The Rab5 Effector Rabankyrin-5 Regulates and Coordinates Different Endocytic Mechanisms

1In mammalian cells multiple mechanisms of endocytosis operate within a single cell to perform nutrient uptake, cellular homeostasis, neurotransmission, signal transduction, antigen presentation, and defence against pathogens. Endocytosis comprises two major categories, phagocytosis and pinocytosis, depending on the uptake of particles or fluid, respectively (reviewed in ). Pinocytosis encompasses various membrane entry routes and mechanisms. Clathrin-mediated endocytosis, the best-studied route at the molecular level to date, primarily serves receptor-mediated uptake. Caveolae assemble on sphingolipid and cholesterol rafts and may internalise molecules partitioning in these lipid microdomains. To a certain extent, ligand-receptor complexes and raft components can follow both entry routes but can also be internalised via clathrin- and caveolae-independent endocytosis. Large volumes of fluid are engulfed by the closure of plasma membrane protrusions in a process termed macropinocytosis. Although it was the first mode of pinocytosis reported, little is known on the molecular mechanisms underlying this physiologically very important endocytic process. Constitutively active in immature dendritic cells, macropinocytosis favors antigen sampling. However, it can be transiently induced in most cells by growth factors, tumor-promoting chemicals such as phorbol 12-myristate 13-acetate (PMA), or oncogenes like H-Ras or v-Src, and it has been proposed to downregulate signalling molecules from the cell surface. Since macropinosomes arise from membrane ruffles, regions of intense actin remodelling which are also a trait of motile cells, they have been implicated in directed cell locomotion. Macropinocytosis is exploited by several invasive pathogens as entry route but differs from phagocytosis with respect to regulation (e.g., receptor mediated) and cargo (e.g., uptake and degradation of opsonised particles). Macropinosomes are distinct from early endosomes, morphologically and biochemically. In epidermal growth factor (EGF)-stimulated A431 cells, whereas macropinosomes can fuse homotypically, they seldom fuse with early endosomes. However, this partition is not absolute, since in dendritic cells macropinosomes do fuse with endosomes. The intracellular trafficking properties of macropinosomes may therefore be governed by cell type-specific mechanisms and fulfil specialised functions. Finally, macropinocytosis shares mechanistic features with apical fluid-phase endocytosis in polarised cells. In kidney, apical pinocytosis contributes an essential function to the physiology of the renal system by contributing to vectorial fluid-phase transport across the membranes.

2Given the dependence on actin remodelling, it is no surprise that Rho GTPases, ARF6, and type 1 phosphatidylinositol-3 kinases (PI3-Ks) are involved in macropinocytosis, presumably through their role in membrane ruffling. Much less clear is the function of two Rab GTPases, Rab5 and Rab34/Rah, both of which are implicated in the formation of macropinosomes. Rab34/Rah colocalises with actin to membrane ruffles and nascent macropinosomes and its overexpression promotes macropinocytosis. Rab5 regulates fluid-phase and receptor-mediated uptake, and its activity has been linked to Ras-stimulated fluid-phase endocytosis. However, Rab5 colocalises to some, but not all, macropinosomes harbouring Rab34/Rah. Furthermore, Rab5 together with its effector PI3-K, hVPS34, regulates the recruitment of a set of cytosolic FYVE-finger effector proteins on early endosomes. These molecules cooperate in the tethering and fusion of clathrin-coated vesicles (CCVs) with early endosomes and homotypic endosome fusion as well as the motility of early endosomes along microtubules. However, in view of the function of Rab5 on the early endosomes, its role in macropinocytosis is difficult to assess. Specifically, it is unclear whether the large vacuoles induced by Rab5 activation, either via the expression of activated mutants or via signalling molecules that stimulate the GTP loading on Rab5, correspond to either macropinosomes or enlarged endosomes originated by increased homotypic early-endosome fusion, or both. Here, the identification of a novel Rab5 FYVE-finger effector, Rabankyrin-5, led us to revisit the role of Rab5 in fluid-phase (macro)pinocytosis in both nonpolarised and polarised cells.

3We have previously purified a large number of Rab5 effectors by an affinity chromatography approach based on glutathione S-transferase (GST)-Rab5-GTPγS (Christoforidis et al.1999a). Among the numerous proteins purified, we focused on a prominent 130-kDa protein (Figure 1A). The protein was digested with trypsin and the amino acid sequences of the tryptic fragments were determined by nanoelectrospray tandem mass spectrometry. The identified peptides matched the mouse sequence of Ankhzn (unpublished data). The corresponding human cDNA was obtained by PCR, using primers derived from the mouse DNA sequence and a random primed HeLa cDNA library as template. The sequence of the human cDNA matched the recently revised hAnkhzn gene sequence (NP 057460.2). The predicted human p130 primary sequence consists of an N-terminal BTB/POZ domain, a C-terminal FYVE-finger, and 21 successive ankyrin (ANK) repeats in between these two domains (Figure 1B). In light of the fact that p130 is a Rab5 effector (this study), we propose to amend its name to Rabankyrin-5 (Rab5 binding and ankyrin repeats containing protein).

