Application of a high-throughput respirometry technique for assessing respiratory metabolism of terrestrial and aquatic invertebrates

Abstract

Respiratory metabolism is one of the key ecological traits of organisms. As such, it is studied by ecologists from a wide range of disciplines, from macroecological studies to evolutionary biology, with the global changes context making the need for its understanding more pressing. Unfortunately, properly measuring respiration is a time- and cost-intensive endeavor, and gathering large datasets can be prohibitive due to the difficulty of measuring it. Here, we describe a fluorescence-based high-throughput closed-system respirometry technique for terrestrial and aquatic invertebrates, able to measure respiration for up to 240 channels in parallel in vials of 2 or 4 ml. We used a setup originally designed to continuously measure oxygen in laboratory cell cultures in liquid medium to measure arthropod respiration. Through a series of experiments, we demonstrate the reliability of the system to measure respiration of arthropods. Even very small arthropods were able to be measured, although Drosophila had to be measured in groups of 2 to 4. Further, the system’s modularity makes it particularly fit for assessing full thermal performance curves of respiratory metabolism in a small amount of time. We provide helpful guidelines and R scripts to make use of this system to create thermal performance curves of insect metabolism.

Introduction

The respiratory metabolic rate is an often-measured phenotypic trait that has consequences for evolutionary biology (Burton et al. 2011), community dynamics, and ecosystem functioning (Yvon-Durocher and Allen 2012). Within a species, the energy cost of self-maintenance (measured as resting metabolic rate) varies remarkably, and is repeatable (Auer et al. 2016) and heritable (Nilsson et al. 2009). Such finite energy budget is subject to tradeoffs from energy allocation between growth, reproduction, and self-maintenance, and thus variation in resting metabolic rate is of large importance for life-history traits and fitness (Burton et al. 2011). In addition, macroecological studies have shown systematic variation in metabolic rates among species (Chown and Gaston 1999). Such systematic variation has long been linked to body mass (Kleiber 1932, Regnault and Reiset 1849). More recently, the metabolic theory of ecology proposes that metabolic rates should both scale allometrically with body mass (Brown et al. 2004, Glazier 2005) and increase exponentially with temperature following the Boltzmann–Arrhenius relationship (Gillooly et al. 2001, Brown et al. 2004, Clarke 2004, Dell et al. 2011). In a global changes context, species vulnerability to climate warming has been linked to thermal performance of their metabolism (Magozzi and Calosi 2015).

Given the centrality of respiratory metabolism in ecology, several techniques have been developed to accurately measure respiration, including flow-through respirometry, intermittent flow respirometry and closed respirometry (Steffensen 1989, Lighton 2008, Clark et al. 2013, Svendsen et al. 2016), each with their own advantages and drawbacks (Lighton and Halsey 2011). In addition, other indirect methods of measuring metabolism have been used in the field, including the doubly labeled water method, heart rate, or body acceleration (Halsey 2011, Treberg et al. 2016). However, methods to estimate metabolism are often time-intensive, with datasets collected over weeks of experiments. Researchers have long aimed at using high-throughput respirometry techniques to measure metabolic rates (Scafaro et al. 2017, Drown et al. 2020).

Closed respirometry involves keeping organisms in a sealed container of known air volume and measuring the change in O₂ (and/or CO₂) composition over a time interval (Vleck 1987, Lighton 2008). Such closed systems can be optimized to measure small organisms through reducing chamber volume or allowing for long measurement times, but they are constrained to one-off measurements from the start to the end of the trial, allowing only for the calculation of an average metabolic rate. More recently, a novel technique using fluorometric quenching of organic ruthenium derivates in the presence of oxygen (Köster et al. 2008) has been developped, allowing for noninvasive sequential, repeated measurements of the same organism inside of the chamber. Such method is interesting for measuring respiration on very small organisms. Further, it allows a parallelization of measurements that can be incomparable with the parallelization achieved with standard respirometry techniques (Scafaro et al. 2017, Drown et al. 2020).

Here, we present an adaptation of a high-throughput closed respirometry technique based on this approach to measure respiration in both terrestrial and aquatic invertebrates, leading to up to 240 measurements at the same time. We used a setup originally designed to continuously measure oxygen in laboratory cell cultures (the PreSens SDR SensorDish Reader) that has been used in particular for biotechnological applications (eg Campbell et al. 2014), or in more ecological settings for the measurement of respiration of bacteria, phytoplankton, or zooplankton in liquid media (Köster et al. 2008, Einum et al. 2019, Tseng et al. 2019, Ghedini et al. 2022, Álvarez-Codesal et al. 2023, García et al. 2023), and very recently to measure respiration in aphids (Clavé et al. 2022) or spiders (Uiterwaal et al. 2022), but without a proper demonstration. Our goal is to demonstrate the reliability of this method, and provide detailed methods. We first demonstrate the stability and repeatability of measurements within and across experimental trials with 2 species of terrestrial arthropods. Second, we show that measured metabolism corresponds to the broad mass–metabolism and temperature–metabolism relationships expected from the metabolic theory of ecology in both terrestrial and aquatic invertebrates. Last, we show a worked example of application to create a full thermal performance curve of metabolism with a terrestrial arthropod. Furthermore, we provide helpful guidelines to measure respiration along with helpful R scripts to help researchers create thermal performance curves of respiratory metabolism (Supplementary Material, Supplementary Appendix 2).

