Peripheral metabolic responses to prolonged weight reduction that promote rapid, efficient regain in obesity-prone rats

Paul S. MacLean, Janine A. Higgins, Matthew R. Jackman, Ginger C. Johnson, Brooke K. Fleming-Elder, Holly R. Wyatt, Edward L. Melanson, James O. Hill


Weight regain after weight loss is the most significant impediment to long-term weight reduction. We have developed a rodent paradigm that models the process of regain after weight loss, and we have employed both prospective and cross-sectional analyses to characterize the compensatory adaptations to weight reduction that may contribute to the propensity to regain lost weight. Obese rats were fed an energy-restricted (50–60% kcal) low-fat diet that reduced body weight by 14%. This reduced weight was maintained for up to 16 wk with limited provisions of the low-fat diet. Intake restriction was then removed, and the rats were followed for 56 days as they relapsed to the obese state. Prolonged weight reduction was accompanied by 1) a persistent energy gap resulting from an increased drive to eat and a reduced expenditure of energy, 2) a higher caloric efficiency of regain that may be linked with suppressed lipid utilization early in the relapse process, 3) preferential lipid accumulation in adipose tissue accompanied by adipocyte hyperplasia, and 4) humoral adiposity signals that underestimate the level of peripheral adiposity and likely influence the neural pathways controlling energy balance. Taken together, long-term weight reduction in this rodent paradigm is accompanied by a number of interrelated compensatory adjustments in the periphery that work together to promote rapid and efficient weight regain. These metabolic adjustments to weight reduction are discussed in the context of a homeostatic feedback system that controls body weight.

  • leptin
  • fuel metabolism
  • metabolic inflexibility
  • adipose cellularity

the incidence of obesity continues to increase at an alarming rate (1, 25), and current weight control strategies have minimal success as long-term solutions (12, 17). One of the underlying problems preventing the success of many weight control strategies is the compensatory metabolic responses to weight reduction. More than a decade ago, studies in animal models indicated that weight loss was accompanied by a profound drive to regain lost weight (9, 14, 15, 29). By following up on these observations, a large body of evidence now suggests that a homeostatic feedback system defending peripheral adiposity is fundamental to this metabolic drive to regain lost weight (22, 36, 42, 65). This communication loop between the brain and periphery is not completely understood, but, in the weight-reduced state, neural control centers in this feedback loop create a large potential energy imbalance, or energy gap, between appetite and expenditure. Clinical studies of the weight-reduced state confirm that this adaptation occurs in humans, because both an increased desire to eat (4, 10, 13, 34) and a reduction in energy expenditure (23, 35, 52, 58, 61, 62) contribute to this perceived gap. The most common approach to maintaining a weight-reduced state is to consciously reduce energy intake (EI) to match the suppressed level of energy expenditure. Overcoming this biological drive to eat excessively while expenditure is suppressed has proven too difficult for most people, and the prevention of weight regain has emerged as the most significant obstacle in our efforts to curtail this growing epidemic (17, 30).

Studying the metabolic contribution to the propensity to regain weight in humans is challenging. Not only is it difficult to control or accurately measure EI (44, 64), it is also next to impossible to control or standardize other factors that influence body weight regulation (17). Environmental and social pressures that vary greatly in humans are thought to alter the homeostatic regulatory system via nonhomeostatic, or “hedonistic,” neural inputs (3). To further complicate matters, genetic variability can impose a wide range of responses in both homeostatic and nonhomeostatic systems. For these reasons, rodent models of human obesity have provided valuable alternatives to specifically examine compensatory adaptations in metabolism that promote weight regain. Rodents can be selected for their polygenic predisposition to become obese under environmental conditions that promote obesity in humans, and the obesity-prone animals can be followed through weight reduction and the relapse to obesity under environmental conditions relevant to humans attempting to lose weight (37, 38, 40, 41). With better control over the genetic and environmental influences, these rodent models of human obesity have emerged as useful tools for characterizing the metabolic adaptations to weight reduction and the strategies that can effectively counter them (37). Recently, we have used these models to show that the energy gap increases with time in the weight-reduced state (40), dispelling our previous hopes that the metabolic drive to regain weight may eventually dissipate if intake could be restricted long enough for the homeostatic system to readjust. We also observed that perturbations in both appetite and expenditure persist during the early part of relapse when the rate of regain is exceptionally high. While a significant amount of work has been done to characterize the compensatory adaptations in the neural control centers of this feedback system, less is known about the concomitant adaptations in the periphery that may alter this feedback system and promote weight regain.

In the present study, we have employed one animal model of human obesity to prospectively examine the resolution of the energy gap, the energetic efficiency of weight gain, and fuel utilization during the relapse to obesity. We have also used a cross-sectional design to characterize whole body and depot-specific adiposity, adipocyte cellularity, and humoral adiposity signals in the weight-reduced state and after weight regain. The observations from this report bring together a number of peripheral adaptations that occur in response to weight reduction that likely work together with neural control centers to promote rapid and efficient weight regain. These adaptations are discussed in the context of the homeostatic feedback system controlling adiposity in an integrative manner that is relevant to the growing global obesity problem.


