PAR2 promotes impaired glucose uptake and insulin resistance in NAFLD through GLUT2 and Akt interference

Abstract

Background and Aims:

Insulin resistance and poor glycemic control are key drivers of the development of NAFLD and have recently been shown to be associated with fibrosis progression in NASH. However, the underlying mechanisms involving dysfunctional glucose metabolism and relationship with NAFLD/NASH progression remain poorly understood. We set out to determine whether protease-activated receptor 2 (PAR2), a sensor of extracellular inflammatory and coagulation proteases, links NAFLD and NASH with liver glucose metabolism.

Approach and Results:

Here, we demonstrate that hepatic expression of PAR2 increases in patients and mice with diabetes and NAFLD/NASH. Mechanistic studies using whole-body and liver-specific PAR2-knockout mice reveal that hepatic PAR2 plays an unexpected role in suppressing glucose internalization, glycogen storage, and insulin signaling through a bifurcating G_q-dependent mechanism. PAR2 activation downregulates the major glucose transporter of liver, GLUT2, through G_q-MAPK-FoxA3 and inhibits insulin-Akt signaling through G_q-calcium-CaMKK2 pathways. Therapeutic dosing with a liver-homing pepducin, PZ-235, blocked PAR2-G_q signaling and afforded significant improvements in glycemic indices and HbA1c levels in severely diabetic mice.

Conclusions:

This work provides evidence that PAR2 is a major regulator of liver glucose homeostasis and a potential target for the treatment of diabetes and NASH.

INTRODUCTION

The liver plays a critical role in regulating fat and sugar metabolism and systemic glucose homeostasis. Poor glucose and insulin control are hallmarks of diabetes, a condition comorbid with other increasingly prevalent metabolic disorders such as obesity and NAFLD.^[1] Individuals with uncontrolled type 1 diabetes mellitus (T1DM) or type 2 diabetes mellitus (T2DM) exhibit persistent hyperglycemia and impaired glycogen storage, which redirects glucose into pathways of denovo lipid synthesis and liver triglyceride accumulation leading to NAFLD.^[2,3] NAFLD and its more severe inflammatory-fibrotic counterpart, NASH, now represent the dominant forms of liver disease with an incidence increasing commensurately with the epidemic of T2DM and metabolic disease worldwide.^[1] Recent studies indicate that poor glycemic control is associated with severity of fibrosis in NASH and that optimizing glucose metabolism in diabetics might modify the risk of NASH severity over time.^[4] This is of import as NASH can progress to liver cirrhosis and the development of hepatocellular carcinoma with limited treatment options. The mechanisms by which NASH develops from NAFLD are still not well understood but appear to be coupled with the development of hepatic insulin resistance caused by liver steatosis.^[2,3]

An emerging new regulator of hepatic metabolism that links liver inflammation with sugar, lipid, and cholesterol metabolism is the G protein–coupled protease-activated receptor 2 (PAR2).^[5] As a cell surface sensor of proteases, PAR2 endows the cell with the ability to respond or overrespond to the rapidly changing proteolytic microenvironment that occurs during tissue remodeling and inflammation.^[6,7] PAR2 is activated by cleavage of its N-terminal extracellular domain to generate a tethered ligand that induces a transmembrane signal to intracellular G proteins.^[8] Tissue factor/Xa–initiated PAR2 activation on macrophages was implicated in the promotion of inflamed visceral adipose tissue and liver in nondiabetic obese rodent models.^[9,10] Hepatic PAR2 was more recently found to activate G_i-Jnk1/2-SREBP-1c pathways to suppress lipid catabolism and increase cholesterol levels in both the liver and blood in high-fat diet (HFD) models.^[5] For poorly understood reasons, whole-body knockout or pharmacologic inhibition of PAR2 had salutary effects on glycemic indices in these nondiabetic obesity models. Despite lower circulating glucose, glucose concentrations were actually elevated in the livers of Par2-deficient mice along with reduced pyruvate concentrations, consistent with suppression of hepatic glycolysis rather than suppression of gluconeogenesis.^[5] Here, we set out to determine the underlying mechanistic basis of how the protease receptor PAR2 unexpectedly impacts liver glucose uptake and utilization and whether PAR2 plays a direct role in liver insulin resistance in the setting of diabetes, NAFLD, and NASH.

MATERIALS AND METHODS

Patient selection and liver biopsies

We used the Duke University Health System NAFLD Clinical Database and Biorepository, an open enrolling database and biorepository, which prospectively collected clinical data and biospecimens of patients who underwent a liver biopsy for evaluation of chronic liver disease. This ancillary study and informed consent procedures were approved by the Duke Institutional Review Board. Liver biopsy slides from patients with histologic diagnosis of diabetes mellitus and NAFLD/NASH, along with non-NAFLD nondiabetic controls, were retrieved for immunohistochemistry staining for PAR2 and GLUT2. NAFLD was defined as the presence of >5% hepatic steatosis on liver biopsy and absence of histologic and serologic evidence for other chronic liver disease in a patient with risk factors for metabolic syndrome. NAFLD cases were randomly selected and stratified by liver fibrosis stage: F0–1, F2, F3–4 (F4 included both compensated and uncompensated cirrhosis), the primary predictor for clinical outcomes, while controlling for age (±5years), sex, and BMI (±3 points) between fibrosis groups. Demographic data (e.g., diabetes) were obtained within 6months of liver biopsy.

Experimental methods

Detailed methods are provided as Supplemental Materials and Methods in the Supporting Information.

Statistical analyses

Bar graphs shown in figures are average ± SEM except where noted. Mouse and hepatocyte data were analyzed by ANOVA followed by Dunnett or Tukey’s multiple comparisons tests for experiments having more than two cohorts. For two cohorts, two-tailed unpaired t test with α = 0.05 was used. Statistical analyses were performed using GraphPad Prism 6.0. Statistical significance was defined as *p < 0.05, **p< 0.01, ***p < 0.001, ****p < 0.0001.

