A Myosin I Is Involved in Membrane Recycling from Early Endosomes

A central problem encountered at each step of the trafficking between intracellular compartments is the effective sorting of membrane components from fluid phase, and of exported proteins from residents. Besides well-characterized signal-based mechanisms, models have been proposed in which the sorting of membrane from fluid phase is achieved on a geometric basis by modulating the surface to volume ratio of the sorting compartments (Linderman and Lauffenburger 1988; Dunn et al. 1989). The role of the cytoskeleton in achieving, regulating, and maintaining the morphology of intracellular compartments, thereby forming the mechanical support for geometry-based sorting processes, has only recently been considered.

A central task of the endocytic pathway is the triage of ingested fluid from membrane-bound molecules (Gruenberg and Maxfield 1995). Although the recycling of most membrane-bound molecules does not require any specific signal, this process is rendered very efficient through the use of small vesicles or tubules with high surface to volume ratio (Mayor et al. 1993; Verges et al. 1999). The kinetics of receptor recycling consists of distinct fast and slow components, indicating that recycling occurs from two different endosomal compartments (Sheff et al. 1999).

During their lifetime, vesicles mediating membrane traffic have to successively recruit the machineries necessary for their formation, their transport, and the selection, tethering, docking, and, finally, fusion with their target membrane. The membrane trafficking machinery has to be coordinated with the cytoskeleton to ensure both efficient and rapid formation of tubular/vesicular structures and the delivery to distant target compartments (Schroer 2000). The role of microtubules and associated motor proteins, kinesin and dynein, has been extensively investigated in the last years (Hirokawa 1998). More recently, the importance of the actin cytoskeleton in the proper working of many steps of the secretory and endocytic pathways is becoming apparent.

Filamentous actin is involved in the endocytosis of fluid phase markers from the apical (Gottlieb et al. 1993) and basolateral membrane of polarized cells (Shurety et al. 1998). An involvement of actin in receptor-mediated endocytic uptake processes was shown in yeast (Kubler and Riezman 1993) and mammalian cells (Durrbach et al. 1996; Lamaze et al. 1997), though its role may be not obligatory in mammalian cells (Fujimoto et al. 2000). Actin was shown to localize also to sorting endosomes in fibroblasts (Nakagawa and Miyamoto 1998). The ADP-ribosylation factor (ARF)-6 GTPase, which was shown to be important in modeling the plasma membrane and cortical actin cytoskeleton (Song et al. 1998), was recently implicated in the targeted delivery of recycling endosomal vesicles to the plasma membrane (D'Souza-Schorey et al. 1998). Despite accumulating evidence for interactions between actin and coat proteins (Kohtz et al. 1990; Gaidarov et al. 1999), much less is known about the involvement of actin-dependent motor proteins. The discovery of a great diversity of myosins associated with many different organelles suggests that, in addition to bona fide transport, the actomyosin cytoskeleton may play a role in sorting processes and the modulation of membrane structures (Coluccio 1997).

At least four classes of myosins, I, II, V, and VI, have been directly involved in membrane trafficking (for review see Mermall et al. 1998). Among these, the double-headed myosin V and myosin VI were shown to transport organelles and particles along actin filaments. Evidence also exists for involvement of myosins in budding of vesicles from intracellular compartments. Nonmuscle myosin II in mammalian cells is recruited to Golgi membranes with features of a coat protein during the budding of transport vesicles (Müsch et al. 1997; Stow et al. 1998). Class I myosins are involved in different aspects of endocytic processes (Tuxworth and Titus 2000). It was proposed that they act at later stages of vesicle formation, like the fission of the endocytic vesicles from the plasma membrane or from internal vacuoles, or power the movement of newly formed vesicles through the cortical actin meshwork (Ostap and Pollard 1996). Myo5 and Myo3, the yeast class I myosins, are required for receptor-mediated endocytosis, possibly for a budding event in the uptake process (Goodson and Spudich 1995; Geli and Riezman 1996). Homologues of these myosins are implicated in endocytosis in many other organisms and may execute a variety of similar tasks. Myosin IB from Dictyostelium discoideum (Jung and Hammer 1990; Wessels et al. 1991), Acanthamoeba castellanii (Jung et al. 1989), and Entamoeba histolytica (Vargas et al. 1997) as well as myosin IA from Aspergillus nidulans (McGoldrick et al. 1995) show highest homology to both Myo5p and Myo3p and somehow participate in endocytic mechanisms. D. discoideum double deletion mutants (myoA⁻/myoB⁻ and myoC⁻/myoB⁻) show conditional defects in fluid phase pinocytosis (Novak et al. 1995; Jung et al. 1996). Other types of class I myosins also function in the endocytic pathway. For example, mammalian myosin Iα is associated with endosomes and lysosomes in rat hepatoma cells (Raposo et al. 1999) and with tubulovesicular structures in cell bodies of cultured rat superior cervical ganglion nerve cells (Lewis and Bridgman 1996).

