PED/PEA-15 induces autophagy and mediates TGF-beta1 effect on muscle cell differentiation.

TGF-beta1 has been shown to induce autophagy in certain cells but whether and how this action is exerted in muscle and whether this activity relates to TGF-beta1 control of muscle cell differentiation remains unknown. Here, we show that expression of the autophagy-promoting protein phosphoprotein enriched in diabetes/phosphoprotein enriched in astrocytes (PED/PEA-15) progressively declines during L6 and C2C12 skeletal muscle cell differentiation. PED/PEA-15 underwent rapid induction upon TGF-beta1 exposure of L6 and C2C12 myoblasts, accompanied by impaired differentiation into mature myotubes. TGF-beta1 also induced autophagy in the L6 and C2C12 cells through a PP2A/FoxO1-mediated mechanism. Both the TGF-beta1 effect on differentiation and that on autophagy were blocked by specific PED/PEA-15 ShRNAs. Myoblasts stably overexpressing PED/PEA-15 did not differentiate and showed markedly enhanced autophagy. In these same cells, the autophagy inhibitor 3-methyladenine rescued TGF-beta1 effect on both autophagy and myogenesis, indicating that PED/PEA-15 mediates TGF-beta1 effects in muscle. Muscles from transgenic mice overexpressing PED/PEA-15 featured a significant number of atrophic fibers, accompanied by increased light chain 3 (LC3)II to LC3I ratio and reduced PP2A/FoxO1 phosphorylation. Interestingly, these mice showed significantly impaired locomotor activity compared with their non-transgenic littermates. TGF-beta1 causes transcriptional upregulation of the autophagy-promoting gene PED/PEA-15, which in turn is capable to induce atrophic responses in skeletal muscle in vivo.

Overproduction of phosphoprotein enriched in diabetes (PED) induces mesangial expansion and upregulates protein kinase C-beta activity and TGF-beta1 expression.

AIMS/HYPOTHESIS: Overproduction of phosphoprotein enriched in diabetes (PED, also known as phosphoprotein enriched in astrocytes-15 [PEA-15]) is a common feature of type 2 diabetes and impairs insulin action in cultured cells and in mice. Nevertheless, the potential role of PED in diabetic complications is still unknown. METHODS: We studied the effect of PED overproduction and depletion on kidney function in animal and cellular models. RESULTS: Transgenic mice overexpressing PED (PEDTg) featured age-dependent increases of plasma creatinine levels and urinary volume, accompanied by expansion of the mesangial area, compared with wild-type littermates. Serum and kidney levels of TGF-beta1 were also higher in 6- and 9-month-old PEDTg. Overexpression of PED in human kidney 2 cells significantly increased TGF-beta1 levels, SMAD family members (SMAD)2/3 phosphorylation and fibronectin production. Opposite results were obtained following genetic silencing of PED in human kidney 2 cells by antisense oligonucleotides. Inhibition of phospholipase D and protein kinase C-beta by 2-butanol and LY373196 respectively reduced TGF-beta1, SMAD2/3 phosphorylation and fibronectin production. Moreover, inhibition of TGF-beta1 receptor activity and SMAD2/3 production by SB431542 and antisense oligonucleotides respectively reduced fibronectin secretion by about 50%. TGF-beta1 circulating levels were significantly reduced in Ped knockout mice and positively correlated with PED content in peripheral blood leucocytes of type 2 diabetic patients. CONCLUSIONS/INTERPRETATION: These data indicate that PED regulates fibronectin production via phospholipase D/protein kinase C-beta and TGF-beta1/SMAD pathways in kidney cells. Raised PED levels may therefore contribute to the abnormal accumulation of extracellular matrix and renal dysfunction in diabetes.

Factor H-binding protein is important for meningococcal survival in human whole blood and serum and in the presence of the antimicrobial peptide LL-37.