4To investigate the function of human Rabankyrin-5, we raised antibodies against the recombinant full-length protein. By Western blotting, these antibodies detected a predominant band of 130 kDa in both bovine brain and HeLa cytosol, in the corresponding GST-Rab5-GTPγS but not GST-Rab5-GDP affinity column eluates (Figure 1C). Evidence for a direct interaction with Rab5 was obtained by incubating in vitro-translated Rabankyrin-5 with beads displaying GST-Rab5, -Rab4, -Rab7, and -Rab11, preloaded with either GDP or GTPγS. Figure 1D shows that human Rabankyrin-5 binds to GST-Rab5-GTPγS, but neither to beads containing GST-Rab5-GDP nor to other Rab proteins. We conclude that Rabankyrin-5 binds to Rab5 specifically, directly, and GTP dependently.

5After EEA1 and Rabenosyn-5, Rabankyrin-5 is the third Rab5 effector containing a FYVE-finger. Since the FYVE-finger is a key determinant for targeting these proteins to early endosomes via its binding to phosphatidylinositol-3-phosphate [PI(3)P], we tested whether this holds true also for Rabankyrin-5. First, recombinant Rabankyrin-5 interacted significantly with PI(3)P in a liposome-binding assay (Figure 1E). Second, endogenous Rabankyrin-5 colocalised significantly with EEA1 and Rab5-positive early endosomes (approximately 80%) in A431 cells by confocal immunofluorescence microscopy analysis (see Figure 3A and 3B below).

6We next investigated whether Rabankyrin-5 plays a role in homotypic early endosome fusion as well as fusion of CCVs with early endosomes, similar to EEA1 and Rabenosyn-5. Quantitative immunodepletion of the cytosolic pool of Rabankyrin-5 using anti–Rabankyrin-5 antibodies (Figure 2A) had only minor effects on early endosome fusion in vitro compared to preimmune serum (Figure 2B). Considering the possibility that the early-endosome–associated pool of Rabankyrin-5 may still be sufficient for fusion, we attempted to neutralise its function by the addition of anti–Rabankyrin-5 antibodies on top of Rabankyrin-5–depleted cytosol. This treatment strongly inhibited early-endosome fusion (Figure 2B). The fusion of CCVs with early endosomes was minimally affected under these conditions (Figure 2C), suggesting that Rabankyrin-5 is not essential in this process, at least in vitro. However, addition of increasing amounts of recombinant Rabankyrin-5 (at concentrations in the range of the endogenous protein; see Materials and Methods) on either immunodepleted cytosol or nontreated cytosol significantly enhanced both homo- and heterotypic fusion activity. We conclude from these experiments that endogenous Rabankyrin-5 colocalises with EEA1 on Rab5 endosomes in vivo, through the interaction with Rab5 (see Figure 1D) and PI(3)P (see Figure 1E). It plays a modulatory role in early-endosome fusion since it can stimulate both homotypic and heterotypic fusion of CCVs with early endosomes. However, it does not appear to be strictly required in the latter reaction in vitro.

7Surprisingly, we noted that some peripheral structures harbouring Rabankyrin-5 overlapped with Rab5 but either poorly or not at all with EEA1 in A431 cells (Figure 3A, arrows). Upon overexpression, Rabankyrin-5 induced the accumulation of several vesicles of variable size in the periphery of the cells, most of them lacking EEA1 (Figure 3B, arrows). The irregular size, intracellular distribution, and lack of endosomal markers (see “Rabankyrin-5 localises to macropinosomes” below) of these structures were suggestive of (macro)pinocytic vesicles. We therefore set out to investigate the putative role of Rabankyrin-5 in macropinocytosis.

8Macropinosomes have been so far defined by a combination of criteria rather than specific molecular markers. To test whether the Rabankyrin-5 vesicles fulfil the previously described requirements of macropinosomes, we determined whether they (1) are stimulated by growth factors, (2) do not promote receptor-mediated endocytosis of transferrin, (3) engulf large amounts of extracellular fluid, (4) depend on PI3-K activity, and (5) follow actin rearrangements at the plasma membrane. A431 and NIH3T3 cells are established experimental systems that exhibit different features in the regulation of macropinocytosis. In A431 cells, macropinocytosis can be transiently induced upon growth factor stimulation, whereas NIH3T3 cells exhibit constitutive macropinocytic activity. We first tested whether EGF can induce the formation of Rabankyrin-5–positive macropinocytic structures. Upon treatment of A431 cells with rhodamine-conjugated EGF, endogenous Rabankyrin-5 colocalised to enlarged structures which, importantly, were largely devoid of simultaneously internalised transferrin (Figure 3C). Interestingly, EGF was not excluded but rather enriched in these structures. In addition, we observed that such vesicles were also enriched in tyrosine-phosphorylated proteins, presumably EGF-receptor and downstream signalling molecules (Figure 3D).

9In NIH3T3 fibroblasts, the staining pattern of endogenous Rabankyrin-5 was similar to that of EGF-stimulated A431 cells. Rabankyrin-5 was detected on enlarged structures, predominantly at the outermost periphery and in membrane protrusions of the cell (Figure 4A, arrows). Consistent with these observations, immunoelectron microscopy revealed that antibodies to Rabankyrin-5 labelled electron-lucent vesicular structures underlying the plasma membrane in the cell periphery and in cell protrusions (Figure 5). The labelled structures were sparse in untransfected cells, but in Rabankyrin-5– overexpressing cells fortuitous sections revealed concentrated areas of Rabankyrin-5–labelled structures. Unfortunately, we failed to distinguish these structures from conventional early endosomes by double immunolabelling experiments due to technical limitations. By light microscopy, the appearance of vacuole-shaped structures devoid of the early endosomal marker EEA1 was dramatically increased upon Rabankyrin-5 overexpression. In addition, after a 3-min pulse, these structures were intensively labelled by a fluid-phase marker (see Figure 4B, arrows) but not by transferrin (unpublished data). Given the potent induction of macropinosomes by Rabankyrin-5 in NIH3T3 cells, we used this experimental system to explore further the regulatory parameters of macropinocytosis.

