scEMC10, a novel circulating inhibitor of adipocyte thermogenesis, is upregulated in human obesity and its neutralizing antibody prevents diet-induced obesity

39 Secreted isoform of endoplasmic reticulum membrane complex subunit 10 (scEMC10) is a 40 poorly characterised secreted protein of largely unknown physiological function. Here we 41 demonstrate that scEMC10 is upregulated in humans with obesity and is positively 42 associated with insulin resistance. Consistent with a causal role for scEMC10 in obesity, 43 Emc10 mice are resistant to diet-induced obesity due to an increase in energy 44 expenditure. Furthermore, neutralization of circulating scEMC10 using a monoclonal 45 antibody reduces body weight and enhances insulin sensitivity in obese mice. 46 Mechanistically, we provide evidence that scEMC10 binds to the catalytic subunit of PKA 47 and inhibits its stimulatory action on CREB while ablation of EMC10 promotes 48 thermogenesis in adipocytes via activation of the PKA signalling pathway and its 49 downstream targets. Taken together, our data identify scEMC10 as a novel circulating 50 inhibitor of thermogenesis and a potential therapeutic target for obesity and its 51 cardiometabolic complications. 52 53 54

and inhibits its stimulatory action on CREB while ablation of EMC10 promotes 48 thermogenesis in adipocytes via activation of the PKA signalling pathway and its 49 downstream targets. Taken together, our data identify scEMC10 as a novel circulating 50 inhibitor of thermogenesis and a potential therapeutic target for obesity and its 51 cardiometabolic complications. 52 53

INTRODUCTION 55
Obesity is the result of a chronic imbalance between energy intake and expenditure 56 and is a major risk factor for metabolic diseases 1,2 such as type 2 diabetes mellitus, 57 cardiovascular disease, and certain types of cancer 1,3 . Brown and beige fat have recently 58 attracted significant interest as tissues that could be leveraged to treat obesity and 59 diabetes. This is due to their ability to consume a considerable amount of glucose and 60 lipids and dissipate the chemical energy from these substrates as heat, in a process called 61 thermogenesis 4-9 . Mouse models with enhanced brown or beige fat content or activity have 62 been previously demonstrated to resist weight gain and exhibit improved metabolic health 63 via activation of adipose thermogenesis 10-13 . 64 However, despite intense investigation and an increasingly detailed understanding 65 of the molecular determinants of adipocyte thermogenesis in cells and mice, modulation of 66 adipocyte thermogenic capacity via either brown fat activation or increasing the amount of 67 thermogenic adipose tissue has not been successfully implemented as a therapeutic 68 strategy in human obesity. The reasons for this are manifold -but of principle importance 69 is that there are important physiological differences in humans and lower organisms with 70 respect to the regulation and functional relevance of adipocyte thermogenesis that may 71 limit translation. Identifying the determinants of adipocyte thermogenic capacity in obese 72 humans using integrated clinical and pre-clinical studies may identify novel therapeutic 73 targets for obesity with increased likelihood of clinical translation. 74 Previously, we identified a novel secreted protein, scEMC10 (secreted isoform of 75 HFD feeding (Supplementary Fig 3C). These observations were confirmed in an 163 independent rodent model of obesity -ob/ob mice (Supplementary Fig 3D). 164 While many key metabolic tissues have unchanged scEmc10 expression during 165 obesity, we observed that scEmc10 transcript was significantly upregulated in liver and 166 pancreatic islets after HFD (Supplementary Fig 3B). To identify the pathophysiological 167 stimuli regulating scEmc10 expression, we used animal models for insulin resistance and 168 hepatic steatosis. We examined livers from acute insulin receptor knockdown (L-IR KD ) 169 mice, a lipodystrophic mouse model (IR FKO ) 25 , and choline-deficient and methionine-170 restricted (CDA) diet-treated mice 26 . Our results showed significant upregulation of 171 scEmc10 in livers from IR FKO and CDA-treated, but not L-IR KD , mice ( Supplementary Fig  172   3E). This suggests that scEmc10 expression could potentially be regulated by hepatic 173 steatosis, but not hepatic insulin resistance. The absence of insulin signaling regulation 174 was further confirmed in insulin receptor KO (bIRKO) b-cells 27 (Supplementary Fig 3F). In 175 addition to pathological conditions, we investigated scEmc10 expression after fasting and 176 during acute refeeding to explore physiological processes that might regulate its 177 expression. We observed that during fasting, scEmc10 expression was dramatically 178 downregulated in the liver compared to the fed state, but completely recovered after 2-4 h 179 of refeeding ( Supplementary Fig 3G). 