Actin Dynamics Regulated by the Balance of Neuronal Wiskott-Aldrich Syndrome Protein (N-WASP) and Cofilin Activities Determines the Biphasic Response of Glucose-induced Insulin Secretion*

Background: Actin dynamics is involved in insulin secretion, but the mechanism is unknown. Results: The G-actin predominant or F-actin remodeling state in pancreatic β-cells, which is regulated by the balance of N-WASP and cofilin activities, determines the biphasic glucose-induced insulin secretion (GIIS). Conclusion: Actin dynamics regulated by N-WASP and cofilin underlie the biphasic GIIS. Significance: The regulation of actin dynamics in β-cells and its role in GIIS are clarified. Actin dynamics in pancreatic β-cells is involved in insulin secretion. However, the molecular mechanisms of the regulation of actin dynamics by intracellular signals in pancreatic β-cells and its role in phasic insulin secretion are largely unknown. In this study, we elucidate the regulation of actin dynamics by neuronal Wiskott-Aldrich syndrome protein (N-WASP) and cofilin in pancreatic β-cells and demonstrate its role in glucose-induced insulin secretion (GIIS). N-WASP, which promotes actin polymerization through activation of the actin nucleation factor Arp2/3 complex, was found to be activated by glucose stimulation in insulin-secreting clonal pancreatic β-cells (MIN6-K8 β-cells). Introduction of a dominant-negative mutant of N-WASP, which lacks G-actin and Arp2/3 complex-binding region VCA, into MIN6-K8 β-cells or knockdown of N-WASP suppressed GIIS, especially the second phase. We also found that cofilin, which severs F-actin in its dephosphorylated (active) form, is converted to the phosphorylated (inactive) form by glucose stimulation in MIN6-K8 β-cells, thereby promoting F-actin remodeling. In addition, the dominant-negative mutant of cofilin, which inhibits activation of endogenous cofilin, or knockdown of cofilin reduced the second phase of GIIS. However, the first phase of GIIS occurs in the G-actin predominant state, in which cofilin activity predominates over N-WASP activity. Thus, actin dynamics regulated by the balance of N-WASP and cofilin activities determines the biphasic response of GIIS.

Insulin secretion from pancreatic ␤-cells plays the central role in maintaining glucose homeostasis; dysfunction of insulin secretion is the major contributor to the pathogenesis and pathophysiology of type 2 diabetes mellitus (1)(2)(3). Insulin secretion is a highly dynamic process regulated by complex mechanisms involving various intracellular signals generated by nutrients and hormonal and neural inputs (4), among which glucose is the most important. Insulin secretion in response to glucose stimulation occurs in a biphasic pattern as follows: an initial component (first phase), which develops rapidly but lasts only a few minutes, followed by a sustained component (second phase) (5)(6)(7). Although the molecular mechanism underlying the biphasic kinetics of glucose-induced insulin secretion (GIIS) 3 is poorly understood, it is thought to include interaction of intracellular signals with distinct insulin granule pools (8). It is generally thought that pancreatic ␤-cells contain at least two pools of insulin secretory granules that differ in release competence as follows: a reserve pool (RP) that accounts for most of the granules, and a readily releasable pool (RRP) that accounts for the other Ͻ5% (9,10). According to the prevailing model of GIIS, the release of insulin granules from RRP comprising predocked granules accounts for the first phase and a subsequent supply of new granules mobilized from RP accounts for the second phase (8). However, we have found that both the first phase and the second phase of GIIS are caused by granules that are newly recruited and immediately fused to the plasma membrane without docking ("restless newcomer") (11). We propose that the first phase results from RRP near the plasma membrane and the second phase from RP located a distance away from the plasma membrane (11,12). It has been suggested that the cortical actin network is also involved in the dynamics of insulin secretion (13,14). It is well known that impaired dynamics of insulin secretion contributes to the pathogenesis and pathophysiology of diabetes (3).
