Factors Predicting Individual Variability in ... - Semantic Scholar

2 downloads 0 Views 231KB Size Report
Oct 13, 2011 - (Charles Austin Pumps) through a sealed Perspex chamber within an incubator (INL-401N-010, Gallenkamp) set at 30 °C. This temperature.
articles

nature publishing group

Integrative Physiology

Factors Predicting Individual Variability in Diet-Induced Weight Loss in MF1 Mice Lobke M. Vaanholt1, Victoria Magee1 and John R. Speakman1 The effectiveness of caloric restriction (CR) as a treatment for obesity varies considerably between individuals. Reasons for this interindividual variation in weight loss in response to CR may lie in pre-existing individual differences and/or individual differences in compensatory responses. Here we studied the responses of 127 MF1 mice to 30% CR over four weeks, and investigated whether pre-existing differences or compensatory changes in body temperature, resting metabolic rate (RMR) and behavior explained the variation observed in body mass (BM) and fat mass (FM) changes. Mice showed considerable variation in BM loss (36–1%), and in the type of tissue lost (FM or fat free mass, FFM). About 50% of the variation in BM and FM loss could be predicted by pre-existing differences in food intake, RMR, and general activity, where BM loss was greater when food intake was lower and activity and RMR were higher. Compensatory changes in activity and body temperature together explained ~50% of the variation in BM and FM loss in both sexes. In models incorporating baseline variables and compensatory changes, food intake, and activity were the strongest predictors of weight loss in both sexes; i.e., lower baseline food intake and increased changes in activity resulted in greater BM and FM loss. Interestingly, increased baseline activity was a significant predictor of weight loss independent of compensatory changes in activity. Identifying factors involved in individual variability in weight loss may give insights into the mechanisms that underlie this variability, and is important to develop individually tailored weight-management strategies. Obesity (2011) 20, 285–294. doi:10.1038/oby.2011.279

Introduction

Energy balance is regulated by processes that influence energy intake (i.e., food consumption) and energy expenditure (EE). Energy intake and the main components of EE, i.e., resting metabolism, physical activity, and thermoregulation, are regulated by an interaction of behavioral, physiological, and molecular mechanisms. Imbalances in energy state can result in health problems, like obesity. The prevalence of obesity and related diseases is rapidly increasing (1) and there is an urgent requirement for effective treatments. The most common intervention currently used is caloric restriction (CR) (i.e., dieting) (2). CR involves adequate intake of micronutrients but reduced energy intake, and in theory should lead to reductions in fat and body mass (BM). A complicating factor, however, is that individuals respond very differently to this intervention. Weight loss in individuals on the same diet intervention has been shown to vary enormously (3–5). For instance, in a dietary intervention consisting of 36 weeks provision of a 4.2 MJ/d low-fat high carbohydrate diet, weight loss varied from 4 to 22 kg (3). This large interindividual variation can result in unclear and contradictory results when effects are studied on a population basis, and an

understanding of the cause of this variation is necessary to develop intervention strategies that are tailored to the individual. Patients showing poor levels of weight loss have historically been classified as noncompliant, but it is becoming apparent that interindividual differences in weight loss may be the result of differences in genetic and/or physiological make-up between individuals (3–6). Studying this variability in weight loss is thus important and may reveal processes of individual regulation of energy balance. Reasons for unsuccessful weight loss may include pre-­existing differences in metabolic parameters between individuals. These may, for instance, include EE, capacity for fat oxidation, insulin sensitivity, and/or behavioral risk factors, such as low general activity or a tendency to gorge (3,5,7–9). Another major factor involved in the variation in weight loss under CR is the amount of compensation that occurs in EE that can offset the caloric deficit. Compensation involves reductions in EE (i.e., resting metabolic rate (RMR) or energy expended on activity), activity, body temperature (e.g., torpor), and/or increases in digestive efficiency (10–14). The development of individually-tailored intervention therapies for obesity depends critically on being able to predict

Integrative Physiology, Institute of Biological and Environmental Sciences, University of Aberdeen, Aberdeen, UK. Correspondence: John R. Speakman ([email protected])

1

Received 30 December 2010; accepted 8 August 2011; published online 13 October 2011. doi:10.1038/oby.2011.279 obesity | VOLUME 20 NUMBER 2 | february 2012