180 Besides pathophysiological regulation, we also examined scEmc10 abundance 181 across various metabolic tissues under physiological conditions such as cold exposure 182 and thermoneutrality, we observed that cold exposure significantly downregulated 183 scEmc10 in BAT and iWAT but not in other tissues (Supplementary Fig 3H) knockout, we subjected mice to either a low-fat diet (LFD) (10% fat by kcal) or high-fat diet 198 (HFD) (60% fat by kcal) and undertook metabolic phenotyping. The KO mice fed LFD for 199 12 weeks exhibited a trend towards reduced body weight (Supplementary Fig 4B). This 200 was mediated by a reduction in total adiposity as assessed by percentage fat mass or 201 adipose tissue weight while lean mass was non-significantly reduced ( Supplementary Fig  202   4C & 4D). 203 The effect of Emc10 KO was accentuated in HFD-fed mice. Emc10 KO mice on 204 HFD were significantly leaner with attenuation in weight gain from as early as 2 weeks 205 after initiation of HFD ( Fig 3A). The lower body weight of Emc10 KO mice was largely 206 accounted for by substantial reduction of both inguinal white adipose tissue (iWAT) and 207 epididymal white adipose tissue (eWAT) weights (Fig 3B), without an alteration in lean 208 body mass (Fig 3C, Supplementary Fig 5A). 209 Our histological analyses revealed that the decrease in the KO fat mass on HFD is 210 likely driven by a reduction in adipocyte size, as demonstrated by a significantly greater 211 frequency of small adipocytes and lower frequency of mid-sized and large adipocytes in 212 both the eWAT and iWAT ( Supplementary Fig 5B & 5C). This is also evidenced by the 213 near-normal appearance of the KO brown adipocytes, compared to the enlarged, lipid-214 laden brown adipocytes harvested from the WT mice after HFD treatment (Supplementary 215 Fig 5B). Similar, but more subtle, changes in the distribution of adipocyte size were also 216 observed in the eWAT and brown adipose tissue (BAT) from the KO mice fed with LFD 217 ( Supplementary Fig 4E & 4F). Consistent with a lean phenotype, Emc10 KO mice fed a 218 HFD exhibited improved glucose tolerance and insulin sensitivity ( Fig 3D) and exhibited 219 lower fasting glucose and insulin levels ( Fig 3E). Increased adipocyte size is positively 220 correlated with leptin production 28 . Consistent with the larger adipocytes in the HFD-fed 221 WT mice, we observed significantly higher leptin levels in the WT mice fed HFD, compared 222 to the KO (Fig 3F). Plasma adiponectin levels are decreased in obesity, insulin resistance, 223 and type 2 diabetes 29 . In line with their leaner phenotype and improved metabolic profile, 224 KO mice on HFD exhibited significantly higher serum adiponectin levels compared to WT 225 controls ( Fig 3F). In addition to improved glucose metabolism, Emc10 KO mice also exhibited significantly lower levels of fed serum triglyceride (TG), cholesterol (CHO), and 227 non-esterified fatty acid (NEFA) levels ( Fig 3G). In contrast, in the chow-fed cohort, we 228 only observed trends towards increased plasma insulin and decreased leptin in the KO 229 mice ( Supplementary Fig 4G). 230 Chronic exposure of mice to HFD causes hepatic steatosis 30 . Feeding a HFD, but 231 not CD, increased liver mass (Fig 3B) and the number of large, lipid-containing vacuoles 232 revealed by H&E staining (Fig 3H, Supplementary Fig 4H) in the WT livers, compared to 233 KO livers. Accordingly, TG content of liver from KO mice was significantly lower following 234 HFD ( Fig 3I). Additionally, adipose inflammation was improved in KO mice fed with HFD 235 evidenced by significantly decreased gene expression of inflammation markers including 236 Mcp-1, Tnfa, and F4/80 in KO eWAT compared to WT (Supplementary Fig 5D). In 237 summary, Emc10 KO protects mice from diet induced obesity. 238 239