In many eukaryotic cells, actin forms a complex and dynamic network beneath the plasma membrane (15). The rapid remodeling of this network plays an important role in the regulation of exocytosis in secretory cells such as chromaffin cells and exocrine cells and their derived cell lines (16,17). An early study showed that F-actin in pancreatic ␤-cells is organized as a cortical network beneath the plasma membrane (18). The cortical actin network was originally thought to be a physical barrier for insulin granules to reach the plasma membrane (19 -21). The role of F-actin in insulin secretion has so far been studied primarily using pharmacological agents. For example, it was shown that F-actin-depolymerizing agents such as cytochalasin B enhance insulin secretion (18). It was also reported that cytochalasin E and Clostridium botulinum C2 toxin, which are F-actin-depolymerizing agents, decrease insulin secretion (22). More recently, latrunculin B, which binds to G-actin and leads to F-actin depolymerization, was shown to potentiate GIIS (23). It was also reported that F-actin is not essential in the metabolic amplifying pathway of GIIS, using cytochalasin B, latrunculin B, and jasplakinolide (23). These disparate findings on the role of F-actin in insulin secretion may partly be due to limitation of the pharmacological approach. Various small GTPases, including Rho (24), Cdc42 (13,25), and Rac (26) and ezrin, radixin, and moesin proteins (27), are known to be involved in F-actin remodeling in GIIS. It has been shown that Cdc42 is activated by glucose stimulation in MIN6 cells (13,25). Cdc42 has recently been demonstrated to be required for the second phase of GIIS from isolated mouse islets and MIN6 cells (13). Cool-1/ ␤-Pix, one of the guanine nucleotide exchange factors against Cdc42 and Rac1, directly converts Cdc42 to the GTP-bound active form in MIN6 cells in a glucose-dependent manner (28). Cdc42 has a large number of effectors that regulate F-actin remodeling (29). N-WASP links Cdc42 to actin polymerization through the actin-related protein 2/3 (Arp2/3) complex, which promotes actin nucleation (30). Cofilin, which is a downstream signal of Cdc42 and other members of the Rho family, is also an important regulator of actin dynamics (31,32). Cofilin promotes actin depolymerization by severing F-actin (33,34). It has been shown that the coordination of N-WASP and cofilin activities in F-actin remodeling is critical for the regulation of invadopodium assembly and maturation in carcinoma cells (35) and migration of endothelial cells (36). However, little is known of the molecular mechanisms of actin dynamics in pancreatic ␤-cells and its role in phasic insulin secretion.
In this study, we show that interaction of N-WASP activated by Cdc42 and Arp2/3 complex is required for GIIS, especially its second phase. We also show that cofilin, which severs F-actin in its dephosphorylated form (active form), is converted to the phosphorylated form (inactive form) upon glucose stimu-lation, thereby promoting the F-actin remodeling necessary for the second phase of GIIS. In contrast, the first phase of GIIS is dependent mostly on the G-actin-predominant process. Thus, actin dynamics is regulated by the balance of N-WASP and cofilin activities, and it is critical in determining the biphasic response of GIIS.
Small Interfering RNA (siRNA)-For N-WASP and cofilin knockdown experiments, siRNAs (ON-TARGET plus SMART pool and ON-TARGET plus nontargeting pool) were purchased from Dharmacon (Lafayette, CO). MIN6-K8 ␤-cells were transfected with siRNAs using DharmaFECT2 transfection reagent (Dharmacon) according to the manufacturer's instructions.
Pulldown Assay for Cdc42-GTP-To specifically detect the GTP-bound Cdc42, the EZ-Detect Cdc42 activation kit was used. MIN6-K8 ␤-cells were incubated for 24 h in DMEM containing 0.1% FBS. Thirty minutes after preincubation in HEPES-balanced Krebs-Ringer bicarbonate (KRB) buffer (119 mM NaCl, 4.74 mM KCl, 2.54 mM CaCl 2 , 1.19 mM MgCl 2 , 1.19 mM KH 2 PO 4 , 25 mM NaHCO 3 , and 10 mM HEPES, pH 7.4) containing 0.1% BSA with 2.8 mM glucose, the cells were stimulated with HEPES/KRB buffer containing 16.7 mM glucose or 60 mM K ϩ for 3 and 15 min. The cell lysates were incubated with 20 g of GST-PAK1-PBD-agarose for 1 h at 4°C. The bound proteins were also subjected to SDS-PAGE, followed by Western blot analysis with anti-Cdc42 antibody. The relative abundance of eluted Cdc42 was determined by densitometry.
