The fate of fluid-phase tracers internalized from the medium in polarized MDCK cells has been studied in detail both in vivo, using confocal and electron microscopy, and in vitro, using a transport assay[4–6]. Apically and basolaterally endocytosed tracers were observed within separate populations of early endosomes, and mixing of these tracers was only detected in the common LE. Recent studies, however, have shown that transferrin (Tf), which is internalized from the basolateral surface of the monolayer, co-localizes, at least in part, with proteins endocytosed from the apical surface[10,47]. Indeed, apically and basolaterally derived tracers were shown to meet in apically located tubular endosomes which are highly reminiscent of the recycling endosomes described by Maxfield and collaborators[12,48] in other cell types.
In vitro fusion properties of MDCK II endosomes
First, we compared the interaction of Tf-labeled recycling endosomes with early endosomes labeled with fluid-phase HRP, using a modified version of the in vitro transport assay we have established previously[33,49]. We used MDCKII cells expressing the human TfR. Previous studies have established that, within minutes after internalization, Tf localizes to early recycling endosomes, consistent with the rapid cycle of its receptor [t1/2=5–10 min; for review see[7,8]]. To ensure efficient labeling of recycling endosomes (ETf) in our in vitro transport assay, biotinylated Tf-HRP conjugate (bTf-HRP) was bound to the basolateral plasma membrane on ice and then endocytosed for 10 min at 37°C. In parallel, the content of apical (Eap) or basolateral (Ebl) early sorting endosomes of a separate cell population was labeled with avidin internalized from the medium corresponding to the appropriate side of the monolayer for 10 min at 37°C, as previously established. Post-nuclear supernatants (PNS) were prepared from each cell population, mixed in the assay in the presence of ATP, and incubated for 40 min at 37°C. If membrane fusion occurred, mixing of the content of the two endosome populations resulted in the formation of an avidin–bTf-HRP complex. The complex was then extracted in detergent, imunoprecipitated with anti-avidin antibodies, and the HRP enzymatic activity of the complex was quantified.
In a first series of control experiments, we tested the fusion properties of apical and basolateral early endosomes, which had been labeled after fluid-phase endocytosis of avidin and biotinylated HRP (bHRP). As previously observed individual elements of each endosomal population exhibited homotypic fusion activity ( Fig. 1a, Eap:Eap and Ebl:Ebl), which is typical of early endosomes both in polarized and non-polarized cells[4,33,49]. Moreover, direct fusion of apical and basolateral early endosomes ( Fig. 1a, Eap:Ebl) did not occur to any significant extent in the assay, consistent with our previous observation. We then used this transport assay to investigate the properties of recycling endosomes (ETf) loaded with Tf internalized from the basolateral surface. As shown inFig. 1a, fusion could be observed between recycling endosomes and both, apical (Eap:ETf) and basolateral (Ebl:ETf) early endosomes. Since direct fusion of apical and basolateral early endosomes did not occur in the assay, yet each population exhibited the capacity to interact with recycling endosomes, one may conclude that the functional properties of recycling endosomes and early sorting endosomes are distinct. However, until now the fusion properties of early endosomes and recycling endosomes in vitro have not been discriminated and both appear to exhibit rabaptin5-dependent fusion activity in non-polarized cells[50,51]. Recycling endosomes were also clearly distinct from LE since they were not fusogenic with LE (LE:ETf,Fig. 1a), whereas late endosomal elements exhibited typical homotypic fusion activity [LE:LE; see].
Figure 1. Fusion properties of endosomes in polarized MDCK cells. (a) PNSs from polarized MDCK cells grown on permeable filter supports were used to study in vitro: 1) the homotypic fusion between elements of early apical (Eap:Eap) or basolateral (Ebl:Ebl) endosomes; 2) the fusion of apical with basolateral early endosomes (Eap:Ebl); 3) the homotypic fusion of late endosomal elements (LE:LE); 4) the fusion of elements derived from Eap or Ebl with Tf-loaded recycling endosomes (ETf); 5) the fusion between ETf and LE. Fusion efficiency corresponds to the fraction of complex formed upon fusion in the assay over the total amounts formed in the presence of detergent. With ATP, black bar; without ATP, open bar. Data are presented as mean±SD (n=5). (b) Time course of Tf internalization (min). In separate cell populations, the content of apical early endosomes was labeled with avidin, whereas recycling endosomes were loaded with bTf-HRP internalized continuously from the medium for the indicated time periods. Acquisition of transfer competence was measured in vitro, as in (a). Black symbol, with ATP; open symbols, without ATP.
