Grp1 acts as an ARF GEF to initiate glut4 vesicle formation
ARF GEFs have been classified functionally into two broad categories based on their sensitivity to pharmacologic inhibition by brefeldin-A (BFA) (Casanova, 2007
; D’Souza-Schorey and Chavrier, 2006
). Glut4 recycling has been shown previously to be insensitive to BFA (Martin et al., 2000
). More recently, another compound (known as SecinH3) has been identified that targets a family of BFA-insensitive GEFs, known as the cytohesins (Hafner et al., 2006
). Thus, to determine whether a member of this family acts in glut4 recycling, we initially examined the effect of adding SecinH3 to differentiated 3T3-L1 adipocyte, which has been the model cell type for physiologic studies on glut4 recycling. A quantitative microscopy-based approach has been widely adopted in recent years as a more precise way of analyzing glut4 recycling, which involves measuring the level of surface glut4 and normalizing to the total level of glut4 (Chen et al., 2007
; Eguez et al., 2005
; Li et al., 2007
; Ng et al., 2008
; Sano et al., 2007
; Williams and Pessin, 2008
). Performing this glut4 recycling assay, we found that the addition of SecinH3 inhibited insulin-stimulated glut4 recycling in adipocytes (). As confirmation, we also performed the more traditional glucose uptake assay in adipocytes. SecinH3 also showed inhibition in this assay (). Thus, these initial results suggested that a member of the cytohesin family acts in glut4 recycling.
However, we also considered that glut4 recycling is regulated by insulin signaling, and members of the cytohesin family has been shown in recent years to have an additional function as early components in this signaling (Hafner et al., 2006
). As this role was first discovered in studying hepatocytes (Hafner et al., 2006
), we examined whether the cytohesins would have a similar role in adipocytes. Initially, examining tyrosine phosphorylation in whole cell lysates, we found that SecinH3 did not induce global changes (Fig S1A
). We next interrogated more specific readouts of insulin signaling. SecinH3 has been observed previously to inhibit the ability of insulin to activate IRS1 in hepatocytes (Hafner et al., 2006
). However, we found that SecinH3 did not have a similar effect on IRS1 in adipocytes, as reflected by phosphorylation at tyrosine residue 612 (Y612) of IRS1 (Fig S1B
). Moreover, whereas SecinH3 was observed previously to inhibit the ability of insulin to activate a distal signaling component, Akt (Hafner et al., 2006
), we found that SecinH3 also did not have a similar effect on Akt in adipocytes, as reflected by phosphorylation at either threonine residue 308 (T308) or serine residue 473 (S473) of Akt (Fig S1C
). We further noted that the depletion of ARF6 by small interfering ribonucleic acid (siRNA) had been shown previously to inhibit insulin signaling in hepatocytes, and thereby also implicating ARF6 to act as an early signaling component (Hafner et al., 2006
). However, we found that siRNA against ARF6 (Fig S1D
) did not induce a global change in the phospho-tyrosine profile of whole cell lysates (Fig S1E
). Moreover, knocking down ARF6 did not have an appreciable effect on the activation of IRS1 (Fig S1F
) or Akt (Fig S1G
). Taken together, these results revealed that a cytohesin member predicted to act in glut4 recycling was unlikely to function as an early component of insulin signaling.
The cytohesin family is predicted to have four members (Casanova, 2007
; D’Souza-Schorey and Chavrier, 2006
). As cytohesin-4 is expressed predominantly in hematopoietic cells (Casanova, 2007
; D’Souza-Schorey and Chavrier, 2006
), we focused on cytohesin-1, cytohesin-2 (also known as ARNO) and cytohesin-3 (also known as Grp1) as candidates. To determine which member acts in glut4 recycling, we sought to target their function by siRNA. We first documented the specificity of each targeting siRNA (). Subsequently, examining glut4 recycling using the quantitative in vivo assay, we found that only siRNA against Grp1 led to appreciable inhibition in glut4 recycling (). Moreover, consistent with our conclusion above that a potential role for a cytohesin member in glut4 recycling was unlikely to involve its role in insulin signaling, we found that siRNA against Grp1 did not have an appreciable effect on the ability of insulin to activate Akt ().
