TOM1 is a PI5P effector involved in the regulation of endosomal maturation

The endocytic route is a complex and tightly regulated pathway. It serves as a sorting platform to differentiate cargoes coming from the cell surface or the extracellular medium and to target them for recycling, retrograde transport or degradation (Huotari and Helenius, 2011). This pathway consists of a succession of discrete compartments, and our understanding of this pathway has evolved rapidly in recent years. The canonical route involves the formation of early endosomes, their maturation into late endosomes and fusion with lysosomes to mediate the degradation of their contents. In this classical pathway, newly internalized cell-surface receptors are thought to be rapidly inactivated in early endosomes to downregulate their signaling. This paradigm has been challenged by the characterization of specific early endosomes, termed signaling endosomes, where signaling from a variety of receptors can continue and even be amplified along the endocytic route (Di Fiore and De Camilli, 2001; Miaczynska et al., 2004; von Zastrow and Sorkin, 2007; Zoncu et al., 2009). This transient APPL1-positive compartment is common to clathrin-dependent and clathrin-independent endocytosis. Upon maturation, these signaling endosomes are converted into classical early endosomes where the signaling of receptors is terminated.

Given the dynamic nature of this compartment, tight regulation is needed to maintain the identity of these organelles. Among the regulatory factors, lipids from the phosphoinositide family are now recognized as master regulators of this route (Di Paolo and De Camilli, 2006; Jean and Kiger, 2012). They are able to recruit specific effectors on the cytosolic face of organelles and coordinate a vast variety of signaling and trafficking processes (Lemmon, 2003; Di Paolo and De Camilli, 2006; Clague et al., 2009). As an example, phosphatidylinositol 3-monophosphate (PI3P) and phosphatidylinositol 3,5-bisphosphate [PI(3,5)P₂], which reside mainly on endocytic compartments, are well known regulators of vesicular trafficking and autophagy that function by recruiting proteins harboring phosphoinositide-binding modules (Balla, 2013). Not much is known about one of the least identified phosphoinositides, phosphatidylinositol 5-monophosphate (PI5P), which is present at low concentrations in cell membranes. However, in recent years, new findings have helped to clarify the function of PI5P in cells. In the nucleus, in response to stress signaling (Jones et al., 2006), PI5P acts as a regulator of chromatin organization and gene expression though its interaction with the PHD (plant homeodomain) domain-containing protein ING2 (Gozani et al., 2003). Outside the nucleus, PI5P has been shown to be implicated in the regulation of membrane dynamics and trafficking (Shisheva, 2013; Viaud et al., 2014a). In this context, three potential pathways lead to an increase in the amount of PI5P: one involves the direct phosphorylation of phosphatidylinositol by the 5-kinase PIKfyve, the second involves a PI3P–PI(3,5)P₂–PI5P route with the concerted action of the PIKfyve and the 3-phosphatases of the myotubularin family (Viaud et al., 2014a) and the third uses PI(4,5)P₂ as a source of PI5P. This latter pathway was originally described in cells infected with the bacterium Shigella flexneri, through the injection of the virulence factor IpgD, a PI(4,5)P₂ 4-phosphatase (Niebuhr et al., 2002). Upon infection with the pathogen, or expression of the bacterial lipid phosphatase, this primary pool of PI5P at the plasma membrane has been shown to induce membrane remodeling and activation of the Akt survival pathway (Niebuhr et al., 2002; Pendaries et al., 2006), to control the modulation of hemichannel opening (Puhar et al., 2013) and to alter the endocytic pathway (Ramel et al., 2011). Namely, increased levels of PI5P induce epidermal growth factor receptor (EGFR) activation independently of its ligand and impair the maturation of early endosomes to late endosomes, delaying the degradation of EGFR to maintain survival signals. PI5P also blocked the bulk endocytosis of fluid-phase markers like dextran, without altering the recycling or the endosome-to-TGN (trans-Golgi network) retrograde pathways (Ramel et al., 2011). As IpgD expression induced a rise in the amount of PI5P on endosomes (Ramel et al., 2011), we speculated that this blockade could be explained by the enrichment of a specific PI5P effector on endosomes, and undertook a fishing proteomic approach to characterize new PI5P effectors that could be responsible for this phenotype. Here, we unravel a new PI5P effector, the protein adaptor TOM1, and show that PI5P-dependent TOM1 recruitment to signaling endosomes blocks endosomal maturation.

