|Home | About | Journals | Submit | Contact Us | Français|
The flow of material from peripheral, early endosomes to late endosomes requires microtubules and is thought to be facilitated by the minus end-directed motor cytoplasmic dynein and its activator dynactin. The microtubule-binding protein CLIP-170 may also play a role by providing an early link to endosomes. Here, we show that perturbation of dynactin function in vivo affects endosome dynamics and trafficking. Endosome movement, which is normally bidirectional, is completely inhibited. Receptor-mediated uptake and recycling occur normally, but cells are less susceptible to infection by enveloped viruses that require delivery to late endosomes, and they show reduced accumulation of lysosomally targeted probes. Dynactin colocalizes at microtubule plus ends with CLIP-170 in a way that depends on CLIP-170’s putative cargo-binding domain. Overexpression studies using p150Glued, the microtubule-binding subunit of dynactin, and mutant and wild-type forms of CLIP-170 indicate that CLIP-170 recruits dynactin to microtubule ends. These data suggest a new model for the formation of motile complexes of endosomes and microtubules early in the endocytic pathway.
The microtubule cytoskeleton provides a dynamic structural framework that underlies a wide variety of subcellular motile events. The steady-state localizations of endomembrane systems such as the endoplasmic reticulum (ER), Golgi apparatus, endosomes, and lysosomes require both intact microtubules and the activities of microtubule-based motor proteins (reviewed by Goodson et al., 1997 ). The extension of ER tubules, the Golgi-to-ER “recycling” leg of the biosynthetic pathway, and the centrifugal movement of lysosomes are thought to require the activity of a plus end-directed motor such as kinesin or a kinesin-related protein (reviewed by Lane and Allan, 1998 ). Cytoplasmic dynein, the predominant cytosolic minus end-directed motor, in conjunction with its activator, dynactin (Gill et al., 1991 ), maintains the Golgi complex, endosomes, and lysosomes in the normal juxtanuclear position (Burkhardt et al., 1997 ; Harada et al., 1998 ). Dynein/dynactin drives membrane transport from the ER to the Golgi complex (Presley et al., 1997 ) and may facilitate inward flow of material within the endocytic pathway, as suggested by in vitro studies of protein transfer from early to late endosomes (Bomsel et al., 1990 ; Aniento et al., 1993 ). However, the contributions of dynein and dynactin to endocytic trafficking in vivo have yet to be determined, because previous dynein antibody microinjection studies (Vaisberg et al., 1993 ; Burkhardt et al., 1997 ) and evaluation of dynein knock-out mice (Harada et al., 1998 ) did not address endosome function in detail. An in-depth exploration of the effects of dynein/dynactin perturbation on endocytic trafficking and dynamics in living cells is clearly warranted.
The cytoplasmic dyneins are a multiprotein family of at least three members (Criswell et al., 1996 ; Vaisberg et al., 1996 ). The most abundant isoform, dynein 1, works in conjunction with a multisubunit activator protein, dynactin. Dynactin is thought to serve as an adapter that mediates dynein binding to a variety of cargo structures, including membranes, chromosomes and microtubules (reviewed by Allan, 1996 ; Holleran et al., 1998 ). It is also proposed to facilitate long-range movement by increasing dynein processivity (King and Schroer, 2000). To allow for these different functions, dynactin has two distinct structural domains (Schafer et al., 1994 ). Its rigid, filamentous backbone contains several proteins whose sequences predict both covalent and noncovalent cargo attachment mechanisms (Eckley et al., 1999 ), whereas its flexible projecting sidearm binds dynein and microtubules (reviewed by Holleran et al., 1998 ). These two domains are thought to be linked by the protein dynamitin which, when overexpressed, causes dynactin’s sidearm to release from the backbone (Echeverri et al., 1996 ), thus decoupling dynein-binding and cargo-anchoring functions. This leads to a variety of defects; mitotic cells arrest in pseudoprometaphase (Echeverri et al., 1996 ), whereas interphase cells show altered steady-state distributions of the Golgi complex, endosomes, and lysosomes (Burkhardt et al., 1997 ). Addition of excess dynamitin to cell extracts (Wittman and Hyman, 1999 ) and purified dynactin (Eckley et al., 1999 ) also disrupts dynactin structure, suggesting that it can also be used to impair dynein-dependent activities in vitro. Dynamitin is thus a powerful tool for analyzing dynactin and, by inference, dynein, function in a variety of contexts.
Endosomes in the cell periphery are believed to be transiently tethered to microtubules via the cytoplasmic linker protein CLIP-170 before centripetal movement (Rickard and Kreis, 1996 ). CLIP-170 contributes to the binding of endosomal membranes to microtubules in vitro (Pierre et al., 1992 ). In cells, CLIP-170 labels the growing ends of microtubules (Perez et al., 1999 ), an appropriate site for its proposed function. Microtubule binding is regulated by phosphorylation (Rickard and Kreis, 1991 ), providing a means for release of CLIP-170 (and associated cargo) from microtubules under circumstances in which a static interaction is no longer needed. This might occur once motors have been recruited to the endosome–microtubule complex.
In the present study, we examine the contribution of the dynein/dynactin motor to endocytic motility and trafficking and explore its links to CLIP-170. Late endosomes located near the center of the cell move bidirectionally over long distances. In cells that overexpress dynamitin, this movement, and that of other organelles, is completely inhibited. Membranes of the endocytic pathway accumulate in the cell periphery, as seen previously. Although early and late endosomes are now in close proximity, forward traffic through the pathway is slowed. The highly conserved dynamitin N terminus is found to be sufficient to inhibit dynactin activity in vivo. In immunolocalization studies, dynactin colocalizes with CLIP-170 at microtubule plus ends, an association that depends on the CLIP-170 C-terminal cargo-binding domain. This suggests that CLIP-170 binds to microtubules first and then recruits dynactin, providing a concerted mechanism for loading endosomes onto microtubules and converting them to a motile pool.