Animal Care and Use

Male Wistar rats (125–150 g) were purchased from Charles River Laboratories (Wilmington, MA). Obesity-prone and obesity-resistant rats were identified by a dietary screening process that predicts future weight gain (6, 26). In short, the rats were standardized to the facility and to the consumption of a low-fat diet (12% kcal fat; cat. no. 11724; Research Diets, New Brunswick, NJ) for 2 wk. They were then switched to a high-fat diet (46% kcal fat; cat. no. 12344; Research Diets) for 1 wk, while weight gain was monitored. Rats were ranked by their rate of weight gain during the high-fat dietary challenge. The top tertile was classified as obesity prone, and the lower tertile was classified as obesity resistant. Rats from the middle tertile were not used for this study. All rats were switched to the low-fat diet for another week before entering the study. Individual housing was provided throughout the study in the University of Colorado at Denver and Health Sciences Center (UCHSC) Center for Laboratory Animal Care (22–24°C; 12:12-h light-dark cycle) with free access to water. These metabolic housing units limit physical activity from a number of perspectives. They are considerably smaller than the large polycarbonate cages used for standard rodent housing, and they eliminate independent foraging because there is no bedding or paper in these wire-mesh cages. In addition, food is more readily accessible from our open food cups that sit adjacent to the animal rather than from the wire-mesh feed holders that are commonly on top of standard housing. This specialized housing and all procedures were approved by the UCHSC Animal Care and Use Committee.

Experimental Design

The present study reports observations from two experimental designs. The first, referred to as experiment 1, is a prospective analysis of weight regain after weight loss from an obese state. The second, referred to as experiment 2, is a cross-sectional analysis of animals in the following metabolic states: preobese, obese, weight reduced, and relapsed obese.

Experiment 1.

Obesity-prone rats were fed a high-fat diet for 16 wk under conditions that limited their physical activity (Obese rats). The rats were then switched to a low-fat, energy-restricted (50–60% of ad libitum calories) diet that induced a 10–15% reduction in body weight over a 2-wk period. This reduced weight was then maintained for 8 wk by giving a limited daily provision of the low-fat diet that reflected 70–80% of the calories consumed in the obese state (Weight-Reduced rats). Energy balance and nonprotein respiratory quotient (NPRQ) was monitored in this weight-reduced state and after calorie restriction was removed on days 1, 3, 7, 14, 28, and 56 of the relapse process. This prospective analysis was compared with a set of obese rats that were similar in age but had never lost weight.

Experiment 2.

The second experimental cohort was a cross-sectional design with groups that reflected various stages of obesity development, weight-reduction, and relapse. We have recently reported a detailed description of energy expenditure and metabolic efficiency in this cohort of animals (40), and we now employ this same cohort to provide a more detailed description of whole body and depot-specific adiposity, adipocyte cellularity, and humoral adiposity signals. This design includes three groups of control rats: 1) Preobese, 2) Obese, and 3) Never Obese. Fourteen obesity-prone rats, designated Preobese, were examined before the development of obesity, immediately after the screening period, and when returned to a low-fat diet for a week. Another group of obesity-prone rats were placed in an obesigenic environment that included free access to a high-fat diet and limited physical activity. These rats, designated Obese, were examined after the development of obesity, either at 16 wk in this obesigenic environment or after continued ad libitum feeding through 42 wk of the study. The final control group consisted of obesity-resistant rats that were placed in similar housing with free access to the low-fat diet. These rats, designated Never Obese, were examined after 16–18 wk under these conditions.

In addition to the control animals, we used two sets of experimental rats. The first consisted of a group of obesity-prone rats that were placed in the obesigenic environment for 16 wk followed by 2 wk of a calorie-restricted (50–60% of ad libitum calories), low-fat diet designed to induce weight loss. These rats were then examined after 0, 8, or 16 wk on an intake-regulated, low-fat diet, designed to maintain the reduced weight (Weight-Reduced rats). The second set of experimental animals were selected and treated identically to the Weight-Reduced rats, except that they were examined after 8 wk of weight regain subsequent to the 0, 8, or 16 wk of weight maintenance (Relapsed-Obese rats). Weight loss was targeted to achieve a 10–15% loss in body weight that would be reflective of targeted adjustments in human weight loss programs. Weight loss and subsequent weight maintenance was achieved in a manner similar to that described for the prospective cohort in experiment 1.

