Rab10 Regulates Membrane Transport through Early Endosomes of Polarized Madin-Darby Canine Kidney Cells
Rab10, a protein originally isolated from Madin-Darby Canine Kidney (MDCK) epithelial cells, belongs to a family of Rab proteins that includes Rab8 and Rab13. Although both Rab8 and Rab13 have been found to mediate polarized membrane transport, the function of Rab10 in mammalian cells has not yet been established. We have used quantitative confocal microscopy of polarized MDCK cells expressing GFP chimeras of wild-type and mutant forms of Rab10 to analyze the function of Rab10 in polarized cells. These studies demonstrate that Rab10 is specifically associated with the common endosomes of MDCK cells, accessible to endocytic probes internalized from either the apical or basolateral plasma membrane domains. Expression of mutant Rab10 defective for either GTP hydrolysis or GTP binding increased recycling from early compartments on the basolateral endocytic pathway without affecting recycling from later compartments or the apical recycling pathway. These results suggest that Rab10 mediates transport from basolateral sorting endosomes to common endosomes.
The function of eukaryotic cells depends upon a highly orchestrated system of membrane transport in which proteins and lipids are sorted and specifically transported to particular membrane compartments. In the case of epithelial cells, this includes sorting of proteins to either the apical or basolateral surface of the plasma membrane. The processes by which proteins are sorted into transport vesicles and vesicles are formed, transported, and fused with specific target compartments is mediated by the concerted actions of monomeric Rab proteins, SNARE (soluble n-ethyl-maleimide–sensitive factor attachment protein receptor) proteins and their effectors (Rodman and Wandinger-Ness, 2000; Deneka et al., 2003).
Rab10, first isolated from a clone of Madin-Darby Canine kidney (MDCK) cells (Chavrier et al., 1990) belongs to a subfamily of Rab proteins that includes Rab8 and Rab13. This subfamily of Rabs represents the closest mammalian relatives of the Sec4p (Chen et al., 1993; Pereira-Leal and Seabra, 2001), one of the first described Rab proteins, which mediates polarized transport from the trans-Golgi network (TGN) to the site of bud formation in yeast (Salminen and Novick, 1987; Walworth et al., 1989). Like Sec4p, Rab8 and Rab13 have also been associated with polarized membrane transport. Rab8 has been found to regulate transport from the TGN to the basolateral membrane of polarized MDCK cells (Huber et al., 1993; Ang et al., 2003), and both Rab8 and Rab13 have been associated with junctional complex dynamics in epithelial cells (Zahraoui et al., 1994; Marzesco et al., 2002; Lau and Mruk, 2003; Kohler et al., 2004; Morimoto et al., 2005).
In previous studies, Rab10 has been morphologically associated with the Golgi complex of fibroblasts (Chen et al., 1993) and sea urchin embryonic cells (Leaf and Blum, 1998). However, in a recent study, Chen et al. (2006) demonstrate that Rab10 localizes to both endosomes and the TGN and regulates an early stage of basolateral endocytic recycling in the intestinal epithelia of Caenorhabditis elegans. We have characterized the intracellular localization and function of GFP chimeras of wild-type and mutant forms of Rab10 in polarized MDCK cells. These studies demonstrate that Rab10 is closely associated with common endosomes, accessible to both the apical and the basolateral recycling pathways. Expression of mutant Rab10 defective for either GTP hydrolysis or GTP binding increased recycling from early compartments on the basolateral endocytic pathway without affecting recycling from later compartments or affecting recycling of apically internalized IgA. These results suggest that Rab10 mediates transport from basolateral sorting endosomes to common endosomes.
MATERIALS AND METHODS
Studies were conducted using PTR cells, MDCK strain II cells transfected with both the human transferrin receptor (TfR) and the rabbit polymeric immunoglobulin (Ig) receptor (pIgR), previously described (Brown et al., 2000; Wang et al., 2000, 2001). Previous studies have demonstrated that the transfected human TfR is expressed on the basolateral membrane of these cells, that Tf and IgA binding are specific and that the kinetics of Tf recycling and pIgR transcytosis agree well with previously published values for the parental cells (Apodaca et al., 1994) and for another MDCK cell line transfected with the human TfR (Odorizzi et al., 1996).
PTR cells were grown in MEM (Invitrogen, Grand Island, NY) with 8% fetal bovine serum, 1% l-glutamine, streptomycin, and 0.05% hygromycin (Calbiochem, San Diego, CA). Cells were passed every 3–4 d and growth medium was changed daily. New cultures of cells were thawed every 4–5 wk. For fluorescence experiments, cells were plated at confluence on the bottoms of collagen-coated Millipore CM 12-mm filters (Millipore, Bedford, MA) and cultured for 4–5 d before experiments, as described previously (Brown et al., 2000; Wang et al., 2000, 2001). Bottom-seeding was accomplished by creating a cup on the underside of the filter units by wrapping parafilm around the base of the filter unit. The filter units were inverted and cells seeded into this parafilm cup overnight. The following morning, the parafilm was removed, the filters were removed to their normal orientation in 24-well dishes, and the cells were then grown hanging from the filter. Previous studies have demonstrated that monolayers grown in this way become impermeable, with transepithelial resistance reaching 450 Ù/cm2 as measured with a Millipore Millicell ERS resistance meter and with cells developing tight junctions (as assessed by immunofluorescence detection of the tight junction protein ZO-1) and a polarized distribution of TfR.
Plasmids and Transient Transfection
Amino terminal GFP chimeras of rab10 were produced using human rab10 cDNA (UMR cDNA Resource Center; www.cDNA.org), which was PCR amplified with primers containing Gateway recombination sequences attB1 and attB2 and cloned into entry vector pDONR201 per manufacturer’s protocol (Invitrogen, Carlsbad, CA). The rab10 cDNA was then transferred into a Gateway destination vector prepared in-house composed of pCDNA3.1 modified with EGFP coding sequences and a Gateway cloning cassette (complete plasmid sequences available upon request). Mutant forms of rab10 were prepared using the Quickchange kit per manufacturer’s protocol (Stratagene, La Jolla, CA).
An amino terminal YFP chimera of rab10 was generated by amplifying the rab10 gene with PfuTurbo polymerase (Stratagene), using the plasmid above as a template. The amplification primers introduced a 5′ HindIII site and a 3′ EcoRI site, which were used to clone the amplicon into the pEYFP plasmid (Clontech, Palo Alto, CA). An amino terminal CFP chimera of rab11 was generated by amplifying the rab11 gene with PfuTurbo polymerase (Stratagene), using pEGFP-rab11 as a template. The amplification primers introduced a 5′ BglII site and a 3′ KpnI site, which were used to clone the amplicon into the pECFP plasmid (Clontech).
The plasmid expressing pEGFP-tubulin was obtained from Clontech (BD Biosciences, Mountain View, CA). The pEGFP-rab11 plasmid is as described previously (Wang et al., 2001). All plasmids were verified by DNA sequence analysis.
Cells were transfected immediately after seeding onto filters using GeneJammer transfection agent (Stratagene), per manufacturer’s protocol. A ratio of 2.8 μl transfection agent to 0.8 μg plasmid DNA was found to be optimal.