10In NIH3T3 cells, Rabankyrin-5–enlarged structures were labelled by an 8-min uptake of fluorescein isothiocyanate (FITC)-dextran (MW, 10.000). Consistent with their macropinocytic nature, these structures were largely depleted of transferrin (see Figure 4C, arrows) and susceptible to the PI3-K inhibitors wortmannin (see Figure 4D) and LY 294002 (unpublished data). These drugs both reduced the formation of peripheral enlarged vesicles and almost completely blocked the internalisation of high-molecular-weight dextran (MW, 70.000) (Figure S1A) that enters the cell almost exclusively via macropinocytosis. Interestingly, we noted that, unlike EEA1, Rabankyrin-5 did not readily dissociate from membranes (see Figure 4D) upon inhibition of PI3-K activity, suggesting that endosome-targeting determinants other than PI(3)P contribute to its localisation. Neither markers of late endosomes (Rab7), nor of caveosomes (caveolin), nor of the clathrin-independent pathway which is used by glycosyl-phosphatidylinositol (GPI)-anchored proteins (GFP-GPI) overlapped significantly with the peripheral enlarged Rabankyrin-5 structures (unpublished data).

11To confirm that the Rabankyrin-5 structures originated from the plasma membrane through an actin-driven process, we employed time-lapse video fluorescence microscopy on NIH3T3 cells transiently coexpressing yellow fluorescent protein (YFP)-Rabankyrin-5 and cyan fluorescent protein (CFP)-β-actin. In the sequence depicted in Figure 4E (see also Video S1), membrane ruffling driven by actin resulted in the formation of a large, plasma membrane-derived vesicle acquiring Rabankyrin-5. Some newly formed vesicles shed off Rabankyrin-5 over time and seemed to regurgitate back to the plasma membrane via an actin comet-tail. Altogether, the data support the hypothesis that the enlarged structures containing Rabankyrin-5 are macropinosomes and can be molecularly distinguished from early endosomes by being mostly, albeit not exclusively, devoid of EEA1. Moreover, and contrary to earlier findings, we found that in A431 cells, EGF but not transferrin could be enriched in macropinosomes.

12The GTP-dependent interaction with Rab5 suggests that Rabankyrin-5 may function as a downstream effector of Rab5 in the macropinocytic pathway. Accordingly, one would expect that Rab5 localises to macropinosomes and, together with Rabankyrin-5, is required for macropinocytosis. We found that Rab5 indeed colocalises with Rabankyrin-5 on macropinosomes (see Figure 3B, 3D, and 6A; unpublished data), which are operationally defined as vesicles (1) enriched in fluid-phase marker but depleted of transferrin receptor, (2) of large size (0,2–2 μm), and (3) negative for EEA1 but positive for Rabankyrin-5. Consistent with a requirement for Rab5 in macropinocytosis, overexpression of RN-tre, a GTPase-activating protein (GAP) for Rab5, significantly reduced (by 60% ± 10%) the uptake of large fluorescent dextran (MW, 70.000) into transferrin- and EEA1-negative structures in NIH3T3 cells (Figure 6B and 6C). Similar results were also obtained when NIH3T3 cells were cotransfected for Rab5S34N and Rabankyrin-5 (unpublished data), implying that Rabankyrin-5 induces macropinocytosis Rab5-GTP dependently.

13Whereas the aforementioned results demonstrate a general requirement for Rab5 and Rabankyrin-5 in pinocytosis, they do not discriminate between clathrin-dependent and-independent endocytosis. To distinguish between the two pathways, we compared the effects of Rabankyrin-5 overexpression on transferrin and horseradish peroxidase (HRP) uptake in NIH3T3 cells, biochemically. A recombinant adenovirus was generated to overexpress Rabankyrin-5 in these cells. In comparison with mock-infected cells, Rabankyrin-5 overexpression increased the uptake of HRP, particularly after longer (10 min) incubation times (Figure 7A). In contrast, the rate of transferrin uptake was unaffected, suggesting that the increase of fluid phase was not due to increased clathrin-mediated endocytosis (Figure 7C). The increased intracellular accumulation of HRP induced by Rabankyrin-5 seems to be due to increased uptake rather than inhibited recycling, as the latter process was only slightly decreased (Figure 7B).

14To determine whether Rabankyrin-5 is also required for macropinocytosis, we attempted to ablate the protein by RNA interference using endoribonuclease-prepared small interfering RNA (esiRNA). Since we optimised this technique in human cells, we conducted these experiments in EGF-stimulated A431 cells where an approximately 55% knock-down of Rabankyrin-5 was achieved (Figure 7D). EGF stimulation of control-treated cells enhanced HRP uptake 2-fold versus control-treated, non-EGF-stimulated cells, as reported previously. Reduction of Rabankyrin-5 expression in esiRNA-transfected cells caused an inhibition of HRP of up to approximately 50% by selectively affecting the burst of pinocytosis induced by EGF (within 5 min) (Figure 7E). Under the same conditions, transferrin uptake was marginally inhibited (Figure 7F). We conclude from these experiments that inhibition of fluid-phase uptake occurs due to an inhibition of (macro)pinocytosis rather than of clathrin-mediated endocytosis.

