Myosin Ib modulates the morphology and the protein transport within multi-vesicular sorting endosomes

Internalised molecules are sorted from resident proteins and lipids along the endocytic pathway, through a complex series of regulated processes. Besides the well-characterized signal motifs, micro-domains on membranes play an important role for efficient sorting of proteins and lipids (Hopkins et al., 1994; Raiborg et al., 2003; Stoorvogel et al., 1987). The transferrin receptor accumulates in tubules before being recycled to the plasma membrane, while the EGF receptor is sequestered in membrane invaginations that give rise to internal vesicles of multi-vesicular endosomes (MVEs, also named multi-vesicular bodies) before its degradation in lysosomes (Hopkins et al., 1994; Raiborg et al., 2003). Compelling evidence from several laboratories indicate that microtubules and associated molecular motors of the dynein and kinesin families are involved in the tubulation of organelles and thereby might contribute to sorting events (Koster et al., 2003; Roux et al., 2002). The participation of actin networks and myosins in different steps of the secretory and endocytic pathways were also investigated but the role of these networks and their associated motors as mechanical devices for sorting processes has only been studied recently.

The acto-myosin generated forces could help to cluster the cargos prior to their transport. For example, the class VI myosin, which binds adaptors proteins containing PDZ domains, such as Dab2 or GIPC, has been proposed to cluster receptors with PDZ-binding motifs in clathrin-coated regions (Hasson, 2003). Acto-myosin tension might also drive morphological changes that could contribute to the formation of micro-domains involved in the sorting process. In this context, in Dictyostelium discoidum the knockout of two genes encoding class I myosins, myoA and myoB, leads to a disordered morphology of sorting recycling endosomes (Neuhaus and Soldati, 2000).

We have previously reported that another myosin from class I, myosin Ib (Myo1b, also named myosin I alpha or Myr1), participates in the cytoplasmic distribution of endosomes and lysosomes and in the delivery of internalised molecules to lysosomes in mammalian cells (Cordonnier et al., 2001; Raposo et al., 1999). Despite the involvement of this myosin in the traffic along the endocytic pathway, Myo1b is unable to move endocytic compartments on actin filaments in vivo (Cordonnier et al., 2001). However, Myo1b, together with actin, might drive morphological changes of the endocytic compartments, similarly to myoA and myoB in Dictyostelium discoidum. It might also act with actin as a dynamic scaffold and be involved in the sorting of specific cargos.

To investigate further the mechanism by which Myo1b contributes to the traffic of internalized molecules towards the lysosomes, we used the pigmented human melanoma cell line MNT-1, which displays a well defined endocytic system. In these cells, endosomes that were traced with gold-labelled bovine serum albumin 15 minutes post uptake and displayed bilayered clathrin coats and internal vesicles, are intermediates for proteins transported to lysosomes and for proteins involved in the biogenesis of melanosomes (Raposo and Marks, 2002; Raposo et al., 2001). These endosomes, which had a defined structure and were involved in targeting of proteins to two different organelles, present the hallmark of sorting endosomes (Raiborg et al., 2003; Raposo and Marks, 2002; Raposo et al., 2001; Sachse et al., 2002). We will refer to these compartments as early sorting MVEs in the present study. We report here that overexpression of Myo1b or the inhibition of actin dynamic by drugs affect the morphology of these early sorting MVEs.

One peculiarity of the MNT-1 cellular model is the expression of Pmel17, a melanosome-specific protein (and the product of murine silver locus, also named GP100) that transits in these early sorting MVEs. After its biosynthesis Pmel17 is delivered to early sorting MVEs where, concomitantly or after its delivery, it is cleaved by a furin-like proprotein convertase (Berson et al., 2001; Berson et al., 2003). This cleavage leads to the formation of intra-lumenal fibrils that characterizes stage-II premelanosomes. Using this unique feature of the MNT-1 model, we investigated the contribution of Myo1b in the trafficking and the processing of Pmel17 in the early sorting MVEs. We report here that, the overexpression of Myo1b delays the processing of Pmel17 and thereby impairs the formation of the intra-lumenal fibrils in stage-II premelanosomes, and that Myo1b associated with endosomes co-immunoprecipitates with Pmel17. All together, these observations indicate that Myo1b and actin are involved in the morphological organisation of the early sorting MVE and in the transfer of cargo proteins such as Pmel17 towards internal vesicles of these endosomes.