Metabolic Monitoring

Energy balance was examined with a metabolic monitoring system developed by the Energy Balance Core Laboratory at the University of Colorado Clinical Nutrition Research Unit, as described previously (41). This monitoring system is composed of a four-chamber indirect calorimeter designed for the continuous monitoring of up to four rats simultaneously, obtaining measurements of V̇o2 and V̇co2 from each chamber every 6 min. Chambers are also equipped for the collection of urine, feces, and food spillage. Rats destined for metabolic monitoring were acclimated to the system for 2–3 days before data collection periods. The rats were then monitored for 23 h, during which time EI was measured. Urine volume was recorded and a portion of it was collected for the measurement of urinary nitrogen levels (ThermoDMA, Louisville, CO). The remaining hour was used to clean the chamber and prepare for the next monitoring period.

Metabolic rate (MR) was calculated with the Weir equation (MR = 3.941·V̇o2 + 1.106·V̇co2 − 2.17·N). Total energy expenditure (TEE) was calculated as the average of all MR measurements taken (every 6 min), throughout the 23-h period and was extrapolated for presentation purposes to reflect that amount of energy expended through 24 h. Resting MR was estimated as an average MR during the latter part of the light cycle, a time in which MR and respiratory quotient indicated minimal physical activity and food intake for the three previous hours (41). Fuel utilization was calculated from the gas exchange data and the urinary nitrogen measurements, using the following equations Math Math Math

Body Composition Analysis

Body composition analyses were performed by dual-energy X-ray absorptiometry (DEXA) using the Lunar DPX-IQ (GE Lunar, Madison, WI) with Lunar's Small Animal Software version 1.0. Corrected fat mass and fat-free mass were calculated from DEXA data and body weights according to the recommendations of Feely et al. (18), who have standardized this approach to chemical analyses of body composition.

Adipocyte Cellularity

Retroperitoneal and epididymal fat pads were removed and weighed, and a portion of each was immediately processed for cellularity characterization. Average cell diameter and cell number per fat pad were characterized by methods described previously (27, 46). In short, a small portion of the pad was digested with collagenase, and adipocytes (150–250 cells/sample) were imaged using an Olympus BX60 microscope and a C-mounted Canon Power Shot G5 digital camera. Images were processed with MATLAB software (courtesy of Dr. William Betz, UCHSC). The distribution and spread of cell diameter (mean, median, kurtosis, skewness, etc.) in each sample was checked for normality, and the number of cells per fat pad was calculated with the average diameter, a density conversion factor (0.915 g/cc), and the mass of the fat pads, as previously described (46).

Plasma Analyses, Urine Analyses, and Whole Body Insulin Sensitivity

For experiment 2, urine was collected over a 23-h period while the animals were in the metabolic monitoring system, and plasma was collected from blood samples when the animals were killed after a 10-h fast. Urinary corticosterone and creatinine was measured by a commercially available ELISA (Diagnostic Systems Labs, Webster, TX) and a colorimetric assay (ThermoElectron, Melbourne, Australia), respectively. The plasma was assayed for insulin, leptin, and a thyroid hormone (free T3) with commercially available ELISAs (Alpco Diagnostics, Windam, NJ), and glucose was measured via colorimetric analysis (43). Whole body insulin sensitivity was estimated in a separate cohort of animals that represented Preobese, Obese, Weight-Reduced, and Relapsed-Obese rats, using the hyperinsulinemic-euglycemic clamp technique (49). With this procedure, the jugular vein and carotid artery were cannulated for each rat in a survival surgical procedure. After 7 days of recovery, a constant-insulin, variable-glucose infusion was initiated after a 12-h fast. The infusion of glucose was adjusted to maintain euglycemia (85–90 mg/dl). The glucose infusion rate was determined over the final 30-min steady-state period of the clamp when glucose infusion did not vary.

Statistical Analysis

Data from the prospective analysis were analyzed with the SPSS software version 13.0 by repeated-measures ANOVA with Fisher’s least significant difference post hoc test when a significant effect was observed. Data from the cross-sectional analysis were analyzed by ANOVA with Duncan's post hoc analysis for homogeneous groups whenever a significant main effect was observed. ANCOVA was employed in this cross-sectional cohort to examine the differences in adiposity signals after adjusting for total body-fat mass. Pearson and partial correlation coefficients were calculated to examine the relationships between parameters of adipocyte cellularity and total body fat. Statistical significance was assumed when P < 0.05.


