The gravitostat protects diet- induced obese rats against fat accumulation and weight gain

The gravitostat is a novel homeostatic body weight- regulating mechanism, mostly studied in mice, and recently confirmed in obese humans. In the present study, we explored the effect of weight loading on metabolic outcomes, meal patterns and parameters linked to energy expenditure in both obese and lean rats. Diet- induced obese (DIO) and lean rats were implanted with capsules weighing either 15% of biological body weight (load) or empty capsules (1.3% of body weight; controls). Loading protected against fat accumulation more markedly in the DIO group. In line with this, the obesity- related impairment in insulin sensitivity was notably ameliorated in DIO rats upon loading, as revealed by the reduction in serum insulin levels and homeostatic model assessment for insulin resistance index scores. Although 24- hour caloric intake was reduced in both groups, this effect was greater in loaded DIO rats than in loaded lean peers. During days 10- 16, after recovery from surgery, loading: (i) decreased meal size in both groups (only during the light phase in DIO rats) but this was compensated in lean rats by an increase in meal frequency; (ii) reduced dark phase locomotor activity only in lean rats; and (iii) reduced mean caloric efficiency in DIO rats. Muscle weight was unaffected by loading in either group. Dietary- obese rats are therefore more responsive than lean

weight (corresponding to 15% of their biological weight); this caused a reduction in biological weight, such that total body weight (biological + capsule) was approximately maintained. 1 Importantly, studies in mice revealed that those of higher body weight (induced by offering a high-fat diet) were more sensitive to the gravitostat body weight gain-preventing effects than lean animals. 3 This turned out to be of considerable clinical importance because obese human subjects had a reduced body weight and body fat in response to increased loading in a very recent randomised clinical trial. 2 The gravitostat sensor is assumed to reside in weight loading bones because mice lacking osteocytes do not reduce their biological weight, nor exhibit decreased fat mass in response to loading. 1 The effector systems of the gravitostat reduce food intake, 1 implying the engagement of brain pathways and potentially also endocrine signals of importance for feeding control. 4 The anorexigenic homeostatic hormone leptin 5 was rapidly ruled out as a mediating system because the body weight reducing effects of weight loading could also be detected in mice that lack leptin. 1 Moreover, obese rodents and humans become leptin resistant, 3,6-8 yet they remain responsive to the gravitostat. 3 Although weight loading does not appear to engage leptin, 1 it possibly targets pathways downstream to those shown to exert leptin's anorexigenic effects. [9][10][11] To better understand the reason behind the lower food intake in loaded animals, we established a gravitostat model in Sprague-Dawley rats, where we more precisely could measure meal size and meal frequency because it has been suggested that such data can provide useful information about the relative importance of the hypothalamus versus brainstem in the regulation of meal patterns. 12 In addition, we aimed to explore the effects of loading on parameters linked to energy expenditure, namely caloric efficiency (CE) and locomotor activity, which arguably are more reliable in rats than in mice. Finally, in addition to verifying beneficial effects of loading on glucose homeostasis and leptin levels, we assessed plasma levels of the orexigenic hormone ghrelin, aiming to determine whether decreased levels could contribute to the weight loss effect.

| Animals
Because loading appears to impact similarly on body weight progression in male and female mice, 1 only adult male Sprague-Dawley rats (10 weeks of age at arrival; Charles River, Sulzfeld, Germany) were used (n = 32). They were kept under standardised non-barrier conditions of a 12:12-hour light/dark photocycle (lights on 07.00 am) at 20-22℃ and approximately 50% relative humidity. Initially, the rats were group-housed (two or three rats per cage) and had ad lib.

| Analysis of body weight, diurnal feeding, meal pattern and CE
Post-loading, body weight and food consumption were monitored regularly. The duration of the study (3 weeks) was informed from studies in mice showing that the loading effect on body weight stabilises during week 2 and persists until the end of the study at 7 weeks. 1 Biological body weight was calculated by subtracting the capsule weight from the total body weight. Food intake was measured by weight (g) and converted into energy intake (kcal) to enable a better comparison between the different diets. Data showing the change in biological body weight during the first 13 days and food intake on day 10 and 16 have been reported previously in load and control obese rats. 1 An automated feeding and drinking monitoring system (TSE LabMaster, Project 4261; TSE Systems, Bad Homburg, Germany) was used to analyse detailed diurnal feeding, drinking and meal pattern. Food hoppers and water bottles were suspended on calibrated sensors that register food and water consumption. HFDfed DIO rats (both load and control groups) were transferred into the cages on day 10 after surgery for detailed 6-day measurements.
Chow-fed lean rats (both load and control groups) were housed in the cages for the entire study period and the same 1-week period was used for diurnal feeding, drinking and meal pattern analysis.
Data for meal analysis were collected as binary data every 10 seconds. Meal analysis was undertaken using LabMaster software (TSE Systems), whereby all meals occurring during the study period were recorded chronologically to allow the evaluation of single feeding bouts. The start of a meal was defined by food removal (≥ 0.5 g) and the meal ended when no further food removal occurred before the end of the inter-meal interval of 10 min. Caloric intake, meal frequency and meal size were summarised over different periods: dark phase (12 hours), light phase (12 hours) and total day (24 hours), and then averaged per rat and group. A 10-minute inter-meal interval and a minimum meal size of 0.5 g are commonly used in defining meals in rats. 13 CE was calculated as: CE = (body weight change/ caloric intake) × 100.

