Dictyostelium EHD associates with Dynamin and participates in phagosome maturation

C-terminal Eps15-homology-domain (EHD) proteins have recently emerged as a new family of proteins with major regulatory functions in the endocytic pathway (Naslavsky and Caplan, 2011). These proteins are characterized by the presence of an ATPase G domain followed by a coiled-coil region, and a single C-terminal Eps15-homology (EH) domain (Daumke et al., 2007; Mintz et al., 1999). The EH domain recognizes proteins with Asn-Pro-Phe (NPF) sequences (de Beer et al., 1998). Structural and functional features have revealed that EHDs are related to the Dynamin superfamily, although EHDs bind to and hydrolyze ATP rather than GTP (Daumke et al., 2007). Hence, as reported for dynamin (Takei et al., 1998), mouse EHD2 has been shown to tubulate liposomes and oligomerize in ring-like structures around these tubules (Daumke et al., 2007). In mammalian cells, there are four homologous (>70% identity) EHD proteins (EHD1–EHD4). Other organisms such as Caenorhabditis elegans and Drosophila have only a single EHD protein. No apparent homolog has been found in Saccharomyces cerevisiae. Several functions have been assigned to EHD proteins, with particular emphasis on protein sorting and membrane trafficking functions along the endocytic pathway (Naslavsky and Caplan, 2011). For instance, one of the most documented roles for EHD1 is in controlling cargo exit from the endocytic recycling compartment back to the plasma membrane (Caplan et al., 2002; Jovic et al., 2007; Lin et al., 2001). This is also observed for RME-1, the C. elegans ortholog of human EHD1, which is required for yolk receptor recycling (Grant et al., 2001). Moreover, EHD3 has been shown to contribute to cargo transport from early endosomes to the Golgi (Naslavsky et al., 2009). Surprisingly, despite EHDs being clearly implicated in cargo recycling from endosomes to several other compartments, to our knowledge, EHDs have never been shown to participate in protein retrieval from phagosomes.

The interaction of EHDs with protein partners mainly provides the molecular basis for the distinct functions of EHD-family members (Grant and Caplan, 2008; Naslavsky and Caplan, 2011; Zhang et al., 2012a). For instance, MICAL-L1, an EHD1 binding partner, has been shown to recruit collapsin response mediator protein-2 (CRMP2), which binds to dynein (Rahajeng et al., 2010). This tripartite interaction (EHD1–MICAL-L1–CRMP2) is proposed to connect EHD1 to dynein motors. Interestingly, EHD partners also include known effectors of Rab GTPases, suggesting that EHDs and Rabs might cooperate to regulate endocytic trafficking (Zhang et al., 2012b). In addition, some EHD partners facilitate EHD-dependent membrane bending, as reported for the BAR-domain-containing protein AMPH1 (also known as amphiphysin and Bin1), which interacts with EHD1 and contributes to tubule formation (Pant et al., 2009). Finally, besides their role in membrane remodeling, EHDs might serve to regulate the proper function of their binding partners. Hence, the single EHD protein in Lampreys has been reported to directly interact with dynamin, and this interaction has been shown to inhibit dynamin polymerization during synaptic vesicle formation (Jakobsson et al., 2011). The existence of dynamin–EHD complexes with similar functions in other organisms is unknown.

The amoeba Dictyostelium discoideum is a professional phagocyte. Genetic, microscopy and biochemical analyses have allowed the fine characterization of the endocytic and phagocytic pathways in these cells (Neuhaus et al., 2002). Similar to those in animal phagocytes, D. discoideum nascent phagosomes containing internalized material undergo a series of maturation events that mainly comprise extensive retrieval of non-phagosomal components (e.g. plasma membrane receptors) and fusion with endosomes and/or lysosomes to gain digestive properties (Gotthardt et al., 2002). A key step in phagosome maturation comprises the delivery of the vacuolar proton ATPase (H⁺-ATPase) and hydrolytic enzymes to phagosomes. Sixty minutes after uptake, retrieval of the H⁺-ATPase from phagosomes results in the formation of neutral post-phagolysosomes, which eventually fuse with the plasma membrane to release undigested material. Several key players have been shown to participate in membrane fusion and fission events that are essential for H⁺-ATPase cycling to phagosomes. Among those, Wiskott-Aldrich syndrome protein and SCAR Homolog (WASH), which drives actin polymerization, has been shown to be essential for H⁺-ATPase recycling (Carnell et al., 2011; Park et al., 2013). In addition, Dynamin A (DymA), a D. discoideum homolog of dynamin notably observed on post-lysosomes (Wienke et al., 1999), is also reported to participate in early steps of the phagosome maturation process (Gopaldass et al., 2012). Indeed, deletion of dymA impairs phagosome acidification and F-actin binding to early phagosomes. Although DymA-driven vesicle scission could account for its role in membrane recycling during phagosome maturation, DymA might also regulate the actin cytoskeleton, as suggested by the reported interaction between dynamin and WASH in mammalian cells (Derivery et al., 2009). Consistent with this hypothesis, in D. discoideum, DymA is reported to interact with the actin-binding protein Abp1, and the recruitment of Abp1 to early phagosomes requires the presence of DymA (Dieckmann et al., 2010; Gopaldass et al., 2012). Interestingly, previous studies in C. elegans have suggested that dynamin might contribute to Rab5 recruitment to phagosomes during phagocytosis of apoptotic cells (Kinchen et al., 2008). More recently, in C. elegans, dynamin has been shown to form a complex with clathrin, clathrin adaptor protein complex 2 (AP2), and LST-4 (the C. elegans homolog of mammalian proteins Snx9, Snx18 and Snx33) to promote phagosome maturation during apoptotic cell corpse clearance (Chen et al., 2013). Therefore, dynamin might have more than one function in phagocytosis. Whether other members of the dynamin superfamily are additionally involved in phagosome maturation is unknown. Because D. discoideum contains a single EHD protein (referred to as EHD), this phagocytic cell is an attractive model to study the possible relationship between dynamin and EHD proteins.