RESULTS

Hepatic PAR2 expression is increased in patients with diabetes with NAFLD and liver fibrosis

We conducted a cross-sectional study of liver specimens from patients with diabetes with the diagnosis of NAFLD or NASH versus nondiabetic controls, to examine the relationship between hepatic PAR2 expression and progression of NAFLD in diabetics. The 75 liver biopsies were stained for PAR2 protein expression and scored for intensity (low vs. high, Figure S1A) in a blinded manner, with baseline clinical parameters shown in Table 1. The nondiabetic control patients had normal liver parenchyma with low levels of PAR2 staining (Figure 1A,B). Hepatic PAR2 protein expression increased in patients with diabetes with NAFLD, with the highest levels in patients with diabetes and NASH (p = 0.0002), including those with more severe fibrosis (Figure 1A, B).

TABLE 1.

Baseline characteristics of patients (n = 75) with liver biopsies

Characteristic	Patients with NAFLD/NASH (n = 55)	Control patients (n = 20)
Age, y	54±9	43±14
Range	31–75	20–63
Male, n	25 (45%)	13 (65%)
BMI, kg/m2	36.9±7.8	28.6±8.2
Range	24.1–62.7	13.7–54.6
Obesitya	46 (84%)	8 (40%)
Diabetes mellitus	55 (100%)	0 (0%)
Race, n
White	45 (82%)	15 (75%)
Black	9 (16%)	3 (15%)
Asian	0 (0%)	2 (10%)
American Indian	1 (2%)	0 (0%)
NASb (0–8)	4.4±1.4	0
Steatosis (0–3)	1.8±0.8	0
Inflammation (0–3)	1.3±0.6	0
Ballooning (0–2)	1.3±0.7	0
Fibrosis Score (0–4)	2.4±1.3	0

Note: Data represent mean±SD where indicated.

^aSubjects with BMI ≥30 kg/m².

^bNAS is the NAFLD activity score.

FIGURE 1.

PAR2 is upregulated in the livers of patients and mice with diabetes and NAFLD and boosts blood glucose levels in both nonobese and diabetic mice. (A) Liver biopsies were obtained from 55 patients with diabetes with the diagnosis of NAFLD (Fib 0–1) or NASH (Fib 2–4), along with 20 control nondiabetic patients with normal liver biopsies. Biopsies were stained with hPAR2-Ab (brown: 3,3′-diaminobenzidine/HRP) and scored in a blinded manner for low or high expression. Data analyzed by Cochran-Armitage test for trend. (B) Photomicrographs (400×) of liver biopsies from control, patients with NAFLD and diabetes, and patients with NASH and diabetes stained with hPAR2-Ab or Masson’s trichrome to show blue collagen matrix deposition and macrovesicular steatosis (white empty spaces), bar = 50 μm. (C) Par2^+/+ (wild type [WT]) C57BL/6 mice were fed either normal chow (NC) or high-fat diet (HFD) for 16weeks. Diabetes was induced with streptozotocin (STZ; 50mg/kg/day for 5days) at week 8 in cohorts of mice fed HFD. Quantitative PCR (ΔΔCT) of Par2 expression from liver obtained at the 16 week endpoint, normalized to Gapdh (liver). (D) Representative photomicrographs of liver sections from WT and Par^−/− mice after 16 weeks of NC, 60%HFD or 60%HFD/STZ, stained with mPAR2-Ab; bar = 50 μm. (E–H) NAFLD activity score (NAS), and individual components of the NAS score of mouse livers (n = 9–11) at week 16 were determined from H&E stained sections using the NASH CRN criteria. (I, K) The total weight and liver weight of each mouse was measured at the 16 week endpoint. (J) Blood glucose levels were obtained after 4h fast at week 14–16 after dietary challenge. (L) Plasma insulin levels following 4h fast were quantified the day before the 16 week endpoint. Data depict mean ± SE. In panels E–L, the comparisons used the Student t test between samples under the same dietary conditions; In panel C, multiple comparisons were made using the Dunnett’s test, ****p < 0.0001, ***p < 0.001, **p <0.01, *p < 0.05.

PAR2 deficiency reduces blood glucose levels independently of obesity or diabetic status

The relationship of hepatic PAR2 expression and glucose metabolism was explored by quantifying Par2 mRNA in mice under different dietary conditions, including induction of T1DM with streptozotocin (STZ). Hepatic Par2 mRNA was increased by 2.5-fold in HFD-fed obese mice and by 8-fold in nonobese diabetic HFD-STZ-treated mice, as compared to nonobese mice fed normal chow (NC) for 16weeks (Figure 1C). PAR2 protein expression increased in the HFD livers compared to NC livers with a further increase in the HFD-STZ livers, with little or no background staining in Par2-deficient (KO) livers (Figure 1D) or kidneys (Figure S1B). The PAR2 staining pattern was particularly evident in the periphery/plasma membrane of hepatocytes in severe NAFLD. Histopathological analyses of livers from HFD ± STZ mice using the criteria for the NAFLD activity score (NAS) documented significant protective effects of Par2 deficiency on the NAS along with reductions in all individual components including steatosis, inflammation, and hepatocyte ballooning (Figure 1E–H). In addition, Par2-deficient mice had significantly lower liver/body weight under all diets tested compared to wild-type (WT) mice, including under NC-fed or lean STZ conditions (Figure 1I). These data indicate that PAR2 contributes to significant liver pathology and steatosis in both obese and nonobese diabetic animals under diverse metabolic stresses.

Previous studies had shown that PAR2 deficiency resulted in suppression of fasting blood glucose levels in the diet-induced obesity HFD model that was ascribed variously to PAR2-expressing inflammatory cells present in visceral adipose tissue or in liver itself.^[9,10] Unexpectedly, we found that nonobese Par2^−/− animals fed an NC diet showed a more significant 26% relative reduction (p <0.01) in fasting blood glucose levels compared to the 12% drop in glucose (p <0.05) in Par2^−/− vs. WT mice fed an HFD for 16weeks (Figure 1J). Par2 deficiency conferred a significant protective effect against weight gain compared to WT mice fed either HFD or NC for 16 weeks (Figure 1K), but no significant effect on daily food consumption (Figure S2A). Although not functionally diabetic (fasting glucose <200mg/dl), the obese HFD-fed WT mice had a significant increase in fasting insulin levels that was completely suppressed to baseline in the Par2-deficient mice (Figure 1L). The nondiabetic Par2-deficient mice exhibited a 10%–15% suppressive effect on blood glucose profiles in glucose-tolerance and insulin-tolerance tests relative to WT mice fed HFD (Figure S2B, C), indicating slightly improved peripheral insulin sensitivity under nondiabetic conditions.

