Endocrinology and Metabolism

Direct calorimetry identifies deficiencies in respirometry for the determination of resting metabolic rate in C57Bl/6 and FVB mice

Colin M. L. Burnett, Justin L. Grobe


Substantial research efforts have been aimed at identifying novel targets to increase resting metabolic rate (RMR) as an adjunct approach to the treatment of obesity. Respirometry (one form of “indirect calorimetry”) is unquestionably the dominant technique used in the obesity research field to assess RMR in vivo, although this method relies upon a lengthy list of assumptions that are likely to be violated in pharmacologically or genetically manipulated animals. A “total” calorimeter, including a gradient layer direct calorimeter coupled to a conventional respirometer, was used to test the accuracy of respirometric-based estimations of RMR in laboratory mice (Mus musculus Linnaeus) of the C57Bl/6 and FVB background strains. Using this combined calorimeter, we determined that respirometry underestimates RMR of untreated 9- to 12-wk-old male mice by ∼10–12%. Quantitative and qualitative differences resulted between methods for untreated C57Bl/6 and FVB mice, C57Bl/6 mice treated with ketamine-xylazine anesthesia, and FVB mice with genetic deletion of the angiotensin II type 2 receptor. We conclude that respirometric methods underestimate RMR in mice in a magnitude that is similar to or greater than the desired RMR effects of novel therapeutics. Sole reliance upon respirometry to assess RMR in mice may lead to false quantitative and qualitative conclusions regarding the effects of novel interventions. Increased use of direct calorimetry for the assessment of RMR and confirmation of respirometry results and the reexamination of previously discarded potential obesity therapeutics are warranted.

  • metabolic rate
  • thermogenesis
  • heat
  • calorimetry
  • respirometry

clinical treatment of obesity is currently restricted to behavioral modification (e.g., exercise) and a short list of pharmaceutical agents (Orlistat, Qsymia, and Belviq). These three FDA-approved compounds/mixtures all function by reducing caloric intake/uptake. Interestingly, no safe or efficacious pharmaceutical treatments have been developed to increase resting metabolic rate (RMR), which could be used as an adjunct or alternative therapy to the intake-suppressing drugs currently available. In the 1930s, a compound that relieves the mitochondrial proton gradient called 2,4-dinitrophenol (DNP) was used extensively as a pharmacological stimulator of metabolic rate. Despite the great effectiveness of this compound to increase resting metabolism, use of DNP can result in cataracts and fatal hyperthermia, and therefore, it is no longer considered safe for human use (2). Substantial ongoing work in the field is aimed at understanding the mechanisms of RMR regulation, but useful pharmacological targets remain elusive. We hypothesize that a major problem facing the field is an inadequate ability to accurately detect small changes in RMR, which leads researchers to prematurely disregard some potential therapeutic targets and wastefully pursue other ineffective targets.

The most common method for assessing RMR in laboratory rodents is respirometry or the estimation of metabolic rate by measuring respiratory gases (oxygen consumption and carbon dioxide production) (3, 6, 8). Based on empirical data from chemical reactions performed in the 1920s (7), an equation relating gas exchange to heat production was constructed, and this equation is used by most major commercially available calorimetry systems. The validity of respirometry results depends on a series of untested and/or untestable assumptions (5, 6, 8). The most important among these are that (5) 1) test animals utilize fuels with the same stoichiometry as farm animals studied in the original experiments from 1900 to 1940, 2) net substrate interconversion (e.g., lipogenesis from glucose, gluconeogenesis from protein, or ketogenesis from triglycerides) is negligible and constant, 3) the total body CO2 pool is constant (e.g., normal and stable blood pH and bicarbonate), 4) energy transfer from protein oxidation is low, constant, and specifiable, and 5) anaerobic metabolism is negligible. One may reasonably expect that genetically modified mice, which are nearly ubiquitous in modern experimental physiology and pharmacology (and especially those models with direct modifications to mitochondrial function), will violate at least one of these many assumptions (5).