In this study, we explored the localization and function of the single D. discoideum EHD protein in the endocytic pathway. For the first time, we report that EHD participates in phagosome maturation. Moreover, we provide evidence that EHD and DymA play non-redundant functions in this cellular process. We also reveal that EHD contributes to the scission of endosomal tubules.

One EHD protein has been previously identified in D. discoideum (Daumke et al., 2007). To identify additional EHD proteins in D. discoideum, we searched the D. discoideum genome sequence database for genes that contain an EH domain (http://dictybase.org/tools/blast). Only three genes were identified, in contrast to the human genome that encodes eleven proteins with EH domains (Confalonieri and Di Fiore, 2002). The first gene, DDB_GO287325, encodes a protein (UniProtKB, Q54KI4) homologous to mammalian epidermal growth factor receptor substrate 15 (EPS15) with three copies of the EH domain in the N-terminal region, and has been previously described (Blanc et al., 2009). Next, the DDB_G0269636 (UniProtKB, Q55DJ6) gene product contains one EH domain in a central position, and was not identified in other species. The third gene, DDB_G0282233 (UniProtKB, Q54ST5), encodes a protein of 568 amino acids with a domain organization similar to that of mammalian EHD-family members and shares over 47% amino acid similarity with all human EHD proteins (EHD1–EHD4). Similar to EHD proteins in other species (Daumke et al., 2007), this D. discoideum protein comprises four domains: two helical domains, a G-domain that shares similarity with the dynamin G-domain and a C-terminal EH domain (Fig. 1A). Therefore, as described in C. elegans (Grant et al., 2001) and D. melanogaster (Olswang-Kutz et al., 2009), D. discoideum contains a single EHD protein, here referred to as EHD.

The crystal structure of mouse EHD2 has been solved and reveals that EHD proteins display a dynamin-like structural fold (Daumke et al., 2007). Based on its homology with mouse EHD2, the secondary structure assignment for D. discoideum EHD has been previously reported (Daumke et al., 2007). The G domain of mouse EHD2 comprises an unstructured loop that contains a KPFxxxNPF motif involved in EHD2 oligomerization (Daumke et al., 2007). D. discoideum EHD contains only the PF motif in this loop (Fig. 1A), like human EHD1 in which this motif is sufficient for oligomerization (Bahl et al., 2015). In mouse EHD2, an additional GPF motif has been identified in the linker region between the G domain and the EH domain and is predicted to bind to the EH domain of the opposing dimeric EHD2 molecule (Daumke et al., 2007). This motif is conserved in all mammalian EHD proteins but is replaced by an NPF motif in D. discoideum. Because NPF motifs are well described to have a high affinity for EH domains, this NPF motif in D. discoideum EHD might support EHD homodimerization.

EHD proteins in other species are known to interact with phosphoinositides (Naslavsky et al., 2007). As a first step towards the characterization of D. discoideum EHD, we thus assayed binding of EHD to lipids. Purified recombinant GST–EHD was incubated with a commercial nitrocellulose membrane that had been spotted with various phosphoinositides. EHD mainly interacted with phosphatidylinositol-3-phosphate [PI(3)P] and associated with other lipids to a lesser degree, including phosphatidylinositol-4-phosphate [PI(4)P], phosphatidylinositol-5-phosphate [PI(5)P], di- and tri-phosphorylated phosphatidylinositides (Fig. 1B). To confirm that EHD selectively binds to PI(3)P, we next performed liposome sedimentation assays with GST–EHD (Fig. 1C,D). In the conditions tested, about 42.5% of total GST–-EHD was observed in liposome pellets containing PI(3)P, whereas 20–25% of GST–EHD sedimented with liposomes containing PI(5)P and phosphatidylinositol-3,5-bisphosphate [PI(3,5)P₂], respectively. Binding to other phosphoinositides or liposomes composed of 30% phosphatidylserine (PS) was weaker. Increasing the concentration of PS to 100% pelleted about 70% of the total GST–EHD, whereas it did not bind to PS that had been immobilized onto nitrocellulose membranes (Fig. 1B). PS that has been incorporated into liposomes might thus better mimic PS topology in cellular membranes. No binding of GST alone to any phospholipid-containing liposomes was observed (Fig. 1C). Because PI(3)P is enriched on D. discoideum early endosomes (Clarke et al., 2010), the interaction with this phospholipid might ensure specific recruitment of EHD to these endomembranes, although other factors might also contribute to this localization. Interestingly, human EHD1 is reported to bind to PS in vitro, and this interaction is essential for human EHD1 recruitment to recycling endosomes (Lee et al., 2015); hence, PS might contribute in a similar way to EHD localization in D. discoideum.

