Activation of brown adipose tissue by a low-protein diet ameliorates hyperglycemia in a diabetic lipodystrophy mouse model

Long-term ad libitum dietary restrictions, such as low-protein diets (LPDs), improve metabolic health and extend the life span of mice and humans. However, most studies conducted thus far have focused on the preventive effects of LPDs on metabolic syndromes. To test the therapeutic potential of LPD, we treated a lipodystrophy mouse model IRFKO (adipose-specific insulin receptor knockout) in this study. We have previously shown that IRFKO mice have profound insulin resistance, hyperglycemia, and whitenng of interscapular brown adipose tissue (BAT), closely mimicking the phenotypes in lipoatrophic diabetic patients. Here, we demonstrate that 14-day of LPD (5.1% kcal from protein) feeding is sufficient to reduce postprandial blood glucose, improve insulin resistance, and normalize glucose tolerance in the IRFKO mice. This profound metabolic improvement is associated with BAT activation and increase in whole body energy expenditure. To confirm, we showed that surgical denervation of BAT attenuated the beneficial metabolic effects of LPD feeding in IRFKO mice, including the ‘browning’ effects on BAT and the glucose-ameliorating results. However, BAT denervation failed to affect the body weight-lowering effects of LPD. Together, our results imply a therapeutic potential to use LPD for the treatment of lipoatrophic diabetes.


Introduction
Metabolic syndrome, a worldwide epidemic with high socioeconomic cost, is a complex disorder driven primarily by obesity 1 . The key features of metabolic syndrome include hyperglycemia, dyslipidemia, insulin resistance, non-alcoholic fatty liver, and many other associated metabolic abnormalities 2 . Great strides have been made toward managing the metabolic pro les of metabolic disease patients, but effective strategies for xing metabolic syndrome still need to be discovered.
Dietary intervention to prevent or control obesity and type 2 diabetes has been used for decades and can be highly effective and affordable. Long-term ad libitum dietary restriction (i.e., 20-40% reduction in food intake) has been shown to improve metabolic health and extend the life spans of mice and possibly humans [3][4][5][6] . Since long-term voluntary dietary/caloric restriction is impractical for most people, diets that alter the levels of speci c macronutrients without decreasing caloric consumption are seen as more sustainable by researchers and the public 7 . For decades, research has focused on the relationship between dietary carbohydrate/fat ratio and the homeostasis of body weight and blood glucose control 8, 9 .
The role of dietary protein in metabolism has begun to be appreciated only recently 5,10−14 .
Recent comprehensive macronutrient analyses have demonstrated that a low-protein (LP), highcarbohydrate diet is as potent as the caloric restriction in preventing aging-induced metabolic dysfunctions in mice 5,15 . Consistently, LPD has been shown to promote weight loss, enhance insulin sensitivity, and increase energy expenditure in wild-type rodents 16,17 . In addition to data from model organisms, multiple long-term prospective cohort studies have shown that LPDs are associated with decreased mortality and cancer in humans. Conversely, high-protein diets are associated with insulin resistance, diabetes, and mortality 6,18−20 . Consistent with these prevention studies, a recent randomized, controlled trial showed that dietary protein restriction (0.8 g of protein/kg body weight) improved metabolic dysfunction in patients with metabolic syndrome 21 , further suggesting a therapeutic potential of LPD. However, the exact bene cial metabolic effects and mechanisms of LPD on diabetes, especially lipodystrophy-associated diabetes, are still unknown.
In this work, we used an adipose-tissue speci c insulin receptor knockout (IR FKO ) mouse model that we have recently developed 22 to mimic the phenotypes observed in lipodystrophy diabetes patients, i.e., hyperglycemia, insulin resistance, glucose tolerance, extremely low body fat, and whitening interscapular brown adipose tissue (BAT). We found that two weeks of LPD feeding profoundly decreased body weight, enhanced BAT browning, reduced postprandial blood glucose, improved insulin resistance, and normalized glucose tolerance in IR FKO mice. Mechanistically, we showed that LPD-induced glucose normalization but not body weight reduction is BAT activation-dependent. Our results imply a therapeutic potential to use LPD for the treatment of severe lipoatrophic diabetes.

