Characterization of Novel Rab6-Interacting Proteins Involved in Endosome-to-TGN Transport


Corresponding author: Solange


Rab6 GTPase regulates intracellular transport at the level of the Golgi complex. Using the yeast two-hybrid screen, we have isolated two clones that specifically interact with the three isoforms of Rab6 present in mammalian cells (Rab6A, A′ and B). The cDNAs encode two proteins of 976 and 1120 amino acids (calculated molecular mass of 112 and 128 kDa, respectively) that we named Rab6IP2A and Rab6IP2B (for Rab6 Interacting Protein 2). The two proteins likely correspond to spliced variants of the same gene. Rab6IP2s have no significant homology with other known proteins, including Rab effectors or partners. They are ubiquitously expressed, mostly cytosolic and found in high molecular mass complexes in brain cytosol. We show that Rab6IP2s can be recruited on Golgi membranes in a Rab6:GTP-dependent manner. The overexpression of any form of Rab6IP2 has no detectable effect on the secretory pathway. In contrast, the retrograde transport of the Shiga toxin B subunit between the plasma membrane and the Golgi complex is partly inhibited in cells overexpressing the Rab6-binding domain of Rab6IP2. Our data suggest that Rab6IP2s is involved in the pathway regulated by Rab6A′.

Rab proteins form the largest recognized subset of Ras-like GTPases with almost 60 members having been identified in humans (1). Present on the cytoplasmic surface of every organelle in the biosynthetic/secretory and endocytic pathways, Rab GTPases play a crucial role in membrane–membrane homotypic fusion as well as in the specific targeting/docking of transport intermediates with their acceptor membranes (2–4). They perform their tasks through interactions with multiple effectors, including molecular motors, protein and lipid kinases and the large docking/tethering complexes that regulate the assembly of v/t SNAREs complexes (2,5).

Three members of the Rab6 family, termed Rab6A, A′ and B, have recently been characterized in mammals. Rab6A is an ubiquitous protein associated with Golgi and TGN (trans-Golgi network) membranes (6,7). Strong evidence exists suggesting that Rab6A regulates a COPI-independent Golgi to endoplasmic reticulum (ER) transport pathway (8,9) and is involved in the ER delivery of internalized Shiga and Shiga-like toxins (8–11). Rab6A′ is generated by alternative splicing of a duplicated exon, and like Rab6A, is ubiquitously expressed and localized to the Golgi complex (12). Rab6A and A′ proteins differ only in three amino acids, the two most important being a TV motif present at the end of the switch II region of Rab6A which is replaced by an AA motif in Rab6A′ (12). Unlike the active form of Rab6A (Rab6A Q72L), Rab6A′ Q72L is not able to redistribute Golgi membranes into the ER when overexpressed in cultured cells. In addition, Rab6A′ does not interact with the Golgi-associated kinesin-like protein Rabkinesin-6, thought to be involved in the movement of Rab6A positive tubular structures between the Golgi and the ER (10). Recent evidence indicates that Rab6A′ is likely to be involved in endosomes-to-TGN transport (13). Rab6B, encoded by a separate gene, displays 91% homology to Rab6A/A′ (14) with the differences mainly located in the C-terminus hypervariable domain of the protein, as previously documented for other isoforms of the Rab family (1). Rab6B also localizes to the Golgi apparatus, but is preferentially expressed in a subset of neuronal cells (14). Whether Rab6B fulfills a specific function in brain tissue is presently unknown.

Several Rab6-interacting proteins have been characterized. They include Rabkinesin-6 and GAPCenA, a Rab6 GTPase activating protein (15,16). Another Rab6 partner is PRA1, part of whose sequence was isolated using a yeast two-hybrid screen with Rab6A Q22V as bait (17). PRA1, which also interacts with Rab3A and Rab1, is preferentially localized to the Golgi apparatus (18,19). PRA1 inhibits the extraction of membrane-bound Rab3A by GDI-1, but its precise function is still unknown. Recently, a new partner of Rab6A and A′, called Rab6IP1/ORF37, has been identified. This protein, of unknown function, possesses two domains termed RUN predicted to adopt an α-fold conformation. Interestingly, the RUN domains are found in a variety of Ras-like GTPase effectors, suggesting that they could be involved in common signaling pathways (20).

