Role of the FYVE Finger and the RUN Domain for the Subcellular Localization of Rabip4*

Rabip4 is a Rab4 effector, which possesses a RUN domain, two coiled-coil domains, and a FYVE finger. It is associated with the early endosomes and leads, in con-cert with Rab4, to the enlargement of endosomes, result-ing in the fusion of sorting and recycling endosomes. Our goal was to characterize the role of these various domains in Rabip4 subcellular localization and their function in Chinese hamster ovary cells. Although the FYVE finger domain specifically bound phosphatidylinositol 3-phosphate and was necessary for the function of Rabip4, it was not sufficient for the protein association with membranes. Indeed a protein containing the FYVE finger and the Rab4-binding site was cytosolic, whereas the total protein was mostly associated to the membrane fraction, whether or not cells were pre-treated with wortmannin. By contrast, a construct corresponding to the N-terminal end, Rabip4-(1–212), and containing the RUN domain was membrane-associated. The complete protein partitioned between the Triton X-100-insoluble and -soluble fractions and a wortmannin treatment increased the amount of the protein in the Triton X-100 fraction. Rabip4-(1–212) was totally Triton X-100-insoluble, In Vitro Binding of GST-Rabip4-(401–600) to Phospholipidic Vesicles— GST-Rabip4-(401–600) was affinity-purified on glutathione-Sepharose beads from Escherichia coli transformed with pGEX-3X- Rabip4-(401–600) (21). The purified protein was resuspended in 20 m M Hepes, pH 7.5, containing 100 m M NaCl, 1 m M MgCl 2 , 1 m M dithiothre- itol, and 0.4 m M phenylmethylsulfonyl fluoride and stored at (cid:1) 80 °C until use. Sucrose-loaded unilamellar phospholipid vesicles were prepared by the extrusion method (22) as modified by Paris et al. (23). Briefly, a dried film of phospholipids composed of 38–40% phosphatidylcholine, 30% phosphatidylserine, 30% phosphatidylethanolamine supplemented or not with 2% phosphoinositides (PtdIns, PtdIns(3)P, PtdIns(3,4)P 2 , and PtdIns(3,5)P 2 ) was formed in a Rotavapor and resuspended in 10 m M Tris, pH 7.5, containing 150 m M sucrose. The suspension was vortexed for 20 min and frozen-thawed five times. Unilamellar vesicles were produced by extrusion through 0.4- (cid:1) m pore size polycarbonate filter (Isopore, Millipore). After extrusion, the sucrose-loaded vesicles were diluted 5 times in 20 m M Hepes, pH 7.5, 100 m M NaCl, and 1 m M MgCl 2 , centrifuged for 20 min at 400,000 (cid:2) g , and resuspended in the same buffer as GST-Rabip4-(401–600).

endocytotic pathway in all cell types (1). In CHO 1 cells, Rabip4 appears as an early endosomal protein, colocalized with early endosome antigen 1 (EEA1), an effector of Rab5 enriched in early sorting endosomes (2,3). It is absent from the recycling and late endosomes. The coexpression of Rabip4 with Rab4 leads to an enlargement of early endosomes, and thus Rabip4 seems involved in early endosomal traffic (1).
Rabip4 is a 600-amino acid protein, in which we noted the presence of two coiled-coil motifs (4) and a C-terminal FYVE (for Fab1p, YOTB, Vac1p, EEA1) finger (5)(6)(7). FYVE finger domains are present in a group of proteins such as EEA1 (8), Fab1p (9), Vps27p (10), and Vac1p (11), which are characterized by an endosomal localization. This cysteine-rich domain specifically binds PtdIns(3)P (6,12,13), a property that is important for the intracellular localization of those proteins. Indeed, using a double FYVE finger construct, Gillooly et al. (14) very elegantly showed an enrichment of PtdIns(3)P on the surface of early endosomes and in the internal vesicles of multivesicular endosomes. In agreement with the association of EEA1 with endosomes through the binding of its FYVE finger to the PtdIns(3)P, a mutation in this domain causes the redistribution of EEA1 to the cytosol (15). However, the endosomal localization of EEA1 was also determined by its ability to bind Rab5 (16). Hrs is another extensively studied FYVE fingercontaining protein, implicated in both membrane trafficking and signal transduction (17,18). It presents an endosomal localization, but results are more conflicting about the exact role of its FYVE finger in this localization (5). In conclusion as reviewed in (5), there are no reports of a FYVE finger alone being sufficient to target a protein to a membrane, and additional interactions might be required.
