Soluble LR11/SorLA represses thermogenesis in adipose tissue and correlates with BMI in humans

Thermogenesis in brown adipose tissue (BAT) is an important component of energy expenditure in mammals. Recent studies have confirmed its presence and metabolic role in humans. Defining the physiological regulation of BAT is therefore of great importance for developing strategies to treat metabolic diseases. Here we show that the soluble form of the low-density lipoprotein receptor relative, LR11/SorLA (sLR11), suppresses thermogenesis in adipose tissue in a cell-autonomous manner. Mice lacking LR11 are protected from diet-induced obesity associated with an increased browning of white adipose tissue and hypermetabolism. Treatment of adipocytes with sLR11 inhibits thermogenesis via the bone morphogenetic protein/TGFβ signalling pathway and reduces Smad phosphorylation. In addition, sLR11 levels in humans are shown to positively correlate with body mass index and adiposity. Given the need for tight regulation of a tissue with a high capacity for energy wastage, we propose that LR11 plays an energy conserving role that is exaggerated in states of obesity.

M urine studies have demonstrated that increasing heat production in brown adipose tissue (BAT) can offset obesogenic effects of a high-fat diet (HFD) 1,2 , whereas reducing thermogenesis exacerbates them 3 . The existence of BAT in adult humans has been 'reconfirmed' by using modern imaging techniques and tissue biopsies [4][5][6] , increasing the importance of understanding BAT's physiological regulation. In addition, levels of detectable BAT correlate negatively with age, body mass index (BMI) and diabetes 7 , and using intermittent cold exposure to increase amounts of BAT is sufficient to increase post-prandial energy expenditure and improve insulin sensitivity 8 . Collectively, these findings support the hypothesis that thermogenesis in BAT is an important factor contributing to human metabolism. Numerous secreted and circulating factors have been identified to date with the ability to enhance the amount and/or activity of BAT 9 . Given the need for tight regulation of a tissue with such a high capacity for energy wastage, it is somewhat surprising that comparatively few negative regulators have been described. Changes in levels of LR11 in humans have primarily been associated with Alzheimer's disease, linked to its potential role in amyloid processing in the brain 10 . More recently, LR11 has been shown to possess the required properties of a functional lipoprotein receptor and the soluble form (sLR11) has been shown to facilitate increased vascular smooth muscle cell migration in response to injury, hinting at an equally important peripheral role [11][12][13] . LR11 has also been shown to regulate lipoprotein lipase (LPL), a key enzyme regulating the supply of lipid to BAT, by coordinating both its anterograde and retrograde vesicular transport 14,15 .
Here we show that mice lacking LR11 are protected from dietinduced obesity because of a hypermetabolic phenotype, facilitated by enhanced thermogenic gene expression in white adipose tissue (WAT) depots. Our data demonstrate that sLR11 is able to bind to bone morphogenetic protein (BMP) receptors and suppress thermogenic signalling via suppression of downstream Smad protein phosphorylation. The expression profile of Lr11 and serum levels of sLR11 in mice increase at times of increased thermogenesis, suggesting a regulatory role of sLR11 in preventing excessive energy wastage. sLR11 levels correlate with adiposity in humans, leading us to conclude that sLR11 may be an important negative regulator of adipose tissue energy expenditure, which is dysregulated in obesity, potentially compromising metabolic responses to nutritional stimuli.

Results
Lr11 À / À mice are hypermetabolic and protected from obesity. We previously generated a mouse lacking LR11 to examine its role in vascular smooth muscle cells in vivo and that of the cleaved soluble form (sLR11) in vitro 11,16 . These studies highlighted a tendency for Lr11 À / À mice to be lighter than their wild-type (WT) littermates, and as such our central hypothesis was that LR11 was in some manner acting as a negative regulator of metabolism. To formally assess differences in body weight, WT and Lr11 À / À mice were fed standard chow or HFD to maximally expose differences in weight gain. The body weight of WT 8-weeks-old mice on chow diet was similar to those of Lr11 À / À mice; however, the size of the retroperitoneal and reproductive fat pads were significantly reduced (Table 1). In contrast, HFD-fed Lr11 À / À mice had markedly lower body weights than WT controls, a large portion of which was accounted for by reductions in the weights of WAT and liver (Table 1 and Supplementary Table 1). Reduced adiposity after 8 weeks of HFD was clearly evident from anatomical examinations ( Fig. 1a-f) and histological analyses revealed that Lr11 À / À mice had smaller lipid droplets in BAT (Fig. 1e) and were protected from hepatic steatosis (Fig. 1f). Analysis of weekly energy balance on chow diet confirmed no significant difference in rate of body weight change or daily food intake in 10-11-week-old mice ( Fig. 1g-i). Longer term monitoring of weight change in a separate study indicated a clear increase in weight gain in WT mice fed a HFD, with a high rate of divergence from the persistently lean Lr11 À / À mice observed by the comparative 10-weeks-old time-point (Fig. 2a). Averaged daily food intakes over the course of this study again confirmed no difference between genotypes (WT ¼ 59.9 ± 1.1, Lr11 À / À ¼ 60.5±1.2 kJ day À 1 ) or HFD (WT ¼ 70.1±3.8, Lr11 À / À ¼ 70.1 ± 2.9 kJ day À 1 ).