In order to characterize new PI5P effectors and gain insight into the molecular mechanisms involved in the blockade of endosome maturation induced by PI5P enrichment, we undertook a proteomic approach on purified early endosomes from control cells or those overexpressing IpgD (PI5P-endosomes). We used a previously described method to separate the early endosomes from the late endosomes on a discontinuous sucrose gradient (Gorvel et al., 1991). Supplementary material Fig. S1A shows the characterization of these fractions with endosomal markers. The purified endosomes were subjected to mass-spectrometric analysis and the relative enrichment of proteins was quantified. Among the hits, the protein TOM1 [target of myb-1 (chicken)] appeared to be a candidate of particular interest, as it was enriched 4.1-fold in PI5P-endosomes compared with control early endosomes (Fig. 1A, upper panel). TOM1 is a 492-residue protein composed of different characterized domains (Fig. 1A, middle panel). It contains a VHS (Vps27-Hrs-STAM) domain and a GAT (GGA and TOM1) domain, both of which are found on several proteins involved in trafficking steps (Bonifacino, 2004; Wang et al., 2010), and a clathrin binding box (Seet and Hong, 2005; Katoh et al., 2006). The two peptides that allowed identification of TOM1 are found in neither of the two TOM1 subfamily members TOM1L1 and TOM1L2 (Fig. 1A, lower panel). In order to confirm this result, we immunoblotted purified early endosomes from control or IpgD-expressing baby hamster kidney (BHK) cells. As shown in Fig. 1B, IpgD induced activation of Akt (monitored by the phosphorylation of the S473 residue) and its accumulation on PI5P-endosomes, as we have described previously in other cell types (Pendaries et al., 2006; Ramel et al., 2011). As nuclear PI5P was shown to regulate gene expression (Gozani et al., 2003; Bua et al., 2013), we verified that TOM1 augmentation was indeed due to the recruitment of the protein to endosomes and not to an increase in its expression level by blotting TOM1 on a total cell lysate (Fig. 1C).

Finally, to demonstrate that the binding of TOM1 to PI5P was responsible for its recruitment to endosomes in IpgD-expressing cells, we sequestered PI5P with the 3xPHD probe, a well-characterized competitor for PI5P binding (Gozani et al., 2003). Expression of the 3xPHD probe interfered with TOM1 recruitment to the endosomes (Fig. 1D,E). We also observed a decrease in the activation of Akt, as monitored by the phosphorylation of S473 (Fig. 1D, 38.33±3.66%; mean±s.e.m.). The overall expression level of TOM1 was not altered in the post-nuclear supernatant (supplementary material Fig. S1B).

The precise localization of TOM1 in the endosomal system is quite debated in the literature. Indeed, TOM1 has been shown to be recruited, through its interaction with endofin (also known as ZFYVE16) or TOLLIP, to EEA1-positive endosomes (Katoh et al., 2004; Seet et al., 2004; Seet and Hong, 2005) or to a later endosomal compartment (Brissoni et al., 2006). More recently, TOM1 has been reported to colocalize with the protein GIPC (Tumbarello et al., 2012), a specific marker of signaling endosomes (Varsano et al., 2006). The antibody that we used to detect TOM1 by western blotting was inefficient for immunofluorescence studies; therefore, we turned to an antibody directed against the closely related isoform TOM1L2, as others did recently (Tumbarello et al., 2012). In our hands, both in human HeLa cells and in mouse embryonic fibroblast (MEF) cells, this antibody detects a single band at the expected size for TOM1 (∼60 kDa), and this band is specifically depleted in cells transfected with a small interfering (si)RNA specifically directed against TOM1 (supplementary material Fig. S1C,F; quantification in Fig. S1D,G). Consistent with what was reported previously for overexpressed and endogenous TOM1 (Katoh et al., 2004; Seet et al., 2004; Seet and Hong, 2005; Brissoni et al., 2006; Tumbarello et al., 2012), this antibody gave a punctate vesicular staining by immunofluorescence. This staining was greatly reduced upon knockdown of TOM1 by siRNA (supplementary material Fig. S1E), demonstrating the specificity of this signal. Therefore, this antibody was further used to determine the localization of TOM1 by immunofluorescence. As shown in Fig. 2A, in MEF cells, TOM1 was localized on peripheral vesicles. Interestingly, only a minor fraction of these vesicles colocalized with EEA1, whereas a stronger colocalization was observed with GIPC. Quantification of the colocalization using the Pearson's coefficient (Fig. 2B) clearly showed that colocalization between TOM1 and EEA1 is indeed low (2.7% of colocalization), whereas the colocalization between TOM1 and GIPC is much higher (32.6%). Interestingly, expression of IpgD does not change the Pearson's coefficients (Fig. 2B), which is consistent with a recruitment of TOM1 to signaling endosomes and not a change in its localization.

Given this, we hypothesized that TOM1 would be recruited to signaling endosomes together with internalized EGFR in IpgD-expressing cells. In control cells, overexpressed EGFR fused to GFP localized mainly at the plasma membrane (Fig. 2C, Control). In IpgD-expressing cells, as we described previously for endogenous EGFR (Ramel et al., 2011), EGFR–GFP was internalized to vesicular structures positive for both TOM1 and GIPC (Fig. 2C; IpgD and highlighted in zoomed boxes in Fig. 2D). Quantification of the colocalization between TOM1 and EGFR–GFP is shown in Fig. 2E. Overall, these results show that EGFR is retained in signaling endosomes that are positive for both TOM1 and GIPC when IpgD is expressed in cells.

