Implications
In the present study, we found evidence establishing a positive correlation between harvest weight and skin pigmentation patterns, including melanin-based spottiness and intensity of the lateral band (LB). Therefore, selection on growth will not impact negatively skin patterns or conversely. The positive relationship between skin pigmentation patterns and BW, and their link with different types of physiological-behavioral individual responses (i.e. coping styles) must be further investigated.
Introduction
Aquaculture has an increasingly important role in global protein production for human consumption, reaching 70.5 million tons in 2012 (Food and Agriculture Organization (FAO), 2016), with rainbow trout (Oncorhynchus mykiss) being one of the most cultivated species. Rainbow trout was the second largest species of salmonid produced in 2014, with an estimated world production of 812 939 tons, valued in the US $3631 million (FAO, 2016). Salmonids are particularly known for their wide genetic variation, life history and their adaptability to different places and environments. An example of variation among individuals is the amount of skin spottiness. In vertebrates, darker eumelanic individuals have been reported to be sexually more active, more aggressive, larger in size, less sensitive to stressful factors, and with a better immune function than lighter individuals of the same population (Ducrest et al., Reference Ducrest, Keller and Roulin2008). In addition, it appears that melanin-based skin spots in rainbow trout is associated with various behavioral (feeding and activity), morphological and physiological characteristics, such as dominance and food consumption, characteristics that have also been associated to different coping styles: low response (LR) and high response (HR) individuals (Kittilsen et al., Reference Kittilsen, Schjolden, Beitnes-Johansen, Shaw, Pottinger, Sørensen, Braastad, Bakken and Øverli2009a). High feed efficiency and feeding motivation of LR fish might have an impact on economically important traits (e.g. harvest weight (HW)), making this kind of fish more desirable for farming purposes. Measuring levels of cortisol to identify LR fish can be expensive and labor-intensive. Pigmentation patterns can be indicative of LR type fish, given the relationship between both the number and size of melanin-based skin spots and the individual’s coping style (Kittilsen et al., Reference Kittilsen, Schjolden, Beitnes-Johansen, Shaw, Pottinger, Sørensen, Braastad, Bakken and Øverli2009a). Thus, to identify and select fish that are more suitable for farming conditions in a simpler manner, the relationship between pigmentation patterns and other desirable characteristics can be exploited.
The success of selective breeding is influenced by levels of additive genetic variation present for the trait and by the genetic correlations with other relevant traits included in the breeding goal, which must not be disadvantageous (Falconer and Mackay, Reference Falconer and Mackay1996). Quantitative genetic studies in several species of farmed fish have analyzed characteristics such as growth, body size, age at maturity and meat quality (Gjedrem, Reference Gjedrem2012). Furthermore, previous studies have identified the presence of significant genetic variation for skin color and a melanin-based spottiness in rainbow trout, indicating the potential for improving these traits by means of selective breeding (Kause et al., Reference Kause, Ritola, Paananen, Eskelinen and Mäntysaari2003 and Reference Kause, Ritola and Paananen2004). The same previous studies have investigated the relationships of skin color and spottiness with growth and body shape in a Finnish population of rainbow trout (Kause et al., Reference Kause, Ritola, Paananen, Eskelinen and Mäntysaari2003 and Reference Kause, Ritola and Paananen2004) and as the genetics of skin color can be environment- (Kause et al., Reference Kause, Ritola and Paananen2004) and population-dependent, we considered necessary to re-evaluate these links in Chilean farmed rainbow trout. These authors did not find significant correlations between body composition and growth with skin color patterns. They judged skin color in terms of commercial influence, which means that silvery color with few spots is desirable. In the present study, skin color is judged in terms of supposed beneficial effects on growth performance, which means that intense coloration and spottiness are advantageous.
In this study, we hypothesized that genetic (co)variation between skin pigmentation patterns and growth exists and it is positive in rainbow trout. The objective of this study was to determine the genetic relationships between skin pigmentation patterns and BW in a breeding population of rainbow trout, to show that selection on growth will not impact negatively skin patterns or conversely. For this, we assess the levels of quantitative genetic variation for different traits related with skin pigmentation patterns in rainbow trout, including dorsal color (DC), the intensity of LB, melanin-based spottiness above (SA) and below (SB) the LB. The genetic correlation among these traits and HW were also determined to demonstrate the potential relationship between skin pigmentation patterns and a growth-related trait.
