RIN3: a novel Rab5 GEF interacting with amphiphysin II involved in the early endocytic pathway

The small GTPase Rab5, which cycles between active (GTP-bound) and inactive (GDP-bound) states, plays essential roles in membrane budding and trafficking in the early endocytic pathway. However, the molecular mechanisms underlying the Rab5-regulated processes are not fully understood other than the targeting event to early endosomes. Here, we report a novel Rab5-binding protein, RIN3, that contains many functional domains shared with other RIN members and additional Pro-rich domains. RIN3 displays the same biochemical properties as RIN2, the stimulator and stabilizer of GTP-Rab5. In addition, RIN3 exhibits its unique intracellular localization. RIN3 expressed in HeLa cells localized to cytoplasmic vesicles and the RIN3-positive vesicles contained Rab5 but not the early endosomal marker EEA1. Transferrin appeared to be transported partly through the RIN3-positive vesicles to early endosomes. RIN3 was also capable of interacting via its Pro-rich domain with amphiphysin II, which contains SH3 domain and participates in receptor-mediated endocytosis. Interestingly, cytoplasmic amphiphysin II was translocated into the RIN3- and Rab5-positive vesicles when co-expressed with RIN3. These results indicate that RIN3 biochemically characterized as the stimulator and stabilizer for GTP-Rab5 plays an important role in the transport pathway from plasma membrane to early endosomes.

biochemical properties of RIN3 are similar to those of other RIN members and further found that it displays unique intracellular localization. RIN3 localizes to Rab5-positive, but not EEA1-positive, vesicles in HeLa cells, and transferrin appears to be transported partly through the RIN3-positive vesicles to early endosomes. Furthermore, we found that amphiphysin II, which mediates receptor-induced endocytosis, interacts with RIN3 and that cytoplasmic amphiphysin II is translocated into the RIN3-and Rab5-positive vesicles. Collectively, these results suggest that RIN3 is a novel player in the transport pathway from plasma membranes to early endosomes.

Materials
Alexa-488 secondary antibody and human Alexa-488 transferrin were purchased from Molecular Probes (Eugene, OR). pFastBacHTa vector was from Invitrogen. pEGFP-C1 and pDsRed-1 vectors and human leukocyte MATCHMAKER cDNA library were obtained from BD Biosciences. Antibody sources were as follows: monoclonal antibody against the Flag epitope (M2) (Sigma); monoclonal antibodies against DsRed and Rab5 (BD biosciences); polyclonal antibody against glutathione-S-transferase (GST) (Santa Cruz Biotechnology).
cDNA constructs A DNA fragment encoding the Flag epitope (MDYKDDDDK) was substituted for 6× histidine affinity tag of pFastBacHTa and constructed into the pFastbac-Flag vector. To express Flag-RIN3 and Flag-RIN2, fragments coding RIN3 and RIN2 were inserted between the EcoRI site and SalI site of the pFastBac-Flag vector. pCMV5-Flag vector was obtained by inserting a DNA fragment encoding a start methionine followed by the Flag epitope between the EcoRI and BamHI sites of pCMV5. Human amphiphysin II was obtained using the human leukocyte MATCHMAKER cDNA library as a template. Deletion mutants of RIN3 and amphiphysin II were constructed using a PCR-based strategy.
Yeast two-hybrid screening A yeast two-hybrid assay was performed according to the method described previously (Hoshino et al., 1999;Kontani et al., 2002). The yeast reporter-strain Hf7c was transformed with pGBT9-Rab5b/Q79L using a lithium-acetate-based method and grown in synthetic medium lacking tryptophan at 30°C for 5 days. The cells were transformed with the human leukocyte MATCHMAKER cDNA library and plated on synthetic medium lacking leucine, tryptophan and histidine at 30°C for 7 days. For interaction analysis, the transformed yeasts were lifted onto filter papers and lysed by brief liquid-nitrogen treatment and incubated with 5-bromo-4-chloro-3-indolyl β-D-galactopyranoside. Library plasmids from positive clones were rescued into Escherichia coli HB101 cells plated on leucine-free medium. True Rab5b/Q79Linteracting clones were sequenced by the dideoxynucleotide chain termination method on both strands. When RIN3 was screened as bait, the same method described above was adopted.
Cloning of the full-length RIN3 cDNA Upstream and downstream sequences of HK281 clone were obtained by 5′ and 3′ rapid amplification of cDNA end (RACE) PCR using the same library as described above, according to the manufacturer's protocols. The RACE PCR fragment was purified and subcloned into pGEM-T easy (Promega). The complete nucleotide sequence was confirmed by isolating and sequencing multiple clones, and designated as RIN3.