We investigated the function of unconventional myosins in the eukaryotic model system D. discoideum, a genetically tractable organism perfectly suited for the investigation of cytoskeleton-related processes in the endocytic pathway (Maniak 1999; Neuhaus and Soldati 1999). Our findings, obtained by a combination of molecular genetics, cell surface labeling, biochemical fractionation, and a variety of microscopy techniques, demonstrate the involvement of MyoB in the recycling of plasma membrane components from an endosomal compartment back to the cell surface. We propose that actin and MyoB are involved in the sorting of membrane from bulk fluid phase by working as mechanical support to shape the endosomal compartment and/or to help in the production of transport intermediates.

D. discoideum cells of wild-type strain AX-2, myoA⁻ (Titus et al. 1993), myoB⁻, and myoA⁻/B⁻ cells (Novak et al. 1995) were grown axenically in HL5c medium (Sussman 1987) on plastic dishes at 22°C. myoA⁻/B⁻ cells (Novak et al. 1995) were maintained with 10 μg/ml G418 (Calbiochem).

The following primary antibodies were used: (a) an mAb against the plasma membrane marker 4C4 (PM4C4) called mAb V4C4F3 (Schwarz et al. 2000), a gift from Dr. J. Garin (CEA, Grenoble, France). This plasma membrane marker has not been molecularly identified, but magnetic fractionation showed that it is present throughout the endolysosomal system (Garin, J., personal communication); (b) a pAb against MyoB (Novak and Titus 1997), a gift from Dr. M. Titus (University of Minnesota, Minneapolis, MN); (c) a pAb against MyoC, a gift from Dr. G. Côté (Queens University, Kingston, Canada); (d) an mAb against myosin II (Pagh et al. 1984), a gift from Dr. G. Gerisch (Max-Planck-Institute for Biochemistry, Martinsried, Germany); (e) a pAb against biotin (Rockland); (f) an mAb 190-340-8 against comitin, an actin binding protein shown to be a component of the Golgi apparatus, from D. discoideum (Weiner et al. 1993), a gift from Dr. A. Noegel (University of Köln, Köln, Germany); (g) an mAb 221-135-1 against D. discoideum protein disulfide isomerase (PDI), an ER resident enzyme (Monnat et al. 1997); (h) an mAb 221-35-2 against the A subunit of the vacuolar H⁺-ATPase (Neuhaus et al. 1998); (i) an mAb 176-3-6 against coronin, an actin binding protein (de Hostos et al. 1993), a gift from (Dr. G. Gerisch). The secondary antibodies were either goat anti–mouse or goat anti–rabbit IgGs conjugated to Cyanine 3.29-Osu (Cy3; Rockland) or to Alexa488 (Molecular Probes).

Cells were plated on coverslips grade 0 (80–100 μm thick; Menzel Gläser) and allowed to adhere for several hours without selection before investigation. Pharmacologically treated cells were incubated with 10 μM cytochalasin A (cytA; Sigma-Aldrich) or 50 mM butanedione monoxime (BDM; Sigma-Aldrich) in HL5c for 15 min before freezing. The coverslips were plunged in methanol at −85°C, and then, using a homemade Dewar-based temperature-controlled apparatus, they were warmed to −35°C. Next, they were plunged in PBS at room temperature, and incubated with PBS containing 0.2% gelatin (Neuhaus et al. 1998). After staining, samples were mounted in ProLong Anti-Fade medium (Molecular Probes). Mounted samples were investigated with a Leica confocal microscope DM/IRB using a 63× objective with NA 1.4. Confocal optical sections were recorded at 0.4 μm per vertical step and eight times averaging; image stacks were imported into Adobe Photoshop^® (Adobe Systems Inc.) for image processing or Image Gauge v3.0 (Fuji Film) for quantitation.

Cells were prepared as described by Neuhaus et al. 1998 by plating on thin sapphire coverslips (Groh+Ripp) and plunged into a liquid ethane slush at −175°C using a guillotine-like device. Samples were freeze-substituted, infiltrated with Lowicryl HM-20 (Bioproducts SERVA), and polymerized at −45°C under UV light. Sections of 100-nm thickness (silver/light gold interference color) were cut horizontally to the plane of the coverslip and placed onto Formvar carbon–coated 100 mesh hexagonal copper grids. Sections were poststained for 10 min with 4% osmium tetroxide and lead citrate.