To empirically test the accuracy of respirometry methods in mice (and thus indirectly assess the validity of the many assumptions that underlie its use), we constructed and utilized a direct calorimeter to assess real-time total heat dissipation. Coupling the effluent air stream from this direct calorimeter apparatus to a respirometer allows for simultaneous measurement of RMR by both techniques. We herein present data that document a substantial underestimation of RMR by respirometry in the widely used C57Bl/6 and FVB mouse strains at baseline and with pharmacological and genetic manipulations of these strains. These data lead us to conclude that although respirometry is effective for detecting large changes in RMR, the quantitative accuracy of respirometry in mice is questionable. We hypothesize that because of the large magnitude of quantitative underestimation of RMR by respirometry (∼10–12%), the effects of novel pharmacological RMR modulators may be missed when this method is used alone.


Direct calorimeter.

A custom gradient layer, Seebeck style, direct calorimetry chamber was fabricated by Dr. Heinz F. Poppendiek of Geoscience. The chamber's inner dimensions are 10 × 10 × 10 cm, and the walls of the chamber are water-jacketed. Thermopile DC output was amplified ×500 with DC high-pass and 5-Hz low-pass filtering (ETH-256; iWorks) before digitization (NI USB-6008; National Instruments). A white light-emitting diode was installed through a pore in the lid of the box to supply normal circadian light cues, and influent and effluent gas channels were centered on opposing faces of the box. Influent and effluent gas compositions were measured using sensors (temperature/humidity: SHT15, Sensiron; pressure: BMP085, Bosch; flow: EM1NL1R0V, Sensiron) interfaced to a custom data acquisition program (programmed in LabView; National Instruments). The mouse was elevated by an acrylic platform that covered gas influent and effluent ports from the mouse and acted as a diffuser to promote mixing of chamber air (Fig. 1).

Fig. 1.

Combined direct calorimeter and respirometer. Air from a standard laboratory supply line was bubbled through distilled water to achieve 100% relative humidity at room temperature. This humidified air was then passed through a series of condenser columns to reduce the temperature to ∼5°C to set the dew point and thus the relative humidity. Air was subsequently reheated to 30°C by passage through a copper coil submerged in a water bath. This conditioning resulted in an air supply with stable temperature (30°C) and relative humidity (∼19%) regardless of time of day and year. After conditioning, the pressure, humidity, and temperature (PHT sensor) and mass flow (EM1 sensor) were then measured before passage into the interior of a 10 × 10 × 10 cm water-jacketed (30°C) chamber lined with thermopiles. Within the chamber, mice are suspended upon an acrylic platform, and light is provided by a white light-emitting diode suspended from the lid. Effluent air PHT and mass flow (EM1 sensor) were again measured, and the positive-pressure air stream [flowing at ∼300 ml/min at standard temperature and pressure (STP)] was sampled by a negative-pressure sampling line (sampling at ∼250 ml/min at STP) with balance vented to the laboratory. The sampled effluent air passed through a 50-ml CaSO4 (Drierite) column before subsequent analyses by carbon dioxide and oxygen analyzers and mass flow (EM1 sensor) confirmation.

Heat dissipation (in kcal/h) was calculated using the following equation (6, 8): Embedded Image where Qm denotes heat detected by the thermopile, Qa denotes heat of dry air, and Qw denotes heat carried by water vapor. Changes in Qa denote sensible heat loss, and changes in Qw denote insensible heat loss.

Qm was calculated from the thermopile output voltage. Calibration was initially performed using a 10.97-Ω resistor (including the resistance from wires connecting the resistor inside the calorimeter) and a variable potentiometer to ensure linearity of the relationship between heat applied and calorimeter output voltage. Specific voltages (0.0000, 0.8374, and 1.6197 V) were then applied to the resistor multiple times (to yield 0.0000, 0.0657, and 0.2457 kcal/h, respectively), and calorimeter output voltage was recorded (slope = 0.0244231 ± 0.0000110 kcal·h−1·μV−1, P < 0.001). As reported previously by Walsberg and Hoffman (10), this relationship was confirmed to be linear (r2 = 0.9999998).

