Effects of temperature and food quality on age and size at maturity in ectotherms: an experimental test with Atlantic salmon

Authors


Summary

  1. The reaction norm between growth rate, age and size at maturity in ectotherms is widely debated in ecological literature. It has been proposed that the effect depends on whether growth is affected by food quality or temperature (called the Berrigan–Charnov puzzle). The present experiment tested this for Atlantic salmon (Salmo salar).
  2. We enhanced growth rates by increasing temperature and ratio of lipids to proteins in the food for groups of Atlantic salmon. Both treatments gave higher percentages of early mature and therefore smaller adults in contrast to the proposed Berrigan–Charnov puzzle. There was a difference between sexes in that males could attain maturity 1 year younger than females when reared under similar environmental conditions.
  3. Males that matured during the first year in sea water were smaller than similar aged immature males. The probability of that Atlantic salmon attained maturity for the first time during their second year in sea increased with growth rate during the preceding winter and if fed a high-lipid diet. Increased summer temperature exhibited no additional effect.
  4. Similar aged fish reared at elevated temperature and fed high-lipid diet attained maturity at a larger body mass and exhibited higher mass-length-ratios than those reared at natural temperature and fed a low-lipid diet, indicating that structural growth has priority over lipid deposits.
  5. Increased growth rate before the onset of maturation, whether this is owing to enhanced lipid content in food or increased water temperature, decreased age and therefore size at maturity. Enhanced lipid relative to protein content in food, but not temperature, had an additive positive effect on early maturation probability, likely due to increased amounts of reserve energy. These results may be general for ectotherm organisms.

Introduction

When will organisms attain maturity? This is a fundamental question in population ecology (Schaffer 1974; Charnov 1993; Roff 2007). Besides the genetic programme influencing when to mature, age at maturity is affected by the somatic resources available at the commencement of the maturation process (Duquette & Millar 1995; McAdam & Millar 1999; Beckerman, Rodgers & Dennis 2010). In fishes, the somatic energy reserves are employed for the development of primary and secondary sexual characters, and activities such as migration to the breeding grounds, courting, competition for mates and breeding sites, copulation and parental care (Jonsson & Jonsson 2011). Therefore, there is a relationship between body size, amount of reserve energy and age at maturity (Plaistow et al. 2004), and the response of age at maturity to change in the growth rate has become an often studied reaction norm (Morey & Reznick 2000; Day & Rowe 2002; Morita & Fukuwaka 2007), not the least in investigations of evolutionary consequences of fisheries (Barot et al. 2004a,b; Morita, Tsuboi & Nagasawa 2009; Beckerman, Rodgers & Dennis 2010). For instance, Olsen et al. (2004) argued that simultaneous decreases in growth rate and age at maturity are evidence of fishery-induced evolution in Atlantic cod (Gadus morhua) following population decrease owing to over-harvesting. This reaction norm is also central when studying thermal effects on ectotherm life histories (Angilletta & Sears 2004).

Theoretically, Berrigan & Charnov (1994) addressed the question about effects of growth on age at maturity. They hypothesized that ectotherms mature earlier at larger size when growth rate is stimulated by high food quality and that they mature earlier at a smaller size when growth rate is stimulated by increased temperature. They based their contention on analysis of the Von Bertalanffy's growth equation and reasoned that this is the expected outcome if there is no correlation between maximum body size and the growth rate as food quality changes, but a negative correlation as temperature changes. Berrigan & Charnov (1994) reviewed studies on invertebrates and fish and found support for that this difference exists, and they concluded that this is a major puzzle for evolutionary and developmental biology.