Membrane recruitment of EHD in vivo was next confirmed by subcellular localization analysis. RFP-tagged EHD (RFP–EHD) was conditionally expressed in cells 24 h before confocal microscopy analysis of fixed cells (Fig. 1E,F). RFP–EHD was excluded from the plasma membrane and the contractile vacuole, as shown by the absence of colocalization with cortical actin and the bulk of the H⁺-ATPase, respectively. Conversely, RFP–EHD colocalized with recycling endosomes (p25 positive) (Charette and Cosson, 2006), lysosomes (H⁺-ATPase positive, p80 positive) and post-lysosomes (H⁺-ATPase negative, p80 positive, actin positive). Identical results were obtained with Myc-tagged EHD (Fig. S1A). The detection of EHD at these multiple locations suggests that EHD might participate in several steps along the endocytic pathway in D. discoideum. To determine the temporal localization of EHD on endosomes and evaluate its possible implication in phagocytosis, phagosomes at different stages of maturation were purified from wild-type (WT) cells, and the presence of endogenous EHD was assessed by western blotting with an antibody against EHD (Fig. 1G). EHD was detected in early phagosomes and rapidly dissociated from maturing phagolysosomes, which were characterized by the accumulation of the endosomal protein LmpA. Previous studies have revealed the transient localization of PI(3)P on early phagosomes, which was detected with the PI(3)P biosensor GFP–2FYVE (Clarke et al., 2010; Guetta et al., 2010) and the arrestin-related protein AdcA (Guetta et al., 2010). Although binding to PI(3)P might explain the temporal profile of EHD phagosomal recruitment, additional interactions with other proteins and/or lipids might also control EHD dynamics. In macrophages, PS is continuously maintained at a high level in phagosomes during the whole maturation process through fusion with PS-enriched endocytic compartments (Yeung et al., 2009). Whether such a constant PS level is observed in D. discoideum phagosomes is unknown and, if verified, it would not match EHD dynamics in phagosomes.

As a first step towards deciphering the role of EHD in the endocytic pathway of D. discoideum, a yeast two-hybrid screen was performed using the EH domain as a bait. Among identified preys (Table S1), DymA was isolated 43 times with a very high confidence score for the interaction. The domain of interaction between EHD and DymA was deduced from sequences of overlapping DymA preys and resided in a peptide of 33 residues (D548 to Q580) that contains an NPF motif recognized by EH domains. This short peptide is included in the Glu-Asp-Ser-rich region of DymA (QNS region), which showed no amino-acid sequence similarity with members of the dynamin family in any other species (Wienke et al., 1999). Thus, D. discoideum DymA appears to be the only dynamin homolog across species with an NPF motif. Although our yeast two-hybrid screen revealed other interesting candidates, we chose to focus on DymA mainly owing to its role in membrane trafficking and the reported interaction between EHD and dynamin in Lamprey (Jakobsson et al., 2011).

To verify the interaction between EHD and DymA, we first performed GST pulldown assays (Fig. 2A). Fusion proteins made between GST and EHD were immobilized onto glutathione agarose beads and incubated with D. discoideum cell lysates. Bound proteins were revealed by western blot with an anti-DymA antibody. Full-length EHD (GST–EHD) and the EH domain alone (GST–EH) interacted with DymA. In contrast, DymA was not detected using GST alone or EHD in which the EH domain had been deleted (GST–EHD-ΔEH), confirming that the EH domain of EHD is responsible for the interaction with DymA. This was further demonstrated by mutation of the NPF motif in GFP-tagged DymA (P573G,F574S double mutation, GFP–DymA-NGS mutant), which fully abolished DymA binding to GST–EHD (Fig. 2B). Next, the interaction between EHD and DymA was assessed by performing immunoprecipitation experiments. Lysates from cells overexpressing GFP–DymA were immunoprecipitated with beads coupled to an antibody against GFP. Immunoblotting of eluted beads with an antibody against EHD revealed the presence of EHD, consistent with an in vivo interaction between EHD and DymA (Fig. 2C). Co-precipitation between endogenous DymA and GFP-tagged EHD was also observed (Fig. S1B).

Taken together, these data suggest that EHD and DymA colocalize in endosomal compartments. To test this hypothesis, RFP–EHD and GFP–DymA were coexpressed in WT cells, and fixed cells were observed with confocal microscopy. As shown in Fig. 2Da, both proteins colocalized around vacuoles as well as in punctate structures in close vicinity to these endomembranes (indicated by arrowheads). Interestingly, despite the physical interaction between EHD and DymA revealed above, both proteins were also present on vacuoles independently of each other. Indeed, in the absence of DymA (dymA⁻ knockout cells), RFP–EHD still localized to p80- and p25-positive vacuoles (Fig. 2Db,Dc). Reciprocally, deletion of the gene encoding EHD (ehd⁻ cells were further characterized hereafter) did not modify GFP–DymA cellular distribution, although a slight diffuse cytoplasmic staining was also observed (Fig. 2Dd). Moreover, the GFP–DymA-NGS mutant, which does not interact with EHD (Fig. 2B), still correctly localized to p80-positive vacuoles when expressed in dymA⁻ cells (Fig. 2De).

