Portal de Periódicos da CAPES

Distinct Rab-binding domains mediate the interaction of Rabaptin-5 with GTP-bound rab4 and rab5

1998; Springer Nature; Volume: 17; Issue: 7 Linguagem: Inglês

10.1093/emboj/17.7.1941

1460-2075

Gaetano Vitale, Vladimir Rybin, Savvas Christoforidis, Per-Öve Thornqvist, Mary McCaffrey, Harald Stenmark, Marino Zerial,

Erythrocyte Function and Pathophysiology

Article1 April 1998free access Distinct Rab-binding domains mediate the interaction of Rabaptin-5 with GTP-bound rab4 and rab5 Gaetano Vitale Gaetano Vitale European Molecular Biology Laboratory, Postfach 10.2209, D-69012 Heidelberg, Germany Search for more papers by this author Vladimir Rybin Vladimir Rybin European Molecular Biology Laboratory, Postfach 10.2209, D-69012 Heidelberg, Germany Search for more papers by this author Savvas Christoforidis Savvas Christoforidis European Molecular Biology Laboratory, Postfach 10.2209, D-69012 Heidelberg, Germany Search for more papers by this author Per-Öve Thornqvist Per-Öve Thornqvist Department of Biochemistry, University College, Cork, Ireland Search for more papers by this author Mary McCaffrey Mary McCaffrey Department of Biochemistry, University College, Cork, Ireland Search for more papers by this author Harald Stenmark Harald Stenmark Department of Biochemistry, The Norwegian Radium Hospital, Oslo, Norway Search for more papers by this author Marino Zerial Corresponding Author Marino Zerial European Molecular Biology Laboratory, Postfach 10.2209, D-69012 Heidelberg, Germany Search for more papers by this author Gaetano Vitale Gaetano Vitale European Molecular Biology Laboratory, Postfach 10.2209, D-69012 Heidelberg, Germany Search for more papers by this author Vladimir Rybin Vladimir Rybin European Molecular Biology Laboratory, Postfach 10.2209, D-69012 Heidelberg, Germany Search for more papers by this author Savvas Christoforidis Savvas Christoforidis European Molecular Biology Laboratory, Postfach 10.2209, D-69012 Heidelberg, Germany Search for more papers by this author Per-Öve Thornqvist Per-Öve Thornqvist Department of Biochemistry, University College, Cork, Ireland Search for more papers by this author Mary McCaffrey Mary McCaffrey Department of Biochemistry, University College, Cork, Ireland Search for more papers by this author Harald Stenmark Harald Stenmark Department of Biochemistry, The Norwegian Radium Hospital, Oslo, Norway Search for more papers by this author Marino Zerial Corresponding Author Marino Zerial European Molecular Biology Laboratory, Postfach 10.2209, D-69012 Heidelberg, Germany Search for more papers by this author Author Information Gaetano Vitale1, Vladimir Rybin1, Savvas Christoforidis1, Per-Öve Thornqvist2, Mary McCaffrey2, Harald Stenmark3 and Marino Zerial 1 1European Molecular Biology Laboratory, Postfach 10.2209, D-69012 Heidelberg, Germany 2Department of Biochemistry, University College, Cork, Ireland 3Department of Biochemistry, The Norwegian Radium Hospital, Oslo, Norway *Corresponding author. E-mail: [email protected] The EMBO Journal (1998)17:1941-1951https://doi.org/10.1093/emboj/17.7.1941 PDFDownload PDF of article text and main figures. ToolsAdd to favoritesDownload CitationsTrack CitationsPermissions ShareFacebookTwitterLinked InMendeleyWechatReddit Figures & Info Rabaptin-5 functions as an effector for the small GTPase Rab5, a regulator of endocytosis and early endosome fusion. We have searched for structural determinants that confer functional specificity on Rabaptin-5. Here we report that native cytosolic Rabaptin-5 is present in a homodimeric state and dimerization depends upon the presence of its coiled-coil predicted sequences. A 73 residue C-terminal region of Rabaptin-5 is necessary and sufficient both for the interaction with Rab5 and for Rab5-dependent recruitment of the protein on early endosomes. Surprisingly, we uncovered the presence of an additional Rab-binding domain at the N-terminus of Rabaptin-5. This domain mediates the direct interaction with the GTP-bound form of Rab4, a small GTPase that has been implicated in recycling from early endosomes to the cell surface. Based on these results, we propose that Rabaptin-5 functions as a molecular linker between two sequentially acting GTPases to coordinate endocytic and recycling