We hypothesized that increased energy expenditure was facilitating the reduction in the rate of weight gain in Lr11 À / À mice fed HFD. Assessment of energy expenditure using indirect calorimetry confirmed that Lr11 À / À mice had a higher metabolic rate than WT mice when fed HFD (30±0.4 vs 27.5±0.3 J min À 1 ) (Fig. 2b,c). Lr11 À / À mice also displayed improved serum biochemistry profiles, with lower fasting glucose, insulin and triglyceride levels, particularly on HFD (Fig. 2d-f). Other circulating factors such as free-fatty acids, T4, leptin and serum cholesterol were not significantly altered ( Fig. 2g-k), nor were any significant alterations in LPL activity observed in Lr11 À / À tissues (Fig. 2l).
Lr11 À / À mice have increased thermogenic markers in WAT. LR11 is known to have a predominantly neuronal expression profile 17 . However, analysis of Lr11 mRNA expression in WT adipose tissue revealed similar levels to those found in brain, with real-time quantitative PCR ct values of 26.9 ± 4.1 in brain versus 26.6 ± 0.6 in BAT and 28.7 ± 0.8 in subcutaneous WAT (scWAT) (n ¼ 8). We therefore examined whether phenotypic changes in adipose tissue may be driving the metabolic shift observed in Lr11 À / À mice. Reduced lipid content in Lr11 À / À BAT on HFD (Fig. 1e) suggested that the tissue was more metabolically, and thus thermogenically, active. Examination of scWAT revealed smaller, less lipid replete adipocytes (Fig. 3a), suggesting either reduced lipid storage or increased lipid oxidation. Previous studies have shown that scWAT has a high propensity for 'browning' in response to pharmacological and nutritional signals 18 . Therefore, we performed immunostaining, which confirmed significantly higher levels of uncoupling protein 1 (UCP1) than in WT tissue (Fig. 3b). Microarray analysis of gene expression revealed significant increases in thermogenic genes, particularly in Lr11 À / À versus WT scWAT (Supplementary  Table 2). Based on histological and microarray data we hypothesized that increased thermogenesis in adipose tissue could explain the increase in metabolic rate in Lr11 À / À mice. Increased expression of known key thermogenic factors was confirmed using quantitative PCR with reverse transcription, including BMP 8b (Bmp8b), which we have previously shown to be an important regulator of diet-induced thermogenesis (Fig. 3c) 3 . Similar gene expression changes were observed in canonical BAT, albeit to a lesser extent (Fig. 3d), likely because of the already high levels of thermogenic gene expression making detection of further modest increases more difficult. Other VPS10p or low-density lipoprotein receptor gene family receptors such as Sortilin 1 (Sort1) and low-density lipoprotein receptor related protein 1 (Lrp1) did not display any significant compensatory regulation in either scWAT or BAT (Fig. 3c,d). No changes to markers of vascular endothelial cells or the smooth muscle cell marker, alpha actin 2 (Acta2), were detected in either BAT or subcutaneous WAT of Lr11 À / À mice (Fig. 3e,f).
Increased thermogenic gene expression in Lr11 À / À mice fed an HFD, coupled with the reduced body weight suggested that Lr11 À / À mice may have increased diet-induced thermogenesis.
To functionally address changes in thermogenic capacity in response to increased energy availability, we assessed the thermogenic capacity of Lr11 À / À mice in response to HFD. To eliminate any effect of environmental temperature, mice were  . Lr11 À / À mice showed a greater increase in energy expenditure from baseline in response to norepinephrine (NE) injection compared with WT littermates, corresponding to B12 J min À 1 (Fig. 4a,b). This demonstrated that HFD-fed Lr11 À / À mice had increased BAT recruitment and thermogenic adaptation to HFD.
