Characterization of the Rab8-specific membrane traffic route linked to protrusion formation

Cell migration is crucial for processes such as embryonic development, metastasis, wound healing and inflammation. During migration, cells polarize and acquire a leading edge and a tracking tail (Lauffenburger and Horwitz, 1996). The leading edge has actin-containing filopodia and lamellipodia that participate in the forward movement (Mitchison and Cramer, 1996), which is driven by the formation of new cell-substrate contacts in the leading edge and the dissociation of cell-substrate contacts in the tracking tail (Laukaitis at al., 2001). It is unknown how the cell coordinates protrusion, traction and detachment, but regulators of these processes include small GTPases of the Rho family (Nobes and Hall, 1999).

The role of membrane vesicle transport in modulation of cell shape is controversial (Heat and Holifield, 1991). It has been suggested that membrane internalized by coated pits is returned to the plasma membrane at the leading edge, where it helps to extend the cell forward (Bretscher, 1996). Whether this membrane flow operates separately from actin dynamics is not yet clear. However, recent studies indicate that an Arf6 modulates cell morphogenesis and motility by modulating actin dynamics and the turnover of adhesion receptors (Palacios et al., 2001; Powelka et al., 2004). Arf6 regulates a clathrin-independent endocytic pathway, through which molecules like the major histocompability complex class I protein (MHCI), CD59 and the interleukin-2 receptor α subunit (Tac) are internalized (Naslavsky et al., 2004). Moreover, Arf6 also participate in processes like exocytosis, phagocytosis and cytokinesis (D'Souza-Schorey and Chavrier, 2006).

Rab GTPases regulate various membrane transport pathways, and each Rab protein has a distinct location corresponding to the pathway that it regulates (Zerial and McBride, 2001). We have shown that Rab8 promotes the polarized transport of newly synthesized membrane proteins in fibroblasts (Peränen et al., 1996), but Rab8 has only a minor effect on the transport kinetics of these membrane proteins (Ang et al., 2003; Peränen et al., 1996). Instead, ectopically expressed Rab8 has a great impact on the cell shape (Armstrong et al., 1996; Peränen et al., 1996), and cell morphogenesis is affected by a Rab8-specific GDP/GTP exchange factor and the Rab8-interacting protein optineurin (FIP-2) (Hattula and Peränen, 2000; Hattula et al., 2002; Rezaie et al., 2002). Furthermore, a mutant Rab8 causes apoptosis of transgenic Xenopus rods, and depletion of Rab8 from neurons inhibits neurite outgrowth (Huber et al., 1995; Moritz et al., 2001).

Here we have investigated the compartmentalization of Rab8 and its role in regulating cell morphogenesis. By observing Rab8 in living cells we found that it is associated with macropinosomes that transform into Rab8-specific tubular structures that move toward the cell center. A rapid Rab8-dependent recycling of these membranes back to the plasma membrane participates in the formation of new cell surface protrusions. The Rab8-specific membrane transport route contained several markers known to undergo internalization and recycling (β1 integrin, transferrin, transferrin receptor, MHCI), or to participate in membrane recycling or secretion (Arf6, JFC1, Rab11).

Overexpression of wild-type or constitutively active Rab8 (Q67L) results in the formation of cell protrusion, but the effect of the dominant-negative mutant on cell morphogenesis and membrane transport is unknown (Peränen et al., 1996). To understand better the role of Rab8 in mediating cell shape changes we stably transfected HT1080 fibrosarcoma cells with a dominant-negative (T22N) Rab8, a constitutively active (Q67L) Rab8 or their N-terminal green fluorescent protein (GFP) fusions (Fig. 1D). Vector-only (pIRES-neo or pEGFP-C1) cells formed loose, semi-ordered contacts with each other (Fig. 1A) while cells expressing dominant-negative Rab8-T22N were cuboidal, and formed close cell-cell contacts (Fig. 1C). Cells expressing constitutively active Rab8-Q67L were spindle shaped, lost contact inhibition and grew on top of each other (Fig. 1B).

To determine if endogenous Rab8 plays a role in cell morphogenesis, HT1080 cells were transiently transfected with small interfering RNAs (siRNA) directed against a control sequence (Bertling et al., 2004) or against the Rab8 sequence (Schuck et al., 2004). Cell transfected with the Rab8-specific siRNA showed a clear decrease in Rab8 protein content (80-90%) when compared with the control siRNA (Fig. 1K). Cells treated with the control siRNA for 2 days grew in an unordered fashion, independently of cell density, and they were asymmetric (Fig. 1E,G). By contrast, cells depleted of Rab8 made close cell-cell contacts, and had a symmetrical morphology (Fig. 1F,H). We noted that cells depleted of Rab8 were devoid of lamellipodia-like structures, but obtained strong actin stress fibers in the middle of the cell (Fig. 1J). Cells transfected with the control siRNA showed lamellipodia-like structures typical for HT1080 cells, and were devoid of actin stress fibers (Fig. 1I). Taken together, our results show that Rab8 is essential for the maintenance of the normal HT1080 cell phenotype: asymmetric shape, protrusive activity and loss of contact inhibition.

