Role of Myotonic Dystrophy Protein Kinase (DMPK) in Glucose Homeostasis and Muscle Insulin Action

Myotonic dystrophy 1 (DM1) is caused by a CTG expansion in the 3′-unstranslated region of the DMPK gene, which encodes a serine/threonine protein kinase. One of the common clinical features of DM1 patients is insulin resistance, which has been associated with a pathogenic effect of the repeat expansions. Here we show that DMPK itself is a positive modulator of insulin action. DMPK-deficient (dmpk −/−) mice exhibit impaired insulin signaling in muscle tissues but not in adipocytes and liver, tissues in which DMPK is not expressed. Dmpk −/− mice display metabolic derangements such as abnormal glucose tolerance, reduced glucose uptake and impaired insulin-dependent GLUT4 trafficking in muscle. Using DMPK mutants, we show that DMPK is required for a correct intracellular trafficking of insulin and IGF-1 receptors, providing a mechanism to explain the molecular and metabolic phenotype of dmpk −/− mice. Taken together, these findings indicate that reduced DMPK expression may directly influence the onset of insulin-resistance in DM1 patients and point to dmpk as a new candidate gene for susceptibility to type 2-diabetes.


INTRODUCTION
Type 2 diabetes is a heterogeneous disease and a major international public health threat. It is widely accepted that type 2 diabetes results from a combination of genetic susceptibility and other risk factors including obesity, increased age, hypertension, and lifestyle [1]. Insulin resistance, which is a major factor in the development of type 2 diabetes [2], is a common metabolic feature in myotonic dystrophy 1 (DM1), an autosomal dominant neuromuscular disorder [3]. DM1 patients frequently exhibit normal basal insulin levels but excessive insulin release after a glucose load, indicating a compensatory betacell response to tissue insulin insensitivity [4,5]. Whole-body glucose disposal in DM1 patients is reduced by 15225% following insulin infusion [6] and experiments with forearm muscle indicate a 70% decrease in insulin sensitivity in skeletal muscle [7]. The DM1 mutation has been identified as the expansion of an unstable CTGrepeat in the 39-untranslated region of a gene encoding DMPK (myotonic dystrophy protein kinase) [8,9]. Insulin resistance in DM1 has been associated with aberrant splicing of the insulin receptor RNA due to a toxic effect of the CUG-expanded repeats, which are transcribed from the mutated dmpk gene but are retained in the nucleus altering the normal metabolism of RNAs [10,11]. However, whether the entire endocrine pathology of DM1 is caused by alterations in RNA processing remains to be seen. Indeed, DM1 patients show a 50% decrease in DMPK expression [12] and studies of dmpk knockout mice indicate that at least some of the features of DM1 result from haploinsufficiency of DMPK [13][14][15]. Interestingly, dmpk gene is located on chromosome 19q13, in which quantitative trait loci (QTLs) for type 2 diabetes-associated phenotypes have been identified by two independent genome-wide linkage scans among large and multiple ethnicity populations [16,17].
DMPK is mainly expressed in muscle [18], which is a key target tissue for insulin-dependent regulation of glucose metabolism [19]. Structurally, DMPK presents homology with protein kinases of the Rho family (Rho-kinase), which have important roles in the organization of the cytoskeleton and several cellular processes including intracellular protein trafficking and metabolism [20,21].
Although little is known about the mechanisms that regulate DMPK activity, it has been described that DMPK is activated in response to G protein second messengers [22] and that the actin cytoskeletonlinked GTPase Rac-1 binds to DMPK, promoting its transphosphorylation activity in a GTP-sensitive manner [23]. Here we examined the role of DMPK in the regulation of insulin action and glucose homeostasis using a DMPK-deficient mouse model [13]. We show that DMPK plays a role in the regulation of whole-body glucose disposal and muscle insulin sensitivity through a mechanism that involves the intracellular trafficking of insulin and IGF-1 receptors.

