Localized Immunosuppression With Tannic Acid Encapsulation Delays Islet Allograft and Autoimmune-Mediated Rejection

WDFY2 Potentiates Hepatic Insulin Sensitivity and Controls Endosomal Localization of the Insulin Receptor and IRS1/2


Abstract

Endosomes help activate the hepatic insulin-evoked Akt signaling pathway, but the underlying regulatory mechanisms are unclear. Previous studies have suggested that the endosome-located protein WD repeat and FYVE domain–containing 2 (WDFY2) might be involved in metabolic disorders, such as diabetes. Here, we generated Wdfy2 knockout (KO) mice and assessed the metabolic consequences. These KO mice exhibited systemic insulin resistance, with increased gluconeogenesis and suppressed glycogen accumulation in the liver. Mechanistically, we found that the insulin-stimulated activation of Akt2 and its substrates FoxO1 and GSK-3β is attenuated in the Wdfy2 KO liver and H2.35 hepatocytes, suggesting that WDFY2 acts as an important regulator of hepatic Akt2 signaling. We further found that WDFY2 interacts with the insulin receptor (INSR) via its WD1–4 domain and localizes the INSR to endosomes after insulin stimulation. This process ensures that the downstream insulin receptor substrates 1 and 2 (IRS1/2) can be recruited to the endosomal INSR. IRS1/2-INSR binding promotes IRS1/2 phosphorylation and subsequent activation, initiating downstream Akt2 signaling in the liver. Interestingly, adeno-associated viral WDFY2 delivery ameliorated metabolic defects in db/db mice. These findings demonstrate that WDFY2 activates insulin-evoked Akt2 signaling by controlling endosomal localization of the INSR and IRS1/2 in hepatocytes. This pathway might constitute a new potential target for diabetes prevention or treatment.

Introduction

The liver is a crucial insulin tissue target for maintaining normal glucose homeostasis: it produces glucose during fasting and stores postprandial glucose (1). In hepatocytes, insulin binds to the insulin receptor (INSR) to activate its intrinsic tyrosine kinase activity. Insulin receptor substrates (IRS) are then phosphorylated on tyrosine residues that serve as anchoring sites for the phosphatidylinositol 3-kinase (PI3K) p85 regulatory subunit (2). PI3K produces PIP3 and subsequent PI34P2, which recruit Akt and mediate its activation (3,4).

Akt2 is the most abundant Akt isoform in insulin-sensitive tissues (5). Once activated in hepatocytes, Akt2 functions in several distinct downstream pathways that modulate metabolism (6). Activated Akt2 phosphorylates FoxO1 to reduce its transcriptional activity on gluconeogenic enzymes (7) and glycogen synthase kinase-3β (GSK-3β) to inhibit glycogen synthesis (8). The endosome has important roles in activating the hepatic insulin-evoked PI3K-Akt pathway (9). In hepatocytes, INSR ligand binding results in INSR tyrosine kinase autophosphorylation and internalization to generate intracellular signaling endosomes; the INSR continues to signal at endosomal compartments, providing the molecular cue to activate downstream IRS proteins and class I PI3K (10–12). IRS1 and IRS2 levels increase by 100%–150% in isolated hepatic endosomes following insulin administration. Akt, especially Akt2, is also recruited to endosomes, where it exhibits higher specific enzymatic activity than on plasma membranes (13,14). These observations support that IRS binding to the INSR in endosomes is essential to initiate the insulin signaling pathway and regulate insulin-dependent energy homeostasis. The underlying mechanism of endosomal targeting of the INSR and IRS1/2, however, is elusive.

Endosomal proteins involved in the insulin signaling pathway have important roles in mediating cell metabolism (15–17). FYVE domain (domain identified in Fab1p, YOTB, Vac1p, and EEA1)–containing proteins bind with high specificity and avidity to phosphatidylinositol 3-phosphate (PI3P) (18). WDFY2 is a highly conserved FYVE domain–containing protein: its expression defines a distinct population of early endosomes found in close proximity to the plasma membrane (19,20). The WD repeat protein family has been implicated in various cellular processes (21). WDFY2 itself promotes adipocyte differentiation and insulin-stimulated glucose uptake, which suggests that WDFY2 might orchestrate the signaling pathways involved in metabolic disorders (22,23). No studies, however, have focused on WDFY2 function in the liver. Here, we investigated the role of WDFY2 in insulin-evoked glucose metabolism using a systemic Wdfy2 knockout (KO) mouse model.

Research Design and Methods

Antibodies and Reagents

An anti-WDFY2 rabbit polyclonal antiserum was raised against a conserved 17–amino acid peptide at the COOH terminus of murine and human WDFY2. The serum was recovered after boosting six times, and the antibody was affinity purified (Medical & Biological Laboratories Co. Ltd., Japan). For commercial antibodies, see Supplementary Table 1. Palmitic, stearic, oleic and linoleic acids, d-glucose, insulin, and puromycin were purchased from Sigma-Aldrich.

Generation of Wdfy2 KO Mice and Genotyping Strategy

A targeting vector was designed to flank Wdfy2 exon 3 with loxP sites. The fragment included a sequence upstream of exon 3 (502 base pairs [bp]), the exon 3–flanked sequence (875 bp), and a sequence downstream of exon 3 (497 bp) of Wdfy2 genomic DNA; it was ligated into the pNLF vector (Beijing Biocytogen Co., LTD, China), which has two loxP sites and an FRT-flanked neomycin (Neo) resistance cassette. The targeting vector covered 5.2 kb sequence upstream of Wdfy2 exon 3 and 7.4 kb sequence downstream of exon 3 and included a diphtheria toxin A negative selection marker; it was subcloned from a BAC clone (RP23-178J15; Invitrogen) from the C57BL/6J mouse genomic BAC library. After linearization, the targeting vector was electroporated into C57BL/6J embryonic stem (ES) cells (Beijing Biocytogen Co., LTD). G418-resistant embryonic stem clones were screened for homologous recombination by PCR, and targeted clones were confirmed by Southern blotting with the indicated probes (Beijing Biocytogen Co., LTD). Two positive clones were injected into BALB/c blastocysts and implanted into pseudopregnant females to generate chimeric mice; chimeric mice were successfully generated from one of the two clones. The chimeras were bred to C57BL/6J mice to obtain F1 mice carrying the recombined allele containing the floxed Neo allele, after which the F1 mice were mated with Cre-deleter mice (C57BL/6J background) to remove exon 3 and the Neo cassette. Obtained heterozygous KO mice were interbred to generate heterozygous KO mice, which were then interbred to yield homozygous Wdfy2 KO and wild-type (WT) mice (littermate controls).