Standard temperature and pressure (STP) correction of flow rate (FRSTP) was performed using the following equation: Embedded Image where FR is the uncorrected flow rate (in ml/min), P is the air pressure (in Pa), and T is the air temperature (in °C).

Qa was determined by the following equation: Embedded Image where Te is the effluent air temperature (in K), Ti is the influent temperature (in K), Cp,m is the constant pressure molar heat capacity of air (29.19 J·mol−1·K−1), and FRi is the STP-corrected influent air flow rate (in ml/min). The unit conversion constant k is 1,344,000 ml·s−1·mol−1·min−1. The output of the Qa equation (in W) was converted to kcal/h by multiplying by 0.86042065 kcal·h−1·W−1. Throughout, we used the thermochemical calorie (1 calorie = 4.184 J), not the international steam table calorie (1 calorie = 4.1868 J).

Qw was determined by the following equation: Embedded Image where FRe is the STP-corrected effluent air flow rate (in ml/min), Pvap,e is the water vapor pressure of the effluent air (in Pa), Pe is the effluent air pressure (in Pa), FRi is the STP-corrected influent air flow rate (in ml/min), Pvap,i is the water vapor pressure of the influent air (in Pa), and Pi is the influent air pressure (in Pa). The unit conversion constant k is 1.34042 × 10−5 g·min−1·ml−1·s−1. The output of the Qw equation (in W) was converted to kcal/h by multiplying by 0.86042065 kcal·h−1·W−1.

Water vapor pressures (Pvap,e and Pvap,i in Pa) were calculated using the following equations, which include calculation of the saturation pressure of water by the Antoine equation: Embedded Image Embedded Image where T is the air temperature (in K) and RH is the relative humidity (in %). The unit conversion constant k is 133.322 (in Pa/mmHg).

Hvap is the heat of vaporization of water (in J/g), which was calculated in real time by the following equation, which is a linear interpolation between 91192.5 Pa, 2,265.65 J/g, and 101,325 Pa, 2,257.92 J/g: Embedded Image where P is the air pressure (in Pa). This value was generally close to 2,260.4 J/g.

Influent air conditioner.

Influent air was supplied under positive pressure from a laboratory supply line, with pressure regulated to adjust air flow through the system (Fig. 1). The air was conditioned by bubbling through water (to achieve 100% RH) and then passage through a series of condenser columns to reduce the temperature to ∼5°C. Air then passed through a copper coil submerged in a water bath maintained at 30°C before entering the direct calorimeter. The desired effect of cooling saturated air (i.e., 100% RH) was to condense out excess water vapor and to set the dew point of the air after it was warmed. This conditioning system thereby supplied a relatively constant influent air stream into the direct calorimetry chamber of ∼30°C with a dew point of 5°C, which results in a ∼19% RH. Air flow into the chamber was roughly controlled through modulation of the pressure supplied to the conditioning system, and actual flow was recorded continuously using an EM1 mass flow meter. By conditioning the air in this manner, the air quality was controlled between experiments regardless of time of day and time of year, which are known to significantly affect air temperature and RH.


Effluent air from the direct calorimetry chamber was sampled at 250 ml/min STP, passed through a Drierite (anhydrous CaSO4) desiccant column, and analyzed sequentially for carbon dioxide (model CD-3A, AEI) and oxygen (model S-3A/II, AEI) content. Mass flow and gas concentration data were logged using a PowerLab with associated Chart software (ADInstruments). Heat production by respirometry was calculated using the equation derived from Lusk (7): Embedded Image

o2 was calculated as the change in O2 content of effluent air while a mouse was in the direct calorimetry chamber vs. baseline, multiplied by the rate of effluent air flow. Respiratory exchange ratio (RER) was calculated as the change in CO2 content divided by the change in O2 content.

Calibration of the oxygen analyzer was performed using a primary standard gas mixture containing 20.50% oxygen (Praxair; paramagnetic certification method). Two-point calibration of the carbon dioxide analyzer was performed using a soda lime column (resulting in 0.0000% CO2) and a primary standard gas mixture containing 5,000 ppm carbon dioxide (Praxair; flame ionization with methanizer certification method).

Nuclear magnetic resonance.