An interesting possibility to explain TOM1 recruitment to PI5P-endosomes would be through a direct interaction with the phosphoinositide. In order to test this possibility, we resorted to lipid overlay experiments using PIP Strips^TM. Consistent with results obtained with Dictyostelium discoideum DdTOM1, the only VHS-domain-containing protein in the amoeba (Blanc et al., 2009), we found that the VHS domain of the human TOM1 interacted with the phosphatidylinositol monophosphates (Fig. 3A, VHS). Phosphatidylinositol bisphosphates and phosphatidylinositol trisphosphates were not recognized or were very weakly recognized, indicating that the interaction with phosphatidylinositol monophosphates was specific and not only based on an accumulation of negative charges. In order to compare the relative avidity of this domain to different phosphatidylinositol monophosphates, we used PIP Arrays^TM. The VHS domain of TOM1 showed a greater avidity for PI5P compared with the other phosphatidylinositol monophosphates, with as little as 25 pmol detected (Fig. 3B, VHS). This is in contrast to DdTOM1, which showed a clear preference for PI3P and no binding to PI5P (Blanc et al., 2009). Preferential binding of the VHS to PI5P was further confirmed in a liposome flotation assay with liposome containing phosphatidylcholine:phosphatidylethanolamine and PI3P, PI4P or PI5P (Fig. 3C).

We next investigated the residues involved in this interaction. The structure of the VHS domain has been solved, and the authors proposed that a patch of positively charged residues mediate potential interaction with membranes (supplementary material Fig. S2A) (Misra et al., 2000). In order to verify this hypothesis, we mutated these residues (K48, R52, K58 and K59) into alanine, yielding the KRKK mutant. Strikingly, these mutations greatly reduced the binding to phosphatidylinositol monophosphates as assessed by lipid overlay assays (Fig. 3A,B, VHS KRKK) and completely abolished the preference towards PI5P in the liposome binding assay (Fig. 3C). We also tested the affinity of the GAT domain of TOM1 for phosphoinositides. As shown in supplementary material Fig. S2B, the TOM1 GAT domain showed a slight preference for PI5P over PI3P and PI4P. However, in the liposome flotation assay, this preference was not confirmed across four independent experiments (supplementary material Fig. S2C). Therefore, we concluded that the GAT domain of TOM1 is not a strong interactor of PI5P. Taken together, these results show that the TOM1 VHS domain preferentially binds to PI5P through the KRKK motif. The existence of this PI5P-binding module on the protein and the fact that IpgD expression yields high levels of PI5P in cells (Niebuhr et al., 2002), and in particular on endosomes (Ramel et al., 2011), are consistent with the PI5P-dependent recruitment of TOM1 to endosomes in IpgD-expressing cells.

TOM1 has been proposed to be part of an endosomal sorting complexes required for transport (ESCRT)-0 complex (Brissoni et al., 2006; Blanc et al., 2009). It was therefore very tempting to postulate that TOM1 would serve as an ESCRT-associated protein, in particular for the regulation of EGFR degradation. To investigate this, we resorted to siRNA-mediated depletion of TOM1 in cells and stimulation with EGF. No effect of TOM1 depletion on EGFR degradation was found either biochemically (supplementary material Fig. S3A) or by immunofluorescence (supplementary material Fig. S3B), indicating that TOM1 does not act as an ESCRT-associated protein for EGFR degradation induced by EGF stimulation. Therefore, we postulated that PI5P enrichment on signaling endosomes might cause an accumulation of TOM1 on these structures and might represent the reason for the delayed EGFR degradation specifically in IpgD-expressing cells.

In order to study the potential role of TOM1 in EGFR trafficking in IpgD-expressing cells, we resorted to an immunofluorescence study to track the degradation of EGFR following EGF stimulation in IpgD-expressing cells. Cells were transfected with either a control siRNA (Fig. 4, siControl) or with a siRNA targeting TOM1 (Fig. 4, siTOM1), and were then transiently transfected to express GFP or GFP–IpgD. After serum starvation, the cells were stimulated with EGF or left unstimulated, fixed and stained with an anti-EGFR antibody. Semi-quantitative confocal imaging was performed, and the intensity of EGFR signal per cell was quantified, reflecting the amount of remaining receptor per cell. As anticipated, in control unstimulated cells, EGFR was restricted to the plasma membrane (Fig. 4A, siControl, −EGF). When stimulated with EGF, the signal for EGFR was lost, as a consequence of its lysosomal degradation. Quantitatively, this resulted in a decrease to 41.20±2.47% (compared with control cells 101±4.2%, Fig. 4C; mean±s.e.m.). As expected, IpgD expression induced EGFR internalization, but did not alter the amount of total EGFR per cell (95.31±5.56%).

By contrast, in IpgD-expressing cells, in the presence of control siRNA, EGFR degradation was delayed as we published previously (Ramel et al., 2011), with 65.5±2.3% remaining EGF (compared with 41.20±2.47% in control EGF-stimulated cells, Fig. 4B,C; mean±s.e.m.). Strikingly, cells overexpressing IpgD and depleted for TOM1 showed a reduction in EGFR level after EGF challenge (35±4%), back to the level of the control. This clearly indicated that TOM1 accumulation is crucial for the blockade of EGFR degradation induced by IpgD. Finally, treating the cells with the lysosomal inhibitor bafilomycin A1 abrogated EGFR degradation under all conditions (for clarity, only one bafilomycin A1 condition is shown in Fig. 4B). These results were confirmed biochemically in HEK293T cells transiently overexpressing EGFR–GFP to monitor its degradation upon EGF stimulation. However, when cells expressed IpgD, we observed a retarded EGFR–GFP degradation. Strikingly, this delay was completely abolished when the cells were silenced for TOM1 expression (Fig. 4D).

