Nardilysin in adipocytes regulates UCP1 expression and body temperature homeostasis

Brown adipose tissue (BAT) dissipates chemical energy as heat through uncoupling protein 1 (UCP1). The induction of mitochondrial reactive oxygen species (ROS) in BAT was recently identified as a mechanism that supports UCP1-dependent thermogenesis. We previously demonstrated that nardilysin (NRDC) plays critical roles in body temperature homeostasis. Global NRDC-deficient (Nrdc–/–) mice show hypothermia due to a lower set point for body temperature, whereas BAT thermogenesis at room temperature (RT) is enhanced mainly to compensate for poor thermal insulation. To examine the primary role of NRDC in BAT thermogenesis, we generated adipocyte-specific NRDC-deficient (Adipo-KO) mice by mating Nrdc floxed (Nrdcflox/flox) mice with adiponectin-Cre mice. Adipo-KO mice showed hyperthermia at both RT and thermoneutrality. They were also more cold-tolerant than Nrdcflox/flox mice. However, UCP1 mRNA levels were significantly lower in Adipo-KO BAT at RT, thermoneutrality, and 4 °C, whereas no significant differences were observed in UCP1 protein levels at RT and 4 °C. We examined the protein stability of UCP1 using the cycloheximide chase assay and found that NRDC negatively regulated its stability via the ubiquitin–proteasome pathway. NRDC may be also involved in ROS-mediated in vivo thermogenesis because the inhibitory effects of N-acetyl cysteine, an ROS scavenger, on β3 agonist-induced thermogenesis were stronger in Adipo-KO mice. Collectively, the present results demonstrate that NRDC in BAT controls adaptive thermogenesis and body temperature homeostasis possibly via the regulation of UCP1 protein stability and ROS levels.

Telemetric body temperature monitoring and respiratory gas analysis. Telemetric body temperature monitoring was performed on conscious mice using a wireless ETA-F10 Mouse Implantable Transmitter System by Data Science International (New Brighton, MN, US) as previously described 21 in the same chamber in which oxygen consumption (VO 2 ) was measured at 23 or 30 °C. Briefly, the transmitter was implanted into the abdominal cavity of mice under anesthesia. Mice then recovered for at least 5 days before data collection. Telemetry data were continuously collected for 2 days and analyzed using HEM-Evolution software (Notocord systems) for a basal core body temperature. Twelve-week-old Nrdc fl/fl and Adipo-KO mice were held individually in a chamber at 23 or 30 °C for 1 week in order to attain a constant respiratory exchange ratio. A gas analysis was performed using indirect calorimetry (Comprehensive Lab Animal Monitoring System: Oxymax/CLAMS, Columbus Instruments, TX, US). A β 3 -adrenergic agonist (BRL-37344) and N-acetyl cysteine (NAC) were injected intraperitoneally at 0.80 mmol/kg and 500 mg/kg, respectively, to examine the effects of these reagents, followed by the measurement of VO 2 for another 100 min.
Respiratory gas analysis and BAT temperature monitoring under anesthesia. Respiratory gas analysis by Oxymax/CLAMS was started immediately after the intraperitoneal injection of urethan (1.3 g/kg). After 100 min measurement of VO 2 , a small incision was made on the back and two temperature sensing probes (Physitemp) were inserted into the interscapular BAT and rectum to measure BAT and rectum temperature. The measurement of BAT and rectum temperature was started at 120 min after urethan injection.
Blood flow measurement. Mice were anesthetized with isoflurane inhalation and blood flow in the soles of the mice was measured using a Laser Speckle Blood Flow Imager (OMEGAZONE OZ-2; OMEGAWAVE, Inc., Tokyo, Japan).

Cold tolerance test.
In the cold tolerance test, 12-week-old Nrdc fl/fl and Adipo-KO mice were individually housed in cages at 4 °C for 5 h. Core body temperature was measured using a rectal temperature probe (Physitemp) immediately after 5 h of cold exposure at an ambient temperature of 4 °C. Mice were then killed by cervical dislocation and BAT was collected.