Rab proteins are known to control receptor-mediated endocytosis and recycling (Zerial and McBride, 2001). To test whether Rab8 is linked to these pathways we studied the colocalization of endogenous Rab8 with different Rab proteins expressed as GFP-fusions. There was no colocalization of endogenous Rab8 with GFP-Rab4 and GFP-Rab5 in HT1080 cells (Fig. 2A,B). Rab5-positive vesicles were concentrated in a region behind the leading edge, whereas Rab8-positive vesicles and macropinosomes were found in the outmost region of the leading edge (Fig. 2B). GFP-Rab11 partially colocalized with endogenous Rab8 on small vesicles (Fig. 2). However, Rab11 was not detected on Rab8-specific tubules (data not shown). Ectopically expressed myc-Rab8-Q67L also showed overlap with GFP-Rab11 (Fig. 2J-L), though unlike Rab11 Rab8 was more frequently associated with the plasma membrane and ruffles.

Partial colocalization of Rab8 with Rab11 suggests that Rab8 might affect the trafficking of transferrin (Tfn) and the transferrin receptor (Tfn-R). Thus, we studied the effect of Rab8 depletion on Tfn internalization and recycling. Initially, we confirmed by immunofluorescence microscopy and by western blotting that Rab8 can be depleted from HeLa cells by Rab8-specific siRNA (Fig. 3). Then HeLa cells, transfected with the control siRNA or the Rab8-specific siRNA, were pulsed (30 minutes) with Alexa-labeled Tfn and then chased (30 minutes) in the presence of serum. After the pulse Tfn had reach the pericentriolar region of the control cells, but in Rab8-depleted cells Tfn was found in small vesicles scattered randomly in the cytoplasm (Fig. 3). Even a longer pulse (60 minutes) did not lead to accumulation of Tfn in the pericentriolar region of Rab8-depleted cells (not shown). After the chase period most of the Tfn was externalized both from the control cells and the Rab8-depleted cells (Fig. 3E-H). Thus, Rab8 is not essential for the internalization and recycling of Tfn, but it is needed for delivery to the pericentriolar region. Next we looked at the distribution of Tfn-R in Rab8-depleted cells (supplementary material Fig. S1). In control HeLa cells about 60% of the cells had TFn-R accumulated in the pericentriolar region, and in 30% of the cells Tfn-R colocalized with Rab8 in protrusions (supplementary material Fig. S1A-D). By contrast, the Tfn-R was randomly located in small vesicles throughout the cytoplasm in cells depleted of Rab8 (supplementary material Fig. S1E,F). Furthermore, there was almost no accumulation of Tfn-R in protrusions. This indicates that Rab8 activity is essential for directing Tfn-R to the pericentriolar region and to cell surface protrusions. We could not see clear differences in the plasma membrane localization of Tfn-R on control versus Rab8-depleted cells. However, the Rab8-depleted Hela cells were more symmetric and showed contact inhibition, whereas the control cells were asymmetric and grew on top of each other (supplementary material Fig. S1G,H). Expression of Rab8-Q67L induced accumulation of Tfn-R in the perinuclear region where Tfn-R colocalized with Rab8-Q67L (supplementary material Fig. S3A-C). By contrast, Rab8-T22N showed no clear colocalization with Tfn-R and in these cells Tfn-R had a more peripheral localization (supplementary material Fig. S4D-F). Thus, activation of Rab8 seems to be important for the transport of both Tfn and Tfn-R to the cell center.

The cholera toxin B subunit (CTxB) binds to the glycosphingolipid GM1 through which it is endocytosed via the Golgi complex to the ER (Kirkham et al., 2005). To see whether Rab8 is essential for the transport of CTxB to the Golgi complex we applied Alexa-594 conjugated CTxB to cells that have been treated with either control RNAi or the Rab8-specific RNAi (Fig. 4). After a 30 minute chase period about 80% of the cells treated with the control RNAi showed Golgi-specific staining, while only 15% of the cells treated with Rab8-specific RNAi had CTxB accumulated in the Golgi complex. In the Rab8-depleted cells there was more CTxB on the plasma membrane and in small peripheral vesicles (Fig. 4G). A longer chase period increased the Golgi-accumulation of CTxB in depleted cells indicating that Rab8 does not inhibit the transport but retards it (data not shown). Partial colocalization of CTxB and Rab8 on vesicles and tubules was also seen indicating that CTxB passes through the Rab8 pathway (supplementary material Fig. S3J-L).

We were able to study the localization of Rab8 in different cell lines because our affinity purified anti-Rab8 antibodies were sensitive enough to detect endogenous Rab8. (Fig. 5). HeLa cells contained Rab8-positive tubular structures; the tubular structures were negative for the transferrin receptor (Fig. 5A), with occasional colocalization of Rab8 and the transferrin receptor in the perinuclear region (Hattula et al., 2002). In polarized HT1080 cells we also observed large macropinosomes containing both Rab8 and β1 integrins at the leading edge, and in the tracking tail (Fig. 5B). The tubular structures were not specific for HT1080 cells but were also present in a human neural cell line (Paju) (Fig. 5C), Huh7, green monkey COS-7, and canine MDCK cells. However, Rab8 was not found in fibroblast-like cells, such as BHK-21, NIH3T3 and WI39, where Rab8b was present. This was also confirmed by RT-PCR (data not shown).

