INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

OBESITY IS AN INCREASINGLY prevalent disorder in the United States, with >50% of all Americans being considered overweight and ~30% considered obese (14). Furthermore, weight loss is difficult to achieve and maintain, with a 95% failure rate of obese individuals for maintaining the weight loss for >4 yr (13). Many insights into the pathophysiology of human obesity have been obtained in mice in which genetic mutations have been identified (2, 20, 26). Several mutations associated with the development of severe obesity include ob/ob (9, 32), db/db (15, 29), cpe/cpe (21), and tubby (12, 22). Strains of mice that are susceptible to obesity when challenged with a high-fat diet have also been identified (30). Moreover, with the advent of transgenic technology to overexpress or knock out genes, there has been an escalation of murine models exhibiting not only obese (18, 24), but also lean, phenotypes (10, 16, 25). Phenotypes alone, however, may not readily identify mechanisms. Measurements of energy balance (energy intake and expenditure) of these murine models are limited and are needed to provide valuable insights into the pathophysiology of human obesity. However, the indirect calorimetry systems used to measure energy expenditure in rodents are often not adequately described or validated in the literature.

Indirect calorimetry involves the measurement of CO₂ produced and O₂ consumed by an organism (4, 5, 17) and assumes that all O₂ is used to oxidize fuels and all CO₂ produced is recovered. These measurements are in turn used to calculate the metabolic rate (MR) and the respiratory quotient (RQ) of the organism (6). MR, an absolute measurement, is a composite of an organism's energy output during periods of rest and activity. RQ, on the other hand, is a relative measurement of carbohydrate and lipid oxidation and has theoretical limits between 0.7 and 1.0, with exceptions being exercise and overfeeding (28). As carbohydrate oxidation increases (i.e., during feeding), RQ approaches 1.0, and as lipid oxidation increases (i.e., during fasting), RQ approaches 0.7 (8). Likewise, animals fed high-carbohydrate diets have higher RQs than those fed high-fat diets (7, 23).

By far the most difficult aspect of indirect calorimetry is the measurement of O₂. Moreover, the abundance of O₂ (20.95%) relative to CO₂ (0.03%) in ambient conditions dictates that measurement errors of O₂ have greater implications on the calculations of MR and RQ. Furthermore, measurements of O₂ can be affected by external (i.e., barometric pressure, humidity, and temperature) and internal factors within a system (i.e., variations in flow rate and pressure). Because compressed air is not feasible for use with animals in long-term experiments, changes in ambient conditions are nearly unavoidable.