Monoclonal antibodies HMB50 and HMB45 directed against the N-terminus of Pmel17 were obtained from Neomarkers (Fremont, CA). Monoclonal antibody directed against early endosome antigen 1 (EEA1) was obtained from BD Biosciences (San Diego, CA), monoclonal antibody directed against green fluorescent protein (GFP) from Molecular Probes (Eugene, OR) and monoclonal anti-tubulin from Amersham (Bioscience Europe GmbH). Polyclonal anti-Rab5 antibodies were a generous gift from M. Zerial (Max-Plank-Institute, Dresden, Germany), anti-Rab7 antibodies a generous gift from P. Chavrier (Institut Curie, Paris, France), αpep13h antibodies (Berson et al., 2001) a generous gift from M. Marks (University of Pennsylvania, Philadelphia, PA), anti-β-actin antibodies a generous gift from C. Chaponnier (Geneva University, Geneva, Switzerland). Two rabbit polyclonal antibodies directed against Myo1b were generated against peptides with the sequence (CSFRTVEAKQEKVSTTLNVAQAT) (SE7995) as previously described by Tang and Ostap (Tang and Ostap, 2001) and the C-terminal sequence (CKRKNNRLLEVAVP) (SE5237).

Cells of the human melanoma MNT-1 cell line were grown in MNT-1 culture medium, consisting of Dulbecco's modifies Eagle's medium (DMEM, Gibco-BRL Life Technology, Paisley, UK) supplemented with 10% AIMV medium (Gibco-BRL Life Technology), 20% foetal calf serum (Gibco-BRL Life Technology), penicillin (10 units/ml) and streptomycin (10 mg/ml) (Gibco-BRL Life Technology). Transfection with the recombinant DNA constructs encoding EGFP, EGFP-Myo1b and EGFP-Myo1b-Tail (Raposo et al., 1999), and isolation of stable transformants were performed as described by Durrbach et al. (Durrbach et al., 2000). Three cellular clones producing EGFP (EGFP cells), three cellular clones producing EGFP-Myo1b (EGFP-Myo1b cells), and four cellular clones producing EGFP-Myo1b-Tail (EGFP-Myo1b-Tail cells) were selected and permanently grown in the MNT-1 culture medium in which penicillin-streptomycin was replaced by 0.7 mg/ml Geneticin (Gibco-BRL Life Technology). These cellular clones were treated overnight with 10 mM sodium butyrate (Sigma-Aldrich, France) before experiments.

HeLa cells were maintained in DMEM supplemented with antibiotics. Stable transfectants were generated with pCI-Pmel17 (Berson et al., 2001; Berson et al., 2003) after transfection using FuGene6 (Roche Biochemicals, Indianapolis, IN) as previously described (Marks et al., 1995).

Wild-type-expressing, EGFP-expressing and EGFP-Myo1b-expressing cells were grown 4 days in a 150 cm² flask, washed with 5 ml of homogenisation buffer (HB) containing 50 mM imidazole pH 7.4, 0.25 M sucrose, 1 mM EDTA, 0.5 mM EGTA, 5 mM MgSO₄, 1 mM DTT, 150 μg/ml casein from bovine milk and protease inhibitors (Sigma Aldrich, France). They were then scraped off into 5 ml HB and homogenized by passing them twice through a cell cracker. Post-nuclear supernatant (PNS) collected after a 7-minute centrifugation at 600 g was then centrifuged for 5 minutes at 2500 g to separate the membrane-rich fraction and the cytosol (S1) from melanosomes. A membrane-rich fraction (P2) was collected after centrifugation of the supernatant (S1) for 20 minutes at 100,000 g.

To isolate magnetic endosomes, MNT-1 cells that had been grown 4 days in a 300 cm² flask, were incubated with 1 mM of Fe nanoparticles (CoFe₂O₄) coated with negatively charged citrate ligand for 1 hour at 4°C. They were then chased for 30 minutes at 37°C as described by Wilhelm et al. (Wilhelm et al., 2002). Cells were homogenized as described above, and PNS was loaded onto a magnetic column (5 ml/high magnetic field gradient). The non-magnetic fraction (NMF) was eluted first. The magnetic fraction (MF) was collected in the absence of magnet, and washed and magnetized twice more in 5 ml HB before its final collection in 200 μl HB.