This puzzle has been discussed by several authors (Atkinson 1994; Perrin 1995; Partridge & Coyne 1997; Angilletta & Dunham 2003), and they present varying views on these relationship between maturation, size, growth and environmental factors. However, few if any author has experimentally tested whether food quality and temperature have contrasting influences on age at size of maturity. We feel that if the observed difference really exists, this may be because temperature and food quality affect energetic deposits or survival rates differently. Recent studies have indicated a decrease in body size of many ectotherms concurrently with climate warming (Daufresne, Lengfellner & Sommer 2009; Ohlberger et al. 2011). On the other hand, a field study on brown trout (Salmo trutta) in rivers located between 59° and 70° N indicated no influence of differing temperature on age at maturity (Jonsson et al. 1991). But feeding and spawning opportunities of brown trout also vary along this environmental gradient, and these variations may overshadow a possible thermal affect. Therefore, there is good reason, under controlled conditions, to test whether age at maturity in ectotherms depends on the environmental cause for the change in growth rate. We did this by use of Atlantic salmon (Salmo salar), a species threatened by climate change (Jonsson & Jonsson 2009), heavy fisheries (Dempson et al. 2001) and changed feeding opportunities in the ocean (Chittenden et al. 2010).

Atlantic salmon are anadromous; they spawn and the young grow up in fresh water, but at a mass of c. 50 g, they smolt and migrate to sea for feeding. They consume most of their food in the ocean and can migrate several thousand kilometres back to fresh water when attaining sexual maturity. During the spawning migration and breeding, they lose from 40 to 70% of their total energy content. The proportion of the somatic resources used increases with the size of the fish, and this rate of increase with size is higher in males than females (Jonsson, Jonsson & Hansen 1997; Jonsson & Jonsson 2003). Experimental evidence suggests that the reproductive success of females increases linearly with body size (Fleming et al. 1996). In males, on the other hand, both small and large males are reproductively fit (Hutchings & Myers 1988). Large males, which attain maturity after having been in the ocean, are fighters, and their reproductive success increases exponentially with increasing body size (Fleming et al. 1996). Small males, attaining maturity prior to smolting and seaward migration, are sneakers at the nest, but their variation in reproductive success is substantial compared with that of the large male spawners (Weir et al. 2005). Thus, it is reasonable to expect different relationships between growth rate and age at maturity in males and females. As the cost of reproduction is high, one may also expect that age at maturity can be constrained by the energy consumption of the fish.

In the experiment, we enhanced the individual growth rate of the fish by raising the temperature during the growth season from April to September by approximately 2 °C, and/or by doubling the lipid density of the food. We expected that (i) increased temperature and increased food quality would increase growth rate. However, increased temperature will reduce size at maturity, whereas increased food quality will increase size at maturity, (ii) that increased food quality will decrease age at maturity owing to a higher lipid energy reserves and (iii) that males had a lower size threshold for maturation than females as reproductive success is less size dependent in males than females. As experimental fish, we used Atlantic salmon reared from wild fish of the River Imsa (Jonsson, Jonsson & Hansen 2003).

Materials and methods

The experimental fish were reared from 20 males and 23 females sampled in the fish trap in the River Imsa (cf. Jonsson & Jonsson 2011). Each female was crossed with one male, but three of the males fertilized the eggs of two females each. The offspring were reared at the NINA Research Station Ims, Norway to 1-year-old smolts, as described by Fleming, Jonsson & Gross (1994). Then, 23 April 2008, eight groups of 100 fish each were individually tagged with PIT-tags after anaesthetization with benzocaine. The groups were placed at random in eight separate tanks. The water was pumped from 30 m depth in the bay outside Ims; mean salinity (±SD) was 32·3 (±1·3) psu (practical salinity units), range of variance 27–35 psu measured once a week with areometer (Sterner Fish Tech, Norway).

Four groups of postsmolt salmon were reared at ambient water temperature (here called low temperature), and four groups were reared in approximately 2 °C warmer water from May to September (here called high temperature). Water temperatures were registered every second hour by water temperature data loggers (HOBO Pro v2) (Fig. 1). At each temperature, two groups of fish were fed Ewos Opal food pellets (29–31·5% lipids, 42–45% proteins; digestible energy density 21 kJ g−1; here called high-quality diet), and two were fed Biomar Classic marine food pellets (15% lipids, 54% proteins, digestible energy density 18 kJ g−1; here called low-quality diet) to satiation. Fish that died during the experiment were not used in the analyses.

Figure 1.