Hence, the goal of this work was to design an indirect calorimetry system that cannot only accurately measure ambient O₂ concentrations and O₂ consumption (O₂) by an animal but, if necessary, adjust and self-correct to changes in ambient conditions. This work presents the design and validation of such a system.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The indirect calorimetry system consists of cages, pumps, flow controllers, valves, and analyzers (Fig. 1) and is computer controlled to sequentially measure the O₂ and CO₂ concentrations in four separate cages on a continual basis. The system operates as follows. Air taken from a common air source is pulled through four metabolic chambers (Metabowl, Jencons Scientific, Bridgeville, PA), a blank cage, and a reference line connected via separate, but identical air-sampling pathways. Mice are placed in four metabolic chambers (1 mouse in each), while the blank cage is used as a reference cage to monitor ambient O₂ and CO₂ concentrations periodically. The cages are designed to separate food, urine, and feces for precise determinations of energy intake, urinary nitrogen, and unabsorbed calories, respectively. The condensation in the air exiting the chambers is removed with electronic sample coolers (Universal Analyzers, Carson City, NV). At this point, the air is pushed with pumps (model 107CAB18, Thomas Industries, Sheboygan, WI) through mass flow controllers (Teledyne Hasting-Raydist, Hampton, VA), which maintain constant, but adjustable airflow (0.25-2.0 l/min). The air then travels to a manifold containing five valves. At predefined intervals, the computer sends a 5-V signal to close a specified valve, corresponding to a particular cage, thereby shunting air through the O₂ and CO₂ analyzers (Oxymat/Ultramat 6, Siemens, Roswell, GA). An optically isolated switch is used between the computer and valves to prevent a voltage surge from traveling back to the computer. Before the air enters the O₂ analyzer, pressure valves are used to regulate the air pressure on the reference line at 3 psi and on the sample line at 1 psi. Because each analyzer is differential and compares the O₂ and CO₂ levels in the cage airstream with levels in a reference line, air is used as the calibration gas to zero the analyzers. The span calibrations for the O₂ and CO₂ analyzer are set using O₂ (1% balance N₂) and CO₂ (0.8%), respectively, as primary gas standards (Air Liquide, Houston, TX). The zero and span calibrations are checked several times until each analyzer reads precisely at the given concentration. The time of measurement, differential CO₂ and O₂ concentrations, flow rate, CO₂ output (CO₂), O₂, RQ, and MR (Weir equation) are calculated and stored in a computer configured with data acquisition hardware (Analogic, Wakefield, MA) and software (Labtech, Wilmington, MA). The software is based on Windows 9X/NT and uses a flexible "drag-and-drop" interface to collect data from the analyzers and flow controllers at 60 Hz for a total of >1 × 10⁶ data points in a given 24-h period. Each cage, with an approximate volume of 1.6 liters, is sampled for 2 min, with a 30-s washout period between cage readings. The transit time of the air from a cage to the analyzer is 15 s. Uniform gas mixing within the cages was observed by flow visualization with smoke. The linear range of each analyzer was limited to a differential reading of 0-1%, with a 4- to 20-mA output. To test for air tightness, all tubing, fittings, and connections between the cages and outlets of the analyzers are checked to maintain a pressure of 3-5 psi. As a preventive measure, each pump is rebuilt or replaced after 6 mo of continual use. Uninterruptible power sources (American Power Conversion, West Kingston, RI) are used to blunt electrical noise by preventing electrical spikes and providing electricity to the system during brownouts.

View larger version (24K): [in this window] [in a new window]   Fig. 1.   Schematic diagram of a self-correcting indirect calorimetry system. Air is pulled from a common air source through 4 cages with animals, 1 blank cage, and a reference line through air dryers, which remove humidity in the airstream. All pathways are separate but identical in length and are checked for leaks by pressurizing the system. Five pumps then push the air through flow controllers, which dynamically maintain airflow at set flow rates. The air from the cages then enters a manifold with pneumatic switches, which vent to the atmosphere when opened and, when closed, direct the airstream to the analyzers. The computer sequentially closes 1 valve at a time for 2.5 min, and, after a 30-s washout period, CO2, O2, and flow rate are measured.

To test the self-correcting ability of the indirect calorimeter system, the following experiments were performed. With the use of a butane burn to give a known RQ (0.62), N₂ gas was infused into the air supply to simulate changes in ambient air concentrations. An additional O₂ analyzer (Magnos 6G, Hartmann & Braun, Bartlesville, OK) was employed to monitor the changes in the ambient O₂ induced by the N₂. Data (CO₂ and O₂) were collected from the two cages (cage with butane burn and an empty cage) for 30 s, with a 30-s washout allowed between readings. Differential measurements of CO₂ and O₂ from the cage containing the butane burn were adjusted in a point-to-point fashion by taking into account the differential measurements from the blank cage, which were transformed using the following smoothing function: y(i) = (1  k) * x(i) + k * y(i  1), where k = 0.85. The adjusted CO₂ and O₂ were then used to correct MR and RQ. This technology was then applied to a typical experiment in which RQ was measured in a mouse for 3 days and a small decrease in the O₂ concentration occurred in the blank cage in the system.