Results
A low-protein diet improves glucose homeostasis in lipodystrophic diabetic mice.
To determine the therapeutic potential of the LPD, we subjected both male and female WT (IR ox/ ox ) and IR FKO (IR ox/ ox , AdipoQ-Cre) lipodystrophic diabetic mice to either a chow or LP diet. We monitored body weight and blood glucose twice a week. Strikingly, we observed that the fed postprandial blood glucose of the male IR FKO mice started to improve one week after the protein restriction and became close to normoglycemic at the end of the two-week treatment compared to those on a chow diet (Fig. 1A). We consistently observed a similar LPD-induced glucose-ameliorating effect in the female IR FKO mice ( Fig. 1B). In contrast to the bene cial effect observed in the lipodystrophic diabetic mice, we did not detect signi cant changes in the postprandial blood glucose in both the male and female WT mice after LPD, which is, however, likely due to the lean and healthy nature of the control mice (Figs. 1A & 1B).
In consistency with previous studies demonstrating body weight loss after LPD treatment 5,11,23,24 , male and female IR FKO mice lost about 14% of their body weight at the end of the two-week LPD treatment (Figs. 1C & 1D). Reduction in body weight could potentially lower blood glucose, improve insulin sensitivity, and ameliorate metabolic health of the IR FKO mice, however, the lower body weight observed in the IR FKO mice on LPD is unlikely to be the major contributing factor for the BG normalization, as the WT mice on the LPD group also lost about 7% of their body weight (Figs. 1C & 1D) without noticeable changes in the postprandial blood glucose levels (Figs. 1A & 1B). Blood glucose-lowering observed in IR FKO after LPD treatment is likely contributed by improved glucose tolerance and insulin sensitivity, as shown by the glucose and insulin tolerance tests (Figs, 1E & 1F). Due to the ectopic lipid accumulation, the IR FKO mice are severely hyperinsulinemic and hyperlipidemic. Interestingly, after the acute short-term protein restriction, in addition to the blood glucose-lowering effect, LPD IR FKO mice also have signi cantly reduced plasma insulin and triglyceride (TG) (Figs. 1G & 1H). Notably, despite reducing TG in the IR FKO mice, LPD did not affect plasma non-esteri ed fatty acids (NEFA) levels in WT and IR FKO mice (Fig. 1I). Taken together, our data showed that acute short-term protein restriction appears to have bene cial metabolic effects even in the severely diabetic lipodystrophic mice model. LPD activates BAT glucose uptake and thermogenesis.
To determine potential contributing factors for blood glucose and body weight reduction in the IR FKO mice after LPD, we euthanized the mice after the dietary treatment and collected tissue samples. As reported previously 22 , loss of white adipose tissue caused organomegaly in the IR FKO mice, as shown by the signi cantly increased BAT and liver tissue weight ( Fig. 2A). In line with the body weight data, we observed that LPD treatment signi cantly reduced BAT, heart, and liver tissue weight in the IR FKO mice ( Fig. 2A). Since LPD treatment has a strong body weight reduction effect in the IR FKO , reduction in the tissue weight could be directionally proportional to the body weight reduction. To con rm, we also determined the body weight-normalized tissue weight. Our data showed that the relative liver and BAT tissue weight in the IR FKO mice after LPD are indeed lowered (Fig. 2B).
Short-term LPD has previously been shown to promote lipid accumulation in the liver of WT rat 25 .
Consistently, our H&E staining showed larger, lipid-containing vacuoles in the IR FKO liver after LPD treatment (Fig. 2C) despite a reduction in the liver tissue weight after LPD (Figs. 2A & 2B). To con rm whether the protein restriction treatment would exacerbate IR FKO mice hepatosteatosis, we extracted lipids from the liver of IR FKO mice before and after LPD treatment. Consistently with previous study 25 and histological data, we observed a signi cant increase of TG in the WT liver after LPD feeding (Fig. 2D). In contrast, unexpectedly, our data showed that LPD did not increase liver TG and NEFA in the IR FKO mice ( Fig. 2D), which is inconsistent with the larger vacuole observed histologically. The reason for this discrepancy requires further studies for clari cation.