Among the described interacting partners of TOM1, endofin has been proposed to play a role in the degradation of EGFR (Seet and Hong, 2001). To investigate whether endofin was also involved, we knocked it down with a specific siRNA. This resulted in a 65% mRNA expression decrease as measured by qRT-PCR (supplementary material Fig. S3C) and a 40% decrease at the protein level (supplementary material Fig. S3D,E). As shown in supplementary material Fig. S3F, we could not detect any role for endofin in the retarded EGFR degradation in IpgD-expressing cells. This clearly indicates that TOM1 itself, and not its partner endofin, is involved in the regulation of EGFR degradation in IpgD-expressing cells.

Given the fact that TOM1 knockdown in IpgD-expressing cells prevents the blockade in EGFR degradation, we wanted next to investigate whether this could perturb the survival signals downstream of EGFR. As shown in supplementary material Fig. S4, knocking down TOM1 indeed reversed PI5P-induced Akt activation, as monitored by S473 phosphorylation.

We showed previously that IpgD expression was also inhibiting fluid-phase bulk endocytosis, as demonstrated by fluorescently labeled dextran uptake (Ramel et al., 2011). In order to confirm whether TOM1 depletion in cells was abrogating this effect, we performed an IgG-uptake experiment, as described elsewhere (Petiot et al., 2003). In control MEF cells not expressing IpgD, internalized IgG was sent to degradation with no remaining signal detected by immunofluorescence compared with bafilomycin-A1-treated cells (Fig. 5A, quantification in Fig. 5B). In IpgD-expressing cells, IgG uptake clearly demonstrated a delayed degradation of IgG, with close to 40% of IgG left. Strikingly, TOM1-depletion completely reversed the phenotype, with no detectable IgG left in these cells. Again, knocking down TOM1 alone did not have any effect on IgG degradation, as we showed for EGFR degradation.

In order to confirm a crucial role of TOM1 in endocytosis blockade in IpgD-expressing cells, we resorted to a recovery experiment. We used siRNA targeting the 3′-UTR region of TOM1 to deplete the protein in control or IpgD-expressing MEF cells, and then overexpressed either the wild-type or KRKK mutant of TOM1 as a GFP fusion protein. The efficiency of siRNA and overexpression was assessed by immunoblotting, as shown in supplementary material Fig. S2D and quantified in supplementary material Fig. S2E. We used the IgG uptake assay to investigate whether overexpression of wild-type or KRKK TOM1 was able to restore the phenotype. As shown in Fig. 6, overexpression of wild-type TOM1 clearly reversed the knockdown phenotype by restoring the blockade in bulk endocytosis induced by IpgD expression, although not completely. It has to be noted that in these experiments, the controls gave undistinguishable results from the ones in Fig. 5. Strikingly, overexpression of the KRKK mutant was not able to restore the phenotype, even at a high expression level (an example is shown in Fig. 6A). These results confirm that PI5P binding to the VHS domain is required for TOM1 to block endocytosis in PI5P-enriched cells.

Taken together, these results show that TOM1 recruitment to signaling endosomes in IpgD-expressing cells is responsible for the endosomal maturation defect and maintenance of survival signals.

Taken together, our data provide compelling evidence that, in IpgD-expressing cells, the recruitment of the adaptor protein TOM1 by PI5P is responsible for the blockade of EGFR degradation and fluid-phase endocytosis. The large increase in PI5P levels on endosomes following the expression of the Shigella flexneri effector IpgD results in the recruitment of TOM1 to signaling endosomes and perturbs the endolysosomal maturation pathway. This effect is specific to IpgD expression and reflects the ability of the bacteria to subvert cellular processes, such as that shown here involving a trafficking machinery component, for its benefit. PI5P is now recognized as a regulator of different cellular processes but PI5P effectors are still sparsely described. The first effector ever described for PI5P was ING2, a nuclear protein involved in chromatin regulation (Gozani et al., 2003). ING2 has a PHD domain, also found on many other nuclear proteins that show affinity for PI5P (Keune et al., 2011). However, PI5P effectors outside the nucleus that might be involved in its function on membranes facing the cytosol are sparse. Very recently, our laboratory has shown that PI5P is able to activate the small GTPase Rac1, therefore regulating actin cytoskeleton dynamics and cell invasion (Viaud et al., 2014b). Our present work identifies TOM1, through its VHS domain, as a PI5P effector regulating endosomal maturation.