To determine if the protective effect of PAR2 deficiency on fasting blood glucose at baseline also occurred in severely diabetic (T1DM) animals, mice were treated with STZ midway through the HFD regimen. The STZ treatment caused a major 2.7-fold increase in fasting blood glucose in the WT mice that was significantly (p < 0.01) suppressed by Par2-deficiency (Figure 1J). Interestingly, Par2-deficiency caused a highly significant (p < 0.001) weight gain in these nonobese STZ mice rising to normal body weight, without increasing systemic insulin levels (Figure 1K, L). The severely diabetic WT STZ-animals showed essentially the same sensitivity to exogenous insulin as Par2-deficient mice (Figure S2D) despite the greatly reduced fasting glucose values found in the Par2-deficient condition. Together, these data suggested that the PAR2-dependent increase in fasting blood glucose levels in both nonobese and obese WT mice was due to hepatic glucose regulation and not to insulin sensitivity of peripheral tissues.

PAR2 suppresses hepatic glucose, glycogen and GLUT2 transporter expression

To determine whether the decrease in baseline blood glucose observed in Par2-deficient mice was due to an increase in hepatic glucose internalization, we measured the levels of glucose and glycogen in livers under the various dietary conditions. Hepatic glucose was increased by 1.6–2.5 fold in both nonobese and obese Par2-deficient mice (Figure 2A). Liver glycogen content was also elevated by 1.4–5.1 fold in the Par2-deficient mice as compared to WT with the greatest difference observed in diabetic (STZ) HFD-fed animals (Figure 2B). Periodic acid schiff (PAS) staining confirmed that increases in hepatocyte glycogen occurred in livers from Par2-deficient mice relative to WT (Figure 2C) regardless of concurrent development of steatosis.

FIGURE 2.

Genetic loss of PAR2 increases hepatic glycogen storage and expression of Glut2. (A, B) Livers were obtained from mice in Figure 1 at the 16 week endpoint. Hepatic glucose and glycogen was isolated and quantified using the KOH method. (C) Representative images of liver sections at 16 weeks stained for glycogen using PAS with hematoxylin counterstain, bar = 25 μm. (D,E) Quantitative PCR (ΔΔCT) of Glut2 and Glut10 expression in livers with values normalized to Gapdh. (F) Quantitative PCR (ΔΔCT) of Par2 expression in HepG2 cells transduced with either shPar2 to knockdown Par2 expression or shScr controls where values were normalized to Actb. (G) Relative mean fluorescence intensity (MFI) for surface expression of PAR2 on HepG2 cells transduced with either shScr or shPar2 was determined by FACS. (H) Change in glycogen content in HepG2 cells following 0 h or 1 h exposure to 5.5 mM glucose after 4 h preincubation period in serum free and glucose free media. Cells exposed to glucose were normalized to baseline glycogen (no serum, no glucose), n = 12. Glycogen was measured using the amyloglucosidase degradation method. Images on right depict representative PAS staining of glycogen in HepG2 cells following 1 h glucose exposure, bar = 10 μm. (I) Quantitative PCR of Glut2 and Glut10 expression in HepG2 cells determined as in panel D. (J) HepG2 hepatocytes (n = 4) were placed in serum free media for 2h with the indicated pharmacologic inhibitor before addition of trypsin (30 nM) or buffer (−). Quantitative PCR for Glut2 expression was then determined as in D. (K) Quantitative PCR for FoxA3 expression in HepG2 cells (n = 8) 2 h after 30 nM trypsin or buffer as in D. (L) FoxA3 promoter/hGlut2-luciferase reporter constructs pGL3-luciferase (Luci), FoxA3-p1-Luciferase, or FoxA3-p1/p2-Luciferase transiently transfected into HepG2 cells in 96-well plates. HepG2 cells were stimulated for 4 h with buffer (−) or PAR2 agonist trypsin (30 nM) and luciferase activity measured. (M) HepG2 cells were stimulated for4h with 30nm trypsin and glucose uptake quantified using a Glo 2DG (2-deoxyglucose)-coupled luciferase assay (Promega). (N) Summary of findings of effects of array of pharmacologic inhibitors in K on PAR2 suppression of Glut2 expression in hepatocytes. All inhibitors were used at 10 μm concentration except for GFX109203X (5 μm) and U73122 (20 μm) and AZ628 (20 μm). The Student t test was used to determine statistical significance, ***p <0.001, **p < 0.01, *p <0.05, ^#p = 0.06.

Because hepatic glucose and glycogen were increased in the Par2-deficient mice, we explored whether PAR2 affected the expression of hepatocyte glucose transporters. The major transporter of glucose into hepatocytes is glucose transporter 2 (GLUT2, SLC2A2) which controls hepatic glucose levels and glycogen storage.^[11,12] As shown in Figure 2D, nondiabetic or diabetic WT mice fed HFD had no significant change in Glut2 mRNA relative to NC-fed mice. However, Par2-deficient mice showed a 1.5–2-fold increase in mRNA and GLUT2 protein expression in liver relative to WT with both normal and HF diets (Figure 2D, Figure S3A, B). In addition, the minor glucose transporter GLUT10 is normally expressed at low levels in liver but increases in chronic liver disease in response to inflammatory stressors such as NAFLD.^[11] Opposite of Glut2, Glut10 levels increased in WT mice fed HFD, but were suppressed in the Par2-deficient animals (Figure 2E), consistent with the reduction in hepatic inflammation in the absence of PAR2 (Figure 1G). shRNA knockdown was used to suppress PAR2 in human HepG2 cells. Cells transduced with shPar2-lentivirus showed a significant decrease in Par2 mRNA and receptor surface expression relative to scrambled shScr control (Figure 2F,G). Following a fast, glycogen accumulation after exposure to glucose increased by 25% more (p < 0.05) in HepG2 cells transduced with shPar2 relative to shScr control (Figure 2H-left panel). The shPar2 cells also had a major increase in cellular PAS staining of glycogen granules (Figure 2H-right panel). Likewise, shPar2 knockdown conferred a significant 2.5-fold increase in Glut2 expression in the HepG2 cells with no effect on Glut10 (Figure 2I) consistent with the findings in the Par^−/− mouse livers.

