ABSTRACT
Trophic interactions are important determinants of the structure and functioning of ecosystems. As the metabolism and consumption rates of ectotherms increase sharply with rising temperature, there are currently major concerns that global warming will increase the strength of trophic interactions, destabilizing food webs, and altering ecosystem structure and function. We used geothermally warmed streams that span a ~10 ºC temperature gradient to investigate the interplay between the thermal response of respiration, local adaptation, and the interaction strength between the keystone gastropod grazer, the wandering snail Radix balthica, and a common algal resource. Populations from a warm stream (~28ºC) had higher maximal metabolic rates and optimal temperatures across all measurement temperatures than those from a colder stream (~17ºC), suggesting local adaptation of metabolic rates. A reciprocal transplant experiment demonstrated that the interaction strength between the grazer and its resource were highest for both populations when transplanted into the warm stream. In line with the thermal response curves for respiration, interaction strengths of the warm-adapted grazers were higher than their cold-adapted counterparts in both the warm and the cold stream. These findings suggest that warming can increase the strength of algal-grazer interactions both through the thermodynamic effects of higher temperatures on physiological rates and through correlated increases in per capita metabolism and consumption as organisms adapt to warmer temperatures.
INTRODUCTION
The strength of consumer-resource interactions (e.g. the effect of a consumer on the population density of its prey) play a critical role in shaping the stability of food webs (May 1973; Paine 1980; McCann et al. 1998; Otto et al. 2007). Grazing is an important class of consumer-resource interaction, determining the flux of energy and materials from autotrophs to heterotrophs. There are currently major concerns that global warming will increase the impact of grazers on algal or plant communities because the ingestion and respiration rates of heterotrophs tend to be more sensitive to rising temperatures than rates of photosynthesis and growth in autotrophs (O’Connor 2009; Gilbert et al. 2014; West and Post 2016). Stronger interactions have the potential to destabilise food webs and consequently, warming induced increases in interaction strengths could have fundamental implications for ecosystem structure and function. For example, elevated grazing rates in aquatic ecosystems, driven by the mismatch in thermal sensitivity between autotrophs and heterotrophs, are a key driver of projected declines in aquatic primary production over the 21st century in models of ocean biogeochemistry (Laufkötter et al. 2015).
The effects of temperature on metabolic rates and traits associated with consumer-resource interactions (e.g. consumption rates, handling times) often follow characteristic unimodal thermal response curves, in which rates increase exponentially to an optimum and decline rapidly thereafter (Dell et al. 2011, 2014; Englund et al. 2011; Rall et al. 2012; Gilbert et al. 2014). Integrating thermal responses for metabolism and interaction-traits with dynamical models of consumer-resource interactions offers a promising framework for predicting food web responses to global warming (Vasseur and McCann 2005; Shurin et al. 2012; Binzer et al. 2016). However, thermal response curves are often evolutionarily flexible (Angilletta et al. 2003; Kingsolver et al. 2004; Deutsch et al. 2008; Kingsolver and Huey 2008) and can shift as organisms adapt to novel thermal environments, meaning that rapid evolution could modulate the effects of rising temperatures on the strength of species interactions. For example, if thermal adaptation serves to down-regulate metabolic rates at higher temperatures (Addo-Bediako et al. 2002), then rapid evolutionary responses to warming could mitigate predicted increases in consumer-resource interaction strength. How adaptation to warming affects rates of metabolism and in turn, the strength of consumer-resource interactions, is largely unknown, limiting our ability to predict how trophic interactions will change in response to warming in the long-term.
