Loss of bone morphogenetic protein-binding endothelial regulator causes insulin resistance

Accumulating evidence suggests that chronic inflammation of metabolic tissues plays a causal role in obesity-induced insulin resistance. Yet, how specific endothelial factors impact metabolic tissues remains undefined. Bone morphogenetic protein (BMP)–binding endothelial regulator (BMPER) adapts endothelial cells to inflammatory stress in diverse organ microenvironments. Here, we demonstrate that BMPER is a driver of insulin sensitivity. Both global and endothelial cell-specific inducible knockout of BMPER cause hyperinsulinemia, glucose intolerance and insulin resistance without increasing inflammation in metabolic tissues in mice. BMPER can directly activate insulin signaling, which requires its internalization and interaction with Niemann-Pick C1 (NPC1), an integral membrane protein that transports intracellular cholesterol. These results suggest that the endocrine function of the vascular endothelium maintains glucose homeostasis. Of potential translational significance, the delivery of BMPER recombinant protein or its overexpression alleviates insulin resistance and hyperglycemia in high-fat diet-fed mice and Leprdb/db (db/db) diabetic mice. We conclude that BMPER exhibits therapeutic potential for the treatment of diabetes.

Insulin resistance is a core defect of T2DM and associated with chronic inflammatory responses.
During chronic inflammation, macrophage plays a crucial role in obese-induced insulin resistance 3,4 .
After microphages are infiltrated into metabolic tissues such as obese adipose tissue, their secreted proinflammatory cytokines inhibit insulin signaling and result in insulin resistance. Another key player of chronic inflammation is vascular endothelium, which coordinates the action of immune cells through the continual adjustment in its own structure and function, including induction of adhesion molecules and pro-inflammatory cytokines, disruption of its barrier function and angiogenesis 5 . Even though endothelial dysfunction and inflammation have been observed during the development of T2DM 6,7 , it remains unknown whether endothelial cells (ECs) play a crucial role in the maintenance of glucose homeostasis and how its dysregulation contributes to insulin resistance and diabetes.
BMP pathway is known as an important regulator in the maintenance of endothelial integrity and the induction of vascular inflammatory responses 8 . BMPER (also called CV-2) binds BMPs and is an extracellular modulator of BMP signaling pathway 9 . We have discovered BMPER plays a pivotal role in BMP-mediated endothelial events and chronic inflammation in the vasculature [10][11][12][13][14][15] . However, the role of BMPER in obesity and insulin resistance has not been studied before.
In this study, we have made a surprising observation that BMPER regulates glucose homeostasis without affecting chronic inflammation. Instead, BMPER is a driver of insulin sensitivity through activating insulin signaling pathway in hepatocytes. This activation requires BMPER endocytosis and interaction with Niemann-Pick C1 (NPC1), an integral membrane protein that transports intracellular cholesterol 16,17 . Both whole-body and EC-specific BMPER inducible knockout (iKO) mice display similar glucose defects, including hyperinsulinemia, insulin resistance and glucose intolerance, suggesting vascular endothelium plays a causal role in glucose homeostasis through secreting metabolic regulators. More importantly, the delivery of BMPER recombinant protein or its adeno-associated viral (AAV) particles dramatically alleviates insulin resistance and hyperglycemia in diabetic mice. Our data reveal that BMPER is a protective regulator of glucose homeostasis and could become a potential therapeutic target for treating diabetes.