Water temperature in the rearing tanks with natural (solid line) and heated (broken line) sea water during the experiment from 23 April 2008 to 15 September 2009.

Natural tip lengths (L mm) and masses (W g) of all the fish were measured at the seawater transfer, and again in October 2008, May 2009 and September 2009. The fish were checked for sexual maturation in the autumn of 2008 and 2009. Mature and immature individuals differ in morphology and colour, and the gonads were inspected on all mature individuals. The fish were sacrificed by use of benzocaine when the experiment was terminated 15 September 2009. Fish that were not yet mature as one-sea-winter salmon would, if they lived, attain maturity 1 year older (two-sea-winter salmon; Jonsson, Jonsson & Hansen 2003). As one-sea-winter fish, they weigh approximately 1 kg, and as two-sea-winter fish they weigh 3–4 kg. Hatchery reared salmon can also mature during the first summer and autumn in sea water, a few hundred grams in mass (Jonsson, Hansen & Jonsson 1993). One year earlier maturation means that they will decrease by two thirds in size. Under natural conditions, such early maturing individuals are expected to remain in fresh water and participate as sneakers on the spawning grounds instead of smolting and migrating to sea. Some of them will re-mature in the subsequent years, others will smolt, migrate to sea for feeding and omit spawning for at least 1 year before attaining maturity again (Hansen et al. 1989).

Growth (g) in mass during defined periods of time (t2 − t1) was calculated as: = (logeW2 − logeW1)/(t2 − t1), where W1 and W2 are the respective masses in grams at time t1 and t2 in number of days. The relationship between mass and length (C, condition factor) was evaluated by the Fulton's formula [= 100(W/L3); L (cm) is body length measured with 0·1 cm precision]. We chose to use the Fulton's formula owing to significant relationships between this condition factor and energy density (E kJ 100 g−1 wet mass) for both sexes of the River Imsa Atlantic salmon (Jonsson & Jonsson 2003).

Overall effects of high temperature and high-quality diet on age at maturity were initially tested using χ2-tests based on numbers. The effects of temperature (T; 0 or 1) and diet (D; 0 or 1) and their interaction (DT) on mass (g) at the end of the first marine growth season were estimated for immature fish by linear models using diet and temperature as explanatory factors. As the fish were similar in size (c. 50 g) at the commencement of the experiment, the size at the end of the growth season reflects the growth during the first summer. Sex was not included in the final model as its effect was not significant (= 0·26). Furthermore, we tested for treatment effects on the reaction norm between the probability of sexual maturation and growth rate using logistic regression with maturation as a binomial response variable using the logit link function for fish offered only two diet and temperature regimes (McCullagh & Nelder 1983). To reduce number of parameters and interactions and to keep the model selection procedure traceable, first-year maturation (males only), and second-year maturation for males and females were modelled separately. Growth during the summer of maturation (Gs) and marine winter growths (Gw) (summer growth only for the first-year maturing salmon) were chosen as a priori explanatory variables. The full model also included the factors diet and temperature treatments as well as two-way interactions. To account for variation among replicates (tank effects), tank was entered as random factor, yielding the full model structure:

display math(eqn 1)

where logit(p) = ln[p/(1 − p] is the logit link function, and β0 to β8 were estimated parameters, and j indicates replicate and i indicates individuals and αi represents the intercepts, which may also differ among replicates (tanks). The mixed effect models were fitted using the lmer-procedure in the nlm4 library (Bates, Maechler & Bolker 2010) in R 2.11.1. The inclusion of the random tank effect was only supported for the model describing maturation probability of first-year maturing males (comparison of log-likelihoods, < 0·001, calculated using restricted maximum likelihood estimation, REML). The tank effect was accordingly excluded from models describing second-year maturation of males and females. Comparisons of different fixed-effects structures were carried out using a backwards selection procedure where different fixed effects were removed sequentially until no further model improvement in terms of AIC (based on maximum likelihood, ML) were attained (Zuur, Ieno & Meesters 2009). Parameter estimates for the final model was obtained using REML.