In contrast, H&E staining showed dramatically reduced lipid droplet size in IR FKO brown adipocytes 2 weeks after LPD treatment (Fig. 2E). Reduction in the lipid storage in IR FKO BAT occurred as early as 3 days after the onset of LPD treatment, and most lipid storage was depleted after one week of protein restriction (Fig. 2E). We know that excessive lipid accumulation in brown adipocytes causes BAT whitening and downregulation of BAT metabolic functions, as seen in the IR FKO mice 22 . To investigate whether reducing lipid storage in brown adipocytes restores IR FKO BAT function, we rst examine BAT in vivo glucose uptake capability. Our in vivo 2-deoxy-glucose uptake assay showed that LPD diet signi cantly increased glucose uptake in the WT BAT and fully restored glucose uptake capacity in the IR FKO mice BAT, which was initially impaired by the excessive lipid accumulation (Fig. 2F). In addition, we also examine thermogenic markers' gene expression. Our results showed that LPD feeding activated and fully restored downregulated UCP1 expression in the IR FKO mice (Fig. 2G). Taken together, our data clearly demonstrated that LPD is a potent activator of BAT metabolic functions. LPD increases energy expenditure in both WT and IR FKO mice.
To determine how the ablation of the insulin receptor in adipose tissues and consumption of LPD modulate energy balance in mice, we subjected both the WT and IR FKO mice to metabolic measurements in metabolic cages (Promethion Metabolic System, Sable). Mice were started on a chow diet and then switched to LPD after 5 days of measurement. Interestingly, we observed that IR FKO mice consistently showed lower oxygen consumption (VO2), carbon dioxide production (VCO2), and energy expenditure (EE) at the onset of the light cycle compared to WT mice on a chow diet (Figs. 3A & 3B). These Consistent with the changes observed in the BAT, IR FKO mice also showed signi cant upregulation of EE after the LPD feeding. However, unlike previous reports on a progressive increase in EE induced by LPD 27-29 , we observed an extremely rapid (within 12 h) increase in EE in both the WT and IR FKO mice after the diet switch. Even though the LPD affected both the light and dark cycles, the LPD-induced EE effect seems more pronounced for the light cycle, especially in the IR FKO mice. Using the hourly data, we performed a paired-wise comparison analysis. Our data showed that LPD induces WT mice EE in both the light and dark cycles but only in the light cycle in the IR FKO mice (Fig. 3C). In addition, we also observed that increases in EE after LPD feeding were not attributed to changes in physical activity (Fig. 3D). In addition, unlike previous reports on LPD-induced hyperphagia 16,30,31 , we did not observe an increase in food consumption in the WT mice on LPD (Fig. 3E). Despite signi cantly lower body weight at the end of the LPD treatment (Figs. 1B & 1C), IR FKO mice tend to consume a similar amount of food before and after the LPD switch. Our data suggests that the reduction in body weight after LPD treatment observed in both WT and IR FKO mice is likely attributed to an increase in EE but not food consumption.
BAT denervation abolished the LPD-induced blood glucose-lowering effect.
To determine whether activation of BAT is required for LPD-induced blood glucose normalization, we performed BAT denervation on IR FKO mice. Upon recovery, sham and denervated IR FKO mice were subjected to chow and LPD treatment. Interestingly, we observed that LDP failed to lower the blood glucose of BAT-denervated IR FKO mice ( Fig. 4A). At the end of the 14-day treatment, LPD-treated IR FKO mice had signi cantly lower blood glucose compared to IR FKO mice on a chow diet and denervated IR FKO mice on LPD. However, unlike blood glucose, BAT denervation did not affect LPD-induced body weight reduction (Fig. 4B).
To determine the impact of BAT denervation on LPD-induced increases in energy expenditure, we subjected an additional cohort of sham and denervated IR FKO mice to metabolic pro le measurements. Similarly, mice were started on a chow diet and switched to LPD after 5 days of measurement. Our data showed that LPD signi cantly increases the sham IR FKO mice EE during the light cycle, but this stimulatory effect was abolished in the denervated IR FKO mice (Figs. 4C & 4D). Consistent with our previous data (Fig. 3C), LPD did not have a signi cant impact on IR FKO mice dark cycle EE, and this is not affected by BAT denervation (Fig. 4E).
