Large changes in environmental conditions put populations at risk of becoming maladapted, thereby leading to negative population growth and potential extinction 1 . Such large changes in environmental conditions may be due to anthropogenic activity 2 , but may also be experienced as individuals invade a novel environment. For ectotherms in general, temperature is considered to be by far the most important external factor, controlling much of the variation seen in embryonic ontogenetic rates 3 4 5 . Following an environmental perturbation a population may be maladapted to the new conditions. If populations cannot rapidly move towards the new evolutionary optimum they may risk extinction 6 .

Modification of traits may happen through purely plastic processes or through evolutionary change 7 8 9 . Phenotypic plasticity is defined as the ability of a single genotype to develop different phenotypes under different environmental conditions 7 . Norms of reaction (i.e., the variation in trait value across environmental gradients) are commonly used to illustrate the phenotypic plasticity of genotypes (reviewed in 10 ). When a perturbation to the environment is outside of the normal range of environmental fluctuations, it is expected that evolutionary change will be required to reach the new fitness optimum 6 . In many situations, for example where there is substantial spatial habitat heterogeneity, the norms of reaction may be the target of selection 11 . Divergence in the slope and/or elevation of reaction norms may develop among populations occupying habitats that differ in their environmental characteristics. Since many traits depend on conditions experienced earlier in life, the reaction norm concept has been extended to include developmental processes 12 . These developmental reaction norms may be the target of selection, where individuals that most optimally modify their developmental trajectories under varying environmental conditions are selected.

A change in the population reaction norm requires genetic change, either through stochastic processes like genetic drift, or through adaptive evolution. Today it is realized that ecological and evolutionary dynamics do not necessarily operate on vastly different timescales 13 14 15 . A mounting number of studies provide evidence of evolutionary change over contemporary time scales 16 17 .

In salmonid fishes, thermal adaptation is thought to be important, especially during the egg-stage and subsequent phases following hatching. For fish larvae in general, size appears to be positively correlated with early-life survival 18 19 20 . This rule has been termed the "bigger-is-better"-hypothesis; for fish larvae, performance is size-dependent to a stronger degree than it is age-dependent, or dependent on any other measure of biological time 21 22 . The size effect is particularly strongly expressed under conditions of high competition, since there tends to be strong dominance hierarchies in juvenile salmonids 23 24 . Strong directional, size-dependent selection should thus lead to increased individual development rates in a population. We here investigate if that is the case in several demes of grayling Thymallus thymallus in an alpine Norwegian lake (Figure 1).

Grayling, a spring spawning salmonid fish, colonized the lake Lesjaskogsvatnet through a man-made connection in the upper reaches of the river Gudbrandsdalslågen that was opened in the late 1880s (20-25 generations ago). Subsequent closing of the connection made further immigration of grayling difficult 25 . Genetic evidence suggests that Lesjaskogsvatnet was colonized by a low number of individuals from Gudbrandsdalslågen 26 . Since the colonization event, the grayling have established more than 20 demes spawning in separate tributaries to the lake 27 . It is likely that grayling return to their natal tributary to spawn. Lacustrine populations of grayling prefer to move to running water to spawn and are known to return to their natal stream with high fidelity 28 .

During spring, the south-facing slopes of Lesjaskogsvatnet receive considerably more sunlight and are less steep than the shaded north-facing slope resulting in more rapid spring warming. In addition, small tributaries tend to warm more rapidly than large ones. As a result of these differences in aspect coupled with size differences among tributaries, tributaries can generally be classified as either large-and-cold (LC) or small-and-warm (SW) 27 . Generally, spring ice-melt commences earlier in SW tributaries than in LC tributaries, and SW tributaries have a higher temperature than LC tributaries through spring and early summer 29 . Typically, mean daily June- July water temperatures differ by 1-1.5 °C among LC and SW streams, which add up to substantial differences in accumulated temperature sums over the stream phase.

