Articleshttps://doi.org/10.1038/s41559-020-1114-9

1Centre for Geometric Biology, Monash University, Melbourne, Victoria, Australia. 2Lund University, Lund, Sweden. 3Queensland University of Technology, Brisbane, Queensland, Australia. *e-mail: dustin.marshall@monash.edu

Development is energetically costly1,2 and selection favours minimizing the costs of development for both parents and offspring. For offspring, minimizing development costs

should maximize the resources available for the nutritionally inde-pendent phase, increasing fitness3,4. For parents, less costly devel-opment minimizes the per capita costs of producing nutritionally independent offspring, thereby allowing for increased fecundity5. The costs of development from a zygote to an independently feed-ing individual can be formally represented as:

C ¼ MR=DR ð1Þ

Anything that decreases the ratio of metabolic rate (MR) to developmental rate (DR) will decrease the costs of development (C) overall3–6. A more intuitive way of presenting this equation is to replace development rate with its inverse, development time D.

C ¼ MR ´D ð2Þ

So, decreasing metabolic rate, developmental time or both will decrease the costs of development.

A key driver of variation in both metabolic rate and develop-mental time is temperature3–5,7. Higher temperatures increase meta-bolic rate (up to a point) and decrease developmental time (again, up to a point). If temperature affected both metabolic rate and developmental rate in the exact same way (that is, they had identical temperature dependencies), then the costs of development would be unaffected by temperature—an increase in one would perfectly counterbalance an increase in the other. However, a recent meta-analysis revealed that the temperature dependencies of metabolic rate and development rate are different3. These differential sensitivi-ties mean that the total cost of development (as in equation (2)) is determined by two functions of temperature.

Within the usual temperature range of an organism, the relation-ship between metabolic rate during development, MR(T), and tem-perature, T, can be described by the function:

MR Tð Þ ¼ δeaT ð3Þ

where δ is a constant and a is the coefficient describing the temperature dependency of metabolic rate. Similarly, the relation-ship between developmental time, D(T), and temperature can be described by the function:

D Tð Þ ¼ de�bT þ c ð4Þ

where b, c and d are parameters that describe the exponential decline of developmental time with temperature (Fig. 1; for a discussion of these parameters see Supplementary Information). The product of equations (3) and (4), using typical parameter values, yields a uni-modal function with the minimum representing the temperature that minimizes the costs of development (Fig. 1)—we can consider this temperature the optimal temperature for development in terms of minimizing energy consumption (Topt). At temperatures cooler than Topt, development is more costly because development time is increased more than metabolic rate is decreased. At temperatures warmer than Topt, development is more costly because metabolic rate increases more than development time decreases. Note that this function considers only the energy required to complete nor-mal development and so only applies within an organism’s natural thermal range; it does not consider survival or any other fitness proxy. Temperatures that are beyond this thermal range will cause high mortality, which we do not consider. Instead, we explore the effects of temperature variation within the range of temperatures where survival is relatively high3 (we discuss this issue further in the Supplementary Information).

Developmental cost theory predicts thermal environment and vulnerability to global warmingDustin J. Marshall   1*, Amanda K. Pettersen   2, Michael Bode1,3 and Craig R. White   1

Metazoans must develop from zygotes to feeding organisms. In doing so, developing offspring consume up to 60% of the energy provided by their parent. The cost of development depends on two rates: metabolic rate, which determines the rate that energy is used; and developmental rate, which determines the length of the developmental period. Both development and metabolism are highly temperature-dependent such that developmental costs should be sensitive to the local thermal environ-ment. Here, we develop, parameterize and test developmental cost theory, a physiologically explicit theory that reveals that ectotherms have narrow thermal windows in which developmental costs are minimized (Topt). Our developmental cost theory-derived estimates of Topt predict the natural thermal environment of 71 species across seven phyla remarkably well (R2 ~0.83). Developmental cost theory predicts that costs of development are much more sensitive to small changes in temperature than classic measures such as survival. Warming-driven changes to developmental costs are predicted to strongly affect popula-tion replenishment and developmental cost theory provides a mechanistic foundation for determining which species are most at risk. Developmental cost theory predicts that tropical aquatic species and most non-nesting terrestrial species are likely to incur the greatest increase in developmental costs from future warming.

NAtuRe eCology & evolutioN | VOL 4 | MArCh 2020 | 406–411 | www.nature.com/natecolevol406

ArticlesNaTure ecOlOgy & evOluTION

Figure 2a,b shows the temperature dependencies of (relative) developmental costs for three very different species, herring (Clupea harengus), krill (Euphausia superba) and a crocodile (Crocodylus johnstoni). While both the curvature of the relationship and the position of Topt vary, a single temperature for each minimizes the cost of development. For some species, the cost of development can vary more than tenfold, depending on how far from Topt an environ-mental temperature may be.