Here, we report the cloning and characterization of two novel proteins, Rab6IP2A and Rab6IP2B, that specifically interact with GTP-bound forms of Rab6A and A′. Rab6IP2 are cytosolic proteins which are recruited onto Golgi membranes in a GTP-dependent manner and are likely to function in endosomes to Golgi transport.


Cloning of Rab6IP2

To identify interaction partners of Rab6A, the GTPase defective mutant Rab6A Q72L was used as a bait in a yeast two-hybrid screen of a mouse embryo cDNA library (17). The characterization of two proteins identified in this screen, Rabkinesin-6 and GAPCenA, has previously been reported (15,16). Here, we focused on a third clone, ‘clone 1’, a 475-base pair cDNA fragment, which interacts with Rab6A Q72L, but not with the dominant-negative mutant Rab6 T27N, nor with the controls Rab5 Q79L, p21Ras and lamin. In contrast to Rabkinesin-6 and GAPCenA, clone 1 interacts with the Rab6A I46E mutant, which carries a point mutation in the effector domain of Rab6A, as well as with the double mutant Rab6A Q72L I46E (data not shown).

Figure 1 shows a Northern blot analysis of various mouse tissues using the clone 1 cDNA fragment as a probe. A doublet corresponding to about 8 kb messenger RNAs was detected in all tissues, with the possible exception of testis in which a fuzzy signal was found at around 7.5 kb. A cloning strategy using clone 1 as a probe (see Experimental procedures) led to the identification of two clones that contained the identical potential initiator ATG codon, preceded 18 nucleotides upstream by a stop codon in the same reading frame and a consensus sequence for translational initiation. These clones encode two proteins of 976 and 1120 amino acids (calculated molecular mass of 112 and 128 kDa, respectively) that we named Rab6IP2A and Rab6IP2B (for Rab6 Interacting Protein 2; Figure 2A). The sequence corresponding to the original clone 1, that we will refer to as the Rab6-binding domain (Rab6BD), is located between residues 860 and 1015. Rab6IP2A bears a 44-amino acid deletion which includes the first 17 residues of this domain. In addition, the C terminus 8 amino acids of Rab6IP2A, which includes the last 3 of Rab6BD, are not found in Rab6IP2B, which has a 108-residue extension. These two major differences between Rab6IP2A and B are likely to be due to at least two alternative splicings of the mRNA of these proteins, one removing an exon encoding a 44-residue insertion, the other accounting for the premature stop codon. Of note, Rab6IP2B contains the integral RabBD originally isolated by the two-hybrid screen, and consequently it was used in the following experiments.

Figure 1.

Figure 1.

Rab6IP2 is ubiquitously expressed. Northern blot analysis was performed on various mouse tissues using the clone 1 as a probe.

Figure 2.

Figure 2.

Amino acid sequence of Rab6IP2A and B deduced from the full-length cDNA sequence.A. Two clones were characterized by screening of mouse cDNA libraries using the clone 1 isolated by the two-hybrid screen as a probe. The clone 1 sequence, representing the Rab6-binding domain (residues 861–1015) of Rab6IP2, is underlined. Numbers count the residues of Rab6IP2B. The differences between Rab6IP2A and B are shown by the dotted line, corresponding to the deleted portion of 44 amino acids, and by the bold letters, indicating the truncated C terminus of Rab6IP2A. B. Prediction of the coiled-coil regions of Rab6IP2A (top) and Rab6IP2B (bottom).

