Myosin Va mediates Rab8A-regulated GLUT4 vesicle exocytosis in insulin-stimulated muscle cells

Interaction of MyoVa with Rab8A, Rab10, and Rab13

Of the MyoV isoforms, only MyoVa was detected in rat L6 myoblast or myotube cDNA by semiquantitative PCR, with no detectable signals for either myosin Vb or Vc (Supplemental Figure S1). Both the –D and +D mRNA splice variants are expressed in both myoblasts and myotube stages of the culture. MyoVa is also expressed in mouse C2C12 muscle cells, 3T3-L1 adipocytes, and mouse and rat skeletal muscles (Supplemental Figure S1). MyoVb was not detectable in any of the foregoing cells or tissues, and MyoVc was only observed in mouse muscle but not in rat muscle or the rat muscle cell line; a trace amount of MyoVc was observed in 3T3-L1 adipocytes and C2C12 myoblasts (Supplemental Figure S1).

MyoVa is an actin-based motor, composed of an N-terminal (head) motor domain and a C-terminal (tail) Rab-binding domain (Figure 1A). Using the +D splice variant cDNA, we created a truncated MyoVa–glutathione S-transferase (GST) fusion protein containing the Rab-binding domain (amino acids 1260–1854; GST–MyoVa-CT) and tested its ability to interact with GFP fusion proteins of Rabs 8A, 10, and 13 transfected into CHO cells stably expressing the insulin receptor (CHO-IR cells) in basal and insulin-stimulated conditions (100 nM, 15 min). CHO-IR cells were chosen for these biochemical experiments, given their proven sensitivity to insulin and their high yield of transfectability necessary to perform pull-down experiments. GST–MyoVa-CT pulled down Rab8A and Rab10 but not Rab13, and only the interaction of Rab8A with MyoVa-CT was elevated (greater than twofold) in an insulin-dependent manner (Figure 1B). Insulin stimulates GTP loading of Rab8A but not Rab10 (Sun et al., 2010), and therefore we tested the ability of GST–MyoVa-CT to interact with wild-type (wt), activated (GTP-locked) or inactivated (GDP-locked) mutants of Rab8A expressed in CHO-IR cells. Figure 1C shows that GST–MyoVa-CT bound wt and GTP-locked mutant Rab8A but not GDP-locked mutant Rab8A. Although a nonspecific band appeared in all input samples that barely separated from GFP-Rab8Awt and more clearly separated from the GTP- or GDP-locked mutants (Figure 1C gel), this band was not present in the pull-down material (Figure 1C, bar graph). Binding of Rab8A-GTP was 2.5-fold higher than that of wt Rab8A when normalized for total GFP-Rab fusion protein expression (Figure 1C), consistent with the hypothesis that MyoVa is an effector of active Rab8A.

Insulin increases Rab8A–MyoVa interaction downstream of PI3K

To examine the MyoVa–Rab8A interaction dependence on insulin signaling in muscle cells, we tested the effect of wortmannin, a PI3K inhibitor that blocks activation of Akt, on GFP-Rab8A pull down by GST–MyoVa-CT. GFP-Rab8A was expressed in CHO-IR cells to equivalent levels in all conditions before stimulation with insulin (100 nM, 15 min) in the absence or presence of wortmannin (100 nM, with 30-min pretreatment). Binding of GFP-Rab8A to GST–MyoVa-CT was 1.9 times greater in lysates of cells treated with insulin than with those of unstimulated, dimethyl sulfoxide (DMSO)–treated control cells, and wortmannin effectively eliminated the insulin-induced response (Figure 1D). These results are consistent with a role of the PI3K axis in regulating the GTP-loaded levels of Rab8A (Sano et al., 2003; Miinea et al., 2005; Ishikura et al., 2007; Zaid et al., 2008).

MyoVa colocalizes with Rab8A in L6 myoblasts

To study the reciprocal localization of MyoVa and Rab8A in muscle cells, we expressed them with individual fluorescent tags into L6 muscle cells. We are aware of the caveat that the information obtained pertains to overexpressed proteins rather than the endogenous ones, but this is a common practice, as regrettably there are no useful antibodies to detect the latter. By confocal fluorescence microscopy of L6 myoblasts cotransfected with full-length (FL) GFP-MyoVa and mCherry-Rab8A (MC-Rab8A), the two proteins showed a very similar punctate perinuclear distribution in untreated (basal) cells (Figure 2A), which suggested a high degree of colocalization. Some GFP-MyoVa fluorescence spread into the cell periphery, where very little MC-Rab8A was detected and barely overlapped with GFP-MyoVa. The ratio of peripheral to total signal of MC-Rab8A fluorescence intensity was quantified in 10–15 cells for basal and insulin-stimulated states and compared with that in control cells expressing GFP plus MC-Rab8A (Figure 2C). Insulin stimulation significantly promoted the distribution of Rab8A to the cell periphery in cells overexpressing GFP-MyoVa (but not in those expressing only GFP), suggesting that the excess motor protein sufficed to mobilize the GTPase toward the cell periphery under the influence of insulin. This resulted in increased colocalization of GFP-MyoVa with MC-Rab8A in the peripheral cytoplasmic region (region of interest excluded the area of intense perinuclear fluorescence of GFP–MyoVa-FL; Figure 2, A, B, and D).

