Capacity for thermal acclimation differs between populations and phylogenetic lineages within a species
Summary
Within-individual plasticity (acclimation) counteracts potentially negative physiological effects resulting from environmental changes and thereby maintains fitness across a broad range of environments. The capacity for the acclimation of individuals may therefore determine the persistence of populations in variable environments.
We determined phylogenetic relationships by Amplified Fragment Length Polymorphism (AFLP) analysis of six populations of mosquitofish (Gambusia holbrooki) from coastal and mountain environments and compared their capacity for thermal acclimation to test the hypotheses that acclimation capacity is greater in more seasonal environments with less diurnal variability, that acclimation is genetically constrained and that demographic processes determine acclimation capacity.
We show that populations are divided into distinct genetic lineages and that populations within lineages have distinct genetic identities. There were significant differences in the capacity for acclimation between traits (swimming performance, citrate synthase and lactate dehydrogenase activities), between lineages and between populations within lineages.
We rejected the hypothesis that climatic conditions (coastal vs. mountain) determined the capacity for acclimation, but accepted the hypotheses that demographic processes and genetic constraint influenced thermal acclimation.
The importance of our data lies in proof of concept that there can be substantial variation in thermal plasticity between populations within species. Similar responses are likely to be found in other species that comprise structured populations. Many predictions of the impact of climate change on biodiversity assume a species-specific response to changing environments. Based on our results, we argue that this resolution can be too coarse and that analysis of the impacts of climate change and other environmental variability should be resolved to a population level.
Introduction
The ability to respond to environmental change determines the persistence of species and populations in variable environments (Stillman 2003; Thompson 2003; Hoffmann & Sgrò 2011). The capacity for reversible within-individual plasticity (acclimation) in particular may render individuals more resilient to environmental change compared to those individuals with fixed phenotypes (Wilson & Franklin 2002; Seebacher et al. 2010). Acclimation counteracts potentially negative physiological effects resulting from thermal and other environmental changes, such as thermodynamic effects on reaction rates, and thereby maintains fitness across a broad range of environments (Wilson & Franklin 2002; Seebacher et al. 2010).
The relationship between a performance trait and temperature is often described by a Gaussian thermal sensitivity curve (an inverted ‘U’ shape) with a characteristic maximum value, or mode and breadth (Gabriel et al. 2005; Angilletta 2006). Acclimation may shift the mode along the temperature axis so that maximal performance coincides with mean environmental temperatures (Gabriel & Lynch 1992), or the breadth of the thermal sensitivity curve may change so that individuals can perform across a broader temperature range (Levins 1968; Huey & Hertz 1984; Lynch & Gabriel 1987). In theory, the capacity for thermal (reversible) acclimation depends on the intragenerational variance in temperature, which provides the selection pressure for acclimation (Gabriel et al. 2005). Populations from environments with little intragenerational variability are predicted to have performance optima that coincide with mean environmental temperatures (Lynch & Gabriel 1987; Lande 2009; Poisot et al. 2011), and it has been suggested that these populations have little capacity for acclimation compared with those from more variable environments (Tewksbury, Huey & Deutsch 2008). However, there are examples that contradict this prediction, and Antarctic fish that live in an extremely stable environment, for example, are well able to acclimate to temperatures well outside the range currently experienced (Franklin, Davison & Seebacher 2007). On the other hand, changes in mean environmental temperatures within generations, such as seasonal changes, may induce acclimation and cause a shift in the mode of the thermal sensitivity curve so that performance is optimized at the new environmental mean (Lynch & Gabriel 1987; Gabriel et al. 2005). Hence, if the capacity for acclimation is sensitive to environmental variability, it may be predicted that in organisms which experience seasonal fluctuations, populations from environments with colder winters should be better able to acclimate to lower temperatures than those from warmer environments. Pronounced short-term (e.g. daily) variation in temperature may mask the longer-term (e.g. seasonal) signal, thereby weakening the environmental induction of plasticity while increasing the width of the thermal sensitivity curve (Huey & Hertz 1984; Ruel & Ayres 1999; Sinclair, Thompson & Seebacher 2006; Lande 2009). These predictions assume that there is a genetic, heritable basis for acclimation. If this were the case, random genetic drift and founder effects could also cause differences in the capacity for acclimation independently from environmental conditions (Lande 1976; Parker, Rodriguez & Loik 2003). Phylogenetic constraints resulting from heritable responses to ancestral environments can also modulate plastic or adaptive responses (Felsenstein 1985; Poisot et al. 2011). Consequently, phylogenetic relationships would predict acclimation capacity. On the other hand, if populations are genetically constrained, and there is no genetic variation for acclimation, there should be no differences in the capacity for acclimation between populations. In this case, the thermal optima of populations would be fixed and coincide with the long-term mean temperatures experienced over evolutionary time, and the thermal breadth should reflect environmental variance (Levins 1968).