Cells were harvested, starved of methionine and cysteine, pulse-labelled for 10 minutes with [³⁵S]methionine-cysteine labelling mix (200 μCi), chased with excess methionine and cysteine for indicated periods of time and washed with PBS. After chase incubation, culture medium was collected and clarified by centrifugation.

Cells that had been grown for 4 days and [³⁵S]methionine-cysteine metabolically labelled or not, were scraped off and lysed on ice in 50 mM Tris pH 7.5, containing 300 mM NaCl, 1 mM EDTA (lysis buffer) and 1% Triton X-100. The cell lysate was cleared by centrifugation for 20 min at 20,000 g, and by absorption with 25 μl of protein G sepharose beads (Amersham Bioscience Europe GmbH). The immuno-complex formed overnight (by incubating the pre-cleared cell lysate with 5 μg of purified antibody at 4°C) was absorbed by 25 μl protein G sepharose, washed 3× with lysis buffer containing 1% Triton X-100, 3×with lysis buffer containing 0.1% Triton X-100, analysed by SDS-PAGE and immunoblotting.

Proteins separated by SDS-PAGE were transferred to nitrocellulose membranes in the presence of 20 mM Tris, 150 mM glycine and 0.05% SDS. Detection of antibodies was performed using the chemiluminescence blotting substrate from Roche Applied Sciences (Meylan, France) or Super Signal West Pico Chemiluminescent substrate from Pierce Biotechnology (Rockford, IL, USA).

Cells were grown 4 days on cover-slips, fixed in PBS containing 3% paraformaldehyde and 0.025% glutaraldehyde, and permeabilized for 10 minutes in PBS containing 0.05% saponin. Cells were then incubated 30 minutes with primary antibodies, followed by a 30-minute incubation with FITC-conjugated or Cy3-conjugated antibodies (Rockland immunochemicals for research, Gilbertsville, PA 19525). Phalloidin (0.5 μg/ml) conjugated with TRITC (Sigma-Aldrich) was used to label F-actin. The cells were viewed with a confocal laser-scanning microscope (Leica, Vienna, Austria).

For conventional electronic microscopy, EGFP-Myo1b cells grown 4 days on cover slips were fixed with 2.5% glutaraldehyde in 0.066M cacodylate buffer for 3 hours at room temperature. After several washes with cacodylate buffer, cells were post-fixed with 2% OsO4 for 45 minutes, dehydrated with increasing concentrations of ethanol and embedded in Epon™. Ultra-thin sections were counterstained with 4% uranyl acetate in methanol for 5 minutes.

Ultra-cryomicrotomy and immunogold labelling was performed as previously described (Raposo et al., 1997). Cells were fixed with a mixture of 2% PFA and 0.125% glutaraldehyde for 3 hours at room temperature. Cells were pelleted, washed with 50 mM PBS-glycine and embedded in 10% gelatine at 37°C. After solidification on ice, small 1-mm³ blocks were prepared, infused in 2.3 M sucrose and frozen in liquid nitrogen. Ultra-thin cryosections were retrieved with a mixture of 2% methyl-cellulose–2.3 M sucrose (vol/vol), thawed and single or double immunogold labelling was performed. Antibodies were detected directly with protein-A gold-conjugates (purchased from Department of Cell Biology, Institute for Biomembranes, Utrecht, Netherlands). Sections were cut with UCT and FCS Leica microtomes, and viewed and photographed with a TEM Philips CM 120 (FEI company, Eindoven, The Netherlands). Negatives were scanned with a Rotatif scanner (Highscan-Eurocore).

Endogenous MYO1B was found to be concentrated in dynamic membrane areas of MNT-1 cells that were enriched with actin cytoskeleton as previously reported for other cell types (Raposo et al., 1999; Tang and Ostap, 2001) (Fig. 1A-C arrows). In addition, MYO1B was associated with punctate structures in the cytoplasm. Some of these structures co-distributed with EEA1 (early endosome antigen 1), one of the Rab5 effectors that is abundant in early endosomes (Fig. 1D-F, see arrows). Thus, consistent with our previous observation in the hepatoma cell line, a pool of MYO1B is associated to the endosomes of MNT-1 cells (Raposo et al., 1999).