The protein sequence of Rab6IP2A and 2B is hydrophilic, and the majority (except the first 143 residues of the N terminus, and a short segment in Rab6IP2B located between residues 993 and 1060) is predicted to be α-helical containing heptad repeats characteristic of coiled-coil domains (Figure 2B). Database searches using the NCBI Blast program indicate that Rab6IP2B shares between 62 and 94% identity with four human sequences of unknown function. Interestingly, two human proteins, KIAA0378 and ELKS (accession numbers BAA20832 and BAA88763, respectively), would correspond to Rab6IP2A in that they bear the internal 44-residue deletion, whereas the two others, KIAA1081 and the ‘unnamed protein product’ (accession numbers BAA83033 and BAA90975) would be Rab6IP2B homologs regarding this segment. However, only BAA90975 ends like Rab6IP2B, the other three human clones having the premature stop codon in the same context as Rab6IP2A. In addition, the Caenorhabditis elegans gene product F42A6.9 (accession number T32623) of unknown function shares 27% identity on an 822-amino acid overlap, and the Drosophila melanogaster gene product CG12933 (accession number AAF58932) is a 290-amino acid-long protein which shares 41% identity with mouse Rab6IP2s on a 190-residue overlap. Each of these four proteins contains a highly conserved domain between residues 168 and 206, which is also present in the unique human clone (KIAA0378) spanning this region. Additional homology domains are found between Rab6IP2A and B and the gene product CG12933 as well as between Rab6IP2A and B and the C. elegans F42A6 gene product. However, the homologous domains do not overlap with each other. Another D. melanogaster gene product (AAF58930) of 1456 amino acids shares 27.5% identity with Rab6IP2B on a 575-residue overlap (from residue 464 of Rab6IP2B). No obvious homologous gene product was found in Saccharomyces cerevisiae.

Rab6IP2 binds to the GTP-bound form of Rab6

The specific interaction between Rab6A:GTP and the full-length protein Rab6IP2 was confirmed in the yeast two-hybrid assay. Rab6IP2 was also found to interact with the GTPase-deficient mutant of Rab6A′, a spliced variant of Rab6A [see Figure 8(A) in (12)], as well as with Rab6B:GTP, a neuronal specific isoform of Rab6 [see Figure 8 in (14)]. To confirm these interactions biochemically, we performed pull-down experiments using glutathione-S transferase (GST) fusion proteins with Rab6A, Rab6A′ and Rab6B coupled to glutathione-Sepharose 4B (Figure 3). A mouse brain postnuclear supernatant was incubated in the presence of the GST-Rab6 fusion proteins loaded with GDP or with GTPγS. Endogenous Rab6IP2 was detected using a specific polyclonal antibody, anti-Rab6BD (see below). Rab6IP2 was found to bind all three GST-Rab6 isoforms in their GTP conformation (Figure 3, upper panel). As controls, we showed that GST alone did not bind Rab6IP2, and that endogenous GDI molecules [GDI-1/Rab3A-GDI and GDIβ migrating at around 60 and 50 kDa, respectively (21)] only bound Rab6:GDP, as expected (Figure 3, lower panel). Rab6IP2s did not interact with GST-Rab11, a Rab protein associated with recycling endosomes (data not shown). It should be pointed out that Rab6A′ bound GDI proteins only weakly. Interestingly, differences were observed in the binding capacity of the three Rab6 isoforms: Rab6B and to a lesser extent Rab6A′ were reproducibly found to bind more Rab6IP2 than Rab6A.

Figure 3.

Figure 3.

Rab6IP2 specifically associates to Rab6 in its GTP bound form. The fusion proteins GST-Rab6A, GST-Rab6A′, GST-Rab6B or the GST protein alone were coupled to Sepharose 4B-glutathione beads, loaded with GDP or GTPγS and incubated in the presence of 50 μl of mouse brain postnuclear supernatant (MBP). After SDS-PAGE and blotting, the nitrocellulose membrane was cut and incubated with the anti-Rab6BD antibody, a rabbit polyclonal antibody raised against the Rab6IP2 Rab6-binding domain, or with anti-GDIβ antibody [the anti-GDIβ antibody recognizes both GDIβ (about 50 kDa) and Rab3A-GDI (approximately 60 kDa)]. These antibodies are indicated on the right of the figure. Molecular weight standards are shown on the left. MBP T indicates a portion (5 μl) of nonincubated MBP.

Expression and distribution of Rab6IP2

To study the localization of Rab6IP2, we raised a rabbit antiserum against the Rab6BD polypeptide isolated in the yeast two-hybrid screen. It is most likely that the anti-Rab6BD antibody recognized both Rab6IP2A and Rab6IP2B since our antibody is polyclonal, and directed against the 155-residue-long Rab6-binding domain, integrally present in Rab6IP2B, and present at 87% in Rab6IP2A. As shown in Figure 4, the affinity-purified antibody recognized by immunoblotting 2 (and sometimes 3) species around 130 kDa in various mouse tissue and cultured cell extracts. This apparent molecular weight of 130 kDa is in agreement with the predicted molecular mass of Rab6IP2. The staining pattern obtained by Western blot shows that the level of expression of both isoforms varies from one cell type or tissue to another. In lung and thymus, the antibody recognized a species migrating slightly more slowly than Rab6IP2 in other extracts. This species, also detected at low levels in spleen and kidney, could correspond to post-translational modifications of the proteins and/or to another isoform. In addition, in some cells (immature dendritic cells) or cell lines (BWTG3) and tissues (kidney, spleen, thymus), the antibody recognized a high molecular weight band (over 200 kDa) which may correspond to a Rab6IP2 dimer. A 27-kDa band common to both brain and thymus is also recognized by anti-Rab6BD antibody.

