average daily gain (ADG) of body weight was calculated for each individual based on BW2 and BW1.
RnA extraction and cDnA synthesis. RNA was extracted from muscle using the Trizol reagent (TAKARA, Japan) and phenol chloroform. The purity and concentration of the RNA was checked using a ND-1000 spectrophotometer (NANO DROP TECHNOLOGIES, USA), and the integrity of the RNA was monitored by electrophoresis on 1.5% agarose gel. Single-stand cDNA was synthesized from total RNA (1 μg) using PrimeScript RT reagent kit (TAKARA, Japan) following the manufacturer's protocol. Then cDNA products were stored at −80 °C until the Real-time quantitative PCR (qRT-PCR).  49 . Primers were tested to ensure specific amplification of single discrete band with no primer dimers. To detect optimal efficiency of the primer pair, the cDNA was analyzed with serial fold dilutions (1, 1/10, 1/20, 1/40 and 1/80). The qRT-PCR was performed in an ABI 7500 Sequence Detection System (ABI, USA) using a SYBR PrimeScript RT reagent Kit (TAKARA, Japan). Each reaction was performed in triplicate with the following conditions: 95 °C for 3 min, followed by 35 cycles of 95 °C for 30 s, 54 °C for 20 s, and 72 °C for 25 s. In order to ensure specificity of PCR product, melting curve determination was set at the end of each PCR reaction. The relative gene expression level of FcMstn for each individual was analyzed using the 2 −ΔΔCT method 50 .
Variance components and heritability estimates. In order to modify the residual distribution, the FcMstn expression data were performed natural logarithm transformation. The descriptive statistics analysis for BW1, BW2, BL2, ADG, and FcMstn expression was carried out using the SPSS software (version 17.0). Means of BW1, BW2, BL2, ADG, and FcMstn expression between females and males were analyzed by independent sample t-test, respectively.
The analysis of variance indicated that sex showed significant effects on BW1, BW2, BL2, ADG, and FcMstn expression (P < 0.001). Age, pond and the initial body weight had no significant effect on BW1, BW2, BL2, ADG, and FcMstn expression (P > 0.05). In addition, body weight has a significant effect on FcMstn expression (P < 0.001), so BW1 as a covariate was included in the analysis model for FcMstn expression. The complete pedigree from G 0 to G 12 was used in the following analysis to account for the genetic relationships among the families and individuals.
The variance components and heritabilities for BW1, BW2, BL2, ADG, and FcMstn expression were estimated using ASREML 4.0 51 at all-individual, female-individual, and male-individual levels, respectively. The fitted model for BW1, BW2, BL2, and ADG was Model 1, and that for FcMstn expression was Model 2, which were showed as follows: where y ij is the observed value of BW1, BW2, BL2, ADG, or FcMstn expression of the j th individual, µ is the overall mean, Sex i is the fixed effect of the i th gender, BW1 j is the covariate of the j th body weight when the individual was sampled for gene expression analysis, b is the regression coefficient, a j is the additive genetic effect of the j th individual, and e ij is the random residual error of the j th individual, e ~ (0, I σ e 2 ). The phenotypic variance (σ p 2 ) was taken as the sum of all of the variance components as follows: σ σ σ = + p a e 2 2 2 . The heritability (h 2 ) was calculated as the ratio between the genetic variance and the total phenotypic variance ( σ σ = h / a p 2 2 2 ). The z-scores were used to test whether the heritability estimates were significantly different from each other at all-individual, female-individual, and male-individual levels for each trait, and whether the heritability estimates were significantly different from zero 52 . where x i and x j are the heritability estimates all-individual, female-individual, or male-individual levels, and σ i and σ j are their respective standard errors. Both x j and σ j were set to be zero when testing whether an estimate was significantly different from zero. The resulting z-score was then tested against a large-sample normal distribution.
the phenotypic and genetic correlation estimates. The phenotypic (r p ) and genetic (r g ) correlations among BW1, BW2, BL2, ADG, and FcMstn expression, and r g between females and males for the five traits were estimated using the bivariate animal model with the ASREML package 51 . The r p and r g were calculated for the