Thermal sensitivity of growth indicates heritable variation in 1-year-old rainbow trout (Oncorhynchus mykiss)

Janhunen e al. Gene Sel E ol (2016) 48:94 DOI 10.1186/s12711-016-0272-3 RESEARCH ARTICLE The mal sensi i i y o g ow h indica es he i able a ia ion in1-yea -old ainbow ou (Onco hynchus mykiss) Ma i Janhunen1*, Juha Koskela2, Nguyễn Hữu Ninh3, Ha i Veh iläinen1, Heikki Koskinen4, An i Nousiainen4 and Ngô Phú Thỏa5 Abs ac Backg ound: Rainbow ou is an impo an aquacul u e species, which has a wo ldwide dis ibu ion ac oss a ious p oduc ion en i onmen s. The di e se loca ions o ou a ms in ol e ema kable a ia ion in en i onmen al ac o s such as wa e empe a u e, which is o majo impo ance o he pe o mance o ish. Thus, obus ish ha could h i e unde di e en and subop imal he mal condi ions is a desi able goal o ou b eeding. Using a spli - amily expe imen al design (40 ull-/hal -sib g oups) o a ainbow ou popula ion de i ed om he Finnish na ional b eeding p og am, we s udied how wo di e en ea ing empe a u es (14 and 20 °C) a ec eed in ake, g ow h a e and eed con e sion a io in 1-yea -old ish. Fu he mo e, we quan i ied he addi i e gene ic (co-) a ia ion o daily g ow h coe icien (DGC) and i s he mal sensi i i y (TS), de ined as he slope o he g ow h eac ion no m be ween he wo empe a u es. Resul s: The ish showed consis en ly lowe eed in ake, as e g ow h and be e eed con e sion a io a he lowe empe a u e. He i abili y o TS o DGC was mode a e ( h2 TS = 0.24 ). The co-he i abili y pa ame e de i ed om selec ion index heo y, which desc ibes he he i able a iance o TS, was nega i e when he in e cep was placed a he lowe empe a u e (−0.28). This esul ed in mode a e accu acy o selec ion. A he highe empe a u e, co-he i abili y o TS was posi i e (0.20). The gene ic co ela ion be ween DGC and i s TS was s ongly nega i e (−0.64) when he in e cep was a he lowe empe a u e and posi i e (0.38) bu no signi ican ly di e en om ze o a he highe empe a u e. Conclusions: The conside able amoun o gene ic a ia ion in TS o g ow h indica es a po en ial o selec ion esponse and hus o a ge ed gene ic imp o emen in TS. The nega i e gene ic co ela ion be ween DGC and i s TS sugges s ha selec ion o high g ow h a e a he lowe empe a u e will esul in mo e empe a u e-sensi i e ish. Ins ead, he co ela ed esponse o TS is less p onounced i he selec ion o a highe DGC occu ed a he highe empe a u e. I seems possible o con ol he co ela ed gene ic change o TS while selec ing o as g ow h ac oss en i onmen s, especially i measu emen s om bo h en i onmen s a e a ailable and b eeding alues o eac ion no m slope a e di ec ly included in he selec ion index. © The Au ho (s) 2016. This a icle is dis ibu ed unde he e ms o he C ea i e Commons A ibu ion 4.0 In e na ional License (h p://c ea i ecommons.o g/licenses/by/4.0/), which pe mi s un es ic ed use, dis ibu ion, and ep oduc ion in any medium, p o ided you gi e app op ia e c edi o he o iginal au ho (s) and he sou ce, p o ide a link o he C ea i e Commons license, and indica e i changes we e made. The C ea i e Commons Public Domain Dedica ion wai e (h p://c ea i ecommons.o g/ publicdomain/ze o/1.0/) applies o he da a made a ailable in his a icle, unless o he wise s a ed. Backg ound Di e en geno ypes, which a e ypically e e ed o as sib-g oups, s ains o popula ions, may di e in hei a e age pe o mance esponse o en i onmen al a i- ables. In wild popula ions, an o ganism’s abili y o mod- i y i s pheno ype in esponse o en i onmen al changes ( e med pheno ypic plas ici y) can i sel be an adap i e li e-his o y ai , which is subjec o na u al selec ion [1–4]. Pheno ypic plas ici y is conside ed synonymous o mac o-en i onmen al sensi i i y, which is a mo e com- monly used e m in he animal b eeding con ex [5]. Fo animal b eede s, mac o-en i onmen al sensi i i y is an impo an aspec due o i s associa ion wi h he animals’ pe o mance ac oss p oduc ion en i onmen s, and wi h hei obus ness (s abili y) and wel a e [6–9]. Fo any measu able pheno ypic ai , he mac o-en i- onmen al sensi i i y o a geno ype can be illus a ed as Open Access G ene ics S elec ion E olu ion *Co espondence: [email p o ec ed] 1 Biome ical Gene ics, Na u al Resou ces Ins i u e Finland (Luke), Mylly ie 1, 31600 Jokioinen, Finland Full lis o au ho in o ma ion is a ailable a he end o he a icle Page 2 o 11 Janhunen e al. Gene Sel E ol (2016) 48:94 he esponse unc ion wi h en i onmen al change [10]. Assuming a linea eac ion no m, he deg ee o sensi- i i y o a geno ype can be quan i ied by he eg ession slope o a geno ype’s pe o mance ac oss an en i onmen- al g adien [11, 12]. The exis ence o mac o-en i onmen- al sensi i i y o a gi en ai is indica ed by slopes ha de ia e om ze o, whe eas la eac ion no ms ac oss he en i onmen al g adien axis e lec s abili y o he ai . Since he eac ion no ms also depic he ex en o e- anking among geno ypes and he change in addi i e gene ic a iance wi h he en i onmen (i.e., wo o ms o geno ype×en i onmen in e ac ion) hey can p o ide in o ma ion abou he capaci y o popula ions and spe- cies o adap o en i onmen al a iabili y [13]. The ainbow ou , Onco hynchus mykiss (Walbaum), is an example o a globally impo an aquacul u e species, which is dis ibu ed ac oss a ious p oduc ion en i on- men s and sys ems, which