PAR2 inhibits GLUT2 expression through a Gq-ERK1/2-FoxA3 pathway

To interrogate which effectors and signaling nodes downstream of PAR2 might be involved in controlling Glut2 levels, PAR2 was activated with agonist in the presence of a large array of inhibitors of putative downstream signaling components (Figure 2J, N). Treatment of HepG2 cells with the PAR2 agonist trypsin caused a significant lowering of Glut2 mRNA by 50% (p < 0.05). This inhibition of Glut2 mRNA levels by trypsin was reversed by the highly specific G_q inhibitor YM254890 (Figure 2J). Blockade of downstream G_q effectors PLCβ and PKC also completely reversed trypsin-mediated Glut2 suppression, however, the intracellular Ca²⁺ chelator BAPTA did not, thereby implying the involvement of a non-calcium-dependent PKC in this PAR2-dependent mechanism. As PKC is known to activate the mitogen-activated protein kinase (MAPK) Raf and MAPK-signaling may modulate Glut2 expression, we found that inhibition of Raf with AZ628, or MEK with UO126 also blocked trypsin-PAR2 inhibition of Glut2 mRNA (Figure 2J). In addition to G_q, PAR2 is an activator of G_i-βγ signaling in liver,^[5] however, inhibition of the G_i-βγ effectors PI3K or PTEN had no effect on Glut2 mRNA. Likewise, inhibition of other potential downstream mediators of PAR2-G_i inflammatory responses such as JNK or p38, showed little or no effect on PAR2 suppression of Glut2 mRNA (Figure 2J). As expected, stimulation of PAR2 with either trypsin or LIGRLO peptide agonist caused activation of phospho-ERK1/2 in the HepG cells (Figure S4A). The PAR2-pERK1/2 signal was suppressed by the G_q inhibitor, YM-254890, but not by the G_i inhibitor, pertussis toxin (Figure S4B).

The dominant transcriptional regulator of Glut2 expression in liver is the pioneer factor FoxA3 (HNF3γ).^[13] We found that activation of PAR2 with trypsin gave a significant 50% decrease in FoxA3 mRNA in the HepG2 cells (Figure 2K). Using the promoter identifier tools JASPR and Ensembl, dual putative FoxA3 consensus promoters (p1, p2) with the highest binding score were putatively identified within close proximity in a 940 bp region in the 5′ UTR of the human Glut2 gene. As shown in Figure 2L, baseline expression of FoxA3-p1 luciferase activity increased by 2-fold, and by 12-fold with the dual FoxA3-p1-p2 promoter from the 5′ UTR of Glut2 in the HepG2 cells. Addition of the PAR2 agonist trypsin caused suppression to baseline (p < 0.0001) for FoxA3-p1 luciferase activity, and by >50% (p < 0.01) for the luciferase activity under the control of the dual FoxA3-p1-p2 promoter of Glut2. Likewise, addition of the PAR2 agonist LIGRLO caused significant suppression of both FoxA3 and Glut2 expression (Figure S4C, D). To show direct inhibition of glucose uptake by PAR2, addition of the PAR2 agonist trypsin cause significant inhibition of glucose uptake over 4 h (Figure 2M). Together, these data indicate that PAR2 suppresses Glut2 expression in hepatocytes through a G_q-FoxA3-GLUT2 dependent mechanism shown schematically in Figure 2N.

Patients with diabetes and NAFLD/NASH with high PAR2 expression have lower levels of GLUT2 in their hepatocytes

Consistent with the data in HepG2 cells and mouse livers, examination of liver biopsies from 52 patients with diabete s and NAFLD/NASH scored for high versus low PAR2 expression in Figure 1A, showed a significant inverse relationship (p = 0.029) between PAR2 and GLUT2 expression (Figure 3A). High GLUT2 staining on the plasma membranes of the hepatocytes (Figure 3B) was more prominent in patient liver biopsies with low levels of PAR2.

FIGURE 3.

Patients with diabetes and NAFLD with high levels of hepatic PAR2 have lower GLUT2 expression. (A) Liver biopsies from 52 patients with the diagnosis of NAFLD/NASH and scored for low versus high hepatic PAR2 protein expression in Figure 1A, were stained for GLUT2 using a monoclonal GLUT-2 Ab and anti-mouse FITC secondary-IgG, along with DAPI. Fisher’s extact test of the medians was performed. (B) Representative photomicrographs of GLUT2 and DAPI staining of liver sections from patients with low PAR2 expression versus high PAR2 expression shown in B, bar = 10 μm.

PAR2 inhibits insulin activation of hepatic Akt in obese and nonobese mouse models

Insulin-mediated Akt phosphorylation at S473 by mTORc2 triggers a signaling cascade that results in activation of multiple metabolic processes in liver including storage of glucose via stimulation of glycogen synthase. Because PAR2 inhibited hepatic glycogen accumulation, we next examined whether PAR2 suppressed insulin-Akt signaling. HepG2 cells stimulated with insulin showed a significant 2-fold increase in Akt S473-phosphorylation whereas activation of PAR2 with trypsin completely suppressed the insulin phospho-Akt signal (Figure 4A, B). Knockdown of Par2 with shPar2 reversed the trypsin suppressive effect on the insulin-Akt signal in the HepG2 cell s a s compared to shScr controls. Likewise, pharmacologic inhibition of PAR2 with the PZ-235 pepducin inhibitor^[14,15] prevented trypsin attenuation of the insulin-Akt signal (Figure 4A–C).

FIGURE 4.

PAR2 suppresses insulin-stimulated Akt phosphorylation in hepatocytes. (A) Phospho-(S474)-Akt western blots from five independent experiments of HepG2 cells transduced with either shPar2 to knockdown Par2 expression or shScr and preincubated with or without the PAR2 pepducin inhibitor PZ-235 (20 μm). PAR2 was then activated with 30 nm trypsin or buffer (−), followed 10 min later by addition of 15 nm insulin and cell lysates harvested after 10 min. (B) Quantification of western blots in A, with phospho (p)-Akt signal intensity normalized to total Akt (n = 5). (C) Normalized comparison of changes in pAkt signal in B between cells pretreated with PAR2 agonist trypsin prior to insulin stimulation. (D) Western blots of p(S474)Akt, total Akt, pGSK3β,and total GSK3β from mouse livers following 16 weeks high-fat diet (HFD) with or without streptozotocin (STZ) to induce diabetes as in Figure 1. (E) Quantification of pAkt/Akt and pGSK3β/GSK3β from D using liver lysates derived from mice in Figure 1 (n = 9–11). Statistical comparisons used the Dunnett’s test for multiple comparisons in B, C and Student t test in E, **p <0.01, *p <0.05.

