Genetic coupling of life-history and aerobic performance in juvenile Atlantic salmon

The physiological underpinnings of life history adaptations in ectotherms are not well understood. Theories suggest energy metabolism influences life history variation via modulation of resource acquisition. However, the genetic basis of this relation and its dependence on ecological conditions, such as food availability, have rarely been characterized, despite being critical to predicting the responses of populations to environmental changes. The Atlantic salmon (Salmo salar) is an emerging wild model species for addressing these questions; strong genetic determination of age-at-maturity at two unlinked genomic regions (vgll3 and six6) enables the use of complex experimental designs and tests of hypotheses on the physiological and genetic basis of life-history trait variation. In this study, we crossed salmon to obtain individuals with all combinations of late and early maturation genotypes for vgll3 and six6 within full-sib families. Using more than 250 juveniles in common garden conditions, we tested (i) whether metabolic phenotypes (i.e., standard and maximum metabolic rates, and absolute aerobic scope) were correlated with the age-at-maturity genotypes and (ii) if high vs. low food availability modulated the relationship. We found that salmon with vgll3 early maturation genotype had a higher aerobic scope and maximum metabolic rate, but not standard metabolic rate, compared to salmon with vgll3 late maturation genotype. This suggests that physiological or structural pathways regulating maximum oxygen supply or demand are potentially important for the determination of age-at-maturity in Atlantic salmon. Vgll3 and six6 exhibited physiological epistasis, whereby maximum metabolic rate significantly decreased when late maturation genotypes were present concurrently in both loci compared to other genotype combinations. The growth of the feed restricted group decreased substantially compared to the high food group. However, the effects of life-history genomic regions on metabolic phenotypes were similar in both feeding regimes, indicating a lack of genotype-by-environment interactions. Our results indicate that aerobic performance of juvenile salmon may affect their age-at-maturity. The results may help to better understand the mechanistic basis of life-history variation, and the metabolic constrains on life-history evolution.