To investigate the role of MYO1B in protein-trafficking along the endocytic pathway, we isolated a cellular clone derived from MNT-1 cells that expressed the EGFP-tagged rat Myo1b isoform c. The EGFP-Myo1b hybrid protein showed similar biochemical properties than the endogenous human MYO1B. Both were associated with cellular fractions enriched in membrane vesicles. Polyclonal anti-Myo1b antibodies raised against a specific peptide of the motor domain (Tang and Ostap, 2001) immunodetected a broad protein pattern with a size of 120 to 130 kDa in the membrane-vesicle-enriched fraction P2, isolated from EGFP-Myo1b cells (Fig. 2A). This protein-mobility pattern very probably corresponds to one or several MYO1B isoforms (Ruppert et al., 1993). Anti-Myo1b antibodies revealed also a protein of the size compatible with the size of the EGFP-Myo1b hybrid protein. Indeed, this protein was also detected by anti-GFP antibodies (data not shown). In agreement with previous observations (Tang and Ostap, 2001), MYO1B was resistant to extraction with Triton X-100 and required ATP in addition to detergent to partially dissociate from the membrane-vesicle-enriched fraction P2 (Fig. 2A). Similarly, EGFP-Myo1b dissociated from P2 after the addition of ATP to the detergent (Fig. 2A).

EGFP-Myo1b accumulated at the plasma membrane of MNT-1 cells in dynamic membrane areas that were enriched with actin cytoskeleton, such as membrane protrusions (Fig. 2B-D see arrow). EGFP-Myo1b was rarely associated with punctate structures in the cytoplasm and co-distribution with endosomes was difficult to determine at light-microscopic level. However, EGFP-Myo1b was detected at EM-level close to EEA1-labelled endosomes and in the magnetic fraction enriched in endosomes (Figs 3C and 9A). Furthermore, expression of EGFP-Myo1b in MNT-1 cells appeared to modify the distribution of the endocytic compartments labelled with anti-EEA1 antibodies. Endosomes accumulated in the perinuclear region in 73% of the cells expressing EGFP-Myo1b and only in 11% of the EGFP cells (Fig. 2, compare F and G with I and J, respectively).

We next analysed how expression of EGFP-Myo1b affected the endosomes of MNT-1 cells at the ultra-structural level. MNT-1 cells exhibit the different compartments of the endocytic pathway that consist of small tubulovesicular endosomes accessible for internalised molecules within 5 to 10 minutes, early sorting MVEs accessible for internalised molecules within 20 minutes, MVEs and also melanosomes at different stages of maturation (Raposo et al., 2001). As determined by conventional electron microscopy, the expression of EGFP-Myo1b in MNT-1 cells, resulted often in the accumulation in the perinuclear region of early sorting MVEs, with a characteristic electron-dense coat at their cytosolic side (Fig. 3A). Consistent with previous reports, immunogold-labelling on ultra-thin cryosections of EGFP-Myo1b cells revealed the presence of EEA1 in the non-coated regions of these early sorting MVEs (Fig. 3B, see arrows). Interestingly, small membrane extensions and vesicles in their vicinity appeared to be more abundant on endosomes from EGFP-Myo1b-expressing cells compared with those of wild-type cells (Fig. 3B versus 3E). EEA1-positive compartments that are redistributed upon EGFP-Myo1b overexpression (Fig. 2F) might correspond to clusters of early sorting MVEs, which display increased amounts of small protrusions (Fig. 3A,B).

We also analysed the distribution of EGFP-Myo1b at the EM level. Low amounts were detected under the above described experimental conditions. However, EGFP-Myo1b appeared restricted to small vesicle-like structures in the vicinity of the early sorting MVEs on ultra-thin cryosections (Fig. 3C). Some of these structures may correspond to cross sections of the small protrusions. To better evaluate the distribution of EGFP-Myo1b, we used whole-mount immunogold labelling on cells in which endosomes had been crosslinked with internalised HRP (Stoorvogel et al., 1996). This method has been used previously to detect Myo1b on endosomes of BWTG3 cells (Raposo et al., 1999). EGFP-Myo1b was detected on the vacuolar domain but appeared enriched on the small tubular membrane extension of the early sorting MVEs (Fig. 3D).