Proteins and Chemicals
Purified dimeric IgA was provided by Prof. J.-P. Vaerman (Catholic University of Louvain, Brussels, Belgium). Human Tf was obtained from Sigma Chemical (St. Louis, MO), and iron was loaded and purified as described in Yamashiro et al. (1984). With the exception of Cy3 and Cy5, which was obtained from Amersham (Arlington Heights, IL), all reactive fluorophores were obtained from Molecular Probes (Eugene, OR). BFA was obtained from Epicenter Technologies (Madison, WI). DiI-LDL was obtained from Biomedical Technologies (Stoughton, MA). Anti-furin antibody was obtained from Alexis Biochemicals (Axxora, San Diego, CA), anti-Rab11a antibody was obtained from Zymed (Invitrogen, Carlsbad, CA) and anti-GP-135 was a gift from Robert Bacallao. TxR-conjugated secondary antibodies were obtained from Jackson ImmunoResearch (West Grove, PA). All other reagents were obtained from Sigma Chemical. Fluorescent ligands were prepared as described previously (Brown et al., 2000; Wang et al., 2000, 2001).
Labeling of Cells with Fluorescent Ligands
For fluorescence labelings, cells were incubated at 37°C on a slide warmer in a humidified chamber for 15 min before addition of fluorescent ligands. All incubations were conducted in medium 1 (150 mM NaCl, 20 mM HEPES, 1 mM CaCl2, 5 mM KCl, 1 mM MgCl2, 10 mM glucose, pH 7.4). As described in each study, cells were labeled with 20 μg/ml fluorescent Tf, 60 μg/ml fluorescent IgA, or 10 μg/ml fluorescent LDL. After incubations with fluorescent ligands, filters were rinsed briefly in phosphate-buffered saline (PBS) at 4°C and then fixed with 4% paraformaldehyde in pH 7.4 PBS at 4°C for 15 min. Filters were then rinsed in PBS. The specificity of receptor-mediated uptake of fluorescent IgA and Tf was demonstrated by its inhibition by competition with excess unlabeled ligands. Fluorescent IgA and Tf were both found to efflux the cells with kinetics similar to those of the radiolabeled ligands. Dual-labeling experiments showed the endosomal distributions of ligands conjugated to different fluorophores to be identical.
All experiments were conducted using a Perkin-Elmer Cetus Ultraview confocal microscope system (Norwalk, CT) mounted on a Nikon TE 2000U inverted microscope, using Nikon 60× NA 1.2 water immersion or Nikon 100×, NA 1.4 oil immersion planapochromatic objectives (Melville, NY). The system is equipped with a Hamamatsu Orca-ER CCD system (Bridgewater, NJ). Image volumes were collected by collecting a vertical series of images, each between 0.2 and 0.6 μm apart. Whenever possible signal saturation was avoided, and objects with saturated pixels were omitted from quantifications.
As mentioned previously, cells were grown on the underside of Millipore filter units. After removing the legs of the filter units, living or fixed cells were observed by placing the entire filter unit on two 50-μm tape spacers attached to the coverslip of a coverslip-bottomed 35-mm dish (Mattek, Ashland, MA) mounted on the stage of an inverted microscope. For live cell studies, incubations were conducted in medium 1 on the microscope stage, with basolateral ligands added to the filter cup, whereas apical ligands were added to the well of the coverslip-bottomed dish. Temperature was maintained with a microscope stage heater, using either a Warner Instruments TC324B (Hamden, CT) or Medical Systems Corp. PDMI-2 (Greenvale, NY). Fixed cells were imaged immediately after fixation in PBS containing 2% DABCO (Sigma Chemical).
For immunolocalizations of furin, Rab11a and gp135, cells were labeled as indicated and then fixed as described above. Cells were then washed three times with 10 mM glycine in PBS, pH 7.4, for 10 min. Cells were then permeablized in 0.025% Saponin for 5 min and nonspecific sites were blocked in 1% BSA + 0.025% Saponin + 0.05% fish skin gelatin in PBS–glycine for 5 min. Cells were incubated with primary antibodies and rinsed three times with PBS followed by the incubation with the TxR-sheep anti-mouse antibody for 1 h at room temperature or with TxR-donkey anti-sheep at room temperature for 1 more hour.
Image Processing for Presentation
Image processing was conducted using Metamorph software (Universal Imaging, West Chester, PA). For images with poor signal-to-noise, images were subsequently averaged spatially using a 3 × 3 low pass filter. Images shown in figures were contrast stretched to enhance the visibility of dim structures, and specific care was taken never to enhance the contrast in such a way that dim objects were deleted from an image. Images to be compared were always contrast enhanced identically. Montages were assembled and annotated using Photoshop (Adobe, Mountain View, CA). Volume renderings were conducted using Voxx, a PC-based image analysis program developed at the Indiana Center for Biological Microscopy (Clendenon et al., 2002) and available at no charge (http://nephrology.iupui.edu/imaging/voxx).
Cross-correlation image analysis was conducted as described previously (Brown et al., 2000; Wang et al., 2001). Three-dimensional (3D) image volumes of each probe were collected. A maximum projection of each volume was constructed, and Pearson’s r between the two images was calculated for a 32-pixel-diameter circular region (corresponding to a 6-μm diameter in the sample) centered over the projected image of the cell. Maximum correlation values were calculated from samples labeled with two colors of Tf (mean r = 0.96), and random values were calculated from these same images in which one region was rotated 90° relative to the other (mean r = 0.03).
Quantification of Rates of Transferrin Recycling
Previous studies have demonstrated that the recycling of transferrin consists of two pathways, a “fast” pathway from an early compartment and a “slow” pathway from the later recycling compartment (Presley et al., 1993; Sheff et al., 1999; Hao and Maxfield, 2000). To evaluate the role of Rab10 in the slow recycling pathway, cells expressing different forms of GFP-Rab10 were incubated with equimolar ratios of basolateral TxR-Tf and Cy5-Tf for 25 min, then in Cy5-Tf alone for an additional 10 min, and then fixed in 4% PFA. For each image volume, each image plane was background corrected by subtracting the median intensity of a 32 × 32 pixel region surrounding each pixel, and then all of the images of the volume were added together. A region-of-interest was drawn around individual transfected or untransfected cells, and the total pixel intensity for each cell was recorded for each probe.
Recycling was quantified as the fraction of the total recycling pool of receptors, as described below. One filter was fixed immediately after the labeling period, and the ratio of the fluorescence of the two different probes was calculated. This ratio was used as a conversion factor to standardize the fluorescence of the two probes into common units. Thus at the beginning of the chase, the ratio of the fluorescence of the two probes was set to 1 via the conversion factor. This factor was then applied to the measurements of cells imaged at the end of the chase period to calculate the total standardized fluorescence for each cell. The fraction of total standardized fluorescence resulting from TxR was then measured, and the recycling rate of each cell was quantified from the departure of this ratio from 50%, the standardized ratio measured for cells imaged before the chase.
To evaluate the role of Rab10 in the fast recycling pathway, cells were incubated with TxR-Tf for 20 min at 37°C, rinsed in ice-cold medium 1 for 10 min, incubated with Cy5-Tf for 2 min at 37°C, rinsed in ice-cold medium 1 for 10 min, incubated with TxR-Tf for 4 min at 37°C, and then fixed in 4% PFA. Image volumes were collected and analyzed as described above. Recycling was quantified from the ratio of Cy5-to-TxR fluorescence, a quantity that corrects for variation in expression of TfR between cells. Because a significant amount of the Cy5-Tf is still on the internalization pathway during the 4-min chase interval, it is not feasible to measure the “starting” Cy5/TxR ratio before recycling. Thus we have quantified recycling relative to that observed at 22°, because recycling is disproportionately inhibited relative to internalization at room temperature (Salzman and Maxfield, 1988; Le et al., 1999; Baravelle et al., 2005).