D. discoideum cells were plated on coverslips and incubated with HL5c medium containing fluid phase markers either for immunofluorescence or electron microscopical experiments. Yellow-green fluorescent nanobeads (FluoSpheres, diameter 20 nm; Molecular Probes) were used as the fluid phase marker for light microscopical investigations, as they turned out to be much better retained after methanol fixation than other available marker molecules such as lucifer-yellow, labeled dextran, and labeled proteins. 14-nm colloidal gold particles, prepared according to Slot and Geuze 1985, complexed with bovine serum albumin (BSA), as described by Griffiths (1993), were used to label endosomes for electron microscopical investigations. To follow membrane uptake, D. discoideum cells were washed with prechilled SBS (Soerensen buffer, 14.7 mM KH₂PO₄ and 2 mM Na₂HPO₄, with 120 mM sorbitol), pH 7.8, and incubated with ice-cold SBS, pH 7.8 containing 2 mg/ml sulfo-NHS-biotin for 30 min at 0°C. Afterwards, excess biotin was quenched by addition of SBS containing 100 mM glycine, and cells were washed with SBS again, resuspended in HL5c medium, and allowed to adhere to coverslips at 22°C to permit trafficking of biotinylated plasma membrane proteins. The cells were then processed as described for the different microscopical techniques.

To visualize acidic compartments, adherent D. discoideum cells were incubated with LysoSensor Green DND-189 (an acidotropic probe that may localize in the membrane of acidic organelles; Molecular Probes Product information sheet) for 5 min, washed in SBS, and imaged with a Zeiss Axiophot 2 microscope using a 100× Achroplan water immersion objective. Pictures were taken with a CCD camera (Imago Sensicam; Till Photonics).

Biotinylation of cell surface proteins was performed using a modified version of the assay of Bacon and coworkers 1994. 5 × 10⁶ D. discoideum cells plated on 6-cm plastic dishes were washed once with prechilled SBS, pH 7.8, and incubated with 1 ml ice-cold SBS, pH 7.8, containing 1 mg/ml sulfo-NHS-SS-biotin (Pierce Chemical Co.) for 10 min on an ice bath. Biotinylation was stopped by washing the cells with ice-cold SBS, followed by ice-cold quenching buffer (SBS containing 100 mM glycine). Cells were then incubated with HL5c medium at 22°C for 5 min to allow internalization of biotinylated plasma membrane proteins. Uptake was stopped by placing the cells on ice and washing them with ice-cold SBS. Biotin was cleaved off of exposed surface components by incubating the cells with 100 mM of the membrane impermeant reducing agent mercaptoethane sulfonic acid (MESNA; Sigma-Aldrich) in SBS, pH 8.2, for 12 min. When cells were kept at 0°C to prevent internalization of biotinylated plasma membrane, this treatment quantitatively stripped all biotin. After stripping, cells were again washed and incubated with HL5c medium at 22°C to allow further trafficking of internalized, labeled membranes. At indicated time points, this incubation was again stopped by washing the cells with ice-cold SBS. Biotin that had been reexposed at the plasma membrane was cleaved off by incubation with MESNA for 12 min. Finally, the cells were washed with ice-cold SBS, collected from the dish with 15 ml SBS, and centrifuged for 2 min at 500 g. The cell pellet was resuspended once in ice-cold Soerensen buffer, centrifuged, lysed in nonreducing sample buffer (30% glycerol, 3% SDS, 125 mM Tris-Cl, pH 6.8), and boiled at 95°C for 3 min. 10⁶ cells were loaded per lane of 7.5% SDS-PAGE gels, transferred to nitrocellulose membrane (Protran; Schleicher & Schuell), and stained with Ponceau S (Sigma-Aldrich). The nitrocellulose membranes were blocked with TBS (150 mM NaCl, 50 mM Tris-Cl, pH 7.4) containing 3% nonfat dried milk (Carnation) and incubated with streptavidin–HRP diluted in 5% BSA (Sigma-Aldrich) in TBS. Detection was performed with ECL plus (Amersham Pharmacia Biotech) using a chemiluminescence imager (Las-1000; Fuji Film). Data quantification was carried out with Image Gauge v3.0 (Fuji Film).

Magnetic fractionation of early endosomes was performed (Adessi et al. 1995) by feeding the cells with dextran-coated iron oxide for 15 min, lysing the cells, and isolating the iron-containing endosomes from the postnuclear supernatant on a steel wool-containing column placed in a magnetic field. The endosomes were eluted, pelleted, and lysed in sample buffer (30% glycerol, 3% SDS, 125 mM Tris-Cl, pH 6.8, 100 mM DTT). Then, they were run on an 8% SDS-PAGE gel and transferred to nitrocellulose membrane (Protran; Schleicher & Schuell). For the ATP-release experiment, a pellet of isolated endosomes was resuspended in buffer containing 25 mM ATP, incubated for 5 min at 20°C, and centrifuged. Equivalent amounts of both fractions, supernatants and pellets, were lysed in sample buffer, run on SDS-PAGE gels, and transferred to nitrocellulose membranes. After blocking, membranes were incubated with primary anti-MyoB, -MyoC, and –myosin II antibodies diluted in 3% milk/TBS, washed, and incubated with secondary anti–rabbit antibodies conjugated to HRP diluted in 3% milk/TBS.