| Locomotor activity
To monitor locomotor activity, 16 operant conditioning chambers were withheld during the habituation and test sessions. Rats were always brought to the testing room 30 minutes prior to starting the measurements. In this paradigm, anxiety is unlikely to influence locomotor activity because, in addition to the 1-hour habituation session, the data were collected over 2 × 3-hour time periods, during which the rats should have adapted to their environment.

| Sacrifice, blood samples and body composition
All rats were sacrificed on day 24 post-surgery after an overnight fast. Rats were anaesthetised with isoflurane. A tail prick blood sample was taken for blood glucose concentration (Accu-Chek ® Compact Plus; Roche Diagnostics Scandinavia AB, Bromma, Sweden). Blood samples, taken by heart puncture, were either processed to serum for analysis of circulating hormones (leptin and insulin) and metabolites (cholesterol and triglyceride) or processed to plasma using 4-(2 -aminoethyl)-benzenesulfonyl-fluoride hydrochloride as a protease inhibitor for analysis of total ghrelin. Post-mortem, white adipose tissue (WAT) was dissected and weighed as a measure of body fat mass (epididymal WAT, s.c. WAT from the hind legs, perirenal including retroperitoneal WAT) and the gastrocnemius muscle from the right leg was dissected and weighed as a measure of lean body mass.
The sensitivity of the assay was 0.2 ng mL -1 for leptin and insulin, and 0.156 ng mL -1 for ghrelin. The intra-assay coefficients of variation (CV) were 6.80%, 2.06% and 5.61% for the leptin, insulin and total ghrelin assays, respectively. Samples from both cohorts were run simultaneously for each assay. The inter-assay CVs were 6.00%, 1.33% and 2.74% for the leptin, insulin and total ghrelin assays, respectively. Cholesterol and triglyceride were assayed in duplicate by colorimetric enzyme assays (cholesterol #TR13421, triglyceride #TR22421; Infinity™; Thermo Scientific, Middletown, VA, USA) using a multiconstituent calibrator (#1E65-05; Abbott Laboratories, Chicago, IL, USA). The intra-assay CVs were 12.18% and 7.83%, respectively. The homeostatic model assessment for insulin resistance (HOMA-IR) index was used for estimation of insulin sensitivity. It was calculated as: HOMA-IR = fasting serum insulin (mIU L -1 ) × fasting blood glucose (mmol L -1 )/22.5.

| Statistical analysis
Analyses were conducted using the SPSS, version 22 (IBM Corp., Armonk, NY, USA). Data were first checked for equal variances and normal distribution. Importantly, to be able to treat DIO and lean rats as separate groups, either two-way repeated measures (RM) ANOVA (for continuous variables) or two-way ANOVAs were always run first. We handled DIO and lean groups separately only when at least one of these three conditions was met to obtain (i) a significant Diet × Loading interaction; (ii) a significant effect of Diet and/ or Loading with a near-significant Diet × Loading interaction (set at P < 0.1); and (iii) only a significant effect of Diet and/or Loading, but always in conjunction with a scientific argument that would justify such data split. After splitting data according to Diet, continuous variables were analysed using one-way RM ANOVAs with Time as the within-subject factor and Loading (load vs control) as the between-subject factor, whereas discrete variables were analysed using one-way ANOVA (Loading). In general, one-way ANOVA tests were conducted when applicable at each data point after RM ANOVA analyses. P < 0.05 was considered statistically significant.
Data are presented as the group mean ± SEM. Statistical details of the main analysis include the P value, its corresponding F ratio value and, in parentheses, the degrees of freedom of the numerator and denominator used to calculate the F ratio.

