Inhibition of Serine Protease Activity Protects Against High Fat Diet-Induced Inflammation and Insulin Resistance

Recent evidence suggests that enhanced protease-mediated inflammation may promote insulin resistance and result in diabetes. This study tested the hypothesis that serine protease plays a pivotal role in type 2 diabetes, and inhibition of serine protease activity prevents hyperglycemia in diabetic animals by modulating insulin signaling pathway. We conducted a single-center, cross-sectional study with 30 healthy controls and 57 patients with type 2 diabetes to compare plasma protease activities and inflammation marker between groups. Correlations of plasma total and serine protease activities with variables were calculated. In an in-vivo study, LDLR−/− mice were divided into normal chow diet, high-fat diet (HFD), and HFD with selective serine protease inhibition groups to examine the differences of obesity, blood glucose level, insulin resistance and serine protease activity among groups. Compared with controls, diabetic patients had significantly increased plasma total protease, serine protease activities, and also elevated inflammatory cytokines. Plasma serine protease activity was positively correlated with body mass index, hemoglobin A1c, homeostasis model assessment-insulin resistance index (HOMA-IR), tumor necrosis factor-α, and negatively with adiponectin concentration. In the animal study, administration of HFD progressively increased body weight, fasting glucose level, HOMA-IR, and upregulated serine protease activity. Furthermore, in-vivo serine protease inhibition significantly suppressed systemic inflammation, reduced fasting glucose level, and improved insulin resistance, and these effects probably mediated by modulating insulin receptor and cytokine expression in visceral adipose tissue. Our findings support the serine protease may play an important role in type 2 diabetes and suggest a rationale for a therapeutic strategy targeting serine protease for clinical prevention of type 2 diabetes.


Results
Baseline characteristics of human study. Table 1 summarizes the demographic and clinical characteristics of the human study subjects. As expected, diabetic patients had significantly higher fasting glucose and hemoglobin A1c (HbA1c) levels, waist circumference, and body mass index (BMI) than did patients without diabetes. Compared with the non-diabetic group, patients with diabetes had elevated levels of inflammatory markers, including IL-6 and high-sensitivity C-reactive protein (hs-CRP), and decreased adiponectin levels. There was no difference in lipid profiles, including total cholesterol, LDL cholesterol (LDL-c), and triglycerides between groups, except that patients with diabetes had lower high-density lipoprotein cholesterol (HDL-c) levels.

Serine protease inhibitor improved the decreased insulin signaling in visceral fat tissue.
HFD-fed mice showed a reduced ratio of protein levels of insulin receptors α and β in visceral adipose (VAD) tissue, but there was no significant effect in muscle or liver tissue (Fig. 3A-E). AEBSF treatment significantly reversed this effect in VAD tissue compared with HFD-fed mice (0.8 ± 0.2 vs. 0.6 ± 0.2, p < 0.05; Fig. 3A-E). Protein levels indicated that phosphorylation of pyruvate dehydrogenase kinase isozyme 1 (PDK1), Akt, and glycogen synthase kinase 3β (GSK3β) was downregulated in VAD tissue, but not in muscle or liver tissue. Unlike Akt that has to be phosphorylated before activation, GSK3β is constitutively active in resting cells, requiring phosphorylation by kinases such as AKT to inactivate it. AEBSF treatment resulted in a modest recovery of downregulation of PDK1, Akt, and upregulation of GSK3β in VAD tissue (all p < 0.05). Thus, insulin signaling was downregulated in VAD tissue in LDLR −/− mice fed a HFD, and AEBSF induced partial recovery from this suppression effect. www.nature.com/scientificreports www.nature.com/scientificreports/ The HFD promoted serine protease activation; IL-6, IL-10, and TNF-α proteins expression; and upregulation of the macrophage-related markers CD11b and F4/80 in VAD tissue ( Fig. 3F-H). AEBSF significantly downregulated these activities, except for IL-10 expression. Together, the HFD induced serine protease activity and triggered inflammatory cytokines secretion, and AEBSF attenuated these effects.