Immunofluorescence analysis of different myosin I–deficient cells showed a striking redistribution of the plasma membrane marker PM4C4, as revealed by antibody staining (Fig. 1). In wild-type cells (Fig. 1, wt) it mainly localized to the plasma membrane and, as the membrane is taken up via endocytosis, to intracellular vacuoles. Labeling intensities on these intracellular vacuoles were weaker than at the plasma membrane, perhaps reflecting the fact that efficient and early recycling of plasma membrane components were taking place. In myoA/B-deficient cells at steady state, this marker was redistributed to endosomal vacuoles; it was found to a lesser extent at the plasma membrane than on intracellular vacuoles (Fig. 1, A-B-). Compared with wild-type cells, the PM4C4-positive vacuoles of myoA/B null cells were more homogenous in size. The altered distribution of PM4C4 likely primarily arose from the absence of MyoB, as myoB single mutants (Fig. 1, B-), but not myoA single mutants (Fig. 1 A-) showed a phenotype similar (but not identical) to the double mutant cells. Nevertheless, as the effect was slightly more pronounced in myoA/B null cells and because strains with multiple, but not single myosin gene deletions have been reported to show endocytosis defects (Novak et al. 1995; Jung et al. 1996), the myoA/B null cells were further investigated.

The redistribution of PM4C4 did not result from a general disturbance of the cytoskeleton, but, more likely, from an impairment of myosin function. Indeed, the staining pattern observed after incubation with the F-actin depolymerizing agent cytA was close to the wild-type labeling pattern (Fig. 1, cytA). In sharp contrast, short treatment of wild-type cells with BDM, a general inhibitor of myosin ATPases, resulted in a redistribution of PM4C4 to intracellular structures (Fig. 1, BDM), similar to what was observed for the myoA⁻/B⁻ and myoB⁻ cells.

To quantify this effect, the ratio of labeling intensities at the plasma membrane and on endosomes was measured in confocal sections. Therefore, line scans were performed through several wild-type and myoA⁻/B⁻ cells, the mean intensity values on the plasma membrane and on intracellular structures were determined, and the ratio of plasma membrane to endosomes per line scan was calculated as shown (Fig. 2). Evaluation of 78 line scans through wild-type and 53 line scans through myosin-deficient cells resulted in plasma membrane to endosome ratios of 2.8 ± 1.43 and 0.34 ± 0.25, repectively. This eightfold concentration of PM4C4 on endosomal vacuoles might be explained by an anomalous flux of membrane components through the endosomal compartments.

To verify that the PM4C4-labeled vacuoles indeed belong to the endolysosomal system, the fate of endocytosis markers was followed (Fig. 3). Fluid phase uptake was visualized by feeding cells with fluorescently labeled nanobeads of 20-nm size (Fig. 3, a and c, nanobeads). Among the many fluid phase markers tested, the nanobeads appeared best suited for live microscopy and for optimal retention during the methanol fixation procedure. The concentration of internalized marker varied from endosome to endosome, in fixed and live cells, and was likely a function of their degree of maturation (Neuhaus, E.M., and T. Soldati, manuscript in preparation). In Fig. 3, a and c, arrows and arrowheads point to vacuoles with low and high concentrations of nanobeads, respectively. After 90 min of feeding, each PM4C4-positive compartment, in wild-type as well as in myoA⁻/B⁻ cells, contained detectable amounts of the fluid phase marker, demonstrating their belonging to the endosomal system (Fig. 3, a and c, PM4C4 and overlay).

In addition to this colocalization of PM4C4 with ingested fluid phase marker, we investigated whether internalized membrane proteins have access to the compartments that accumulate PM4C4 (Fig. 3b and Fig. d). First, plasma membrane proteins were labeled at 0°C with a biotin-conjugated cross-linker. Then, before fixation, the cells were incubated in medium at 22°C to allow trafficking of biotinylated proteins. Internalized plasma membrane proteins (Fig. 3b and Fig. d, biotin) colocalized with each of the PM4C4-labeled structure in wild-type and myoA⁻/B⁻ cells (Fig. 3b and Fig. d, PM4C4 and overlay). In addition, myoA⁻/B⁻ cells accumulated higher concentrations of biotinylated plasma membrane proteins on intracellular PM4C4-positive vacuoles than wild-type cells (Fig. 3b and Fig. d, PM4C4). The relative intensities of biotin and PM4C4 labeling on intracellular vacuoles did not always correspond to each other: some vacuoles had higher intensities of the biotin labeling (Fig. 3b and Fig. d, arrows) and others had higher intensities of PM4C4 labeling (Fig. 3b and Fig. d, arrowheads).