traffic. Introduction Vesicular transport between organelles occurs with high fidelity and ensures that the biochemical composition of each intracellular compartment is accurately controlled. In order to maintain organelle homeostasis, the traffic of vesicles which fuse with a given compartment needs to be balanced with the flow of vesicles which bud from it. Early endosomes function as a primary sorting station in the endocytic pathway (Gruenberg and Maxfield, 1995; Mellman, 1996). Endocytosed molecules which enter early endosomes can be directed to the degradative pathway or be recycled back to the plasma membrane, either directly or through the perinuclear recycling endosomes (Yamashiro and Maxfield, 1984; Ghosh et al., 1994; Hopkins et al., 1994; Ullrich et al., 1996). These processes are mediated by a continuous traffic of vesicular and tubular intermediates which needs to be coordinated to ensure proper progression of cargo through the different compartments while preserving their identity, proportions and dynamic properties. Many molecules have been identified that function at different steps of membrane transport (Rothman, 1994). Among others, members of the Rab family of small GTPases play a central role in regulating vesicular traffic (Pfeffer, 1994; Novick and Zerial, 1997). Several Rab family members have been localized to distinct compartments of the endocytic pathway and play different roles in endocytosis and recycling (Chavrier et al., 1990; van der Sluijs et al., 1991; Lombardi et al., 1993; Lütcke et al., 1993; Olkkonen et al., 1993; Ullrich et al., 1996). For example, Rab5 and Rab4 both localize to early endosomes but exert opposite effects on the uptake of membrane-bound proteins such as the transferrin receptor. Overexpression of Rab5 increases the rate of fluid phase and receptor-mediated endocytosis and concomitantly induces an increase in the size of early endosomes, suggesting that this GTPase regulates the clathrin-coated pathway of receptor internalization and transport into the early endosomes (Bucci et al., 1992). In contrast, cells overexpressing Rab4 exhibited normal kinetics of endocytosis but displayed a reduced uptake of fluid-phase markers and of transferrin-bound iron, and a striking accumulation of transferrin receptor on the cell surface (van der Sluijs et al., 1992). In the light of these data, Rab4 has been implicated in the regulation of membrane recycling from the early endosomes to the recycling endosomes or directly to the plasma membrane (Daro et al., 1996). It is not yet clear whether Rab4 and Rab5 reside on the same early endosome, or localize to distinct endosomes or subcompartments of this organelle. Nevertheless, given the complementary roles of Rab4 and Rab5 an interesting question posed by these studies is whether the two GTPases function independently or have coordinated activity. Elucidating the functional mechanism of Rab5 and Rab4 requires the identification and molecular characterization of their regulators and effectors. To this end, we have previously identified a Rab5-interacting protein named Rabaptin-5 (Stenmark et al., 1995b). Several lines of evidence indicate that Rabaptin-5 functions as an effector protein for Rab5. First, Rabaptin-5 binds directly to the GTP-bound form of Rab5 and is recruited to early endosomes by Rab5 in a GTP-dependent manner (Stenmark et al., 1995b). Secondly, Rabaptin-5 stabilizes Rab5 in the GTP-bound active form by down-regulating GTP hydrolysis (Rybin et al., 1996). Thirdly, its overexpression causes morphological alterations of the early endosomal compartment similar to those induced by Rab5 (Stenmark et al., 1995b). Finally, it is required for the homotypic fusion between early endosomes as well as for the heterotypic fusion of clathrin-coated vesicles with early endosomes in vitro (Stenmark et al., 1995b; Horiuchi et al., 1997). The importance of Rabaptin-5 in endocytic transport is underlined by the finding that fragmentation of endosomes and inhibition of the endocytic pathway in apoptosis is induced by the selective cleavage and inactivation of Rabaptin-5 by proteases of the caspase family (Cosulich et al., 1997). In addition, we have recently shown that cytosolic Rabaptin-5 is complexed to a 60 kDa protein, Rabex-5, which