Murine phenotypes were confirmed in both males and females at the specified age. Genotyping was performed by PCR using genomic DNA isolated from tails using the primers detailed in Supplementary Table 2.

Animal Models With Insulin Resistance and Dietary Intervention

Male db/db mice (3 months old) and lean control mice were purchased from the Nanjing Model Animal Research Center. Dietary intervention with a high-fat diet (HFD) (60% calories from fat; Research Diets) or chow diet (11.4% calories from fat; Beijing Vital River Laboratory Animal Technology Co., Ltd) was conducted for 12 weeks from 1 month of age in WT C57BL/6 mice. All animal protocols were approved by the Committee for Animal Research of Peking University, Beijing, China, and conformed to the Guide for the Care and Use of Laboratory Animals (National Institutes of Health [NIH] publication no. 86-23, revised 1985). All mice were maintained in a temperature-controlled barrier facility with a 12-h light/dark cycle and were given free access to food and water in the Center for Experimental Animals at Peking University.

Animal Experiments

Male mice were used in this study. Glucose tolerance tests (GTT), insulin tolerance tests (ITT), and pyruvate tolerance tests (PTT) were performed as previously described (16). The serum concentrations of metabolic factors were measured with an Ultrasensitive Mouse Insulin ELISA kit (Mercodia), a triglyceride assay kit (FUJIFILM Wako Pure Chemical Corporation), and a cholesterol assay kit (FUJIFILM Wako Pure Chemical Corporation). Hepatic glycogen was measured with a Liver/Muscle Glycogen Assay Kit (Nanjing Jiancheng Bioengineering Institute).

Histological Analysis

Liver tissue was isolated from euthanized mice, fixed in 4% paraformaldehyde overnight, embedded in paraffin, and cut into sections of 5-μm thickness. The sections were analyzed by hematoxylin-eosin or periodic acid Schiff (PAS) staining.

Plasmid Construction

The GFP-INSR (mouse) (MG51062-ACG) plasmid was purchased from Sino Biological. The Myc-WDFY2 plasmid was provided by Dr. Karin Moelling (Institute of Medical Virology, University of Zurich, Zurich, Switzerland). INSR (human) cDNA was amplified and cloned into a p3xFLAG-CMV-10 vector (Addgene). Full-length WDFY2 and fragments (WD1–4 domain, 1–195 aa; WD5–7 domain, 196–400 aa) were cloned into pGEX-6p1 or pET28a vectors (Addgene). The INSR cytoplasmic domains (962–1,382 aa) were cloned into the pGEX-6p1 vector. Retroviruses containing PITA-Flag-WDFY2 or a WD5–7 fragment were produced using the 293T packaging cell line (ATCC).

Cell Culture and DNA Transfection

HepG2 (human hepatic cancer cells), H2.35 (murine hepatocytes), and 293T (human renal cells) cell lines were purchased from ATCC and cultured according to ATCC guidelines. All cells were maintained in a 37°C incubator with a humidified, 5% (v/v) CO2 atmosphere. HepG2 and 293T cells were transfected with Lipofectamine 2000 according to the manufacturer’s instructions. H2.35 cells were transduced by a retrovirus system (22).

Generation of Wdfy2 KO Cell Lines

CRISPR/Cas9-based gene-editing Wdfy2 KO H2.35 cells were generated via the Lipofectamine 2000–mediated transfection of single guide RNA (sgRNA) constructs in an SpCas9-2A-Puro vector (Addgene) (24). The following Wdfy2 sgRNAs were designed using online software (http://crispr.mit.edu): sequence 1, 5′-GGGTGTCATCAGCGTCTCGG-3′; sequence 2, 5′-ATCCTGCTGCAGCGGGTCGA-3′.

Protein Extraction

Murine tissues were homogenized in radioimmunoprecipitation assay lysis buffer (50 mmol/L Tris-HCl, pH 7.4; 150 mmol/L NaCl; 1% NP-40; and 0.1% SDS) supplemented with a 1% protease inhibitor cocktail (Roche) and 1% phosphatase inhibitors (APPLYGEN). Whole cell extracts were isolated as previously described (25).

Subcellular Fractionation

Plasma membrane and endosomal fractions were isolated from cell lysates by discontinuous sucrose-gradient ultracentrifugation (26).

PI3K Assay

Extracted proteins were immunoprecipitated with the indicated antibodies as previously described (16). PI3K activity was determined using a PI3K ELISA kit (Echelon Biosciences) according to the manufacturer’s instructions.

Western Blotting and Immunoprecipitation

Western blotting and immunoprecipitation were performed as previously described (25).

Protein Purification and In Vitro Binding Assay

Protein purification and in vitro binding assay were performed as previously described (25).

Coomassie Brilliant Blue Staining and Mass Spectrometry

HepG2 cells were transfected with Flag-WDFY2 or an empty vector for 48 h. Anti-Flag immune affinity columns were prepared using an anti-Flag M2 affinity gel (Bimake), following the manufacturer’s instructions. The Flag peptide (0.2 mg/mL) (Bimake) was applied to the column to elute the Flag protein complex. Fractions of the bead volume were resolved by SDS-PAGE, and the gels were stained with Coomassie brilliant blue stain (Solarbio); the bands were excised for liquid chromatography–tandem mass spectrometry sequencing.

RNA Extraction and Real-time PCR

Total RNA was isolated using TRIzol reagent (APPLYGEN). cDNA was synthesized from 2 μg RNA using a Quantscript RT kit (TOYOBO). Relative gene expression was determined by real-time PCR (25) using the primers detailed in Supplementary Table 3.

Immunofluorescence

Immunofluorescence was performed as previously described (19).

Adeno-Associated Virus Construction for WDFY2 Overexpression

cDNAs encoding full-length human WDFY2 or the WD5–7 fragment were inserted into a pHBAAV2/9-TBG-GFP vector and then subcloned in Escherichia coli. For packaging of the adeno-associated virus (AAV), the AAV vectors were transfected into AAV-293 cells using Lipofectamine 2000. After propagation, the recombinant AAVs were purified on a column (Biomiga) and the titer was determined by end point assay (Hanbio Biotechnology Co., LTD).