It is unlikely that the vacuoles accumulating PM4C4 belong to the contractile vacuole, the other prominent endomembrane system in D. discoideum, as staining of myosin-deficient cells for the contractile vacuole markers vacuolar H⁺-ATPase (Fig. 4 a, v-H⁺-ATPase) and calmodulin (data not shown) did not differ from the wild-type staining patterns. Moreover, the morphology of other endomembrane compartments was undisturbed and the constitutive secretory pathway was not altered in the mutant cells (Temesvari et al. 1996). The PM4C4-positive vacuoles in myoA⁻/B⁻ cells did not stem from anomalous ER or Golgi cisternae, as staining for the ER resident enzyme PDI (Fig. 4 a) and for comitin (Fig. 4 a, comitin), an actin-binding protein associated with the Golgi apparatus, did not differ from the wild-type.

In D. discoideum, it has been shown that as early as 1–2 min after ingestion and release of the cytoskeletal coat consisting of actin and coronin (Maniak et al. 1995), the endosomal pH drops significantly (Maniak 1999), reaches a minimum at ∼15 min, and slowly rises afterwards (Aubry et al. 1993; Padh et al. 1993). No significant change was detected in the number of intracellular coronin-positive vacuoles in myoA⁻/B⁻ cells, indicating that the macropinosomes likely shed their cytoskeletal coat at normal rates (Fig. 4 b, coronin). Acidic compartments can be visualized in live cells by the pH-sensing reagent LysoSensor, a membrane-permeant weak base that is trapped in acidic compartments upon protonation and shows a pH-dependent increase in fluorescence. Wild-type and myoA⁻/B⁻ cells were incubated with LysoSensor Green (Fig. 4 b, lysosensor). Whereas in wild-type cells labeled compartments exhibited a large size variation, they were relatively big and homogenous in size in myoA⁻/B⁻ cells. The uniform morphology, the size, and the high abundance of these acidic vacuoles suggest that they are identical to the compartment accumulating PM4C4.

The observed concentration of PM4C4 on endosomal vacuoles in myoA⁻/B⁻ cells might be explained by a membrane recycling defect. To directly test this hypothesis, we adapted a protocol primarily used to monitor membrane uptake and based on reversible biotinylation of plasma membrane proteins (Bacon et al. 1994; Novak et al. 1995; and illustrated in Fig. 5 a). Control experiments for each of the multiple steps of this recycling assay are shown in Fig. 5 b.

Plasma membrane proteins were biotinylated in adherent D. discoideum cells at 0°C with a biotin-conjugated cross-linker containing a disulfide bridge. Western blotting of D. discoideum cell extracts with streptavidin-HRP before and after biotinylation (control lane and +biotin lane, respectively, in Fig. 5 b, biotinylation) showed a prominent band at ∼80 kD corresponding to an endogenous biotin-containing protein. The major band resulting from biotinylation appeared at ∼120 kD and corresponds to gp126, a cell surface protein that is internalized at room temperature (Bacon et al. 1994). This signal disappeared after biotin was cleaved off with the membrane-impermeant reducing agent MESNA (Fig. 5 b, biotinylation).

After uptake of labeled plasma membrane into endosomes (Fig. 5 a, uptake), internalized biotin should become inaccessible to cleavage by MESNA (Fig. 5 a, cleavage 1). As expected, increasing amounts of biotinylated 120-kD protein became inaccessible to cleavage by MESNA with time (+MESNA lanes in Fig. 5 b, uptake wt). Note that the total amount of biotin incorporated in the samples (−MESNA lanes in Fig. 5 b, uptake wt) remained constant, excluding that the 120-kD labeled protein was significantly degraded during the experiment. Similarly, in myoA⁻/B⁻ cells a portion of the biotinylated marker protein was internalized and became resistant to MESNA cleavage (Fig. 5 b, uptake myoA⁻/B⁻). The membrane uptake rates estimated from the first time points of these experiments revealed that internalization of one cell surface equivalent occurs every 18 ± 4 min in wild-type cells and every 28 ± 4 min in myoA⁻/B⁻ cells. These data confirm quantitatively the earlier qualitative observation (Novak et al. 1995) that myoA⁻/B⁻ cells internalize membrane components slower than wild-type cells. As the rates of fluid phase uptake were similar in the two strains (data not shown), the reason for this decrease could reside in a difference in the average size of the primary pinosomes, but was not further investigated in the present study.

After a 5-min uptake pulse followed by MESNA cleavage at 0°C, further trafficking of biotinylated membrane proteins was allowed at 22°C (Fig. 5 a, chase). The total amount of endocytosed biotinylated 120-kD protein (cleavage 1 resistant) remained constant over the duration of the experiment (Fig. 5 b, chase wt), demonstrating that neither was it degraded, nor was the biotin cleaved off during passage in the endolysosomes.

