SNARE [soluble N-ethylmaleimide-sensitive factor (NSF) attachment protein (SNAP) receptor] proteins control the membrane-fusion events of eukaryotic membrane-trafficking pathways. Specific vesicular and target SNAREs operate in specific trafficking routes, but the degree of specificity of SNARE functions is still elusive. Apical fusion requires the polarized distribution at the apical surface of the t-SNARE syntaxin 3, and several v-SNAREs including TI-VAMP and VAMP8 operate at the apical plasma membrane in polarized epithelial cells. It is not known, however, whether specific v-SNAREs are involved in direct and indirect routes to the apical surface. Here, we used RNAi to assess the role of two tetanus-neurotoxin-insensitive v-SNAREs, TI-VAMP/VAMP7 and VAMP8, in the sorting of raft- and non-raft-associated apical markers that follow either a direct or a transcytotic delivery, respectively, in FRT or Caco2 cells. We show that TI-VAMP mediates the direct apical delivery of both raft- and non-raft-associated proteins. By contrast, sorting by means of the transcytotic pathway is not affected by TI-VAMP knockdown but does appear to be regulated by VAMP8. Together with the specific role of VAMP3 in basolateral transport, our results demonstrate a high degree of specificity in v-SNARE function in polarized cells.
The plasma membrane of polarized epithelial cells is divided into two domains, apical and basolateral, differing in their protein and lipid compositions and specialized functions (Mostov, 2003; Rodriguez-Boulan and Powell, 1992). Intracellular sorting occurs at the level of the trans-Golgi network (TGN) and endosomes where proteins are segregated into distinct vesicles upon recognition of specific apical or basolateral sorting signals (Folsch, 2005; Matter and Mellman, 1994; Mostov et al., 2000). The proteins are delivered to the apical or basolateral surface (Rodriguez-Boulan and Powell, 1992; Keller et al., 2001; Kreitzer et al., 2003; Wandinger-Ness et al., 1990) by means of a direct or an indirect (transcytotic) route that passes first through the opposite membrane domain. The use of the direct or transcytotic pathways seems to be both cell and protein specific (Rodriguez-Boulan et al., 2005). Although direct and indirect sorting signals are quite well defined for transmembrane (TM) basolateral proteins, this is not the case for apical proteins, and in particular for GPI-anchored proteins (GPI-APs). These proteins are anchored to membranes by a posttranslational lipid modification, the glycosylphosphatidylinositol (GPI) anchor (Bangs et al., 1985; Ferguson et al., 1985; Ikezawa, 1963; McConnell et al., 1981). A relevant feature of GPI-APs is their association with lipid rafts (Brown, 1994; Brown and Waneck, 1992; Brown, 1992; Brown and Rose, 1992). GPI-APs are apically sorted in several epithelial cell lines (Brown et al., 1989; Lisanti et al., 1989) and use their GPI anchor to associate with rafts (Brown and Waneck, 1992; Brown, 1992). It has therefore been proposed that the GPI anchor acts as an apical sorting determinant by mediating raft association (Simons and Ikonen, 1997; Simons and van Meer, 1988).
The direct sorting of GPI-APs to the apical surface of Madin-Darby canine kidney (MDCK) cells (Hua et al., 2006; Paladino et al., 2006) requires their oligomerization into detergent-resistant membrane domains (DRMs) at the Golgi level (Paladino et al., 2004). This reinforces the hypothesis that apical sorting of GPI-APs occurs intracellularly before arrival at the plasma membrane (Hua et al., 2006; Paladino et al., 2006). However, not much is known about the molecular machinery involved in the direct targeting of GPI-APs to the apical surface. Various components of the exocytic machinery that transport apical and basolateral proteins have been identified, but the mechanisms that regulate their localization and function have been challenged recently (reviewed in Rodriguez-Boulan et al., 2005). Polarized protein trafficking to the apical and basolateral plasma membranes requires different sets of SNARES, a family of proteins specifically involved in the fusion of vesicles with their target membranes (Lafont et al., 1999; Low et al., 1996; Low et al., 1998; Steegmaier et al., 2000; Weimbs et al., 2003). SNAREs on vesicular cargo are called v-SNAREs, and those on target membranes t-SNAREs (Sollner et al., 1992). The formation of a SNARE complex between v-SNAREs and t-SNAREs (Schiavo et al., 1997; Sollner et al., 1993) mediates the specific recognition and subsequent fusion (McNew et al., 2000) of vesicles with their appropriate target membrane. SNAREs are assisted by several partners and regulators including the small GTPase Rab proteins (Grosshans et al., 2006; Novick and Zerial, 1997; Zerial and McBride, 2001).