Northern blot analysis
Human 12-lane multiple tissue northern blot containing poly(A) mRNA was obtained from Clontech. An [α-32 P] dCTP-labelled fragment encoding amino acids 434-678 of RIN3 was hybridized to the membrane overnight at 65°C in ExpressHyb Solution (BD Biosciences). The membrane was washed twice with 2× SSC (sodium chloride/sodium citrate) containing 0.1% SDS for 10 minutes each and twice with 1× SSC containing 0.1% SDS for 20 minutes each. The filter was autoradiographed for 12 hours at -80°C and analysed by BAS1800 (FujiFilm).

Assay for the in vitro association between Rab5 and RIN3
The association between Rab5 and Flag-tagged RIN3 was assayed as described previously (Saito et al., 2002). The GDP-or GTPγS-bound Rab5 was incubated with agarose resin (10 µl) conjugated with Flag-RIN3 at 30°C for 60 minutes. The resin was washed, and proteins Rab5-GEF RIN3 interacts with amphiphysin II were eluted from the resin with 30 µl of the washing buffer containing 100 ng ml -1 of Flag peptide. After centrifugation, the supernatant (24 µl) was mixed with 8 µl of 4× SDS sample buffer, boiled for 5 minutes and subjected to SDS-PAGE. Immunoblotting was performed with anti-Flag monoclonal and anti-Rab5 polyclonal antibodies.

Confocal microscopy
HeLa cells transiently expressing red fluorescent protein (RFP)-RIN3 were cultured on a polylysine-coated glass coverslip (15-mm diameter) and washed three times with PBS before fixation with 4% paraformaldehyde in PBS for 15 minutes at 4°C. After treatment with 0.1 mM glycine in PBS for 15 minutes, the cells were permeabilized with 0.1% Triton X-100 in blocking solution [3% bovine serum albumin (BSA) in PBS] before incubation with a primary antibody (1 µg ml -1 diluted with blocking solution) for 1 hour at room temperature. The cells were washed three times with PBS and incubated for 1 hour with Alexa-488-conjugated secondary antibodies diluted with the blocking solution. After washed three times with PBS, the coverslip was mounted onto a glass slide in Permafluor-mounting medium (Immunon) and viewed on a Carl Zeiss confocal microscope with LSM510 software using excitation wavelengths of 488 nm or 546 nm. The images were merged using Photoshop (Adobe Systems, Mountain View, CA). In co-expressing experiments of green fluorescent protein (GFP), yellow fluorescent protein (YFP) and RFP fusion proteins, the transfected HeLa cells were cultured for 48 hours in a glass-based dish (35-mm diameter, Iwaki) and examined by confocal microscopy.