Quantification of Transferrin Internalization Rates
Cells were incubated for 30 min with Cy3-Tf and then with an equimolar ratio of both Cy3-Tf and Cy5-Tf for an additional 4 min and then fixed in 4% PFA. 3D image volumes were collected and analyzed as described above. The amount of fluorescence detected per mole of each probe was standardized using a factor measured from cells labeled with an equimolar mixture of the two probes, imaged after 30 min of uptake with settings identical to those used for the internalization assay. The rate of Tf uptake was quantified as the fraction of total steady state cell-associated Tf fluorescence resulting from the Cy5-Tf internalized during the pulse period.
Quantification of Rates of Apical IgA Recycling
Unlike TfR, which efficiently recycles, pIgR is both recycled and cleaved from the apical plasma membrane. Thus, unlike TfR, the concept of labeling the cellular pool of recycling pIgR to steady state is not justified. For this reason, IgA recycling studies were designed such that the same cells could be evaluated at the beginning and end of a chase period. Cells were incubated for 30 min with Cy5-IgA in the apical chamber, rinsed, and placed in medium 1, with 100 μg/ml trypsin included in the apical and basolateral media. Multiple fields were then immediately imaged using a computerized microscope stage, and the same fields were imaged again after 12 min of incubation. Analyses of fields collected in rapid succession demonstrated that image collection reduced fluorescence by only 1%, indicating that photobleaching minimally affected quantifications. Image volumes were analyzed as described above, except that background correction was accomplished by subtraction of a local background measured manually. The total cell-associated Cy5 fluorescence was then measured at the initial and final time points, and recycling for each cell was quantified from the fraction of the original fluorescence lost before collection of the final volume. In control studies, apical trypsin was found to rapidly cleave apically exposed pIgR-IgA complexes without affecting cell viability or adhesion, consistent with previous studies (Apodaca et al., 1994; Brown et al., 2000)
Quantification of Rates of Efflux of Basolaterally Internalized IgA
Studies were conducted and analyzed as described above, except that cells were incubated with Cy5-IgA in the basolateral chamber for 20 min, and image volumes were collected at both the beginning of the chase and after 20 min in medium 1.
To investigate the role of Rab10 in polarized MDCK cells, we evaluated polarized MDCK cells expressing amino-terminal chimeras of GFP and wild-type or mutant Rab10. In studies of yeast, a similar chimera of GFP and the highly homologous yeast Sec4p protein was targeted and trafficked like endogenous Sec4p and was capable of rescuing the growth defect in cells expressing defective Sec4p (Schott et al., 2002). For this purpose we developed a method for transiently expressing plasmids encoding GFP chimeras of wild-type and mutant Rab proteins in filter-grown cells (see Materials and Methods). Using this technique, relatively high frequencies of expression of transfected proteins are obtained (10–20%) in fully polarized cells.
To compare the distribution of Rab10 with well-characterized markers of intracellular trafficking, studies were conducted in the PTR line of MDCK cells, because this cell line is transfected with both the human TfR and the rabbit pIgR. As described in our previous work (Brown et al., 2000; Wang et al., 2000, 2001), the polarity and intracellular kinetics of these receptors is similar to that reported in other MDCK cell lines and these transfected receptors permit sensitive labeling of both the basolateral and apical endocytic pathways. The transfected TfR is polarized to the basolateral membrane, allowing the use of fluorescently labeled Tf as a probe of the basolateral recycling pathway. Although newly synthesized pIgR is directed to the basolateral membrane of transfected MDCK cells, from which it mediates apical transcytosis, a significant fraction of pIgR is expressed at the apical plasma membrane, from which it is internalized and recycled with greater than 95% fidelity (Breitfeld et al., 1989; Leung et al., 1999; Rojas et al., 2001). Thus, the pIgR ligand IgA can be used as a sensitive probe of both the basolateral-to-apical transcytotic pathway and the apical recycling pathway (Apodaca et al., 1994; Brown et al., 2000; Wang et al., 2000, 2001).
GFP-Rab10 Associates with Endosomes in Polarized MDCK Cells, But Not with the trans-Golgi Network
In polarized MDCK cells, GFP-Rab10 is found in vesicular compartments throughout the cell. In contrast to previous studies (Chen et al., 1993; Leaf and Blum, 1998), we find that these vesicular compartments do not correspond to the trans-Golgi network (TGN). 3D image volumes collected of MDCK cells expressing GFP-Rab10 and processed for furin immunofluorescence show that GFP-Rab10, the vesicular distribution of GFP-Rab10 is clearly distinct from peri-nuclear tubular distribution of furin in the TGN (Figure 1, A and B). Similar results were obtained using an antibody against gamma-adaptin, another protein enriched in the TGN (our unpublished data).
To determine whether the vesicular structures associated with Rab10 are endosomes, GFP-Rab10–expressing cells were incubated with fluorescent Tf on the basal side of the monolayer. Figure 1, C and D, shows XY and XZ projections of an image volume of GFP-Rab10 in a cell incubated with fluorescent Tf. Comparison of the GFP and Tf images shows that the vast majority of the vesicles associated with GFP-Rab10 contain internalized Tf (arrows in the figures indicate a few examples). Although these figures demonstrate that GFP-Rab10 colocalizes closely with internalized Tf but not with furin, the relative distributions of these three probes can be better appreciated in the 3D renderings of the image volumes presented in Video 1. (Please note that videos will play more smoothly if downloaded and played rather than played directly through the browser.)
Similar results were found over a range of levels of expression of GFP-Rab10. Figure 1, E and F, shows XY and XZ projections of a field containing cells expressing a fourfold range of GFP-Rab10. In both the low- and high-expressing cells, GFP-Rab10 predominantly associates with Tf-containing endosomes. Furthermore, comparison with nontransfected cells shows that the morphology of endosomes containing Tf appears unaltered in cells expressing GFP-Rab10. Although the cells shown in this figure are fixed, identical results were obtained in cells imaged while living (our unpublished data).
GFP-Rab10 Associates with Common Endosomes, at the Junction of Apical and Basolateral Recycling Pathways
Tf has been associated with a variety of different endocytic compartments in polarized epithelial cells, variously referred to as basolateral sorting endosomes, basolateral recycling endosomes, common endosomes, and the subapical compartment (e.g., Apodaca et al., 1994; Odorizzi et al., 1996; Futter et al., 1998; Sheff et al., 1999; Brown et al., 2000; Wang et al., 2000, 2001; Hoekstra et al., 2004). Although some of this variety reflects nomenclature and/or experimental differences, the basolateral recycling pathway followed by Tf clearly includes functionally distinct compartments (Sheff et al., 1999; Brown et al., 2000). To better understand the function of Rab10, further studies were conducted to identify the nature of the endosomes with which it associates.
Our previous studies have demonstrated that Tf is initially internalized into basolateral sorting endosomes, along with ligands such as LDL and IgA (Brown et al., 2000). Subsequently, Tf is directed to common endosomes from which it is recycled to the basolateral plasma membrane (Wang et al., 2000). Two sets of studies were conducted to address the question of whether Rab10 associates with sorting endosomes. In the first, cells were incubated for 2 min with basolateral Tf, a period in which internalized Tf is restricted to the earliest endosomes, located immediately adjacent to the lateral membrane. In the cells shown in Figure 2, Tf is found in lateral compartments in the apical (A) and basal portions of the cells (B). GFP-Rab10 is associated with relatively few of these compartments; the majority of GFP-Rab10 is associated with apical compartments away from the lateral membrane.