Results
Loss of BMPER results in glucose dysregulation. BMPER null mice die at birth due to lung defects 12 , which limits our understanding of BMPER's functions during pathological conditions at adulthood. To study the role of BMPER in glucose homeostasis, we generated BMPER flox/flox mice using CRISPR-cas9 gene editing (Fig. 1a). A BMPER inducible knockout KO (iKO) mouse model was created by crossing BMPER flox/flox and CAG-CreER +/transgenic mice (Fig. 1a), which allows temporal depletion of BMPER upon tamoxifen injection. BMPER depletion was confirmed in plasma and a variety of tissues of BMPER iKO compared to their littermate control (WT, BMPER flox/flox ; CAG-CreER -/-) mice (Fig. 1b).
Interestingly, BMPER iKO mice displayed hyperinsulinemia and higher homeostasis model assessment for insulin resistance (HOMA-IR) scores than WT mice at four months after tamoxifen injection (Fig. 1c-e, Supplementary Table 1). In addition, BMPER iKO mice were more glucose intolerant and insulin resistant ( Fig. 1f-g). To further explore insulin sensitivity, hyperinsulinemic-euglycemic clamp studies were performed. As shown in Fig. 1h-i, the amount of exogenous glucose required for maintaining euglycemia (glucose infusion rate, GIR) and the glucose disposal rate (GDR) were markedly lower in BMPER iKO mice than WT mice. Both the ability of insulin to suppress hepatic glucose production, which reflects hepatic insulin sensitivity, and the insulin-stimulated glucose uptake in skeletal muscle, heart and brown adipose tissue, which reflects insulin sensitivity in peripheral tissues, were significantly decreased in BMPER iKO mice (Fig. 1j-k). We also evaluated the expression of liver genes that reflect glucose output. BMPER depletion led to a significant induction of gluconeogenic enzymes, including G6Pase (glucose-6-phosphatase) and PEPCK (phosphoenolpyruvate carboxykinase), and a lipogenic transcription factor SREBP1 (sterol regulatory element-binding transcription factor 1; Supplementary   Fig. 1a). However, no change was observed with the glycolytic enzyme-glucokinase (GK; Supplementary Fig. 1a). Taken together, these results suggest BMPER depletion results in glucose dysregulation through regulating insulin sensitivity, gluconeogenesis and lipogenesis.
Inflammation is not observed in BMPER iKO mice. Given that vascular inflammation also disrupts function of metabolic tissues such as liver and WAT 18-20 , we examined whether BMPER depletion resulted in an inflammatory response in these tissues. In the liver of BMPER iKO and WT mice, the induction of the inflammatory cytokine IL1β (Supplementary Fig. 1b) was not changed while others-IL6 and TNFα were not detectable (data not shown). Their levels did not change in WAT either ( Supplementary Fig. 1c). These results suggest that inflammation does not contribute to the disruption of glucose homeostasis by BMPER depletion.
BMPs play a role in obesity through regulating adipogenesis and energy storage partitioning 21-24 , we asked whether BMPER also regulates body weight through modulating BMP signaling. Surprisingly, there was no body weight difference between BMPER iKO and WT mice (Supplementary Table 1).