Coimmunoprecipitation Experiments-Coimmunoprecipitation experiments were performed using ExactaCruz (Santa Cruz Biotechnology) according to the manufacturer's instructions. One hour after preincubation, MIN6-K8 ␤-cells were treated with HEPES/KRB buffer containing 2.8 or 16.7 mM glucose for 3 and 15 min. The cell lysates were incubated with anti-Cdc42 antibody conjugated with agarose beads. After incubation at 4°C for 1 h, the beads were washed three times with PBS, and the bound proteins were subjected to SDS-PAGE, followed by Western blot analysis with anti-N-WASP and anti-PAK1 antibodies.
Insulin Secretion Experiments-Mouse pancreatic islets were isolated from 12-to 16-week-old C57BL/6J male mice by collagenase digestion and cultured for 2 days as described previously (40). Insulin released in the incubation buffer and cellular insulin content in mouse islets and MIN6-K8 ␤-cells were measured by insulin assay kits from Medical Biological Laboratories (Nagoya, Japan) and CIS Bio international (Gif sur Yvette, France), respectively. The amount of insulin secretion was normalized by cellular insulin content.
TIRFM Analysis of Actin Dynamics-Primary cultured ␤-cells isolated from mouse pancreatic islets were infected with adenovirus carrying AcGFP1-actin together with mCherry, mCherry-N-WASP WT, or mCherry-N-WASP DN for 2 days. After 30 min of preincubation with 2.8 mM glucose, the fluorescence intensity of AcGFP1-actin near the plasma membrane was monitored using TIRFM (IX71, Olympus, Tokyo, Japan) in the cells stimulated with 16.7 mM glucose.
TIRFM Analysis of Insulin Granule Exocytosis-Primary cultured ␤-cells isolated from mouse pancreatic islets were infected with adenovirus carrying insulin-Venus together with mCherry-N-WASP WT or mCherry-N-WASP DN for 2 days and were subjected to TIRFM analysis as described previously (11). The number of insulin-Venus-containing granules docked to the plasma membrane was measured by G-Count software (G-Angstrom K.K, Miyagi, Japan) as described previously (41).
Immunocytochemical Analysis-MIN6-K8 ␤-cells were infected with adenovirus carrying mCherry-CAAX, which targets to the plasma membrane, and cultured for 2 days. After 30 min of preincubation, MIN6-K8 ␤-cells were stimulated by 16.7 mM glucose for 3 and 15 min or 60 mM K ϩ for 3 min. The cells were fixed in 3.7% formaldehyde and treated with 0.2% Triton X-100 and 10% normal donkey serum. They were incubated with rabbit anti-N-WASP antibody at 4°C overnight, followed by Alexa Fluor488-labeled donkey anti-rabbit IgG antibody at room temperature for 30 min. After immunostaining, the cells were counterstained with NucRed Live 647 ReadyProbe reagent at room temperature for 30 min. The stained cells were observed by confocal laser scanning microscopy (FV1000, Olympus).
mRNA Expression-Total RNA from MIN6-K8 ␤-cells was isolated using the RNeasy kit (Qiagen, Hilden, Germany). For reverse transcription, ReverTra Ace-␣-kit (Toyobo, Osaka, Japan) was used. Real time PCR was performed using TaqMan probe with a model 7000 real time thermal cycler from Applied Biosystems (Foster City, CA). Measurement of the expression levels of hypoxanthine phosphoribosyltransferase was used as internal control.
Perifusion Experiments-Perifusion experiments on insulin secretion of MIN6-K8 ␤-cells were performed as described previously (42). Briefly, the cells were seeded on coverslips. The following day, the cells were infected with adenovirus carrying mCherry, mCherry-N-WASP WT, mCherry-N-WASP DN, cofilin WT, or cofilin S3D or transfected with ON-TARGET plus nontargeting pool siRNA (control) or ON-TARGET plus SMART pool siRNAs against N-WASP and cofilin. After 2 days, the cells were perifused in HEPES/KRB buffer containing 2.8 mM glucose for 5 min, and the perfusate was then switched to HEPES/KRB buffer containing 16.7 mM glucose. Each fraction was collected at 1-min intervals, and the released insulin was measured by insulin assay kit (CIS Bio international). The amount of secreted insulin was normalized by cellular insulin content.