72
Physiological processes control how life-history diversity emerges from resource allocation and acquisition trade-offs (Ricklefs & Wikelski 2002). The rate of aerobic energy metabolism 74 is a pivotal mechanism contributing to life-history variation -it modulates resource 75 acquisition, provides cells with ATP, and constrains energy allocation to different body 76 components and functions. Theories, such as the metabolic theory of ecology and the pace-77 of-life syndrome theory (Brown et al. 2004; Dammhahn et al. 2018), suggest metabolic rate 78 covaries with life-history variation within and among species. This covariation may have a 79 genetic basis, consequently constraining trait evolution (Roff 1997), yet only a few studies 80 have demonstrated intraspecific genetic covariation or co-evolution between metabolic rate 81 and life-history traits (Boratynski et  history traits (Thorpe 2007). Life-history traits, such as the timing of maturation and 107 migration, are determined by adaptive body-size thresholds (Roff 1994;Salminen 1997; 108 Thorpe et al. 1998;Theriault et al. 2007), and metabolic phenotypes are often correlated with 109 growth rate, albeit in a context-dependent manner (Burton et al. 2011;Metcalfe et al. 2016 In anadromous (sea-migrating) salmonids, the number of years the fish spends at sea before 125 the first spawning, i.e., sea age-at-maturity, has a dramatic effect on its size-at-maturity 126 (Fleming & Einum 2011): individuals spending one year at sea typically weigh 1-3 kg 127 compared with 10-20 kg after 3 or more years. Increased size in late maturing individuals 128 also translates to marked gains in reproductive investment in both sexes (Fleming & Einum 129 vgll3 (vestigial-like 3) gene on chromosome 25 (Ayllon et al. 2015; Barson et al. 2015). In 142 addition, variation in another locus on chromosome 9, six6, (after six homeobox 6 gene in this 143 region), is a strong predictor of mean age-at-maturity among populations (Barson et al. 144 2015), and associated with early maturation in an aquaculture strain of salmon (Sinclair-145 Waters et al. 2020). Vgll3 and six6 are also associated with size-at-maturity, with the alleles 146 conferring late maturation being associated with larger age-specific body size especially after 147 multiple years at sea (Barson et al. 2015). In the last few decades, many Atlantic salmon 148 populations have been maturing, on average, at younger ages (Chaput 2012), which is 149 associated with an increase in the frequency of the early maturation allele in vgll3, in some 150 cases (Czorlich et al. 2018 The strong effects of the six6 and vgll3 genomic regions on life-history variation provide an 158 opportunity for the genetic covariation between age-at-maturity and energy metabolism to be 159 studied prior to maturation, i.e., at the juvenile stage, by genetic prediction. This approach 160 makes controlled, empirical settings more feasible, as salmon require several years to reach 161 maturation. In this study, we use genetic prediction of the age at maturity of Atlantic salmon 162 and test if (i) genetic covariation exists between life-history and metabolic phenotypes in 163 juveniles, and (ii) resource limitation can induce genotype-by-environment interactions in 164 metabolic phenotypes. Specifically, we test the hypothesis that resource availability and age-165 at-maturity genotypes interact in their effects on metabolism; under high food availability, 166 salmon with early maturation genotypes are predicted to show a higher SMR and AS than 167 fish with late maturation genotypes, but under low food availability, the effects of genotypes 168 are predicted to be weaker. Feed rations were calculated assuming feed conversion efficiency of 0.8, using growth 198 predictions from Elliott & Hurley (1997). Fish were initially fed with 0.2mm commercial 199 feed (Vita, Veronesi), 4 times d -1 , and transferred to 0.5mm feed and more frequent feeding, 200 (allowing food to be available in the water column for a maximum amount of time while 204 minimizing the amount of leftover feed). Tanks were cleaned by scrubbing surfaces and 205 siphoning excess food once week -1 until June 1 st , then approximately twice week -1 until 206 August 14 th , then once week -1 until the end of the experiment. Mortality from first feeding in 207 March until PIT-tagging in July was approximately 8%. Water temperature during this time 208 increased from ca. 4.5 °C to 11°C (Fig. S1). , and a small fin clip was collected from their caudal fin using a scalpel. Fish were 216 allowed to recover in aerated buckets briefly after anaesthesia, after which they were returned 217 to the rearing tanks. Total mortality due to tagging/anaesthesia was approx. 5%. After 218 tagging, environmental enrichment was provided to the tanks in the form of stones (diameter 219 approx. 8cm) placed in square, stainless-steel baskets (mesh size 2cm, width and length 220 20cm). Three baskets with three stones in each were provided to each tank. At minimum two weeks after PIT-tagging, fish from each family were divided into two tanks 232 at roughly even densities (mean 1.29 g L -1 , N = 145-152 tank -1 ). The experiment (see Fig. 1, 233 created with BioRender.com) started in August 2020, at least three weeks after PIT-tagging 234 for each family. In the beginning of August, the relative age of the fish was approximately One tank from each family was assigned to a feed restriction (hereafter low food) treatment. 245 Immediately before the treatment, these fish had been measured once for SMR (as described 246 below, for a separate study), with in total 3 days fasting, weighed to nearest 0.01g, and 247 measured to nearest mm. During the low food treatment, fish were fed twice week -1 using 248 automatic feeders, which distributed the estimated whole daily ration of food to the tank 249 within 2h in 10 doses. Intermittent feeding to satiation was preferred over constant low ration 250 to minimize the formation of strong dominance hierarchies in tanks (Ward et al. 2006), and 251 because individual feeding as in, e.g., Auer et al. (2015b) was not feasible. Parallel to the 252 low-food treatment, the other tank for each family was assigned to a high food treatment. 253 These fish were weighed and measured before the treatment begun, including 2d fasting and 254 anaesthesia. The high food treatment consisted of the total estimated daily ration, delivered 255 daily in eight doses distributed equally during an 8h period (9:00-17:00) using the automatic 256 feeders. Further details provided in online supplemental material. 28 days after the feeding 257 treatments commenced (range 28-31d because measurements took 2-3 days for each tank), 258 (Table S2). Densities of fish biomass in the tanks at the end of the treatments were on average 262 1.8 and 2.8 g L-1 in the low food and high food treatments, respectively. 263

