Role of EHD1 and EHBP1 in perinuclear sorting and insulin-regulated GLUT4 recycling in 3T3-L1 adipocytes.

Insulin stimulates glucose transport in muscle and adipose tissues by recruiting intracellular membrane vesicles containing the glucose transporter GLUT4 to the plasma membrane. The mechanisms involved in the biogenesis of these vesicles and their translocation to the cell surface are poorly understood. Here, we report that an Eps15 homology (EH) domain-containing protein, EHD1, controls the normal perinuclear localization of GLUT4-containing membranes and is required for insulin-stimulated recycling of these membranes in cultured adipocytes. EHD1 is a member of a family of four closely related proteins (EHD1, EHD2, EHD3, and EHD4), which also contain a P-loop near the N terminus and a central coiled-coil domain. Analysis of cultured adipocytes stained with anti-GLUT4, anti-EHD1, and anti-EHD2 antibodies revealed that EHD1, but not EHD2, partially co-localizes with perinuclear GLUT4. Expression of a dominant-negative construct of EHD1 missing the EH domain (DeltaEH-EHD1) markedly enlarged endosomes, dispersed perinuclear GLUT4-containing membranes throughout the cytoplasm, and inhibited GLUT4 translocation to the plasma membranes of 3T3-L1 adipocytes stimulated with insulin. Similarly, small interfering RNA-mediated depletion of endogenous EHD1 protein also markedly dispersed perinuclear GLUT4 in cultured adipocytes. Moreover, EHD1 is shown to interact through its EH domain with the protein EHBP1, which is also required for insulin-stimulated GLUT4 movements and hexose transport. In contrast, disruption of EHD2 function was without effect on GLUT4 localization or translocation to the plasma membrane. Taken together, these results show that EHD1 and EHBP1, but not EHD2, are required for perinuclear localization of GLUT4 and reveal that loss of EHBP1 disrupts insulin-regulated GLUT4 recycling in cultured adipocytes.

Insulin stimulates glucose transport in skeletal muscle and adipose tissues by promoting the translocation of the glucose transporter GLUT4 from intracellular pools to the plasma membrane (1)(2)(3)(4)(5)(6). Although this phenomenon was documented for the first time Ͼ20 years ago (7,8), the mechanisms by which insulin regulates GLUT4 recycling and the cellular processes that retain GLUT4 within intracellular storage membranes in the basal state are still largely unknown. In the absence of insulin, GLUT4 cycles slowly between the plasma membrane and some intracellular compartments distinct from the constitutive recycling pathway, with the vast majority of the transporter residing within the cell interior near the juxtanuclear region. Activation of the insulin receptor tyrosine kinase triggers a cascade of signaling events that impact the GLUT4 recycling system, leading ultimately to the recruitment of intracellular GLUT4-containing vesicles to the cell surface (1,2). This process appears to involve both motor-driven movements of these vesicles on microtubule-and actin-based cytoskeletal tracks (9 -11) as well as fusion of the GLUT4-containing vesicles with the plasma membrane (12). Substantial data indicate that activation of phosphatidylinositol 3-kinase and the downstream protein kinases Akt2 and protein kinase C/ may be required for this insulin-mediated process (13). Thus, blockade of phosphatidylinositol 3-kinase activity (14) or depletion of Akt2 mediated by small interfering RNA (siRNA) 1 -based gene silencing in 3T3-L1 adipocytes (15,16) or knockout of protein kinase C (17) impairs insulin signaling to GLUT4. A Rab GTPase-activating protein substrate of Akt protein kinase has also recently been discovered, which may in part mediate insulin action on GLUT4 trafficking (18). However, the identities of the Rab protein or proteins involved in this mechanism are unknown, and it is likely that additional components required for GLUT4 regulation are yet to be discovered.
A number of fundamental questions about the membrane trafficking pathway of recycling GLUT4-containing membranes also remain unanswered. These include the following. 1) How does GLUT4 relate to the constitutive recycling pathway that is traversed by such proteins as the transferrin receptor? 2) What is the mechanism that retards movement of GLUT4 out of the perinuclear compartment in the basal state? 3) What are the site(s) in the GLUT4 recycling pathway regulated by insulin? In attempts to address these questions, several groups have reported that the trafficking of GLUT4 appears to partially overlap with the general endocytic recycling pathway, and several endosomal proteins have been identified in GLUT4 vesicles (19 -21). Similar to the transferrin receptor, GLUT4 is internalized from the cell surface to early endosomes via clathrin-mediated endocytosis (22)(23)(24) and then moves from early endosomes to a perinuclear endosomal recycling compartment (ERC). Evidence for a role of dynein during this latter step has been published recently, consistent with the hypothesis that the cytoskeleton plays a role in this process (10,11,21,25). In the ERC, however, GLUT4 appears to be segregated away from the other recycling proteins that constitutively recycle back to the cell surface. The sorting and segregation of GLUT4 from the ERC into a specialized sequestration compartment are strongly supported by studies employing both biochemical and morphological techniques, indicating that GLUT4 is targeted to a discrete intracellular compartment compared with other proteins such as the transferrin receptor and GLUT1 (20, 26, 27, 29 -31). Consistent with the hypothesis that GLUT4-containing vesicles bud off from the ERC, a role for Rab11 in GLUT4 vesicle biogenesis and recycling has been proposed recently (32). There is also indirect evidence that GLUT4 recycling through the trans-Golgi network occurs (33)(34)(35), although this appears to conflict with other data suggesting that a distinct compartment harbors insulin-sensitive GLUT4 (36).