To test for a possible co-recruitment of EHD and DymA to phagosomes, we next analyzed the dynamics of these proteins during phagosome maturation. Cells coexpressing RFP–EHD and GFP–DymA were incubated with latex beads and imaged by using time-lapse video microscopy (Fig. 3). As observed in Fig. 3 (see Movie 1) (Fig. 3B), RFP–EHD and GFP–DymA proteins appeared simultaneously and were recruited extremely rapidly to phagosomes, reaching a peak at about 5 min, and then concomitantly released during phagosome maturation. These similar dynamics of both proteins on maturing phagosomes might thus reflect coordinated functions of these two proteins.

To determine the function of EHD in the endocytic pathway and to further explore the possible interplay between EHD and DymA, we decided to inactivate the gene encoding EHD both in WT and dymA⁻ cells. Systematic morphological and functional analysis of the endocytic pathway did not reveal any major defect resulting from the absence of EHD in WT cells or aggravation of dymA⁻ phenotypes in dymA⁻ehd⁻ cells (Figs S1C–G,S2 and S3).

DymA has been previously shown to play a major role in maturation of early phagosomes (Gopaldass et al., 2012). The interaction of EHD with DymA prompted us to test whether EHD might share similar functions. We first measured proteolytic activity inside phagosomes in living cells. Cells were fed with beads coupled to Alexa-Fluor-594 and to bovine serum albumin (BSA) labeled with DQ Green at a self-quenching concentration. In this assay, BSA proteolysis in phagosomes releases DQ Green and subsequent dequenched fluorescence emission is recorded (Sattler et al., 2013). We observed that protease activity was very decreased in ehd⁻ cells compared to that in WT cells (Fig. 4A). This defect was even more marked than that in dymA⁻ cells and was further increased in dymA⁻ehd⁻ double mutants. Measuring the activity of various intracellular hydrolases (Fig. S3E–H) failed to reveal any decrease between mutants and WT cells. On the contrary, in mutants lacking DymA, the overall activity was higher than that measured in WT cells. Therefore, the defect in phagosomal proteolytic activity is likely not to be the result of hydrolases being absent from the cells, but can be explained either by a suboptimal phagosomal pH or the mis-trafficking of the proteases that are normally delivered to maturing phagosomes.

An essential step in phagosome maturation is the rapid acidification of newly formed phagosomes, followed by their neutralization before exocytosis of undigested material (Maniak, 2003). Thus, the endosomal and phagosomal pH has been shown to reach values around 4 after 45–50 min, and reneutralize to about 6 after 120–180 min (Aubry et al., 1993; Maniak, 2003; Marchetti et al., 2009; Sattler et al., 2013). To determine whether this pH profile during phagosome maturation is affected in the absence of EHD, we monitored phagosomal pH as previously described (Sattler et al., 2013). Monolayers of cells in 96-well plates were fed with beads labeled with the pH-sensitive fluorochrome fluorescein isothiocyanate (FITC) and the pH-insensitive fluorochrome Alexa-Fluor-594. The fluorescent emissions of both fluorochromes were recorded with a plate reader, and relative pH kinetics were analyzed by plotting the ratios of FITC and Alexa-Fluor-594 fluorescence values as a function of time (Fig. 4B). The observed ratios in WT cells correspond to an initial acidification minimum of around pH 4 that was reached after 53 min, followed by a reneutralization over the next 3 h (Fig. 4B). In ehd⁻ cells, this profile was substantially different, because the population of maturing phagosomes was less acidic, reaching its minimum only after 63 min, but the kinetics of reneutralization were similar to those in WT cells. Interestingly, this pH profile was less affected than that in dymA⁻ cells (Fig. 4B) (Gopaldass et al., 2012). However, the absence of both EHD and DymA proteins resulted in a dramatic decrease in the rate and extent of acidification because a minimum similar to that measured in ehd⁻ cells was reached after 149 min, and full neutralization did not occur even after 4 h.

The cumulative defects in phagosomal proteolysis and pH kinetics in dymA⁻ehd⁻ double mutants suggested that DymA and EHD play non-redundant functions in these two processes. This hypothesis was confirmed by the fact that overexpression of GFP–DymA in ehd⁻ cells did not restore normal phagosomal proteolysis and the normal pH profile but, instead, worsened these defects to an extent even more dramatic than that observed in dymA⁻ehd⁻ cells (Fig. 4C,D). Synthetic aggravation of a phenotype in a mutant cell through overexpression of another wild-type gene is widely exploited in genetic screens and usually indicates that two gene products function in the same pathway or associate in a protein complex (Prelich, 2012). The enhanced inhibition observed here is thus consistent with DymA and EHD co-functioning during phagosome maturation. Next, to determine the role of the interaction between DymA and EHD, we assessed whether a DymA mutant that was unable to bind to EHD (GFP–DymA-NGS) could rescue defects in dymA⁻ cells. Phagosomal proteolysis and pH profile analyses revealed that both GFP–DymA-NGS and GFP–DymA rescued the dymA⁻ phenotype to equal extents (Fig. 4E,F). Thus, the physical interaction between DymA and EHD appears to be dispensable for DymA functions in the phagosomal maturation processes analyzed, although we cannot formally exclude that the DymA–EHD interaction, undetected under our experimental conditions, might still occur independently of the NPF motif in DymA.

