Hepatokine ERAP1 Disturbs Skeletal Muscle Insulin Sensitivity Via Inhibiting USP33-Mediated ADRB2 Deubiquitination

The number of people with type 2 diabetes is increasing rapidly worldwide, and the disease is often associated with many metabolic complications, such as hypertension and hyperlipidemia (1). Insulin resistance is a hallmark in the pathology of type 2 diabetes, and chronic inflammation in metabolic organs, such as liver, adipose tissue, and skeletal muscle (SM), is a major cause of this pathological change (2). These metabolic organs can also interact with each other to regulate systemic insulin sensitivity (2); however, mechanisms underlying the crosstalk remain largely unknown. The liver, as a large endocrine organ, may release hundreds to thousands of proteins, termed “hepatokines,” into circulation to regulate tissue and systemic metabolism (3), such as fibroblast growth factor 21 (4), apolipoprotein J (5), glycoprotein nonmetastatic melanoma protein B (6), and selenoprotein P (7). Supposedly, these proteins are important for metabolic control in SM and adipose tissue (4–7). Moreover, hepatokines are regulated by inflammatory status (8), suggesting hepatokines are probably important in the pathological processes of inflammation-induced insulin resistance.

Endoplasmic reticulum aminopeptidase 1 (ERAP1) is a multifunctional enzyme belonging to the M1 family of zinc metallopeptidases (9,10). It has the potential to trim peptide antigens to optimal lengths for binding to MHC class I molecules in the endoplasmic reticulum (ER) (11) and plays a vital role in many autoimmune diseases (12–14). ERAP1 is abundantly expressed in many tissues, including the liver (15); reportedly, the lack of hepatic ERAP1 promotes hepatocellular carcinoma growth in immune-deficient mice and reduces the efficacy of adoptive T-cell therapy in mice (16). The subcellular localization of ERAP1 is considered to be in the ER (11); particularly, it is detected in the culture medium of primary hepatocytes (PHs), suggesting it is a hepatokine (3). Moreover, because ERAP1 can be released after an inflammatory stimulus (17), we hypothesized that it might be involved in mediating inflammation-induced insulin resistance.

Previous studies have shown that the SM is responsible for the majority of insulin-induced glucose uptake and disposal under normal conditions, and it is considered the most important organ for whole-body glucose homeostasis (18). β-Adrenergic receptors (ADRBs) are G protein–coupled receptors expressed in most tissues (19) and play important roles in regulating lipid and glucose metabolism (20). ADRB2 is the predominant isoform expressed in the SM (21). Stimulating ADRB2 can activate protein kinase A (PKA) and subsequently potentiate insulin-stimulated protein kinase B (AKT) activation, which is a key node in the insulin signaling pathway (22–24). ADRB2/PKA signaling is typically regulated by its agonists (e.g., formoterol and isoproterenol) or its antagonist ICI118,551 (25); however, whether there are other extracellular signals participating in its regulation is largely unknown. The aim of our study was to investigate the possible role of the hepatokine ERAP1 in the pathological processes of inflammation-induced insulin resistance and explore the underlying mechanisms.

Male or female C57BL/6J wild-type (WT) mice were obtained from Shanghai Laboratory Animal Co, Ltd (Shanghai, China) and leptin-receptor–mutated (db/db) mice from the Model Animal Research Center of Nanjing University (Nanjing, China). ERAP1 hybrid knockout (Erap1^+/−) mice were generated using CRISPR/Cas9 technology and the guide RNAs were targeted at introns 2 and 3 of Erap1 (Shanghai Model Organisms Center, Shanghai, China). Erap1^+/− mice were crossed to generate global Erap1-knockout (Erap1-KO) mice. For all experiments, littermates of the same sex were randomly assigned to experimental groups. For the high-fat diet (HFD) feeding experiment, 4-week-old WT mice were fed either a control diet or 60% HFD (Research Diets, New Brunswick, NJ) for 4 months. For the ERAP1 inhibitor amastatin (26) experiments, mice were intraperitoneally injected with amastatin (5 mg/kg) (27) or PBS for 10 days. For neutralizing antibody injection, mice received a single injection of ERAP1-neutralizing antibodies or IgG (1 mg/kg; R&D Systems, Minneapolis, MN) 30 min before experiments. Mice were maintained in a 12-h/12-h light/dark cycle at 23°C. All animal experiments were performed in accordance with the guidelines of the Institutional Animal Care and Use Committee of Shanghai Institute of Nutrition and Health, Chinese Academy of Sciences.

