How to quantify thermal acclimation capacity?

Abstract

An experienced change in environmental temperature may result in a phenotypic trait value to deviate from optimality. This may be caused by an acute phenotypic change (i.e. passive phenotypic plasticity), and/or a shift in the optimal value for that trait. This in turn triggers an acclimation response that may last from days to months, whereby the organisms attempt to regain optimality of phenotype. Their capacity to acclimate will influence their ability to cope with ongoing global changes in thermal regimes (Stillman, 2003). To gain insights into the sources of variation in acclimation capacity Rohr et al. (2018) reanalyzed the data of Seebacher, White, and Franklin (2015). However, we believe that their approach introduces two problems: 1. Data analyzed by Rohr et al. originate from studies that were primarily (322 of 333 cases) conducted by measuring traits (mostly physiological/biochemical) in 2×2 factorial design experiments, with two acclimation temperatures and two measurement temperatures (Figure 1a). Yet, they only use the data obtained when measuring the traits at the temperature that the individuals were acclimated to ("post-acclimation response"), and define a high acclimation capacity as a flat post-acclimation response, independent of changes occurring during acclimation (Figure 1a,b). This may introduce a bias, because it requires more pronounced changes in the quantity and quality of cellular biochemistry and structures to obtain a flat post-acclimation response for traits that show a steep acute response (Figure 1a vs. 1b). Evidence of such a bias is revealed by a strong negative correlation between the index of acclimation capacity used by Rohr et al. and the estimated mean acute response measured in the same studies (i.e. high mean acute responses are associated with low acclimation capacity values, Figure 2, R = −0.55, df = 320, p < 0.001; cases selected as below). Therefore, variation in acclimation capacity as calculated by Rohr et al. is largely driven by variation in the acute response. 2. The approach used by Rohr et al. follows the implicit assumption of Seebacher et al. (2015) that all acclimation responses cause a reduced thermal response in a trait post-acclimation compared to the acute response (Figure 1a,b). We contend that any measure of acclimation capacity should acknowledge that acclimation may also result in organisms showing an increased trait response after acclimation is complete (Figure 1c). This will occur if the optimal value of a trait increases with temperature, and animals need time to produce this altered phenotype. For example, when the zooplankter Daphnia magna is exposed to a high temperature they gradually (over ca. 5 days) increase their hemoglobin concentration to allow for oxygen supply to match demand (Seidl, Pirow, & Paul, 2005). This allows them to maintain fitness across temperatures by increasing the thermal response of a trait through acclimation. To evaluate the prevalence of this type of acclimation, we considered the data subset of Seebacher et al. an...

Distinguishing between active plasticity due to thermal acclimation and passive plasticity due toQ10effects: Why methodology matters

“…Complete and over‐compensation may be rare because most animals may not be able to completely compensate for changes due to passive plasticity. It is unclear at this time what evolutionary mechanisms drive these different responses, but our results highlight recent criticisms of assuming one particular response is common across diverse organisms and experiments (Einum et al, ).…”

“…Although the diversity of acclimation responses can be complex (Figure ), there is renewed interest in quantifying thermal acclimation and understanding the selection pressures that generate variation in acclimation capacity. Such studies are increasingly being used to predict responses to global climate change and investigate environmental adaptation (Deutsch et al, ; Duarte, ; Einum et al, ; Gunderson & Stillman, ; Kingsolver, ; Pörtner & Knust, ; Rohr et al, ; Seebacher et al, ; Wythers, Reich, & Bradford, ). For example, the climate variability hypothesis suggests that animals living in thermally variable high latitudes should have broader thermal niches and exhibit stronger thermal acclimation responses than those from the less variable tropics (Angilletta, Condon, & Youngblood, ; Compton, Rijkenberg, Drent, & Piersma, ; Ghalambor, Huey, Martin, Tewksbury, & Wang, ; Janzen, ; Shah et al, ).…”

“…A fundamental challenge is that acute exposure causes a passive Q 10 response, while continued exposure might result in an entirely different Q 10 due to active acclimation effects. Although relatively rare, others have also differentiated between acute Q 10 effects (passive plasticity) and thermal acclimation (active plasticity), which is paramount to predicting the effects of climate change (see Cossins & Bowler, ; Einum et al, ; Payne & Smith, ; Schulte et al, ; Seebacher, White, & Franklin, ).…”

“…only a strictly passive Q 10 effect is observed); (b) Partial compensation —the passive rate change is counteracted by acclimation and the acclimated rate approaches the original rate; (c) complete compensation— the acclimated rate returns to the original rate, neutralizing the passive response (referred to as ‘metabolic homeostasis’ by Cossins & Bowler, ); (d) over‐compensation —the acclimated rate ‘overshoots’ the original rate, more than compensating for the passive response; and (e) inverse compensation —the passive rate change is amplified with acclimation. Indeed, Einum et al () recently criticized studies of thermal acclimation for not distinguishing among these acclimation responses.…”

Different metrics of thermal acclimation yield similar effects of latitude, acclimation duration, and body mass on acclimation capacities

“…Our analyses demonstrated consistency in the effects of body size, latitude, and methodological factors on thermal plasticity across these datasets. Einum et al () argue that a metric we used to assess thermal acclimation responses in one of four datasets (Rohr et al, ) was less ideal than one that they propose.…”