Upregulation of circulating EMC10 promotes obesity 240
To determine the metabolic consequences of increasing circulating EMC10, we 241 performed intravenous injections of adeno-associated virus (AAV) encoding human 242 scEMC10 (hscEMC10) or LacZ to deliver full-length scEMC10 or LacZ construct to the 243 liver of 7-wk-old C57BL/6 male mice. Animals were subjected to either CD or HFD feeding 244 one week after the injection. This method generally results in robust expression of the 245 protein in the liver and potential secretion into the plasma 31 . Ten days post-injection, we 246 observed a ~10-fold increase in liver EMC10 protein abundance and ~5-6-fold increase in 247 plasma EMC10 levels, as detected by western blotting with an anti-EMC10 antibody 248 (Supplementary Fig 6A). As suggested by our human data, we observed that mice over-249 expressing hscEMC10 gained significantly more weight than LacZ-expressing controls, 250 even on chow diet ( Fig 3J). The higher body weight of hscEMC10 mice was largely 251 contributed by increased WAT mass (Fig 3K, Supplementary Fig 6B). Consequently, the 252 heavier hscEMC10 mice were also more glucose intolerant, insulin resistant (Fig 3L), and 253 had significantly higher levels of plasma insulin and leptin, consistent with their obese 254 phenotype ( Fig 3M, Supplementary Fig 6C). 255 Consistent with the chow diet data, we observed that increased circulating EMC10 256 promotes diet-induced obesity in mice as early as two weeks after introduction of HFD 257 (Supplementary Fig 6D). In line with the body weight phenotype, we observed that mice expressing hscEMC10 are also more glucose intolerant and insulin resistant 259 Fig 6E). The increased body weight observed in hscEMC10 mice is 260 largely contributed by a significant increase in adipose tissue weight ( Supplementary Fig  261   6F). In addition to increased fat mass, hscEMC10 over-expressors exhibited 262 hyperinsulinemia, hyperleptinemia, hyperlipidemia, and showed significantly lower 263 circulating adiponectin (Supplementary Fig 6G). Taken together, our gain and loss of 264 function experiments in mice demonstrate that scEMC10 is a novel regulator of energy 265 balance and provides supportive evidence that the associations between serum EMC10 266 and BMI observed in obese humans are causal. 267 268 EMC10 ablation promotes energy expenditure through the activation of adipose 269

tissue thermogenesis 270
To determine how EMC10 alters energy balance in mice, we determined the food 271 intake, gene expression of anorexigenic and orexigenic peptides in the hypothalamus, and 272 the ability of the intestine to absorb fat in both WT and KO mice 32 . We found that the total 273 food intake, hypothalamic anorexigenic, and orexigenic peptide gene expression, and 274 intestinal fat absorption between the WT and KO mice were unchanged (Supplementary 275 Fig 7A-7C). However, when the food intake data were normalized to body weight, the 276 Emc10 KO mice were hyperphagic relative to wildtype controls (Supplementary Fig 7A). 277 Reduction in adipose tissue mass without significant alterations in energy intake 278 suggested a potential increase in energy expenditure in the KO mice. Indeed, 279 measurement of oxygen consumption (VO2) and carbon dioxide elimination (VCO2) rates 280 over 48 hours, including two cycles of light and dark phases, revealed significant increases 281 in oxygen consumption and carbon dioxide production in the HFD-fed KO mice (Fig 4A), 282 which could not be attributed to changes in activity (Supplementary Fig 7D). Consistent 283 with the higher VO2 and VCO2, Emc10 KO mice also generated more heat ( Fig 4B) and 284 exhibited a ~ 0.8° C higher basal core body temperature ( Fig 4C). To confirm, we 285 reanalyzed the data with analysis of covariance (ANCOVA) 33 . Consistent with our prior 286 analysis, the differences in energy expenditure remained significant even when lean mass 287 was used as a covariate ( Fig 4D). The generally higher respiratory exchange ratio (RER = 288 VCO2/VO2) in the HFD-fed Emc10 KO mice indicated higher carbohydrate utilization than 289 the more obese WT controls, despite both groups being fed HFD for 12 weeks (Fig 4E). 290 Taken together, our data indicate that increased thermogenesis in the Emc10 KO mice is 291 the primary mechanism contributing to their resistance to diet-induced obesity. 