To visualize the impact of BAT denervation on LPD-induced morphological changes, we collected BAT from sham and denervated IR FKO mice after 2 weeks of LPD feeding. Our histological analysis showed that unlike the sham IR FKO mice that lost most of the ectopically accumulated lipid due to lipodystrophy after LPD treatment (Figs. 2E & 4F), BAT of denervated IR FKO mice remained lipid-lled even after LPD feeding (Fig. 4F).
To determine the effects of BAT denervation on thermogenesis, we examined the mRNA expression of well-characterized thermogenic markers. Consistent with the metabolic and histological data, our gene expression analysis showed that BAT denervation completely abolished LPD-induced UCP1, Dio2, and Previous studies demonstrated that LPD-induced metabolic bene ts are at least partially mediated by FGF21 27,28,32 . Since circulating FGF21 is primarily produced by the liver, we rst examined FGF21 mRNA expression in the liver. Consistently, we observed a signi cant increase in FGF21 in WT and IR FKO liver after 2 weeks of LPD feeding (Fig. 5A). Interestingly, we also observed a tendency of increase in liver FGF21 induced by IR FKO under basal condition (Fig. 5A).
Since our data so far supports a crucial role of BAT in mediating LPD-induced blood glucose-lowering effects, we also examined BAT FGF21 expression. Despite extremely low basal expression, we observed a signi cant increase in the mRNA expression of BAT FGF21 in both the WT and IR FKO mice (Fig. 5B). To con rm that increase in liver and BAT FGF21 expression will indeed increase circulating FGF21 levels, we examined plasma FGF21 in both WT and IR FKO mice at different LPD-treatment time points. Our immunoassay data con rmed that LPD indeed increases circulating FGF21 levels in both the WT and IR FKO mice ( Fig. 5C). Intriguingly, LPD-fed IR FKO mice showed signi cantly higher circulating FGF21 levels compared to the LPD-fed WT mice in all the time points examined (Fig. 5C).
To determine the impact of BAT denervation on LPD-induced FGF21 expression, we examined FGF21 expression of BAT in IR FKO -denervated mice. Our data showed that BAT denervation signi cantly reduces LPD-induced FGF21 expression in BAT and circulating FGF21 levels (Figs. 5D & 5E). Taken together, our current data suggests that FGF21 potentially mediates the LPD-induced blood glucose normalization in the IR FKO mice.

Discussion
In the past decades, the alarming rise in the prevalence of metabolic diseases, including insulin resistance and diabetes, has invigorated interest in developing complementary therapeutic strategies. Reducing dietary proteins has recently been shown to promote or preserve metabolic health in young mice and rats 29,30,33 and obesity-induced metabolic dysfunction rodent models [34][35][36] . This study tested the therapeutic potential of protein restriction (LPD) on a diabetic lipodystrophic mouse model (adipose tissue-speci c insulin receptor knockout, IR FKO ) 22 . The IR FKO mice display lipodystrophy associated with white and brown adipose tissue dysfunction, hepatosteatosis, and profound insulin resistance characterized by severe hyperglycemia. These phenotypes are consistent with other published pre-clinical models of lipodystrophy-associated diabetes and lipodystrophy diabetes in human patients [37][38][39] , making it a unique model for studying lipodystrophy diabetes. Moreover, since insulin signaling plays critical roles in various metabolic processes, including glucose uptake, lipolysis, and lipogenesis in adipocyte [40][41][42] , our model will also directly test the requirement of adipocyte insulin signaling in the LPD-induced bene cial metabolic effects.
Most studies conducted so far have demonstrated bene cial metabolic effects at least after 4 weeks of LPD treatment 15,35,43,44 . Unexpectedly, we observed a signi cant reduction in blood glucose as early as 10 days after initiating LPD treatment. By the end of 2 weeks of treatment, protein restriction normalized IR FKO mice hyperglycemia, reduced plasma insulin, improved glucose tolerance, and enhanced insulin sensitivity of IR FKO mice to WT levels. These parameters also improved in WT mice after 2 weeks of shortterm protein restriction, consistent with previous studies 34 . However, we did not observe glycemic variation in WT mice, likely due to intact mechanisms of glucose homeostasis in this genotype.