We can formally calculate Topt for a species as the temperature at which the first derivative of the product of the equations (3) and (4) is equal to zero:

Topt¼1bln

d b� að Þac

� �ð5Þ

Equation (5) shows that all four parameters affect Topt but the temperature dependency of development time (parameter b) has the relatively strongest effect on Topt because it sits outside the logged part of the equation. Using equation (5) we can calculate the Topt for species for which we have data (Fig. 2c shows the estimated value of Topt for 71 species).

We used these data to examine how Topt covaries with environ-mental temperature in nature to test the hypothesis that the tem-perature dependencies of metabolic rate and developmental time have evolved to minimize the cost of development. We then exam-ined how estimated values of Topt differ among species that develop in aquatic and terrestrial environments (recent studies suggest that these two systems differ systematically in their sensitivity to tem-perature8). We found that the cost of development is very sensitive to deviations from optimal temperature, and that, for most species, there is remarkable congruence between environmental temperature

and Topt. We then examined whether the developmental costs are more sensitive to temperature deviations than traditional measures of thermal performance such as survival.

ResultsThe temperature that minimizes the costs of development is strongly related to the mid-environmental temperature for that species (R2 = 0.83; F1,69 = 344.3, P < 0.0001; Fig. 2c). Phylogeny explained a significant proportion of the residual variation (λ = 0.294 [95% con-fidence interval (CI): 0.08, 0.66]) in Topt. The slope of the overall relationship (1.02 ± 0.08 s.e.m.) was not significantly different from 1:1 and the intercept (0.83 ± 1.43) was not significantly different from zero (t69 = 0.27, P = 0.78; t69 = 0.58, P = 0.56).

The apparent isometric relationship between the environmental mid-temperature and Topt belies systematic differences in this rela-tionship for organisms that develop aquatically, in terrestrial nests or terrestrially but not in nests (development environment: F2,67 = 6.30, P = 0.003; Fig. 2c). Phylogeny did not explain a significant propor-tion of the residual variation (λ = 0.049 [95% CI: 0, 0.403]). These patterns can best be visualized by calculating the difference between Topt and Tmid for the different developmental environments (Fig. 3). Topt was systematically higher than Tmid in species that developed in terrestrial nests whereas, for non-nesting terrestrial species, in all but two instances Topt was lower than Tmid. For aquatic species, the value of Topt relative to Tmid depended on the temperature but for around 45% of species, Tmid was equal to or greater than Topt.

In terrestrial species that do not nest, the relationship between Topt and Tmid was supralinear (scaling exponent ± s.e.m. = 1.3 ± 0.08) whereas in aquatic environments, it was sublinear (scaling expo-nent ± s.e.m. = 0.82 ± 0.07). On the basis of these nonlinear relation-ships between Tmid and Topt for terrestrial non-nesters and aquatic species, we can calculate the temperature ranges in which Tmid already exceeds Topt. For aquatic species, those in the tropics with Tmid’s greater than ~22 °C have Topt’s that are lower than Tmid. For ter-restrial non-nesting species, only tropical species have Topt’s that are higher than Tmid.

Sensitivity to deviations from Topt. Terrestrially developing species had systematically higher sensitivity (estimated as curvature at the origin via the second derivative) to deviations from the Topt than aquatic developing species (aquatic mean ± s.e.m. = 7 ± 1.3; terres-trial mean ± s.e.m. = 60 ± 21; F1,69 = 48.8, P < 0.001; λ = 0.03 [95% CI: 0.00, 0.28]). For example, a 20% shift (relative to the thermal range) in temperature above the optimum for the terrestrially developing C. johnstoni increases development costs by 15%, whereas in the aquatically developing Danio rerio, a similar temperature increase only increases development costs by 5% (see Supplementary Table 4 for full list). Applying the same percentage increase in temperature for all species, terrestrial species show much greater development costs than aquatic species (mean aquatic, 5.8% ± 1; mean terrestrial, 12.0% ± 1.6; F1,69 = 12.4, P < 0.001; λ = 0 [95% CI: 0.00, 0.19]).

