Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
1. The ‘temperature-size rule’ (TSR) is a widely observed phenomenon within ectothermic species: individuals reared at lower temperatures grow more slowly, but are larger as adults than individuals reared at warmer temperatures. Although the TSR is common and of widespread ecological importance, it is not known whether there is a general physiological mechanism causing the TSR or even if species share a similar pattern of thermal response across ontogeny.
2. We constructed a conceptual model to show that binary division forces growth (g) and development (D) rates to return to a fixed ratio in unicellular organisms exposed to a change in temperature. After a period of decoupling during thermal acclimation, these rates must be restored to maintain a fixed ratio of adult:progeny size. However, the relationship between adult and progeny size need not be fixed in multicellular organisms at different temperatures, and hence growth and development rates need not have a fixed ratio either.
3. We conducted a meta-analysis on data of metazoan ontogenetic responses to temperature which demonstrates that adult size shows a much stronger temperature–size response than progeny size, and reveals variation in size response among other life cycle phases.
4. This study shows fundamental differences in the operation of the TSR in unicellular and multicellular organisms, suggesting that a general physiological mechanism causing the TSR is unlikely. Our findings also reveal the value of analysing shifts in size through the life cycle and across generations: these will yield a more complete quantitative description of how, and potentially provide clues to why, body size responds to temperature.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Body size is fundamental to the functioning of all organisms, impacting on all aspects of life including growth, reproduction and mortality (Kingsolver & Huey 2008). Therefore, understanding what drives species body size is a critical aspect of ecology. One widespread pattern of body size in ectothermic organisms is the ‘temperature–size rule’ (TSR). The TSR refers to how, within a species, lower rearing temperatures leads to increased size at a given developmental stage (Atkinson 1994). This phenotypically plastic response has been found to occur in the majority of ectotherms (83% of those examined) including a bacterium, protists, plants and animal groups including molluscs, arthropods, amphibians and fish (Atkinson 1994). Furthermore, a meta-analysis of protist data (Atkinson, Ciotti & Montagnes 2003) found that for each 1 °C increase in temperature, cell volume decreased by 2.5% of their volume at 15 °C. Changes in size have been described as the ‘third universal ecological response to global warming’ (Daufresne, Lengfellner & Sommer 2009), alongside shifts in species’ range and changes in phenology. A feature of current climate change is the predicted increase in frequency and intensity of heat-waves (IPCC, 2007); therefore, understanding how organisms will respond to increasing temperature in the short-term and longer term and the mechanisms underpinning these responses is critical.
Growing to a smaller final size at warmer temperatures seems counterintuitive and has been termed a ‘life-history puzzle’ (Sevenster 1995). One might expect that as organisms have faster growth rates at higher temperatures, they should delay maturation to exploit the increase in fecundity, survival and mating success associated with larger size (Sibly & Atkinson 1994; Kingsolver & Huey 2008). Indeed, increased growth rate associated with improved food conditions results in larger adults, whereas the increased growth rate associated with higher temperature results in reduced adult size (Kindlmann, Dixon & Dostalkova 2001). Attempts to explain why temperature differences result in body size changes often consider the problem with respect to the maximization of fitness (population growth rate, r, and offspring production, R0), and include the interplay of multiple traits such as growth, fecundity, development and mortality (Sibly & Atkinson 1994; Kozłowski, Czarnoleski & Danko 2004; Kiørboe & Hirst 2008). Rather than focusing on why the TSR occurs, others have instead focused on the question of how body size changes. For example, Davidowitz & Nijhout (2004) formulated a physiological (endocrine-based) model for holometabolous insects. However, we still lack a general model to explain the TSR (Angilletta, Sears & Steury 2004): such a general model would account for differences across taxa, changes in size during ontogeny and changes in size across generations. Here, we will explore how critical differences between methods of reproduction, and in growth and developmental responses to temperature between different ontogenetic stages, affect attempts to derive a universal mechanistic TSR model.
