Adipocyte-specific ablation of the Ca2+ pump SERCA2 impairs whole-body metabolic function and reveals the diverse metabolic flexibility of white and brown adipose tissue

Objective Sarco/endoplasmic reticulum Ca2+-ATPase (SERCA) transports Ca2+ from the cytosol into the endoplasmic retitculum (ER) and is essential for appropriate regulation of intracellular Ca2+ homeostasis. The objective of this study was to test the hypothesis that SERCA pumps are involved in the regulation of white adipocyte hormone secretion and other aspects of adipose tissue function and that this control is disturbed in obesity-induced type-2 diabetes. Methods SERCA expression was measured in isolated human and mouse adipocytes as well as in whole mouse adipose tissue by Western blot and RT-qPCR. To test the significance of SERCA2 in adipocyte functionality and whole-body metabolism, we generated adipocyte-specific SERCA2 knockout mice. The mice were metabolically phenotyped by glucose tolerance and tracer studies, histological analyses, measurements of glucose-stimulated insulin release in isolated islets, and gene/protein expression analyses. We also tested the effect of pharmacological SERCA inhibition and genetic SERCA2 ablation in cultured adipocytes. Intracellular and mitochondrial Ca2+ levels were recorded with dual-wavelength ratio imaging and mitochondrial function was assessed by Seahorse technology. Results We demonstrate that SERCA2 is downregulated in white adipocytes from patients with obesity and type-2 diabetes as well as in adipocytes from diet-induced obese mice. SERCA2-ablated adipocytes display disturbed Ca2+ homeostasis associated with upregulated ER stress markers and impaired hormone release. These adipocyte alterations are linked to mild lipodystrophy, reduced adiponectin levels, and impaired glucose tolerance. Interestingly, adipocyte-specific SERCA2 ablation leads to increased glucose uptake in white adipose tissue while the glucose uptake is reduced in brown adipose tissue. This dichotomous effect on glucose uptake is due to differently regulated mitochondrial function. In white adipocytes, SERCA2 deficiency triggers an adaptive increase in fibroblast growth factor 21 (FGF21), increased mitochondrial uncoupling protein 1 (UCP1) levels, and increased oxygen consumption rate (OCR). In contrast, brown SERCA2 null adipocytes display reduced OCR despite increased mitochondrial content and UCP1 levels compared to wild type controls. Conclusions Our data suggest causal links between reduced white adipocyte SERCA2 levels, deranged adipocyte Ca2+ homeostasis, adipose tissue dysfunction and type-2 diabetes.


INTRODUCTION
The white adipose tissue (WAT) is important for maintenance of metabolic homeostasis in the body. Besides storing excess energy in the form of triglycerides, WAT also regulates whole-body metabolism by secreting a large variety of bioactive molecules, referred to as adipokines. Adipokines may be cytokines (e.g. interleukins and tumour necrosis factor alpha) whereas others, including adiponectin, leptin and resistin, are protein hormones. Obese adipose tissue is commonly dysfunctional with decreased triglyceride storage capacity and disturbed adipokine release, leading to ectopic lipid deposition, chronic low-grade inflammation, and systemic insulin resistance [1,2]. The endoplasmic reticulum (ER) constitutes the folding compartment for proteins entering a secretory pathway and is also a Ca 2þ reservoir, important for controlling the dynamic regulation of Ca 2þ -dependent cellular functions including proliferation and differentiation, contraction, gene transcription, hormone secretion and apoptosis. Disrupted ER function causes mis-or unfolded proteins to accumulate within the ER and activates a process known as unfolded protein response (UPR). A high concentration of Ca 2þ within the ER is essential for the normal function of chaperones and enzymes involved in protein folding and thus for the ability of UPR to resolve ER stress [3]. Moreover, intracellular Ca 2þ has been shown to regulate several adipocyte-specific processes, such as lipolysis [4e6] and glucose uptake [7,8]. Ca 2þ is also important for regulation of the secretion of adipocyte hormones, such as the appetite-controlling hormone leptin [9e11]. In earlier studies, we defined an important regulatory role of Ca 2þ for secretion of the adipocyte-specific insulin-sensitizing hormone adiponectin; Ca 2þ potently augments cAMP-stimulated adiponectin exocytosis and is also required for sustained adiponectin release over prolonged time periods [12,13]. The regulation of Ca 2þ fluxes between ER and the cytosol is largely dependent on sarco/endoplasmic reticulum Ca 2þ ATPase (SERCA) pumps that transport Ca 2þ from the cytosol into the ER. There are three paralogous genes that code for SERCA: Serca1 (Atp2a1), Serca2 (Atp2a2), and Serca3 (Atp2a3). SERCA activity has also been linked to uptake of Ca 2þ in the mitochondria, another important intracellular Ca 2þ storage compartment [14]. The sequestration of Ca 2þ within mitochondria impacts oxidative phosphorylation as well as cytosolic Ca 2þ signals and excessive mitochondrial Ca 2þ can trigger cell death (as reviewed in [15,16]). Interestingly, studies in the heart, the liver and in pancreatic insulin-secreting beta-cells have affirmed SERCA pumps as key players in the pathophysiology of type-2 diabetes (T2D) [17e21]. However, diseases associated with disrupted ER Ca 2þ dynamics have received insufficient attention and the role of SERCA for adipose tissue functionality is largely uninvestigated. Here we show that SERCA2 is the paralog chiefly expressed in white mouse adipocytes isolated from both subcutaneous and visceral depots. Our study also demonstrates that SERCA2 is downregulated in adipocytes isolated from high fat diet (HFD)-induced obese mice and from human subjects with obesity and T2D. Moreover, genetic ablation of this Ca 2þ pump results in a disturbed secretion of adipocyte hormones and impaired mitochondrial function in white and brown adipocytes, associated with whole-body glucose intolerance. Our study highlights an important causal link between a disturbed Ca 2þ homeostasis in adipocytes and metabolic disease. Life Technologies) and differentiated into mature adipocytes as previously described [13]. Briefly, cells were grown to confluence (day 0) and thereafter incubated in medium 1 supplemented with 1 mM dexamethasone, 850 nM insulin, and 0.5 mM 3isobutyl-1-methylxanthine (IBMX). After 48 h (day 2) the medium was changed to medium 1 supplemented with insulin only. After an additional 48 h (day 4) the medium was then replaced by medium 1 alone and was thereafter freshly replaced every 2 days. Experimental studies were performed in mature 3T3-L1 adipocytes between days 8 and 10 from the start of differentiation. IWAT stromal vascular fraction (SVF) were isolated as previously described [22]. IWAT SVF were differentiated into mature adipocytes following the same protocol used for 3T3-L1 preadipocyte, but with the addition of 1 mM rosiglitazone (Sigma) in the first differentiation step.