We next sought further confirmation for the specificity of the targeting siRNA, as a siRNA-resistant form of wild-type Grp1 could rescue the inhibition of glut4 recycling induced by the oligonucleotides used for siRNA against Grp1 (). We also found that a catalytic-dead point mutation in Grp1 (E161K) prevented this rescue (), indicating that the catalytic activity of Grp1 is needed for its role in glut4 recycling. Characterizing the effect of this mutation further, we found that it localized to the glut4-positive endosomal compartment similarly as the wild-type form (Fig S1H
). Thus, we concluded that the mutant Grp1 could act in a dominant negative manner to inhibit glut4 recycling, by localizing properly but not being able to activate ARF6 at the target site.
For the above results, we generated stable cell lines to express transfected forms of Grp1, because transient transfection of ARF GEFs that leads to their overexpression has been known to hyper-activate their activity in vivo (Santy and Casanova, 2001
). Thus, to avoid the possibility that this enhanced GEF activity might convert glut4 recycling to a constitutive process (instead of being regulated by insulin), we sought to express modest levels of the rescue constructs through stable transfection using a lentiviral expression system. Moreover, because this approach resulted in levels of transfected Grp1 less than that of the endogenous form (), we also sought to reduce the level of endogenous Grp1 by siRNA, so that the potential effects to the transfected forms could be more readily manifested. Consequently, the expression of transfected wild-type form of Grp1 only supported a 5-fold induction in the surface expression of glut4 upon insulin stimulation, as compared to a 9-fold increase in control cells that expressed the higher level of endogenous Grp1 (see ). We also ascertained that the use of the lentiviral expression system did not appreciably impact on the ability of the targeting siRNA to deplete endogenous Grp1 (Fig S1J
We next examined the distribution of Grp1 in adipocytes. By confocal microscopy, we detected a pool of Grp1 that colocalized with internal glut4 in unstimulated cells (). We also compared the distribution of Grp1 with that of ACAP1 and clathrin, which has been shown to act as the coat complex in glut4 recycling (Li et al., 2007
), and also that of ARF6, which has been shown to recruit this coat complex to the glut4-positive endosomal compartment (Li et al., 2007
). We could detect a pool of Grp1 colocalizing with ARF6, ACAP1 and clathrin heavy chain (CHC) (). Because the GEF acts to recruit the cognate ARF and coat complex to initiate vesicular transport (Casanova, 2007
; D’Souza-Schorey and Chavrier, 2006
), we next examined whether the depletion of Grp1 by siRNA would reduce the localization of ARF6, ACAP1, and CHC to the internal glut4-positive compartments. We found that ARF6, ACAP1, and CHC all showed decreased localization to this compartment (Fig S2A
), which was further confirmed by quantitation (). The decreased localization could not be attributed to the affected proteins having been degraded, as the protein levels of ARF6, ACAP1, and CHC were not altered by the siRNA treatment (Fig S2B
). Further confirming that Grp1 acted upstream to ARF6 and the coat components, we also found that targeting the downstream factors by siRNA did not affect the localization of Grp1 to the internal glut4 compartment (Fig S2C
), which was also confirmed by quantitation (). Consistent with these results, siRNA against either ACAP1 or CHC also did not affect the ability of Grp1 to associate with ARF6 (). Thus, the observations altogether led us to conclude that Grp1 acted upstream of ARF6, ACAP1, and clathrin in its role in glut4 recycling.
Grp1 regulates ARF6 and the clathrin-ACAP1 coat in glut4 recycling.