TOM1 was originally described as a target of the v-myb gene (Burk et al., 1997). Although their cellular functions are still not totally understood, TOM1 and its two orthologs, TOM1L1 and TOM1L2, are proposed to play a role in membrane trafficking (Wang et al., 2010). TOM1L1 shares 31% identity with TOM1 (Seroussi et al., 1999) and has been implicated in src family kinase mitogenic activity (Li et al., 2005; Collin et al., 2007) and EGFR endocytosis (Liu et al., 2009) The function of TOM1L2, which shares 59% identity with TOM1, is less well defined, although a study with Tom1l2 gene-trapped mice suggests an important role in the immune response linked to a potential tumor suppressor activity (Girirajan et al., 2008). TOM1 was recently proposed to play a role in autophagy as an adaptor protein targeting myosin VI to signaling endosomes to facilitate autophagosome maturation (Tumbarello et al., 2012). Moreover, TOM1 has been shown to interact with Hrs, an ESCRT-0 component (Brissoni et al., 2006), although this is debated in the literature (Seet et al., 2004). This suggests that TOM1 might serve as an ESCRT component, or at least be associated with the ESCRT machinery, an essential multi-protein complex mediating the degradation of several receptors (Katzmann et al., 2001; Wollert and Hurley, 2010). It has been proposed that TOM1 and Hrs could play parallel roles in clathrin recruitment and sorting of ubiquitylated cargoes on early endosomal membranes, for their subsequent delivery to the lysosomes. Recent phylogenetic analysis shows that TOM1 family proteins form an ancestral ESCRT-0 complex in the amoeboflagellate Breviate anathema (Herman et al., 2011), as well as in Dictyostelium discoideum (Blanc et al., 2009). The authors demonstrated that in Dictyostelium, DdTOM1 interacts with Esp15, a clathrin adaptor protein involved in EGFR internalization (Benmerah et al., 1998) and degradation by the ESCRT complex (Roxrud et al., 2008). However, the data presented here and by others (Tumbarello et al., 2012) clearly demonstrate that, in mammals, TOM1 does not act as an ESCRT protein for EGFR, but might present some specificity towards a specific cargo, as highlighted by its role in the degradation of IL-1 receptor (Brissoni et al., 2006). By contrast, we show that in cells expressing IpgD, PI5P-mediated TOM1 recruitment to endosomes is responsible for the delayed degradation of EGFR, and this is accompanied by an increase in downstream survival signals. In the context of the infection by Shigella flexneri, such a subversion of the cellular components represents a massive advantage for the invading bacteria.

The VHS domain of TOM1 was originally described as a protein–protein interaction domain with the potential to bind to lipid membranes thanks to the presence of a positively charged stretch (Misra et al., 2000). We show here that the mammalian VHS domain of TOM1 can bind directly to PI5P. We propose that in our system, where PI5P levels are increased, the balance between PI3P and PI5P could be modified, which would favor the preferential direct binding of TOM1 to PI5P. We also describe a patch of positively charged residues (K48, R52, K58 and K59) in the TOM1 VHS domain that are important for phosphoinositide binding. We confirm the involvement of these residues in the binding of TOM1 VHS to membranes as proposed by Misra and collaborators (Misra et al., 2000). It has to be noted that the binding was not completely abolished by the mutations, suggesting that other residues in the surroundings might contribute to the binding. For example, R57 has been suggested to be close enough to participate in this contact (Misra et al., 2000; supplementary material Fig. S3). The TOM1 GAT domain could also contribute to the affinity of TOM1 for phosphoinositides. The GAT domain has several ubiquitin-interacting motifs and is involved in the recognition of ubiquitylated protein (Shiba et al., 2004; Akutsu et al., 2005). In TOM1, it differs from the GAT domain of the Golgi-localized, γ-ear, Arf-binding proteins (GGA), a family of adaptors in the TGN. First, it lacks the residues involved in Arf1 binding, which explains its lack of interaction with these small GTPases of the Arf family (Shiba et al., 2004; Wang et al., 2010). Second, the two basic residues involved in GGA1 GAT binding to the phosphoinositide PI4P (R260, R281) are not conserved in TOM1 GAT. In Dictyostelium, the GAT domain of DdTOM1 was also shown, like the VHS domain, to bind to phospholipids, mainly to PI(3,5)P₂. However, the residues involved in the lipid binding were not defined (Blanc et al., 2009). Although we show here that human TOM1 GAT domain is able to bind to monophosphorylated phosphoinositides, it does not show any specificity towards PI5P. Our results clearly demonstrate that, in mammalian cells, the TOM1 VHS domain has a dominant role over the GAT domain in phosphoinositide binding and the regulation of cellular trafficking, as demonstrated in our rescue experiments.

It was initially shown that TOM1 is recruited to endosomes through protein–protein interaction with the PI3P-binding proteins tollip (Katoh et al., 2004) and endofin (Seet et al., 2004), endofin being known to be involved in EGFR degradation (Seet and Hong, 2001). It was therefore tempting to postulate that endofin or another binding partner was acting upstream of TOM1 in IpgD-expressing cells to control the degradative pathway. To rule out this possibility, we propose two arguments. First, we excluded the involvement of endofin in the control of EGFR degradation in IpgD-expressing cells. Second, given the fact that the knockdown of TOM1 alone totally abrogates IpgD effects on the degradative pathway, it is very unlikely that an interacting partner plays a role upstream of TOM1. In addition, this demonstrates that neither TOM1L1 nor TOM1L2 have redundant functions with TOM1 in this system.