functions as a guanine-exchange factor for Rab5, suggesting that the complex couples nucleotide exchange to effector recruitment on the membrane (Horiuchi et al., 1997). In order better to understand the functional role of Rabaptin-5, we have characterized the molecule by mapping some of its structural determinants. We show that both cytosolic and recombinant Rabaptin-5 exist in a homodimeric state and that its coiled-coil predicted sequences serve as self-interacting determinants. Furthermore, we report that the Rab5-binding domain (R5BD) is located in the C-terminus of Rabaptin-5 and demonstrate that it is necessary and sufficient for Rab5-mediated recruitment of the protein on early endosomes. Unexpectedly, we identified a second Rab-binding domain (RBD) in the N-terminus of Rabaptin-5 which mediates the direct interaction with the GTP-bound, active form of Rab4. Our findings describe a novel type of structural organization for a Rab protein effector and suggest a role for Rabaptin-5 in coordinating both import to, and export from, the early endosomal compartment. Results N- and C-terminal coiled-coil regions mediate Rabaptin-5 homodimerization The structural analysis of the Rabaptin-5 primary amino acid sequence by computer-assisted predictions reveals that its N- and C-terminal regions contain heptad repeats characteristic of coiled-coil domains (Stenmark et al., 1995b). Two regions at the N-terminus of Rabaptin-5 (residues 12–93, CC1–1; residues 191–262, CC1–2) and two at the C-terminus (residues 551–656, CC2–1; residues 672–812, CC2–2) display the highest probability of forming coiled-coil structures (Figure 1A). Coiled-coil domains consist of seven amino acid repeats where positions one and four are predominantly non-polar and positions five and seven are mainly polar. Their overall secondary structure is α-helical and the interaction between two helices result in efficient burial of the hydrophobic side-chains at their interface. Figure 1.(A) Schematic organization of the putative coiled-coil sequences of Rabaptin-5, and (B) sequence alignment between Rabaptin-5 and three related protein sequences at their respective C-termini. (A) Histogram indicating the probability of forming α-helical coiled-coil structures in Rabaptin-5 as determined using the Paircoil program (Berger et al., 1995). The y-axis shows the predicted probability (0–1 represents 0–100%) of α-helical coiled-coil formation, while the x-axis indicates the amino acid number for Rabaptin-5. Rabaptin-5 regions with a predicted probability >50% of forming coiled-coil structures are indicated as grey areas in the scheme below the histogram. (B) Alignment of Rabaptin-5 (1_Rbp), with the protein translations of the DDBJ/EMBL/GenBank database sequences U34932 (2_Fra), U13070 (3_Cel), and U17585 (4_Aca) at their respective C-termini. The sequence alignment was performed with the program PileUp, and displayed with the PrettyPlot program of the GCG package. Numbers represent amino acid positions. Download figure Download PowerPoint We verified the presence of α-helices in Rabaptin-5 experimentally by analysing its overall secondary structure in aqueous solution using circular dichroism (CD) spectroscopy. Under these conditions recombinant Rabaptin-5 reveals a high content of α-helical secondary structure. The helical population calculated by the method of Cheng et al. (1974) from the ellipticity value at 222 nm is ∼70% (Figure 2A). This confirms that the overall Rabaptin-5 secondary structure is compatible with the presence of large coiled-coil domains. Figure 2.(A) CD analysis of recombinant His6::Rabaptin-5, and (B) SDS–PAGE electrophoresis and anti-Rabaptin-5 immunoblot of cross-linked products from cytosol and recombinant His6::Rabaptin-5. (A) CD spectrum was recorded from 1 μM His6::Rabaptin-5, 10 mM phosphate-buffered solution (pH 7.2) in the presence of 0.5 mM 2-mercaptoethanol; the y-axis shows the ellipticity value ϑ (expressed in deg cm2/dmol), while the x-axis indicates wavelength (in nm). (B) 20 μg of HeLa cytosol (lanes 1–5) and 40 ng of recombinant His6::Rabaptin-5 (lane 6) were incubated with the indicated concentrations of the covalent cross-linker BS3 and subsequently analysed by SDS–PAGE followed by immunoblotting using affinity-purified