Analysis of transferrin uptake and internalization
For the steady-state internalization of Alexa-488 transferrin, HeLa cells transiently expressing RFP-RIN3 (or RFP-mock) were cultured in the 35-mm glass dish. The cells were incubated at 37°C for 60 minutes in internalization medium (IM) consisting of DMEM plus 20 mM Hepes-NaOH (pH 7.4) and 2 mg ml -1 BSA to deplete endogenous transferrin, and further incubated at 4°C for 45 minutes in IM containing 10 µg ml -1 Alexa-488/transferrin. The cells were washed and incubated in IM at 25°C for the indicated times to allow internalization.
Interaction between RIN family and amphiphysin II in transfected HeLa cells HeLa cells that had been co-transfected with pCMV5 containing RIN3 and Flag-tagged amphiphysin II (the full-length or with the SH3 domain deleted (∆SH3)) were solubilized with 2 ml buffer A. The supernatants were pre-cleared and immunoprecipitated with agarose resin (10 µl) conjugated with 0.5 µg of the anti-Flag monoclonal antibody. After incubation at 4°C for 90 minutes, the resin was washed three times with 100 µl TBS containing 0.1% (w/v) NP-40 and three more times with 100 µl buffer C, consisting of 75 mM Tris-HCl (pH 7.5), 100 mM NaCl, 1 mM EDTA and 0.1% (w/v) NP-40. Proteins were eluted from the resin with the buffer C containing 50 ng ml -1 Flag peptide. After centrifugation, the supernatant (24 µl) was mixed with 8 µl 4× SDS sample buffer, boiled for 5 minutes and subjected to SDS-PAGE. Immunoblotting was performed with anti-RIN3 polyclonal and anti-Flag monoclonal antibodies.

Western blot analysis
Proteins were transferred to a polyvinylidene difluoride membrane (BioRad), blocked in 5% BSA in TBS for 1 hour and incubated with primary antibodies for 1 hour at room temperature. After washing three times with 0.2% Triton X-100/TBS, the membrane was further incubated with a horseradish-peroxidase-conjugated secondary antibody for 1 hour. The immuno-positive signal was visualized in the presence of luminol (Pierce).

DDBL/EMBL/GenBank accession number
The accession number for the full-length RIN3 is AB081753.

Results
Complementary DNA cloning of the Rab5-binding protein RIN3 To identify proteins that interact with GTP-bound form of Rab5, a human leukocyte cDNA library was screened with the GTPase-deficient mutant Rab5b/Q79L in the yeast two-hybrid system. Screening of 5×10 7 transformants yielded five positive clones that interacted strongly with Rab5b/Q79L. Three of these were the same as a clone encoding Rab5c, indicating the biochemical properties of Rab5 dimerization (Daitoku et al., 2001), and one contained a clone coding for the human homologue of Rabaptin-5β (Gournier et al., 1998). The remaining clone, termed HK281, consisted of a 1575-bp cDNA encoding 525 amino acids. An additional 0.7-kb fragment, which contained an in-frame stop codon and the poly(A) tail, was produced by the 3′-RACE method. The upstream 5′-end was obtained by 5′-RACE, which gave an additional 1.5-kb fragment with an initiation codon based on the Kozak consensus sequence. Analysis of the total cDNA revealed an open reading frame that encoded a 985-amino acid protein in which the original HK281 corresponded to amino acids 435-959.
The isolated cDNA has a sequence similar to two independent clones, Ras-interaction/interference 1 (RIN1) and RIN2 (see supplementary Fig. S1, http:// jcs.biologists.org/ supplemental/). Therefore, this protein was designated RIN3. Although the present RIN3 and previous RIN2 (Saito et al., 2002) were isolated as Rab5b-binding proteins, they also interact with other members of Rab5 (Rab5a and Rab5c) but not with Rab4, Rab7 or Rab11 in the yeast two-hybrid system (data not shown). Thus, the RIN family could be characterized as a binding partner at least specific to Rab5 group of the Rab small GTPases. The RIN family shared a Src homology 2 (SH2) domain, a RIN-homology (RH) domain (Saito et al., 2002), a Vps9 domain conserved in the catalytic domains of the GEFs Vps9p and Rabex-5, and a Ras-association (RA) domain from their N termini (Fig. 1A). RIN1 has a prolinerich domain (PRD) that binds to the c-Abl SH3 domain (Afar et al., 1997;, and RIN2 and RIN3 further contain one and two PRDs (the class II motif of PXXPPR; Yu et al., 1994), respectively, which are lacking in RIN1.