We also reported that the N-terminal part of Rabip4 presents 40% analogy with RPIP8 (Rap2-interacting protein 8) (19) and with the two KIAA 0871 and 1537 ((1) and data not shown). These proteins have been shown very recently to contain a RUN domain (for RPIP8, UNC-14, and NESCA) (20). The RUN domain is organized into six conserved blocks, which are predicted to constitute the "core" of a globular structure. Although the role of this RUN domain is not known yet, it appears in several proteins that are particularly linked to the functions of GTPases of the Rap and Rab families (20).
The purpose of this work was to clarify the roles of the FYVE and RUN domains present in Rabip4. We found that the FYVE finger of Rabip4, although it specifically binds PtdIns(3)P in vitro, was insufficient for the endosomal localization of the protein. Furthermore, the N-terminal part containing the RUN domain appears to play a key role in directing the protein toward its cellular localization. Our results suggest that, within an endosome, Rabip4 divides between phospholipidic membrane microdomains through its FYVE finger and another type of microdomains characterized by their insolubility in a non-ionic detergent, perhaps through the RUN domain.

EXPERIMENTAL PROCEDURES
Materials-Egg phosphatidylcholine, egg phosphatidylethanolamine, and brain phosphatidylserine were from Sigma. The phosphoinositides were from Echelon (Research Laboratories Inc. Salt Lake City, UT). Monoclonal antibodies against the Myc epitope (9E10) were from Santa Cruz Biotechnology. Antibodies against caveolin, EEA1, and Rab4 were from Transduction Laboratories. Anti-Rabip4 antibodies (1) were purified by affinity chromatography. Texas Red-coupled phalloidin was from Molecular Probes. Goat Texas Red-coupled antibodies to mouse immunoglobulins were from Amersham Pharmacia Biotech. Chemicals were from Sigma, and molecular biology reagents were from Biolabs, Life Technologies, Inc., and Qiagen.
cDNA Constructs and Mutagenesis-pEGFP-GPI encoded for a fusion protein between GFP and the glypiation signal of the decay-accelerating factor. The cDNAs for Rab4 Q67L, Rabip4, Rabip4-(1-212), and Rabip4-(401-600) were amplified by polymerase chain reaction and subcloned in a pcDNA3-myc vector (1) to allow for the expression of Myc-tagged proteins at their N terminus. pEGFP-C1-Rabip4 was used for expression of Rabip4 as a C-terminal fusion with the green fluorescent protein (GFP) (1). The His 554 and His 555 of Rabip4 were replaced by leucine residues using the QuickChange site-directed mutagenesis kit (Stratagene) in the construct encoding for GFP-Rabip4. Mutations were verified by sequencing.
GST-Rabip4-(401-600) (3 M) was incubated for 15 min at 25°C with 30 g of sucrose-loaded lipid vesicles in a final volume of 75 l. Following incubation, vesicles were recovered by ultracentrifugation at 400,000 ϫ g for 15 min (TL100 Beckman centrifuge). The amount of Rabip4-(401-600) associated with the lipid vesicles was determined by Coomassie Blue staining following protein separation by SDS-PAGE.
Cells and Transfections-CHO cells were cultured in Ham's F-12 medium supplemented with 10% fetal calf serum and transiently transfected by electroporation. Cells were cultured for 2 days before use. A CHO cell line stably expressing GFP-Rabip4 was obtained by selection with G418 (500 g/ml) and limited dilution of cells transfected with pEGFP-Rabip4 (1). Cells were serum-starved for 16 h before each experiment.
Immunofluorescence Confocal Analysis-Cells grown on glass coverslips were treated as indicated in the legends to figures, washed in phosphate-buffered saline, and fixed in 4% paraformaldehyde. When non-GFP fusion proteins have to be detected, cells are permeabilized in phosphate-buffered saline containing 0.1% Triton X-100 and 1% fetal bovine serum. Coverslips were successively incubated with the primary antibodies and the appropriate Texas Red-coupled anti-species antibodies before they were mounted in Mowiol (Hoechst Pharmaceuticals). The cells were examined by sequential excitation at 488 nm (GFP) and 568 nm (Texas Red) using a confocal microscope (TCS SP, Leica, Deerfield, IL) and a PL APO 63 ϫ 1.40 oil objective (Leica). The images were then combined and merged by using Photoshop (Adobe Systems, Mountain View, CA). Subcellular Fractionation-CHO cells (2 preconfluent 100-mm dishes) were homogenized with a Thomas potter (type AA) in 300 l of Tris-HCl, pH 7.4, containing 1 mM EDTA, 250 mM sucrose, 125 mM KCl, and a protease inhibitor mixture (Complete TM , Roche Molecular Biochemicals). Homogenates were centrifuged at 100,000 ϫ g for 1 h (Optima X100 ultracentrifuge, Beckman Instruments) to separate the cytosol and the total membrane pellet. The pellet was then resuspended and incubated for 1 h at 4°C in 300 l of Tris-HCl, pH 7.4, 1 mM EDTA, 250 mM sucrose, and protease inhibitors with 1% Triton X-100 or 60 mM octylthioglucoside or 1 M NaCl or 4 M urea or 10 mM Na 2 CO 3 , pH 10.5. The suspensions were then respun at 100,000 ϫ g for 1 h to separate the solubilized and the insoluble fractions. Proteins from equal volumes of homogenate, cytosol, or membrane-associated fractions were resolved by SDS-PAGE and analyzed by immunodetection.