The MyoVa-CT–Rab8A interaction mislocalizes Rab8A and reduces insulin-dependent GLUT4 translocation

Because full-length MyoVa and Rab8A showed partial colocalization and the GST–MyoVa-CT fragment was able to pull down Rab8A, we next examined whether MyoVa-CT and Rab8A can interact in situ. GFP–MyoVa-CT was coexpressed with MC–Rab8A-wt, -GTP, or -GDP constructs in CHO-IR cells (used here again due to their high rate of transfection), and complexes were immunoprecipitated with anti-mCherry antibody. MC-Rab8A coprecipitated with GFP–MyoVa-CT, and, of note, the magnitude of this response was higher in the insulin-stimulated state (Figure 3A). Further, GFP–MyoVa-CT coprecipitated MC–Rab8A-GTP, whereas there was only minimal coprecipitation of MC–Rab8A-GDP above the background level observed when using control immunoglobulin G instead of anti-mCherry antibody (Figure 3B). These findings are consistent with the GST pull-down results in Figure 1.

The interaction of GFP–MyoVa-CT with Rab8A was next verified in L6 muscle cells. On confocal fluorescence microscopy of L6 myoblasts, MC–Rab8A-wt displayed a typical, dispersed vesicle–like, perinuclear distribution when coexpressed with GFP (Figure 3C). This disperse, small punctate morphology was observed in all cells transfected with GFP and MC-Rab8A (50 cells imaged). When cotransfected with GFP–MyoVa-CT, however, MC-Rab8A was considerably mislocalized into vastly enlarged perinuclear depots that colocalized with GFP–MyoVa-CT (Figure 3C). This pattern was observed in 76% of cells imaged (38 of 50 cells), and similarly collapsed depots but of somewhat smaller size were observed in the remaining 12 of 50 cells. Because MyoVa-CT lacks its motor domain and cannot interact with actin filaments, it probably interacts with MC-Rab8A on vesicles or compartments but fails to connect them with actin filaments (Wu et al., 2001). In contrast, there was no obvious change in the distribution of MC-Rab13 when coexpressed with either GFP or GFP–MyoVa-CT in muscle cells (Supplemental Figure S2), consistent with the inability of GFP-Rab13 to bind GST–MyoVa-CT (Figure 1B).

Given that Rab8A is required for insulin-dependent GLUT4 translocation in muscle cells (Ishikura et al., 2007; Ishikura and Klip, 2008; Sun et al., 2010) a prediction of the mislocalization of Rab8A by GFP–MyoVa-CT is that the latter would impair insulin-dependent GLUT4 translocation. Indeed, the insulin-induced gain in GLUT4myc at the muscle cell surface was vastly reduced in cells expressing GFP–MyoVa-CT (Figure 3D). Thus Rab8A and MyoVa-CT can interact in situ, and this leads to abnormal localization of Rab8A and inhibition of insulin-stimulated translocation of GLUT4 to the muscle cell surface.

Mutations on MyoVa-CT largely reduce the interaction with Rab8A and restore insulin-stimulated GLUT4myc translocation

We hypothesized that the inhibition of insulin-stimulated GLUT4myc translocation by MyoVa-CT is related to its ability to interact with and mislocalize Rab8A. Thus we sought to identify the binding sites for Rab8A on MyoVa-CT. Goldenring and coworkers (Roland et al., 2011) showed that binding of Rab8A to another MyoV isoform, MyoVb, requires Q1300 and Y1307, but the sites on MyoVa are unknown. Aligning the C-terminal sequences of MyoVa and MyoVb revealed potential equivalence of DVQ¹³⁰⁰ in MyoVa to VDQ¹³⁰⁰ in MyoVb and of AY¹³¹²IG in MyoVa with AY¹³⁰⁷HG in MyoVb. Moreover, Q1300 and Y1312 in MyoVa were conserved in mice, rats, and humans (Supplemental Figure S3); therefore, we introduced the mutations Q1300L (M1) and Y1312C (M2) on MyoVa-CT to generate the double mutant (2M) in both GST and GFP expression vectors of MyoVa-CT. In pull-down assays with CHO-IR cell lysates, very little transfected GFP-Rab8A associated with GST–MyoVa-CT(2M) compared with GST–MyoVa-CT (Figure 4A), suggesting the Rab8A-MyoVa interaction requires Q1300 and Y1312 in MyoVa. In contrast, GST–MyoVa-CT(2M) bound transfected GFP-Rab10 to a similar extent as did GST–MyoVa-CT. These results indicate that Q1300 and Y1312 are required to bind Rab8A but not Rab10 to MyoVa (Supplemental Figure S4). In line with the inability of GST–MyoVa-CT(2M) to bind Rab8A in CHO-IR cell lysates, coexpression of GFP-MyoVa(2M) with MC-Rab8A in muscle cells did not cause MC-Rab8A proteins to aggregate as did GFP–MyoVa-CT (Figure 4B).