The cell lines 293T (ATCC, CRL-11268), Hep G2 (ATCC, HB-8065), C3H/10T1/2 (ATCC, CCL-226), and C2C12 (ATCC, CRL-1772) were purchased from Cell Bank of Shanghai Institute of Cell Biology, Chinese Academy of Sciences. Mouse PHs and non-PHs were prepared by collagenase perfusion (6,28). The differentiation of C2C12 myoblasts and C3H/10T1/2 preadipocytes was induced as previously described (29,30). To detect insulin signaling, cells were incubated with 100 nmol/L insulin for 20 min (29,31,32). For inhibiting endogenous ubiquitination, C2C12 myotubes were incubated with 50 μmol/L PYR41 for 24 h (33) (MedChemExpress, Shanghai, China). C2C12 myotubes were incubated with forskolin (34) (100 μmol/L, 24 h; Beyotime, Shanghai, China) or ICI 118,551 (25) (0.5, 1, and 3 μmol/L, 6 h; MedChemExpress).

Adenovirus expressing Erap1 (Ad-Erap1), Adrb2 (Ad-Adrb2), and ubiquitin-specific peptidase 33 (Usp33) were generated using the AdEasy Adenoviral Vector System (pAdEasy-EF1a-MCS-CMV-EGFP vector; Qbiogene, Irvine, CA). AAV8-TBG-GFP and AAV8-TBG-ERAP1 were purchased from Vigene Bioscience (Shandong, China). The negative control adenovirus or adenovirus expressing shRNA specific for mouse Erap1 (sequence: 5′-CCAGCACCATTATTATGCATAGTCA-3′) was generated using the BLOCK-iT Adenoviral RNAi Expression System (BLOCK-iT U6 and pAd/BLOCK-iT-DEST vectors; Invitrogen, Waltham, MA), according to the manufacturers’ instructions. Purified stocks of amplified recombinant adenoviruses were diluted in PBS and injected into the tail vein (1 × 10⁹ pfu/mouse) (28). For in-site adenovirus injection in SM, each mice was injected with Ad-GFP or Ad-Adrb2 at 5 μL/site for six sites (10⁹ pfu/mL).

Hep G2 cells were infected with Ad-GFP or Ad-Erap1 at a dose of 5 × 10⁸ pfu/well in 12-well plates (28). Conditioned medium was collected at 48 h and 72 h, and then centrifuged at 10,000 rpm for 5 min. Supernatants were filtered and stored at −80°C. For protein isolation from conditioned medium, 300 μL of conditioned medium was mixed with 450 μL of isopropyl alcohol, and then followed by manufacturer’s instructions from TRIzol (Invitrogen).

Glucose uptake was determined by measuring 2-deoxy-2-[(7-nitro-2,1,3-benzoxadiazol-4-yl) amino]-d-glucose (2-NBDG) uptake in C2C12 myotubes, PHs, and C3H10T1/2 adipocytes as previously described (35). Before treatment, cells were cultured in glucose- and serum-free DMEM overnight and then exposed to 2-NBDG (100 μmol/L) for 1 h at 37°C. Different concentrations of rmERAP1 and 100 nmol/L insulin were added at the last 20 min. Furthermore, cells were washed twice with PBS and suspended in 100 μL of PBS. Fluorescence (excitation: 485 nm; emission: 535 nm) was detected by a microplate reader (PerkinElmer, Waltham, MA) within a few minutes.

Blood glucose level was measured with a Glucometer Elite monitor (Optium Xceed) and serum insulin with the Insulin ELISA kit (ALPCO Diagnostic, Salem, NH) according to the manufacturers’ instructions. The glucose tolerance tests (GTTs), insulin tolerance tests (ITTs), HOMA of insulin resistance (HOMA-IR) index assessment, and in vivo insulin signaling analysis were conducted as previously described (28,32).

Serum triglyceride, aspartate aminotransferase, and alanine aminotransferase levels were measured using commercial kits (Shensuo), and serum tumor necrosis factor α or interleukin 6 levels were measured using ELISA kits (Proteintech, Rosemont, IL), according to the manufacturers’ instructions. Leucine aminopeptidase activity was measured as described previously (17).