Sequential delivery and recycling of the vacuolar H⁺-ATPase to and from phagosomes accounts for the acidification and neutralization steps accompanying maturation (Carnell et al., 2011; Clarke et al., 2010). Thus, impaired acidification of phagosomes in ehd⁻ cells might be due to either defective H⁺-ATPase targeting to phagosomes or faster retrieval from phagosomes. To distinguish between these two possibilities, we next analyzed the dynamics of GFP-tagged M subunit of the H⁺-ATPase (GFP–VatM) during maturation of phagosomes with time-lapse video microscopy (Fig. 5). We observed that GFP–VatM delivery to phagosomes in ehd⁻ and WT cells occurred at similar rates. In contrast, GFP–VatM retrieval from phagosomes had already started, on average, 15 min after delivery in ehd⁻ cells, instead of at 25 min in WT cells, and was considerably faster in mutant cells (Fig. 5). These results strongly suggest that premature and faster rates of H⁺-ATPase retrieval from phagosomes might explain the abnormal pH profile observed during phagosome maturation in the absence of EHD.

To determine whether the defect in the dynamics of phagosomal H⁺-ATPase transport reflects a more general defect in phagosome maturation in ehd⁻ cells, we isolated phagosomes from WT and ehd⁻ cells at different steps of the maturation process. The temporal profiles of several phagosomal proteins were then monitored by western blotting analysis (Fig. 6). As previously reported (Gopaldass et al., 2012), DymA and MyoB mainly associated with nascent and newly formed phagosomes, and were rapidly dissociated during phagosome maturation in WT cells (Fig. 6). In the absence of EHD, DymA and MyoB were still efficiently recruited onto nascent phagosomes, further confirming EHD-independent membrane recruitment of DymA, as first suggested by the results of confocal microscopy studies (Fig. 2D). In contrast, DymA and MyoB were more rapidly released from maturing phagosomes in ehd⁻ cells than in WT cells. Because it has been suggested that MyoB and DymA control the recycling of the H⁺-ATPase from phagosomes (Gopaldass et al., 2012), the early departure of MyoB and DymA from phagosomes in ehd⁻ cells might thus explain the accelerated retrieval of the H⁺-ATPase and the aberrant acidification observed in ehd⁻ cells (Figs 4B and 5). Surprisingly, EHD showed similar release from phagosomes in dymA⁻ and WT cells (Fig. S3I), indicating that the transport profile of EHD is not affected by the absence of DymA. Note that deletion of dymA has been reported to result in massive recruitment of MyoB to maturing phagosomes (Gopaldass et al., 2012) in contrast to the MyoB early release observed here in ehd⁻ cells.

Interestingly, analysis of proteins recruited during late stages of phagosome maturation (Fig. 6) revealed earlier phagosomal delivery of the lysosomal hydrolase Cathepsin D in ehd⁻ cells, whereas the delivery of LmpA was very similar in ehd⁻ and WT cells (although earlier delivery of LmpA was noticed in some experiments). Several delivery and recycling steps occur during phagosome maturation, and notably different subsets of hydrolases are delivered in this process. Accordingly, LmpA and Cathepsin D display different temporal delivery profiles, indicating that these proteins are delivered through fusion with distinct endolysosomal populations (Gotthardt et al., 2002). Hence, the heterogeneity of lysosomes might explain why Cathepsin D and LmpA delivery to phagosomes were not affected to the same extent in the absence of EHD. Taken together, these results indicate that the overall phagosome maturation process is accelerated in the absence of EHD. Thus, EHD might control the appropriate timing of sequential molecular events (e.g. vesicle scission) during phagosome maturation.

DymA has been previously shown to control the scission of tubules formed at the surface of endosomes (Gopaldass et al., 2012). Because Dynamin and EHD proteins share structural and functional features (Daumke et al., 2007), we next assessed the interplay between DymA and EHD in the dynamics of endosomal tubules. Cells were incubated overnight in culture medium containing tetramethylrhodamine B isothiocyanate (TRITC)–dextran to load all endo-lysosomal compartments and observed by using spinning disk video microscopy. Because endosomal tubules are highly dynamic structures, a high frame rate of one image per second, limiting the observation to a single focal plane, was used. In addition, to allow the observation of a high number of tubules in one focal plane, cells were flattened by overlaying them with a thin agar sheet. As previously described (Gopaldass et al., 2012), dymA⁻ cells showed an increased number of tubules compared to WT cells, and these tubules were longer and persisted for extended periods of time (Fig. 7A–D; Movies 4 and 5). Remarkably, similar defects were also observed in ehd⁻ cells (Fig. 7A–D; Movie 6). These alterations were not intensified upon deletion of both dymA and ehd genes (Fig. 7A–D; Movie 7). The absence of cumulative defects in dymA⁻ehd⁻ cells suggested that DymA and EHD might share similar functions in the scission of endosomal tubules. Note that tubules emanating from Lucifer-Yellow-filled pinosomes were labeled by RFP–EHD (Fig. 7E; Movie 8), in agreement with a role of EHD in endosomal tubule dynamics. Finally, transmission electron microscopy analysis of endosomal tubules did not reveal any marked ultrastructural differences in the tubular structures observed in WT and mutant cells (Fig. S4).

In this study, we report the function of the single EHD protein expressed in the model organism D. discoideum. We provide the first evidence that EHD regulates phagosome maturation and endosomal tubule formation. Dynamin (DymA) has previously been shown to control these two processes (Gopaldass et al., 2012). We therefore analyzed the functional relationship between EHD and DymA. We demonstrate here that these proteins physically interact and play non-redundant independent functions in phagocytosis.