Statistical Analyses

The data represent the means ± SD of at least three independent experiments. Statistical significance was calculated by two-tailed unpaired Student t test for two groups or one or two-way ANOVA followed by Tukey post hoc test for more than two groups. P < 0.05 was considered statistically significant.

Data and Resource Availability

The data sets generated and/or analyzed during the current study are available from the corresponding author upon reasonable request.

Results

Hepatic WDFY2 Expression Is Decreased by Saturated Fatty Acids and High Glucose Levels

To investigate the role of WDFY2 in metabolism, we first detected WDFY2 expression levels in insulin target tissues. WDFY2 was highly expressed in the liver and adipose tissues but expressed at lower levels in skeletal muscles (Fig. 1A and B). We thus aimed to determine whether physiological metabolites influence WDFY2 expression in the liver and visceral white adipose tissue (Vis. WAT). We fed 1-month-old C57BL/6 mice an HFD or a normal diet (chow) for 12 weeks and then detected WDFY2 protein and mRNA levels in the liver and Vis. WAT. WDFY2 expression was significantly decreased in the livers of mice with HFD-induced obesity (Fig. 1CE, left lanes). WDFY2 was also reduced in the Vis. WAT of mice with HFD-induced obesity but to a lesser extent than observed in the liver (Supplementary Fig. 1AC, left lanes). The same pattern of changes was detected in the livers and Vis. WAT of 3-month-old diabetic (db/db) mice (Fig. 1CE and Supplementary Fig. 1AC, right lanes). These results indicate that WDFY2 expression levels in the liver and Vis. WAT are decreased in metabolic disorder mice, but the level of this decrease is most pronounced in the liver.

Figure 1

Hepatic WDFY2 expression is decreased by saturated fatty acids and high glucose levels. Western blot (A) and real-time PCR (B) analysis of WDFY2 expression in peripheral insulin target tissues. The data are normalized to β-actin. BAT, brown adipose tissue; SM, skeletal muscle; Sub. WAT, subcutaneous white adipose tissue. (n = 3.) Western blots (C) and quantified WDFY2 expression levels (D) in livers isolated from mouse models of insulin resistance and metabolic disorders. The comparisons are made with respective age-matched and sex-matched controls (n = 6 for all groups). E: Real-time PCR analysis of Wdfy2 mRNA levels in mouse models of metabolic disorders. The data are normalized to β-actin (n = 6 for all groups). Western blots (F) and quantified WDFY2 expression levels (G) in H2.35 hepatocytes treated with 100 μmol/L palmitic, stearic, oleic, or linoleic acids for 24 h, respectively (n = 3). H: Real-time PCR analysis of Wdfy2 mRNA levels in H2.35 hepatocytes treated with free fatty acids. The data are normalized to β-actin (n = 3). Western blots (I) and quantified WDFY2 expression levels (J) in H2.35 hepatocytes treated with different concentrations of glucose for 24 h (n = 3). K: Real-time PCR analysis of Wdfy2 mRNA levels in H2.35 hepatocytes treated with increasing concentrations of glucose. The data are normalized to β-actin (n = 3). The data represent the means ± SD. P values were calculated by Student t test (CE) and one-way ANOVA with Tukey post hoc test (FH) (*P < 0.05, **P < 0.01, ***P < 0.001 vs. the respective control). N.S., not significant.

As the liver is critical for regulating the impact of insulin on metabolic homeostasis, we focused our subsequent studies on the liver. Mice with HFD-induced obesity and db/db mice exhibited elevated levels of serum metabolic factors, including glucose and nonesterified fatty acids (27). We queried whether these metabolites affected WDFY2 expression in the H2.35 hepatocyte cell line. We thus treated H2.35 cells with exogenous palmitic, stearic, oleic, and linoleic acids and then detected WDFY2 protein levels. Interestingly, only the saturated palmitic and stearic fatty acids reduced WDFY2 expression levels compared with untreated cells (Fig. 1FH). We then exposed H2.35 cells to different glucose concentrations and found that WDFY2 expression increased in response to low glucose but decreased in response to high glucose (Fig. 1IK). These observations demonstrate that WDFY2 is enriched in the liver and its expression is downregulated by high levels of glucose or saturated fatty acids. WDFY2 might, therefore, be involved in the pathogenesis of metabolic diseases, such as type 2 diabetes.

Loss of Wdfy2 Leads to Insulin Resistance and Glucose Intolerance in Mice

To determine whether WDFY2 is involved in the pathogenesis of metabolic diseases, we generated systemic WDFY2-deficient mice (Supplementary Fig. 2A and B). Crossing WDFY2 heterozygous mice produced litters with the expected Mendelian ratios and normal body size. Furthermore, Wdfy2 KO mice were viable and fertile, with no differences in body weight (Supplementary Fig. 2C). GTT, however, showed impaired glucose reduction in Wdfy2 KO compared with WT mice (Fig. 2A and B). Consistently, ITT showed significantly reduced insulin sensitivity in Wdfy2 KO than in WT mice (Fig. 2C and D). Moreover, Wdfy2 KO mice had significantly higher serum insulin levels than WT mice under fasting conditions (Fig. 2E). These changes were detectable in mice aged 3 and 8 months old, indicating that WDFY2 has a continuous influence on glucose metabolism (Fig. 2AE). These results demonstrate that loss of Wdfy2 at the whole body level leads to impaired insulin sensitivity and glucose intolerance.

Figure 2
Figure 2

Loss of Wdfy2 leads to insulin resistance and glucose intolerance in mice. GTT were performed in 3-month-old (A) and 8-month-old (B) WT and Wdfy2 KO mice at the indicated time points (left). The area under the curve (AUC) for GTT is shown (right) (n = 5–6 mice for each genotype). ITT were performed in 3-month-old (C) and 8-month-old (D) WT and Wdfy2 KO mice at the indicated time points (left). The area under the curve for ITT is shown (right) (n = 5 mice for each genotype). E: Serum insulin levels of 3-month-old and 8-month-old WT and Wdfy2 KO mice were measured after 6 h of fasting (n = 5–10 mice for each genotype). Serum triglycerides (F) or cholesterol levels (G) in 3-month-old and 8-month-old WT and Wdfy2 KO mice maintained under random fed conditions (n = 5 mice for each genotype). The data represent the means ± SD. P values were calculated by Student t test (*P < 0.05, **P < 0.01, ***P < 0.001 vs. the respective control). N.S., not significant.