In a previous study, we showed that PI5P produced by IpgD specifically modulates the degradative pathway and has no effect on the recycling or retrograde pathway (Ramel et al., 2011). Here, we unravel the molecular mechanisms used by PI5P to affect the endolysosomal pathway. We clearly demonstrate that the recruitment of the adaptor TOM1 to signaling endosomes by PI5P is responsible for this perturbation of the endolysosomal pathway, affecting both receptor and soluble cargo degradation and suggesting a role for PI5P and TOM1 at an early step of endocytosis. PI3P has also been shown to control the endolysosomal pathway. However, whether PI3P can affect both transport and sorting is not clear. PI3P was reported to affect receptor sorting, through its interaction with the ESCRT-0 Hrs protein, but not bulk transport (Petiot et al., 2003). However, in podocytes from the phosphatidylinositol 3-kinase Vps34-invalidated mice, both receptor-mediated endocytosis and fluid-phase uptake are compromised at the level of Rab5 and EEA1-positive early endosomes (Bechtel et al., 2013). It is interesting to note that, in our model, contrary to a loss of PI3P, a strong increase in PI5P leads to the blockade. Moreover, although signaling endosomes were not documented in the Bechtel et al. study, PI5P seems to affect an upstream compartment, i.e the signaling APPL1 endosomes, the precursors of the Rab5 and EEA1-positive PI3P-rich endosomes (Zoncu et al., 2009). PIKfyve, an enzyme that produces PI5P either by direct phosphorylation of phosphatidylinositol or together with myotubularins from the PI3P and PI(3,5)P₂ pathway (Sbrissa et al., 1999; Tronchère et al., 2004), controls fluid-phase endocytosis but has no effect on receptor tyrosine kinase (RTK) receptors, like the EGFR (Ikonomov et al., 2003). The PIKfyve effect was attributed to PI(3,5)P₂ acting at a later stage of endocytosis. In this study, a potential role for PI5P derived from dephosphorylation of PI(3,5)P₂ was not evoked. In our system, PI5P originates from the plasma membrane pool of PI(4,5)P₂. This difference could explain the blocking of the cargoes at very early stages (i.e. signaling endosomes) of endolysosomal maturation.

Taken together, the results in this study suggest an unprecedented role of PI5P in endosomal maturation and in the regulation of discrete trafficking steps in the endosomal network. We propose a model in which expression of Shigella flexneri phosphatase IpgD induces PI5P enrichment in and TOM1 recruitment to signaling endosomes. As a result, the endosomal maturation is blocked, impeding the signaling pathway of RTKs and fluid-phase cargoes (Fig. 7). Therefore, PI5P appears to be a crucial coordinator of multiple fundamental cellular functions, such as the regulation of gene expression, actin cytoskeleton, membrane trafficking and signaling. It is also closely related to the development of several pathologies, such as cancer, diabetes or shigellosis.

siRNAs were from Ambion. The negative control used was the Silencer® Select Negative Control No. 1 (#4390843). Two different oligos targeting human TOM1 were used – Silencer® Select Pre-Designed siRNA s19514 and s19515. Initial characterization showed that both siRNAs were able to efficiently silence TOM1 in HeLa cells (human origin), and gave indistinguishable results throughout the study. Owing to the differences in mRNA sequence between the human and mouse TOM1, only s19514 was efficient to deplete TOM1 in MEF cells (mouse origin) and was therefore used in this study on MEF cells. The siRNA targeting the mouse 3′-UTR of TOM1 (referred as siTOM1-UTR) was custom designed and synthesized according to the sequence (sense) 5′-UCCCAUCCUGCUAACGACUAUGAUU-3′. The siRNA s18852 (Ambion) was used to silence endofin (also known as ZFYVE16). The antibodies used in this study were as follows: anti-TOM1 (ab99356, Abcam), anti-TOM1L2 (ab96320, Abcam), anti-Akt (clone H-136, sc-8312, Santa Cruz Biotechnology), anti-pAkt (S473) (clone D9E, #4060, Cell Signaling Technology), anti-EEA1 (#610457, BD Biosciences), anti-actin (A5441, Sigma-Aldrich), anti-GST (clone B-14, sc-138, Santa Cruz Biotechnology), anti-Hsp90 (sc-13119, Santa Cruz Biotechnology), anti-RhoGDI (06-730, Upstate-Millipore), anti-Hrs (clone A-5, #ALX-804-382-C050, Enzo Life Sciences), anti-endofin (#H00009765-D01, Abnova). Antibodies against EGFR were from BD Biosciences (#555996) or from Cell Signaling Technology (clone D38B1, #4267). Control rat IgG was from Abcam (ab37361). Secondary horseradish peroxidase (HRP)-coupled antibodies were from Promega, and Alexa-Fluor-coupled antibodies were from Life Technologies. Other chemicals were from Sigma-Aldrich. The peGFP-C2-TOM1 was kindly provided by Dr Emmanuel Lemichez (INSERM U895, Nice, France) and was as described previously (Visvikis et al., 2011).