Factor H-binding protein (fHBP; GNA1870) is one of the antigens of the recombinant vaccine against serogroup B Neisseria meningitidis, which has been developed using reverse vaccinology and is the basis of a meningococcal B vaccine entering phase III clinical trials. Binding of factor H (fH), an inhibitor of the complement alternative pathway, to fHBP enables N. meningitidis to evade killing by the innate immune system. All fHBP null mutant strains analyzed were sensitive to killing in ex vivo human whole blood and serum models of meningococcal bacteremia with respect to the isogenic wild-type strains. The fHBP mutant strains of MC58 and BZ83 (high fHBP expressors) survived in human blood and serum for less than 60 min (decrease of >2 log(10) CFU), while NZ98/254 (intermediate fHBP expressor) and 67/00 (low fHBP expressor) showed decreases of >1 log(10) CFU after 60 to 120 min of incubation. In addition, fHBP is important for survival in the presence of the antimicrobial peptide LL-37 (decrease of >3 log(10) CFU after 2 h of incubation), most likely due to electrostatic interactions between fHBP and the cationic LL-37 molecule. Hence, the expression of fHBP by N. meningitidis strains is important for survival in human blood and human serum and in the presence of LL-37, even at low levels. The functional significance of fHBP in mediating resistance to the human immune response, in addition to its widespread distribution and its ability to induce bactericidal antibodies, indicates that it is an important component of the serogroup B meningococcal vaccine.

The PEA15 gene is overexpressed and related to insulin resistance in healthy first-degree relatives of patients with type 2 diabetes.

AIMS/HYPOTHESIS: Overexpression of the gene encoding phosphoprotein enriched in astrocytes 15 (PEA15), also known as phosphoprotein enriched in diabetes (PED), causes insulin resistance and diabetes in transgenic mice and has been observed in type 2 diabetic individuals. The aim of this study was to investigate whether PEA15 overexpression occurs in individuals at high risk of diabetes and whether it is associated with specific type 2 diabetes subphenotypes. SUBJECTS AND METHODS: We analysed PEA15 expression in euglycaemic first-degree relatives (FDR) of type 2 diabetic subjects. RESULTS: The expression of PEA15 in peripheral blood leucocytes (PBLs) paralleled that in fat and skeletal muscle tissues. In PBLs from the FDR, PEA15 expression was two-fold higher than in euglycaemic individuals with no family history of diabetes (control subjects), both at the protein and the mRNA level (p < 0.001). The expression of PEA15 was comparable in FDR and type 2 diabetic subjects and in each group close to one-third of the subjects expressed PEA15 levels more than 2 SD higher than the mean of control subjects. Subjects with IFG with at least one type 2 diabetes-affected FDR also overexpressed PEA15 (p < 0.05). In all the groups analysed, PEA15 expression was independent of sex and unrelated to age, BMI, waist circumference, systolic and diastolic BP, and fasting cholesterol, triacylglycerol and glucose levels. However, in euglycaemic FDR of type 2 diabetic subjects, PEA15 expression was inversely correlated with insulin sensitivity (r = -557, p = 0.01). CONCLUSIONS/INTERPRETATION: We conclude that PEA15 overexpression represents a common defect in FDR of patients with type 2 diabetes and is correlated with reduced insulin sensitivity in these individuals.

Activation and mitochondrial translocation of protein kinase Cdelta are necessary for insulin stimulation of pyruvate dehydrogenase complex activity in muscle and liver cells.

In L6 skeletal muscle cells and immortalized hepatocytes, insulin induced a 2-fold increase in the activity of the pyruvate dehydrogenase (PDH) complex. This effect was almost completely blocked by the protein kinase C (PKC) delta inhibitor Rottlerin and by PKCdelta antisense oligonucleotides. At variance, overexpression of wild-type PKCdelta or of an active PKCdelta mutant induced PDH complex activity in both L6 and liver cells. Insulin stimulation of the activity of the PDH complex was accompanied by a 2.5-fold increase in PDH phosphatases 1 and 2 (PDP1/2) activity with no change in the activity of PDH kinase. PKCdelta antisense blocked insulin activation of PDP1/2, the same as with PDH. In insulin-exposed cells, PDP1/2 activation was paralleled by activation and mitochondrial translocation of PKCdelta, as revealed by cell subfractionation and confocal microscopy studies. The mitochondrial translocation of PKCdelta, like its activation, was prevented by Rottlerin. In extracts from insulin-stimulated cells, PKCdelta co-precipitated with PDP1/2. PKCdelta also bound to PDP1/2 in overlay blots, suggesting that direct PKCdelta-PDP interaction may occur in vivo as well. In intact cells, insulin exposure determined PDP1/2 phosphorylation, which was specifically prevented by PKCdelta antisense. PKCdelta also phosphorylated PDP in vitro, followed by PDP1/2 activation. Thus, in muscle and liver cells, insulin causes activation and mitochondrial translocation of PKCdelta, accompanied by PDP phosphorylation and activation. These events are necessary for insulin activation of the PDH complex in these cells.