Recently, results derived from several laboratories relating to a family of Eps15 homology (EH) domain-containing proteins, denoted EHD1-4 (37)(38)(39)(40)(41)(42)(43), have converged to suggest a role for these proteins in regulated membrane recycling. Thus in Caenorhabditis elegans, a protein similar to EHD1, denoted RME-1, was found localized in perinuclear recycling membranes, and its disruption caused both redistribution of the endocytic recycling compartment and impaired transferrin recycling to the cell surface (38). In mammalian cells, EHD1 was found to reside in a recycling membrane compartment related to exocytosis of major histocompatibility complex class I and the cystic fibrosis transmembrane conductance regulator (39,40) and to interact with EHD3 in a tubular membrane recycling compartment (43). Findings in our laboratory showed that the related protein EHD2 localizes near the plasma membrane in actin-rich areas and is required for endocytosis of transferrin and the GLUT4 glucose transporter (37). In this study, we therefore sought to determine whether EHD1 plays a role in GLUT4 recycling similar to its role in transferrin receptor trafficking and to compare the functions of EHD1 and EHD2 in cultured 3T3-L1 adipocytes. Here, we present evidence that EHD1 and its interacting protein EHBP1 (EH domain-binding protein-1), but not EHD2, are required for perinuclear sorting and insulin-regulated GLUT4 recycling in these cells. Loss of function of EHD1, mediated by expression of a dominant-negative construct or siRNA-based gene silencing, markedly enlarged the ERC, dispersed GLUT4 membranes, and impaired translocation and fusion of GLUT4 vesicles with the plasma membrane in response to insulin. Based on these results, we postulate that EHD1 is part of the cellular machinery necessary for normal sequestration and regulation of GLUT4-containing vesicles.

EXPERIMENTAL PROCEDURES
Material and Chemicals-Anti-EHD1 antibody was a gift from Dr. Mia Horowitz (Tel Aviv University, Tel Aviv-Jaffa, Israel). Rabbit anti-EHD2, affinity-purified anti-EHBP1, and anti-hemagglutinin (HA) polyclonal antibodies were generated as described (37). Rabbit anti-Arp3 and mouse anti-phosphotyrosine (clone 4G10) antibodies were from Upstate Biotechnology, Inc. Goat anti-GLUT4 polyclonal (C-20) and mouse anti-insulin receptor monoclonal antibodies were from Santa Cruz Biotechnology, Inc. Mouse anti-HA monoclonal antibody was from Covance. Anti-phospho-Ser 473 Akt and anti-Akt antibodies were from Cell Signaling Technology. Rabbit polyclonal antibody against Acrp30 (adipocyte complement-related protein of 30 kDa) was from Affinity BioReagents. Mouse anti-Myc epitope monoclonal antibody (clone 9E10) was conjugated to rhodamine red with a protein labeling kit (Molecular Probes, Inc.) following the manufacturer's specifications. Rhodamine-conjugated phalloidin and rhodamine-conjugated transferrin were from Molecular Probes, Inc. Cy3-conjugated mouse anti-goat and fluorescein isothiocyanate-conjugated mouse anti-rabbit secondary antibodies were from Jackson ImmunoResearch Laboratories, Inc. Anti-␣-tubulin monoclonal antibody was from Amersham Biosciences.
DNA Constructs-The murine expressed sequence tag encoding mouse EHD1 (GenBank TM /EBI accession number BC037094) was obtained from American Type Culture Collection. The plasmid DNA was isolated and sequenced, and full-length mEHD1 was used to generate the different constructs in this study. The plasmid expressing yellow fluorescent protein (YFP)-tagged ⌬EH-EHD1 (amino acids 1-442) was constructed by PCR amplification using primers creating XhoI and HindIII sites at the 5Ј-and 3Ј-ends, respectively. The PCR product was subcloned in-frame with a pEYFP-C3 vector. The plasmid expressing full-length HA-EHD1 was constructed using a linker (5Ј-A-TAAGCTTGATGTTCAGCTGGGTGAGCAAGGATGCCCGCCGCAAG-AAGGAGCCGGAGCTC-3Ј) with HindIII and SacI sites at the 5Ј-and 3Ј-ends, respectively, designed to place Ehd1 cDNA in-frame with the 3XHA vector. Full-length EHD1 in pCMV-sport6.0 was digested with SacI and XbaI, and this fragment was ligated with the linker into the HA vector to yield full-length HA-EHD1. The plasmid expressing HA-⌬EH-EHD1 was constructed by excising the ⌬EH-EHD1 fragment (amino acids 1-442) from the YFP-⌬EH-EHD1 vector described above and subcloned in-frame with the 3XHA vector using the linker described above. The plasmids expressing YFP-⌬EH-EHD2, HA-EHD2, and HA-⌬EH-EHD2 were constructed as described (37). For in vitro pull-down assays, full-length EHD1 or EHD2 or fragments were expressed as glutathione S-transferase (GST) fusion proteins in Escherichia coli as described (37). EHD2 (amino acids 1-543), ⌬EH-EHD2 (amino acids 1-444), ⌬EH⌬A-EHD2 (amino acids 1-428), and the EH domain (amino acids 444 -543) were PCR-amplified using primers generating BamHI and XhoI restriction sites at the 5Ј-and 3Ј-ends, respectively. ⌬EH-EHD1 (amino acids 1-442) constructs were PCR-amplified using primers generating SalI and NotI restriction sites at the 5Ј-and 3Ј-ends, respectively. The PCR products were subcloned in-frame with a pGEX5ϫ3 vector. The GST-VCA-expressing plasmid was similarly constructed by amplifying the region encoding amino acids 398 -501 in rat neural Wiskott-Aldrich syndrome protein. All constructs were sequenced for verification prior to transfection. Cell Culture, Transfections, Myc-GLUT4-CFP Translocation, and Transferrin Uptake Assay-CHO-T cells expressing the human insulin receptor were grown in nutrient mixture F-12 supplemented with 10% fetal bovine serum. 3T3-L1 fibroblasts were differentiated into adipocytes as described (37). Transfection of 3T3-L1 adipocytes, COS-1 cells, and CHO-T cells with ⌬EH-EHD1 or ⌬EH-EHD2 was performed as described (37). Myc-GLUT4-CFP translocation assay and rhodamineconjugated transferrin uptake assay have been described previously (37).
siRNA-induced Degradation of EHD1 and EHBP1-The siRNA species purchased from Dharmacon were designed to target the following cDNA sequences: scrambled, 5Ј-CAGTCGCGTTTGCGACTGG-3Ј; and EHD1 siRNA, 5Ј-CAGCCGAGGTTATGACTTT-3Ј; EHBP1 siRNA1, 5Ј-AAGCTCTTGCCACCAGCAGCATT-3Ј; EHBP1 siRNA2, 5Ј-AAGAG-GAGAAGGCGGCAAAAATT-3Ј. 20 nmol of scrambled siRNA, 20 nmol of EHD1 siRNA, or 10 nmol of each of the EHBP1 siRNA species were electroporated into 3T3-L1 adipocytes as described (37). Briefly, adipocytes were detached from culture dishes with 0.25% trypsin in phosphate-buffered saline (PBS), washed twice, and resuspended in PBS. Half of the cells from one 150-mm dish were mixed with siRNA and then delivered to the cells by a pulse of electroporation with a Bio-Rad Gene Pulser II system at a setting of 0.18 kV and 950-microfarad capacitance. Using a Cy3-tagged siRNA for lamin A/C, we showed previously that Cy3-tagged siRNA was introduced with virtually 100% efficiency into cells using this method and that nearly all cells showed loss of nuclear lamin A/C (15).