Haugen and Vøllestad 30 conducted a common-garden study of several early growth and developmental traits of grayling in the same area, but on an inter-lake scale. They investigated differences between the Lesjaskogsvatnet population and two other populations recently derived from Lesjaskogsvatnet grayling. In their study, they discovered apparent adaptations in many of the traits. Based on the same data, evidence for divergence in time to hatching was found between two of the populations 31 . These results demonstrate that in the absence of gene flow local adaptation may progress significantly over less than a 20-generation time span. Here, we use a common environment approach at three different temperature conditions to investigate how temperature reaction norms for early life history traits vary for grayling from two cold and two warm streams where gene flow among streams is possible.

In high latitude mountain lakes like Lesjaskogsvatnet the growth season is short. For north-temperate fishes the first winter is usually a period of energy deficit 32 , potentially selecting for faster growth. The selection intensity is expected to be strongest for the demes with the shortest growth season (i.e. LC demes). Individuals from these demes are expected to compensate for the shorter growth season with a higher general capacity for growth 33 .

As grayling occupy divergent habitats during development it is possible that evolutionary divergence could result. However, one should bear in mind that this divergence is observed under conditions of continuing gene flow and potential meta-population dynamics 34 35 . The grayling could have responded to the differing environmental conditions by evolutionary changes; alternatively, the grayling could have a plastic response common to individuals of all demes 9 . If the latter is the case, reaction norms for all demes should appear more or less identical. If reaction norms differ this would be indicative of genetic effects.

In this study, we test if there is evidence for evolution in early life-history traits in larval grayling from four different spawning demes that are experiencing different environmental selection pressures. To contrast the development of larval grayling from different demes a common-environment experiment was conducted. In the experiment, individuals from four tributaries (two LC, two SW) were subjected to three experimental temperatures and studied from fertilization to swim up, the stage when the juveniles emerge from the gravel and become free-living 36 .

The results presented in this study demonstrate genetic differences in growth-related traits between grayling from four demes within Lesjaskogsvatnet. Grayling from different demes responded differently to changes in developmental temperature, with respect to growth in length and yolk-to-body-size transition efficiency. This appeared as differentiated temperature-related reaction norms for these traits. A major part of the among-deme variation in growth-related traits is likely to be genetically based since the results were obtained from a common-environment experiment. Parallelism of developmental phenotypes between demes belonging to similar thermal groups indicates that the differentiation is consistent with local adaptation.

We found significant among-deme variation in growth, estimated as the change in length over time, and in how yolk was allocated to growth. As expected from previous work ( 38 and references therein), these traits were also significantly influenced by temperature. However, the different demes responded differently to the treatment temperatures. Haugen and Vøllestad 30 also discovered similar differences when testing three grayling populations having the same ancestry, but experiencing vastly different environmental conditions.

The differences in the growth rate estimates for LC deme individuals relative to the SW individuals may, to some extent, be interpreted as indicative of local adaptation. Larvae from the LC demes grow faster than SW deme individuals under experimental conditions comparable to what is observed in the LC deme habitats (i.e., the cold treatment temperature, Figure 2). LC deme individuals also have more efficient yolk-to-body size transition efficiencies under cold and medium temperature conditions (Figure 3). All grayling demes originate from fish spawning downstream in the river Gudbrandsdalslågen. Overall, temperatures in Gudbrandsdalslågen is comparable to the temperatures in the SW streams, indicating that only the LC deme individuals have been exposed to temperatures leading to strong selection. The full picture emerging from the data, however, does not fully correspond with predictions for the "bigger-is-better"-hypothesis or counter-gradient variation hypothesis 39 40 : at the medium and warm treatment temperatures, the rank order of demes appears not to follow a pattern that can be ascribed to conditions in their natal tributaries. Overall, there is strong evidence for genetically based differences in growth rate among demes. However, since the optimal strategy is unknown it is premature to decide whether the result is due to local adaptation or not. Further, there is a general trend in all demes for growth rates to increase with temperature, much in line with results from comparable studies 38 .