Parameter variation. The temperature dependency of metabolic rate (parameter a) and development time (parameter b) decreased with environmental temperature (a: t68 = −4.93, P < 0.001; b: t68 = −2.50, P = 0.01), but did not differ among habitats (a: t68 = −1.12, P = 0.27; b: t68 = −1.60, P = 0.11). Neither parameter was associated with a strong phylogenetic signal (λ = 0)

Relative sensitivity of costs of development and survival. Costs of development were far more sensitive than survival to a small deviation from the optimal environment temperature (paired t-test: t41 = 4.01, P < 0.001). The costs of development increased by ~8% on average whereas mortality increased by less than 3% for the same temperature deviation (Fig. 4). For just over 50% of the species for which we could find data, a temperature deviation of 20% from

Met

abol

icra

te

Developm

enttim

e

Cos

t of

deve

lopm

ent

Temperature (°C)

ba

c

10 20 30 40

C(T ) = MR(T ) × D(T )

D(T ) = de–bT + cMR(T ) = δeaT

d(b–a)acTopt =

1b

In( )

Fig. 1 | graphical representation of temperature-dependent developmental cost theory. a,b, Temperature affects both metabolic rate (a) and development time (b) but, for most species, development time is more sensitive than metabolism. c, The product of development time and metabolic rate yields the total costs of development; a single temperature (Topt) is predicted to minimize the cost of development. See main text for definitions of parameters.

NAtuRe eCology & evolutioN | VOL 4 | MArCh 2020 | 406–411 | www.nature.com/natecolevol 407

Articles NaTure ecOlOgy & evOluTION

Topt caused no change in mortality relative to the optimum but the same temperature deviation increased the costs of development for all species in the dataset. The costs of development are therefore more likely than survival to change in response to changes in tem-perature, and the costs of development are more sensitive than sur-vival when temperatures differ slightly from the optimum for that species. Survival tended to be more sensitive to deviations in tem-perature in terrestrial species relative to aquatic species (F1, 40 = 3.88, P = 0.056, λ = 0 [95% CI: 0.00, 0.17]).

DiscussionAcross a wide array of taxa, the temperature that minimizes the costs of development is very similar to the temperature at which they are likely to develop. The cost of development is close to minimum across only a relatively small thermal window—above or below which results in increasingly costly development. Our data suggest that ectotherms have minimized developmental costs by evolving combinations of temperature dependencies of metabolic rate and developmental time that ensure strong concordance between envi-ronmental temperature and Topt. Our model predicts that future temperature increases will increase the costs of development for some species but could decrease costs for others. Our model also suggests that ectotherms may be far more sensitive to future warm-ing than traditional measures of thermal performance would imply.

Our findings have worrying implications for the impacts of global temperature increases and identify those organisms that are at greater risk. Importantly, there seems to be no obvious ‘safety fac-tor’ buffering organisms from deviations in temperature—the rela-tionship between Topt and Tmid had an intercept that was not different from 0. Thus, any temperature deviation, where Tmid is either cooler or warmer than Topt generates additional developmental costs.

For almost half of all aquatic species, Tmid’s are already higher than Topt. Tropical aquatic species, in particular, appear to live in temperatures that often exceed Topt but it is also worth noting that two Antarctic invertebrates have Topt’s that are lower than Tmid. For non-nesting terrestrial species, that are living in temperatures cooler than ~17.5 °C (temperate and polar species), Tmid already exceeds Topt; any further temperature increase is predicted to increase devel-opmental costs. Higher costs of development require mothers to invest more resources in each offspring to compensate for these costs3. If maternal resources remain unchanged, then fecundity must decrease as per capita offspring investment increases. Thus, for species that are already above Topt, our model predicts an insidi-ous twofold impact of further warming. First, development will become less efficient—offspring will achieve nutritional indepen-dence with a lower proportion of energy reserves than they would at cooler temperatures—this is probably to reduce the performance of these offspring9. Second, parents may increase the per capita

investment in offspring to compensate for this increase in develop-mental costs and suffer reduced fecundity as a consequence3. Assuming density dependence is not very strong, and populations are recruitment limited, both of these effects could reduce the recruit-ment of adults into the population and reduce population robustness. An important next step will be to examine how thermal develop-ment environments affect postdevelopment performance. Higher

0 10 20 30 40

Environmental temperature, Tmid(°C)

–10

0

10

20

30

40 1:1

0.2

0.4

0.6

0.8

1.0

b

c

0.2

0.4

0.6

0.8

1.0a

Rel

ativ

e M

R a

nd d

evel

opm

ent

Rel

ativ

e co

st o

f dev

elop

men

tO

ptim

al te

mpe

ratu

re, T

opt

(°C

)

Fig. 2 | Development costs across temperatures for developing ectotherms. a, The relationship between temperature and relative metabolic rate and development time for three different species: C. harengus (dark blue) E. superba (light blue) and C. johnstoni (yellow). b, Illustrating how tempera-tures dependencies combine to affect the relative costs of development across temperatures for the three species and that a single temperature (Topt) minimizes the costs of development for each species. c, The relation-ship between Topt and Tmid (the estimate of the temperature species experience in nature, see main text for details) for 71 species across seven phyla including the three species shown in a and b (silhouettes indicate their points). Species with aquatic development are shown in blue, species that develop in terrestrial nests are shown in yellow and species that develop terrestrially but not in nests are shown in green. The grey dotted line indicates a 1:1 ratio between Topt and Tmid. Credit: silhouettes adapted from http://phylopic.org/ (crocodile, krill and mullet) (https://creativecommons.org/licenses/by/3.0/).