The influence of GST–MyoVa-CT(2M) on GLUT4 traffic was then tested. As before, insulin promoted a GLUT4myc translocation to the cell surface (by 2.7-fold) in control L6 muscle cells expressing GFP that was markedly impaired by transfected GFP–MyoVa-CT, whereas GFP–MyoVa-CT(2M) allowed an almost complete (88%) insulin-stimulated GLUT4myc response (Figure 4C). These experiments strongly suggest that preventing Rab8A binding to the MyoVa-CT fragment in situ eliminates the ability of MyoVa-CT to interfere with GLUT4 traffic to the cell surface. Of note, MyoVa-CT(2M) could still bind Rab10 like the nonmutated MyoVa-CT fragment, suggesting that this GTPase is not responsible for the differential effect of these fragments on GLUT4 translocation.

Given that the binding of MyoVa-CT to Rab8A correlated with the ability of MyoVa-CT to interfere with GLUT4myc translocation to the cell surface, these findings implicate MyoVa as an effector of Rab8A required for insulin-stimulated GLUT4 traffic.

Rab8A colocalizes with GLUT4 in perinuclear regions but not in the TIRF zone of muscle cells

The foregoing results indicate that MyoVa interacts with Rab8A and that this interaction is required for GLUT4 translocation to the plasma membrane. To begin to address the cellular locus where the input of Rab8A:MyoVa takes place during the itinerary of GLUT4, we examined the subcellular localization of Rab8A vis-à-vis GLUT4. Using spinning-disk confocal microscopy, we found GFP-GLUT4 and MC-Rab8A to partially colocalize at the perinuclear region in both basal and insulin pretreated states (Figure 5A). By this approach, Rab8A is not detected near the cell surface; however, L6 myoblasts are relatively flat cells, and hence we examined the localization of GFP-GLUT4 and MC-Rab8A within ∼100 nm of the plasma membrane using two-color TIRF microscopy (Mattheyses et al., 2010). Imaging fixed muscle cells, we readily detected GFP-GLUT4 in the TIRF zone (Figure 5B), where we previously characterized its mobility and response to insulin (Boguslavsky et al., 2012). Although molecules on GLUT4 vesicles can be detected in the TIRF region, such as the nonprocessive myosin 1c (Boguslavsky et al., 2012), Rab8A was not present (Figure 5B), despite prominent overlap with GLUT4 in perinuclear regions. Based on the predominantly perinuclear colocalization of Rab8A with MyoVa-FL (Figure 2A), the higher overlap of Rab8A with MyoVa at the cell periphery after insulin stimulation (Figure 2B and C), and the perinuclear colocalization of Rab8A with GLUT4 (Figure 5A; Sun et al., 2010), we envisage that the Rab8A:MyoVa complex directs GLUT4 traffic before it reaches the cell periphery.

A prediction of this scenario is that GLUT4 vesicles would be mobilized along with MyoVa, possibly positioned along actin filaments. We therefore determined the position of red fluorescent protein (RFP)–GLUT4 vis-à-vis full-length GFP-MyoVa in insulin-stimulated L6 myoblasts over time, using real-time imaging. Figure 5C shows several time-lapse images of the localization of the transporter and the motor protein and illustrates that RFP-GLUT4 puncta—likely vesicles—organize sequentially along cytosolic GFP-MyoVa structures.

Knockdown of MyoVa reduces insulin-stimulated GLUT4myc translocation

To ascertain the functional contribution of MyoVa in enabling insulin-dependent GLUT4 traffic, we reduced the expression of endogenous MyoVa via small interfering RNA (siRNA)–mediated knockdown. With an achieved knockdown efficiency of ∼60% (Figure 6A and B), the net insulin-stimulated GLUT4myc gain at the cell surface was reduced by ∼47% compared with the response in cells expressing a nonrelated (NR) control siRNA. Under these conditions, there was no reduction in insulin-dependent phosphorylation of Akt (Figure 6A). These results demonstrate the participation of endogenous MyoVa in insulin-mediated GLUT4 exocytic movement.