| Effect of weight loading on biological body weight and food intake
The body weights before weight loading surgery were 586.6 ± 6.5 g (n = 16) for the DIO group and 559.3 ± 9.1 g (n = 16) for the lean group (Loading: F 1,30 = 5.928, P = 0.021). With these data, we were able to link the effectiveness of loading to different body weights and adiposity levels determined at the start of the procedure. The first two-way RM ANOVA "filter" for the entire period (day 0 to day 20 post-surgery) revealed a Time × Diet interaction (F 9,20 = 7.459, P < 0.001), a Time × Loading interaction (F 9,20 = 5.363, P = 0.001), and general effects of the factors Diet (F 1,28 = 9.644, P = 0.004) and Loading (F 1,28 = 4.501, P = 0.043) on body weight progression.
Collectively, these results suggest that following the implantation of the capsules, all rats initially lost weight regardless of the weight of the implanted capsules, including those in both control groups, as a response to the surgical trauma. However, the control groups recovered faster from surgery than the load groups and, after approximately 2 weeks, the control groups reached or were above their start weight, whereas load groups remained in a negative weight change for the whole study period ( Figure 1A,B). Taking into consideration the body weight curves ( Figure 1A (Table 1). Additionally, we also detected a Diet × Loading interaction (F 1,28 = 9.582, P = 0.004) and a general effect of the factor Loading (F 1,28 = 36.456, P < 0.001) on body weight change from pre-surgery and day 20 after loading surgery ( Figure 1C).
Following capsule implantation, all rats showed a very low caloric intake, including the control rats, while recovering from surgery   perirenal WAT (P = 0.002) ( Figure 2B), whereas lean rats only had a reduction in s.c. WAT (P = 0.020) (Figure 2A). The sum of the three WAT pads was reduced by loading in the DIO (P = 0.001) but not the lean group ( Figure 2C). The loading-dependent decrease in weight of the epididymal WAT (−36.5% for DIO rats vs 9.3% for lean rats; P < 0.001) and total WAT (−49.2% for DIO rats vs −24.8% for lean rats; P = 0.009) was greater in DIO than lean rats ( Table 1). The weight of the gastrocnemius muscle was not affected by either Diet or Loading factors ( Figure 2D-E).

| Effect of weight loading on diurnal feeding, meal pattern and CE
Food intake microstructure was analysed on days 10 to 16 postloading. Two-way ANOVA analyses for the entire 24-hour period re- Caloric efficiency (CE) was calculated as: CE = (body weight change/caloric intake) × 100. Data are presented as the mean ± SEM. *P < 0.05 and **P < 0.01 by one-way ANOVA tests (load vs control), n = 8 per group P = 0.035), but not the frequency of meals, as well as significant effects of the factor Loading on caloric intake (F 1,28 = 16.352, P < 0.001), meal size (F 1,28 = 12.754, P = 0.001) and meal frequency (F 1,28 = 5.481, P = 0.027). Likewise, we found a Diet × Loading interaction on caloric intake (F 1,28 = 6.729, P = 0.015). Although we did not detect Diet × Loading interactions for either meal size or meal frequency, we nevertheless split data according to Diet because, arguably, diet and/or loading-dependent differences in caloric intake should have been caused by changes in at least one of these two variables (meal size or frequency). In the DIO load group, there was a decrease in dark phase feeding (P = 0.032), light phase feeding (P = 0.006) and in 24-hourr feeding (P = 0.003) ( Figure 3B). Only dark phase food intake was significantly suppressed in the lean group by loading (P = 0.023) ( Figure 3A). Additionally, the decrease in mean 24-hour caloric intake as a result of loading was greater in the DIO (−31.7% vs DIO controls) than in the lean group (−8.8% vs lean controls) (P = 0.017) ( Table 1). Loading also impacted upon meal pattern. Rats in the DIO load group had a reduced meal size during the light phase (P = 0.010) ( Figure 3D). In the lean group, loading decreased meal size during the dark phase (P = 0.018), the light phase (P = 0.001) and the total day (P = 0.006) ( Figure 3C). Meal frequency (ie, meal number) in the lean group, unlike that of the DIO group, was increased by loading during the light phase (P = 0.001) ( Figure 3E-F).  Figure 3H). By contrast, CE was unaffected by loading in the lean group ( Figure 3G). The effect of loading on mean CE was greater in the DIO (−66.7% vs DIO controls) than in the lean group (−3.6% vs lean controls) (P = 0.047) ( Table 1).