Expression of RIN3 mRNA in human tissues
To investigate the expression pattern of RIN3 in human tissues, northern blot analysis was performed. Poly(A)-selected RNAs from several human tissues were hybridized with a radiolabelled probe containing the 729-base coding sequence of RIN3. A 4-kb transcript was detected in a variety of tissues, with the highest expression level in the peripheral blood cells (Fig. 1B). This expression pattern is slightly different from those of other RIN members, because RIN1 and RIN2 mRNAs have been reported to be abundant in brain  and in heart, kidney and lung (Saito et al., 2002), respectively, although they are widely expressed. Furthermore, the 4-kb transcript was also detected in human cell lines including THP-1 and Jurkat cells (data not shown).
Unique biochemical properties of RIN3 RIN3 contains Vps9 domain (amino acids 724-864), which was conserved in the RIN family and other characterized GEFs for Rab5 (Hama et al., 1999;Horiuchi et al., 1997), so we first investigated whether this molecule also functions as a Rab5-GEF. For the analysis, Flag-tagged RIN3 and RIN2 proteins were purified from baculovirus-infected Sf9 cells ( Fig. 2A) and assayed for their ability to stimulate [ 3 H]GDP release from prenylated Rab5. The rate of GDP-GTP exchange on Rab5 was extremely slow at physiological concentrations (mM) of Mg 2+ (Fig. 2B). However, GDP release from Rab5 was markedly accelerated by the simultaneous addition of RIN3 or RIN2. We also examined the effects of RIN3 and RIN2 on [ 35 S]GTPγS binding to Rab5. As expected, RIN3 and RIN2 markedly enhanced the GTPγS binding to Rab5 (Fig. 2C), although RIN3 and RIN2 themselves had no detectable GTPγS-binding activity. These results clearly indicate that both RIN3 and RIN2 act as GEFs for Rab5.
We next investigated whether the nucleotide-bound state of Rab5 exerts its influence on the interaction with RIN3 by an in vitro binding assay. Flag-tagged RIN3 immobilized to resins was incubated with GDP-or GTPγS-bound form of Rab5, and the Rab5 binding was estimated by immunoblotting with an anti-Rab5 antibody. As had been observed in RIN2 (Saito et al., 2002), RIN3 associated more tightly with GTPγS-bound Rab5 than its GDP-bound form (Fig. 2D). Thus, RIN3, like RIN2, appears to act biochemically not only as a GEF for Rab5 but also as a stabilizer for GTP-Rab5.

Intracellular localization of RIN3 in transfected HeLa cells
We next examined the intracellular localization of RIN3 in mammalian cells. HeLa cells transiently expressing RFP-RIN3 were immunostained with anti-Rab5 and anti-EEA1 antibodies, and their localization analysed by confocal microscopy. RIN3 localized to punctate vesicles scattered around the cytoplasm (Fig. 3A). We found that the RIN3positive vesicles also contained endogenous Rab5 in some degree (Fig. 3A, top), although most Rab5 localized to perinuclear early endosomes, as previously described (Chavrier et al., 1990). The co-localization of RIN3 and Rab5 was more evident when both proteins were expressed in HeLa cells (Fig. 3B). However, the RIN3 fluorescence did not overlap with the early endosomal marker EEA1 at all (Fig. 3A,  bottom), indicating that the RIN3-positive vesicles are different from early endosomes. When the N-terminal region (amino acids 1-586) of RIN3, which lacks the Rab5-binding site (Saito et al., 2002), was expressed in HeLa cells, it distributed over the cytoplasm (Fig. 3C, middle). However, the C-terminal region (amino acids 587-985) of RIN3 exhibited its vesicular localization (Fig. 3C, right) similar to the full-length form. These results suggest that the translocation of RIN3 into the vesicles requires its association with Rab5. Essentially the same results were obtained in A431 cells, a human epidermal Journal of Cell Science 116 (20) It has been reported that the overexpression of RIN1 in a stable CHO cell line expressing Rab5a causes the enlargement of Rab5a-positive endosomes and that a proportion of RIN1 localizes with the endosomes (Tall et al., 2001). Therefore, intracellular localization of other RIN members was also investigated under the same conditions. When RFP-RIN2 was transiently expressed in HeLa cells, it localized to punctate vesicles, as observed with RIN3 (data not shown). However, RIN1 was distributed over the cytoplasm of HeLa cells (Fig.  3D). We confirmed that RFP-RIN1 and RIN3 fusion proteins were certainly produced in HeLa cells, using western blot analysis with an anti-RFP antibody (Fig. 3E).
To identify the entity of the RIN3-positive vesicles, we monitored the vesicular trafficking of transferrin in HeLa cells transiently expressing RIN3. HeLa cells that had been transfected with RFP-RIN3 were pulse-chased with Alexatransferrin and further incubated at 25°C to allow its internalization. The fluorescently labelled transferrin was first observed in the plasma membrane (Fig. 4A), and it partly moved to the RIN3-positive vesicles 4-10 minutes after its internalization (Fig. 4B,C). However, the fluorescence of transferrin was no longer observed in the RIN3-positive vesicles at 15 minutes and it localized mostly to perinuclear EEA1-positive early endosomes (Fig. 4D). There was no significant difference between the time-dependent trafficking of Alexa-transferrin in the mock-transfected control and RIN3expressing HeLa cells, suggesting that the RIN3-positive vesicles are not artefacts of transfection. These results strongly suggest that RIN3 is specifically involved in a transport pathway from plasma membranes to early endosomes.