NAtuRe eCology & evolutioN | VOL 4 | MArCh 2020 | 406–411 | www.nature.com/natecolevol408

ArticlesNaTure ecOlOgy & evOluTION

temperatures during development can either increase or decrease subsequent performance9, thus offsetting or exacerbating the effects we describe here.

We found that the costs of development are systematically rela-tively more sensitive to temperature deviations from the optimal than survival differences. Relying on temperature effects on survival to estimate the consequences of climate warming therefore risks sys-tematically underestimating changes in population replenishment. For example, a 1.6 °C temperature increase is predicted to reduce hatching success of Atlantic salmon (Salmo salar) by 8%, whereas our model predicts that development will become 17% more costly for the same temperature increase. Similarly, a 2.2 °C temperature increase will decrease the hatching success of Rana pipiens tadpoles by only 3.5% but increase the costs of development by ~30%. If mothers raise their per capita investment in offspring to compen-sate for such a temperature increase, fecundity may decrease by a concomitant amount10, reducing the total fecundity of the popula-tion overall. An interesting next step will be to examine how tem-perature affects fecundity. Also, note that if mortality rates during the embryonic phase are extremely high, then warming-induced reductions in development time could still enhance fitness, despite increasing developmental costs overall11. For now, however, we pre-dict that the effect of warming on population replenishment via development cost-mediated effects on fecundity could exceed the direct effect of temperature on survival.

Our results imply that there are probable beneficiaries of further warming, at least from a developmental costs perspective, though cautious interpretation is necessary. Developmental cost theory predicts that species that develop in terrestrial nests should have lower development costs at higher temperatures—Topt was always higher than Tmid in this group. However, such an interpreta tion assumes that nests are not systematically warmer than general temperature conditions already—given non-random nest site selection

and the thermoregulatory role of some nests, such assumptions seem unlikely. So our theory predicts that terrestrial nesting spe-cies should benefit from warmer temperatures, if and only if nest temperatures are not warmer than local conditions already. If on the other hand, nests are already systematically warmer than local conditions, then nesting species must either alter their nest site selection or construction, or face rising developmental costs. It is also important to note that for the species considered here, increases in temperature increase the costs of development eventually. As such, even if minor temperature increases decrease the costs of develop ment initially, further warming will always increase the costs of development once Topt is exceeded. Thus, we suggest caution when making inferences about any possible benefits of continued temperature increases for the costs of development.

We show development is more costly whenever embryos develop in temperatures that deviate from optimum but the mag-nitude of these increases differed among environments. We found that increases in development costs were much greater in terrestri-ally developing species relative to aquatically developing species (estimated both as curvature or changes in development costs at set temperature deviations from the optimum). Interestingly, we found similar trends when we examined survival at different tem-peratures. Our results contrast with a much more comprehensive, recent comparison of thermal performance between terrestrial and aquatic species8. Our focus here was on developing embryos and larvae, whereas Pinsky et al.8 examined adult performance. Earlier life-history stages tend to be the most sensitive and so it could be that different stages are systematically more or less sensitive when comparing terrestrial and aquatic species. Alternatively, differences in the way in which sensitivity was measured between our study and Pinsky et al.8 may drive differences in the findings—they focused on thermal safety margins rather than physiological sensitivity per se.

We suspect the difference in sensitivity between animals that develop in terrestrial and aquatic environments occurs because species with terrestrial development have more opportunities to regu late the temperature their offspring experience12. Terrestrial

Cha

nge

(%)

10

20

30

40

50

Survival Developmental costs

Performance measure

Fig. 4 | Change in performance associated with a 20% increase (relative to the range experienced by that species) in temperature relative to the thermal optimum. On the left, performance is measured in terms of percentage change in embryo survival. On the right, performance is measured in terms of percentage change in developmental costs. Boxplots summarize the two measures. Each line represents an individual species and connects the relative effects in terms of survival and developmental costs for that particular species. Blue lines, sloping upward from left to right indicate that developmental costs associated with temperature increases are greater than reductions in survival (n = 42), while red lines (n = 4) show the converse.