How does body size change?
The TSR indicates that when juveniles grow in cooler environments, they develop into larger adults (Atkinson 1994); consequently, although the rate of development (passing through life stages) and growth (accumulation of mass) from embryo to adult decreases with decreasing temperature, there must be a relatively larger decrease in the development rate. Although this seems obvious, previous general models of how size changes with temperature have often not explicitly indicated that these two rates are decoupled [e.g. explanations based on the von Bertalanffy growth equation (Von Bertalanffy 1957; Perrin 1995)].
Van der Have & de Jong (1996) argued that the TSR must be a result of mismatch in the temperature dependence of growth rate and development rate (which they use synonymously with differentiation rate). They then built a biophysical model, assuming development and growth rates to have independent thermal reaction norms under non-limiting food conditions. van der Have & de Jong (1996) state, ‘As a proximate model, the biophysical model applies to all ectotherms, including protists in which ‘differentiation’ consists only of cell divisions.’van der Have & de Jong (1996) also suggest that progeny size may be impacted by different rates of differentiation and growth of oocytes (e.g. vitellogenin synthesis in insects; (Ernsting & Isaaks 1997)). From this assumption, they argued that the same effect of temperature on oocyte production as on the size at metamorphosis could be predicted; that is, larger eggs will be produced at lower environmental temperature.
Van der Have & de Jong (1996) made an important advance in analysing the TSR by explicitly treating growth and development as separate rates. However, any general mechanistic model needs to explain fundamental differences in the establishment of the TSR in different organisms. In this study, we show how reproduction by binary division (i.e. cell dividing into two equally sized progeny) results in fundamental differences in the operation of the TSR between unicellular and multicellular organisms. Secondly, we perform a meta-analysis to measure the effects of temperature on adult vs. progeny size in metazoans, and show that progeny size in multicellular organisms does not follow the same response as adult size. Thirdly, we explore size responses to temperature across ontogeny in multicellular organisms, to examine whether these organisms exhibit systematic changes in size throughout the whole life cycle or whether size responds mostly during specific stages after which these changes are maintained. Analysis of shifts in size throughout the life cycle and across generations will yield a more complete quantitative description of how, and potentially provide clues to why, body size responds to temperature.
We specifically addressed the following questions:
1 How do constraints of a unicellular vs. multicellular life cycle affect the adjustments of size and therefore growth and development associated with the TSR?
2 Does the TSR affect adult and progeny mass of multicellular organisms equally?
3 Does the TSR have consistent effects throughout ontogeny?
The conceptual model
We constructed a model building on the linear equation used by van der Have & de Jong (1996) to link growth and development with adult and progeny size:
Where g = mean juvenile growth rate (day−1), t = development time (days, e.g. egg hatch to maturation, or time between subsequent divisions in unicellular organisms), MA = mass of adult, and MP = mass of a single progeny. We use the term ‘progeny’ to refer to young at the point of inception. This is the daughter cell just after binary division of the mother in a unicellular organism, or the newly produced egg, or the propagule at the point of budding in a multicellular organism. Using the inverse of development time t in eqn 1 converts this parameter to a mean rate of development, D (D = 1/t). Individual growth from progeny to adult can then be expressed as:
Thus, examining the ratio of adult to progeny mass provides a straightforward way to determine the effect of temperature on two fundamental biological rates (growth and development rate) and to test van der Have & de Jong (1996) hypothesis: that development and growth rates have independent thermal reaction norms under non-limiting food conditions. We consider the implications of this growth equation to the TSR in unicellular and multicellular organisms.