These results demonstrate that little Rab10 appears to be associated with the earliest basolateral endosomes. In previous studies of polarized MDCK cells, we have demonstrated that Tf and LDL are both internalized into the same compartments, from which Tf is rapidly sorted (Brown et al., 2000). In this and previous work in fibroblasts (Dunn et al., 1989; Dunn and Maxfield, 1992), we have identified these compartments as sorting endosomes. To determine whether Rab10 could be detected in sorting endosomes, as defined by the combined presence of internalized Tf and LDL, image volumes were collected of Rab10-expressing cells incubated with fluorescent LDL and Tf. The results of these studies are shown in Figure 2, C and D. As in previous figures, the distribution of Rab10 closely matches that of internalized Tf. Close inspection also shows the presence of sorting endosomes, containing both Tf and LDL (indicated with arrows). Consistent with the studies shown in Figure 2, A and B, relatively few of these sorting endosomes are also associated with Rab10 (indicated with blue arrows).
Taken together, our studies suggest that Rab10 is predominantly associated with a downstream recycling compartment. In previous studies, we have demonstrated that this downstream recycling compartment consists of common endosomes that are accessible to apically internalized ligands (Brown et al., 2000; Wang et al., 2000). To address the question of whether Rab10 compartments are accessible to the apical endocytic pathway, cells were incubated for 20 min with fluorescent Tf on the basal side and fluorescent IgA on the apical side of the monolayer. The XY and XZ projections of the image volumes collected of these cells (Figure 3, A and B) show that the Tf-containing compartments with which GFP-Rab10 associates are accessible to apically internalized IgA, thus identifying them as common endosomes.
Previous studies have demonstrated that apical sorting endosomes, the first compartment on the apical endocytic pathway, are minimally accessible to basolaterally internalized Tf (Leung et al., 2000). Therefore, the close association of Rab10 with Tf-containing endosomes suggests that Rab10 is not associated with apical sorting endosomes. The question of whether Rab10 can be found associated with apical sorting endosomes was addressed by incubating cells for 14 min with basolateral TxR-Tf, with Cy5-IgA added to the medium on the apical side of the cells for the final 4 min of the incubation. At this point, the majority of internalized IgA is found in punctate vesicles at the very apex of the cells (yellow arrows in Figure 3C) with relatively little IgA present in lower focal planes (Figure 3D). GFP-Rab10 is not associated with these punctate, apical compartments but, similar to previous figures, is closely associated with medial compartments containing basolaterally internalized Tf. The absence of GFP-Rab10 from the apical IgA-containing endosomes is especially obvious in the projected XZ images shown in Figure 3E. Similar to previous studies (Wang et al., 2000), even after incubations of only 4 min, a fraction of apically internalized IgA accesses Tf-containing common endosomes which, consistent with the previous figure, are associated with Rab10 (blue arrows in Figure 3, D and E).
In Figure 3E, we have outlined the apical extent of the cells, as determined separately from the diffuse cytosolic GFP fluorescence. It is clear in these images that the punctate IgA fluorescence arises from within the cells, rather than from surface-bound IgA aggregates. To ensure that surface-bound IgA does not confuse identification of apical early endosomes, studies were conducted in which cells were incubated for 4 min with apical IgA and then treated with 100 μg/ml trypsin at 4° before fixation, a treatment determined to efficiently strip surface-bound IgA. The results of these studies, shown in Figure 3F, are identical to those shown in the previous figures. IgA is largely restricted to sharply punctate apical endosomes that lack detectible Rab10.
Taken together, our data indicate that Rab10 is not detectibly associated with apical sorting endosomes and is associated with only a fraction of basolateral sorting endosomes. The majority of Rab10 associates with common endosomes.
Unlike Rab11a, GFP-Rab10 Is Not Enriched in the Apical Recycling Endosome
The apical recycling endosome (ARE) is an endocytic recycling compartment that consists of vesicular and tubular compartments that are condensed around the microtubule organizing center just below the apical plasma membrane (Apodaca et al., 1994; Brown et al., 2000; Leung et al., 2000; Hoekstra et al., 2004). The ARE is a distinct compartment in MDCK cells that appears to function specifically in the apical recycling pathway; the ARE is depleted of basolaterally recycling Tf, but enriched in internalized IgA, which accumulates in the ARE before transcytotic delivery or recycling to the apical plasma membrane (Apodaca et al., 1994; Brown et al., 2000; Leung et al., 2000; Wang et al., 2000). The ARE is also strongly associated with Rab11a, which regulates apical membrane traffic (Casanova et al., 1999; Brown et al., 2000; Leung et al., 2000). To determine whether Rab10 associates with the ARE, studies were conducted in which the distribution of GFP-Rab10 was compared with that of basolaterally internalized IgA and Rab11a.
Figure 4, A and B, shows XY and XZ projections of cells incubated with basolateral Tf and IgA for 20 min and then fixed. As in previous figures, comparison of the distribution of GFP-Rab10 and Tf shows a close correspondence between the two. Although some of these compartments also contain internalized IgA (blue arrows, especially in the XZ image projection), IgA is largely associated with the ARE, the collection of compartments located at the tops of the cells (indicated with yellow circles in A and yellow arrows in B). Close inspection of the ARE, particularly in the high-magnification inset, shows that although the distributions of GFP-Rab10 and Tf are very similar, they are both absent from the ARE, as labeled with internalized IgA.
The difference between the distributions of GFP-Rab10 and IgA in the ARE is more apparent when compared with the close correspondence in the distributions of IgA and Rab11a, an ARE-associated protein. Figure 4, C and D, shows XY and XZ projections of a field of cells expressing GFP-Rab11a that were incubated with fluorescent IgA and Tf. In contrast to the images shown above, these cells show a striking colocalization of GFP-Rab11a with internalized IgA, especially in the ARE (indicated with yellow circles in C and yellow arrows in D). As in the previous figure, little or no Tf is detected in the ARE, as labeled with GFP-Rab11a.
A direct comparison between Rab10 and Rab11a is shown in Figure 4, E and F, in which GFP-Rab10–expressing cells were incubated with fluorescent Tf and then processed for Rab11a immunofluorescence. These studies likewise demonstrate that the preponderance of immunolocalized Rab11a associates with the ARE, which lacks both GFP-Rab10 and Tf, whereas GFP-Rab10 closely colocalizes with Tf.
These studies indicate that Rab10, unlike Rab11a, does not associate with the ARE. However, the distributions of multiple probes are difficult to resolve visually, particularly because the morphology of the ARE is variable. To quantify the relationship between the distributions of Rab10, Rab11a, Tf, and IgA, an image cross-correlation analysis was conducted. For this analysis, a projection of the image volume of the cells was produced for a circular region corresponding the position of the ARE. A Pearson’s r was then calculated, comparing the intensities of the two fluorescent probes at each pixel in the region. Consistent with our visual analysis, we find that the correlation between internalized IgA and GFP-Rab10 is approximately half that between IgA and GFP-Rab11a (Figure 5A). Corresponding results are found in the quantitative comparison of the distributions of GFP-Rab10 and GFP-Rab11a with internalized Tf; whereas the distribution of internalized Tf is highly correlated with GFP-Rab10, it is negatively associated with the distribution of GFP-Rab11a in the region of the ARE, consistent with its exclusion from the ARE (Figure 5B).
The validity of this correlation analysis is demonstrated in Figure 5C. When cells are labeled with Tf conjugated to two different fluors, the two probes are expected to codistribute almost perfectly. In fact we measure a mean correlation of 0.96 between the projected images of two different colors of internalized Tf. Random colocalization was modeled by comparing the two images after rotating one by 90°. Analysis of such image pairs reduced the mean correlation to 0.03.