Purification of Lipid Rafts by a Flotation Assay-Purification of lipid rafts was performed as described (24). Briefly, cells (2 preconfluent 100-mm dishes) were lysed in 300 l of ice-cold buffer containing 25 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM EGTA (TNE buffer), and 1% Triton X-100 and protease inhibitors. After 1 h at 4°C, sucrose was added to get a 40% (w/v) solution. This mixture (800 l) was sequentially overlaid with 2 ml of 30% (w/v) sucrose and 1 ml of 4% sucrose and centrifuged at 200,000 ϫ g for 16 h in an MLS 50 rotor (Optima Max centrifuge, Beckman Instruments). Fractions (300 l) were recovered from the top of the tube and analyzed by SDS-PAGE and Western blotting for Rabip4 and caveolin.

RESULTS
Structural Characteristics of Rabip4 -Rabip4 is an endosomal protein containing the sequence C(2X)C(8X)(R/K)(R/K) HHCRXC(4X)C(2X)C(16X)C(2X)C characterizing a FYVE finger motif (shown in red in Fig. 1). This double zinc finger domain of Rabip4 is 18 -52% identical to the FYVE finger of the mammalian proteins EEA1, Hrs, Ankhzn, SARA, Fgdl, PIKfyve, and Rabenosyn-5. The highest homology was found with EEA1 and the lowest one with Rabenosyn-5. The cysteine residues of the core motif involved in zinc binding are conserved. The essential nucleotides involved in the binding of EEA1 with PtdIns(3)P-1369 RKHHC 1373 and the arginine residues at positions 1374 and 1399 (shown in green)-and the ␤1 and ␤2 strands are conserved with regard to those of EEA1 and thus likely form a ␤-hairpin (25,26). While the ␤4 strands of Rabip4 and EEA1 were identical, the ␤3 strand, proposed to form the second hairpin, was different (25).
By studying Rabip4 sequence with Pfam (Protein family) data base, we detected the presence of a RUN domain between amino acids 33 and 163. We found the six conserved blocks characterizing the RUN domains (20) that are also present in KIAA 1537, the likely human homologue of Rabip4, the Rap2 effector RPIP8 (19) and the Rab6-interacting protein ORF37 (27) (Fig. 1B). As described for other RUN domains, the percentage of identity between Rabip4/KIAA 1537 and RPIP8 or ORF37 is not important but the blocks are enriched in hydrophobic amino acids in conserved positions (20). The secondary structure of the RUN domain was predicted to be essentially ␣-helices. Furthermore, the conserved basic residues in this predominantly all-␣-fold might reveal a conserved tridimensional conformation that could be important for the function of the RUN domain.
The FYVE Finger Motif of Rabip4 Binds PtdIns(3)P and Was Required for the Function of Rabip4 -Since Rabip4 possesses a putative FYVE finger, we next determined whether it was able to bind PtdIns(3)P. Therefore, we measured the ability of GST-Rabip4-(401-600), a fusion protein between GST and the amino acids 401-600 of Rabip4, to bind to phospholipidic vesicles containing the different phosphatidylinositols ( Fig. 2A). GST-Rabip4-(401-600) bound to vesicles only when they contained PtdIns(3)P, whereas GST alone did not interact with these vesicles (data not shown). This binding was highly specific, since GST-Rabip4-(401-600) did not recognize vesicles containing the same amount of PtdIns, PtdIns(3,4)P 2 , or PtdIns(3,5)P 2 . Thus, the FYVE finger motif of Rabip4 specifically interacts with PtdIns(3)P.