Statistical Analysis-Data are presented as mean Ϯ S.E. Statistical analysis was performed by unpaired, two-tailed Student's t test or one-or two-way analysis of variance with Tukey's post hoc test. Differences between groups were considered statistically significant when p Ͻ 0.05.

Glucose Activates N-WASP, a Downstream Target of
Cdc42-Rho GTPase Cdc42 is known to participate in actin dynamics in many eukaryotic cells. To clarify the involvement of Cdc42-mediated F-actin remodeling in GIIS, we first examined the effect of glucose on Cdc42 activation in MIN6-K8 ␤-cells. The level of Cdc42-GTP was very low at basal state (2.8 mM glucose). Cdc42-GTP was increased 3 min after glucose stimulation (16.7 mM) in MIN6-K8 ␤-cells and remained at the same level 15 min after stimulation (Fig. 1A). In contrast, Cdc42 was not activated by 60 mM K ϩ stimulation (Fig. 1B). Nifedipine, an L-type voltage-dependent Ca 2ϩ channel blocker, did not inhibit glucose-activated Cdc42 in MIN6-K8 ␤-cells (Fig. 1C).
These results indicate that Cdc42 in pancreatic ␤-cells is activated by glucose metabolism, but not by Ca 2ϩ .
Because N-WASP is a downstream target of Cdc42 in PC12 cells, COS-7 cells, and breast carcinoma cells (43), we investigated possible glucose inducement of interaction of Cdc42 and N-WASP in MIN6-K8 ␤-cells. The interaction of N-WASP and Cdc42 was significantly increased 1 min after glucose stimulation, and the increased interaction was maintained during the Fig. 1D, upper panel). The interaction was not enhanced by 60 mM K ϩ stimulation (Fig. 1D, lower left). These results indicate that glucose simulation promotes interaction of Cdc42 with N-WASP in pancreatic ␤-cells (Fig. 1D, lower right).

Critical Role of Actin Dynamics in Biphasic Insulin Secretion
Interaction of N-WASP and Cdc42 induces a conformational change of N-WASP, leading to its activation (30). It has been shown that N-WASP at tyrosine 253 (Tyr-253) is phosphorylated by Src family kinases, which induce a conformation change of N-WASP that is important for its activation (44). Indeed, we found that N-WASP at Tyr-253 was phosphorylated at a low level at basal state in MIN6-K8 ␤-cells and that the phosphorylation was significantly decreased by PP2, a specific inhibitor of the Src family kinases (Fig. 1E, upper panel). We also found that the phosphorylation was not increased either 3 or 15 min after glucose stimulation (Fig. 1E, lower panel). These results indicate that N-WASP is not activated by phosphorylation during glucose stimulation.
We then examined subcellular localization of endogenous N-WASP in MIN6-K8 ␤-cells and found it to be diffusely localized in MIN6-K8 ␤-cells at basal state, whereas it was colocalized with mCherry-CAAX, a plasma membrane marker, under glucose stimulation (Fig. 1F). Stimulation by 60 mM K ϩ did not alter subcellular localization of N-WASP. Thus, N-WASP is translocated to the ␤-cells plasma membrane by glucose stimulation and not by Ca 2ϩ . This is consistent with the finding that interaction of N-WASP and Cdc42 is induced by glucose but not by Ca 2ϩ .