Energy Balance During Weight Regain

In experiment 1, a prospective examination of the energetics of weight regain was performed on 11 rats, which were similar with respect to their polygenic predisposition to become obese, the conditions under which they became obese, and the method to reduce weight and maintain the lower weight for 8 wk. The extended period of time with ad libitum high-fat feeding and limited physical activity led to the development of severe obesity in these obesity-prone rats. Before the weight reduction phase of the study, the rats weighed 717 ± 17 g and had a body fat of 39 ± 2%. Switching to an energy-restricted low-fat diet induced a 14% loss in total body weight that was maintained by feeding ∼80% of the calories consumed before weight loss. In experiment 2 and in our previous studies with this paradigm (40, 41), we have observed that weight loss with this approach is primarily due to a reduction in fat from adipose depots rather than lean body mass (Table 1). The reduction in the food provision was necessary so that intake of the low-fat diet was equivalent to the level to which energy expenditure was suppressed (Fig. 1, AB). Following the 8-wk maintenance period, this restriction on intake was removed, and the Weight-Reduced rats were allowed to eat the low-fat diet ad libitum. Within the first 2 days of relapse, consumption increased to a level that was ∼20% higher than the caloric intake before weight loss, and the small increase in expenditure did little to offset the exceptionally large positive energy imbalance and the substantial weight gain (Fig. 1, AC). Intake remained elevated for the first 4 wk of the relapse period, while expenditure remained suppressed even after all of the weight had been regained. Consequently, the energy imbalance was sustained, but gradually declined over the 56 days of relapse. Compared with the Obese rats, the rate of weight gain was 15-fold higher in the first few days of relapse and remained four- to sixfold higher during the first week of relapse. By the end of the 56-day monitoring period, the relapsing rats had not only regained all of the lost weight, but they had also accumulated an additional amount that represented 20% of what was originally lost (Fig. 1C).

Fig. 1.

Energy balance and the rate of weight gain during the relapse to obesity. Obese rats (n = 11) were weight reduced with a calorie-restricted, low-fat diet and were maintained at this reduced weight for 8 wk by limiting the daily provision of the low-fat diet. These Weight-Reduced rats were examined prospectively before and after the imposed restriction of the low-fat diet was removed and the weight was regained. Body weights and energy intake were measured daily, whereas energy expenditure (EE) was measured over a 24-h period on days 0, 1, 3, 7, 14, 28, and 56 of the relapse period. Energy intake (EI) data at these time points represent an average of several days around the time point (except for the single measurement on day 1). The prospective analysis was compared in the context of age-matched Obese group controls that were adapted to the low-fat diet for at least 1 wk (n = 9). Energy intake and energy expenditure (A), energy balance (B), and the percentage of lost weight that was regained (C) are expressed as means ± SE, and groups that are not significantly different have the same letter designation.

View this table:
Table 1.

Morphometric, humoral, and urine characteristics in the cross-sectional analysis of the development, treatment, and relapse to obesity

Feed Efficiency and Fuel Metabolism of Weight Gain

In experiment 1, the feed efficiency was calculated throughout the relapse period as the ratio of the number of calories eaten to the amount of weight gained (Fig. 2A). On day 1, feed efficiency dramatically increased to a level that was 10-fold higher than that found in the Obese rats. Within a few days this effect was dramatically reduced, but it did remain higher throughout the first 4 wk of relapse. A fair portion of this effect in the first few days of relapse can be attributed to glycogen repletion, water retention, and luminal filling of the intestine. However, feed efficiency remains elevated well into the relapse process when the contribution of these factors would be minimal. Given the minimal resolution in TEE, these observations imply that the energetic cost of weight gain was lower during the early stages of the relapse process and that more weight was gained for a given amount of food consumed. We have hypothesized that a portion of this reduction in the cost of weight gain during relapse is due to a shift in fuel utilization that favors the oxidation of carbohydrate (CHO) and preferentially diverts ingested lipid to adipose depots (39). To address this issue, we used the gas exchange data with 24-h urinary nitrogen measurements to examine how the NPRQ changed after weight reduction and during the relapse to obesity (Fig. 2B). NPRQ did not change with prolonged weight reduction, but it was significantly elevated from the obese state during the 2 wk of the relapse period and gradually normalized over the remaining period of the study. When expressed as substrate disappearance (Fig. 2, CD), these observations indicate that the shift in fuel metabolism is most profound in the first week of relapse, during a time in which close to 40% of the lost weight was regained. It should be noted that this shift in NPRQ may not only arise from the preferential oxidation of CHO that diverts ingested lipid to storage, but it also may involve the induction of de novo lipogenesis that promotes the deposition of excess CHO and protein as fat. However, this pathway of lipid deposition carries a considerable energetic expense and would be expected to decrease, rather than increase, the energetic efficiency of weight regain.

Fig. 2.

Feed efficiency, nonprotein respiratory quotient (NPRQ), and fuel disappearance. Prospective data regarding the energetic cost of weight gain and fuel utilization are shown for the cohort of animals described in Fig. 1. A: the feed efficiency of weight gain represents the amount of weight gain for a given amount of intake. It is calculated as the weight gain (g/day) divided by energy intake (kcal/day). B: NPRQ was calculated from the gas exchange and urinary nitrogen data acquired during 24-h indirect calorimetry, and these same data were employed to calculate carbohydrate (CHO; C) and lipid (D) disappearance throughout the relapse period. Data are expressed as means ± SE, and groups that are not significantly different have the same letter designation.