We then compared the behavior of a Rabip4 protein mutated on the FYVE finger (Rabip4 His 554 -His 555 ) to the wild type Rabip4. We chose to mutate the His 554 and His 555 because the corresponding mutations in EEA1 abolished its ability to interact, in vitro, with the zinc molecules required for PtdIns(3)P binding and its endosomal localization (15,28). We compared confocal images obtained in CHO cells that expressed GFP-Rabip4 and GFP-Rabip4 His 554 -His 555 to a similar level (Fig.  2B, top panels). Both GFP fusion proteins labeled punctated structures. However, the size of these vesicles was smaller when the FYVE finger was not functional (Rabip4 His 554 -His 555 ) than with Rabip4 WT. Rabip4 overexpression was shown previously to induce the appearance of enlarged vesicles (1), and the mutation within the FYVE finger altered the ability of Rabip4 to induce this phenomenon. To strengthen further the hypothesis that the FYVE finger was involved in the function of Rabip4, we analyzed the ability of Rabip4 His 554 -His 555 to induce an enlargement of endosomal structures when coexpressed with active Rab4 (1). Although the coexpression of Rabip4 with myc-Rab4 Q67L (a constitutively active form of Rab4) led to the expansion of endosomal structures in CHO cells (Fig. 2B, right panels), such a similar enlargement was not induced by the expression of GFP-Rabip4 His 554 -His 555 with active Rab4. It should be noted, however, that Rabip4 His 554 -His 555 and active Rab4 are colocalized in slightly enlarged punctated structures (Fig. 2B, left panels). Taken together, these observations suggest that the FYVE finger plays a role in the function of the protein since the mutant protein did not induce a marked enlargement of the endosome, but that overexpression of the mutated protein can partially compensate for this loss of function.
Rabip4 Was a Membrane-associated Hydrophilic Protein-To determine the cellular distribution of Rabip4, CHO cells were homogenized and the membrane-associated fraction (pellet) and the cytosol were obtained by ultracentrifugation (Fig.  3A). Equal relative amounts of homogenate and both fractions were separated by SDS-PAGE before immunodetection of Rabip4. This antibody detects an endogenous protein of 68 kDa migrating with the same apparent molecular weight as the overexpressed protein (Fig. 3A, top panel). Despite its hydrophilic character, endogenous or overexpressed Rabip4, similar to Rab4, was mainly associated with the membrane fraction, with only low amounts detected in the cytosol.
To determine the nature of the association between Rabip4 and membranes, the pellet fractions obtained from control or Rabip4 overexpressing cells were submitted to various treatments, i.e. 1% Triton X-100, 60 mM ␤-octylthioglucoside, 1 M NaCl, 4 M urea, or 10 mM Na 2 CO 3 , pH 10.5 (Fig. 3A). Rabip4 was partly resistant to the detergents (Triton X-100 and ␤-octylthioglucoside), whereas Rab4 was efficiently solubilized. The interactions between Rabip4 or Rab4 and the membrane fraction were highly resistant to high ionic strength and chaotropic agents. By contrast, a Na 2 CO 3 , pH 10.5, treatment detached Rabip4 from the pelleted fraction, whereas the geranylgeranyl membrane-anchored protein Rab4 was still present in the membrane fraction. These observations indicated that Rabip4 is a peripheral protein tightly associated with membranes as well as with Triton X-100-insoluble structures.
Since the Rabip4 FYVE finger was able to bind PtdIns(3)P, we determined that interaction was needed for targeting Rabip4 to the membrane fraction. Therefore, CHO cells were transiently transfected with pcDNA3 myc-Rabip4-(401-600), a construct that contains the FYVE finger and the Rab4-binding domain (Fig. 3B). Following homogenization, myc-Rabip4-(401-600) was found predominantly in the cytosol, although a small but significant part of myc-Rabip4-(401-600) was associated with the membrane fraction. Similar observations were made with Rabip4-(540 -600) that contains only the FYVE finger (data not shown). These results indicate that the interactions with Rab4 and PtdIns(3)P are not sufficient to stabilize the protein association with the membranes.