Simultaneously, LPD decreased the body weight of IR FKO and WT mice, consistent with previous studies 16, 21,29,34,35 . Since LPD reduces body weight in both genotypes and decreases glycemia only in IR FKO mice, we infer that the bene cial effects of LPD on glucose homeostasis are independent of body weight reduction in the IR FKO model.
Based on protein leverage hypothesis [45][46][47] , animals are expected to increase food consumption to compensate for the protein requirement. However, we did not observe altered food intake in the LPD experimental groups, which is consistent with recent studies that showed even protein levels of 1 and 2.5% did not induce hyperphagia 17,34,48 . Based on their data, they concluded that energy intake is linked to dietary fat and not protein or sucrose in C57BL/6 mice 34 . Nevertheless, it is important to note that other studies in mice and rats 29,48 did observe a small increment in energy intake in the animals which are protein restricted. Differences in energy intake could be due to the length of dietary treatment, the nutritional status of the experimental model before the treatment (fasted, calorie-restricted, ad-libitum fed, type of diet fed), macro and oligo-nutrient ratios of the diet, and accuracy of energy intake measurements. More accurate feeding behavior experiments are required to clarify this discrepancy.
Interestingly, WT and IR FKO mice showed increased EE the day after we introduced LPD with no energy intake or movement changes. While previous studies have also indicated an increment for EE with a 5% protein diet, this effect was observed after 3 days of treatment 16 . This discrepancy could be due to the differences in the intrinsic characteristics of the chosen indirect calorimetry system. Current commercial available indirect calorimetry systems have been shown to differ in their assessments of EE, food intake, and RER in mice, presumably due to differential sensitivities of the systems (resolution of endpoint detection, endpoint detection limits, or time-resolution between measurements, etc.), effects of system design upon the cage microenvironment ( oor type, air ow rate, noise levels and types, food hopper designs, environmental enrichment, etc.), and study design (diets supplied, ambient temperature, mouse strains, etc.) 49 . Nevertheless, our indirect calorimetry results suggest that the decreased body weight induced by LPD feeding is primarily due to increased EE but not reduced food intake. In line with this hypothesis and other previous reports 50, 51 , LPD feeding in WT and IR FKO mice strongly actives the BAT thermogenesis, a critical contributor to adaptive energy expenditure [52][53][54] .
It is worth noticing that the contribution of BAT to the LPD-induced bene cial metabolic effects could be exaggerated in IR FKO mice compared to WT mice. Due to their lipodystrophy, IR FKO mice have extremely low levels of subcutaneous white adipose tissue, which also signi cantly contribute to LPD-induced increases in energy expenditure through browning 55 . Based on these observations, we set out to study if sympathetic activation of BAT is required for the positive effects of LPD in IR KFO mice. It is important to note that besides thermogenesis, BAT also plays a critical role as a metabolic sink for glucose, lipid, and branch-chained amino acids (BCAA) 14,56−59 . Consistently, BAT surgical denervation attenuated LPDinduced metabolic changes, including the 'browning' effects on BAT, energy expenditure-promoting response, and the glucose-ameliorating results, but not the body weight-lowering effects. These results suggest that sympathetic BAT activation induced by LPD positively affects both energy and glucose homeostasis in IR KFO mice. Considering the genetic nature of the IR KFO model, the LPD-induced BAT glucose uptake is at least partially mediated by an insulin-independent mechanism, different from the previous view that LPD lowers blood glucose primarily through enhancing BAT insulin sensitivity 60,61 . This is in line with a previous study that has identi ed a novel sympathetic/β3-adrenoceptor/mTORmediated iBAT glucose uptake, independent of the classical insulin/phosphoinositide 3-kinase/Akt Notably, we found that FGF21 mRNA gene expression in the liver and BAT, as well as FGF21 plasma levels, are signi cantly upregulated after LPD feeding in both WT and IR FKO mice. Interestingly, in IR FKO mice, blunted-sympathetic input to BAT through denervation effectively reduced BAT expression and plasma levels of FGF21, suggesting that intact sympathetic stimulation is necessary for IR FKO BAT to produce FGF21 in response to protein restriction. Considering the essential role of FGF21 in glucose utilization, lipid metabolism, and whole-body energy balance 28,44,63 , the denervation-induced drop of BAT and plasma FGF21 could potentially contribute to the attenuated bene cial metabolic effects induced by LPD feeding.