Activated Akt inhibits basal GSK3β suppression of glycogen synthase by phosphorylating GSK3β thereby allowing glucose to be polymerized and stored as glycogen. Therefore, we quantified the effect of Par2-deficiency on hepatic pAkt-pGSK3β using liver lysates from HFD-fed animals treated with or without the diabetes-inducing agent STZ. Par2-KO animaly exhibited significant increases in Akt phosphorylation under both metabolic conditions as compared to WT (Figure 4D,E). Similar effects were observed with significantly elevated pGSK3β in Par2-KO versus WT animals (Figure 4D,E). The observed increases in pAkt and pGSK3β in the livers of Par2-KO mice indicate that PAR2 suppresses Akt pathways that enhance glucose storage under dietary/metabolic stress.

PAR2-Gq evoked calcium signaling suppresses insulin activation of Akt

Calcium-dependent signaling pathways have been shown to modulate insulin signaling in liver cells.^[16] Trypsin and the PAR2 specific agonist, LIGRLO, caused a rapid and robust rise in intracellular calcium in HepG2 cells (Figure 5A, B). The PAR2 calcium signal was significantly attenuated (p < 0.001) by shPar2 knockdown or with the PAR2 pepducin, PZ-235 (P2pal-18S^[17]). To delineate the mechanism of how PAR2 inhibits insulin-dependent Akt signaling, we first examined the role of the major downstream effector of PAR2-G_q, namely PLCβ. The PAR2 calcium signal could be completely blocked by the G_q inhibitor, YM254890 and the PLCβ inhibitor, U73122 (Figure 5C, D). To directly demonstrate that intracellular calcium was responsible for the observed PAR2-dependent suppression of Akt signaling, calcium chelation with BAPTA caused a complete reversal of the trypsin agonist inhibitory effect on pS473-Akt, whereas cells treated with the calcium ionophore A23187 gave complete suppression of the insulin-stimulated pS473 Akt signal (Figure 5E). Likewise, the trypsin-induced suppression of insulin-Akt phosphorylation on S473 wa s reversed by the PLCβ inhibitor, U73122. A major effector of PLCβ signaling is the IP₃ receptor that causes release of intracellular Ca²⁺ from internal stores. Accordingly, the IP₃ receptor inhibitor, 2-APB, also blocked the ability of trypsin and LIGRLO-activated PAR2 to stimulate calcium release (Figure 5C, D) and reversed the PAR2 inhibition of the insulin-pS473 Akt signal (Figure 5E).

FIGURE 5.

PAR2 attenuates insulin-Akt signaling by a calcium-CaMKK2 dependent mechanism. (A–D) Intracellular calcium flux signals evoked by PAR2 agonists 30 nM trypsin or 1 μm LIGRLO in HepG2 cells transduced with shPar2 to suppress PAR2 expression or shScr ± 20 μm PZ-235 (A-B), or untransduced cells preincubated with 20 μm U73122, 50 μm 2-APB, 0.3 μm YM254890 (C, D). Each calcium trace represents the average signal with AUC normalized relative to the shScr groups (n = 4). Statistical comparisons were made using ANOVA with the Dunnett’s test, ****p <0.0001. (E) Representative western blots (n = 2–5) of the pAkt signals at Ser 473 and Thr 308 5 min after addition of PAR2 agonist 30 nM trypsin and/or 10 nm insulin in HepG2 cells preincubated for 45 min with 10 μm BAPTA-AM, 10 μm A23187, 20 μm U73122, 50 μm 2-APB, 25 μm Sto-609, or 10 μm FK-506. (F) Akt phosphorylation at Ser 473 and Thr 308 in response to trypsin and insulin treatment in HepG2 cells as in E with siRNA knockdown of β-Arrestin2 vs scrambled siRNA control. (G) Detection of CaMKK2 in immunoprecipitates of HepG2 cells following total Akt antibody or β-arrestin1/2 antibody pulldowns after 5 min stimulation with 30 nM trypsin or buffer (−). (H) Proposed mechanism of PAR2 inhibition of insulin-dependent Akt signaling in hepatocytes.

PAR2-CaMKK2 suppresses insulin activation of pS473-Akt

Intracellular calcium flux activates calmodulin which in turns activates both the phosphatase calcineurin, and the calmodulin-dependent kinase-kinase 2 (CaMKK2), the latter of which has been shown to enhance plasma glucose in HFD models in mice.^[18] We found that inhibition of CaMMK2 with Sto-609, but not calcineurin with FK-506, eliminated the ability of trypsin-activated PAR2 to inhibit the insulin-pS473-Akt response (Figure 5E). Maximal activation of Akt by insulin typically occurs through a two-step phosphorylation pathway: first the insulin receptor activates PI3K production of PIP3 which colocalizes the PH domain-containing Akt with the kinase PDK1 causing phosphorylation at T308 of Akt. In a second step, the pT308-Akt is fully activated by phosphorylation at S473 by mTORc2. Interestingly, activation of PAR2 with trypsin did not suppress the first insulin-dependent phosphorylation step at T308 by PDK1, nor did any of the above inhibitors of calcium signaling impact the PKD1 pT308-Akt signal (Figure 5E). Together, these data indicate that PAR2 negatively regulates insulin signaling at the level of pS473-Akt/mTORc2 in a calcium/CaMKK2-dependent manner.

Previous studies^[19] showed that PAR2 prevented CaMKK2 activation of AMPK through a β-Arrestin 2 (β-Arr2) scaffolding mechanism in fibroblasts. Knockdown of β-Arr2 with an siβ-Arr2 reversed the trypsin inhibition of the insulin-pS473-Akt signal but had no effect on T308 phosphorylation as compared to scrambled siScr (Figure 5F). Immunoprecipitation of trypsin-activated HepG2 cells with antibodies to either Akt or β-Arr1/2 co-precipitated CaMKK2 which was not appreciable in the basal (non-PAR2 activated) state (Figure 5G). These data indicate that CaMKK2 can associate with Akt and β-Arrestin together in a complex following activation of PAR2-G_q to inhibit activation of pAkt at S473 by mTORc2 a s depicted in the proposed mechanism of Figure 5H.