Constitutively active Rab8-Q67L (GFP- or myc-tagged) colocalized with β1 integrins on intracellular vesicles and on large tubular structures in transient and stable transfectants as shown for endogenous Rab8 (supplementary material Fig. S2A,B). The tubular structures were approximately 10-30 μm in length, and extended into cell protrusions. Dominant-negative Rab8-T22N (GFP- or myc-tagged) showed reduced colocalization with β1 integrins, instead β1 integrins accumulated in large vacuoles that were surrounded by a reticular structure containing the dominant-negative Rab8-T22N (supplementary material Fig. S2C). Co-expressing the dominant-negative Rab8-T22N with the constitutively active Rab8-Q67L inhibited the formation of the cell surface extensions and tubular structures normally induced by Rab8-Q67L (data not shown). These results indicate that the formation of cell surface extensions is coupled to the activity of a Rab8/β1 integrin-positive tubular membrane compartment.

Rab8 has been regarded as a regulator of exocytosis (Huber et al., 1993). However, our finding that Rab8 is also found on large tubular structures and on macropinosomes raises questions about the actual membrane traffic route of Rab8. To unravel the nature of the Rab8-specific route we followed the transport of GFP-Rab8 or GFP-Rab8b in live cells. HT1080 cells expressing GFP-Rab8-Q67L are highly polarized with a clear leading edge and show strong protrusive activity by formation of filopodia, lamellipodia and ruffles. By time-lapse video microscopy we saw that membrane ruffling at the leading edge is associated with the formation of Rab8-67L positive macropinosomes and vesicles that are transported inwards to the cell center (Fig. 6A; supplementary material Movie 1). The macropinosomes are closely associated with tubular structures, and sometimes smaller vesicles are seen to fuse with these tubules. The tubules are also seen to independently attach to and detach from the plasma membrane. In addition, there was traffic of rapidly moving small vesicles from the cell center toward the leading edge, probably representing recycling vesicles, but due to their small size they were more difficult to observe (Fig. 6A,C; supplementary material Movies 1 and 2). In motile NIH3T3 cell GFP-Rab8b-wt was predominantly seen on membrane ruffles and on dynamic tubular structures that formed at the ruffling area (Fig. 6B; supplementary material Movie 2). The tubules moved mostly to the cell center, although sometimes they also moved in a forward direction into protrusions. A closer look at the leading edge showed an example where Rab8-containing membrane is taken in from a ruffling area and then redirected to a large vacuole from which membrane was transported to the leading edge to promote lamellipodia forward movement (Fig. 6C; supplementary material Movie 3). Taken together, we show that Rab8-mediated protrusive activity is associated with membrane recycling at the leading edge.

About 10% of sparsely seeded HeLa cells contain prominent Rab8-specific tubular structures (Fig. 7K). To see whether tubule formation was dependent on actin dynamics we incubated cells with 0.1 μM cytochalasin D for 30 minutes. We observed a dramatic increase in the number of cells containing tubular structures that stained for endogenous Rab8 (Fig. 7K). Since cytochalasin D depolymerize actin we asked whether RhoA mediated polymerization of actin would have an effect on tubule formation and Rab8 distribution. HeLa cells were transfected with either the dominant-negative GFP-RhoA-T19N or constitutively active GFP-RhoA-G14V (promotes actin stress fiber formation), and then treated with cytochalasin D (Fig. 7A-D). GFP-RhoA-T19N had no effect on the formation of Rab8-specific tubules, whereas RhoA-G14V had a negative effect on the formation of these structures. In GFP-RhoA-G14V cells the tubules were replaced by vesicular structures that were mainly perinuclear. In cells not treated with cytochalasin D GFP and GFP-RhoA-T19N expressing cell showed Rab8 accumulating in cell protrusions, whereas in GFP-RhoA-G14V expressing cells Rab8 was redistributed into a perinuclear region (supplementary material Fig. S4). There was also a slight reduction of the number of cells exhibiting tubules after RhoA-G14V treatment (not shown). Thus, actin polymerization and depolymerization are critical in the biogenesis of Rab8-specific tubules and in the polarized distribution of Rab8 to cell protrusions. To see if microtubules are needed for the formation of Rab8-specific tubules we pre-treated cells with nocodazole for 30 minutes and incubated the cells with both nocodazole and cytochalasin D for an additional 30 minutes. In these cells the formation of tubules was strongly inhibited, and Rab8 re-localized to a perinuclear region (supplementary material Fig. S4G,H). The relocalization of Rab8 to a perinuclear region after both RhoA-G14V expression and microtubule disruption was associated with inhibition of protrusion formation indicating that Rab8 translocation to the cell periphery is coupled to protrusion formation.