N-WASP Mediates Glucose-induced Actin Dynamics-Although changes in the cortical actin network have been studied by morphological analyses of fixed cell preparations, there has been no report of visualization of actin dynamics in living pancreatic ␤-cells. We visualized N-WASP-mediated actin dynamics beneath the plasma membrane in pancreatic ␤-cells by using G-actin fused to green fluorescent protein AcGFP1 (AcGFP1actin). A previous study of fixed ␤-cells using phalloidin-fluorescein isothiocyanate (FITC), which labels F-actin specifically, suggested that glucose induces F-actin remodeling beneath the plasma membrane of MIN6 cells (45). To determine whether N-WASP mediates actin polymerization in ␤-cells, we utilized a dominant-negative mutant N-WASP lacking the VCA region (N-WASP DN), which cannot bind to either the G-actin or Arp2/3 complex ( Fig. 2A). We introduced N-WASP DN together with AcGFP1-actin into primary cultured mouse pan-creatic ␤-cells by adenovirus-based gene transfer. Actin dynamics was then visualized by monitoring the fluorescence intensity of AcGFP1-actin detected within 40 nm from the cell surface, by using TIRFM (Fig. 2B). The fluorescence intensity of AcGFP1-actin near the plasma membrane was enhanced in mCherry-introduced ␤-cells (control) and in N-WASP WTintroduced ␤-cells in the presence of 16.7 mM glucose, whereas introduction of N-WASP DN into ␤-cells completely blocked enhancement of the fluorescence intensity (Fig. 2B). These results suggest that the interaction of N-WASP and Arp2/3 complex mediates glucose-induced actin dynamics beneath the plasma membrane in pancreatic ␤-cells.
N-WASP Regulates the Second Phase of Glucose-induced Insulin Secretion-Glucose induces insulin secretion in a biphasic manner. To clarify the role of N-WASP-mediated actin polymerization in biphasic insulin secretion, we performed perifusion experiments using MIN6-K8 ␤-cells (Fig. 4,  A and B). Knockdown of N-WASP suppressed GIIS, especially in second phase (Fig. 4A). In addition, there was no significant difference in biphasic insulin secretion between mCherry-introduced MIN6-K8 ␤-cells and N-WASP WT-introduced MIN6-K8 ␤-cells. The first phase of GIIS in N-WASP DN-in- troduced MIN6-K8 ␤-cells was similar to that in N-WASP WTintroduced cells (Fig. 4B). In contrast, the second phase of GIIS in N-WASP DN-introduced MIN6-K8 ␤-cells was significantly reduced, compared with that in mCherry-introduced cells or N-WASP WT-introduced cells (Fig. 4B). We further examined the role of N-WASP-induced actin polymerization in insulin granule exocytosis in primary cultured mouse pancreatic ␤-cells (Fig. 4, C-F), using the TIRFM system (11). Insulin granule exocytosis occurs in three different modes based on the dynamics of the insulin granules as follows: 1) old face, granules that are predocked to the plasma membrane and fused to the membrane by stimulation; 2) restless newcomer, granules that are newly recruited and immediately fused to the plasma membrane by stimulation without docking; and 3) resting newcomer, granules that are newly recruited, docked, and fused to the plasma membrane by stimulation. There was no significant difference in the number of docked granules between N-WASP WT-introduced ␤-cells and N-WASP DN-introduced ␤-cells (Fig. 4C). Both phases of glucose-induced insulin granule exocytosis were caused mainly by restless newcomer in N-WASP WT-introduced ␤-cells (Fig. 4 D, upper panel, and F). The number of fusion events in the first phase in N-WASP DN-introduced ␤-cells was similar to that in N-WASP WT-introduced ␤-cells (Fig. 4, D-F). In contrast, the number of fusion events in the second phase was significantly reduced in N-WASP DNintroduced ␤-cells (Fig. 4, D-F), a finding consistent with those obtained by perifusion experiments (Fig. 4B). These results demonstrate that actin polymerization regulated by Cdc42/N-WASP signal is required in the second phase of GIIS by facilitating recruitment of restless newcomer.