Vero (CCL81; American Type Culture Collection, Manassas, VA) and HeLa (CCL2; American Type Culture Collection) were grown in minimum essential medium plus 1% glutamine and 5 or 10% FCS. Cos7 cells were grown in Dulbecco’s modified Eagle’s medium and 10% FCS.
Rabbit Abs were CLIP-170 (Ab 55; Pierre et al., 1992 ), influenza hemagglutinin (HA; Daro et al., 1996 ), MPR300 (a gift from O. Rosorius, University of Geneva), early endosomal antigen 1 (EEA1) (Simonsen et al., 1998 ), vesicular stomatitis virus glycoprotein (VSV-G; Griffiths et al., 1985 ), pAb p150Glued (Vaughan et al., 1999 ), and pAb dynamitin antiserum to dynamitin gel purified from bovine dynactin.
Mouse mAbs were anti-HA-epitope (Daro et al., 1996 ), lysobisphosphatidic acid (LBPA mAb 6C4 (Kobayashi et al., 1998 ), anti-transferrin (Tfn)-R mAb OKT9 (Sutherland et al., 1981 ), anti-giantin (Linstedt and Hauri, 1993 ), mAb P5D4 against the VSV-G cytoplasmic tail (Kreis, 1986 ), anti-galactosyltransferase (Kawano et al., 1994 ), anti-myc epitope mAb 9E10 (Evan et al., 1985 ), CD63 mAb 1B5 (a gift from M. Marsh, University College, London, United Kingdom), and CLIP-170 mAbs 2D6 and 4D3 (Rickard and Kreis, 1991 ; Pierre et al., 1992 ). The anti-dynamitin mAb 50A (chicken specific) and anti-p150Glued mAb 150B were isolated in the same monoclonal screen that yielded mAbs 45A and 62B (Schafer et al., 1994 ). Anti-β-galactosidase (β-gal) was from Promega (Madison, WI).
A λgtll chick embryo library (B. Vennstrom, European Molecular Biology Laboratory, Heidelberg, Germany) was screened using mAb 50A. cDNAs from immunopositive clones were isolated, digested with EcoRI, and ligated into the EcoRI site of pBluescript KS II+ (Stratagene, La Jolla, CA) for later manipulations. Nested ExoIII deletions of the longest dynamitin cDNA (1.5 kb) were sequenced using Sequenase (Amersham Life Science, Arlington Heights, IL), and the sequence from both strands was assembled using the MacDNasis DNA analysis program (Hitachi Software Engineering, San Bruno, CA).
The dynamitin cDNA insert was subcloned into two cytomegalovirus (CMV)-based expression vectors: pGW1-CMV (Quintyne et al., 1999 ) and pCB6 containing an upstream HA tag (Balda et al., 1996 ). To facilitate subcloning, the first 108 bp of the dynamitin cDNA was amplified by PCR using an AlfII site introduced at the dynamitin initiator codon and a unique, downstream SacI site. The remainder of dynamitin was cloned from the SacI and XbaI sites, and the two fragments were religated into pCB6-HA. The dynamitin N terminus (amino acids 1–87) cDNA was made by reverse transcriptase-PCR from the original λgt11 clone and subcloned in pTA (Invitrogen, Carlsbad, CA). The insert was verified by sequencing and cloned into pGW1-CMV. A λgt10 insert encoding full-length chicken p150 was cloned as described (Quintyne et al., 1999 ). Plasmid pSG5-myc-CLIP-170 wild type (wt) contained the EcoRI insert from pGEM-myc-CLIP (Pierre et al., 1994 ) subcloned into pSG5; pSG5-myc-CLIPΔ1240 was generated by truncating pSG5-myc-CLIP-170 wt using XhoI and BamHI followed by blunting and religation (both were gifts from J. Rickard, University of Geneva). pSG5-Sar1p contained Sar1p cDNA (from B. Balch, Scripps Institute, La Jolla, CA; Aridor et al., 1995 ) subcloned into pSG5 (a gift from M. Gomez, University of Geneva). CMV-based vectors encoding humanized green fluorescent protein (GFP) and a red-shifted GFP (pCGFP2) were gifts from W.J. Nelson (Stanford University, Stanford, CA) and D. Shima and G. Warren (Imperial Cancer Research Fund, London, United Kingdom), respectively.
HeLa and Vero cells, plated on 10-mm2 coverslips 24 h before transfection, were transfected (CaPO4) with Qiagen (Santa Clarita, CA)-purified plasmid DNAs (15 μg/dish) and analyzed 24–36 h later. Cos7 cells plated on 31-mm round or 22-mm2 square coverslips 24–48 h before transfection were transfected (LipofectAMINE; Life Technologies, Gaithersburg, MD) with CsCl-purified plasmid DNAs (dynamitin, 4 μg/ml; dynamitin N terminus; GFP, 1 μg/ml) and analyzed 20–30 h later. DNA was microinjected into Vero cell nuclei using an automated microinjection system (AIS; Zeiss, Thornwood, NY) as described (Scales et al., 1997 ). Microinjected cells were incubated at 37°C for 3–4 h before virus infection.
Vero cells (36–40 h after transfection or 3–4 h after microinjection) were infected with temperature-sensitive Orsay 45 (ts-O45) VSV for 1 h at 17–20°C, 1.5 or 2 h at 20°C, or 1 or 2 h at 37°C and then incubated at the nonpermissive temperature (39.5°C) for 3 h as described (Kreis, 1986 ).