Jensen et al. 41 found significant divergence in traits such as larvae length and growth rate of brown trout juveniles in a common-garden experiment, and suggested that this was due to local adaptation. However, as for the present study, fitness was not estimated so conclusions in this aspect are premature. In our study, differences in traits related to growth and energy utilization do not unambiguously correspond with predictions for local adaptation. Any lack of adaptation might be caused by evolutionary constraints, resulting from the history of the Lesjaskogsvatnet grayling. It is possible that the demes under investigation have not had equal opportunities for environmental adaptation and that equilibria have not been reached 34 35 . When investigating potential genetic effects in divergence and adaptation studies one should never overlook the potential importance of random genetic drift. Through the colonization process, the grayling of Lesjaskogsvatnet have gone through various severe bottlenecks 34 35 . Normally, bottlenecks are expected to reduce genetic variability and hence impede response to selection (e.g. 42 43 ). However, because founder populations may contain a high fraction of the heterozygosity and additive genetic variance in quantitative characters that existed in a large source population, prospects for adaptation may not necessarily have to be impeded 44 . Potentially, genetic drift could complicate the interpretation of results since the outcome of evolution by genetic drift cannot be predicted (e.g. 45 ). Further, in young systems such as in the Lesjaskogsvatnet grayling system gene flow and drift may vary in strength, leading to large among year variation in strength of selection 34 . Gene flow most often is considered to hinder local adaptation 46 47 , however, this is not necessarily always the case 48 49 . Gene flow among inbred demes may counteract the effect of drift and increase the evolvability of these demes by introducing genetic variation, and as such enhance short-term adaptation 43 . However, in most situations where you have heterogeneous habitats and genotype-environment interaction effects, gene flow is predicted to constrain local adaptation 49 50 .

The different traits studied here may be influenced by maternal effects. If such effects are present, and if they are physiological side effects and not consequences of adaptive evolution, they have the potential to confound our results. Potential maternal effects might be egg provisioning of energy and other important nutrients 51 . This cannot be directly accounted for in a study like ours, where the maternal environment is not controlled. One complicating factor is the time delay between maturation of gonads for LC and SW demes, potentially leading to quality differences in the eggs. In Haugen and Vøllestad's 30 study on grayling from the same area, egg size was found to have a significant contribution to the total female effect for three out of four early-life size traits and also on early survival rate. Our study design did not allow for family-level resolution. Egg size - a main determinant of maternal effects - did differ among the four demes. However, these differences cannot explain the parallelism in growth rates and energy use as both the largest and the smallest eggs were observed in the SW demes.

We have shown that several early life history traits have diverged among the four grayling demes studied here, and some of the divergence is consistent with adaptive predictions. Taken together with several related studies from the same area 29 30 this indicates that these traits can evolve rapidly and in spite of conditions that are generally assumed to constrain adaptation. We have shown that most of the demes studied have low genetic variability and have experienced bottlenecks and founder effects 34 35 . Gene flow among demes is strong and highly variable 34 . Additionally, some demes are small indicating that genetic drift may be strong. These conditions can potentially lead to strong genetic drift, a process that can constrain adaptive evolution 43 46 47 . A recent theoretical study does, however, show that dispersal, mutation and sexual reproduction may accelerate local adaptation in growing populations 49 , especially when populations are small. However, these same processes may reduce local adaptation in the long run.

In summary, we find evidence for significant among-deme variation in important early life-history traits in grayling from the Lesjaskogsvatnet. These differences have evolved during 20-25 generations in the face of gene flow, recent bottlenecks, and the fact that the grayling live in sympatry during most of their life. Some of these results are consistent with adaptation expectations as inferred from signs of parallelism, but we cannot rule out that other evolutionary or non-evolutionary processes have influenced the results.

TOH and LAV designed the study, all authors took part in field and laboratory work, but all measurements have been collected by GT. Analyses of the results were conducted by GT and TOH. All have taken part in the writing of the paper. All authors have read through and approved the last version.