Further CLAMS (comprehensive lab animal monitoring system) studies demonstrated a decrease in respiration exchange ratio (RER), a mild increase in physical activity and a non-significant trend of increase in food intake of BMPER iKO mice compared to WT mice ( Supplementary Fig. 2). In addition, we observed no significant changes with the expression of BMPs and BMP receptor 2 (BMPR2) in the liver of BMPER iKO mice compared to WT mice ( Supplementary Fig. 1d). It suggests, unlike BMPs, BMPER does not play a significant role in body weight control.
Given that BMPER iKO mice spontaneously developed hyperinsulinemia, insulin resistance and glucose intolerance without weight gain ( Fig. 1c- Fig. 3a-c). However, BMPER iKO mice exhibited more severe glucose intolerance and insulin resistance than WT mice ( Supplementary Fig. 3d-e). These observations suggest BMPER depletion exacerbates obesity-induced insulin resistance and glucose intolerance.
BMPER plasma level decreases in humans with metabolic syndrome. Since BMPER serum level was lower in BMPER iKO mice that spontaneously developed hyperinsulinemia and insulin resistance, we examined whether there is an association between BMPER level and insulin resistance. To this end, we observed that BMPER plasma level was reduced by ~50% in humans with metabolic syndrome (MS) 25 , a complication often coupled with insulin resistance and increased risk for T2DM 26 , compared to healthy individuals ( Fig. 2a, b). In addition, BMPER levels negatively correlated with body weight and plasma levels of insulin and triglycerides (TG; Fig. 2c-f), suggesting an association between decreased BMPER levels and conventional serum markers of insulin resistance. Similarly, a ~70% decrease of BMPER plasma level was observed in high-fat diet (HFD)-fed mice ( Fig. 2g-h). These results suggest that BMPER plasma level is negatively regulated by metabolic stress.
BMPER activates insulin signaling pathway through IR. To understand the underlying mechanism by which BMPER regulates insulin sensitivity and glucose tolerance, we focused on its impact on the canonical insulin signaling pathway that governs glucose output 27,28 . As expected, we observed BMPER increased activation of insulin signaling pathway in the liver, skeletal muscle and heart, indicated by the phosphorylation of insulin receptor substrate1 (IRS1) and AKT (Fig. 3a). To determine whether BMPER regulates insulin signaling through modulating BMP signaling pathway, we used the BMPR2 kinase-dead (BMPR2-KD) mutant construct that inhibited BMP2-induced Smad 1, 5, 8 phosphorylation in primary hepatocytes ( Supplementary Fig. 6a). However, BMPR2-KD mutant failed to inhibit BMPER-promoted IRS1 phosphorylation (Fig. 3b), suggesting BMPR2 is not required for the activation of IRS1 by BMPER. On the other hand, IRS1 phosphorylation induced by BMPER or insulin was blocked in IR-depleted hepatocytes isolated from IR-iKO mice (Fig. 3c). Thus, our data suggest that IR, but not BMP signaling, is required for BMPER activation of insulin signaling.