Reagents and tools
Brown fat SVF was isolated and differentiated as previously described [23].

Tissue triglyceride content
The triglyceride content was measured following instructions of manufacturer (Randox Laboratories, Dublin, United Kingdom). Briefly, frozen tissue (30e100 mg of liver) was homogenized using a tissue lyser (3 min at 25 Hz) in chloroform:methanol (2:1), washed with NaCl 0.9% and evaporated overnight. Lipid samples were resuspended in isopropanol and trigycerides were measured.
2.3.6. Isolation of mouse adipose tissue mitochondria Mitochondria of IWAT or BAT were isolated by differential centrifugation, based upon [27]. Freshly dissected adipose tissue was minced with scissors in w10 volumes of ice cold MSHE þ BSA (70 mM sucrose, 210 mM mannitol, 5 mM HEPES, 1 mM EGTA and 0.5% (w/v) fatty acidfree BSA, pH 7.2, 4 C). The tissue was disrupted using a drill-driven Teflon glass homogenizer with 4e5 strokes. Homogenate was centrifuged at 800 g for 10 min at 4 C, the lipid layer was carefully aspirated, and the remaining supernatant was separated and centrifuged at 8000 g for 10 min at 4 C. The pellet was resuspended in MSHE, and the centrifugation was repeated. The final pellet was resuspended for further analysis of oxygen consumption or western blot. Total protein was determined using Bradford Assay reagent (Bio-Rad).

Pancreatic islets isolation
Pancreata were injected through the portal vein with 3e5 mL cold Hanks' buffer containing 0.95 mg/mL collagenase V (Sigma, USA). The collected pancreata were incubated at 37 C for 7 min. The released islets were washed with Hanks' buffer (0.1% BSA) 3e5 times and then hand-picked under a dissecting microscope to ensure that the islet preparation was pure. The islets were first cultured for 1 h in RPMI 1640 containing 10% FBS, 100 U/mL Penicillin/Streptomycin, 10 mM glucose. After washing the islets once with KRBH buffer þ 0.1% BSA, 10e12 islets of similar size were handpicked in 0.3 mL KRBH buffer (0.1% BSA) containing 1 mM glucose. The islets were then pre-incubated KRBH buffer (0.1% BSA) containing 1 mM glucose for 30 min at 37 C, followed by 1 h incubation in 0.3 mL KRBH buffer (0.1% BSA) containing (in mM): 1, 6 or 20 glucose, 20 glucose þ 0.11 Tolbutamide or 20 glucose þ 10 Arginine. Insulin was measured in the collected medium.

Measurements of mitochondrial content and intracellular ROS
The mitochondrial DNA (mtDNA), an index of mitochondrial content (mtDNA copy number), was carried out in brown adipose tissue as previously validated and described [31]. Briefly, DNA was extracted using QIAzolÒ (Qiagen) protocol. Real-time PCR was performed to amplify the mitochondrial gene 16S which is exclusive of mtDNA, and a nuclear gene (mtDNA/nDNA ratio). In cultured brown adipocytes, Mitotracker Green (Thermo Fisher Scientific, Waltham, MA, USA) was used at 200 nM for 90 min, meanwhile Hoechst (Thermo Fisher Scientific, Waltham, MA, USA) was used at 2 mg/ml for 10 min followed by wash-out before imaging (Zeiss LSM 710 confocal microscope). Mitotracker Green was excited by a 488 nm laser, while 361 nm laser was used for detection of Hoechst.
Imaging was performed with a 40Â objective. The dye signal was quantified with the Image J analysis software and normalized to the number of cells (Hoechst dye). The ROS assay was performed as previously described [32]. Values were expressed as the ratio between the fluorescence of the oxidized CM-H2DCFDA (Thermofisher) and Hoechst, after subtraction of the signal of not-stained cells.

