This is the first report of persistent diet-induced obesity in mice. Two previous studies also using C57BL/6 mice did not observe persistent obesity following removal of high fat diets 
. But these previous studies used younger mice (4–6 weeks of age vs. 3 months old in the present study) that were housed in groups of 5 animals per cage as opposed to individual housing. These factors, along with differences in the composition of the diets, likely contributed to our unique observation that returning mice to a non-obesigenic diet after 7 weeks of exposure to two different obesigenic diets resulted in sustained elevations of body fat by 40–50% when compared with mice that were never exposed to the obesigenic diets.
Of course, the persistence of obesity for 12 weeks following the return to a non-obesigenic diet does not prove that these mice would have permanent alterations of body weight and composition. Nevertheless, this period of time represents a significant fraction of the average lifespan of these mice (about 10 percent) and was more than 70% longer than the period of obesity induction. Furthermore, unlike the previous observations of persistent diet-induced obesity in rats that occurred on a background of significant growth 
, our chow fed mice increased their fat-free mass by less than 5% over the 19 week study. Thus, beginning the study with 3 month old mice achieved relative stability of lean tissue mass in the control group and allowed diet interventions of sufficient duration while avoiding confounding factors of advanced age.
Interestingly, the different average body weights and fat masses attained after 7 weeks of HF versus EN diets did not impact the final level of persistent obesity (). This suggests that both diets resulted in sufficient weight gain to cross a threshold for inducing a fixed increment of persistent body weight gain. In contrast, a graded effect would be expected to result in different levels of persistent obesity for varying degrees of diet-induced weight gain. This did not happen. Rather, our observations are consistent with a previously suggested mechanism of inducing multiple body weight set points generated by acquired leptin resistance during the high energy diets 
. Future studies could potentially locate the threshold weight gain required to achieve persistent obesity by varying the dietary fat percentage to scan a range of body weight gains over the 7 week obesigenic period.
To help understand the energy imbalances underlying our results, we used the principle of energy conservation to calculate the average energy output rate as a function of the average energy intake rate (). This analysis revealed that HF and EN mice had decreased average energy output relative to their average energy intake. Decreased average ambulatory activity in the HF and EN mice () likely contributed to this relative reduction of energy output and similar behavior was recently observed in C57BL/6 mice fed with a high fat diet 
. The mice in the diet switched groups, HF-C and EN-C, had average energy output rates that fell on the same line as the C group implying that their average energy output was comparable to the control group.
The HF-C and EN-C mice cumulatively consumed more energy than the control mice, but there were no statistically significant differences between the energy intake rates of the HF-C, EN-C, and C groups over the last several weeks of the study () despite the steady elevation of body weight. While not statistically significant, the energy intake rates of the HF-C and EN-C mice at the end of the study were numerically greater than the C group, but within 1 kcal/d. Significant differences of body weight can be maintained by very small differences of energy intake thereby putting severe constraints on the ability to experimentally detect such differences. For example, the slope of the regression line relating energy intake and body weight at the end of our experiment () was only 0.21±0.02 kcal/g/d which means that less than 1 kcal/d of additional energy intake would be required to maintain the 4 g of additional body weight observed in the HF-C and EN-C groups at the end of the study. While this difference of body weight is easily measured, detecting such a small difference of energy intake or output is technically challenging. Similar considerations suggest the potential importance of alterations of gut absorption 
resulting in changes of energy excretion in urine and feces which were not directly measured in the present study, but were included in our calculations of energy output.
These observations highlight an important fact that is often overlooked in studies of obesity: changes of body weight indicate past cumulative energy imbalance and understanding the genesis of body weight differences requires knowledge about this history in addition to measurements made at isolated time points 
. Because the relationship between steady-state energy expenditure versus body weight has such a shallow slope, isolated measurements of energy intake or output rates may be unable to detect statistically significant differences between groups that have clearly different body weights. This difficulty may lead to the potentially erroneous conclusion that the lack of a statistically significant difference implies that there is in fact no difference and therefore altered energy efficiency must underlie the body weight differences – often illustrated by somehow normalizing the measurements for body weight 
. In contrast, by taking into consideration the history of energy imbalance, the present study suggests that very small increases of energy intake likely explain the persistent obesity observed in the HF-C and EN-C mice.