Next, we performed a biochemical analysis of lipid metabolism. We detected no changes in serum triglyceride but a mild increase in serum cholesterol levels in 3-month-old Wdfy2 KO mice (Fig. 2F and G). However, we detected significantly increased serum triglyceride and cholesterol levels in 8-month-old Wdfy2 KO mice (Fig. 2F and G). WDFY2 thus has a slight impact on lipid metabolism in young mice that becomes more prominent with age.

WDFY2 Inhibits Gluconeogenesis but Potentiates Glycogen Synthesis in Liver

We next aimed to determine liver metabolic function in Wdfy2 KO mice. The liver contributes to postprandial circulating glucose by suppressing hepatic gluconeogenesis and increasing the rate of glycogen synthesis (28). Results from a PTT in 3-month-old Wdfy2 KO mice showed a significant increase in hepatic glucose output compared with WT mice (Fig. 3A). Consistently, loss of Wdfy2 markedly attenuated the suppressive effects of insulin on two gluconeogenic genes: glucose 6-phosphatase (G6pc) and phosphoenolpyruvate carboxykinase (Pck1) (Fig. 3B). We also observed reduced positivity to PAS staining in 3-month-old Wdfy2 KO liver sections (Fig. 3C), indicative of decreased glycogen deposits compared with WT mice. The results of an anthrone reaction confirmed these findings (Fig. 3D).

Figure 3
Figure 3

WDFY2 inhibits gluconeogenesis but potentiates glycogen synthesis. A: PTT were performed in 3-month-old WT and Wdfy2 KO mice at the indicated time points (left). The area under the curve (AUC) for the PTT is shown (right) (n = 5 mice for each genotype). B: Real-time PCR analysis of G6pc and Pck1 mRNA levels in WT and Wdfy2 KO livers isolated from mice fasted overnight (fasted) or fasted and then refed for 4 h (refed). The data are normalized to β-actin (n = 6 mice in each group). C: Representative histological liver sections from 3-month-old WT and Wdfy2 KO mice stained with PAS, evidencing carbohydrate deposits in hepatocytes. Scale bars, 50 μm (n = 8 mice for each genotype). D: Hepatic glycogen content, determined by anthrone reaction in livers obtained from 3-month-old mice in fasted and refed conditions (n = 8 mice in each group). The data represent the means ± SD. P values were calculated by Student t test (A) and two-way ANOVA with Tukey post hoc test (B and D) (*P < 0.05, **P < 0.01, ***P < 0.001 vs. the respective control).

Hepatic glucose metabolism disorders lead to lipid heterotopic deposition with age (27). Consistently, we observed that 8-month-old Wdfy2 KO mouse livers were heavier than those of WT mice of the same age (Supplementary Fig. 3A), and hematoxylin-eosin staining showed mild but obvious hepatic steatosis (Supplementary Fig. 3B). Accordingly, we found increased triglyceride and cholesterol content in the livers of Wdfy2 KO mice compared with WT mice (Supplementary Fig. 3C). These data show that WDFY2 helps regulate hepatic glucose metabolism by decreasing hepatic glucose production and increasing hepatic glycogen accumulation.

WDFY2 Activates Insulin-Evoked Akt Signaling in the Murine Liver and in H2.35 Hepatocytes

Given the important role of the PI3K-Akt axis in the insulin signaling pathway (3), we hypothesized that WDFY2 might be involved in activating this axis in hepatocytes. We first studied Akt phosphorylation in response to insulin in liver extracts from Wdfy2 KO and WT mice. Loss of WDFY2 attenuated insulin-stimulated Akt phosphorylation at Ser473 and Thr308 and reduced FoxO1 phosphorylation at Ser253 and GSK-3β phosphorylation at Ser9 compared with WT controls; total Akt, FoxO1, and GSK-3β expression remained unchanged (Fig. 4A and B). We observed the same patterns in phosphorylation in response to feeding (Supplementary Fig. 4A and B). We then generated Wdfy2 KO H2.35 hepatocytes and again detected inhibited kinase activation compared with WT cells (Fig. 4C and D), whereas Erk1/2 phosphorylation was unaffected (Supplementary Fig. 4C and D). WDFY2 thus potentiates the insulin-evoked activation of Akt and its substrates FoxO1 and GSK-3β in murine livers and hepatocytes.

Figure 4
Figure 4

WDFY2 specifically activates insulin-evoked endosomal Akt2 signaling. A and B: WT and Wdfy2 KO 3-month-old mice were fasted overnight and then injected intraperitoneally with either insulin (0.75 units/kg body wt) or saline for 15 min. Representative Western blots (A) and the associated quantifications (B) show phosphorylated (p-)Akt, FoxO1, and GSK-3β levels in liver lysates (n = 10 mice in each group). C and D: WT and Wdfy2 KO H2.35 hepatocytes were serum starved overnight and then treated with 10 nmol/L insulin for 15 min. Representative Western blots (C) and the associated quantifications (D) show phosphorylated Akt, FoxO1, and GSK-3β levels in hepatocyte lysates (n = 5). E and F: WT and Wdfy2 KO 3-month-old mice were fasted overnight and then injected intraperitoneally with either insulin (0.75 units/kg body wt) or saline for 15 min. Representative Western blots (E) and the associated quantifications (F) show phosphorylated Akt1 and Akt2 levels in the liver lysates (n = 8 mice in each group). G and H: WT and Wdfy2 KO H2.35 hepatocytes were serum starved overnight and then treated with 10 nmol/L insulin for 15 min. Representative Western blots (G) and the associated quantifications (H) show phosphorylated Akt1 and Akt2 levels in hepatocyte lysates (n = 5). I–L: Analysis of total and phosphorylated Akt1 and Akt2 levels in the endosomes (ENs) and plasma membrane (PM) from WT and Wdfy2 KO H2.35 hepatocyte lysates. Representative Western blots (I) and the associated quantifications (J–L) show phosphorylated Akt1 and Akt2 levels in the endosomes and plasma membrane fraction (n = 6). The data represent the means ± SD. P values were calculated by two-way ANOVA with Tukey post hoc test (**P < 0.01, ***P < 0.001 vs. the respective control). N.S., not significant.