Cells were fixed in −20°C methanol or 3% paraformaldehyde in PBS and then permeabilized with 0.05% saponin, or with 3.7% formaldehyde in PBS and then permeabilized with 0.1% Triton-X-100, before being stained with antibodies and observed on Zeiss Axiovert 35 or TV135 microscopes. Images were recorded as described (Pepperkok et al., 1993 ).
Dynactin integrity was assayed as previously described (Echeverri et al., 1996 ). Briefly, cells on three 10-cm dishes were transfected and grown for 30 h. Transfection efficiency was monitored by including, in each dish, a coverslip that was fixed and stained at harvest. The remaining cells were harvested using PBS-EDTA, washed, and solubilized in an equal volume of lysis buffer (Echeverri et al., 1996 ). Cell lysates were clarified by centrifugation, the supernatants were centrifuged into sucrose gradients, and gradient fractions were analyzed on immunoblots.
α2-Macroglobulin (α2-M; Calbiochem, La Jolla, CA) or Tfn (Sigma, St. Louis, MO) was conjugated with Cy3 or FITC/Fluor-X (Amersham) according to the manufacturer’s protocol. Dye–protein conjugates were isolated on desalting columns; dye:protein ratios of 1:1.25 were obtained. Cells were incubated with Cy3-Tfn or α2-M for analysis of probe uptake, accumulation, or motility.
Before Tfn uptake, cells were serum starved (≥1 h at 37°C). Cells labeled with both C5-dimethyl BODIPY (DMB)-ceramide (Molecular Probes, Eugene, OR) and Tfn were serum starved before Golgi labeling. Cells were then incubated at 37°C with Cy3-Tfn (10–15 μg/ml) for 10 min or Cy3-α2-M (60–200 μg/ml) for 20 min. For uptake and accumulation studies, cells were rinsed for 2 min at 37°C and then fixed (Tfn and α2-M) or chased for 90 min before fixation (α2-M only). For each cell, labeling intensity (bright, dim, or no label) was scored, and the predominant location of fluorescence (e.g., central, random, or peripheral) was determined by eye. For motility studies, coverslips were rinsed by dipping 2 × 50 ml of 37°C HEPES-buffered medium plus 0.2% BSA before imaging.
Cells were serum starved for 1 h at 37°C in medium containing glutamine and 0.5% BSA, then labeled with 25 μg/ml FITC-Tfn for 30 min to 1 h. The coverslips were washed for 1 min and then fixed in 3% paraformaldehyde followed by −20°C methanol. Cells were labeled with Cy3-α2M for 10–30 min, washed, and then chased for 5 min or 1 h at 37°C.
To examine uptake, cells grown on coverslips were serum starved as above and then pulse labeled for 2.5, 5, 10, 20, or 60 min with 50 μg/ml FITC-Tfn. To examine recycling, coverslips were pulse labeled with 50 μg/ml FITC-Tfn for 1 h and then washed for 30 s and chased for 5, 10, 20, or 60 min in 2 ml in medium containing 10% FCS. At the end of each labeling and chase interval, coverslips were washed for 30 s and then fixed in 3% paraformaldehyde followed by −20°C methanol. Cells overexpressing dynamitin were identified using Abs the HA epitope tag. To quantify Tfn fluorescence, the cells were imaged using a charge-coupled device camera with a fixed data collection time. Images were saved as nonByte images. The periphery of each cell analyzed was defined manually. Data were analyzed using IPLab software (Scanalytics, Fairfax, VA) as described (Scales et al., 1997 ).
For video-enhanced fluorescence microscopy (VEFM), cells were grown on 31-mm round glass coverslips. For video-enhanced differential interference contrast (VEDIC) microscopy, cells were grown on 22-mm2 glass coverslips. The cells were transfected, stained with C5-DMB-ceramide (Pagano et al., 1991 ), and then loaded with endocytic tracers as above. To verify that Golgi morphology was a reliable indicator of dynamitin overexpression, cells were stained for Golgi (giantin mAb) and dynamitin (pAb dynamitin). Of 3000 cells evaluated in three separate experiments, <6% in the controls had disrupted Golgi apparatus compared with 96% of dynamitin overexpressers. Cells overexpressing GFP were identified by direct observation.
Cells in HEPES-buffered medium (without phenol red) plus 5% FCS were observed by VEDIC microscopy at room temperature. For VEFM, coverslips were mounted in 37°C medium in a heating stage (Biophysica Technologies, Towson, MD) on a Zeiss Axiovert 35TV microscope and kept covered to minimize evaporation. Temperature was monitored continuously and remained between 32 and 37°C. Cells could be kept on the microscope for longer than 2 h; particle movements continued, and the cells did not develop vacuoles or retraction fibers.
Cells exhibiting bright fluorescence were imaged through a 2× optovar using a silicon-intensified target camera (C-2500; Hamamatsu, Hamamatsu City, Japan) mounted on an intensifier (VideoScope International, Washington DC) to provide additional intensification and magnification. Data was recorded on a frame-addressable Hi8 videocassette recorder (EVO-9650; Sony, Tokyo, Japan). Video fields could be viewed continuously for several minutes with little reduction in particle motility.
Particle tracks were traced from the video monitor onto transparency sheets. Run lengths and start and stop frames for each movement were used to calculate velocities. A movement was defined as a saltation at a single velocity of ≥0.3 μm. For particles undergoing multiple movements, each individual saltation was scored separately. The frustule spacing of the diatom Pleurosigma angulatum provided a magnification standard. Multiple coverslips were labeled with each marker, and 5–10 cells were viewed per coverslip per experiment. For analysis, cells were selected that had similar sizes and shapes. Multiple experiments were performed.