Insulin receptor substrate-2 phosphorylation is necessary for protein kinase C zeta activation by insulin in L6hIR cells.

We have investigated glycogen synthase (GS) activation in L6hIR cells expressing a peptide corresponding to the kinase regulatory loop binding domain of insulin receptor substrate-2 (IRS-2) (KRLB). In several clones of these cells (B2, F4), insulin-dependent binding of the KRLB to insulin receptors was accompanied by a block of IRS-2, but not IRS-1, phosphorylation, and insulin receptor binding. GS activation by insulin was also inhibited by >70% in these cells (p < 0.001). The impairment of GS activation was paralleled by a similarly sized inhibition of glycogen synthase kinase 3 alpha (GSK3 alpha) and GSK3 beta inactivation by insulin with no change in protein phosphatase 1 activity. PDK1 (a phosphatidylinositol trisphosphate-dependent kinase) and Akt/protein kinase B (PKB) activation by insulin showed no difference in B2, F4, and in control L6hIR cells. At variance, insulin did not activate PKC zeta in B2 and F4 cells. In L6hIR, inhibition of PKC zeta activity by either a PKC zeta antisense or a dominant negative mutant also reduced by 75% insulin inactivation of GSK3 alpha and -beta (p < 0.001) and insulin stimulation of GS (p < 0.002), similar to Akt/PKB inhibition. In L6hIR, insulin induced protein kinase C zeta (PKC zeta) co-precipitation with GSK3 alpha and beta. PKC zeta also phosphorylated GSK3 alpha and -beta. Alone, these events did not significantly affect GSK3 alpha and -beta activities. Inhibition of PKC zeta activity, however, reduced Akt/PKB phosphorylation of the key serine sites on GSK3 alpha and -beta by >80% (p < 0.001) and prevented full GSK3 inactivation by insulin. Thus, IRS-2, not IRS-1, signals insulin activation of GS in the L6hIR skeletal muscle cells. In these cells, insulin inhibition of GSK3 alpha and -beta requires dual phosphorylation by both Akt/PKB and PKC zeta.

Protein kinase C (PKC)-alpha activation inhibits PKC-zeta and mediates the action of PED/PEA-15 on glucose transport in the L6 skeletal muscle cells.

Overexpression of the PED/PEA-15 protein in muscle and adipose cells increases glucose transport and impairs further insulin induction. Like glucose transport, protein kinase C (PKC)-alpha and -beta are also constitutively activated and are not further stimulatable by insulin in L6 skeletal muscle cells overexpressing PED (L6(PED)). PKC-zeta features no basal change but completely loses insulin sensitivity in L6(PED). In these cells, blockage of PKC-alpha and -beta additively returns 2-deoxy-D-glucose (2-DG) uptake to the levels of cells expressing only endogenous PED (L6(WT)). Blockage of PKC-alpha and -beta also restores insulin activation of PKC-zeta in L6(PED) cells, with that of PKC-alpha sixfold more effective than PKC-beta. Similar effects on 2-DG uptake and PKC-zeta were also achieved by 50-fold overexpression of PKC-zeta in L6(PED). In L6(WT), fivefold overexpression of PKC-alpha or -beta increases basal 2-DG uptake and impairs further insulin induction with no effect on insulin receptor or insulin receptor substrate phosphorylation. In these cells, overexpression of PKC-alpha blocks insulin induction of PKC-zeta activity. PKC-beta is 10-fold less effective than PKC-alpha in inhibiting PKC-zeta stimulation. Expression of the dominant-negative K(281)-->W PKC-zeta mutant simultaneously inhibits insulin activation of PKC-zeta and 2-DG uptake in the L6(WT) cells. We conclude that activation of classic PKCs, mainly PKC-alpha, inhibits PKC-zeta and may mediate the action of PED on glucose uptake in L6 skeletal muscle cells.

Insulin-activated protein kinase Cbeta bypasses Ras and stimulates mitogen-activated protein kinase activity and cell proliferation in muscle cells.