To understand how Rab8-specific tubules are formed after addition of cytochalasin D, live cells expressing either GFP-Rab8-wt or GFP-Rab8b-wt were imaged. Non-motile cells lacking endogenous tubules were chosen to better observe the effect of CD. In NIH3T3 cells GFP-Rab8b-wt was first observed on small vesicular structures scattered around the cytoplasm (Fig. 7E; supplementary material Movie 4). After addition of CD, small and larger tubular structures were formed in the periphery of the cell and rapidly transported to a perinuclear region where they fused into a tubular network (Fig. 7F; supplementary material Movie 4). Similar observations were made in HeLa cells expressing GFP-Rab8-wt and treated with CD (Fig. 7G-H, supplementary material Movie 5). In these cells we also looked closer on the region at the plasma membrane and could observe small vesicles budding from the plasma membrane forming into larger tubules or fusing with already existing tubules (Fig. 7I-J, supplementary material Movie 6). This membrane internalization took place in the cell periphery and was associated with retraction of cell protrusions. Interestingly, we also saw rapid transport of vesicles to the cell periphery, where the vesicles entered and fused with forming filopodia, indicating that Rab8 is found on both outgoing and incoming vesicles (Fig. 7I,J; supplementary material Movie 6). In conclusion, our data show that actin and microtubules are essential for the formation of Rab8-specific tubules.

Arf6 is known to regulate clathrin-independent internalization and recycling of membranes through an actin-dependent pathway, and Arf6-positive tubular structures form when cells are treated with cytochalasin D (Brown et al., 2001; Radhakrishna and Donaldson, 1997). Treatment of Arf6-wt transfected HeLa cells with 0.1 μM cytochalasin D lead to colocalization of Arf6-wt with endogenous Rab8 on tubular structures, suggesting that Arf6 and are found on same tubular structures (Fig. 8A,B). Moreover, anti-MHCI antibodies were internalized to Rab8-specific tubules as earlier shown for Arf6 (Brown et al., 2001) (supplementary material Fig. S3G-I). Cells expressing the dominant-negative mutant of Arf6(T27N) also showed prominent formation of Rab8-specific tubules after CD treatment, whereas expression of the constitutively active mutant of Arf6(Q67L) strongly inhibited the formation of Rab8-tubules (Fig. 8C-F). In Arf6-Q67L cells Rab8-tubules were mostly absent or found fragmented as small vesicles. Since expression Rab8-Q67L promotes the formation of long cell protrusions we wanted to see whether co-expression of Arf6 proteins with Rab8-Q67L affects the protrusion formation in HT1080 cells (Fig. 9A,B). We found that Arf6-wt and Rab8-Q67L localized to tubular structures in the leading lamella in cells with prominent cell surface extensions. Arf6-T27N and Rab8-Q67 colocalized on vesicles and tubules, and Arf6-T27N partially inhibited the formation of Rab8-Q67L-induced cell surface extension. Arf6-Q67L promoted accumulation of large vacuoles and rounding up of cells, as described earlier for other cell lines (Brown et al., 2001; Santy, 2002). Arf6-Q67L also inhibited Rab8-Q67L-induced cell surface extensions and redistributed Rab8-Q67L to the perinuclear region, Rab8-T22N's usual location. This suggests Arf6-Q67L inhibits membrane from reaching the Rab8-specific recycling compartment. Taken together, these data indicate that both Arf6 and Rab8 participate in the formation of recycling tubules and cell protrusions.

To find new Rab8-specific effector molecules we used Rab8-wt as bait to screen a human kidney yeast two-hybrid cDNA library (Hattula and Peränen, 2000). We found several identical clones that corresponded to the open reading frame encoding the human JFC1 protein, the human homolog of the mouse synaptotagmin-like protein 1 (Slp1) (Fukuda and Mikoshiba, 2001; MacAdara-Berkowitz et al., 2001). JFC1 specifically interacted with wild-type Rab8 and the active GTP-bound mutant Rab8-67L, but not with the dominant-negative GDP-bound mutant Rab8-22N (Fig. 10A). By contrast, the JFC1 bound to both mutants (T23N and Q78L) of Rab27a, but not to Rab2-Q65L or to laminin (Fig. 10A). Using in vitro binding assays we confirmed the specific interaction of Rab8 to JFC1 (Fig. 10C). In this assay Rab8 bound more efficiently JFC1 than Rab27a, and with higher specificity. Moreover, GST-JFC1 precipitated very efficiently endogenous Rab8, GFP-Rab8-Q67L from cell lysates, but not GFP-Rab8-T22N (Fig. 10D). A co-transfection approach was used to show in vivo interaction of JFC1 to Rab8 (Fig. 10E) (Chen et al., 2002). In the transfected cells JFC1 is produced together with GST, GST-Rab8-T22 or GST-Rab8-Q67L. Precipitation of the extracted proteins with glutathione-Sepharose beads resulted only in precipitation of JFC1 by the GST-Rab8-Q67L protein. Thus, JFC1 shows a clear specificity for Rab8-GTP in vivo. When Hela cells were transfected with constructs encoding GFP-JFC1 or myc-JFC1 we frequently observed JFC1 colocalizing with Rab8 on tubular structures (Fig. 10B). To unravel the possible role of JFC1 in tubule formation we treated GFP and GFP-JFC1 expressing cells with cytochalasin D (supplementary material Fig. S5). We observed that expression of GFP-JFC1, in contrast to GFP, increased the association of Rab8 with vesicles and tubules, and this was associated with a decrease in the number of Rab8-specific tubules, and in an increase in the number of Rab8-specific vesicles and vacuoles (supplementary material Fig. S5).