RESULTS
Dmpk 2/2 mice exhibit insulin signaling defects in skeletal and cardiac muscles We used dmpk-null mice to explore the involvement of DMPK in insulin action. We first analyzed cardiac and skeletal muscles, in which DMPK is preferentially expressed [18]. Dmpk 2/2 mice showed normal expression of the insulin receptor and other components of the insulin signaling pathway, such as the protein kinase Akt and glycogen synthase kinase 3b (GSK3-b) ( Fig. 1 A). However, abnormalities were found in the activation of the insulin signaling pathway in both cardiac and skeletal muscle. Insulininduced autophosphorylation of the insulin/IGF-I receptor (Tyr1150/1151-InsR; Tyr1135/1136-IGF-1R) was substantially decreased in dmpk 2/2 mice after insulin treatment in vivo (54610% decrease in cardiac muscle; 44610% decrease in skeletal muscle) ( Fig. 1 A). Phosphorylation of other components of the insulin signaling pathway such as Ser473-Akt and Ser9-GSK3-b was also Figure 1. DMPK regulates muscle insulin signaling. (A) Cardiac and skeletal muscles (soleus and gastrocnemius); (B) white adipose tissue; and (C) liver homogenates from dmpk 2/2 and wild type mice were analysed by Western blot. Caveolin 3, caveolin 1 and b-actin were used as loading controls. Heart homogenate was loaded as positive control for DMPK expression in B and C. (D) C2C12 myoblasts transduced with recombinant adenovirus for myc-DMPK or green fluorescent protein (GFP) as control were analyzed by Western blot. Caveolin 3 was used as loading control. Data are means6SEM. *P,0.05 vs control values (n = 3-5). doi:10.1371/journal.pone.0001134.g001 severely reduced in cardiac muscle from dmpk 2/2 mice (3965% decrease in Akt phosphorylation; 6769% decrease in GSK3b phosphorylation) ( Fig. 1 A). In contrast, insulin-induced autophosphorylation of the insulin/IGF-I receptor and phosphorylation of Ser473-Akt were preserved in adipose tissue and liver ( Fig. 1 B and C, respectively), in which DMPK is not expressed. These results suggest that insulin signaling defects in dmpk 2/2 mice are restricted to DMPK-expressing tissues. A role of DMPK in muscle insulin signaling was corroborated by overexpression of DMPK in C2C12 skeletal muscle cells. In myoblasts transduced with myc-tagged wild-type DMPK-adenovirus, insulin-stimulated phosphorylation of Ser473-Akt and Ser9-GSK3-b was increased 1.860.2-and 5.562.6fold, respectively, compared to control cells (Fig. 1 D).

Metabolic alterations in dmpk 2/2 mice
To examine whether dmpk 2/2 mice display muscle insulin resistance, we first measured insulin-stimulated glucose transport in cardiac and skeletal muscle. In myocytes isolated from left ventricles and soleus muscles from dmpk 2/2 mice, insulin-stimulated glucose transport was decreased compared with wild-type mice ( Fig. 2 A and C, respectively). The alterations in glucose uptake were not due to a decrease in the insulin receptor or the glucose transporter GLUT4 expression levels ( Fig. 2 B and D). Moreover, expression of caveolin 3, a specific muscular marker of caveolae that is altered in other forms of muscular dystrophies [24] and is required for insulin-stimulated glucose uptake [25], was also normal in cardiac and skeletal muscles from dmpk 2/2 mice. We further studied the role of DMPK in muscle insulin sensitivity by analyzing GLUT4 translocation, a critical muscle response to insulin. We performed subcellular fractionations of cardiac muscle membranes from wild-type and dmpk 2/2 mice. Endosomes and sarcolemmal membrane fractions were separated by successive spins of homogenates (Fig. 2 E). Low density microsome (LDM) fractions were enriched in GLUT4 and contained very low levels of Na + /K + -ATPase, indicating that such fractions were mostly devoid of plasma membrane. In wild-type mice, GLUT4 content in plasma membrane (PM) fraction was increased 51615% by insulin injection, at the expense of the LDM pool of GLUT4, which showed 2364% reduction upon insulin treatment. Insulin-induced translocation of GLUT4 from LDM to PM was not detected in dmpk 2/2 mice (Fig. 2 F).
Glucose tolerance tests performed at 4 weeks of age were similar for dmpk 2/2 and wild-type mice (Fig. 3 A). In contrast, at 8-10 weeks of age glucose tolerance in both male and female dmpk 2/2 mice was altered ( Fig. 3 B and C, respectively). Plasma insulin levels measured during glucose tolerance tests were also elevated in dmpk 2/2 as compared to wild-type mice (Fig. 3 D). However, dmpk 2/2 mice showed normal fasting glucose and glycemia returned to baseline 2 h after glucose injection in both groups. Fasted dmpk 2/2 mice showed normal insulin, triglyceride and free fatty acid levels ( Table 1) while these parameters in fed animals were higher than those from control animals (Fig. 3 E). No increases in fat-cell mass or fat-cell number were detected in dmpk 2/2 mice (Fig. 3 F). The glucose intolerant response of dmpk 2/2 mice placed on a high-fat diet for 8 weeks was more severe than that of dmpk 2/2 mice on a standard chow diet. Indeed, this metabolic stress significantly increased fasting blood glucose levels as well as glycemia 2 h after the glucose overload in dmpk 2/2 compared to wild type mice (Fig. 3 G).