‘Too hot’

‘Too cold’

Tm

id–T

opt

–20

–10

0

10

Terrestrial(non-nesting)

Terrestrial(nesting)

Aquatic

Fig. 3 | effect of developmental environment on mismatch between Topt and Tmid for 71 ectotherms. Values greater than zero indicate that environmental mid-temperatures are higher than those that minimize the costs of development and values less than zero indicate that mid-temperatures are less than those that minimize the costs of development. Species with aquatic development are shown in blue, species that develop in terrestrial nests are shown in yellow and species that develop terrestrially but not in nests are shown in green. Boxplots show interquartile ranges where boxes include first two quartiles of the data and whiskers include data within 1.5× the interquartile range. Credit: silhouettes adapted from http://phylopic.org/ (crocodile, bug and tuna) (https://creativecommons.org/licenses/by/3.0/).

NAtuRe eCology & evolutioN | VOL 4 | MArCh 2020 | 406–411 | www.nature.com/natecolevol 409

Articles NaTure ecOlOgy & evOluTION

species can take advantage of thermal microclimates via the selec-tion of oviposition site and the thermal regulatory potential of nests13. In contrast, aquatic habitats are relatively more homo-geneous in temperature, rendering the selection of thermal micro-habitat a less effective strategy for thermoregulation14. Aquatically developing species might therefore need to be more robust to tem-perature variation, such that the costs of development increase less sharply when temperatures deviate from the optimum relative to terrestrially developing species.

Developmental cost theory reveals that a single temperature optimum that minimizes the costs of development emerges from the nature of the relationships between temperature, metabolic rate and development time. Similar exponential relationships exist for another ubiquitous but costly (from an energy perspective) onto-genetic transition—metamorphosis15. Transitioning from the larval or pupal phase to the juvenile/adult phase can cost up to 60% of energy reserves in some species2. An interesting next step would be to explore whether there is a particular temperature that minimizes the costs of metamorphosis for species with complex life cycles. If so, then this this represents another pathway by which temperature affects the energetics and performance of species with complex life histories. Across a wide variety of taxa, energy reserves at the com-pletion of metamorphosis predict postmetamorphic performance9. Temperatures that deviate from the optimum that minimize the costs of metamorphosis will therefore either reduce the performance of the adult stage or require organisms to spend longer in the (often dangerous) larval stage so as to accumulate sufficient resources.

Importantly, we expect the temperature dependencies of develop-mental and/or metabolic rate to evolve in response to global change to reduce these fitness costs incurred at higher temperatures. Previous work has shown that broad-scale climate influences the thermal sensitivity and plasticity of physiological traits, including metabolic rate16. Indeed, our own data suggest that these parameters do evolve in response to local conditions: higher Tmid’s were associated with lower temperature dependencies of both metabolism and developmental time. Laboratory natural selection experiments, on the other hand, suggest that the metabolic rate of adult Drosophila melanogaster has limited capacity to evolve in response to temperature change17,18 but comparable studies of development, and the metabolic rates of devel-oping embryos, are presently lacking. Our framework allows the a priori identification of which parameters are likely to drive changes in the costs of development across different temperatures. On the basis of equation (6), we would expect b, the temperature dependence of development rate, would be under the strongest selection as this parameter has the greatest influence on the relative costs of develop-ment and Topt. An important next step will be to test whether our expectation matches findings from experimental evolution studies that have manipulated temperature in species with complex life cycles.

We found that, for a given species, a single temperature mini-mizes the costs of development for a wide array of ectotherms and that this predicted value is very similar to the temperature in which each species develops. The unimodal nature of the relationship between temperature and developmental costs means that devel-opment is inevitably more costly when temperature deviates from the optimum—this represents a previously unrecognized, near uni-versal, vulnerability of metazoan ectotherms. Developmental cost theory links life-history theory with metabolic theory to show how changes in function (the costs of development) drive these perfor-mance curves in a physiologically explicit framework. Our results strongly suggest that life histories and investment strategies will change under future temperature increases, though the impacts of these changes may vary systematically across habitats and latitudes.

MethodsDefinitions of development and data compilation. We define development as the phase between fertilization and when independent feeding begins. So, for

some species (for example, zebra fish), once they hatch into feeding larvae, we would describe development as complete, while for other species (for example, lecithotrophic marine invertebrates such as Bugula), development would include the non-feeding larval stage all the way through metamorphosis into a juvenile. Importantly, this definition does not suppose that all parentally derived resources have been exhausted once feeding begins; rather, this is simply the end of the phase when offspring rely exclusively on these resources.