In free moving cells endogenous Rab8 is associated with cell protrusions but when cells form contacts Rab8 re-localizes to a diffuse perinuclear region (Hattula et al., 2002). The Rab8-specific GEF (Rabin8) associates with actin containing lamellipodia and ruffle-like structures at the plasma membrane indicating that Rab8 activation occurs there (Hattula et al., 2002). However, whether Rab8 is found on outgoing or incoming vesicular structures is not known.

Here we show that Rab8-specific vesicles and macropinosomes are formed at ruffling area behind the leading edge, from where they move centripetally toward the cell body. The macropinosomes and vesicles fuse to or form tubular structures that are transported back to the plasma membrane to form new cell protrusions. Inhibiting this membrane traffic route, by Rab8-depletion or dominant-negative mutants of Rab8, abolishes protrusive activity by enhancing actin stress fiber formation and cell-cell adhesion. This indicates that Rab8 regulates the formation of dynamic actin containing cell surface structures. The inverse is also true, because actin modulates the formation of Rab8-specific membrane organelles. We show that actin depolymerization by cytochalasin D promotes the formation of Rab8-specific vesicles and tubules in the cell periphery. The tubules accumulated in the cell center and membrane recycling is inhibited, leading to cell retraction. Removal of cytochalasin D restores protrusions and the distribution of Rab8 to the cell periphery (J.P., unpublished). However, we also show that tubule formation is microtubule-dependent, because if microtubules are disrupted before and during treatment with cytochalasin D no Rab8-specific tubules are formed. Thus, microtubules are needed for the formation of the tubules, while actin depolymerization probably initiates the process of tubulation.

The small GTPase Arf6 is also found on tubular structures after actin depolymerization with cytochalasin D, and behaves in many respects like Rab8 (Radhakrishna and Donaldson, 1997; Prignet et al., 2003) (this study). Both Arf6 and Rab8 induce the formation of protrusions (Peränen et al., 1996; Radhakrishna and Donaldson, 1997). It is known that overexpression of Arf6-Q67L results in internalization and trapping of actin-coated vacuoles that are unable to recycle back to the cell surface (Brown et al., 2001). This leads with time to inhibition of both ruffling, endocytosis and cell polarity (Brown et al., 2001; Santy, 2002; Hashimoto et al., 2004). Here we show that Arf6-Q67L also inhibits the formation of Rab8-tubules, and decreases the formation of Rab8-induced cell protrusions. One explanation for this inhibitory action of Arf6 on Rab8 is that Arf6-Q67L acts upstream of Rab8 by inhibiting membrane from reaching the Rab8 recycling compartment. How Arf6 and Rab8 cross-talk is currently under investigation.

It is well established that polarized exocytosis of newly synthesized membrane proteins occurs towards the leading edge in migrating cells (Bergman et al., 1983). In addition, directional recycling of Tfn-R and integrins to the leading edge has been reported (Hopkins et al., 1994; Laukaitis et al., 2001). Rab8 may be a major player in these processes, because activation of Rab8 induces polarized membrane transport of newly synthesized proteins to cell extensions, and Rab8 depletion inhibits delivery of Tfn-R to cell protrusions (Peränen et al., 1996) (this study). Activation of Rab8 also missorts VSV-G to the apical surface domain in MDCK cells, while the dominant-negative mutant of Rab8 has no effect on sorting (Ang et al., 2003). Interestingly, the dominant-negative mutant of Rab8 and Rab8-depletion enhance cell-cell adhesion suggesting that Rab8 activity is perhaps not important in maintaining proper sorting in fully polarized epithelial cells. We have shown that Rab8 is linked to huntingtin via optineurin (Hattula and Peränen, 2000). Sahlender et al. (Sahlender et al., 2005) have demonstrated that optineurin links myosin VI to the Golgi and is involved in Golgi organization and exocytosis. They also found that Rab8 recruits myosin VI onto Rab8-positive vesicles and tubules. Myosin VI is not only found in the Golgi region but also on ruffles, endocytic and secretory vesicles as shown here for Rab8 (Buss et al., 2002). Moreover, Rab8 is essential for the targeting of AMPAR to the spine surface and receptor recycling at the postsynaptic terminal (Gerges et al., 2004). Thus, Rab8 seems to play an important role in targeting molecules to specific cell surface domains.

JFC1 is a member of the synaptotagmin-like protein (Slp) family that contains a Rab-binding domain (RBD) in the amino-terminal and two tandem-C2 domains in the carboxy-terminus. It interacts with Akt, Rab27a and the NADPH oxidase (Johnson et al., 2005a; Johnson et al., 2005b; MacAdara et al., 2001). JFC1 is known to regulate androgen-dependent secretion of prostate-specific antigen and prostatic-specific acid phosphatase (Johnson et al., 2005b). Here, we show that the JFC1 protein not only interacts with Rab27a but also with Rab8 in a nucleotide specific manner suggesting that these Rab proteins may functionally overlap. This is also supported by the fact that Rab8 and Rab27 colocalize on dense-core vesicles and on melanosomes, and they both participate in the actin-dependent movement of the melanosomes (Chabrillat et al., 2005; Fukuda et al., 2002). Furthermore another Slp (Slp4-a) is also known to bind both Rab8 and Rab27a (Fukuda, 2003). We show that JFC1 colocalizes on tubular and vesicle structures containing endogenous Rab8 suggesting that JFC1 is linked to the Rab8-specific pathway. Moreover, ectopically expressed JFC1 seems to increase the association of Rab8 with tubular and vesicular structures after actin disruption, and this is associated with a decrease in the number of tubules. At the moment we do not know whether overexpression of JFC1 inhibits the transformation of incoming vesicles into tubules or promotes disassembly of existing tubules.