Coimmunoprecipitation and protein ubiquitination assay were performed as previously described (36). For the ubiquitination assay, 293T cells were transfected with hemagglutinin (HA)-tagged ubiquitin together with the indicated plasmids. Twenty-four hours after transfection, the cells were treated with 10 μmol/L MG132 for an additional 6 h to block proteasomal degradation of ADRB2 before being lysed with denaturing lysis buffer. Pulled-down samples were subjected to immunoblotting using anti-HA antibodies to visualize polyubiquitinated ADRB2 levels. For endogenous ubiquitination assay, C2C212 myotubes were treated with 10 μmol/L MG132 for 6 h, incubated with rmERAP1 for 2 h, and stimulated with 10 μmol/L isoprenaline for 30 min.

Western blot was performed as previously described (32). The primary antibodies are listed in Supplementary Table 1.

Total RNA was extracted from mouse tissue samples or PHs using TRIzol (Invitrogen) as previously described (28). Consequently, quantitative PCR (qPCR) was conducted with the primers described in Supplementary Table 2. The resulting products were sequenced and analyzed using Illumina HiSeq 4000 system and the I-Sanger Cloud Platform (Shanghai Majorbio, Shanghai, China), respectively.

Statistical analysis was performed using GraphPad Prism, version 8.0 (GraphPad Software, San Diego, CA). All data are expressed as mean ± SEM. Significant differences were assessed by an unpaired 2-tailed Student t test (comparing two groups), 1-way ANOVA, or 2-way ANOVA followed by the Student–Newman–Keuls test (for multiple comparisons). For GTTs and ITTs, the data were analyzed for statistical significance using Student t test or 1-way ANOVA to compare the differences between or among different groups of mice at each time point examined. P < 0.05 was considered statistically significant.

The data sets generated during and/or analyzed during this study are available from the corresponding author upon reasonable request. The accession numbers of RNA-sequencing (RNA-seq) data are PRJNA635985 and PRJNA693992.

Hepatocytes account for ∼90% of liver volume, but there are still other cell types (37). To identify hepatocyte-specific factors that are dysregulated under inflammation, we isolated PHs from WT mice fed either a control diet or HFD for 16 weeks. At the time of PH isolation, HFD-fed mice exhibited increased body weight and epididymal and subcutaneous fat mass, higher blood glucose levels, higher serum interleukin 6 and tumor necrosis factor α levels compared with control diet–fed mice (Supplementary Fig. 1A–E). RNA-seq of PHs discovered 1,650 differentially expressed genes between the two groups (fold change >1.4; P < 0.01) (Fig. 1A and Supplementary Data 1). Among them, 242 genes were predicted to translate into secreted proteins (those with N-terminal signal peptides or annotated as secreted in UniProt by searching on keywords Gene Ontology Cellular Component or subcellular location). Of the 242 genes, 55 were upregulated (Fig. 1B and Supplementary Data 1).

We aimed to discover a hepatokine that was induced by inflammation and with unknown metabolic functions. Among the 55 upregulated genes, the metabolic function of three genes—Erap1, a disintegrin-like and metallopeptidase with thrombospondin type 1 motif 14 (Adamts14), and hyaluronan and proteoglycan link protein 4 (Hapln4)—have not been reported previously, to our knowledge. We then examined the expression and secretion of these three genes under inflammatory factors treatments. Although the RNA-seq results corroborated with those of RT-qPCR (Fig. 1C, Supplementary Fig. 1F), the hepatic and serum levels of ERAP1 and ADAMTS14, but not HAPLN4, were increased in HFD-fed mice (Fig. 1D, Supplementary Fig. 1G and H). We then examined the levels of ERAP1 and ADAMTS14 in PHs and culture medium upon incubation with interferon-γ (IFN-γ), which represents an inflammatory stimulus (17), and found only ERAP1 levels were elevated (Fig. 1E and F, Supplementary Fig. 1I).