Experimental Design

Materials

The SDR SensorDish Reader (PreSens Precision Sensing, Regensburg, Germany) is a small 24-channel reader for noninvasive detection of oxygen through optical fluorescence-based respirometry (Fig. 1). In the standard configuration, the SensorDish can measure oxygen values in 2 or 4 ml reusable glass vials (SensorVial SV-PSt5-2 ml or SV-PSt5-4 ml) of 14.9 mm in diameter and 27.6 and 47 mm in total length (with cap), respectively, with an opening of 8.2 mm inner neck diameter, and a maximum filling volume of 2.8 and 4.9 ml, respectively. In addition, a special mask (SDR-MSV12-20ML) allows using 20-ml SensorVials (SV-PSt5-20 ml), of 27.4 mm in diameter and 60 mm in length, with an opening of 17.4 mm inner neck diameter, and a maximum filling volume of 25 ml, although the SDR then fits half of the number of vials (12 20-ml vials versus 24 4 ml vials in the standard configuration). The SensorVials contain at the bottom an oxygen sensor spot (type PSt5) consisting of an oxygen-sensitive foil with an immobilized fluorescent dye (ruthenium derivate) and an optical isolation. The measuring principle of the optical-based sensors is based on fluorescence quenching of organic ruthenium derivates in the presence of oxygen (Köster et al. 2008). The range of measurements is from +15 to +45 °C. The vials come precalibrated, but it is possible to create user-defined calibrations (eg due to working at low temperatures).

Fig. 1.

The SensorDish reader. A) SensorDish with 2 ml SensorVials inside an incubator. B–D) Example pictures of either 1 firebug (B) or 3 fruit flies (C) inside a 2 ml SensorVial, or (D) spiders inside 4 ml SensorVials. E) Multidevice setup of the SensorDish reader and one extension set. F) Screen capture of the SDR software graph window. Credits: pictures A to E: Elvire Bestion. Screen capture F: PreSens Precision Sensing.

The SensorDish basic set of one 24-channel reader can be extended with up to 9 SDR extensions sets, leading to up to 240 samples monitored in parallel, controlled with one computer. Such features makes it ideally suited for highly parallelized measurements.

To perform the respiration experiment, researchers will need one (or more) SensorDish reader, 24 (or more) SensorVials (either 2, 4, or 20 ml depending on invertebrate body size), a connected Windows 10 or 11 computer equipped with the SDR Software v4.0.0, and one (or more) incubator (in our case Panasonic MLR-352-PE, temperature control range: 0 to +50 °C, temperature control fluctuation ±0.3 °C) to maintain a constant temperature.

Technical Validation Protocol

We performed a series of experiments testing (i) the long-term stability and repeatability of measurements over a 5-h period and at 2 temperatures with crickets Acheta domesticus, (ii) the repeatability of measurements over 6 separate measurements over 3 d in firebugs Pyrroccoris apterus, (iii) the mass- and temperature-dependency of respiration across species with 7 species of terrestrial arthropods, and (iv) the mass-dependency of respiration in aquatic settings with amphipods Echinogammarus sp (Table 1).

Table 1.

Summary of the different experiments with species, samples sizes, temperatures, vial size, and acclimation time

Experiment	Species, Nindiv, Nb tests	Temperatures	Vial size (ml)	Acclimation time (min)
Experiment 1: Long-term stability of respiration	Acheta domesticus (36 individuals, 18 per temperature, 1 test per individual)	20, 30 °C	4	60
Experiment 2: Repeatability of respiration	Pyrrocoris apterus (40 individuals, 6 tests per individual)	20 °C	4	60
Experiment 3: Mass- and temperature–respiration relationship in terrestrial invertebrates	Pyrrocoris apterus (N = 6), Pardosa sp (N = 6), Iberoformica subrufa (N = 4), Armadillidium vulgare (N = 2), Pomatias elegans (N = 2), Acheta domesticus (N = 6), Drosophila melanogaster (N = 12 rep), 2 tests per individual, 1 per temperature	20, 30 °C	2	60
Experiment 4: Mass–respiration relationship in aquatic invertebrates	Echinogammarus sp (40 individuals, 1 test per individual)	16 °C	4	60
Experiment 5: Thermal performance curve of respiration	Otiorhynchus pyrenaeus (N = 35 individuals, 11 tests per individual, 1 per temperature)	18, 22, 26, 29, 32, 35, 37, 38, 39, 40, 41 °C in a semirandomized sequence	4	90

The experiments were done at the Station d’Ecologie Théorique et Expérimentale between July 2022 and July 2025. We tested the SensorDish Reader with one extension set with diverse terrestrial and aquatic invertebrates either sourced from lab husbandries (fruit flies and crickets) or captured from the wild near the Station (all other species). Experiments followed French ethics regulations on animal use. See Supplementary Appendix 1 for more information about invertebrate care.

Long-Term Stability and Repeatability of Respiration

Experiment 1: Long-term stability of respiration.

We used 36 juvenile crickets (A. domesticus) to test for the stability of the respiratory metabolism measurement over time, at 2 temperatures (20 or 30 °C) using 4 ml SensorVials and 2 SensorDish readers (the basic set plus one extension). Crickets were introduced individually in SensorVials before the experiment by gently pushing them with a paint brush. For each SensorDish, we prepared 18 replicates containing one cricket and 6 control replicates without crickets. SensorDishes were placed inside 2 Panasonic MLR-352-PE incubators set at the 2 temperatures in dark conditions, and we measured the pO₂ (% air saturation) values every minute for 5 h. After that, we stopped the measurement, removed the vials, and checked for dead or unwell individuals. Two individuals died during the test and were removed from the analyses.