ange om o sho e ne cages o land-based e-ci cula ion acili ies. The di e se geog aphi- cal loca ions o ainbow ou a ms may in ol e consid- e able a ia ion in many abio ic (e.g., wa e empe a u e, salini y, and pho ope iod) and bio ic (quali y o eed, pa h- ogens and pa asi es) ac o s, which a e o majo impo - ance o he pe o mance o ish. High g ow h capaci y may be conside ed wo ldwide as he single mos economi- cally impo an ai o be imp o ed by selec i e b eeding [14], bu he capaci y o ish o exp ess he selec ed g ow h po en ial unde a iable o subop imal en i onmen al con- di ions may be cons ained. The e o e, a mo e obus ish ma e ial wi h s abile g ow h would be an eligible p oduc o b eeding unde a iable en i onmen al condi ions. Being na i e o cool, empe a e egions o he no he n hemisphe e, he ainbow ou , like all o he salmonids, is adap ed o ela i ely low wa e empe a u es [15, 16]. S a- bili y o g ow h is o special impo ance in ou a ming a eas, whe e ea ing empe a u es emain cons an ly high o whe e s ong seasonal wa ming occu s. Fu he mo e, due o global wa ming, i is likely ha he e will be an inc easing demand in he ish a ming sec o o popula- ions o mo e hea - ole an ou in he u u e. To assess whe he he mal sensi i i y (TS) o g ow h has he po en- ial o be changed by selec ion, an es ima e o he addi- i e gene ic componen in he slopes o eac ion no ms is needed. The exis ence o gene ic a ia ion in g ow h esponses as he empe a u e changes would enable he de elopmen o mo e empe a u e- ole an o locally- adap ed popula ions o di e en he mal condi ions. In his s udy, we i s in es iga ed a a gene al (popula- ion) le el how wo di e en ea ing empe a u es (14°C, namely ‘low’, and 20 °C, namely ‘high’) in luence eed in ake, g ow h and eed con e sion a io in 1-yea -old ainbow ou . Second, by using a spli - amily design o he expe imen , we quan i ied he addi i e gene ic (co-) a ia ion o g ow h a e (daily g ow h coe icien , DGC) and i s TS, he la e ai being de ined as he slope o he eac ion no m be ween he wo empe a u e condi ions. Me hods S udy ma e ial The ish used in his s udy we e de i ed om he Finnish na ional b eeding p og am ha is main ained a he Te o ish a m (b eeding nucleus) by he Na u al Resou ce Ins i u e Finland (Luke). The pheno ypic da a comp ised 800 indi iduals om 40 amilies, which we e c ea ed in Ap il 2013 using a pa ial ac o ial ma ing design o 35 si es and 26 dams. Each si e was ma ed o an a e age o 1.1 dams ( anging om 1 o 3) and each dam o an a e - age o 1.5 si es ( anging om 1 o 3). The a e age num- be o o sp ing was 22.9 pe si e ( anging om 20 o 60) and 30.8 pe dam ( anging om 20 o 60). The pa en al ish we e selec ed using a mul i- ai selec ion index wi h he main weigh on imp o ed g ow h (50% o he index). The pedig ee ile included 1661 indi iduals and nine gen- e a ions acing back o he base popula ion es ablished in 1989 and 1990 (see Addi ional ile1). Rea ing p o ocol In his s udy, he p o ocols used we e app o ed by he FGFRI Animal Ca e Commi ee, Helsinki, Finland. The i s 6mon hs o ea ing ook place in he b eeding nucleus, whe e he ull-sib amilies we e ea ed sepa a ely in ound 150-L indoo anks un il agging. Va ia ion in ea ing empe a u e ollowed a ia ion in ambien wa e - way h oughou ha pe iod ( anging om 0 o 20°C om he s a o ha ching un il he s a o id- agging). Du ing he pe iod be ween Janua y 10 and 21 2014, 25 andomly chosen ish om each s udied amily (73.6±14.4g, mean weigh ±SD) we e indi idually agged wi h passi e in e- g a ed ansponde s (Bioma k, Inc., Boise, Idaho, USA). The agged ish we e anspo ed in o he communal pool a he Laukaa ish a m, whe e hey we e ea ed unde ambien empe a u e (1–10°C) and ligh condi ions (day leng h 16.30–21.30h du ing he las mon h p io o he expe imen ) un il s a o he expe imen . In June 2014, 20 agged ish pe amily we e andomly sampled o he empe a u e ial. The ea ing empe a u es we e g adu- ally inc eased o he expe imen al empe a u es (14 and 20°C) in he cou se o 3days. To cons uc a spli - amily design, each o he 40 amilies was i s andomly spli in o wo g oups o be ea ed a low and high empe a- u es. These g oups we e e enly dis ibu ed o e 4+4 (low empe a u e) and 4+4 (high empe a u e) ound 0.4-m3 g een plas ic anks ( wo eplica e anks pe am- ily). In o al, he empe a u e ial began wi h a o al o 800 ish, each ank con aining 50 ish ( en amilies and i e ish om each amily). Page 3 o 11 Janhunen e al. Gene Sel E ol (2016) 48:94 The ial was conduc ed om June 3 o Augus 12 2014. The ish we e ed adlibi um 6h pe day (4.00 am o 10.00 am) using bel eede s wi h comme cial ou die (Rai- sioag o L d, Finland Vi al p o LP; chemical composi ion gi en by he manu ac u e 3.5/5.0 mm; c ude p o ein 43.0/40.0%, c ude a 28.0/30.0%, c ude ib e 1.5/1.5%, ash 6.5/6% and g oss ene gy 24.4/24MJkg−1). F om days 1 o 10, he ish we e ed wi h 3.5-mm pelle s, ollowed by a mix u e (1:1) o 3.5- and 5.0-mm pelle s om day 11 o 20, and he ea e wi h 5.0-mm pelle s un il he end o ial. Du ing he expe imen , daily eeding amoun was inc eased so ha he sha e o was e eed anged om 0 o 35% o he le el o daily eed. The numbe s o unea en pelle s we e collec ed a he ank ou le in a box wi h a mesh bo om. The daily numbe o was e pelle s was cal- cula ed, and hei weigh was es ima ed by mul