Loss or pharmacologic inhibition of PAR2 reduces hepatic insulin resistance in Db/Db mice

We next tested whether Par2-deficiency would improve insulin resistance in type 2 diabetic Db/Db mice. We crossed Par2^−/− mice with Db/Db mice to create a double deficient strain. Par2^−/− Db/Db mice had a striking 45% reduction in fasting blood glucose levels from 325mg/dl to 175mg/dl (p < 0.001) compared to Par2^+/+ Db/Db mice (Figure 6A), but with no suppression of weight gain (Figure 6B). However, Par2 deficiency gave a significant 32% (p < 0.0001) protective effect in reducing the high liver/body weight ratio that occurs in the obese Db/Db mice (Figure 6C). Similar to results above with the WT mice fed HFD (±STZ), there was a significant increase in both the transcript levels of Par2 mRNA and hepatocyte staining of PAR2 protein in Par2^+/+ Db/Db livers relative to WT mice fed NC, but with no effect on Par2 mRNA levels in visceral white adipose tissue (gWAT) (Figure 6D,E).

FIGURE 6.

PAR2 deficiency or pepducin inhibition improves hepatic insulin sensitivity and glycemic parameters in obese-diabetic (Db/Db) mice. (A) Blood glucose levels were quantified after a 4 h fast from 8 week old Db/Db;Par2^−/− (n = 11) and littermate control wild-type (wt) PAR2 Db/Db (n = 10) mice in the C57BL/6 background. (B, C) Mean body weight and liver weight of mice in panel A at 8 weeks. (D) Quantitative PCR (ΔΔCT) for Par2 expression in liver and fat from wt C57BL/6 versus Db/Db mice (n = 5) with values were normalized to Gapdh (Liver) or Eef2 (Fat). (E) Representative images of PAR2-Ab stained (brown: 3,3′-diaminobenzidine/HRP) liver sections from Db/Db and Db/Db;PAR2^−/− mice, bar = 50 μm. (F) Western blots of pAkt in liver and gWAT from fasted mice (n = 4–5) following 5 min stimulation with ip insulin or buffer (−). Quantification is shown on the right panel, fold activation was determined by normalizing pAkt to total Akt for each mouse. (G) Western blot of pAkt from fasted Db/Db mice following 5 min stimulation with ip insulin or buffer (−) receiving PZ-235 (sc 10mg/kg) PAR2 pepducin inhibitor or vehicle 1 h prior to insulin. (H) Blood glucose levels (normalized to baseline) of 6 week Db/Db mice receiving a single sc dose of PZ-235 (10 mg/kg) or vehicle over a 6 h fasting period (n = 9). (I) Average absolute drop in blood glucose levels (mg/dl) in Db/Db mice at the end of the 6 h fasting period shown in panel H. (J) Mean % glycated hemoglobin (HbA1c) levels in whole blood over a 3 week period in Db/Db mice receiving daily sc doses of PZ-235 (10mg/kg/day) or vehicle (n = 9). (K) Fasting plasma insulin levels from mice in J at the end of the 3 week period. Statistical comparisons made use of the Student t test (A–D, F), repeated measures ANOVA (H, J), ***p < 0.001, **p <0.01, *p < 0.05.

The Par2^+/+ Db/Db mice exhibited severe liver insulin resistance as evidenced by a paradoxical drop in pS473-Akt levels in liver below baseline after challenge of the mice with exogenous insulin (Figure 6F). Par2-deficiency reversed this insulin-resistant phenotype and instead caused a significant 3-fold increase (p < 0.05) in liver pAkt as compared to the Par2^+/+ Db/Db mice following insulin challenge (Figure 6F). In both genotypes there was no significant change in the level of phosphorylated Akt in visceral adipose tissue in response to exogenous insulin. Resensitization of the liver to appropriate insulin-Akt signaling occurred following acute treatment of the Par2^+/+ Db/Db mice with the PAR2 pepducin PZ-235 prior to administration of insulin (Figure 6G).

To test whether chronic pharmacologic inhibition of PAR2 could lower blood glucose levels inT2DM, PZ-235 was then administered once daily for 3weeks to Db/Db mice versus vehicle in cohorts that were randomized to have identical average initial plasma blood glucose levels. PZ-235 treated mice showed a greater reduction in blood glucose levels relative to vehicle with a significant 15% reduction (p < 0.01) in their relative level of glucose exposure over 6 h (Figure 6H). Treatment of Db/Db mice with PZ-235 led to a 3-fold greater absolute reduction (p < 0.05) in blood glucose levels compared to vehicle at the end of the 6h fast (Figure 6I). Daily administration of PZ-235 to Db/Db mice over the 3week period also significantly suppressed the relative increase in HbA1c by 29% (Δ2.0% vs. Δ2.8 %Hba1c, p < 0.01) (Figure 6J). Three weeks of PZ-235 treatment gave a 40% numerically lower mean reduction in fasting plasma insulin levels as compared to the vehicle Db/Db cohort (Figure 6K). Together, these data indicate that pharmacologic block de of PAR2 or deficiency greatly improves hepatic insulin resistance in severely diabetic mouse models.

Hepatocyte-specific PAR2-KO mouse

As PAR2 is expressed by many tissues and cell types and a conventional whole-body knockout hay been used to date, it is unclear how much of the KO phenotype or effects of PAR2 inhibition on steatosis, metabolism (glucose, lipid), and weight loss^[5,9,14] is mediated only at the level of the hepatocyte and how much is due to loss of PAR2 at other prominent sites of expression such as inflammatory cells in visceral adipose tissue or liver macrophages as suggested by bone marrow transplant methods.^[9,10] To address the in vivo role of hepatocyte PAR2 in the observed effects, we developed a hepatocyte-specific PAR2-KO mouse, PAR2^ΔHep, using a floxed (fl) PAR2 allele, with the knockout driven by the Alb-cre gene in a transgenic mouse strain (Figure 7A). As shown in Figure 7B,C, the floxed PAR2 and hepatocyte deletion of PAR2 in PAR2^ΔHep mice was confirmed at the genomic and mRNA level with 90% suppression of Par2 mRNA. The mouse PAR2 Ab readily detected PAR2 expression in the livers of WT Alb-cre⁻ littermate controls with very low protein staining in the PAR2^ΔHep mouse livers (Figure 7D), which in turn were significantly protected against NAFLD and individual components of NAS including steatosis, inflammation, and hepatocyte injury after HFD (Figure 7E).