To further assess the importance of the inclusion of a blank cage in the system for monitoring ambient conditions, mice (n = 95) of various ages, genetic strains (FVB/N, C57BL/6J, and C57BL/6Jx129/SVJae), body compositions, and nutritional states (i.e., fed high-carbohydrate or high-fat diets) and under various degrees of hypo- and hyperthyroidism were examined in the system. For each mouse, the coefficient of variation (standard deviation/mean * 100) was calculated for RQ and MR with and without correction for the data gathered in the blank cage. Because each mouse served as its own control, a paired t-test was used to analyze the coefficients of variation (SigmaStat version 2.03, SPSS, Chicago, IL).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The abundance of O₂ (20.95%) relative to CO₂ (0.03%) in ambient conditions dictates that measurement errors of O₂ have greater implications on the calculations of MR and RQ. To fully understand how small changes in ambient O₂ impact the calculations of MR and RQ, the equations for MR and RQ are given below. To illustrate this, two hypothetical conditions were tested. In the first condition, MR and RQ were calculated by holding constant all the variables (i.e., cage CO₂ concentration = 0.275%, cage O₂ concentration = 20.6%, ambient CO₂ concentration = 0.03%, flow rate = 0.5 l/min) and varying only the ambient O₂ concentration from 20.95 to 20.85%. When ambient O₂ concentration decreased by 0.5%, MR decreased in a linear fashion, by ~24% (Fig. 2A). Conversely, when the same conditions were applied to the equation for RQ, there was an inverse relationship between ambient O₂ concentration and RQ, such that a change of 0.5% in ambient O₂ concentration resulted in the full theoretical range of RQ (0.7-1.0; Fig. 3A). In the second condition, MR and RQ were calculated by holding constant all variables as given above, except ambient CO₂ concentration was varied from 0.0300 to 0.0299% and ambient O₂ concentration was set at 20.95%. In both equations, varying ambient CO₂ concentration over the same magnitude of change as that of ambient O₂ concentration resulted in negligible changes in MR (Fig. 2B) and RQ (Fig. 3B). The MR (Weir equation; Ref. 3) is calculated as follows


where PCO_2cage and PCO_2amb are cage and ambient CO₂, PO_2cage and PO_2amb are cage and ambient O₂, k₁ = 1.106, and k₂ = 3.941. 

View larger version (16K): [in this window] [in a new window]   Fig. 2.   Theoretical plots of variations in ambient O2 (A) and CO2 (B) concentrations and corresponding impacts on the calculations of metabolic rate (MR). MR was calculated by holding constant all the variables (i.e., cage CO2 = 0.275%, cage O2 = 20.6%, ambient CO2 = 0.03%, flow rate = 0.5 l/min) and varying only the ambient O2 from 20.95 to 20.85%, or ambient CO2 was varied from 0.0300 to 0.0299%.

View larger version (18K): [in this window] [in a new window]   Fig. 3.   Theoretical plots of variations in ambient O2 (A) and CO2 (B) concentrations and corresponding impacts on the calculations of respiratory quotient (RQ). RQ was calculated by holding constant all the variables (i.e., cage CO2 = 0.275%, cage O2 = 20.6%, ambient CO2 = 0.03%, flow rate = 0.5 l/min) and varying only the ambient O2 from 20.95 to 20.85%, or ambient CO2 was varied from 0.0300 to 0.0299%.

The respiratory quotient is calculated as
A representative experiment in which ambient O₂ concentrations were changed is presented in Fig. 4, and mean data for RQ and MR are presented for four experiments in Table 1. Figure 4A shows the change in ambient O₂ as N₂ was infused into the general air supply for the system. RQ, adjusted and unadjusted, is plotted before, during, and after the infusion of N₂. In Fig. 4B, the differential PO₂ (corrected and uncorrected) and the differential PCO₂ for the empty cage are graphed.

View larger version (23K): [in this window] [in a new window]   Fig. 4.   Self-correcting ability of the indirect calorimeter system to measure RQ during a butane burn while N2 was infused into the system to decrease ambient O2 concentrations. A: change in ambient O2 (right ordinate) as N2 was infused into the general air supply for the system; RQ [corrected () and uncorrected ()] is plotted before, during, and after infusion of N2 (left ordinate). B: differential PO2 [untransformed () and transformed ()] and differential PCO2 for the empty cage. Transformed PO2 was smoothed using the following function: y(i) = (1  k) × x(i) + k × y(i  1), where k = 0.85. Each time point represents a 1-min interval.