Differences in size (mass) of mature fish and condition among treatments were tested with linear mixed effect models (using the lme function in the nlme package; R Development Core Team 2009), with tank included as random factor to account for variation among replicates. Models were run separately for males maturing during the first year in sea water, males maturing during the second year and females (no female attained maturity during their first marine year). For condition factors, initial screening indicated no sex difference and the full models were run without sex as independent variable. All of the full models included the effect of high- or low-quality diet and temperature, as well as the two-way interaction. Inclusion of the random tank effect was supported in all models (comparison of log-likelihoods, < 0·001 using REML) and therefore retained throughout the model selection procedure (Zuur, Ieno & Meesters 2009). The model selection procedure was carried out as described previously under the GLMM analyses.

Results

Maturation effect of treatment

Some of the males attained maturity during the first summer in sea water (Fig. 2), but the proportion varied among treatments (χ2 = 17·55, d.f. = 3, < 0·001). The proportion of mature males was higher when the fish were reared at high temperature and offered high-quality diet than when given any of the other treatments. There was, however, no significant difference in per cent mature between males given high temperature/low-quality diet and given low temperature/high-quality diet (χ2 = 0·06, d.f. = 1, = 0·8). However, there was a significant difference between proportions of mature males in the two replicate groups treated with heated water/low-quality diet (χ2 = 12·5, d.f. = 1, < 0·001), but not between any of the other replicate groups (all > 0·05).

Figure 2.

Mean percentage (±mean of 2 replicates) of mature Atlantic salmon. Dotted line: males attaining maturity during the first year in sea water, solid line: males attaining maturity during the second year in sea water and broken line: females attaining maturity during the second year in sea water. HT, high summer temperature; NT, low temperature; HL, high-quality diet; LL, low-quality diet.

One year later, 80–85% of the males, which were immature during the first year, attained maturity as one-sea-winter fish (Fig. 2). The proportions of mature males did not differ significantly among experimental treatments (χ2 = 1·08, d.f. = 3, = 0·3). There was, however, a significant difference between the two replicate groups reared at low temperature and fed low-quality diet (χ2 = 4·53, d.f. = 1, = 0·03).

Females were older than males at maturity. No female attained maturity during the first year in sea water, whilst a large part matured during the second year. The treatment effect was significant in that almost all females matured when reared in heated water and fed high-quality diet, whereas relatively fewer attained maturity in the other treatment groups (Fig. 2, χ2 = 29·6, d.f. = 3, < 0·001). There was no significant difference between any of the replicate groups (all > 0·05), or between groups reared in heated water and fed low-quality diet and at low temperature fed high-quality diet (χ2 = 0·52, d.f. = 1, = 0·5).

Growth, size and maturation

On average, immature Atlantic salmon more than tripled their mass during each year in sea water (Fig. 3). Consequently, the younger the fish attained maturity, the smaller they were.

Figure 3.

Mean wet mass (±SD) of (a) mature (filled circles, solid line) and immature (open circles, broken line) female Atlantic salmon in September 2009 and (b) males being immature in both 2008 and 2009 (open triangles, broken and dotted line), immature in 2008 and mature in 2009 (filled triangles, solid line), mature in both 2008 and 2009 (filled squares, broken line) and mature in 2008 and immature in 2009 (open squares, dotted line).

There was a significant treatment effect on growth. For instance, mean mass (±SD) of immature fish given high-quality diet and reared at high temperature was higher than of those given high-quality diet and lower temperature (300·4 ± 59·5 g vs. 269·1 ± 53·5 g), both were larger than of those treated with low-quality diet and high (198·7 ± 75·3 g) and low temperature (203·7 ± 47·8 g). The latter two did not differ significantly. The positive effect of high-quality diet on size was highly significant (= 10·1, < 0·001), and high temperature interacted positively with high diet quality (= 3·7, < 0·001). The effect of temperature alone was not significant (= 0·70, = 0·5). Mean mass (M g) of immature salmon at the end of the first growth season was as follows:

display math(eqn 2)

Growth influenced age at maturity. Females maturing in their second marine year (2009) grew faster during the first marine winter, but slower during the subsequent summer than those which remained immature in 2009 (Fig. 3a). Males maturing in their first marine year (2008) were small (Fig. 3b, Table S1), and the most slow-growing ones did not re-mature in 2009. Most of the immature males in 2008 attained maturity in 2009. In October 2008, mean body size in this group differed little from the group that remained immature through 2009.