anti-Rabaptin-5 antibodies. Download figure Download PowerPoint Given that proteins such as myosins (Harrington and Rodgers, 1984) and lamins (Heitlinger et al., 1991) form stable homodimers through their coiled-coil domains, we used chemical cross-linking to test the ability of cytosolic Rabaptin-5 to dimerize. Upon Bis(Sulfosuccinimidyl) suberate (BS3) cross-linking, SDS–PAGE and immunoblot analysis (Figure 2B) a major cross-linked product of Rabaptin-5 from cytosol migrated at the apparent Mr of ∼200 kDa, as expected for a homodimer. Cross-linking of Rabaptin-5 dimers was titratable and quantitative up to 1.5 mM BS3. A band of similar size was also detected with purified recombinant Rabaptin-5, indicating that both the recombinant and the cytosolic protein exist in a dimeric state. At the highest concentration of cross-linking reagent a number of bands at molecular mass >200 kDa appeared both with cytosolic and recombinant Rabaptin-5. These bands may correspond to inter- and intra-molecular cross-linked products. The existence of Rabaptin-5 homodimers was confirmed by fractionating bovine brain cytosol and recombinant Rabaptin-5 using a combination of sedimentation on a linear glycerol gradient and gel-filtration chromatography on a Superose-6 column. From the estimated Stokes radius Rs = 12 nm and S20,w = 6.5 we have previously calculated the Mr of cytosolic Rabaptin-5 to be ∼330 kDa (Horiuchi et al., 1997). Using the same approach we estimated the Stokes radius Rs = 9 nm and S20,w = 4.8 for recombinant Rabaptin-5 alone, corresponding to a Mr of ∼200 kDa, consistent with the formation of homodimers. The difference between the molecular mass of cytosolic and recombinant Rabaptin-5 is in agreement with the fact that in vivo, Rabaptin-5 is complexed to other cytosolic proteins such as Rabex-5 and p50 (Horiuchi et al., 1997). Thus, the hydrodynamic properties we measured provide further support for the homodimerization of Rabaptin-5. To test the role of the individual predicted coiled-coil sequences in Rabaptin-5 dimerization, we made use of the yeast two-hybrid system. We cloned Rabaptin-5 full-length cDNA as well as the four putative coiled-coil sequences (CC) in both ‘bait’ (as C-terminal fusion with the LexA DNA-binding domain) and ‘prey’ (as C-terminal fusion with the Gal4 activation domain) two-hybrid vectors and transformed them in the yeast reporter strain L40. As a control for specificity we used a LexA fusion bait with the Coil1B domain of lamin C (McKeon et al., 1986). The trans-activation of the HIS3 reporter upon bait–prey interaction was revealed by the ability of the yeast transformants to grow on histidine-lacking medium. As shown in Figure 3A, Rabaptin-5 interacts with itself but not with the Coil1B domain of laminC. The individual CC sequences could interact both with full-length Rabaptin-5 (Figure 3A) and with themselves (Figure 3C), whereas no interaction was detected with the Coil1B domain of lamin C (Figure 3A). These results confirm that Rabaptin-5 can homodimerize and further indicate that the predicted CC sequences function as self-interacting determinants. Figure 3.Identification of Rabaptin-5 self-interacting determinants by the yeast two-hybrid system. HIS3 reporter gene activation caused by specific interactions between various bait and prey Rabaptin-5 fusions in the two-hybrid system. L40 reporter yeast cells co-transformed with Gal4 RNA polymerase II activation domain (Gal4AD::) fusions in the 2μ/LEU2 pGAD10 plasmid and LexA DNA-binding domain fusions in the 2μ/TRP1 pLexA plasmid (LexA::) of the proteins indicated were spotted on synthetic complete medium lacking tryptophan and leucine (his+) or lacking tryptophan, leucine and histidine (his−). (A) Rabaptin-5 self-interaction and interaction between the full-length molecule and the four individual coiled-coil predicted sequences; a LexA bait fusion (LexA::) with the Coil1B domain of laminC was used as control for non specific coiled-coil interactions. (B) Self-interaction between the individual coiled-coil predicted sequences. Download figure Download PowerPoint 73 amino acid residues in Rabaptin-5 are necessary and sufficient for interaction with Rab5 and for Rab5-dependent recruitment on early endosomes Rabaptin-5 was originally identified in a