As these factors are predicted to be involved in vesicle formation, we next sought confirmation by pursuing a previously established glut4 vesicle reconstitution system (Xu and Kandror, 2002
). Briefly, fractions that contained either cytosol or compartmental membrane were collected from adipocytes in the basal condition, which ensured that glut4 would reside mostly at internal endosomes rather than at the PM. The two fractions were then incubated in the presence of an ATP-regenerating system. Glut4 vesicle formation was reflected by the redistribution of the transmembrane glut4 from the pellet fraction (which contained the larger compartmental membranes) to the supernatant fraction (which contained the smaller vesicular membranes). When membrane and cytosol fractions were collected from cells that had been treated with siRNA against Grp1 for incubation in the reconstitution system, we found that the redistribution of glut4 was inhibited (). Further validating this reconstitution system, we found that the redistribution of glut4 was also inhibited when membrane and cytosol fractions were collected from cells that had been treated with siRNA against ARF6, ACAP1, or CHC (). Thus, these results supported the conclusion that Grp1 acts as the ARF GEF to initiate glut4 vesicle formation.
Insulin signaling regulates Grp1 by phosphorylating key residues
We next explored whether upstream insulin signaling regulates this role of Grp1. As Akt has been predicted to be a key distal component of the insulin signaling cascade that interfaces with downstream transport factors of glut4 recycling (Ng et al., 2008
), we initially pursued an algorithm that predicted residues on substrates targeted by different kinases (Yaffe et al., 2001
). This prediction algorithm suggested three serine/threonine residues in Grp1 as potential targets of Akt – serine 155 (S155), threonine 220 (T220), and threonine 280 (T280) (Fig S3A
). To determine whether any of these sites regulated the role of Grp1 in glut4 recycling, we first examined whether mutating each residue to alanine (which abrogates potential phosphorylation at the site) would inhibit the ability of insulin to induce glut4 recycling. Technically, we again sought to express modest levels of transfected Grp1 by generating stable cell lines (see ), and also reducing the level of endogenous Grp1 through siRNA to allow the effect of the mutant forms to be manifested more clearly. Subsequently, assessing the effects of mutation at each of the three residues, we found that mutations at either S155 (S155A) or T280 (T280A) reduced the ability of insulin to induce glut4 recycling, and as control, mutation at T220 (T220A) did not have a similar effect ().
Insulin regulates Grp1 through Akt phosphorylation.
We next considered the intriguing possibility that mutations to mimic constitutive phosphorylation at the two key residues (S155 and T280) in Grp1 may convert glut4 recycling from a regulated (insulin-dependent) to a constitutive (insulin-independent) process. Indeed, when both residues were mutated to aspartates (S155D and T280D), we found that glut4 recycling occurred despite no insulin having been added, and as control, a similar mutation at T220 (T220D) did not stimulate glut4 recycling (). We also noted that a positive feedback loop has been suggested recently in describing how the cytohesin members are regulated (Stalder et al., 2011
). As this type of regulation is predicted to magnify the effect of the phosphorylation at key residues, we next sought confirmation by mutating only one residue to mimic constitutive phosphorylation (S155D or T280D), and then examining for effects on glut4 recycling in the basal condition. Indeed, either mutation was able to induce glut4 recycling nearly to the same extent as that seen for the “double D” mutation ().
We next sought confirmation that the two key residues affected the activation of ARF6 in vivo. First however, we noted that Grp1 has been suggested to be a more potent activator of ARF1 than ARF6, based on the in vitro assessment of GEF activity (Klarlund et al., 1998
), but in vivo studies also suggest that Grp1 acts on ARF6 (Langille et al., 1999
). Thus, to resolve this issue more definitely with respect to glut4 recycling, we examined the effect of siRNA against ARF1, and found that it did not affect glut4 recycling (Fig S3B
). In contrast, we noted that siRNA against ARF6 had been shown previously to inhibit glut4 recycling (Li et al., 2007
). Thus, we next focused on determining whether the two key residues in Grp1 regulated its ability to activate ARF6 in vivo. A general approach has been to use an effector domain to detect the binding of the target small GTPase in its active form, as exemplified by the previous use of a GGA effector domain to detect the activated form of ARF6 in vivo (Santy and Casanova, 2001
). Pursuing this approach, we initially could not detect a significant change in ARF6 activation in cells that expressed the different mutant forms of Grp1 (). We next considered that a component of the coat complex involved in glut4 vesicle formation is ACAP1, which also acts as the GAP to deactivate ARF6. As such, ARF6 activation is predicted to be followed by its rapid deactivation during glut4 vesicle formation. As this situation would thwart the ability to detect any significant accumulation of activated ARF6 in vivo, we sought to overcome this hurdle by treating adipocytes with siRNA against ACAP1, and the examining for the effect of stimulating Grp1. In this circumstance, ARF6 activation could be detected when cells were stimulated with insulin, or transfected with the activating mutations of Grp1 ().