Endosomal pH was measured using an FITC pH ratio imaging technique (Kim et al., 1996 ). Cells overexpressing dynamitin were identified on the basis of Golgi morphology (Texas Red ceramide staining; Molecular Probes), loaded with FITC-α2-M (0.4 mg/ml) for 20 min at 37°C, then mounted in the heating stage of the microscope. Excitation at 440 and 490 nm was provided by a xenon lamp and two monochromators (Deltascan 4000; Photon Technology International, South Brunswick, NJ). The fluorescence intensity (emission cutoff, 535 ± 20 nm) of random fields was imaged with a slow scan charge-coupled device camera (VME200A; Photometrics, Tucson, AZ) and the 490:440 ratio was calculated (Ratiotool software; Inovision, Durham, NC). At the end of each experiment, a calibration curve of fluorescence intensity versus pH was generated in situ (Kim et al., 1996 ) and used to estimate the pH of individual fluorescent particles.
To identify conserved and potentially functionally important domains of dynamitin, we cloned and sequenced the chicken homologue. Its primary sequence predicts a protein of 45,052 Mr and a pI of 4.75 with 80% overall identity to the human protein. Comparison with available dynamitin sequences (human, bovine, budding yeast Jnm1p, C. elegans putative 37.2-kDa protein, and mouse and Drosophila melanogaster expressed sequence tags [ESTs]) reveals several predicted coiled coils and a possible DNA-binding motif (Echeverri et al., 1996 ) that are conserved between species. Dynamitins show the highest similarity (97% human vs. chicken and 65% human vs. fly) in the amino-terminal region (Figure (Figure1)1) and complete conservation of the first 35 amino acids among vertebrates, suggesting that this part of the protein might also be important for function.
Overexpression of dynamitin in cultured cells causes the dynein-binding p150Glued sidearm to be released from the cargo-binding, actin-related protein 1 (Arp1) filament (reviewed by Allan, 1996 ; Schroer, 1996 ; Vallee and Sheetz, 1996 ; Holleran et al., 1998 ). The ensuing perturbation of dynein targeting causes mitotic defects (Echeverri et al., 1996 ) and altered endomembrane organization in interphase cells (Burkhardt et al., 1997 ). We found membranes of the cis-Golgi network/Golgi apparatus (ER-Golgi intermediate compartment-positive/GFP-N-acetyl glucosamine transferase; our unpublished results) and trans-Golgi network (TGN; mannose-6-phosphate receptor-positive; Figure Figure2h)2h) to be disrupted and dispersed upon overexpression of chicken dynamitin. The distributions of endocytic organelles were also altered (Figure (Figure2),2), although not all endosome subcompartments were affected in the same way. Late endosomes (lysobisphosphatidic acid-positive; Figure Figure2d)2d) and lysosomes (CD63-positive; our unpublished results) accumulated in clusters at the cell periphery. Early and recycling endosomes (positive for the early endosome antigen EEA1; Figure Figure2b;2b; Tfn-R; Figure Figure2f)2f) also redistributed away from the center of the cell. Some Tfn-R-labeled structures were seen at the periphery, and some appeared to be clustered, but this redistribution was not as dramatic as for late endosomes and lysosomes. In contrast, endosomes stained for EEA1 were dispersed evenly throughout the cell, which suggests a difference in the behaviors of early and recycling endosomes. No effect was detected on ER structure (calnexin-positive; our unpublished results).
Overexpression of the highly conserved N-terminal 87 amino acids of dynamitin (Figure (Figure1)1) had similar disruptive effects on endosome distribution (our unpublished results) and Golgi complex organization (Figure (Figure3A;3A; 69% of cells; n = 675) to the full-length protein. However, sedimentation analysis of cytosolic dynactin revealed no effect on dynactin structure (Figure (Figure3B),3B), unlike the full-length dynamitin control (Echeverri et al., 1996 ). Apparently, the dynamitin N terminus disrupts membrane organization without altering dynactin structure.
Dynamitin overexpression results in an accumulation of late endosomes and lysosomes at the cell periphery (Burkhardt et al., 1997 ; Figure Figure2),2), whereas microtubule poisons cause these organelles to become randomly distributed. This observation led Burkhardt et al. (1997) to suggest that dynamitin overexpression selectively inhibits dynein-based motility but not movement driven by kinesin and other kinesin family members (the “balance of power” model for endosome and lysosome distribution). This hypothesis is supported by the observation that ER structure, which depends on ongoing plus end-directed motility, is not affected. However, other studies have shown that plus and minus end-directed movements of bidirectional particles are tightly coupled, so both are enhanced or inhibited in unison (Hamm-Alvarez et al., 1993 ; Welte et al., 1998 ). In our analysis, we observed that only some endosomal subcompartments were concentrated at the extreme periphery of dynamitin-overexpressing cells (Figure (Figure2),2), suggesting that the balance of power model might only hold for a subset of endosomes. To gain a clearer understanding of the impact of dynamitin overexpression on bidirectional organelle motility, we turned to live cells in which the movements of endosomes and other subcellular particles could be imaged directly.
Cells were observed using VEDIC and VEFM, and the movements of two classes of organelle were quantified. Cells overexpressing dynamitin were identified by their disrupted Golgi complexes using the vital Golgi dye C5-DMB-ceramide (Pagano et al., 1991 ). As controls, untransfected cells, cells overexpressing cytosolic GFP, and cells in the transfected population that were not overexpressing dynamitin were examined.