If the 120-kD proteins become reexposed at the plasma membrane, they are again accessible to MESNA cleavage at 0°C (Fig. 5 a, cleavage 2) and only the biotin remaining protected in intracellular compartments is detected by Western blotting. In wild-type cells, the intensity of the intracellular, cleavage 2–resistant 120-kD band continuously decreased with increasing time of chase at 22°C (Fig. 5 c, wt), demonstrating for the first time that early membrane recycling occurs in D. discoideum. In sharp contrast, except for the decrease detected in the first 5 min, in myoA⁻/B⁻ cells the 120-kD signal remained constant for >40 min (Fig. 5 c, A-/B-). The intensity profiles of the 120-kD band from at least three independent experiments for each time point and cell line were quantified and the resulting curves were plotted (Fig. 5 d). Due to the difference in membrane internalization rates between both cell lines, the starting points of the recycling experiments were normalized to the amount of biotin taken up by the cells after 5 min. Wild-type D. discoideum cells recycled ∼90% of a 5-min pulse of biotinylated plasma membrane proteins in 30–40 min, and >40–50% of the signal was recycled in the first 5 min. In myoA/B-deficient cells, only 30–40% of a 5-min pulse of biotinylated plasma membrane proteins was recycled after 40 min, most of it in the first 5 min. Compared with wild-type cells, the intensity of the intracellular signal in myoA⁻/B⁻ cells did not significantly diminish, a strong indication of a severe defect in membrane recycling from the endosomes to the cell surface.

The presence of MyoB on early endosomes was investigated. D. discoideum cells were fed with colloidal iron for 15 min, and the iron-containing endosomes were magnetically purified (see Materials and Methods for details). This method has been extensively used in D. discoideum (Rodriguez-Paris et al. 1993; Nolta et al. 1994; Temesvari et al. 1994; Adessi et al. 1995; Souza et al. 1997) and results in a high enrichment for endolysosomal markers with very low contamination of nonendocytic organelles (plasma membrane, mitochondria, and ER are all <1%). The early endosomal proteins recovered by this procedure represented ∼1% of the total amount of protein in the starting material or 2% of the particulate fraction of the postnuclear supernatant, as this fraction contains 50% of the total cell protein (data not shown).

MyoB was present in the early endosomal fraction, whereas MyoC, a closely related myosin I used as a control, was not detected in this fraction (Fig. 6 a). About 1% of total cellular MyoB was found in the early endosomal fraction. But, since 90% of total MyoB is present in a large cytosolic pool of inactive myosins, and only 10% is present in the particulate fraction of the cell lysate, this means that 10% of the MyoB present on the particulate fraction is on early endosomes. As early endosomal proteins form only 2% of the particulate fraction, this represents at least a fivefold enrichment of MyoB on the early endosomes.

The nature of the association of MyoB with the early endosomes was further examined. In the absence of ATP, myosins bind to actin filaments via their motor domain (rigor mortis), and this binding can be released by addition of ATP. Furthermore, myosins can bind cargos via their tail domain, which is the most variable part of the molecule and has been suggested to reflect the adaptation of myosins to specific functions (Mermall et al. 1998). In MyoB, as in many class I myosins, the tail is composed of three subdomains (for review see Coluccio 1997): a basic phospholipid binding domain, a domain shown in vitro to contain a secondary ATP-independent actin-binding site, and finally, an Src homology 3 domain. To test whether MyoB is associated with the early endosomes via its motor domain or whether it is specifically associated via its tail domain, an ATP-release experiment was performed (Fig. 6 b). The small fraction of myosin II present on the endosomes served as a control. Myosin II was efficiently released from the endosomal fraction after incubation with ATP and was detected in the supernatant (Fig. 6 b, ATP-release), indicating that it was bound via its head domain to endosome-associated actin filaments. MyoB was not released from the endosomes by incubation with ATP, indicating that binding was likely mediated by its tail domain. The tail domain could either bind to endosome-associated actin filaments via the ATP-insensitive actin-binding site or directly to the endosomal membrane via the membrane-binding site (Fig. 6 c). The specific binding of MyoB to biochemically purified early endosomes via its tail domain may be indicative of a direct involvement in the recycling of membrane components from this compartment.

To determine the precise place and mode of action of MyoB, the ultrastructure of endocytic vacuoles in myosin I–deficient cells was investigated (Fig. 7). Wild-type and myoA⁻/B⁻ cells were fed with BSA-covered colloidal gold particles. Freshly prepared gold particles are monodispersed, but clump together after the protein coat is digested by lysosomal enzymes (Bright et al. 1997). Clumping of particles in endosomal compartments can therefore be used as a marker for the different stages of maturation/degradation. Early endosomal vacuoles have intermediate electron density and contain only few gold particles. Concentration of the fluid phase during maturation of the endosomes results in a more electron-dense appearance of the endosomal lumen and in the accumulation of clumped gold particles. Particles also appear in compartments that have vesicular membranes in their lumen (Neuhaus, E.M., and T. Soldati, manuscript in preparation).

Rapidly frozen wild-type cells commonly exhibited one small (∼110–120 nm) vesicle per 100-nm section of micrometer-sized early endosomal vacuole (Fig. 7 a, wild-type), suggesting that the whole vacuole was surrounded by <10 small vesicles. In myosin-deficient D. discoideum cells (Fig. 7 a, myoA-/B-), on average ∼6 small vesicles were found around early endosomal vacuoles, suggesting that the vacuoles were surrounded by 50–100 vesicles.