In each of the experiments, ambient O₂ concentrations and RQ were stable in the indirect calorimetry system before the introduction of N₂ into the system. However, when the ambient O₂ concentrations decreased suddenly at the beginning of the N₂ infusion by ~1%, RQ dropped. This decrease in RQ was due to a slight increase in the differential measurement of O₂ in the blank cage (Fig. 4B). Likewise, when the N₂ was turned off, there was a corresponding increase in RQ, again because of a slight decrease in the differential measurement of O₂. During the infusion of N₂, there were minute changes in RQ, but these changes appeared to be related to the large fluctuations in the ambient O₂ concentrations and not to the absolute ambient O₂ concentration. There were no significant changes in the mean RQ of the butane burn before, during, and after the introduction of N₂ into the indirect calorimetry system (Table 1). Also, RQs from the butane burn were generally higher than the predicted value for combustion of butane (i.e., 0.62); however, when the RQs were adjusted for the slight offsets seen in the differential measurements of O₂ and CO₂ in the blank cage, RQs became appropriate. This further suggests that inclusion of a blank cage in the system can mathematically adjust for small variations in calibration of the analyzers and potential small drifts in the differential readings of O₂ and CO₂. As expected, the calculations of MR were also not impacted significantly by the alteration of ambient O₂ (Table 1). Because CO₂ and O₂ decreased during the butane burn, the mean MR before and after the N₂ infusion was compared with the MR during the N₂ infusion. Differential CO₂ was extremely stable and unaffected, despite the large perturbations in ambient conditions in these experiments.

The self-correcting ability of the indirect calorimetry system was then tested in a typical experiment with mice. A representative experiment is given in Fig. 5. Figure 5A shows the uncorrected RQ of a mouse in which measurements of RQ were normal until the 2nd day, when RQ increased abruptly. When decreases in O₂ observed in the blank cage on the 2nd day (Fig. 5B) were used to adjust O₂ in the cage with the mouse, RQ was normalized (Fig. 5C). Furthermore, typical diurnal changes in RQ, not seen in the uncorrected data (Fig. 5A), were now clearly evident in the corrected data (Fig. 5C). To further validate the importance of the blank cage and monitoring ambient conditions, the coefficients of variation were calculated for each mouse (n = 95) for RQ and MR with and without correction of the individual mouse data (i.e., O₂ and CO₂) with readings obtained in the blank cage. The coefficient of variation for RQ was significantly decreased when the blank cage readings were used to correct the data (9.9 vs. 8.3%, P < 0.00001). The coefficient of variation for MR was also significantly decreased by 2.3% (16.5 vs. 14.2%, P < 0.001).

View larger version (27K): [in this window] [in a new window]   Fig. 5.   A representative experiment showing RQ (uncorrected and corrected) of a mouse measured over a 3-day period. A: uncorrected RQ of a mouse in which measurements of RQ increased abruptly on the 2nd day. C: corrected RQ of the mouse when decreases in PO2 in the blank cage on the 2nd day were taken into account (B). Typical diurnal changes in RQ, not seen in the uncorrected data (A), are now clearly evident in the corrected data (C). A point-to-point correction was used to correct the O2 and CO2 measured in the cage housing the mouse by taking into account the differential measurements from the blank cage, which were transformed using the smoothing function in Fig. 4 legend. A line representing the smoothed data for PCO2 and PO2 is graphed as a thick line for each variable (B).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The technical requirements for an indirect calorimeter system are strict. As outlined by Ferrannini (6), these include: 1) an airtight system with appropriate flow to give O₂ and CO₂ concentrations within the linear range of the analyzers, 2) sensitive and stable O₂ and CO₂ analyzers for continuous measurements of expired air, 3) a method to remove moisture from the expired air before analysis, 4) computer hardware and software to capture and analyze the data, and 5) a calibration method using standard gas mixtures. As presented in our study, the calibration procedure should also include the periodic measurement of ambient O₂. In our system, this is accomplished with the use of differential analyzers that dynamically adjust for changes in ambient O₂ and CO₂ along with inclusion of a blank cage to adjust for instrumental drift. Several indirect calorimeter systems designed for human use have incorporated ambient air and calibration gas measurements into their routine operation (11, 27, 31). Although not evaluated for their ability to self-correct, indirect calorimeter systems for rats using absolute (19) and differential analyzers (1) along with monitoring of ambient conditions have also been described. However, many indirect calorimeter systems overlook the implementation of this critical measurement in their systems.