DMPK is required for insulin receptor targeting to the plasma membrane
Regarding the molecular mechanism whereby DMPK activity could influence insulin receptor signaling, previous observations led us to analyze the role of DMPK in stress fiber formation: (i) DMPK function has been associated with the regulation of cytoskeleton in lens cells [26], and (ii) it has recently been shown that disruption of the actin cytoskeleton leads to alterations in insulin receptor localization and signaling [27]. We used two mutants of DMPK: myc-K110ADMPK, a kinase deficient form mutated at the ATP-binding site [28], and myc-DMADMPK, which lacks C-terminal residues 550-629 but retains kinase activity [22]. HeLa cells were transiently transfected with wild-type myc-DMPK, myc-K110ADMPK or myc-DMADMPK. Transfected cells were analyzed after 3-h starvation to determine the stress fiber content in steady state ( Fig. 4 A and B). The stress fiber pattern in cells expressing myc-WTDMPK was indistinguishable from that of the surrounding untransfected cells (Fig. 4 A, a-c and  Fig. 4 B). Expression of the kinase-dead mutant myc-K110ADMPK resulted in a disassembly of stress fibers (Fig. 4 A, d-f and Fig. 4 B) while myc-DMADMPK induced a gross condensation of actin filaments within the cell (Fig. 4 A, g-i and Fig. 4 B). These results were consistent with those previously reported for DMPK in lens cells and for the DMPK homolog ROKa in HeLa cells [26,29]. Having verified the effect of DMPK mutants in stress fiber formation in HeLa cells, we analyzed their effect in the intracellular trafficking of the insulin receptor. HeLa cells were transiently co-transfected with yellow fluorescent protein-tagged InsR (YFP-InsR) along with myc-WTDMPK,  myc-K110ADMPK, myc-DMADMPK or empty vector as control. Co-transfected cells were analyzed by confocal immunofluorescence after 3-h starvation to determine the localization of YFP-InsR in steady state. In cells expressing either empty vector or myc-WTDMPK, YFP-InsR was targeted to the cell surface (Fig. 5 A, a-d and e-h, respectively). Quantification analysis showed that the overexpression of myc-WTDMPK induced a significant increase in the receptor density at the plasma membrane compared to control cells (Fig. 5 B). In contrast, in the presence of either myc-K110ADMPK or myc-DMADMPK mutants, YFP-InsR was retained in intracellular structures (Fig. 5 A, i-l and m-p, respectively), with no evident co-localization with the DMPK mutants ( Fig. 5 A, k and o). For both mutants, the percentages of YFP-InsR at the plasma membrane were significantly decreased compared to control cells (Fig. 5 B). To identify the intracellular compartments in which the YFP-InsR was retained, HeLa cells were transfected with the vector encoding this protein and myc-WTDMPK or myc-K110ADMPK and subjected to staining with anti-myc antibody and a series of markers for different organelles.
Co-localization studies revealed that in the presence of myc-K110ADMPK, the intracellularly retained receptor partially coincides with the Golgi matrix protein GM130 (Fig. 6). No evidence of co-localization was detected with early endosomes (EEA1) or to recycling endosomes (transferrin receptor) (not shown). To further analyze DMPK role in insulin receptor trafficking, we measured the insulin binding activity in cardiomyocytes isolated from dmpk 2/2 and wild type mice. 125 I-insulin binding assays were carried out at 12uC on cells from 16-h fasted mice to analyze cell surface binding in the absence of internalization. Cells from dmpk 2/2 mice showed a 3668% (n = 4, p,0.05) decrease in 125 I-insulin binding at 0.033 nM insulin concentration compared to controls (Table 2). This decrease was associated with a reduced number of receptors at the cell surface with no apparent changes in receptor affinity for the ligand as the percentage of 125 I-insulin displacement at saturating unlabeled insulin concentrations was similar in both groups (58613% and 6268% displacement for wild type and dmpk 2/2 mice, respectively). To better understand the metabolic phenotype of dmpk 2/2 mice, we also analyzed the impact of DMPK in the IGF1-receptor intracellular trafficking.
The results obtained by co-transfection of GFP-tagged IGF-1 receptor along with myc-WTDMPK, myc-K110ADMPK, myc-DMADMPK or empty vector as control were similar to those obtained for the insulin receptor ( Fig. 7 A and B).