WDFY2 Selectively Regulates Akt2 but Not Akt1 Signaling in Endosomes

We next queried whether other functions under Akt control, such as growth and proliferation, are affected by Wdfy2 loss. We assessed the activity of TSC2 and PRAS40—two downstream Akt effectors that control protein synthesis and are involved in cell growth (29)—and found that phosphorylation of these proteins was unaffected by Wdfy2 loss. These data suggest that cell growth–related substrates are not influenced by Wdfy2 (Supplementary Fig. 4C and D).

Both Akt1 and Akt2 are expressed in the liver, and Akt2 accounts for ∼85% of total liver Akt (30). Akt2 is the predominant Akt isoform to mediate insulin’s control of glucose metabolism. Akt2−/− mice exhibit a strong metabolic phenotype with type 2 diabetes–like symptoms that are not seen in Akt1−/− mice (31,32). We thus examined the effects of WDFY2 loss on isoform-specific Akt phosphorylation in Wdfy2 KO mice liver and Wdfy2 KO H2.35 hepatocytes. Only Akt2 phosphorylation levels were significantly suppressed in the insulin-stimulated Wdfy2 KO liver and H2.35 hepatocytes compared with WT models (Fig. 4EH). Of note, insulin-stimulated Akt2 phosphorylation was more dramatically reduced than total Akt phosphorylation in Wdfy2 KO models. Total Akt1 and Akt2 protein and mRNA levels remained unchanged (Supplementary Fig. 4G and H), and Akt2 expression was unaffected in Wdfy2 KO WAT (Supplementary Fig. 4I and J).

Because WDFY2 mainly localizes to the early endosome membrane by binding PI3P via its FYVE domain (18), we investigated the effects of WDFY2 on Akt subcellular distribution in WT and Wdfy2 KO H2.35 cells. Compared with Akt2 phosphorylation in Wdfy2 KO H2.35 cells, Akt2 in the endosomal fraction of WT cells was markedly phosphorylated at Ser473 in response to insulin (Fig. 4I and J). However, both WT and KO cells showed a similar level of elevated Akt2 phosphorylation in the plasma membrane in response to insulin treatment (Fig. 4I and K). Akt1 was exclusively phosphorylated in the plasma membrane after insulin stimulation, and Akt1 phosphorylation was unaffected by loss of Wdfy2 (Fig. 4I and L). WDFY2 thus selectively regulates Akt2 signaling in endosomes.

WDFY2 Affects Insulin-Evoked Signaling Events Upstream of Akt

WDFY2 colocalizes and interacts with Akt in 3T3-L1 adipocytes (23), but whether this interaction exists in hepatocytes is unknown. After transfecting H2.35 cells with Myc-tagged WDFY2, we found no detectable association between Myc-WDFY2 and Akt1 or Akt2 in hepatocytes (Supplementary Fig. 4K). WDFY2 might, therefore, indirectly regulate Akt phosphorylation in hepatocytes.

Insulin induces Akt activation by stimulating IRS1 and IRS2 tyrosine phosphorylation, which in turn triggers PI3K activation (33). We investigated whether WDFY2 regulates these upstream signaling events in hepatocytes. We found that the insulin-induced elevation of PI3K activity was markedly reduced in Wdfy2 KO H2.35 cells compared with WT cells (Fig. 5A). Although WDFY2 protein deficiency impaired IRS1 and IRS2 tyrosine phosphorylation in response to insulin, it had no effect on the insulin-stimulated tyrosine phosphorylation of INSR and total INSR, IRS1, and IRS2 protein expression in hepatocytes (Fig. 5B and C). Decreased tyrosine phosphorylation of IRS1/2 was also observed in Wdfy2 KO mice liver tissue (Supplementary Fig. 5A and B). We also investigated the effects of WDFY2 on subcellular IRS1/2 phosphorylation in WT and Wdfy2 KO H2.35 cells. We found that insulin-stimulated IRS1/2 tyrosine phosphorylation was decreased in the endosomal fraction, but not in the plasma membrane, of Wdfy2 KO cells compared with WT cells (Supplementary Fig. 5CG). These results suggest that WDFY2 promotes insulin signaling by acting downstream of the INSR along the PI3K-Akt signaling pathway.

Figure 5
Figure 5

WDFY2 interacts with the INSR via the WD1–4 fragment. A: WT and Wdfy2 KO H2.35 hepatocytes were serum starved overnight and treated with 10 nmol/L insulin for 10 min. PI3K activity associated with IRS1 or IRS2 is expressed as the fold change relative to the baseline levels in the WT group (n = 4). B and C: WT and Wdfy2 KO H2.35 hepatocytes were serum starved overnight and treated with 10 nmol/L insulin for 5 min. Representative Western blots (B) and the associated quantifications (C) show the insulin-induced tyrosine phosphorylation levels of the INSR, IRS1, and IRS2 in hepatocyte lysates (n = 3). D and E: HepG2 cells were transiently transfected with Myc-WDFY2 and Flag-INSR or empty vectors. The interaction between exogenous WDFY2 and INSR was analyzed by immunoprecipitation (IP) assay. F: The interaction between endogenous WDFY2 and the INSR in liver and hepatocytes was analyzed by IP assay. G and H: HepG2 cells were transiently transfected with Myc-WDFY2 and GFP-INSR. Colocalization of WDFY2 (red) with the INSR (green) appeared as yellow fluorescence in the merged images (G) and was quantified using Manders’ colocalization coefficient (H). n = 3 with 50 cells per condition. Scale bars, 20 μm. I(a): Schematic of full-length and truncated WDFY2 (FYVE: domain identified in Fab1p, YOTB, Vac1p, and EEA1). WD, tryptophan-aspartic acid dipeptide. I(b): Schematic of the INSR β subunit. The CD contains the juxtamembrane (JM), tyrosine kinase (TK), and CT (C-terminal) domains. FnIII, fibronectin domains. J: GST, GST-WDFY2, GST-WD1–4, or GST-WD5–7 was incubated with whole-cell lysates of H2.35 cells. All the samples were resolved by SDS-PAGE, followed by Coomassie brilliant blue (CBB) staining or immunoblotting (IB) with the indicated antibodies (*specific protein bands). K: GST or GST–INSR CD was incubated with H2.35 whole-cell lysates. All samples were resolved by SDS-PAGE, followed by Coomassie brilliant blue staining or Western blotting with the indicated antibodies (*specific protein bands). The data represent the means ± SD. P values were calculated by two-way ANOVA with Tukey post hoc test (AC) or Student t test (H) (*P < 0.05, **P < 0.01, ***P < 0.001 vs. the respective control). N.S., not significant.