Mutations in the GTP-binding and GTP-Hydrolysis Domains Alter the Localization of Rab10
The function of Rab proteins depends upon a cycle of GTP binding and hydrolysis, with the GTP-bound form typically considered “active” and the GDP-bound form considered “inactive.” To further evaluate the potential functions of Rab10, plasmids encoding canonical site-directed mutant forms of Rab10 were created. One mutant encoded the change of a critical threonine in the GTP-binding domain into an asparagine, creating the T23N mutant. The corresponding mutations in Ras, Rab7, Rab8, Rab11a, and Rab25 have been shown to decrease the affinity of the protein for GTP, resulting in the accumulation of a GDP-bound form (Feig and Cooper, 1988; Peranen et al., 1996; Bucci et al., 2000; Wang et al., 2000). This mutation frequently results in a protein with dominant-negative effects on Rab function; overexpression has been found to alter the function of Rab8 (Moritz et al., 2001) and yeast Sec4p (Walworth et al., 1989), as well as Rab4 (McCaffrey et al., 2001), Rab5 (Stenmark et al., 1994), Rab7 (Feng et al., 1995; Bucci et al., 2000), Rab11 (Ullrich et al., 1996; Wang et al., 2000a), Rab14 (Jununtula et al., 2004), and Rab15 (Zuk and Elferink, 2000). A second mutant was also created, encoding a change from glutamine into leucine at position 68 (Q68L). The corresponding mutation has been found to impair GTP hydrolysis in Ras, Sec4p, Rab11a, and Rab25, thus increasing the amount of GTP-bound form of the protein (Walworth et al., 1992; Frech et al., 1994; Wang et al., 2000). This mutation has been found to have dominant-active effects on the function of a variety of GTPases, with overexpression altering the function of Rab5 (Stenmark et al., 1994), Rab7 (Bucci et al., 2000), Rab8 (Ang et al., 2003), Rab14 (Jununtula et al., 2004), Rab15 (Zuk and Elferink, 1999), and Rab17 (Zacchi et al., 1994).
In contrast to the wild-type protein, GFP-Rab10-T23N was found distributed in the cytosol as well as to intracellular compartments that are not accessible to internalized Tf. Figure 6A shows two such cells, in which the tubular distribution of GFP-Rab10-T23N is distinct from that of internalized Tf. Comparison of these cells with their neighbors shows that expression of the T23N mutant Rab10 has no obvious effect on endosome morphology, nor on the amount of Tf taken up by the cells.
Because the tubular morphology of the compartments associated with GFP-Rab10-T23N resembled that of the TGN, the distribution of the GFP-Rab10-T23N was compared with that of immunofluorescently labeled furin. Figure 6B shows that the tubular compartments with which GFP-Rab10-T23N associates contains furin, indicating an association with the TGN.
Cells expressing GFP-Rab10-Q68L are shown in Figure 6, C and D. These images show that although GFP-Rab10-Q68L remains associated with Tf-containing endosomes, a large fraction is strikingly relocalized to a condensed apical compartment lacking internalized Tf. This compartment is distinct from the TGN, as determined by furin immunofluorescence (Figure 6E).
The redistribution of the Q68L mutant is more easily appreciated in Video 2, which shows 3D renderings of image volumes of Tf-labeled cells expressing either the wild-type or Q68L mutant form of GFP-Rab10.
Although the images shown here were collected from fixed cells, identical redistributions of the T23N and Q68L mutant forms of GFP-Rab10 were obtained in studies where image volumes of living cells were collected (our unpublished data).
GTP-Hydrolysis Mutant Q68L-Rab10 Is Associated with the ARE
When the distribution of GFP-Rab10-Q68L was compared with that of basolaterally internalized Tf and IgA, it was found to closely associate with the ARE, as identified as the apical compartment heavily labeled with internalized IgA, but lacking internalized Tf (Figure 7, A and B). The relocalization of Rab10 to the ARE induced by the Q68L mutation in these cells is more obvious in Video 3, which shows 3D renderings of internalized IgA (labeled red) in cells expressing either wild-type or Q68L forms of GFP-Rab10. Note that, whereas much of the GFP-Rab10-Q68L is associated with the ARE, the remainder appears to associate with Tf in common endosomes.
To compare the distribution of Rab10 with the ARE-associated Rab11a, plasmids encoding YFP chimeras of wild-type and Q68L Rab10 were constructed and cotransfected into polarized PTR cells along with plasmids encoding CFP-Rab11a. As shown in Figure 7, C and D, YFP-Rab10 is distributed to distinct vesicles below the level of the CFP-Rab11a–labeled ARE. In contrast, YFP-Rab10-Q68L is almost completely coincident with CFP-Rab11a in the ARE. This comparison is more easily appreciated in Video 4, which shows 3D renderings of image volumes of cells expressing CFP-Rab11a and either the wild-type or Q68L mutant form of YFP-Rab10. The ARE localization of GFP-Rab10-Q68L was also indicated by its codistribution with immunolocalized Rab11a in an apical compartment lacking Tf (our unpublished data).
The effect of the Q68L mutation on the distribution of GFP-Rab10 was quantified, using the image cross-correlation analysis described above. As shown in Figure 7E, the Q68L mutation nearly doubled the correlation of the distribution of Rab10 with IgA in the region of the ARE, increasing the correlation to match that between Rab11a and IgA (Figure 5A). The redistribution of GFP-Rab10 to the ARE induced by the Q68L mutation was also reflected in a more than twofold increase in the correlation between Rab10 and immunolocalized Rab11a. As expected, the redistribution of Rab10 to the ARE induced by the Q68L mutation decreased its correlation with the distribution of internalized Tf by around fourfold.
Expression of Mutant Forms of Rab10 Has No Effect on Membrane Polarity in Polarized MDCK Cells
The common endosomes have been previously associated with the process of sorting of membrane proteins to the basolateral membrane (Odorizzi et al., 1996; Futter et al., 1998; Sheff et al., 1999; Wang et al., 2000), a process mediated by Rab8, the Rab protein most closely related to Rab10 (Ang et al., 2003). This, combined with the redirection of Rab10 to the ARE by the Q68L mutation, suggested that expression of mutant Rab10 may alter endocytic sorting and/or the basolateral polarity of the Tf receptor.
To quantitatively evaluate the effects of expression of the different forms of Rab10 on intracellular sorting of TfR, we conducted a cross-correlation analysis between the projected images of internalized Tf and IgA in cells expressing the different constructs. In a previous study, we found that brefeldin A blocked basolateral sorting of internalized Tf from IgA and that the resulting mis-direction of Tf to the ARE was reflected in an increased correlation between the distribution of internalized IgA and Tf. The results of these studies are summarized in Figure 8. Similar to our previous studies (Wang et al., 2001), treatment of cells with brefeldin A more than tripled the correlation between the distributions of internalized Tf and IgA, reflecting BFA-induced transport of Tf to the IgA-enriched ARE.
No such dramatic effects were observed in cells expressing the various forms of Rab10. Expression of GFP-Rab10-Q68L had no significant effect, and expression of the GFP chimera of the wild-type protein or T23N mutant had small, but statistically significant effects on the correlation, causing a slight increase and a slight decrease, respectively.