Most unicellular organisms reproduce by binary division (Adolph 1931), a term we use to encapsulate binary fission in prokaryotes, and mitosis in unicellular eukaryotes. In binary division, an ‘adult’ cell (of mass MA) divides into two ‘daughter’ cells (of mass MP), each with a mass half that of the adult, i.e. MA = 2MP. Thus, at a fixed temperature across generations, and with other conditions constant, MA/MP = 2, eqn 2 then simplifies to:
At a fixed temperature, unicellular organisms must have a fixed ratio of growth to development rate, thus, referring to Fig. 1, g/D(cold) = g/D(warm) = 2. This is in clear disagreement with the assumption of the van der Have and de Jong model, that development and growth rates have independent thermal reaction norms under nonlimiting food conditions. In fact, binary division imposes strict limits on adult and progeny size ratio and forces g/D to return to a fixed ratio of 2. However, most unicellular organisms obey the TSR, becoming larger at cooler temperatures and smaller at warmer temperatures (Atkinson 1994; Montagnes & Franklin 2001; Atkinson, Ciotti & Montagnes 2003). Therefore, within generations, the size of unicells must change when exposed to a new temperature and g/D must become temporarily decoupled (see Fig. 1). After g/D adjustment is complete, the rates must become coupled once more; these rates cannot be considered independent as binary division requires that total temperature compensation occurs (i.e. Eqn 3 is restored) to prevent cells continuing to get smaller or larger ad infinitum. These conclusions are not qualitatively affected by altering the growth function from linear increase in mass per unit time, to exponential or von Bertalanffy (see Appendix S1 in Supporting Information).
Application of eqn 2 is more complex for multicellular organisms. As they do not replicate by simple binary division of the adult, the progeny mass is not restricted to be a fixed proportion of adult mass and in many species, individual organisms are able to produce progeny that can vary in size (Blanckenhorn 2000; Atkinson et al. 2001; Fischer et al. 2003; Fischer, Zwaan & Brakefield 2004). Therefore, unlike unicells, individual progeny are not so strictly constrained by maternal size, thus the ratio MA/MP need not be fixed across different temperatures and consequently growth and development rates would not need to return to a fixed ratio (eqn 2). If this were the case, growth and development rates could change independently with temperature, which supports the assumption of the biophysical model applied by van der Have & de Jong (1996). There is much evidence confirming a temperature–size response in adults (see review in Atkinson 1994), but less evidence for eggs (see review in Atkinson et al. 2001). We show this potential temperature independence of g and D in Fig. 2 in which adult mass is assumed to change with temperature more than progeny mass, thus MA/MP(cold) > MA/MP(warm). If adult and progeny mass show different temperature–size responses in metazoans, this would suggest a fundamental difference between the TSR in unicellular and multicellular organisms: unicellular organisms living at a fixed temperature must have a constant ratio of g/D, whereas multicellular organisms have a variable ratio of g/D.
Is this supported by experimental data in the literature? We conducted a meta-analysis on metazoan adult and progeny size data for a wide range of species, and tested whether the thermal responses of these data sets differ. Where data were available, changes in mass were also examined separately throughout ontogeny, described in the Methods below.
Materials and methods
Detailed accounts of methods are provided in Supporting Information (Appendix S2). We collected published data to assess the extent to which mass of both adults and progeny of ectotherms varies with temperature (after allowing a period for size acclimation). We only include data where individuals were grown at constant temperatures and at food conditions believed to be saturated. We designated a minimum acclimation period for both progeny and adult data (Appendix S2) to ensure that sizes were acclimated to the temperature at which they were recorded. Initially, we included data when presented as masses for either progeny, adults or both from single studies (Tables S1 and S2 respectively in Supporting Information). To test for differences in the response of adult and progeny size to temperature more rigorously, and act as direct comparison with size-acclimated unicellular organisms where MA/MP is fixed, we analysed a sub-set of this entire data set, specifically those data in which adult and progeny sizes are described on single species by the same study group: we term this ‘paired data’ (Appendix S2 and Table S3). We also consider the period of acclimation more closely in this set (Appendix S2).