After 72 h, cells were harvested, and equal amounts of protein from different lysates were resolved by SDS-PAGE and analyzed by immunoblotting with the indicated antibodies. A portion of these cells were analyzed by immunofluorescence microscopy.
Immunofluorescence Microscopy Analysis-For immunofluorescence microscopy analysis, differentiated 3T3-L1 adipocytes were transfected by electroporation as described (37). The cells were then replated and allowed to recover for at least 24 h. To visualize endogenous EHD1, GLUT4, actin, and microtubules, 3T3-L1 adipocytes were fixed with 4% formaldehyde in PBS; permeabilized; and blocked with PBS containing 0.05% Triton X-100, 0.05% Tween 20, and 0.1% bovine serum albumin. Cells were incubated with the indicated primary antibody followed by a secondary antibody or rhodamine-conjugated phalloidin to detect Factin. For analysis of co-localization of endogenous GLUT4 and EHD1, z-series stacks of images were deconvoluted and reconstituted using specialized software as described (37). The co-localization of endogenous EHD1 with perinuclear GLUT4 (see Fig. 6A) and that of ⌬EH-EHD1 with GLUT4 (see Fig. 7A) were measured using the Metamorph software package. A rectangular region of interest in a single z-plane of the EHD1 and GLUT4 Recycling deconvolved images of cells costained with anti-EHD1 and anti-GLUT4 antibodies (Figs. 6A and 7A) was drawn. An intensity threshold was set in each image, and the percentages of EHD1 and truncated EHD1 that overlapped with GLUT4 were measured in the boxed region. The percentages of GLUT4 that overlapped with EHD1 and truncated EHD1 were also similarly measured.
Analysis of Interaction of the Arp2/3 Complex and EHD Protein Variants-GST or GST-tagged EHD variants (25 g) immobilized on glutathione beads were mixed with 0.3 mg of cytosolic fractions from 3T3-L1 adipocytes in buffer A (PBS containing 1% Triton X-100, 1 mM phenylmethylsulfonyl fluoride, 5 g/ml leupeptin, and 5 g/ml aprotinin). The samples were incubated at 4°C end-over-end, and the beads were washed five times with buffer A and then boiled in SDS sample buffer. The Arp2/3 complex was detected by immunoblotting using anti-Arp3 polyclonal antibody after resolution by SDS-PAGE.
2-Deoxyglucose Uptake Assay-Insulin-stimulated glucose transport in 3T3-L1 adipocytes was estimated by measuring 2-deoxyglucose uptake as described (9,15). Briefly, siRNA-transfected cells were reseeded on 24-well plates, cultured for 72 h, washed twice, and starved for 2 h with Krebs-Ringer Hepes buffer (130 mM NaCl, 5 mM KCl, 1.3 mM CaCl 2 , 1.3 mM MgSO 4 , and 25 mM Hepes, pH 7.4) supplemented with 0.5% bovine serum albumin and 2 mM sodium pyruvate. Cells were then stimulated with insulin for 30 min at 37°C. Glucose uptake was initiated by addition of 2-[1,2-3 H]deoxy-D-glucose to a final assay concentration of 100 M for 5 min at 37°C. Assays were terminated by three washes with ice-cold Krebs-Ringer Hepes buffer; the cells were solubilized with 0.4 ml of 1% SDS; and 3 H was determined by scintillation counting. Nonspecific deoxyglucose uptake was measured in the presence of 20 M cytochalasin B and subtracted from each determination to obtain specific uptake.

Distinct Functions of EHD1 Versus EHD2 in Recycling
Endosomes-To compare the cellular localization of EHD1 and EHD2 proteins, HA-tagged constructs of these proteins were expressed at low levels in COS-1 cells (Fig. 1A). Detection of these tagged proteins with anti-HA antibody showed that expressed HA-EHD1 appears to be associated with both the plasma membrane and perinuclear endosomes, whereas expressed HA-EHD2 localizes mostly at the cell surface and is not detected near the nucleus (Fig. 1A). The apparent divergent cellular localization of EHD1 versus EHD2 is remarkable in light of the high degree of sequence identity (71%) in these proteins, as depicted in Fig. 1B. Both EHD1 and EHD2 proteins present a P-loop motif close to the N terminus, a central coil-coiled region, and the EH domain at the COOH terminus. However, a detailed comparison of the mEHD1 and mEHD2 FIG. 1. Cellular localization of EHD proteins. A, COS-1 cells were transfected with plasmid encoding HA-EHD1 or HA-EHD2, fixed after 24 h, permeabilized, and stained with anti-HA polyclonal antibody to visualize HA-tagged proteins. HA-EHD1, but not HA-EHD2, localized in the perinuclear region in COS-1 cells. Arrows and arrowheads indicate HA-EHD1 localized at the plasma membrane and in the perinuclear region, respectively. Shown are representative digital microscopy images taken with an Olympus IX-70 inverted microscope with a CCD camera and then processed using Metamorph software. B, shown is a comparison of the overall structures and deduced sequences of mEHD1 and mEHD2. They share 71% identity over their full lengths. The predicted proteins have a P-loop motif, a central coiled-coil region, and an EH domain at their C termini. The putative Arp2/3-binding domain in EHD2 is shown in red. Extra acidic amino acids were observed in this domain in EHD2, but not in EHD1.
protein sequences revealed that mEHD1 lacks several acidic residues within a region just outside of the EH domain (Fig.  1B). As we have previously shown, this acidic region is similar to the Arp2/3-binding motif in Wiskott-Aldrich syndrome protein-like proteins and appears to be required for EHD2-mediated cortical actin reorganization (37). Thus, the lack of this acidic domain suggested to us that EHD1 may not be able to modulate cortical actin dynamics as does EHD2.