These results suggest that overexpression of GFP chimeras of wild-type and mutant Rab10 has little effect on intracellular sorting of internalized Tf from IgA. To more sensitively evaluate the effect on Tf receptor sorting, we examined the polarity of TfR, as reflected by uptake of Tf from the apical and basolateral sides of the monolayer. In previous work, we found that treatment of cells with brefeldin A increased the expression of Tf receptor at the apical plasma membrane, as reflected in a 10-fold increase in endocytosis of Tf from the apical side of the monolayer (Wang et al., 2001). This result was reproduced in the study shown in Figure 9A, which shows that treatment of cells with brefeldin A results in a significant internalization of Tf from the apical plasma membrane, which is undetectable in control cells. Similar studies of cells expressing wild-type and mutant GFP-Rab10 showed no effects on polarized uptake of Tf; apical uptake of Tf was undetectable in all cases (Figure 9, B–D).
That the expression of GFP chimeras of wild-type and mutant Rab10 had no effect on membrane polarity is also indicated by the observation that the apical polarity of the apical membrane protein GP135 (canine podocalyxin) was, in all cases, unaffected by overexpression of the GFP-Rab10 constructs (Figure 9E).
Expression of Wild-type and Mutant Forms of Rab10 Has No Effect on Basolateral Recycling of Transferrin in Cells Labeled to Steady State
Because both apical and basolateral endocytic pathways traverse common endosomes, proteins associated with common endosomes may regulate the trafficking kinetics of either pathway. To study the effects of expression of wild-type and mutant GFP-Rab10 on endocytic kinetics, we developed single-cell fluorescence methods that are independent of transfection efficiency and also allow correlation of effects with levels of protein expression. These methods are described more fully in Materials and Methods.
To measure the steady state rate of basolateral recycling, cells were incubated in the presence of TxR-Tf and Cy5-Tf for 25 min and then in Cy5-Tf alone for an additional 10 min. Because 35 min is sufficient to label all Tf receptors on the recycling endocytic pathway (Apodaca et al., 1994; Odorizzi et al., 1996; Brown et al., 2000), the sum of the TxR-Tf and Cy5-Tf fluorescence (after standardization) is proportional to the total number of Tf receptors in the cells. Thus the ratio of the fluorescence of TxR-Tf to the sum of TxR-Tf and Cy5-Tf fluorescence can be used to quantify a specific recycling rate (see details in Materials and Methods). Although similar results are obtained if recycling is quantified simply from the mean TxR-Tf fluorescence of the population of cells, the ratiometric quantification has the virtue of standardizing for the number of Tf receptors in the cell. Because the number of Tf receptors varies between cells, this quantification reduces the coefficient of variation in kinetics measurements up to fourfold.
The results of these studies are shown in Figure 10A, which shows that expression of mutant and wild-type GFP-Rab10 had no effect on rates of Tf receptor recycling. Closer analysis showed no correlation between recycling rates and the amount of transfected protein (as measured from total cellular GFP fluorescence), over a 12–30-fold range (our unpublished data). The validity of this assay is demonstrated by the fact that quantitation of these results indicates a half-time of Tf receptor recycling of 7.4 min, very similar to our previous measurement of 8 min using a biochemical assay of 125I-Tf efflux (Brown et al., 2000).
The morphological studies presented in Figures 6, 7, and 9 indicated that expression of mutant forms of Rab10 had no effect on rates of Tf uptake. An approach similar to that described above was taken to evaluate this question quantitatively. For these studies, cells were incubated for 30 min in the presence of Cy3-Tf and then in the presence of both Cy3-Tf and Cy5-Tf for an additional 4 min. As in the recycling assay, the sum of the Cy3-Tf and Cy5-Tf fluorescence (after standardization) is proportional to the total number of Tf receptors in the cells. Internalization is then quantified as the ratio of Cy5-Tf fluorescence divided by the sum of standardized Cy3-Tf and Cy5-Tf fluorescence. Because the rate at which cells acquire Cy5-Tf fluorescence depends not only on the rate of Cy5-Tf internalization, but also on the rate at which receptors reappear at the plasma membrane, this assay is also sensitive to variations in recycling rates. However, because the previous studies showed nearly superimposable recycling kinetics for the different conditions, this assay will report only differences in internalization.
The results of these studies are summarized in Figure 10B, which shows that expression of mutant forms of Rab10 had little effect on the rate of Tf uptake. Expression of GFP-Rab10 caused a small, but statistically significant, 14% decrease in the rate of Tf internalization. The decrease in internalization rate was not related to the amount of cellular GFP-Rab10 over a 25-fold range, as measured from total cellular GFP fluorescence (our unpublished data).
Expression of Mutant Forms of GFP-Rab10 Increases Basolateral Recycling from an Early Compartment
Previous studies have demonstrated that there are two recycling pathways followed by internalized Tf, a slow pathway involving passage through a late compartment and a faster pathway in which Tf recycles directly from an earlier compartment (Presley et al., 1993; Sheff et al., 1999; Hao and Maxfield, 2000). The recycling assay described above will primarily detect differences in recycling from the later compartments, which dominates the efflux kinetics of cells labeled to steady state. To assay the role of Rab10 in recycling from early compartments, the efflux of a brief pulse of fluorescent Tf was measured. In this assay, cells were incubated with TxR-Tf for 20 min at 37°C, rinsed in ice-cold medium 1 for 10 min, incubated with Cy5-Tf for 2 min at 37°C, rinsed in ice-cold medium 1 for 10 min, incubated with TxR-Tf for 4 min at 37°C, and then fixed. Recycling was then quantified from the ratio of Cy5-to-TxR fluorescence in individual cells. Because a significant amount of the Cy5-Tf is still on the internalization pathway during the 4-min chase interval, it is not feasible to measure the “starting” Cy5/TxR ratio before recycling. Thus we have quantified recycling relative to that observed at 22°, a condition under which recycling is slowed (Salzman and Maxfield, 1988; Le et al., 1999; Baravelle et al., 2005). Thus the quantities obtained will provide conservative estimates of recycling, but estimates that are useful for relative comparisons.
As shown in Figure 10C, similar rates of Tf recycling were found in untransfected cells and in cells expressing GFP-Rab10, both of which retained ∼70% of the Cy5-Tf retained by cells incubated at 22 μ. However, the rates of recycling were significantly increased in cells expressing GFP-Rab10-Q68L, and especially in cells expressing GFP-Rab10-T23N, which recycled Cy5-Tf at nearly twice the rate of untransfected cells.
These results suggest that expression of mutant forms of Rab10 increases recycling from an early compartment on the basolateral recycling pathway, a compartment we refer to as the “sorting endosome.” Although as described above, MDCK cells sort basolaterally internalized Tf from IgA, Sheff et al. (1999) demonstrated that recycling via this pathway is relatively nonspecific, mediating the rapid basolateral recycling of basolaterally internalized Tf and IgA with nearly equivalent kinetics. Thus, if expression of mutant forms of Rab10 increase recycling from basolateral early (sorting) endosomes, we would expect to see similar effects on the efflux of basolaterally internalized IgA.
Because only a fraction of pIgR recycles, it is not possible to label the complement of recycling pIgR to steady state. Thus a different approach was used to assay IgA efflux in which the IgA content of the same fields of labeled cells was quantified at the beginning and end of a chase period. The assays of basolateral efflux of IgA are complicated by the fact that although Tf is efficiently recycled to the basolateral membrane, basolaterally internalized IgA is effluxed to both apical and basolateral membranes. Insofar as the study of transiently transfected cells requires single-cell analyses, we cannot distinguish these two pathways. However, similar to the assays of recycling of Tf from early endosomes, the relative contribution of the transcytotic pathway can be decreased by limiting the duration of labeling. The brevity of labeling for these studies is limited by the relatively dim fluorescence of IgA in endosomes, which requires longer incubations to generate sufficient signal, especially to quantify signal decreases. Because the half-time of IgA transcytosis is two to four times slower than the half-time of TfR recycling, with half-times of 20–50 min (Futter et al., 1998; Henkel et al., 1998; Leung et al., 1999; Brown et al., 2000), a 20-min incubation was used, a duration sufficient to provide adequate signal with relatively little IgA in downstream transcytotic compartments. Cells were incubated with fluorescent IgA in the basolateral medium, washed, and then incubated in the absence of IgA for an additional 20 min. Image volumes of the same cells were collected at the beginning and end of the chase interval, and recycling was assayed by the fractional decrease in fluorescence.