DISCUSSION
In this study, we addressed the molecular and metabolic function of DMPK, a poorly characterized serine/threonine protein kinase. Consistent with the preferential expression of DMPK in muscle tissues [18], we show that dmpk 2/2 mice exhibit insulin signaling defects in cardiac and skeletal muscle. Our data indicate that dmpk 2/2 mice present a significant degree of metabolic alteration, reflected in elevated glucose levels in glucose tolerance tests and increased circulating fed insulin and lipid levels. As observed in DM1 patients [4], dmpk 2/2 mice show higher concentrations of plasma insulin than wild-type mice in the glucose tolerance tests. These mice exhibit impaired glucose uptake and GLUT4 translocation indicating that decreased insulin sensitivity in muscle could be at the basis of the observed metabolic alterations. Indeed, dmpk 2/2 mice show normal adiposity and insulin signaling in adipose tissue and liver, in which DMPK is not expressed.
Insulin-induced autophosphorylation of insulin/IGF1-R in response to insulin is influenced by the number of receptors at the cell surface. Correct intracellular trafficking of the InsR is critical for insulin sensitivity and it has been shown that mutations in the InsR gene that impair the transport of the receptor to the plasma membrane lead to type 2 diabetes in humans [30]. After insulin binding, InsR is rapidly internalized and either sent to lysosomes for degradation, or recycled to the plasma membrane  for another round of binding, activation, and internalization [31].
Our results show that: (i) the insulin receptor expression levels in dmpk 2/2 mice are normal; (ii) the insulin binding to the plasma membrane of dmpk 2/2 cardiomyocytes is decreased; (iii) the overexpression of kinase-deficient and C-terminal truncated DMPK mutants leads to the retention of the InsR in intracellular compartments; and (iv) the overexpression of wild-type myc-DMPK increased the percentage of YFP-InsR at the cell surface. Altogether, these data indicate that DMPK is involved in InsR intracellular trafficking. However, this molecular mechanism cannot fully explain the metabolic alterations observed in dmpk 2/2 mice, especially considering the previous characterization of the muscle-insulin receptor knockout (MIRKO) mice which do not show glucose intolerance [32]. For this reason, we analyzed whether DMPK also regulated the targeting of IGF-1 receptor to the plasma membrane. Indeed, functional inactivation of the IGF-1 and insulin receptors in skeletal muscle (MKR mice) leads to type 2 diabetes phenotype [33]. Interestingly, the glucose tolerance tests performed in 4-week-old MKR mice are very similar to those observed in 8-10-week-old dmpk 2/2 animals. However, in contrast to MKR mice, older dmpk 2/2 mice do not develop type 2 diabetes. The milder phenotype of dmpk 2/2 mice compared to MKR mice is consistent with a mechanism involving a reduction of InsR and IGF-1R function rather than full inactivation of these receptors. We show that the kinase and C-terminal domains in DMPK are positive and negative regulators, respectively, of the cytoskeleton reorganization. Similar functions were previously found for DMPK in lens cells [26] and for homologous domains in Rho kinase a [29]. Both kinase and C-terminal domain mutants of DMPK alter insulin and IGF-1 receptor targeting to the plasma membrane and stress-fiber formation; however, whether these effects are functionally related remains to be determined. One possibility is that DMPK could be involved in the biogenesis of Golgi-derived transport carriers through regulation of actin cytoskeleton dynamics. Indeed, when actin dynamics is impaired by a variety of actin toxins that depolymerize or stabilize actin filaments, the Golgi complex shows significant structural changes [34]. Interestingly, we detected that the receptor intracellular accumulation caused by DMPK kinase-deficient mutant partially co-localizes with the Golgi matrix protein GM130.
In summary, our study provides in vivo and in vitro evidence for the role of DMPK in the regulation of insulin action and glucose homeostasis. Taken together, these findings indicate that reduced DMPK expression may directly influence the onset of insulinresistance in myotonic dystrophy 1 patients and suggest that DMPK could represent a susceptibility gene to type 2-diabetes.