Statistical analysis
Statistical analysis was performed with GraphPad Prism 8.0 (GraphPad Software, CA, USA). Statistical parameters, including n values, are noted in figure legends. Data are presented as means AE standard error of the mean (SEM) and a p < 0.05 was considered significant. Statistical analysis of data with more than 2 groups was assessed with one-way ANOVA. Data with two groups was analysed by two-tailed Student's t test.

RESULTS
3.1. White adipocyte SERCA2 is down-regulated in HFD-induced obese mice and in obese type-2 diabetic humans By DNA microarray analysis, we found that Serca2 (Atp2a2) and -3 (Atp2a3) are the most abundant Serca-paralogs in mouse inguinal, gonadal, and mesenteric WAT (IWAT, GWAT and MWAT) with expression both in the adipocyte and in the stromal vascular fraction (SVF), while Serca1 (Atp2a1) was detected at extremely low levels. In isolated adipocytes, Serca2 was the predominant paralog, although Serca3 was also highly expressed (Suppl. Fig. 1A). SERCA2 expression and/or activity are reduced in liver, pancreatic islets, and cardiomyocytes in animal models of obesity and T2D [18,33,34]. This prompted us to investigate whether HFD-induced obesity reduces Serca2 also in adipose tissue. Indeed, in both IWAT and GWAT Serca2 mRNA levels were reduced from the fourth week in HFD-induced obese mice. In contrast, Serca3 mRNA levels were unaltered in both fat depots until week 16 of HFD, when the expression was dramatically decreased in IWAT but not in GWAT (Figure 1AeD). This HFD-induced decrease in WAT Serca2 was confirmed at the protein level (Suppl. Fig. 1BeC) and was due to reduced levels specifically in adipocytes, while SVF levels of Serca2 remained unaffected (Figure 1Ee  F). The expression of Serca3 was unaltered in the isolated IWAT adipocyte fraction and tended to be downregulated in SVF from this adipose tissue depot ( Figure 1G). Serca3 mRNA levels were however upregulated in GWAT adipocytes and unchanged in the SVF fraction ( Figure 1H). In agreement with data from HFD-challenged adipocytes, 24 h palmitate treatment downregulated the SERCA2 protein levels in  Fig. 1D). In contrast to findings in WAT, the brown adipose tissue (BAT) Serca2 expression was slightly elevated in HFD-induced obese mice while Serca3 expression was unaltered ( Figure 1I). To determine whether adipocyte SERCA2 is relevant also for human pathophysiology, we measured SERCA2 protein in subcutaneous adipocytes obtained from subjects with and without obesity-associated T2D. The SERCA2 levels were w20% lower in adipocytes from the obese individuals compared to the nonobese controls ( Figure 1J) and correlated inversely with the plasma glucose levels of adipocyte donors ( Figure 1K). As hypoxia downregulates SERCA2 in cardiomyocytes [35e37], we hypothesized that a similar mechanism may be at play in enlarged hypoxic adipocytes of insulin-resistant subjects [38,39]. Indeed, hypoxia (24 h) led to reduced SERCA2 protein and mRNA levels in 3T3-L1 adipocytes ( Figure 1L, Suppl. Fig. 1E). Serca3 was also reduced by hypoxia but was expressed at a 3000-fold lower level than Serca2 (Suppl. Fig. 1F).
In conclusion, our results pinpoint Serca2 as the major SERCA paralog that is downregulated in white adipocytes from both diet-induced obese mice and obese/T2D patients.