To determine the effect that silencing MyoVa had on GLUT4 intracellular localization, we cotransfected sh-GFP-MyoVa (or control sh-expression vector encoding GFP) along with RFP-GLUT4 in L6 myoblasts. Silencing MyoVa resulted in the collapse of RFP-GLUT4 into a tight and dense location at the perinuclear region, which contrasted with the punctate perinuclear distribution observed in the control condition (Figure 7). This finding suggests that MyoVa normally allows proper positioning of GLUT4 in the perinuclear region, from where it is dynamically recruited to the cell periphery. In the absence of MyoVa, GLUT4 is unable to leave this region and instead coalesces in a tight perinuclear locale. The fact that insulin no longer allows its translocation to the membrane in cells depleted of MyoVa (Figure 6A and B) suggests that MyoVa serves the function of allowing GLUT4 departure (possibly motored by MyoVa) from its normally punctate perinuclear compartment.

Finally, we assessed the specificity of MyoVa toward GLUT4 vis-à-vis its possible effects on the traffic or distribution of other membrane proteins. Unlike the mislocalization of GLUT4 caused by silencing MyoVa shown earlier, sh-GFP-MyoVa did not alter the localization of the transferrin receptor, furin, or syntaxin-6 (Figure 8). Similarly, their localization was not altered by expression of the C-terminal-fragment GFP–MyoVa-CT (Figure 8). Moreover, unlike inhibition of GLUT4 translocation caused by silencing MyoVa or expressing GFP-MyoVa-CT, these two strategies failed to appreciably affect transferrin recycling (Supplemental Figure S5).

Overall these results support participation of MyoVa in the mobilization of GLUT4 toward the muscle cell surface promoted by Rab8A binding.

Function of Rab8A in defined intracellular traffic events

Rab8A has gained interest in recent years, as it has been increasingly shown to participate in directional endosomal and exocytic events, as summarized in Table 1. These collective functions are extended here to the regulation of GLUT4 traffic in muscle cells. Of note, Rab8A associates with tubular membranes that are connected to kinesin and dynein motors promoted by Rab35, EHD1, and MICAL-L1 (Roland et al., 2007; Giridharan et al., 2012). This finding suggests the possibility that Rab8A may act at a point of transference of vesicular cargo from microtubules (mobilized by kinesin or dynein) to actin filaments. This scenario may also play out in GLUT4 vesicle mobilization, as Rab35, EHD1, and kinesin have all been implicated in this process (Guilherme et al., 2004; Davey et al., 2012). We entertain a model by which Rab8A would “hand over” GLUT4 vesicles from microtubules to actin filaments, where they would be propelled by MyoVa, as this molecular motor has been shown to functionally couple microtubules to actin filaments (Cao et al., 2004), and insulin promotes MyoVa association with actin filaments (Yoshizaki et al., 2007). This model should be the subject of future experimental enquiry.

Rab GTPases mobilize GLUT4 in muscle cells and adipocytes

Rab8A and Rab10 are both inactivated by the Rab-GTPase activating protein AS160, which is inhibited by phosphorylation in response to insulin (Sano et al., 2007). Whereas in muscle cells overexpression of Rab8A rescues GLUT4 translocation from the inhibition imposed by the constitutively active AS160-4A (Sun et al., 2010), in 3T3-L1 adipocytes Rab10 is the more relevant AS160-regulated GTPase (Sano et al., 2007; Chen et al., 2012; Sadacca et al., 2013). It is therefore important to explore whether there are similarities and differences in the functions of these GTPases, as both are expressed in each cell type (Sun et al., 2010).

To begin with, the subcellular localizations of Rab8A and Rab10 are markedly different. Whereas Rab8A in muscle cells is largely located in perinuclear/cytoplasmic endomembranes and it cannot be detected in the TIRF-imaged zone (Figure 5B), Rab10 in adipocytes is readily detected in this region, where it contributes to GLUT4 vesicle docking with the membrane (Chen et al., 2012). Of note, in muscle cells, Rab13 is the insulin-stimulated GTPase that reaches the TIRF-imaged zone, where it interacts with proteins that may enact GLUT4 vesicle docking (Y.S. and A.K., unpublished data).