Body composition was assessed using nuclear magnetic resonance (NMR; Bruker LF90II) either ≥1 day preceding or on the afternoon following measurements in the combined calorimeter. Mice were restrained but not anesthetized for the rapid (∼1 min) analysis and then returned to home cages.


Wild-type male C57Bl/6J mice were obtained from The Jackson Laboratory, and male FVB/NCrl mice were obtained from Charles River Laboratories at 5–8 wk of age and allowed to acclimate to housing conditions before analyses. Animals were group housed in forced-ventilation racks in a room with ambient temperatures at 25°C and maintained on a 12:12-h light-dark cycle (6 AM to 6 PM) with ad libitum access to tap water and Teklad 7013 chow. Mice with genetic disruption of the angiotensin II type 2 receptor gene [AT2-knockout (KO) mice] maintained on the FVB background were obtained through a gift from Drs. Victor J. Dzau and Richard E. Pratt (Duke University). Notably, the “FVB” analysis group reflects n = 3 inbred mice obtained from vendors plus n = 3 wild-type littermate controls from the AT2-KO breeding colony, which has been extensively backcrossed onto the FVB background strain. All studies were approved by the University of Iowa Animal Care and Use Committee.

Stability of body masses was assessed by measuring body masses weekly in the various mice. C57Bl/6 mice exhibited an average gain of 0.040 ± 0.017 g/day for the 2 wk preceding the tests. FVB mice exhibited an average body mass gain of 0.105 ± 0.018 g/day for the 3 wk preceding the tests. AT2-KO mice on the FVB background gained an average of 0.085 ± 0.025 g/day for the 3 wk preceding the tests. On the day of analysis, animals were removed from their home cage (between 9 and 10 AM), and abdomens were massaged to evacuate the urinary bladder. Mice were then weighed and placed into the combined calorimeter. Heat and gas exchange were monitored continuously for 6 h without food or water. Data recorded during bouts of sleep (typically between 3 and 5 h of recording for conscious animals; Fig. 2) were then analyzed. A subset of animals was removed from the chamber and injected with a ketamine-xylazine anesthetic mixture and returned immediately to the chamber to assess the effect of anesthesia. Following recordings, animals were returned to home cages with ad libitum access to food and water.

Fig. 2.

Representative tracings of metabolic variables recorded by combined calorimetry. Variables and computed resting metabolic rate (RMR) values by respirometry and direct calorimetry are shown for representative mice from the 4 experimental groups. RMR calculated by respirometric methods (QLusk) relies upon 3 variables: rate of air flow through the test chamber, the rate of oxygen consumed (V̇o2; calculated by the change in effluent oxygen concentration multiplied by the flow through the chamber), and the respiratory exchange ratio (RER; calculated by dividing CO2 produced by O2 consumed). RMR calculated by direct calorimetry (QTotal) relies upon the sum of 3 heat components: heat conducted through the chamber wall (Qm), sensible heat (Qa; reflecting changes in dry air temperature), and insensible heat (Qw; reflecting evaporative water loss).


Comparisons of results from respirometry and direct calorimetry within subjects of each distinct treatment group were compared by paired t-test. Ages, body masses, lean masses, and RERs were compared with C57Bl/6 mice by one-way ANOVA followed by Tukey multiple-comparison procedures. The effects of anesthesia were assessed by two-way ANOVA (anesthesia, method) with repeated measurements, followed by Tukey multiple-comparison procedures. Treatment effects were also assessed using ANCOVA to account for body mass differences between groups. All comparisons were performed as two-tailed tests, with P < 0.05 considered significant.



Substantial effort was devoted to confirm accurate calibration of both the direct calorimeter and the respirometer. In brief, the direct calorimeter was calibrated, and calibrations were verified multiple times throughout the study using a 10.97-Ω resistor (including wire resistance), with stepwise voltages applied to produce power dissipation over the expected RMR range of mice (0.0–0.3 kcal/h). Oxygen and carbon dioxide analyzers were calibrated using primary standard gas mixtures (0.0000% CO2 and 0.5000% CO2-20.50% O2). Gas analyzer calibrations were confirmed by combustion of ethanol (observed RER 0.6662 ± 0.0041, 99.93% of the expected 0.6667, P = 0.915 by 1-sample t-test). All gas flows were STP-corrected by measurements of temperature, pressure, and mass flow in sampling lines during recording sessions.