Moreover, in several studies, researchers have discovered that ERAP1 is upregulated upon stimulation by inflammatory factors in macrophages (17), but no reported work appears to indicate the relationship between ADAMTS14 and inflammation. Therefore, we proceeded with ERAP1 but not ADAMTS14. Hepatic and serum ERAP1 levels also were increased in leptin-receptor-mutated (db/db) mice (Supplementary Fig. 1J), another commonly used insulin-resistance model (38). These results suggested that ERAP1 was induced by inflammatory factors, which might play an important role in inflammation-induced insulin resistance.

To test the function of ERAP1, we generated global Erap1-KO mice and observed that ERAP1 levels were completely absent in the liver (Supplementary Fig. 2A). The body weight, food intake, body composition, and serum AST, ALT, triglyceride, tumor necrosis factor α, and interleukin 6 levels were not different from those of WT mice (Supplementary Fig. 2B–I). Erap1-KO mice exhibited improved glucose metabolism with decreased fasting blood glucose levels and HOMA-IR index, as well as improved glucose clearance and insulin sensitivity, as evaluated by GTTs and ITTs, respectively (Supplementary Fig. 2J–N). These results suggested that global Erap1-KO mice had improved systemic insulin sensitivity.

To further explore the function of hepatic ERAP1, male HFD-fed mice were injected with negative control adenovirus or adenovirus expressing shRNA specific for mouse Erap1 (Ad-shErap1) via the tail vein. ERAP1 expression was knocked down in the livers of mice injected with Ad-shErap1 compared with that in control mice (Fig. 2A and B). The downregulation of ERAP1 blocked HFD-increased blood glucose and serum insulin levels under both fed and fasting conditions, as well as HOMA-IR index (Fig. 2C–E). GTTs and ITTs were all improved in mice injected with Ad-shErap1 compared with results in control mice (Fig. 2F and G).

Next, we conducted an in vivo insulin signaling assay by examining the insulin-stimulated p-AKT and phosphorylated glycogen synthase kinase 3β (p-GSK3β) levels, which are key nodes in insulin signaling (32). Surprisingly, we found that SM insulin signaling was significantly enhanced, as demonstrated by the increased p-AKT and p-GSK3β levels (Fig. 2H). Insulin signaling in the liver or epididymal white adipose tissue (eWAT) was not affected (Supplementary Fig. 3). Similar results were also seen in female HFD-fed mice (Supplementary Fig. 4). Furthermore, Ad-shErap1 also improved glucose metabolism in control diet–fed WT or in the db/db mice (Supplementary Figs. 5 and 6). These results suggested that liver-specific knockdown of Erap1 ameliorates SM insulin resistance.

To verify the role of hepatic ERAP1 in regulating physical glucose homeostasis, WT mice were injected with adenovirus expressing control green fluorescent protein (Ad-GFP) or Erap1 (Ad-Erap1). First, we observed that ERAP1 was overexpressed in the livers of mice injected with Ad-Erap1 (Fig. 3A and B, Supplementary Fig. 7A). The aforementioned basic parameters were not affected by Ad-Erap1 (Supplementary Fig. 7B–G). Ad-Erap1 increased blood glucose and serum insulin levels under both fed and fasting conditions, as well as HOMA-IR index (Fig. 3C–E). Furthermore, glucose tolerance and insulin tolerance were inhibited in mice injected with Ad-Erap1 compared with those in control mice (Fig. 3F and G).

Consistently, insulin signaling in the SM was significantly impaired, as demonstrated by decreased p-AKT and p-GSK3β levels compared with those in control mice (Fig. 3H). Insulin signaling also was not altered in the eWAT or livers (Supplementary Fig. 7H and I). The effect of hepatic ERAP1 were further confirmed by injecting WT mice with AAV8-TBG-ERAP1 or injecting Erap1-KO mice with Ad-Erap1. Interestingly, AAV8-TBG-ERAP1 mice showed similar phenotypes to Ad-Erap1 mice (Supplementary Fig. 8). Additionally, all of these improved parameters in Erap1-KO mice disappeared, including SM insulin sensitivity, when mice were injected with Ad-Erap1 (Fig. 4). These results suggest the important role of hepatic ERAP1 in regulating SM insulin sensitivity.