One additional aspect of fuel metabolism observed in the rats from experiment 1 was the diurnal fluctuation in substrate disappearance, which is an estimate of substrate oxidation when energy balance is achieved (Fig. 3, AC). Rats are nocturnal feeders and spend most of their waking hours during the 12-h dark cycle, whereas they are more prone to be inactive and sleep during their 12-h light cycle. Even so, Obese and Relapsed-Obese rats exhibited very little diurnal variation in respiratory quotient and fuel utilization derived from NPRQ. Although a minor shift in CHO disappearance was observed for both of these groups, neither altered lipid disappearance over the 24-h period. In contrast, the Weight-Reduced rats exhibited a dramatic shift in both CHO and lipid disappearance in the transition from their dark cycle to their light cycle. This fluctuation is a function of the limited food provision that is given at the beginning of the dark cycle. The provision is generally gone by the first few hours of the subsequent light cycle, leaving restricted rats with an extended period of time (∼10 h) with little or no food. Therefore, this study design forced Weight-Reduced rats into diurnal cycles in which fuel utilization was dramatically altered, first favoring CHO and then favoring lipid, according to the availability of exogenous fuel.

Fig. 3.

Diurnal flux in CHO and lipid utilization. Diurnal flux of respiratory quotient (V̇co2/V̇o2) (A), as well as of both CHO (B) and lipid (C) disappearance is shown for Obese, Weight-Reduced, and Relapsed-Obese (after 56 days of relapse) rats. Data are expressed as means ± SE of substrate disappearance over 12 h of the dark cycle and 12 h of the light cycle. Obese and Relapsed-Obese rats show little fluctuation in lipid disappearance between the light and dark cycles, while Weight-Reduced rats are forced to adjust lipid utilization as they cycle between periods with and without exogenous fuel availability. aSignificantly different from its paired 12-h dark cycle value. RQ, respiratory quotient.

Adipose Tissue Cellularity and Weight Regain

In a previous study, we performed a cross-sectional examination of metabolic efficiency and energy expenditure in a number of rats in this model of obesity recidivism, representing Preobese, Obese, Never-Obese, Weight-Reduced, and Relapsed-Obese rats (40). In experiment 2, we used this same cohort of animals to report whole body and depot-specific adiposity, adipocyte cellularity, and humoral adiposity signals in the weight-reduced state and after weight regain. A summary of the morphometric characteristics for the animals in experiment 2 is shown in Table 1. Body fat, estimated either from DEXA-determined body composition analyses or from the combined weight of retroperitoneal and epididymal fat pads, indicated that weight regain after weight loss was primarily due to lipid accumulation in adipose depots (Fig. 4A). This characteristic of weight regain was consistent, regardless of the amount of time in weight maintenance (data not shown). To further characterize the repletion of fat in this cohort of rats, we examined adipose tissue cellularity in both the retroperitoneal (Fig. 4B) and epididymal (Fig. 4C) fat pads. The data were consistent in both fat pads in that the development of obesity and calorie-restricted weight loss was reflected in predictable fluctuation in cell diameter. However, the rapid and efficient gain during relapse was accompanied by a 30% increase in the total number of cells per fat pad. After weight regain, the relapsed animals returned to their previous level of adiposity with a larger number of adipocytes at an equivalent average diameter, effectively increasing the total adipose capacity for lipid storage. These data would suggest that in this model, the development and progression of obesity is linked with adipocyte hypertrophy, whereas the weight gain associated with regaining lost weight is associated with adipocyte hyperplasia.

Fig. 4.

Body fat and adipocyte cellularity with weight regain. This cohort represents a cross-sectional analysis of 1) Preobese rats and Obesity-Prone rats examined before the development of obesity (n = 14); 2) Obese and Obesity-Prone rats examined after 16 wk of high-fat feeding and before or after 26 subsequent weeks on the low-fat diet (n = 12); 3) Never-Obese rats and Obesity-Resistant rats examined after 16–18 wk on a low-fat diet (n = 11); 4) Weight-Reduced rats and Obesity-Prone rats that had developed obesity with 16 wk of high-fat feeding, but then underwent weight loss on a intake-regulated low-fat diet and were maintained at this reduced weight for 0, 8, or 16 wk before examination (n = 19); and 5) Relapsed-Obese and Obesity-Prone rats that had developed obesity underwent weight loss, which was maintained at this reduced weight for 0, 8, or 16 wk, and were then allowed free access to the low-fat diet for eight subsequent weeks before examination (n = 21). A: total body fat was estimated by dual-energy X-ray absorptiometry (DEXA) and by the combined weight of epididymal (EPI) and retroperitoneal (RP) fat pads. Data are expressed as means ± SE, and groups with the same letter designation are not significantly different (analysis yielded the same results for both parameters). Adipocyte cellularity was examined for retroperitoneal (B) and epididymal (C) fat pads and is expressed as average cell diameter and total number of cells per fat pad. Data are expressed as means ± SE, and groups with the same letter (diameter) or symbol (number) designation are not significantly different.