Glucose Enhances Phosphorylation of Cofilin That Initiates F-actin Remodeling-Because PAK1 is also a downstream target of Cdc42 (43), we examined the effect of glucose stimulation on the interaction of Cdc42 and PAK1 in MIN6-K8 ␤-cells. The interaction was significantly enhanced 3 and 15 min after glucose stimulation (Fig. 5A, left and middle panels), although the interaction was not enhanced by 60 mM K ϩ stimulation (Fig.  5A, right panel). PAK1 activates LIM kinase 1 (LIMK1) by its phosphorylation (46), which phosphorylates cofilin at serine 3. Phosphorylation of cofilin (inactive form) inhibits its actin depolymerizing activity by binding to actin monomers (47,48). Dephosphorylated cofilin (active form) increases the rate of actin depolymerization and maintains a pool of actin monomer (33). We found that phosphorylation of LIMK1 is induced by glucose stimulation but is not phosphorylated by K ϩ stimulation (Fig. 5B). We therefore examined whether glucose stimulation (16.7 mM) increases phosphorylation of cofilin in MIN6-K8 ␤-cells. The level of phosphorylated cofilin was grad-  ually increased by glucose stimulation in a time-dependent manner (Fig. 5C, left panel). In contrast, the level of phosphorylated cofilin was decreased by K ϩ stimulation (Fig. 5C, right  panel). Cofilin is also known to be phosphorylated at serine 3 by LIMK2, which is an isoform of LIMK1 (46). LIMK2 is activated by its phosphorylation of Rho kinase ROCK (49). However, it is not known whether cofilin is phosphorylated by ROCK/LIMK2 in pancreatic ␤-cells stimulated with glucose. We examined  mRNA expression levels of LIMK1 and LIMK2 in MIN6-K8 ␤-cells by real time PCR analysis. We found that LIMK1 and LIMK2 are equally expressed in MIN6-K8 ␤-cells (Fig. 5D). We then examined whether ROCK/LIMK2 phosphorylates cofilin in a glucose-dependent manner. For this purpose, we utilized ROCK inhibitor Y27632 and PAK1 inhibitor IPA3 (Fig. 5E,  upper panel). Phosphorylation levels of cofilin were inhibited by IPA3 in MIN6-K8 ␤-cells stimulated with glucose (16.7 mM), whereas they were not inhibited by Y27632, suggesting that cofilin is not phosphorylated by ROCK/LIMK2 (Fig. 5E, lower  left). These results indicate that glucose stimulation induces phosphorylation and inactivation of cofilin through activation of Cdc42/PAK1/LIMK1 signaling (Fig. 5E, lower right). Cofilin Regulates the Second Phase of Glucose-induced Insulin Secretion-To clarify the role of cofilin in insulin secretion, we performed insulin secretion experiments, using cofilin siRNA. Knockdown of cofilin in MIN6-K8 ␤-cells significantly reduced GIIS (Fig. 6A). To clarify the role of cofilin in dynamics of GIIS, we performed perifusion experiments using MIN6-K8 ␤-cells. The first phase of GIIS in cofilin knockdown MIN6-K8 ␤-cells was similar to that in control cells (Fig. 6B). In contrast, the second phase of GIIS in cofilin knockdown MIN6-K8 ␤-cells was significantly reduced, compared with that in control cells (Fig. 6B).
We next examined whether cofilin activity is involved in insulin secretion. For this purpose, we utilized cofilin S3D mutant-mimicking phosphorylated form (i.e. inactive form) (38). Cofilin S3D mutant competitively acts on the binding of phosphorylated endogenous cofilin to its specific phosphatase, thereby inactivating endogenous cofilin, which inhibits actin polymerization (50). In cofilin S3D-introduced MIN6-K8 ␤-cells, GIIS was significantly reduced, compared with that in  cofilin WT-introduced MIN6-K8 ␤-cells (Fig. 6C). The effect of cofilin S3D mutant on the dynamics of GIIS was also examined by perifusion experiments. The first phase of GIIS in cofilin S3D-introduced MIN6-K8 ␤-cells was similar to that in cofilin WT-introduced MIN6-K8 ␤-cells, whereas the second phase in cofilin S3D-introduced MIN6-K8 ␤-cells was decreased signif-  icantly (Fig. 6D), indicating that activity of cofilin is critical for the second phase of GIIS.