Polarized epithelial cells represent an interesting model to study the specific function of SNAREs because plasma membrane t-SNAREs are differentially localized in these cells. The t-SNARE syntaxin 3 (Stx3) is at the apical plasma membrane, whereas syntaxin 4 (Stx4) is expressed predominantly at the basolateral membrane domain of MDCK cells (Low et al., 1996). By contrast, SNAP23 (ubiquitously expressed homologue of SNAP25) and syntaxin 2 (Stx2) are present in both membrane domains (Low et al., 1998). Stx3 and SNAP23 constitute the apical t-SNARE complex, and they interact with the v-SNARE tetanus-neurotoxin-insensitive vesicle-associated membrane protein (TI-VAMP) in Caco2 cells (Galli et al., 1998). The strong interaction between TI-VAMP, Stx3 and SNAP23 has been further demonstrated by yeast two-hybrid assays (Martinez-Arca et al., 2003). In MDCK cells, overexpression of Stx3 inhibits biosynthetic transport from the TGN to the apical membrane and the endocytic recycling pathway from apical endosomes of a mutant form of the polymeric immunoglobulin receptor (pIgR), signal-less pIgR (SL-pIgR) (Casanova et al., 1991), but not the basolateral delivery of the wild-type form of pIgR (WT-pIgR) (Low et al., 1998). Furthermore, overexpression of Stx3 strongly inhibits the apical delivery of TM sucrase isomaltase (SI) and the secreted protein α-glucosidase without any effect on basolateral delivery in Caco2 cells (Breuza et al., 2000). Furthermore, inhibition of TI-VAMP with specific antibodies affects apical delivery of haemagglutinin (HA) but has no effect on the basolateral route (Lafont et al., 1999). In addition, Stx3 and TI-VAMP associate with DRMs in post-TGN apical carriers (Lafont et al., 1999). These data have demonstrated that Stx3 and TI-VAMP are important for apical transport of transmembrane and secretory proteins both in MDCK and Caco2 cells. However, apical sorting shows further levels of complexity: (1) there are at least two pathways for apical delivery, one raft dependent and one raft independent (Benting et al., 1999; Lipardi et al., 2000); (2) there are at least two routes followed by apical proteins, direct and transcytotic (Gilbert et al., 1991; Rodriguez-Boulan et al., 2005; Zurzolo et al., 1992a); (3) it is not clear whether apical sorting of proteins following different pathways occurs at the same intracellular site and (4) utilizes the same machinery.
Although the results mentioned above on the apical SNAREs indicate a clear involvement of TI-VAMP in the direct sorting of transmembrane and secreted apical proteins, it is not clear whether TI-VAMP is also involved in the direct sorting of GPI-APs and whether it participates in the transcytotic pathway, for which the specific v-SNARE has not yet been identified. Therefore, we first asked whether TI-VAMP is involved in the apical sorting of GPI-APs. Then, in order to discriminate between the direct and transcytotic pathways, we used two different cell lines: Fisher rat thyroid cells (FRT) and human colorectal cancer cells (Caco2) that, respectively, use predominantly the direct and the transcytotic pathway to deliver apical proteins to the plasma membrane (Le Bivic et al., 1990; Matter et al., 1990; Zurzolo et al., 1992a). By using an RNA interference (RNAi) approach targeting TI-VAMP in both cell lines, we found that TI-VAMP is necessary for the correct localization at the apical membrane of both GPI-APs and TM proteins that use a direct route to the apical membrane independently of their sorting mode (i.e., raft dependent or raft independent). Furthermore, TI-VAMP is not involved in apical sorting of proteins that use a transcytotic pathway. Instead, the apical transcytotic pathway is regulated by VAMP8, another tetanus-neurotoxin-insensitive v-SNARE, which can pair with both apical and basolateral t-SNAREs (Imai et al., 2003; Pombo et al., 2003; Wang et al., 2007). Thus, we demonstrate that at least two different VAMPs are involved in the regulation of the two alternative routes to the apical plasma membrane. This mechanism appears to be independent of the cargo protein per se but dependent on the pathway that the protein is following in the different epithelia.
TI-VAMP is necessary for the correct sorting of both GPI-AP and TM apical proteins in fully polarized FRT cells
We set up a specific RNAi approach to study the role of TI-VAMP in the apical sorting of a model GPI-AP, placental alkaline phosphatase (PLAP), in two different epithelial cell lines, FRT and Caco2 cells, that respectively use predominantly a direct or an indirect pathway to target proteins to the apical membrane. Several isoforms of TI-VAMP have been described (Martinez-Arca et al., 2003), but only one isoform of TI-VAMP is present in FRT cells [sequence NM_021659 (Wang et al., 2005)].
To decrease endogenous levels of TI-VAMP in FRT cells, we used a specific small interfering RNA (siRNA) that has been described previously (Alberts et al., 2003). This siRNA and an unrelated siRNA targeting human β-globin, used as a negative control, were transiently transfected into FRT cells stably expressing PLAP (FRT-PLAP) (Lipardi et al., 2000). TI-VAMP expression was assayed by western blotting of total cell lysates from cells grown on filters for 4 days in conditions favoring polarization (Fig. 1A). Compared with control cells, siRNA-treated FRT cells showed a strong decrease of the western blot signal corresponding to TI-VAMP [∼88%; normalization was performed in comparison with calreticulin; (Fig. 1A)], thus indicating an efficient silencing of the expression of TI-VAMP in polarized FRT cells.