The simplest interpretation of our observations is that endocytosed Tf was rapidly transferred from early sorting endosomes, which contained the bulk of endocytosed fluid-phase tracers, to specialized recycling elements, as was shown in other cells[12,48,52,53]. In fact, electron microscopy studies have shown that the tubular endosomal elements, characteristic of recycling endosomes, are only poorly accessible to endocytosed HRP in vivo , whereas they are very rapidly labeled with internalized Tf[48,55]. Indeed, basolaterally internalized Tf arrives within 5 min ( Fig. 1b) in a compartment that is fusion competent with the apical sorting endosome, reaching a plateau after 15 min. These data indicate that internalized Tf is rapidly transferred from basolateral sorting endosomes, which are not fusogenic with apical endosomes, to recycling endosomes, which are competent to undergo fusion with apical endosomes.
The relationship between early sorting endosomes and recycling endosome is still poorly understood, but it becomes increasingly clear that they represent separate compartments [for discussion, see]. The mechanisms of Tf trafficking from sorting to recycling endosomes remain to be elucidated. In vivo experiments in non-polarized cells, however, suggest that these endosomes may communicate by vesicular traffic, perhaps via tubular intermediates. In the text, we will use the term ‘transfer’ when referring to interactions between early sorting endosomes and recycling endosomes, to avoid confusion with homotypic fusion events.
Role of CaM in homotypic fusion and transfer
Only few proteins that regulate the recycling pathway have been identified. Although, CaM antagonists may have multiple targets in vivo, their use suggests that the recycling of receptors depends on CaM in polarized cells[27,32,56]. Similarly, transcytosis and transport to lysosomes were also proposed to be CaM-dependent. Therefore, we used confocal laser immunofluorescence microscopy to determine whether CaM localizes to one of the compartments studied in the fusion assay. For this purpose Alexa™ 488-Tf was internalized from the basolateral medium and HRP through the apical medium for 10 min. Because of the large cytoplasmic pool of CaM, it was necessary to cross-link the internalized HRP first by DAB reaction and extract cytoplasmic CaM prior to fixation (see Materials and Methods). Thereby, the membrane-associated fraction of CaM was analyzed together with the internalized endocytic tracers in a triple labeling experiment ( Fig. 2a–e). CaM partially co-localized with Alexa™ 488-Tf in the supranuclear region underneath the apical plasma membrane ( Fig. 2d) and, to a significantly lesser extent, with HRP that labels apical sorting endosomes and has only limited access to the recycling endosomes ( Fig. 2e). These data suggest that a fraction of CaM is membrane-associated and localizes to recycling endosomes containing endocytosed Tf.
Figure 2. CaM co-localizes with Tf on recycling endosomes. Cells were grown on permeable filter supports. Tf-Alexa™ 488 (a) and HRP were internalized from basolateral and apical, respectively, for 10 min. CaM (b) and HRP (c) were detected with specific antibodies. To reduce the complexity of the image, co-localization of Tf with CaM (d) and HRP with CaM (e) throughout the cells was calculated and is shown as projection of the entire stack of sections together with vertical views of the entire stack (extended focus, filters are indicated by arrowheads). Note that CaM shows considerable co-localization with Tf, particularly in the subapical region of the cells, but much less with apically internalized HRP.
These data prompted us to investigate in our cell-free fusion assay the functional involvement of CaM in the transfer from apical (Eap:ETf) or basolateral (Ebl:ETf) early endosomes to the common recycling endosomes. As shown inFig. 3, in vitro transfer from apical (Eap:ETf) or basolateral (Ebl:ETf) early endosomes to the common recycling endosomes was inhibited by the CaM antagonist W13. In contrast, no inhibition was observed when using the analogue W12, which is 5–10-fold less potent than W13 and has been used as a control for non CaM-mediated effects of the drugs in vivo . These data were further supported by the observation that transfer from either early endosome population to recycling endosomes was also inhibited by the addition of a well-characterized anti-CaM monoclonal antibody ( Fig. 3), known to block many CaM-dependent processes in vitro and in vivo [37,46].