Material and methods
Population and traits
This work was carried out in a rainbow trout (O. mykiss) breeding population from a genetic improvement program established in 1998 by Aguas Claras S.A. in Puerto Montt, Chile (Yoshida et al., Reference Yoshida, Bangera, Carvalheiro, Correa, Figueroa, Lhorente and Yáñez2018a and Reference Yoshida, Carvalheiro, Rodriguez, Lhorente and Yáñez2018b). A total of four generations representing the years 1998, 2001, 2004 and 2007 were included in the study, reflecting a total of 40 429 pedigree-recorded fish. Each generation with a mean of 90 families were generated according to a nested design in which one male was mated on average to three females. Eggs from each family were incubated independently in one tank per family. Thus, eyed-eggs were transferred to family tanks for hatching. Every year during May fish from each family were passive integrated transponder tagged at about 5 to 7 g to keep pedigree traceability. Table 1 summarizes pedigree structure for the breeding population across years. Fish were then transferred to fresh water facilities to begin smoltification. An average number of 60.5 (SD=5.7) to 205.2 (SD=58.1) individuals from each family, depending on generation, were randomly divided into two fresh water tanks. After smoltification, fish were transferred to sea cages, keeping the tank distribution defined during fresh water rearing. The farming process finished at an average of 22 months post-spawning and at ~3.7 kg, in which HW and skin color scores were recorded for all fish. Within the same year, fish were harvested within a period of 1 to 2 months. Four skin color pattern scores were generated to visually classify fish into different categories. Dorsal color was divided into three categories according to the color at the back (blue (0), gray (1) or light green to brown (2)), intensity of the LB (none (0), tenuous (1), median (2) or marked (3)) and intensity of SA and SB the LB were divided into four categories according to the number of melanin spots and intensity (none (0), limited (1), median (2) and abundant (3)). The skin color scores were always given by the same person within generations. The descriptions of the different categories for each score are shown in Table 2. The dataset was comprised phenotypes for a total of 20 542, 20 560, 20 541, 20 541 and 19 615 records for DC, LB, SB, SA and HW, respectively.
Pedigree structure of the breeding population of rainbow trout (Oncorhynchus mykiss) by year
Appearance scores for each skin color category for dorsal color (DC), intensity of lateral band (LB) and intensity of spots above (SA) and below (SB) lateral band in a rainbow trout (Oncorhynchus mykiss) population
Statistical analysis
To estimate the variance and co-variance components for DC (y 1), LB (y 2), SA (y 3), SB (y 4) and HW (y 5), we used the following multi-trait linear animal model:
$$\eqalignno{ \left[ {\matrix{ {y_{1} } \cr {y_{2} } \cr {y_{3} } \cr {y_{4} } \cr {y_{5} } \cr } } \right]&{\equals}\left[ {\matrix{ {X_{1} } \cr 0 \cr 0 \cr 0 \cr 0 \cr } \matrix{ 0 \cr {X_{2} } \cr 0 \cr 0 \cr 0 \cr } \matrix{ 0 \cr 0 \cr {X_{3} } \cr 0 \cr 0 \cr } \matrix{ 0 \cr 0 \cr 0 \cr {X_{4} } \cr 0 \cr } \matrix{ 0 \cr 0 \cr 0 \cr 0 \cr {X_{5} } \cr } } \right] \left[ {\matrix{ {b_{1} } \cr {b_{2} } \cr {b_{3} } \cr {b_{4} } \cr {b_{5} } \cr } } \right]{\plus} \cr& \hskip10pt \left[ {\matrix{ {Z_{1} } \cr 0 \cr 0 \cr 0 \cr 0 \cr } \matrix{ 0 \cr {Z_{2} } \cr 0 \cr 0 \cr 0 \cr } \matrix{ 0 \cr 0 \cr {Z_{{3 }} } \cr 0 \cr 0 \cr } \matrix{ 0 \cr 0 \cr 0 \cr {Z_{4} } \cr 0 \cr } \matrix{ 0 \cr 0 \cr 0 \cr 0 \cr {Z_{5} } \cr } } \right] \left[ {\matrix{ {u_{1} } \cr {u_{2} } \cr {u_{3} } \cr {u_{4} } \cr {u_{5} } \cr } } \right]\cr&{\plus} \left[ {\matrix{ {W_{1} } \cr 0 \cr 0 \cr 0 \cr 0 \cr } \matrix{ 0 \cr {W_{2} } \cr 0 \cr 0 \cr 0 \cr } \matrix{ 0 \cr 0 \cr {W_{{3 }} } \cr 0 \cr 0 \cr } \matrix{ 0 \cr 0 \cr 0 \cr {W_{4} } \cr 0 \cr } \matrix{ 0 \cr 0 \cr 0 \cr 0 \cr {W_{5} } \cr } } \right] \left[ {\matrix{ {c_{1} } \cr {c_{2} } \cr {c_{3} } \cr {c_{4} } \cr {c_{5} } \cr } } \right]{\plus} \left[ {\matrix{ {e_{1} } \cr {e_{2} } \cr {e_{3} } \cr {e_{4} } \cr {e_{5} } \cr } } \right]$$
where y 1, y 2, y 3, y 4 and y 5 are vectors of phenotypic records for DC, LB, SA, SB and HW, respectively; b i the vectors of fixed effects for trait i (=1, 2, 3, 4 and 5), all including the contemporary group of sea cage:sex:year as factor and harvest age (HA) as covariate; u i , c i and e i the vectors of random animal genetic effects, common environment effects related to full-sib families and residual effects, respectively, for trait i; and X i , Z i and W i the design matrices for trait i.
For all traits, the animal, common environment, and residual effects were assumed random:
$$\eqalignno{ & u{\equals}\left[ {u_{{1{\rm }}}^{{\rm '}} \,u_{{2{\rm }}}^{{\rm '}} \,u_{{3{\rm }}}^{{\rm '}} \,u_{{4{\rm }}}^{{\rm '}} \,u_{{5{\rm }}}^{{\rm '}} {\rm }} \right]{\rm '}\,\sim\,N(0,{\rm }G_{0} \,\otimes\,A), \cr & c{\equals}\left[ {c_{{1{\rm }}}^{{\rm '}} \,c_{{2{\rm }}}^{{\rm '}} \,c_{{3{\rm }}}^{{\rm '}} \,c_{4}^{{\rm '}} \,c_{{5{\rm }}}^{{\rm '}} {\rm }} \right]{\rm '}\,\sim\,N(0,{\rm }C_{0} \,\otimes\,I_{{\rm C}} ), \cr & e{\equals}\left[ {e_{{1{\rm }}}^{{\rm '}} \,e_{{2{\rm }}}^{{\rm '}} \,e_{{3{\rm }}}^{{\rm '}} \,e_{{4{\rm }}}^{{\rm '}} \,e_{{5{\rm }}}^{{\rm '}} } \right]{\rm '} \,\sim\,N(0,{\rm }R_{0} \,\otimes\,I_{{\rm N}} ) $$
where A is the additive genetic relationship matrix constructed from the pedigree records, I C and I N are identity matrices with dimension C and N, respectively, ⊗ indicates the Kronecker product. G 0, C 0 and R 0 are 5×5 co-variance matrices of animal additive genetic, common environment and residual effects, respectively. The random common environment effect related to full-sib families was assessed preliminarily based on a likelihood ratio test carried out by means of fitting single-trait models. This effect was statistically significant (P<0.05) for all traits. The ASREML software (Gilmour et al., Reference Gilmour, Gogel, Cullis, Thompson and Butler2009) was used to fit the multi-trait model described above to estimate the variance and co-variance components for DC, LB, SA, SB and HW.