In order to analyse whether TI-VAMP has a role in apical sorting of TM proteins in FRT cells, we investigated the effect of TI-VAMP transient knockdown on the localization of TM apical and basolateral markers endogenously expressed by these cells, respectively dipeptidyl peptidase IV (DPPIV) and antigen 35/40 kDa (Ag 35/40 kDa) (Zurzolo et al., 1991; Zurzolo et al., 1992a; Zurzolo et al., 1993; Zurzolo et al., 1992b). Control cells (FRT-PLAP transfected with siRNA targeting β-globin) and FRT-PLAP cells transfected with siRNA targeting TI-VAMP were grown on filters for 4 days and subjected to immunofluorescence using specific antibodies against the two TM proteins [A. Quaroni, personal gift (Zurzolo et al., 1992a)] (Fig. 1B). As expected, in control cells, DPPIV was localized at the apical membrane, whereas Ag 35/40 kDa was localized at the basolateral membrane (Fig. 1B) (Zurzolo et al., 1992a). By contrast, in cells transfected with TI-VAMP siRNA, DPPIV was mis-sorted to the lateral membrane, whereas the basolateral localization of Ag 35/40 kDa was unaffected (Fig. 1B). These results obtained by gene silencing in FRT cells confirm the specific involvement of TI-VAMP in apical sorting of apical TM proteins that was previously demonstrated by antibody inhibition in MDCK cells (Lafont et al., 1999) and suggest that the role of TI-VAMP is conserved across different epithelial cells. Next, we investigated the effect of TI-VAMP transient knockdown on the localization of PLAP, our model apical GPI-AP (Lipardi et al., 2000) (Fig. 1C). Compared with the control condition where PLAP accumulated at the apical membrane, in cells transfected with siRNA targeting TI-VAMP it was also mislocalized to the lateral membrane (Fig. 1C). These results indicate that TI-VAMP is necessary also for the correct apical localization of GPI-APs.
The above result is particularly noteworthy because PLAP and DPPIV, although both following a direct route, are sorted to the apical membrane by means of two different mechanisms, respectively raft dependent and non-raft dependent (Benting et al., 1999; Lipardi et al., 2000). In order to quantify and better characterize the role of TI-VAMP, we produced stable TI-VAMP knockdown FRT clones (see Materials and Methods). Selected stable knockdown clones (FRT si TI-VAMP::PLAP), which by western blot had a markedly decreased signal corresponding to TI-VAMP while expressing PLAP at a level similar to that of control FRT PLAP cells (Fig. 2A), were analyzed for the localization of our apical and basolateral endogenous TM markers by indirect immunofluorescence and confocal analysis, as shown in Fig. 2B. As expected, in FRT si TI-VAMP::PLAP cells, DPPIV lost its predominant apical localization and was mislocalized to the lateral membrane, whereas the basolateral localization of Ag 35/40 kDa was unchanged (Fig. 2B). Next, we examined the localization of PLAP both by indirect immunofluorescence and confocal analysis (Fig. 3A) and by surface biotinylation (Fig. 3B). Similar to the transient knockdown, PLAP was mislocalized to the lateral membrane of stable FRT knockdown cells (compare Fig. 3A with Fig. 1C). By selective surface biotinylation, we calculated that ∼80% of surface-expressed PLAP was localized at the apical membrane in control cells, whereas PLAP was equally distributed to the apical and basolateral membrane (∼50% apical and ∼50% basolateral) in FRT si TI-VAMP::PLAP cells (Fig. 3B).
Thus, both the transient and stable knockdowns showed that the apical v-SNARE TI-VAMP is necessary for the correct apical localization of DPPIV and PLAP, two proteins directly sorted to the apical membrane of FRT cells through two different mechanisms [respectively raft independent and raft dependent (Lipardi et al., 2000)].
Loss of TI-VAMP does not impair the development of a polarized monolayer
Because of the effect of TI-VAMP knockdown on both raft- and non-raft-mediated apical sorting, we asked whether the mistargeting of PLAP and DPPIV could just be a consequence of the loss of the cells' overall ability to acquire a polarized phenotype. Because the kinetics of formation of tight junctions has been frequently used as a measure of the ability of epithelial cells to polarize, we assessed whether transient or stable knockdown of TI-VAMP affected the `normal' development of transepithelial resistance (TER), which measures the ability of a monolayer to impede ion flow – this measure should increase with the establishment of a polarized monolayer (Fig. 4A). The TER of FRT PLAP cells transiently transfected with siRNA targeting β-globin versus FRT PLAP cells transiently transfected with siRNA targeting TI-VAMP was monitored during a 4-day period (Fig. 4A). Both curves start at a TER between 200 and 400 Ω cm–2 and reach after 4 days 1400 Ω cm–2 (the TER of an empty filter is represented by the yellow curve and reaches 75 Ω cm–2). The same observation was made for FRT-PLAP cells versus the stable knockdown clone FRT si TI-VAMP::PLAP cells (Fig. 4A). The higher TER of FRT si TI-VAMP::PLAP cells compared with FRT-PLAP suggests a clonal selection effect because both curves exhibit the same slope, indicating that the two cell lines behave similarly on filters during the establishment of the polarized epithelial phenotype (Fig. 4A). In order to confirm that the monolayer was well established, we also analysed by immunofluorescence the distribution of ZO-1, a major protein of the tight junctional complex (Fig. 4B). Both siRNA-treated cells and control cells grown on filters in conditions favoring polarization for 4 days exhibit the `chicken wire'-like immunostaining characteristic of the tight junction protein ZO-1 (Fig. 4B), confirming that the knockdown of TI-VAMP in FRT cells does not induce any significant alteration in the tight junctional complex.