Inhibition of SERCA in adipocytes is associated with altered
Ca 2þ homeostasis and defective adiponectin exocytosis We next tested how SERCA inhibition affects the adipocyte Ca 2þ homeostasis by performing ratiometric time-lapse recordings of intracellular Ca 2þ concentrations. Pre-treatment of 3T3-L1 adipocytes with 20 mM of the SERCA inhibitor cyclopiazonic acid (CPA) reduced the ER calcium store by w50%, as shown by the diminished Ca 2þ increase in response to acute addition of thapsigargin (another SERCA inhibitor) during the recording (Figure 2AeB). A similar reduction of stored ER calcium was found in cultured adipocytes genetically ablated for SERCA2 (Figure 2CeE). The reduced Ca 2þ stores in SERCA2-ablated adipocytes were associated with the transcriptional upregulation in IWAT of two key players of cellular Ca 2þ influx mediated via storeoperated Ca 2þ entry (SOCE): the stromal interaction molecule 1 (Stim1) that recognized when ER Ca 2þ levels are low, and the calcium release-activated calcium channel protein 1, Orai1 [40] (Suppl. Fig. 2A). We also observed a downregulation of the Ryanodine Receptor channel (Ryr3) that mediate Ca 2þ release from ER to cytosol [41] (Suppl. Fig. 2A). It has been demonstrated that the extracellular application of ATP rapidly elevates cytosolic Ca 2þ in adipocytes via purinergic P2Y2 receptors and STIM1/ORAI1-mediated activation of SOCE [42,43]. We thus tested the effect of ATP on cultured SERCA2-ablated adipocytes. Mitochondria can accumulate a 10-fold higher concentration of Ca 2þ than that measured in the cytosol and Ca 2þ is transferred between ER and the mitochondria via mitochondria-associated membranes (MAMs), specialised regions of communication between the two organelles [44]. Mitochondrial Ca 2þ levels were therefore measured in parallel. In agreement with [42,43], extracellular addition of ATP (100 mM) elevated [Ca 2þ ] i in control adipocytes while the ATP-induced [Ca 2þ ] i peak was reduced in SERCA2deficient adipocytes (Figure 2FeG). The ATP-triggered mitochondrial Ca 2þ elevations were similarly affected in these knockout adipocytes (Suppl. Fig. 2BeC). This finding was associated with a transcriptional upregulation of the mitochondrial calcium uniporter (Mcu) that mediates electrophoretic Ca 2 ⁺ uptake into the mitochondrial matrix [45] in IWAT of adipocyte-specific SERCA2 deficient mice (Suppl. Fig. 2A). Thus, SERCA2 regulates Ca 2þ in both the mitochondria and the ER of white adipocytes.
To further explore the importance of SERCA for metabolic regulation, we studied the role of these Ca 2þ pumps in adiponectin secretion. Ca 2þ is important for sustained adiponectin secretion [12,46] and both mitochondrial dysfunction and ER stress have been linked to reduced adiponectin production in adipocytes [47,48]. The effect of SERCA inhibition on regulated adiponectin exocytosis has however not previously been studied. CPA treatment induced ER stress in 3T3-L1 adipocytes (as verified by increased Bip and Xbp1s expression), reaching significance after 4 h (Suppl. Fig. 2DeE). Adiponectin secretion stimulated by a combination of forskolin and IBMX (elevates intracellular cAMP and also slightly increases Ca 2þ in adipocytes; [13]) was blunted in CPA-treated cells ( Figure 2H) whereas basal (unstimulated) adiponectin release was unaffected (not shown). Adiponectin secretion triggered by forskolin/IBMX or the beta 3 adrenergic receptor agonist CL-316,243 (CL) was similarly blunted also in primary adipocytes genetically ablated for SERCA2 ( Figure 2I). Basal adiponectin release was not altered in SERCA2 null adipocytes compared to wild type ( Figure 2J).

Adipocyte-specific SERCA2 knockout causes ER stress in IWAT and BAT
To define the role of SERCA2 for adipocyte functionality and whole-body metabolic health, we generated an adipocyte-specific Serca2 knockout mouse model. The knockouts displayed w60e70% reduced Serca2 levels in IWAT, GWAT and BAT, while the SVF Serca2 expression was unaffected ( Figure 3A). SERCA2 protein levels were reduced to a similar degree in isolated Serca2 knockout adipocytes ( Figure 3B). There was a compensatory upregulation of Serca3 expression in the adipocyte fraction of WAT and BAT, and this Serca3 increase was particularly pronounced in GWAT adipocytes ( Figure 3C). Downregulation of Serca2 was associated with increased expression of the UPR-markers Xbp1s and Bip and the inflammatory markers Emr1 (F4/80), Ccl2 (Mcp1) and Tnfa in IWAT and BAT, indicating ER stress (Figure 3DeE). In support of a higher degree of IWAT inflammation, there was also an increased abundance of crown-like structures ( Figure 3F). Interestingly, the GWAT expression of UPR and inflammation markers was unaffected by knockdown of Serca2 (Suppl. Fig. 3), possibly due to the compensatory elevation of Serca3 in this fat depot ( Figure 3C).