Identification of amphiphysin II as a RIN3-binding protein
To uncover the function of RIN3, we further searched for RIN3binding proteins by using the yeast two-hybrid system. A human leukocyte cDNA library was screened with full-length RIN3 as bait. Screening of 5×10 6 transformants yielded seven positive clones that strongly interacted with RIN3. Two and four of them contained clones coding for Rab5b and Rab5c, respectively. The remaining one was composed of a cDNA encoding the partial sequence of amphiphysin II. Amphiphysin I and II are nerveterminal-enriched proteins containing SH3 domains that interact with dynamin and synaptojanin. The amphiphysins not only function in synaptic vesicle endocytosis by targeting dynamin and synaptojanin to endocytic buds through their interactions with clathrin and AP2, but also contribute to T-tubule organization and function Lee et al., 2002;Leprince et al., 1997;Ramjaun et al., 1997;Wigge et al., 1997). Among the multiple amphiphysin II-splicing variants, amphiphysin II/BIN1 (whose exons 10, 12 and 13 are alternatively spliced) has been reported to be ubiquitously expressed (Wechsler-Reya et al., 1997). This amphiphysin II variant is a 409-amino-acid protein and the isolated clone as a RIN3-binding protein corresponded to amino acids 28-409.

N-terminal region of RIN3 interacts specifically with the SH3 domain of amphiphysin II
To identify the interacting regions between amphiphysin II and RIN3, the full-length and SH3 domain of amphiphysin II were purified as GST-fused proteins and subjected to GST pull-down assays. The full-length and SH3 domain of amphiphysin II were capable of binding to RIN3 (Fig. 5A). The importance of the SH3 domain was also investigated in HeLa cells expressing the full-length and SH3-deleted form (∆SH3) of Flag-amphiphysin II. When the cell lysate was immunoprecipitated with the anti-Flag antibody, the full-length form but not amphiphysin II/∆SH3 co-precipitated RIN3 (Fig.  5B). Various forms of RIN3 were also expressed in HeLa cells as Flag-tagged proteins and were subjected to a GST pulldown assay with GST-fused amphiphysin II. Amphiphysin II appeared to interact with the full length of RIN3 and its N-terminal region but not with its Cterminal region (Fig. 5C). These results indicate that the SH3 domain of amphiphysin II and the N-terminal region of RIN3 that contains PRDs are sufficient for their direct interaction.
We further investigated whether the interaction of amphiphysin II is specific for RIN3, because other RIN members (RIN1 and RIN2) have also contained PRDs between the SH2 and RH domains (Fig. 1A). Each member of the RIN family was purified from Sf9 cells and incubated with GST-fused amphiphysin II that had been immobilized on glutathione Sepharose beads. Amphiphysin II associated with RIN3 (and RIN2, data not shown) but not with RIN1 (Fig. 5D). RIN3 and RIN2 were also capable of binding to the SH3 domain derived from amphiphysin II in vitro (data not shown). Thus, the binding of amphiphysin II appeared to be rather specific for a certain type (probably the class II) of PRDs present in RIN3 and RIN2 but not in RIN1.