two-hybrid screen as clone L1_46, which encodes the C-terminus of the molecule (amino acids 551–862) and is responsible for the interaction with GTP-bound Rab5 (Stenmark et al., 1995b). In order to gain more information about this interaction, we mapped the sequence within the C-terminus of Rabaptin-5, which interacts with Rab5. By databank homology searches, we found that Rabaptin-5 shares homology with three sequences, one from rat and two from the nematodes Caenorhabditis elegans and Angiostrongylus cantonensis (Figure 1B), with a pronounced degree of identity (∼75%) at their respective C-termini (residues 780–836 for Rabaptin-5). The rat sequence corresponds to a Rabaptin-5-like molecule that has been independently identified as a Rab5-interacting protein (Horiuchi et al., 1997; Gournier et al., 1998). The C.elegans open reading frame has not been characterized so far, while the sequence from A.cantonensis encodes a muscle-associated protein specifically expressed in adult female worms (Joshua and Hsieh, 1995). The high sequence conservation of this region and its position in the C-terminus of Rabaptin-5 made it a candidate for a R5BD. To test this possibility, we analysed a series of N-terminal L1_46 deletion mutants (Figure 4a), both in the two-hybrid system for their ability to interact with the GTPase-deficient Rab5Q79L mutant and for recruitment on early endosomes as C-terminal fusions with a cytosolic reporter, the Green-Fluorescent protein (GFP) of Acquorea victoria. By the yeast two-hybrid system, we detected interaction between Rab5Q79L and a C-terminal fragment of Rabaptin-5 corresponding to residues 789–862 (Figure 4a). A deletion within this sequence (L1_46 Δ809–832) abolished the interaction, indicating that this region contains structural elements which are necessary for the association with Rab5. We confirmed this interaction by fluorescence microscopy in HeLa cells co-expressing Rab5Q79L and fusions between GFP and Rabaptin-5 C-terminal sequences. GFP alone gave a diffuse cytosolic localization when co-expressed with Rab5Q79L (Figure 4b, panel A). In contrast, GFP localized on enlarged early endosomes induced by the expression of the Rab5 mutant when fused to the L1_46 protein (Figure 4b, panel B). Among the L1_46 deletions tested (Figure 4b, panels C–F), the Rabaptin-5 789–862 mutant, which could interact with Rab5Q79L in the yeast two-hybrid system (Figure 4a), was also recruited on early endosomes (Figure 4b, panel E). Interestingly, a further deletion of the 30 most C-terminal amino acids (Rabaptin-5 789–832) did not impair membrane association. Although the region comprising residues 789–832 did not support interaction in the yeast two-hybrid system (Figure 4a), it appears that this fragment nevertheless allows recruitment on endosomal membranes in vivo. All GFP fusion proteins showed diffuse cytosolic staining when expressed alone (not shown). Altogether, the combination of yeast two-hybrid system and fluorescence microscopy data indicates that the R5BD lies within a 73 amino acid sequence in the C-terminus of Rabaptin-5. This region is conserved among other Rabaptin-like proteins from rat and nematodes, and is sufficient for Rab5-dependent recruitment of the protein on early endosomal membranes. Figure 4.(a) Mapping of the R5BD of Rabaptin-5 by the yeast two-hybrid system, and (b) confocal fluorescence microscopy of cells co-expressing Rab5Q79L and GFP or GFP::Rabaptin-5 fusions. (A) HIS3 reporter gene activation caused by specific interactions between the indicated Rabaptin-5 deletions Gal4AD fusions in pGADGH and LexA::Rab5Q79L, tested as growth in synthetic medium lacking tryptophan, leucine and histidine (his−). (b) BHK cells infected with T7 RNA polymerase recombinant vaccinia virus (vT7), were co-transfected with Rab5Q79L (panels A–F), and: (panel A) the Green Fluorescent protein (GFP); (panel B) GFP::Rabaptin-5 551–862; (panel C) GFP::Rabaptin-5 739–862; (panel D) GFP::Rabaptin-5 789–862; (panel E) GFP::Rabaptin-5 789–832; or (panel F) GFP::Rabaptin-5 807–862. The cells plated on 11 mm glass coverslips were infected and transfected for 3.5 h as described previously (Stenmark et al., 1994b), incubated at 30°C for 30 min, fixed and viewed with the confocal microscope developed at EMBL, also as