The aim of this study was to determine the capacity for acclimation in different populations of mosquitofish Gambusia holbrooki. We tested the following alternative hypotheses based on the theoretical background briefly outlined above: (i) if environmental means and variances determine acclimation, populations from cooler climates at higher altitudes have greater capacity for cold acclimation, and acclimation response to seasonal changes is weaker in environments with greater daily temperature fluctuations; (ii) if demographic processes are important, phylogenetic relationships of populations determine the capacity for acclimation; and (iii) if populations are genetically constrained, the capacity for acclimation is low and does not differ between populations, but the thermal optima for performance differ between mountain and coastal populations, reflecting differences in long-term climatic conditions.
To test these hypotheses, we analysed thermal responses of mosquitofish (Gambusia holbrooki) from four different coastal and two different mountain locations in NSW, Australia. In the twentieth century, mosquitofish were introduced into many parts of the world from their native range in the south-eastern United States. In many places, the species has become invasive and has greatly expanded its range into new habitats (Pyke 2008). These characteristics make mosquitofish ideally suited to test the hypotheses proposed here. We show that phylogenetic lineages within the species determine the capacity for acclimation, which means that different populations possess different capacities to compensate for thermal change.
Materials and methods
Study animals and sites
We collected animals in the wild from six locations that were within 100 km of their original release site at the Royal Botanical Gardens in Sydney: two populations [Lake Catalina (33°42′41″S, 150°18′14″E, altitude = 978 m), Glenbrook Lagoon (33°45′37″S, 150°36′46″E, altitude = 200 m)] originated from the Blue Mountains which rise from the coastal plain c. 50 km inland. The other four populations were collected from locations on the coastal plain where greater Sydney is situated [Homebush (33°51′51″S, 151°04′56″E), Victoria Park (33°53′7″S, 151°11′35″E), Manly Dam (33°46′19″S, 151°14′45″E) and Royal Botanical Gardens (i.e. the original release site 33°51′53″A, 151°13′00″E)]. The coastal sites all were within 20 km of each other and at elevations of <50 m. We measured water temperatures of the mountain sites (Lake Catalina, Glenbrook Lagoon), and a representative coastal site (Manly Dam) by deploying temperature data loggers (iButton; Maxim Integrated Products Inc., San Jose, CA, USA) programmed to record every 30 min at depths at which fish were most commonly observed (between 0·1 and 0·5 m); note that some of the raw data were lost so that we presented and analysed daily means for Lake Catalina only.
Fish were caught with hand nets at the same time of year (late spring/summer) and maintained at 22 °C for 1–2 weeks before acclimation treatments were started. We used only male fish for all experimental measurements, but we kept fish in mixed shoals to facilitate normal behavioural interactions (Sinclair, Ward & Seebacher 2011). The total length of all experimental fish (except for those used for AFLP analysis) from the tip of the snout to the terminal end of the peduncle was determined with callipers to the nearest 0·1 mm. The average length of fish was 16·17 (0·022 SE) mm and all experimental fish were size matched to within 2·0 mm of each other.
AFLP analysis
Animals were euthanized in a solution of MS222 (250 mg L−1 buffered to pH 7·0; Sigma, Sydney, Australia), and DNA was isolated from the tails of mosquitofish (26–32 from each location, n = 172 total) using a standard phenol/chloroform extraction procedure (Hillis, Moritz & Mable 1996). Determination of individual haplotypes (banding patterns) was made according to the original AFLP protocol (Vos et al. 1995) with slight modifications as follows. Two to 700 ng of DNA was digested using EcoRI and MseI restriction enzymes for 2 h at 37 °C in 1X NEB4 buffer [New England BioLabs (NEB), Ipswich, MA, USA]. Adapters (see Table S1 in Supporting Information) were then ligated to the cut sites, using T4 DNA ligase (100 U/reaction) in 1X NEB T4 ligase (1 mm adenosine triphosphate, ATP) buffer, by incubating DNA fragments for 3 h at 37 °C. Preselective amplification was performed using the primers E01 and M02 (Table S1) complementary to the adapters, but with one base extension. Each PCR was performed using 1 μL of ligation product with 1·25 U Thermopol Taq DNA polymerase, 0·25 mm DNTP each, primers (0·5 μm) and sufficient ddH2O and 1X NEB Thermopol buffer to give a final reaction volume of 20 μL, and run using the 1st PCR cycling sequence given in Table S1. The initial extract, the digest and preselective PCR products were visualized on 1·5% agarose gels to assess the success of the procedure and quality and quantity of the products.