Other endomembrane compartments did not show an accumulation of small vesicles (Fig. 7 b). Later endosomal vacuoles containing big aggregates of gold particles (Fig. 7 b, arrowheads) and/or membranous structures in their lumen (Fig. 7 b, arrows) were not surrounded by small vesicles, either in wild-type or in myoA⁻/B⁻ cells. Also the newly formed macropinosomes, identified by their actin coat, were not surrounded by small vesicles in myoA⁻/B⁻ cells (Fig. 7 b, MP). Translucent bladders free of gold particles and closely apposed to the plasma membrane most likely belong to the contractile vacuole complex, an osmoregulatory system consisting of interconnected tubules, vacuoles, and cisternae (Heuser et al. 1993). No accumulation of vesicles was discernible around the contractile vacuoles of myoA⁻/B⁻ and wild-type cells (Fig. 7 b, CV).

Many of the small vesicles in myoA⁻/B⁻ cells appeared to be directly attached to the vacuoles by a tether of unidentified nature (Fig. 7c, arrowheads, and a, myoA⁻/B⁻, small arrowheads point to a ∼1-μm-long tether). Interestingly, some of the vesicles had a relatively electron-dense periphery, possibly resulting from the presence of a coat (Fig. 7 c, arrows). It is tempting to speculate that the observed structures result from incomplete or nonproductive budding of recycling vesicles.

Class I myosins have been implicated in endocytosis in a variety of organisms, from yeast to mammalian cells. The simple eukaryote D. discoideum, which is perfectly suited as a model system for the investigation of cytoskeleton-related processes in the endocytic pathway (for reviews see in Maniak 1999; Neuhaus and Soldati 1999), expresses seven class I myosins (Uyeda and Titus 1997; Schwarz et al. 1999). The precise place and mode of action of these myosins are slowly being unraveled. Here, we present evidence for the function of myosin IB in the process of membrane recycling from the endosomes back to the cell surface.

In D. discoideum, significant amounts of fluid phase are taken up via macropinocytosis (Hacker et al. 1997). Therefore, the primary pinosomes are rather big vacuoles. Maturation of this primary pinocytic compartment starts immediately after release of the cytoskeletal coat and, as early as 1–2 min after ingestion, the pH drops significantly and delivery of a subset of lysosomal enzymes occurs (Aubry et al. 1993; Padh et al. 1993; Maniak 1999). As very little or no early fluid phase recycling was observed in D. discoideum (Aubry et al. 1997), endocytic trafficking was proposed to be a linear process (Maniak 1999), ending with the exocytosis of food remnants from postlysosomes (Jenne et al. 1998; our unpublished observations). But membrane components not destined for degradation have to be recycled back to the plasma membrane. Moreover, homeostasis of the size of the pinocytic compartment implies that delivery of membranes containing the proton pumps and lysosomal enzymes has to be compensated by an equally rapid and efficient retrieval of membrane. Here, we show that this indeed happens in D. discoideum. Membrane recycling from endosomes occurs with a complex kinetics: ∼50% of a 5-min pulse of biotinylated plasma membrane proteins is recycled in 5 min, and the remaining marker reaches the cell surface in the next 30–40 min. Mammalian cells show a biphasic recycling kinetics with direct, fast recycling from the early sorting endosome and slow recycling from the recycling endosome (Ghosh et al. 1994; Sheff et al. 1999). myoA⁻/B⁻ cells show recycling of 30–40% of a 5-min uptake pulse in the first 5 min, but show no further recycling in the next hour. As a result of this defect, myoA⁻/B⁻ cells accumulate the plasma membrane marker PM4C4 on intracellular endocytic vacuoles. As myoB⁻, but not myoA⁻ cells accumulate this marker in a very similar manner, we conclude that MyoB might be responsible for membrane recyling in D. discoideum. Interestingly, there is recent evidence that class I myosins play a role in endocytic membrane trafficking along the recycling pathway in polarized epithelial cells (Durrbach et al. 2000; Huber et al. 2000).

An increased number of small vesicles was found in the immediate vicinity of early endocytic vacuoles in myoA⁻/B⁻ cells. Many of these vesicles were tethered to the vacuole surface. Some of these vesicles have a relatively electron-dense periphery, suggesting the presence of a coat. Clathrin-coated vesicles have been involved in the recycling process in mammalian cells (Stoorvogel et al. 1996; Futter et al. 1998). A close relationship between clathrin-coated vesicle formation, motility, and the actin cytoskeleton was recently shown in D. discoideum (Damer and O'Halloran 2000). These findings suggest that MyoB could help maintain the proper morphology of the endosomal compartment involved in recycling, thereby enabling efficient sorting of membranes from fluid phase. One might speculate that, together with actin filaments, MyoB could be involved in the formation of a contractile ring or spiral around the vesicle neck, thereby helping it to pinch off. Alternatively, actin and MyoB could act as a coat and provide the driving force to bud off a vesicle, or it may function as a bona fide motor and transport vesicles along the actin filaments away from the donor organelle. Indeed, possible roles in the process of vesicle budding and/or scission have been discussed for the yeast MyoB homologues, Myo5p and Myo3p, that are involved in receptor-mediated endocytosis (Geli and Riezman 1998).