15To confirm the requirement of Rabankyrin-5 for fluid-phase endocytosis we used an independent method: quantification of the uptake of fluorescent dextran in NIH3T3 cells microinjected with affinity-purified anti–Rabankyrin-5 antibodies. Following microinjection, FITC-dextran uptake was internalised for 30 min to label macropinosomes. Microinjection of anti–Rabankyrin-5 antibodies inhibited FITC-dextran uptake by 53% ± 5% when compared with cells injected with preimmune serum (Figure S2), whereas the uptake of rhodamine-transferrin was not significantly changed. Altogether, these results suggest that Rabankyrin-5 is a Rab5 effector required for the formation of macropinosomes.

16Apical pinocytosis in polarised epithelial cells and growth factor-induced macropinocytosis share mechanistic properties, like dependence on actin and PI3-K activity. In view of the role of Rab5 in endocytic trafficking of polarised cells such as epithelial cells and neurons, we explored the role of Rabankyrin-5 in apical endocytosis in polarised epithelial cells. First, we determined the intracellular localisation of Rabankyrin-5 on cryosections of adult mouse kidney, since the epithelial cells in this organ, especially proximal tubule cells, exhibit high levels of apical pinocytosis under physiological conditions. Apical endocytic vesicles were pulse-labelled by injecting mice with HRP followed by fixation of the kidney through perfusion after 5 min. Prominin-1 was used as marker for the apical brush border of proximal tubules in combination with affinity-purified anti–Rabankyrin-5 antibodies. Basement membrane staining was demonstrated by immunolabelling of endogenous mouse immunoglobulin G (IgG). As illustrated in Figure 8, Rabankyrin-5 immunoreactivity was mainly detected on vesicle-like structures underneath the apical, prominin-1–labelled brush border. Most of the vacuole-like Rabankyrin-5 structures contained HRP, suggesting that they correspond to apical endocytic structures (Figure 8D, arrows). By further investigating the nature of these Rabankyrin-5– and HRP-labelled structures, we observed that some vesicles were immunoreactive for LAMP-1 (Figure 8D, arrowheads), suggesting that Rabankyrin-5 (macro)pinocytic vesicles can acquire characteristics of late endocytic compartments.

17This was confirmed and extended by immunogold electron microscopy on ultrathin frozen sections of mouse proximal tubule. Rabankyrin-5 showed weak labelling of the apical microvilli-covered surface of proximal tubule cells but labelled large electron-lucent structures underlying the apical surface (Figure 9A and 9B). Consistent with the immunofluorescence analysis, double labelling with LAMP-1 antibodies confirmed that many of these structures were LAMP-1 negative or weakly labelled. However, a significant number of Rabankyrin-5–positive structures were also LAMP-1 positive. Rabankyrin-5 immunoreactivity was also observed in distal tubule cells and other segments of the nephron.

18To functionally address the role of Rabankyrin-5 in pinocytosis, we used filter-grown Madin-Darby canine kidney (MDCK) cells as an established system to study polarised trafficking. Alexa 488 dextran was internalised for 13 min from either the apical or the basolateral chamber. Although endogenous Rabankyrin-5 exhibited partial colocalisation with fluid-phase marker internalised from the basolateral side (unpublished data), a striking degree of overlap between the internalised tracer and Rabankyrin-5 on often ring-shaped structures underlying the apical region was observed (Figure 10A). These large vesicles were neither detectable in the nuclear area (Figure 10B) nor close to the basolateral domain (Figure 10C). We next tested whether Rabankyrin-5 overexpression could stimulate fluid-phase uptake from the apical plasma membrane domain. Rabankyrin-5 was overexpressed using the recombinant adenovirus system in MDCK cells grown on coverslips and the cells stained for Rabankyrin-5 and EEA1. As in A431 and NIH3T3 cells, and similar to endogenous Rabankyrin-5 in MDCK cells (Figure 11A), the overexpressed protein exhibited significant colocalisation with endogenous EEA1, but in addition, it induced the formation of enlarged structures devoid of EEA1 in the periphery of the cell (Figure 11B). The recombinant adenovirus was next applied to infect filter-grown MDCK cells. In these cells, it induced enlarged Rabankyrin-5 structures predominantly underlying the apical domain that contained dextran internalised from the apical side. Some of these structures were EEA1 positive, whereas others lacked this marker of early endosomes (see Figure 10D, arrows). These structures were also detectable to some extent in the nuclear region (see Figure 10E) and basolateral membrane domain (see Figure 10F). In addition, there were some enlarged Rabankyrin-5 vesicles that were labelled for neither dextran nor EEA1. These may correspond to macropinosomes that were formed prior to the labelling with the fluorescent fluid-phase marker.