Wild-type C57Bl/6J mice.

Under thermoneutral testing conditions (30°C), adult male C57Bl/6 mice at 12 wk of age exhibited an RMR of 0.160 ± 0.004 kcal/h by direct calorimetry. In contrast, an RMR of 0.139 ± 0.005 kcal/h was estimated by respirometry (P = 0.0005 by paired t-test). This equates to a 13.3 ± 2.7% underestimation of RMR by respirometry (Table 1). Normalization of RMR values to total or lean body mass has no effect on this relationship, as both results are divided within subject by the same factor.

View this table:
Table 1.

Body compositions, RER, and rates of underestimation by respirometry in male mice on C57BI/6 or FVB background

Changes in core temperature can positively or negatively influence heat dissipation measurements by direct calorimetry. Therefore, we determined the magnitude of core temperature drift by colonic probe in mice during typical analysis periods (3–5 h in the chamber) under similar conditions (singly housed, no food or water access, acrylic floor, and in an opaque enclosure). Core temperatures were determined during exposure to 30.0°C (n = 8; 29.02 ± 0.88 g) or room temperature (∼25°C) in littermate-matched mice (n = 16; 29.56 ± 0.87 g, P = 0.698 vs. 30°C group). After 3 h of exposure, when steady-state conditions appear to be established by calorimetric methods (Figs. 2 and 3A), core temperatures in each group were similar (25°C group, 37.49 ± 0.24 vs. 30°C group, 37.22 ± 0.36°C, P = 0.527). To examine the core temperature changes in the mice during the expected steady-state phase (between 3 and 5 h of exposure; Figs. 2 and 3A), we again measured core temperatures in the same mice. Following a total of 5 h of exposure, core temperatures were again very similar between groups (25°C group, 37.41 ± 0.23 vs. 30°C group, 37.06 ± 0.23°C, P = 0.348). Importantly, comparing the change in core temperature within each animal, there were no statistically significant differences in the changes within or between groups (25°C group, −0.09 ± 0.11 vs. 30°C group, −0.16 ± 0.52°C; group effect P = 0.377, time effect P = 0.540, and group × time interaction P = 0.862).


A cohort of mice was anesthetized by intraperitoneal injection of ketamine (400 mg/kg) and xylazine (50 mg/kg) before analysis in the combined calorimeter. The heat production by anesthetized mice as estimated by respirometry reflected an 8.5 ± 5.2% (P = 0.08) underestimation of the heat detected by direct calorimetry (Table 1). Importantly, anesthesia had a significant suppressive effect on RMR when assessed by direct calorimetry, but no significant effect of anesthesia was detected by respirometric methods (Fig. 3).

Fig. 3.

Comparison of RMR measurement methods in C57Bl/6 mice in conscious and anesthetized states. A: sample simultaneous RMR values calculated by respirometry and direct calorimetry for a C57Bl/6J male mouse. B: sample simultaneous RMR values calculated by respirometry and direct calorimetry for a C57Bl/6J male mouse following induction of anesthesia by intraperitoneal injection of ketamine and xylazine. C: correlation of RMR values by respirometry vs. direct calorimetry from conscious (n = 12) and anesthetized (n = 10) C57Bl/6J male mice. Dashed line highlights expected function if results were identical between methods. Note the general leftward, as opposed to downward, shift from the conscious to anesthetized state. D: data from C redrawn to illustrate both the underestimation of RMR by respirometry regardless of treatment and the failure of respirometry to detect a significant effect of anesthesia. All grouped data are presented as means ± SE. *P < 0.05 vs. conscious mice within analysis method; †P < 0.05 vs. respirometry within conscious or anesthetized state.

Wild-type FVB/NCrl mice.