Overall, these results demonstrate that the mislocalizations of apical non-raft and raft associated proteins in cells transiently and stably knockdown for TI-VAMP are not due to a general perturbation of the development of the epithelium. They also indicate that TI-VAMP dependent apical pathways are not involved in the events necessary for tight junction formation during the development of a polarized monolayer.
TI-VAMP is also involved in the direct apical sorting of PLAP but not in the transcytotic delivery of DPPIV in Caco2 cells
In contrast to FRT and MDCK cells, which use mainly a direct pathway to sort both apical and basolateral proteins to their respective membrane domains after their intracellular sorting, other epithelial cell types can use different pathways (Rodriguez-Boulan et al., 2005). A study in intestine has shown that certain apical plasma membrane proteins follow an indirect pathway to the cell surface that passes first through the basolateral domain (Hauri et al., 1979). These results were confirmed in intestinal cell lines (e.g. Caco2 cells) cultured on filters under conditions favoring polarization (Le Bivic et al., 1990; Matter et al., 1990).
Because TI-VAMP is also present at the apical membrane of Caco2 cells (Galli et al., 1998), this appears to be a very good model to analyse its role both in the direct and indirect apical pathways. To this end, we followed the sorting of DPPIV and PLAP, which are both endogenously expressed by Caco2 cells and are sorted, respectively, by means of a direct (PLAP) and a transcytotic (DPPIV) route to the apical plasma membrane under knockdown conditions for TI-VAMP.
To decrease the endogenous levels of TI-VAMP in Caco2 cells, we designed a siRNA targeting human TI-VAMP (siRNA 8), which was transiently transfected in Caco2 cells. TI-VAMP expression levels were measured after 4 days of culture on filters by western blotting on total cell lysates in a manner similar to that described for FRT cells in Fig. 1. By using a specific mouse antibody against human TI-VAMP (Muzerelle et al., 2003) in western blots, we found a marked decrease (<70%) of the TI-VAMP signal in siRNA-transfected cells compared with control cells transfected with an unrelated siRNA targeting human β-globin (normalization was done in comparison with calreticulin) (Fig. 5A), indicating that there was a significant transient knockdown of TI-VAMP in polarized Caco2 cells. Similar results were obtained using a second siRNA (siRNA 7) (see supplementary material Fig. S1). We then investigated the effect of transient TI-VAMP knockdown on the apical localization of PLAP by immunofluorescence and confocal analysis of filter-grown Caco2 cells (Fig. 5B). Compared with control cells (siRNA targeting β-globin expression) where PLAP accumulates at the apical membrane, in Caco2 cells transfected with siRNA targeting TI-VAMP, PLAP was also mislocalized to the lateral membrane (Fig. 5B), similar to our finding with FRT si TI-VAMP::PLAP cells (Fig. 1 and Fig. 3A). In order to rule out intracellular accumulation of the protein below the plasma membrane, we have also performed the surface staining in non-permeabilized conditions where we obtained the same results (see supplementary material Fig. S2A). These results clearly indicate that PLAP is mislocalized to the lateral membrane in si-RNA-treated Caco2 cells. As expected, TI-VAMP knockdown did not have any effect on the basolateral localization of the basolateral marker Ag525 (Fig. 5B), indicating that the overall polarity of the monolayer was not affected. In order to eliminate the possibility of an off-target effect of the siRNA, we performed a rescue experiment by double transient-transfection of the siRNA targeting TI-VAMP together with a rat cDNA encoding rat brain TI-VAMP [sequence NP_445983 (Hibi et al., 2000)] that is resistant to the siRNA targeting the human sequence (see supplementary material Fig. S3). In approximately a third of the cells, PLAP was not mislocalized to the lateral membrane like in control conditions, whereas in two-thirds of the monolayer the mislocalization of PLAP to the lateral membrane was less prominent compared with the one observed in the TI-VAMP knockdown cells (see supplementary material Fig. S3). This experiment showed a partial rescue of the human TI-VAMP knockdown by the rat TI-VAMP cDNA, demonstrating that the effects on apical sorting were specific.