To test this possibility, we expressed in COS-1 cells a mutant construct of EHD2 devoid of its EH domain (YFP-⌬EH-EHD2), which we previously found induces a massive cortical actin phenotype (37), or the corresponding truncated EHD1 construct (YFP-⌬EH-EHD1). Consistent with our previous observations (37), COS-1 cells expressing YFP-⌬EH-EHD2 displayed marked actin rearrangements and many filopodia and spikes over the cell surface detected with rhodamine-conjugated phalloidin (Fig. 2B), but no changes in recycling endosome distribution as visualized by transferrin receptor staining (37). In contrast, neither untransfected cells nor cells expressing YFP-⌬EH-EHD1 showed significant cortical actin rearrangement, filopodia, or microspikes ( Fig. 2B). Instead, cells expressing the truncated EHD1 protein, but not the full-length protein (Figs. 1A and 2A, left panel), presented a marked enlargement of perinuclear membrane compartments, presumably the ECR ( Fig. 2A, right panel). These results are consistent with previous observations that expression of a dominantnegative EHD1 construct causes morphological changes in the ERC (38 -40).
To determine whether EHD2 interacts with the Arp2/3 complex, we examined the association of GST-tagged EHD2 proteins ( Fig. 3A) with Arp2/3 from the cytosol in an in vitro pull-down assay. As shown in Fig. 3B, the Arp2/3 complex bound the GST-tagged peptide derived from the neural Wiskott-Aldrich syndrome protein (GST-VCA) as well as GST-⌬EH-EHD2, but not GST-⌬EH-EHD1, GST-bound beads, or GST-EH alone, indicating that EHD2 can interact with the Arp2/3 complex. Interestingly, deletion of the EH domain of EHD2 markedly increased the interaction of the EHD2 protein with Arp2/3. In contrast, this increase in Arp2/3 interaction was reversed when a GST-⌬EH-EHD2 construct devoid of the acidic region was evaluated (Fig. 3B). Thus, these data are consistent with previous results (37) that show a requirement of the acidic region for EHD2 to induce actin rearrangements and that ⌬EH-EHD2 is more effective in induction of actin rearrangement than the full-length EHD2 protein. These data also indicate that EHD2, but not EHD1, is involved in cortical actin remodeling processes through interaction with the Arp2/3 complex, suggesting functional differences between these proteins.
Recently, we demonstrated that expression of the ⌬EH-EHD2 construct blocks the endocytosis and movement of transferrin and GLUT4 away from the cell surface and toward the perinuclear recycling compartment (37). We next examined whether the expression of YFP-⌬EH-EHD1 also disrupts endocytosis by monitoring uptake of rhodamine-conjugated transferrin in COS-1 cells. As shown in Fig. 2C, 15 min after addition of labeled transferrin, most of the fluorescence signal observed in untransfected cells was already localized in the juxtanuclear region. In cells expressing YFP-⌬EH-EHD1, the labeled transferrin co-localized with the truncated protein in the enlarged perinuclear compartment. In contrast, consistent with our previous report (37), cells expressing YFP-⌬EH-EHD2 displayed no detectable rhodamine-conjugated transferrin in the perinuclear area (Fig. 2C). Thus, EHD2 appears to function in early steps of endocytosis as well as in cortical actin remodeling, whereas expressed EHD1 localizes to the perinuclear region and apparently functions differently from EHD2.
We also investigated whether the expression of ⌬EH-EHD1 affects the recycling of transferrin from the ERC to the cell surface. In these experiments, we labeled COS-1 cells transfected with ⌬EH-EHD1 and neighboring untransfected cells with rhodamine-conjugated transferrin for 1 h, and then we subjected them to a chase for 90 min. After the chase, rhodamine-conjugated transferrin was markedly depleted from the untransfected cells. In contrast, cells expressing ⌬EH-EHD1 retained a significant amount of rhodamine-conjugated transferrin in the enlarged perinuclear compartment (data not shown), consistent with the notion that expression of ⌬EH-EHD1 affects the recycling of the transferrin receptor back to the plasma membrane.
The localization of endogenous EHD proteins in 3T3-L1 adipocytes was then investigated using biochemical and morphological approaches with anti-EHD1 and anti-EHD2 antibodies. Since mEHD1 and mEHD2 share high structural identity and homology (Fig. 1B), a critical issue was the degree to which antibodies against these proteins display specificity for each isoform. To address this question, equal amounts of the HAtagged mEHD1 and mEHD2 proteins expressed in COS-1 cells were immunoprecipitated from lysates with anti-HA monoclonal antibody. Immunoblot analysis of the precipitates with anti-HA, anti-EHD1, and anti-EHD2 antibodies was then performed. As depicted in Fig. 4A, the anti-EHD1 antibody reacted strongly with HA-EHD1, whereas the anti-EHD2 antibody reacted strongly with HA-EHD2. Only very weak cross-reactivity was exhibited by these antibodies, suggesting a high degree of specificity (Fig. 4A).
To further investigate the specificity of these antibodies with respect to endogenous EHD proteins, we also used siRNAmediated gene silencing to decrease specifically the expression level of endogenous EHD1 protein, followed by immunoblot analysis of adipocyte cell lysates with anti-EHD1 or anti-EHD2 antibody. As shown in Fig. 4B (upper panel), electroporation of 3T3-L1 adipocytes with siRNA targeted against EHD1 mRNA markedly depleted the EHD1 protein, detected with anti-EHD1 antibody (by Ͼ90%), and this depletion was not observed using scrambled siRNA. In contrast, no difference in EHD2 protein expression levels, detected by anti-EHD2 antibody, was observed in lysate samples depleted of EHD1 (Fig. 4B, lower  panel). Thus, regardless of the high expression level of EHD2 proteins in adipocyte lysates depleted of EHD1, the anti-EHD1 antibody does not appear to react with the EHD2 antigen. Conversely, despite the marked decrease in the expression level of EHD1 in siRNA-treated cells, the anti-EHD2 antibody detected essentially the same signal of the EHD2 antigen as in control lysates from scrambled siRNA-treated cells. These results confirm a high degree of specificity and low cross-reactivity for each of the anti-EHD protein antibodies used in these studies, thus validating the specificity of anti-EHD1 and anti-EHD2 antibodies to detect cellular localizations of endogenous EHD1 and EHD2 proteins in cultured adipocytes.