Western blot analysis. The preparation of tissue or cell extracts and a Western blot analysis were performed as previously described 14,21 . Rabbit polyclonal anti-UCP1 (Sigma), anti-PGC1α (Merk-Millipore), anticatalase (Abcam) and anti-SOD1 antibody (ProteinTech) were used as the primary antibodies. Rat monoclonal anti-NRDC antibodies (#1, #135) were raised in our laboratory 21 . Anti-α-tubulin (ProteinTech) and anti-β actin Histological procedures and immunohistochemistry. The immunohistochemical staining of mouse sections was performed as previously described 18 . An anti-UCP1 (Sigma) antibody was used as the primary antibody, the specificity of which was validated by staining BAT sections from UCP1-deficient mice (Fig. 7A). The sections of wild-type and UCP1-deficient mice were kindly provided by Dr. Yuko Okamatsu & Masayuki Saito (Hokkaido University). Images were obtained by fluorescence microscopy (BZ9000, Keyence) and confocal microscopy (SP8, Leica).

Measurement of lipids.
Lipid droplets were isolated from BAT using the Lipid Droplet Isolation Kit (Abcam) in accordance with the manufacturer's protocol. Triglycerides and total cholesterol in isolated lipid droplets were measured using LabAssay Triglyceride and LabAssay Cholesterol (Fujifilm/Wako), respectively.
Real time-PCR. Total RNA extraction, cDNA synthesis, and a quantitative real-time PCR analysis were performed as previously described 21 . Specific primers used for the real-time PCR analysis are shown in Supplementary Table 1.
Cell culture and transfection. Brown preadipocytes were isolated from neonatal Nrdc −/− mice and immortalized as previously described 21 . These cells were stably transfected with a Tet-on system vector for the doxycycline-dependent expression of NRDC to establish Nrdc −/− ; Tet-on-NRDC cells. Differentiation to mature adipocytes was performed as described elsewhere 25 . Briefly, cells were grown to confluence in DMEM (4.5 g l −1 glucose) supplemented with 10% fetal bovine serum, antibiotics, 17 µM pantothenic acid, 33 µM biotin, 100 µM ascorbic acid, 1 µM octanoic acid, and 50 nM 3,3′,5-triiodo-l-thyronine (T3) (adipocyte culture medium). Differentiation was induced by treating confluent cells for 48 h in adipocyte culture medium supplemented with 10 mg ml −1 insulin and 2.5 µM dexamethasone. After a 48-h induction, the culture medium was changed to adipocyte culture medium supplemented only with 10 mg ml −1 insulin and cells were cultured for another 96 h. All experiments with brown adipocytes were performed following full differentiation unless otherwise indicated. In the cycloheximide chase assay, mature brown adipocytes were treated with or without cycloheximide (50 μg/ml) in combination with or without MG132 (10 μM), followed by sample collection 0, 2, and 4 h after the treatment. The transfection of plasmid DNA was performed using HilyMax (Dojindo) according to the manufacturer's protocol.
Statistical analysis. Data are presented as mean values ± the standard deviation (SD) or standard error (SE) as specified. Differences between treatments, groups, and genotypes were analyzed using an unpaired 2-tailed Student's t-test for 2 groups or a one-way ANOVA (the post hoc Tukey-Kramer HSD test) for more than 3 groups. Mixed model repeated-measures analysis was used to analyze repeated measurements. Statistical analyses were conducted using the JMP software package (SAS institute).