19We next measured fluid-phase uptake biochemically by quantifying the internalisation of HRP either from the apical or the basolateral side for various periods of time. Filter-grown MDCK cells were either infected with adenovirus encoding Rabankyrin-5 or control virus prior to HRP internalisation. While internalisation of HRP from the basolateral plasma membrane did not significantly vary (Figure 11D), uptake from the apical membrane was increased up to 65% in cells overexpressing Rabankyrin-5 when compared with control cells (Figure 11C). Given that Rabankyrin-5 was overexpressed in only 50–60% of the cells, this value may be an underestimate. Since the kinetic profiles resembled the ones obtained in NIH3T3 cells, we also tested whether the increase in HRP uptake may have been due to an inhibition of recycling or transcytosis. Neither of these trafficking pathways was significantly affected (unpublished data). To verify that the increase in fluid-phase uptake was not due to stimulation of clathrin-mediated endocytosis, we measured the internalisation of FcLR 5–27, a chimeric receptor between the IgG Fc receptor and the LDL receptor which, when expressed in MDCK cells, is targeted to, and internalised from, both apical and basolateral plasma membrane domains. Rabankyrin-5 overexpression did not affect significantly the internalisation of FcLR 5–27, as revealed by the uptake of a biotinylated Fab-fragment of the 2.4G2 monoclonal anti-FcRII antibody (Figure 11E and 11F). Taken together, these findings indicate that Rabankyrin-5 is predominantly localised to apical endocytic structures and specifically stimulates apical, non-clathrin-mediated fluid-phase endocytosis upon overexpression in polarised epithelial cells.

20Since the pioneering work of almost two decades ago, it has been established that certain oncogenes and signalling molecules induce actin-dependent membrane ruffling and macropinocytosis. Subsequent work has advanced our understanding of the signalling pathway that led to growth factor-dependent actin remodelling and membrane ruffling. Despite macropinocytosis being the most ancient form of pinocytosis and an endocytic process of high physiological importance, progress in determining the molecular mechanisms that regulate the generation and trafficking of macropinosomes has advanced to a lesser extent. It has been demonstrated that macropinocytosis differs from clathrin-mediated endocytosis and that the two pathways lead to distinct endocytic structures. Rab5, an established regulator of clathrin-mediated endocytosis and endosome dynamics, has also been implicated in macropinocytosis. This hypothesis was based on the expression of mutants that induced or inhibited fluid-phase endocytosis and formation of giant endocytic vesicles. In those studies, however, an accurate assessment of whether the enlarged compartments corresponded to macropinosomes or were produced by the coalescence of multiple endosomes by homotypic fusion was not determined.

21Here, we provide evidence that Rabankyrin-5 is a novel Rab5 effector which, in addition to being localised to early endosomes, is associated with macropinosomes and promotes their formation according to previously established criteria. First, Rabankyrin-5 localised to enlarged vesicles that were stimulated by growth factors (e.g., EGF). Second, these vesicles differed from early endosomes by the lack of constitutively endocytosed molecules, such as transferrin, and components of the transport machinery, such as EEA1. Third, the vacuoles originated from the plasma membrane and engulfed extracellular fluid. Fourth, their formation depended on PI3-K activity, and fifth, they followed actin rearrangements at the plasma membrane.

22Macropinosomes are formed by the closure of membrane protrusions generated upon actin-mediated membrane ruffling. Although membrane ruffling is required for macropinocytosis, it seems not to be sufficient for macropinosome formation. Our data suggest that Rabankyrin-5 is a novel regulator of this endocytic mechanism. One possibility is that it could play an active role in the generation of macropinosomes (see Figure 7). The deduced primary structure of Rabankyrin-5 predicts the existence of several protein-protein interaction motifs, suggesting a role as a multifunctional adaptor protein. The N terminus of Rabankyrin-5 contains a BTB/POZ motif, which is present in proteins involved in signalling, development, and tumorigenesis and mediates homo- and heterodimerisation. It contains 21 ANK repeats which, by analogy with the function of ANKs, could interact with different proteins on the membrane, including Rab5, and mediate the assembly of a multiprotein complex. Such a scaffolding function has been postulated to bridge the membrane to the actin cytoskeleton but also to link proteins involved in endocytosis and signal transduction. Rabankyrin-5 could assemble a membrane-cytoskeleton scaffold committing the ruffling membrane to generate a pinocytic vesicle or directly participate in the closure of the plasma membrane sealing the vesicle. Alternatively, Rabankyrin-5 could be recruited onto macropinosomes concomitantly or following their formation to prevent their back-fusion with the plasma membrane or regulate their onward trafficking. This possibility receives support from our time-lapse video microscopy analysis, where the dissociation of Rabankyrin-5 from the membrane precedes what appears to be the regurgitation of macropinosomes back to the plasma membrane (as shown in Video S1). Interestingly, it has recently been proposed that ANK repeats may not be simply anchoring domains but may be part of a mechanical sensory system that transmits tension from the cytoskeleton to ion channels in the membrane. In the case of Rabankyrin-5, the ANK repeats may confer a sensory function to detect and adapt to rearrangements of the actin cytoskeleton during membrane ruffling and membrane closure. Interactions of the ANK repeats with ion channels may sense changes in water-salt homeostasis and pH, and respond to them via modulation of macropinocytosis and macropinosome trafficking.