There is evidence from studies across naturally occurring thermal gradients over large spatial scales, that local thermal adaptation can play an important role in shaping the strength of species interactions (Barton 2011; De Block et al. 2012). While these studies provide important insights into how consumer-resource interactions are shaped by evolution across thermal gradients (Fukami and Wardle 2005), their usefulness for understanding the mechanisms underlying responses to rapid climate warming might be limited, because other factors, such as day length, light intensity and precipitation, tend to be confounded with temperature along such broad scale spatial gradients. Furthermore, the timescales over which local adaptation has occurred in such broad scale studies could be much longer than the rapid evolutionary change required to keep pace with climate warming (Loarie et al. 2009; Hoffmann and Sgrò 2011). Here, we focus on the interplay between the effects of local thermal adaptation on metabolism and the strength of a keystone grazing interaction (the gastropod Radix balthica, which grazes algal biofilms in streams) in naturally warmed Icelandic geothermal streams spanning a gradient of 11ºC. Critically, temperature is the main abiotic factor that varies among streams in the catchment and is not correlated with pH, conductivity or inorganic nutrient concentrations (see Table 1), and the streams are thought to have been subject to geothermal heating for at least the last century (O’Gorman et al. 2012). This system therefore provides the opportunity to investigate the mechanisms that shape how temperature and local adaptation influence species interactions in a natural system, where the effects of confounding factors are minimised. Specifically, we ask (i) can the underlying responses of metabolism to temperature explain the magnitude of the effects of warming on the strength of algal-grazer interactions? (ii) Can local thermal adaptation dampen the direct effects of warming on the strength of consumer-resource interactions?
METHODS
Study site
The streams are located North of the Icelandic Hveragerði valley, in the south east of the Hengil high temperature geothermal field (N64° 0’ 2.944" W21° 11’ 17.451") and consist of a catchment of 11 streams spanning a temperature gradient of approximately 20 ºC (see Figure 1 and SI Figure 1). Two streams, stream 5 (17.5 ºC ± 4.5 ºC, hereafter ‘cold stream’) and stream 11A (28.3 ºC ± 1.3 ºC, hereafter ‘warm stream’), were chosen for experiments due to their close proximity, large temperature difference and the abundance of the keystone grazer, Radix balthica. The grazer plays an important functional role geothermal stream ecosystems, where grazer biomass as well as grazing rates are strongly influenced by temperature (O’Gorman et al. 2012). The two streams are similar in all other measured physical and chemical characteristics but differ in average temperature by 11 ºC (see Table 1), and hence present an opportunity to investigate how the effects of warming and local adaptation interact to shape the thermal dependence of consumer-resource interactions.
Grazer metabolism
To quantify whether local adaptation to the different thermal regimes in the two adjacent streams resulted in divergence in metabolic traits of R. balthica we measured the acute responses of respiration to a broad gradient in temperature. We collected 33 individuals of similar weight and length from each stream, which were cleaned from any algal debris to avoid carry-over of a food source into the tank or subsequent respiratory measurements on the oxygen electrode. The snails were kept overnight in aerated tanks at the average stream temperature of origin and in the absence of a food source to minimise any potential effects of differences in food quantity or quality between streams. Respiration was quantified as the rate of oxygen consumption in a Clark-Type oxygen electrode, measured between 4 – 44 ºC in 4 ºC increments (11 temperatures in total). At each temperature, respiration was measured for 3 individuals, and a different set of individuals was measured at each temperature (i.e. each animal was only subjected to a single assay). Individuals were allowed 15 minutes at the assay temperature prior to the measurements. The subsequent thermal responses of respiration were quantified using a modification of the Sharpe-Schoolfield equation (see Schoolfield et al. (1981) for the original equation): where b(T), is the per capita metabolic rate (µmol O2 L-1 h-1) at temperature T in Kelvin (K), k is Boltzmann’s constant (8.62×10-5 eV K-1), Ea is an apparent activation energy (in eV) for the metabolic process, ln (b(Tc)) is the rate of metabolism normalised to an arbitrary reference temperature, Tc = 18 ºC, where no low or high temperature inactivation is experienced. Mi is the mass (g) of an individual i, α is the allometric scaling exponent that characterises the power-law relation of mass and metabolic rate (Brown et al. 2004). Eh characterizes temperature-induced inactivation of enzyme kinetics above Th where half the enzymes are rendered non-functional. Differentiating equation (1) and solving for the global maxima yields an expression for the optimum temperature
Equation (1) differs from the Sharpe-Schoolfield equation (Sharpe & DeMichele 1977; Schoolfield et al. 1981) in a number of ways. First, we account for the power law relation between body mass and metabolic rate, Mα (Brown et al. 2004). Second, we exclude parameters from Eq. (1) used to characterize low-temperature inactivation due to insufficient data to quantify this phenomenon in our analysis. Second, rather than characterize temperature effects below Topt using the Eyring (1935) relation, , we instead use the simpler Boltzmann factor, . This simplification enables an explicit solution for Topt (Eq. 2) and facilitates more direct comparison with previous work on the temperature dependence of metabolism using metabolic theory (e.g. Gillooly et al 2001; Savage et al 2004; Brown et al 2004; Allen et al. 2005).