Because PLAP is directly sorted to the apical membrane both in FRT and Caco2 cells (Le Bivic et al., 1990; Paladino et al., 2006), these data indicate that TI-VAMP is involved in direct apical sorting of TM and GPI proteins in different cells. In order to test whether TI-VAMP is also involved in the transcytotic apical pathway, we analyzed the effect of its knockdown on the sorting of DPPIV, which in Caco2 cells transcytoses through the basolateral surface before being inserted in the apical membrane (Gilbert et al., 1991; Matter et al., 1990). Surprisingly, in contrast to the results obtained in FRT cells (Fig. 2), we found no effect of TI-VAMP knockdown on the apical localization of DPPIV (Fig. 5C and supplementary material Fig. S2B). The fact that TI-VAMP knockdown has such a different effect on the same protein in two different cell lines indicates that this effect is dependent upon the pathway followed by DPPIV in the two cell lines (Gilbert et al., 1991; Zurzolo et al., 1992a). Thus our data strongly suggest that the apical v-SNARE TI-VAMP is involved in direct apical sorting of TM and GPI-AP proteins independently of their raft association but does not function in the apical transcytotic pathway.
VAMP8 is involved in the transcytotic pathway of DPPIV in Caco2 cells
The v-SNAREs that control the transcytotic pathway have not been identified. In order to understand the mechanisms that control the transcytosis of DPPIV in Caco2 cells, we focused our studies on two possible v-SNARE candidates (VAMP3 and VAMP8) that have been described as being involved in endosomal recycling (Advani et al., 1998; Breton et al., 2000; Galli et al., 1994; Wong et al., 1998). We reasoned that, if one of these two v-SNAREs functioned in apical transcytosis, the localization of the v-SNARE should be disturbed by the TI-VAMP knockdown because of the expected crosstalk between the direct and indirect apical pathways. Interestingly, we observed that VAMP3 localization was not affected in TI-VAMP knockdown of Caco2 cells (Fig. 6A, left panel), whereas VAMP8 partially accumulated basolaterally and was less concentrated in intracellular vesicles and on the apical surface (Fig. 6A, right panel). Quantitative analysis of fluorescence by applying a threshold in order to select intracellular objects of a specific level of intensity allowed calculation (using the ImageJ® software) of the number of objects and the areas occupied by each object (Fig. 6B). This analysis showed that there are fewer and much smaller objects in the TI-VAMP knockdown condition compared with the control (Fig. 6B). Together with the more prominent surface staining of VAMP8 (Fig. 6A), these experiments show that the localization of VAMP8 is altered and that it is less concentrated in vesicular structures and more at the lateral membrane in TI-VAMP knockdown cells, thus indicating an involvement of VAMP8 in apical sorting. We therefore decided to study the effects of VAMP8 knockdown on the direct and indirect apical pathways that are followed by PLAP and DPPIV.
To interfere with VAMP8 expression, we used a siRNA targeting human VAMP8 [Hs_VAMP8_3 HP-validated siRNA (SI02653245) from Qiagen] that was transiently transfected in Caco2 cells. VAMP8 expression levels were measured after 4 days of culture on filters by western blotting on total cell lysates (Fig. 7A). By using a specific rabbit antibody against human VAMP8 in western blots, we found a marked decrease (>90%) of the signal corresponding to VAMP8 in knockdown cells compared with control cells (transfected with human β-globin siRNA) (Fig. 7A). We then investigated the effect of this knockdown on both the direct delivery of PLAP and the indirect delivery of DPPIV in filter-grown Caco2 cells (Fig. 7B,C).
Interestingly, while both the direct apical delivery of PLAP and direct basolateral delivery of Ag525 were not affected (Fig. 7B), DPPIV transcytosis to the apical membrane was strongly perturbed and the protein became mislocalized to the basolateral surface (Fig. 7C and supplementary material Fig. S2B). In order to analyze the mechanism of action of VAMP8 in more detail, we performed an immunofluorescence-based assay to follow DPPIV internalization from the basolateral to the apical plasma membrane (see supplementary material Fig. S4). After adding the antibody against DPPIV from the basolateral side of cells grown on filters for 4 days, we followed its internalization at different times. We first labeled surface DPPIV (non-permeabilized conditions) using a secondary antibody coupled to fluorescein isothiocyanate (FITC). After extensive washing and quenching, we added a secondary antibody conjugated with TRITC in permeabilized conditions to label internal DPPIV. In control conditions, the basolateral surface staining of DPPIV disappeared very quickly with time, followed by intracellular staining at ten minutes, consistent with a transcytosis of DPPIV from the basolateral towards the apical membrane, which was stained after 30 minutes (see supplementary material Fig. S4B). In VAMP8 transient-knockdown experiments, the basolateral staining of DPPIV remained unchanged with time, and no signal appeared intracellularly after ten minutes (supplementary material Fig. S4) or at the plasma membrane at later times (data not shown), thus demonstrating a role forVAMP8 in the endocytosis of DPPIV from the basolateral to the apical surface.
In summary, these data indicate that VAMP8: (1) does not cooperate with TI-VAMP to effect direct apical delivery, (2) is not responsible for the basolateral insertion of directly sorted basolateral proteins and (3) is a major player in the transcytotic apical pathway in polarized epithelial cells.