Rab8 also modulates the traffic of incoming cargo molecules. Rab8 partially colocalizes with Rab11 suggesting that it is connected to the endocytic recycling compartment (ERC). However, they seem to control different steps of recycling. Rab11 regulates Tfn recyling from the ERC to the PM, whereas Rab8 controls the transport of Tfn/Tfn-R to the ERC (Ullrich et al., 1996). In Rab8 depleted cells Tfn are seen in small vesicles, probably representing sorting endosomes (SEs). However, Rab8 depletion does not significantly inhibit Tfn recycling, which probably occurs from SEs by a Rab4-dependent pathway. Overexpression of the myosin-Vc tail leads to colocalization with the Tfn-R and Rab8, and also perturbs Tfn trafficking (Rodriguez and Cheney, 2002). Thus, the role of Rab8 might be indirect, perhaps by regulating cytoskeleton-based movement of vesicular structures (Chabrillat et al., 2005) (this study). This is supported by the fact that Rab8 depletion promotes the formation of actin stress fibers, whereas activation of Rab8 has the opposite effect. Rab8 also regulates microtubule dynamics, and its localization is dependent on the integrity of microtubules (Peränen et al., 1996) (this study). Thus, Rab8 might play a role in the organization of the ERC and in facilitating polarized vesicle movement along microtubules to cell surface domains.

Cholera toxin B (CTxB) uses different modes to enter the cell (Shogomori and Futerman, 2001; Massol et al., 2004; Kirkham et al., 2005). We show here that Rab8 depletion inhibits the retrograde transport of CTxB to the Golgi compartment, suggesting that Rab8 functions in a step between the plasma membrane and the Golgi. Arf6, Rab22a and dynamin mutants also inhibit the transport of CTxB to Golgi. (Kirkham et al., 2005; Mesa et al., 2005). The fact that Rab8 colocalizes with Arf6, Rab22b and to some extent with Rab22a indicates that these proteins define a common clathrin-independent pathway through which CTxB is transported (Rodriguez-Gabin et al., 2001) (this study; J.P., unpublished).

How does Rab8 control cell shape? Our results support a model where Rab8 activity is needed to rapidly take in membrane from the cell surface (ruffles) regions that no longer support adhesion. This membrane that contains structural (adhesion receptors) and regulatory (Arf6, Rho and Ras) components is then recycled back to the plasma membrane via a recycling compartment for the formation of new protrusions (Radhakrishna and Donaldson, 1997; Ng et al., 1999; Furuhjelm and Peränen, 2003; Deretic et al., 2004; Powelka et al., 2004; Schlunck et al., 2004). According to this, a shift from exocytosis to endocytosis (macropinocytosis) of membranes would stop protrusive activity and result in retraction of cell surface domains (tracking tail). Thus an appropriate equilibrium between exocytosis and endocytic recycling could be the driving force in the formation and the destruction of cell surface domains. A cessation of this membrane-recycling pathway would promote plasma membrane stabilization and cell-cell adhesion. This process is likely to be evolutionary well preserved because a dominant-negative mutant of the Rab8-like protein of Dictyostelium discoideum, Sas1, also promotes cell-cell adhesion (Powell and Temesvari, 2004). Our hypothesis is in accordance with the role of Rab8 in neurite outgrowth and development of photoreceptors (Huber et al., 1995; Kametani et al., 2003; Moritz et al., 2001; Deretic et al., 2004). Future studies should unravel the role of Rab8 in processes like cell polarity, cell migration, epithelial-mesenchymal transition, and cell division.

The pEGFP-Rab8-wt, pEGFP-Rab8-T22N, pEGFP-Rab8-Q67L, pEGFP-Rab8b-wt, pEGFP-Rab8b-T22N, Rab8b-Q67L, pEGFP-RhoA-wt, pEGFP-RhoA-T19N and pEGFP-RhoA-G14V constructs have been previously described (Hattula et al., 2002; Peränen and Furuhjelm, 2000; Vartiainen et al., 2000). pEGFP-Rab8-N121I was constructed as described for the other Rab8 mutants (Peränen and Furuhjelm, 2000). The open reading frames of Rab8-T22N and Rab8-Q67L were also cloned into pcDNA4/TO (Invitrogen) and pIRES (Clontech). The GST-tagged fusions of Rab8 were obtained by cloning Rab8-T22N and Rab8-Q67L into the pEBG-SrfI vector (Chen et al., 2002). The human Rab4, Rab5, Rab11 coding regions were amplified from human kidney cDNA by specific primers, and cloned into pGEM-1 (Promega) and pEGFP-C1 vectors (Clontech). The human Rab27a was similarly cloned into pGEM-1, and site-directed mutagenized to create pGEM-Rab27a-T23N and pGEM-Rab27a-Q78L. Arf6-wt was amplified by PCR from human muscle cDNA and site-directed mutagenized to obtain Arf6-T27N and Arf6-Q67L. The Arf6 genes were cloned into pGEM-1, pcDNA4/TO, and pEGFP-N1 (Clontech). The open reading frame of JFC1 was cloned into pEGFP-C1, pcDNA4/TO, and pGEX-2T (Amersham). All amplified genes were sequenced to confirm that no mutations had been introduced during PCR. Details on the constructs are available upon request.