3.4.
Adipocyte-specific SERCA2-deficient mice are mildly lipodystrophic and display whole-body metabolic dysfunction Male and female adipocyte-specific SERCA2 knockout mice grew normally after weaning, but displayed reduced WAT mass, while their liver, pancreas and BAT masses were increased (Figure 4AeC). Heart (0.183 AE 0.006 vs. 0.186 AE 0.007 g in males) and skeletal muscle (Gastrocnemius; 0.202 AE 0.008 vs. 0.212 AE 0.015 g in males) masses were unchanged. This lipodystrophic phenotype was associated with impaired whole-body metabolic function, as judged by slightly reduced glucose tolerance and hyperinsulinemia (Figure 4De G), as well as a trend towards increased glucagon levels (2.61 AE 0.33 vs. 4.44 AE 0.97 pmol/L, p ¼ 0.122). The enlarged pancreata of adipocyte-specific SERCA2 knockout mice displayed normal morphology (judged by haematoxylin and eosin staining), and the islet density and size distribution per analysed section were similar between genotypes (Suppl. Fig. 4AeC). Insulin secretion induced by high glucose, arginine and tolbutamide was elevated in adipocyte-specific SERCA2 knockouts (Suppl. Fig. 4D). The increased BAT weight appeared to primarily result from increased fat deposition, as judged by the whiteish BAT appearance ( Figure 4H). The hepatic triglyceride content was similar between genotypes (Figure 4IeJ). When fed HFD (to induce obesity), control and adipocyte-specific SERCA2 knockout mice exhibited similar weight gain (Suppl. Fig. 4E), and the difference in WAT, BAT and liver mass persisted between genotypes (Suppl. Fig. 4F). In this metabolically challenged setting, the hepatic lipid deposition was increased in the  (Figure 4IeJ). The glucose intolerance was aggravated by HFD, but the differences between genotypes were similar to those observed at chow conditions (Suppl. Fig. 4GeJ).
To elucidate the underlying reasons for the reduced WAT mass, we measured IWAT adipocyte size. During chow conditions, the adipocytespecific SERCA2 knockouts displayed smaller IWAT adipocytes (although the difference did not reach significance). Despite w60% decreased IWAT weight, the SERCA2 knockout adipocytes of HFD-fed mice were w50% larger than littermate wild type control adipocytes (Figure 4KeL). This increase in average adipocyte size was primarily due to a doubling of the percentage of larger adipocytes (>175 mm in diameter), while the fraction of smaller adipocytes was lower (Suppl. Fig. 4K). These data indicate that SERCA2 deficiency does not lead to impaired lipid storage capacity of an individual adipocyte, but that there is an increase in HFD-induced adipocyte death (as implied by the increased occurrence of crown-like structures, Figure 3F) and/or a decreased formation of new adipocytes in these mice. Given that only adiponectin-expressing cells (i.e. mature adipocytes), are SERCA2deficient in this mouse model and that HFD-induced expansion largely depends on hypertrophy [49], we argued that increased adipocyte death is more likely to explain the fewer but larger IWAT adipocytes in HFD-fed mice. Indeed, the IWAT levels of cleaved caspase 3 were increased in adipocyte-specific SERCA2 knockout mice ( Figure 4M) and the adipogenic capacity was similar between genotypes as judged by ex vivo differentiation of IWAT SVF cells (Suppl. Fig. 4L).
3.5. Levels and secretion of adipocyte hormones are altered in adipocyte-specific SERCA2 knockout mice As SERCA2-deficient adipocytes display abrogated regulated release of adiponectin (Figure 2HeI), we determined whether adipocyte-specific SERCA2 knockouts have altered levels of the adipose tissue hormones adiponectin, resistin and leptin. Both male and female adipocytespecific SERCA2 knockout mice had much reduced levels of circulating adiponectin and this was associated with decreased adiponectin protein and gene expression in IWAT and GWAT (Figure 5AeB, Suppl. Fig. 5A). In contrast, serum HMW adiponectin levels were similar between adipocyte-specific SERCA2 knockouts and littermate wild type controls. However, HFD-feeding reduced both total and HMW adiponectin in both genotypes ( Figure 5C). Because of the unaltered HMW adiponectin levels, the HMW to total adiponectin ratio was significantly higher in the male knockouts with a similar trend in females (Suppl. Fig. 5B). The ER chaperones disulphide-bond A oxidoreductase-like protein (DsbA-L), ER membrane-associated oxidoreductase (Ero1-La) and its associated protein Erp44 have been identified as critical players in the assembly and secretion of HMW adiponectin [50,51]. We found increased gene expression of Ero1la in IWAT from both male and female knockouts, and higher gene expression of Erp44 in IWAT from the female SERCA2 null mice (Suppl. Fig. 5C).
Resistin has been suggested to contribute to the pathogenesis of metabolic diseases, although its exact role is not fully elucidated [52e 54]. Similar to adiponectin, resistin has a complex structure with multimeric assembly [55] and the two hormones are in mouse adipocytes co-released from the same vesicles [56]. As expected, circulating resistin levels were dramatically reduced both at chow and HFD-fed conditions ( Figure 5D) and the stimulated resistin secretion was blunted in primary SERCA2-ablated adipocytes, while the basal release was increased ( Figure 5EeF); IWAT and GWAT resistin (Retn) mRNA levels were also much reduced (Suppl. Fig. 5D). In contrast to adiponectin and resistin, leptin is released as a monomer in proportion to the degree of adiposity. We therefore expected a less  potent effect of SERCA2 ablation on leptin levels. In line with this assumption, circulating leptin levels as well as IWAT and GWAT leptin (Lep) mRNA levels were similar between genotypes in HFD-fed mice ( Figure 5G and Suppl. Fig. 5E). Circulating leptin and IWAT Lep mRNA levels were however reduced in chow-fed knockouts, in agreement with their lower fat mass ( Figure 5G and Suppl. Fig. 5E).
3.6. Opposite effect of adipocyte-specific SERCA2 ablation on white and brown adipose tissue glucose uptake We speculated that the phenotype of SERCA2 knockout mice is due to the systemic impact of impaired metabolic function of white and brown adipocytes. To investigate this, we used [ 14 C]-labelled glucose to measure tissue glucose uptake and de novo lipogenesis 2 h after an oral glucose load. Contrary to our expectations, we found that the glucose uptake was increased in WAT (represented by IWAT, GWAT and MWAT), while only BAT displayed a clear reduction in glucose uptake ( Figure 6A). The enhanced glucose uptake in knockout WAT was primarily due to increased [ 14 C]-counts in the aqueous phase, likely reflecting glucose oxidation ( Figure 6A). In contrast, [ 14 C]counts in the organic phase (representing de novo lipogenesis) dominated in BAT ( Figure 6A). Glucose uptake was also augmented in liver, heart, skeletal muscle, and pancreas, largely due to an increase in the aqueous phase ( Figure 6A). Thus, the impaired glucose uptake in BAT appears to be largely compensated for by increased glucosestimulated insulin release from pancreas leading to enhanced glucose uptake in insulin-sensitive tissues. This provides an explanation for the nearly normal whole-body glucose tolerance ( Figure 4D).