Adiposity Signals Before and After Weight Regain

In experiment 2, we also measured the two humoral factors, leptin and insulin, that have emerged as key signals from the periphery to the central nervous system relaying information about adiposity levels (2, 5, 48). Having several measurements of whole body and tissue-specific fat accumulation in conjunction with both leptin and insulin in this large cohort of rats provided an opportunity to study what aspect of peripheral fat was being conveyed with their signal. To address this issue, we examined the relationship of both leptin and insulin to whole body parameters of adiposity and to depot-specific parameters of adipose cellularity. Because the relationship between insulin and fat mass was not linear, log-transformed values of insulin were used in the analyses. As expected, both leptin and log insulin were closely linked to total body fat, whether estimated by DEXA or by fat-pad weights (Table 2). Both adiposity signals were more closely linked to the average cell diameter than to the total cell number in the retroperitoneal fat pad (Table 2). In addition, a significant portion of the relationship between leptin and total body fat could be explained by the variability in cell diameter, but adjusting for both size and number had a greater impact on this relationship. In contrast, only cell size appeared to be a critical factor in the relationship between insulin and peripheral fat mass. Similar relationships were observed with epididymal cellularity (data not shown).

View this table:
Table 2.

Relationship of adiposity signals to fat mass and RP pad cellularity

In experiment 2, the development and progression of obesity was associated with an increase in both leptin and insulin and a suppression in peripheral insulin sensitivity (glucose infusion rate, Table 1), all of which were favorably affected with weight reduction. With weight reduction, leptin and insulin levels were reduced to a greater extent than would be expected for the loss in total body fat. This is most clearly seen when comparing Weight-Reduced rats to Never-Obese rats. Although Weight-Reduced rats still had >40 g more body fat (Fig. 4A), their plasma leptin and insulin values were similar to those in the Never-Obese group (Fig. 5, AB). This reduction does not resolve even with 16 wk of weight maintenance at the reduced weight. Another expression of this same effect can be observed by ANCOVA with DEXA fat mass or fat-pad weights as a covariate. With the leptin comparison, controlling for the variation in body fat yields a level of leptin in Weight-Reduced rats that appears to underestimate peripheral adiposity in these animals (Fig. 5A). With the insulin comparison, the results of this adjustment would suggest that the level of insulin in both Preobese and Weight-Reduced rats underestimates the level of peripheral adiposity (Fig. 5B). A cursory examination of the corticosterone and free T3 did not reveal a significant role for these factors in the propensity to regain weight in this paradigm (Table 1).

Fig. 5.

Plasma leptin and insulin levels with the development, treatment, and relapse to obesity. The cross-sectional analysis of the cohort described in Fig. 4 was employed to measure the adiposity signals of leptin (A) and insulin (B) in the weight-reduced state and after weight regain, after a ∼10-h fast. The data were reanalyzed by ANOVA with DEXA-determined fat mass (FM) as a covariate and are expressed as means ± SE (black bars), with groups designated by the same letter or symbol not being significantly different. Given that the relationship between insulin and fat mass was not linear, ANCOVA for this parameter was performed on log-transformed values of insulin.


This study reports several novel aspects about the metabolic propensity to regain weight after weight loss. First, these observations extend our previous study of energy balance in early relapse by examining the resolution of the energy gap throughout the entire relapse process. Even after the rats gained more than what was originally lost, TEE had not completely resolved to levels observed in the obese state. Second, the early part of relapse was characterized by 1) weight gain that was both rapid and energetically efficient and 2) a sustained shift in fuel utilization favoring the suppression of fat oxidation. Third, weight-reduced rats exhibited an improvement in insulin sensitivity and a diurnal fluctuation in fuel utilization, both of which are sustained with prolonged weight reduction and imply a reversal of the metabolically inflexible state found with obesity. Fourth, regain during the relapse process increased adipose tissue fat mass via hyperplasia rather than hypertrophy. Finally, two key humoral factors that convey information about adiposity to the central nervous system underestimate peripheral adiposity, even with long-term weight reduction. We believe these compensatory adjustments in metabolism are part of an interrelated group of adaptations in the homeostatic feedback loop between the periphery and the central nervous system that controls body weight (Fig. 6). The adiposity signals complete the feedback loop by altering the architecture of key neural pathways in the central nervous system that control energy balance, promoting the elevated drive to eat and reduced expenditure that are fundamental to the weight-reduced state. Although this compilation of compensatory adjustments is by no means complete, it provides a basic picture of the metabolic state accompanying weight reduction and the early period of relapse and suggests how these peripheral adjustments may work coordinately to promote rapid and efficient regain.

Fig. 6.