The Rab8-wt gene was used as bait to screen a human kideney cDNA yeast two-hybris library for possible Rab8 effectors as previously described (Hattula and Peränen, 2000). The full-length open reading frame of JFC1 was cloned into pB42AD (Clontech) and was tested against Rab8, Rab2, Rab27a and their corresponding mutants that had been cloned into pGilda (Clontech).

GST-JFC1 was expressed in E. coli at 15°C, purified on glutathione beads and immunized into rabbits as described previously (Peränen and Furuhjelm, 2000). Affinity purification of antibodies was done by using nitrocellulose strips containing GST-JFC1 (Peränen et al., 1996).

HT1080 fibrosarcoma cells were grown and maintained in MEM-medium supplemented with 10% (v/v) fetal bovine serum (Gibco BRL). To obtain stable cell lines, HT1080 cells were grown on 100 mm plates and transfected with 10 μg of pIRES (control), pIRES-Rab8-T22N, pIRES-Rab8-Q67L, pEGFP (control), pEGFP-Rab8-T22N or pEGFP-Rab8-Q67L plus 40 μl each of Lipofectamine in Optimem™ (Gibco BRL). Cells were split the next day and plated at different dilutions in medium containing 700 μg ml^-1 G418. Clones were collected 12 days later and the presence of the Rab8 proteins was immunologically confirmed. The cell morphology of stable cell lines was monitored at low magnification (25×) after fixing the cells with 4% paraformaldehyde and staining with aqueous 1% crystal violet. Stable NIH3T3 transfectants expressing GFP-Rab8b-wt were obtained similarly. The growth conditions for the HeLa and Paju cells were as described earlier (Peränen and Furuhjelm, 2000; Zhang et al., 1996). To study the influence of Rab8 on cell morphogenesis we transfected HT1080 cells with Rab8 constructs alone or together with different Arf6 constructs in a molar ratio of 1:1. Fugene 6 (Boehringer Mannheim) was used as transfection reagent, and all cells were studied 20 hours after transfection. Cells containing neurite-like surface extensions longer than one cell body were considered extension-positive. Quantifications were done by examining a total of 50 cells from three independent experiments, and data were mean ± s.d.

The target sequence used for synthesizing Rab8-specific RNAi was GACAAGTTTCCAAGGAACG (Schuck et al., 2004), and the sequence for the corresponding control RNAi was AGCTTCATAAGGCGCATGC (Bertling et al., 2004). Hela cells were seeded in 35 mm plates 1 day prior to transfection. Semi-confluent cells were transfected using the magnet assisted transfection kit (MATra, IBA, Göttingen, Germany). In brief, 60 pmol of siRNA synthesized by Dharmacon (Lafayette, CO) was mixed into 500 μl of Optimem™. Then 1 μl of MATra beads were added to the 500 μl, mixed, and incubated at room temperature for 30 minutes. Next, 500 μl of growth medium from the cell plate was removed and replaced with the MATra/siRNA solution, and mixed. The cell plates with the transfection mix were placed on a magnetic plate for 15 minutes at 37°C in a cell incubator. After 4 hours in the incubator the cells obtained fresh growth medium, and were grown for 2 days before assaying Rab8 depletion.

The cholera toxin uptake was performed essentially as described by Hölttä-Vuori et al. (Hölttä-Vuori et al., 2000). In brief, cells that had been treated for 2 days by either control RNAi or Rab8-specific RNAi (see above) were incubated with 1 μg ml^-1 Alexa 594-conjugated cholera toxin (CTxB; Molecular Probes) in serum-free culture medium (MEM) for 30 minutes on ice. After washing three times with serum-free medium supplemented with 0.02% BSA, the cell were incubated in the same medium for 30 minutes at 37°C and fixed. To quantify the fraction of CTxB not associated with Golgi, the fraction of cells where CTxB colocalized with the Golgi marker anti-GM130 (BD transduction), stained by anti-mouse IgG allophycocyanin (Molecular Probes), were subtracted and the result expressed as a percentage of total number of cells.

Two days after Rab8-depletion cells grown on coverslips in 35 mm plates were washed in serum-free DMEM supplemented with 0.2% BSA and preincubated for 2 hours at 37°C in serum-free medium. The cells were then incubated with 10 μg ml^-1 alexa488-transferrin in serum-free medium (Molecular Probes) for 30 minutes at 37°C (pulse). The cells were then washed twice with DMEM supplemented with 10% FCS, and then incubated in the same medium for 30 minutes at 37°C (chase). The cells were rinsed for 45 seconds with 0.5% acetic acid, 0.5 M NaCl pH 3.0 to remove Tfn from the cell surface, washed three times with PBS, fixed with paraformaldehyde and processed for immunofluorescence as described previously (Peränen et al., 1996). Incubation of HeLa cells with anti-MHCI antibodies (W6/32; Novocastra) were as previously described (Naslavsky et al., 2004).