Experiment 2: Repeatability of respiration

We tested the repeatability of respiratory metabolic rate across days on 40 adult firebugs (P. apterus). Individuals were weighted with a 10⁻⁵ g precision scale to get biomass and to derive biovolume from a biomass–biovolume relationship (see Supplementary Appendix 1). After 1 d of acclimatization to laboratory conditions, we measured respiratory metabolism twice a day (at 9 AM and 13:30 PM) for 3 d on the same individuals (ie 6 measurements per individual) using 4 ml SensorVials. Fourty firebugs were distributed into the 2 SensorDish readers, with 4 vials per SensorDish serving as control replicates without invertebrates. SensorDishes were placed inside a Panasonic MLR-352-PE incubator set at 20 °C in dark conditions, and we measured the pO₂ (% air saturation) values every minute for 2 h. After that, we stopped the measurement, removed the vials, and checked for dead or unwell individuals. No individual died or became unwell during the tests. Individuals were released into their housing falcon tubes to rest until the afternoon session, for which we used a different set of SensorVials. After the afternoon experiment, we cleaned all vials with ethanol, rinsed them with osmosed water, predried them with a paper towel, and finished drying overnight inside a 45 °C oven.

Mass–Temperature–Respiration Relationships in Terrestrial and Aquatic Arthropods

Experiment 3: Mass–temperature–respiratory metabolism relationship

We tested for the mass- and temperature-dependency of respiratory metabolism across several arthropod species using 2 ml SensorVials. We captured firebugs Pyrroccoris apterus, wolf spiders Pardosa sp, ants Iberoformica subrufa, woodlice Armadillidium vulgare, snails Pomatias elegans and sourced fruit flies Drosophila melanogaster and crickets Acheta domesticus from cultures. We measured respiration for N = 40 individuals (or groups of individuals in the case of fruit flies, for which we created 6 replicates with one individual, 2 with 2 individuals, 1 with 3 individuals and 1 with 4 individuals to check for the ability to detect very low oxygen concentration changes for these very small species). For each of the 2 SensorDish readers, we prepared 20 replicates containing invertebrates (1 snail, 1 woodlouse, 3 crickets, 3 firebugs, 3 wolf spiders, 2 ants, and 6 fruit fly replicates), and 4 control replicates without invertebrates. SensorDishes were placed inside 2 Panasonic MLR-352-PE incubators set at either 20 or 30 °C in dark conditions, and we measured the pO₂ (% air saturation) values every minute for 2 h. In the afternoon, we repeated the measurements exchanging vials between temperatures so that individuals that were measured in the morning at 20 °C were then measured at 30 °C and vice versa. The afternoon reading lasted only for 1 h and a half. After each reading, we checked for dead or unwell individuals. Only fruit fly replicates presented some dead individuals, probably because the fruit fly cultures were quite old. Individuals from the other species were all alive and well. Individuals were weighted with a 10⁻⁵ g precision scale in the evening of the experiment to get biomass and to derive biovolume from a biomass–biovolume relationship (see Supplementary Appendix 1).

Experiment 4: Mass–respiratory metabolism relationship in an aquatic arthropod

In order to assess whether this method works similarly for terrestrial and aquatic invertebrates, we used respirometry data collected on 40 individuals of amphipods Echinogammarus sp sampled from 4 populations in the Adour river basin (43°22′32″N, 0°36′39″W, 43°22′28″N, 1°41′10″W, 43°19′55″N, 0°52′23″W, 43°21′16″N, 1°38′17″W). Over 1 d of experiment, we measured the respiratory metabolism of 40 amphipods at 16 °C. Individuals were placed in 4 ml SensorVials vials filled with Lez river water and added onto 2 SensorDish readers set inside a Panasonic MLR-352-PE incubators set at 16 °C. We devoted 4 vials per SensorDish to be controls empty of individuals filled with Lez river water, randomizing their position within the SensorDish. The control vials are particularly important in aquatic settings to control for background microbial respiration of microorganisms in the water. Respiratory metabolism was measured in dark conditions as O₂ (µmol/l) every minute for 2 h. After the measurements, individuals were oven-dried at 60 °C and weighted with a 10⁻³ g precision scale to get biomass and to derive biovolume from a biomass–biovolume relationship (see Supplementary Appendix 1).

Respiratory Metabolic Rate Calculations

Data files were exported from PreSens SDR software v4.0.0, and analyzed in R v4.4.1, using the tidyverse package collection v2.0.0. We first eyeballed the relationships between oxygen concentration and time for each sample to detect any inconsistency (eg sudden drop or rise in oxygen concentration that breaks the oxygen-time tendency, Supplementary Fig. S1). Samples with large inconsistencies were removed from the dataset. From each sample, we removed the first hour of measurement from our analysis as the acclimation period, allowing for the air in the vials to reach the temperature in the incubator, and the individuals to acclimate to their new thermal environment (Supplementary Figs S1 and S2).