BMPER-activated insulin signaling requires NPC1 and BMPER internalization.
To further understand how BMPER increased insulin signaling, we tested whether BMPER binds IR. However, we failed to observe the direct interaction of BMPER and recombinant IR, indicating BMPER regulates insulin pathway via unknown mediators. By performing co-immunoprecipitation-combined with liquid chromatography-mass spectrometry proteomic techniques, we identified NPC1, an endosome and lysosome-residing membrane protein 16 , as an interacting protein of BMPER ( Supplementary Fig. 6b, Table 2). NPC1 loss of function mutants develop devastating lysosomal cholesterol accumulation 29,30 . NPC1 also influences insulin signaling in adipocytes and NPC disease mouse models 31,32 . However, its exact role in insulin action remains unclear. Therefore, we tested the notion that NPC1 is required for BMPER to regulate insulin signaling. We confirmed NPC1 and BMPER were in the same complex and BMPER co-localized with endogenous NPC1 mainly in vesicles inside of hepatocytes ( Fig. 4a-b). In addition, NPC1 knockdown markedly blocked BMPER, but not insulin, induced IRS1 phosphorylation ( Fig. 4c-d), suggesting NPC1 specifically mediates BMPER-promoted insulin signaling. By performing immunoprecipitation studies with membrane fractions purified from hepatocytes, we discovered that BMPER was associated with endogenous IR in vivo and NPC1 knockdown markedly blocked the BMPER and IR complex formation in hepatocytes (Fig. 3e, Supplementary Fig. 6c). These observations suggest NPC1 is required for BMPER to interact with IR and promote insulin signaling. Our previous studies demonstrate BMPER can be internalized into endothelial cells 15 . In hepatocytes, BMPER was also internalized and the internalized BMPER was degraded with a half-life at 38.45 min during a chase period with cold media (Supplementary Fig. 6d). Its internalization was inhibited by chlorpromazine (CPM), an inhibitor of clathrin-dependent endocytosis but not by methyl-β-cyclodextrin (MCD), an inhibitor of the caveolin-dependent endocytosis ( Fig. 3f-g). In addition, CPM abolished IRS1 phosphorylation induced by BMPER (Fig. 3h, Supplementary Fig. 6e), suggesting BMPER internalization is required for IRS1 activation.
BMPER supplementation improves glucose homeostasis in diabetic mice. Since BMPER depletion resulted in defective glucose metabolism, we hypothesized that BMPER supplementation improves glucose homeostasis in diabetic mice. We injected adeno-associated virus (AAV) of BMPER, or AAV-GFP as the control, into mice and then fed them with HFD for eight weeks. We observed AAV-BMPER injection recovered BMPER plasma level in HFD-fed mice back to the level seen in control chow-fed mice ( Fig. 5a) and had no impacts on appetite and weight gain (Supplementary Fig. 7a-b). However, plasma insulin and glucose levels were decreased in AAV-BMPER-injected mice compared to AAV- Supplementary Fig. 7c-d). Moreover, responses of glucose clearance and insulin sensitivity were improved in AAV-BMPER-injected mice following HFD feeding, although no difference was observed in CC-fed mice (Fig. 5d, Supplementary Fig. 7e-f).
We also tested whether BMPER alleviates insulin resistance in a Lepr db/db (db/db) monogenic mouse model of severe diabetes. BMPER plasma level was markedly lower than wild-type (C57BL/6) mice and AAV-BMPER injection recovered BMPER level in db/db mice (Fig. 5e). We observed no differences in food intake and body weight between AAV-BMPER and AAV-GFP-injected mice (Supplementary Fig. 8a-b). However, AAV-BMPER injection led to the normalization of hyperinsulinemia and hyperglycemia ( Fig. 5f-g, Supplementary Fig. 8c-d) and improved glucose clearance and insulin sensitivity in db/db mice (Fig. 5h). Polyuria and glucosuria are important hallmark clinical features of T2DM. We observed dramatic decreases in urinary glucose and albumin levels and urine volume in AAV-BMPER-injected db/db mice (Fig. 5i-k).
The liver-specific IR knockout mice display severe insulin resistance and defects in insulin clearance 33 . Our signaling studies suggest that IR is required for BMPER-promoted insulin signaling (Fig. 3c). Therefore, we tested whether AAV-BMPER could improve glucose response in IR-iKO mice that developed insulin resistance and glucose intolerance (Fig. 5l). Although AAV-BMPER improved glucose response in diet-induced obese (DIO) and db/db mice ( Fig. 5a-k), it failed to improve insulin resistance or glucose clearance in IR-iKO mice (Fig. 5l). It suggests that BMPER protects glucose homeostasis through an IR-dependent mechanism.
We also tested the feasibility of delivering recombinant BMPER protein in vivo. Following the injection of recombinant BMPER into HFD-fed mice, we detected significant normalization of hyperinsulinemia and improved glucose and insulin tolerance responses (Fig. 6). Taken together, our results suggest BMPER supplementation through AAV or recombinant protein delivery can improve T2DM.