iplying he numbe o was e pelle s by he ai d y weigh o a pelle . Be o e calcula ions, i e 100-pelle subsamples we e aken om each o he die s o measu e ai d y weigh o pelle s. The daily in ake o a ank’s popula ion was calcula ed as he di e ence in weigh be ween he ed and was e eed. Tanks o he low- empe a u e g oup we e supplied wi h esh lake wa e and anks o he high- empe a u e g oups we e supplied wi h wa e ia semi-in ensi e eci cula - ing aquacul u e sys ems (RAS; low a e o makeup wa e 4–6m3kg−1 eed). Pu e oxygen was added o incoming wa e o bo h empe a u e g oups o imp o e wa e oxy- gen con en and baking soda was added o he RAS o main ain he pH be ween 6.7 and 7.0. Du ing he expe imen , he wa e empe a u e was au oma ically eco ded hou ly (low- empe a u e g oup 14.1±1.0°C and high- empe a u e g oup 20.4±1.8°C; mean±SD). Wa e oxygen sa u a ion (%) was eco ded once e e y second week (low- empe a u e/high- em- pe a u e; ank inle : 96.5±7.8/105.7±4.9, ank ou - le : 78.6±6.8–83.3±7.1/82.0±3.5–86.8±5.2) and o he wa e quali y pa ame e s we e eco ded weekly (high empe a u e pH 6.9±0.1, o al ammonia mg L−1 (NH3+NH4) 0.05±0.04, un-ionized ammonia mg L−1 (NH3) <0.001, ni i e mg L−1 (NO2) 0.07±0.01, ni a e mg L−1 (NO3) 4.45±0.7). A 24-h whi e ligh was p o- ided wi h led lamps on he ank co e . Measu emen s andcalcula ions Indi idual ag numbe and body weigh ( o he nea es g) we e eco ded a he beginning (129±28g; mean±SD, n=800 ish) and end o he ial (516±98g, n=785). Fi een ish died du ing he ial, hus only he ini ial body weigh was a ailable o hese. Daily g ow h coe - icien s (DGC, % day−1) o 785 ish we e calcula ed as [17]: DGC =  BW1/3 2−BW1/3 1  /  × 100, whe e BW1 and BW2 a e he body weigh o he ish a he s a and end o he expe imen , and is he du a ion o he expe imen (69–70days) (see Addi ional ile2). Unlike speci ic g ow h a e (SGR), ano he widely used measu e o ish g ow h a e, he DGC is independen o ish body weigh and ime in e al be ween weighings a a gi en empe a u e [17]. This was also alida ed o ainbow ou [18]. The mean eed in ake o a ank’s popula ion pe day (FImean, gday−1) was calcula ed as FIcum/ 1, whe e FIcum is cumula i e eed in ake (g) o a ank’s popula ion du - ing he pe iod o eed in ake measu emen s di ided by he numbe o measu emen days ( 1=57d). The ela- i e eed in ake (FI % biomass−1day−1) was calcula ed as ollows: 100 × FImean/[(biomass1 + biomass2)/2] , whe e biomass1 and biomass2 a e he ini ial and inal ank biomasses (g), espec i ely (see Addi ional ile 3). Feed con e sion a io (FCR) was calcula ed as FImean × /(biomass2 − biomass1) . S a is ical analyses The di e ence in DGC means be ween empe a u e ea men s was es ed o indi idual da a using es ic ed maximum likelihood me hod in SAS® 9.4 (MIXED p o- cedu e; SAS® Ins i u e, Ca y, NC, USA). The model used was: whe e y is he obse a ion o he i h indi idual, ea - men j is he ixed e ec o empe a u e ea men (j=1–2), ankj is he andom e ec o ea ing ank du - ing he ial (k=1–16), nes ed wi hin ea men , and eijk is he andom e o e m. E o a iances we e modelled sepa a ely o each empe a u e condi ion. In addi ion, deg ees o eedom o he es o he ixed e ec we e co ec ed using he me hod o Kenwa d and Roge [19]. Fo he ela i e eed in ake and FCR, ank popula ion alues we e used as obse a ions (n=8 pe ea men g oup) and he analysis o a iance (ANOVA) was used o compa e he di e ences in means be ween he ea - men g oups. No co a ia e was used in hese models. Gene ic (co) a iance o TS ( eg ession slope) o DGC was es ima ed using a linea andom eg ession model (also e med as a eac ion no m model). (Co) a iance componen s we e es ima ed by es ic ed maximum like- lihood in ASReml 3.0 [20]. App oxima e s anda d e o s we e calcula ed wi h ASReml acco ding o Fishe e al. [21]. The linea andom eg ession model was as ollows: whe e βin is he ixed eg ession coe icien o he popula ion in e cep (in ) and βsl is he o e all ixed eg ession slope (sl) o he ai on he h- h le els o an (1) yijk = ea men j+ ank( ea men )k + eijk, (2) yhij = βin + βslXh + ai,in + ai,slXh + ehij, Page 4 o 11 Janhunen e al. Gene Sel E ol (2016) 48:94 en i onmen al g adien Xh. Xh is he eg esso o he en i onmen s in which he in e cep was placed ei he on he low o high empe a u e en i onmen (a Xh=0). The alue o Xh was 6 o high empe a u e and −6 o low empe a u e when hese en i onmen s we e no used a he in e cep . The scale o X is equi alen o he di - e ence in expe imen al empe a u es, applying a uni o 1°C. The alues o 0 and 6 (o −6) we e used ins ead o he ac ual empe a u e alues (14 and 20°C) in o de o ha e an app op ia e in e p e a ion o he in e cep . ai is he andom gene ic e ec o he in e cep and slope o eac ion no m,  a, in a, sl  ∼MVN[0,A⊗G ] , whe e MVN is a mul i a ia e no mal dis ibu ion, A is he addi i e gene ic ela ionship ma ix de i ed om he pedig ee aced back o he base popula ion, and G is he addi i e gene ic co a iance ma ix: G =  σ 2 a,in σa,in ,sl σa,in ,sl σ2 a,sl  , whe e σ2 a,in and σ2 a,sl a e he addi i e gene ic a iances o he in e cep and slope, espec i ely, and σa,in ,sl is he addi- i e gene ic co a iance be ween he in e cep and slope. e ∼N  0,  Iσ 2 e10 0Iσ2 e2 is he andom esidual e ec o indi idual i in en i onmen h whe e I is he iden i y ma ix wi h a di e en esidual a