The parameters, ln b(Tc), α, Ea, Eh, Th, and Topt, in Eqs. (1) & (2) represent traits characterising the metabolic thermal response that we expect to be under selection in R. balthica inhabiting the hot and cold streams. We tested for differences in each of the parameters between the populations of R. balthica by fitting the respiration data to Eq. (1) using generalised non-linear least squares regression (within the ‘gnls’ function in the ‘nlme’ package for R, package version 3.1-128) and including ‘origin’ as a two level factor (i.e. ‘cold’ and ‘warm’ stream). We tested for differences between populations for each parameter by sequentially removing the effect of ‘origin’ on each parameter and comparing the Akaike information criterion for small sample sizes (AICc) for all possible models (see SI Table 1 and SI Table 2) using the ‘aictab’ and ‘modavg’ functions from the AICcmodavg package (package version 2.1-0). The model chosen for further exploration was that with the lowest (AICc) value. Model averaging was carried out when models fell within 2 AICc units of each other, and the conditional averages of the parameters were used for curve fitting and interpretation (see also Table 2). The relative importance of the fixed factors in the averaged model was determined using the sum of their relative weights.
Reciprocal transplant experiment
The reciprocal transplant experiments to assess the effects of temperature and local adaptation on algal-grazer interactions were carried out by placing snails in microcosms consisting of a tissue culture flask on which diatom biofilms had been established. Diatoms of the genera, Acnanthes, Nitzschia, Navicula, and Gomphonema are common in streams across the Hengill volcanic area (Guðmundsdóttir et al. 2012) were ordered from culture collections (Culture collection of algae and protozoa and Sciento) and grown in the laboratory in mixed assemblages to yield common resource for testing the effects of temperature and local adaptation on grazing. The diatom assemblages were inoculated into Corning plastic translucent flasks (maximum volume 1L) with 20 mL COMBO medium (Kilham et al. 1998), and brought to a salinity of 5-10 (equivalent to approximately 5-10 g salts/kg water) to match the slightly elevated salinity and conductivity found in these thermal stream environments (Guðmundsdóttir et al. 2012). The flasks were turned onto their sides to allow for a larger area of biofilm growth on the base (~ 60 cm2 in total per flask) and communities were left to grow for 14 days prior to the experiment. After 14 days all flasks had substantial biofilm development on the base and were used as microcosms for the in situ reciprocal transplant experiment. Analysis of control flasks (no grazer) showed that growth of the diatom lawn per se did not differ significantly for flasks placed in hot or cold streams (SI Figure 2). Thus, any changes to the biofilm biomass in the experiment can be attributed to the per capita effects of the grazer.
The experiment consisted of 3 treatments (each with 6 replicate microcosms placed in each of the 2 streams): (i) a control microcosm in which a biofilm was present and no R. balthica were added, (ii) an ‘origin’ treatment in which R. balthica that were resident in the stream were added to microcosms, and (iii) a ‘transplanted’ treatment in which R. balthica that were from the adjacent stream were added to microcosms. R. balthica individuals were collected from the 2 streams the day before the experiment and were starved for 24h in the laboratory in aerated tanks at the average temperature of the stream of origin. There was no significant difference in average snail weight between the two streams (see SI Figure 3). Microcosms were assembled by adding 3 snails of similar body dimensions (0.35 ± 0.03 g of R. balthica weight reported as blotted fresh weight throughout) and 100 mL of 0.4 µm filtered water from the stream in which the microcosm was to be placed. This resulted in a grazer density of 5 individuals m−2, which was comparable to the average in situ density in the streams (see SI Figure 4). This design was preferred to a set-up with each microcosm holding a single grazer, which attempt to exclude the effects of mutual interference on feeding behaviour (e.g. Lang et al. (2011), Skalski and Gilliam (2001),Rall et al. (2010); Vucic-Pestic et al. (2011)), because (i) the experimental densities are representative of natural conditions; and (ii) the consumption rates of a single individual were insufficient to detect a significant change in algal biomass. The microcosms were submerged in each stream and the snails were left to graze for 48 hours. We observed no grazer mortality over the experimental period.