23How does the function of Rab5 and Rabankyrin-5 in macropinocytosis relate to early endosome trafficking? We have previously proposed a model whereby Rab5, via the recruitment of its effectors, generates and maintains a spatially restricted and functionally specialised membrane domain on the early endosomes. The Rab5 effector hVps34 is a PI3-K that generates PI(3)P. Together with Rab5, this phosphoinositide serves as a binding determinant for the FYVE-finger Rab5 effectors EEA1 and Rabenosyn-5 on the early-endosome membrane. The Rab5 effectors form oligomeric complexes and play distinct but cooperative roles in membrane tethering, fusion, and motility of early endosomes along microtubules. Our data extend the list of PI(3)P-binding Rab5 effectors to Rabankyrin-5. Rabankyrin-5 colocalises with Rab5 and EEA1 to early endosomes and plays a role in homotypic early-endosome fusion, further underscoring the activity of Rab5 as organiser of an endosomal Rab5 domain enriched in PI(3)P-binding effector proteins. Rabankyrin-5 played a minor modulatory role in the fusion of CCVs with early endosomes, suggesting mechanistic differences between homotypic early-endosome fusion and the heterotypic fusion of CCVs with early endosomes. Clearly, Rabankyrin-5 exerted the most striking effects on fluid-phase rather than clathrin-mediated endocytosis.

24Clathrin-dependent endocytosis and macropinocytosis are independent but interconnected pathways. Whereas macropinosomes and early endosomes remain segregated in EGF-stimulated A431 cells, they can fuse in other cells, such as dendritic cells and macrophages. Rabankyrin-5 alone may not be sufficient for macropinosomes to fuse with endosomes, and this activity may require the additional recruitment of EEA1 and Rabenosyn-5. The extent to which macropinosomes can recruit Rabankyrin-5, release or retain it, and further acquire late endocytic components or other Rab5 effectors acting in endosome tethering and fusion may depend on the cellular context. In kidney cells, for example, several Rabankyrin-5–positive structures appeared also to contain EEA1 and late endocytic markers (LAMP-1), suggesting that macropinosomes can communicate with other endocytic organelles and undergo maturation as it has been described for phagocytosis. Also intriguing is the weak interaction of Rabankyrin-5 with PI(5)P (see Figure 1E) and recent findings that BTB domain-containing proteins interact with ubiquitin ligases. Since both components have been implicated in the biogenesis of multivesicular bodies, Rabankyrin-5 may function in the trafficking to late endocytic components, as it was described previously for PIKfyve, another FYVE-finger–containing protein.

25Stimulation of epithelial cells with either EGF or phorbol esters increases fluid-phase uptake while reducing internalisation via clathrin-mediated endocytosis. Our results provide a possible explanation for how the two endocytic routes can be quantitatively balanced. Since Rab5 appears to be rate limiting for both receptor-mediated and fluid-phase endocytosis, its activity could be shifted between these two endocytic pathways, depending on the stimuli and the cell type. The shared activity of Rab5 and Rabankyrin-5 would ensure coordination between endosome trafficking and macropinocytosis, regulate the kinetics, and limit the extent of both endocytic processes, thus preserving plasma membrane homeostasis.

26In addition to macropinocytosis in nonpolarised cells, we have found that Rabankyrin-5 regulates apical, non-clathrin-mediated pinocytosis in polarised epithelial cells. The role of apically stimulated pinocytosis is crucial for the physiology of various organs. For example, it plays an important role in the reabsorption of proteins from the glomerular filtrate in the renal proximal tubule. Rabankyrin-5 may be a regulator of this process and this possibility is supported by our findings, which show that it (1) localises to subapical compartments in kidney proximal tubules and (2) induces fluid-phase, clathrin-independent uptake from the apical but not basolateral side in polarised MDCK cells. As for macropinocytosis, apical endocytosis depends on actin remodelling mediated by Rho family GTPases and ARF6. Since macropinocytosis appears to occur primarily from the apical side, the involvement of epidermal growth factor receptor as demonstrated for nonpolarised cells is unlikely since it is primarily localised to the basolateral domain of polarised epithelial cells. However, apical pinocytosis depends on other components of receptor tyrosine kinase signalling pathways such as PI3-K and protein kinase C and can be induced by oncogenes (e.g., v-Src) that stimulate macropinocytosis in nonpolarised cells. It is not known which PI3-K subtype functions in apical pinocytosis, but an attractive candidate is PI3-Kβ, as it has been identified as a Rab5 effector. Its activity could be regulated by Rab5 constitutively and/or be subjected to stimulation by apical or axonal signalling molecules. In neurons, Rac- and actin-dependent endocytosis of Eph receptor-ephrin complexes is required to control repulsive versus attractive cell movement during tissue patterning in embryonic development and may play a role in the structural plasticity of synapses. It will be interesting to determine whether Rabankyrin-5 and the macropinocytic machinery play a role in this event.

27In conclusion, the identification of Rabankyrin-5 opens new opportunities for investigating the molecular principles underlying macropinocytosis and its regulation by signalling molecules. The dual role of Rab5 and Rabankyrin-5 in endosome and macropinosome function argues for a role in coordinating these two different endocytic mechanisms. Future work is required to identify the molecular partners of Rabankyrin-5 and to establish the mechanisms responsible for its membrane targeting and its role in endocytic membrane dynamics.