sLR11 inhibits O 2 consumption in brown adipocytes. Based on the thermoneutral HFD study data we hypothesized that a lack of LR11 allowed animals to mount a disproportionately high thermogenic response to increased sympathetic nervous system stimulation following an increase in caloric intake. Primary brown adipocytes (which normally have high endogenous thermogenic capacity) from WT and Lr11 À / À mice were cultured and differentiated in the presence or absence of purified sLR11 (ref. 20). Functional thermogenic capacity was assessed by NE-induced cellular respiration, as NE is the major signal from the sympathetic nervous system that stimulates thermogenesis in brown adipocytes 21 . Differentiated WT and KO adipocytes showed similar basal respiration levels but primary adipocytes from BAT of Lr11 À / À mice showed a greater response to NE stimulation compared with cells from WT mice (Fig. 4c). Conversely, treatment of Lr11 À / À cells with sLR11 almost completely ablated the enhanced effect of NE stimulation on oxygen consumption (Fig. 4d) and sLR11 significantly reduced lipolysis in WT brown adipocytes, as indicated by glycerol release in response to maximum stimulation with NE (Fig. 4e). Opposite to tissues lacking LR11, sLR11 treatment of WT primary brown adipocytes resulted in a significant reduction in expression of multiple genes essential for thermogenesis, including Ucp1, Pgc1b and b3 adrenergic receptor (Adrb3) (Fig. 4f). Indirect calorimetry expressed as a scatter plot showing the correlation between energy expenditure and body weight for singlehoused, 12-week-old wild-type (R 2 ¼ 0.75) and Lr11 À / À (R 2 ¼ 0.74) mice fed HFD (60% fat) from 8 to 16 weeks; and in (c) also expressed as an adjusted mean energy expenditure for an adjusted mean body weight of 29.29 g, P-value obtained using analysis of covariance, n ¼ 8-11. (d) Fasting glucose, (e) fasting insulin, (f) fasting serum triglycerides following 8 weeks of diet administration from 8-16 weeks old. (g) Fasting free-fatty acids following 1 and 4 weeks of HFD from 8-weeks-old mice and (h-k) serum measurements of hormones and lipids in 16-week-old male wild-type and Lr11 À / À mice fed HFD from 8 to 16 weeks, n ¼ 5-6. (l) Lipoprotein lipase activity in conditioned media from mature adipocyte cultures from murine BAT of WT and Lr11 À / À mice, using free-fatty acid production as a readout (n ¼ 3-4). Data are presented as means ± s.e.m. w, z and # denote significant effect of genotype, diet and genotype*diet interaction respectively, obtained using 2-way ANOVA. (See also Supplementary Table 2). FFA, free-fatty acid. ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms9951 sLR11 inhibits thermogenesis via the BMP/Smad pathway. Bmp7 and its downstream signalling cascade, which requires phosphorylation of Smad proteins, has been identified as a key mechanism regulating the formation of BAT and recruitment of 'beige' adipocytes within WAT [22][23][24][25] . To examine whether LR11 was interacting with this mechanism to regulate the observed browning of scWAT, we differentiated primary adipocytes from scWAT of WT and Lr11 À / À mice in the presence of recombinant BMP7. BMP7 is a known driver of thermogenic development and, as a member of the TGFb superfamily, signals through BMP receptors (BMPR) to increase phosphorylation of intracellular Smad proteins 25 . Cells lacking LR11 displayed a significantly greater increase in thermogenic gene expression in response to BMP7, and this response was completely blocked in cells of both genotypes by treatment with sLR11 ( Fig. 5a-c). Phosphorylation of Smads 1, 5 and 8 following BMP7 treatment of primary adipocytes was also greatly reduced by sLR11 co-treatment (Fig. 5d). sLR11 treatment was also blocked downregulation of the BMP receptor Bmpr1b, observed after BMP7 treatment (Fig. 5e,f) and significantly reduced phosphorylation of SMAD3 in adipocytes following stimulation with TGFb (Fig. 5g). Further evidence for the ability of sLR11 to   ARTICLE directly inhibit the BMP signalling pathway was provided by an immunoprecipitation assay, which showed that sLR11 was able to bind BMPR1a and BMPR1b, but not BMPR2 (Fig. 5h). Finally, reduced expression of either BMPR1a or BMPR1b in cultured brown adipocytes was sufficient to negate the repressive action of sLR11 on Ucp1 expression (Fig. 5i). The cytoplasmic domain of LR11 has also been shown to function as a transcription factor 26 , leading us to test whether LR11 could regulate the transcriptional activity of known thermogenic response elements. Overexpression of LR11 resulted in a small but significant reduction in transcription driven by the Bmp8b promoter, a thyroid response element and the Ucp1 promoter enhancer, but not by the Ucp1 core promoter ( Supplementary Fig. 1). We next examined the physiological regulation of LR11 and found its gene expression to be regulated in line with thermogenic activation, decreasing in both BAT and scWAT as environmental temperature rose from cold to thermoneutrality (Fig. 6a). Serum sLR11 levels followed the same pattern, reaching their highest point at 5°C when thermogenesis is maximally activated in mice (Fig. 6b). Lr11 mRNA was also regulated by changes to energy availability in a pattern identical to thermogenesis, increasing in the fed state and declining after fasting (Fig. 6c).