Interaction strength
At the end of the experiment, algal biomass in each of the microcosms was quantified via methanol chlorophyll extraction (modified from Holm-Hansen & Riemann (1978)). Here, the walls of the microcosms were scrubbed until all biofilm particles were in suspension. The solution was filtered onto a 0.4µm GF/F filter, which was then ground in methanol for 5 minutes. The samples were centrifuged at 3500 rpm for 15 minutes and the absorbance of the supernatant was measured at 632nm, 665nm, and 750nm. Total chlorophyll content in µg mL−1 was then calculated as described in Holm-Hansen & Riemann (1978). The per capita interaction strength in each microcosm was then estimated by calculating the dynamic index (DI, see also Berlow et al. 1999; 2004 for a technically similar set-up): where N is total chlorophyll (sum of Chl a + Chl c) content of control, D total chlorophyll in the grazed microcosm, Y is the grazer biomass (g C), and t is time in hours. Snail blotted wet weight was converted to carbon mass (in grams) using conversion factors that assume dry weight to be 7.5% of the blotted wet weight (Ricciardi & Bourget 1998) and a carbon content of 22% dry weight (Burgmer et al. 2010).
We carried out two analyses using the data from the reciprocal transplant experiment. The first analysis, used a generalised linear model (GLM), with ‘interaction strength’ as the response variable and ‘origin’ (‘cold’ or ‘warm’ stream) and ‘transplant temperature’ (17.5 and 28.3 ºC) as potentially interacting factors. We used this analysis to determine (i) whether interaction strengths differed between snails that originated from the warm or cold streams (e.g. a main effect of ‘origin’); (ii) whether interaction strengths were temperature dependent (e.g. a main effect of ‘temperature’); and (iii) whether the temperature dependence of interaction strength differed between the snails from the cold and warm streams (e.g. interaction between ‘origin’ and ‘temperature’). The design of the reciprocal transplant experiment also enabled us to disentangle short-term temperature responses attributable to acclimation (e.g. responses to the temperature in the ‘transplanted’ stream) from those reflecting processes operating over longer, evolutionary time scales (e.g. adaptation to the temperature in the stream of ‘origin’). The second GLM included ‘interaction strength’ as the response variable and ‘timescale’ (‘short’ or ‘long’) and ‘transplant temperature’ (17.5 and 28.3 ºC) as potentially interacting factors. Here, ‘short-term’ temperature responses were characterised as the change in interaction strength between the stream of origin and the transplant stream. By contrast, the ‘long-term’ temperature response was characterised as the change in interaction strength comparing measurements made only when the snails were in their stream of origin. We re-express the transplant temperature data as Boltzmann temperatures so that the coefficients of the model yield activation energies in units of eV (see E. (1)). In this analysis, a significant interaction between ‘transplant temperature’ and ‘timescale’ would demonstrate that the temperature dependence of interaction strength differs between the ‘short-term’ (Eshort), and ‘long-term’ (Elong). We assume that Eshort captures rapid physiological plasticity (e.g. acclimation) in interaction strength in response to a change in temperature and Elong captures both acclimation and adaptation (evolution). Consequently, the component of the temperature sensitivity attributable to evolution is given by Eevol = Elong - Eshort.