Discussion
The study results suggest that circulating serine protease activity is significantly elevated and associated with HBA1c, HOMA-IR, TNF-α, and adiponectin levels in patients with type 2 diabetes. Animal experiments in LDLR −/− mice confirmed these findings that administration of HFD produced progressive hyperglycemia and insulin resistance associated with upregulation of serine protease activity in plasma and VAD tissue. By showing a reduced ratio of protein levels of insulin receptors-α and β in VAD, we suggest that HFD will produce excessive proteolytic cleavage of the extracellular domain of insulin receptors in VAD (not in liver nor muscle tissue), thereby further advancing the knowledge in this field. It is worth noting that the treatment of the selective serine protease inhibitor AEBSF significantly attenuated the diabetogenic effects of HFD through the suppression of systemic inflammation, enhancement of adiponectin concentration, and improvement of insulin sensitivity in this animal model.
The pathogenesis of obesity-induced diabetes is multifactorial and complex. α1-Antitrypsin is known to play a pivotal role in blocking the early stage of proinflammatory cytokine production in mice with HFD-induced obesity 5 . Serine protease activation is likely involved in the inflammation of HFD-induced obesity. These studies clarified serine protease expression in subjects with diabetes mellitus and its action in diet-induced obesity. We showed that patients with diabetes have greater total and serine protease activities, in association with higher BMI and systemic inflammation. Moreover, plasma serine protease activity was positively correlated with BMI, HBA1c, fasting glucose levels, HOMA-IR, TNF-α, and also negatively with adiponectin concentration. However, the total protease activity showed no significant correlation with any of the above parameters. These findings provide novel evidence that serine protease activity was associated more significantly than total protease activity in patients with type 2 diabetes mellitus. www.nature.com/scientificreports www.nature.com/scientificreports/ Obesity has been established as an important factor in the etiology of insulin resistance and type 2 diabetes 13 . Tissue factor protease-activated receptor 2 signaling was shown to promote diet-induced obesity and insulin resistance 3 . In the present study, glucose level and HOMA-IR were significantly elevated in HFD-fed than in chow-fed LDLR −/− mice. These results suggest that the glucose challenge mediated insulin secretion in the circulation, but the increased insulin had only a minor effect on the promotion of glucose uptake by peripheral tissue. These phenomena were confirmed by the IPGTT, in which the glucose challenge increased blood glucose and insulin levels and reduced glucose clearance in the circulation. Besides, HFD-induced increases in systemic inflammation and serine protease activity were also observed in LDLR −/− mice. Inhibition of serine protease by AEBSF diminished the effects of the HFD on HOMA-IR, fasting glucose level, and blood glucose and insulin levels in the IPGTT, confirming that inhibition of serine protease reduced HFD-induced insulin resistance in LDLR −/− mice. Our data are in line with previous reports that reduced α1-antitrypsin levels were correlated negatively with BMI in obese humans 5 , and HFD-induced insulin resistance is attenuated by adipose tissue-derived serpins 9 .
Proteases derived from diverse sources are important mediators of HFD-induced obesity and diabetes. An HFD appears to cause protease and antiprotease imbalances in circulation 5 . Our study showed that inhibition of serine protease recovered HFD-related impaired insulin signaling in VAD tissue in LDLR −/− mice, which suggests specific suppression of serine protease activity may prevent the destruction of the extracellular domain www.nature.com/scientificreports www.nature.com/scientificreports/ of insulin receptor-α, and protects insulin receptors. These findings support previous reports about the soluble insulin-receptor ectodomain elevated in human obesity and type 2 diabetes mellitus 10,11 . Additionally, we extended previous findings showing that HFD produced excessive proteolytic cleavage of extracellular domain of insulin receptors in VAD not in liver nor muscle tissues. The excessive cleavage resulted in decreased detection of insulin receptor-α and impaired insulin signaling cascade in VAD in this study. Little is known about insulin receptor-α in animal study. Recently, it was reported that hyperinsulinemia would promote proteolytic cleavage of insulin receptor-α in rat hepatocyte culture 14 . They found increased proteolytic cleavage of insulin receptor in culture medium with higher insulin concentration 14 . HFD induced more complex changes in mice than simply hyperinsulinemic condition. Elevated serine protease activity in VAD, not liver, seems to play an important role in HFD-induced obesity in our study.