The slope S of the oxygen concentration over time (O₂ proportion airsat/h) were calculated using a linear model for each replicate. We then calculated the average slope of the control replicates $\overline{S_{\mathrm{cont}}}$ for each trial and SensorDish reader. Such $\overline{S_{\mathrm{cont}}}$ measurements allow for removing background respiration, particularly important for aquatic samples for which there could be microbial respiration in the water. Finally, based on the ideal gas law, we calculated respiratory metabolic rates $I$ in µmol O₂ h⁻¹ using

$$I=\frac{P_{0}V(-1\times (S-\overline{S_{\mathrm{cont}}}\left)\right)}{RT}$$

where $P_0$ is the partial pressure of ambient O₂ in Pa, thus 20.95% of the atmospheric pressure of 1.01325 × 10⁵ Pa at sea level, and here for demonstration, also corrected for altitude (thus 96% of atmospheric pressure, as the station is located at 350 m of altitude), $V$ is the volume of the sample tube in m³ minus the biovolume of the assessed individual (see Supplementary Appendix 1), $R$ is the gas constant (8.314 J K⁻¹mol⁻¹) and $T$ is the temperature in Kelvin (Scafaro et al. 2017). Note that the equation is approximate, as it does not take into account the dilution of oxygen partial pressure by the addition of CO₂ partial pressure from the animal’s metabolism, nor by water vapor produced by the animal. Note also that the correction of $P_0$ for altitude is mainly for demonstration here as the effect of altitude would be negligible compared to daily variation of air pressure due to weather variables, but could be more important for researchers doing experiments at higher altitudes. For some very rare samples, negative respiratory metabolic rate measurements could be obtained when the signal is too low to detect any respiration (eg fruit flies in experiment 3, 2 one-individual samples at 20 °C). Those samples were removed from further analyses. Last, we removed samples for which individuals died or became visibly unwell during the respiratory metabolism measurement.

Statistical Analyses

In experiment 1, we investigated the stability of respiratory metabolism measurements over 5 h of acquisition (4 h after exclusion of the acclimation phase) at 20 and 30 °C. Slopes of oxygen concentration (% air saturation) were calculated for eight 1-h time windows, each starting 30 min after the previous one. We calculated the repeatability of both these individual slopes and control slopes over the moving periods, discarding the first 2 periods as acclimation periods, using rptR v0.9.22 package. We included the type of slope (crickets versus control) as a fixed effect the variance of which was removed from repeatability estimates. Confidence intervals were estimated through 1,000 parametric bootstrap iterations, with 100 permutations being performed. Since repeatability at 20°C was found to be low, we corrected slopes of oxygen concentration over time by the slopes of measured temperature inside of the incubator that were found to vary over time. To do so, we subtracted from the slopes of oxygen concentration the predicted slopes from a linear model explaining slopes of oxygen concentration by temperature slopes. Repeatability was reassessed afterward. A second type of correction was correcting the slope of the cricket samples for the slope of the control samples. Repeatability was reassessed afterward. We further investigated whether time and temperature affected the stability of measurement by fitting a linear mixed model using the lme4 v1.1-35.5 and lmerTest v3.1-3 packages with oxygen concentration slope value as the response. Fixed effects included interactions between moving period and test temperature (20 or 30 °C) and between moving period and type of slope (crickets versus control) to test whether within-individual variation stemmed from mechanical or biological sources. We included all centered and scaled simple effects and we fitted individual identity as a random intercept.

In experiment 2, we calculated the repeatability of respiratory metabolism over several days. The rptR model accounted for a Gaussian distribution and was adjusted by including the batch number and the test number as fixed effect factors, the variance of which were excluded from the computation of repeatability. Individual ID was fitted as a random effect. In a second time, we mass-corrected the respiratory metabolic rate as fully described below for experiment 3, and computed repeatability on this mass-corrected variable.

In experiment 3, we studied the dependence of respiratory metabolism to body mass. We calculated the allometric scaling exponent α from the expected relationship between respiratory metabolism and body mass $I=I_0{M}^{\alpha }$ (Brown et al. 2004, Glazier 2005), where α is the scaling exponent and $I_0$ a normalization constant independent of body size, through a linear model of log respiratory metabolic rate by log body mass. We then mass-corrected respiratory metabolic rates $I_{\mathrm{cor}}$ through multiplying $I$ by $M^{-\alpha }$. Further, respiratory metabolic rates increase exponentially with temperature following the Boltzmann–Arrhenius model, such as the joint effect of body size and temperature T (in K) can be described as $I=I_0{M}^{\propto }{e}^{-E/kT}$, with E being the activation energy, and k the Boltzmann’s constant (Brown et al. 2004, Clarke, 2004). Thus, we calculated the activation energy of through a linear model on the form: $\ln \left(I{M}^{-\propto }\right)=\ln \left(I_{\mathrm{cor}}\right)=-E\left(\frac{1}{kT}\right)+\ln \left(I_0\right)$. We also calculated $Q_{10}$ values as $Q_{10}={\left(I_{\mathrm{cor}2}/I_{\mathrm{cor}1}\right)}^{10/(T_2-T_1)},$ where $I_{\mathrm{cor}2}$ and $I_{\mathrm{cor}1}$ correspond to the mass-corrected respiration values at the temperatures $T_2$ (here 30 °C) and $T_1$ (here 20 °C) for which we assessed respiration.

In experiment 4, we investigated the relationship between mass and respiratory metabolic rates in an aquatic arthropod. The allometric scaling exponent α was estimated following the method in experiment 3.