Using VEDIC microscopy we analyzed the movement of the conspicuous, highly motile, cytoplasmic particles present in many fibroblasts. Nile Red staining revealed these to be lipid droplets (our unpublished results). Under normal conditions, lipid droplets undergo long- and short-range microtubule-based movements toward and away from the cell center and, on occasion, parallel to the cell margin (Hamm-Alvarez et al., 1993 ; Bulinski et al., 1997 ) These behaviors indicated that lipid droplets are capable of both plus end-directed and minus end-directed movement. High levels of motility were seen in untransfected cells and in the two transfection controls (Table (Table1).1). In contrast, lipid droplets in cells overexpressing dynamitin showed almost no movement. All motility was inhibited, suggesting that more than one motor had been affected. The lipid droplets did not accumulate at the cell periphery (our unpublished results) suggesting that, for these organelles, the balance of power model did not hold.
We then investigated the movement of endosomes, which are known to rely on microtubules and dynein for their steady-state distribution in vivo (Matteoni and Kreis, 1987 ; Burkhardt et al., 1997 ). Microtubules and dynein also contribute to endocytic trafficking and fusion in vivo and in vitro, suggesting that endosome motility might contribute to function as well as steady-state localization (Gruenberg et al., 1989 ; Bomsel et al., 1990 ; Aniento et al., 1993 ). We visualized early endosomes using Tfn conjugated with the brilliantly fluorescent, photo-stable dye Cy3. Cells were briefly loaded with the probe and immediately viewed by VEFM. In control Cos7 and HeLa cells, Tfn first appeared in irregularly shaped structures (0.2–2 μm diameter) in the periphery (“sorting endosomes”) and near the nucleus (“recycling endosomes”). This distribution was similar to that of endosomes stained for Tfn-R (e.g., Figure Figure2f;2f; Gruenberg and Maxfield, 1995 ) or endosomes stained for the peripherally associated early endosomal protein EEA1 (Mu et al., 1995 ; Figure Figure2b).2b). A qualitative assessment of Tfn uptake in dynamitin overexpressers indicated that most cells had internalized the probe (Figure (Figure4B).4B). However, Tfn was delivered to structures that were more peripherally distributed than in controls (Figures (Figures4A,4A, right panel, and 5D). Although these organelles showed some overlap with late endosomes, the two compartments did not completely colocalize (Figure (Figure4A).4A). Quantitative analysis of the Tfn cycle in individual cells (Figure (Figure4C)4C) revealed that dynamitin overexpressers took up and recycled Tfn as efficiently as controls, despite the altered localization of the recycling compartment. This finding corroborates recent work in Madin–Darby canine kidney cells showing that receptor recycling can occur from early endosomes in the periphery (Sheff et al., 1999 ).
In both control and dynamitin-overexpressing Cos7 cells, Cy3-Tfn-labeled particles exhibited mostly short-range oscillations (<0.5 μm; Table Table1;1; Ghosh and Maxfield, 1995 ; Schrader and Schroer, unpublished results). These movements did not appear to be actomyosin based, because they were not inhibited by treatment of cells with the actin poison, latrunculin A, or the broad-spectrum myosin inhibitor butanedione monoxime (our unpublished results). Longer-range (>0.5-μm), presumably microtubule-based, movements were occasionally observed in control cells, but these were not common, and rarer in cells overexpressing dynamitin (Table (Table11).
Unlike early endosomes, late endosomes undergo robust motility that is thought to use the dynein/dynactin motor complex. To test this hypothesis directly, control and dynamitin-overexpressing Cos7 cells were loaded with Cy3-labeled α2-M. This classic endocytic marker (Willingham and Pastan, 1978 ) enters cells by receptor-mediated uptake but, unlike Tfn, dissociates from its receptor in early endosomes (Yamashiro et al., 1989 ) and is trafficked to late endosomes. At early times of observation (0–25 min) Cy3-α2M appeared in structures that resembled those labeled with FITC-Tfn (Figure (Figure5A,5A, middle). After ~25 min, the fluorescent particles in control cells had a variety of sizes and shapes, including round structures (0.1–4.5 μm diameter) and branched or elongated tubules (0.7–4 μm long). These translocated over long distances (up to 6 μm) in curved or straight trajectories (motility is quantified Table Table1;1; Schrader and Schroer, unpublished results). Although the labeled structures were most abundant in the center of the cell, they moved in all directions (inward, outward, and parallel to the cell margin), and many moved bidirectionally, suggesting that they were powered by both plus and minus end-directed motors. Motility persisted for >1 h.
As expected, cells overexpressing dynamitin showed a dramatically different labeling pattern from controls. After a short pulse of α2-M, the intensity (Figure (Figure5C,5C, left) and morphology of labeled particles were similar to control cells, suggesting that uptake and delivery to early endosomes was unaffected. However, the labeled endosomes were either randomly distributed or concentrated at the periphery (Figure (Figure5D),5D), and many colocalized with early endosomes stained for EEA1 (Figure (Figure5A).5A). For the first hour of observation, neither the randomly distributed nor peripheral structures moved to an appreciable degree (Table (Table1).1). Importantly, outward (i.e., plus end-directed) movements were not observed. Similar effects on Cy3 α2-M uptake, localization, and particle motility were seen in most cells overexpressing the dynamitin N terminus (our unpublished results).