described previously (Bucci et al., 1992). Arrows indicate typical membrane structures formed upon Rab5Q79L expression and decorated (panels B–E) with the GFP::Rabaptin-5 fusion proteins. Download figure Download PowerPoint Rab4 interacts with Rabaptin-5 via a distinct RBD Next we tested whether the C-terminal R5BD is the only RBD present in Rabaptin-5. The interaction specificity of the L1_46 protein was previously analysed with a panel of different Rab baits and found to be highly specific for Rab5 (Figure 5A; Stenmark et al., 1995b). Surprisingly, when we extended this analysis to full-length Rabaptin-5 as prey (Figure 5B), we found that the protein strongly interacts with Rab5Q79L but also with the GTPase-deficient Rab4Q67L mutant. Importantly, besides a weak interaction with Rab3a (the significance of which is not clear at the moment), full-length Rabaptin-5 failed to interact with other Rab proteins of the endocytic or biosynthetic pathway (Rab6, Rab7, Rab17 and Rab22; also see below) indicating that Rabaptin-5 does not interact indiscriminately with all family members. Figure 5.Interaction specificity of Rabaptin-5 and C- or N-terminal deletions with Rab family members in the yeast two-hybrid system. Interaction between Rabaptin-5 (A) C-terminal, (B) full-length or (C) N-terminal Gal4BD prey fusions in pGADGH, and various bait LexA::Rab fusions assayed as lacZ reporter gene activity in liquid β-Gal assay. L40 reporter yeast cells transformed with the 2μ/TRP1 pLexA plasmids encoding for the indicated LexA::Rab fusion and with 2μ/LEU2 pGADGH plasmids encoding Gal4 fusions of Rabaptin-5 and the C- or N-terminal deletions, were grown to OD600 values of ∼1.0 in synthetic medium lacking tryptophan and/or leucine. β-galactosidase (β-Gal) activity was then measured, using O-nitrophenyl-β-D-galactoside (Sigma) as a substrate (Guarente, 1983). β-Gal activities (in relative units) are presented in boxes as mean values ±SEM (bars) obtained with three independent transformants. Download figure Download PowerPoint The observation that Rabaptin-5 interacts with Rab4 and Rab5 is intriguing in the light of previous studies showing that both GTPases reside on early endosomes (Chavrier et al., 1990; van der Sluijs et al., 1992). We therefore investigated this interaction further and mapped the Rab4-binding domain (R4BD). The yeast two-hybrid analysis indicated that the deletion mutant Rabaptin-5 551–862 failed to interact with Rab4Q67L (Figure 5A). Conversely, an N-terminal prey construct (amino acids 5–547) tested against the same panel of Rab baits showed an efficient interaction with Rab4 but a very weak or undetectable association with Rab5 and other Rab proteins (Figure 5C; see below). It thus appears that the sequences interacting with Rab4 and Rab5 are located in topologically distinct regions of Rabaptin-5, i.e. in the N- and C-terminus, respectively. In order to map more precisely the localization of the R4BD, we tested several Rabaptin-5 C-terminal deletions in the two-hybrid system for their ability to interact with Rab4Q67L. As shown in Figure 6A, a Rabaptin-5 fragment corresponding to amino acids 5–135 retained the ability to interact with Rab4Q67L. Interestingly, this region does not share significant sequence homology with the region harbouring the R5BD. Figure 6.(A) Mapping of the R4BD of Rabaptin-5 by the yeast two-hybrid system, (B) in vitro binding of cytosolic or recombinant His6::Rabaptin-5 to MBP::Rab4 and (C) cofractionation of Rab4, Rab5 and Rabaptin-5 by size exclusion chromatography. (A) HIS3 reporter gene activation caused by specific interactions between various indicated Rabaptin-5 deletions, Gal4AD fusions and LexA::Rab4Q67L, tested as growth in synthetic medium lacking tryptophan, leucine and histidine (his−). (B) Aliquots of Affi-Gel 15 beads containing immobilized MBP::Rab4 or MBP::Rab6 pre-loaded with 100 μM GDP (lanes 2 and 5), GTPγS (lanes 3 and 6) or buffer alone (free form, lanes 1 and 4) were incubated in the presence of 5 μg His6::Rabaptin-5 or with 200 μg of bovine brain cytosol. After washing, Rabaptin-5 associated with the beads was detected by SDS–PAGE followed by immunoblotting. (C) Following pre-loading with GTPγS, recombinant His-Rab4 and His-Rab5 were incubated with Rabaptin-5 and the