RESULTS
Metabolic thermal response curves
The allometric scaling coefficient, α, and the apparent activation energy, Ea, were consistent between the populations of R. balthica from the cold and warm streams (see Table 2 for model comparison and estimated parameter values). The temperature normalised rate of respiration, ln b(Tc), and Th (the temperature at which respiration was 50% inactivated) were both higher in the population of R. balthica from the warm stream. Because the optimum temperature, Topt, depends strongly on Th (see Eq. (2)), Topt was higher in R. balthica from the warmer stream (Topt warm = 38.25 ± 0.6 ºC; Topt cold = 33.05 ± 1.5 ºC). As ln b(Tc) and Topt were both higher, the warm populations of R. balthica had elevated per capita metabolic rates across the full range of measurement temperatures (Fig. 2).
Local adaptation of interaction strength
Interaction strength increased with elevated transplant temperature for the populations of R. balthica from both the warm and the cold streams (Fig. 3; main effect of ‘transplant temperature’ (GLM, t1,21=2.56; p<0.01). Furthermore, interaction strengths were consistently higher for the populations of R. balthica from the warm stream in both transplant temperatures (Fig. 3; GLM main effect of ‘origin’ t1,21 = 2.90; p <0.005). These findings are consistent with the higher respiration rates observed in the warm population (Fig. 2) and highlight the association between metabolism and interaction strength.
Disentangling the effects of acclimation and adaptation on interaction strength
Our experimental design enabled us to compare temperature sensitivities that capture short-term thermal acclimation (e.g. changes in interaction strength in response to the reciprocal transplant) as well as the long-term temperature sensitivity, which also includes effects of local adaptation (e.g. changes in rates between warm and cold populations quantified in the stream of origin). We found that interaction strength increased with temperature in both the short- and the long-term (Fig. 3). However, the magnitude of the temperature response was significantly larger in the long-term (Fig. 3; interaction between ‘transplant temperature’ and ‘timescale’ on interaction strength; GLM t1,18= −2.19; p < 0.05), where, the average Eshort was 0.46 eV, while Elong was significantly higher at 0.99 eV. This divergence between the short- and long-term temperature sensitivities implies a non-trivial contribution of evolution in amplifying the effects of temperature on interaction strength in situ, with the contribution of Eevol of 0.51 and 0.53 eV in the cold and warm adapted populations respectively.
DISCUSSION
Understanding how global warming will affect the strength of consumer-resource interactions and the stability of aquatic food webs is a fundamental challenge in evolutionary ecology that requires insight on the short-term effects of temperature on metabolism and interaction traits as well as how these processes are modulated by evolution over longer time scales. There is evidence from terrestrial (Rall et al. 2010; Vucic-Pestic et al. 2011; Barton 2011; Brose et al. 2012), freshwater (Kratina et al. 2012) and marine ecosystems (Sanford 1999), that warming is likely to increase the strength of consumer-resource interactions, at least in the short-term, owing to the exponential effects of temperature on the consumption rates of mobile ectothermic consumers (Dell et al. 2014; Gilbert et al. 2014). What is less clear however, is how rapid evolutionary adaptation to rising temperatures will modulate the direct effects of warming on species interactions. Space-for-time substitutions across broad spatial scales indicate that local adaptation to different thermal regimes can play an important role in shaping species interactions, often compensating for the direct effects of temperature on interaction traits (Barton 2011; De Block et al. 2012). Here, we build on this work by investigating the effects of temperature and local adaptation on the interaction between the gastropod grazer, R. balthica, and its algal resource. Our study contributes novel insights in a number of ways. First, we explore patterns of local adaptation over a relatively small spatial scale (m as opposed to km). The two streams in our experiment are separated by approximately 500 m but differ in temperature by 11 ºC. Because dispersal, gene flow and genetic divergence among populations in this species are strongly related to geographic distance (Johansson et al. 2016), our study over a relatively small spatial scale, provides insight into how closely related natural populations evolve in response to warming and is therefore directly relevant for understanding adaptation to climate change (Richter-Boix et al. 2010; Keller et al. 2013; Merilä and Hendry 2014). Second, we quantified the effects of temperature on both metabolic and consumption rates to determine the mechanisms underpinning patterns of thermal adaptation and their influence on the strength of consumer-resource interactions.