Heritability and genetic correlations
For each trait i (=1, 2, 3, 4 and 5) the heritability (
$h_{i}^{2} $
) and common environment (
$c_{i}^{2} $
) were calculated as follows:
$$\eqalignno{ & h_{i}^{2} {\equals}{{\sigma _{{G_{i} }}^{2} } \over {\sigma _{{G_{i} }}^{2} {\plus}\sigma _{{C_{i} }}^{2} {\plus}\sigma _{{E_{i} }}^{2} }} \cr & c_{i}^{2} {\equals}{{(\sigma _{{C_{i} }}^{2} )} \over {(\sigma _{{G_{i} }}^{2} {\plus}\sigma _{{C_{i} }}^{2} {\plus}\sigma _{{E_{i} }}^{2} )}} $$
where
$\sigma _{{G_{i} }}^{2} $
,
${\rm }\sigma _{{C_{i} }}^{2} $
and
$\sigma _{{E_{i} }}^{2} $
are the additive genetic, common environment and residual variances from G
0, C
0, and R
0 matrices, respectively. The phenotypic (r
p) and genetic (r
g) correlations between traits x and y were determined as (Falconer and Mackay, Reference Falconer and Mackay1996):
$$r_{{xy}} {\equals}{{\sigma _{{x,y}} } \over {\sqrt {\sigma _{{x{\rm }}}^{2} \sigma _{{y{\rm }}}^{2} } }}$$
where
$\sigma _{{x,y}} $
is the additive genetic or phenotypic co-variance between x and y,
$\sigma _{{x{\rm }}}^{2} $
the additive genetic or phenotypic variance of x and
$\sigma _{{y{\rm }}}^{2} $
is the additive genetic or phenotypic variance of y.
Results
Phenotypic variation
We found a considerable phenotypic variation for skin color scores. The number of records, absolute and relative frequencies of the skin color scores for DC, LB, SA and SB recorded at harvest time are shown in Table 3. For DC the relative frequencies varied between 1.6% to 7.2%, 3.9% to 40% and 52.8% to 94.6% across the 4 years for categories 0, 1 and 2, respectively. Considerable phenotypic variation was also present for HW. Table 4 shows the number of records and descriptive statistics for HW and HA divided by year. Overall, the mean (and standard deviation) for HW and HA were 3.4±1.10 kg and 682.92±73.13 days, respectively, across 19 615 records measured during the 4 years. The minimum and maximum values for HW and HA across the 4 years ranged from 0.3 to 7.2 kg and 564 and 818 days, respectively.
Sample size (n), absolute frequency (AF) and relative frequency (RF) for each skin color category for dorsal color (DC), intensity of lateral band (LB) and intensity of spots below (SB) and above (SA) lateral band in a rainbow trout (Oncorhynchus mykiss) population per year
Summary statistics for harvest weight (HW) and harvest age (HA) in a rainbow trout (Oncorhynchus mykiss) population per year
Min=mínimum; Max=máximum.
a Number of fish included in the analysis after removing outliers by interquartile range rule and discarding missing values.
Heritabilities and common environment effect
The estimated variance components and heritabilities for DC, LD, SA, SB and HW are shown in Table 5. We identified significant additive genetic variation for all analyzed traits. Moderate values of heritability were estimated for LB (0.17±0.03), SA (0.17±0.04) and HW (0.24±0.04). A low and a relatively high heritability were found for DC (0.08±0.02) and SB (0.44±0.04), respectively. Regarding skin color traits, the common environment effect ranged from 3.4% to 11.4% of phenotypic variance for DC and SA, respectively. In addition, the common environment effect for HW accounted for 8.7% of the phenotypic variance.
Estimate variance components and heritabilities for dorsal color (DC), intensity of lateral band (LB) and intensity of spots above (SA) and below (SB) lateral band and weight at harvest (HW) in a rainbow trout (Oncorhynchus mykiss) population
a Heritability (h 2).
b Effect of the common environment (c 2).
c
Additive variance (
$\sigma _{{\rm a}}^{2} $
).
d
Common environment variance (
$\sigma _{{\rm c}}^{2} $
).
e
Residual variance (
$\sigma _{{\rm e}}^{2} $
).
f
Phenotypic variance (
$\sigma _{{\rm p}}^{2} $
).