Although it has been difficult to detect the activity status of Rab10 in response to insulin in adipocytes (Sano et al., 2008), silencing DENND4C (a Rab10 GEF) reduces the insulin-dependent gain in surface GLUT4, and DENND4C overexpression promotes a gain in surface GLUT4 (Sano et al., 2011; Sadacca et al., 2013). DENND4C is tonically active in adipose cells (Sadacca et al., 2013), and in addition, GDI-1 may control Rab10 functional availability (Chen et al., 2009). In muscle cells, the activated state of Rab8A has been directly measured upon binding of a photoactivatable GTP (Sun et al., 2010), and its activation is supported here by our GST-dependent pull-down approach by which GTP-bound Rab8A binds to the CT domain of MyoVa (Figure 1C). This finding further suggests that MyoVa-CT may be used in pull-down experiments as an indicator of Rab8A activity, akin to the use of domain effectors of Rho-family GTPases (e.g., the Cdc42/Rac interactive binding domain of PAK for Rac and the Rho-binding domain of rhotekin for Rho). Of note, binding of Rab8A to this CT domain rose in response to insulin, whereas Rab10 binding did not (Figure 1B). Finally, of importance, we identified two amino acids required for Rab8A binding to MyoVa-CT (Q1300 and Y1312) that were dispensable for binding Rab10 (Supplemental Figure S4). Because mutating these residues in MyoVa-CT abrogated the inhibitory action of the transfected MyoVa fragment on GLUT4 translocation in muscle cells, we conclude that Rab8A but not Rab10 binding to MyoVa is important for insulin-regulated GLUT4 traffic in muscle cells.

Additional differences have been reported in the function of Rab8A and Rab10 in GLUT4 traffic. In muscle cells, silencing Rab8A largely reduces the insulin-dependent gain in surface GLUT4, and this is not observed upon silencing Rab10 expression (Ishikura et al., 2007). In contrast, the reverse is observed in 3T3-L1 adipocytes (Sano et al., 2007), although a partial contribution for Rab8A in these cells has also been reported (Brewer et al., 2011). In keeping with AS160 acting as a retaining signal, silencing this Rab-GAP elevates surface GLUT4 levels in both muscle cells and adipocytes, an effect selectively counteracted by concomitantly silencing Rab8A in muscle cells (Ishikura and Klip, 2008) and Rab10 in adipose cells (Sano et al., 2007).

Finally, in both cell types, Rab14 silencing reduces GLUT4 traffic, and emerging findings place this GTPase at a step before GLUT4 sorting into the insulin-sensitive compartment (Reed et al., 2013; Sadacca et al., 2013), potentially controlled more by TBC1D1 than by AS160 (Ishikura and Klip, 2008).

Based on the foregoing analysis, the function of Rab8A in GLUT4 traffic in muscle cells is likely not the same as that of Rab10 in adipocytes, and more studies are required to elucidate the cohort of Rab GTPases functioning downstream of AS160 that contribute to GLUT4 traffic in each cell type.

Function of MyoVa in vesicle traffic as an effector of Rab GTPases

The interaction of noncontractile myosin motors with Rab GTPases has emerged as a canonical pathway for vesicle traffic (Hume et al., 2001), as illustrated for the MyoV family in Table 2. Here we report that muscle cells only express MyoVa (Supplemental Figure S1). Of note, pigment granule transport propelled by MyoVa requires actin filament dynamics (Semenova et al., 2008), and insulin triggers actin filament dynamics (branching and severing) in muscle cells (Chiu et al., 2013). MyoVa was first linked to GLUT4 traffic in 3T3-L1 adipocytes upon demonstration that insulin promotes its association with actin filaments and silencing the protein expression reduces insulin-dependent GLUT4 translocation (Yoshizaki et al., 2007). Here we provide mechanistic insight into this role by showing that MyoVa and Rab8A colocalize in perinuclear/cytosolic endomembranes (Figure 2) and that the C-terminal fragment of MyoVa can coimmunoprecipitate with or pull down insulin-activated Rab8A (Figures 1 and 3). Further, upon expression of this truncated fragment in muscle cells, Rab8A becomes mislocalized, and insulin-dependent GLUT4 translocation is aborted (Figure 3D). We take these collective findings as evidence of the participation of MyoVa in GLUT4 traffic in muscle cells and link it to the upstream activation of Rab8A in response to insulin. Buttressing this conclusion, we find that a MyoVa-CT altered by site-directed mutagenesis to eliminate its binding to Rab8A is no longer inhibitory of insulin-dependent GLUT4 translocation. We surmise that normally the endogenous MyoVa links Rab8A-containing GLUT4 vesicles to actin filaments and that the overexpressed MyoVa-CT displaces the endogenous motor protein from Rab8A, preventing the link of GLUT4 vesicles to actin filaments. Whether these actin filaments are stress fibers or the more cortically located branched filaments induced by insulin remains to be determined. It is tantalizing to propose that, by increasing actin filament dynamics (Chiu et al., 2013), insulin may promote binding of MyoVa to actin filaments. Based on literature reports, the Rab8A-MyoVa interaction may allow its vesicular cargo to switch from microtubules to actin filaments (Cao et al., 2004) and/or be mobilized along the latter (Semenova et al., 2008).