Data on the temperature dependency of development time and the temperature dependency of metabolic rate were mostly compiled from an earlier study3, supplemented with additional data from the literature. Due to the paucity of data on these rates under varying temperature regimes, we could not rely on search terms. Rather we combed the literature haphazardly using ISI Web of Science (http://www.webofknowledge.com/WOS) for a range of search terms and followed relevant citations to identify as many studies as possible. We parameterized the MR(T) and D(T) functions, on the basis of the temperature ranges experienced by each species during the early life history when offspring are non-feeding and dependent on maternally derived energy (throughout the ‘dependent phase’). For two species (Gynaephora groenlandica and Erynnis propertius), we could not find data on the temperature dependence of metabolic rate of the embryo stage but it was available for the early larval stage. We chose to use these data because data for terrestrial developers under cool temperatures (<20 °C) were extremely rare otherwise (our results were qualitatively identical whether these species were included or not). In total, we parameterized the relationships between temperature and development time (D(T)), temperature and metabolic rate (MR(T)) from studies of 71 species from seven phyla (Supplementary Table 1). We also collected embryo/larval survival data for as many species as we could find and, in most cases, these survival data came from the same study as that of the metabolism and development data.

We also collected data on the thermal regime in which our species of interest developed. Where temperature range was not specified for the studies that measured MR(T) and D(T), we searched the literature for studies reporting temperature ranges for each species in similar locations. Throughout our analyses, we used the temperature midway (Tmid) between the minimum and maximum temperatures recorded for that species—our results were qualitatively unchanged by using alternative estimates of environmental temperature (Tmid was strongly positively correlated with mean, maximum and median temperature; R2 > 0.70 in all cases; see Supplementary Information).

Meta-analysis methods. Most studies did not report sample size associated with measurements; for those studies that did report both sample size and estimates of error, we used generalized least squares using the ‘gls’ function within the ‘nlme’ R package19 to determine whether sample size was related to a higher precision of estimates. This approach has been suggested by Nakagawa and Lagisz20 for use in meta-analyses that measure only the absolute magnitude (mean) of the effect size on the response variable (effect of temperature on mean D and MR). We used phylogenetic generalized least squares to fit models where residuals are correlated (correlation structure taken from phylogeny) using the ‘pgls’ function in the R package ‘caper’13,21. The tree topology was taken from the open tree of life using the ‘rotl’ package in R v.3.3.2 (refs. 22,23). The methods used to produce a phylogenetically controlled meta-analysis were closely followed as per guidelines presented in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement24—for details on methodological considerations and publication bias for these data see ref. 3. Phylogenetic signal is presented as Pagel’s25 lambda (λ), which represents the proportion of variance—conditioned on the fixed effects—attributable to phylogeny26.

Analysis of Topt − Tmid relationships. We first analysed our data in a phylogenetically controlled framework where environmental temperature (Tmid) was the continuous predictor and Topt was the response variable. We then included a second, categorical predictor—the habitat in which offspring development took place. Species were classified as either ‘aquatic’ or ‘terrestrial’ according to where development took place, irrespective of where adults of the species lived. For example, turtles and crocodiles, which live in aquatic habitats as adults, were classified as terrestrial because their offspring develop out of water. Likewise, any species that had aquatically developing offspring but terrestrial adult stages (for example, amphibians and many insects) were classified as aquatic. For species that develop terrestrially, we further classified them according to whether the offspring develop in nests or not. We did this because such species may have greater capacity for thermoregulation during development and are likely to experience different temperature regimes to terrestrial species that do not nest. Because we did not have fully overlapping temperature ranges across all our groups, we excluded the two coldest aquatic species from our formal comparisons of habitats27.

We next examined whether systematic deviations between the predicted Topt and Tmid varied with the environment (aquatic, terrestrial nesting and terrestrial non-nesting) in which species developed by calculating Tmid – Topt as a response variable with the same predictors as above. Positive values of Tmid – Topt indicate conditions that are warmer than the temperature that minimizes the costs of development, while negative values indicate the converse.

NAtuRe eCology & evolutioN | VOL 4 | MArCh 2020 | 406–411 | www.nature.com/natecolevol410

ArticlesNaTure ecOlOgy & evOluTION

We also examined how specific parameters varied with environmental temperature—specifically we examined how/if the temperature dependency of metabolic rate (a) and development (b) varied across environmental temperature.

Analysis of curvature. Describing the costs of development using a relatively simple set of equations allows us to formally estimate the sensitivity of develop-mental costs to deviations in temperature away from the optimal temperature. We calculated the second derivative of our cost function to yield the equation:

d2CdT2

¼ ea Tð Þ a2cþ a� bð Þ2de�bT� � ð6Þ

and we then evaluated this equation at Topt to calculate the rate at which the cost changes around this optimum.

Relative sensitivity of costs of development and survival. Our results suggested that the costs of development were extremely sensitive to environmental temperature and we wanted to compare the relative sensitivity of developmental costs to the sensitivity of survival of developing embryos under relatively small temperature changes. We did not collect data on adult survival, as meta-analyses suggest that later life-history stages are far more robust to a range of temperatures3. Ideally, we would have compiled thermal performance curves where performance was defined as embryo survival for each species but data were too scarce. We could only find hatching success or embryo survival at different temperatures for 46 of 71 species (64%) in our database and, for many of these, hatching success or embryo survival was measured at only two temperatures. We therefore fit simple linear relationships between survival at Topt and survival at a set deviation from Topt (20% of the reported temperature range for that species in °C). We then compared how both survival and costs of development changed across this standardized temperature deviation. We compared these values using phylogenetically controlled analyses to test whether these two measures of sensitivity to temperature differed from each other. We found no phylogenetic signal and so report only the results from a simple paired t-test where changes in costs and changes in survival were the paired measures for each species.