If Myo1b is driving morphological changes of these endocytic compartments then it might do this together with actin. Interestingly, EEA1-labelled endosomes that were observed at the light microscopic level co-distributed partially with actin filaments (Fig. 4A-C see arrows) and a consistent pool of actin was found in the close vicinity of these endosomes when they were observed at the electron microscopic level (Fig. 3E see long arrows). Particularly, actin was detected at the opposite side of the clathrin coated areas, where the small membrane vesicles and small tubular extensions were observed (Fig. 3E see small arrows). Furthermore, the treatment of MNT-1 cells with cytochalasin D induced the re-distribution of EEA1-labelled endosomes in the perinuclear region in close vicinity of actin patches (Fig. 4D,E,F see arrows) and facilitated the visualisation of the tubular extensions at the electron microscopic level (Fig. 3F see arrows).

These observations indicate that a small fraction of EGFP-Myo1b is associated with EEA1-labelled endosomes. Some of these endosomes are early sorting MVEs, as judged by electron microscopy. Furthermore, overexpression of EGFP-Myo1b, or cytochalasin D treatment impairs the cellular distribution and the morphology these compartments.

The delocalization and the morphological changes of the early sorting MVEs induced by the expression of EGFP-Myo1b may consequently perturb the transport of proteins that are passing through these endocytic compartments. Therefore, we analysed the cellular distribution and the maturation of Pmel17, which is processed in early sorting MVEs in their transition to stage-II melanosomes (Berson et al., 2001; Berson et al., 2003; Raposo et al., 2001). Intracellular compartments labelled with HMB50, an antibody that detects Pmel17, were dispersed from the cell periphery to the perinuclear region in all EGFP cells analysed, while they were mostly concentrated in the perinuclear region in 56% of EGFP-Myo1b cells (Fig. 5A versus 5B). However, consistent with a previous report (Raposo et al., 2001), the distribution of Pmel17 and specific markers of endosomes did not overlap when cells were examined through a light microscope. In that report, the majority of Pmel17 in MNT-1 cells was associated with premelanosomes and mature melanosomes (Raposo et al., 2001). Thus, the Pmel17-positive compartments, which were concentrated in the perinuclear region of EGFP-Myo1b cells, might correspond to premelanosomes and melanosomes. By contrast, it has been reported that the majority of Pmel17 is associated with a subset of late endosomes in HeLa cells transiently expressing Pmel17 (Berson et al., 2001). We isolated a cellular clone, derived from HeLa cells and expressing Pmel17. After 20 minutes of internalisation, Pmel17 co-distributed with EGF (data not shown). Interestingly, the transient expression of EGFP-Myo1b in HeLa cells expressing Pmel17 induced the perinuclear re-distribution of Pmel17 (Fig. 5 compare C with D). Overexpression of EGFP-Myo1b alters, therefore, the distribution of Pmel17 associated with endosomes in HeLa cells, and with mature and immature melanosomes in MNT-1 cells. We hypothesised that the disrupted localisation of Pmel17 associated with mature and immature melanosomes in MNT-1 cells is a consequence of the altered trafficking of Pmel17 in endosomes.