A variety of immune cells have been shown to participate in the complex intracellular communication network that organizes the chronic inflammatory response to obesity 4 . TNF-α and IL-6 are considered to be contributors to insulin resistance and diabetes mellitus development 15,16 . In this study, AEBSF attenuated increases in TNF-α and IL-6 levels in HFD-fed mice. Serine protease could be considered to be an active participant in obesity-induced inflammation and insulin resistance. These results are consistent with those of previous studies showing that polymorphonuclear neutrophil-derived serine proteases hydrolyze peptide bonds of TNF-α precursors and produce mature cytokine molecules 17 . Targeting inflammatory pathways could possibly be a component of the therapeutic strategies to prevent and control insulin resistance and diabetes.
Furthermore, adiponectin and leptin, an adipokine produced exclusively by adipocytes, are known to have critical effects on body weight regulation 18 . In this study, circulating adiponectin levels were significantly lower in participants with diabetes and HFD-fed LDLR −/− mice compared with controls. Selective inhibition of serine protease attenuated HFD-suppressed adiponectin levels. These findings are in agreement with the previous demonstration of improved insulin sensitivity in vaspin-suppressed obese mice fed high-fat, high-sucrose chow, as reflected by leptin and TNF-α downregulation and adiponectin upregulation 19 . Lower circulating adiponectin levels have been observed in patients with obesity and type 2 diabetes compared with lean subjects, and have been associated with insulin sensitivity, lipid and glucose metabolism, and inflammation 20,21 . AEBSF attenuated HFD-induced circulating IL-6 accumulation and recovered HFD-suppressed circulating adiponectin production. It also attenuated HFD-increased IL-6, CD11b, F4/80, and TNF-α expression and serine protease activity in VAD tissue. These data suggest that serine protease activity inhibition reduces adipose tissue inflammation, improving insulin sensitivity, in HFD-fed LDLR −/− mice. Endogenous serine protease inhibitor may exert insulin-sensitizing effects on white adipose tissue in various states of obesity 19 .
Previous studies have shown that administration of an endogenous serine protease inhibitor is correlated negatively with BMI and overcomes insulin resistance in obese mice 5,19 . However, details of the underlying mechanisms remained largely unknown. To clarify whether and how a serine protease inhibitor could modify insulin sensitivity, three kinases in the insulin signaling cascade (PDK1, Akt, and GSK3β) were examined in liver, muscle, and VAD tissues in HFD-fed LDLR −/− mice. Glucose and insulin levels were higher in HFD-fed mice after the IPGTT glucose challenge, and were reduced by AEBSF. In addition, AEBSF recovered the baseline phosphorylation status of all three kinases in VAD tissue, but not in liver or muscle tissue, in HFD-fed mice, indicating the selective tissue-specific improvement of insulin sensitivity. AEBSF mediated the recovery of insulin receptor-α/β, phosphorylation of PDK1, Akt, and GSK3β protein levels, contributing, at least partially, to resistance to HFD-induced insulin signaling loss in LDLR −/− mice. Our findings provide initial evidence suggesting that insulin receptors in VAD tissue are the main targets of serine protease inhibition, which reduces systemic inflammation and insulin resistance in experimental diabetes. VAD is known to play a critical pathophysiological  www.nature.com/scientificreports www.nature.com/scientificreports/ role in clinical metabolic syndrome and type 2 diabetes. Interestingly, a recent study reported that the amount of biologically insulin receptor active is regulated by the cleavage of its ectodomain, by the β-site amyloid precursor protein cleaving enzyme 1 (BACE1), in a glucose concentration-dependent manner 22 . They demonstrated that BACE1 (an aspartyl protease, not serine protease) regulated the cleavage of insulin receptor and insulin signaling mainly in db/db mouse livers 22 . In contrast, we confirmed upregulation of serine protease in VAD (not liver or muscle) play some role in HFD-induced insulin resistance in LDLR −/− mice in concordance with the previous studies 5, 19 . We speculated that different pathophysiological mechanisms may exist in different diabetic models (HFD-induced diabetes vs. db/db mice). Further studies are warranted to validate this speculation.