SMR and MMR measurements 265
Two days before their SMR measurement started, fish for each batch were caught by netting 266 and their genotype was identified using PIT-tags, after which they were moved into an 267 acclimation tank (Fig. S3). Each batch contained 16 individuals from the same family and 268 tank, and was balanced for all homozygous genotype-sex-combinations (in most cases N = 2 269 group -1 batch -1 , with fish heterozygous for either locus used when homozygous fish were not 270 available). Fish of desired genotypes were randomly picked from the rearing tanks by netting, 271 apart from the high food treatment from the 4 th family where some fish had grown too large 272 for the respirometers. All fish from this tank were anaesthetised, weighed, and measured four 273 days before their respirometry trials, and appropriate size fish (max. length 82mm) were 274 selected for the trials from each genotype (vgll3 and six6 genotype frequencies were not 275 different between these two size groups, Chi 2 = 5.5, df = 7, p = 0.6). (example slope in Fig. S5A). Second, slopes for MMR were extracted using a derivative of a 320 polynomial curve fitted on each measurement (function smooth.spline, df=10). This is the 321 'spline-MMR' method (Fig. S5B). The slopes were then used to calculate MMR in mg O2 h -1 322 using FishResp-package function calculate.MR. MMR values calculated by the 1-min respR 323 and spline-MMR approaches were highly correlated (Pearson-r = 0.98, 95% confidence 324 interval 0.98 -0.99). We selected the spline-MMR data for further analysis. Absolute aerobic the spline-MMR data. For further details on SMR and MMR analyses, see online 327 supplementary information and data availability. 328

Statistical analyses 330
To test for the effects of treatment and genotype on metabolic variables, we ran separate 331 linear mixed models using SMR, MMR, and AS as response variables. This mixed model 332 framework allowed us to account for variation arising from i) fish body size, ii) family effects 333 (background genetic variation between families and tank effects), iii) timing of experiments 334 in relation to photoperiod and temperature during rearing, i.e., batch effects, iv) and technical 335 variation related to MMR measurements. Only individuals homozygous for both vgll3 and 336 six6 were included in the analysis because individuals heterozygous for either locus were not 337 measured from all families. 338

339
The response variables and body mass as a covariate were log10-transformed to account for 340 allometric scaling of metabolic rate. We included treatment, vgll3 and six6 genotypes, and 341 sex as fixed effects in all models. Random effects for SMR included family and measurement 342 batch. For MMR and AS they included family and the chamber where MMR was measured. 343 The chamber effect accounted for random variation among persons performing the chase 344 (half of the trials were consistently performed by the same person, J.P., the rest were divided 345 among three people), and variation due to the timing of the chase test (chamber order was 346 always the same when placing fish into chambers with only a few exceptions). Inclusion of 347 the name of person performing the chase as a random effect did not change the results of the 348 analysis or decrease the residual variance. Batch was not included as a random effect in 349 MMR and AS models due to model singularity. To test if genotype-specific metabolic rates 350 were affected by sex -because females mature later than males in both early and late age-at-351 maturity genotypes-and by food availability, i.e., to test for genotype-by-environment 352 interaction, we fitted additional interactions into the models, including pairwise interactions 353 between vgll3 and six6 genotypes, between genotypes and treatment, and between genotypes 354 and sex. Further, the interaction of log10 body mass with treatment was included to test for 355 potential treatment-specific allometric scaling of metabolic rate, which may be related to 356 body composition differences between treatments. The full models were reduced by omitting 357 non-significant interactions in a stepwise process based on Type III test p-values to obtain 358 estimates for significant interaction effects. When interactions between fixed genotype effects 359 were significant, p-values for the pairwise differences between genotypes were obtained by 13 post hoc analysis using package emmeans (Lenth 2020). A summary of model parameters is 361 shown in Table S3.

378
Low food treatment decreased both the specific growth rate and condition factor of the fish 379 compared to high food treatment (Fig. S6). The mean body length of fish was 70.6 ± 4.5 and 380 66.2 ± 4.9mm (SD), and the mean body mass was 4.2 ± 0.8 and 3.3 ± 0.8g after the high and 381 low food treatment, respectively. 382 383

Standard metabolic rate 384
There was no significant genotype, food availability or sex effect on SMR (Table 1, Fig. 2). 385 There was a marginally significant interaction effect of six6 and food availability on SMR (p 386 = 0.045 in the full model, p = 0.055 in the simplified model, Table S4), but none of the 387 pairwise contrasts were significant (the largest effect being: six6 EE-genotype, high food vs. 388 low food, t25.6 = -2.37, p = 0.11). The metabolic scaling exponent, b, i.e., the slope of log 389 maturation genotype (Fig. 2, Table 1). Vgll3 genotype also interacted with six6, such that 395 MMR was decreased when late maturation genotypes of the two loci cooccurred compared to 396 other genotype combinations (Fig. 2, Table 1). None of the treatment-genotype or sex-397 genotype interactions or the main effects of sex or food availability had a significant effect on 398 MMR (Table S5). Unlike in SMR, the metabolic scaling of MMR was not significantly 399 affected by food treatment (b = 0.86, R 2 = 0.76). 400 401