sLR11 levels correlate with BMI and adiposity in humans. Finally, to test the relevance of our findings to human metabolism and as LR11 is highly expressed in adipose tissue, we examined the relationship between adiposity and sLR11 in two cohorts of patients. In 156 subjects with sleep apnoea (Supplementary  Table 4), circulating sLR11 levels were positively correlated with age, BMI, visceral (VFA) and subcutaneous fat areas (SFA), and HbA1c and negatively correlated with smoking and high-density lipoprotein (HDL) cholesterol (Fig. 6d-f and Table 2). Multiple  (c) Cellular respiration in differentiated primary adipocytes from BAT of wild-type and Lr11 À / À mice in response to 100 nM NE, n ¼ 9-10. (d) Cellular respiration in primary adipocytes from wild-type BAT treated with vehicle or 10 ng ml À 1 of soluble LR11, n ¼ 9 or 10. (e) Lipolysis measured by glycerol release from differentiated brown adipocytes in response to increasing concentrations of norepinephrine þ / À 10 ng ml À 1 of sLR11 over 2 h, n ¼ 4. (f) Thermogenic gene expression in wild-type primary brown adipocytes, differentiated in the presence of vehicle or 100 ng ml À 1 soluble Lr11 (n ¼ 4 triplicate experiments). Data presented as mean±s.e.m., * ¼ Po0.05 using oneway ANOVA followed by Tukey post-hoc test where appropriate. w, z and # denote a significant effect of genotype, treatment and genotype*treatment interaction respectively, using 2-way ANOVA, for (d) w ¼ effect of sLR11 treatment, z ¼ effect of NE treatment and # ¼ treatment interaction. regression analysis of these variables showed that sLR11 levels were not significantly associated with any of them but tended to associate most with VFA, independently of other factors ( Table 2). A second study of 25 patients with type 2 diabetes mellitus or glucose intolerance-associated metabolic disorders (Supplementary Table 5) showed a similar significant correlation between BMI and sLR11 levels (Fig. 6g). Importantly none of the other measured characteristics of these subjects, such as age, HbA1c or cholesterol levels, showed any correlation with sLR11 (Supplementary Table 5). This further suggested that overall body weight, and thus adiposity, was the major determinant of sLR11 levels in humans. In a third study of 14 obese patients who underwent bariatric surgery, post-operative sLR11 levels were significantly reduced after 12 months, concomitantly with reductions to BMI, VFA and SFA ( Table 3). The percentage reduction in sLR11 was positively correlated with that of VFA and SFA (Fig. 6h,i). Multiple regression analysis of the two significant correlates, VFA and SFA, showed no significance between their respective contributions to the relationship ( Table 3), suggesting that overall fat mass, regardless of its distribution, was driving the correlation.

Discussion
Our data demonstrate a role for sLR11 as a negative regulator of thermogenesis, acting through the BMP/Smad pathway in adipose tissue to suppress activation of thermogenic machinery and inhibit transcription of key thermogenic genes.
Pgc1α Cideα (a-c and e,f) Gene expression in differentiated primary adipocytes from subcutaneous WAT of wild-type and Lr11 À / À mice incubated with 3.3 nM BMP7 for the first 2 days and 10 ng ml À 1 sLR11 throughout the incubation (n ¼ 6). (d) Levels of phosphorylated Smad proteins 15 min after treatment with vehicle or 3.3 nM BMP7 in primary adipocytes subcutaneous WAT of wild-type mice, differentiated À / þ 10 ng ml À 1 sLR11 (n ¼ 6). (g) Western blot for phospho-Smad3 in mature adipocytes cultured from subcutaneous WAT of wild-type mice treated for 30 min with 2 ng ml À 1 TGFb in the presence or absence of 10 ng ml À 1 sLR11. (h) Western blot showing denoted proteins immunoprecipitated from membrane protein extracts from differentiated human adipocytes from scWAT with an antibody to LR11. (i) Ucp1 mRNA expression following 2 h stimulation with 10 ng ml À 1 sLR11 in mature adipocytes cultured from wild-type BAT and transfected with the indicated siRNA for BMP receptors. Data presented as mean ± s.e.m., * ¼ Po0.05 using one-way ANOVA followed by Tukey post-hoc test. w, z and # denote a significant of genotype, treatment and genotype*treatment interaction respectively, using 2-way ANOVA. Blot protein weights shown in kDa. (see also Supplementary Fig. 2 Tables 2 and 3). scWAT, subcutaneous white adipose tissue; gWAT, gonadal adipose tissue. Initial correlations derived using Pearson correlation test. Multiple regression analysis performed in patients to examine the seven characteristics more strongly correlated with sLR11 levels using beta weights to detect significant differences in the contribution of each factor, n ¼ 156.