We next examined how altered ERAP1 levels in the liver affected SM insulin sensitivity. Because ERAP1 was detected in the culture medium of PHs, suggesting it is a hepatokine (3), we speculated that circulating ERAP1 might regulate SM insulin sensitivity. In addition to db/db and HFD-fed mice (Supplementary Fig. 1J and Fig. 1D), serum ERAP1 levels were also elevated in Ad-Erap1 mice, and decreased in Ad-shErap1 mice (Fig. 5A and B). Interestingly, serum ERAP1 levels in Erap1-KO mice were globally reversed by Ad-Erap1 (Fig. 5C). Liver-specific ERAP1 overexpression did not affect eWAT, brown adipose tissue, or SM ERAP1 levels (Supplementary Fig. 9A).

To confirm whether SM insulin sensitivity was disturbed by ERAP1, C2C12 myotubes were incubated with recombinant mouse ERAP1 (rmERAP1). Insulin-stimulated glucose uptake and insulin signaling were impaired in rmERAP1-treated C2C12 myotubes (Fig. 5D and E). In contrast, rmERAP1 had no influence on insulin signaling and insulin-stimulated glucose uptake in PHs or C3H/10T1/2 adipocytes (Supplementary Fig. 9B–E). Interestingly, ADRB2 antagonist ICI 118,551 (25) selectively disturbed C2C12 insulin signaling (Supplementary Fig. 10), which may explain the tissue-specific role of ERAP1 in regulating SM insulin sensitivity.

We also assessed the role of the secreted ERAP1, using high ERAP1 conditioned medium (CM-ERAP1) in the presence or absence of ERAP1-neutralizing antibodies. We first confirmed that Ad-Erap1 infection increased ERAP1 levels in HepG2 culture medium (Supplementary Fig. 11A). As expected, CM-ERAP1 impaired insulin signaling in C2C12 myotubes, and this effect was reversed by ERAP1-neutralizing antibodies (Fig. 5F and Supplementary Fig. 11B). Furthermore, we used ERAP1-neutralizing antibodies to block the activity of circulating ERAP1 in mice and found that a single injection of the ERAP1-neutralizing antibodies ameliorated insulin resistance in HFD-fed mice, as demonstrated by ITTs and SM insulin signaling, with no effect on the liver or eWAT (Fig. 5G and H, Supplementary Fig. 11C–E), further confirming an important role for serum ERAP1 in SM insulin resistance.

To investigate the cause of SM insulin resistance mediated by serum ERAP1, we performed RNA-seq of the SM in WT mice injected with Ad-GFP or Ad-Erap1. Kyoto Encyclopedia of Genes and Genomes enrichment analysis revealed that the cAMP pathway was enriched remarkably in Ad-Erap1 mice (Fig. 6A). Because ADRBs are important for regulating the cAMP pathway, which potentiates insulin-stimulated AKT activation, an important node of insulin signaling (22–24), we speculated that an ADRB-dependent pathway might be involved in ERAP1-regulated insulin signaling in SM. To test this possibility, we examined the expression of ADRB1, ADRB2, and ADRB3, the three subtypes of ADRBs (39), in control and Ad-Erap1 mice. The expression of ADRB1 and ADRB3 remained unchanged; interestingly, though, Adrb2 mRNA was increased, whereas its protein levels were downregulated, in the SM of Ad-Erap1 mice (Fig. 6B and Supplementary Fig. 12), suggesting inhibited ADRB2 signaling. Consistently, the levels of p-PKA substrate, which can be regulated by ADRB2 (23), were also reduced (Fig. 6B). Moreover, ADRB2 and p-PKA substrate expression was increased in Ad-shErap1 mice and decreased in C2C12 myotubes treated with rmERAP1 (Fig. 6C and D).

We then explored the possible reasons leading to the reduced ADRB2 expression by ERAP1. That Adrb2 mRNA was increased in the SM of Ad-Erap1 mice suggested that decreased ADRB2 protein levels involved a posttranslational regulation. The most characterized posttranslational regulation of ADRB2 is ubiquitination (40). To test this possibility, we first examined whether ERAP1 and ADRB2 could interact with each other. 293T cells were transfected with plasmids expressing HA-tagged ERAP1 or HA-vector, and/or MYC-tagged ADRB2 or MYC-vector, as indicated. Coimmunoprecipitation analysis revealed that ERAP1 could directly bind to ADRB2 (Fig. 7A). To investigate whether ERAP1-induced downregulation of ADRB2 was ubiquitin dependent, we treated C2C12 cells with or without the E1 ubiquitin–activating enzyme antagonist PYR41 (33) in the presence or absence of mERAP1. Consistently, rmERAP1-reduced ADRB2 expression was blocked by PYR41 (Fig. 7B). We then examined the ubiquitination levels of ADRB2 in 293T cells transfected with HA-tagged ubiquitin, in the presence or absence of MG132, as shown previously for ubiquitination analysis (36). Cells were transfected with FLAG-tagged ADRB2, incubated with or without rmERAP1, and stimulated with isoproterenol, the agonist for ADRB2 (25). ADRB2 ubiquitination levels were increased by rmERAP1 in 293T cells (Fig. 7C). Endogenous ADRB2 ubiquitination levels were also increased by rmERAP1 in C2C12 myotubes (Fig. 7D).