Technical Validation Results

Long-Term Stability and Repeatability of Respiration

Respiration was both repeatable over a 5-h long measurement session at 2 temperatures (R [CI] = 0.25 [0.08, 0.40] and 0.77 [0.61, 0.86] at 20 and 30 °C, respectively, Supplementary Fig. S2), and over 6 separate measurements sessions over 3 d at 20 °C (r = 0.79 [0.69, 0.86], Supplementary Fig. S3). Mass-correcting respiration rates through a linear model of log respiration by log body mass for the 3-d experiment led to a lower, but still strong repeatability (r = 0.41 [0.26, 0.54], Supplementary Fig. S3). This suggests that while an important part of the intraspecific variation in respiration rates is due to differences in body mass, individual differences subsist when considering such variation. During the 5-h measurement session, the repeatability at 20 °C was lower than the other repeatabilities, which was due to variation in incubator temperature over time (Supplementary Fig. S2). Correcting for the slope of temperature variation led to higher repeatability (R[CI] = 0.70 [0.51, 0.81], Fig. 2, Supplementary Fig. S2). Such effect of temperature variation highlights the need for control samples, as correcting the respiration of invertebrate samples for the slope of the control samples also yielded a very strong repeatability (R [CI] = 0.71 [0.50, 0.83], Supplementary Fig. S2). Further, the 5-h session showed a trend toward lower oxygen consumption over time in insect samples, which was particularly marked at high temperatures (Fig. 2, Supplementary Tables S1 and S2), while control samples stayed close to an expected zero slope. This suggested a downregulation of metabolism of crickets over time due to acclimation to the high temperature.

Fig. 2.

Stability of temperature-corrected oxygen concentration slopes in crickets Acheta domesticus in experiment 1 across time. Points represent 1-h moving reverse slope of respiratory metabolism of separate samples, with measurements for the same samples being linked by a line. One-hour windows start at 30 min intervals, with the first hour of measurement being removed as an acclimation period (Supplementary Fig. S3). More positive values represent stronger respiratory metabolism.

Mass–Temperature–Respiration Relationships in Terrestrial and Aquatic Arthropods

Respiratory metabolism of terrestrial arthropods depended on their body mass such that the allometric coefficient α was 0.48 ± 0.06 SE (t = 7.94, df = 59, P = 7e⁻¹¹, R² = 0.52, Fig. 3A). Removing the snails (Po. elegans), for which the calcified shell represented an important, yet inert, part of body mass, led to a higher allometric coefficient (α  =  0.56 ± 0.06 SE, t = 8.97, df = 55, P = 2e⁻¹², R² = 0.59). Respiratory metabolism of aquatic amphipods showed an allometric coefficient α of 0.39 ± 0.08 SE (t = 4.91, df = 32, P = 3e⁻⁵, R² = 0.43, Fig. 3C).

Fig. 3.

Body mass and temperature dependency of respiration over several species of terrestrial (A, B) and aquatic (C) invertebrates. A) Relationship between respiratory metabolism and body mass expressed in log–log scale over a range of terrestrial invertebrate species. B) Relationship between mass-corrected respiratory metabolism and temperature in the same terrestrial species. C) Relationship between respiratory metabolism and body mass expressed in log–log scale in aquatic Echinogammarus sp. Temperature is expressed as $\left(\frac{1}{kT}\right)$ with k being the Boltzmann’s constant and T the temperature in Kelvin. Points represent the respiratory metabolism of separate individuals (or groups of individuals for fruit flies Drosophila melanogaster).

Mass-corrected respiration of terrestrial arthropods increased with temperature, such that the activation energy E was 0.54 ± 0.15 (t = 3.62, df = 59, P = 6e⁻⁴, R² = 0.18, Fig. 3B), with a $Q_{10}$ value of 2.1.

Protocol for a Thermal Performance Curve of Respiration

We present a working example using the SensorDish reader to create thermal performance curve of metabolism using snout beetles. We provide methods and results, and present pitfalls and guidelines to help guiding researchers through creating thermal performance curves of metabolism for terrestrial invertebrates.

Methods

We captured 35 adult snout beetles (Otiorhynchus pyrenaeus) from a semi-natural mesocosm facility located nearby the Station (the Metatron (Legrand et al. 2012)). Individuals were weighted with a 10⁻⁵ gram precision scale to get biomass and to derive biovolume from a biomass-biovolume relationship (see Supplementary Appendix 1). After a week of acclimation to the lab conditions, we measured thermal performance curve of respiratory metabolism.

We tested each individual for respiratory metabolism at 11 temperatures (18, 22, 26, 29, 32, 35, 37, 38, 39, 40, 41 °C) over 12 days using 4 ml SensorVials. The sequence of temperatures was chosen to place the very stressful high temperatures at the end of the test sequence, with 3 d of rest between 40 and 41 °C, 2 d of rest between 39 and 40 °C, and 1 d of rest between 38 and 39 °C, while the milder temperatures were done 2 per day in a sequence of 29–32, 18–35, 26–37, 22–38. Each day at 9 AM, snout beetles were introduced individually in 4 ml SensorVials by gently pushing them with a paint brush. Thirty-five snout beetles were distributed into the 2 SensorDish readers, with 6 vials per SensorDish serving as control replicates without invertebrates and one spot left empty due to a missing vial. Control vial position was randomized within the SensorDish so as to avoid systematic bias due to potential temperature gradients within the incubator. SensorDishes were placed inside 2 Panasonic MLR-352-PE incubators set at the desired temperatures in dark conditions, and we measured the pO₂ (% air saturation) values every minute for 2 h. After that, we stopped the measurement, removed the vials, and checked for dead or unwell individuals. In the afternoon, we repeated the experiment with a second temperature (except for high temperature tests). After the afternoon experiment, we cleaned all vials with ethanol, rinsed them with osmosed water, predried them with a paper towel, and finished drying overnight inside a 45 °C oven.