Translocation of cytoplasmic amphiphysin II into RIN3positive vesicles in transfected HeLa cells
To investigate how the interaction between amphiphysin II and RIN3 exerts its influence on the distribution of their proteins in intact cells, HeLa cells were transfected with GFPamphiphysin II and/or RFP-RIN3 and subjected to confocal microscopic analysis (Fig. 6A). As previously observed (Lee et al., 2002), amphiphysin II exhibited a diffused pattern of cytoplasmic distribution, regardless of whether RFP-mock was co-expressed or not. By contrast, amphiphysin II was entirely targeted to and concentrated in the RIN3-positive vesicles and not present in the cytoplasm upon co-expression with RIN3. The same translocation of amphiphysin II into RIN-positive vesicles was observed with the expression of RIN2 but not with RIN1 (data not shown). We also investigated whether this targeting depends on the SH3 domain of amphiphysin II, because its deletion form failed to interact with RIN3 (Fig. 5B).
Journal of Cell Science 116 (20) Fig. 3. RIN3 transfected into HeLa cells co-localizes with Rab5 but not EEA1. (A) HeLa cells were transiently transfected with RFP-RIN3 and cultured for 48 hours. The cells were fixed and labelled with antibodies to Rab5 (top) or EEA1 (bottom). The fluorescence of RFP-RIN3 (left) and Alexa-488 secondary antibody (middle) was visualized by confocal microscopy, and merged images of the two signals are displayed in yellow (right). Arrowheads indicate the co-localization of RIN3 and Rab5. (B) HeLa cells transiently expressing RFP-RIN3 and GFP-Rab5b were subjected to confocal microscopy, and the fluorescence of RFP-RIN3 (left) and GFP-Rab5b (middle) was visualized by confocal microscopy as described in (A). (C) HeLa cells transiently expressing the full-length (FL, left), N-terminal (N, middle) or C-terminal (C, right) form of RFP-RIN3 were subjected to confocal microscopy. (D) HeLa cells transiently expressing RFP-RIN1 (left) or RFP-RIN3 (right) were subjected to confocal microscopy. (E) Lysates from the transfected cells (RFP-RIN1, left; RFP-RIN3, right) were separated by SDS-PAGE and immunoblotted (IB) with an anti-RFP antibody. Scale bars, 10 µm. Rab5-GEF RIN3 interacts with amphiphysin II Amphiphysin II/∆SH3 distributed in the cytoplasm (Fig. 6B). This distribution was not altered by the co-expression with RIN3, although the formation of RIN3-positive vesicles was certainly observed. These data indicate that the targeting of amphiphysin II to the RIN3-positive vesicles depends on its SH3 domain. We finally investigated whether the ternary complex RIN3/Rab5/amphiphysin II might be observed in HeLa cells by means of the co-expression of RFP-RIN3, GFPamphiphysin II and YFP-Rab5. Both amphiphysin II and Rab5 thoroughly localized in the RIN3-positive vesicles (Fig. 6C), and this localization was not markedly altered by the nucleotide-bound forms of Rab5 (data not shown). RIN3, Rab5 and amphiphysin II form a complex, but Rab5 does not directly interact with amphiphysin II in vitro (data not shown). Collectively, these data suggest that the ternary complex RIN3/Rab5/amphiphysin II is involved in the early endocytic transport pathway.