Results
Hyperthermia in adipocyte-specific NRDC-deficient mice. To specifically assess the role of NRDC in adipose tissue, we generated adipocyte-specific NRDC-deficient (Adipo-KO) mice by mating Nrdc flox/flox 22 mice with adiponectin-Cre transgenic mice 24 . In Adipo-KO mice, NRDC mRNA levels were significantly reduced in both BAT and WAT, but not in the liver or heart (Fig. 1A). We also confirmed that the protein expression of NRDC was markedly decreased in BAT (Fig. 1B). Adipo-KO mice were born at the expected Mendelian frequency and showed no overt phenotypes. Under a normal diet, no significant differences were observed in body weight between Nrdc flox/flox and Adipo-KO mice until the age of 12 weeks (Fig. 1C).
To characterize the thermogenic phenotype of Adipo-KO, we initially monitored their core body temperature using the telemetry system. Body temperature measured every minute at RT (23 °C) is shown in Fig. 2A, and hourly averages are described in Fig. 2B. At most of the time points indicated, Adipo-KO mice showed a significantly higher core body temperature than Nrdc flox/flox mice, whereas the half day average was significantly higher only in the light time (Fig. 2C). We then measured body temperature at thermoneutrality (30 °C), at which mice maintain their body temperature by basal metabolism at rest and adjustments to skin blood flow 26 . Adipo-KO mice showed hyperthermia in the thermoneutral zone; however, the difference between Nrdc flox/flox and Adipo-KO mice was smaller at 30 °C ( Fig. 2D-F).
Unchanged VO 2 , but lower physical activity in Adipo-KO mice. Since the higher core body temperature in Adipo-KO mice suggested hypermetabolism, we measured VO 2 using indirect calorimetry. No significant differences were observed in VO 2 at 23 or 30 °C between Nrdc flox/flox and Adipo-KO mice ( Fig. 3A-C). Furthermore, no significant differences were noted in the respiratory exchange rate, suggesting no significant change in the energy source in the two genotypes ( Fig. 3D-F). We then analyzed the physical activity of mice because major constituents of energy expenditure are adaptive thermogenesis and physical activity. The physical activity of Adipo-KO mice was significantly reduced at 23 °C, indicating that adaptive thermogenesis was enhanced in Adipo-KO mice ( Fig. 3G-I).
Since body temperature is influenced by the balance between heat production and dissipation, hyperthermia may be due to better insulation. The level of insulation may be shown by the slope of increasing VO 2 with decreasing ambient temperature. The lack of a significant difference in VO 2  www.nature.com/scientificreports/ and Adipo-KO mice indicated no differences in insulation. We also examined skin blood flow as an indicator of heat dissipation in a hindlimb and found no significant differences between the genotypes (Fig. 3J). These results suggested that the higher core body temperature in Adipo-KO mice was due to enhanced thermogenesis.

Increased VO 2 and BAT temperature in Adipo-KO mice under anesthesia.
To exclude the effect of physical activity on energy expenditure, we next measured VO 2 and BAT temperature in Nrdc flox/flox and Adipo-KO mice under anesthesia by urethan. VO 2 in Adipo-KO mice was higher all the time after urethan injection (time 0) and there was a significant difference in the interaction between genotype and time between 50 and 100 min after the anesthesia, during which body movements ceased and the oxygen consumption values stabilized ( Fig. 4A-C). After the measurement of VO 2 , we measured BAT temperature by inserting a probe with the simultaneous measurement of the core body temperature (rectum). BAT temperature was consistently higher than the rectum temperature by 1-1.5 °C, and BAT and core temperatures changed in parallel.