Statistics

We calculated metabolic rates following the “respiratory metabolic rate calculations” section above. A difference with the previous experiments was the length of the acclimation period for which we removed data. Indeed, because the temperatures were quite high, acclimation to the temperature was longer and we thus removed the first 90 min from our analysis instead of the first 60 min. We further tested the sensibility of our results to the acclimation cutoff by varying the acclimation window. Another difference with the methods above was the removal of unwell or dead individuals. Indeed, we did not remove those individuals for the thermal performance curve as we expected that temperatures past the thermal optimum for respiratory metabolic rate would be very close to the critical thermal maximum and removing these individuals would remove the drop in respiratory metabolic rate.

Extracted respiratory metabolic rates were then mass corrected by calculating the allometric scaling exponent α from the expected relationship between respiratory metabolism and body mass $I=I_0{M}^{\alpha }$ (Brown et al. 2004, Glazier 2005), where α is the scaling exponent and $I_0$ a normalization constant independent of body size, through a linear model of log respiratory metabolic rate by log body mass. We then mass-corrected respiratory metabolic rates $I_{\mathrm{cor}}$ through multiplying $I$ by $M^{-\alpha }$.

We calculated thermal performance curves of mass-corrected respiratory metabolism. We fitted the Sharpe–Schoolfield model with high temperature inactivation only (Schoolfield et al. 1981) to the thermal performance curve, which can be described as follows:

$$I_{\mathrm{cor}}=\frac{I_{\mathrm{ref}}{e}^{\frac{E}{k}(\frac{1}{T_{\mathit{ref}}}-\frac{1}{T})}}{1+e^{\frac{E_h}{k}(\frac{1}{T_h}-\frac{1}{T})}}$$

where I_cor is the mass-corrected respiratory metabolic rate, T is the temperature in Kelvin, k is the Boltzmann’s constant, E is the activation energy, E_h characterises the temperature-induced inactivation of enzyme kinetics above T_h where half the enzymes are rendered non-functional and, I_ref is the rate normalized as an arbitrary reference temperature T_ref of 15 °C where no low or high temperature inactivation is experienced. The following equation yields a maximum respiratory metabolic rate at an optimum temperature where metabolic rates are fastest:

$$T_{\mathrm{opt}}=\frac{E_h{T}_h}{E_h+k{T}_h\mathrm{l}\mathrm{n}(\frac{E_h}{E}-1)}$$

To do that, we used the nls_multstart function from nls.mulstart v1.3.0 package with the sharpeshoolhigh_1981 from the rTPC v1.0.4 package, and we bootstrapped the model refit with the minpack.lm function from the minpack.lm v1.2-4 package with the Boot function from the car v3.1-3 package to calculate confidence intervals.

Results

The thermal performance curve of respiration of the snout beetles was well represented by a Sharpe–Schoolfield model with high temperature inactivation (Fig. 4, Supplementary Table S3). The model led to estimate an activation energy E = 0.36 [0.30, 0.45], with a thermal optimum T_opt = 39.33 [39.26, 39.40] °C and a sharp inactivation energy past the thermal optimum of E_h = 4.60 [1.90, 6.83]. Rerunning the model with different acclimation windows gave very similar results (Supplementary Table S4).

Fig. 4.

Thermal dependency of respiratory metabolism of snout beetles Otiorhynchus pyrenaeus on the range of 18 to 41 °C. Points represent the respiratory metabolic rate of each individual at each temperature, and the line and confidence interval represent the fit from a Sharpe–Schoolfield model (Supplementary Table S3).

Designing a Thermal Performance Curve Experiment: Guidelines and Pitfalls

Designing a thermal performance curve experiment does come with some pitfalls. A first step is the choice of temperatures. Ideally, researchers should have an idea of the critical thermal maxima of the studied species. If this is not the case, previously test some individuals at different high temperatures for a few hours to get an idea of thermal maxima. Researchers should choose at least 8 separate temperature points, ideally more, to be able to fit thermal performance curves. Because thermal optima of respiratory metabolism are often quite close to critical thermal maxima, and because the exponential phase is often more easy to capture than the thermal optimum, it will be important to sample more finely close to the expected thermal optimum.

A second important choice is the order of the temperature sequence. Although researchers often assume that the shape of the thermal performance curve is invariant to prior thermal experience, with both an independence of the TPC to the order of temperature exposure, to the time of exposure, and no effect of prior exposure to temperatures above the T_opt or close to the CT_max and CT_min on TPC, these assumptions are often found to be false (Sinclair et al. 2016). Nevertheless, a perfect design does not exist, and researchers may have to make practical choices to run experiments. When researchers measure performance of the same individual at multiple temperatures, the order of the temperatures is often randomized, although the highest temperatures are often chosen last to avoid heat-shock responses affecting performance at other temperatures (Feder and Hofmann 1999, Ruiz-Aravena et al. 2014, Sinclair et al. 2016). Here, we chose to both randomize the order of lower temperatures, put the higher temperatures at the end of the sequence, and increase the time of recuperation from half a day between milder temperatures, to 1 d for hot temperatures and up to 3 d for the hottest temperature. This avoided overexerting individuals. We further removed individuals that were in a bad shape prior to the experiment, although we allowed for keeping information from individuals that died during the test as we considered that their reached their CT_max. We believe that such an approach may mitigate effects of prior temperature experience, although it may be interesting to test for the effect of several randomization techniques. Reassuringly, a study comparing “repeated static-temperature respirometry,” in which the same individual is measured at different temperatures, “independent static-temperature respirometry,” in which independent groups of individuals are measured at each temperatures, and “dynamic temperature respirometry,” in which the animal is exposed to a continuous increase or decrease of temperature during measurement, finds that crickets exposed to those 3 methods present similar Q10 values (Lake et al. 2013).