Phenotypic and genetic correlations
The phenotypic and genetic correlations between DC, LD, SA, SB and HW are shown in Table 6. All the genetic correlations were significant and positive, except for the genetic correlations between DC and SA and DC and HW, which were not significantly different from zero. The highest and the lowest significant genetic correlations were found between SB and SA (0.73±0.07) and SB and HW (0.25±0.11), respectively. Thus, the magnitude of all of the significant genetic correlations were from moderate to high values. All the phenotypic correlations were significant and positive, except for the phenotypic correlations between DC and HW, which was significantly different from zero of low magnitude and negative, and SB and HW that was not significantly different from zero.
Phenotypic correlations (below diagonal) and genetic correlations (above diagonal) for dorsal color (DC), intensity of lateral band (LB) and intensity of spots above (SA) and below (SB) lateral band and weight at harvest (HW) in a rainbow trout (Oncorhynchus mykiss) population
Discussion
The genetic analysis performed in the present study revealed significant genetic variation for skin color traits. There are few previous studies aimed at determining heritability values for skin color traits in farmed rainbow trout. For instance, low to moderate heritability values for skin color, ranging from 0.13±0.03 to 0.29±0.10, were found when the trait was scored in three different categories: silver shining, dark silver and dark color (Kause et al., Reference Kause, Ritola, Paananen, Eskelinen and Mäntysaari2003 and Reference Kause, Ritola and Paananen2004). In addition, moderate to high heritability for melanin-based skin spottiness, ranging from 0.45±0.12 to 0.6±0.05, were found when the trait was categorized as few/little, moderate and many/large spots (Kause et al., Reference Kause, Ritola, Paananen, Eskelinen and Mäntysaari2003 and Reference Kause, Ritola and Paananen2004). In the present study, we found a low heritability for DC (scored gray, blue or light green to brown), moderate values for intensity of the LB and spottiness above it (SA), and a moderately high heritability for spottiness below the LB (SB) (Table 5). These results confirm the existence of genetic variation for skin color-related appearance traits in farmed populations of rainbow trout using different trait definitions involving different areas of the skin surface. It is important to mention that, there is a high variation between years for DC, with an extreme frequency distribution in 2004, with 94% of fish in the same category. This could be a bias due to the person recording the trait or a high environmental effect on this trait during this particular year and may be explaining the low heritability value for DC. We also found a significant common environment effect for all skin color traits, which can be associated with the separate rearing of full-sib families before tagging, ranging from c 2=0.03 to 0.11. This common environment effect was also present for skin color traits in previous studies, however, the range of values for this effect was slightly smaller (c 2=0.01 to 0.04) (Kause et al., Reference Kause, Ritola, Paananen, Eskelinen and Mäntysaari2003 and Reference Kause, Ritola and Paananen2004).
Previous studies have demonstrated the existence of significant genetic variation for BW in rainbow trout. For example, low to moderate heritability values have been estimated for BW ranging from 0.16±0.02 to 0.37±0.04 (Henryon et al., Reference Henryon, Jokumsen, Berg, Lund, Pedersen, Olesen and Slierendrecht2002; Pante et al., Reference Pante, Gjerde, McMillan and Misztal2002; Haffray et al., Reference Haffray, Bugeon, Pincent, Chapuis, Mazeiraud, Rossignol, Chatain, Vandeputte and Dupont-Nivet2012; Janhunen et al., Reference Janhunen, Kause, Vehviläinen and Järvisalo2012; Sae-Lim et al., Reference Sae-Lim, Kause, Janhunen, Vehviläinen, Koskinen, Gjerde, Lillehammer and Mulder2015; Flores-Mara et al., Reference Flores-Mara, Rodríguez, Bangera, Lhorente, Neira, Newman and Yáñez2017). Here, we found heritability for BW at harvest of moderate magnitude (0.23±0.04) and within the range previously reported. The presence of a significant common environment effect for BW measured at different times in salmonid species, which might lead to an overestimation of the genetic variance for this trait in a particular population if not included in the analysis has also been demonstrated previously. Here, we also found a significant effect associated with the common environment for HW due to the common rearing of full siblings before tagging. The magnitude of the common environment effect for HW expressed as a ratio of the phenotypic variance is slightly higher (c 2=0.09) than those previously reported for rainbow trout: c 2=0.02 to 0.08 (Pante et al., Reference Pante, Gjerde, McMillan and Misztal2002; Janhunen et al., Reference Janhunen, Kause, Vehviläinen and Järvisalo2012; Sae-Lim et al., Reference Sae-Lim, Kause, Janhunen, Vehviläinen, Koskinen, Gjerde, Lillehammer and Mulder2015; Flores-Mara et al., Reference Flores-Mara, Rodríguez, Bangera, Lhorente, Neira, Newman and Yáñez2017), and coho salmon and other salmonid species from the same genus, c 2=0.02 to 0.06 (Gallardo et al., Reference Gallardo, Lhorente and Neira2010; Dufflocq et al., Reference Dufflocq, Lhorente, Bangera, Neira, Newman and Yáñez2017; Yáñez et al., Reference Yáñez, Bangera, Lhorente, Barría, Oyarzún, Neira and Newman2016a).