Discussion
Taken together, our data unravel an unexpected role for BMPER in glucose homeostasis and indicate BMPER supplementation (i.e. gene therapy or recombinant protein delivery) can be a potential approach to treat T2DM and insulin resistance. Notably, the requirement of NPC1 and endocytosis for BMPER's action suggests a new perspective for transactivating insulin signaling. Although the exact role for NPC1 in BMPER IR complex formation and downstream insulin signaling events remains to be further characterized, our data strongly support that NPC1 can directly impact insulin signaling pathway through recruiting BMPER to IR. The endothelium, by virtue of its location, tightly controls metabolic exchange between the circulation and surrounding tissues. Endothelial dysfunction and inflammation have been observed during the development of T2DM and insulin resistance 6,7 , however, it remains a question how important endothelial cells are in the maintenance of metabolic homeostasis and how its dysregulation contributes to metabolic dyshomeostasis. Here, our data indicate BMPER is downregulated in humans with metabolic syndrome and DIO mice (Fig. 2), suggesting endothelial secretory function might also be disrupted during metabolic stress conditions and contribute to the progression of T2DM and insulin resistance.

Reagents and antibodies
All the chemicals and antibodies are listed in the main reagent table (Supplementary Table 3).

Mice
BMPER flox;flox mice were generated with standard CRISPR-cas9 gene editing technique. C57BL/6 embryos were microinjected with Cas9 protein, each guided RNA and a circular donor vector. Designed guide RNAs targeted intron 3 and intron 4 of the bmper gene for insertion of loxP sites flanking Exon 4 ( Fig. 1a). Guide RNAs were evaluated in vitro using purified Cas9 and PCR-amplified target region.
Functional guide RNAs were identified and used for model generation. Two founder animals were identified by specific PCR assays as positive for both 5' and 3' loxP site insertions. PCR assays spanning each loxP site followed by sequencing was also performed to confirm the loxP sequences.
Spanning PCR assays across both loxP sites was performed to confirm cis orientation. Further validation to rule out random integration of the donor vector was performed, including long PCR assays across 5' and 3' homology arms and PCR assays specific to the donor vector backbone. In addition, off-target analysis was performed on the original founders and no hits in any of the potential off-target sites screened were identified. Male chimeras were mated to wild-type C57BL/6 females to establish an isogenic line, and then backcrossed to C57BL/6 mice to obtain BMPER flox/flox mice on C57BL/6 background. All experiments were conducted on the resulting C57BL/6 background. Genomic DNA of BMPER flox;flox mice was isolated as described previously 12 and subjected for standard PCR assays to identify wild-type and targeted alleles. A PCR assay has been developed to genotype the pups. PCR We used the mating of BMPER flox;flox and CAG-CreER +/or BMPER flox;flox and Cdh5-CreER +/mice to generate the BMPER flox;flox ; CAG-CreER +/-(WT or iKO) mice or BMPER flox;flox ; Cdh5-CreER +/-(eWT or eKO) male mice for control chow (CC, 14.7% calories from fat) or high-fat diet (HFD, 60% calories from fat)-induced diabetes studies. In addition, IR-iKO and their littermate control IR-WT mice were generated from the mating of IR flox/flox and CAG-CreER +/mice followed by tamoxifen injection. Blood serum was obtained before and after they were fed with different diets. Primary hepatocytes were isolated from 5~8 weeks mice. For insulin signaling experiments, mice were injected with insulin at 0.5 U/kg (for CC fed mice) or 1.0 U/kg (for HFD and db/db mice) and blood glucose levels were monitored.
For adeno-associated virus (AAV)-transduced experiments, C57BL/6 and db/db mice at five weeks old were injected intravenously with the AAV-GFP or AAV-BMPER respectively at the titer at ~10 12 per 25 g mice. C57BL/6 mice were then fed CC or HFD and db/db mice were fed CC as indicated. Metabolic studies were performed with these mice and tissues and serum were collected for further analysis.

Subjects
Plasma samples were obtained from the study 25 that was approved by the Institutional Review Board of Baylor College of Medicine. Male and female volunteers were recruited at the Center for Cardiometabolic Disease Prevention at Baylor College of Medicine, Houston, Texas, or by advertisement.

Cell lines and primary cells
HEK293 cells were grown in DMEM supplemented with 10% FBS and antibiotics (100 U/ml penicillin, 68.6 mol/L streptomycin). Primary hepatocytes were isolated using collagenase perfusion method from C57BL/6 mice based on previously published paper with modifications 35 . Briefly, the liver from 5 ~8 weeks old mice was perfused through the inferior vena cava with 7 ml pre-warmed (37°C) liver perfusion medium at 1 ml/min. Then, the liver was constantly digested with 5 ml pre-warmed (37°C) collagenase (1mg/ml) at 1 ml/min. The perfused liver was chopped gently in DMEM and centrifuged at 72 g for 2 minutes. Cells were washed with 20 ml hepatocyte wash medium and purified with 20 ml 45% (v/v) Percoll solution at 72 g for 2 min. Cell viability and number were measured through Trypan blue exclusion and manual counting. Hepatocytes were cultured in William's E medium containing 5% FBS, 100 units/ ml of penicillin and streptomycin, and primary hepatocyte maintenance supplements at 37°C in a 5% CO 2 incubator. For NPC1 depleted cells, primary hepatocytes were transduced with lentiviral NPC1 shRNA or control virus for 3 days. For signaling experiment, hepatocytes were starved overnight and treated with insulin at 100 nM and BMPER at 100 nM for 1 hour in all the experiments.

Analysis of endocrine hormones and metabolites
For fasting blood serum, mice were fasted 4 hours or overnight and then blood was collected through submandibular bleeding using a lancet. Plasma and urinary values for glucose were measured with a mouse endocrine multiplex assay, and insulin and albumin with ELISA kits.