This study has several limitations. First, many proteases are active in blood 23 . Serine protease was shown to exert a crucial role in HFD-induced diabetes development, but interaction by other proteases in this study cannot be ruled out. Second, we investigated total serine protease activities in blood and VAD but not specific protease such as vaspin. The identity of the serine protease involved in obesity-associated diabetes development remains unclear; future studies should use activity-based probes to identify potential candidates. Third, the influence of serine hydrolase, which has a similar active site as serine protease 23 , was not investigated in the present study; its function may also be blocked by AEBSF. That is, the attenuation of serine hydrolase may have contributed to the improvement in lipid metabolism after AEBSF treatment in LDLR −/− mice 24 . Also, it would be important to understand if there is a combined action to generate the improvement in lipid metabolism after AEBSF treatment. In addition, we explored the insulin signaling pathway by investigating tissue insulin receptor-α and three kinases in the signaling cascade, PDK1, Akt, and GSK3β. The other important parameters (insulin receptor substrate 1 and glucose transporter type 4 in signaling cascade and plasma soluble insulin receptor-α) may provide further information about the roles of insulin receptor-α/β in insulin receptor function and signalling. The study is still preliminary and the nature of human and mouse plasma and VAD proteases need to be established. Furthermore, the potential side effect of AEBSF is unclear. In this study, we found significantly elevated levels of total bilirubin in AEBSF treatment group compared to chow diet and HFD group. www.nature.com/scientificreports www.nature.com/scientificreports/ In conclusion, serine protease activity is increased in clinical and experimental diabetes, which may be critical for type 2 diabetes development. As depicted in Fig. 4, the specific serine protease inhibitor AEBSF attenuated systemic inflammation, obesity, and insulin resistance in diabetic mice, probably by modulating insulin receptor and cytokine expression in VAD tissue. Our findings support the potential role of serine proteinase as the therapeutic target for clinical prevention of type 2 diabetes. Further clinical studies are required to verify this concept.

Methods
Human study: population, design, and measurements. We conducted a single-center, cross-sectional study with 87 subjects from Taipei Veterans General Hospital, Taipei, Taiwan. Study subjects were divided into two groups: type 2 diabetes mellitus and no diabetes. Diabetes mellitus was diagnosed by anti-hyperglycemic agent use or American Diabetes Association criteria 25 . All subjects apart from those reporting the prior diagnosis of diabetes underwent a standard 75-g oral glucose tolerance test, which included the measurement of fasting glucose. Subjects were excluded using the following criteria: (1) age < 18 years or >80 years, (2) receipt of insulin or insulin analog, (3) renal and/or liver function impairment, and (4) HbA1c > 8.0% (64 mmol/mol). The study protocol was approved by Institutional Review Board of Taipei Veterans General Hospital, Taipei, Taiwan (VGHIRB no. 201001004IC). All participants provided written informed consent before entering the study. In addition, all methods were performed in accordance with relevant guidelines and regulation.
Blood samples were collected after a ≥12-h fast and examined using biochemical tests to determine creatinine, alanine aminotransferase (ALT), glucose, HbA1c, and lipid profiles. To examine the role of circulating proteases in type 2 diabetes, plasma total and serine protease activities were determined. Protease-related markers, such as MMP-9 and MMP-13, were analyzed by Quantikine human ELISA kits (R&D Systems, Minneapolis, MN).

HFD-induced diabetes mouse model.