Phenotypic and genetic correlations among skin color traits and spotting were mostly positive and of moderate to high magnitude, except for the correlation between DC and SA which was not significantly different from zero (Table 6). These results indicate positive genetic relationships among different skin color traits. Such correlations suggest either positive pleiotropy among the different skin color traits or linkage disequilibrium between the regions controlling them. The pleiotropic effects of key regulators of melanogenesis have been previously suggested as responsible of a widespread association between melanin-based coloration and other behavioral and phenotypic traits in vertebrates, such as growth (Ducrest et al., Reference Ducrest, Keller and Roulin2008). From a genetic improvement perspective, these results also indicate that skin color traits can be simultaneously improved by artificial selection.
In the present study, we used mixed linear models to estimate genetic parameters on the observed scale for color skin patterns traits, which are traits of a categorical nature. This can have implications on the genetic parameters estimation. For instance, Kause et al. (Reference Kause, Ritola, Paananen, Eskelinen and Mäntysaari2003 and Reference Kause, Ritola and Paananen2004) found that the heritability values of skin color patterns presented in the underlying scale and estimated using a threshold model were always higher than the heritability values presented on the observed scale and estimated using a linear model. Thus, the heritability values for the categorical traits presented here may be underestimated if compared to values obtained using threshold models. We used a linear multi-trait approach to assess the genetic relationships between all the skin color pattern traits and BW. This approach was chosen in order to compare the results against the previous studies by Kause et al. (Reference Kause, Ritola, Paananen, Eskelinen and Mäntysaari2003 and Reference Kause, Ritola and Paananen2004), in which the authors did not fit multi-trait threshold models because of computing limitations. The genetic correlations presented here are on the observed scale, and they can be directly compared against those obtained by Kause et al. (Reference Kause, Ritola, Paananen, Eskelinen and Mäntysaari2003 and Reference Kause, Ritola and Paananen2004).
The amount of melanin-based spots and the intensity of the LB were positively correlated with HW. However, the correlation between DC and HW was not significantly different from zero. Kause et al. (Reference Kause, Ritola, Paananen, Eskelinen and Mäntysaari2003) found that correlations among skin color patterns and conformational body traits, including condition factor, body shape and body mass measured at 2 and 3 years of age were not significantly different from zero. It is important noting that when a fish has few or many spots after smoltification, these will remain almost invariable during growth and under different rearing environments (fresh water and sea) while the skin color may be more flexible (Kause et al., Reference Kause, Ritola and Paananen2004). It has also been reported that rapid growth is correlated with the silvery appearance of salmonid smolts (Rodgers et al., Reference Rodgers, Ewing and Hall1987). Thus, our results differ from previous studies and indicate that melanin-based spottiness and intensity of LB have a positive genetic relationship with growth in rainbow trout. These differences may be due to the fact that we used a different population of farmed fish, implying a different genetic background and consequently different allele frequencies and linkage disequilibrium between causative variants controlling skin color patterns and growth. It is also important noting that we used a larger dataset compared to previous studies, which can have had an impact on statistical power to detect significant correlations between skin color patterns and BW. Another potential explanation for the differences between the present and previous studies are differences in trait definitions; for instance, Kause et al. (Reference Kause, Ritola and Paananen2004) used body shape, skin color and skin spots as appearance traits each with three scoring categories, Kause et al. (Reference Kause, Ritola, Paananen, Eskelinen and Mäntysaari2003) used the same trait definitions plus body mass measured at age of 2 and 3 years and condition factor, while in the present study we used appearance scores with three categories for DC and four categories for intensity of LB and intensity of spots above and below LB and BW measured at harvest. The differing results between this work and previous studies must be considered, and further evidence must be carefully evaluated before making general conclusions on the relationships between skin color patterns and growth in salmonids.