Glucose/ insulin tolerance tests
Glucose tolerance tests (GTTs) and Insulin tolerance tests (ITTs) were performed following our established protocol 36 . Briefly, GTTs were performed after an overnight (for CC and HFD-fed mice or db/db mice) fasting. Blood glucose was measured before and 15, 30, 60, 120 min after an i.p. glucose injection (1 g/kg) with a Freestyle Glucose Monitoring System (Abbott Laboratories). ITTs were performed after 4 hours fasting. Blood glucose was measured before and indicated time periods after an i.p. insulin injection.

Hyperinsulinemic-euglycemic clamp studies
The hyperinsulinemic-euglycemic clamp studies were performed in unrestrained mice using the insulin clamp technique (with constant insulin dose) in combination with [ 3 H] glucose and [ 14 C]2-deoxyglucose following our established protocol. In summary, mice were cannulated as described previously 37 and allowed to recover for 4 to 7 days before the clamp. After an overnight fasting, mice received a primed dose of [ 3 H]glucose (10 µCi) and then a constant rate intravenous infusion (0.1 µCi/min) of [ 3 H]glucose using a syringe infusion pump for 90 minutes. Blood samples were collected for the determination of basal glucose production. After 90 minutes, mice were primed with a bolus injection of insulin followed by a 2-hour continuous insulin infusion. Simultaneously, 25% glucose was infused at an adjusted rate to maintain the blood glucose level at 100-140 mg/dl. Blood glucose concentration was determined every 10 minutes by a glucometer. At the end of a 120-minute period, blood was collected for the measurements of hepatic glucose production and peripheral glucose disposal rates. For tissue specific uptake, we inject 2-deoxy-D-[1,- 14

Indirect calorimetry and metabolic cage studies
This assay was performed following our established protocol 36 with minor modification. Mice were individually housed in metabolic chambers maintained at 20-22ºC on a 12-h light/dark cycle. Metabolic measurements (oxygen consumption, food and water intake, locomotor activity) were recorded using an Oxymax/CLAMS (Columbus Instruments) open-circuit indirect calorimetry system. Food and water intake were also monitored. To determine the amount of urinary albumin excretion, individual mice were separated in a metabolic cage, where urine was collected and measured for 24 h. The urinary volume and its glucose and albumin concentration was determined.

Membrane fractionation
Membrane fractionation was performed following our published protocol with small modification 15 .
Briefly, tissue or cells were washed twice with cold PBS and suspended in buffer A (10 mM HEPES, pH 7.9; 10 mM KCl; 0.1 mM EDTA, 1 mM DTT, protease inhibitor). The mixture was homogenized and incubated on ice for 10 min, and then centrifuged at 4°C at 10,000 g for 3 min. The supernatant was collected and further centrifuged at 100,000 g at 4 °C for 1 hour. The cloudy pellet was washed and dissolved in the lysis buffer as the membrane fraction, which was used for further analysis.

Immunoblotting and immunoprecipitation
Immunoblotting experiments were performed based on our previously published paper 38 . Briefly, cells were harvested in lysis buffer (1% Triton X-100, 50 mM Tris, pH 7.4, 150 mM NaCl, 10% glycerol and protease and phosphatase inhibitors) and clarified by centrifugation at 8000 g for 5 min. Equal amounts of protein were loaded on the SDS-PAGE gel and subjected for Western blotting. For endogenous immunoprecipitation (IP) experiments, protein A/G Plus-agarose was used to pull down antibody complexes following our established methods 13

Gene expression analysis (real-time PCR)
Total RNAs were reversely transcribed into cDNAs with iScript TM cDNA synthesis kit. The specific pairs of primers used for the real-time PCR are listed in Supplementary Reaction mixtures were incubated at 95°C for 10 min followed by 55 cycles at 95°C for 10 sec and 60°C for 30 sec. β-Actin was used as the housekeeping gene.

Quantification and statistical analysis
No statistical methods were used to predetermine the sample size. No randomization was used as all mice used were genetically defined, inbred mice. Statistical data were drawn from normally distributed group with similar variance between groups. Data are shown as the mean ± SEM. Differences were analyzed with Student's t-test, 1-way or 2-way ANOVA and followed by a Fisher's LSD test unless otherwise specifically stated. Values of P≤0.05 were considered statistically significant.