Correcting the RQ and MR for changes in ambient conditions monitored in the blank cage significantly increased the precision of the indirect calorimetry system. This represented a 2.3% increase in precision for the measurement of MR, and for an average mouse, this would represent 0.4 kcal/day. Evidence supporting the importance of this increased precision can be obtained in a study by West et al. (30), in which nine strains of mice were evaluated for their susceptibility to gain weight when fed a high-fat diet of condensed milk. Two strains in particular, although they consumed the equivalent amounts of food (~900 kcal/7 wk), gained different amounts of weight when fed the high-fat diet. The AKR/J strain mice of mice was obesity prone, gaining 13.3 g of body weight over the 7-wk diet period, while the SJL/J mice were obesity resistant and gained only 6.2 g. With the assumption that the absorption of food was identical between these two strains of mice, a difference in MR of 0.43 kcal/day could explain the differences in body weight. Another example of how small changes in MR could account for changes in body composition is seen in male transgenic mice overexpressing skeletal muscle lipoprotein lipase (10). In these experiments, an MR difference of only 0.28 kcal/day during the 13-wk period of high-fat feeding would explain the prevention of 3 g of carcass lipid seen in the transgenic mice compared with nontransgenic littermates. The increased precision of this system gained by correcting CO₂ and O₂ by any slight variations in ambient conditions would make it possible to measure these small, but very important differences in MR.

As previously mentioned, perhaps the most overlooked aspect of indirect calorimetry is the measurement of ambient O₂, and as we have shown, small variations in ambient O₂ have large impacts on the calculations of MR and RQ. Ironically, the most difficult aspect of indirect calorimetry is probably the measurement of O₂. Although both technologies used to measure O₂ and CO₂ (i.e., paramagnetic and infrared, respectively) are affected by environmental factors such as humidity, temperature, and barometric pressure, accurate measurements of O₂ in the range of O₂ measurements made during a typical calorimetry experiment may be more difficult to achieve (personal communication, P. Good, Siemens Process Analyzers, Rosswell, GA). Furthermore, the abundance of O₂ (20.95%) relative to CO₂ (0.03%) in ambient conditions dictates that the magnitude of change of O₂ and CO₂ in a typical experiment is not on the same scale. This is illustrated in the following hypothetical example. If an animal had an RQ of 0.80 and decreased ambient O₂ by 0.6% (i.e., from 20.95 to 20.35%), this would represent a decrease of only 3% from normal ambient O₂ concentrations. However, in the same example, CO₂ produced by the animal would change ~16-fold over ambient CO₂ concentrations (i.e., from 0.03 to 0.48%).

The indirect calorimetry system presented here was self-correcting in two ways. First, the differential nature of the analyzers was self-correcting even during severe changes in ambient O₂ concentrations, far greater than would be expected to occur during an experiment. Second, the inclusion of a blank cage in the system offered a means by which fluctuations in the differential measurement itself, because of slight interexperimental calibration settings and/or instrumental drift, could be mathematically adjusted for in the equations of MR and RQ. With this system, long-term measurements of the energy balance of the many murine models of leanness and obesity can be accurately assessed. Perhaps with this information, new insights into the treatment of human obesity can also be developed.