Making use of several protein–protein interaction assays, we reveal here a physical interaction between DymA and EHD. This observation is in agreement with the initial report of a direct binding between EHD and dynamin in Lamprey axons (Jakobsson et al., 2011), and suggests that this interaction might also exist in other organisms. What are the biological consequences of the formation of DymA–EHD complexes? We first examined whether this interaction might control the membrane association of both DymA and EHD. Here, we demonstrated that DymA and EHD are recruited to endosomal membranes independently of each other. Indeed, both proteins are properly localized in the absence of the other. Membrane recruitment of both DymA and EHD might occur through direct binding to phosphoinositides. Indeed, DymA has been shown to bind to several phosphoinositides including PI(3)P, PI(4)P, PI(5)P, PI(3,5)P₂, phosphatidylinositol-4,5-bisphosphate [PI(4,5)P₂] and phosphatidylinositol-3,4,5-trisphosphate (Klockow et al., 2002). Although EHD bound to the same variety of phosphoinositides as DymA, EHD displayed a stronger interaction with PI(3)P (Fig. 1B–D), and this selective binding might thus explain its recruitment to early phagosomes. Still, the temporal profile of EHD recruitment to phagosomes only approximately matched the reported dynamics of PI(3)P on early phagosomes. Indeed, PI(3)P labeling is lost a few minutes after phagosome closure (Clarke et al., 2010), whereas it took about 20 min to observe a 50% reduction in EHD presence on phagosomes (Fig. 1G). This implies that other proteins might also control EHD membrane recruitment. Taken together, these results suggest that DymA–EHD complexes might attach to membranes through either of the proteins (or even both), or through the interaction with other proteins such as Abp1, which has been shown to bind to DymA with high affinity (Dieckmann et al., 2010). The yeast two-hybrid candidate interaction between EHD and AbpF (Table S1) might support this hypothesis. Further time-lapse video microscopy studies revealed that DymA and EHD are simultaneously recruited and retrieved during phagosome maturation. Although technical limitations (e.g. rate of acquisition) might prevent assessment of the actual sequence of events, this result strongly suggests that DymA and EHD might be recruited together to phagosomes and, afterwards, simultaneously released from membranes, thus ruling out any clear control of membrane attachment or release by either protein.

Next, we examined whether the DymA–EHD interaction ensures a possible functional interdependence between these proteins. Analysis of ehd⁻ cells revealed a role of EHD in phagosome maturation. Because DymA has also been shown to control this process (Gopaldass et al., 2012), the physical interaction between DymA and EHD indicates possible interconnected functions shared between these proteins. Contrary to this hypothesis, several lines of evidence demonstrate that DymA and EHD display non-redundant and independent functions in phagosome maturation. Indeed, analysis of the intraphagosomal pH profile, the dynamics of the vacuolar H⁺-ATPase and the content of key proteins in phagosomes revealed that ehd⁻ cells showed faster maturation of phagosomes compared to WT cells (Figs 4–6). Accordingly, ehd⁻ cells exhibited faster release of DymA and MyoB from maturing phagosomes (Fig. 6). In contrast, dymA deletion was reported to inhibit early stages of phagosome maturation resulting in phagosome acidification defects (Gopaldass et al., 2012). This premature and fast phagosome maturation in the absence of EHD suggests that the actual function of EHD is to regulate key steps in phagosome maturation. Finally, we showed that DymA could not substitute for EHD (Fig. 4C,D) and that EHD binding to DymA is not required for DymA functions because mutation of the NPF motif in DymA impaired EHD binding without any functional defects (Figs 2B and 4E,F). Taken together, these results indicate no apparent functional connections between these proteins in phagosome maturation despite their physical interaction.

What is the actual function of EHD in phagosome maturation? Phagosomes undergo a series of finely tuned vesicular fusion and fission events during maturation (Fairn and Grinstein, 2012; Gotthardt et al., 2002). Therefore, EHD might control the timely delivery and/or retrieval of key proteins involved in phagosome maturation. Owing to inferred membrane remodeling properties typically observed for EHDs in other organisms, D. discoideum EHD might be directly involved in the budding of vesicles that recycle proteins from maturing phagosomes. Alternatively, EHD might play an indirect role by controlling the function of EHD-binding partners that regulate the retrieval of cargo from phagosomes. Several protein complexes, including CORVET and HOPS tethering complexes and retromer, are thought to contribute to phagosomal protein recycling (Fairn and Grinstein, 2012). For instance, the retromer complex is reported to participate in phagocytosis of apoptotic cells in C. elegans (Chen et al., 2010). Interestingly, EHD1 has been shown to regulate endosomal retrograde transport mediated by the retromer complex through its interaction with rabankyrin-5 (Zhang et al., 2012c). Accordingly, we hypothesize that D. discoideum EHD controls retromer-mediated retrieval through interaction with an as-yet unidentified functional homolog of Rank-5. Furthermore, our yeast two-hybrid screen suggests that EHD might interact with a protein that is similar to human VPS33B-interacting protein (Table S1). The VPS33B-interacting protein (also known as VIPAS39, C14orf133, SPE-39, VIPAR and VPS16B) is part of the CORVET complex and regulates lysosomal delivery (Zhu et al., 2009). Interestingly, mutation in the full-of-bacteria (fob) gene, which encodes Drosophila VPS16B, results in defects in the fusion of phagosomes with lysosomes (Akbar et al., 2011). Thus, the hypothetical role of EHD in the regulation of the D. discoideum homolog of VPS16B and CORVET during phagosome maturation might be worth exploring.