The results of these studies, shown in Figure 10D, are very similar to those of the studies of Tf recycling from early endosomes. As in those studies, similar rates of IgA efflux were found in untransfected cells and cells expressing GFP-Rab10, but rates of IgA efflux were significantly increased in cells expressing GFP-Rab10-Q68L, and especially in cells expressing GFP-Rab10-T23N.
Expression of Mutant Forms of GFP-Rab10 Has No Effect on Rates of Apical IgA Recycling
The similar effects of mutant forms of Rab10 on the efflux of basolaterally internalized IgA and on the efflux of Tf from early compartments suggests that the mutants alter basolateral efflux of IgA from basolateral sorting endosomes. However, these studies do not exclude the possibility that expression of the mutants increases efflux of basolaterally internalized IgA from downstream compartments on the transcytotic pathway, such as the common endosomes or apical recycling endosomes. Our previous studies have demonstrated that these later compartments are shared with the apical recycling pathway, as labeled with IgA internalized from the apical plasma membrane (Wang et al., 2000). Thus, if expression of mutant forms of Rab10 increases apical transport of IgA through common endosomes and apical recycling endosomes, we would predict that expression of mutant Rab10 should likewise increase recycling of apically internalized IgA. In contrast, if mutant forms of Rab10 increase basolateral efflux of IgA from basolateral sorting endosomes, which are not included on the apical recycling pathway, they should have no effect on recycling of apically internalized IgA.
For these studies, cells were incubated with fluorescent IgA in the apical medium for 30 min, washed, and incubated in medium lacking IgA for an additional 12 min. Image volumes of the same cells were collected at the beginning and end of the chase interval, and recycling was assayed by the fractional decrease in fluorescence. Trypsin, 100 μg/ml, was included in the apical and basolateral media during the chase, to ensure release of recycled IgA (Apodaca et al., 1994; Brown et al., 2000).
The results of these studies are presented in Figure 10E. Transfection of cells with either wild-type or mutant forms of Rab10 had no significant effect on the rates of efflux of apically internalized IgA. Half-times of recycling varied between 30 min for cells expressing GFP-Rab10 up to 37 min for cells expressing GFP-Rab10-Q68L, but none of the differences were statistically significant. These rates of efflux are similar to, but slightly longer than those previously reported using biochemical methods (∼30 min, Henkel et al.1998; Leung et al., 1999; Rojas et al., 2001). The fact that expression of mutant forms of Rab10 had no effect on apical recycling is consistent with the hypothesis that expression of the mutants increased efflux of basolaterally internalized IgA from sorting endosomes, rather than increasing the rate of IgA transcytosis.
Rab10 Associates with Highly Dynamic Transport Vesicles That Carry Tf between Endosomes
The results above suggest that Rab10 negatively regulates rapid recycling from basolateral sorting endosomes. One possible mechanism might be that Rab10 mediates transport of Tf and IgA from basolateral sorting endosomes to common endosomes. When this function is inhibited, as by the overexpression of mutant Rab10, Tf and IgA may accumulate in sorting endosomes, increasing their direct efflux to the basolateral plasma membrane. According to this model, Rab10 would be expected to associate with transport vesicles. The yeast homologue of Rab10, Sec4p, is predominantly associated with motile secretory vesicles that mediate transport of newly synthesized proteins to the exocyst, the site of new bud formation (Guo et al., 1999; Schott et al., 2002). The studies presented above demonstrate a close association of Rab10 with common endosomes, but provide little evidence of Rab10 in small transport vesicles. To more sensitively detect vesicular Rab10, high-speed confocal imaging of living cells was conducted.
For these studies, 3D image volumes of polarized MDCK cells expressing GFP-Rab10 were collected at a rate of 22 images per second, so that image volumes were collected nearly at the rate of one per second. These studies demonstrate surprisingly dynamic vesicular traffic of GFP-Rab10. They also demonstrate the importance of viewing transfected cells in time series because the vesicles, which are quite obvious in time series, are so dimly fluorescent that they are difficult to distinguish in static images. Figure 11A shows a projected image of the first 3D image volume from a time series collected at the rate of 0.85 image volumes (each consisting of 26 image planes) per second over a period of 60 s. The entire time series is shown as a series of projected images in Video 5a or as a series of stereo anaglyphs in Video 5b, each of which plays at 15 times the rate of collection. As in the figures of fixed cells, GFP-Rab10 can be seen bounding relatively large endosomes, but the movies also show GFP-Rab10 on rapidly moving vesicles.
To determine whether these transport vesicles mediate transport of Tf between endosomes, high-speed imaging was conducted on living PTR cells expressing GFP-tubulin and GFP-Rab10, incubating in TxR-Tf. For these studies, cells were grown on coverslips, which results in a more horizontal presentation of the cell that is more conducive to capturing vesicular dynamics by confocal microscopy. Confocal images were collected, alternately collecting GFP or TxR fluorescence at a rate of four pairs per second over a period of 50 s. The resulting movie, which plays at seven times the actual rate, is shown in Video 6, a single frame of which is presented in Figure 11B. As in the figures of fixed cells, GFP-Rab10 can be seen surrounding relatively large, Tf-containing endosomes, but the movie also shows GFP-Rab10 on rapidly moving vesicles, frequently moving along microtubules between endosomes.
At the end of Video 6, the contrast in one subregion is increased to alternately present the TxR-Tf image and the GFP-Rab10 image. As the time series of this inset is toggled forward and backward, one can see that the GFP-Rab10–associated vesicle contains internalized TxR-Tf and that this vesicle is trafficking between two endosomes.
The association of Rab10 transport vesicles with microtubules is more obvious in Figure 11C, which shows a portion of a cell expressing GFP-tubulin and GFP-Rab10. Confocal images of GFP fluorescence were collected at a rate of five frames per second over a 30-s period. In this figure, the tubulin signal was enhanced by averaging a series of 34 sequential frames, and the sequential positions of two vesicles imaged at 15, 16, 19 and 21 s are displayed in purple, red, yellow, and green, respectively. In the resulting time series (Video 7, which plays at 3.5 times the actual rate) numerous vesicles can be observed clearly moving along microtubules.
Although Rab10 is associated with motile vesicles, expression of either mutant form of Rab10 had no observable effect on the dynamics of endosomes, nor on the association of endosomes with microtubules, suggesting that Rab10 function is not necessary for microtubule-based vesicle motility.