292 Since increased adipose thermogenesis and metabolism can augment whole-body 293 energy expenditure, we examined expression levels of markers of adipocyte 294 differentiation, lipolysis, lipogenesis, and thermogenesis in both BAT and iWAT, as both 295 types of fat play protective roles during obesity. We observed that ablation of Emc10 296 robustly upregulated lipolytic, lipogenic, and thermogenic marker expression in BAT (Fig  297   4F). This was also observed in the expression of lipolytic and thermogenic markers in 298 iWAT harvested from HFD-fed Emc10 KO mice, compared to WT controls ( Fig 4G). As 299 HFD/obesity alone is known to have an impact on the expression of many of these 300 markers, to further delineate which adipocyte function is the primary mechanism 301 contributing to the obesity-resistant phenotype of the KO mice, we examined BAT and 302 iWAT collected from CD-fed mice. As expected, not all markers identified in the HFD-fed 303 mice showed similar expression changes; however, the markers of thermogenesis did (Fig  304   4H). To confirm changes in adipose tissue function, the oxygen consumption of BAT and 305 iWAT harvested from the CD-fed mice was measured using the Clark electrode. 306 Consistent with the gene expression data, we observed that the ablation of Emc10 307 significantly increased oxygen consumption in both the BAT and iWAT (Fig 4I). 308 A key physiological difference in mice and humans is the proportion of energy 309 devoted to maintaining body temperature when housed at ambient room temperature. This 310 difference has key implications for the translation of mouse biology to humans, particularly 311 with respect to thermogenesis and energy expenditure 34 . To demonstrate the robustness 312 of our findings in this regard, we repeated our analysis of mouse body weight at 313 thermoneutrality (30ºC). Both WT and KO mice were weaned and subsequently subjected 314 to HFD treatment at thermoneutrality. Our data demonstrated that the KO mice were 315 significantly leaner than the WT controls at thermoneutrality (Fig 4J). 316 The differences in body weight at thermoneutrality and the effects of Emc10 KO on 317 adipose tissue phenotype suggest that EMC10 augments energy expenditure by 318 increasing adipocyte thermogenic capacity and increasing non-shivering thermogenesis. 319 To confirm the role of EMC10 in non-shivering thermogenesis, we measured oxygen 320 consumption in WT and KO mice at thermoneutrality in response to a β3 adrenergic 321 agonist -CL316,234 -which is expected to activate thermogenesis only in brown adipose tissue where β3-adrenoreceptors are highly expressed 35 . Consistent with the obesity-323 resistant phenotype, the KO mice showed a significantly enhanced response to CL316,243 324 compared to WT control ( Fig 4K). 325 Our data demonstrate that loss of EMC10 enhances thermogenic capacity of 326 adipocytes, increases energy expenditure and protects mice from diet-induced obesity. Pgc1a expression in Emc10 KO brown, inguinal, and epididymal adipocytes compared to 337 controls (Fig 5A-5C). 338 To determine whether increased adipocyte thermogenesis is truly mediated by loss 339 of EMC10, we treated the differentiated KO brown adipocytes with exogenous 340 recombinant scEMC10. We observed that the basal Ucp1 and Pgc1a upregulation in 341 Emc10 KO adipocytes were diminished by increasing doses of recombinant scEMC10 in 342 the culture media ( Fig 5D). Consistent with the transcript data, western blotting also 343 showed dramatically increased UCP1 protein in the differentiated KO adipocytes 344 compared to WT and the UCP1 protein levels were reduced by increasing amount of 345 exogenous recombinant, but not heat-inactivated scEMC10 ( Fig 5E). As Emc10 KO 346 adipocytes appeared to be highly responsive to b3 agonist stimulation, we first determined 347 whether basal upregulation of Ucp1 and Pgc1a is dependent on key downstream effectors 348 of b3-adrenoceptors cAMP/PKA by using the PKA-specific inhibitor, H89. Our results 349 showed that inhibiting the kinase activity of PKA completely abolished the upregulation of 350 thermogenic markers in Emc10 KO brown adipocytes (Fig 5F). 