Total RNA from transfected cells were prepared using the GeneElute Total RNA Mammalian Miniprep kit from Sigma. To prepare cDNA, 1 µg was reverse-transcribed using Superscript III reverse transcriptase (Life Technologies), following the manufacturer's recommendations. Primers were as follows: endofin, 5′-AGGTAGGATGGACAGTTA-3′ and 5′-TGAGCAGTGGTTAGAATC-3′; GAPDH, 5′-ACATCGCTCAGACACCATG-3′ and 5′-TGTAGTTGAGGTCAATGAAGGG-3′. The expression of endofin mRNA was normalized against GAPDH mRNA expression.

Truncated forms of TOM1 were amplified by PCR using peGFP-C2-TOM1 as a template and cloned in frame into the pGEX-KG vector using BamHI and EcoRI restriction sites (underlined in the primer sequences below). The VHS domain was amplified using the following primers: 5′- CGGGATCCATGGACTTTCTCCTGGG-3′ and 5′-CGGAATTCTCACATGGGGAACTCCAGGC-3′; the GAT domain was amplified using the following primers: 5′-CGGGATCCACTGACCTGGACATGC-3′ and 5′-CGGAATTCCTCATTTCCGTTCATGGCGC-3′. Site-directed mutagenesis was performed to introduce the KRKK mutations in the VHS domain of TOM1. All constructs were verified by double-strand sequencing.

MEF cells stably expressing IpgD or Inp54p were generated using standard methods. Briefly, MEF cells validated for the Tet-Off system from Clontech were cultured in G418-containing medium (500 µl/ml), and were transfected with the previously described pTRE2-hyg-2myc vector including the coding sequence for IpgD or Inp54p (Ramel et al., 2011) using Lipofectamine LTX with Plus Reagent (Life Technologies) according to the manufacturer's instructions. After 48 h, hygromycin (150 µg/ml) was added to the cells with doxycycline (2 µg/ml) to suppress the transgene expression. Transformants were selected and single-cell-cloned with the limiting dilution method. Expression of IpgD was induced by removal of doxycycline for 24 h.

BHK and HeLa cells were cultured as described previously (Ramel et al., 2011). cDNA transfection was performed using Effectene (Qiagen). MEF cells were transfected using Lipofectamine LTX with Plus Reagent (Life Technologies) according to the manufacturer's instructions. siRNA transfection was performed using Lipofectamine® RNAiMAX (Life Technologies) using the reverse transfection protocol, according to the manufacturer's instructions. When needed, the cells were serum-starved for at least 3 h and stimulated as indicated with EGF in serum-free medium for the indicated time.

Endosome purification was performed as described previously (Gorvel et al., 1991; Ramel et al., 2011). Briefly, BHK cells were serum-starved overnight, gently scraped in ice-cold PBS on ice and homogenized in homogenization buffer (250 mM sucrose, 3 mM imidazole, pH 7.4). The post-nuclear supernatant was adjusted to 40.6% sucrose, loaded at the bottom of a SW41 tube (Beckman Coulter) and layered with 35% and 25% sucrose, 3 mM imidazole pH 7.4 and homogenization buffer. After centrifugation at 35,000 rpm (210,000 g at rmax) for 90 min at 4°C in a SW41 rotor (Beckman Coulter), early endosomes were collected at the interface of the 35–25% sucrose layers and the late endosomes at the interface of the 25% sucrose-homogenization buffer layers. The proper isolation of early endosomes from the late endosomes was verified by western blotting using antibodies specific for each compartment, as described previously (Ramel et al., 2011).

Proteins from the early endosomal fractions were precipitated using methanol-chloroform. Protein samples were then reduced in Laemmli buffer (final composition 25 mM DTT, 2% SDS, 10% glycerol, 40 mM Tris-HCl pH 6.8) for 5 min at 95°C and alkyled with iodoacetamide. Proteins were loaded on a 1D SDS-PAGE gel containing 4–12% acrylamide, made in-house (Mini-Protean, BioRad), and electrophoretic migration was stopped as soon as the protein sample entered the separating gel. The gel was briefly stained with Coomassie Blue and a single band, containing the whole sample, was excised. Proteins were digested overnight with a solution of modified trypsin (20 ng/µl, sequence grade, Promega) at 37°C. The resulting peptides were dissolved in 2% acetonitrile, 0.05% trifluoroacetic acid (TFA). Three independent nano-LC-MS/MS analyses were performed for each sample using an Ultimate 3000 system (Dionex) coupled to an LTQ-Orbitrap Velos mass spectrometer (Thermo Fisher Scientific). Each sample was loaded on a C-18 precolumn (300 µm ID×5 mm, Dionex) at 20 µl/min in 5% acetonitrile, 0.05% TFA. After 5 min of desalting on the precolumn, peptides were separated on a 75 µm ID×15 cm C18 column (packed in-house with Reprosil C18-AQ Pur 3 µm resin, Dr Maisch; Proxeon Biosystems). Peptides were eluted using a 5–50% linear gradient of solvent B over 110 min [solvent A was 0.2% formic acid in 5% acetonitrile (ACN) and solvent B was 0.2% formic acid in 80% ACN]. The LTQ-Orbitrap Velos was operated in data-dependent acquisition mode with the XCalibur software. Survey scan mass spectra were acquired in the Orbitrap on the 300–2000m/z range with the resolution set to a value of 60,000. The 20 most intense ions per survey scan were selected for collision-induced dissociation (CID) fragmentation and the resulting fragments were analyzed in the linear ion trap (LTQ). Dynamic exclusion was employed within 60 s to prevent repetitive selection of the same peptide.