FIGURE 7.

Liver-specific knockout of PAR2 suppresses NAFLD, reduces weight gain, and improves glycemic parameters in mice fed high-fat diet (HFD). (A,B) Strategy and confirmation of targeted mutation in the Par2 gene (F2rl1) in C57BL6 mice. Mice expressing Cre-recombinase under control of the Albumin promoter, Alb-Cre, were crossed with floxed F2rl1^tm1c mice in order to generate PAR2^ΔHep. Mouse livers (wild-type [WT] PAR2^fl/fl vs PAR2^ΔHep littermates) were analyzed for (C) Par2 mRNA, (D) PAR2 immunostaining, (E) NAFLD activity score (NAS), and individual components of the NAS score (n = 8–12) after 12weeks HFD. (F) Body weight and (G) daily food consumption in WT (fl/fl) vs PAR2^ΔHep mice over 12weeks HFD. (H) Liver weight after 12weeks of HFD. (I) Hepatic glycogen was isolated and quantified by the KOH method in representative livers. (J) Fasting plasma glucose in WT (fl/fl) vs PAR2^ΔHep mice. (K) Glucose tolerance test (GTT, 2g glucose ip/kg) in 7–9 WT vs PAR2^ΔHep mice at the 12 week endpoint. (L) Plasma insulin levels to mice in I were quantified prior to GTT. (M) Western blots of pS473 AKT, total AKT, and β-actin in representative livers from 12week HFD WT vs. PAR2^ΔHep mice with quantification of all samples in the lower panel. Data depict mean ± SE. Student t test or multiple comparisons were made using ANOVA, ****p <0.0001, ***p < 0.001, **p < 0.01, *p < 0.05.

The PAR2^ΔHep mice gained significantly less body weight over 3–12weeks on HFD while eating the same amount of food (Figure 7F,G) and had lower gross liver weight compared to WT (Figure 7H). As was observed in the global PAR2-KO, the PAR2^ΔHep mice had significantly more liver glycogen but lower levels of fasting plasma glucose (Figure 7I, J). In the glucose tolerance test, the mice with hepatic-specific deletion of PAR2 had a significantly lower glycemic index with lower plasma insulin levels than WT (Figure 7K,L). As was observed with the global PAR2-KO, the livers of the PAR2^ΔHep mice had major increases in pAKT-S473 (Figure 7M) and increases in GLUT2 levels (Figure S3A).

DISCUSSION

Hepatic insulin resistance sits at a critical nexus between poor glycemic control and diabetes that is correlated with the progression of NAFLD into the severe fibrotic-inflammatory condition of NASH.^[4] Given the dramatic increase in prevalence of NAFLD (55%) and NASH (37%) in the diabetes population in the US and worldwide in the past two decades, new drugs are urgently needed to simultaneously target both diabetes and NASH progression.^[1] In this study, we now link the protease receptor PAR2, with hepatic insulin resistance and glucose metabolism. We discovered that Par2-deficient mice exhibit reduced blood glucose levels regardless of obesity, diabetic status, or liver steatosis. The PAR2-dependent elevation of systemic glucose levels occurs through a two-pronged mechanism centering on G_q signaling in hepatocytes (Figure 8). Through one mechanism, PAR2 is responsible for suppressing FoxA3-dependent expression of the major liver glucose transporter, GLUT2, which is essential for hepatic glucose uptake.^[20] We found that PAR2 inhibition of FoxA3-dependent transcription of Glut2 occurs through a G_q-PKC-mediated activation of a Raf-MAPK cascade. PAR2 expression was inversely correlated with GLUT2 expression in the livers of patients with diabetes and NAFLD/NASH and in mice. Through a second mechanism, PAR2 actively promotes hepatic insulin resistance by reducing the level of insulin-mediated Akt-S473 phosphorylation in both HFD and diabetic disease states. As a result, Par2-deficient animals had increased levels of phospho-GSK3β with greatly enhanced hepatic glycogen content in both obese and diabetic animals. PAR2 inhibition of insulin-mediated Akt-S473 phosphorylation occurred through a G_q-calcium signaling mechanism requiring CaMKK2. A liver-homing pepducin, PZ-235,^[14,21] blocked PAR2-G_q signaling and afforded significant improvements in glycemic indices and HbA1c levels in severely diabetic, obese mouse models. These data point to an unexpectedly important role for the protease receptor PAR2 in modulating glucose homeostasis and hepatic insulin signaling under a range of metabolic conditions spanning from lean/non-obese, to obese, and severely diabetic. As PAR2 has also been shown to promote inflammation, fibrosis and hypercholesterolemia in NASH,^[5,14] therapeutic targeting of this receptor may afford broad beneficial effects in the diabetic NASH population.

FIGURE 8.

Mechanism of PAR2 in suppression of hepatic glucose storage and insulin signaling. Activation of the PAR2 protease receptor inhibits the expression of the major hepatic glucose transporter, Glut2, by G_q–MAPK suppression of FoxA3, and reduces the ability of insulin to fully activate phosphoT473-Akt via G_q-calcium-CaMKK2 interference. This results in hepatic insulin resistance, lowered glucose uptake and glycogen storage, with a commensurate increase in circulating glucose levels.

Interestingly, the major transporter of glucose in and out of the liver in humans and mice, namely GLUT2, is regulated independently of insulin. Instead, its expression is controlled by the hepatic pioneer transcription factor, FoxA3^[22] whose promoters we define here for the first time for the Glut2 gene. As a major control point of sugar utilization and blood glucose levels during feeding and fasting periods in liver, GLUT2 removes or enhances systemic glucose^[12] and facilitates glycogen synthesis during feeding and the efficient use of hexose sugars as a primary energy source during the postadsorptive period.^[23] For instance, following duodenal-jejunal bypass surgery in diabetic Goto-Kakizaki rats, there were significant increases in Glut2 mRNA and protein expression in liver along with a 50% drop in blood glucose, but with no effect on insulin levels.^[24] Conversely, patients with Fanconi-Bickel and genetic loss of function of their GLUT2 transporter, have high postprandial glucose with severe derangements of hepatic glucose.^[25] The liver-specific Glut2-KO mouse has a loss of glucose uptake but not release, hepatomegaly and an increase in liver weight relative to WT.^[12] This is opposite of the effects we found in both the global Par2-deficient and the liver-specific PAR2^ΔHep mice which have major increases in hepatic GLUT2 levels and glycogen, but smaller livers.