Discussion
In the present study, we have identified a novel Rab5-binding protein, RIN3, which contains SH2, proline-rich, RH, Vps9 and RA domains from the N terminus. This domain structure is conserved in other RIN members, RIN1 (Han et al., 1997) and RIN2 (Saito et al., 2002). RIN3 stimulates guanine nucleotide exchange reaction on Rab5 (Fig. 2B,C) but preferentially interacts with the GTP-bound form of Rab5 (Fig. 2D). The same properties are observed in RIN1 and RIN2, except for the preferential binding of RIN1 to GDPbound Rab5 (Saito et al., 2002;Tall et al., 2001). These biochemical properties are equivalent to the combined actions of Rabex-5 and Rabaptin-5, which form a complex in early endosomes (Horiuchi et al., 1997;Lippe et al., 2001). Rabex-5 is a Rab5 GEF involved in the homotypic fusion of early endosomes, and Rabaptin-5 specifically associates with the GTP-bound form of Rab5 to stabilize its nucleotide form (Stenmark et al., 1995). Thus, RIN3, together with RIN2, appears to function as both the stimulator and the stabilizer for Rab5 in an endocytic transport pathway just as the Rabex-5/Rabaptin-5 complex does in the homotypic fusion of early endosomes.
Although RIN1, RIN2 and RIN3 appear to belong to the same family in terms of their domain structure and biochemical properties for Rab5, several differences are highlighted from the present study. First, the distributions of mRNAs are different among the RIN family in spite of their wide expression: RIN1, RIN2 and RIN3 mRNAs are abundant in brain , in heart, kidney and lung (Saito et al., 2002), and in peripheral blood cells (Fig.  1B), respectively. Second, the intracellular localizations of the three RFP-RIN members transiently expressed in HeLa cells are not identical to one another (Fig. 3): RIN3 and RIN2 localize to endocytic vesicles, whereas RIN1 exhibits a cytoplasmic distribution. RIN3 (and RIN2) appears to localize with Rab5 in the vesicles, especially when both proteins were expressed in the cells (Fig. 3A,B). However, it has been reported that RIN1 could partially co-localize with Rab5positive vesicles upon their co-expression, and stimulates epidermal-growth-factor-receptor-mediated endocytosis (Tall et al., 2001). The different localizations of the RIN members might be derived from the types of cells used, but it is likely that RIN1 also localizes to endocytic vesicles if the cells are stimulated by membrane receptors. Third, there is selectivity in the association of RINs with amphiphysin II: RIN3 and RIN2, but not RIN1, can interact with amphiphysin II (Fig.  5D). The N terminus of RIN3 that contains PRDs directly associates with the SH3 domain of amphiphysin II (Fig. 5A-C). The class-II PRD present in both RIN3 and RIN2 but not in RIN1 appears to be responsible for the association, because a mutation of proline in the class-II PRD failed to interact with amphiphysin II (data not shown).
Another unique feature of the RIN family is the existence of an RA domain in their C termini, which was initially identified as a region interacting with H-Ras (Hofer et al., 1994;Ponting and Benjamin, 1996). This suggests that RINs are capable of Fig. 4. Endocytic transferrin is transported through RIN3-positive vesicles to early endosomes in HeLa cells. HeLa cells expressing RFP-RIN3 were pulse-chased with Alexa-488/transferrin at 4°C, shifted to 25°C to allow its internalization and further incubated for the indicated times. The fluorescence of RFP-RIN3 (left) and transferrin (middle) was visualized by confocal microscopy, and merged images of the two signals are displayed in yellow (right). Arrowheads indicate the co-localization of RIN3 and transferrin. Scale bars, 10 µm.
binding not only to Rab5 but also to the Ras-family GTPases. In this regard, there is an interesting report showing that the GEF activity of RIN1 for Rab5 is enhanced by its interaction with Ha-Ras (Han and Colicelli, 1995;Tall et al., 2001). However, Ha-Ras failed to interact with RIN3 or RIN2 to stimulate their GEF activities in conditions under which it certainly binds to RIN1. Instead, several Ras members other than Ha-Ras could interact with RIN2 (K.S., unpublished). These results suggest that each RIN interacts with certain types of Ras members. In addition, the RIN-family members also contain an SH2 domain in their N-terminal regions, suggesting that receptor-linked tyrosine kinases might regulate the functions of RINs through their interactions with tyrosinephosphorylated receptors and/or adaptors. We are currently investigating how the Ras members and/or receptor stimulation exert their influences on the GEF activities of RINs.