Selective amplification of diluted (1 : 9 in ddH2O) preselective PCR products was made using eight primer combinations (Table S1) selected on the basis of their ability to discriminate between samples and their consistency. Each PCR was performed using 1 μL of diluted preselective amplification product, 1·25 U Thermopol Taq DNA polymerase, 0·25 mm DNTP each, 0·5 μm fluorescently labelled (6-FAM, VIC, NED, PET) EcoR1-primer, 0·5 mm MseI-primer and sufficient ddH2O and 1X NEB Thermopol buffer to give a final reaction volume of 20 μL and run using the 2nd PCR thermo-cycling sequence given in Table S1. Reactions with different fluorescent labels were subsequently pooled, 1 : 1 : 1 : 2 (6-FAM:VIC:NED:PET) and analysed using an Applied Biosystems 3730 DNA Analyzer with GeneScan-500 (LIZ) (Applied Biosystems, Foster City, CA, USA) as the size standard. Fragment analysis was performed using SequentiX GelQuest software version 2.6.5 (SequentiX, Klein Raden, Germany) with bands being scored in a binary fashion as present (1) or absent (0). Bands were assigned to bins based on 3 bp size intervals and only bands above a preset level of intensity (>150) were scored. To ensure that the results were reliable and replicable, two separate preselective amplifications were performed on three samples from each population and subsequently processed as separate samples. In addition, a positive and negative control was used throughout all reactions.
Acclimation treatments and performance
To test whether there was a significant relationship between genetic identity and animal function, we exposed fish from each population to different thermal acclimation treatments and determined performance over an acute temperature range. We chose sustained swimming as a whole animal performance measure that is closely related to fitness (Le Galliard, Clobert & Ferriere 2004; Husak et al. 2006), plus biochemical indicators of anaerobic (lactate dehydrogenase, LDH activity) and aerobic (citrate synthase, CS activity) metabolic energy (ATP) production capacity to characterize the thermal dependence and plasticity of animal performance. Lactate dehydrogenase is an indicator of the capacity for glycolytic, anaerobic ATP production and CS controls flux through the citric acid cycle and its maximal activity therefore provides a measure of capacity of mitochondrial, oxidative ATP production. Fish from each population were acclimated to either 10 °C (n = 30 per population) or 25 °C (n = 30 per population) and cold and warm treatment tanks were interspersed randomly within the same constant temperature room. The rooms were set at 10 °C and aquarium heaters were used to achieve the 25 °C treatment; 10 °C tanks also contained aquarium heaters but these were not switched on. Acclimation temperatures were chosen to represent the lower and upper range of mean daily temperatures experienced by all populations. At the beginning of the experiment, temperatures were changed by 2 °C per day from the initial 22 °C until the desired acclimation temperatures were reached, and fish were kept for 4 weeks at the respective acclimation temperatures.
We determined critical sustained swimming speed (Ucrit; Kolok 1999), as a measure of fitness-related aerobic performance. Ucrit was determined according to published protocols (Sinclair, Ward & Seebacher 2011). Briefly, fish were swum in a swimming flume consisting of a 15 cm × 5 cm diameter clear Perspex tube tightly fitted over the intake end of a submersible inline pump (12V DC, iL500, Rule, King Pumps Inc., Miami, FL, USA). A plastic grid separated the Perspex tube from the pump and a bundle of hollow straws at the inlet helped maintain laminar flow; the flume and pump were submerged in a plastic tank (645 × 423 × 276 mm). We used a variable power source (MP3090; Powertech, Perth, Australia) to adjust the flow speed, which was calibrated using a flow meter (FP101; Global Water, Gold River, CA, USA). The initial velocity was 8 cm s−1 for 20 min, after which flow was increased by 2 cm s−1 every 5 min until fish fatigued. Ucrit was calculated as Ucrit = Uf + (Tf/Ti)Ui), where Uf is the highest speed increment maintained for the full 5 min interval, Tf = the time swum at the highest speed increment, Ti is the time interval between speed increments (5 min), and Ui is the speed increment (2 m s−1). We determined Ucrit for n = 9 fish from each acclimation treatment and each population at each of three acute test temperatures (10 °C, 15 °C and 25 °C). In case fish did not swim for the entire initial 20-min period (as was the case for some warm acclimated fish swum at 10 °C), we estimated Ucrit by setting Uf = 0, Tf = the time swum at 8 cm s−1 and Ti = 20 min.