Notably, BAT temperature in Adipo-KO mice was significantly higher than that of control mice, indicating that BAT thermogenesis is enhanced in Adipo-KO mice (Fig. 4D). Heart rate under anesthesia was measured as an index of sympathetic activity, but no significant difference was observed between Nrdc flox/flox and Adipo-KO mice Reduced mRNA expression of UCP1 in Adipo-KO BAT despite hyperthermia. To assess the thermogenic state in the BAT of Adipo-KO mice, we examined its histology and the expression levels of thermogenic genes. Despite the hyperthermic phenotype, BAT histology at thermoneutrality and RT showed no significant differences between control and Adipo-KO mice (Fig. 5A,B). Furthermore, UCP1 mRNA levels were signifi- www.nature.com/scientificreports/ cantly lower in Adipo-KO BAT than in control mice at 30 and 23 °C (Fig. 5C). Unexpectedly, a critical upstream regulator of UCP1 transcription, PGC1α mRNA levels at 30 °C, were significantly higher in Adipo-KO, while no significant differences were observed between Nrdc flox/flox and Adipo-KO mice at RT (Fig. 5C). Consistent with mRNA levels, the protein expression levels of UCP1 in Adipo-KO BAT were significantly lower than those in Nrdc flox/flox BAT at 30 °C (Fig. 5D,E). At 23 °C, which is a mildly cold environment for mice, UCP1 and PGC1α protein levels in Adipo-KO BAT were similar between control and Adipo-KO mice (Fig. 5F,G).

The cold induction of thermogenic genes is impaired in Adipo-KO mice.
To clarify the role of NRDC in adipocytes on cold-induced BAT adaptive thermogenesis, we examined the cold tolerance of Adipo-KO mice at 4 °C for 5 h. The measurement of core body temperature showed that Adipo-KO mice were significantly more tolerant to the cold environment than control mice (Pre-exposure: Nrdc flox/flox 37.04 ± 0.34 °C versus Adipo-KO 37.12 ± 0.46 °C; no significant difference, Post-exposure: 35.10 ± 0.4 °C versus 36.32 ± 0.54 °C; p = 0.0025) (Fig. 6A). In control mice, the mRNA levels of Ucp1, PGC1α, and other thermogenesis-related genes were markedly induced by cold exposure. In contrast, the cold induction of these thermogenic genes, except for the β3-adrenergic receptor, was significantly weaker in Adipo-KO mice than in Nrdc flox/flox mice (Fig. 6B). However, UCP1 and PGC1α protein expression, measured semi-quantitatively by Western blotting, was similarly induced by cold in control and Adipo-KO mice, indicating a discrepancy in the mRNA and protein levels of UCP1 (Fig. 6C,D). UCP-1 protein expression was also evaluated by immunofluorescent staining. First, we vali-  www.nature.com/scientificreports/ dated the anti-UCP1 antibody for immunohistochemistry using BAT sections from UCP1-deficient mice, and found that the staining is highly specific for UCP1 (Fig. 7A). The immunohistochemical staining demonstrated an equivalent level of staining in Nrdc flox/flox and Adipo-KO mice at 23 and 4 °C (Fig. 7B,C). We also examined lipid contents, triglycerides, and total cholesterol in the BAT extract, and found no significant differences in lipid levels between Nrdc flox/flox and Adipo-KO BAT (Supplementary Fig. 1). Our observation of UCP1 expression at 4 and 23 °C suggests that the translated UCP1 protein is more stable in Adipo-KO mice than in control mice.

NRDC promotes the protein degradation of UCP1. To assess whether NRDC contributes to UCP1
protein stability, we established brown adipocyte cell lines in which the expression of NRDC may be temporarily regulated. Specifically, brown preadipocytes isolated from Nrdc −/− mice and immortalized 21 were stably transfected with the Tet-on system vector for the doxycycline-dependent reintroduction of NRDC (Nrdc −/− ; Tet-on-NRDC cells). Nrdc −/− ; Tet-on-NRDC cells, treated with or without doxycycline, were fully differentiated into mature brown adipocytes and then treated with or without cycloheximide for 4 h (Fig. 8A). In this experimental setting, doxycycline / the presence or absence of NRDC did not affect the differentiation state of adipocytes, as demonstrated by the appearance (Fig. 8B) and mRNA expression of Pparg2 and Fabp4 (Fig. 8C). The cycloheximide chase assay revealed that UCP1 protein expression levels were significantly higher in NRDC-depleted cells than in doxycycline-treated NRDC-positive cells (Fig. 8D). These results suggest that NRDC promotes the protein degradation of UCP1. We then examined the effects of MG132, a potent proteasome inhibitor, on NRDCinduced UCP1 protein degradation. The simultaneous treatment of fully differentiated Nrdc −/− ; Tet-on-NRDC cells with MG132 and cycloheximide clearly prevented cycloheximide-induced decreases in UCP1 expression, suggesting that NRDC promotes the degradation of UCP1 through the ubiquitin-proteasome pathway (Fig. 8E).