Mice of the C57Bl/6 background strain are known to be highly sensitive to diet-induced obesity and harbor a single copy of the renin gene (renin-1c). In contrast, the FVB strain is resistant to diet-induced obesity and harbors two distinct versions of the renin gene (renin-1d and renin-2). Therefore, we compared the RMR of these two strains when analyzed by respirometry and direct calorimetry. Heat production by FVB mice was underestimated by ∼10.1 ± 2.4% by respirometry (Table 1 and Fig. 4). RMR of FVB mice was substantially higher than C57Bl/6J mice by both methods despite similar total and lean body masses (Fig. 4B). To account for possible differences in RMR due to body mass or body composition variances between strains, metabolic rates were compared after normalization to lean mass (Fig. 4C) or by ANCOVA (Fig. 4D), but the difference in results between strains and between methods remained significant regardless of normalization procedure.

Fig. 4.

Comparison of RMR measurement methods in C57Bl/6 and FVB mice. A: sample simultaneous RMR values calculated by respirometry and direct calorimetry for an FVB/NCrl male mouse. B: RMR measurements by respirometry and direct calorimetry for FVB/NCrl (n = 5) and C57Bl/6J mice (n = 12). C57Bl/6J mice replotted from Fig. 3. C: RMR values for mice in B normalized against lean body mass (determined by NMR). D: RMR analyses for mice in B adjusted using ANCOVA procedures. All grouped data are presented as means ± SE. *P < 0.05 vs. FVB/NCrl mice within analysis method; †P < 0.05 vs. respirometry within strain.

Genetic disruption of the angiotensin II type 2 receptor.

RMRs of mice with genetic disruption of the AT2 receptor (AT2-KO) maintained on the FVB background were compared with age- and lean-mass matched wild-type FVB mice (Table 1). Respirometry underestimated RMR by ∼5.7 ± 3.7% in AT2-KO mice (Fig. 5 and Table 1). Most notably, respirometry failed to detect a significant reduction in RMR by these animals compared with the FVB background strain. This failure was present in raw RMR values (kcal/h; Fig. 5B), RMR values normalized to lean body mass (Fig. 5C), or RMR values adjusted by ANCOVA procedures (Fig. 5D).

Fig. 5.

Comparison of RMR measurement methods in mice with genetic disruption of the angiotensin II type 2 receptor [AT2-knockout (KO)]. A: sample simultaneous RMR values calculated by respirometry and direct calorimetry for an AT2-deficient male mouse on the FVB/NCrl background strain. B: RMR measurements by respirometry and direct calorimetry for FVB/NCrl (n = 5) and AT2-KO (n = 6) mice. FVB/NCrl mice replotted from Fig. 4. C: RMR values for mice in B normalized against lean body mass (determined by NMR). D: RMR analyses for mice in B adjusted using ANCOVA procedures. All grouped data are presented as means ± SE. *P < 0.05 vs. FVB/NCrl mice within analysis method; †P < 0.05 vs. respirometry within strain.


The need for validation of respirometry methods and practical issues with regard to the application of “total” calorimetry (simultaneous use of direct plus respirometry methods) were recently reviewed thoroughly by Kaiyala and Ramsay (5), who argued in favor of the reevaluation of the methods of assessing RMR. In a carefully controlled study, Walsberg and Hoffman (10) examined the accuracy of respirometry in multiple species, including the kangaroo rat (Dipodomys merriama Mearns), dove (Columbina inca Lesson), and quail (Coturnix communis Linnaeus), by comparing simultaneous outputs from animals studied with both direct and respirometry methods. Those authors concluded that when disparate species were studied under various conditions that estimations of heat production by RER-based respirometry calculations led to errors averaging 21% for kangaroo rats, 15% for doves, and 17% for quail. Here, we add to this list an average of ∼10–12% underestimation of RMR by respirometry for the commonly used C57Bl/6 and FVB substrains of laboratory mice (Mus musculus Linnaeus). Perhaps more importantly, we have also demonstrated that respirometry can lead to errors in both quantitative and qualitative conclusions regarding the effects of pharmacological and genetic interventions upon RMR even within substrains of a single species.