Besides its hypothetical role in retromer-dependent retrieval, EHD might regulate other protein recycling mechanisms that are at play during phagosome maturation. For instance, beyond their role in clathrin-independent endocytosis, new trafficking functions have emerged for flotillins (Meister and Tikkanen, 2014). Remarkably, flotillins have been shown to regulate sorting and recycling of transferrin receptor and E-cadherin from tubulovesicular recycling compartments that are positive for Rab11a, SNX4 and EHD1 back to the plasma membrane (Solis et al., 2013). Although flotillins are recruited onto mature phagosomes both in mammals and D. discoideum (Gotthardt et al., 2006; Jenne et al., 1998), their role in phagosome maturation has never been reported. Owing to the unique sorting functions of flotillins, the interplay between flotillins and EHD might be an interesting future research avenue.

Sorting and transport of cargo from endocytic recycling compartments back to the plasma membrane mainly relies on tubular membranes budding from these endosomes (Maxfield and McGraw, 2004). Mammalian EHDs are reported to either contribute to the formation of these endosomal tubules or to control their scission (Naslavsky and Caplan, 2011). Here, we observed that the length and the time of persistence of endosomal tubules were increased in EHD-depleted cells (Fig. 7). Furthermore, although RFP–EHD localized to the length of short-lived tubules (Fig. 7E), overexpression of RFP–EHD had no major effects on endosomal tubule formation. These results are consistent with a mechanistic role of D. discoideum EHD in vesiculation rather than in tubulation of endosomes. What is the molecular basis for this functional specificity? As previously proposed for human EHD1 and EHD3, which show opposite vesiculation and tubulation functions (Cai et al., 2013), EHD-mediated vesiculation in D. discoideum might originate from the three-dimensional structure of EHD that is likely to favor vesiculation rather than tubulation. This propensity for vesiculation might be also linked to EHD putative binding partners that provide this feature. No EHD-binding proteins with known membrane tubulation activities were identified in our yeast two-hybrid screen. Although additional experiments would be required to confirm this result, the absence of D. discoideum EHD-binding partners with membrane-bending functions might explain its restricted role in vesiculation. Finally, it is unknown whether identical mechanisms are employed for the biogenesis of tubules and vesicles emanating from phagosomes and endosomes. Our results indicate that both DymA and EHD are shared elements for membrane remodeling activities in these different compartments. However, we reveal here that DymA and EHD play specific roles in phagosome maturation, whereas they might have redundant functions in endosomal vesiculation. The molecular basis of these possible differences will have to be further dissected.

D. discoideum strain AX2 was grown at 22°C in HL5c medium. The laboratory strain of Klebsiella pneumoniae was also used to grow D. discoideum, as reported previously (Alibaud et al., 2008). Antibodies used in this study were mouse monoclonal antibodies against p80 (H161, not commercial; dilution 1:800) (Ravanel et al., 2001), p25 (H72, not commercial; undiluted hybridoma culture supernatant) (Mercanti et al., 2006), vacuolar H⁺-ATPase (221-35-2, not commercial; undiluted hybridoma culture supernatant) (Neuhaus et al., 1998) and rabbit polyclonal antibodies against Alix (1:500; Mattei et al., 2005), CatD (1:2000; Journet et al., 1999), DymA (1:2000; Wienke et al., 1999), EHD (anti-p64) (1:5000; Dias et al., 2013), LmpA (1:5000; Karakesisoglou et al., 1999) and MyoB (1:500; Novak et al., 1995). For immunofluorescence analysis, cells were fixed and treated as detailed previously in Dias et al. (2013). The Alexa-Fluor-647-coupled anti-p80 antibody was a kind gift from Pierre Cosson (Centre Médical Universitaire, Geneva, Switzerland). Actin was stained with TRITC-labeled phalloidin (Sigma-Aldrich). Cells were observed with confocal microscopy (Zeiss LSM 510). For nuclei staining, fixed cells were incubated with DAPI for 30 min and observed with a Zeiss AxioImager Z1 photomicroscope. Internalization of fluid phase (Alexa-Fluor-647–dextran; Molecular Probes) or phagocytic particles (YG-fluorescent 1-µm latex beads; Polysciences) was performed as previously described (Lima et al., 2012) and analyzed by performing flow cytometry (FACS Calibur, Becton Dickinson). All flow cytometry data were normalized to protein content.

The EHD coding sequence was cloned into an extrachromosomal inducible expression vector (Veltman et al., 2009). After electroporation, transfectants were selected and maintained in 25 μg/ml hygromycin. Expression was induced by adding 10 μg/ml doxycycline 24 h before analysis. Plasmids encoding VatM–GFP (Clarke et al., 2002) or GFP–DymA (Wienke et al., 1999) were transfected into cells by electroporation, as reported previously (Alibaud et al., 2003). Mutant GFP–DymA-NGS (mutations P573G,F574S) was obtained by PCR mutagenesis, sequenced and cloned into the GFP–DymA expression vector. To obtain the ehd (dictyBase ID, DDB_G0282283) knockout vector, the 5′ fragment was amplified from genomic DNA with sense (5ʹ-GGATCCATGAAGAAACTTAATGTTCAAGAACAA-3ʹ) and antisense (5ʹ AAGCTTTTATCAATACAATTCTAATCTTTTCATCAT-3ʹ) oligonucleotides, and cloned into pBlueScript vector (Stratagene). The 3′ fragment was obtained with PCR using sense (5′-AAGCTTTAATACGATTGGTTCATTCTGGAGTGGACC-3ʹ) and antisense (5′-CTCGAGTTATTGTTGTTGTTGTTGTTGTTGTTGTTG-3ʹ) oligonucleotides, and cloned into pBlueScript containing the 5′ fragment. After sequencing, the knockout vector was completed by inserting the blasticidin or neomycin resistance cassettes between the two 5′ and 3′ fragments. The resulting plasmids were linearized through digestion with restriction enzymes (KpnI and NotI) and electroporated into AX2 or dymA⁻ cells. Transformants were selected in the presence of 10 µg/ml blasticidin and/or 10 µg/ml G418. Individual colonies were tested by performing PCR, and the absence of expression of EHD was verified by western blotting with an anti-EHD polyclonal antibody.