Mouse experiments
All animal studies were performed in accordance with the guidelines and under approval of the Institutional Review Committee for the Animal Care and Use of the University of Barcelona and by the animal welfare regulations of the University of California, San Diego. The dmpk 2/+ heterozygous mice on 129SV background were generated by Reddy et al. [13]. These mice were mated to produce littermates that were homozygous for intact DMPK allele (WT) and homozygous for the null DMPK allele (KO). Except when indicated, 3-month-old male mice were used in experiments shown. Mice were backcrossed for at least 9 times maintaining the animals as congenic in the colony. Female mice were also analyzed in most of the studies presented with similar results.

Physiological assays
The following measurements were performed on randomly fed or on 16-h fasted animals when indicated. We measured blood glucose levels on whole venous blood using an automatic glucose monitor (One Touch Basic, Lifescan). Plasma insulin levels were measured by ELISA, using rat insulin as a standard (Crystal Chem). Free fatty acids and triglyceride levels were quantified from plasma using kits from Wako (NEFA-C Ki and Triglyceride L-Type, respectively). We performed glucose tolerance test on 16 h fasted mice injected intraperitoneally with D-glucose (2 g/kg body weight, Sigma). For high-fat diet treatment, mice were individually placed on a fat-adjusted diet (60 kcal% from fat, Research Diets, Inc., New Brunswick, NJ) for 8 weeks.

Cardiomyocyte isolation and glucose uptake assays
Left ventricle myocytes from 3-month-old mice were prepared by collagenase digestion as described previously [35] except that cells were finally resuspended in D-glucose-free DMEM (Gibco) supplemented with sodium pyruvate (0.22 mg/ml) and 0.2% BSA. Cardiomyocyte suspensions at 10% cytocrit were incubated with or without 100 nM insulin at 35uC for 30 min. The transport assay was initiated by the addition of 2-deoxy-D-glucose (1 mM final concentration, containing 0.5 mCi of 2-deoxy-D-[ 3 H]glucose, Amersham Pharmacia Biotech). Glucose uptake was terminated after 20 min by transferring the cell suspension to microfuge tubes and immediately centrifuged at 1006g for 30 s. Cell pellets were washed 3 times in ice-cold 50 mM D-glucose in PBS. Background activity was determined by measuring the transport in a solution that contained 50 mM D-glucose. Cells were lysed with 1 ml of ice-cold 0.1 N NaOH with 0.1% SDS and aliquots were taken for determination of radioactivity and protein levels.
For skeletal muscle glucose uptake assays, mice were sacrificed by cervical dislocation and soleus muscles were rapidly dissected. Muscles were then allowed to recover for 15 min in flasks containing 2 ml of incubation medium (D-glucose-free DMEM (Gibco) supplemented with 0.22 mg/ml sodium pyruvate and 0.2% BSA), continuously oxygenated with 95% O 2 , 5% CO 2 in a shaking water bath (35uC). After recovery, glucose transport was performed by adding 2-deoxy-D-[ 3 H]glucose (1 mM, 1 mCi/ml) and ( 14 C)mannitol (19 mM, 0.3 mCi/ml) for 20 min and then, muscles were frozen, weighed, and digested in 1 ml of 0.5 N NaOH. Radioactivity was determined by liquid scintillation counting for dual labels and the extracellular and intracellular spaces were calculated as described [36].

Subcellular membrane fractionation
Separation of cardiac sarcolemmal and endosomal membranes from mouse hearts was performed by differential centrifugation as previously described [37]. Successive spins at 1006g, 5,0006g, 20,0006g, 50,0006g and 100,0006g were performed. Membranes pelleting at 20,0006g [plasma membrane fraction, PM] were enriched in the plasma membrane marker Na + /K + -ATPase. To obtain a fraction containing the intracellular GLUT4 pool, the 20,000 g supernatant was centrifuged for 30 min at 50,0006g, resulting in the separation of a high-density microsome fraction [highly contaminated with plasma membranes] and a low-density microsomes (LDM) fraction in the supernatant. This supernatant was finally ultracentrifuged for 60 min at 100,0006g to obtain the LDM pellet. Combining the quantitated signals from LDM and PM fractions indicates that 8365% of Na + /K + ATPase is found in PM fraction, while 7762% of GLUT4 is in LDM (n = 3, p,0.05).