After a 60-min chase, most dynamitin-overexpressing cells still contained detectable amounts of α2-M, like the controls. The probe now colocalized with late endocytic markers (Figure (Figure5B),5B), suggesting that delivery to late endosomes had occurred. However, some cells were not labeled, and an increased percentage showed reduced levels of accumulation (Figure (Figure5C).5C). Because nearly all cells took up the probe originally, this finding suggested that early to late endocytic trafficking might be subtly perturbed. If traffic to late endosomes were slowed, α2-M in early endosomes might have more opportunities to be released from the cell.
In most dynamitin-overexpressing cells Cy3-α2-M accumulated in long-lived structures that colocalized with late endosome markers. To further characterize this compartment, we measured its lumenal pH (Figure (Figure5E).5E). For this analysis, cells were allowed to traffic FITC-α2-M to late endosomes for ≥30 min. In both dynamitin-overexpressing and control cells, the labeled structures had a lumenal pH in the range of 5.0–5.5, the expected value for late endosomes. The mean pH increased slightly with increasing chase time, but the pH distribution was always the same as in controls, and the broad range of values obtained made any differences statistically insignificant. Acidification of endosomes containing α2-M did not appear to be appreciably altered by dynamitin overexpression.
Although dynamitin overexpression has profound effects on endosome localization and motility, our findings suggest it does not dramatically alter compartment function. Cells can still take up endocytic probes and recycle them to the surface or deliver them to late endosomes with the appropriate pH. This may not be surprising, because both early and late endosomes in dynamitin-overexpressing cells are near the cell periphery, which would obviate the need for long-range movement between the two compartments. However, some cells showed reduced accumulation of α2-M, suggesting that early-to-late endocytic traffic might be slowed. To determine the effects on trafficking of another endocytic cargo, we examined the cells’ susceptibility to VSV, a virus that must be endocytosed and delivered to a low pH compartment to be infective. Virus infection is inhibited by endosome alkalization or conditions that interfere with endocytic trafficking, such as inhibition of coatomer function by mutation (Daro et al., 1997 ) or antibody microinjection (Whitney et al., 1995 ). For the present experiments, Vero cells were incubated with a temperature-sensitive VSV strain (ts-O45 VSV), and infection efficiency was measured by staining for newly synthesized viral glycoprotein (ts-O45-G; Figure Figure6A).6A). Cells were infected at room temperature and then maintained at the restrictive temperature (39.5°C), which causes ts-O45-G to accumulate in the ER (Bergmann et al., 1981 ). After infection for 1 h, only ~10% of Vero cells microinjected with the dynamitin expression construct showed detectable levels of ts-O45-G in the ER, compared with 80% of cells microinjected with a control vector. Cells containing dynamitin introduced by transfection were also resistant to virus infection; only ~25% of the cells had detectable levels of ts-O45-G protein expression compared with 80% of the transfection controls (Figure (Figure6B).6B). To determine whether virus infection was completely blocked or just delayed, cells were infected for longer times (1.5–2 h at 20°C) or at a higher temperature (1–2 h at 37°C). Under these conditions, nearly 100% of control cells were infected. An increase (to 60–80%; Figure Figure6B)6B) in infected dynamitin-overexpressing cells was also observed. This supports our hypothesis that trafficking through the endocytic pathway is delayed but not completely inhibited by dynamitin overexpression.
Our data verify that the dynein/dynactin motor is a major contributor to microtubule-based movement of endosomes. However, live cell imaging studies reveal that, for several minutes after entering the cell, endocytic probes in early (“sorting”) endosomes remain fairly stationary and do not undergo long-range, microtubule-dependent translocations (Ghosh and Maxfield, 1995 ; Schrader and Schroer, unpublished results). This delay most likely reflects the time required for the probe to transit to the appropriate endosomal subcompartment (endocytic carrier vesicle or late endosome). However, once this has occurred, an additional mechanism is required to bind the endosome to the microtubule track. CLIP-170 has been proposed to serve the initial docking role, because this protein can interact with both endosomes (Pierre et al., 1992 ) and the distal ends of microtubules (Rickard and Kreis, 1990 ). Short arrays of CLIP-170 treadmill at microtubule plus ends as they grow toward the periphery (Perez et al., 1999 ), providing a potential mechanism for attachment of peripheral endosomes and other structures (Dujardin et al., 1998 ).
The p150Glued subunit of dynactin contains a CLIP-170-related microtubule-binding motif that may contribute to dynein processivity by transiently stabilizing the enzyme–cargo link (King, and Schroer, 2000). Given the structural similarities of CLIP-170 and p150Glued (Pierre et al., 1992 ) and their overlapping roles in endosome–microtubule interactions, these two proteins may work in concert at microtubule plus ends. CLIP-170 might promote the formation of a nonmotile, microtubule–endosome complex, which would first recruit dynactin and then dynein. As a first test of this model, we compared the subcellular localizations of CLIP-170 and dynactin in HeLa cells. Previous dynactin localization studies showed a fine, punctate staining throughout the cell and accumulation at centrosomes (Gill et al., 1991 ; Paschal et al., 1993 ; Waterman-Storer et al., 1995 ; Holleran et al., 1996 ). Some of this staining colocalizes with CLIP-170 (Vaughan et al., 1999 ). In double-labeling studies, we also observed dynactin staining that colocalized with CLIP-170 in short linear arrays in the cell periphery (Figure (Figure7,7, a and b).
Both CLIP-170 and dynactin p150Glued are able to bind microtubules directly via their N-terminal microtubule binding sites (Pierre et al., 1992 ; Waterman-Storer et al., 1995 ). According to our model, CLIP-170 binds microtubules first. To test this hypothesis, we overexpressed different dynactin subunits and determined their effects on CLIP-170 localization. If intact dynactin were required, dynamitin overexpression would be expected to alter CLIP-170 distribution. When overexpressed at low levels, dynamitin colocalized with CLIP-170 along microtubules (Figure (Figure7,7, c and d). Even high-level dynamitin overexpression had no detectable effect on CLIP-170 at microtubule plus ends (Figure (Figure7e7e and f), corroborating a recent report (Vaughan et al., 1999 ). These results indicate that CLIP-170 plus end binding does not require intact dynactin.