The Mascot Daemon software (version 2.2.0, Matrix Science) was used to perform searches in the SwissProt-Trembl database (Uniprot release 2010_09 protein database). Mascot results were parsed with a software developed in-house – Mascot File Parsing and Quantification (MFPaQ, version 4.0.0; Mouton-Barbosa et al., 2010) – and an identified protein was considered as a hit if it was identified with at least two peptides with a score greater than the significance threshold score for a probability P<0.05 or at least one peptide with a score greater than the significance threshold score for P<0.001, as determined by the Mascot Search program. Quantification of proteins was performed using the label-free module implemented in the MFPaQ v4.0.0 software. In order to perform protein relative quantification, a Protein Abundance Index (PAI) was calculated, defined as the average of XIC (extracts ion chromatograms) area values for the three most intense tryptic peptides identified for this protein. Treated:control ratios were determined by the sum of PAI values in three replicate analyses of the treated condition over the sum of PAI values in three replicate analyses of the control condition. For differential conditions, a Student's t-test on the PAI values was used for statistical evaluation of the significance of expression level variations. A twofold change and a P-value of 0.05 were used as combined thresholds to define biologically regulated proteins.

Among the 1807 proteins quantified on purified early endosomes, 87 were found to be enriched on early endosomes from IpgD-expressing cells and 43 proteins were lost or greatly reduced. Finally, 1677 proteins were unaltered by IpgD expression. A complete table of the obtained results is available upon request.

Typically, cells were washed in ice-cold PBS and lysed in buffer [50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% Triton X-100, 1 mM EDTA, 1 mM orthovanadate sodium, 50 mM β-glycerophosphate, protease inhibitor Cocktail Set V (Calbiochem)] for 30 min on ice. Cell lysates were clarified by centrifugation at 10,000 g for 10 min at 4°C, and quantified using the BioRad protein assay. Lysates were denatured in Laemmli sample buffer and the proteins were separated by SDS-PAGE, before western blotting on Immobilon-P membranes (Merck-Millipore). Immunoreactive bands were detected by chemiluminescence with the SuperSignal West Pico detection system (Thermo Scientific) on a ChemiDoc MP acquisition system (BioRad).

Immunofluorescence was performed essentially as described previously (Boal et al., 2010). Cells grown on glass coverslips were fixed with 4% paraformaldehyde for 15 min followed by permeabilization with 0.1% Triton X-100 in PBS for 5 min. The cells were then blocked using PBS containing 1% BSA (PBS-BSA) and probed with primary antibodies and secondary antibodies. Coverslips were mounted in Mowiol and imaged by confocal microscopy on either a LSM510 or a LSM780 confocal microscope (Zeiss). For quantification of EGFR signal per cell, the imaging parameters were set to prevent any saturation of the signal. Single cells were selected as regions of interest (ROIs), and the intensity of the EGFR channel reflected the amount of EGFR present in the cell. For quantification of colocalization, Pearson's coefficient was calculated using Volocity software (Perkin Elmer). Briefly, single cells were selected as ROIs, and the Pearson's coefficient was calculated using the autothresholding function.

Rat IgG uptake was performed as described previously (Petiot et al., 2003). Briefly, MEF cells stably expressing IpgD or control MEFs were seeded on glass coverslips, and expression of the transgene was induced by the removal of doxycycline. Nonspecific rat IgG (40 µg/ml) was incubated overnight in culture medium in the presence or absence of bafilomycin A1 (100 nM). The cells were then fixed and processed for immunofluorescence using a fluorescent anti-rat-IgG antibody. Cells were imaged by confocal microscopy and z stacks were acquired throughout the cells. The number of labeled vesicles per cell was counted for each condition and expressed as a percentage of those counted in bafilomycin-A1-treated cells.

Recombinant protein purification was essentially as described previously (Boal et al., 2011) using gluthatione–Sepharose beads according to the manufacturer's instructions (GE Healthcare). For the lipid-protein overlay assay, PIP Strips^TM or PIP Arrays^TM (Echelon Biosciences) were rehydrated in TBS containing 0.1% Tween 20 (TBST), blocked in TBST containing 0.2% BSA (fatty-acid free) and incubated overnight with 15 µg/ml GST fusion proteins in TBST. After extensive washes, bound recombinant proteins were detected using an anti-GST antibody followed by an HRP-coupled anti-mouse-IgG antibody. Liposome flotation assays were essentially performed as described previously (Bigay and Antonny, 2005). Briefly, purified recombinant proteins were mixed with extruded liposomes (100 nm, phosphatidylcholine:phosphatidylethanolamine:NBD–phosphatidylethanolamine:PIP 64∶28∶2∶6 mol%)</emph>.

Statistical analysis was performed using the unpaired Student's t-test. The results are presented as the mean±s.e.m. and P-values considered to be significant are indicated in the figure legends.

We would like to thank all the past and present members of the Payrastre laboratory, and Andrea Puhar for helpful discussions relating to this project. We are grateful to Véronique Pons for her initial help with the purification of endosomes.