The pattern of body fat accumulation in obesity is believed to play a role in disease risk 
and we found interesting dynamics of body fat accumulation in the various fad pads during the development of obesity (). Both obesigenic diets led to dramatic increases of all fat pads at weeks 7 and 19 versus the modest increases observed in the chow group. However, not all fat pads monotonically increased in size on the obesigenic diets. Rather, the epididymal fat pads decreased by more than 25% between weeks 7 and 19 in the HF and EN groups despite increases of the other fat pad masses as well as overall body fat. This observation is consistent with a recent report showing a non-monotonic relationship between epididymal fat mass and body weight that was associated with adipose tissue remodeling during the development of diet-induced obesity 
. In the mice that were switched from the obesigenic diets back to chow, all fat pad masses were decreased at week 19 in comparison to their corresponding masses at week 7. However, in reflection of their persistent obesity, the HF-C group retained significant elevations of mesenteric, epididymal, and inguinal fat pad masses versus the C group at week 19 and a similar trend was observed in the EN-C group.
Both obesigenic diets resulted in increased serum leptin, insulin, and glucose, but triglycerides were paradoxically lower. After the switch to chow, all serum measurements returned to control values at week 19, with the exception of leptin which tended to remain increased in the persistent obese HF-C and EN-C mice versus the control group (). The HF and EN mice that continued their obesigenic diets for the entire 19 weeks had further deterioration of their metabolic state with increased serum glucose, insulin, and leptin concentrations while curiously maintaining lower serum triglycerides ().
Paradoxically decreased circulating triglycerides (TG) have been previously found in high fat fed C57BL/6 mice despite their insulin resistant state 
. While reduced dietary carbohydrate might explain these previous observations and why we also found lower serum TG in the HF group versus the C group, the EN group consumed the same amount of carbohydrate as the C group but also had decreased serum TG. Since both the EN and HF groups consumed similarly high amounts of dietary fat, perhaps the increased fat intake (rather than the reduced carbohydrate intake) caused the observed lowering of circulating triglycerides either by suppressing TG production and/or increasing TG clearance. Further support for this hypothesis comes from the study by Meugnier et al. who measured decreased circulating TG concentrations after adding 550 kcal/d of fat to the daily diet of lean men while keeping carbohydrate intake constant 
A possible alternative mechanism for the reduced serum TG levels could have been the elevated insulin concentrations observed in both the HF and EN groups since insulin is known to acutely inhibit hepatic VLDL secretion 
and stimulate adipose TG uptake via LPL 
. However, chronic elevations of insulin consequent to insulin resistance typically go along with high VLDL secretion and correspondingly high circulating TG 
. Furthermore, both HF and EN groups had increased rates of whole body lipolysis as reflected by elevated glycerol appearance rates () and increased serum glycerol and FFA concentrations (). Thus, adipose tissue was not responding to the elevated insulin and the resulting increased FFA supply to the liver would be expected to increase VLDL secretion and thereby increase rather than decrease serum TG. Clearly, more work is required to understand the mechanism for lowering serum TG during high fat feeding.
The cause of the obesity epidemic is typically attributed to an environment that promotes excessive food intake and limits physical activity 
. Thus, a possible solution to the obesity problem involves re-engineering the obesigenic environment towards one that would not have resulted in obesity. The present study raises the possibility that even such a dramatic change of the environment may not be sufficient to completely reverse obesity. Nevertheless, the return to the non-obesigenic environment allowed the persistently obese mice to normalize their previously increased serum concentrations of glucose, insulin, leptin, free fatty acids, and glycerol, while their paradoxically low serum triglycerides increased to normal values. While the energy intake rate of the persistently obese mice was within 1 kcal/d of the control mice at the end of the study, we believe that such a small increase likely explains the persistent elevation of body weight and fat mass.