We previously used the C-terminal fragment of the MyoVb isoform to similarly mislocalize Rab8A and impair GLUT4 translocation (Ishikura and Klip, 2008). The fragment also mislocalized Rab10, but only silencing the expression of Rab8A and not Rab10 prevented insulin-dependent GLUT4 translocation (Ishikura et al., 2007). Because MyoVb is not expressed in either mature muscle or muscle cells (Supplemental Figure S1), the previous result using C-terminal fragment cannot be taken to indicate participation of MyoVb in GLUT4 traffic, and instead the fragment must have acted by sequestering away Rab8A. This suggests that both the MyoVa-CT and MyoVb-CT fragments act as scavengers that mislocalize Rab GTPases and thereby prevent their cargo from achieving its destination within the cell. Consistent with this view, Chen et al. (2012) transfected a MyoVa short-tail fragment (virtually equivalent to MyoVa-CT) into 3T3-L1 adipocytes and found that it delayed and reduced the maximal insertion into the membrane of the GLUT4 surrogate protein IRAP-phluorin. On the basis of colocalization of the MyoVa fragment with Rab14 and Rab10 at endosomal compartments and the TIRF-imaged zone, respectively, these authors proposed that the endogenous MyoVa may have a dual role, promoting GLUT4/IRAP endosomal traffic and vesicle positioning for docking/fusion (Chen et al., 2012). Hence, in adipocytes, MyoVa is engaged by Rabs 10 and 14, whereas in muscle cells it is engaged by Rab8A.

In summary, we provide evidence of a functional link between a molecular switch, Rab8A, and a molecular motor, MyoVa, in the signaling relay triggered by insulin to mobilize GLUT4 to the muscle cell membrane. The results are consistent with the following model. Insulin-activated Akt phosphorylates/inactivates AS160, allowing GTP-loaded Rab8A on perinuclear/cytosolic GLUT4 vesicles to prevail and engage with MyoVa, effectively linking the vesicles to actin filaments as part of the mechanism that relays the vesicles to the plasma membrane.

Reagents and DNA constructs

DMSO and cocktail proteinase inhibitor were from Sigma-Aldrich (St. Louis, MO). Isopropyl-β-d-thiogalactoside (IPTG) was from Fermentas (Glen Burnie, MD), DSP PI22585 cross-linker was purchased from Thermo Scientific (Waltham, MA). Wortmannin was from Enzo Life Science (Farmingdale, NY). Humulin human insulin was from Eli Lilly (Indianapolis, IN). Mouse anti-myc (9E1) antibody was purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Rabbit anti-pAkt (473) and anti-myosin Va were from Cell Signaling Technology (Danvers, MA). Polyclonal anti-GFP antibodies were from Abcam (Cambridge, MA) and Clontech (Mountain View, CA). Monoclonal anti-mCherry antibody was from Clontech. Alexa 555–conjugated goat anti-mouse antibody was from Life Technologies (Carlsbad, CA). Horseradish-peroxidase (HRP)–conjugated goat anti-rabbit and goat anti-mouse secondary antibodies were from Jackson ImmunoResearch (West Grove, PA). Glutathione–Sepharose 4B and protein G beads were from GE Healthcare (Uppsala, Sweden).

GFP-MyoVa-FL (full-length) and GFP-MyoVa-CT (melanocyte specific) were kind gifts from M. C. Seabra (Imperial College, London, United Kingdom). The constitutively active, GTP-bound Rab8AQ67L and the dominant-negative, GDP-bound Rab8AT22N mutant cDNA constructs were kind gifts from I. Mellman (Genentec, South San Francisco, CA); the mCherry-Rab8A (MC-Rab8A), as well as its mutants (MC-Rab8A-GTP, MC-Rab8A-GDP), were kind gifts from J. Goldenring (Vanderbilt University, Nashville, TN). The C-terminal melanocyte-specific MyoVa C-terminal codons (1260–1854) were PCR amplified from the GFP–MyoVa-FL construct and subcloned into the pGEX-6p-2 (Clontech) vector by BamHI/EcoRI to generate GST–MyoVa-CT. Double mutagenesis (Q1300L and Y1312C) was performed on GFP/GST–MyoVa-CT by using the QuikChange Site-Directed Mutagenesis Kit (Stratagene, Santa Clara, CA) to generate GFP–MyoVa-CT(2M) and GST–MyoVa-CT(2M). Predesigned siRNA oligomers targeting rat MyoVa or a nontargeting control siRNA (siNR) were from Qiagen (Mississauga, Canada). A pSuper-basic vector encoding a GFP-tagged short hairpin RNA to the same sequence as the siRNA oligomer targeting rat MyoVa (sh-GFP-MyoVa) was a kind gift from A. El-Husseini (University of British Columbia, Vancouver, Canada). A pSuper-basic vector tagged with GFP but without the short hairpin sequence was used as control.