regulated in tissues with inherent/adaptive thermogenic capacity, in response to changes in environmental temperature and diet, but not in other metabolic tissues such as liver (Fig. 6a). In absence of any endogenous LR11, nutritional stimuli that induce thermogenesis in scWAT are left unchecked, leading to increased recruitment of brown adipocytes, reduced lipid accumulation, hypermetabolism and overall leanness. Our in vitro data support a cell-autonomous role of sLR11 in regulating adipose tissue responses to increased energy availability. This is suggested by the peptide's ability to suppress the known thermogenic signalling pathways and the lack of any significant change to energy balance on a standard chow diet. Neurological roles for LR11 have been previously described, although we hypothesize that if LR11 is required to maintain a central circuit regulating metabolic rate, it would be expected that its ablation would yield a consistent phenotype across environmental conditions. Instead metabolic rate changes only become apparent in Lr11 À / À mice fed HFD, when sympathetic tone to BAT and WAT depots increases and adaptive changes occur to strike a new balance between lipid oxidation and storage. In the absence of LR11 and thus sLR11, increased energy availability results in increased DIT in adipose tissue but without the appropriate anabolic response. Where WT mice are able to store additional calories as triglyceride, Lr11 À / À mice can only increase the metabolic arm of the response and burn them. However, even on a chow diet there is a slight tendency for Lr11 À / À mice to be lighter and consume slightly less food (Fig. 1g-i and Fig. 2a) and it is also highly likely that DIT is itself regulated by discrete neuronal pathways that mediate alterations to sympathetic activation of thermogenic tissues. Therefore we cannot rule out based on our current data, that LR11 may also play a central role in the initiation and maintenance of metabolic responses and energy balance. LR11 has also been previously shown to increase vascular smooth muscle cell migration in response to injury 11 . We did not observe any alteration to vasculogenesis or angiogenesis in Lr11 À / À mice, suggesting that perturbations in these processes are not causative of the hypermetabolic phenotype. Similarly, the documented ability of LR11 to regulate LPL activity 15 was not evident as a mechanism by which its deletion led to hypermetabolism in our model (Fig. 2l). Our data also suggest that sLR11 levels are dependent on adipose mass in humans, increasing in line with fat mass. Clinical interventions such as bariatric surgery that lead to reductions in adiposity, also lower sLR11 levels (Table 3 and Fig. 6h,i), potentially contributing to the observed increase in metabolic rate in such patients 27,28 . The fact that the strongest correlation in all human subjects is between sLR11 and visceral and SFA, rather than BMI is further evidence that adipose tissue is the main site of sLR11 production and secretion.
We identify LR11 as a key metabolic regulator, which functions to maintain the adequate balance between lipid storage and oxidation in response to changing environmental conditions and particularly when thermogenic activity is increased in capable tissues. This mechanism might be exploited for anti-obesity therapies or conversely, for the treatment of conditions such as anorexia nervosa or cachexia. The apparent anabolic role of sLR11 may also be relevant for understanding its prior implication in cardiovascular disease and Alzheimer's disease 29,30 . While an increasing number of factors and pharmacological molecules exist that either increase the number of brown adipocytes or their thermogenic capacity 9,31 , only a small number of negative regulators are apparent, namely nuclear receptor-interacting protein 140 (Rip140) and the diurnally regulated Rev-erba (refs 32,33). It seems likely that other negative regulators must exist to ensure effective control of this energy-expensive mechanism and to 'fine tune' the thermogenic response. Before our study, a lipoprotein receptor-mediated mechanism for regulation of BMP signalling in the regulation of thermogenesis has not been described.

Methods
Animals and diets. Unless otherwise stated, all data are from work on males (8-16 weeks of age). C57Bl6/J mice were purchased from Charles River. Lr11 À / À mice were generated as previously described 11 on a C57Bl6/J background and compared to WT littermates. Unless stated, mice were housed in a temperature-controlled room (22°C) with a 12-h light/dark cycle with free access to diet and water. Standard chow or HFD was administered ad libitum from weaning until indicated. High-fat diet consisted of 60% calories from fat (D12492, Research Diets, NJ, USA) unless otherwise stated. Thermoneutrality involved housing mice at 30°C in a normal home cage environment. All animal procedures were approved by either the UK Home Office or the Special Committee on Animal Welfare, School of Medicine, at the Inohana Campus of Chiba University.
Calculation of energy expenditure. Animals were placed in a monitoring system based on their home cages (Creative Scientific, Canterbury, UK) with the ability to measure oxygen and carbon dioxide concentrations using a system (Minimox) designed by Peter Murgatroyd (University of Cambridge) 34 . Oxygen consumption and carbon dioxide production was measured, with samples taken at 18 min intervals over a 48-h period. Energy expenditure was then calculated using indirect calorimetry with the Elia and Livesey constant 35 . Energy expenditure is expressed as J min À 1 per mouse using an adjusted mean body weight. This was obtained using analysis of covariance with weight as the covariate, a robust method for Microarray. The four RNA sources, WT scWAT, Lr11 À / À scWAT, WT BAT and Lr11 À / À BAT, were prepared from subcutaneous WAT and BAT from WT mice and Lr11 À / À mice, respectively (n ¼ 3 each). Microarray analyses were performed by Takara Bio (Japan). Briefly, cyanine-3-labelled cRNA was prepared from 0.1 mg total RNA using the Low Input Quick Amp Labeling kit (Agilent) according to the manufacturer's instructions, followed by RNAeasy column purification (QIAGEN, Valencia, CA). Dye incorporation and cRNA yield were checked with the Nano-Drop ND-2000 Spectrophotometer. An amount of 0.6 mg of cyanine-3-labelled cRNA was then fragmented at 60°C for 30 min in a reaction volume of 25 ml containing 1 Â Agilent fragmentation buffer and 2 Â Agilent blocking agent following the manufacturer's instructions. On completion of the fragmentation reaction, 25 ml of 2 Â Agilent hybridization buffer was added to the fragmentation mixture and hybridized to Agilent Mouse GE 8x60K Microarray for 17 h at 65°C in a rotating Agilent hybridization oven. After hybridization, microarrays were washed 1 min at room temperature with GE Wash Buffer 1 (Agilent) and 1 min at 37°C GE Wash buffer 2 (Agilent), then dried immediately by brief centrifugation. Slides were scanned immediately after washing on the Agilent DNA Microarray Scanner (G2565CA) using one colour scan setting for 8 Â 60 k array slides (scan area 61 Â 21.6 mm, scan resolution 3 mm, dye channel is set to Green PMT to 100%). The scanned images were analysed with Feature Extraction Software 10.10 (Agilent) using default parameters to obtain background-subtracted and spatially detrended processed signal intensities. The gene-tree analysis revealed six clusters based on the expression pattern of the four RNA sources. The gene expression pattern with Cluster 6: WT scWAToLr11 À / À scWAToor ¼ WT BAT ¼ oroLr11 À / À BAT was dominant compared with other patterns. The top 100 genes of Lr11 À / À scWAT4WT scWAT in cluster 6 were listed and summarized (Supplementary Table 2).