In L6 muscle cells expressing wild-type human insulin receptors (L6hIR), insulin induced protein kinase Calpha (PKCalpha) and beta activities. The expression of kinase-deficient IR mutants abolished insulin stimulation of these PKC isoforms, indicating that receptor kinase is necessary for PKC activation by insulin. In L6hIR cells, inhibition of insulin receptor substrate 1 (IRS-1) expression caused a 90% decrease in insulin-induced PKCalpha and -beta activation and blocked insulin stimulation of mitogen-activated protein kinase (MAPK) and DNA synthesis. Blocking PKCbeta with either antisense oligonucleotide or the specific inhibitor LY379196 decreased the effects of insulin on MAPK activity and DNA synthesis by >80% but did not affect epidermal growth factor (EGF)- and serum-stimulated mitogenesis. In contrast, blocking c-Ras with lovastatin or the use of the L61,S186 dominant negative Ras mutant inhibited insulin-stimulated MAPK activity and DNA synthesis by only about 30% but completely blocked the effect of EGF. PKCbeta block did not affect Ras activity but almost completely inhibited insulin-induced Raf kinase activation and coprecipitation with PKCbeta. Finally, blocking PKCalpha expression by antisense oligonucleotide constitutively increased MAPK activity and DNA synthesis, with little effect on their insulin sensitivity. We make the following conclusions. (i) The tyrosine kinase activity of the IR is necessary for insulin activation of PKCalpha and -beta. (ii) IRS-1 phosphorylation is necessary for insulin activation of these PKCs in the L6 cells. (iii) In these cells, PKCbeta plays a unique Ras-independent role in mediating insulin but not EGF or other growth factor mitogenic signals.

The IR1152 mutant insulin receptor selectively impairs insulin action in skeletal muscle but not in liver.

In patients harboring the IR1152 mutant insulin receptor, hepatic glucose production was normally suppressed by insulin. Hepatocytes without the insulin receptor gene and expressing IR1152 (Hep(MUT)) also showed normal insulin suppression of glucose production and full insulin response of glycogen synthase. In contrast, expression of the IR1152 mutant in skeletal muscle maximally increased glucose uptake and storage, preventing further insulin stimulation. IRS-1 phosphorylation was normally stimulated by insulin in both intact Hep(MUT) and L6 skeletal muscle cells expressing the IR1152 mutant (L6(MUT)). At variance, IRS-2 phosphorylation exhibited high basal levels with no further insulin-dependent increase in L6(MUT) but almost normal phosphorylation, both basal and insulin-stimulated, in the Hep(MUT) cells. In vitro, IR1152 mutant preparations from both the L6(MUT) and the Hep(MUT) cells exhibited increased basal and no insulin-stimulated phosphorylation of IRS-2 immobilized from either muscle or liver cells. IR1152 internalization in liver and muscle cells closely paralleled the ability of this mutant to phosphorylate IRS-2 in vivo in these cells. Block of receptor internalization (wild-type and mutant) in the liver and muscle cells also inhibited IRS-2, but not IRS-1, phosphorylation. Thus, the mechanisms controlling insulin receptor internalization differ in liver and skeletal muscle cells and may enable IR1152 to control glucose metabolism selectively in liver. In both cell types, receptor internalization seems necessary for IRS-2 but not IRS-1 phosphorylation.

In L6 skeletal muscle cells, glucose induces cytosolic translocation of protein kinase C-alpha and trans-activates the insulin receptor kinase.

In L6 skeletal muscle cells expressing human insulin receptors (L6(hIR)), exposure to 25 mM glucose for 3 min induced a rapid 3-fold increase in GLUT1 and GLUT4 membrane translocation and glucose uptake. The high glucose concentration also activated the insulin receptor kinase toward the endogenous insulin receptor substrates (IRS)-1 and IRS-2. At variance, in L6 cells expressing kinase-deficient insulin receptors, the exposure to 25 mM glucose elicited no effect on glucose disposal. In the L6(hIR) cells, the acute effect of glucose on insulin receptor kinase was paralleled by a 2-fold decrease in both the membrane and the insulin receptor co-precipitated protein kinase C (PKC) activities and a 3-fold decrease in receptor Ser/Thr phosphorylation. Western blotting of the receptor precipitates with isoform-specific PKC antibodies revealed that the glucose-induced decrease in membrane- and receptor-associated PKC activities was accounted for by dissociation of PKCalpha but not of PKCbeta or -delta. This decrease in PKCalpha was paralleled by a similarly sized increase in cytosolic PKCalpha. In intact L6(hIR) cells, inhibition of PKCalpha expression by using a specific antisense oligonucleotide caused a 3-fold increase in IRS phosphorylation by the insulin receptor. This effect was independent of insulin and accompanied by a 2.5-fold increase in glucose disposal by the cells. Thus, in the L6 skeletal muscle cells, glucose acutely regulates its own utilization through the insulin signaling system, independent of insulin. Glucose autoregulation appears to involve PKCalpha dissociation from the insulin receptor and its cytosolic translocation.