An appropriate model for the response of adult and progeny size to temperature was required which could be applied across all species. There is conflicting opinion as to the form that the body mass thermal reaction norm should take within a species, and many different equation forms have been proposed (Karan et al. 1998; Atkinson, Ciotti & Montagnes 2003; de Jong 2010). We therefore applied a range of equation forms (linear, exponential, Arrhenius and allometric) to the full data set to determine which best described the empirical data, using a linear mixed effects model (see Appendix S3 for details and in-depth results). An information theoretic approach was used to determine which model best fit the adult and progeny data; Akaike weights were used, which determine the best fit while accounting for the complexity of the model. We next analysed both the full data set and the paired data to examine if the slopes of the temperature–size response in adults and progeny were significantly different, comparing the mean slopes for the full data set and conducting a paired t-test on individual species in the paired data set.
Beyond examining progeny and adult masses, an appreciation of where in the development schedule changes in the mass to temperature relationship occur in metazoans will provide insight into the causes of these changes. Within the data, two studies had individual masses and times for multiple larval stages between egg and adult (including prior acclimation), which allowed us to examine how the response of mass to temperature varies throughout ontogeny. These were both for copepods, Acartia tonsa (Leandro, Tiselius & Queiroga 2006), and Calanus finmarchicus (Campbell et al. 2001), and included egg, six nauplii stages and six copepodite stages, the final stage being the adult. To determine the mass vs. temperature relationship for each stage, the best-fit model type, as shown from our analysis of all progeny and adult data, was applied to each species individually (see Appendix S3).
Adult to progeny size ratios of multicellular organisms
For the larger unpaired dataset, we collected progeny data for 33 different species (Table S1) and adult data for 85 different species (Table S2). Within this larger set, there were adult and progeny paired data for 15 (sub)species which fulfilled the more rigorous requirements (Table S3). We found that an exponential model with species-specific intercepts and slopes provided the best fit to both the adult and progeny mass vs. temperature responses (see Appendix S2), with the basic form:
where M = mass, T = temperature, a is the mean intercept and b is the mean slope term.
According to the fitted slopes for this best fit model, adult mass had a significantly more negative slope (b = −2.60 × 10−2, 95% CIs = ±0.57 × 10−2) than progeny mass (b = −0.90 × 10−2, 95% CIs = ±0.61 × 10−2) across the entire data set (t-test, t = 6.19, P < 0.001, Fig. 3, Appendix S3). This is equivalent to a 0.9% decrease in mass °C−1 in progeny, but a 2.5% decrease in mass °C−1 in adults, with the magnitude of size change in adults being similar to that seen in protists (Atkinson, Ciotti & Montagnes 2003). Similar results were found for paired data; the mean slope for progeny mass vs. temperature was −0.14 × 10−2 (95% CIs = ±0.71 × 10−2, Figs 3 and 4a), but was −2.32 × 10−2 for their adult mass (95% CIs = ±1.21 × 10−2, Figs 3 and 4b). The ratio of MA/MP was calculated for the paired data using species-specific slope terms (Fig. 4c). The species-specific slope parameters for the paired data were used to test whether these slopes were significantly different, by conducting a pair-wise t-test for adult vs. progeny slopes. This indicated that adult mass has a stronger temperature dependence than progeny mass, (paired t-test, t = 4.34, P = 0.001), and consequently the ratio of MA/MP is not fixed within single species at different temperatures. Comparison of the slopes of ln mass vs. temperature for the paired data shows that the mean slope for progeny mass is not significantly different from zero (mean = −0.14 × 10−2, 95% CIs = −0.86 × 10−2, 0.57 × 10−2), whereas the mean slope for adult mass is significantly negative (mean = −2.32 × 10−2, 95% CIs = −3.53 × 10−2, −1.10 × 10−2). Thus, after allowing time for size acclimation, the ratio MA/MP does not return to a fixed temperature-independent constant, but this ratio is generally larger at low temperatures and smaller at high temperatures, as shown in Fig. 2. These changes in MA/MP were extremely large in some cases; the ratio of MA/MP at the highest experimental temperature was half of that at the lowest experimental temperature in Pseudocalanus newmani (Fig. 4c). Further, more than half of the species in the paired data (8/15) showed changes in the ratio of MA/MP of >30% over their thermal range (Fig. 4c). A consequence of this is that multicellular organisms must have a g/D ratio, which varies substantially across temperatures, i.e. growth and development rates have a different temperature dependence. In contrast, in unicells reproducing by binary division, we know that g/D must be fixed (eqn 3).