Virtually every intracellular membrane fusion event is mediated by a SNARE machinery (Mostov et al., 2003; Weimbs et al., 1997). Although it has been demonstrated that only a perfect match between the v-SNARE and its corresponding t-SNARE leads to successful membrane fusion (McNew et al., 2000; Scales et al., 2000), it is still debated how SNAREs contribute to specify trafficking and at which step (i.e. targeting or fusion) they regulate these events.
Studies of SNARE function in epithelial cells have shown a polarized distribution of t-SNAREs highly conserved among epithelial cell types and support their role in the establishment and maintenance of epithelial polarity (Li et al., 2002; Low et al., 1996; Low et al., 1998; Low et al., 2002). The t-SNARE Stx3 localizes to the apical plasma membrane of MDCK cells (Low et al., 1996), and both Stx3 and the v-SNARE TI-VAMP operate at the apical plasma membrane of Caco2 cells (Breuza et al., 1999; Galli et al., 1998). Overexpression of Stx3 reduces apical transport by a factor of 20 to 50 (Low et al., 1998) and does not cause mistargeting to the basolateral surface but, instead, results in the accumulation of apical proteins in intracellular vesicles (Low et al., 1998), consistent with the formation of non-productive SNARE complexes that impair fusion.
However, the apical pathway is rather complicated because it can be either raft dependent or raft independent and there are two ways to reach the apical surface, one direct and the other indirect (Paladino et al., 2006; Rodriguez-Boulan et al., 2005). Previous work has demonstrated the involvement of Stx3 in the apical sorting of two specific transmembrane proteins HA and SI (both associated with DRMs) that are directly sorted to the plasma membrane (Breuza et al., 2000; Lafont et al., 1999; Low et al., 1998). Consistent with these findings, both apical v- and t-SNARE TI-VAMP and Stx3 are present in post-TGN carriers in DRMs, and TI-VAMP forms apical SNARE complexes with Stx3 (Galli et al., 1998). Although overexpression of Stx3 does not alter the transcytosis of WT-pIgR in MDCK cells (Low et al., 1998), the role of TI-VAMP in the transcytotic pathway followed by apical proteins, especially in intestinal cells, has not been directly established (Breuza et al., 2000). Furthermore, it was not known whether raft- and non-raft-associated apical proteins use the same exocytic mechanism at the apical surface. Our data clearly demonstrate that TI-VAMP is necessary for the apical localization of raft- and non-raft-associated proteins, implying a role for TI-VAMP in the direct apical delivery for both pathways (Figs 1, 2 and 5). We also confirmed previous data that TI-VAMP is not involved in basolateral delivery (Figs 1 and 5). A quantitative and statistical analysis of these data was performed by means of a surface biotinylation assay in stable FRT clones lacking TI-VAMP (Fig. 3). Furthermore, by measuring the TER and studying the morphology of tight junctions, we also showed that mislocalization of apical non-raft and raft-associated proteins in knockdown cells for TI-VAMP was not due to a general perturbation of the epithelia (Fig. 4B). These results also suggest that the TI-VAMP-dependent apical targeting pathway is not involved in the formation of tight junctions and is not essential for the development of polarized epithelia.
Next, in order to analyse directly whether TI-VAMP was also involved in the indirect pathway to the apical membrane, we repeated the same knockdown experiments in Caco2 cells that, unlike FRT cells, target endogenous DPPIV to the apical surface via a transcytotic pathway (Gilbert et al., 1991). As control for the effect of TI-VAMP on the direct pathway, we used again PLAP, which is endogenously expressed by Caco2 cells and is targeted directly to the apical membrane (Le Bivic et al., 1990). As in FRT cells, TI-VAMP knockdown impaired the direct apical delivery of PLAP; however, it had no effect on both the transcytotic delivery of DPPIV and, as expected, basolateral delivery (Fig. 5). These data clearly demonstrate that TI-VAMP is involved only in the apical direct pathway and not in the apical indirect pathway and that this mechanism is conserved in different epithelia.
Interestingly, by specific knockdown of TI-VAMP, we not only impaired the direct apical delivery but also found mislocalization of the apical proteins to the basolateral surface, an effect that was not observed previously when the expression of TI-VAMP was perturbed using either an overexpression or an antibody block approach (Breuza et al., 2000; Galli et al., 1998; Low et al., 1998). These data suggest that, in the absence of TI-VAMP, apical carriers can fuse with the basolateral membrane. This could be explained by two hypotheses: either TI-VAMP has a prominent function in the sorting of the proteins at the level of the TGN, or another v-SNARE allowing basolateral fusion is present on post-TGN apical carriers and it is unmasked only after TI-VAMP knockdown. Because it is possible that this `recessive' VAMP would also be involved in transcytotic apical delivery, we focused our attention on two v-SNAREs that could be able to overcome TI-VAMP loss in our knockdown experiments. The first of these was VAMP3 (also known as cellubrevin or Cb) (McMahon et al., 1993), which is involved in the recycling of the transferrin (Tf) receptor (Galli et al., 1994) and in early endosomal pathways such as the apical transport of H+-ATPase (Breton et al., 2000), and the other was the v-SNARE VAMP8/endobrevin (Advani et al., 1998; Wong et al., 1998), which might mediate the endocytic apical recycling pathway in polarized cells (Antonin et al., 2000; Mullock et al., 2000; Steegmaier et al., 2000; Wong et al., 1998) and colocalizes also with the Tf receptor (Wong et al., 1998). To test the possible involvement of these two v-SNAREs in the apical pathway, we examined their localization in TI-VAMP knockdown in Caco2 cells (Fig. 6). Interestingly, the localization of VAMP3 was not affected by TI-VAMP knockdown (Fig. 6A, left panel). This is consistent with recent findings that VAMP3 co-immunoprecipitates mainly with Stx4, thus suggesting that it is not involved in the apical pathway but instead functions in the basolateral sorting of AP1B-dependent cargos, as recently demonstrated (Fields et al., 2007). By contrast, in TI-VAMP knockdown cells, VAMP8 appeared to be less concentrated in intracellular vesicular structures and localized more at the lateral membrane (Fig. 6A, right panel and Fig. 6B), indicating that it might be involved in the apical pathway.