The positive relationship between skin pigmentation patterns and BW, and their link with different types of physiological-behavioral individual responses (i.e. coping styles) must be further investigated (Castanheira et al., Reference Castanheira, Conceicao, Millot, Rey, Bégout, Damsgard, Kristiansen, Höglund, Øverli and Martins2015). A previous study have shown that rainbow trout with more melanic pigmentation have a reduced cortisol response (Kittilsen et al., Reference Kittilsen, Ellis, Schjolden, Braastad and Øverli2009b). Lower cortisol release associated with LR individuals has also been reported to be associated with other beneficial characteristics for aquaculture such as higher feed intake and feed motivation (Øverli et al., Reference Øverli, Pottinger, Carrick, Øverli and Winberg2002), higher disease resistance (Kittilsen et al., Reference Kittilsen, Johansen, Braastad and Øverli2012) and ease in routine formation (Ruíz-Gomez et al., 2011). On the other hand, HR individuals, that have lighter pigmentation, have expressed higher mortality and lower capacity to cope with multiple stressors (Fevolden et al., Reference Fevolden, Røed and Fjalestad2003), such as those present in aquaculture systems. Feed intake differences have also been reported between LR and HR individuals. For example, Kittilsen et al. (Reference Kittilsen, Schjolden, Beitnes-Johansen, Shaw, Pottinger, Sørensen, Braastad, Bakken and Øverli2009a) reported a generally quicker resumption of normal feeding behavior in spotted (LR) salmon after confronted with a stressor (isolation), and also less time moving during acute stress, decreasing energy loss. Based on the literature, it seems that there is an association between skin and response to farm conditions, which lead us to support the hypotheses that the favorable genetic correlation between skin color patterns and growth might be explained by genetic links between coping styles and both appearance and growth traits. Thus since in our study, there is a moderate correlation between skin color and growth, it is possible that selection for growth will impact positively or at least not negatively on response to farm conditions. However, this hypothesis should be confirmed by estimating genetic correlations between growth and response to farm conditions. Thus, the genetic relationship between skin color patterns and different coping styles (HR and LR), and its potential impact on growth-related traits in farmed rainbow trout still remains to be proven in further studies.
The commercial implications of skin color and spottiness need to be taken into account; since they are important production traits that affect consumers’ acceptance (Colihueque, Reference Colihueque2010). Although to our knowledge no studies have addressed Chilean consumers perception of these traits, because one of the main markets is the Japanese one, this external aspect is critical since Japanese consumers demand specimens with few spots and silvery (Taub and Palacios, Reference Taub and Palacios2003). Thus, the correlated response of selection for growth towards more spotted fish could have a negative commercial impact for Chilean exports of rainbow trout products to the Japanese market.