Figure 3. CaM regulates in vitro fusion of endosomes. Effects of CaM inhibitors on homotypic early endosome fusion processes (Eap:Eap and Ebl:Ebl), on heterotypic fusion of apical with basolateral early endosomes (Eap:Ebl), and on transfer from early endosomes to recycling endosomes (Eap:ETf and Ebl:ETf). Internalization was carried out for 10 min as described in Materials and Methods. To compare different experiments, the result of each fusion reaction is expressed as a percentage of the untreated control (+ATP) level. Fusion of apical with basolateral endosomes, which normally does not occur, is expressed as a percentage of the mean value obtained for the +ATP controls (Eap:Eap, Ebl:Ebl, Eap:ETf and Ebl:Etf) measured in the same experiments (see alsoFig. 1a). W12, W13, and mAb-anti CaM (anti-CaM antibody); other symbols as above. Hatched bars indicate parallel fusion reactions in the presence of 30 nM of exogenous CaM added to the reaction mixture. Data are presented as mean±SD (n=5).
Interestingly, both treatments (W13 and anti-CaM monoclonal antibody) had stimulatory, rather than inhibitory, effects on homotypic early sorting endosome fusion events (Eap:Eap and Ebl:Ebl,Fig. 3), pointing to the existence of a distinct CaM-dependent mechanism in early endosome dynamics. Moreover, both W13 and anti-CaM antibodies enabled apical and basolateral early endosomes to fuse with each other (heterotypic fusion,Fig. 3), a process which otherwise does not occur ( Fig. 1a). We demonstrated that these effects were specific to early endosomes, since the treatments did not cause Tf-containing endosomes to fuse with LE (data not shown). The fact that early endosome fusion and transfer reactions were affected in an opposite manner by both antibodies and antagonists further demonstrates that these two processes are different. Whereas further work is needed to understand these differences, the effects of antagonists and antibodies were clearly specific. Confirmation that CaM mediates the effects was obtained by addition of purified exogenous CaM (Ca2+ was present at physiological levels from rat cytosol), which completely reversed both inhibitory and stimulatory effects induced by drugs or antibodies ( Fig. 3, hatched bars), but had no effect on control, untreated reactions (not shown). We also observed that neither transfer reactions, nor the inhibitory effects of W13 and anti-CaM antibodies, were affected by the so-called slow Ca2+ chelator EGTA (data not shown, see also) suggesting a Ca2+-independent regulatory function of CaM. Similar observations by using EGTA have been obtained recently by another group; however, fusion was inhibited when the fast Ca2+ chelator BABTA was applied. BABTA is a membrane-permeant chelator that therefore also reaches internal Ca2+ pools. These observations led the authors to conclude that fusion of early endosomes, participating in recycling of synaptic vesicles, requires local release of Ca2+ from the endosomal interior. The effects of CaM inhibitors in our in vitro assay agree well with in vivo observations that CaM antagonists increase the size of endosomes and cause mixing of apical and basolateral endosomes to occur, but inhibit TfR recycling[27,56].
Myr4 on MDCK endosomes
Our next step was to identify possible target(s) of CaM in our assay. Since CaM is known to bind (as light chain) to unconventional myosins in a Ca2+-independent manner, we reasoned that a member of this superfamily may function as a target for CaM in our experiments.
Fractions containing the small GTPase rab5, a marker of early endosomes, and the TfR (not shown), were prepared by flotation in a sucrose gradient solubilized under mild conditions (0.2% NP-40), and then incubated in the presence or absence of Ca2+ with purified CaM immobilized on Affi-gel 15 agarose. The resin was washed and eluted either with high salt (and EGTA for Ca2+-independent binding) or with EGTA (for Ca2+-dependent binding). A major protein with an apparent mobility of 110 kDa was retained on the column in a Ca2+-independent manner, and was then identified as the unconventional myosin myr4 using specific antibodies ( Fig. 4a). Indeed, myr4 is known to bind CaM as a light chain, and to contain one Ca2+-independent CaM binding site, in addition to a Ca2+-dependent site. We then tested whether other unconventional myosins were also present in the same endosomal fraction obtained from the gradient, and could detect, in addition to myr4 also myr 1, myr2 and myr3, but not myr5 and myr6 ( Fig. 4b). Consistent with these biochemical observations, confocal laser scanning immunofluorescence microscopy showed that myr4 was present on recycling endosomes labeled with Tf-Alexa™ 488 internalized for 10 min, mostly in the supranuclear region underneath the apical plasma membrane ( Fig. 5). However, the co-localization of myr4 seems also to extend towards the basolateral sorting endosomes along the lateral plasma membrane (compare withFig. 2).