iance o each en i- onmen . In addi ion, he andom e m ankk × ullsibl , accoun ing o he in e ac ion e ec o expe imen al ank and ull-sib amily (modelled wi hou he e ec on he slope; k=1–8 a low empe a u e and 11–18 a high empe a u e, l=1–40), was es ed. This a iance pa ame e ook he pe manen en i onmen e ec s in o and ano he i e si es wi h close o ze o EBV o he slope ( he leas sensi i e) o d awing he eac ion no ms. Since he e we e only wo en i onmen s, he EBV o DGC o he high empe a u e en i onmen could be de i ed om a simple equa ion: The gene ic alues o he slope a e mul iplied by 6 o adjus he eac ion no ms o he scale o X (change o 6°C). Fo he in e cep o he eac ion no ms, he i abili y ( h 2 in ) was calcula ed as: h2 in =  σ 2 a,in /  σ 2 P,in , whe e  σ2 P,in is he pheno ypic a iance o DGC in he in e cep en i onmen when X=0 (equal o he sum o addi i e gene ic and esid- ual a iance in he in e cep en i onmen :  σ2 a,in +  σ 2 e,in ). Because he e is no pheno ypic a iance o he slope, he s ic sense he i abili y canno be calcula ed. The e- o e, wo al e na i e pa ame e s we e used o desc ibe he gene ic cha ac e is ics o TS. Following Sae-Lim e al. [23], he he i abili y o TS ( h2 TS ) was calcula ed as: whe e  σ2 a,sl ×  σ 2 a,X is he addi i e gene ic a iance o he slope mul iplied by he a iance o X, espec i ely.  σ2 X is equal o 18 in his s udy, since he alues o X a e 0 and 6. The s anda dized nume a o in Eq.(4) is equi alen o he a iance o he geno ype by en i onmen (GxE) in e - ac ion, which is independen om he di e en scales o an en i onmen al a iable X . The denomina o  σ2 P,To al was de ined as ollows: (3) EBV DGC,highT = EBV in ,lowT + 6 × EBV sl . (4) h 2 TS =  σ 2 a,sl ×  σ 2 a,X  σ2 P,To al , � σ 2 P,To al =   (nlowT −1) � σ2 PDGC,lowT + � nhighT −1 �� σ2 PDGC,highT +nlowTnhighT � GlowT −GhighT �2 /nlowT +nhighT nlowT +nhighT −1  , accoun , which we e caused by di e en ea ing anks be ween and wi hin amilies. Howe e , based on he like- lihood a io es , inclusion o he ankk × ullsibl e m did no a ec he i o he model (χ2=0.001, p=0.486). The e o e, he esul s a e only p esen ed o he models in which his a iance pa ame e was omi ed. Bo h he magni ude and sign o a gene ic co ela ion be ween he in e cep and slope, as well as he gene ic a iance o he ai , depend on he en i onmen o which he in e cep is de ined [22]. The e o e, he andom eg es- sion model was un wice, ei he wi h low (Xh1 = 0) o high empe a u e ea men (Xh2 = 0) as he in e cep en i onmen . The co a iance be ween es ima ed b eeding alues (EBV) o DGC a in e cep (in ) and slope (sl) was g aphically illus a ed by choosing i e si es wi h he high- es absolu e EBV (i.e., gene ically he mos sensi i e si es) whe e n is he numbe o indi iduals wi h a eco d o an animal ai ,  σ2 PDGC is he es ima ed pheno ypic a iance o he ai and G is he aw pheno ypic mean o DGC. The denomina o was de i ed om Scheine ’s app oach, whe e he o al pheno ypic a iance ac oss en i onmen s was calcula ed om an analysis o a iance [24]. How- e e , i is impo an o no e ha  σ2 P,To al is no he pheno- ypic a iance o TS, and hus h2 TS is a desc ip i e a he han a p edic i e pa ame e [24]. The de ini ion o he i - abili y in Eq.(4) does no co espond wi h he con en- ional de ini ion o he i abili y, which is he eg ession o b eeding alue on pheno ype. Following Sae-Lim e al. [23], an al e na i e measu e ha is called co-he i abili y was de ined on he basis o selec ion index p inciples. Co-he i abili y is exp essed as he eg ession coe icien (b) o he b eeding alue Page 5 o 11 Janhunen e al. Gene Sel E ol (2016) 48:94 o slope on he pheno ype P and de ines he he i able gene ic a iance o TS o DGC when he selec ion c i e- ion is DGC in one en i onmen . The pheno ypic a i- ance ( σ2 P ) o a ai is: Co-he i abili y can a y in magni ude and ha e bo h posi i e and nega i e alues, depending on which en i- onmen is se as in e cep en i onmen (X=0). The sign o he co-he i abili y explains he change in co e- la ed esponse o TS when mass selec ion o highe phe- no ypic alues o DGC is p ac iced in one en i onmen [25]. He e, he selec ion on DGC was assumed o be pe - o med ei he a low o high empe a u e (in e cep en i- onmen ; X=0), which gi es he co-he i abili y b in he ollowing equa ion: whe e  σ2 Ph is he pheno ypic a iance o DGC in he selec- ion en i onmen h. Co-he i abili y is p edic i e o esponse o selec ion. The gene ic co ela ion be ween in e cep and slope  G(in ,sl) was calcula ed as: Finally, he accu acy ( IH) o EBV o TS when DGC is used as a selec ion c i e ion in one o he en i onmen s is equal o: whe e σP is he pheno ypic s anda d de ia ion and σa,sl is he slope s anda d de ia ion. Equa ion(4) is equi alen o ha de i ed by Kolmodin and Bijma [25]. Resul s Means o g ow h a e, eed in ake and eed con e sion a io The low- empe a u e g oup had sligh ly bu signi ican ly highe DGC means han he high- empe a u e g oup (Table1). Fo 25 o he 40 amilies, he aw (unco ec ed) DGC mean was highe a he lowe empe a u e (Fig.1). In con as , bo h eed in ake (FI) and eed con e sion a io (FCR) means we e signi ican ly lowe in he low- empe a u e g oup (Table1). (5) σ2 P =σ 2 a,in +2Xσ a,in ,sl +X 2 σ 2 a,sl +σ 2 e, (6) b = 6 σ a,in ,sl  σ2 a,in +  σ2 e,in = 6 σ a,in ,sl  σ2 Ph , (7) G(in ,sl)=  σain ,sl  