Figure 4.

Figure 4.

Rab6IP2 is present in various amounts and isoforms according to the source tested. Equivalent amounts of various mouse tissue PNS (A) or of cell line total extracts (B) were tested for the presence of the Rab6IP2 by Western blot using anti-Rab6BD antibody. Molecular weight markers are indicated on the left. The cell lines used are as follows: MNT1 is a human melanocyte cell line; BWTG3 a mouse hepatic secretory cell line; LLC-PK1 and MDCK are polarized cell lines derived from pig and dog kidney, respectively; NIH3T3 is a mouse embryo fibroblast cell line; PtK2 is a cell line derived from rat kangaroo kidney; HeLa are human epithelial cells; BHK is a baby hamster kidney-derived cell line; IDc are primary cultures of human immature dendritic cells.

Fractionation of HeLa cells into 100 000 ×g supernatant (cytosol) and pellet (membrane) showed that Rab6IP2 was mainly cytosolic (Figure 5A). Similarly, the affinity-purified antibody gave a weak diffuse immunofluorescence staining throughout the cytoplasm of HeLa cells (Figure 5B, left panel). No membrane staining was detected, except in MDCK cells (Figure 5B, right panel) in which part of Rab6IP2 colocalizes with the medial Golgi marker CTR 433 (data not shown). Fractionation of bovine brain cytosol by gel filtration through a Superdex 200 column (Figure 5C) indicated that Rab6IP2 was present in the form of monomer but also in high molecular weight complexes (around 650 kDa), suggesting that Rab6IP2 molecules either formed oligomers, or were complexed with other protein(s). No detectable amount of Rab6 was found in these complexes (Figure 5C).

Figure 5.

Figure 5.

Rab6IP2 is mainly cytosolic, and forms high molecular mass complexes of around 650 kDa.A. Two subclones of HeLa cells were used to prepare a PNS, which was further fractionated into membrane (mb) and cytosolic (cyt) fractions by centrifugation at 200 000 ×g. Equivalent amounts of the 3 fractions were analyzed by SDS-PAGE, processed for immunodetection using anti-Rab6BD antibody. B. HeLa (left) and MDCK (right) cells were fixed with paraformaldehyde and incubated with anti-Rab6BD antibody. Scale bar indicates 10 μm. C. Bovine brain cytosol was gel filtered using a Superdex 200 column. Fractions were analyzed by SDS-PAGE, blotted on nitrocellulose membrane, which was cut into 3 parts; from top to bottom, these parts were revealed with anti-Rab6BD, with anti-GDIβ, and with anti-Rab6 antibodies. Molecular weight standards are indicated on the left. Size markers indicated with arrows on the top of panel C were used to calibrate the column. The antibodies used for immunodetection are shown on the right. T indicates the nonfractionated brain cytosol.

Rab6IP2 is recruited on Golgi membranes byRab6:GTP

To address the function of Rab6IP2, we constructed a green-fluorescent protein (GFP)-tagged Rab6IP2B, that was transiently expressed in HeLa cells. Like the endogenous protein, GFP-tagged Rab6IP2B was predominantly present in the cytoplasm of transfected cells. However, when cells were cotransfected with the GTPase-deficient mutant Rab6A Q72L, a fraction of GFP-Rab6IP2B colocalized with Rab6 on Golgi membranes (Figure 6 bottom). The same result was obtained by cotransfection of cells with GFP-Rab6IP2B and Rab6A′ Q72L, and when GFP-Rab6IP2B was present in cells overexpressing wild-type Rab6A (data not shown). In contrast, no Golgi localization of GFP-Rab6IP2B was detected when cells were cotransfected with the dominant-negative mutant Rab6A T27N (Figure 6 top). This suggests that Rab6IP2B could be recruited on Golgi membranes in a GTP-dependent manner by Rab6A and A′.

