Reply to Yamada et al.: Questions and answers to the validity of the doubly labeled water method in high-fat and sucrose-feeding mice irrespective of obesity proneness.

REPLY: we have reported recently that 1) mice fed a diet rich in saturated fat and sugar (HFS), irrespective of whether they are obesity prone or resistant, have an overestimated rate of CO2 production (rCO2) when calculated by the doubly labeled water (DLW) method compared with actual gas exchange using indirect calorimetry (IC) equipment and 2) a two-pool model approach, depending on the timing of sampling the initial and the final, showed the least discrepancies for calculating rCO2 (8). These findings are challenged in a letter to the editor by Yamada et al. (25). A key issue in our debated study (8), which was not addressed at all in the letter by Yamada et al. (25), is that we did not find inconsistencies in rCO2 between the two methods in mice fed a low-fat (LF) diet (8). Although this is one indication that our methodology is sound, it is possible that the overestimation of rCO2 in the HFS condition is applicable only to mice. But what if it were applicable to larger animals, including humans, too? In light of these findings, we concluded that “caution needs to be addressed when using the DLW method in humans and animals feeding a HF diet, because MR may be overestimated” (8). Below, we give a point-by-point reaction to the issues raised by Yamada et al. (25). The first point refers to our observation that a two-pool model “fits the data better” than a one-pool model. As originally described by Lifson and McClintock (14), an assumption underlying the DLW method is that the isotopically labeled water (H and O), after being introduced into the body, does not dilute beyond the body water pool. Since the dilution space of H is typically 1–4% larger than the dilution space of O and both are larger than the total body water pool when assessed by desiccation, this assumption is apparently not met, and other dilution pools and/or routes of elimination are thus in play. Therefore, one would expect that a two-pool model equation is a priori superior to a one-pool model, knowing that a two-pool model weighs the magnitudes of two separate routes of elimination in the final rCO2 estimation rather than one route of elimination. In small animals ( 1 kg), however, researchers tend to find that a one-pool model “fits the data better” than a two-pool model, although the differences are relatively small (19, 20) or nonexistent (2, 6). The answer to why differences are found relative to expectations remains elusive to date. If the underpinnings of this “rule of thumb” are not clear in the first place, it is of course quite impossible for us to give an answer as to why we find conditions in which the two-pool model approach is shown to be least discrepant in estimating rCO2 in our study (8). Rather than acknowledging the fact that this “rule of thumb” is apparently not rock solid, Yamada et al. (25) point out that we “did not mention this, test, or even discuss possible reasons for this discrepancy” and suggest that our data are “specific to [our] own laboratory”. We contend that more validation studies aimed at the heart of the matter should be done to clarify mechanisms underlying inconsistencies in the first place (20). Second, the suited approach to validate rCO2 estimated by the DLW method is to compare it with levels obtained by gas exchange measurements in an IC setting. We strongly oppose the view of Yamada et al. (25) that our IC “reference method is questionable” and that our IC data “are so discrepant from the rest of the literature.” We performed IC measurements exactly according to the methodology reported by Visser et al. (22), and that study was cited by us in our debated report (8). Specifically, the mass-flow controllers were calibrated with a soap foam flow meter (Bubble-O-Meter, La Verne, CA) before and after the trials, showing little variation over time (i.e., 1%). The infrared respiration gas analyzer was calibrated daily with two certified gas standards (AGA Gas, Amsterdam, The Netherlands), spanning the observed CO2 gas concentrations between 0 and 0.5%. Daily adjustments of the span of the CO2 gas analyzer were very small and were typically 1% of the certified CO2 concentration. Therefore, we estimate the maximum overall error of our gas respiration method to be 2%. Comparable methodologies were applied for O2 calibration. We did not perform the ethanolburning procedure for calibration as proposed by Yamada et al. (25), which is commonly applied in human studies (24). Although such methodology has certain advantages, downscaling the calibration procedure of quantitative ethanol combustion to small animal IC equipment can be problematic (21). We do agree with Yamada et al. (25) that the respiratory quotient (RQ) of the mice in our study is higher than would be expected based on the estimated food quotient (FQ). Elevated RQ levels are, among others, reported as a consequence of exercise (10), lipogenesis (1), and stress (23), and these conditions are not included in estimation of the FQ. Particularly the latter two may have been relevant for our study. First, the frequent tail blood sampling procedure probably induced an increased level of discomfort in the mice, which subsequently increased RQ in all diet groups. Increased de novo lipogenesis would be expected to occur particularly in mice fed a HFS diet (i.e., the RQ for the conversion of glucose to fat is 5.55, whereas the RQs for oxidation of fat and glucose are 0.7 and 1, respectively), which would increase RQ further in this diet group above the FQ. Address for reprint requests and other correspondence: G. van Dijk, Nijenborgh 7, 9747AG, Groningen, The Netherlands (e-mail: [email protected]). Am J Physiol Endocrinol Metab 305: E1181–E1183, 2013; doi:10.1152/ajpendo.00504.2013. Letter To The Editor