Reporting Summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availabilityThe data used in these meta-analyses are available from Dryad: https://doi.org/10.5061/dryad.zw3r2284x.

Received: 2 August 2019; Accepted: 14 January 2020; Published online: 2 March 2020

References 1. Mueller, C. A., Joss, J. M. P. & Seymour, R. S. The energy cost of embryonic

development in fishes and amphibians, with emphasis on new data from the Australian lungfish, Neoceratodus forsteri. J. Comp. Physiol. B 181, 43–52 (2011).

2. Bennett, C. E. & Marshall, D. J. The relative energetic costs of the larval period, larval swimming and metamorphosis for the ascidian Diplosoma listerianum. Mar. Freshwat. Behav. Physiol. 38, 21–29 (2005).

3. Pettersen, A. K., White, C. R., Bryson-Richardson, R. J. & Marshall, D. J. Linking life-history theory and metabolic theory explains the offspring size-temperature relationship. Ecol. Lett. 22, 518–526 (2019).

4. Kamler, E. Early Life History of Fish—An Energetics Approach, Vol. 4 (Chapman and Hall, 1992).

5. Kamler, E. Parent–egg–progeny relationships in teleost fishes: an energetics perspective. Rev. Fish Biol. Fish. 15, 399–421 (2005).

6. Zuo, W. Y., Moses, M. E., West, G. B., Hou, C. & Brown, J. H. A general model for effects of temperature on ectotherm ontogenetic growth and development. Proc. R. Soc. B 279, 1840–1846 (2012).

7. Gillooly, J. F., Charnov, E. L., West, G. B., Savage, V. M. & Brown, J. H. Effects of size and temperature on developmental time. Nature 417, 70–73 (2002).

8. Pinsky, M. L., Eikeset, A. M., McCauley, D. J., Payne, J. L. & Sunday, J. M. Greater vulnerability to warming of marine versus terrestrial ectotherms. Nature 569, 108–111 (2019).

9. Pechenik, J. A. Larval experience and latent effects—metamorphosis is not a new beginning. Integr. Comp. Biol. 46, 323–333 (2006).

10. Marshall, D. J., Pettersen, A. K. & Cameron, H. A global synthesis of offspring size variation, its eco-evolutionary causes and consequences. Funct. Ecol. 32, 1436–1446 (2018).

11. Vance, R. R. On reproductive strategies in marine benthic invertebrates. Am. Nat. 107, 339–352 (1973).

12. Shine, R. Manipulative mothers and selective forces: the effects of reproduction on thermoregulation in reptiles. Herpetologica 68, 289–298 (2012).

13. Goller, M., Goller, F. & French, S. S. A heterogeneous thermal environment enables remarkable behavioral thermoregulation in Uta stansburiana. Ecol. Evol. 4, 3319–3329 (2014).

14. Steele, J. H. A comparison of terrestrial and marine ecological systems. Nature 313, 355–358 (1985).

15. Melampy, R. M. & Willis, E. R. Respiratory metabolism during larval and pupal development of the female honeybee (Apis mellifica L.). Physiol. Zool. 12, 302–311 (1939).

16. Seebacher, F., White, C. R. & Franklin, C. E. Physiological plasticity increases resilience of ectothermic animals to climate change. Nat. Clim. Change 5, 61–66 (2014).

17. Berrigan, D. & Partridge, L. Influence of temperature and activity on the metabolic rate of adult Drosophila melanogaster. Comp. Biochem. Physiol. A 118, 1301–1307 (1997).

18. Alton, L. A., Condon, C., White, C. R. & Angilletta, M. J. Jr Colder environments did not select for a faster metabolism during experimental evolution of Drosophila melanogaster. Evolution 71, 145–152 (2017).

19. Pinheiro, J., Bates, D., DebRoy, S., Sarkar, D. & R Core Team. nlme: Linear and Nonlinear Mixed Effects Models. R package version 3.1-101 (2011).

20. Nakagawa, S. & Lagisz, M. Visualizing unbiased and biased unweighted meta-analyses. J. Evol. Biol. 29, 1914–1916 (2016).

21. Orme, D. The caper Package: Comparative Analysis of Phylogenetics and Evolution in R. R package version 5 (2013).

22. Michonneau, F., Brown, J. W. & Winter, D. J. rot1: an R package to interact with the Open Tree of Life data. Methods Ecol. Evol. 7, 1476–1481 (2016).