EHD1, but Not EHD2, Co-localizes with Perinuclear GLUT4 in 3T3-L1 Adipocytes-We next used these reagents to examine the localization of EHD1 and EHD2 proteins in cultured adipocytes and whether they associate with perinuclear GLUT4containing compartments using immunofluorescence microscopy. In these experiments, 3T3-L1 adipocytes were costained with anti-EHD1, anti-EHD2, and anti-GLUT4 antibodies before acquisition of z-series images by digital imaging microscopy and performance of deconvolution analyses. Fig. 5 shows initial images taken by focusing at the adherent bottom membrane of 3T3-L1 adipocytes stained with anti-EHD1 or anti-EHD2 antibody. Consistent with our previous observations (37), EHD2 appears to localize with ring-shaped, actin-based structures at the cell surface, as depicted in the lower panel.
However, EHD1 appears to associate with small punctate structures distinct from those positive for EHD2, although it was also localized close to or at the adipocyte cell surface. A substantial amount of anti-EHD1 signal was detected associated with the cytosol as well (data not shown). Consistent with these results, subcellular fractionation followed by Western blot analysis revealed EHD1 protein present in the plasma membrane, microsomal membranes, and cytosol (Fig. 4C). The cytosolic localization of EHD1 also contrasts with EHD2, which appears to be associated mostly with adipocyte plasma membranes and was hardly detected in the cytoplasm (Fig. 4C) (37).
As depicted in Fig. 6A, a complete z-series analysis of EHD1 in these adipocytes showed a pattern consistent with the cellular distribution of HA-EHD1 expressed in COS-1 cells (Figs.  1A and 2A, left panel). Endogenous EHD1 appears to localize not only with vesicles close to or at the adipocyte cell surface, but also with membranes in a perinuclear compartment, presumably the ERC. Interesting, we found that EHD1 partially co-localized with perinuclear GLUT4, as shown in Fig. 6A (Merge). In contrast, no EHD2 protein could be detected colocalizing with GLUT4 in the juxtanuclear region (Fig. 6B).
These results are consistent with recent studies reporting the localization of EHD1 with the ERC (38,40). Thus, EHD1, but not EHD2, appears to associate with juxtanuclear GLUT4containing membranes, suggesting a potential role for EHD1 in GLUT4 recycling.

Depletion of EHD1 Disperses GLUT4 from the Perinuclear
Region in 3T3-L1 Adipocytes-It has been shown recently that loss of function of endogenous EHD1 by expression of a dominant-negative construct of this protein disrupts ERC morphology and function (38,40). In an attempt to investigate whether EHD1 function is critical for localization of the GLUT4-containing compartment and GLUT4 recycling, we expressed a truncated form of EHD1 missing the EH domain and fused with YFP (YFP-⌬EH-EHD1), predicted to function as a dominantnegative inhibitor of full-length endogenous EHD1 protein (38,40,44). Adipocytes were then fixed and stained with anti-GLUT4 antibody to visualize the distribution of endogenous GLUT4-containing membranes, followed by z-series immunofluorescence microscopy analysis. As depicted in Fig. 7A, expression of YFP-⌬EH-EHD1 induced a remarkable dispersion of GLUT4 membranes from the perinuclear region of 3T3-L1 adipocytes. In cells transfected with truncated EHD1, GLUT4 was detected in ring-shaped, tubular, enlarged compartments at the cell periphery, possibly reflecting morphologically altered ERC. Such compartments co-localized with YFP-⌬EH-EHD1 (Fig. 7A, Merge). Interestingly, this remarkable phenotype was observed upon expression of YFP-⌬EH-EHD1, but not YFP-⌬EH-EHD2 or the full-length EHD1 construct (Fig. 7B), illustrating a high degree of specificity for EHD1 function in perinuclear GLUT4 sorting.
The integrity of the microtubule-based cytoskeleton has been reported to be a requirement for proper perinuclear localization of GLUT4 in 3T3-L1 adipocytes (10,11,21). Expression of EHD1 in HeLa cells induces tubular membrane formation, and this phenotype appears to require microtubule integrity (39). Also, we previously reported that expression of GFP-⌬EH-EHD2 or the EH domain of EHD2 induces a remarkable rearrangement of the cortical actin-based cytoskeleton in cultured adipocytes (37). Thus, one possible interpretation for the results depicted in Fig. 7 is that disruption of perinuclear GLUT4 by expression of YFP-⌬EH-EHD1 is due to disassembly of the microtubule network in 3T3-L1 adipocytes. Experiments were conducted to examine this possibility. 3T3-L1 adipocytes transfected or not with YFP-⌬EH-EHD1 were fixed and stained with anti-tubulin and anti-GLUT4 antibodies to visualize microtubules and GLUT4 distribution, respectively. No significant changes in the integrity of microtubules were detected in cells expressing YFP-⌬EH-EHD1, whereas a marked disruption of perinuclear GLUT4 was observed in these cells (data not shown). These results suggest a regulatory role for EHD1, but not EHD2, in sorting GLUT4 into its juxtanuclear membrane compartment. These data are also consistent with recent reports showing morphological changes in the perinuclear ERC of cells expressing a dominant-negative EHD1 construct (38,40).
We further investigated by a second approach whether endogenous EHD1 functions in localizing perinuclear GLUT4 in cultured adipocytes. In these experiments, we used siRNAmediated gene silencing to attenuate EHD1 protein expression levels. 3T3-L1 adipocytes mock-transfected or transfected with EHD1-targeted siRNA were fixed after 72 h and then costained with anti-EHD1 and anti-GLUT4 antibodies to visualize EHD1 and GLUT4 distribution. In mock-transfected cells, much of the GLUT4 appears clustered in the perinuclear region, as expected (Fig. 8, left panels). Consistent with the results depicted in Fig. 6A, EHD1 also partially co-localized with perinuclear GLUT4 in these control adipocytes. In contrast, a marked disruption in the juxtanuclear GLUT4 compartment was observed in cells depleted of EHD1 though siRNA-based gene silencing (Fig. 8, right panels). In these cells, GLUT4 appears associated with tubular structures, similar to what was seen in cells positive for YFP-⌬EH-EHD1 (see also Fig. 7A), and in small vesicles dispersed through the cytosol. Thus, impairment of EHD1 function by expressing a dominant-negative construct (Fig. 7) or by depleting endogenous EHD1 protein (Fig. 8) appears to disrupt perinuclear localization of GLUT4 in cultured adipocytes.