351 To identify further downstream signaling targets, we examined PKA downstream 352 signaling molecules that are known regulators of Ucp1 and Pgc1a. Western blotting revealed robust upregulation of phospho-CREB and phospho-p38MAPK in Emc10 KO 354 brown adipocyte lysates (Fig 5G). To confirm that activation of the p38MAPK pathway is 355 required for thermogenic marker induction, we treated KO brown adipocytes with the 356 p38MAPK inhibitor, SB203580. Our results showed that the upregulation of Ucp1 and 357 Pgc1a gene expression was completely abolished by SB203580 treatment in the KO 358 adipocytes (Fig 5H). Similarly, treatment with the CREB inhibitor, HY-101120, also 359 abolished upregulation of basal Ucp1 and Pgc1a expression in the KO brown adipocytes 360 ( Fig 5I). 361 To gain further mechanistic insight, we investigated whether scEMC10 is capable of 362 directly regulating PKA activity. PKA is a holoenzyme composed of two regulatory subunits 363 and two catalytic subunits. For both subunits, there are several isoforms. Among the 364 isoforms of catalytic subunits, PKA catalytic alpha (PKA Ca) is the predominant isoform 365 expressed in adipocytes 36 . Firstly we performed co-immunoprecipitation (Co-IP) of 366 scEMC10 with PKA Ca in 293T cells. We observed that using AKT1 as a control, 367 exogenous scEMC10 directly binds to exogenous PKA Ca, but not exogenous AKT1 (Fig  368   5J). Furthermore, we found that exogenous scEMC10 also directly binds to endogenous 369 PKA Ca in 293T cells (Fig 5K). Next we performed an in vitro kinase assay to confirm the 370 impact of scEMC10 on PKA activity. Our assays showed that recombinant, but not the 371 heat-inactivated, scEMC10 inhibited CREB phosphorylation by PKA (Fig 5L). Our data 372 suggest that PKA signaling pathway is modulated by direct interaction with scEMC10. 373 Taken together, our results show that under basal conditions, scEMC10 suppresses 374 the p38MAPK and CREB signaling pathways, leading to inhibition of adipocyte 375 thermogenesis. 376 377

Circulating EMC10 neutralization reduces body weight in obese mice 378
Our in vivo studies so far are limited in that our loss of function model cannot 379 differentiate the effects of mEMC10 and scEMC10 and the effects in our overexpression 380 paradigm may due to hepatic scEMC10 rather than circulating scEMC10. To reconcile 381 these issues and investigate whether pharmacological inhibition of scEMC10 action could 382 reduce body weight and improve diet-induced metabolic dysfunction, we generated and 383 screened multiple clones of mouse anti-scEMC10 monoclonal antibodies for the ability to 384 neutralize scEMC10 inhibitory effect on CREB phosphorylation. Using our in vitro assay, 385 we identified two monoclonal antibodies (4C2 and 4B12-1) that could repeatedly block 386 scEMC10-mediated CREB inhibition (Supplementary Fig 8A). 387 To confirm the 4C2 antibody-neutralizing efficacy, we used an AAV-based 388 scEMC10 over-expressor model since the methodology to detect and quantify circulating 389 mouse scEMC10 is currently not readily available. Wildtype C57BL/6J mice were 390 intravenously injected with AAV-scEmc10. Thirteen days after the AAV injection, mice 391 were treated with either isotype-matching IgG control or 4C2 neutralizing antibody. 392 Following our earlier experimental design, mice over-expressing scEMC10 were injected 393 with a second dose of IgG or 4C2 antibodies 2 days later. Our results showed that even 394 with elevated circulating scEMC10 levels, administration of 4C2 neutralizing antibody was 395 still able to almost completely diminish its target in the blood (Supplementary Fig 8B). 396 With the newly generated monoclonal antibodies, we treated the C57BL/6J mice 397 after 6 wk of HFD feeding. Obese B6 mice were injected with either isotype-matching IgG 398 control, an anti-EMC10 monoclonal antibody (1F12) that could not neutralize scEMC10 399 effect in our in vitro assay or one of the two anti-EMC10 neutralizing antibodies twice a 400 week. We observed that immediately following the immunological treatment, mice treated 401 with both the neutralizing antibodies (4C2 and 4B12-1), but not the non-neutralizing 1F12 402 antibody or IgG control, demonstrated reduced body weights with stronger effects 403 observed in mice treated with 4C2 antibody (Fig 6A, Supplementary Fig 8C). To confirm 404 the effect of 4C2 antibody on body weight loss, we performed an antibody-swapping 405 experiment where