In the next experiment, we treated cells with doxycycline after the completion of Nrdc −/− ; Tet-on-NRDC cell differentiation (Fig. 8F). The 1-day induction of NRDC significantly decreased UCP1 protein expression levels, which were restored by the MG132 treatment (Fig. 8G). Collectively, these results suggest that the impaired cold induction of UCP1 mRNA was compensated by its protein stabilization in Adipo-KO BAT.

NRDC is involved in the ROS-mediated regulation of thermogenesis.
Cold exposure induces the generation of mitochondrial ROS, which is indispensable for UCP1-dependent thermogenesis in BAT 10 . To test whether NRDC is involved in the regulation of ROS, we analyzed the mRNA levels of ROS-related genes in BAT. Among these genes, sod1, sod2, and catalase, which are important for the ROS defense system, were significantly lower in Adipo-KO BAT after cold exposure (Fig. 9A). Consistently, protein expression of catalase and SOD1 showed a tendency to decrease in Adipo-KO (Fig. 9B,C). These results suggest that cold-induced ROS production was increased in Adipo-KO BAT. To assess the involvement of NRDC in ROS-mediated in vivo thermogenesis, we measured VO 2 in Nrdc flox/flox and Adipo-KO mice in the presence or absence of NAC, a ROS scavenger. While BRL37344, a β3 agonist, similarly increased VO 2 in control and Adipo-KO mice, the inhibitory effects of the simultaneous treatment with NAC were significantly stronger in Adipo-KO mice (Fig. 9D,E). These results indicated that the dependence on ROS by BAT thermogenesis was greater in Adipo-KO mice than in Nrdc flox/ flox mice.

Discussion
In the present study, we demonstrated that the core body temperature of adipocyte-specific NRDC-deficient mice was significantly higher than that of control mice. Body temperature in wild-type mice is strictly maintained by the balance between heat production and dissipation. Therefore, hyperthermia in Adipo-KO mice may be attributed to excessive heat production or reduced heat dissipation. We assessed VO 2 as total energy expenditure and physical activity, and found that the latter was reduced in Adipo-KO mice. Since thermogenesis and physical activity are major constituents of energy expenditure, these results indicate that thermogenesis was enhanced in Adipo-KO mice. Moreover, the measurement of VO 2 and BAT temperature under anesthesia directly indicated that BAT thermogenesis is enhanced in Adipo-KO mice. We also evaluated insulation in Adipo-KO mice by measuring VO 2 at different ambient temperatures of 23 and 30 °C. No significant differences were observed in insulation because VO 2 did not markedly differ at either ambient temperature between control and Adipo-KO mice. Collectively, these results suggest that hyperthermia in Adipo-KO mice was due to increased heat production. Our previous findings revealed higher oxygen consumption in NRDC whole-body knockout (Nrdc -/-) mice at 23 and 30 °C, albeit the difference in energy expenditure between wild-type (Nrdc +/+ ) and Nrdc -/mice was markedly smaller at 30 °C 21 . Since the slope of increasing metabolism with decreasing ambient temperature reflects the level of insulation, these results suggest that Nrdc -/mice are less insulated than Nrdc +/+ mice. The present results showed no significant differences in insulation between Nrdc flox/flox and Adipo-KO mice, indicating that NRDC in non-adipose tissue is responsible for the regulation of thermal insulation.