Lipid dot blot (Echelon Biosciences, Salt Lake City, UT), GST–EHD fusion protein production in bacteria and GST pulldown assays were performed as described previously (Blanc et al., 2009). Anti-GFP immunoprecipitations were performed using the commercial GFP-Trap system (ChromoTek GmbH, Martinsried, Germany) according to the manufacturer's instructions. Lipids were purchased from Avanti Polar Lipids (Alabaster, AL, USA) – egg L-α-phosphatidylcholine (EPC); 1,2-dioleoyl-sn-glycero-3-phospho-L-serine (DOPS); 1,2-dioleoyl-sn-glycero-3-phospho-(1′-myo-inositol-3′-phosphate) [PI(3)P]; 1,2-dioleoyl-sn-glycero-4-phospho-(1′-myo-inositol-4′-phosphate) [PI(4)P]; 1,2-dioleoyl-sn-glycero-5-phospho-(1′-myo-inositol-5′-phosphate) [PI(5)P]; 1,2-dioleoyl-sn-glycero-3-phospho-(1′-myo-inositol-3′,5′-bisphosphate) [P(3,5)P₂]; and 1,2-dioleoyl-sn-glycero-3-phospho-(1′-myo-inositol-4′,5′-bisphosphate) [PI(4,5)P₂]. Small unilamellar vesicles (SUVs) were prepared as described previously (Daumke et al., 2007). For binding assays, 1 µM GST–EHD and 1.5 mg/ml SUVs were incubated in liposome buffer (20 mM HEPES pH 7.2, 150 mM NaCl, 1 mM MgCl₂) for 1 h at room temperature. SUVs were then centrifuged at 20,000 g for 20 min at room temperature; supernatant and pellet fractions were recovered and analyzed by performing SDS-PAGE electrophoresis.

Glycosidase activities in cell pellets and extracellular fluid were assessed as previously described (Froquet et al., 2008; Le Coadic et al., 2013). For phagosomal pH measurement, 3-µm carboxylated silica beads coupled to the pH-sensitive fluorophore FITC and to the pH-insensitive fluorophore Alexa-Fluor-594 were used. For phagosomal proteolysis activity analysis, 3-µm carboxylated silica beads coupled to Alexa-Fluor-594 and to BSA labeled with DQ Green at a self-quenching concentration were used. Upon BSA digestion, DQ Green fluorescence increases owing to dequenching. Phagosomal pH and proteolysis activity measurements were performed as described previously (Sattler et al., 2013) using a fluorescence plate reader (Synergy Mx, Biotek, Bad Friedrichshall, Germany). Phagosomes were isolated at different times of maturation, as previously described (Dieckmann et al., 2008). Immunoblots were performed as described previously (Gotthardt et al., 2006), and band intensities were measured using the ImageJ software. For each temporal profile, intensities were normalized by setting the maximal signal obtained to 100%.

Cells were plated at 30% confluence in a 35-mm iBIDI dish and incubated overnight in HL5c medium. 30 min before imaging, HL5c medium was replaced with low fluorescence medium (LoFlo medium). Then, 3-µm latex beads were washed three times in HL5c and added to the cells at a ratio of two beads per cell. To initiate bead uptake, beads were centrifuged onto the adhered cells for 1 to 2 min at 500 g at 4°C. Excess medium was removed, and a 1-mm thin 2% agar sheet was placed on top of the cells in order to improve imaging (Sattler et al., 2013). Unless otherwise mentioned, images were then acquired every 30 s to 1 min with a spinning disk confocal system (Intelligent Imaging Innovations, Göttingen, Germany) mounted on an inverted microscope (Leica DMIRE2; Leica, Heerbrugg, Switzerland). To visualize tubules, cells were incubated overnight in 0.5 mg/ml of TRITC-labeled dextran. Cells were then overlaid with a 1-mm thin agar sheet, imaged every second for 3 min as described above. Images were processed using the PureDenoise plugin of the ImageJ software.

Cells were fixed for 1 h in 2% glutaraldehyde in HL5c culture medium and then stained for 30 min in 2% osmium with 0.1 M imidazole, as described previously (Barisch et al., 2015). After staining, cells were washed with PBS. Samples were then dehydrated, embedded in Epon resin and processed for conventional electron microscopy. Grids were examined with a Tecnai transmission electron microscope (FEI, Eindhoven, The Netherlands).

We thank Dr Jason King (University of Sheffield, UK) for critical reading of the manuscript.