after one-week treatment of 4C2 antibody or control IgG, the two 406 antibodies swapped with each other to treat mice for 4 days followed by swapping back to 407 original respective antibody for another 3-day treatment. We observed that as expected, 408 one-week treatment of 4C2 antibody significantly decreased mouse body weights. After 409 exchanging 4C2 for control IgG, the body weights of mice in the same group significantly 410 increased, before going down after 4C2 antibody was re-instated ( Fig 6B). Similarly, in the 411 control IgG group, the crossover to 4C2 antibody decreased mouse body weights, which 412 then subsequently increased when 4C2 was withdrawn (Fig 6B). These observations 413 clearly demonstrate that scEmc10 inhibition promotes weight loss in mice. 414 Consistent with our mouse genetic studies, the lower body weights of the 4C2-415 treated mice were the result of decreased fat mass with a smaller contribution from changes in liver mass ( Supplementary Fig 8D & 8E). Our histological analyses showed 417 that the decrease in the fat mass in 4C2-treated mice is driven by the reduction in 418 adipocyte size in both the WAT and BAT (Supplementary Fig 8F). We observed Emc10 419 KO improved steatosis caused by HFD (Fig 3H & 3I). Similarly, 4C2 antibody treatment 420 prevented mice from hepatic steatosis, as evidenced by histologically reduced ectopic lipid 421 accumulation and decreased triglyceride content in liver (Fig 6C & 6D). In addition to 422 effects on body composition, scEMC10 neutralization also leads to significantly improved 423 glucose tolerance and insulin sensitivity in obese mice (Fig 6E, Supplementary Fig 8G). 424 We observed beneficial effects of scEMC10 neutralisation on other metabolic parameters, 425 including significantly lower fasting blood glucose, ALT, and fed TG and NEFA, and higher 426 adiponectin in obese mice treated with 4C2 antibody, and significantly lower fed plasma 427 TG and NEFA in 4B12-1 antibody-treated mice when compared with control antibodies 428 ( Fig 6F-6H, Supplementary Fig 8H & 8I). In Emc10 KO mice fed with HFD, we observed 429 increased mRNA levels of thermogenic markers in BAT, which likely accounts for the 430 enhanced thermogenesis observed in these mice (Fig 4B & 4F). Similar to the 431 observations in the KO mice, mRNA levels of several thermogenic markers including 432 Ucp1, Pgc1a, Dio2, and Cox8b, were significantly increased in obese mice treated with 433 4C2 antibody (Fig 6I). In agreement with the transcript data, protein levels of UCP1 and 434 PGC1a markedly increased in BAT of these mice (Fig 6J). Moreover, metabolic cage 435 analysis revealed increased oxygen consumption, carbon dioxide elimination, and heat 436 production in 4C2 antibody-treated obese mice, suggesting neutralization of circulating 437 EMC10 promotes thermogenesis and energy expenditure (Fig 6K). The cAMP/PKA/CREB axis has broad regulatory actions in a range of tissues and, 489 therefore, it is likely that scEMC10 exerts regulatory actions beyond adipose tissue that 490 are independent of its effects on thermogenesis or energy expenditure. Elucidating these 491 functions is beyond the scope of this study, but we expect this is fertile ground for further 492 discovery and will have important implications for any development of scEMC10 as an 493 obesity therapeutic. However, it is worth noting that PKA dependent signaling modulates 494 glucoregulatory processes in liver whereby the net effect is to elevate blood glucose 46 . 495 This is not consistent with the beneficial effects we see on glucose homeostasis when 496 scEMC10 is inhibited in mice. It may simply be that the beneficial effects on adiposity 497 exceed any acute glucose raising effects or glucose-lowering actions of PKA in other 498 tissues outweigh the effects in liver. These explanations notwithstanding -these findings 499 raise the intriguing possibility of tissue specific actions of scEMC10 mediated via selective 500 uptake, differential expression of unidentified inhibitors of scEMC10 or varying sensitivity 501 of different PKA isoforms to scEMC10. 