| Effect of weight loading on locomotor activity
Two-way ANOVA revealed a Diet × Loading interaction for the dark-

| Effect of weight loading on circulating hormones, circulating metabolites and insulin sensitivity
Two-way ANOVA revealed significant effects of the factor Diet on

Loading decreased serum levels of leptin similarly in both DIO
(P = 0.004) and lean (P = 0.034) rats ( Figure 5A). In DIO animals, loading also decreased plasma insulin levels (P = 0.037) ( Figure 5B) and HOMA-IR index scores (P = 0.042) ( Figure 5D), whereas it increased plasma total ghrelin levels (P = 0.021) ( Figure 5C). In lean rats, loading did not affect total ghrelin levels ( Figure 5C), insulin levels or HOMA-IR index scores ( Figure 5B, D). Nonetheless, direct comparisons of the effects of loading between both DIO and lean rats did not reach statistical significance for any of the blood parameters studied ( Table 1). As expected, non-loaded DIO rats displayed a metabolic profile that was in accordance with their obese-like phenotype. Specifically, they exhibited higher leptin (P = 0.005), insulin (P < 0.001), total ghrelin (P = 0.018), glucose (P = 0.008) and  (Table 2).

| D ISCUSS I ON
The recent discovery that challenging the gravitostat mechanism through artificial weight loading reduces body fat and body weight in obese human subjects 2 provides an entirely new research field with respect to the pursuit of novel targets for the treatment of obesity in humans. To introduce a new therapy in humans, it is of utmost importance to first thoroughly investigate the mechanisms behind the anti-obesity effect, which requires work in experimental animals. Here, we demonstrate that the reduction of food intake caused by weight loading appears to reflect especially a decrease in meal size. This is important because meal size is especially regulated by pathways involved in satiation in the hindbrain, 12 thereby implicating these pathways in the effects of loading. We also found that CE was suppressed by loading in DIO rats, implying increased energy expenditure. Taken together with our finding that loading does not impact on locomotor activity in DIO rats, we may infer that loaded rats use more calories to move a given distance.
Changes in circulating levels of leptin, insulin, and ghrelin as well as insulin sensitivity that were consistent with weight loss and energy deficit.

| The gravitostat decreases body fat and food intake more effectively in obese than in lean rats
Comparing weight loss curves between DIO and lean rats, it appears that those animals with the greatest amount of body fat benefit most from the protective effect of loading on weight gain. Post mortem weighing of dissected tissues showed that the effects of loading were the result of a decrease in fat mass rather than a decrease in lean mass, again with the largest effect in the DIO group that had more fat to begin with. The plasma leptin levels, which reflect fat mass, 5 mirrored the body fat data, confirming a loss of fat by loading. Insulin levels in DIO rats were more than double those of the lean group (probably linked to the greater need for insulin in obese animals that gain insulin resistance) and these could also be reduced by loading in the DIO group. In addition, loading of DIO rats F I G U R E 5 Effect of weight loading on terminal circulating hormones and insulin sensitivity. Effect of weight loading on logarithmised concentrations of (A) serum leptin, (B) serum insulin, (C) total plasma ghrelin and (D) homeostatic model assessment for insulin resistance (HOMA-IR) in diet-induced obese (DIO) rats on high fat diet and Lean rats on chow. Circulating hormones were measured in the terminal blood sample after an overnight fast on day 24 after weight loading surgery. Data are presented as the mean ± SEM. *P < 0.05 and **P < 0.01 by one-way ANOVA tests (load vs control), n = 8 per group decreased the HOMA-IR index, indicating increased insulin sensitivity. Collectively, these data suggest that the gravitostat mechanism can be recruited by loading for weight loss and improved glucose metabolism in obese animals.
The duration of the present study in rats was 3 weeks. Our previous studies in mice have shown that the loading effect on body weight stabilises during week 2 and persisted until the end of that study at 7 weeks. 1 Further studies are required to determine how long the loading effect persists in obese rats and mice. What is known is that removal of the artificial weight reverses the suppression of biological body weight, supporting the assumption that the loading effect is not a result of stress.

| Gravitostat-induced decrease in fat mass is mainly caused by decreased food intake
In all loading studies performed in rodents to date, including the present study, the loss in fat mass (and body weight) appears to reflect a reduced food intake rather an increased energy expenditure.
Indeed, it was previously shown that DIO mice pair-fed to the load group lose the same amount of body weight as seen by loading, 1 suggesting that reduced food intake is the main factor causing the weight loss. Our finding that loading suppressed food intake, an effect that was greater in DIO rats, supports this conclusion. However, we demonstrate here that CE is decreased by loading in DIO rats, which could be indicative of enhanced energy expenditure, a logical effect of carrying an extra load. 14 Thus, loading decreases fat mass mainly by decreasing food intake, although the increase in energy expenditure likely contributes to this effect.