Quantitative PCR with reverse transcription. Total RNA was isolated from cells or using Buffer RLT and purified by RNeasy Mini columns (Qiagen). RNA was isolated from ground tissues using STAT-60 reagent (TEL-TEST) followed by chloroform extraction and isopropanol precipitation. Complimentary DNA was generated from 500 ng of RNA using M-MLV reverse transcriptase and master mix (Promega) in a 20 ml reaction mixture with 2.5 mM MgCl 2 , 1.25 mM dNTPs and 5 mg ml À 1 random hexamers at 37°C for 1 h. Complementary DNA was diluted 75-fold and 5 ml of diluted complementary DNA was used in a 12 ml real-time PCR reaction mixture using TaqMan primers and probes or SYBR green reagent (Applied Biosystems) according to manufacturer's instructions. Reactions were run in duplicate for each sample and quantified in the ABI Prism 7900 sequence detection system (Applied Biosystems). Data are expressed as arbitrary units where expression of target genes are corrected using the geometric average of four housekeeping genes: 18S, b2-microglobulin, b-actin and 36B4 using Bestkeeper (freeware). Sequences of primers and probes used are listed in Supplementary  Table 3.
Tissue collection and histology. All animal tissues for protein or RNA extraction were snap frozen in liquid nitrogen at the time of collection unless otherwise stated and later ground to a fine powder using a sterile pestle and mortar on liquid nitrogen. Samples for histology were placed in 10% buffered formalin overnight before transfer to 70% ethanol and later embedding in paraffin. Multiple sections were stained with haematoxylin and eosin for morphological analysis.
Immunohistochemistry. Paraffin-embedded sections (5 mm) were used for immunohistostaining. Deparaffinized sections were pretreated with 0.3% H 2 O 2 to inactivate endogenous peroxidase. Slides were stained in the presence of 3% bovine serum albumin (BSA) with antibody against UCP1 (1:50 dilution) at 23°C for1 h followed by horseradish peroxidase-conjugated anti-rabbit IgG secondary antibodies (Molecular Probes) at 1:200 dilution. The slides were counterstained with haematoxylin. Controls with non-immune rabbit IgG were conducted in parallel with each immunoassay procedure.
Western blot. Powdered tissue or collected cultured cells were resuspended in lysis buffer (20 mM Tris-HCL, 150 mM NaCl, 1 mM EGTA, 1 mM EDTA, 1% Triton X-100, pH7.5) with added protease and phosphatase inhibitor cocktails according to manufacturers instruction (Sigma). After lysis, lysates were cleared by centrifugation at 10 000g for 10 min at 4°C. Protein concentrations of the supernatants were determined by D c Protein assay (Bio-Rad). Proteins were diluted in Laemmli buffer with 1% 2-mercaptoethanol. Proteins (20 mg) were separated by SDS-polyacrylamide gel (10%) electrophoresis and transferred to Immobilon-P membrane using a semi-dry transfer apparatus (Biorad). Membranes were blocked for 1 h at room temperature and incubated overnight at 4°C with the indicated antibody. Bound primary antibodies were detected using peroxidase-coupled secondary antibodies and enhanced chemiluminescence (Amersham). Relative quantification of band intensities was calculated by digitally photographing exposed films and using Genesnap and Genetools software (Syngene). Uncropped scans of blots are shown in Supplementary Fig. 2.
SMAD assays. Cultured fully differentiated adipocytes derived from the subcutaneous fat tissues of wild-type or Lr11 À / À mice in the presence of BMP7 for the first 2 days were prepared. After serum depletion for 16 h, the cells were incubated with or without 3.3 nM BMP7 or 2 ng ml À 1 TGF-b in the presence or absence of 10 ng ml À 1 sLR11 for 30 min, and then collected for the following western blot analyses. Cells were washed three times with phosphate-buffered saline (PBS) and harvested in RIPA buffer containing 0.01 mM APMSF. For immunoblotting, equal amounts of solubilized protein was subjected to 10% SDS À polyacrylamide gel after heating to 95°C for 5 min under reducing conditions and transferred to a nitrocellulose membrane.