We collected tail muscle tissue (n = 9 fish per acclimation treatment and population) for assays of LDH and CS activities. After euthanasia in a solution of MS222 (250 mg L−1 buffered to pH 7·0; Sigma), fish were skinned and tails were removed and immediately frozen in liquid nitrogen and stored at −80 °C for later analysis. Tissue samples (0·05–0·1 g) were homogenized in 9 vols of ice-cold extraction buffer (50 mm imidazole, 2 mm MgCl2, 5 mm EDTA, 0·1% Triton and 1 mm glutathione, pH 7·5 at 20 °C). Enzyme activities were determined according to published protocols (Seebacher & Walter 2012) in a spectrophotometer with a temperature-controlled cuvette holder (Ultrospec 2100 pro UV; GE Healthcare, Sydney, Australia). Saturating substrate concentrations were ascertained in pilot assays, and enzyme activity was expressed as micromolar of substrate converted min−1 g−1 wet tissue mass. Enzyme activities in muscle tissue from each fish were determined in duplicate at 10, 15 and 25 °C.
Statistical analyses
We used Kruskall–Wallace tests to compare mean daily water temperatures between sites within seasons and to compare daily water temperature ranges between Glenbrook Lagoon and Manly Dam in summer and in winter.
We calculated a matrix of Euclidean (unsquared) distances from the AFLP fragments and subjecting the resulting matrix to a principal coordinates (PCO) analysis using MVSP (Kovach Computing Services, Anglesey, Wales) to visualize the results. To investigate the structure between populations, we performed an analysis of molecular variance (amova) using the GeneAlex 6 package (Peakall & Smouse 2006) with a squared Euclidean distance matrix (default setting for binary data). About 10 000 iterations were run to assess the significance of the variance components within and between populations for various groupings and for calculating the significance of individual pairwise genetic distances (Φpt). Significance levels for the pairwise comparisons were adjusted for multiple comparisons by a sequential Bonferroni correction procedure. Various different models were tested, based on the PCO plot, to investigate the relationship between the populations and the model that was both statistically valid and gave the least within-population variance selected. The best fitting model was obtained by dividing the populations into two lineages each comprising three populations: Catalina + Glenbrook + RGB and Manly + Victoria + Homebush (see 3). Mantel tests were used to examine the correlation between the geographical distances between the populations and the Φpt values from amova, and Nei's genetic distance (Nei 1972), calculated using AFLP-SURV (Vekemans 2002).
Swimming performance and enzyme activities were analysed by permanova (Primer 6; Primer-e software, Ivybridge, UK), and our analysis strategy followed from the results of the AFLP analysis. Hence, we first analysed data with phylogenetic lineage, acclimation treatment and test temperatures as fixed factors to test the hypothesis that phenotypes differ between lineages. Secondly, to test the hypothesis that population phenotypes within lineages are distinct, we used populations within lineages, acclimation treatment and test temperature as fixed factors. Lastly, to test the hypothesis that the phenotypic traits acclimate within each population (i.e. are not genetically constrained), we used acclimation treatment and test temperatures as fixed factors within each population.
We used the truncated product method (Zaykin et al. 2002) to assess the effect of multiple comparisons on the validity of P-values. Briefly, the truncated product method considers the distribution of P-values from multiple hypothesis tests to provide a tablewide P-value for the overall hypothesis that P-values were not skewed, leading to type 1 errors. Multiple hypothesis testing did not bias the statistical results presented here (P < 0·0001).