To test this hypothesis, we followed the processing of Pmel17 in MNT-1 cells. Pmel17 is an integral membrane protein. It is synthesised in the endoplasmic reticulum as precursor P1, O-linked-glycan- and N-linked-glycan-modified in the Golgi complex to result in the precursor P2, and cleaved in a post-Golgi pre-lysosome compartment into two disulfide-linked subunits. These subunits include a C-terminal 28 kDa Mβ subunit that contains part of the lumenal domain, the entire transmembrane and cytoplasmic domains and an 80 kDa Mα subunit with the remaining N-terminal region of the lumenal domain (Fig. 6A) (Berson et al., 2001). We compared the processed forms of Pmel17 in EGFP cells with that in EGFP-Myo1b cells. P1 and P2 precursors and the two disulfide-linked subunits Mα and Mβ were immunoprecipitated together with a monoclonal antibody directed against the N-terminus of Pmel17, HMB50. The amounts of Mα and Mβ immunoprecipitated from EGFP-Myo1b cells in these experimental conditions appeared to be lower than the amount immunoprecipitated from EGFP-expressing cells (Fig. 6B). Conversely, the amount of the precursor P2 was slightly increased in EGFP-Myo1b cells compared with that of EGFP cells (Fig. 6B). To confirm this first observation, which suggests that Myo1b can control the transport of Pmel17 leading to its processing, we compared the posttranslational processing of Pmel17 in EGFP-Myo1b cells with that in EGFP-expressing cells. In agreement with the previous report by Berson and colleagues (Berson et al., 2001), the endoplasmic reticulum precursor P1 immunoprecipitated 10 minutes after pulse-labelling, whereas P2 immunoprecipitated as a broad protein pattern that overlapped with the P1 precursor after 10 minutes of pulse-labelling and 30 minutes of chase (Fig. 6C). Furthermore, Mα, which predominantly immunoprecipitated in a diffuse protein pattern co-migrating with an alternatively spliced form of Pmel17 and Mβ, was detected after 1 hour of chase (Fig. 6C). Consistent with the observations at the steady state, Mβ levels were significantly lower and Mα was hardly detectable in EGFP-Myo1b cells, while levels of the precursor P2 appeared to slightly increase after 1 and 2 hours of chase in the EGFP-Myo1b-expressing cells (Fig. 6C and D).

We next examined whether the delayed cleavage of Pmel17 perturbed the formation of the intralumenal fibrils of stage-II premelanosomes in EGFP-Myo1b cells as previously reported, when cleavage of Pme17 was inhibited by expressing a variant of α1-antitrypsin inhibitor (Berson et al., 2003). Indeed, abortive fibrils were frequently observed in the immature melanosomes of the EGFP-Myo1b cells (Fig. 7). All together these observations suggest that MYO1B controls the processing of Pmel17, leading to the formation of the intralumenal fibrils of stage-II premelanosomes.

MYO1B might regulate the trafficking of proteins that pass through the early sorting MVE by controlling the morphology of these compartments as reported above. It might also act as a dynamic scaffold in interacting directly or indirectly with these proteins. Thus, we next investigated whether MYO1B can interact with Pmel17. Anti-GFP antibodies immunoprecipitated EGFP or EGFP-Myo1b from EGFP and EGFP-Myo1b cell lysates respectively (Fig. 8B). In addition, these antibodies precipitated a polypeptide (120 kDa) recognised by anti-Pmel17 antibodies (pep13h) from EGFP-Myo1b cell lysates but not from EGFP cell lysates (Fig. 8B). This polypeptide, having an apparent molecular mass of 120 kDa, could be the Golgi-modified O- and N-linked glycan precursor of Pmel17, P2 (see Fig. 6A). By contrast, other endosomal proteins, such as Eps15, that have been proposed to interact with actin or EEA1 and clathrin were not detected in these immunoprecipitates (data not shown). Conversely, when immunoprecipitates were prepared with the monoclonal antibody HMB50 against Pmel17, MYO1B and EGFP-Myo1b were detected in precipitates from EGFP-Myo1b cell lysates but EGFP was not detected in precipitate from EGFP-expressing cell lysates (Fig. 8C). The tail-domain at the C-terminus of Myo1b has been postulated to be involved in the interaction of Myo1b with membrane domains (Fig. 8A) (Raposo et al., 1999; Tang and Ostap, 2001). Consistent with this hypothesis, we detected the EGFP-Myo1b-Tail in anti-Pmel17 immunoprecipitates prepared from lysates of MNT-1 cells that overexpressed EGFP-Myo1b-Tail (Fig. 8C).