Earlier investigations demonstrated that myoB⁻ cells exhibit a reduced efficiency of chemotactic aggregation and cell motility in only 51% of the of wild-type cells (Jung and Hammer 1990; Wessels et al. 1991). The reduction in cell motility of myoB⁻ cells could be a result of the impaired membrane flow from the endosomes back to the plasma membrane. Membrane flow models for cell locomotion, in which the advance of the leading edge is provided by membrane from internal pools, have been proposed (Abercrombie et al. 1970; Bretscher 1996). The delivery of receptors to the leading edge of migrating cells (Hopkins et al. 1994; Lawson and Maxfield 1995; Pierini et al. 2000) closely link membrane recycling and cell movement.

We estimated initial rates of membrane uptake in wild-type cells from the first time point of the biotinylation experiment and showed that internalization of one cell surface equivalent occurs every 18 ± 4 min. Using a fluorimetric assay, Aguado-Velasco and Bretscher recently measured that D. discoideum amebas endocytose one cell surface equivalent every 4–10 min (Aguado-Velasco and Bretscher 1999). Earlier studies, which followed cell surface proteins labeled with radioactive galactose, reported a plasma membrane internalization rate of 45 min (Thilo and Vogel 1980; Thilo 1985). It was shown that D. discoideum cells take up fluids mainly via macropinocytosis. We calculated that formation of five macropinosomes per minute with an average diameter of 1.6 μm (Hacker et al. 1997) leads to the uptake of one cell surface equivalent every 24 min, which is in the range of our experiments. These high uptake rates may be well correlated with the elevated endocytic membrane flow necessary for rapid directed movement during chemotaxis (Aguado-Velasco and Bretscher 1999).

More evidence for links between motility and membrane flow comes from the investigation of several mutants, for example D. discoideum cells lacking clathrin heavy chain or coronin (for review see Gerisch et al. 1999). The role of MyoB in membrane recycling could therefore be one factor contributing to the reported reduction in cell motility of myoB⁻ cells (Jung and Hammer 1990; Wessels et al. 1991).

Furthermore, defects in the uptake of yeast were observed in myoA⁻/B⁻ cells assayed on substratum (our unpublished observations), although other studies found no reduction of phagocytosis in suspension-grown myoA⁻/B⁻ cells (Novak et al. 1995). Phagocytosis was found to be reduced in myoB⁻, but not myoA⁻ cells (Jung and Hammer 1990). These defects are interesting in light of the recent discussion emphasizing the tight mechanistic relationship between phagocytosis and cell motility (Mellman 2000). The requirement for vectorial insertion of membrane at sites of membrane protrusions may link both processes to the accurate and polarized recycling of plasma membrane from the endosomes back to the cell surface (Bajno et al. 2000; Mellman 2000).

The molecular mechanisms of membrane trafficking in the simple eukaryote D. discoideum show extensive similarities to mammalian cells (Neuhaus and Soldati 1999). Nevertheless, the linearity of the major endosomal fluid phase flux was thought to illustrate a major discrepancy between these organisms. Here we show that D. discoideum cells, like mammalian cells, efficiently recycle plasma membrane components from an early endocytic compartment, indicating that major transport steps are conserved between these organisms. We further show that a myosin I is involved in this membrane recycling step. This work highlights that, besides the extensively studied signal-based sorting mechanisms, actin and myosin may be implicated in the sorting of membrane and contents. We propose that, together with the formation of specialized lipid domains (for review see Mukherjee and Maxfield 2000), the actin–myosin cytoskeleton may build the mechanical support to shape some intracellular compartments and produce transport carriers, thereby providing the foundation for geometry-driven sorting processes.

We especially thank Dr. M. Titus for the gift of myosin-deficient cell lines, antibodies, and for lots of helpful comments, and also Dr. J. Garin for performing the magnetic fractionation experiments and for the gift of the plasma membrane marker antibody. We further thank Dr. M. Maniak (Kassel University, Kassel, Germany) for helpful discussions and donation of antibodies. Drs. G. Côté, G. Gerisch, and A. Noegel kindly provided antibodies. Thanks also to Drs. M.I. Geli, H.-H. Gerdes (both University of Heidelberg), and E. Schwarz for their helpful comments on the manuscript.

The work was supported by a grant from the Deutsche Forschungsgemeinschaft (SFB 352) and by the Max-Planck Gesellschaft (doctoral fellowship to E.M. Neuhaus).