σ2 a,in ×  σ2 a,sl . (8) IH =  σa,in ,sl +X  σ 2 a,sl  σ P  σ a,sl , Gene ic (co) a iance o DGC andi s he mal sensi i i y The he i abili y o he slope ha de ines TS was mod- e a e ( h2 TS = 0.24), which indica es ha he addi i e gene ic a ia ion in TS cons i u es qui e a la ge p opo - ion o he o al pheno ypic a ia ion in DGC ac oss en i onmen s. The co-he i abili y o TS was mod- e a ely nega i e (−0.28) when he lowe empe a u e was assigned as he in e cep en i onmen . This indi- ca es ha a mode a e accu acy o selec ion o TS o g ow h can be achie ed when indi idual selec ion o DGC is p ac iced in a low- empe a u e en i onmen ( IH=0.44). Ins ead, he co-he i abili y es ima e was posi i e (0.20) when a high- empe a u e en i onmen was used o he in e cep . This esul ed in lowe accu- acy (0.25), compa ed o ha o selec ion a he lowe empe a u e. Fo he DGC a he in e cep , he he i - abili y es ima es we e simila and mode a e ( h2 in =0.46) a bo h empe a u es (Table 2). Thus, he he i able Table 1 E ec o  ea ing empe a u e (lowT 14.1 °C andhighT 20.4°C) ondaily g ow h coe icien (DGC), eed in ake (FI) and eed con e sion a io (FCR, in ake/gain) d =deg ees o eedom The alues a e aw pheno ypic mean±S.D. DGC was analysed using da a om indi idual ish, whe eas ank alues we e used as obse a ions o FI and FCR T ea men DGC % day−1FI % biomass−1day−1FCR LowT 4.35 ± 0.63 1.66 ± 0.05 0.88 ± 0.01 HighT 4.19 ± 0.53 1.74 ± 0.05 0.97 ± 0.02 F- a io 5.13 8.40 115.9 d 1, 13.5 1, 14 1, 14 p alue 0.041 0.012 < 0.001 Di e e nce in DGC (highT - lowT) Family numbe -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 010203 04 0 Fig. 1 Di e ence in daily g ow h coe icien means o ainbow ou amilies ea ed a low (14.1 °C) and high (20.4 °C) empe a- u es. Families wi h a nega i e alue g ew be e a low han a high empe a u e Page 6 o 11 Janhunen e al. Gene Sel E ol (2016) 48:94 po en ial o g ow h a e did no change be ween he wo empe a u e condi ions. The gene ic co ela ion be ween DGC and i s TS was signi ican and s ongly nega i e (−0.64) when he lowe empe a u e en i onmen was used o he in e - cep (Table2). This implies ha a high DGC a he lowe empe a u e is gene ically associa ed wi h inc eased TS ac oss en i onmen s. Howe e , Fig.2 shows ha he geno ypes (si es) wi h s eep slope EBV we e no con- sis en ly hose ha showed he as es g ow h po en ial in he in e cep en i onmen . In ac , some geno ypes associa ed wi h slow g ow h also showed conside able sensi i i y o he mal change, al hough in he opposi e di ec ion. When he in e cep was placed a he highe empe a u e, he es ima ed gene ic co ela ion was posi i e (0.38) al hough no s a is ically di e en om ze o (based on he la ge s anda d e o ; Table2). Hence, he g ow h po en ial o geno ypes a he highe em- pe a u e may no show an associa ion wi h di e ences in TS when he indi iduals a e mo ed o he lowe empe a u e. Discussion The empe a u e di e ence o 6°C in ou expe imen was su icien o cause subs an ial a ia ion in he g ow h eac ion no ms among 1-yea -old ainbow ou amilies, which indica ed he i able di e ences in hei TS. Indeed, he TS o DGC in ol ed a conside able amoun o gene ic a ia ion, which indica es a po en ial o esponse o a - ge ed selec ion on TS. Bo h he desc ip i e pa ame e h2 TS and he co-he i abili y, which explain he he i able a i- ance o TS o g ow h, p o ed o be mode a e. Ou ind- ing is no consis en wi h p e ious da a in he li e a u e, including 18 s udies on ish g ow h ai s, which showed ha he he i abili y o mac o-en i onmen al sensi i - i y is gene ally low [23, 24]. Fo example, Sae-Lim e al. [23] who used a mul igene a ion da ase on Finnish ain- bow ou did ind subs an ial addi i e gene ic a ia ion in mac o-en i onmen al sensi i i y o body weigh , bu he es ima ed he i abili y was low (0.07). In hei s udy, he wo mac o-en i onmen s we e a eshwa e b eed- ing nucleus and a sea es s a ion, which had disc e e loca ions and di e ged in se e al en i onmen al ac o s, including wa e salini y and empe a u e, and ea ing sys em as a whole (ea h-bo omed aceways s. sea ne cages, ish densi y, die and eeding in ensi y). Ou esul s suppo he idea ha gene ic a ia ion in mac o-en i on- men al sensi i i y would be smalle han ha in he phe- no ypic alue o he ai a he in e cep poin [24]. Table 2 Gene ic pa ame e s and gene ic co ela ions (± hei app oxima e s anda d e o ) be ween in e cep andslope ob ained om he andom eg ession models o daily g ow h coe icien (DGC) when he in e cep was placed ei he in he low o high empe a u e en i onmen  σ2 a,in =gene ic a iance o DGC a he in e cep poin ;  σ2 e,in = esidual a iance o DGC a he in e cep poin ;  σ2 a ,sl =gene ic a iance o he eac ion no m slope;  σ2 P,To al = o al pheno ypic a iance o DGC ac oss en i onmen s; h2 in =he i abili y o DGC a he in e cep ( σ 2 a,in /  σ 2 P,in ) , whe e  σ2 P,in is he pheno ypic a iance o DGC; co-he i abili y o TS  6  σa,in ,sl  σ2 a,in +  σ2 e,in  , whe e  σa,in ,sl is he addi i e gene ic co a iance be ween he in e cep and slope h2 TS =he i abili y o he slope  σ2 a,sl×  σ2 a,X  σ2 P,To al  , whe e  σ2 a,sl ×  σ 2 a , X is he addi i e gene ic a iance o he slope mul iplied by he a iance o en i onmen al alues X , espec i ely; G(in ,sl) =gene ic co ela ion be ween he in e cep and slope  σain ,sl  σ2 a,in ×  σ2 