We hypothesize that the primary flaw in simple respirometric methods [measuring only oxygen and carbon dioxide exchange, which is undoubtedly the most common commercially available method, (3, 6, 8)] stems from the inability of this method to detect anaerobic and protein metabolic rates. Estimations of protein metabolism are typically made through the measurement of nitrogenous waste in urine and feces, but accurate assessments require complicated chemical analyses or expensive elemental analyzers that are beyond the capabilities of the gross majority of respirometry users. Furthermore, urine/fecal measurement of these nitrogenous products lacks sufficient temporal resolution, and therefore, it is impossible to assess “resting” vs. “total” protein metabolism. Finally, anaerobic metabolism, which is completely ignored by respirometry, is also of growing interest with the recent appreciation of the multifaceted impacts that the gut microbiome, which operates in a largely anaerobic environment, has upon whole body metabolism.

There are two major design issues that complicate the use and interpretation of data collected with our current calorimeter system. First, we specifically designed our system so that animals would not have access to food or water during the testing session. These restrictions were made to simplify the calculations of heat loss due to evaporation (QW) and to ensure similar fasting states of the animals tested. Measurements of RMR were typically calculated between the 3rd and 5th hours of the recording session (sessions were always performed during the middle of the light phase, when food intake is normally low), which would therefore represent a mild fast. From real-time calculations of RER (Fig. 2A and Table 1), we also know that the animals were in a metabolic steady state during this recording time frame. Second, to mimic a high-throughput paradigm (thereby more likely to be adopted by the wider obesity research field), we chose not to implant core body temperature radiotelemeters. Unfortunately, by not implanting such devices, we are incapable of assessing changes in heat retention by the animal during the testing period in real time.

However, using the same reasoning of Walsberg and Hoffman (10), we conclude that core body temperatures of the animals after several hours of recording at thermoneutrality are likely to be in a steady state and that the observed differences in RMR by direct calorimetry and respirometry are too large to be explained by changes in core temperatures alone. For example, the observed RMR differences between methods here would require a core body temperature change of −1.86°C/2 h. This value is calculated first by multiplying the steady-state 0.022 kcal/h difference between direct calorimetry and respirometry (Fig. 2D and Table 1) by 2 h of recording time (between the 3rd and 5th hour of recording), which yields a total difference of 0.044 kcal. Assuming the specific heat capacity of a mouse with ∼10% body fat is 0.874 kcal/kg K [interpolated from Blaxter (1)], and using the average group mass of 27.01 g (Table 1), this yields a difference of −1.86°C in 2 h. In contrast, we have examined empirically the total core temperature change over the same time frame of a subset of C57Bl/6 mice exposed to 30.0°C and determined that the average change is −0.16°C, which represents only 8.6% of the required −1.86°C change. Furthermore, this may be an overestimation of the core temperature loss caused by exposure to a 30°C ambient temperature, as mice exposed simultaneously to standard room temperature conditions (thereby capturing normal circadian changes) lost 0.09°C. We conclude that core temperature changes could at most account for a very small fraction (∼0.0019 kcal/h, or 8.6%) of the observed 0.022 kcal/h difference in RMR by the two methods. Altogether, subtracting the 0.0019 kcal/h contribution of core temperature changes from the 0.022 kcal/h difference between methods would change our reported rate of “RMR underestimation by respirometry” from 13.3 to 12.3%. Nonetheless, we acknowledge the advantages that simultaneous core temperature measurements would yield, including both possibly explaining a very small fraction of the underestimation by respirometry and providing further proof that animals are in a thermal steady state. Advancements in core temperature measurement technology (to reduce cost, surgical requirements, etc.) would be welcome for the ultimate commercialization and widespread adoption of this methodology. Regardless, we advocate real-time assessments of core temperature changes during recording sessions (via radiotelemetry) for all future studies. A listing of other major pros and cons for the use of respirometry vs. direct calorimetry is presented in Table 2.

View this table:
Table 2.