Immunoprecipitation. An amount of 100 mg of membrane protein was mixed with 1 mg of anti-human LR11 antibody M3 or control IgG, and rotated for 1 h at 4°C. The LR11/BMPRs/antibody complex was precipitated by protein A Sepharose. The proteins were released into 25 ml SDS sample buffer by heating to 95°C for 10 min, and applied for western blot analysis using anti-BMPR1A, BMPR1B or BMPR2 antibody as described.
Culture of adipocytes. Mouse primary adipocytes were obtained from either BAT or subcutaneous WAT by excision of the fat pad followed by mincing with sterile surgical scissors and digestion at 37°C with shaking for B35 min (10 ml Hanks buffered salt solution, 2% BSA, 2 mg ml À 1 Type 2 collagenase (C6885, Sigma). Digests were filtered through a 100 mM filter and left to sit on ice for 10 min. The upper fat layer was removed and then the upper 2/3 of the supernatant washed twice in PBS before being plated. Cells were maintained and differentiated in Dulbecco's modified Eagle's medium with 10% fetal bovine serum, 20 mM L-glutamine, 100 units ml À 1 penicillin and 100 mg ml À 1 streptomycin with 150 mM sodium ascorbate and 4 nM insulin in the presence of 1 nM 3, 3 0 , 5-Triiodo-Lthyronine at 37°C in 5% CO 2 . After 8 days, cells were fully differentiated (validated by gene expression profiles and morphological appearance), and if BMP7 and/or sLR11 were added, 3.3 nM BMP7 was present in the culture media for the first 2 days and/or 10 ng ml À 1 sLR11 was present throughout the culture period. Cultures were routinely tested for the presence of mycoplasma contamination and decontamination/rederivation performed when necessary. Human primary preadipocytes were prepared from the stromal vascular fraction of fat tissues and cultured as described 37 . The procedures were approved by the Ethical Committee on human study, Toho University Sakura Medical Center.
Hek 293 cells were obtained from Sigma (product number 85120602). While cells were not independently authenticated, this was not a major concern as they were used only for the luciferase reporter assays and have been routinely used in our laboratory for multiple previous similar experiments without problems. Grossly the cells behaved as expected in terms of their growth and morphology and their viability and transfection efficiency were found to be excellent. As for primary cultures, cultures were routinely screened for mycoplasma contamination.
LPL activity assay. This assay was performed as previously described 38 . Briefly, the culture medium for adipocytes cultured for 16 h without fetal bovine serum were collected and the LPL activity was measured as below using triolein (Sigma, St Louis, MO) as the substrate. The substrate solution was prepared as follows: 100 mg of triolein and 7.5 ml of 0.2% Triton X-100 at a final volume of 7.5 ml of 1 mole l À 1 Tris-HCl (pH 8.0) were sonicated on ice for 10 min. The LPL activity assay reaction mixture contained 50 ml of substrate solution, 25 ml of 20% fatty acidfree bovine serum albumin (pH 8.0; Sigma), 5 ml of HDL (3 mg protein ml À 1 ) as a source of apo C-II and 100 ml of sample. After incubation for 2 h at 37°C, the enzyme reaction was terminated by the addition of 10 ml of diisopropylfluorophosphate. The amount of free-fatty acids released in the mixture was measured at 4°C using an enzymatic method (Nescauto NEFA V2; Alfresa Pharma, Osaka, Japan).
Cellular respiration. Differentiated primary adipocytes were placed in a XF24 bioanalyser (Seahorse Bioscience, MA, USA) in the manufacturer's designated culture/assay plate (each plate was seeded with an equal number of cells from a pool of primary adipocytes from two animals). Basal cellular respiration was measured over a period of 15 min using the oxygen consumption rate (OCR) measurement assay template. An amount of 100 nM NE was then added to the assay medium (XF assay medium, Seahorse) and respiration measured for a further 15 min using OCR measurement template. Basal and stimulated oxygen consumptions reported are the mean of three separate OCR measurements per well.
NE-induced thermogenesis. Two individually housed mice (litter mates) were monitored simultaneously at an airflow of 1 l min À 1 . HFD-fed mice housed at thermoneutrality for 8 weeks before assessment were anaesthetized with pentobarbital (45 mg kg À 1 i.p.). The oxygen consumption and carbon dioxide production of the mice was measured at 30°C to obtain basal values during 15 min. Thereafter the mice were removed from the metabolic chambers for 6 min and injected subcutaneously with NE (1 mg kg À 1 ) before being returned to the metabolic chambers for 45 min. Energy expenditure was calculated as described above but presented as the fold increase or mean increase, thus not requiring covariate analysis.