285

articles Integrative Physiology how individuals will respond to CR. It follows that para­meters measured prior to a CR intervention may be of more practical use than parameters that vary during compensation to CR. Although the pre-existing differences and compensatory responses may not be completely independent. The aim of the present study was to explore the relative importance of preexisting and compensatory responses on realised weight loss in response to a CR intervention. Animal models have proven useful tools in the study of obesity (15). For instance, animal models have been developed to predict whether rats are obesity prone or resistant to investigate physiological and behavioral factors that are involved in weight gain (16,17). These models have been successful in predicting weight gain on a fat diet and these rats show similar changes in metabolic profiles seen in obese humans. Here we used an outbred strain of mice, MF1, to investigate individual variability in diet-induced weight loss. Due to their genetic heterogeneity, MF1 mice vary considerably in their responses to CR and they develop age-related obesity on standard low-fat laboratory diets (10% kcal from fat); i.e., fat content is ~30% of BM at 6 months of age, compared to 10% in adult mice at 10 weeks of age (10). These mice therefore provide a suitable model to investigate variability in diet-induced weight loss when diet interventions are started at an older age. A major benefit of using animal models over humans is that because food supply is rigorously controlled by the experimenter there are no issues of diet compliance to contend with. Methods and Procedures Animals and housing Male and female outbred MF1 mice were obtained from Harlan, UK at 4 weeks of age (parental generation, n = 47) or bred in house (first generation of offspring from parental generation, F1, n = 80). Mice were maintained in a temperature controlled room (21 ± 1 °C) under a 12:12-h light–dark cycle, with lights on at 0500 h and a “dawn/dusk” period of 20 min at either end of the light period. After a breeding event at 10 weeks of age that occurred in all mice, mice were individually housed in standard cages with ad libitum food (D12450B, 10% kcal fat, 18.24 kJ/g, Research Diets, New Brunswick, NJ) and water. All mice (n = 127) were implanted intraperitoneally with temperature transmitters (PDT-4000 E-Mitter; Mini Mitter Bend, OR) under general anesthesia (mixture of isoflurane and oxygen). Males were implanted at 14 weeks of age and females at 17–18 weeks of age at least 10 days after their litters had been weaned. Mice were allowed at least 12 days to recover from the surgery before the start of the experiment. All procedures concerning animal care and treatment were approved by the ethical committee for the use of experimental animals of the University of Aberdeen, and licensed by the UK Home Office. Experimental procedure Baseline measurements started at the age of 19–20 weeks and were taken over a period of 4 weeks (day −28 to −1). During this time mice had ad libitum access to food and water. BM and food intake were measured each day between 1600–1700 h (1 hour before lights off) throughout the experimental period, including baseline. Food intake of all mice was then restricted to 70% of their individual baseline food intake (calculated in grams over the last week of baseline) for a period of 28 days (CR; day 0–28). Food rations were weighed and delivered daily between 1600 and 1700 h after animals were weighed. Supplementary Figure S1 online shows how BM and food intake changed during the experimental procedures and includes a diagram showing the timing of the different procedures that were performed. 286

Body temperature and general activity Mice in their home cages were placed onto transponder energizers (ER-4000 Receiver; Mini Mitter) allowing us to noninvasively monitor body temperature and general activity throughout the protocol. The VitalView Data Acquisition System (Mini Mitter) was used to collect the data in 1 min intervals (for a detailed description see (18)). Food anticipatory activity (FAA) was determined by calculating the amount of activity in the 3 h prior to food provision during CR (1300–1600 h) and was expressed as a % of total 24 h activity. RMR RMR was determined in all animals during baseline (−19 ± 2 days) and after 16 ± 2 days on CR. All measurements took place during the light phase between 0700 and 1600. RMR was measured in an open-flow respiratory system (as described in ref. 19). In short, fresh air was pumped (Charles Austin Pumps) through a sealed Perspex chamber within an incubator (INL-401N-010, Gallenkamp) set at 30 °C. This temperature was chosen because it lies in the thermo-neutral zone for these mice. When mice are exposed to CR some animals may enter into torpor to conserve energy (13) and these effects on metabolic rate were excluded by measuring animals under thermo-neutral conditions. Mass-flow controllers (MKS Instruments UK, Cheshire, UK) provided 500–700 ml O2/min which was monitored using an Alexander Wright DM3A flow meter. Air leaving the animal chamber was dried using silica gel and 150 ml/min was passed through a gas analyzer (Servomex Xentra). CO2 was not absorbed prior to gas analysis as this maximizes the accuracy of EE measures (20). Gas concentrations were measured continuously, and averaged values were stored every 30 s for 180 min. RMR was quantified as the oxygen consumption over the lowest 20 consecutive values (10 min interval) and corrected for ambient temperature and pressure, using the appropriate equation (21). The data (in ml O2/min) were converted to energy equivalents using an oxycalorific value of 21.117 J/ml O2, derived from the Weir (22) equation for an RQ of 1 (23). Mean BM was calculated from mass before and after each run. Body composition Fat mass (FM) and fat free mass (FFM) of mice was determined twice during the experimental protocol using dual energy X-ray absorptometry (DXA; PIXImus2 Series Densitometers with software version 1.46.007; GE Medical Systems Ultrasound and BMD, Bedford, UK); at day −8 of baseline and day −28 of CR. Mice were anesthetized using a face mask which provided a mixture of isoflorane and oxygen for the duration of the scan (~3 min). The software enabled a region of interest to be created to exclude the head with the mask from analysis (10). Data were corrected with a calibration formula specific to our machine that has been generated by the linear regression of fat content determined by DXA analysis, with the fat content as determined by soxhlet chemical extraction (see ref. 24 for detailed description of the procedure). Data analysis All data were tested for normality using the Kolmogorov–Smirnov test in SPSS (version 18), and when necessary data were log-transformed to obtain a normal distribution. Stepwise linear regressions were performed with changes in BM or FM as the dependent variable and several independent variables (i.e., baseline variables like RMR, general activity, body temperature etc., or compensatory changes that occurred in these variables in response to CR) to establish a model that predicted weight loss on CR. Changes in BM and FM were expressed as the residuals from regressions between mass at baseline and CR or as absolute change in grams (mass at CR – mass at baseline). Regressions models were performed including individuals of both sexes (where sex was included as a predictor variable) and for each sex separately. Independent predictors were only included in the final model when they significantly contributed to the model, i.e., P < 0.05. Data were investigated for unusual outliers (standardized residual >2) and influential data (leverage >(2k + 2)/n, where k is the number of predictors in the model and n is the number of observations). One significant outlier VOLUME 20 NUMBER 2 | february 2012 | www.obesityjournal.org