A third difficulty may be measuring biovolume. Depending on the aim of the experiment, and the researchers’ need for nondestructive sampling or not, biovolume may be measured destructively for each individual at the end of the experiment, or assessed from a biomass–biovolume regression from a subset of individuals as we did here. If researchers do not need nondestructive sampling, we believe that the best way would be euthanize and weigh the individuals, then measure water displacement (see Supplementary Appendix 1 for an example).

Another caveat, particularly when researchers have multiple extension sets, and can measure up to 240 individuals at a time, is that the preparation of each individual takes time. This should lead to differences in acclimation time between individuals. Researchers should mitigate this by working in batches, and taking into account the time of preparation of each batch when removing the acclimation period.

Although the respirometer readings were quite precise, it is important to include empty control vials to control for changes in measured O₂ concentration values other than those caused by respiration. This is important for 2 reasons. One, because oxygen sensors are temperature-dependent, changes in temperature lead to changes in measured oxygen concentration values. Thus, the SDR will only show correct oxygen values if the measurement temperature entered by the user matches the actual sample temperature. Thus, particularly when vials are prepared at room temperature and then put at higher temperatures, some time is necessary for the vial air temperature to reach the desired temperature. Not allowing for enough acclimation time would thus lead to incorrect slopes detected by control samples (Supplementary Figs S1 and S3). Second, and particularly for aquatic invertebrates, background respiration can occur when oxygen is consumed by other microorganisms present in the chamber, such as bacteria in the water (Svendsen et al. 2016, Killen et al. 2021). It is thus very important to have control vials with the same water as the water in the measured samples to correct for this background respiration. However, and particularly for very high temperatures where bacterial growth may be of importance even within a relatively short 2-h time span, subtracting the respiration in the controls may not be enough to properly remove background bacterial respiration in water. This is because microbial growth in the controls may differ from the growth in vials with organisms, as organisms release nutrients. It may be wise in this case to remeasure baseline respiration of water samples after removing the invertebrates to properly correct for microbial respiration (Killen et al. 2021). Lastly, and again for aquatic trials, it is important to verify that the cleaning is thorough so as to avoid organic residues inside the vials affecting subsequent microbial growth. A good practice would be to run regular “blank” trials between the runs, with all vials treated as controls filled with water to detect any bacterial growth that would bias the results.

We propose as an appendix (Supplementary Appendix 2) and as Supplementary Material a set of guidelines, complete with templates and R scripts, to measure thermal performance curves. We further propose the full datasets and R scripts for the 5 experiments proposed here (4 technical validation experiments and 1 thermal performance curve experiment) for readers to try the method (Supplementary Material).

Discussion

Our study demonstrates the applicability of a commercial system originally designed to measure respirometry of laboratory cell cultures (the SDR) to the measurement of respiration of a wide range of small terrestrial and aquatic arthropods after minimally adapting the protocols. All of the results obtained with this setup are consistent with general aerobic metabolic models. For example, the mass–metabolism relationships showed allometric coefficients α of 0.39 to 0.56 that were within the broad distribution of values found for invertebrate species (Glazier 2005), although a bit low compared to the 0.75 coefficient expected by allometric scaling theories (Kleiber 1932, Brown et al. 2004). The temperature–metabolism relationship found an activation energy E of 0.54 ± 0.15, close to the range (0.60–0.70 eV) of activation energies commonly reported for aerobic respiration (Gillooly et al. 2001), with a more recent meta-analysis showing a mean activation energy of 0.66 ± 0.15 eV with a median of 0.55 eV (Dell et al. 2011). Q₁₀ value was 2.11, which was in line with values often found for insects (Hoffmann 1985). Further, the system was able to capture the full thermal performance curve of metabolism, up to the drop in respiration past the thermal optimum.

The experimental setup was reliable in measuring respiratory metabolism, showing both intrameasurement repeatability during a 5-h measurement period and across-measurement repeatability through 6 measurements over 3 d. Repeatability over 3 d was very high (0.78), which was due in part to intraspecific variation in firebug body size, with a mass corrected repeatability of 0.41, in line with results from a meta-analysis on endotherms (Auer et al. 2016). The authors found that repeatability of free-living animals (0.42 [0.30, 0.54] median [95% CI]) was lower than those from laboratory-housed animals (0.65 [0.52, 0.80]), with repeatability declining with increasing interval duration (−0.22 [−0.36, −0.08]). The 5-h measurement period showed high repeatabilities as well, although the repeatability at 20 °C was lower than at 30 °C. This lower repeatability stemmed from a variation in temperature inside of the 20 °C incubator over time, leading to changes in measured oxygen concentration values as oxygen sensors are temperature-dependent (Supplementary Fig. S3). Correcting for the slope of temperature variation during the period led to much higher repeatability. This underlines the importance of using the system in very controlled temperature settings to capture correct oxygen concentration values, and thus the need for a good incubator, or at least to properly measure temperature variation to correct for it later at the analysis step. However, it is to note that here we measured repeatability of raw oxygen slope data, while in practice, we would measure actual air respiration by subtracting the control values. Thus, in practice, this issue would be mitigated by the fact that the “control” vials would be subjected to exactly the same temperature as the “invertebrate” vials, barring any spatial variation within the incubator that can be further mitigated by randomizing the position of the controls within a plate. Subtracting the controls thus allows to account for such microvariations in temperature over time that would change oxygen readings, as we demonstrate here. A second observation was that cricket respiration tended to decrease over time during the 5-h window, particularly at 30 °C. It is likely that the downward trend was due to an acclimation to temperature over time, with individuals downregulating their metabolism after the initial shock. Another important possibility to consider for the lowered respiration over time would be hypoxia. Indeed, closed respirometry systems such as this one can lead to both hypoxia and hypercapnia over time. Thus, they may not be suited for longer measurements, particularly in aquatic ecosystems and if the animal is used to generate hypoxic conditions (Steffensen 1989, Svendsen et al. 2016). Here, the lowest value of air saturation was close to 75% (Supplementary Fig. S3), making hypoxia a less likely factor. Depending on the biological question tested, it is thus important to consider both the potential for hypoxia and for acclimation to temperature when measuring respiratory metabolism. However, choosing relatively short-term duration of experiments, such as in our experiments, with vials with enough free air volume compared to the volume of the individual measured, and checking the decrease of air saturation values during the experiment should prevent any decrease of metabolism due to hypoxia and CO₂ build-up. Indeed, when testing for the effect of different acclimation cutoffs in our thermal performance curve experiment on snout beetles, we demonstrate that even at very high temperatures, measuring respiration early or late during the experiment did not meaningfully change the slope of the thermal performance curve, suggesting that metabolism was not suppressed.