VAMP8 is the best candidate to be involved in the transcytotic pathway because it is able to interact both with the basolateral t-SNARE Stx4 and with the apical t-SNARE Stx3. It has been shown that VAMP8 and Stx3 can form complexes with SNAP23 (Pombo et al., 2003). Furthermore, it has also been proposed that VAMP8, Stx4 and SNAP23 act together (Wang et al., 2007). Finally VAMP8, Stx3 and Stx4 were co-immunoprecipitated in parotid acinar cells (Imai et al., 2003). In addition, it was recently shown that, although the major localization of VAMP8 is endosomal in fixed MDCK cells, in live conditions it appears to be present in fine, long tubular structures emanating from endosomes and to be present at the apical surface of the cell (Wakabayashi et al., 2007). This is consistent with the role of VAMP8 in sorting to the apical plasma membrane, and, because VAMP8 participates in endocytosis and apical recycling in MDCK cells (Steegmaier et al., 2000), but not in apical direct delivery (Lafont et al., 1999), all these data suggest that VAMP8 could operate in the apical transcytotic pathway.
To confirm this hypothesis, we analysed the effect of VAMP8 knockdown on the direct and transcytotic apical delivery pathways in Caco2 cells (Fig. 7). As expected, VAMP8 knockdown had no effect both on the direct apical delivery of PLAP and on the direct basolateral delivery of Ag525 (Fig. 7B). However, in these conditions, the transcytotic apical delivery of DPPIV was remarkably affected and the protein was partially mis-sorted to the basolateral surface. Interestingly, we found an increase in the total amount (data not shown) and of the plasma membrane fraction of DPPIV in in VAMP-knockdown cells (Fig. 7). This could also be consistent with an impairment of the endocytic apical recycling pathway in VAMP8 knockdown. However, the fact that, in VAMP8-knockdown cells, DPPIV remains on the basolateral surface (supplementary material Fig. S4) is consistent with a role in endocytosis from the basolateral surface to the apical membrane.
Overall, our data indicate that VAMP8 is involved in the transcytotic pathway. Furthermore, because the localization of Ag525 was not impaired by VAMP8 knockdown (Fig. 7B), we can also conclude that direct basolateral delivery and the transcytotic apical pathway use different sets of v-SNAREs (Fig. 8). This hypothesis finds additional support from recent data showing the involvement of VAMP3 in the basolateral sorting of AP-1B-dependent cargos (Fields et al., 2007). Interestingly, when we examined Stx3 and Stx4 localizations in TI-VAMP-knockdown Caco2 cells, they were not perturbed compared with the control, suggesting that t-SNAREs and v-SNAREs do not play the same role in apical sorting (see supplementary material Fig. S5).
In conclusion, our data reveal specificity in the sorting of apical cargos that could be explained with two different models (Fig. 8). We cannot exclude that each cargo vesicle is equipped with more then one v-SNARE and the specificity is governed by their relative amount and stoichiometry. However, our data support the hypothesis that there is a specific v-SNARE for each post-TGN cargo (Fig. 8) and that mistargeting in the absence of one v-SNARE arises from the loading of cargo on an alternative pathway. Further work will be necessary to understand how v-SNAREs and cargos are sorted together at the level of the Golgi apparatus and endosomes.
Materials and methods
Reagents and antibodies
Cell culture reagents were purchased from Invitrogen. Antibodies against PLAP and ZO-1 were purchased from Rockland Bioscience and Zymed. The antibodies against rat 35/40 kDa and DPPIV were gifts from A. Quaroni (Dept of Biomedical Sciences, VRT 8004, Cornell University, Ithaca, NY). The antibodies against human Ag525, DPPIV and Patj were used as described previously (Le Bivic et al., 1988; Michel et al., 2005; Quaroni and Isselbacher, 1985). The antibody against TI-VAMP was used as described previously (Muzerelle et al., 2003). VAMP 3, VAMP 8, syntaxin 3 and syntaxin 4 were purchased from Transduction Laboratories. Biotin and protein A sepharose were obtained from Pierce Chemical Co. and Amersham. Horseradish peroxidase (HRP)-linked antibodies and streptavidin were purchased from Amersham GE Healthcare. All other reagents were purchased from Sigma-Aldrich.