Another interesting observation is that surgical denervation only partially reduced FGF21 gene expression compared to sham mice. This result indicates that protein restriction can partly stimulate the BAT FGF21 expression gene independently of insulin signaling and sympathetic stimulation. This could be explained by intrinsic amino acid sensing of BAT, including mTOR, GCN2, AMPK, ATF4 and/or endocrine signals to BAT from other tissues like the liver 33,67−69 .
Consistent with previous reports in metabolic syndromes induced by aging 28,44 , our results discussed here focusing on a lipodystrophy diabetes mouse model consistently showed the bene cial metabolic effects of protein restriction on glucose balance and energy homeostasis.
Taken together, our data suggests that protein restriction could be a potential therapeutic strategy for the treatment of severe metabolic syndromes.

CONTACT FOR REAGENT AND RESOURCE SHARING
Further information and requests for resources and reagents should be directed to and will be ful lled by the Lead Contact, Chong Wee Liew (cwliew@uic.edu) Experimental Model And Subject Details Animals Adipoq-Cre (#028030) and IR / (#006955) mice were initially obtained from the Jackson Laboratory.
Both lines are on a C57BL/6 background. IR / mice were crossed with Adipoq-Cre to generate a fatspeci c insulin receptor knockout (IR FKO ) mouse model. IR ox/ ox mice containing the Adipo-Cre allele were bred with IR ox/ ox littermates lacking the Adipo-Cre allele to generate the mice for the IR FKO experiments. IR ox/ ox mice with the Adipoq-Cre allele were termed IR FKO mice while littermates lacking the Cre allele were used as control mice in IR FKO experiments. Mice were housed in environmentally controlled conditions with a 12-h light/dark cycle and had free access to standard rodent pellet food and water. The animal protocols were approved by the Institutional Animal Care and Use Committee (IACUC) of University of Illinois at Chicago. Animal care was given in accordance with institutional UIC and ARRIVE guidelines. 6-8 week-old mice were used for all the experiments. Body weights and glycemia were measured every 3 days from the start of treatment until tissue collection.

Diets
Control mice were fed chow diet (17% fat, 25% protein and 58% carbohydrate by kcal; #7012, Envigo, Indianapolis, Indiana, USA). Low protein-amino acid de ned diet (TD. 140918) was designed for a reduction of total protein content to 5% (kcal/from). Complete macronutrient composition: kcal from protein 5%, carbohydrates 76.4% and fat 18.5%. Caloric density: 3.9 kcal/g. Diet was color coded orange by manufacturer. Detailed information on diet composition can be seen on Table 1 for chow diet and  Table 2 for LPD.   Surgical BAT denervation was performed as previously described 70,71 . Eight-week-old mice were used for this experiment. On the day of surgery, each mouse was weighed and anesthetized with iso urane. The mouse was shaved and secured on a warm surgical table. Following a standard skin disinfection procedure with ethanol and iodine swabs, a lateral incision was made to expose the interscapular fat pads. On both sides, all ve branches of intercostal sympathetic nerves connecting to the right and left BAT fat pads were identi ed, carefully isolated, and sectioned. After denervation, the interscapular fat pads were returned to their original positions. Mice were allowed to recover for 7 days post-surgery before being exposed to the experimental conditions.
In vivo glucose uptake assay Glucose uptake assay protocol was adapted from C. Ronald Kahn's group (Joslin Diabetes Center) method 72 . In short, overnight fasted mice were anesthetized using Tribromoethanol (Avertin) 20 minutes prior to tail vein injection of glucose tracer (14C deoxy glucose, 0.1 uCi/g, PerkinElmer  Values are expressed as average grams consumed by day for absolute values and % of BW consumed per day for relative values. N=5-8 as per group. All data are presented as the mean ± standard error of mean (SEM). Energy expenditure in gure (C) was analyzed using paired two-tailed student's T test comparing the average of day/night period values for each mouse before and after diet switch. Food consumption measurements in gure E were analyzed by one-way ANOVA test with a threshold for signi cance of 0.05 adjusted P value. *P>0.05; **P>0.01. Figure 4 BAT denervation abolished the LPD-induced blood glucose-lowering effect.