Figure 6.

Figure 6.

Rab6IP2B is recruited by the GTPase defective mutant Rab6A Q72L in vivo. The GFP-Rab6IP2B fusion protein was transiently coexpressed with Rab6A T27N (top), or with Rab6A Q72L (bottom), in HeLa cells. Twenty hours after transfection, cells were fixed, incubated with anti-Rab6 and with a secondary Texas Red coupled anti-rabbit antibody. Scale bar represents 10 μm.

Overexpression of the Rab6-binding domain inhibits the endosome-to-TGN transport pathway

We then tested whether the overexpression of full-length Rab6IP2 or of the Rab6BD would affect Rab6A and/or Rab6A′ function in transport.

The biosynthetic/secretory pathway was shown to be a target of Rab6 proteins: overexpression of the GTPase-deficient mutants of Rab6A and A′ resulted in a decrease of intracellular transport and release in the extracellular medium of the secreted form of alkaline phosphatase (SEAP) (12,22). No significant effect on release of SEAP in the extracellular medium was observed after overexpression of Rab6IP2B or of Rab6IP2 Rab6BD (data not shown). The maturation and intracellular transport of newly synthesized SEAP also remained unaffected; the morphology of the Golgi apparatus was unchanged (data not shown).

Rab6A′ was recently shown to be involved in the direct transport pathway connecting early/recycling endosomes to the TGN (13). As a transport marker for this pathway, we used the B subunit of the Shiga toxin bearing two sulfation sites (B-(Sulf)2) to monitor its arrival into the TGN. As shown by Figure 7, sulfate incorporation into B-(Sulf)2 was greatly reduced in cells overexpressing Rab6A′ T27N compared to cells transfected by the vector alone; interestingly, Rab6BD overexpression was also able to produce a similar effect but to a lesser extent. 61.1 ± 5.5% (n = 6) vs. 35 ± 10% (n = 7) B-(Sulf)2 sulfation was left in Rab6BD and Rab6A′ T27N-overexpressing cells, respectively. No significant change in B-(Sulf)2 sulfation was observed in cells overexpressing Rab6IP2A (Figure 7) or Rab6IP2B (not shown).

Figure 7.

Figure 7.

Overexpression of the Rab6-binding domain of Rab6IP2 (Rab6BD) partly inhibits the retrograde transport of Shiga toxin B subunit to the TGN. HeLa cells were transfected using the vaccinia virus system with plasmids encoding the various proteins as indicated at the bottom of the figure. After transfection, B-(Sulf)2 was allowed to bind for 30 min at 4 °C, cells were then incubated at 37 °C for 30 min in the presence of 35S sulfate. Cell lysates were immunoprecipitated with an antibody against the B subunit of Shiga toxin, samples analyzed by Tris/Tricine gel electrophoresis followed by autoradiography. Results are expressed as a percentage of sulfated B-(Sulf)2 compared to the control (empty vector) condition. The means (± SE) of 6 (Rab6BD) and 7 (Rab6A′ T27N) independent experiments are shown; an average of 2 experiments is indicated for Rab6IP2A.


Here, we report the cloning and characterization of two novel proteins, Rab6IP2A and B, that specifically interact with Rab6. The two proteins differ in 144 amino acids and probably correspond to spliced variants of the same gene. In addition, the fact that several species were detected by Western blotting using affinity-purified antibody suggests either that other variants exist, or that Rab6IP2s are post-translationally modified.