a,sl  *Es ima e ha is signi ican ly di e en om 0 (95% CI does no include ze o) Pa ame e In e cep en i onmen LowT HighT  σ2 a ,in 0.190 (0.063) 0.130 (0.043)  σ2 e ,in 0.221 (0.041) 0.156 (0.028)  σ2 a ,sl 0.005 (0.002) 0.005 (0.002)  σ2 P,To al 0.354 0.354 h2 in 0.463 (0.123) 0.455 (0.120) h2 TS 0.244 0.244 Co-he i abili y o TS −0.284 (0.124) 0.197 (0.141) G(in ,sl) −0.643* (0.147) 0.376 (0.214) Si e EBV o DGC LowT High -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 Fig. 2 Gene ic eac ion no ms o en ainbow ou si es ac oss wo empe a u e en i onmen s. Ten si es wi h he highes absolu e EBV ( i e dashed lines) and close o ze o EBV ( i e solid lines) o he slope a e ep esen ed. The lines connec he si e EBV o daily g ow h coe - icien (DGC) ac oss he wo empe a u e en i onmen s, when he in e cep was placed a he lowe empe a u e (X = 0) Page 7 o 11 Janhunen e al. Gene Sel E ol (2016) 48:94 As expec ed, in ou s udy, he ainbow ou showed, on a e age, lowe eed in ake, as e g ow h and be e eed con e sion a io a he lowe (14°C) han a he highe ea ing empe a u e (20 °C). The pe cen age dec ease in leas squa e means o DGC was 3.5 om low o high empe a u e. Unde condi ions, wi h a no mal oxygen con en in he wa e and a su icien amoun o ood p o- ided, he ju enile ainbow ou a e known o pe o m well a empe a u es o 14 o19 °C, he op imum em- pe a u e o g ow h being a ound 17 °C; empe a u es highe han 20°C cause as decline in g ow h a e [15, 16, 26, 27]. Fu he mo e, in gene al, eed in ake has a highe op imum empe a u e and eed con e sion a io a lowe op imum empe a u e han g ow h a e o ish [28]. These ac s a e likely esponsible o he obse ed di e ences in eed in ake, g ow h and eed con e sion a io be ween he expe imen al empe a u es. Du ing he expe imen , he wa e quali y pa ame e s ha we e analyzed i.e. oxygen con en , pH and ni ogen compounds we e a a o able le - els o salmonid aquacul u e unde bo h empe a u e con- di ions [29–31]. Thus, i is unlikely ha o he wa e quali y pa ame e s han empe a u e a ec ed he obse ed esul s. Ou esul s show ha he es ima es o he gene ic pa ame e s o DGC we e no d ama ically in luenced by he empe a u e en i onmen s in which he ish we e ea ed. En i onmen al s ess was epo ed o bo h inc ease and dec ease he addi i e gene ic a ia ion in impo an li e-his o y and mo phological ai s [32, 33]. On he one hand, he amoun o en i onmen ally- induced a ia ion may ypically inc ease unde un a- o able condi ions, and hus dec ease he p opo ion o gene ic a ia ion and in u n, he es ima ed he i abili y e.g. [34, 35]. On he o he hand, he opposi e e ec was epo ed in some animal b eeding s udies; challenged en i onmen s inc ease he gene ic a iance o a ai mo e han he esidual a iance, which esul s in highe he i abili y e.g. [36]. In his s udy, he he i abili y o DGC (a he in e cep ) was simila o he wo empe a u es es ed. Bo h he gene ic and esidual a iance p o ed o be la ge a he lowe empe a u e, which was p esum- ably a milde en i onmen in e ms o g ow h. One can expec ha he di e ence in he i able po en ial o g ow h (mean) mani es s i sel only a mo e ex eme empe a- u es abo e o below he op imum. Mul i- ai models ha e been used in many s udies on ainbow ou o show he p esence o G×E in e ac- ions in g ow h [37–41]. Mul i- ai and andom eg es- sion models a e equi alen when he dimension o he gene ic co a iance ma ix and he ixed e ec s included in he models a e he same [7, 23]. The gene ic co ela- ion o DGC be ween empe a u e en i onmen s was equal o 0.47 (SE 0.20) when a bi a ia e model was used wi h ou da a (see “Appendix”), which con i ms he s ong e- anking o amilies ac oss he wo empe a u es es ed. The same pa ame e es ima e can also be ob ained om he andom eg ession model using he gene ic (co) a iances. Consequen ly, selec ion in ei he o he em- pe a u e en i onmen s will p esumably esul in lowe - han-expec ed gene ic gains in he o he en i onmen , i he G×E in e ac ion is no aken in o accoun [41–43]. Al hough he gene ic co ela ion ha is es ima ed om a mul i- ai model does exp ess he deg ee o e- anking among amilies, i does no desc ibe how mac o-en i on- men al sensi i i y o he ai can ac ually e ol e ac oss en i onmen s. One ad an age o he andom eg ession model is ha he gene ic pa ame e s o mac o-en i on- men al sensi i i y o a gi en ai can be ob ained di ec ly, which allows implemen a ion o mac o-en i onmen al sensi i i y as a ai in he selec ion index [5, 12, 23]. Ye , an assump ion o his in ou s udy is ha he slopes, es i- ma ed om he wo en i onmen poin s, a e linea . En i onmen al sensi i i y was gene ally shown o inc ease in esponse o selec ion o high pheno ypic alues when G × E in e ac ion is p esen [5, 6, 44]. Acco ding o Jinks and Connolly [45], his should be ue especially when selec ion o a high pheno ypic alue occu s in an en i onmen ha p oduces a pheno ype wi h a highe alue compa ed o ano he en i onmen (syne gis ic selec ion). Co espondingly, selec ion o dec eased en i onmen al sensi i i y may gene ally esul in educed mean p oduc i i y in mo e a o able en i on- men s [46]. In his s udy, he gene ic co ela ion be ween DGC (a he in e cep ) and i s TS (slope) was ma kedly nega i e, when he in e cep was placed a he lowe em- pe a u e. This inding is consis en wi h he assump ion o Jinks