Figure 4. Myr4, a CaM binding protein on endosomes. (a) Sucrose gradient purified endosomal fractions were detergent solubilized (0.2% NP-40) and tested for CaM binding activities by CaM affinity chromatography. Fractions were incubated with CaM Affi-gel 15 agarose either in the presence (100 μM CaCl2) or absence (2 mM EGTA) of Ca2+. Retained proteins were eluted as indicated, precipitated with acetone and applied to SDS-PAGE (10%) followed by western blotting. A typical elution profile is shown. The protein retained on the column after Ca2+-independent binding was identified by immunoblotting as myr4. (b) Other unconventional myosins (myr1-6) were mapped in endosomal fractions by western blotting with specific rabbit polyclonal antisera (seeTable 1). MDCK homogenate (H), PNS (P) and endosome fractions (E). The band which migrates similar to myr4 is a breakdown product of myr3 since the myr3 protein is easily degraded at the C-terminus. Antibodies used are Tü29, Tü49, FML5, Tü61, Tü66 and SA1354 ( Table 1).
Figure 5. Myr4 and internalized Tf co-localize in the apical region of MDCK cells. MDCK Cells were grown on permeable filter supports. Tf-Alexa™ 488 was internalized through the basolateral medium for 10 min and myr4 detected by indirect immunofluorescence. Images represent projections of the most apical sections showing internalized Tf (a) and myr4 (b). Co-localization of the two signals is shown as projection of the entire stack in the x/y and x/z direction (a:b) as indicated. The filter level is indicated by an arrowhead. Size bar, 10 μm.
Myr4 regulates transfer from early endosomes to recycling endosomes in vitro
Using a panel of anti-myr antibodies ( Table 1), we then tested whether the myr proteins which were enriched in our gradient fraction were involved in the transport reactions we have studied. As shown inFig. 6, the transfer from apical or basolateral early endosomes to recycling endosomes (Eap:ETf and Ebl:ETf) was strongly inhibited by the anti-myr4 antibodies (Tü12 and Tü13, seeTable 1). The peptide these antibodies were raised against corresponds to a specific extension in the myosin I tail homology motif, which is proposed to mediate membrane association (seeTable 1). In contrast, no inhibitory effects were observed in the assay when using antibodies against other regions of myr4, or against other members of the myosin I family (not shown; seeTable 1), including an antibody blocking the actin-activated ATPase activity of myr3, which was also present in our fractions. None of the antibodies, including anti-myr4 antibodies, had any effect on the homotypic fusion processes (Eap:Eap and Ebl:Ebl), demonstrating that the effects of anti-myr4 antibodies on the transfer reactions from sorting to recycling endosomes were specific. This also suggests that the effects of CaM inhibitors on homotypic fusion processes (seeFig. 3) are myr4-independent. These observations indicate that the actin-based mechanoenzyme myr4, which can bind two CaM molecules as light chains, is involved in trafficking from apical and basolateral early endosomes to recycling endosomes in MDCK cells.
Table 1. List of anti-myr antibodies
|G371||Cons. of brush border myosin I||YLGLLENVRVRRAGYAYRQ||AA 611-629 (head domain)|||
|Tü29||Myr1||MAKKEVKSSLLDNM||Head domain||[40,41] Tü30 |
|Tü25||Myr2||His-fusion protein||AA 585-685 (head domain)||Unpub.|
|Tü49||Myr2||CFDKSELSDKKRPET||AA 540-564 (head domain)|||
|FML5||Myr3||Fusion protein||AA 821-1107 (tail domain)|||
|Tü58||Myr3||RQMDSKWGGKSESIHVT||AA 326-342 (head domain) blocks Actin-activated ATPase activity|||
|Tü12||Myr4||GTFVPVANELKRKDKYMN||AA 839-856 (tail domain)|||
|Tü13|| || || || |
|Tü14||Myr4||AA 513-1006||C-term half (incl. Tail domain)|||
|Tü59||Myr4||TGRDIIDKQHTEQE||AA 322-335 (head domain)||Unpub.|
|Tü61|| || || || |
|482||Myr4||DFTKNRSGFILSVPGN||AA 991-1006 (C-terminus, tail)||Unpub.|
|G427||Myr4||ARRFHGVKNMRDYGKHVK||AA 744-761 (regulatory domain)||Unpub.|
|SA25||Myr4||DFTKNRSGFILSVPGN||AA 991-1006 (C-terminus, tail)||Unpub.|
|Tü66||Myr5||AA 1857-1980||C-terminal region of protein|||
Figure 6. Myr4 regulates transfer from early endosomes to recycling endosomes. Effect of anti-myr antibodies on homotypic fusion between early apical (Eap) and basolateral (Ebl) endosomes and on transfer from Eap or Ebl to recycling endosomes (ETf). Experimental conditions are as inFig. 1a. Only anti-myr4 antibodies Tü12 and Tü13 were inhibitory in our assay measuring transfer from early endosomes to recycling endosomes in vitro. Antibodies against other myosin I molecules, including myr1 (Tü29, Tü30), myr2 (Tü25, Tü49), and myr3 (FML5, Tü58) were without effects. The myr5 protein was not detected in MDCK cells using the SA1354 antibody, and myr6 was not enriched in endosomes, using the Tü66 antibody ( Fig. 3b). To compare different experiments, values are expressed as percentages of the controls without antibodies (+ATP). Symbols as inFig. 1B; data are presented as mean±SD (n=3). Antibodies used were Tü29, Tü25, Tü58, Tü12, Tü13 ( Table 1).