No obvious homology was found between Rab6IP2s and the other Rab-interacting proteins identified thus far (3), illustrating the fact that Rab GTPases are able to interact with a wide variety of proteins. However, Rab6IP2s belong to the class of cytosolic Rab partners and effectors that can be recruited on membranes in a GTP-dependent manner. In mammalian cells, these factors include Rabaptin-5, p115 and RLIP, a recently characterized Rab7 effector (23–25). Rabaptin-5 is recruited on endosomal membranes by Rab5:GTP, and participates in the assembly of a multiprotein complex which includes EEA1 and phosphatidylinositol 3-kinase, and which promotes fusion of early endosomes (26,27). Rab1 was shown to mediate the recruitment of p115 into a cis-SNARE complex during the budding of COPII vesicles from the endoplasmic reticulum (24). Another example of a large cytosolic multiprotein complex interacting with a Rab GTPase (Sec4p) in a GTP-dependent manner is the exocyst, involved in the tethering of post-Golgi vesicles with the plasma membrane in S. cerevisiae (28). It is worth noting that Rab6IP2s, like for instance Rabaptin-5 and p115, exhibit a high content of α-helical secondary structure that may provide a backbone for the assembly of a multiprotein complex; this is also suggested by the gel filtration profile of endogenous Rab6IP2s in brain cytosol. However, no interaction in the yeast two-hybrid assay was found between Rab6IP2s and other Rab6-interacting proteins previously identified, including Rabkinesin-6, GAPCenA and Rab6IP1/ORF37 (data not shown), neither could any component of the Rab6IP2 complexes be identified so far by immunoprecipitation with the anti-Rab6BD antibodies (our unpublished results).

Rab6IP2 appears to interact better in vitro with Rab6A′ than with Rab6A. Moreover, overexpression of the Rab6IP2 Rab6-binding domain inhibits the Shiga toxin B subunit transport to the TGN, and not the intracellular transport of the secretory protein SEAP. Recent work suggests that Rab6A′ is involved in the regulation of endosome-to-TGN transport in HeLa cells (13). Altogether, this suggests that Rab6IP2 is involved in the endosome-to-TGN Rab6A′ regulated pathway. Rab6A′ appears to fulfill the same function as Ypt6p (the only copy of Rab6 expressed in S. cerevisiae). Ypt6p is thought to be required for fusion of endosomal/prevacuolar vesicles with the late Golgi apparatus (29), and would act by recruiting the Vps52/53/54 tethering complex (VFT, Vps Fifty Three complex) onto TGN membranes (30). This complex can directly bind the Golgi t-SNARE Tlg1p, which is also present on endosome-derived vesicles, and might promote SNARE pairing and fusion (31). Syntaxin 6 (the human homolog of Tlg1p) seems to be a component of the machinery involved in the Shiga toxin transport step between endosomes and TGN (13). However, the mechanism of action of Rab6A′ and its interaction partners, particularly Rab6IP2, in the retrograde pathway between endosomes and TGN awaits further studies. No obvious homology was found between Rab6IP2 and any component of the VFT complex, but it cannot be excluded that Rab6IP2-containing complexes themselves might trigger the same function as VFT. Another multiprotein complex, the retromer, was suggested to play a role of a vesicular coat in prevacuolar to Golgi transport in yeast (32). The study of interactions between the human counterparts of yeast proteins involved in the endosome-to-late Golgi pathway, together with the use of an in vitro assay, should help to elucidate the molecular machinery of this vesicular transport step.

Materials and Methods

Cloning of Rab6IP2

A two-hybrid screen was performed as previously described (17), using pLexA-Rab6 Q72L as a bait and a mouse embryo cDNA library kindly provided by A. Vojtek (33). Among other clones, a 475-bp insert (‘clone 1’) was characterized and used to screen a mouse brain cDNA library in Lambda ZAPII vector (Stratagene). Four clones were found which contained portions of the clone 1 sequence. Overlapping regions allowed reconstitution of two cDNA species: a long segment of 2154 bp, and a smaller one bearing a 132-bp deletion, probably the product of alternative splicing. It must be noted that this deletion includes the 5′ portion of clone 1, and results in a 17-residue truncation out of the 154 amino acids of clone 1. In addition, a 12-bp divergence was observed between these two clones and the 3′ end of the original clone 1. Further screening was then needed, first to define an open reading frame, and second to find clones containing the 3′-most 12 bp present in the two-hybrid clone 1. By screening a mouse skeletal muscle ′Stretch Plus cDNA library (Clontech, Palo Alto, CA, USA) using as a probe the 5′-most 220 bp of the available clones, two overlapping clones were found, containing an identical open reading frame starting at position 553. Moreover, a new screen was performed by PCR using a mouse spleen cDNA library (Stratagene, La Jolla, CA, USA) primed with a DNA oligonucleotide corresponding to the region located upstream from the clone 1 5′ end sequence, and, at the 3′ end, with an antisense oligonucleotide hybridizing to the cloning vector of the library. A segment was found, which contained the entire sequence of clone 1, including the last three residues, and an additional extension of 105 amino acids. Rab6IP2A refers to the shortest version of the reconstituted clones, has the deletion encompassing the 5′ end of the clone 1 sequence, and the premature stop codon. Rab6IP2B is the long version. The accession numbers of Rab6IP2A and Rab6IP2B are AF340028 and AF340029, respectively.