Two possibilities may contribute to the increased ADRB2 ubiquitination levels: ERAP1 reduced ADRB2 expression by increasing E3 ligase–dependent ubiquitination or decreasing deubiquitinating enzyme–dependent deubiquitination. Neural precursor cell expressed, developmentally downregulated 4 (NEDD4) is an important E3 ligase participating in the regulation of ADRB2 ubiquitination (40). Ubiquitin-specific peptidase (USP) 20 and USP33 are two deubiquitinating enzymes regulating ADRB2 deubiquitination (41). To determine which was possibly involved in the effect of ERAP1, we examined the influence of ERAP1 on the binding between ADRB2 and these enzymes in 293T cells and/or C2C12 myotubes. rmERAP1 inhibited the binding between ADRB2 and USP33, whereas it had no influence on the interaction of NEDD4 or USP20 with ADRB2 (Fig. 7E and F, Supplementary Fig. 13A and B). Furthermore, rmERAP1-induced ADRB2 reduction was reversed when C2C12 cells were infected with Ad-USP33 (Fig. 7G).

The role of ADRB2 signaling in ERAP1-induced SM insulin resistance was then confirmed by examining insulin signaling in SM-specific ADRB2 overexpression mice. WT mice were injected with Ad-GFP or Ad-Erap1 via the tail vein. Four days later, the soleus muscle in the right leg was in situ injected with Ad-GFP and the left side was injected with Ad-Adrb2. We found that overexpression of ADRB2 in the SM reversed insulin-stimulated p-AKT and p-GSK3β levels in liver-specific Erap1 overexpression mice (Fig. 7H). We also conducted some in vitro experiments: C2C12 myotubes were infected with Ad-Adrb2 or Ad-GFP or were stimulated with the adenylyl cyclase activator forskolin (34) in the presence or absence of rmERAP1. Overexpression of ADRB2 or treatment with forskolin reversed the suppressive effect of rmERAP1 on insulin signaling and p-PKA substrate levels in C2C12 myotubes (Supplementary Fig. 13C and D). These results suggested ADRB2 signaling played a vital role in ERAP1-induced SM insulin resistance.

Finally, ERAP1 inhibitors were used to verify the possibility that ERAP1 could be exploited as drug targets for insulin resistance. We assessed the effects of amastatin (26) on glucose metabolism in HFD-fed mice. Amastatin inhibited leucine aminopeptidase activity in HFD-fed mice (Fig. 8A). Moreover, we found that amastatin significantly ameliorated insulin resistance, as depicted by the corresponding changes in blood glucose and serum insulin levels, HOMA-IR index, as well as GTTs and ITTs (Fig. 8B–F).

Chronic inflammation–induced insulin resistance is one of the major causes of type 2 diabetes. Metabolic organs may release secreted proteins into circulation to regulate metabolism systemically and in specific tissues. Notably, some hepatokines are regulated by inflammatory status (8), suggesting that hepatokines are important in the pathological processes of inflammation-induced insulin resistance. We discovered that ERAP1 was as an inflammation-induced hepatokine, because its expression and secretion were increased in insulin-resistant mice and stimulated by IFN-γ in PHs. In support of our results, it has been shown that IFN-γ induces the expression of interferon regulatory factor 1, which is a transcription factor that increases the expression of Erap1 (42). However, authors of a previous study reported that ERAP1 levels are not changed in the conditioned medium of PHs from mice fed an HFD for 6 weeks (3); it could be possible that the duration of HFD was not long enough to cause significant inflammation in PHs, compared with 16 weeks in our study and other reports (43). Because ERAP1 has no ER retention signal, it is reported to accumulate in the ER by binding to ER-resident protein 44 (ERp44) (44). When the binding is interrupted or ERp44 is depleted, ERAP1 is released into extracellular space (44). We found that increased hepatocyte ERAP1 expression was associated with increased ERAP1 secretion. Possibly, there was not enough ERp44 to retain elevated ERAP1 levels in the ER. Because ERAP1 was also expressed in the SM and WAT, we speculated that ERAP1 might be also secreted by SM or WAT under specific conditions, for example, when ERAP1 was elevated in these tissues. However, this idea needs more exploration.