Actin dependence of transfer from early to recycling endosomes in vitro
Since myr4 is known to interact with F-actin, our observations raise the possibility that the myr4-dependent transfer from apical or basolateral early endosomes to recycling endosomes also depends on actin filaments. A fluorescence analysis after FITC-phalloidin staining revealed that actin filaments, in contrast to polymerized microtubules, were indeed present under our in vitro conditions (not shown). Formation of filaments in the assay could be prevented by monomer sequestration in a 1:1 complex with DNase1 [not shown; see]. Alternatively, existing filaments could be severed with recombinant gelsolin or with the S1-3 active domain of gelsolin [not shown; see[42,61]]. S1-3 binds two actin molecules in the presence or absence of Ca2+ and it severs and caps filaments like full-length gelsolin although it does not nucleate polymerization. Under both conditions, transfer from either early endosomal population to common recycling endosomes was significantly inhibited ( Fig. 7, Eap:ETf and Ebl:ETf). Therefore, our data suggest that, after disruption of the cellular organization in our assay, dynamic interactions between endosomal elements can be facilitated by the presence of polymerized actin filaments. Similarly, taxol-stabilized microtubules were shown to facilitate in vitro interactions between endosomes at late stages of the pathway[4,45]. Our in vitro findings agree well with the in vivo observations that actin filaments facilitate Tf recycling in a hepatoma cell line and with actin sequestering experiments performed in perforated A431 cells. This role of actin filaments also agrees with their in vivo localization to a sub-cortical region in MDCK cells, which contains early endosomes and recycling endosomes[10,15,27]. We stained MDCKII cells grown on tissue culture filters were for CaM and myr4 using specific antibodies and for F-actin using phalloidin-Alexa™ 568. All three markers co-localized in the supranuclear region underneath the apical plasma membrane (data not shown). Altogether, these observations strongly suggest that actin filaments are involved in the endocytic/recycling pathway.
Figure 7. Actin dependence of fusion and transfer events. Conditions of internalization and in vitro assay were as inFig. 1a. DNase1 (1, 5, 10 μg), gelsolin (15 μM), S1-3 domain (15 μM) or dialysis buffer were added prior to the assay and incubated for 30 min on ice in the assay mixture containing endogenous amounts of Ca2+. To compare different experiments, fusion is expressed as a percentage of the untreated control level (+ATP, black bar). Data are mean±SD (n=5).
In conclusion, our data show that, in MDCK cells, recycling endosomes can communicate with both apical and basolateral early endosomes, and that this process differs at the molecular level from the homotypic fusion process within the early endosomal compartments. Transfer reactions depend on actin, CaM and myr4, an actin-based mechanoenzyme, in contrast to the transport between early and LE, which depends on microtubules and other mechanochemical motors[4,45].
It is unclear at present whether myr4 in our system acts as a motor protein and moves endosomal vesicles on actin cables, or whether the protein tethers or anchors endosomal membrane domains on actin filaments and thereby contributes to the dynamics or the functional organization of the early endosomal membrane system. These possible functions of unconventional myosins in membrane dynamics are still debated, and have been discussed in detail by Coudrier et al. and Mooseker and Cheney. Recently, evidence was provided that the vesicle delivery function of Rho3 in budding yeast is mediated by the unconventional myosin Myo2. In addition, molecular genetic analysis of the Dictyostelium unconventional myosin VII reveals that this motor protein plays a specific and significant role in phagocytosis. While this paper was under review, Durrbach et al. reported the involvement of an acto-myosin-driven mechanism in trafficking of basolaterally internalized molecules to the apical plasma membrane. Brush border myosins might also be involved in membrane trafficking occurring between endosomes and lysosomes in unpolarized cells. These observations, together with our results, indicate that unconventional myosins underlie the elusive crosstalk between endosomes and the actin cytoskeleton that is required for membrane trafficking along the recycling pathway in polarized epithelial cells.