Timing of size adjustment during ontogeny
The effect of temperature on the size of specific larval stages in the copepod species Acartia tonsa and Calanus finmarchicus shows some variation between these two species (Fig. 5); however, there are general patterns in these responses. There is no discernible effect of temperature on size of progeny (represented by early larval stages) in either species. The mass vs. temperature relationships exhibit a generally increasing negative trend throughout ontogeny when examined in relation to time (Fig. 5a, b). When examined with respect to mass, the majority of the temperature-dependence of size has been completed by approximately 20% of the adult mass (Fig. 5c, d). The majority of the temperature–size effect has been completed before the last 3–4 larval stages, despite these stages accounting for the majority of mass accrual (approximately 80% of total mass) due to the exponential nature of mass accrual with time exhibited in copepod species (Escribano & Mclaren 1992).
Using a conceptual model, we have shown that unicellular organisms acclimated to different temperatures must have a ratio of growth to development rate which is a constant; this is due to the constraints of binary division. When a unicellular organism that follows the TSR is exposed to a new thermal environment, any decoupling of growth and development rate is constrained within a period of acclimation, and g/D must return to a fixed value of 2. If g/D did not return to this value, at increased temperature, cells would get progressively smaller with each division. Although this specifically only applies to those organisms which divide by binary division, this is the major reproductive strategy in prokaryotes (Angert 2005), and in many unicellular eukaryotic cells (Sleigh 1991; Reynolds 2006). Therefore, the thermal sensitivity of growth and development rates are not independent in the majority of unicellular organisms.
Our synthesis of adult and progeny mass in multicellular organisms shows that the ratio of these is not constant at different temperatures. Combining this evidence with our conceptual model shows that unlike in unicells, multicellular organisms can maintain different temperature dependence for growth rate relative to development rate. Our meta-analysis demonstrates that progeny mass shows a reduced response to temperature compared with adult mass. Although both show a negative response, adult mass has a significantly more negative slope than that of progeny. Furthermore, when the paired data were compared, the ratio of MA/MP was consistently negative across the 15 (sub)species, with the average progeny temperature–size response not being significantly different from zero. This novel finding suggests that it is incorrect to assume a temperature–size effect on progeny size in multicellular organisms that is of similar magnitude to adults. van der Have & de Jong (1996) proposed that ‘the same effect of temperature on oocyte production as on size at metamorphosis could be predicted, that is, smaller eggs will be produced at higher environmental temperature’. This statement requires clarification: we find that the magnitude of this change is consistently larger in adults than in progeny. This is the case even after allowing for acclimation of both adult and progeny size. Referring to eqn 2, this means that growth and development rates have a different temperature dependence in multicellular organisms, with development being more temperature-sensitive than growth.