Cell culture and transfections
FRT cells (Ambesi-Impiombato and Coon, 1979) were grown in F12 Coon's modified medium containing 5% fetal bovine serum (FBS). A stable clone expressing PLAP was previously obtained (Lipardi et al., 2000). Cells were transiently transfected using the GenePulserXCell (Bio-Rad) by mixing 100 pmol siRNA with 400 μl RPMI medium and 107 freshly trypsinized cells. Cells (2×106) were seeded on Transwell filters (24 mm diameter, Corning, NY) and TER was measured each day with a MilliCell apparatus (Millipore Corporation, Bedford, MA). Cells were stably transfected using Lipofectin® (Invitrogen) by mixing 3-5 μg DNA with 10 μg Lipofectin® and 3 ml Optimem, and selected in hygromycin (250 μg/ml) and/or G418 (10 μg/ml) for two weeks. A clone of Caco2 (TC7 cells) was grown in DMEM containing 10% FBS (Chantret et al., 1994). Cells were transfected using the Amaxa device, T solution and T20 program (Amaxa Biosystems, Germany) by mixing 60 pmol siRNA with 100 μl buffer T and 1.8×106 freshly trypsinized cells. Cells (Caco2, 1.8×106 and FRT, 2×106) were seeded on Transwell filters (24 mm diameter, Corning) and TER was measured with a MilliCell apparatus (Millipore Corporation).
Three siRNAs targeting TI-VAMP were designed: one against rat (5′-CCTCGTAGATTCGTCCGTC-3′) and two against human (5′-TGCCATTAAATTGAAATTATA-3′ and 5′-CTGCCAAGACAGGATTGTATA-3′). DNA corresponding were introduced as a hairpin behind a U6 promoter added to peGFP-N2 (BD Bioscience Clontech, Palo Alto, CA) containing a neomycin-resistance cassette, using the HindIII and BglII restriction sites (a kind gift from A. Le Bivic, Luminy, Marseilles, France) and sequenced. The siRNA targeting β-globin (bgloE1; 5′-GGUGAAUGUGGAAGAAGUUtt-3′) was a kind gift from N. Sauvonnet (Institut Pasteur, Paris, France). The siRNA targeting VAMP8 [Hs_VAMP8_3 HP-validated siRNA (SI02653245)] was purchased from Qiagen.
FRT cell extracts were prepared and analysed by western blotting as previously described (Lipardi et al., 2000), with primary antibodies and the corresponding secondary antibodies coupled to peroxidase (Amersham, GE Healthcare). FRT cells were selectively biotinylated and processed as described (Paladino et al., 2006). Lysate were immunoprecipitated with specific antibodies and biotinylated antigens were revealed with streptavidin. Caco2 cell extracts were prepared and analyzed by western blotting as described (Lemmers et al., 2002) with primary antibodies and the corresponding secondary antibodies coupled to peroxidase (Amersham, GE Healthcare). Samples were run on SDS-PAGE and bands were revealed using the ECL kit (Amersham, GE Healthcare) and quantified using the ImageJ (NIH) software.
FRT and Caco2 cells grown on Transwell filters for 4 days were washed with PBS containing CaCl2 and MgCl2, fixed with 4% PFA, and quenched with 50 mM NH4Cl. Cells were permeabilized with 0.075% saponin. Primary antibodies were detected with FITC or TRITC-conjugated secondary antibodies. Images were collected using a laser scanning confocal microscope (LSM 510; Carl Zeiss MicroImaging Inc.) equipped with a plan apo 63× NA-1.4 oil-immersion objective lens (Carl Zeiss MicroImaging Inc.). Fluorescent images were converted using ImageJ® software to 8 bit (256 grey levels) and a threshold selection was applied between 80 and 255 grey levels (top panels, red). Measurements of the area of the fluorescent objects was obtained using ImageJ software (bottom panels, white ellipses in Fig. 6B).
We are grateful to Lucien Cabanié for the purification of the mAb Cl158.2 hybridoma and TG18 rabbit serum and to Philippe Casanova for technical help. We are grateful to Véronique Proux-Gillardeaux and Lydia Danglot for helpful discussion of the data. This work was supported by Grants to C.Z. from MURST (PRIN 2005), from the European Union (LSH-2004-1.2.2-4) and from Agence Nationale de la Recherche ANR (05-BLAN 296-01) and to T.P. from the Weizmann-Pasteur Foundation. Work in the T.G. group was supported in part by grants from INSERM (Avenir Program), the European Commission (`Signalling and Traffic' STREP 503229), the Association pour la Recherche sur le Cancer, and the Fondation pour la Recherche Médicale. A.L.B. is supported by Ligue Nationale contre le Cancer and UMR 6216 CNRS/Université Aix Marseille II.