PED/PEA-15: an anti-apoptotic molecule that regulates FAS/TNFR1-induced apoptosis.

PED/PEA-15 is a recently cloned 15 kDa protein possessing a death effector domain (DED). In MCF-7 and HeLa cells, a fivefold overexpression of PED/PEA-15 blocked FasL and TNFalpha apoptotic effects. This effect of PED overexpression was blocked by inhibition of PKC activity. In MCF-7 and HeLa cell lysates, PED/PEA-15 co-precipitated with both FADD and FLICE. PED/PEA-15-FLICE association was inhibited by overexpression of the wild-type but not of a DED-deletion mutant of FADD. Simultaneous overexpression of PED/PEA-15 with FADD and FLICE inhibited FADD-FLICE co-precipitation by threefold. Based on cleavage of the FLICE substrate PARP, this inhibitory effect was paralleled by a threefold decline in FLICE activation in response to TNF-alpha. TNFalpha, in turn, reduces PED association with the endogenous FADD and FLICE of the cells. Thus, PED/PEA-15 is an endogenous protein inhibiting FAS and TNFR1-mediated apoptosis. At least in part, this function may involve displacement of FADD-FLICE binding through the death effector domain of PED/PEA-15.

Differential role of insulin receptor substrate (IRS)-1 and IRS-2 in L6 skeletal muscle cells expressing the Arg1152 --> Gln insulin receptor.

In L6 muscle cells expressing the Arg1152 --> Gln insulin receptor (Mut), basal tyrosine phosphorylation of insulin receptor substrate (IRS)-1 was increased by 35% compared with wild-type cells (WT). Upon exposure to insulin, IRS-1 phosphorylation increased by 12-fold in both the Mut and WT cells. IRS-2 was constitutively phosphorylated in Mut cells and not further phosphorylated by insulin. The maximal phosphorylation of IRS-2 in basal Mut cells was paralleled by a 4-fold increased binding of the kinase regulatory loop binding domain of IRS-2 to the Arg1152 --> Gln receptor. Grb2 and phosphatidylinositol 3-kinase association to IRS-1 and IRS-2 reflected the phosphorylation levels of the two IRSs. Mitogen-activated protein kinase activation and [3H]thymidine incorporation closely correlated with IRS-1 phosphorylation in Mut and WT cells, while glycogen synthesis and synthase activity correlated with IRS-2 phosphorylation. The Arg1152 --> Gln mutant did not signal Shc phosphorylation or Shc-Grb2 association in intact L6 cells, while binding Shc in a yeast two-hybrid system and phosphorylating Shc in vitro. Thus, IRS-2 appears to mediate insulin regulation of glucose storage in Mut cells, while insulin-stimulated mitogenesis correlates with the activation of the IRS-1/mitogen-activated protein kinase pathway in these cells. IRS-1 and Shc-mediated mitogenesis may be redundant in muscle cells.

In NIH-3T3 fibroblasts, insulin receptor interaction with specific protein kinase C isoforms controls receptor intracellular routing.