There is evidence from two copepod species that early larval stages (i.e. beyond egg stage) show no size response to temperature, whereas later stages show strong negative relationships. The data suggest that thermal selective pressures act increasingly during the maturation of the two copepods (Fig. 5) and the unequal effect of temperature on growth and development rates only begins acting on size during post-embryonic growth. Although temperature–size effects are cumulative during ontogeny, the majority of the temperature–size response is established by the point at which approximately 0.2 of the adult weight has been achieved. This is in contrast to larval development in the butterfly Lycaena tityrus, where the TSR is only established during the final larval stage associated with the largest (approximately 80%) increase in mass (Karl & Fischer 2008). Furthermore, ontogenetic size changes in Fig. 5 indicate that these copepod species may be seen as adjusting size in every generation: changes in size are effectively being reset or considerably muted at egg/progeny stage. This, again, is not the case in butterfly species, which show marked changes in egg size at different temperatures (Fischer et al. 2003; Fischer, Bauerfeind & Fiedler 2006). This suggests that although the TSR applies to the majority of metazoa (Atkinson 1994), there may be taxon-specific changes in size with temperature, which impact on different life stages to different extents.
How do these differences between unicellular and multicellular organisms impact on the potential causes of the TSR? The results of our conceptual model, combined with the meta-analysis reveal that unicellular organisms are restricted in the adjustment of their rates of growth and development. The ratio of g/D must return to a constant of two in a species living at a fixed temperature, thus any temperature-induced changes in this ratio is limited to a temporary acclimation phase. In multicellular organisms, the ratio of g/D need never be a fixed constant when comparing across different temperatures because these organisms alter their adult:progeny size ratio (see eqn 2 and Fig. 4c). Although there must be limits imposed on size changes set by physical constraints, such as maternal ovipositor/birth canal diameter (Atkinson et al. 2001), this does not impose strict limits on the ratio of MA/MP, and therefore g/D, as it does in unicells. Indeed, the results from the paired meta-analysis show the ratio of MA/MP can change substantially over a species’ thermal range in multicellular organisms. Consequently, there can be large alterations in the ratio of g/D (eqn 2). For example, MA/MP data for Pseudocalanus newmani show that development rate must increase by more than twice the rate of growth over this copepod’s total thermal range.
It is important to note that despite the different restrictions imposed by reproductive method in unicellular and multicellular organisms, both follow the TSR. Therefore, despite the restrictions imposed by binary division, rates of g/D must temporarily decouple to facilitate size change, even if they must eventually return to a fixed ratio. This suggests that there must be significant fitness benefits to this thermal plasticity, as it occurs in different groups through different means. Thus, although the proximate mechanism for the TSR differs between these two groups, the ultimate explanation for the TSR may still be the same. Despite many hypotheses having been proposed (Angilletta, Sears & Steury 2004; Atkinson, Morley & Hughes 2006; Walters & Hassall 2006; Kingsolver & Huey 2008), we are yet to find a general, ultimate cause for the phenomenon of the TSR. To understand the variation in size responses to temperature, we propose that more attention be directed to fuller quantitative descriptions of responses throughout the period of population growth in unicells, and ontogeny in multicellular organisms. In unicells, for example, by identifying the number of cell generations until g/D adjustment is complete and the amount of g/D adjustment per cell cycle per °C, we can partition variation in size responses among species to the different mechanisms (average thermal sensitivity of size per cell cycle per °C, number of cell divisions to complete acclimation), and seek patterns in these among taxa and ecological niches. Likewise, in multicellular organisms, differences between species in the period of g/D adjustment, as shown in Fig. 5 for two species of copepod, can help identify variation, or indeed similarities, between species and taxa. Another potential benefit from quantifying trends in TSR across ontogeny is to identify particular stages or size ranges when selection for size response to temperature may be particularly intense. Berven & Gill (1983) suggest that temperature-dependent variation in adult size in Rana sylvatica may be a consequence/correlate of temperature-dependent selection on offspring size plasticity. By quantifying which developmental phases actually show a size response to temperature (Fig. 5), particular parts of the life cycle may be examined to see whether or not there are particular temperature-dependent selection pressures that affect those developmental phases or size classes.
Any universal mechanism explaining the TSR must be applicable to all ectothermic groups. We have shown that fundamental differences exist between unicellular and multicellular organisms in the way size changes are brought about. This suggests that there is no universal physiological mechanism to explain the TSR.