Reportedly, ERAP1-deficient mice have fewer type 1 regulatory T cells and develop skeletal and intestinal features of ankylosing spondylitis, an autoimmune disease characterized by spinal ankylosis, osteoporosis, and spinal inflammation (45). This finding suggested that KO of ERAP1 induced inflammation, which is a main cause of insulin resistance. However, in the present study, we found that ERAP1 KO mice exhibited improved insulin sensitivity, indicating the effects of knocking out ERAP1 in regulating glucose metabolism were immune or inflammation independent.

However, several questions remain unanswered. We noticed there was a distinct role of serum ERAP1 in regulating insulin signaling in different tissues; we found that ERAP1 only attenuated insulin signaling in the SM but had no significant effect in the eWAT or liver. This possibly was due to the preferential and tissue-specific expression of ADRB isoforms (46). ERAP1 had significant effects on the expression of ADRB2, the dominant isoform expressed in the SM; notably, eWAT mainly expressed ADRB3, whereas the liver mainly expressed ADRB1 and ADRB2 (46). On the other hand, the different roles of ADRB2 on insulin sensitivity in different tissues led to these differences (46). For the liver, increased ERAP1 levels inside the hepatocytes may have different effects on insulin signaling, compared with the circulating ERAP1, which might cause the different results of insulin sensitivity between liver and SM. However, these possibilities need additional study. Furthermore, we showed that ERAP1 caused SM insulin resistance by inhibiting the interaction between ADRB2 and USP33, which decreased ADRB2 deubiquitination levels and caused ADRB2 downregulation. We speculate that ERAP1 was bound to ADRB2 and changed its conformation, which disturbed the association of USP33 and ADRB2. However, this needs further investigation, as do the binding sites of ERAP1 and ADRB2. Furthermore, we noticed that ERAP1-neutralizing antibodies and the ERAP1 inhibitor amastatin improved HFD-induced insulin resistance, so we speculated that the active sites of ERAP1 were important for its interaction with ADRB2. However, this also needs more investigation.

Taken together, our findings demonstrate that proinflammatory factors induce the expression and secretion of the hepatokine ERAP1 that interacts with ADRB2 and inhibits USP33-mediated deubiquitination of ADRB2, thereby interrupting PKA/AKT signaling and causing SM insulin resistance (Fig. 8G). We appear to have discovered a new function of the hepatokine ERAP1 in regulating SM insulin sensitivity, which also provides valuable insights into the molecular mechanisms underlying SM insulin resistance. In addition, our results suggest that the inhibition of ERAP1 might provide a new therapeutic strategy for treating insulin resistance–related diseases, including type 2 diabetes. Finally, our results provide new insights into the regulation of ADRB2 expression, the key receptor mediating signals of sympathetic nervous system, as well as many physiological processes.

Funding. This work was supported by the National Natural Science Foundation (grants 91957207, 31830044, 81870592, 82170868, 81970731, 81770852, 81970742, and 82000764); The National Key R&D Program of China (grant 2018YFA0800600); Novo Nordisk-Chinese Academy of Sciences Research Fund (grant NNCAS-2008-10); and Natural Science Foundation of Shanghai “Science and Technology Innovation Action Plan” (grant 21ZR1475900).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. F.G. and Y.N. planned and supervised the experimental work and data analysis. Y.N. performed the experiments and wrote the manuscript. H.J., H.Y., F.W., and X.H. researched data and provided technical support. R.H. and S.C. provided research materials. B.P. and Y.S. provided research materials and contributed to discussion. Z.L. provided technical support. F.G. directed the project, contributed to discussion, and wrote, reviewed, and edited the manuscript. All authors revised and approved the manuscript. F.G. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.