Insulin increased protein kinase C (PKC) activity by 2-fold in both membrane preparations and insulin receptor (IR) antibody precipitates from NIH-3T3 cells expressing human IRs (3T3hIR). PKC-alpha, -delta, and -zeta were barely detectable in IR antibody precipitates of unstimulated cells, while increasing by 7-, 3.5-, and 3-fold, respectively, after insulin addition. Preexposure of 3T3hIR cells to staurosporine reduced insulin-induced receptor coprecipitation with PKC-alpha, -delta, and -zeta by 3-, 4-, and 10-fold, respectively, accompanied by a 1.5-fold decrease in insulin degradation and a similar increase in insulin retroendocytosis. Selective depletion of cellular PKC-alpha and -delta, by 24 h of 12-O-tetradecanoylphorbol-13-acetate (TPA) exposure, reduced insulin degradation by 3-fold and similarly increased insulin retroendocytosis, with no change in PKC-zeta. In lysates of NIH-3T3 cells expressing the R1152Q/K1153A IRs (3T3Mut), insulin-induced coprecipitation of PKC-alpha, -delta, and -zeta with the IR was reduced by 10-, 7-, and 3-fold, respectively. Similar to the 3T3hIR cells chronically exposed to TPA, untreated 3T3Mut featured a 3-fold decrease in insulin degradation, with a 3-fold increase in intact insulin retroendocytosis. Thus, in NIH-3T3 cells, insulin elicits receptor interaction with multiple PKC isoforms. Interaction of PKC-alpha and/or -delta with the IR appears to control its intracellular routing.

Abnormal glucose transport and GLUT1 cell-surface content in fibroblasts and skeletal muscle from NIDDM and obese subjects.

Glucose transport and GLUT1 expression were studied in fibroblasts from 7 lean and 5 obese non-insulin-dependent diabetic (NIDDM) subjects with at least 2 NIDDM first-degree relatives and from 12 lean and 5 obese non-diabetic subjects with no family history of diabetes. The obese individuals also had a strong family history of obesity. Fibroblasts from all of the subjects exhibited no difference in insulin receptor binding, autophosphorylation, and kinase and hexokinase activity. At variance, basal 2-deoxyglucose (2-DG) uptake and 3H-cytochalasin B binding were 50% increased in cells from individuals with NIDDM (p < 0.001) and/or obesity (p < 0.01) as compared to the lean non-diabetic subjects. Insulin-dependent (maximally stimulated-basal) 2-DG uptake and cytochalasin B binding were decreased three-fold in cells from the diabetic and/or obese subjects (p < 0.01). GLUT1 mRNA and total protein levels were comparable in fibroblasts from all the groups. However, basal GLUT1 cell-surface content was 50% greater in fibroblasts from the NIDDM and/or obese subjects as compared to the lean non-diabetic individuals while insulin-dependent GLUT1 recruitment at the cell surface was diminished three-fold. Increased basal GLUT1 content in the plasma membrane was also observed in skeletal muscle of 4 NIDDM and 3 non-diabetic obese individuals (p < 0.05 vs the lean non diabetic subjects). Basal 2-DG uptake in fibroblasts from diabetic/obese individuals and lean control subjects strongly correlated with the in vivo fasting plasma insulin concentration of the donor. A negative correlation was demonstrated between the magnitude of insulin-dependent glucose uptake by the fibroblasts and plasma insulin levels in vivo. We conclude that a primary abnormality in glucose transport and GLUT1 cell-surface content is present in fibroblasts from NIDDM and obese individuals. The abnormal GLUT1 content is also present in skeletal muscle plasma membranes from NIDDM and obese individuals.

Receptor-mediated internalization of insulin. Potential role of pp120/HA4, a substrate of the insulin receptor kinase.

pp120/HA4 is a hepatocyte membrane glycoprotein phosphorylated by the insulin receptor tyrosine kinase. In this study, we have investigated the role of pp120/HA4 in insulin action. Transfection of antisense pp120/HA4 cDNA in H35 hepatoma cells resulted in inhibition of pp120/HA4 expression and was associated with a 2-3-fold decrease in the rate of insulin internalization. Furthermore, insulin internalization in NIH 3T3 fibroblasts co-transfected with insulin receptors and pp120/HA4 was increased 2-fold compared with cells expressing insulin receptors alone. In contrast, no effect on internalization was observed in cells overexpressing a naturally occurring splice variant of pp120/HA4 that lacks the phosphorylation sites in the intracellular domain. Insulin internalization was also unaffected in cells expressing three site-directed mutants of pp120/HA4 in which the sites of phosphorylation by the insulin receptor kinase had been removed (Y488F, Y488F/Y513F, and S503A). Our data suggest that pp120/HA4 is part of a complex of proteins required for receptor-mediated internalization of insulin. It is possible that this function is regulated by insulin-induced phosphorylation of the intracellular domain of pp120/HA4.