For the body temperature homeostasis, the central set point is another critical factor. We previously reported the lowered set point of body temperature in Nrdc -/mice. Although Nrdc -/mice showed normal thermogenic capacity at around the thermoneutrality, the body temperature stayed 1.5 °C lower than that of Nrdc +/+ mice at the ambient temperature 30 °C, and even at 33 °C 21 . A shift in the set point could be one reason for the higher body temperature in Adipo-KO mice. Certainly, although there should be an extra capacity for heat dissipation at 23 °C, body temperature in Adipo-KO mice was still maintained higher than the control mice, suggesting that the set point of BT is higher in Adipo-KO mice. Since NRDC deficiency in Adipo-KO mice is restricted to adipose tissue, if it affects the set point, the remote factors from adipose tissue would be important.
It currently remains unclear why the physical activity of Adipo-KO mice is reduced. Although NRDC in neuronal cells critically regulates myelination and axonal maturation 18  www.nature.com/scientificreports/ in neurons is not likely because Cre recombinase in adiponectin-Cre mice is exclusively expressed in adipose tissue 27 . While the contribution of dysregulated adipokine secretion cannot be ruled out, another possibility is that hyperthermia itself is the cause of hypokinesia. The difference in core body temperature between control and Adipo-KO mice was smaller at thermoneutrality, at which no significant difference was observed in physical activity. These results suggest the importance of the primary effect of hyperthermia on physical activity. Adipo-KO mice showed increased cold tolerance; however, the cold induction of the mRNA levels of UCP1 and several other thermogenesis-related genes was weaker. In the present study, we identified two possible factors under NRDC-deficient conditions that induced thermogenesis despite the low mRNA levels of UCP1; the enhanced stability of the UCP1 protein and increased ROS production. While a few studies suggested that UCP1 protein stability is regulated through the ubiquitin-proteasome pathway 8 , the underlying molecular mechanisms remain unclear. The results of the cycloheximide chase assay in vitro revealed that NRDC was involved in UCP1 protein stability via the MG132-sensitive ubiquitin-proteasome pathway. Since NRDC in Drosophila localizes to mitochondria and functions as a co-chaperone 28 , this metalloendopeptidase in mammals may also regulate UCP1 protein quality control. In mice, UCP1 protein stabilization by the deficiency of NRDC appears to be limited to when the ambient temperature is between room temperature and low temperature. At the thermoneutrality (30 °C), both mRNA and protein levels of UCP1 are significantly lower in Adipo-KO BAT, and there is no dissociation between mRNA and protein levels. To avoid the risk of hyperthermia, UCP1 has to be rapidly downregulated when the ambient temperature rises. Therefore, we speculate that the organism may be preparing more UCP1 degradation pathways at high ambient temperatures. These pathways may compensate for the effect of NRDC deficiency and degrade UCP1 as in normal conditions.
Recent studies have highlighted the importance of ROS production in UCP1-dependent thermogenesis. ROS levels are elevated by cold exposure or a β-adrenergic stimulus, which then activate thermogenesis in mouse BAT. Furthermore, the pharmacological depletion of ROS in vivo has been shown to impair the capacity for BAT thermogenesis 10 . The present results indicated that the ROS dependency of BAT thermogenesis was greater in Adipo-KO mice. Although we found the down-regulated expression of several genes responsible for the ROS defense system in Adipo-KO mice, the molecular mechanisms by which NRDC regulates the synthesis of ROS warrant further study.
In the present study, we focused on the role of NRDC in adipocytes on UCP1-dependent thermogenesis in BAT. However, recent studies indicated that there are many alternative systems for UCP1-independent adaptive thermogenesis, such as sarco/endoplasmic reticulum Ca 2+ -ATPase-dependent and creatine metabolismdependent pathways 9 . The involvement of NRDC in UCP1-independent thermogenesis cannot be excluded at this point. The more detailed characterization of Adipo-KO mice will deepen our understanding of the molecular system of BAT thermogenesis.