We then tested whether the RUN domain in the N-terminal region of Rabip4 was involved in this process. To approach the role of this domain, we looked at the subcellular distribution of myc-Rabip4- (1-212), a protein corresponding to the N-terminal part of Rabip4, which was found associated to the pellet fraction (Fig. 3B). This result indicates that the N-terminal region Supernatant is referred as cytosol (Cy), and the pellet (P) contained all the membrane-associated proteins. Equal amounts of the pellet fraction were further treated as indicated with 1% Triton X-100, 60 mM octylthioglucoside (OTG), 1 M NaCl, 4 M urea, or 10 mM Na 2 CO 3 , pH 10.5, for 1 h at 4°C. The non-extracted proteins, recovered by ultracentrifugation, were separated by SDS-PAGE together with equal amounts of homogenate (H), cytosol (Cy), and of the non-treated membrane-associated fraction (P). Endogenous Rabip4 and Rab4 as well as the overexpressed myc-Rabip4 were detected using antibodies to Rabip4 (upper panel), Myc (middle panel), and Rab4 (lower panel). Homogenates of mock-transfected cells or cells transiently transfected with pcDNA3-Rabip4 were run in parallel (left columns, top panel). B, the N-terminal part but not the FYVE finger triggers Rabip4 to the membrane-associated fraction. CHO cells were transiently transfected with pcDNA3-myc-Rabip4 (for the expression of the total protein or of the constructs schematized in the right part). Cells were fractionated as described in A. Equal cellular amounts of proteins from homogenate (H), cytosol (Cy), and membrane-associated fraction (P) were separated by SDS-PAGE, and the Myc-tagged protein domains were detected using anti-Myc antibodies. The levels of overexpression were identical in the three conditions. of Rabip4 was involved in the subcellular localization of Rabip4.
To confirm that the FYVE finger was not sufficient for the targeting of Rabip4 to the membrane-associated fraction, CHO cells expressing myc-Rabip4 were treated or not with wortmannin, an inhibitor of PtdIns(3)kinase. Cells were then either fractionated into cytosol and membrane pellets or analyzed by confocal microscopy. Both analyses indicated that wortmannin did not modify the association of Rabip4 with the membrane fraction, whereas it induced the redistribution of EEA1 from the pellet to the cytosol (Fig. 4A). We reported previously (Fig.  3) that the myc-Rabip4 protein associated with the membrane pellet was only partly detergent-soluble. We studied whether a wortmannin treatment modified this property. The membraneassociated fraction obtained from control or wortmannintreated cells was treated with 1% Triton X-100 before recovering the soluble and insoluble fractions by ultracentrifugation and analyzing equal amounts of both fractions by SDS-PAGE. As visible in Fig. 4, A and B, a wortmannin treatment of the cells yielded a marked enrichment of myc-Rabip4 in the Triton X-100-insoluble fraction. Thus, the association between Rabip4 and the Triton X-100-soluble material (probably the lipid membrane bilayers) requires an interaction between Rabip4 FYVE finger and PtdIns(3)P since it is sensitive to wortmannin. In parallel, we observed, by confocal microscopy, that wortmannin modified the labeling given by GFP-Rabip4. In control cells, GFP-Rabip4 labeled punctated structures uniformly distributed throughout the cytoplasm, as also observed in Fig. 2B. By contrast, in wortmannin-treated cells, GFP-Rabip4 was present in smaller structures, similar to those observed with GFP-Rabip4 His 554 -His 555 (Fig. 2B), forming sort of "honeycomb" network (Fig. 4C).

The N-terminal Part of Rabip4, Containing a RUN Domain, Is Involved in Its Association with the Triton X-100-insoluble
Fraction-To determine which domain of Rabip4 was required for its association with the Triton X-100-insoluble material, we examined whether Rabip4-(1-212), which appears to trigger Rabip4 association to the membranes, was also responsible for the partial Triton X-100 insolubility. CHO cells expressing myc-Rabip4-(1-212) were submitted to subcellular fractionation, and the pellet was treated with 1% Triton X-100. Solubilized and non-solubilized material was then analyzed by Western blot using an anti-Myc antibody. The totality of the membrane-associated myc-Rabip4-(1-212) was recovered in the Triton X-100-insoluble fraction (Fig. 5A). To strengthen the fact that the amino acids 1-212 of Rabip4 were involved in the targeting of Rabip4 to its intracellular localization, we looked for the colocalization of Rabip4 and Rabip4-(1-212) when overexpressed in CHO cells (Fig. 5B). We observed that GFP-Rabip4 was colocalized with myc-Rabip4-(1-212) in punctated structures as indicated by their yellow color, with very few green structures containing only GFP-Rabip4. Further myc-Rabip4-(1-212) labeling defined a filamentous network giving rise to a honeycomb appearance as emphasized in the enlarged square of Fig. 5B and already visible in Fig. 4C in cells treated with wortmannin. Taken together, these results suggest that the N-terminal part of Rabip4 first determines the subcellular localization of the protein, whereas other domains, such as the FYVE finger, might be secondary to other types of interactions around its previous localization.