Results
Water temperatures
There were significant differences in mean daily temperatures and daily variation between the sites in summer and in winter (Fig. 1). Daily variation was significantly less at the mountain site compared with the coastal site (summer χ2 = 34·47, P < 0·0001; winter χ2 = 48·86, P < 0·0001), and Lake Catalina and Glenbrook Lagoon had lower daily mean temperatures than the coastal site (summer χ2 = 50·38, P < 0·0001; winter χ2 = 73·99, P < 0·0001; Fig. 1).
AFLP analysis
Analysis of the genetic identity of the six populations (see Appendix S1 and Table S2 in Supporting Information) showed that populations were genetically distinct from each other and that they were divided into two distinct lineages [Analysis of molecular variance (AMOVA) P < 0·001; see Table S3 in Supporting Information; Fig. 2]. The first lineage included the original release site of mosquitofish in Australia in 1926 at the coastal Royal Botanical Gardens in Sydney (Lloyd, Arthington & Milton 1986) plus the two mountain populations (Lake Catalina and Glenbrook Lagoon). The second lineage comprised three coastal sites (Homebush, Manly Dam, Victoria Park; Fig. 2). Between lineage differences accounted for 45% of the total variance (amova: Φrt = 0·499; P < 0·001). Variation between populations within a lineages explained 7% of the total variance (Φpr = 0·128; P < 0·001), and the remaining 48% of the variance was within populations (Φpt = 0·519, P < 0·001; Table S3). Surprisingly, individual population identity was much stronger in the second lineage that comprised only coastal populations (Φpt ≥ 0·118) compared with the first lineage that comprised both mountain population and coastal population (Φpt ≥ 0·036; Table S4 in Supporting Information). Mantel's test indicates that there was no association between geographic distance and genetic distance between the different populations (Table S4).
Swimming performance and enzyme activities
The thermal sensitivity of swimming performance (Fig. 3), LDH activities (Fig. 4) and CS activities (Fig. 5) differed significantly between the two phylogenetic lineages (Botanical Gardens/Glenbrook/Lake Catalina vs. Victoria Park/Homebush/Manly Dam) as indicated by the phylogeny × test temperature interaction for all traits (Table 1). The capacity for acclimation of CS activity also differed between the two lineages (phylogeny × acclimation interaction, Table 1), and the different acclimation treatments changed the thermal sensitivity of swimming performance in both lineages (acclimation × test temperature interaction, Table 1).
| U crit | LDH | CS | ||||
|---|---|---|---|---|---|---|
| F d.f. | P | F d.f. | P | F d.f. | P | |
| Phylogeny | 3·561,312 | 0·21 | 9·721,312 | 0·071 | 14·771, 312 | 0·094 |
| Acclimation | 2·351,312 | 0·29 | 3·101,312 | 0·32 | 21·451,312 | 0·091 |
| Test temp. | 83·072,312 | <0·001 | 80·802,312 | <0·001 | 127·822,312 | <0·001 |
| Ph*T | 4·332,312 | <0·01 | 3·382,312 | <0·02 | 3·632,312 | <0·01 |
| Ph*A | 2·191, 312 | 0·23 | 0·791, 312 | 0·48 | 12·931,312 | <0·05 |
| A*T | 4·322,312 | <0·01 | 0·242,312 | 0·90 | 0·142,312 | 0·96 |
| Ph*A*T | 1·102,312 | 0·36 | 0·132,312 | 0·97 | 0·292,312 | 0·83 |
- Pseudo-F values, degrees of freedom (Fd.f.), and permutation probabilities (P) are shown, and significant results are highlighted in bold.
Populations within lineages differed significantly in their thermal responses (Table 2). In the first lineage comprising Lake Catalina, Glenbrook Lagoon and Royal Botanical Gardens, the three-way interaction between population × acclimation × test temperature for Ucrit indicates that both the thermal sensitivity and the capacity for acclimation of swimming performance differs between these populations (Table 2; Fig. 3); The population × test temperature interaction for LDH indicates that the thermal sensitivity of LDH activity differed between populations (Table 2; Fig. 4). The main effects of population, acclimation and test temperature on CS activities indicate quantitative differences between populations, as well as quantitatively different responses to acclimation and acute test temperature changes (Table 2; Fig. 5). In the second lineage comprising Victoria Park, Homebush, and Manly Dam, population of origin significantly affected all traits (Table 3). Additionally, acclimation responses of LDH and CS activities differed significantly between populations (population × acclimation interactions; Table 2), whereas thermal sensitivity of Ucrit significantly changed (acclimation × test temperature interaction).