In the present study, we found evidence establishing a positive correlation between HW and skin pigmentation patterns, including melanin-based spottiness and intensity of the LB. Melanin has been linked to the physiological stress response and behavior in different vertebrates (Ducrest et al., Reference Ducrest, Keller and Roulin2008). For example, the melanin-like pigment is positively correlated with dominance in mammals, birds and salmonids (Kittilsen et al., Reference Kittilsen, Ellis, Schjolden, Braastad and Øverli2009b; Backström et al., Reference Backström, Heynen, Brännäs, Nilsson and Magnhagen2015), suggesting that those salmon most densely spotted are more aggressive, and more dominant, so that the feed consumption after the stress is not affected. In addition, it has been reported that those salmon with a large number of skin spots are more resistant to stress and ectoparasites (Kittilsen et al., Reference Kittilsen, Schjolden, Beitnes-Johansen, Shaw, Pottinger, Sørensen, Braastad, Bakken and Øverli2009a; Øverli et al., Reference Øverli, Nordgreen, Mejdell, Janczak, Kittilsen, Johansen and Horsberg2014). Alternatively, Pérez et al. (Reference Pérez, Cánepa, Fossati, Fernandino, Delgadin, Canosa, Somoza and Vissio2012), in an in vitro assay with Cichlasoma dimerus showed that the concentration of melanin is involved in the regulation of skin color in the fish and is also responsible for regulating the expression and synthesis of growth hormone. Furthermore, melanin can regulate food intake in mammals and fish, acting as an appetite stimulant (Takahashi et al., Reference Takahashi, Tsuchiya, Yamanome, Amano, Yasuda, Yamamori and Kawauchi2004; Yamanome et al., Reference Yamanome, Amano and Takahashi2005). Our results clearly show that higher pigmentation is genetically associated with higher HW in trout. If this increased BW is genetically associated with the effect that melanin triggers at the physiological and cellular level in farmed rainbow trout still remains to be proven.
Further studies aimed at unraveling the molecular mechanisms of the genetic relationships between skin pigmentation patterns and growth found in this study are needed. With the advent of high throughput genome-wide single nucleotide polymorphism genotyping methods, already available for salmonid species (Houston et al., Reference Houston, Taggart, Cézard, Bekaert, Lowe, Downing, Talbot, Bishop, Archibald, Bron, Penman, Davassi, Brew, Tinch, Gharbi and Hamilton2014; Palti et al., Reference Palti, Gao, Liu, Kent, Lien, Miller, Rexroad and Moen2015; Yáñez et al., Reference Yáñez, Naswa, López, Bassini, Correa, Gilbey, Bernatchez, Norris, Neira, Lhorente, Schnable, Newman, Mileham, Deeb, Di Genova and Maass2016b), and next-generation sequencing technologies, a better understanding of the molecular basis of complex traits at the genomic level is expected in the near future (Yáñez et al., Reference Yáñez, Newman and Houston2015). In fact, genomic technologies have recently allowed the identification of genomic regions involved in growth traits in different salmonid salmonid species by means of genome-wide association studies (Gutierrez et al., Reference Gutierrez, Yáñez, Fukui, Swift and Davidson2015; Tsai et al., Reference Tsai, Hamilton, Tinch, Guy, Gharbi, Stear, Matika, Bishop and Houston2015; Gonzalez-Pena et al., Reference Gonzalez-Pena, Gao, Baranski, Moen, Cleveland, Kenney, Vallejo, Palti and Leeds2016; Yoshida et al., Reference Yoshida, Lhorente, Carvalheiro and Yáñez2017). The discovery of functional variants for this and other traits could be facilitated by the international initiative on Functional Annotation of All Salmonid Genomes (Macqueen et al., Reference Macqueen, Primmer, Houston, Nowak, Bernatchez, Bergseth, Davidson, Gallardo-Escárate, Goldammer, Guiguen, Iturra, Kijas, Koop, Lien, Maass, Martin, McGinnity, Montecino, Naish, Nichols, Ólafsson, Omholt, Palti, Plastow, Rexroad, Rise, Ritchie, Sandve, Schulte, Tello, Vidal, Vik, Wargelius and Yáñez2017). Further studies are required in order to identify genomic regions associated to skin pigmentation patterns. These studies will be crucial to confirm and provide further insights into the genetic relationship between skin pigmentation patterns and HW. Further research should aim study in detail the association between these traits and the implications they can have for response to stress and fish welfare in aquaculture systems.
Acknowledgments
The authors would like to acknowledge Aguas Claras S.A. for providing the data used in this study. F. H. R. wants to acknowledge Beca Presidente de la República from the Government of Peru. This study was partially funded by FONDECYT REGULAR N° 1171720. José M. Yáñez is supported by Nucleo Milenio INVASAL funded by Chile’s government program, Iniciativa Cientifica Milenio from Ministerio de Economia, Fomento y Turismo.
Declaration of interest
The authors declare no conflict of interest.
Ethics statement
The research protocol was approved by bioethics committee of Chile University.
Software and data repository resources
The data sets and programs used in the current study are available from the corresponding author on reasonable request.