DISCUSSION
Actin dynamics has recently emerged as a key event in vesicle trafficking in exocytosis and endocytosis (17,51). In chromaffin cells and their derived PC12 cells, actin forms a complex and dynamic network of filaments beneath the plasma membrane, and rapid remodeling of the actin network is required for exocytosis of catecholamine (52). In 3T3-L1 adipocytes, remodeling of cortical actin at the plasma membrane is a necessary and rate-limiting step for fusion of GLUT4 vesicles (53). Although roles of the actin network in insulin secretion from pancreatic ␤-cells have been investigated by pharmacological and morphological approaches, the dynamics of actin in living pancreatic ␤-cells has not been described. In this study, we investigated the mechanism of glucose regulation of actin dynamics regulated by N-WASP and cofilin, which are the major regulators of actin dynamics, and its role in the biphasic response of GIIS. We first established a system using a combination of AcGFP1-labeled actin and TIRFM to visualize actin dynamics of living ␤-cells. We found that glucose induces actin dynamics in primary cultured pancreatic ␤-cells that is completely blocked by inactivation of N-WASP. In PC12 cells, the Ca 2ϩ signal initiates the interaction of Cdc42 and N-WASP, leading to actin polymerization and triggering of catecholamine release (54). However, the Ca 2ϩ signal in pancreatic ␤-cells does not activate Cdc42 and does not induce interaction of Cdc42 with N-WASP or PAK1. In contrast, glucose induces interaction of Cdc42 with N-WASP or PAK1. Cdc42 activation also requires glucose metabolism (13). Thus, glucose metabolism rather than the Ca 2ϩ signal is essential for activation of the Cdc42/N-WASP/Arp2/3 complex signal and the Cdc42/PAK1/LIMK1/ cofilin signal in pancreatic ␤-cells.
Phosphorylation of N-WASP at tyrosine 253 by Src family kinases has been shown to cause its activation in neuronal cells (55). However, our data reveal that N-WASP is not activated by phosphorylation at tyrosine 253 by glucose stimulation in MIN6-K8 ␤-cells. We also found that Cdc42 activates PAK1 by glucose stimulation. It has been shown that activated PAK1 phosphorylates LIMK1 (46) and that phosphorylated LIMK1 phosphorylates cofilin (i.e. inactivates cofilin), which decreases its actin depolymerizing activity (47,48). In addition, we found that phosphorylations of LIMK1 and cofilin are induced by glucose stimulation in MIN6-K8 ␤-cells.
Because Cdc42 binds to N-WASP and PAK1, the three proteins might form a complex. However, it has been shown that WASP and PAK bind to the same region in Cdc42 (56), suggesting that the complex cannot be formed. The distinct pools comprising Cdc42/N-WASP/Arp2/3 complex signal and Cdc42/ PAK1/LIMK1/cofilin signal could well be involved in control of glucose-induced actin dynamics.
We recently found that both the first phase and the second phase of GIIS are caused by insulin granules that are newly recruited and immediately fused to the plasma membrane without docking (restless newcomer) (11,41), but the mechanisms underlying biphasic insulin secretion remain unclear. At basal state (2.8 mM glucose, nonstimulated condition), most of the cofilin is present in its dephosphorylated form (active form), which severs F-actin. As a result, at basal state, G-actin predominates over F-actin. Glucose stimulation drastically alters the states of G-actin and F-actin by changing cofilin and N-WASP activities in a time-dependent manner. In the early period (Ͻ ϳ5 min) during glucose stimulation, which corresponds to the first phase of GIIS, the level of phosphorylated cofilin (inactive form) is gradually increased, but the majority of cofilin is still dephosphorylated in this period. The activity of cofilin therefore predominates over the activity of the N-WASP/Arp2/3 complex, thereby severing the remodeled F-actin and resulting in a G-actin predominant state. In accord with this, a recent study reports that glucose stimulation (20 mM) decreases the amount of F-actin near the plasma membrane within 5 min, compared with that at basal state (glucose-free), and recovers it in 30 min (45). Considered together, these findings suggest that, in contrast to the later period, insulin granules in first phase glucose stimulation undergo undirected trafficking from RRP, independent of the actin network, and are immediately fused to the plasma membrane. This may account for the characteristic dynamics of first phase of GIIS, which occurs rapidly, massively, and transiently. Indeed, actin-depolymerizing agents such as cytochalasin B and latrunculin B sustain massive insulin secretion from perifused pancreatic islets during