Plasmids and transfection experiments

Wild-type Rab6A, Rab6A Q72L, Rab6Aí Q72L, and Rab6A T27N cDNAs were all subcloned in the pcDNA3 plasmid (Invitrogen, Carlsbad, CA, USA). Rab6IP2A and B were subcloned into the pGEM1 between EcoR1 and SalI restriction sites, and into the mammalian expression vectors pcDNA3 and pEGFP-C2 (Clontech) between EcoR1 and EcoRV, and EcoR1 and SalI, respectively. Both, as well as the Rab6-binding domain (Rab6BD) represented by the original clone 1 isolated in the yeast two-hybrid screen (see above), were also inserted into the pGEMmyc4 vector plasmid (a generous gift of Dr M. Zerial). Transfection of HeLa cells was performed in two ways. For transport experiments, we overexpressed indifferently the nontagged or the Myc-tagged Rab6IP2A and Rab6IP2B, and the Myc-tagged Rab6BD, using the vaccinia virus system as previously described (22). For the other experiments, cDNAs were introduced into cells by electroporation. Briefly, cells were trypsinized and the cell suspension was pulsed by a discharge of 230 V and at 950 μFaraday in the presence of 10 μg total DNA, 30 μg salmon sperm carrier DNA and 210 mm NaCl. For Rab6IP2B-GFP expression, a lower amount (2 μg) of cDNA was used, the other 8 μg representing either the empty pEGFP-C2 vector cDNA, or other cDNAs to be tested. Cells were plated and processed for immunofluorescence 20–24 h after transfection. It must be noted that the electroporation of Rab6A Q72L cDNA in the pcDNA3 plasmid led to a moderate overexpression of the protein, which did not result in the dispersion of the Golgi apparatus as observed when using the vaccinia virus transfection system (22).

Preparation of the recombinant fusion proteins GST-Rab6A, GST-Rab6A′ and GST-Rab6B

The construction of the plasmid cDNAs encoding the fusion proteins GST-Rab6A, A′ and B have been described elsewhere (12,14) The fusion proteins were purified from bacteria induced with 1 mm IPTG for 3–5 h. Bacteria were harvested and frozen at − 80 °C until use. One liter bacteria culture pellet was resuspended in 20 ml of phosphate buffer saline (PBS) supplemented with a mixture of protease inhibitors, 50 μm GTP, 0.1% Tween 20, 1 mm DTT, incubated with 1 mg/ml lysozyme and centrifuged at 20 000 ×g for 20 min The supernatant was saved and incubated in the presence of 750 μl of dry glutathione-Sepharose 4B beads (Pharmacia). After 1 h incubation at 4 °C under rotation, the beads were washed twice, and the fusion protein was eluted in a column using 500-μl fractions of elution buffer containing 50 mm Hepes pH 7.5, 150 mm NaCl, 1 mm MgCl2, 10 μm GTP and 15 mm reduced glutathione. The recombinant proteins were stored at − 80 °C until use.

Pull-down experiments

About 10 μg GST or GST-Rab6A, -Rab6A′ and -Rab6B fusion proteins were coupled back to 30 μl glutathione beads (corresponding to 15 μl of dry beads) by incubation in loading buffer (25 mm Tris pH 7.5, 10 mm EDTA, 5 mm MgCl2) for 1 h at 4 °C under rotation. Beads were washed in the same buffer. The GST fusion proteins coupled to glutathione-Sepharose 4B beads were then loaded with GDP or GTPγS by incubation for 1 h at 37 °C in 200 μl loading buffer supplemented with 200 μm GDP or GTPγS. The buffer was removed after a 5-min centrifugation at 500 ×g and replaced by 200 μl of interacting buffer (25 mm Tris pH 7.5, 10 mm MgCl2, 50 mm NaCl, 0.1% Triton X100) containing 200 μm GDP or GTPγS to which 50 μl (about 500 μg of protein) of PNS was added. Incubation was performed at room temperature for 90 min under rotation. Beads were washed 4 times in interacting buffer without nucleotides. Samples were directly (without elution) submitted to SDS-PAGE and the gel processed for Western blotting.