In accordance with the fact that the FYVE finger of Rabip4   FIG. 4. Effects of a wortmannin treatment on Rabip4 localization. A and B, wortmannin induces the redistribution of Rabip4 from the Triton X-100-soluble fraction to the Triton X-100-insoluble fraction. CHO cells transiently transfected with pcDNA3-myc-Rabip4 were treated without (control) or with wortmannin (100 nM) for 30 min. Cells were homogenized and submitted to subcellular fractionation as described under "Experimental Procedures" to obtain the cytosolic (Cy) and the membrane-associated (P) fractions. These pellet fractions were further treated for 1 h at 4°C with 1% Triton X-100 before recovering the soluble (T X-100-s) and the insoluble fractions (T X-100-i) by ultracentrifugation. Equal cellular amounts for each fraction were analyzed by Western blotting using anti-Myc and anti-EEA1 antibodies. Representative autoradiograms are shown (left), and the distribution of myc-Rabip4 between the Triton X-100-soluble and -insoluble fractions was quantified in four independent experiments (right). C, wortmannin modifies GFP-Rabip4 labeling. CHO cells stably expressing GFP-Rabip4 were treated without (control) or with 100 nM wortmannin for 30 min. Cells were analyzed by confocal fluorescent microscopy as described under "Experimental Procedures." The bar represents 1 m.

FIG. 5. The N terminus of Rabip4 is involved in the Triton X-100 insolubility of the protein.
A, the membrane-associated myc-Rabip4-(1-212) is detergent-insoluble. CHO cells were transiently transfected with pcDNA3-myc-Rabip4-(1-212). Cells were homogenized and submitted to subcellular fractionation and Triton X-100 solubilization as described in Figs. 3 and 4. The presence of myc-Rabip4-(1-212) was detected in each fraction by using anti-Myc antibodies. B, Myc-Rabip4-(1-212) is present in GFP-Rabip4-containing vesicles. CHO cells were cotransfected with pcDNA3-myc-Rabip4-(1-212) and pEGFP-Rabip4. Cells were treated for confocal analysis as described under "Experimental Procedures." myc-Rabip4-(1-212) was detected using an anti-Myc antibodies followed by Texas Red-coupled anti-mouse antibodies. The figure shows the merged image of GFP-Rabip4 (in green) and myc-Rabip4-(1-212) (in red); the yellow color results from the overlay of green and red. The bar represents 1 m. In the inset an enlargement of a cellular part shows the punctated structures aligned on a network with a honeycomb appearance.
was not the only determinant for its endosomal localization, we noticed that EEA1 and GFP-Rabip4 define different microdomains at the level of the same endosomal vesicles (Fig. 6). Although GFP-Rabip4 was present together with EEA1 in the same endosomes (nearly all the vesicles were positive for both GFP-Rabip4 and EEA1), we observed at a higher magnification that EEA1 (red) and GFP-Rabip4 (green) labeled different regions within endosomes. By contrast, GFP-Rabip4 and Rab4 uniformly labeled the endosomes that were enlarged by the overexpression of wild type Rab4. Similar types of observations have already led to the conclusion that in the same endosomal vesicle Rab5 (whose effector is EEA1) and Rab4 were present in distinct microdomains (29).
Rabip4 Was Not Present in Lipid Rafts-Lipid rafts are microdomains found at the plasma membrane and intracellular membrane organelles, which are experimentally recovered in a detergent-insoluble fraction (30). To test whether Rabip4 was present in these microdomains, we first searched for a colocalization of Rabip4 with proteins known to be enriched in lipid rafts (Fig. 7). By immunofluorescence confocal analysis, we found that GFP-Rabip4 (green) did not colocalize with endogenous caveolin (red) and that myc-Rabip4 (red) did not colocalize either with a GFP-GPI-anchored protein (green). Furthermore, when lipid rafts were purified by a flotation assay, endogenous caveolin floated on the sucrose cushion while Rabip4 remained at the bottom of the tube (Fig. 7B).