Pharmacological inhibition of SERCA is associated with mitochondrial dysfunction in 3T3-L1 adipocytes
The increased glucose uptake and metabolism in WAT from adipocytespecific SERCA2 null mice urged us to investigate the effect of SERCA inhibition on mitochondrial function. Imaging of Rhod-2 loaded 3T3-L1  adipocytes showed that the basal mitochondrial Ca 2þ accumulation was increased by CPA treatment ( Figure 6B), indicating transfer of Ca 2þ from the ER to the mitochondria [44]. The increase in mitochondrial Ca 2þ was accompanied by a higher ROS production and reduced oxygen consumption rate (OCR; Figure 6CeD). As expected, administration of the mitochondrial Ca 2þ uptake inhibitor RU360 reduced both basal and CPA-induced mitochondrial Ca 2þ and ROS levels ( Figure 6BeC). As shown in Figure 6E, the cell mitochondrial content (judged by Mitotracker green staining) was unchanged by CPA but reduced by RU360 treatment. Taken together, our results indicate that acutely disturbed intracellular Ca 2þ homeostasis affects the mitochondrial function in adipocytes.
3.8. Adipocyte-specific SERCA2-deficient mice display altered mitochondrial dysfunction in IWAT and in BAT Lack of SERCA2 in white adipocytes may result in enhanced metabolism (as indicated by the increased WAT glucose uptake) through a mitohormesis-like process. We explored this possibility by analysing the expression of browning markers and found a trend for increased IWAT Ucp1 levels in the adipocyte-specific SERCA2 knockout mice, while the browning markers Prdm16 and Dio2 were downregulated (Suppl. Fig. 6A). Notably, UCP1 protein levels in isolated IWAT mitochondria were w1.5-fold higher in SERCA2 ablated mice than in littermate wild type controls, while electron transport chain protein levels were similar between genotypes ( Figure 6F). Furthermore, the OCR of isolated IWAT mitochondria was much higher (Figure 6GeH). This increase in mitochondrial OCR was primarily due to a lower coupling efficiency in SERCA2-deficient adipocytes ( Figure 6H). To test if this effect of SERCA2 ablation on mitochondrial function is cell autonomous, we analysed the mitochondrial function in in vitro differentiated adipocytes from IWAT SVF. Indeed, also cultured SERCA2-deficient adipocytes displayed elevated uncoupled OCR and the ATP production-linked OCR was slightly increased ( Figure 6I). The altered expression of browning markers was similar in BAT and IWAT; BAT Prdm16 and Dio2 levels were reduced, while Ucp1 was unchanged in adipocyte-specific SERCA2 knockouts compared to littermate wild type controls (Suppl. Fig. 6B). Moreover, the level of total electron transport proteins was reduced along with increased UCP1 protein levels in isolated mitochondria from BAT of adipocytespecific SERCA2 knockout mice ( Figure 6J). In sharp contrast to SERCA2-deficient white adipocytes, mitochondria of BAT adipocytespecific SERCA2 knockout mice displayed lower OCR ( Figure 6K). Accordingly, OCR was decreased in cultured SERCA2-deficient brown adipocytes, as shown by the reduction of basal and ATP-productionlinked respiration and other key mitochondrial respiration parameters, ( Figure 6L). Surprisingly, this reduction in mitochondrial OCR was associated with elevated markers of mitochondrial content in both BAT and differentiated brown adipocytes (Figure 6MeN). However, the capacity to increase the respiration in response to norepinephrine was unaffected (Suppl. Fig. 6C).
Several studies have shown that ER stress can lead to an adaptive increase in fibroblast growth factor 21 (FGF21) [57,58], a metabolic hormone capable of inducing thermogenesis in BAT and promoting a brown fat-like thermogenic program in white adipocytes [59,60]. In line with the increased UPR marker expression ( Figure 3D), we found that the expression of FGF21 was elevated in adipocyte-specific SERCA2 knockout IWAT (Figure 6OeP). However, such an ERstress/UPR-associated induction of FGF21 was not observed in BAT ( Figures 3E and 6OeP). As expected, the hepatic Fgf21 mRNA and serum protein levels were similar between genotypes ( Figure 6O, Suppl. Fig. 6D). Thus, our data suggest that genetic ablation of SERCA2 leads to different adaptations in white and brown adipocytes.