and Connolly [45]: selec ion o as g ow h in a mo e a o able empe a u e en i onmen should a o inc eased sensi i i y, ha is, geno ypes ha ing s eep nega i e slopes in eac ion no ms. Howe e , gene ically, he mos sensi i e geno ypes wi h he s eepes slope EBV may no consis en ly ha e highe gene ic po en ial o g ow h a he lowe empe a u e, compa ed o he leas sensi i e geno ypes wi h la slope EBV (Fig.2). In ac , some o he geno ypes wi h a slow g ow h a he lowe empe a u e also seem o exhibi p onounced sensi i - i y o he mal change, bu in he posi i e di ec ion (i.e., g owing as e a he highe empe a u e). Nega i e gene ic ela ionships be ween p oduc ion ai s and TS ha e also been epo ed in e es ial a m animals. Fo example, in pigs, a gene ic co ela ion o −0.5 be ween ca cass weigh and sensi i i y o hea s ess was epo ed [47]. In dai y ca le, gene ic co ela ions be ween milk yield and hea ole ance anged om −0.30 o −0.45 [48, 49], whe eas, in sheep, his gene ic co ela ion was equal o −0.8 [50]. The obse ed nega i e gene ic co ela ion be ween DGC and i s TS and he p esence o s ong Page 8 o 11 Janhunen e al. Gene Sel E ol (2016) 48:94 G×E in e ac ion sugges ha selec ion decisions should be based on mo e han one he mal en i onmen only. Howe e , by applying a es ic ed selec ion c i e ion, he app op ia e index weigh s ha p oduce he desi ed gene ic esponses in bo h g ow h a e and i s TS can be ob ained [51]. I would hen be possible o simul ane- ously imp o e he g ow h a e ac oss en i onmen s and cons ain he gene ic change in TS. In addi ion o he nega i e gene ic co ela ion be ween DGC and i s TS, he es ima ed co-he i abili y o TS was also nega i e when a lowe empe a u e was used as he in e cep en i onmen . Co-he i abili y is an app oxima e measu e o he inhe i ance o he associa ion be ween DGC and i s TS, when he selec ion c i e ion is DGC in one empe a u e en i onmen . The co-he i abili y has he same sign as he co ela ed esponse o di ec selec ion, and, unlike h2 TS and gene ic co ela ion, his pa ame e also e lec s he accu acy o selec ion [52, 53]. Ou esul s a e in line wi h a ecen su ey on aquacul u e s udies by Sae- Lim e al. [23], which showed ha he g ow h o ainbow ou in one en i onmen is gene ically ela ed o mac o- en i onmen al sensi i i y ac oss en i onmen s. Howe e , he co ela ed esponse o TS is less p onounced i selec- ion o imp o ed g ow h a e occu s a he highe em- pe a u e. In his case, he es ima ed co-he i abili y was also associa ed wi h ela i ely la ge s anda d e o s, which sugges s ha i should be ea ed wi h cau ion. The magni- ude o he co-he i abili y gene ally inc eases, i espec i e o i s sign, wi h an inc ease in G×E in e ac ion [23]. In his s udy, mac o-en i onmen al sensi i i y o each geno ype ( ish amily) was de ined as he di e ence in DGC be ween wo empe a u es. The eac ion no m slope is no an indi idual measu e, bu i s b eeding alue can only be es ima ed based on he g ow h eco ds o el- a i es in wo (mac o-)en i onmen s. Because he mac o- en i onmen al sensi i i y is basically a p ogeny ai , he accu acy o selec ion is ac ually highe o he pa en s han o hei o sp ing (which a e used as b eeding can- dida es). Rainbow ou , as many o he aquacul u e spe- cies, p oduces la ge amilies, which e ec i ely con ibu e o he gene ic analysis o mac o-en i onmen al sensi i - i y. Since he co-he i abili y o TS p o ed o be mode a e, ansla ing his in o mode a e accu acy o selec ion, e y la ge amily sizes a e no equi ed o each mode a e- o- high p ecision in slope EBV. Fo a ainbow ou b eede , a s ock pe o ming well ac oss mul iple empe a u es is he mos desi able ou - come. The leas sensi i e amilies wi h la slopes o eac ion no ms could be selec ed when he b eeding goal is o ob ain obus ish ha h i e unde a iable empe a- u e condi ions (inc eased s abili y). Al e na i ely, amilies wi h posi i e g ow h esponses a highe (o lowe ) em- pe a u es can be chosen when de eloping a locally-adap ed popula ion o a ce ain en i onmen . In he la e case, TS can be iewed a he as an ad an ageous cha ac e o be used by selec i e b eeding when imp o ing he ‘ i ’ be ween he selec ed ish and he he mal en i onmen in which hey a e ea ed [54]. Ei he way, i is likely ha he e will be a high demand o mo e hea - ole an popula ions o ain- bow ou in he u u e since empe a u es will con inue o inc ease a ound he wo ld due o global wa ming. This s udy was unde aken as pa o a capaci y build- ing p ojec wi h he Resea ch Ins i u e o Aquacul u e No. 1 (RIA-1) in No he n Vie nam whe e a na ional b eeding p og am o ainbow ou was ecen ly es ablished om he Finnish b oods ock (Resea ch Cen e o Cold Wa e Aqua- cul u e Species, RIA-1). Vie nam is es ima ed o be one o he wo ld’s mos ulne able a eas o he nega i e impac s due o clima e change. This is he case, in pa icula , wi h cold wa e aquacul u e, o which empe a u e and wa e a ailabili y a e he main limi ing ac o s [55]. The ish used in his expe imen sha ed ela edness wi h he RIA-1’s b ood- s ock. Combining ou esul s wi h he eco ds o ish pe - o mance in Vie nam unde highe empe a u es will enable he assessmen o selec i e b eeding possibili ies in o de o dec ease TS in he nex b eeding gene a ions. This will aid in expanding ainbow ou p oduc ion o lowe la i udes and al i udes wi h be