EHD1 Associates with the EH domain-binding Protein EHBP1-EHBP1 is a recently described protein that contains a calponin homology domain and five NPF motifs and that appears to associate with EHD2 through its EH domain (37). We conducted experiments to test whether EHD1 interacts physically with EHBP1 through its EH domain. In these experiments, equal amounts of the expressed HA-tagged proteins were immunoprecipitated from mock-transfected, HA-EHD1transfected, or HA-⌬EH-EHD1-transfected COS-1 lysates. Immunoprecipitated samples were then immunoblotted with anti-EHBP1 antibody to detect the presence of endogenous EHBP1 protein in these precipitates. As depicted in Fig. 9, the anti-HA antibody efficiently immunoprecipitated similar amounts of the HA-tagged proteins from the lysates, with very little HAtagged protein remaining in the supernatant (Fig. 9, lower  panel). In contrast, only full-length HA-EHD1, but not HA-EHD1 devoid of its EH domain (HA-⌬EH-EHD1), coprecipitated endogenous EHBP1 protein. A small but significant amount of endogenous EHBP1 was also detected in anti-HA-EHD2 immunoprecipitates. These results indicate that through its EH domain, EHD1 physically interacts with the EHBP1 protein in intact cells and suggest that the EHD1-EHBP1 complex may function during membrane recycling.
Expression of Dominant-negative EHD1 Blocks Insulin-stimulated GLUT4 Recycling-The marked dispersion of GLUT4containing membranes observed in adipocytes depleted of EHD1 or expressing YFP-⌬EH-EHD1 (Figs. 7 and 8) prompted us to investigate whether these morphological changes are associated with alterations in insulin-regulated GLUT4 recycling. In these experiments, 3T3-L1 adipocytes were electropo- FIG. 5. Both EHD1 and EHD2 localize close to the adipocyte surface, but in distinct membrane structures. Shown is the plasma membrane localization of EHD1 and comparison with EHD2. Differentiated 3T3-L1 adipocytes were fixed, permeabilized, and stained with rabbit anti-EHD1 (upper panel) and anti-EHD2 (lower panel) antibodies and fluorescein isothiocyanate-conjugated secondary antibody to visualize endogenous EHD1 and EHD2 proteins, respectively. Whereas EHD1 associated with punctate membrane structures close to the cell surface (upper panel), EHD2 localized in ring-shape membrane structures close to the cell surface (lower panel). These ringshaped structures containing EHD2 colocalized with cortical F-actin, but not cortactin or gelsolin proteins (Supplemental Fig. S2). Images were acquired by focusing at the bottom of the cell with an Olympus IX-70 inverted microscope and deconvoluted using a software as described under "Experimental Procedures." Enlargements of the boxed regions are depicted in panels a and b. rated with Myc-GLUT4-CFP with or without YFP-⌬EH-EHD1, and translocation of the GLUT4 construct to the plasma membrane was assayed by staining for anti-Myc antibody in unpermeabilized cells, as described previously (37,45). The effect of YFP-⌬EH-EHD2 or full-length EHD1, which did not interfere with perinuclear GLUT4 localization (Fig. 7B), on Myc-GLUT4-CFP recycling was also examined. 24 h after transfection, adipocytes were serum-starved and stimulated or not with insulin, and then Myc-GLUT4-CFP translocation was measured by detection of Myc rims at the cell surface. As depicted in Fig. 10A, expression of YFP-⌬EH-EHD1, but not YFP-⌬EH-EHD2 or full-length EHD1 (data not shown), caused a marked inhibition of insulin-stimulated GLUT4 translocation. Surprisingly little or no effect of EHD1 depletion due to the siRNA silencing method on insulin-stimulated 2-deoxyglucose transport or on Myc-GLUT4-CFP translocation to the plasma membrane was observed, as shown in Fig. 10 (B and C). This may be due to other proteins that can compensate for the loss of EHD1. The dominant inhibitory EHD1 construct likely blocks the action of all EHD isoforms, whereas the siRNA depletes only EHD1. Thus, these data are consistent with the concept that EHD1 and other members of the EHD family appear to be required for insulin action on GLUT4 translocation. Based on the data described above that EHD1 significantly interacts with EHBP1 and may function in the mechanism of GLUT4 translocation, we tested whether EHBP1 may be re-FIG. 6. Perinuclear GLUT4 compartments in 3T3-L1 adipocytes contain EHD1, but not EHD2. A, shown is the cellular localization of EHD1 and comparison with EHD2. Differentiated 3T3-L1 adipocytes were fixed, permeabilized, and stained with goat anti-GLUT4 and anti-EHD1 antibodies and fluorescein isothiocyanate-or Cy3-conjugated secondary antibody to visualize endogenous EHD1 and GLUT4 proteins, respectively. The arrowheads indicate the presence of EHD1 partially co-localized with GLUT4 in the juxtanuclear region. 49 and 78% of the EHD1 protein in boxed regions a and b co-localized with GLUT4, respectively. B, fixed and permeabilized 3T3-L1 adipocytes were costained with anti-EHD2 and anti-GLUT4 antibodies followed by fluorescein isothiocyanate-or Cy3-conjugated secondary antibody to visualize endogenous EHD2 and GLUT4, respectively. EHD2 localized mostly close to/or at the plasma membrane, but no detection at the perinuclear region was observed. The overlay image (Merge) shows in yellow the co-localization of EHD1 and GLUT4 (A) or lack of co-localization between EHD2 and GLUT4. The number at the top right of each panel indicates the distance (in micrometers) from the bottom. Enlargements of the indicated perinuclear regions in the merged pictures are depicted in panels a-c. quired for insulin-mediated GLUT4 translocation. Strikingly, treatment of 3T3-L1 adipocytes with siRNA directed against EHBP1 almost completely abolished insulin action on deoxyglucose uptake and GLUT4 movements (Fig. 11, A and B). This inhibition of the ability of insulin to modulate GLUT4 translocation and hexose transport due to EHBP1 depletion failed to attenuate insulin signaling to Akt protein kinase (Fig. 11C), indicating that the role of EHBP1 in regulating hexose transport is downstream