Cell culture and transfection

Rat L6 muscle cells stably expressing GLUT4 with an exofacial myc epitope (L6-GLUT4myc) were cultured in α-MEM supplemented with 10% fetal bovine serum and antibiotics as described previously (Wang et al., 1998; Foster et al., 2001). Transfection of cDNA constructs was performed with FuGENE HD (Promega, Madison, WI) or X-treme GENE HP (Roche, Mannheim, Germany) according to the manufacturer's protocol. Briefly, cDNA/reagent mixture was applied to the cells in serum-free medium for 4 h without antibiotics, medium was changed, and cells were grown for 24, 48, or 72 h, as indicated.

Reverse Transcriptase-PCR

Total RNA was isolated from L6-GLUT4myc muscle cells, 3T3-L1 adipocytes, C2C12 myoblasts, and mouse and rat skeletal muscles using RNeasy Mini Kit (Qiagen). cDNAs were prepared using Superscript III Reverse Transcriptase (Life Technologies). Specific primer pairs were designed to detect rat/mouse MyoVa, Vb, or Vc expression in L6 cells, using glyceraldehyde-3-phosphate dehydrogenase expression as a reaction control.

Transferrin recycling

L6 GLUT4myc myoblasts were transfected with sh-GFP-MyoVa-FL, GFP-MyoVa-CT, or sh-GFP-control plasmids for 48 h, followed by serum deprivation for 1.5 h. Cells were incubated with 50 μg/ml Alexa 546–conjugated transferrin in serum-free medium containing 1% bovine serum albumin (BSA) for 10 min at 37°C. Some cells were then chased with 500 μg/ml holotransferrin (Sigma-Aldrich) for 10 min at 37°C (intensity without chasing was the initial total signal). Cells were placed on ice, washed two times with ice-cold phosphate-buffered saline (PBS), acid buffer (0.15 M NaCl, 0.1 M glycine, pH 3.0), and PBS, then fixed with 4% paraformaldehyde in PBS at room temperature for 90 min. Samples were quenched with 50 mM NH₄Cl for 10 min, followed by washing with PBS, and then processed for fluorescence microscopy of Alexa 546–conjugated transferrin using a Zeiss LSM 510 laser-scanning confocal microscope (Carl Zeiss, Jena, Germany). Cells were scanned along the z-axis, and collapsed xy-projections were assembled from optical stacks taken at 1-μm intervals. The pixel intensity of each cell (≥15 cells per condition) was quantified by ImageJ software (National Institutes of Health, Bethesda, MD).

Cell lysis, immunoprecipitation, and immunoblotting

After experimental treatments, cells were washed and lysed in lysis buffer (1% Triton-X100, 50 mM Tris, 140 mM NaCl, 1:500 dilution of Sigma-Aldrich cocktail proteinase inhibitor, 20 mM NaF, 2 mM Na₃VO₄, 5 mM MgCl₂, pH 7.4). For coimmunoprecipitation, CHO-IR cells were grown in 10-cm-diameter dishes and transfected for 48 h, followed by serum deprivation for 3 h. During the last 30 min, cells were treated with 1.5 mM dithiobis succinimidyl propionate (DSP), a cell-permeant thiol-cleavable and amine-reactive cross-linking agent, followed by treatment with or without insulin. Lysates were passed through a 19-gauge syringe needle 10 times, and supernatants were collected after centrifugation at 11,000 × g at 4°C for 10 min. Supernatants were incubated with protein G beads conjugated to 1 μg of antibody at 4°C for 4 h, followed by three washes with washing buffer (0.4% Triton, 10 mM Tris-HCl, 140 mM NaCl, pH 7.4). Coimmunoprecipitated complexes were eluted with 2× Laemmli buffer, then separated by SDS–PAGE. Immunoblotting was conducted on polyvinylidene difluoride (PVDF) membrane (Bio-Rad, Richmond, CA) with respective primary antibodies and visualized by HRP-conjugated secondary antibodies by enhanced chemiluminescence (GE Life Sciences).

Expression of GST constructs and pull-down experiments

GST–MyoVa-CT or GST–MyoVa-CT(2M) was transformed in BL21 bacteria, and a single clone was picked to grow overnight in 3.5 ml of LB medium containing ampicillin. The culture was expanded into 15 ml during the day and then diluted into 300 ml, followed by induction with 0.3 mM IPTG overnight at room temperature under agitation (100 rpm). For pull-down assays, bacteria expressing GST–MyoVa-CT pellets were lysed by sonication in 1% Triton-X100/TBS buffer (20 mM Tris, 140 mM NaCl, pH 7.4), followed by binding with 35 μl of glutathione–Sepharose beads (GSHS, 60% slurry) at 4°C for 1 h. Lysates expressing different GFP-Rabs or GFP from CHO-IR cells were incubated with similar amounts of GST–MyoVa-CT- or GST–MyoVa-CT(2M) bound to GSHS beads at 4°C for 4 h, and after washes, bound complexes were eluted with 2× Laemmli buffer. Eluates were separated by SDS–PAGE, and the upper section of the gels was stained with Coomassie blue to reveal loading density of GST-MyoVa (95 kDa) and the lower sections transferred to PVDF for immunoblotting of GFP-Rabs using anti-GFP antibody.