If the direct or indirect interaction of MYO1B with Pmel17 is involved in the transport of Pmel17, this interaction should take place on endosomes, where sorting and maturation of Pmel17 occurs. Taking advantage of a new method to isolate magnetic endosomes we determined whether MYO1B and Pmel17 could co-immunoprecipitate from a fraction enriched in endocytic compartments. Magnetic endosomes were isolated after internalising super paramagnetic nano-particles for 30 minutes (Wilhelm et al., 2002). This magnetic fraction was enriched in specific markers for early endosomes, such as Rab5, and exhibited low amounts of specific markers of late endosomes, such as Rab7, confirming the efficiency of this purification method (Fig. 9A). Endogenous MYO1B was enriched in the magnetic fractions isolated from wild-type and EGFP-Myo1b cells and EGFP-Myo1b was associated to some extend with the fraction isolated from EGFP-Myo1b cells (Fig. 9A). Precursor and mature forms of Pmel17 were also detected in the two magnetic fractions isolated from wild-type or EGFP-Myo1b cells (Fig. 9A). From the magnetic endosomes isolated from EGFP-Myo1b cells, anti-GFP antibody immunoprecipitated two proteins: one with the apparent molecular mass similar to the Pmel17 precursor P2 detected in the post nuclear supernatant (Fig. 9B compare lanes PNS and IP GFP) and one with higher molecular mass than the P2 precursor. This second protein is probably a degradation product because its mobility on the gel was slower than that of the P1 precursor when using the same antibodies for detection in post nuclear supernatant fraction (see lane PNS versus lane IP GFP Fig. 9B). Together, these observations indicate that, MYO1B and the Pmel17 precursor P2 are associated with the endocytic compartments (which are enriched in Rab5 protein), and bind directly or indirectly to each other. This interaction might occur through the C-terminal tail-domain of MYO1B.

Members of several classes of myosins have been implicated in endocytosis in mammals. Myosin VI is required for the traffic of endosomes and clathrin coated vesicles at the cell periphery (Buss et al., 2001; Hasson, 2003). Myosin V is involved in transferrin recycling and Myo1b contributes to the delivery of internalised molecules to lysosomes (Hales et al., 2002; Raposo et al., 1999; Rodriguez and Cheney, 2002). In contrast to myosin V and myosin VI, which form dimers in solution and are processive, Myo1b is monomeric and non-processive (Stafford et al., 2005). In yeast, class-I myosins control endocytosis by regulating actin nucleation, whereas in Dictyostelium discoidum one member of this class is involved in membrane recycling, most probably by regulating the morphology of these compartments (Geli et al., 2000; Neuhaus and Soldati, 2000). Here, we present evidence for the role of Myo1b in the morphology of endosomes and in the trafficking of molecules in early sorting MVEs.

It is generally accepted that the cellular distribution of the endocytic compartments depends upon the balance between anterograde and retrograde microtubules-motors that bind to these compartments. However, we have previously reported that Myo1b contributes to the distribution of lysosomes and transferrin receptor labelled endosomes (Cordonnier et al., 2001; Raposo et al., 1999). Furthermore, we have also reported that Myo1b and actin filaments cooperate with microtubules for the movement of lysosomes (Cordonnier et al., 2001). Consistent with these observations we show here that Myo1b and actin filaments control the distribution of EEA1 labelled endosomes. Myo1b and the actin networks might act as part of the regulatory machinery that transiently retains the endocytic compartments on actin filaments during their travelling on microtubules.

EEA1 is associated with early endosomes and early sorting MVEs with a bilayered clathrin coat. Early sorting MVEs are important sorting stations for the proteins that are delivered to lysosomes or involved in the biogenesis of lysosomes related organelles such as melanosomes (Raposo and Marks, 2002; Raposo et al., 2001; Sachse et al., 2002). We report here that, overexpression of EGFP-Myo1b or treatment with drugs that perturbe the dynamics of the actin cytoskeleton highlighted the presence of numerous small membrane extensions at the surface of these endosomes. We therefore suggest that MYO1B and actin contribute to the morphology of these endocytic compartments. Several lines of evidence suggest that MYO1B binds to the membrane of the endosomes. Using electron microscopy, we have previously reported that Myo1b was associated with endosomes and lysosomes from an hepatoma cell line and report here that EGFP-Myo1b is found on EEA1 labelled early sorting MVEs. Myo1b was also detected in fractions enriched with endosomes isolated from the hepatoma cell line (Raposo et al., 1999) and from MNT-1 cells (this study). Furthermore, we have previously shown that Myo1b was released from membrane fractions in the presence of the tail-domain of this molecular motor, suggesting that the interaction of Myo1b with membranes is mediated by this domain. In parallel, recent observations suggest that in solution, Myo1b regulates the organisation of actin filaments (Stafford et al., 2005). At low concentrations of Myo1b actin filaments are straighter and at high concentrations they form bundles. Modification of the cellular organisation of the actin network was never observed in the presence of EGFP-Myo1b at light microscopic level. However, we cannot exclude that MYO1B regulates the organisation of actin filaments at the surface of the early sorting MVE. In interacting with the membrane of endosomes on one hand and actin filaments on the other, MYO1B might transiently retain endosomes on actin filaments and exert tension on the endosomal membrane to form the small membrane extensions observed at the EM level.