Aerobic scope 402
Fish with the vgll3 early maturation genotype had an approximately 4.5% higher AS 403 compared to fish with the late-maturation genotype (Fig. 2, Table 1, predicted means 458.9 404 and 439.2 mg O2 kg -1 h -1 for early and late maturation genotypes, respectively). AS was 405 increased by the low food availability compared to high food availability, but only in smaller 406 fish (interaction p = 0.049, Table 1); scaling exponent b = 0.94 (R 2 = 0.57) in the high food 407 and 0.90 (R 2 = 0.68) in the low food treatment (Fig. S7b). The vgll3 and food treatment 408 effects were also significant when mass adjusted SMR was included as a covariate in the 409 model (Table S6), indicating that the genotype effect was independent of SMR. The six6 or 410 sex effects were not significant, and there were no significant interaction effects between 411 genotypes and treatment or sex on AS (full model in Table S7).   Table 2. Pearson's correlation coefficients between metabolic phenotypes in high food (above diagonal) and low food (below diagonal) treatments. P-values given in parentheses. simultaneously, at least at this developmental stage. Such pathways may be related to oxygen 478 demand by tissues or its supply (uptake, transport, or unloading) during stress and/or 479 exhaustive exercise. For example, structural and functional variation in the heart (i.e., cardiac 480 output) or muscle (Gamperl & Farrell 2004 between metabolic phenotypes and age-at-maturity. Against our predictions, our experiments 493 did not reveal a change in SMR or MMR, or genotype-by-environment interactions, due to 494 feed restriction, despite a strong decrease in growth rate. The average increase in AS 495 observed in the low food compared to high food availability was specific to small size 496 classes, and possibly related to lower relative amount of adipose tissue that may affect the 497 body mass -adjusted AS. The lack of genotype-by-environment interactions indicates that 498 salmon age-at-maturity genotypes coped equally well with the resource variation we applied. 499 The lack of SMR response likely indicates that the metabolic activity of tissues did not 500 approximately 3 days of fasting in between feeding to satiation, similar to a "feast and 504 famine" feeding strategy (Armstrong & Schindler 2011). A lack of metabolic response to 505 reduced food availability may be beneficial if it allows the individual to maximize acquisition 506 via food assimilation when this strategy is used. 507 508 Salmon in the wild are increasingly experiencing higher than optimal temperatures due to 509 climate change (Friedland et al. 2009), and MMR is typically less plastic than SMR in aerobic scope between vgll3 early and late maturation genotypes may be reflected in the 512 resource acquisition and growth of juvenile salmon at elevated temperatures. Under these 513 conditions, higher aerobic scope may enable higher feeding capacity (Auer et al. 2015a), 514 because a temperature-related reduction in aerobic scope due to increased SMR and specific 515 dynamic action (the post-feeding increase in metabolic rate) could reduce appetite in fishes, it was also shown that salmon females have a higher size threshold for maturation after one 536 year at sea compared to males (Tréhin et al. 2021), and further studies are required to link the 537 performance differences and sex-specific maturation schedules of salmon at sea. 538 539 Our experiment focussed on the genetic component determining age-at-maturity -rearing 540 fish until maturation was out of our scope, excluding the possibility to evaluate the 541 environmental component. The presence of genetic covariation between aerobic scope and 542 age-at-maturity constrains evolution because selection acting on either trait would alter the 543 phenotypic variation of the other (Lande 1979;Roff 1997). For example, natural selection 544 favouring later age-at-maturity would indirectly constrain the aerobic scope of juveniles to a 545 lower level, even if that may be a suboptimal phenotype. On the other hand, the genetic 546 covariation between aerobic scope at the juvenile stage and age-at-maturity may help 547 maintain optimal trait variation in age-at-maturity, e.g., by constraining the potentially 548 maladaptive environmentally induced variation in age-at-maturity (e.g., De Jong 1999;Tufto 549 2000). River geophysical properties are important determinants of the optimal age structure 550 of populations at maturity, whereby populations in smaller tributaries have a younger and 551 populations in large, fast-flowing rivers have an older age structure (Fleming & Einum 2011). 552 Therefore, forecasting the optimal age-at-maturity from juvenile phenotypic performance 553 (i.e., growth) would be maladaptive for salmon individuals if covariation between 554 performance traits and age at maturity was explained entirely by environmental effects.