We used a murine model to clarify the role of serine protease in obesity and dysglycemia development. All experimental procedures and protocols involving animals were approved by the institutional animal care committee of National Yang-Ming University, Taipei, Taiwan [IACUC approval no. IACUC-1030306 (YM)], and were carried out in accordance with the Guidelines for the Care and Use of Laboratory Animals. Thirty 8-week-old male LDLR −/− mice (C57BL/6 J background) were obtained from The Jackson Laboratory (Sacramento, CA, USA), and housed under standard 12/12-h light/dark conditions. They were divided randomly into a chow diet control group, a HFD group (positive control), and a HFD + AEBSF group. The chow diet (Picolab Rodent Diet 20; 4 kcal/g, 2% cholesterol, 5.7% fat) and HFD (Teklad diet TD 88137; 4.5 kcal/g, 0.2% cholesterol, 21.2% fat; Harlan Tackle Co.) were administered for 10 weeks. An HFD is known to trigger the development of obesity and type 2 diabetes mellitus in LDLR −/− mice 12 . Mice in the HFD + AEBSF group received AEBSF (5 mg/kg/day; once a day, 5 days/week, intraperitoneal injection), and those in the HFD control group received PBS (100 μl/day, 5 days/week, intraperitoneal injection), for 10 weeks. AEBSF (Sigma, St. Louis, MO, USA), which inhibits serine protease activity by irreversible covalent bonding to the enzyme, was administered to clarify whether the pharmacological inhibition of serine protease could reverse obesity and improve insulin sensitivity in HFD-fed mice. Western blot analysis. Skeletal muscle, liver, and VAD tissue lysates were prepared by dissection and homogenized in buffer (25 mM HEPES, pH 7.4; 1% Nonidet P-40; 137 mM NaCl; 1 mM phenylmethylsulfonyl fluoride, 10 μg/ml aprotinin, 1 μg/ml pepstatin, 5 μg/ml leupeptin) using a PRO 200 homogenizer (PROScientific, Oxford, CT, USA). The samples were centrifuged at 14,000 × g for 20 min at 4 °C, and supernatant protein concentrations were determined using a protein assay kit (Bio-Rad Laboratories, Inc., Hercules, CA, USA). Supernatants (50 μg) were resolved by SDS-PAGE and subjected to immunoblotting. Protein abundance was detected with antibodies against insulin receptor-α, insulin receptor-β, Akt1, phospho-Akt1, PDK1, phospho-PDK1, GSK3β, phospho-GSK3β, and β-actin. All proteins were detected with HRP-conjugated secondary antibodies using Chemiluminescence Reagent Plus (PerkinElmer Life Science, Boston, MA, USA), and quantified with a densitometer. All proteins were normalized by β-actin, and specific protein phosphorylation was normalized by corresponding proteins. casein protease activity assay. To assess the ability of plasma proteases to digest large globular proteins, we utilized the Enzchek protease assay kit (Enzchek BODIPY, casein derivative, catalogue no. E-6638; Molecular Probes, Carlsbad, CA, USA; cleaved by metallo-, serine, acid, and sulfhydryl proteases), which consists of casein internally quenched with Texas FITC fluorophores (Ex/Em: 505/513 nm) reconstituted in a digestion buffer. Plasma samples were tested simultaneously for overall protease activity. Protease activity levels were determined from fluorescent intensity after peptide cleavage following 18 h incubation at 37 °C (SpectraMax Gemini XS; Molecular Devices, Sunnyvale, CA, USA).
Serine protease activity assay. Human and mouse plasma serine protease activities were measured with suc-AAPF-pNA, a highly specific synthetic substrate for serine protease, according to methods described previously 28 . Mouse plasma and VAD tissue were thawed by addition to 1 ml tissue lysis buffer (2 mM Tris, pH 7.2; 100 μM NaN 3 ; 50 mM NaCl; 0.1% IGEPAL). Tissue was homogenized and centrifuged at 12,000 × g. Supernatants were then centrifuged in 0.22-μm SPIN-X filter tubes. For the protease activity assay, suc-AAPF-pNA was dissolved in dimethyl sulfoxide, and 1 μl 40-mM solution was added to a 100-μl volume of 10 μl mouse plasma or 0.5 mg/ml lysate protein in tissue lysis buffer. Cleavage of suc-AAPF-pNA was followed by absorbance at 405 nm for 18 h incubation. Statistical analysis. Continuous variables were expressed as means ± standard deviations, and categorical variables were presented as frequencies and percentages. Differences between groups were analyzed using two-tailed Student's t tests. Subgroup comparisons of categorical variables were performed using the chi-squared test and Fisher exact test. Correlations of plasma total and serine protease activities with variables in the study groups were calculated using Pearson's product-moment correlation analysis. All tests were two sided and p < 0.05 was considered to be statistically significant. Statistical analyses were performed with SPSS software (version 17; SPSS Inc., Chicago, IL, USA).