We found significant variation in the thermal response curves for respiration between the populations of R. balthica from the warm and cold streams. The optimum temperature (Topt) for respiration was higher in the warm population (i.e. metabolic rates peaked at higher temperatures). Furthermore, the inactivation energy (Eh) was lower in the warm population, indicating that declines in the rate of respiration after the optimum (i.e. at high temperatures) were less pronounced than in grazers from the cold stream, where metabolic rates peaked at lower temperatures and declined markedly at temperatures above Topt. These divergences in metabolic traits suggest that the metabolism of the warm and cold populations of R. balthica reflect local adaptation to the different thermal regimes in these streams. Whilst the higher Topt and lower Eh in the warm population were in line with expectations assuming local thermal adaptation, we found no evidence that metabolic performance at high temperature was traded-off against performance at low temperature. Instead, metabolic rates were higher for R. balthica from the warm stream across all measurement temperatures. These results are in broad agreement with the “hotter is better” hypothesis, which proposes that maximal performance of organisms with higher optimal temperatures should be greater than those with lower optimum temperatures because of the thermodynamic constraints imposed by high temperatures on enzyme kinetics (Huey and Kingsolver 1993; Kingsolver et al. 2004; Angilletta et al. 2010). Indeed maximal respiration rates in the population from the warm stream were greater than those from the cool (warm stream: 8.26 ± 0.41 µmol O2 L−1 h-1, cool stream: 7.3 ± 0.22 µmol O2 L-1 h-1). The lower Eh, (i.e. the steepness of the decline of the thermal reaction norm past the optimum), and higher ln b(Tc), i.e. the rate of respiration normalised to 18 ºC, in the warm population also meant that the thermal response curve for R. balthica from the warm stream was broader. In agreement with previous work (e.g. on bacteriophages, Knies et al.(2009)), our data for the gastropod R. balthica indicate that adaptation to higher temperatures resulted in both greater maximal metabolic performance and a broader metabolic thermal reaction norm.
The general patterns observed in the metabolic traits were also reflected in the effects of temperature on interaction strength. Interaction strength was higher for individuals placed in the warm stream, irrespective of their stream of origin. These findings suggest that elevated temperatures increase consumption rates though the effects of temperature on respiratory physiology, but local adaptation to warmer environments also results in a correlated increase in metabolism and interaction strength at low temperature. This may have important wider implications for the effects of warming on the structure, functioning and stability of aquatic food webs (Rall et al. 2010; O’Connor et al. 2011; Vucic-Pestic et al. 2011; Dell et al. 2014; Fussmann et al. 2014; Gilbert et al. 2014, Fussmann et al. 2017). If adaptive responses to increasing temperature give rise to higher maximal rates of metabolism and consumption as well as elevating rates at lower temperatures, then the effects of warming on the strength of consumer-resource interactions in the long-term could be greater than previously anticipated (Gilbert et al. 2014). Indeed, work on experimental warming of aquatic ecosystems has shown that increases in the strength of top-down control can have profound effects on community structure and ecosystem processes (Burgmer and Hillebrand 2011; Kratina et al. 2012; Yvon-Durocher et al. 2015). Elevated grazing rates at warmer temperatures can have a wide range of impacts in aquatic systems, with evidence for both increases (Yvon-Durocher et al. 2015) and decreases (Burgmer and Hillebrand 2011) in algal species richness, biomass and productivity.