By controlling the morphology of the sorting MVE, MYO1B might participate in the trafficking of those proteins that exist in these endocytic compartments. The peculiar processing of Pmel17 allowed us to study the impact of the expression of EGFP-Myo1b in the traffic of proteins that are passing through these endocytic compartments. We found that, the expression of EGFP-Myo1b impaired the cellular distribution of Pmel17, delayed its proteolytic cleavage and, thereby, the maturation of melanosomes (Berson et al., 2001; Berson et al., 2003; Sachse et al., 2002). Furthermore, MYO1B associated to endosomes that co-immunoprecipitated with the precursor P2 of Pmel17. Consistent with the previous observations, which suggest that the interaction of Myo1b with membranes is mediated by its tail, this domain co-imunoprecipitates also with Pmel17. MYO1B might bind directly to the cytoplasmic domain of Pmel17 through its tail-domain or, these two proteins might belong to the same protein complex at the surface of the early sorting MVE. The interaction with the precursor P2 of Pmel17 might allow MYO1B to transiently retain P2 on the external membrane of the early sorting MVE and thereby prevent it from proteolytic cleavage. The interaction with Pmel17 on one hand and with actin filaments on the other might allow MYO1B to act as a dynamic scaffold that clusters proteins to be transported towards lysosomes through internal vesicles.

Altogether, the experimental evidence reported here strongly suggests that, MYO1B and actin filaments at the surface of MVEs act as a mechanical device that contribute to the morphology of these endocytic compartments and to the sorting of proteins that are transported to the internal vesicles of MVEs.

The machinery involved in sorting of protein cargos towards internal vesicles of MVEs has been recently unravelled in the case of ubiquitinated proteins. Cargo proteins are recognized at the cytosolic side of endosomes by a protein complex, called ESCRT-0, which contains HRS (mammalian hepatocyte receptor tyrosine kinase substrate) STAM (signal transduction adaptor molecule) and Eps15 (Raiborg et al., 2003). In this model HRS recruits Tsg101, a component of the ESCRT-I complex that in turn recruits the protein complexes ESCRT-II and ESCRT-III. These protein-complexes act together to sequester ubiquitinated cargo proteins into the inward-budding vesicles of the MVEs (Bache et al., 2003; Lu et al., 2003). Some non-ubiquitinated proteins, however, are also delivered to the internal vesicles of MVEs (Reggiori and Pelham, 2001). A recent report indicates that HRS, but not Tsg101, is required for the transport of one of these proteins towards lysosomes (Hislop et al., 2004). Pmel17 might belong to this second class of proteins because we have not been able to detect an ubiquitin motif on this protein (data not shown). The acto-Myo1b-based mechanism might represent a part of the alternative machinery to sort non-ubiquitinated proteins in specific membrane domains at the cytosolic side of endosomes. Alternatively – or additionally – the retention of the proteins to be sorted in specific membrane domains at the surface of endosomes by an acto-Myo1b-based mechanism, might constitute an additional step in the trafficking events occurring in sorting MVEs.

We thank Catherine Martin for performing electronic microscopic analysis of actin in the vicinity of early MVE, Sophie Neveu for kindly preparing the magnetic nanoparticles and Marie-Thérèse Prospéri for preparing anti-MYO1b antibodies. We are grateful to Matthew Morgan for critical reading of the manuscript, and Michael Marks for fruitful discussions. This work was supported by a grant to E.C. from the association pour la recherche sur le Cancer (ARC n°4623). F.Y. and L.C.S. were successively the recipient of a grant from the Institut Curie.