Clinical study. Study 1 subjects consisted of 156 consecutive patients with sleep apnoea syndrome and study 2 consisted of 25 consecutive patients with type 2 diabetes mellitus or glucose intolerance-associated metabolic disorders at Toho University Sakura Medical Center or Yanbian University Hospital. Blood samples were collected in the morning after an overnight fast, and the biochemical data were measured using standard hospital laboratory techniques. The potential risk factors for diabetes and atherosclerosis were analysed, including age, sex, BMI, VFA and SFA, blood lipids and HbA1c for both studies; smoking, medication of diabetes, medication of hypertension, medication of dyslipidemia, history of coronary artery disease and history of cerebrovascular diseases for Study 1; fasting plasma glucose and systolic and diastolic blood pressures for Study 2. In Study 3, 14 obese subjects received bariatric surgery at Toho University Sakura Medical Center, and their VFA and SFA were measured before and 12 months after the surgery. The body weights of all patients decreased 1 year after surgery compared with those before surgery. Visceral fat areas and SFA were measured at umbilical levels of patients using computed tomography. Hypertension was defined as systolic pressure of more than 140 mm Hg or diastolic pressure of more than 90 mm Hg. Diabetes mellitus was defined as a fasting blood glucose level of more than 126 mg dl À 1 , glycosylated haemoglobin (HbA1c) of more than 6.2%, or both. Dyslipidemia was defined as serum total cholesterol of more than 220 mg dl À 1 and triglycerides of more than 150 mg dl À 1 in the fasting state, or both, and HDL cholesterol o40 mg dl À 1 , or a combination thereof. The patients suffering from possible malignant diseases, Alzheimer's disease, renal diseases or hepatic diseases were excluded in the study subjects. The procedures of the above human studies were approved by the Human Investigation Review Committee on human study, Toho University Sakura Medical Center or Yanbian University Hospital. Informed consent was obtained from all subjects. sLR11 measurement. Serum sLR11 concentrations were determined using a sandwich ELISA system. 12.5 ml of human and 2.5 ml of murine serum were used for the measurement of sLR11 by respective ELISA. Human serum sLR11 concentrations were determined using a revised procedure based on sandwich ELISA system previously reported 39 . In brief, 12.5 ml of human serum were diluted eightfold with 1% BSA in PBS containing 0.1% Tween 20 (PBST), reacted with the capture anti-LR11mouse monoclonal antibody M3 F(ab 0 )2 for 2 h, then incubated with a biotinylated capture anti-LR11 rat monoclonal antibody R1 for 1 h. A murine serum soluble LR11 ELISA was performed using two mouse monoclonal antibodies, Mab93222 and Mab93213 that were established by immunization of Lr11 À / À mice using synthetic peptides against different epitopes of VPS10p domain of LR11 protein 40 . Briefly, 2.5 ml of murine serum were diluted 20-fold with 1% BSA in PBST, reacted with the capture Mab93222 for 2 h, and then incubated with the biotinylated capture Mab93213 for 1 h. The LR11-antibody complex was quantified with horseradish peroxidase-conjugated streptavidin using purified human and rabbit LR11 protein as a standard.
Power calculations. Group sizes were calculated using STPLAN software created by Barry Brown from the University of Texas MD Anderson Cancer Center. We have used mean values and standard deviations obtained from preliminary data from studies in Lr11 À / À mice. In this instance we expected effect sizes ranging down to 30% and aimed to obtain statistical significance of Po0.05 at a power of 0.9. With the expected s.d. of up to 12% and thus planned to include a minimum of seven animals/group for in vivo studies.
Exclusion criteria. Animals were excluded if their phenotypic characteristic fell more than two s.d. from the mean of the group. In practice only two animals were excluded from two separate parts of this study.
Randomization and blinding. No formal randomization was applied, but for diet studies, animals of each genotype were grouped so as to match mean starting bodyweights of chow and HFD groups. The researchers were not blinded as to the genotype of the animals as this was not practical because of the weight divergence phenotype. All measurement of gene expression and serum biochemistry was conducted by a researcher who was blinded for the experimental groups' identities.
Statistics. Statistical analysis was performed with SPSS statistics version 23 (IBM) or Stat Flex v. 5.0 (Artech). Data was tested for normality and a Levene test for equality of variance was also applied to all data sets before parametric tests were applied. P values were not applied to exploratory initial studies where only the null hypothesis was being tested, only where an a priori hypothesis had been defined and informed the study. Associations of sLR11 levels with risk factors were examined by Pearson correlation analysis for continuous variables and by unpaired t-test for categorical variables between groups (sex, smoking, medication of diabetes, hypertension or dyslipidemia and history of coronary artery disease or cerebrovascular disease). The study subjects were principally separated to two groups based on the averaged values. In clinical studies, multiple linear regression analyses were used to calculate coefficient beta for sLR11 by controlling for age, smoking, HDL-cholesterol, HbA1c, BMI, VFA and SFA for study 1, and VFA and SFA for the study of patients that received bariatric surgery (study 3), which are significantly correlated with sLR11.