articles Integrative Physiology a

20

Pre-existing differences

All variables measured under baseline conditions varied considerably between individuals (Table 1). Food intake varied from 3 to 5 g (Table  1). Forty-two percent of this variation could be explained by differences in BM (Figure  2a, linear regression, F1,126 = 89.1, P < 0.001). For FM, RMR and general activity, the variation in BM explained 55, 10, and 11% of the individual variation, respectively and this relationship was positive for FM and RMR and negative for general activity (linear regression, Figure 2b: F1,126 = 154.1, P < 0.001, F1,126 = 13.2, Figure 2c: P < 0.001 and Figure 2d: F1,122 = 15.6, P < 0.001, respectively). Predicting weight loss with baseline variables

To establish whether pre-existing differences (i.e., variability in baseline variables) between mice could predict subsequent changes in BM and FM on CR stepwise linear regression was applied (for detailed methods see paragraph on “Data analysis”). Sex (for models including both sexes), BM, food intake (g/d), general activity (× 103 N counts/d), body temperature (ºC), RMR (kJ/d), and FM (g) were added as independent obesity | VOLUME 20 NUMBER 2 | february 2012

15

10

5

0 0–5 5–10 10–15 15–20 20–25 25–30 30–35 35–40 Weight loss category (%)

b

Results Variability in weight loss

4 2 0 −2

Fat mass change (g)

CR resulted in an immediate decrease in BM that levelled off towards the end of the restriction period (Supplementary Figure S1 online). There was a large variation in the amount of BM lost on 30% CR in both male and female mice (Figure 1a). BM loss ranged from 36 to 1% of baseline BM, equivalent to 17.3–0.3 g in absolute mass. Mice made adjustments to their body composition as they lost mass, but individual mice used very different strategies (i.e., there was no correlation between FM loss and FFM loss, P > 0.5, Figure 1b). Most mice reduced both FM and FFM mass to varying degrees (FFM; bottom left corner, Figure 1b), but some mice increased their FM while decreasing FFM (Top left corner, Figure 1b) or increased FFM while decreasing FM (Bottom right corner, Figure 1b).

25 Females Males

Number of mice

(standardized residual >3) was removed before final analysis, because there were confounding factors affecting the data for this individual. Some other statistical outliers were identified, but these represented true biological variation and were therefore not removed. Several tests were performed in SPSS to confirm that our models met the assumptions of linear regression. Linearity of predictors was tested by investigating partial regression plots. Standardized residuals for each model were calculated and tested for normality (Kolmogorov–Smirnov test). Homogeneity of variance (homoscedasticity) was inspected by plotting stardardized residuals against predicted variables. Finally, multicollinearity was tested by estimating tolerance and the variance inflation factor for each of the predictors. These tests showed that the assumptions of linear regression were not violated in our models; i.e., there were no nonlinear effects of predictors, residuals were normally distributed (P > 0.05) and homoscedastic (i.e., variance of the residuals was homogeneous across levels of the predicted values), and no multicollinearity was observed (variance inflation factor was