We also sought confirmation that the two key residues in Grp1 regulated its role in glut4 vesicle formation by re-visiting the glut4 vesicle reconstitution system. Membrane and cytosol fractions were again collected from adipocytes in the basal condition that had been treated with siRNA against Grp1. The recombinant forms of different mutant Grp1 were then added to the reconstitution system. We found that S155D and T280D promoted glut4 vesicle formation, while S155A and T280A had markedly less effect (). We also sought in vivo confirmation for this result. A subcellular fractionation approach has been used previously to track glut4 vesicle formation from compartmental membrane of endosomes. Briefly, this involves fractionating cytoplasmic membranes by velocity sedimentation to segregate large (compartmental) membrane from small (vesicular) membrane, followed by immunoblotting for glut4 (Li et al., 2007)
. By this fractionation approach, we found that glut4 resided mostly in the vesicular fraction (). Notably, upon inhibiting Grp1 by siRNA or the expression of the alanine mutants of Grp1, we found that glut4 was redistributed to the compartmental membrane fraction (). Confirming that this redistribution likely represented inhibition at the step of glut4 vesicle formation, we found that siRNA against Myo1c or Sec10, which acted in the translocation and docking steps of glut4 recycling respectively (Bose et al., 2002
; Chen et al., 2007
; Inoue et al., 2003
; Yip et al., 2008
), did not induce a similar redistribution (). We further noted that siRNA against ARF6, ACAP1, or CHC had been shown to induce a similar redistribution of glut4 using this fraction approach (Li et al., 2007
). Thus, when taken altogether, the observations further supported a key role for Grp1 in glut4 vesicle formation, and also S155 and T280 in regulating Grp1 in this process.
We next examined the distribution of glut4 among intracellular compartments upon the inhibition of Grp1. In control cells, we found by confocal microscopy that glut4 showed little colocalization with TGN38 [which marked the trans-Golgi network (TGN)], and Lamp1 (which marked the late endosome and lysosomes), while having modest colocalization with transferrin receptor (TfR) (which marked the recycling endosome) (Fig S3C
). However, in cells treated with siRNA against Grp1, glut4 showed enhanced colocalization with TfR, but not with the other organellar markers (Figs S3C
). These results were also quantified (). We also confirmed that the lack of colocalization between glut4 and Lamp1 was not due to a population of glut4 having been degraded at the lysosome, as immunoblotting of whole cell lysates did not reveal a decrease in the protein level of glut4 when Grp1 was inhibited (Fig S3D
). Thus, these results were consistent with a current view that glut4 vesicles are derived from the recycling endosome (Foley et al., 2011
We next sought confirmation that Akt acted on these two residues in Grp1. Initially, we took advantage of an antibody that detects sites on proteins that are phosphorylated by Akt (Manning et al., 2002
). We found that this phospho-Akt substrate antibody showed increased detection of Grp1, when it was isolated from insulin-stimulated cells (). Supporting this finding, we also found that inhibition of Akt, through siRNA, or targeting phosphatidylinositol-3 (PI3) kinase or Akt by pharmacologic inhibition, resulted in reduced detection of Grp1 by this antibody (). The specificity of these inhibitions was reflected by effects on the phosphorylation of Akt and ribosomal protein S6 (S6P) (Manning and Cantley, 2007
) (Fig S4A
). We also sought confirmation that Akt could act directly on Grp1 by performing an in vitro kinase assay. Upon incubation of purified activated Akt with recombinant Grp1, we detected phosphorylation of Grp1 using the phospho-Akt substrate antibody (). We also found that this phosphorylation could only be abolished by mutating both residues in Grp1, S155 and T280, to alanines (). Thus, these results suggest that we have identified at least two residues in Grp1, S155 and T280, which can be direct targets of Akt.