DISCUSSION
Here we show that diet-induced obesity in mice, T2D in humans, and exposure to palmitate or hypoxia are associated with lower SERCA2 levels in white adipocytes. SERCA2-deficient white adipocytes display altered calcium homeostasis, hormone release and mitochondrial function, and adipocyte-specific SERCA2 knockout mice present with hyperinsulinemia, glucose intolerance and lower adiponectin levels, likely driven by both white and brown adipocyte dysfunction. In brief, the WAT of this mouse model fails to expand properly, presumably because of increased adipocyte death. However, the remaining SERCA2-deficient white adipocytes show accelerated metabolism, as judged by increased glucose uptake and increased OCR. In contrast, BAT is enlarged but less metabolically active, as demonstrated by its whiter appearance, decreased glucose uptake, and decreased OCR. Furthermore, this mouse model illustrates how genetically normal tissues to some extent can compensate for adipose tissue defects. For instance, the glucose uptake is dramatically increased in liver, heart, skeletal muscle and pancreas, and the glucose-stimulated insulin release is enhanced, both in vivo and ex vivo in isolated islets.

Maintained Ca 2þ dynamics and ER function are essential for adipocyte hormone secretion
The ER is the first compartment of the secretory pathway, where proteins are folded, oligomerized and finally packaged into vesicles that continue their route towards the plasma membrane. Protein folding and oligomerization in ER is controlled by Ca 2þ -dependent chaperones [3] and obesity-associated ER stress has been shown to impair adipocyte hormone secretion [1]. In agreement with this, we found that circulating levels of adiponectin are reduced in the adipocyte-specific SERCA2 null mice ( Figure 5A). However, the HMW form of adiponectin remained unaffected in serum from SERCA2 ablated mice ( Figure 5C). A few studies suggest that reduced circulating levels of specifically the HMW form of adiponectin is connected to development of obesity-associated metabolic disease and that higherorder adiponectin complexes protect from progress to T2D [61,62]. Thus, the adiponectin aberrations observed in the adipocyte-specific SERCA2 null mice do not fully concur with adiponectin alterations reported in metabolic disease. The found increase in Erp44 and Ero1-La, two ER chaperones belonging to the protein-disulphide isomerase family may underly this difference [63]. Disulphide bonds form and stabilize the higher-order adiponectin complexes [64] and Erp44 and Ero1-La are critically important for the proper ER-located modifications and the secretion of adiponectin. Cycloheximide-induced inhibition of protein synthesis, which leads to retention of proteins within the ER, resulted in production of more HMW adiponectin, at the expense of smaller adiponectin forms [65]. Thus, we suggest that disturbed Ca 2þ dynamics and ER stress in SERCA2-ablated adipocytes result in retention of adiponectin within the ER. This, together with the concomitant upregulated expression of Erp44 and Ero1-La, promote the assembly of HMW adiponectin at the expense of smaller adiponectin forms. Furthermore, circulating adiponectin levels may not only reflect the secretory function of individual adipocytes, but rather the sum of the functionality of different fat depots (as well as effects of negative feedback and hepatic clearance). GWAT and IWAT display different adipocyte hormone release patterns [66,67] and in obese subjects, small sized visceral adipocytes were positively correlated to serum HMW adiponectin [68]. The finding that SERCA2 knockout GWAT adipocytes displayed a dramatic increase in Serca3 expression ( Figure 3B) points to a less pronounced phenotype in this fat depot. Thus, the higher HMW/total adiponectin in adipocyte-specific SERCA2 knockout mice, may result also from GWAT adipocytes remaining more functional and contributing more to the levels of higher-order adiponectin complexes. The circulating levels of resistin are dramatically reduced in the adipocyte-specific SERCA2 knockouts ( Figure 5D). This is not surprising as resistin, like adiponectin, circulates in differently sized molecular forms [69] and maintained ER chaperone activity is likely essential for its assembly and secretion. Moreover, it has been suggested that resistin itself functions as a chaperone and is retained within the ER during ER stress to protect from cell apoptosis [70]. Leptin, on the other hand, is released from adipocytes as molecular monomers and yet circulating leptin levels are also lower in adipocyte-specific SERCA2 knockouts. We attribute this difference in leptin levels primarily to the reduced adipose tissue mass in the adipocyte-specific SERCA2 knockouts [71]. In support of this, serum leptin levels were similar in HFD-fed wild type and adipocyte-specific SERCA2 knockout mice, the latter with increased adipocyte size which is strongly associated with increased leptin secretion and levels [72,73]. Thus, our study indicates that loss of adipocyte SERCA2 in obesity/T2D can mediate alterations in adipokine levels, although genetic SERCA2 ablation causes more dramatic effects than those observed in metabolic disease where the reduction in adipocyte SERCA2 is less dramatic ( Figure 1E, F, J).