Metabolic state after weight-reduction (A), early in relapse (B), and after relapse (C). The compensatory adaptations in the periphery, regarding 1) energy balance, 2) energetic cost of gain, 3) lipid accretion, and 4) adiposity signals that facilitate rapid and efficient weight regain are summarized in the context of the homeostatic feedback system that defends peripheral fat stores. A: the metabolic state after weight reduction is diagrammed. The anabolic central nervous system profile promotes a large energy gap with an increased drive to eat and a suppressed expenditure of energy. The weight-reduced state is maintained only if intake is forcefully restricted to the level that energy expenditure is suppressed. Peripheral tissues are forced to be flexible with their use of fuels because of intermittent exogenous fuel availability. Adipose depots are stable, but depleted, and humoral adiposity signals are lower than would be expected for the level of peripheral fat. This blunted signal contributes to the adapted regulatory loop by lowering the sensitivity of neural control centers to hunger and satiety signals in a manner that promotes the overall anabolic output (for more discussion, see Refs. 11, 36, and 65). B: with uncontrolled intake, the energy gap is realized. Excess fuels and the accompanying insulin excursions, in the context of insulin-sensitive tissues that have the cellular infrastructure to suppress fat oxidation, promote a dramatic shift in fuel utilization whereby exogenous fat is preferentially diverted to storage depots. Ingested energy is rapidly and efficiently diverted toward the repletion of adipose stores. C: the pressures driving regain gradually resolve as the lost weight is regained. The metabolic state gradually shifts to one that attempts to prevent further weight gain. The energy gap is minimized at a higher overall energy flux, and peripheral tissues become resistant to insulin's actions and metabolically inflexible. The energetic efficiency of further gain is elevated. The regulatory system achieves a level of equilibrium similar to what is seen before weight loss except that the adipose depots have increased their total capacity to store fat, an effect that may have implications on the continued progression of obesity. In the face of both hyperinsulinemia and hyperleptinemia, the neural control centers exhibit what would appear to be a selective resistance to regulating intake, as both intake and expenditure remain high.

Our observations are consistent with a considerable amount of evidence in both animals (8, 15, 32, 37, 40, 41) and humans (28, 50, 52, 58, 62) that, in response to calorie-restricted weight reduction, there exists a large positive potential energy imbalance with intake-regulated weight maintenance that does not dissipate over time. The underlying imbalance, or energy gap, is the result of an overall anabolic output from the neural pathways in the brain that control energy balance (36, 65), increasing the drive to consume while suppressing energy expenditure. Our previous reports with this rodent paradigm indicate that, not only does this energy gap persist with prolonged weight reduction, but it may also become more profound as the length of time in weight maintenance increases and the defended body weight drifts upward (40, 41). We have shown that the suppression in TEE is due, at least in part, to an enhancement in metabolic efficiency that persists through the early part of relapse, thereby contributing to the high rate of weight regain. In the present study, we show that both EI and TEE remain perturbed throughout the entire relapse process and contribute to the large energy gap. After more weight was gained than what was originally lost, EI had normalized, whereas TEE remained suppressed. These observations emphasize one of the difficult aspects of maintaining a weight-reduced state: resisting this metabolic drive to eat in great excess while burning less (36, 39). To maintain this reduced weight, one must eat what might be considered a pittance when it is compared with the amount of consumption that the brain is strongly encouraging.

When the forced restriction on intake is removed, the body is challenged with excess calories. We and others have asserted that the improved insulin sensitivity and diurnal flux of metabolism observed in the weight-reduced state is likely contributing to the rapid, efficient regain (42, 57). From one perspective, this metabolic adaptation could be perceived as a practical expression of metabolic flexibility coined by Kelley and Mandarino (33). The infrastructure to turn fat oxidation on and off, which is not working or not employed in Obese and Relapsed-Obese rats, is utilized on a daily basis in Weight-Reduced rats as they cycle through periods when exogenous energy sources are unavailable. Over-consumption following a short period of energy restriction leads to the activation of acetyl-CoA carboxylase, the production of high levels of malonyl-CoA, and the inhibition of carnitine palmitoyl transferase I in the liver and muscle, all of which work to suppress fat oxidation (7, 19, 63). When faced with caloric excess, Weight-Reduced rats are minimizing fat oxidation, burning CHO to meet the basic energy needs, and directing exogenous lipid to the depleted fat stores. From an energetic perspective, it costs less to accumulate fat from ingested lipid than it does from ingested CHO or protein (20). The high energetic efficiency of regain and the sustained elevation in NPRQ early in relapse in the present study and similar findings in other studies (7, 23) support the notion of this shift in fuel utilization. Although the conversion of glucose to lipid via de novo lipogenesis could also contribute to the elevation in NPRQ, this pathway of lipid deposition is energetically expensive and would make weight gain less efficient. Even so, characterizing the respective contributions on a temporal basis to the change in fuel metabolism and the efficiency of weight gain during relapse will require the use of isotopic substrate tracers in future studies. An important consideration in this regard is that fuel utilization during relapse will undoubtedly be affected by the composition of the diet. Relapse on standard rodent chow (∼5% fat) has very little fat to traffic, and the contribution of de novo lipogenesis to the process of weight regain is likely to be high. In contrast, relapse on a high-fat diet allows the trafficking of a substantial amount of fat and may yield weight regain that is very energetically efficient.