28Phospholipids were purchased from Sigma (St. Louis, Missouri, United States) and Calbiochem (San Diego, California, United States). FITC- or Texas Red-conjugated dextran (lysine fixable; MW, 10.000 and 70.000), rhodamine-conjugated EGF, and rhodamine-conjugated transferrin were purchased from Molecular Probes (Eugene, Oregon, United States). Wortmannin and LY 294002 were from Calbiochem. A431 and NIH3T3 cells were grown in DMEM (high glucose) supplemented with 10% (v/v) heat-inactivated foetal calf serum, 100 U/ml penicillin, 100 μg/ml streptomycin, and 2 mM L-glutamine. MDCK II cells were grown in MEM plus 10% FCS. MDCK II cells stably transfected for FcLR 5–27 were a gift of Professor Ira Mellman.

29The 130-kDa protein band was excised from gels and enzymatically digested. The tandem mass spectroscopy protein sequencing procedure was performed as described previously. Bovine peptides determined from Rabankyrin-5 were from bovine brain and were used to identify corresponding expressed sequence tags using BLAST similarity searches at NCBI. A random primed HeLa cDNA library was used in a PCR reaction with primers based on the mouse cDNA to obtain the full-length clone of Rabankyrin-5.

30Recombinant proteins of Rabankyrin-5 were expressed in SF plus insect cells according to the manufacturer's instructions using Rabankyrin-5 cDNA subcloned into pfastBAC vector or subcloned into pGEX-6P series and expressed in BL21 cells.

31Human anti-EEA1 serum (1:10.000) was a gift from Ban Hok Toh (Monash Medical School, Adelaide, Australia), whereas a mouse monoclonal EEA1 antibody, a sheep polyclonal anti-HRP antibody and an antibody against phosphotyrosine (α-4G10) were purchased from BD Bioscience (Heidelberg, Germany), Abam (Cambridge, United Kingdom), and Upstate Biotechnology (Lake Placid, New York, United States), respectively. Fluorescently labelled secondary antibodies were purchased from Molecular Probes. Recombinant full-length Rabankyrin-5 was used to raise a polyclonal antibody in rabbit. Human full-length Rabankyrin-5 was cloned into pEYFP vectors containing an N-terminal tag (pEYFP-C1; Clontech, Palo Alto, California, United States). Recombinant adenovirus encoding full-length Rabankyrin-5 was generated according to the manufacturer's protocol (AdEasy). cDNAs encoding Rab5wt were fused to the amino termini of ECFP or EYFP, as previously described. The antibody for the FcLR 5-27 chimeric receptor was purified from 2.4G2 hybridoma supernatant by ammonium sulfate precipitation. For the preparation and biotinylation of Fab fragments of 2.4G2, 6 mg of purified 2.4G2 IgG was digested with insoluble papain enzyme as described by the manufacturer (Sigma). Fab fragments were biotinylated with NHS-LC-biotin or NHS-SS-biotin (Pierce, Rockford, Illinois, United States).

32The GST-Rab5 affinity chromatography was performed as in. 35S-Methionine–labelled proteins were transcribed and translated in vitro using a TnT-coupled transcription-translation kit (Promega, Madison, Wisconsin, United States). For Rab5 effector recruitment assays, in vitro-translated proteins were incubated with glutathione-sepharose beads complexed with GST-Rab5-GTPγS or GST-Rab5-GDP and eluted as described previously.

33Cells were transfected with plasmids using FuGENE6 according to the manufacturer's instructions (Roche, Basel, Switzerland). After 16–24 h, transfected cells were washed twice with PBS and fixed with 3% paraformaldehyde and immunofluorescence labelling was performed according to standard procedures. Cells were mounted in moviol and examined on a confocal microscope (Leica TCS SP2; Leica, Wetzlar, Germany) with a 100×/1.40 plan-Apochromat objective. Fluorescent images were collected using the Leica IM500 Image Manager and processed using Adobe Photoshop v5.5 (Adobe Systems, San Jose, California, United States).

34NIH3T3 cells grown on glass coverslips and transfected with plasmids encoding RN-tre were incubated with 3 mg/ml FITC-labelled dextran (MW, 70.000) in DMEM supplemented with 1% FCS and 20 mM HEPES (pH 7,4), fixed, and stained with rhodamine-conjugated antirabbit secondary antibody. Dextran uptake was determined by quantifying grey values of thresholded, fluorescent images of at least 15 cells using MetaVue 6.1r3 and ImageJ (NIH).

35In vitro fusion assays were performed using early endosomes labelled with biotinylated transferrin or antitransferrin antibody as well as CCVs labelled with biotinylated transferrin prepared from HeLa cells as described previously. The concentration of Rabankyrin-5 in HeLa cell cytosol in the in vitro early-endosome fusion assay was estimated to be approximately 50 nM, as determined by comparison with a serial dilution of recombinant Rabankyrin-5 analysed by SDS-PAGE, followed by Western blotting analysis. Recruitment of cytosolic proteins to early endosomes was performed as described in. Recombinant Rabankyrin-5 was recruited to phosphoinositides as described previously using liposomes (98% phosphatidylcholine, 2% phosphoinositides; 2 mg/ml final concentration in 30 mM HEPES-NaOH [pH 7,2], 120 mM NaCl, 0,5 mM EGTA) prepared as reported previously.