Dynamitin overexpression causes p150Glued to dissociate from dynactin’s Arp1 backbone (Echeverri et al., 1996 ). The released p150Glued may still be able to interact with potential binding partners, including microtubules. In dynamitin-overexpressing cells, p150Glued is still seen at microtubule ends (Vaughan et al., 1999 ). When p150Glued by itself is overexpressed in cells, it binds microtubules along their length (Waterman-Storer et al., 1995 ; Figure Figure8b).8b). CLIP-170 in these cells is still present in peripheral, linear arrays (Figure (Figure8a),8a), suggesting that CLIP-170 binds microtubule plus ends independently of p150Glued.
Parallel experiments were performed to determine whether CLIP-170 overexpression affected dynactin localization. At low levels, overexpressed CLIP-170 forms small microtubule-associated structures (Figure (Figure8e),8e), and at higher levels, CLIP-170 accumulates in patchy aggregates (Figure (Figure8c;8c; (Pierre et al., 1994 ). These structures also stained for dynactin subunits (p150Glued, Figure Figure8d;8d; Arp1, Figure Figure8f).8f). The localizations of the Golgi complex, late endosomes, and lysosomes were normal in cells overexpressing CLIP-170 (Figure (Figure9).9). This suggests that the actions of dynein and dynactin contribute more significantly than CLIP-170 to the final distributions of these organelles.
In addition to its microtubule-binding domain, CLIP-170 contains a metal-binding motif (Pierre et al., 1992 ) that is implicated in cargo interactions (Pierre et al., 1994 ; Dujardin et al., 1998 ). Deletion of the protein domain containing this motif yields a mutant species that can still bind microtubules but does not accumulate in aggregates (Pierre et al., 1994 ; Figure Figure8g).8g). The mutant protein does not recruit dynactin (Figure (Figure8h),8h), suggesting that the putative cargo-binding domain is required for the interaction. Moreover, dynactin is no longer detected in linear stretches at microtubule plus ends. Quantitative determination of the prevalence of the peripheral, linear dynactin staining revealed it in nearly all of cells in the control population (97%; n = 191) but in only a minority (18%; n = 38) of those overexpressing the mutant CLIP-170 species. This reinforces the idea that CLIP-170 binds microtubule plus ends independently of dynactin and again suggests that the CLIP-170 putative metal-binding domain contributes to dynactin interactions.
Our data provide new insights into the roles of dynein, dynactin, and CLIP-170 in endosome function. The clear but limited effects of dynamitin overexpression on endocytic trafficking refine our understanding of the contributions of microtubule-based movement to this process. CLIP-170 and dynactin colocalize at microtubule plus ends, and binding is found to be CLIP-170 dependent, suggesting a hierarchy of binding. We have also identified the highly conserved dynamitin N terminus as a novel potential dynein-binding element. On the basis of our findings, we propose a series of molecular interactions that may underlie endosome docking and movement. Endosome-associated CLIP-170 provides the initial link to microtubules and recruits dynactin, which then binds dynein. Once dynein is bound and/or activated, CLIP-170 releases its grip on the microtubule, and long-range endosome motility begins.
The discovery that overexpression of the dynamitin N terminus perturbs endomembrane dynamics without affecting dynactin structure was unexpected. Although p150Glued is the only dynactin subunit that has been shown to bind dynein directly (reviewed by Allan, 1996 ; Schroer, 1996 ; Holleran et al., 1998 ), our results suggest that dynamitin may also play a role. p150Glued and dynamitin are tightly associated within the projecting dynactin sidearm (Eckley et al., 1999 ) that is proposed to serve as the dynein-binding site. Dynamitin may stabilize the dynein–dynactin interaction by binding dynein directly. We find (Quintyne et al., 1999 ) that overexpression of different dynactin subunits can have a variety of effects, including Golgi complex and endosome dispersion, disorganization of the interphase microtubule array, and disruption of dynactin structure. Dynamitin induces all these effects, whereas other dynactin subunits such as p150Glued disrupt Golgi structure and microtubule organization without affecting dynactin integrity. These overexpressed dynactin subunits, as well as the p150Glued released by dynamitin overexpression, most likely perturb dynactin function by competing for binding sites on dynein or cargo. The dynamitin N terminus may act in a similar manner.
We suspect that the immediate effect of dynamitin overexpression is to inhibit dynein-based movement selectively, allowing plus end-directed movement to predominate for a short time, in support of a previous hypothesis (Burkhardt et al., 1997 ). However, we observe no plus end-directed particle movements in dynamitin-overexpressing cells, suggesting that, at steady state, kinesin and/or kinesin-related protein-based motility is also inhibited. Late endosomes and lysosomes might possess a latent microtubule plus end binding activity (e.g., CLIP-170) that would explain their microtubule-dependent retention at the periphery (Burkhardt et al., 1997 ; Harada et al., 1998 ). Regardless of the mechanism for endosome relocalization, our results suggest that cells overexpressing dynamitin have arrived at a new steady-state condition in which endosome movement has stopped. Under normal conditions, overall membrane flux is kept in balance so that export parallels import (Steinman et al., 1976 ). The bidirectional movement of individual organelles (e.g., lipid droplets) is also held in balance, because motility in both directions is altered in parallel when cells are subjected to physiological (Hamm-Alvarez et al., 1993 ) or genetic manipulation (Welte et al., 1998 ). Endosome movements may be subject to similar controls. Current models of the mechanism underlying coordinated organelle movement invoke a shared motor receptor (Sheetz et al., 1989 ; Vallee and Sheetz, 1996 ), although other mechanisms are possible. The use of dynamitin overexpression and other dynein-selective inhibitors should prove useful in further studies of this important question.