23. Hinchliff, C. E. et al. Synthesis of phylogeny and taxonomy into a compre-hensive tree of life. Proc. Natl Acad. Sci. USA 112, 12764–12769 (2015).

24. Moher, D., Liberati, A., Tetzlaff, J., Altman, D. G. & The, P. G. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. PLoS Med. 6, e1000097 (2009).

25. Pagel, M. Inferring the historical patterns of biological evolution. Nature 401, 877–884 (1999).

26. Hadfield, J. D. & Nakagawa, S. General quantitative genetic methods for comparative biology: phylogenies, taxonomies and multi-trait models for continuous and categorical characters. J. Evol. Biol. 23, 494–508 (2010).

27. Quinn, G. P. & Keough, M. J. Experimental Design and Data Analysis for Biologists (Cambridge Univ. Press, 2002).

AcknowledgementsWe thank B. Comerford, M. Parascandalo and L. Morris for assistance. This work was supported by funding to the Centre for Geometric Biology, Monash University.

Author contributionsD.J.M. conceived the study, collected and analysed the data and wrote the first draft. M.B. developed the mathematical theory. D.J.M. and A.K.P. collected the data. D.J.M. and C.R.W. analysed the data. All authors contributed to subsequent drafts.

Competing interestsThe authors declare no competing interests.

Additional informationSupplementary information is available for this paper at https://doi.org/10.1038/s41559-020-1114-9.

Correspondence and requests for materials should be addressed to D.J.M.

Reprints and permissions information is available at www.nature.com/reprints.

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© The Author(s), under exclusive licence to Springer Nature Limited 2020

NAtuRe eCology & evolutioN | VOL 4 | MArCh 2020 | 406–411 | www.nature.com/natecolevol 411

1

nature research | reporting summ

aryO

ctober 2018

Corresponding author(s): Dustin J. Marshall

Last updated by author(s): Jan 6, 2020

Reporting SummaryNature Research wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency in reporting. For further information on Nature Research policies, see Authors & Referees and the Editorial Policy Checklist.

StatisticsFor all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.

n/a Confirmed

The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement

A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly

The statistical test(s) used AND whether they are one- or two-sided Only common tests should be described solely by name; describe more complex techniques in the Methods section.

A description of all covariates tested

A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons

A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals)

For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted Give P values as exact values whenever suitable.

For Bayesian analysis, information on the choice of priors and Markov chain Monte Carlo settings

For hierarchical and complex designs, identification of the appropriate level for tests and full reporting of outcomes

Estimates of effect sizes (e.g. Cohen's d, Pearson's r), indicating how they were calculated

Our web collection on statistics for biologists contains articles on many of the points above.

Software and codePolicy information about availability of computer code

Data collection The data were collected for a previous meta-analysis (Pettersen et al. 2019)

Data analysis Systat Vers 13, R Vers. 3.3.2.

For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors/reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Research guidelines for submitting code & software for further information.

DataPolicy information about availability of data

All manuscripts must include a data availability statement. This statement should provide the following information, where applicable: - Accession codes, unique identifiers, or web links for publicly available datasets - A list of figures that have associated raw data - A description of any restrictions on data availability

/doi.org/10.5061/dryad.zw3r2284x

Field-specific reportingPlease select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.

Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences

For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf

2

nature research | reporting summ

aryO

ctober 2018

Ecological, evolutionary & environmental sciences study designAll studies must disclose on these points even when the disclosure is negative.

Study description A study of 71 species with regards to the temperature-dependence of development and metabolic rate

Research sample 71 species of ectotherm that lack extensive post-developmental parental care.

Sampling strategy PRISIMA meta-analysis guidelines were followed

Data collection Dustin Marshall and Amanda Pettersen were principally responsible for data collection.

Timing and spatial scale This was a meta-analysis based on studies spanning the last 80 years

Data exclusions See Pettersen et al (2019), on whihc this study was based for data flow description.

Reproducibility No experimental findings.

Randomization No experiments were conducted

Blinding This was a meta-analysis so no blinding was possible or necessary.

Did the study involve field work? Yes No

Reporting for specific materials, systems and methodsWe require information from authors about some types of materials, experimental systems and methods used in many studies. Here, indicate whether each material, system or method listed is relevant to your study. If you are not sure if a list item applies to your research, read the appropriate section before selecting a response.

Materials & experimental systemsn/a Involved in the study

Antibodies

Eukaryotic cell lines

Palaeontology

Animals and other organisms

Human research participants

Clinical data

Methodsn/a Involved in the study

ChIP-seq

Flow cytometry

MRI-based neuroimaging