| U crit | LDH | CS | ||||
|---|---|---|---|---|---|---|
| F d.f. | P | F d.f. | P | F d.f. | P | |
| (a) Lake Catalina, Glenbrook Lagoon, Botanical Gardens | ||||||
| Population | 18·052,144 | <0·001 | 85·962,144 | <0·001 | 14·052,144 | <0·001 |
| Acclimation | 10·141,144 | <0·001 | 1·561,144 | 0·53 | 5·511,144 | <0·02 |
| Temperature | 43·652,144 | <0·001 | 40·232,144 | <0·001 | 43·432,144 | <0·001 |
| P*A | 3·462,144 | <0·001 | 1·822,144 | 0·14 | 1·862,144 | 0·15 |
| P*T | 5·394,144 | <0·001 | 6·114,144 | <0·001 | 1·694,144 | 0·12 |
| A*T | 3·112,144 | <0·02 | 0·372,144 | 0·87 | 0·332,144 | 0·81 |
| P*A*T | 2·314,144 | <0·02 | 0·194,144 | 0·99 | 0·384,144 | 0·88 |
| (b) Victoria Park, Homebush, Manly Dam | ||||||
| Population | 3·72 2,144 | <0·01 | 5·952,144 | <0·002 | 19·902,144 | <0·001 |
| Acclimation | 4·461,144 | <0·03 | 1·971,144 | 0·19 | 0·251,144 | 0·66 |
| Temperature | 71·402,144 | <0·001 | 222·862,144 | <0·001 | 188·52,144 | <0·001 |
| P*A | 0·812,144 | 0·49 | 3·372,144 | <0·04 | 8·452,144 | <0·001 |
| P*T | 1·634,144 | 0·094 | 1·094,144 | 0·38 | 1·154,144 | 0·33 |
| A*T | 3·902,144 | <0·006 | 0·342,144 | 0·83 | 0·0882,144 | 0·98 |
| P*A*T | 0·914,144 | 0·48 | 0·644,144 | 0·70 | 0·794,144 | 0·53 |
- Pseudo-F values, degrees of freedom (Fd.f.), and permutation probabilities (P) are shown, and significant results are highlighted in bold.
| Population | Factor | U crit | LDH | CS | |||
|---|---|---|---|---|---|---|---|
| F | P | F | P | F | P | ||
| Lake Catalina | A | 6·18 | <0·002 | 4·52 | <0·04 | 2·64 | 0·11 |
| T | 27·22 | <0·001 | 44·32 | <0·001 | 20·64 | <0·001 | |
| A*T | 5·51 | <0·001 | 1·09 | 0·35 | 0·56 | 0·59 | |
| Glenbrook L. | A | 6·85 | <0·002 | 0·55 | 0·52 | 0·13 | 0·82 |
| T | 26·15 | <0·001 | 18·41 | <0·001 | 20·32 | <0·001 | |
| A*T | 3·01 | <0·03 | 0·12 | 0·96 | 0·17 | 0·89 | |
| Royal B. Gardens | A | 5·04 | <0·02 | 1·14 | 0·29 | 5·30 | <0·02 |
| T | 11·96 | <0·001 | 10·03 | <0·001 | 9·35 | <0·001 | |
| A*T | 1·52 | 0·19 | 0·13 | 0·97 | 0·37 | 0·78 | |
| Homebush | A | 5·69 | <0·01 | 2·34 | 0·15 | 5·99 | <0·02 |
| T | 55·76 | <0·001 | 56·95 | <0·001 | 85·88 | <0·001 | |
| A*T | 5·32 | <0·004 | 0·53 | 0·62 | 0·62 | 0·53 | |
| Victoria Park | A | 1·50 | 0·22 | 2·51 | 0·10 | 12·74 | <0·002 |
| T | 13·56 | <0·001 | 97·09 | <0·001 | 69·89 | <0·001 | |
| A*T | 1·15 | 0·31 | 0·42 | 0·66 | 1·10 | 0·34 | |
| Manly Dam | A | 1·17 | 0·29 | 3·65 | <0·05 | 0·38 | 0·55 |
| T | 29·79 | <0·001 | 79·20 | <0·001 | 46·32 | <0·001 | |
| A*T | 1·58 | 0·18 | 0·62 | 0·59 | 0·11 | 0·92 | |
- Pseudo-F values, and permutation probabilities (P) are shown, and significant results are highlighted in bold.