WDFY2 Interacts With the INSR via a WD1–4 Fragment

We next examined whether WDFY2 interacts with insulin signaling molecules. By affinity purification and mass spectrometry, we found that the INSR was a major WDFY2-associated protein (Supplementary Fig. 5H). Exogenous and endogenous coimmunoprecipitation assays indicated that the INSR could interact with WDFY2 in the liver and in hepatocytes (Fig. 5DF). We also investigated whether WDFY2 and the INSR colocalized under basal and insulin-stimulation conditions. Under serum-starved conditions, we observed that the INSR was mainly located on the plasma membrane and showed only a weak level of colocalization with WDFY2. After insulin stimulation, the INSR internalized and distinctly colocalized with WDFY2 (Fig. 5G and H).

To determine whether the interaction between WDFY2 and INSR is direct, we performed a glutathione S-transferase (GST) pull-down assay. His-tagged INSR interacted with GST-tagged WDFY2 but not with GST alone (Supplementary Fig. 5I). WDFY2 contains seven WD40 repeats and an FYVE domain, and the deletion of single amino acids, larger fragments, or individual blades elicits no marked reduction in its binding capability with proteins (34,35). We therefore generated truncation mutants embracing larger portions of WDFY2 [Fig. 5I(a)] and found that the INSR interacted with the C-terminal deletion mutant GST-WD1–4 but not with the N-terminal deletion mutant GST-WD5–7 (Fig. 5J).

The INSR is a transmembrane protein, and the cytoplasmic domain of its β subunit can interact with several intracellular proteins (36). We next generated a construct consisting of the INSR cytoplasmic domain (INSR CD), which included part of the juxtamembrane segment, the entire tyrosine kinase domain, and the C-terminal tail [Fig. 5I(b)]. WDFY2 interacted with the INSR CD peptide (Fig. 5K). Binding assays with purified proteins (Supplementary Fig. 5J) also showed that only full-length WDFY2 could interact with the endogenous INSR (Supplementary Fig. 5K). These data support an intracellular association between WDFY2 and the INSR and suggest that WDFY2 can directly interact with the INSR via the WD1–4 fragment.

WDFY2 Interacts With the INSR and Localizes It in Early Endosomes to Recruit and Activate IRS1/2

We next performed coimmunoprecipitation experiments to explore whether WDFY2 interacts with direct INSR substrates, IRS1/2. Although IRS1/IRS2 and WDFY2 did not interact in H2.35 cells under basal conditions, they did coimmunoprecipitate after insulin administration (Supplementary Fig. 6A). GST pull-down assay, however, showed that the association between WDFY2 and nonphosphorylated IRS1/2 was indirect (Supplementary Fig. 6B). These data suggest that WDFY2 regulates IRS1/2 activation via a mechanism other than directly binding both the substrate and the kinase.

After insulin binding, the insulin receptor kinase (IRK) β-subunit is phosphorylated, resulting in IRK activation and rapid internalization into endosomes (13). We thus examined the localization of the INSR upon insulin stimulation. We extracted endosome fractions from WT and Wdfy2 KO mice livers with or without insulin stimulation and then detected the INSR levels. In the WT liver, the endosomal INSR was significantly increased after insulin stimulation; this increase was suppressed in Wdfy2 KO mice (Fig. 6A and B). We also found that the interaction between WDFY2 and the INSR occurred in endosomes and significantly increased after insulin stimulation (Fig. 6C and D). Because IRS binding with the INSR and the phosphorylation of IRS occurs both on the plasma membrane and in endosomes upon insulin stimulation (13), we explored whether the IRS1/2 recruitment to endosome was affected by WDFY2 depletion. We found that IRS1/2 recruitment to endosomes following insulin treatment was markedly suppressed in the Wdfy2 KO liver, yet IRS1/2 localization on the plasma membrane was unaffected (Fig. 6EI). These data indicate that WDFY2 interacts with the INSR and localizes it to endosomes after insulin stimulation; then, INSR recruits IRS1/2 to endosomes.

Figure 6
Figure 6

WDFY2 interacts with the INSR and localizes it in early endosomes (ENs) to recruit and activate IRS1/2. AD: The endosomes fraction obtained from WT and Wdfy2 KO 3-month-old mice that were fasted overnight and injected intraperitoneally with either insulin (0.75 units/kg body wt) or saline for 15 min. Representative Western blots (A) and the associated quantifications (B) show the INSR levels in the liver endosomes fraction (n = 4 mice in each group). C: The stimulation-dependent interaction between WDFY2 and the INSR was analyzed by immunoprecipitation (IP) with the indicated antibodies. D: Quantification of the relative levels of the interaction between WDFY2 and the INSR (n = 3 mice in each group). EI: The plasma membrane (PM) and endosomes fraction obtained from WT and Wdfy2 KO 3-month-old mice that were fasted overnight and injected intraperitoneally with either insulin (0.75 units/kg body wt) or saline for 15 min. E: Subcellular fractions enriched in endosomes and the plasma membrane were analyzed by Western blotting with the indicated antibodies. FI: Quantification of the relative levels of IRS1 and IRS2 in the endosomes and plasma membrane fractions (n = 3 mice in each group). J and K: Wdfy2 KO H2.35 hepatocytes were transduced with an empty vector, Flag-WDFY2, or Flag-WD5–7 by retrovirus system for 48 h. WT and Wdfy2 KO H2.35 hepatocytes were stimulated with 10 mmol/L insulin for 5 min or unstimulated before lysis. Representative Western blots (J) and the associated quantifications (K) show the tyrosine phosphorylation levels of IRS1 and IRS2 and phosphorylation level of Akt in hepatocyte lysates (n = 3). The data represent the means ± SD. P values were calculated by Student t test (C and D), one-way ANOVA with Tukey post hoc test (J and K), or two-way ANOVA with Tukey post hoc test (A, B, and FI) (**P < 0.01, ***P < 0.001 vs. the respective control). IB, immunoblotting; N.S., not significant; T-Akt, total Akt.