Major pros and cons for the use of direct calorimetry vs. respirometry for the assessment of RMR

Another procedural complication that may have a minor effect on our data interpretation stems from the fact that we did not explicitly “train” the mice by repeated or prolonged pretest exposure(s) to the calorimetry chamber. To assess whether serial testing sessions would alter the observed differences in RMR by respirometry and direct calorimetry, a conscious male C57Bl/6J mouse was repeatedly tested daily for 5 consecutive days. On the 1st testing day, a 0.0233 kcal/h (11.8%) difference between methods was observed (direct calorimetry, 0.1968 kcal/h; respirometry, 0.1734 kcal/h). On the 5th consecutive day of testing, RMR by direct calorimetry had decreased by 0.0139 kcal/h, and RMR by respirometry had decreased by 0.0081 kcal/h. In addition, the RER slowly drifted upward over the five serial recording days (day 1, 0.84 vs. day 5, 0.94), suggesting both a shift in substrate utilization toward carbohydrates and a concomitant reduction in V̇o2. Therefore, after 5 consecutive days of testing, the difference between methods decreased by 15% to 0.0175 kcal/h (indicating “only” a 9.6% underestimation by respirometry on the 5th serial recording day compared with 11.8% on the 1st day). The mouse studied lost on average 0.135 g of total body mass per day for the 5 testing days, and this trend continued for at least 3 days following the recording on the 5th day. In contrast, lean body mass did not change over this same period. These data also support the presence of a stress response to serial testing and highlight the confounding effect that serial measurements have on data interpretation. We conclude that lack of training may contribute a small fraction of the observed quantitative differences in methods but that respirometry faithfully underestimates RMR even when substantial training is performed. These findings support “training” for further improved accuracy of RMR measurements by either method and bolster the concept that psychological stress (e.g., exposure to a novel testing chamber) can have a minor but measurable effect on metabolic rate even during sleep. Future technical refinements that would allow for housing the animal in the testing chamber continuously for several days may help to address and better understand the effect of training procedures on RMR measurements.

The major findings of this study are summarized as follows. First, respirometric methods based only upon oxygen and carbon dioxide exchange measurements underestimated the RMR of C57Bl/6 and FVB mice by ∼10–12%. Second, the inaccuracy of respirometry may lead to incorrect quantitative and qualitative assessments of the effect of various interventions (and novel therapeutic compounds) upon RMR in vivo. Clinically, 2,4-dinitrophenol (a dangerously strong pharmacological stimulant of RMR) can be used to cause ∼2 lbs. weight loss/wk in humans at a dose of 500 mg/day, which results from a roughly 50–60% increase in RMR (9). This should be seen as an extreme weight loss regimen given the potentially fatal side effects (2, 9), and thus a more reasonable target of pharmacological stimulation of RMR would be a 5–10% increase. Here, we demonstrate that the rate of inaccuracy of respirometry is roughly 10–12% and posit that this magnitude of inaccuracy, given the target range, is unacceptably large. We conclude that the challenges faced by the obesity therapeutics research community in identifying or validating novel therapeutic targets in mice (and likely other species as well) may be compounded by the inappropriate yet almost universal and sole reliance upon respirometry. Because construction of our current direct calorimeter system can cost substantially less than the high-quality respirometry equipment used herein, direct calorimetry may also be a more affordable technology for new users, especially if complete systems become commercially available. More widespread use of direct calorimetry and the reexamination of previously discarded novel therapeutics for the modulation of RMR by this method are warranted.


C. M. L. Burnett was supported through the University of Iowa Medical Student Research Program. This work and J. L. Grobe were supported through a National Heart, Lung, and Blood Institute K99/R00 award (HL-098276) and Program Project Grant (HL-084207).


The authors have no conflicts of interest, financial or otherwise, to disclose.


C. M. B. and J.L.G. contributed to conception and design of the research; C.M.B. and J.L.G. performed the experiments; C.M.B. and J.L.G. analyzed the data; C.M.B. and J.L.G. interpreted the results of the experiments; C.M.B. and J.L.G. prepared the figures; C.M.B. and J.L.G. drafted the manuscript; C.M.B. and J.L.G. edited and revised the manuscript; C.M.B. and J.L.G. approved the final version of the manuscript.


We gratefully acknowledge husbandry support by the University of Iowa's Office of Animal Resources.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
View Abstract