application (23,45,57). It is also established that insulin is immediately, massively, and transiently secreted by K ϩ stimulation (6,8). We found that K ϩ stimulation did not cause activation of Cdc42 in pancreatic ␤-cells, confirming a previous report (25), and that N-WASP was not activated by K ϩ stimulation. Interestingly, in contrast to the effect of glucose stimulation, the level of phosphorylated cofilin was decreased significantly by K ϩ stimulation, compared with its basal level, indicating that the F-actin-severing activity of cofilin is enhanced during stimulation. These findings indicate that a G-actin-predominant state is associated closely with immediate, massive, and transient response of insulin granule exocytosis. However, in the late period (Ͼ ϳ5 min) during glucose stimulation, which corresponds to the second phase of GIIS, the level of phosphorylated cofilin reaches nearly maximum 5 min after stimulation and is sustained thereafter. In contrast, N-WASP is rapidly activated, and its activity is sustained throughout glucose stimulation. Thus, activity of the N-WASP/Arp2/3 complex signal predominates over that of cofilin in the late period, thereby effecting F-actin remodeling. Indeed, introduction of an N-WASP DN into MIN6-K8 ␤-cells or N-WASP knockdown disclosed a marked decrease of the second phase of GIIS, as assessed by perifusion experiments. These findings are consistent with the recent studies showing that the second phase of GIIS was decreased in perifused Cdc42 knockdown islets and islets isolated from PAK1 knock-out mice (13,58). A cofilin DN is known to inhibit the supply of G-actin that is essential for actin polymerization by blocking endogenous cofilin (50). We found that introduction of cofilin DN into MIN6-K8 ␤-cells also decreases the second phase of GIIS. Given that both N-WASP DN and cofilin DN inhibit the second phase of GIIS, F-actin remodeling is clearly regulated by activities of N-WASP and cofilin in pancreatic ␤-cells and is essential for the second phase of GIIS. The second phase of GIIS is caused almost exclusively by restless newcomer (11); F-actin remodeling might control or otherwise facilitate trafficking of these insulin granules to the plasma membrane, where fusion events can occur.
We propose a model of the relationship between actin dynamics and the biphasic response of GIIS (Fig. 7). The first phase of GIIS is associated with insulin secretion mostly by the G-actin predominant process, in which the Cdc42/PAK1/ LIMK1/cofilin signal is predominant. However, the second phase of GIIS is associated mostly with insulin secretion caused by the F-actin remodeling predominant process, in which the Cdc42/N-WASP/Arp2/3 complex signal is predominant. Thus, actin dynamics regulated by the balance between N-WASP and cofilin activities is critical in determining the biphasic response of GIIS.
As impaired dynamics of insulin secretion is a characteristic feature of diabetes (59,60), it is important to clarify the roles of N-WASP and cofilin in the pathogenesis and pathophysiology of diabetic ␤-cells. FIGURE 7. Model for the relationship between actin dynamics and biphasic insulin secretion by glucose. A, glucose simulation promotes interaction of Cdc42 with both PAK1 and N-WASP and activates them through glucose metabolism in pancreatic ␤-cells. PAK1 activates LIMK1 by its phosphorylation. Activated LIMK1 phosphorylates and inactivates cofilin, which severs F-actin. N-WASP induces actin polymerization through its interaction with the Arp2/3 complex. Glucose stimulation triggers F-actin remodeling through Cdc42/PAK1/LIMK1/cofilin signal and Cdc42/N-WASP/Arp2/3 complex signal. B, glucose induces insulin secretion by G-actin predominant process (yellow) and F-actin remodeling predominant process (blue). In the early period (Ͻ ϳ5 min) after glucose stimulation, cofilin activity, which severs F-actin, predominates over N-WASP activity. In this period, insulin secretion from readily releasable pool (RRP) near the plasma membrane occurs rapidly, massively, and transiently. In the late period (Ͼ ϳ 5 min), N-WASP activity predominates over cofilin activity, thereby effecting the F-actin remodeling that facilitates trafficking of insulin granules from RP to the plasma membrane. G-actin predominant process and F-actin remodeling predominant process occur concurrently after glucose stimulation. The first phase of GIIS consists of the sum of insulin secretion due to both G-actin predominant process and F-actin remodeling predominant process, the former being major. However, the second phase of GIIS consists primarily of insulin secretion caused by the F-actin remodeling predominant process.