For females, the probability of second-year maturation increased with growth in the preceding winter and decreased with the subsequent summer growth (Table 1). High-quality diet exhibited an additional positive effect. There was also significant interaction between diet quality and winter and summer growth. The respective ΔAICs were 2·91 and 3·61 in favour of models with winter and with summer growth to lipid treatment interaction compared with models without; comparison of log-likelihoods, both < 0·026, based on maximum likelihood (ML). There was no significant effect of temperature on the probability of maturation other than indirectly through its effect on growth rate [cf. eqn. (eqn 2)]. High-quality diet during winter increased the probability of maturation (Fig. 4a), whereas the effect of high diet quality in the subsequent summer was small (Fig. 4b).

Figure 4.

Predicted probability of second-year maturation [p(m)] in sea water for Atlantic salmon (a) females against winter growth with summer growth standardized to mean values, (b) females against summer growth with winter growth standardized at mean values and (c) males against winter growth with summer growth standardized at mean values in high-quality (black lines) and low-quality diets (grey lines). Regression coefficients are given in Table 1.

Table 1. Parameters for best fit model after model selection on full model (eqn 1) describing the probability of first-time maturation as function of winter growth (Gw), last summer growth (Gs) and high-quality diet (HL) of Atlantic salmon (a) females attaining maturity second year in sea water, (b) males attaining maturity second but not first year in sea water, and (c) males attaining maturity first year in sea water. Parameters given for HL represent additive effect of high-lipid treatment with low-lipid treatment given by the intercept. Estimated parameters for the random effects in (a) are the SD of the random intercepts in the models
VariableParameters (±SE)z-Value P
(a) Second-year maturing females
Intercept−1·70 (±1·89)−0·890·368
G w 1998·15 (±418·43)4·77<0·001
G s −1278·61 (±349·91)−3·65<0·001
HL12·20 (±4·44)2·74<0·001
Gs × HL−1732·31 (±873·34)−1·980·047
Gw × HL−1677·49 (±752·16)−2·230·025
(b) Second-year maturing males
Intercept3·39 (±1·52)2·230·025
G w 759·07 (±301·17)2·520·017
G s −1177·37 (±241·15)−4·88<0·001
HL2·29 (±1·39)1·630·101
Gw × HL−971·76 (±395·90)−2·450·014
(c) First-year maturing males
Tank (random intercept)2·40 (± 1·54)  
Intercept6·79 (±0·98)6·89<0·001
G s −926·82 (± 99·82)−9·32<0·001

For males, the probability of second-year maturation increased with winter growth and decreased with the subsequent summer growth (Table 1). There was a weak positive effect of high-lipid treatment and a significant interaction between high diet quality and winter growth (ΔAIC was 7·15 in favour of the model with interaction compared with the model without; comparison of log-likelihoods, = 0·003, based on ML). Thus, maturing males grew overall faster during winter and early spring, but less during the subsequent summer, than those which remained immature (Fig. 4c). The effect of winter growth was modified by diet treatment. (Table 1).

For males, the probability of first-year maturation decreased with increasing summer growth (ΔAIC 139·17 in favour of model with summer growth only; comparison of log-likelihoods, < 0·001, based on ML). Diet and temperature treatment exhibited no additional effect (Table 1).

Mass at maturity of first-year maturing males was little affected by treatment with only a random tank effect (Fig. 5; ΔAIC 1·02 between next best model including effect of high-quality diet and model with random factor only; comparison of log-likelihoods, = 0·383, based on ML). The mass of second-year maturing males was higher for those treated with high than low-quality diet. There was a significant positive effect of high-quality diet and a random tank effect (ΔAIC 2·83 in favour of model including high-quality diet compared with model without; comparison of log-likelihoods, = 0·028, based on ML). The corresponding female masses were not significantly affected by diet treatment (Fig. 5; ΔAIC 1·17 between model including high-quality diet compared with low-quality diet; comparison of log-likelihoods, = 0·827, based on ML).