I-insulin binding assays
Insulin binding in suspended cells was measured as described [38]. Human biosynthetic insulin was a kind gift from Eli Lilly Co. (Indianapolis, IN). A 14 -125 I-human insulin was purchased from Perkin-Elmer (Boston, MA). Cells were exposed to a tracer concentration of A 14 -125 I-labeled human insulin and varying concentrations of unlabeled human insulin for 4 hr at 12uC. Reactions were terminated by layering duplicate aliquots of the binding reaction over dibutyl phthalate in microcentrifuge tubes and centrifuging at 14,0006g for 30 sec. The supernatant was aspirated off and the radioactivity in cell pellets determined. Specific binding was calculated by subtracting non-specific binding measured in the presence of a large excess (1.67 mM) of unlabeled insulin.

Weight and DNA content of WAT
Perirenal, perigonadal and subcutaneous adipose tissues were carefully dissected and weighed to determine the fat pad weight as percentage of total body weight. To determine fat pad cell number, perirenal adipose tissue was resected, weighed, and immediately frozen in liquid nitrogen. About 50 mg of tissue were homogenized, and genomic DNA was extracted using Quant-iT PicoGreen (Invitrogen) as described by the manufacturer. DNA was measured by fluorimetric method.

Biochemical analyses
16-h fasted mice were injected intraperitoneally with D-glucose (1 g/kg body weight) and 10 min later, with insulin (5 U/kg body weight). After 10 min, heart, skeletal muscle, adipose tissue and liver were rapidly extracted, freeze clamped in liquid nitrogen and homogenized as described previously [28]. Control tissues from untreated fasted mice were obtained in parallel. Immunoblotting analyses were performed following standard procedures. Antibodies to mouse DMPK were from Zymed Laboratories. Antibodies to insulin receptor b-subunit, caveolin 3 and caveolin 1 were from BD Transduction Laboratories. All other antibodies used were from Cell Signaling Technology. Immunoblots were scanned and signals were quantified using HP Precisionscan Pro and Syngene Gene Tools software. Shown are representative immunoblot data from at least 3 independent experiments which were quantified and expressed as the mean6SEM relative ratio of phosphoprotein to total protein between untreated and insulintreated mice.
Gene transfer by adenovirus vectors Adenoviruses expressing myc-tagged human DMPK were generated by homologous recombination as described [28]. Subconfluent C2C12 cells were infected for 2 h with 100 plaque-forming units/cell of adenovirus vector encoding either myc-DMPK or green fluorescent protein (GFP) before the addition of a suitable volume of myogenic culture media (DMEM with 5% horse serum and antibiotics). After 2 days, myotubes were treated with or without 100 nM insulin for 30 min at 37uC, harvested, lysed and analyzed by immunoblotting as described above.
Cell transfection and confocal microscopy For cell transfection experiments, HeLa cells on coverslips were transiently transfected using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Plasmids used were: myc-tagged wtDMPK and myc-tagged DMADMPK [22]; myc-tagged K110ADMPK [28]; yellow fluorescent protein-tagged insulin receptor (YFP-InsR) [39]; and green fluorescent proteintagged IGF-1 receptor (GFP-IGF-1) [40]. Thirty-six hours after transfection, cells were serum starved for 3 h, then washed in PBS, fixed for 30 minutes in 4% w/v paraformaldehyde and processed for immunofluorescence as described above. Antibodies used were anti-myc rabbit polyclonal antibody (Upstate), anti-myc mAb 9E10 (ATCC), mAb anti-GM130 [Amersham Biosciences], antimouse Cy5-conjugated IgG (Jackson), anti-rabbit Alexa Fluor 555 IgG (Molecular Probes) and anti-mouse Alexa Fluor 594 IgG [Molecular Probes]. For staining of actin filaments, the coverslips were incubated with phalloidin conjugated to Alexa Fluor 594 (Molecular Probes). Cells were examined using a Zeiss LSM510 confocal laser microscope with an oil immersion 636/NA1.3 objective. Micrographs shown are representative optical sections imaged through the centre of the cell. At least 20 cells for each condition of 3 independent experiments were examined and subjected to quantification analyses using Image J software.

Statistical analysis
Data are presented as mean6standard error. Statistical analysis was performed using a two-tailed unpaired t-test. Two-way analysis of variance (ANOVA) was applied for multiple comparisons, followed by the Bonferroni post hoc test. Values of P,0.05 were considered as statistically significant.