of this phosphatidylinositol 3-kinase-dependent step. EHD1 has recently been suggested to play a role in insulinlike growth factor-1 receptor endocytosis and signaling to Akt and mitogen-activated protein kinase (41). Experiments were thus conducted to examine whether loss of function of EHD1 affects insulin receptor cellular distribution and signaling to Akt activation. In these experiments, EHD1 constructs were expressed in CHO-T cells, and their effects on insulin receptor localization as well as signaling through Akt and phosphotyrosine were analyzed. As depicted in Supplemental Fig. S1A, no differences in insulin receptor localization and tyrosine or Akt phosphorylation stimulated by insulin were detected in cells expressing EHD1 constructs. Furthermore, depletion of endogenous EHD1 by siRNA-mediated gene silencing was without detectable effect on insulin stimulation of Akt phos-phorylation in cultured adipocytes (Supplemental Fig. S1B). Taken together, our results provide strong support for the hypothesis that EHD1 and EHBP1, perhaps acting together, play a required role in GLUT4 trafficking and responsiveness to insulin. DISCUSSION A growing list of recent results has implicated EH domaincontaining proteins and their interacting proteins in a variety of cellular processes, including endocytosis, intracellular signal transduction, actin cytoskeleton remodeling, and vesicle recycling (28,46,47). A key finding of this study is that the highly related proteins EHD1 and EHD2 exhibit remarkably divergent properties when analyzed in intact cells. Although these proteins share a similar domain structure, exhibit nearly 71% sequence identity, and function in membrane trafficking, they apparently operate at quite different stages in the overall membrane recycling pathway. The divergent characteristics of EHD1 versus EHD2 revealed in this study include highly distinct patterns of cellular localization. EHD2 resides primarily just beneath the plasma membrane in vesicles associated with cortical actin filaments, whereas the relatively small amount of EHD1 that is localized near the plasma membrane resides in punctate structures that are not associated with actin (Figs. 2 and 5). In keeping with this distinctive relationship to cortical actin, expression of EHD2 constructs that bind Arp2/3 in vitro (Fig. 3B) leads to actin reorganization (37), whereas EHD1 or truncated versions of the protein fail to modulate actin structures (Fig. 2) or to interact with Arp2/3 (Fig. 3B). Importantly, EHD1 is mostly localized in the perinuclear region of Chinese hamster ovary cells (38) and COS-1 cells and 3T3-L1 adipocytes (Figs. 2 and 6). In this juxtanuclear region of cultured adipocytes, it partially co-localizes with GLUT4-containing membranes even though much of the EHD1 is in membranes devoid Differentiated 3T3-L1 adipocytes were mock-transfected (left panels) or transfected with 20 nM siRNA to target EHD1 mRNA (right panels), After 72 h, cells were fixed, permeabilized, and stained with goat anti-GLUT4 or rabbit anti-EHD1 antibody to visualize endogenous GLUT4 or EHD1 protein, respectively. Images in z-series were then taken and deconvoluted as described under "Experimental Procedures." Nearly 100% of cells electroporated with EHD1 siRNA, but not scrambled siRNA, showed a marked loss of EHD1 as detected by immunofluorescence microscopy with anti-EHD1 antibody. Dispersion of juxtanuclear GLUT4 was observed in cells depleted of EHD1. The number at the top right of each panel indicates the distance (in micrometers) from the bottom.
FIG . 10. Expression of ⌬EH-EHD1, but not ⌬EH-EHD2, blocks insulin-stimulated Myc-GLUT4-CFP translocation to the plasma membrane. A, differentiated 3T3-L1 adipocytes expressing Myc-GLUT4-CFP alone (CONTROL) or with YFP-⌬EH-EHD1 or YFP-⌬EH-EHD2 were serum-starved, left untreated (Basal) or treated with insulin (ϩ Insulin) for 30 min, fixed, and incubated with rhodamine-conjugated anti-Myc monoclonal antibody to detect Myc-GLUT4 rims at the cell surface. Truncated EHD1, but not truncated EHD2, impaired insulin stimulation of Myc-GLUT4-CFP translocation. Shown are representative digital microscopy images taken at an optical plane near the middle of the cells where the Myc signal around the rim of the cells was the brightest. Cells depicted in the images were counted for Myc rims at the cell surface, and data are represented in a bar graph (mean Ϯ S.E. of three different experiments). B and C, shown is the effect of EHD1 depletion on insulin-stimulated 2-deoxyglucose uptake and Myc-GLUT4-CFP translocation to the plasma membrane in 3T3-L1 adipocytes, respectively. B, 72 h after transfection with the indicated siRNAs, cells were serum-starved and treated or not with insulin for 30 min, and then 2-deoxyglucose uptake was assayed. Depicted is the -fold insulin stimulation of glucose uptake in scrambled siRNA or EHD1-directed siRNA-treated cells. Cell lysates were also prepared, and total protein was resolved by SDS-PAGE and immunoblotted with anti-EHD1 antibody (inset) to examine the amounts of EHD1 protein in 3T3-L1 adipocytes transfected with scrambled siRNA or EHD1 siRNA. The protein band is indicated by the arrow. The data represent the mean Ϯ S.E. of five different experiments. C, 3T3-L1 adipocytes were transfected with Myc-GLUT4-CFP plasmid DNA (50 g) together with either scrambled siRNA (20 nmol) or EHD1 siRNA (20 nmol). 72 h after transfection, cells were serum-starved, treated or not with insulin for 30 min, and then fixed and incubated with rhodamine-conjugated anti-Myc monoclonal antibody to detect Myc-GLUT4 rims at the cell surface. Depicted is the percentage of transfected cells showing a Myc-GLUT4-CFP rim on the cell surface. The data represent the mean Ϯ S.E. of two independent experiments. of detectable GLUT4 (Fig. 6). This is consistent with previous results indicating that EHD1 localizes to the endosome recycling compartment, through which transferrin and other proteins constitutively recycle back to the plasma membrane (38,40).