We further confirmed the role of the WDFY2-INSR interaction in evoking downstream IRS1/2 and Akt activation by performing a rescue test. Here, overexpressing full-length WDFY2 fully rescued the impaired interaction between INSR and IRS in Wdfy2 KO H2.35 hepatocytes, whereas overexpressing the WD5–7 mutant plasmid had no effect on the interaction (Supplementary Fig. 6C and D). Inhibition of IRS1/2 and Akt phosphorylation as a result of Wdfy2 KO was also alleviated upon the overexpression of full- length, but not WD5–7 mutant, WDFY2 (Fig. 6J and K). These results indicate that after insulin stimulation, the INSR is unable to localize to the endosome and phosphorylate IRS1/2 without associating with WDFY2.

Hepatic WDFY2 Overexpression Attenuates Hyperglycemia and Insulin Resistance in Wdfy2 KO Mice and db/db Mice

We finally evaluated whether we could rescue the metabolic changes observed in our Wdfy2 KO mice with adeno-associated viral WDFY2 or WD5–7 mutant. Ectopic full-length and mutant WDFY2 expression was detectable in the liver tissue of Wdfy2 KO mice at day 21 (Fig. 7A) but not in the skeletal muscle and adipose tissue (Supplementary Fig. 7A). Interestingly, GTT and ITT showed that AAV-mediated WDF2 delivery alleviated both insulin resistance and glucose intolerance in Wdfy2 KO mice compared with Wdfy2 KO mice infused with a control AAV. WD5–7 mutant expression, however, could not rescue the impaired glucose and insulin tolerance (Fig. 7B and C). Similarly, in Wdfy2 KO mice with viral WDFY2 overexpression, the circulating insulin levels at baseline were much lower than those in the control Wdfy2 KO and Wdfy2 KO mice with viral WD5–7 overexpression (Fig. 7D). We also found that the expression of gluconeogenic genes (G6pc and Pck1) decreased and the accumulation of liver glycogen increased after refeeding only in the Wdfy2 KO mice with viral WDFY2 overexpression (Supplementary Fig. 7B and C).

Figure 7
Figure 7

Hepatic WDFY2 overexpression attenuates hyperglycemia and insulin resistance in Wdfy2 KO mice and db/db mice. A: Immunoblotting was performed to detect the protein levels in the liver tissue of 2-month-old Wdfy2 KO mice at day 21 after infection with an AAV encoding WDFY2 and WD5–7 mutant (n = 5 mice for each group). B and C: GTT (B) and ITT (C) were performed in 3-month-old Wdfy2 KO mice, Wdfy2 KO mice with WDFY2 overexpressed, and Wdfy2 KO mice with WD5–7 overexpressed at the indicated time points (left). The area under the curve (AUC) for the GTT is shown (right) (n = 5 mice for each group). D: Serum insulin levels in 3-month-old Wdfy2 KO mice, Wdfy2 KO mice with WDFY2 overexpressed, and Wdfy2 KO mice with WD5–7 overexpressed under fasting conditions (n = 5 mice for each group). EH: Two-month-old db/db mice were infected with AAV encoding WDFY2 or WD5–7 for 42 days. The fasting glucose (E), fasting serum insulin (F), fasting serum triglyceride (G), and fasting serum cholesterol (H) levels were determined (n = 4 mice for each group). The data represent the means ± SD. P values were calculated by one-way ANOVA with Tukey post hoc test (*P < 0.05, **P < 0.01, ***P < 0.001 vs. the respective control). CTR, empty vector control; N.S., not significant.

We also infused db/db mice (with frank diabetes) with the full-length or mutant WDFY2 AAV. Here, full-length WDFY2 overexpression in the liver alleviated fasting hyperglycemia, hyperinsulinemia, and hyperlipidemia, while overexpression of the WDFY2 truncation mutant had no obvious effect (Fig. 7EH). These data indicate that AAV-mediated WDFY2 overexpression can alleviate insulin resistance and glucose intolerance in mice with metabolic disorders.

Finally, we infused WT mice with WDFY2 AAV and monitored WDFY2 expression relative to WT WDFY2 mice after 21 days (Supplementary Fig. 7D). The GTT and ITT showed that AAV-mediated WDFY2 delivery enhanced both insulin resistance and glucose intolerance in WT mice (Supplementary Fig. 7E and F). Furthermore, the fasting insulin levels of WT mice overexpressing hepatic WDFY2 were significantly decreased (Supplementary Fig. 7G). These results demonstrate that increased WDFY2 expression in the liver ameliorates glucose utilization in healthy animals.

Discussion

WDFY2 interacts with the INSR via its WD1–4 domain and localizes it in early endosomes to recruit downstream IRS1/2 following insulin stimulation. In this way, WDFY2 mediates normal insulin signaling transduction in the liver and compromising the INSR-WDFY2 interaction ultimately impairs downstream PI3K-Akt signaling. Subsequently, phosphorylation of Akt2 and its direct substrates is decreased, hepatic glucose production is inappropriately promoted, and glycogen accumulation is suppressed. This novel mechanism explains how endosomal WDFY2 activates the hepatic insulin-evoked Akt2 signaling pathway (Fig. 8).

Figure 8
Figure 8

A model for the role of WDFY2 in insulin-evoked Akt2 signaling. After insulin stimulation, the INSR internalizes into endosomes in hepatocytes. Here, WDFY2 interacts with the INSR to localize it to early endosomes so that downstream IRS1/2 can be recruited to the endosomal INSR. If WDFY2 is absent, the INRS loses its localization in the early endosomes to interact and phosphorylate its substrates and mediate downstream metabolism-associated Akt2 activation in endosomes. P, phosphorylation.

WDFY2 protein is highly expressed in the liver (37), but prior studies had not analyzed WDFY2 function in the liver in depth. We found that hepatic WDFY2 is dynamically suppressed in the liver of dietary and genetic mouse models of diabetes. Hepatic WDFY2 levels decreased upon treatment with saturated but not unsaturated fatty acids, providing rationale for a link between WDFY2 and metabolic diseases. Using our systemic Wdfy2 KO mouse model, we found that mice lacking WDFY2 exhibited insulin resistance, glucose intolerance, and impaired hepatic glucose regulatory function when fed a normal diet. Whether these described metabolic changes occur in female mice now needs further study.

Insulin resistance in Wdfy2 KO male mice is mainly due to WDFY2 loss in the liver. Restoring WDFY2 liver but not adipose expression improved glucose homeostasis and insulin sensitivity. Interestingly, WDFY2 loss in adipose tissue can influence glucose metabolism, as WDFY2 knockdown in 3T3-L1 adipocytes impairs glucose uptake and adipogenesis (22,23). Adipocyte dysfunction causes decreased adipokine production, which can potentiate insulin sensitivity (38,39).