Accompanying this shift in fuel metabolism during regain is a preferential deposition of fat in adipose tissue and a change in the cellularity characteristics of the adipose depots. Previous studies have consistently shown that regain favors fat deposition over that of lean body mass (16, 41), and while it is not unanimously reported, hyperplasia has been seen after ad libitum feeding following prolonged caloric restriction in other rodent models (24). Several studies are now indicating that both apoptosis and differentiation of adipose cells occur regularly in mature humans and rodents (24, 51, 55, 56, 59). This turnover appears to be a regulated process that is sensitive to nutritional and hormonal factors (51, 53). Whereas further studies are required to delineate whether increased differentiation or suppressed apoptosis underlies the hyperplasia observed in the present study, the anabolic humoral conditions accompanying relapse are more consistent with the promotion of preadipocyte differentiation. New, small adipocytes would be more prone to accumulate lipid and, by their nature, have a greater capacity to do so (21, 45). The turnover of adipose cells under the conditions early in relapse may then be shifted toward increasing the number of cells that are not only available for lipid storage but preferentially increasing cells that are very prone to take up and store excess lipid diverted from oxidation. This effect may not only contribute to the high rate of regain, but also may provide a larger capacity for lipid storage and the further progression of obesity once the lost weight has been regained. Clearly more studies are required to substantiate and characterize the effect of regain on adipocyte cellularity within this paradigm.

Finally, to regulate energy homeostasis, the peripheral tissues communicate to the central nervous system via neural, nutrient, and endocrine pathways that integrate both satiety and adiposity signals (11, 65). Insulin and leptin have clearly emerged as key humoral signals that convey information to the central nervous system about peripheral adiposity (2, 5, 48). In the Weight-Reduced rats, leptin and insulin appear to be underestimating peripheral adiposity, an effect that has been shown with leptin in weight-reduced humans (31, 54). Our observations would suggest that this is not a transient phenomenon, but one that persists with prolonged weight reduction. Although consistent with other reports that suggest that leptin expression is a function of adipocyte size rather than number (60), the observations in the present study with adipocyte cellularity do not provide a clear explanation for this suppression of adiposity signals. It is likely that other neural, nutrient, or endocrine factors may play a more critical role in altering the levels of these adiposity signals during prolonged weight reduction. In any case, these adiposity signals are critical in that they can work together to affect not only the neural pathways controlling energy balance (65), but also peripheral pathways controlling fuel metabolism (47). The dual impact of this compensatory response may be that relatively low levels of these factors foster the potential energy gap in the weight-reduced state and then mediate the shift in fuel metabolism that promotes rapid and efficient weight regain early in relapse (42). Relevant to this issue is the temporal resolution of the adjustment in these adiposity signals during relapse. The chronic caloric excess early in relapse could lead to a sustained increase in insulin and leptin, limiting their relevance to the neural regulation of energy balance and peripheral fuel utilization to the initial stages of relapse. Clearly, further examination of these and other adiposity signals during the process of relapse would help to address this issue.

In summary, weight reduction is accompanied by a number of compensatory metabolic adjustments that work together to promote rapid and efficient weight regain. Defense of a high body weight, even with prolonged weight reduction, speaks to the persistent nature of these adaptations in promoting weight regain. The more profound pressure to regain early in the relapse process may explain how short-term lapses in maintenance strategies can have profound consequences. We have examined a number of peripheral metabolic changes in a rodent model that resembles the development, treatment, and relapse of obesity in humans, and placed them in the context of the homeostatic feedback system that controls body weight. Although our representation of the peripheral adjustments to weight reduction is categorized into the energy gap, fuel utilization, tissue-specific lipid accretion, and humoral adiposity signals, this picture of the weight-reduced state is fairly simplistic and by no means complete. Further studies are needed not only to incorporate other metabolic adaptations that may be contributing to the metabolic drive to regain weight, but also to examine how these metabolic adjustments resolve as the weight is regained during the relapse period. By extensively characterizing this model of weight regain, we may be able to use it to identify nutritional, behavioral, and pharmacological strategies that counter the propensity to regain weight and facilitate long-term weight reduction in obese and overweight humans.


This work was supported by National Institute of Diabetes and Digestive and Kidney Diseases Grants DK-38808 (to J. O. Hill) and DK-67403 (to P. S. MacLean). We also acknowledge the generous support from the Energy Balance and Metabolic Core Laboratories within the Colorado Clinical Nutrition Research Unit, which also received funding from National Institute of Diabetes and Digestive and Kidney Diseases Grant DK-48520 (to J. O. Hill).


We thank Dr. William Betz and Steve Fadul for their assistance with the imaging and analysis of isolated adipocytes. We appreciate the technical assistance from Neal Beeman, Dana Higbee, and Ling Bai.


  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.


View Abstract