Figure 5.

Mean wet mass (±SD) of male Atlantic salmon attaining maturity in 2008 (dotted line), 2009 (solid line) and females attaining maturity in 2009 (broken line). HT, high summer temperature; NT, low temperature; HL, high-quality diet; LL, low-quality diet.

The condition factor of the fish decreased from the autumn 2008 to the spring 2009 (Fig. 6). The best model for immature fish in the autumn 2008 included an effect of high-quality diet and a random tank effect. Both the effects of sex and temperature treatment were not significant and excluded (ΔAIC 2·58 in favour of the model including high-quality diet compared with model with low-quality diet; comparison of log-likelihoods, = 0·032, based on ML). There was also difference in condition factor between treatments in May 2009 (Fig. 6). The best model describing spring condition factor in May 2009 included both high temperature and high-quality diet as main factors, and a random tank effect (ΔAIC 6·07 in favour of models including both high-quality diet and high temperature compared with models without both or one of the factors; comparison of log-likelihoods, < 0·005, based on ML).

Figure 6.

Condition factor [(wet mass)/(body length)3] (±SD) of immature Atlantic salmon in October 2008 and May 2009. Solid line, reared at high temperature and fed high-quality diet; broken and dotted line, reared in high temperature and fed low-quality diet; broken line, reared in low temperature and fed high-quality diet; dotted line, reared in low water temperature and fed low-quality diet.

Discussion

Based on the Berrigan & Charnov (1994) hypothesis (cf. 'Introduction'), predicted and observed age at maturity were similar, whereas the expectations for size at maturity and immature growth were less precise (Table 2). Furthermore, the results did not exhibit the predicted contrasting effects of temperature and diet quality on growth. Rather, high food quality was a prerequisite for fast growth, and high temperature influenced the ability of the fish to utilize the food. High food quality (i.e. lipid density) increased the probability of early age and therefore smaller size at maturity, and enhanced temperature exhibited an additive influence through its effect on growth in interaction with food quality. The effect probably lasts until low oxygen content in the water reduces growth, for instance owing to a very high temperature (Pörtner, Mark & Bock 2004). For young Atlantic salmon, maximum growth is at temperatures between 18° and 20 °C (Jonsson et al. 2001), much higher than usually encountered by feeding salmon in the North Atlantic. Thus, this exothermic organism may be compared with an engine using food as fuel and temperature as accelerator. The higher the temperature, the more surplus energy can be produced from high-quality food. There was no significant effect of temperature alone, but the term was negative, and obvious, growth is terminated given low enough temperature. For young Atlantic salmon, this is at c. 7 °C (Jonsson et al. 2001; Finstad & Jonsson 2012).

Table 2. Cross table of predicted and observed effects of high and low diet quality and high and low water temperature during summer on relative age and size at maturity and immature growth rate. High diet quality alone was predicted to give fish maturing at an early age and large size and that the immature growth rate would be high. High temperature alone was predicted to give fish maturing at an early age and small size and that the immature growth rate would be high. Opposite effects were predicted for low diet quality and low temperature treatments. When the predicted effects of the two treatments differed, we expected the sum-effect to be intermediate
TraitTreatment
High temperatureLow temperature
PredictedObservedPredictedObserved
High-quality diet
AgeYoungYoungMediumMedium
SizeMediumSmallLargeMedium
GrowthHighHighMediumHigh
Lower quality diet
AgeMediumMediumOldOld
SizeSmallMediumMediumLarge
GrowthMediumLowLowLow

The probability of maturation during the second year in sea water increased with growth rate during the winter, and there was a significant interaction between winter growth and diet. This variation is probably environmental as the per cent mature salmon varied among treatments. Large body size and stored lipids give reserve energy needed for sexual maturation (Jonsson & Jonsson 2003). Lipid deposition is associated with maturation in organisms as different as copepods and birds (Marker et al. 2003; Reddish, Nestor & Lilburn 2003), and there is a positive association between lipid storage and steroid production inducing maturation (Cerdá et al. 1995; Marker et al. 2003).