Consistent with the divergent cellular locations in which EHD1 and EHD2 reside, functional data provided here demonstrate divergent roles of these proteins in membrane trafficking. Disruption of EHD2 function by expression of a dominant inhibitory form of the protein selectively inhibits endocytosis of transferrin, whereas disruption of EHD1 function by a similar strategy is without effect (Fig. 2C). These results clearly show that loss of EHD2 activity prevents or delays the arrival of internalized transferrin to perinuclear membranes, whereas loss of EHD1 function does not disrupt such movement to perinuclear membranes. Conversely, loss of EHD1 function through the same dominant inhibitory approach or by depletion of the protein through siRNA-based gene silencing leads strikingly to dispersion of perinuclear GLUT4 into membranes that are spread throughout the cytoplasm (Figs. 7 and 8). In contrast, disruption of EHD2 function fails to affect perinuclear GLUT4 localization under conditions in which endocytosis is severely compromised (Figs. 2 and 7). Thus, it seems likely that EHD2 functions near the plasma membrane to facilitate the endocytic process at an early stage, whereas EHD1 likely acts much later in the recycling pathway of cultured adipocytes to facilitate regulated exocytosis of GLUT4. These data are consistent with previous results in other cell types showing that EHD1 appears to co-localize with recycling endosomes and functions during the exit of proteins such as the transferrin receptor, cystic fibrosis transmembrane conductance regulator, and major histocompatibility complex class I from this com- FIG. 11. Depletion of endogenous EHBP1 mediated by siRNA blocks insulin-stimulated 2-deoxyglucose uptake, but not Akt phosphorylation, in 3T3-L1 adipocytes. A, shown is the dose dependence of insulin-stimulated deoxyglucose uptake. Cells were transfected with either 40 nmol of scrambled siRNA or 20 nmol of each of the EHBP1 siRNA species (1 ϩ 2) by electroporation, reseeded, and cultured for 72 h. The serum-starved cells were then stimulated with insulin, and deoxyglucose uptake was assayed as described under "Experimental Procedures." Depicted is the -fold insulin stimulation of glucose uptake in scrambled siRNA or EHBP1-directed siRNA-treated cells. Cell lysates were also prepared, and total protein was resolved by SDS-PAGE and immunoblotted with anti-EHBP1 antibody (inset) to examine the amounts of EHBP1 protein in 3T3-L1 fibroblast cells transfected with scrambled siRNA or EHBP1 siRNA. The protein band is indicated by the arrow. The numbers indicate the amount of protein (in micrograms) loaded onto the gel. The data represent the mean Ϯ S.E. of four different experiments. B, 3T3-L1 adipocytes were transfected with Myc-GLUT4-CFP plasmid DNA (50 g) together with either 40 nmol of scrambled siRNA or 20 nmol of each of the EHBP1 siRNA species (1 ϩ 2) by electroporation and reseeded. 72 h after transfection, cells were serum-starved, treated or not with insulin for 30 min, and then fixed and incubated with rhodamine-conjugated anti-Myc monoclonal antibody to detect Myc-GLUT4 rims at the cell surface. The percentage of transfected cells showing a Myc-GLUT4-CFP rim on the cell surface is depicted. The data represent the mean Ϯ S.E. of two independent experiments. C, representative Western blot images for phospho-Ser 473 Akt (p-Akt), Akt, and Acrp30 protein levels in total lysates from 3T3-L1 adipocytes transfected with the indicated siRNAs as described above; stimulated or not with 100 nM insulin 72 h post-transfection; and harvested for Western blot analysis. Data are representative of three independent experiments. partment back to the plasma membrane (38 -40).
A remarkable result from this study is the virtual abolition of insulin-stimulated GLUT4 translocation to the cell surface upon expression of a dominant inhibitory EHD1 construct (Fig.  10). No detectable effect on this same process was observed when EHD2 function was disrupted, reinforcing the selective roles of EHD1 versus EHD2 in GLUT4 trafficking. Interestingly, the alternative approach of depleting the EHD1 protein did not result in a marked inhibition of GLUT4 translocation (Fig. 10C). This may reflect the fact that expression of a dominant inhibitory construct of EHD1 may have inhibitory effects on multiple EHD isoforms, whereas siRNA-mediated depletion of the EHD1 protein leaves expression of other isoforms intact and functional. Thus, other EHD isoforms or other proteins may compensate in part for the loss of EHD1 under these conditions. Further experiments will be necessary to test this hypothesis. Nonetheless, an EHD1-interacting protein that also localizes in the perinuclear region, EHBP1, is indeed required for GLUT4 translocation in response to insulin (Fig. 11). Thus, depletion of EHBP1 by siRNA-mediated gene silencing virtually abolishes insulin action on hexose transport in 3T3-L1 adipocytes. Taken together, our results indicate that EHD1 and EHBP1 are required for optimal responsiveness of GLUT4 to insulin signaling, whereas EHD2 does not function in this process.
What might be the molecular basis of EHD1 and EHBP1 function in regulated GLUT4 exocytosis? Two clues related to this question are that loss of EHD1 function results in 1) inhibition of transferrin exocytosis (this study and Ref. 38) as well as GLUT4 translocation (Fig. 10) and 2) dispersion of perinuclear GLUT4 (Figs. 7 and 8). It is thought that GLUT4 is internalized through the same pathway as transferrin and that the two proteins co-localize through their transit to the endosome recycling compartment, where GLUT4 then segregates into a specialized membrane system. Thus, one hypothesis that would explain the data obtained in this and other studies is that EHD1 and EHBP1 play a role in sorting GLUT4 from the recycling endosome to the specialized insulin-regulated GLUT4 sequestration compartment. This model would explain the requirement of EHD1 for insulin action on GLUT4 recycling. Disruption of recycling endosome function, involving sorting from this compartment, might also account for dispersion of this organelle and apparent GLUT4 dispersion throughout the cytoplasm. This hypothesis will be useful in future experiments designed to refine our understanding of the role of EHD1 and EHBP1 in GLUT4 recycling.