e and mo e s able wa e esou ces, and he ish a ming sec o o adap o clima e change. Conclusions We ound ha he 1-yea -old ainbow ou exhibi sub- s an ial gene ic a ia ion in g ow h esponses ac oss di e en ea ing empe a u es. In e ms o g ow h and eed con e sion e iciency, he ish pe o med be e p edominan ly unde he lowe (14°C) han he highe empe a u e condi ions (20°C). Owing o la ge addi i e gene ic a ia ion, pe manen changes in TS o g ow h a e possible in he s udied popula ion. The e is a ade-o be ween g ow h a e and i s TS, since s ong selec ion o as e g ow h a he lowe and mo e a o able empe a- u e will p esumably esul in less empe a u e- ole an ish. Howe e , he co ela ed gene ic change in TS could be e ec i ely con olled while selec ing o high g ow h ac oss en i onmen s, especially i slope EBV a e inco po- a ed in o he selec ion index wi h app op ia e weigh ing. Page 9 o 11 Janhunen e al. Gene Sel E ol (2016) 48:94 Au ho s’ con ibu ions All au ho s pa icipa ed in planning he s udy. MJ, HV, NHN, HK and AN con- ibu ed o da a collec ion. MJ and JK pe o med da a analysis. MJ was mainly esponsible o d a ing he manusc ip , and JK, HV, NHN, AN and NPT ook pa in w i ing and commen ing he ex . All au ho s ead and app o ed he inal manusc ip . Au ho de ails 1 Biome ical Gene ics, Na u al Resou ces Ins i u e Finland (Luke), Mylly ie 1, 31600 Jokioinen, Finland. 2 Aquacul u e, Na u al Resou ces Ins i u e Finland (Luke), Su on ie 9 A, 40500 Jy äskylä, Finland. 3 Resea ch Ins i u e o Aqua- cul u e No. 3 (RIA-3), Nha T ang, Khanh Hoa, Vie nam. 4 Te o Fish Fa m, Na u al Resou ces Ins i u e Finland (Luke), Huuh ajan ie 160, 72210 Te o, Finland. 5 Resea ch Ins i u e o Aquacul u e No. 1 (RIA-1), Dinh Bang, Tu Son, Bac Ninh, Vie nam. Acknowledgemen s This expe imen was ca ied ou as a pa o he ICI p ojec “Capaci y building o he de elopmen o selec i e b eeding p og am in Vie nam, special ocus on global change and en i onmen al sus ainabili y”. The s a s a he Te o and Laukaa ish a ms p o ided he expe ise o main ain he expe imen . MJ would like o hank Panya Sae-Lim o his ad ice on he gene ic analysis in ASReml and Timo Pi känen o his ad ice conce ning he s a is ics. Fou anonymous e e ees ga e aluable commen s on he p e ious d a s o he manusc ip . The inancial suppo o he s udy was p o ided by he Minis y o Fo eign A ai s o Finland. Compe ing in e es s The au ho s decla e ha hey ha e no compe ing in e es s. A ailabili y o da a and ma e ials The da ase (s) suppo ing he conclusions o his a icle is(a e) included wi hin he a icle (and i s addi ional ile(s)). Appendix In a mul i- ai model, a ai ha is eco ded in di e - en en i onmen s is ea ed as sepa a e ai s (see Addi- ional ile4). Then, he magni ude o he e- anking o he gene ic g oups can be quan i ied by calcula ing he gene ic co ela ion ( G) o he ai in each pai o he en i onmen s [42, 56]. A mul i- ai animal mixed model was used in his s udy and un using ASReml 3.0. The bi a ia e model was as ollows: whe e yij is he obse a ion o he i h indi idual o he k h ai (DGC being eco ded in wo he mal en i on- men s), μ is an o e all ai mean, aiis he andom addi- i e gene ic e ec o he i h indi idual, and eij is he andom esidual e ec . I was assumed ha he andom a iable a is mul i-no mally dis ibu ed wi h a mean o 0 yij =µ+ai+eij, and a iance A⊗G , whe e A is he addi i e gene ic ela- ionship ma ix de i ed om he pedig ee aced back o he base popula ion and G is he addi i e gene ic (co) a i- ances ma ix. In his s udy: whe e σ2 a,lowT and σ2 a,highT a e he addi i e gene ic a i- ances o DGC measu ed a low and high empe a u es, espec i ely, and σa,lowT,a,highT is he addi i e gene ic co a iance be ween low and high empe a u es. The esidual (co) a iance ma ix is: whe e σ2 e,lowT and σ2 e,highT a e he esidual a iances a low and high empe a u es, espec i ely. Because each indi- idual ish was p esen only in one empe a u e en i on- men , he esidual co a iance be ween empe a u es was se o 0. Fo bo h DGC ai s, he he i abili y was calcula ed as: whe e σ2 P is he pheno ypic a iance (sum o addi i e gene ic and esidual a iances), i.e. σ2 a +σ 2 e . The gene ic co ela ion ( G ) be ween DGC measu ed in wo empe a u e en i onmen s was calcula ed as: whe e σa,lowT,a,highT is he co a iance be ween addi i e gene ic alues measu ed a low and high expe imen- al empe a u es, and σ2 a,lowT and σ2 a,highT a e he addi- i e gene ic a iances o DGC measu ed a low and high empe a u es, espec i ely. The gene ic pa ame e s es ima ed using a bi a ia e animal model a e shown in Table3. G =  σ2 a,lowT σa,lowT,a,highT σa,lowT,a,highT σ2 a,highT , R =  σ2 e,lowT 0 0σ2 e,highT , h2 =σ 2 a /σ 2 P, G= σ a,lowT,a,highT  σ2 a,lowT ×σ2 a,highT , Table 3 Gene ic pa ame e es ima es (± hei app oxi- ma e s anda d e o s) o  DGC in wo he mal en i on- men s using a bi a ia e animal model Pa ame e DGC lowT DGC highT σ2 a 0.130 (0.043) 0.190 (0.063) σ2 e 0.156 (0.028) 0.221 (0.041) σ2 P 0.286 (0.026) 0.411 (0.038) h20.455 (0.120) 0.463 (0.123) G 0.468 (0.197) Addi ional iles Addi ional ile1. Pedig ee o he s udied ainbow ou . Addi ional ile2. Indi idual measu emen da a o uni a ia e gene ic analyses. Addi ional ile3. Tank mean alues o FI and FCR. Addi ional ile4. Indi idual measu emen da a o mul i a ia e gene ic analysis.