reaction mixture was separated on a Superdex-200 column. Proteins in the fractions were detected by measuring absorbance at 280 nm and Western blot analysis (inset for fractions 9–13) using anti-Rabaptin-5, -Rab4 and -Rab5 rabbit polyclonal antibodies. The peak of A280 in fractions 25–30 corresponds to a Mr of 25–27 kDa and contains the excess of free GTPases. Download figure Download PowerPoint To confirm that Rabaptin-5 can bind Rab4 in a nucleotide-dependent manner, a Rab4 N-terminal fusion with maltose-binding protein (MBP::Rab4) was covalently bound on Affi-Gel 15 beads, pre-loaded with GDP or GTPγS, or in a nucleotide-free state, and assayed for binding of recombinant (Figure 6B, upper panel) or cytosolic Rabaptin-5 (Figure 6B, lower panel). In both cases Rabaptin-5 interacted with MBP::Rab4 loaded with GTPγS. Consistent with the two-hybrid analysis shown in Figure 5, only background levels of Rabaptin-5 binding were detected with GDP-loaded MBP::Rab4 or in the absence of nucleotide, or with an MBP::Rab6 fusion, indicating that the interaction with Rab4 is specific and depends on the GTP-bound conformation. Independent biochemical evidence that Rabaptin-5 can interact with both Rab5 and Rab4 in a GTP-dependent manner was obtained by gel-filtration chromatography. Rab5 and Rab4 were pre-loaded with either GDP or GTPγS and then incubated in the presence of recombinant Rabaptin-5. The proteins were fractionated on a Superdex-200 column and analysed by SDS–PAGE and Western blot. The position of Rabaptin-5 on the chromatogram was consistent with the Mr corresponding to the homodimer. As shown in Figure 6C, the fractions corresponding to the Rabaptin-5 homodimer peak also contained both Rab5 and Rab4 pre-loaded with GTPγS. However, no cofractionation was detected when the proteins were pre-loaded with GDP (data not shown). Finally, we examined the subcellular localization of Rab4 and Rabaptin-5 by confocal immunofluorescence microscopy. Since the available antibodies against Rab4 failed to detect endogenous levels of the protein, we co-expressed VSV-G tagged Rab4Q67L or Rab5Q79L with either full-length myc-tagged Rabaptin-5 or various N- and C-terminal deletions in BHK cells, using the T7 RNA polymerase recombinant Vaccinia virus system. As shown in Figure 7A, VSV-G::Rab4Q67L localized to vesicular structures dispersed throughout the cytoplasm which were also labelled for the human transferrin receptor (hTfR), in agreement with previous studies (van der Sluijs et al., 1991; Daro et al., 1996). However, when VSV-G::Rab4Q67L was co-expressed with myc::Rabaptin-5 the two proteins accumulated on enlarged endosomes whose expansion is caused by the overexpression of Rabaptin-5 alone (Figure 7B; Stenmark et al., 1995b). Thus, as predicted on the basis of their interaction, Rabaptin-5 and VSV-G::Rab4Q67L co-localize to the same endosomal compartment. If Rabaptin-5 can bind both Rab5 and Rab4 on the same endosome the two Rab proteins would be expected to co-localize substantially upon overexpression of Rabaptin-5. Indeed, Figure 7C shows that while structures containing predominantly Rab5 or Rab4 proteins are visible, the large endosomes induced by the overexpression of Rabaptin-5 contain both VSV-G::Rab4Q67L and endogenous Rab5. Moreover, when co-expressed, VSV-G::Rab4Q67L and myc:: Rab5Q79L were also found to co-localize to enlarged endosomal structures (Figure 7E). As control, Rab7 (a late endosomal-specific small GTPase) and myc::Rabaptin-5 were clearly found segregated in different compartments (Figure 7D). We did not observe membrane recruitment of two Rabaptin-5 C-terminal deletions (myc::Rabaptin-5 Δ547–862, Figure 7F; and myc::Rabaptin-5 Δ215–862, data not shown) by VSV-G::Rab4Q67L despite the presence of the R4BD. These results suggest that the interaction between Rab4 and the R4BD on the membrane is either not stable or it requires the full-length molecule (see Discussion). Conversely, VSV-G::Rab5Q79L was able to recruit the C-terminal region of Rabaptin-5 containing the R5BD (myc::Rabaptin-5 with the N-terminal Δ1–550 deletion; Figure 7G) on endosome membranes, but not the N-terminus containing the R4BD (myc::Rabaptin-5 Δ551–862; Figure 7H). These data provide further support for the conclusion that the membrane recruitment