There was considerable variation in the capacity to acclimate the different traits between populations (Table 3). Ucrit acclimated in all populations except for Victoria Park and Manly Dam (Table 3; Fig. 3). LDH activity acclimated only in fish from Lake Catalina and Manly Dam (Table 3; Fig. 4), while CS activity acclimated in fish from Royal Botanical Gardens, Homebush, and Victoria Park (Table 3; Fig. 5).
Discussion
We have shown that the capacity for thermal acclimation differs between populations and phylogenetic lineages within a species, thereby providing experimental evidence that there is variation within species in within-individuals thermal plasticity. The importance of our data lies in proof of concept that there can be substantial variation in thermal plasticity between populations within species. There was surprisingly little gene flow and strong structuring of our mosquitofish populations, considering that the populations occurred within a relatively small geographic area. Similar values of Φ for proportion of genetic divergence because of lineage and populations as we found for mosquitofish usually occur in populations across much broader geographic ranges (Thacker et al. 2008; Liu et al. 2009). Additionally, genetic distance between populations of mosquitofish was not related to geographic distance, which makes it unlikely that genetic differences between populations can be explained by differential selection pressures when animals radiated to mountain habitats from their coastal introduction site. It is possible that the two genetic lineages of mosquitofish resulted from separate introductions. Nonetheless, even within each lineage populations had a distinct genetic identity.
There was no indication that environmental conditions exerted selection pressures that could explain genetic divergence. The coastal populations, which occupied similar climates, were distinct from each other, and the Royal Botanical Gardens population was closer related to the mountain populations than to the other coastal populations. In other words, the phylogenetic relationships did not align with the different environments occupied by the populations. It seems likely that the observed genetic divergence between populations resulted from genetic drift and founder effects (Lande 1976; Barton & Charlesworth 1984) resulting from separate introductions, which could explain differences between lineages, or from few fish being transported to new water bodies within the same geographic area, which could explain divergence of coastal populations. Additionally, populations may evolve in response to biotic selection pressures that are independent from climatic differences. For example, populations of guppies (Poecilia reticulata), which are closely related to mosquitofish, diverged rapidly within 10–100 generations in response to selection exerted by predators (Magurran 1998).
Variation in thermal responses, both to changes in acute test temperature and to longer-term acclimation, between populations was surprisingly large, and our data cannot be explained by a single hypothesis. All populations showed thermal acclimation in at least some of the traits, so that we reject the hypothesis that acclimation per se is genetically constrained. However, the generality of this conclusion is limited because some traits did not acclimate in some populations. Lack of acclimation may indicate that a particular trait is not limiting for fitness, so that there is no selection for acclimation responses. However, if this were the case, responses should be similar in populations from the similar coastal or mountain climates, which was not the case. Hence, it appears that genetic constraint of acclimation varies between populations.
We predicted that fish from colder environments that showed less daily temperature fluctuations should acclimate better to cold conditions than fish from populations experiencing warmer temperatures and greater short-term variability. Again, there is no clear pattern that would allow us to accept this hypothesis. Fish from the coldest environment at Lake Catalina showed the most pronounced cold acclimation response of Ucrit and LDH activity, but did not acclimate CS activity. Additionally, some of the coastal populations showed similar significant cold acclimation, which means that daily variability in temperature does not preclude acclimation. The similarities between mountain and coastal populations, and the differences between the coastal populations in the thermal sensitivity and acclimation capacity of the phenotypic traits indicate that environmental variability alone does not determine thermal responses. Rather, the combination between demographic processes, genetic constraint, and environmental selection pressures determine thermal plasticity.
Developmental plasticity (Ho & Burggren 2010) may have caused differences between populations from different climates. Environmental conditions experienced during development can determine phenotypes later in life by changing DNA and histone methylation patterns and thereby gene expression profiles (Klose & Bird 2006; Greer & Shi 2012). Developmental conditions can alter phenotypes of natural populations (Niewiarowski & Angilletta 2008; Lind & Johnasson 2011). Little is known, however, about the underlying mechanisms in wild populations, although different species of fish occupying different thermal habitats do show consistently different DNA methylation patterns (Bucciarelli et al. 2009). Developmental processes could explain differences between coastal and mountain populations, but not differences between coastal populations that experienced similar developmental conditions.