Genetic ablation of SERCA2 impairs calcium homeostasis and induces different thermogenic adaptations in white and brown adipocytes
Adipocyte-specific SERCA2 knockout BAT and IWAT both display reduced expression of Prdm16 and Dio2 (Suppl. Fig. 6AeB) and increased mitochondrial UCP1 levels ( Figure 6F, J). However, BAT glucose uptake and mitochondrial respiration are reduced while we observed accelerated metabolism in IWAT of the adipocyte-specific SERCA2 knockout mice ( Figure 6A, G, H, K). These differences in mitochondrial respiration between genotype are seen also in cultured white and brown adipocytes ( Figure 6I, L) arguing for a cell autonomous mechanism. Similar differences in IWAT and BAT mitochondrial responses to stressors have been reported previously [26,74,75]. Moreover, the OCR was recently reported to increase in subcutaneous adipocytes from insulin resistant compared to insulin sensitive obese subjects [76], which is consistent with our findings (Figures 1JeK and  6GeH). In contrast to genetic SERCA2 ablation, acute pharmacological SERCA inhibition in 3T3-L1 adipocytes increases the mitochondrial Ca 2þ levels associated with increased ROS production and reduced OCR (Figure 6BeD). Thus, the browning-like effect in SERCA2-deficient adipocytes appears to be part of an adaptive ER stress-induced mitohormetic that does not occur upon acute SERCA inhibition. Based on several reports [77e81], we propose that the upregulation of UCP1 ( Figure 6F, J) is triggered by increased mitochondrial ROS levels, to suppress further increases in the ROS production and Ca 2þ overload in SERCA2-deficient adipocytes. It is also possible that the already high levels of UCP1 and endogenous antioxidant enzymes prevent the mitohormetic response in BAT, as has been suggested previously [75]. The different adaptations of white and brown adipocytes to SERCA2 ablation may also be explained by differential induction of FGF21; FGF21 was increased in IWAT but not in BAT ( Figure 6N). However, the reduction in mitochondrial metabolism may also signpost a specific role of SERCA2 in mitochondrial function of brown adipocytes that cannot be compensated for by SERCA1 and 3. Indeed, the SERCA-ryanodine receptor (RyR)-pathway has been suggested to be an evolutionary conserved mechanism for non-shivering thermogenesis [82]. The amount of heat released by SERCA-catalysed ATP hydrolysis varies dependent on how much of the energy is used to pump Ca 2þ . For instance, sarcolipin enhances muscle thermogenesis by uncoupling SERCA-mediated ATP hydrolysis from Ca 2þ transport [83e85] and SERCA2 has been implicated in ATP-dependent thermogenesis by Ca 2þ cycling that can compensate for loss of UCP1 in beige adipose tissue [82]. The finding that the norepinephrine-induced increase in OCR was normal in SERCA2-deficient brown adipocytes (Suppl. Fig. 6C) is thus in line with that UCP1 is essential for norepinephrine-induced thermogenesis in BAT [86]. It is however possible that adrenergically stimulated thermogenesis is slightly blunted in SERCA2-deficient white adipocytes. Alternatively, the upregulation of UCP1 that we observe ( Figure 6F) serves to compensate for the lack of this thermogenic mechanism in SERCA2-deficient adipocytes.

SERCA2 in fat accumulation
The lipodystrophic phenotype of adipocyte-specific SERCA2 knockout mice is in agreement with the proposed role of SERCA and Ca 2þ homeostasis in fat storage [87]. However, our data suggest that fat storage and the capacity for de novo lipogenesis is normal in SERCA2 deficient adipocytes ( Figure 6A). This proposes that SERCA2 is not essential for fat storage in mouse adipocytes, and that loss of SERCA2 rather makes the white adipocytes less viable, as supported by the increased number of crown-like structures, the increased Caspase 3 activation ( Figures 3F and 4M) and the fewer, but larger, adipocytes under HFD-fed obese conditions (Figure 4KeL). Thus, both increased apoptosis and accelerated metabolism (as discussed above) provide plausible explanations for the reduced fat mass in these mice.

Conclusion
Collectively, our data emphasizes the importance of SERCA2 for preserved calcium homeostasis associated with proper ER and mitochondrial function in white and brown adipocytes. In obesity, reduction of adipocyte SERCA2 may thus contribute to adipose tissue dysfunction and the pathogenesis of metabolic disorders.

AUTHOR CONTRIBUTIONS
IWA and CSO conceived the idea, supervised this work, interpreted data, and wrote the manuscript. They have contributed equally to this study and are shared senior authors. MBT and EB performed experiments and interpreted data, made figures, and wrote parts of the manuscript. BC, EP, YW, SM and CJ performed experiments and assisted in data interpretation. CJ and PS provided human adipocyte samples and clinical data. PR and PS interpreted data and gave valuable feedback to this work. All authors have approved the final version of this manuscript.