36Time-lapse epifluorescence video microscopy was performed using an Olympus IX70 (Olympus, Hamburg, Germany) inverted microscope equipped with a polychrome II monochromator (TILL Photonics, Martinsried, Germany), a custom filter block for simultaneous visualization of YFP and CFP (AHF Analysentechnik, Tübingen, Germany), a 100× oil-immersion objective (NA 1.35, Olympus) attached to a PIFOC z-SCAN (Physik Instrumente, Waldbronn, Germany), an incubation chamber (37 °C), and a 12-bit CCD digital camera IMAGO (0,134 μm pixel-1; 2 times 2 binning) (TILL Photonics), controlled by TILLvisION v3.3 software (TILL Photonics). Time-lapse video sequences of YFP and CFP images were merged as RGBs using TILLvisION v3.3. They were exported as single TIFF files and either further processed using Adobe Photoshop 5.5 and Illustrator 8.0 or converted into QuickTime movies using ImageJ (NIH).

37Anaesthetized mice were perfusion fixed with 4% PFA, 100 mM HEPES (pH 7,25), and 0,2% sucrose either directly or 5 min after injection with 5 mg/g body weight HRP (Sigma) into the femoral vein. Kidneys were excised and postfixed 6 h at 4 °C. Some tissue pieces were cryoprotected in 2,3 M sucrose overnight at 4 °C, frozen in liquid nitrogen, and cryosectioned on a Leica UCT.

38MDCK cells were grown on 12-mm, 0,4-μm pore Transwell polycarbonate filters (Corning Costar, Cambridge, Massachusetts, United States). Ninety-six hours after seeding, cells were infected from the apical side for 4 h at 37 °C with adenoviruses in 250 μl of complete medium. After changing medium, cells were incubated for 18 h at 37 °C and then used for biochemical assays or microscopy. NIH3T3 and MDCK cells grown on coverslips in 24-well plates were infected in 500 μl of medium. HRP uptake was performed as described previously. In brief, cells were incubated in incubation medium (IM; DMEM, 1% FCS, 24 mM HEPES [pH 7,4]) at 37 °C for 1 h. For internalisation experiments coverslips or Transwell filters (Costar) were incubated with IM containing 5–10 mg/ml HRP, 2 μg/ml biotinylated transferrin, or 5 μg/ml biotinylated Fab (2.4G2). In the case of MDCK cells, HRP and Fab were applied from either the apical or the basolateral side. After internalisation, cells were briefly washed with warm IM followed by three washes with ice-cold PBS supplemented with 1 mM CaCl2 and 1 mM MgCl2 (PBS++) containing 2 mg/ml BSA. Cell surface-bound transferrin was efficiently removed (95%) by three incubations over 10 min in IM (pH 3,4). Extracellular HRP was quenched in the cold by incubating the cells for 20 min with 20 mM MESNA in 100 mM NaCl, 50 mM Tris (pH 8,7), followed by two washes with ice-cold PBS and a further incubation for 10 min with ice-cold 50 mM iodoacetamide in PBS. For recycling experiments, cells were transferred again to 37 °C in fresh medium that was collected at the indicated time points. Thereafter, cells were washed once more with PBS and extracted for 15 min at 4 °C with 300 μl of lysis buffer (1% w/v Triton X-100, 0,1% w/v SDS, 20 mM HEPES [pH 7,4], and 100 U/ml DNase). Total HRP activity present in the cells was determined in duplicates out of at least three independent experiments using the previously described enzymatic method and standardised to the protein content of each well. Fab and transferrin uptake assay was performed as described previously.

39A431 cells were seeded in 24-well plates the day prior to transfection. Cells were transfected using oligofectamine (Invitrogen, Carlsbad, California, United States) with either 300 ng/ml double- or single-stranded (control) esiRNA derived from the I.M.A.G.E. clone (IMAGE:258664) according to the protocol of. After 4 d, cells were starved in MEM containing 2 mg/ml BSA and 24 mM HEPES (pH 7,3) for 1 h and stimulated with 50 ng/ml EGF in complete medium containing 24 mM HEPES (pH 7,3), 5–10 mg/ml HRP, and 2 μg/ml biotinylated transferrin for the indicated time points (see “Adenovirus infection and continuous uptake measurement with HRP, biotinylated transferrin, and biotinylated Fab (2.4G2)” above).

40Cells or tissues were fixed with 8% PFA in PHEM buffer and then processed for frozen sectioning according to published methods. Thawed sections were either single or double labelled. For single labelling a polyclonal anti–Rabankyrin-5 antibody followed by protein A 10 nm gold (University of Utrecht) was used. For double labelling the polyclonal anti–Rabankyrin-5 antibody was incubated together with rat anti–LAMP-1 (1D4B; courtesy of Professor M. Desjardins, University of Montreal, Montreal, Canada). The sections were then incubated with a mixture of anti-rabbit 10 nm gold and anti-rat 5 nm gold (BBI International, Cardiff, United Kingdom).

41The LocusLink (http://www.ncbi.nlm.nih.gov/LocusLink) accession numbers of proteins discussed in this paper are H-Ras (ID 3265), v-Src (ID 6714), Arf6 (ID 382), hVps34 (ID 5289), p110β (ID 5291), Rab34 (ID 83871), Rab5 (ID 5868), Ankhzn (ID 51479), Rab4 (ID 5867), Rab7 (ID 7879), Rab11 (ID 8766), Rabenosyn-5 (ID 64145), EEA1 (ID 8411), Caveolin-1 (ID 857), RN-tre (ID 9712), Transferrin (ID 7018), EGF (ID 1950), LAMP-1 (ID 16783), Prominin-1(ID 8842), PIKfyve (ID 200576), RhoA (ID 387), and PKC-α (ID 5578).