Most theories attempting to explain the evolution of plasticity in fact refer to developmental plasticity. Hence, the theoretical predictions derived from quantitative genetic models (Lande 2009) may not be applicable to reversible acclimation because the mechanistic bases of these two forms of plasticity are likely to be different. The capacity for reversible acclimation may be modulated by developmental processes, but acclimation by itself induces molecular changes that alter phenotypic traits. For example, cold acclimation in animals that developed under similar conditions causes an increase in the expression of transcriptional regulators of metabolism (e.g. PGC1α and PPARs), which is paralleled by increased enzyme activities (e.g. CS and cytochrome c oxidase) similar to those we report for mosquitofish (Le Moine, Genge & Moyes 2008; Seebacher, Murray & Else 2009). The signalling cascade that links changes in environmental temperatures to cellular responses could be central via peripheral thermal sensing proteins that send afferent signals to the brain to produce an efferent sympathetic response to tissues and cells (Seebacher 2009). Alternatively, a temperature-induced imbalance in cellular energy status (ATP/AMP ratio) would elicit an AMP-activated protein kinase (AMPK)-mediated response to adjust metabolic status (Hardie 2008). Both AMPK and sympathetic responses can cause an increase in cellular metabolism by stimulating transcriptional regulators, such as PGC-1α and its target transcription factors (Lin, Handschirm & Spiegelman 2005; Hardie 2008). Additionally, the thermal sensitivity of mitochondrial enzymes is modulated by membrane fatty acid composition, which is influenced by environmental conditions including by thermal acclimation (Hulbert & Else 1999; Seebacher, Murray & Else 2009; Cooper et al. 2012). Reversible acclimation may therefore be mediated by protein evolution and hence genetic constraint (Gilchrist & Lee 2007; Mitchell-Olds, Willis & Goldstein 2007; Bremer & Moyes 2011), as well as by epigenetic mechanisms (Mattick 2010), and signal transduction pathways (e.g. AMPK mediated) that provide an intrinsic capacity of the cell to respond to environmental change.
Many species comprise distinct populations (Ward, Woodwark & Skibinski 1994; Song et al. 2006; Liu et al. 2009) and thermal acclimation is widespread among fish (Guderley 2004) and other vertebrates, and invertebrates (Piersma & Drent 2003), so that the interactions between genetic structure and thermal responses we showed for mosquitofish are likely to be a more widespread phenomenon. For example, one of the best studies and commercially valuable fish species, Atlantic cod Gadus morhua, comprise genetically structured populations with regionally varying dynamics in the North Sea and Atlantic Ocean (Hutchinson, Carvalho & Rogers 2001; Hutchinson 2008). At the same time, it appears that different populations show varying capacities for thermal acclimation (Lucassen et al. 2006; Sylvestre et al. 2007; Lurman, Bock & Poertner 2009). Although the genetic or demographic processes that led to this variation in acclimation capacity between populations may vary between species, the outcome is similar and may affect population persistence in variable climates.
Acclimation buffers the effect of environmental variation on biological functions, and it can therefore increase the resilience of individuals to thermal change, including that resulting from climate change (Franklin, Davison & Seebacher 2007; Chown et al. 2010). A species that comprises populations with diverse physiological phenotypes will be more resilient to climate change because of the increased likelihood that at least some populations will thrive under the changed conditions.
Understanding the potential for animals to move into novel environments and to withstand climate change requires an understanding of their ability to acclimate performance such as enzyme activities and locomotion. Predicting the future impact of climate change therefore requires an understanding of acclimation (Seebacher & Franklin 2012), and our data show that this needs to be resolved to a population level. The capacity to modulate the thermal sensitivity of physiology and behaviour can alter population dynamics (Gaston 2009) and change ecological interactions with other species (Grigaltchik, Ward & Seebacher 2012). The importance of thermal plasticity of these last two aspects remains largely unexplored. At the current state of knowledge, there are no universal mechanisms that explain geographic distributions of species (Gaston 2009; Calosi et al. 2010), and understanding the connections between thermal plasticity, population dynamics and ecological interactions will be important.
Acknowledgements
This research was supported by funding from the Australian Research Council to FS. Collection and experimental procedures were approved by the University of Sydney Animal Ethics Committee (approval no. L04-3-2008-3-4769), and by NSW Industry and Investment (Scientific Collection Permit no. P08/0028-3.0).
