Rat energy balance study
EI, wheel use, and body weight
Energy intake (EI; kcal/d) in the AN group decreased by approximately 50 % from day −2 to diagnosis of AN (day 0), and AN rats were consuming significantly fewer calories on day 0 than controls (p = 0.0002; Fig. 1a). Wheel running increased significantly in the 4 days leading up to AN diagnosis (Fig. 1b). This combination of increased physical activity and decreased energy intake precipitated the decline to an anorexic body weight, defined as 75 % of initial weight (90.1 ± 2.8 g), in 4 to 7 days (Fig. 1c). 75 % of initial body weight was chosen as the threshold for AN because we previously found that this was the mean percent ideal body weight of female adolescent AN patients who presented for inpatient medical care [16]. When given the same timed access to food without running wheels, control animals remained weight stable over the 7 day lead-in period (Fig. 1c).
In the AN group, refeeding was modeled after the staged refeeding protocol currently used in adolescent AN patients at our hospital. Thus, on days 1 and 2 of refeeding, rats were provided 200 % of their intake on d0. This was increased to 300 % on d3, and ad libitum intake for the remainder of the study. Conversely, control rats were allowed ad libitum food access for the entire refeeding period. In a mixed model analysis, a model consisting of group indicator, linear, quadratic terms and linear by group interaction as fixed effects best described the weight regain (Fig. 1d). Both AN and control rats gained a significant amount of weight over the weight re-gain period, but the rate of weight gain was significantly greater in the AN rats than controls (p = 0.0188). By d19 of refeeding, the body weights of the AN rats (180.9 ± 6.0 g) were no longer statistically different from controls (194.3 ± 5.7 g; Fig. 1c).
Energy expenditure and energy balance
Total energy expenditure (TEE) gradually declined in the 5 days leading to the AN diagnosis (Fig. 2a), a result of a declining resting energy expenditure (REE, Fig. 2b). During this time, nonresting energy expenditure (NREE) increased (Fig. 2c) as physical activity increased (Fig. 1b), until energy intake precipitously declined (Fig. 1a). At the time of AN diagnosis, REE reached a distinct nadir, while NREE remained relatively high. AN rats were in a negative energy imbalance during the entire lead-in period. The adaptive response to decrease REE minimized the energy imbalance in AN rats to −10 kcal at the time of diagnosis. In addition, protein disappearance increased during the days leading up to AN diagnosis (Fig. 2e), implying that these animals are breaking down lean body mass in attempt to maintain body weight. In contrast, the control group were in relative energy balance throughout the entire lead-in period (Fig. 2d) with little variation in TEE, REE, and NREE, consistent with stable body weight (Fig. 1c).
Over the first 4 days of refeeding, TEE increased in both the AN and control groups, but TEE was lower in AN rats compared to controls (Fig. 2a). This decrease in TEE is explained primarily by the lower REE in the AN group (Fig. 2b). NREE did not change significantly in either the AN or control group during the first few days of refeeding, despite the fact that wheel running activity was stopped. This suggests that the increased food intake during this time, and the resulting increase in thermic effect of food (TEF), offset the decrease in activity thermogenesis, resulting in no change in NREE (Fig. 2c). As a result, EB increased in both the AN and control group during the refeeding period, with no difference between the groups (Fig. 2d).
EI was lower in AN than controls rats (EI = 33, 35, and 71 % of control on days 1, 2, and 3 of refeeding, respectively; p < 0.0001) resulting in lower thermic effect of food in the AN group during this time. Despite this lower EI in AN rats, TEE was so much lower that energy balance was the same for the AN and control groups on days 2 and 3 of refeeding (Fig. 2d). The differences in TEE between AN and control rats during refeeding was primarily due to suppression of REE (Fig. 2b) with a much lower contribution from NREE (Fig. 2c).
Hormone and metabolite responses
The ABA protocol induced a significant stress response, as demonstrated by significantly higher levels of urinary corticosterone in AN rats, compared to controls (P < 0.0001; Fig. 3a). Corticosterone levels began to decline immediately with the cessation of exercise and increased food intake, and were not significantly different than controls by d1 of refeeding (Fig. 3a).
Plasma glucose, insulin and leptin concentrations were measured in the fasted state at three time points: 1) at baseline, prior to the start of the AN protocol; 2) during the early refeeding period (d2 to 14 of refeeding); and 3) at end of the study. Blood draws were not performed at AN diagnosis (day 0) as the animals were too fragile. Glucose levels increased over time (Fig. 3b; p < 0.001), as was expected given that all rats were overfeeding and in a positive energy imbalance, but there were no differences between groups. Plasma insulin levels also increased over time in the control, but not AN rats, and, at the end of study, insulin levels were significantly lower in the AN group when compared to controls (Fig. 3c; p < 0.05).
As expected, leptin levels increased in the control animals (Fig. 3d) as body fat increased (Fig. 4a). AN rats, however, had lower leptin levels in the early refeeding phase, despite having the same body weight and fat mass as control rats. At the end of the study, leptin was 3.7 fold lower in the AN rats compared to control rats, which is lower than what would be expected based on the 1.6 fold lower body fat.
Body composition
At the time of AN diagnosis (day 0) there was no difference in body fat or lean body mass between the two groups (Fig. 4a-b) despite a significantly lower body mass in the AN animals. Thus, while both access to a running wheel and timed food access are required for the development of AN, timed food access alone (control rats) led to a decrease in body fat without a significant change in total body weight. During the weight regain period, body fat increased significantly in both the AN and control groups (Fig. 4a-b). However, fat gain plateaued in the AN group during the second week of refeeding, but continued to increase in the control animals. As a result, AN rats gained significantly less fat over the follow-up period (day 0 vs 28, p < 0.001), perhaps an indication that they are “defending” their lower body weight. At the time of sacrifice, the amount of subcutaneous fat was not different between control and AN rats but visceral fat was significantly lower in the AN rats relative to controls (p = 0.02, Fig. 4d), suggesting that a decrease in visceral fat was primarily responsible for the lower total body fat in the AN group. Refeeding increased liver fat by approximately 40 % in AN rats relative to controls (0.7 vs 1.9 g in AN and controls, respectively; p < 0.05, Fig. 4e).
Clinical feasibility and proof of concept study
Clinical acceptance of a diet higher in fat and protein
The average macronutrient content of the meal plan (selected by parents from a specialized menu, with dietician support and oversight) in the clinical program for adolescents with AN was 15 % protein, 26 % fat, and 59 % carbohydrate. Meals accounted for 64 % of total calories (16 % protein, 31 % fat, and 53 % carbohydrate), while snacks accounted for the remaining 36 % of total calories (11 % protein, 21 % fat, and 68 % carbohydrate).
To follow up on the finding of increased liver fat observed during the rat refeeding model, and to explore the possibility of feeding a different “test diet” to patients with AN, which might decrease the risk of accumulation of liver fat during refeeding, we conducted feasibility studies utilizing a moderate fat (MF) diet. The current LF (15 % protein, 26 % fat, 59 % carbohydrate) and test MF (20 % protein, 40 % fat, 40 % carbohydrate) diets were ranked the same for both taste and palatability (Fig. 5a and b). There was no difference in the energy consumption between the two meals as participants consumed all food presented to them under both conditions. These data suggest that feeding a MF diet would be possible in this population. In addition, a single inpatient admitted to the clinical program was fed a dietician-directed, self-selected MF diet for 10 consecutive days proving that it is indeed possible to plan meals with a higher fat, higher protein and lower carbohydrates clinically in adolescents with AN during the refeeding period (Fig. 5c-d).