Rabip4 Was Not Directly Associated with Actin but Vesicles Containing Rabip4 Are Linked to Actin Cytoskeleton-Actin cytoskeleton is characterized by its Triton X-100 insolubility. To determine whether Rabip4 was associated with actin filaments, CHO cells expressing GFP-Rabip4 were treated or not with the F-actin depolymerizing agent, latrunculin B (31). As shown in Fig. 8A, although latrunculin B led to the disappearance of F-actin, GFP-Rabip4 labeled, as in control cells, punctated structures scattered throughout the cytoplasm of cells, which appeared smaller due to the disruption of actin cytoskeleton (Fig. 8A, right). Latrunculin B did not modify either the subcellular localization of Rabip4 or the amount of the protein found in the Triton X-100-insoluble fraction (Fig. 8A, left). Furthermore, all attempts to copurify actin filaments and Rabip4 were negative (data not shown). This series of results argues strongly against a direct interaction between F-actin and Rabip4. By contrast to latrunculin B, cytochalasin D does not totally disrupt F-actin, but induces its fragmentation in small filaments and its clustering in several regions of the cytoplasm (31) (Fig. 8B). In parallel, we observed a redistribution of GFP-Rabip4-containing vesicles when cells were treated with cytochalasin D. Furthermore, GFP-Rabip4-containing structures are enriched in the same region where F-actin was clustered.
Taken together, these observations indicate that, although Rabip4 does not directly interfere with actin-filaments, Rabip4containing vesicles might be linked to actin cytoskeleton. DISCUSSION We have characterized the implication of various domains of Rabip4 for the protein subcellular localization and function. A series of evidences suggest that the FYVE finger of Rabip4 and the Rab4-binding region are not sufficient to trigger the protein to its endosomal localization, although they play a role in its ability to enlarge the size of early endosomes. First, the protein was associated to endosomes, even when null mutations were FIG. 7. Rabip4 is not associated with "lipid rafts." A, Rabip4 was not colocalized with caveolin or a GPI-anchored protein. CHO cells were transiently transfected with pEGFP-Rabip4 (left) or cotransfected with pEGFP-GPI and pcDNA3-myc-Rabip4. Cells were treated for confocal analysis as indicated under "Experimental Procedures." Endogenous caveolin was detected using anti-caveolin monoclonal antibodies, and myc-Rabip4 was detected with anti-Myc monoclonal antibodies followed by Texas Red-coupled antibodies to mouse IgG. The bar represents 1 m. B, Rabip4 does not copurify with lipid rafts on a flotation gradient. CHO cells were transiently transfected with pcDNA3 myc-Rabip4 and solubilized in ice-cold TNE buffer containing 1% Triton X-100. Lipid rafts were purified on a bottom-loaded sucrose step gradient as described under "Experimental Procedures." Fractions were recovered from the top of the gradient and analyzed by Western blotting using anti-Myc or anti-caveolin antibodies. The total lysate (1/3 of the total material) was analyzed in parallel. A typical autoradiogram representative of 3 independent experiments is shown.
FIG. 6. GFP-Rabip4 is not located in the same microdomains as EEA1 in early sorting endosomes. CHO cells were transfected with pEGFP-Rabip4 and pcDNA3-mycRab4 WT (upper panel) or pEGFP-Rabip4 alone (lower panel). Cells were treated for confocal analysis as described under "Experimental Procedures." myc-Rab4 was detected using anti-Myc antibodies, and endogenous EEA1 was detected using specific monoclonal antibodies against EEA1, both followed by Texas Red-coupled anti-mouse antibodies. The figure shows the images obtained for GFP-Rabip4 (in green), mycRab4 or EEA1 (in red), and the merged images. An enlargement of the vesicles in the square is shown in the right panel.
performed into the FYVE finger, as evidenced by confocal microscopy or following fractionation (data not shown). However, these mutations are not silent since the inactivation of the FYVE finger decreases the ability of Rabip4 to enlarge endosomes when expressed together with active Rab4. Second, the C terminus of Rabip4, Rabip4-(401-600), which contains the FYVE finger and the Rab4-binding domain, is cytosolic. Third, a wortmannin treatment that decreases the intracellular level of PtdIns(3)P does not increase the amount of Rabip4 in the cytosol, whereas a similar treatment inhibits the endosomal localization of EEA1 (16,32) and . These observations indicate that the FYVE finger of Rabip4 was not the major determinant for its endosomal localization. Thus we observed that Rabip4 and EEA1 are not totally colocalized within a single endosomal vesicle. This highly suggests that another domain(s) of Rabip4 was responsible for its association with endosomal microdomains that do not necessarily contain PtdIns(3)P. It should be noted that this endosomal localization was not due either to the protein region implicated in the Rab4 interaction since, as noted above, Rabip4-(401-600) contains both the FYVE finger and amino acid sequences needed for Rab4 binding (1) and is cytosolic.