Lysates from stable cell lines and Rab8-depleted cells were run on SDS-PAGE and transferred to nitrocellulose as described (Peränen et al., 1996). Rab8 was detected by a monoclonal anti-Rab8 antibody (code 66320, BD transduction) or an affinity purified anti-Rab8 (Peränen and Furuhjelm, 2000). GST-fusions were detected by anti-GST (Cell Signalling Technology). The secondary antibodies were goat anti-mouse HRP or anti-rabbit HRP (Jackson), and detection was by ECL (Amersham).

Recombinant GST-JFC1 was expressed in Escherichia coli strain BL21(DE3), and purified onto gluathionine Sepharose® beads (Amersham) according to the manufacturer (Amersham). Rab2, Rab8-wt. Rab8-T22N, Rab8-Q67L, Rab27a-T23N and Rab27a-Q78L were translated in vitro from corresponding plasmids using a TNT Quick kit (Promega) according to the manufacturer's instructions. The in vitro translation products were then incubated with GST-JFC1 attached to glutathione-Sepharose® beads as previously described (Hattula and Peränen, 2000). In vivo produced GFP-Rab8 was also used for binding assays. Briefly, Hela cells expressing GFP-Rab8-T22N or GFP-Rab8-Q67L were lysed in binding buffer (50 mM Tris pH 7.5, 150 mM NaCl, 2 mM MgCl₂, 1% Triton X-100). The lysate was cleared by centrifugation (15,000 g, 15 minutes, +4°C). The cleared lysate was incubated with GST-JFC1 beads for 2 hours at 4°C. The beads were washed four times with the incubation buffer during 30 minutes. Bound material was eluted from the beads with Laemmli sample buffer and loaded onto a 12% SDS-polyacrylamide gel. As a control, 1/5 of the input in vitro material was also loaded onto the gel. Bound and input material was assayed by western blot using anti-Rab8 antibody (code 66320, BD transduction) as described above.

In vivo binding was done by co-transfection of pcDNA-mycJFC1 with pEBG-SrfI, pEBG-Rab8-T22N or pEBG-Rab8-Q67L in HEK293 cells using Fugene 6 according to the manufacturer (Roche). One day latter the cells from a 5 cm plates were homogenized in a buffer containing 30 mM Tris-HCl pH 7.5, 150 mM NaCl, 1.0% Triton-X 100, 0.5 mM GTPγS and 0.4 mM PMSF and solubilized for 20 minutes at 4°C. Insoluble material was removed by centrifugation at 15,000 g for 15 minutes. The obtained supernatant was incubated with glutathione-Sepharose beads (30 μl beads) at 4°C for 2 hours. The beads were washed four times with 30 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.2% Triton-X 100, and 0.4 mM PMSF. The bound proteins were eluted by Laemmli sample buffer, and subjected to 12% SDS-PAGE, followed by immunoblotting with anti-JFC1 antibodies as described above. To detect GST, GST-Rab8-T22N and GST-Rab8-Q67L the blot was stripped and reprobed with anti-GST (Cell Signaling).

Cells were grown on coverslips, fixed with 4% paraformaldehyde, permeabilized with 0.1% Triton X-100 and processed as described (Peränen et al., 1996). The anti-Rab8 rabbit antiserum was affinity purified as previously described (Peränen et al., 1996; Peränen and Furuhjelm, 2000). Other antibodies used were anti-Arf6 (6ARF01; Neomarkers), anti-β1-integrin (clone P4C10; Gibco BRL), anti-transferrin receptor (Roche), anti-myc (Roche), anti-γ-catenin (TransLab). The secondary antibodies were goat anti-rabbit IgG lissamine rhodamine, goat anti-mouse IgG lissamine rhodamine and goat anti-mouse IgG fluorescein isothiocyanate (FITC) from Jackson ImmunoResearch. TRITC-conjugated phalloidin (Molecular Probes) was used to visualize actin. Fluorescence of fixed cells was observed either with the Bio-Rad MRC-1024 confocal system linked to a Zeiss Axiovert 135 M microscope or with an Olympus AX70 fluorescence microscope.

Living HeLa, HT1080 and NIH3T3 cells expressing different GFP-Rab8/Rab8b proteins were observed with an inverted IX71 Olympus microscope (Olympus, Tokyo, Japan) through a 40 β1.5N objective, equipped with a Polychrome IV monochromator (TILL Photonics, Eugene, OR) with appropriate filters. Images were recorded in intervals of 15 or 30 seconds and processed using TILL Imaging system software.

We thank P. Auvinen, S. Falck, and P. Lappalainen and for comments on the manuscript. We also thank L. C. Andersson for the Paju cells. The work was supported by the Academy of Finland, Helsinki University Foundation, Oskar Öflunds Foundation, Jenny and Antti Wihuri foundation and by the Research and Science Foundation of Farmos.