One of the important findings of this study is the identification of amphiphysin II/BIN1 as another binding partner for RIN3. Amphiphysin II forms heterodimer with amphiphysin I (Wigge et al., 1997) and they are involved in endocytosis, particularly in synaptic vesicle recycling (Wigge and McMahon, 1998). Many amphiphysin II isoforms appear to arise through alternative splicing of a single gene (Butler et al., 1997;Kadlec and Pendergast, 1997;Leprince et al., 1997;Ramjaun et al., 1997;Sakamuro et al., 1996;Sparks et al., 1996;Tsutsui et al., 1997). Among them, the isoform that was cloned to interact with RIN3 in the present study lacks exon 10 (encoding the nuclear localization signal), exon 12 (which mediates interactions with clathrin and AP-2) Ramjaun and McPherson, 1998) and exon 13 (a part of the c-myc binding domain). This form, also called BIN1-10-13, is ubiquitously expressed (Wechsler-Reya et al., 1997), although its exact role is unclear. In the present study, we showed that the C-terminal SH3 domain of amphiphysin II is necessary and sufficient for the interaction with RIN3 (Fig. 6), and that this domain is present in all the isoforms. Thus, it is very likely that RIN3 associates with all spliced forms of amphiphysin II, including the one that can interact with clathrin and AP-2.
We also revealed here that amphiphysin II is recruited to RIN3-positive vesicles in the endocytic transport pathway. This translocation appears to depend on the association between the PRDs of RIN3 and the SH3 domain of amphiphysin II (Figs 5,6). Amphiphysin II has been reported to interact via its SH3 domain with dynamin and synaptojanin (Leprince et al., 1997;Ramjaun et al., 1997;Wigge et al., 1997). Dynamin functions in the fission steps of clathrin-coated vesicles from the plasma membrane. Synaptojanin is a PtdIns-5-phosphatase, which Journal of Cell Science 116 (20) hydrolyses PtdIns(4,5)P2, and PtdIns(3,4,5)P3 (Chung et al., 1997;McPherson et al., 1996;Woscholski et al., 1997), and is enriched in nerve terminals. This phosphatase has also been implicated in the uncoating of clathrin at a step closely related to the action site of dynamin (Cremona et al., 1999;McPherson et al., 1994). Thus, amphiphysin II appears to interact not only with dynamin and synaptojanin in the fission step of clathrincoated vesicles but also with RIN3 after the fission in the endocytic transport pathway. In addition, we observed here that the RIN3-positive vesicles contain Rab5 but not the early endosomal marker EEA1 (Fig. 3A). Furthermore, transferrin appeared to be partly transported through the RIN3-positive vesicles to early endosomes (Fig. 4). Taken together, these results suggest that the unique vesicles identified by the colocalization of RIN3, Rab5 and amphiphysin II might participate in an intermediate process from endocytosis to early endosomes.
Although the molecular mechanism whereby Rab5 regulates the early endocytic pathway is still unclear, it is tempting to speculate that a regulatory system similar to the homotypic fusion process of early endosomes might also operate on the endocytic pathway. In this regard, it has been reported that Rab5 can interact with the class-Ia PtdIns-3(OH)-kinase p110β to stimulate the production of PtdIns(3,4,5)P3 (Kurosu and Katada, 2001). Interestingly, this PtdIns-3(OH)-kinase subtype is markedly enriched in clathrin-coated vesicles compared with the class-II PtdIns-3(OH)-kinase hVps34, which produces PtdIns(3)P in the early endosomes (Christoforidis et al., 1999). Taken together, the different distributions of PtdIns-3(OH)kinases and Rab5-GEFs might be responsible for the strict regulation of Rab5-dependent processes involved in many membrane trafficking pathways. Although further experiments are apparently necessary to elucidate the molecular mechanisms underlying the Rab5-regulated processes, our present data strongly suggest that the RIN family interacting with Rab5 and amphiphysin II is involved in the early endocytic pathway.