Based on a field study, Morita & Morita (2002) reported that slow-growing white-spotted charr (Salvelinus leucomaenis) attained maturity at an older age and smaller size than fast-growing conspecifics. In eastern fence lizards (Sceloporus undulates), on the other hand, Angilletta et al. (2004) found that the body size was larger in colder environments owing to delayed maturation. Both these reaction norms were covered by Alm's (1959) experiments with a number of freshwater fishes showing a dome-shaped reaction norm for size-to-age at maturity. Size at maturity was smaller in both very fast-growing and very slow-growing fish relative to those with intermediate growth.

We interpret our results on age at maturity as a function of adaptive plasticity possibly resulting from decreased longevity in warmer and richer environments. This may be because the oxidative stress increases with temperature and growth rate (Mair et al. 2003; Metcalfe & Monaghen 2003). Thus, the inherited reaction norms relative to increased temperature and increased lipid consumption may have been evolved as a response to this cost (Jonsson et al. 1991; Karlsson & Wicklund 2005; Johnsson & Bohlin 2006).

In a number of other commercially harvested fishes, such as Atlantic cod (Gadus morhua) and Norwegian spring-spawning herring (Clupea harengus), field studies have indicated that growth-related plasticity has been responsible for changes in age at maturity (Barot et al. 2004a; Engelhard & Heino 2004). Angilletta, Steury & Sears (2004) maintained that this theory applies well to some, but not to other ectotherms. This may depend on whether the temperature decreases from above or below the optimal temperature for growth (Jonsson et al. 2001). Growth rate increases with temperature below and decreases above the optimal temperature. However, temperature and food quality are not the only variables influencing ectotherm life histories. For instance, variable predation pressures should not be ruled out when investigating ongoing changes in age at maturity in nature, as reported for a number of species, such as American plaice (Hippoglossoides platessoides; Barot et al. 2004b), North Sea sole (Solea solea; Mollet, Kraak & Rijnstorp 2007), Pacific salmon (Oncorhynchus spp.; Morita & Fukuwaka 2007) and Atlantic salmon (Consuegra et al. 2005).

Maturation of one-sea-winter salmon appeared to commence during winter or early spring, half a year or more before spawning. Thorpe et al. (1998) suggested that there are two critical time windows for the commencement of the maturation process in Atlantic salmon, in November and April, that is 12 and 6 months prior to spawning-time. At these times, body size, adiposy and the rate of change may be compared with respect to genetically determined threshold levels. Duston & Saunders (1999) found that food deprivation in either early winter, late winter or both decreased the proportion of the salmon population attaining maturity during the subsequent year. For one-sea-winter salmon, our analysis supports that growth rate during winter and spring is important for whether or not the fish will mature later during the year. The similar body sizes in the first marine autumn of immature and maturing one-sea-winter males make the autumn decision relative to body size less obvious. The reduced growth rate during the summer they matured probably mirrored a lower allocation to somatic growth of maturing fish owing to increased energy costs of maturation.

Sexes can exhibit genetically different size-at-age thresholds for maturation (Thorpe & Morgan 1980; Saunders, Henderson & Glebe 1982; Baum et al. 2004). In Atlantic salmon there appears to be smaller nutritional requirements for male than female gonadal development (Adams & Thorpe 1989; Jonsson & Jonsson 2005). On the other hand, females use less energy for courting and aggressive interactions on the breeding grounds than males (Fleming 1996; Fleming et al. 1996). The total energetic requirements for reproduction in males and females are similar (Jonsson, Jonsson & Hansen 1991; Jonsson & Jonsson 2003). But selection for large body size may be stronger in females than males because reproductive success is more directly related to size in females. Both small and large males are reproductively fit (Jones & King 1952; Myers & Hutchings 1987; Jones & Hutchings 2002), while female breeding success chiefly depends on the size dependent egg production and ability to build and defend nests. From a series of experiments under semi-natural conditions, Fleming et al. (1996) reported that female size accounted for more than 80% of the variance in breeding success, for males it was less (61·5%), even in absence of mature male parr (sneakers). This may explain the higher size and growth thresholds for age at maturity in females than males and that only males matured during the first year in sea water. A higher threshold size for female than male maturation is general for salmonids (Jonsson & Jonsson 2011).