Our results point to an important role of the N-terminal domain of Rabip4 in the subcellular localization of the protein. We show that the N-terminal part of Rabip4 (amino acids 1-212) not only associates with endosomes containing Rabip4 but also with a network of filaments presenting a honeycomb organization. It was also responsible for the partial insolubility of Rabip4 by non-ionic detergents. This part of the molecule contains the newly described RUN domain, which was predicted to constitute the "core" of a globular structure that possibly interacts with other partners. Since it was found in RPIP8, a protein that interacts with the small GTPase Rap2 (19), in a Rab6 effector (ORF37) (27) and in proteins containing FYVE finger (including Rabip4), it has been proposed that the RUN domain might play a role in relation to small GTPase signaling (20). It was suggested that it could be of use to bind small GTPases, since the RUN domain of RPIP8 was interacting with Rap2. It is perhaps not the function of a RUN domain to bind a GTPase, since the RUN domain of Rabip4 does not bind any form of Rab4 (1), Rab5, and Rab11, 2 the three Rabs enriched in early endosomes, nor Rap2 (1). In view of our results, we suggest that the RUN domain might be responsible for an interaction with a filamentous network of an unidentified nature but is probably linked to the actin cytoskeleton, since the disorganization of actin by cytochalasin D profoundly modified the labeling obtained with Rabip4. The proteins that form these filaments are not known yet but would also be present at the surface of endosomes since Rabip4-(1-212) was colocalized with Rabip4 WT on endosomes. This colocalization was not due to a dimerization of the two proteins. Indeed, while Rabip4 homodimerized in the yeast two-hybrid system, Rabip4 WT and (1-212) did not. 2 As discussed above, the FYVE finger and the Rab4-binding site are not sufficient for the association of Rabip4 with endosomes, although those domains play a role in the partitioning of the proteins between the detergent-soluble and -insoluble fractions. In the presence of wortmannin, most of the protein is Triton X-100-insoluble. Furthermore, both Rabip4 mutated on the FYVE finger and Rapib4⌬-(507-517), a form of Rabip4 unable to interact with active Rab4 (1), are found only in the Triton X-100-insoluble fraction. 2 The amounts of Rabip4 present in the membrane detergent extracts are tightly dependent on the presence of PtdIns(3)P and probably of Rab4. Both types of interaction are required since the destruction of only one of these interactions renders Rabip4 totally detergent-insoluble. This would indicate that either the binding of Rabip4 to FIG. 8. Rabip4 is not directly associated with actin, but structures containing Rabip4 are linked to actin cytoskeleton. A, the actin-depolymerizing agent latrunculin B does not modify the distribution of Rabip4. Left, CHO cells were transiently transfected with pcDNA3-myc-Rabip4 and were treated without (control) or with latrunculin B (2 M) for 3 h. Cells were homogenized and submitted to subcellular fractionation and detergent extraction as described in Figs. 3 and 4. Equal amounts of the obtained fractions (H, homogenate; Cy, cytosol; P, pellet, membrane-associated fraction; s, Triton X-100-soluble pellet; I, Triton X-100-insoluble pellet) were separated by SDS-PAGE, and myc-Rabip4 was immunodetected with an anti-Myc monoclonal antibody. Right, CHO cells stably overexpressing GFP-Rabip4 were treated or not with 2 M latrunculin B for 3 h. Cells were treated for confocal analysis as described under "Experimental Procedures." F-actin is labeled using Texas Red-phalloidin. The bar corresponds to 1 M. B, the actin-depolymerizing agent cytochalasin D induces a coclustering of actin and GFP-Rabip4 containing vesicles. CHO cells stably overexpressing GFP-Rabip4 were treated or not with 1 M cytochalasin D for 2 h. Cells were treated as in Fig. 7A. The figure shows GFP-Rabip4 (green), F-actin (red), and the merged image. The bar corresponds to 1 M. PtdIns(3)P through its FYVE finger is required for a subsequent interaction with Rab4, as described for Rab5 and EEA1 (16), or that the binding of Rabip4 with active Rab4 stabilizes the interaction between the FYVE domain and PtdIns(3)P.
We suggest that Rabip4 is triggered in endosomal microdomains enriched in some unidentified cytoskeletal elements. This localization would allow for an efficient recruitment of Rabip4 to the PtdIns(3)P-enriched area of the endosomes (14) and for a concomitant recruitment of active Rab4 to permit the enlargement of endosomes. We don't know yet whether the filamentous structures are necessary for some intracellular movements of endosomes or if they are only needed for a correct targeting of Rabip4. More studies will be necessary to get a better knowledge of the potential partners and/or roles of the RUN domains.