Cell-surface GLUT4myc detection

Cell-surface GLUT4myc detection was as described previously (Wang et al., 1998; Ishikura and Klip, 2008). Briefly, myoblasts were grown in 24-well plates and serum starved for 3 h and treated with or without 100 nM insulin for 15 min, followed by rinsing with cold PBS, fixation with 3% (vol/vol) paraformaldehyde (PFA), quenching with 50 mM NH₄Cl, and blocking with 3% (vol/vol) goat serum for 30 min on ice. Cells were next incubated with polyclonal anti-myc antibody (1:200) for 1 h on ice, followed by three washes of 5 min with ice-cold PBS. Cells were incubated with HRP-conjugated goat anti-rabbit secondary antibody (1:1000) for 1 h on ice. After three washes with ice-cold PBS, the cells were incubated with o-phenylenediamine (1.0 ml/well of 0.4 mg of o-phenylenediamine/ml in 50 mM NaH₂PO₄, 25 mM citric acid, pH 5.0) and allowed to develop for 20–30 min at room temperature in dark. The reaction was stopped with 3 M HCl (0.25 ml/well). Supernatants were collected, and absorbance was measured at 490 nm, with background values lacking anti-myc antibody subtracted from all values.

Surface GLUT4myc was also measured by immunofluorescence in L6 myoblasts plated on glass coverslips transiently transfected with cDNA constructs, as previously described (Khayat et al., 2000; Ishikura and Klip, 2008). Briefly, cells were serum starved for 3 h and then treated with or without 100 nM insulin for 15 min, rinsed with PBS, fixed with 3% PFA, quenched with 50 mM NH₄Cl, and blocked with 2% (vol/vol) BSA in PBS. Surface GLUT4myc was detected by monoclonal anti-myc (1:100) for 1 h. After three washes, cells were incubated with Alexa 555–conjugated anti-mouse secondary antibody in the dark for 1 h, rinsed with PBS, and mounted on slides. Fluorescence images were taken with a Zeiss LSM 510 laser-scanning confocal microscope. Cells were scanned along the z-axis, and single collapsed images (collapsed xy-projection) were assembled from optical stacks taken at 1-μm intervals. The pixel intensity of each cell (≥25 cells per condition) was quantified by ImageJ software.

Spinning-disk confocal fluorescence microscopy

GFP-GLUT4 and MC-Rab8A, or GFP-MyoVa-FL and MC-Rab8A, were cotransfected in L6 muscle cells for 24 h, followed by serum deprivation for 3 h. Cells were then treated with or without insulin for 15 min and fixed with 3% PFA in PBS. Cells expressing MyoVa-FL and GFP-GLUT4 were also subjected to real-time imaging. In some experiments, fixed cells were incubated with antibodies to transferrin receptor (1:500), furin (1:300), or syntaxin-6 (1:300), followed by reaction with respective secondary antibodies conjugated to Alexa 555 (1:1000). Fluorescence images were acquired with a Leica spinning-disk confocal microscope equipped with Volocity 5.4.1 software (PerkinElmer, Waltham, MA). Cells were scanned along the z-axis from top to bottom at 0.25-μm stack intervals. Images of single optical planes from stacks were exported into Photoshop 7.0 (Adobe, San Jose, CA) for processing. The intensity of the MC-Rab8A fluorescence signal in the cell periphery and the colocalization of GFP-MyoVa-FL (green) with MC-Rab8A (red) were analyzed in raw images of spinning-disk fluorescence microscopy using Volocity 6.1.2 software. The periphery cytoplasmic area signal (avoiding the perinuclear area) of GFP–MyoVa-FL was selected as shown and quantified relative to the total cell signal or for its colocalization with MC-Rab8A. The degree of colocalization was assessed from the Manders coefficient (MC-Rab8A to GFP-MyoVa-FL). Mean values were analyzed by Prism software (GraphPad, La Jolla, CA) and are presented as mean ± SE.

TIRF microscopy

After 24 h of transfection with GFP-GLUT4 and MC-Rab8A, myoblasts were serum deprived for 3 h in 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid–containing RPMI-1640 medium and then treated with or without insulin (100 nM) for 15 min and examined by TIRF microscopy using an Olympus IX81 microscope (Olympus, Tokyo, Japan) equipped with a dual-laser (488, 561 nm using a 150×/1.45 numerical aperture objective), and Volocity 4.0 software was used to visualize GFP-GLUT4 vesicles with possible red fluorescent MC-Rab8A within 100 nm of the cell membrane.