The smallest of the early maturing males do not re-mature during the subsequent year. They may not have the required amount of reserve energy to complete re-maturation (cf. Adams & Thorpe 1989; Hansen et al. 1989; Berglund et al. 1991). Jonsson, Hansen & Jonsson (1993) observed that in Atlantic salmon, some stocked males attained maturity during the summer they move to sea and entered rivers for breeding several hundred km from the river they left as smolts. Thus, Atlantic salmon males can attain early maturity after having been in sea water for a few months only, similar to jacks in Pacific salmon (Heath et al. 2002; Koseki & Fleming 2006), but their tendency to stray to other rivers indicates that there is strong selection against such individuals. Fitness in foreign rivers is probably low. They may be too large to be effective sneakers (Gross 1985), lack prior experience from the river they enter (Jonsson, Jonsson & Hansen 1990) and are not genetically adapted to the conditions present (Taylor 1991). This may all be reasons why the probability of maturation is reduced during the first year in sea water relative to corresponding fish retained in fresh water (Lundqvist & Fridberg 1982). We do not know why this early maturation process is not fully inhibited during the first year in sea water. Possibly, some maturing fish have passed a point of no return and therefore continue the maturation process after entering the sea (Metcalfe 1998; Thorpe et al. 1998; Ando, Swanson & Urano 2003).

In all treatment groups, more than 80% of the males, which were immature during the first year in sea water, attained maturity during the subsequent year (one-sea-winter salmon). There was no significant difference among treatment groups. This percentage of maturing one-sea-winter salmon is similar to the proportion which has been observed in the natural River Imsa population (Jonsson, Jonsson & Hansen 2003) and may be the inherited level of maturation after one winter in sea water when fed to satiation.

As individuals allocate more energy to permanent structural growth (body size), they necessarily reduce their allocation to lipid reserves and gonads (Zera & Harshman 2001). In birds such as European shag nestlings (Phalacrocorax aristotelis) structural growth appears to have priority over lipid deposits (Moe et al. 2004), and our findings support that a similar allocation rule may hold for Atlantic salmon. This view is based on the observation that the condition factor, and consequently energy and lipid densities (cf. Jonsson, Jonsson & Hansen 1997; Todd et al. 2008), was highest for the salmon reared in warm water and given high-quality food. However, according to Johansson & Andersson (2009), such a difference can also be influenced by differences in activity. With improved feeding opportunities, activity decreases. Reduced activity subsequently induces a deeper body and thus a higher condition factor.

The present results have relevance to studies of climate change effects on wild fish. For instance, body size of one-sea-winter salmon in the north Atlantic has decreased during recent years (Jonsson & Jonsson 2004). Todd et al. (2008) attributed the decline to climate change. The present results indicate that this may be a combined effect of higher temperature and poor feeding opportunities in the ocean, restraining the fish from reaching their growth potentials.

In all, this study supports the view that both temperature and feeding conditions influence maturation in Atlantic salmon in sea water. The norm of reaction is gender specific in that females have a higher size and growth thresholds for attaining maturity than males. There was no contrasting effect of high temperature and high feeding rate on size at maturity in this ectotherm vertebrate. Water temperature affected growth rate through interaction with diet quality, and thereby age and size at maturity. Diet quality alone also affected growth and the deposition of energy reserves, which affected age and size at maturity.

Acknowledgements

We are grateful for help provided by the manager Knut Bergesen and his staff at the NINA Research Station Ims. The project received financial support from the NORKLIMA programme of the Norwegian Research Council (NFR 185109/S30) and the strategic institute programme (SIS) on ecological effects of climate change of the Norwegian Institute for Nature Research. Three reviewers gave helpful comments.

Ancillary

Advertisement