502 Our studies in humans reveal a striking correlation between indices of adiposity and 503 circulating EMC10 in both European and Chinese Han cohorts -thus scEMC10 is a novel 504 biomarker of adipose tissue mass. While we initially reasoned that adipose tissue was the 505 probable source of scEMC10 upregulation in obesity, this hypothesis is not consistent with the fact that scEMC10 is actually decreased in the subcutaneous adipose tissue of obese 507 mice and humans. Given the breadth of expression of scEMC10 it is difficult to ascertain 508 the relative contribution of each tissue to circulating EMC10 and it is possible that different 509 tissues contribute variable amounts depending on the nature of the stimuli. Development 510 of assays to measure circulating mouse EMC10 with adequate sensitivity and tissue 511 specific Emc10 knockout mice will likely be pre-requisite tools to answer this question in 512 future studies.  Table 1). All subjects had a stable weight, defined as the absence of 591 fluctuations of > 2% of body weight for at least 3 months before surgery. 592

Group 2 593
A total of 186 Chinese subjects who were recruited for diabetes screening which 594 were either lean (BMI < 24 kg/m², n=32), overweight (BMI 24-28 kg/m², n=115) or obese 595 (BMI > 28 kg/m², n=39) were also enrolled in the cross-sectional study (Supplementary  596   Table 2). Chinese subjects with the following conditions were excluded: histories of 597 diabetes, acute or chronic inflammatory disease, heart, liver or renal failure, cancer, or 598 active use of oral hypotensive, hypolipidemic, anti-diabetic medications. Serum EMC10 599 levels were investigated in the subjects of this cohort. 600

Group 3 601
In two interventional studies, we measured circulating EMC10 before and 12 602 months after a combined exercise and calorie restricted diet study (n=50), before and 12 603 months after bariatric surgery (n=50) (Supplementary Table 3). We defined the following 604 exclusion criteria: 1) Thyroid dysfunction, 2) alcohol or drug abuse, 3) pregnancy, 4) 605 treatment with thiazolidinediones. 606 All studies were approved by the ethics committee of the University of Leipzig 607 (approval numbers: 159-12-21052012 and 017-12-23012012), or human research ethics 608 committee of Huashan hospital, following the principles of the Declaration of Helsinki. All 609 subjects gave written informed consent before taking part in the study. 610

Generation of mouse monoclonal antibodies against human scEMC10 611
Mouse monoclonal antibodies against human scEMC10 were generated using 612 hybridoma methodologies (Phrenzer Biotechnology, Shanghai, China). Briefly, human 613 scEMC10 was expressed in 293 cells. Recombinant scEMC10 was purified from the supernatant medium and 100 μg was used to immunize each female BALB/c mouse at the 615 age of 6-8 weeks every 2-3 weeks for 4 times. Lymphocytes were subsequently isolated 616 from spleens of the immunized mice and fused with Sp2/0-Ag14 cells to form hybridoma 617 cells. The supernatants of the hybridoma cells were used to react with scEMC10 by ELISA 618 for screening out positive hybridoma cells. scEMC10-specific cells were sorted and then 619 underwent further subcloning. scEMC10-specific hybridoma cells were injected 620 intraperitoneally into BALB/c mice. Ascites was collected in which antibodies were 621 subsequently purified using antigen affinity chromatography. In total, eight mouse 622 monoclonal antibodies against human scEMC10 were validated by ELISA, among which 623 mAb 6B9 and 1F12 were selected as coating and detecting antibody for the sandwich 624 CLIA (chemiluminescent immunoassay) to detect scEMC10 in human serum (Phrenzer 625 Biotechnology, Shanghai, China), respectively. 626

Measurement of scEMC10 in human serum using double sandwich CLIA 627
The CLIA kits for detecting scEMC10 in human serum were obtained from Phrenzer 628 Biotechnology. Briefly, 96-well immunoplates were coated overnight at 2-8 °C with mouse 629 anti-scEMC10 mAb 6B9 at 500 ng/well. Each well was then blocked with blocking buffer 630 (PBS, 0.5% bovine serum albumin, 10% sucrose) at 37°C for 2 hours. After aspirating 631 each well and drying at room temperature for about 24 hours, the immunoplates were 632 ready for use. For performing a CLIA experiment, firstly, add 50 μl of human serum 633 sample or scEMC10 standard to each well. Dispense 50 μl (dilution at 1:50,000) of 634 scEMC10 mAb IF12 HRP-conjugate into each well. Seal the immunoplates with acetate 635 plate sealers. Mix all the wells gently with a shaker at 300-400 rpm for 15 seconds.