| The gravitostat and locomotor activity
We did not detect any effect of loading on locomotor activity in the DIO group, in line with data from mice. 1 Rats in the lean control group were more active during the dark phase than those in the DIO control group, which could contribute to the low fat mass of the lean group. Loading decreased locomotor activity in lean but not in DIO rats, and it might be speculated that this is a counter-regulatory effect that partly explains why fat mass is less decreased by loading in lean rats.

| The gravitostat in relation to caloric restriction
An extensively studied animal model of weight loss is caloric restriction in which animals have a limited supply of food over an extended time period. It typically ranges from 50% to 80% of ad lib. feeding. 15,16 In addition to reduced body weight, rats on caloric restriction have reduced body fat mass 16 and reduced insulin and leptin, 17 as also seen here after loading. One interesting difference is that the caloric restriction decreases energy expenditure, 18 an effect for which no evidence could be found after loading. Putting these observations together, it would appear that the reduction in food intake by loading is insufficient to cause energy-saving adaptations (eg, increased CE and/or reduced locomotor activity), However, it cannot be excluded that increased effort when carrying a weight could mask a decrease in energy expenditure caused by weight loss.

| The gravitostat in relation to ghrelin
Loading changed the levels of several circulating hormones. This included an increase in ghrelin in addition to decreases in circulating leptin and insulin. These effects are consistent not only with a reduced fat mass, but also a state of energy deficit and heightened hunger. Ghrelin is a stomach-derived hormone 19 that is released during hunger 20 and drives food intake and food motivated behaviour. [21][22][23][24] Heightened ghrelin in the absence of feeding in loaded rats resonates with that observed in models of anorexia nervosa 25,26 and could indicate resistance to ghrelin's orexigenic effects. Arguing against ghrelin resistance in loading, however, is our observation that ghrelin receptor knockout mice respond normally to weight loading with a decreased body weight and reduced food intake. 1

| The gravitostat neural circuitry
The neural identity of the circuits mediating the gravitostat mechanism remains elusive. Ideally, these circuits would be identified (by activity and/or gene expression studies) during the time window when body weight homeostasis is adjusting to loading but, unfortunately, this partly corresponds to a period of recovery from surgical trauma, which could also impact on these pathways. Candidate

| The gravitostat and the leptin system as guards against obesity
Our data in the present study seem to point to a more effective gravitostat-body-weight-regulating mechanism in DIO rats compared to leaner chow-fed rats. This in accordance with our previous finding that the gravitostat is more effective in obese than in lean mice. 3 In line with this, there were only small and inconsistent effects of loading in lean animals in some early studies. 30,31 Very recently, we also found that loading is effective in suppressing body weight in obese humans. 2 It could be speculated that obese animals (even severely obese ones) would be even more obese in the absence the gravitostat mechanism. As one example, Kim et al 32 have shown that severely obese leptin deficient ob mice can be manipulated to become even more obese. It could also be speculated that the gravitostat mechanism must be suppressed or reset during pregnancy, such that the mother can sustain her body weight at the same time as supporting foetal growth. Other examples of physiological roles of the gravitostat are regulation of prepubertal growth and possibly the timing of puberty, as we recently reported. 27,33 However, a remaining question concerns which pathophysiological events inactivate the gravitostat during adulthood, and thereby contribute to obesity.
Ablation of the fat-derived hormone leptin causes severe obesity that can be reversed by leptin replacement in animals 8,34 and humans. 35 This indicates that leptin can exert powerful biological effects and that it is needed to prevent grave obesity.
In conclusion, the leptin system and the gravitostat are the only two known homeostatic regulators of body fat. It will be of great interest to investigate how combined manipulations of these two powerful homeostatic systems may provide the long sought after cure of obesity.

ACK N OWLED G EM ENTS
We thank research engineer Staffan Berg for manufacturing the weight capsules, Dr Pol Solé-Navais for statistical advice and Jakob Bellman for assistance with blood lipid analysis. interpretation of data or the writing of the report, as well as the decision to submit the article for publication.

CO N FLI C T O F I NTE R E S T S
The authors declare that they have no conflicts of interest.

PE E R R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/jne.12997.

DATA AVA I L A B I L I T Y S TAT E M E N T
The datasets generated during and/or analysed during the present study are not publicly available, but are available from the corresponding author upon reasonable request.