Our results suggest the additional intriguing possibility that endosome function is governed by a control mechanism that links trafficking with compartment architecture. Precedent is seen in two trafficking-defective Chinese hamster ovary cell lines that exhibit distinct endocytosis phenotypes (McGraw et al., 1993 ; Daro et al., 1997 ). Both show peripheral accumulations of early and late endosomes without any obvious alteration to microtubules. Similar endosome rearrangements are seen in chloroquine-treated chick embryo fibroblasts (Lippincott-Schwartz and Fambrough, personal communication). Apparently, disruption of endocytic traffic and endosome rearrangement are tightly coupled. Although the primary defect is different from cells overexpressing dynamitin (e.g., the ldlF cell line encodes a mutant ε-COP; Daro et al., 1997 ), mutant Chinese hamster ovary cells show alterations in late endosome function similar to those we observe. Endocytic cargoes such as VSV (Daro et al., 1997 ) and ricin (McGraw et al., 1993 ) do not pass from acidic endosomal compartments to the cytoplasm, and delivery of epidermal growth factor to lysosomes is impaired (Daro et al., 1997 ). In dynamitin-overexpressing cells and in mutant cell lines, short-range cycling of material between early endosomes and the plasma membrane continues. Late endosomal membranes may be induced to cycle in parallel when traffic is disrupted. The perturbation of normal mechanisms for forward or inward movement (e.g., budding or dynein-driven motility) might then allow the endosomal compartments participating in these loops to accumulate in the cell periphery near sites of uptake.
Early events in the endocytic pathway, such as ligand uptake and receptor recycling, are found to occur normally in dynamitin-overexpressing cells. However, trafficking to late endosomes is slowed. Under normal conditions, microtubules have been proposed to expedite transfer of material from early to late endosomes, perhaps via endocytic carrier vesicles (Gruenberg et al., 1989 ). Why then, in cells in which these compartments are near each other, should endocytic traffic be impaired? One possibility is that dynein-based motility is required to transport endocytic vesicles in the periphery through actin-rich cortex (Marsh and Bron, 1997 ). The spatial segregation that results from microtubule-based movement may also be required to maintain the distinct functions of different endocytic compartments (Gruenberg and Maxfield, 1995 ; Mellman, 1996 ). Endosomes that have been relocated to the cell periphery may fuse promiscuously with each other, which would result in membrane mixing unless balanced by sorting and retrieval mechanisms. Early and late endosome markers remain distinct in dynamitin-overexpressing cells, indicating that the two compartments are not completely randomized. However, inappropriate exchange of functionally important components or inhibitory factors not examined here might lead to the trafficking delays we observe.
Early and late endosomes are both highly pleiomorphic organelles, yet no clear relationship between structure and function has been established. The membrane deformations induced by microtubule motor activity may, in fact, contribute to membrane fusion. In pure lipid bilayer systems, high degrees of membrane curvature facilitate fusion (Chernomordik, 1996 ). The enhancement of early and late endosome content mixing in vitro seen in the presence of microtubules (Aniento et al., 1993 ) may be another reflection of this phenomenon.
Whatever other roles it may play in trafficking, it is clear that the dynein/dynactin motor is critical for the translocation of endosomal membranes on microtubules. What remains an open question is how endosomes switch from the short-range, oscillatory movements seen early on to the long-range, bidirectional translocations seen at later times. The discovery that dynactin colocalizes with CLIP-170 at microtubule ends suggests an order of assembly of the microtubule–endosome complex. Microtubules extending into the periphery may contact an endosome and become docked via CLIP-170. Once the endosome is tethered in this manner, CLIP-170 can recruit dynactin. At this point, dynactin may simply provide a dynein-binding site, or it may transiently stabilize the endosome–microtubule assembly via its own cargo- and microtubule-binding sites. To switch from this stable binding configuration to one that allows motility, the CLIP-170-microtubule link must be severed, perhaps by phosphorylation. This model provides many hypotheses to be tested in future studies.
We thank C. Crego and N. Jeangunat for technical assistance, Dr. M. Gomez for the microinjection studies, M. Cheng, Dr. M. Eckley, Dr. S. King, and N. Quintyne for Figure Figure3,3, and N. Quintyne for the Nile Red results. Thanks to Drs. J. Gruenberg, H.-P. Hauri, A. Helenius, A. Linstedt, M. Marsh, I. Mellman, O. Rosorius, K. Simons, and T. Suganuma for antibodies and Drs. W. Balch, M. Gomez, W.J. Nelson, J. Rickard, and G. Warren for expression vectors. We thank members of the Kreis and Schroer laboratories for helpful discussions and valuable comments on the manuscript. C.V. was supported by a European Molecular Biology Organization long-term fellowship and Telethon grant 411/bi; D.M.W. was supported by a Howard Hughes Undergraduate Summer Research Fellowship and a Johns Hopkins University Provost’s Award; M.S. was supported by a grant from the Deutsche Forschungsgemeinschaft; T.E.K. was supported by the Fonds Nationale Suisse and the Canton de Genève; and T.A.S. was supported by National Institutes of Health grants GM-44589 and DK-44375 and a Lucile and David Packard Fellowship for Science and Engineering.