Phosphorylation at distinct residues of Grp1 regulates distinct activities.
We next sought insight into how the two residues in Grp1 affected its function. Grp1 has two main functional domains, a Sec7 domain that catalyzes ARF activation and a PH domain that binds to target membrane (see Fig S3A
). First, to assess the GEF activity of Grp1, we performed an in vitro GEF assay, which involved incubating recombinant forms of different point mutants of Grp1 with recombinant ARF6. We found that S155D showed enhanced ability to activate ARF6, but the other point mutations did not have a similar effect ( and Fig S4B
). Second, to assess the recruitment of Grp1 to target membrane, we isolated a membrane fraction from adipocytes that was enriched for internal glut4 (Fig S4C
). Upon incubation of this membrane fraction with different point mutants of Grp1 as recombinant proteins, we found that only T280D showed enhanced binding to this membrane fraction ( and Fig S4D
). Thus, we concluded that S155 regulated the GEF activity, while T280 regulated the localization of Grp1. We also sought in vivo support for these findings. In one experiment, we isolated the different forms of Grp1 that had been transfected in adipocytes, and then assessed their ability to active ARF6 in the GEF assay. The results confirmed that mutations at S155 affected the GEF activity, while mutations at T280 did not have a similar effect (). In another experiment, we pursued confocal studies and found that mutations at T280 affected the ability of Grp1 to localize to the glut4-positive compartment, while mutations at S155 did not have a similar effect (). Thus, these results further confirmed that one key residue (S155) controlled the GEF activity, while the other key residue (T280) controlled the localization of Grp1 to the glut4-positive endosomal compartment.
We also noted that, although the above effects of Grp1 mutants in the reconstitution system suggested that its phosphorylation was critical for glut4 vesicle formation, we had found above that membrane and cytosol fractions collected from adipocytes in the basal condition (which should result in low level of phosphorylated Grp1) supported a robust level of glut4 vesicle formation (see ). In considering a reconciling explanation, we noted that an ATP-regenerating system was required to reconstitute glut4 vesicle formation (see ). Moreover, the in vitro kinase assay indicated that the presence of ATP was sufficient for Akt to phosphorylate Grp1 (see ). Taken together, these observations suggested that glut4 vesicle formation occurred in the reconstitution system, because the ATP-regenerating system promoted the ability of Akt to phosphorylate Grp1 in this reconstitution. Confirming this explanation, we isolated Grp1 from the reconstitution system and found that it only became phosphorylated by Akt when the ATP-regenerating system was provided ().
Grp1 activates ARF6 to promote the later steps of glut4 recycling
We also noted that key factors have been identified that actively suppressed the translocation, docking and fusion of glut4 vesicles, which become relieved upon insulin stimulation (Eguez et al., 2005
; Jewell et al., 2011
; Yamada et al., 2005
). As such, how could the activating mutations (S115D and T280D) of Grp1 induce the surface expression of glut4 in the basal condition (see ), if Grp1 acted only to promote glut4 vesicle formation? Thus, we next explored whether activation of Grp1 also promoted the later steps of glut4 recycling. Taking a systematic approach, we first noted that the regulation of glut4 recycling could be viewed mechanistically as a hierarchy, with insulin signaling acting at the top, the major transport effectors at the bottom, and key small GTPases in between (Fig S5A
). Thus, we first examined whether the Grp1 activating mutants could somehow feedback to activate the insulin signaling cascade. However, assessing the activation status of Akt as a downstream readout for this signaling, we found that the Grp1 mutants did not induce the activation of Akt in the basal (no insulin) condition (). We also confirmed that insulin signaling was intact in cells that expressed the Grp1 mutants, as adding insulin resulted in the activation of Akt (). Thus, we concluded that glut4 recycling induced in the basal condition by the Grp1 mutants did not occur through feedback to activate the insulin signaling cascade.