Mechanistically, insulin stimulation activates Akt-mediated FoxO1 phosphorylation, resulting in FoxO1 nuclear exclusion and reduced transcriptional activity (40). GSK-3β phosphorylation relieves the inhibitory effect on downstream glycogen synthase activity (41). This pathway explains the increased levels of gluconeogenesis and decreased glycogen synthesis seen in the Wdfy2 KO mouse liver. Akt2 has a role in glucose homeostasis (32,42), preferentially associating with PI34P2 at the plasma membrane and in endosomes, while Akt1 and Akt3 associate with PI345P3 exclusively at the plasma membrane (43,44). PI34P2, which is abundant in early endocytic membranes, promotes Akt2 activation in endosomes (45). WDFY2 has a distinct and crucial role in endosomal Akt2 signaling activity and specifically regulates metabolism-related substrates, such as FoxO1 and GSK-3β (43,46). As a result of WDFY2 endosomal localization, Akt2 signaling in endosomes is preferentially impaired such that the associated metabolic processes are mainly influenced in Wdfy2 KO H2.35 cells. Interestingly, Wdfy2 KO has no effect on the phosphorylation of Akt substrates, which are mainly involved in cell growth and proliferation. This finding is in line with previous studies showing that their phosphorylation is not regulated by endosomal Akt2 (47,48). Insulin-mediated phosphorylation of Akt1 might be slightly affected by WDFY2 deficiency; further studies are required to clarify this point. A previous study showed that Akt2 could bind to WDFY2 and was protected from degradation in adipocytes (23); our results, however, indicated that total Akt protein levels were unaltered in the Wdfy2 KO mouse liver and Wdfy2 KO H2.35 hepatocytes. We did not detect an association between WDFY2 and Akt2 in hepatocytes. This discrepancy could be due to differences in the models used, tissues investigated, and/or approaches used to detect the interaction. Most crucially, changes in WDFY2 protein expression might be tissue specific.

Although we did not detect an association between WDFY2 and Akt by coimmunoprecipitation in H2.35 cells, it is still possible that WDFY2 regulates hepatic insulin-evoked Akt signaling. Insulin induces Akt activation by stimulating INSR and IRS tyrosine phosphorylation, which in turn triggers PI3K activation (49). We found that INSR activation was not affected by Wdfy2 KO in H2.35 hepatocytes but IRS1/2 or PI3K activation was significantly impaired. In particular, insulin-stimulated IRS1/2 phosphorylation was decreased in the endosomes, but not plasma membrane, in Wdfy2 KO cells. WDFY2 might thus help regulate IRS1/2 activation by the INSR. The WDFY2 WD repeats preferentially bind the INSR in response to insulin stimulation. After insulin stimulation, the INSR on the plasma membrane is phosphorylated and internalized to endosomes. The INSR continues to signal at endosomal compartments, providing the molecular cue to activate IRS1/2 and class I PI3K to generate PIP3 at endosomes (11,12). PIP3 is a short-lived lipid messenger that generates PI34P2 in endosomes dephosphorylated by inositol phosphatases (43). Endosomes are the major site of IRS1/2 activation by the INSR (50). As an INSR-binding partner, WDFY2 retains the INSR within the early endosome so that downstream IRS1/2 can be recruited to the endosomal INSR and activated. Under basal conditions, nonphosphorylated IRS1/2 is distributed in the cytoplasm but is rapidly recruited to endosomes to interact with the INSR upon insulin administration. We found that the INSR cytoplasmic domain bound to the WDFY2 WD1–4 domain; a WD5–7 truncation mutant could localize on the endosome membranes but could not interact with the endosomal INSR. Overexpressing this mutant in Wdfy2 KO H2.35 hepatocytes and murine liver could not rescue impaired insulin signaling in cells or insulin resistance in mice. Given the function of WDFY2 in endocytosis, WDFY2 might also have a broader role in internalization, such as the internalization of plasma membrane–derived PIP3 or PI34P2. A defect in the internalization of plasma membrane–derived PIP3 or PI34P2 to endosomes also selectively impacts the endosomal pool of phosphorylated PI34P2 and thus Akt2 (19,43).

It is worth noticing that our study only demonstrated the function of WDFY2 in the liver; thus, insulin-stimulated Akt2 activation may not require WDFY2 in skeletal muscle. In addition, the mechanism of WDFY2 transcriptional regulation by metabolites, and its regulation in other organs, now remains to be studied. Because WDFY2 overexpression in the liver can reverse hyperglycemia, hyperinsulinemia, insulin resistance, and glucose intolerance in db/db mice, our data could have translational relevance. Indeed, increasing WDFY2 liver expression might constitute a new therapeutic concept for metabolic disorders. Future work will now assess the role of WDFY2 in human patient samples.

Article Information

Acknowledgments. The authors thank Dr. Ying Zhao (Peking University) and Dr. George Liu (Peking University) for critical comments on the manuscript. The authors also appreciate Dr. Karin Moelling (University of Zurich, Switzerland) and Silvia Corvera (University of Massachusetts Medical School) for sharing the WDFY2 plasmids. Finally, the authors thank Dr. Jessica Tamanini (ETediting, Shenzhen, and Shenzhen University, China) for proofreading the manuscript.

Funding. This work was supported by grants from National Key R&D Program of China (2017YFA0503900), National Natural Science Foundation of China (81720108027, 81530074), Science and Technology Program of Guangdong Province in China (2017B030301016), Shenzhen Municipal Commission of Science and Technology Innovation (JCYJ20170818092450901), Discipline Construction Funding of Shenzhen [(2016)1452], and Shenzhen Bay Laboratory (SZBL2019062801011).

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

Author Contributions. L.Z. and W.-G.Z. conceived the project and designed most of the experiments. L.Z. performed most of the experimental work. X. Li and N.Z. contributed to protein purification. X.Y., T.H., and L.W. contributed to animal study. W.F. and F.Y. performed construction of plasmids. H.Z. and X. Lu helped with the design of some experiments. L.Z. and W.-G.Z. wrote the manuscript. W.-G.Z. supervised the project. L.Z., H.W., Y.T., X. Lu, and W.-G.Z. critically read the manuscript. W.-G.Z. and L.Z. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

  • Received July 17, 2019.
  • Accepted June 4, 2020.



Source link