We discovered that PAR2-G_q activated a PKC-Raf–MEK-ERK1/2 signaling pathway that significantly attenuated FoxA3-Glut2-expression. Genetic knockdown of Par2 caused an increase in Glut2 with commensurate increases in glucose uptake and storage as glycogen granules in cell culture. Stimulation of PAR2 with trypsin decreased Glut2 mRNA levels and glucose uptake, and inhibited luciferase reporter expression driven by the two FoxA3 consensus promoters located in the 5′ UTR of Glut2. These data explain our in vivo observations that livers lacking PAR2 have significantly higher levels of hepatic glucose and glycogen accumulation along with increases in the level of GLUT2 protein and mRNA relative to WT mice.

The PAR2-dependent impairment in the conversion of glucose to glycogen by glycogen synthase further implied a role for PAR2 in the inhibition of insulin-Akt signaling that contributes to hepatic insulin resistance. The phosphoinositide-3-phosphate kinase (PI3K)-Akt pathway has been implicated as the key signaling integrator that controls the effects of insulin on liver glucose and fat metabolism in humans and other animals. Following localization to the plasma membrane through interaction of its PH domain with the PIP₃ lipid product of the PI3K reaction, Akt is first partially activated by phosphorylation of T308 by the phosphoinositide-dependent protein kinase 1 (PDK1). Full activation of Akt then requires phosphorylation on S473 by the mechanistic target of rapamycin complex 2 (mTORc2). Knockout of either PI3K or Akt1/2 isoforms, or pharmacologic inhibition results in hyperglycemia and severe insulin resistance.^[26] Once fully activated, pS473-Akt then induces glycogen synthesis by inhibiting GSK-3β, thereby stimulating hepatic storage of excess glucose under fed conditions. We found that PAR2-G_q activated a calcium-CaMKK2-dependent pathway that had no effect on the first PDK1-phosphorylation step at T308, but suppressed Akt phosphorylation at S473 by mTORc2 (Figure 8). Activation of PAR2 induced formation of a protein complex between CaMKK2, Akt and the PAR2-effector, β-Arr2. This PAR2 calcium-driven Akt-interfere nee mechanism mediated by CaMKK2 may help explain earlier studies that showed that either genetic deficiency^[27] or pharmacologiy inhibition of CaMKK2 activity^[18] regresses NAFLD and improves hepatic insulin signaling by an unknown mechanism.

We showed that PAR2 activation elicited a robust calcium signal through G_q-PLCβ and that PAR2 expression itself significantly increases in the livers of patients with diabetes with NAFLD/NASH and in mice with NAFLD and diabetes compared to controls. G_q is enriched by >15-fold in cholesterol-enriched microdomains in hepatocytes, and GLUT2 and the insulin receptor (IR) are significantly localized in these same microdomains with no GLUT2 detectable in cholesterol-poor microdomains.^[28] As PAR2 has also recently been linked to elevated systemic and hepatic cholesterol production,^[5] this may serve to locally organize PAR2-G_q inhibitory signaling in proximity to the two main controllers of glucose homeostasis in liver, namely GLUT2 for transport and the IR for glucose synthesis and metabolism.

Previous studies showed that PAR2 was involved in the development of obesity and NAFLD and ascribed these effects to PAR2 in inflammatory cells of visceral adipose tissue and liver.^[9,10] We also found that globally Par2-deficient mice are effectively protected against the development of obesity, NAFLD, and hyperglycemia under HFD-conditions. However, we showed that mice with specific deletion of PAR2 in hepatocytes, PAR2^ΔHep, also had significantly lower HFD-induced weight gain and NAFLD, and were protected against hyperglycemia. This would indicate that PAR2 in the hepatocytes was actually responsible for the majority of the observed systemic and liver-specific metabolic effects. Moreover, we discovered that systemic glucose, liver and body weight were also significantly lower in lean PAR2-KO animals with minimal visceral adipose tissue fed an NC diet, thereby uncoupling the effects of PAR2 on glucose metabolism from comorbid obesity. Conversely, genetic deficiency of PAR2 in severely obese, Db/Db mice had no effect on obesity but still had significantly increased hepatic insulin sensitivity. Moreover, the increases in PAR2 expression we observed in liver were not detected in visceral adipose tissue in the Db/Db strain. Together, these data are consistent with the notion that PAR2’s role in the development of hepatic insulin resistance and worsening glycemic indices is mainly due to direct effects of hepatocyte PAR2 on liver function.

A recent analysis of 713 patients with NAFLD/NASH showed that glycemic control predicted severity of hepatic fibrosis,^[4] and that a SNP (rs8192675; C allele) associated with lower expression of SLCA2 (Glut2) was highly prevalent in patients that derived more benefit from metformin anti-diabetic therapy in lowering HbA1c.^[29] We found that patients with diabetes and NAFLD/NASH with high expression levels of liver PAR2 had significantly lower GLUT2 in livers with a higher fibrotic score. It remains to be determined if blockade of PAR2 translates into superior glycemic control and better clinical outcomes in patients as indicated by significant drops in HbA1c in severely diabetic mice that received a PAR2 pepducin inhibitor. Given the multipronged beneficial effects of the liver-homing pepducin on glucose homeostasis, fibrosis, and steatosis, PZ-235 and its derivatives will be explored for therapeutic purposes^[21] and may offer certain advantages over current monotherapies, especially in the diabetic NASH population.

Abbreviations:

CaMKK2

calmodulin-dependent kinase-kinase 2

GLUT2

glucose transporter 2

HFD

high-fat diet

NAS

NAFLD activity score

NC

normal chow

PAR2

protease-activated receptor 2

STZ

streptozotocin

T1DM

type 1 diabetes mellitus

T2DM

type 2 diabetes mellitus

WT

wild type