Elucidating how the Grp1 mutants induce glut4 recycling in the basal condition.
Next, to examine whether the Grp1 mutants exerted their effect at the level of key small GTPases that governed glut4 recycling, we examined the effect of depleting these small GTPases by siRNA (Fig S5B
). We found that only siRNA against ARF6 prevented the Grp1 mutants, either S155D () or T280D (Fig S5C
), from inducing glut4 recycling. We also confirmed that siRNA against RalA and Rab10 were effective, as targeting either small GTPase in control cells inhibited glut4 recycling induced by insulin (Fig S5D
). Thus, we concluded that glut4 recycling induced by the activating mutants of Grp1 required ARF6, but not RalA or Rab10.
We then noted that activated ARF6 has been shown previously to interact with the exocyst to promote endocytic recycling in the context of membrane ruffle formation (Prigent et al., 2003
). Thus, we pursued the possibility that activation of ARF6 interacted with key effectors for the later steps of glut4 recycling in explaining how the Grp1 mutants could promote all the steps of glut4 recycling. First, we confirmed that glut4 recycling induced by the Grp1 mutants in the basal condition required the exocyst, as siRNA against Sec10 reduced the ability of either S155D () or T280D (Fig S5E
) to induce glut4 recycling. We then sought to detect the activated form of ARF6 in association with key effectors that acted in the later steps of glut4 recycling. For this goal, we again used the GGA domain to probe cell lysates, reasoning that it could potentially detect activated ARF6 in association with another effector, because ARF small GTPases have been suggested to be capable of interacting with more than one effector partner simultaneously (Goldberg, 1999
). Indeed, we found that the GGA domain could detect ARF6 in association with either Sec10 or Myo1c (). Importantly, these in vivo interactions required ARF6, as siRNA against ARF6 prevented these interactions (). Thus, these results suggested that Grp1 also promoted the later steps of glut4 recycling by activating ARF6 to engage core effectors that mediated these steps.
We also noted that two recycling pathways exist in adipocytes, an insulin-responsive one (which transports cargoes such as glut4) and a largely insulin-independent one (which transports “generic” cargoes such as TfR) (Govers et al., 2004
; Xiong et al., 2010
). We initially confirmed that endogenous Grp1 acts in glut4 recycling, as siRNA against Grp1 did not affect TfR recycling (). We then also sought confirmation that the activating mutants of Grp1 did not promote the surface expression of glut4 through diversion into the “generic” recycling pathway. For this goal, we noted that the Vamp2 SNARE acts in glut4 recycling (Williams and Pessin, 2008
), while Vamp3 (also known as cellubrevin) acts in TfR recycling (Daro et al., 1996
), which marks the generic pathway. We first confirmed that siRNA against Vamp2 in control adipocytes inhibited insulin-stimulated glut4 recycling (Fig S6A
), while siRNA against Vamp3 did not affect this recycling (Fig S6B
). Conversely, TfR recycling was inhibited by siRNA against Vamp3, but not against Vamp 2 (Fig S6C
). We then found that the ability of the Grp1 mutants to induce glut4 recycling in the basal condition became inhibited, when adipocytes were treated with siRNA against Vamp2 (). In contrast, siRNA against Vamp3 did not have a similar effect (). We also found that TfR recycling was not enhanced by the Grp1 mutants (). Thus, the results altogether led us to conclude that Grp1 activation increased the surface expression of glut4 by promoting the specialized (insulin-regulated) recycling pathway in adipocytes, rather than diverting internal glut4 to the generic (constitutive) recycling pathway.
Grp1 mutants do not induce glut4 recycling through the generic recycling pathway.