Antibody, immunodetection and immunofluorescence

A polyclonal antibody was raised against the fusion protein GST-Rab6BD. Before use, the antiserum was depleted of the anti-GST antibodies by filtration through a GST-coupled NHS column (Pharmacia). For immunodetection of Rab6IP2 in tissue and cell extracts, as well as in immunofluorescence experiments, anti-Rab6BD antibody was affinity purified using the Rab6BD protein generated by thrombin digestion of the GST-Rab6BD fusion protein. Immunofluorescence was performed as previously described (34).

Tissue and cell extracts

Mouse tissue extracts were prepared from frozen samples (kindly provided by Dr Sylvie Robine, Institut Curie). Tissues were homogenized using a Potter homogenizer in 50 mm Hepes pH 7.2, 90 mm KCl. Homogenates were diluted twice using 40% sucrose (w/v in water) and centrifuged at 1800 ×g. The protein content of these postnuclear supernatants was measured. The same amount of each was loaded on the gel, and then processed for immunoblotting.

Cell extracts were made by solubilizing equivalent amounts of PBS-washed cells in lysis buffer (20 mm Tris pH 8, 150 mm NaCl, 5 mm EDTA, 1% Triton X100, 0.2% BSA), except HeLa S3 and another batch of HeLa cells (see Figure 5A), which were processed like tissues (a Dounce homogenizer was used instead of a Potter homogenizer). A portion of these cell PNS was centrifuged a second time at 1800 ×g for 5 min and separated into the soluble, cytosolic fraction and the membrane fraction by ultracentrifugation at 200 000 ×g for 30 min Equivalent amounts of each fraction were analyzed by Western blot for their content in Rab6IP2. Bovine brain cytosol (BBC) was prepared by homogenizing bovine brain devoid of meninges and cerebellum in 115 mm KAc, 25 mm Hepes pH 7.4, 2.5 mm MgCl2, 1 mm DTT, and protease inhibitors, using 1 ml of buffer per gram of wet tissue, firstly in a blender, and secondly in a Polytron homogenizer. Homogenate was centrifuged at 4000 r.p.m. for 30 min twice, supernatant was then ultracentrifuged at 100 000 ×g for 1 h twice. Gel filtration was performed on a 50-μl aliquot of BBC (about 1 mg of total protein) loaded on a 2.5-ml Superdex 200 (Pharmacia) in PBS. A 10-μl aliquot out of 50 μl fraction was analyzed by SDS-PAGE and used for immunodetection of various proteins.

Retrograde transport assay

A Shiga toxin B subunit bearing two sulfation sites [(35), B-(Sulf)2] was used to study retrograde transport between plasma membrane and Golgi in HeLa cells overexpressing various proteins using the vaccinia virus system (22). After 5 h 15 min transfection, cells were depleted of endogenous sulfate by incubation in PBS containing 5.6 mm glucose for 45 min. B-(Sulf)2 binding was realized at 4 °C for 30 min in the same buffer; internalization was performed in the presence of 0.5 mCi/ml radioactive 35S sulfate for 30 min at 37 °C. Cell lysates were processed as described in (13). Results are expressed as a percentage of B-(Sulf)2 sulfation measured in the control condition (cells transfected with the empty pGEM vector).

Other procedures

A multiple tissue Northern blot (Clontech) was hybridized with the clone 1 insert as a probe labeled with 32P using the oligolabeling kit (Pharmacia, Peapack, NJ, USA). Alignment of protein sequences was performed using the Blast program of NCBI (36). The research of coiled-coil domains was done using the paircoil program described by Berger et al. (37).


We thank Lucien Cabanié for performing gel filtration experiments, and for the anti-Rab6BD antibody purification, Frédéric Mallard for his help in the Shiga toxin experiments, Franck Perez for his help in computer work. We are grateful to Ray McDermott for critical reading of the manuscript. This work was supported by grants from the Centre National de la Recherche Scientifique, the Institut Curie and the Association de la Recherche contre le Cancer (ARC no. 5294 and no. 9028).