We further showed that the system was fit to accommodate respiration readings for a large range of arthropods, both terrestrial and aquatic. The system was even able to detect respiration of fruit flies. Although measurements of one individual alone were too low to reliably detect a signal, grouping from 2 to 4 individuals together led to measurable respiratory activity in line with predicted mass–metabolism relationships. While individual readings of the smallest species may be difficult, grouping individuals together may be a reasonable accommodation. Such method is commonly used for small arthropods, such as for the mite Stratiotelaps scimitus (600 to 700 µm in length) with groups of 25 individuals in stop-low respirometry to get a signal (Meehan et al. 2022). However, keeping animals together may compromise standard metabolic rate calculation as animals may interact, with potential increases in metabolism due to stress. Conversely, it could be argued that for gregarious species, such as Drosophila melanogaster (Billeter et al. 2025), testing individuals alone may increase stress, and testing individuals as groups may be the more conservative approach. Correctly choosing the size of the vial may reduce pain points (Svendsen et al. 2016). An inconvenient of the system is that it lacks flexibility in chamber size, with only 2, 4, and 20 ml vials provided by PreSens. An option may be to introduce an inert death volume inert glass balls inside of the vials, for instance glass balls, to reduce air volume for the smallest invertebrates and allow for more flexible chamber sizes. Further, in the standard 2 and 4 ml configurations, the opening of the vials is only 8.2 mm, thus only arthropods smaller in diameter than that could be measured with this method. However, the 20 ml vial configuration allow for a larger opening (17.4 mm inner neck), although with a lower number of vials fitting all at once on the system (12 vials per SDR instead of 24).

One big advantage of the system is the possibility to parallelize measurements. For instance, a recent method paper for a high-throughput intermittent flow respirometer used for small aquatic animals ran only 20 respirometers overnight (Drown et al. 2020). In contrast, the SensorDish basic set of one 24-channel reader can be extended with up to 9 SDR extensions sets, leading to up to 240 samples monitored in parallel, reducing the total time needed to achieve a large sample size. We believe that the technology presented here will make feasible to gather intensive data on arthropod respiratory metabolism, making it possible to answer pressing ecological and evolutionary questions in a context of global climate change. Because of the modular nature of the system, with extension sets able to be placed at several temperatures in different incubators, a particularly interesting application is the creation of full thermal performance curves of respiratory metabolisms in only 1 or 2 d of measurements. We provide a helpful guideline and accompanying R scripts to help researchers perform experiments to calculate thermal performance curves of terrestrial invertebrates.

Author Contributions

Nicolas Mouret (Formal analysis [equal], Investigation [equal], Writing—original draft [equal], Writing—review & editing [equal]), Laura Kouyoumdjian (Investigation [equal], Writing—review & editing [supporting]), Chloé Chabaud (Investigation [supporting], Writing—review & editing [supporting]), Maxime Stanislawek (Investigation [supporting], Writing—review & editing [supporting]), Emma Couranjou (Investigation [supporting], Writing—review & editing [supporting]), Pau Molina-Vila (Investigation [supporting], Writing—review & editing [supporting]), and Elvire Bestion (Conceptualization [lead], Data curation [lead], Formal analysis [equal], Funding acquisition [lead], Investigation [equal], Methodology [lead], Supervision [lead], Writing—original draft [equal], Writing—review & editing [equal])

Supplementary Material

Supplementary material is available at Journal of Insect Science online.

Funding

This work was supported by funding from the Agence Nationale de la Recherche to EB (ANR-23-CE02-0009) and by the LABEX TULIP (ANR-10-LABX-0041).

Conflicts of Interest

The authors declare that the research was performed in the absence of any commercial or financial relationships that could be construed as a conflict of interest.

Data Availability

Raw data and R scripts are available on Zenodo, at https://doi.org/10.5281/zenodo.21295638.

The data share is consituted of two zipped folders:

- Analyses_exp1_to_5: folder containing data and code to re-create analyses in the manuscript

- Example_code_for_analyses: folder containing example scripts and protocols for researchers who want to adapt our methodology to their study.

Data should be cited as: Mouret et al. 2026.