In our experiments, the thermal sensitivities of metabolic rates were much larger than those of interaction strengths in the short-term (e.g. 0.96 and 0.45 eV respectively), in line with findings in other invertebrate systems (Rall et al. 2010; Vucic-Pestic et al. 2011; Fussmann et al. 2014). These findings suggest that rates of grazing and metabolism were clearly linked, but became decoupled when individuals experience rapid changes in temperature that depart substantially from those in their local environment. In the short-term, if increases in metabolic demands with temperature are greater than those of consumption rates (as found here), then less energy will be transferred from the resource to the consumer (i.e. more is lost through respiration, see also Rall et al. 2010). If such imbalances are maintained over long periods of time then starvation of the consumer can ultimately result in a decline in top-down control on the resource (Fussmann et al. 2014, Binzer et al. 2016). However, when consumers’ feeding rates are more sensitive to temperature than metabolic rates, interaction strengths can become amplified in warmer environemnts, leading to faster resource depletion and eventually driving either the resource or the consumer to extinction (Vasseur &McCann 2005). Long-term effects of temperature on interaction strengths have so far only been explored using food web models, parameterised using temperature sensitivities derived from short-term experiments (Vasseur &McCann 2005; Rall et al. 2010; Fussmann et al. 2014). Consequently, such analyses don’t capture the capacity for thermal adaptation to modulate per capita rates. Our results highlight substantial differences between the short- and long-term effects of temperature on interaction strength; implying that thermal adaptation plays an important role in maintaining the balance between metabolic and consumption rates over the long-term.
We quantified the effects of local adaptation (evolution) on interaction strength by comparing the short-and long-term effects of temperature in the reciprocal transplant experiment. The short-term temperature response (Eshort) captures the effects of physiological plasticity over the 48h experiment. Conversely, the long-term response (Elong) also accounts for processes operating over longer, evolutionary timescales. The Elong value was higher than Eshort, implying a significant role for evolution in shaping the effects of temperature on in situ interaction strengths. Notably, the higher Elong was driven both by elevated grazing rates in the warm populations in the warm stream and lower rates in the cold populations in the cold stream. These results diverge from expectations based on the metabolic cold adaptation hypothesis (Addo-Bediako et al. 2002) which would predict adaptation to higher temperatures should dampen the acute effects of temperature on metabolic rates. On the contrary, our results suggest that adaptation to warming amplified the effects of temperature on metabolic as well as grazing rates. The lower interaction strengths in the population of R. balthica locally adapted to the colder stream highlight that evolution can have unexpected effects on species interactions. The evolutionary maintenance of lower than anticipated grazing rates in the cold stream could be selected for since lower grazing rates might result in greater food chain stability and/or stoichiometric homeostasis (Sterner & Elser 2002, Cross et al. 2005; 2015) under the prevailing temperature regime. Thus, understanding the impacts of environmental change on the strength of consumer-resource interactions over timescales that are relevant to the rate of climate change (e.g. gradual warming over decades) will require an appreciation both of the direct effects of rising temperatures on species interactions and the reciprocal feedback between ecological and evolutionary dynamics (Fussmann et al. 2007; Gravel et al. 2010; Loeuille 2010; Urban 2013; Barraclough 2015)
Conclusions
We used a natural geothermal temperature gradient to investigate how warming influences the strength of algal-grazer interactions via the direct effects of temperature on metabolism and consumption, and indirect feedbacks through evolutionary adaptation. Metabolic rates and interaction strength increased with temperature in the same way for both the warm- and cold-adapted populations of R. balthica, suggesting that rapid changes in temperature have a consistent effect on interactions between mobile consumers and sessile resources, mediated by the effects of temperature on consumer metabolic rates (Dell et al. 2014). However, the warm-adapted populations had higher metabolic and grazing rates across all measurement temperatures compared to their cold-adapted counterparts. These findings are consistent with the ‘hotter is better and broader’ hypothesis (Huey and Kingsolver 1993; Knies et al. 2009; Angilletta et al. 2010) (e.g. adaptation to warming gives rise to higher maximal metabolic rates and broader thermal reaction norms). In consequence, our results suggest that warming could increase the strength of algal-grazer interactions, which are often ‘keystone’ interactions in aquatic systems, both via the thermodynamic effects of higher temperatures on enzyme kinetics and through correlated increases in per capita metabolism and consumption as organisms adapt to warmer temperatures.
Conflict of interest
The authors declare no conflict of interest
Acknowledgments
The authors thank Eoin O’Gorman for comments on an earlier version of this manuscript. This study was funded by a Leverhulme Trust research grant (RPG-2013-335), and an ERC starting grant (ERC-StG 677278) awarded to GYD; and the University of Exeter.
Footnotes
Data accessibility statement: All data will be made available as supporting information should the manuscript be accepted. R-Code for analysis will be available from the authors on request.