Rab10 Associates with Common Endosomes, Distinct from the ARE
In contrast to previous studies of fibroblasts (Chen et al., 1993), sea urchin embryonic cells (Leaf and Blum, 1998), and C. elegans intestinal epithelia (Chen et al., 2006), the studies presented here demonstrate that Rab10 does not associate with the trans-Golgi network of polarized MDCK cells, but rather with endosomes accessible to both the apical and basolateral recycling pathways. Although the number and nature of the compartments on the recycling pathway of MDCK cells is controversial, quantitative morphological analyses demonstrate that Rab10 associates with common endosomes, distinct from the ARE. Our previous studies demonstrated that the ARE can be distinguished from common endosomes by its accumulation of internalized IgA and depletion of internalized Tf (Brown et al., 2000; Wang et al., 2000, 2001). Consistent with this, a correlation analysis demonstrates that the distribution of the ARE marker Rab11a is highly correlated with that of internalized IgA, but negatively correlated with that of internalized Tf. The same analysis of Rab10 shows that the distribution of Rab10 is highly correlated with that of internalized Tf but not with that of internalized IgA, nor with Rab11a. These studies thus identify Rab10 as a marker of common endosomes and also demonstrate that the ARE and common endosomes can be clearly distinguished by their strong associations with Rab11a and Rab10, respectively.
To evaluate the function of Rab10, we have utilized an approach in which we overexpressed GFP chimeras of mutant and wild-type forms of Rab10. Although this approach must be treated cautiously because the behavior of the transfected protein may be altered by overexpression and/or addition of GFP, it has proved useful in the study of numerous Rab proteins, including the closely related Rab8 (Ang et al., 2003), Rab13 (Marzesco et al., 2002), and yeast Sec4p (Schott et al., 2002). Although abnormal behaviors induced by overexpression cannot be excluded, our morphological and kinetic analyses demonstrated that cells expressing a GFP chimera of the wild-type Rab10 behaved very similarly to untransfected cells and that the effects of the different GFP-Rab10 constructs were consistent over a wide range of expression levels and were restricted to a subset of membrane transport pathways. We were unable to immunolocalize endogenous Rab10 with available antibodies, but the appropriate localization of GFP-Rab10 was supported by observations of identical distributions for exogenously expressed HA- or FLAG epitope–tagged Rab10 (our unpublished data).
Rab10 Mediates Transport from Basolateral Sorting Endosomes to Common Endosomes
Analyses of the endocytic kinetics of Tf and IgA demonstrated that expression of mutant forms of Rab10 increased recycling from early compartments. These results contrast with studies of Rab4, a protein associated with recycling from basolateral early endosomes (Sheff et al., 1999) in which expression of a GTP-binding mutant decreased the rate of Tf recycling (McCaffrey et al., 2001). Although the inhibition of recycling induced by expression of the mutant Rab4 indicates that Rab4 is necessary for recycling, the enhanced recycling induced by expression of mutant Rab10 suggests that Rab10 somehow negatively regulates recycling from basolateral sorting endosomes. Although this may be true, an alternative model for Rab10 function involves regulation of an alternative efflux pathway from basolateral sorting endosomes. It has been previously demonstrated that there are two efflux pathways from basolateral sorting endosomes, one directly to the basolateral plasma membrane and the other to common endosomes (Sheff et al., 1999). To the degree that Rab10 is necessary for transport from sorting endosomes to common endosomes, expression of mutant forms of Rab10 might be expected to disrupt this transport pathway, thereby increasing the amount of Tf in sorting endosomes and increasing its efflux via the alternative, direct pathway to the basolateral membrane. Thus we suggest that Rab10 may mediate transport from sorting endosomes to common endosomes (Figure 12).
This model is consistent with several observations from this and previous studies. First, we predict that recycling from common endosomes should be unaffected by expression of mutant Rab10. Our data indicate that although transport to common endosomes may be slowed in cells expressing mutant Rab10, it is not blocked altogether. Thus even in cells expressing mutant Rab10, the bulk of Tf internalized for an extended period is downstream of sorting endosomes, a distribution that would emphasize efflux from common endosomes. Consistent with our model, we find that expression of mutants has no detectible effect on recycling of Tf from cells labeled to steady state.
Second, Sheff et al. (1999) demonstrated that recycling from early basolateral compartments is nonselective, so that both IgA and Tf efflux via this pathway. Thus we expect that expression of mutant forms of Rab10 should likewise increase efflux of basolaterally internalized IgA. Indeed, we found a pattern similar to the effects of the mutants: as in the studies of Tf recycling from early compartments, expression of Rab10-T23N dramatically increased efflux of basolaterally internalized IgA, with cells expressing Rab10-Q68L showing an intermediate effect.
Third, we have previously demonstrated that the apical and basolateral recycling pathways converge at the level of the common endosomes (Figure 12). Thus the apical recycling pathway should be independent of any effects of mutant Rab10 on transport from basolateral sorting endosomes to common endosomes. Consistent with this, we find that expression of mutant forms of Rab10 has no significant effect on the rates of recycling of apically internalized IgA.
Fourth, we find Rab10 associated with common endosomes and to a lesser extent, with sorting endosomes. Studies of living cells demonstrate that Rab10 associates with Tf-containing vesicles trafficking between endosomes, consistent with its suggested role in sorting endosome-to-common endosome transport.
Finally, these studies are highly consistent with results of a recent study of the role of Rab10 in C. elegans (Chen et al., 2006), who likewise identify a role for Rab10 in the early basolateral recycling pathway of intestinal epithelia. This study demonstrated that Rab10 is associated with endosomes on the basolateral recycling pathway and that Rab10-null mutants accumulate basolaterally internalized cargo and accumulate swollen Rab5-positive basolateral early endosomes, while losing downstream recycling compartments. As in the studies presented here, these observations are most consistent with a model in which Rab10 mediates transport from basolateral sorting endosomes to common endosomes.
Taken together, these studies strongly indicate that Rab10 mediates transport from basolateral sorting endosomes to common endosomes. A variety of Rab proteins have been characterized in polarized MDCK cells (Rodman and Wandinger-Ness, 2000), but Rab10 is the first of these proteins associated with this transport step.
Alternative Model—the Role of the ARE in Basolateral Membrane Recycling
In our model, we have interpreted that early recycling occurs from sorting endosomes, with the later recycling pathway arising from common endosomes (Figure 12). However, there is some controversy about these designations, specifically about whether the common endosomes are in fact distinct from the ARE and thus whether the slow component of Tf recycling occurs via the ARE (Sheff et al., 1999). In large part, we believe that this controversy reflects differences in how the ARE is defined. Previous studies have identified a condensed, peri-centriolar compartment that accumulates Tf in MDCK cells, referring to this compartment as the “recycling endosome” (Ang et al., 2003). Although we likewise observe an accumulation of Tf in apically localized endosomes, we also detect an additional compartment that is enriched in IgA internalized from either the apical or basolateral plasma membrane, regulated to a higher lumenal pH, and depleted of internalized Tf (Brown et al., 2000; Wang et al., 2000). In addition, the studies described here distinguish this compartment by its association with Rab11a, but not with Rab10, which associates with common endosomes. The low concentration of Tf in the ARE is incompatible with a model of slow recycling from the ARE, which would be expected to result in an accumulation of Tf in the ARE, similar to the accumulation of Tf in the peri-nuclear recycling compartment, which mediates slow recycling in CHO cells (Yamashiro et al., 1984; Presley et al., 1993). Rather, the small amounts of Tf in the ARE are consistent with the rates of apical mis-sorting of TfR (Brown et al., 2000), suggesting that the ARE plays no role in basolateral membrane transport. This conclusion is supported by studies demonstrating that proteins associated with the ARE as identified above (Rab11a, Rab25, Myosin Vb) regulate apical but not basolateral membrane transport (Wang et al., 2000a; Lapierre et al., 2001).
Mechanism of the Inhibitory Effects of Mutant Forms of GFP-Rab10 on the Function of Endogenous Rab10
Early endosome membrane transport was significantly disrupted by overexpression of Rab10-T23N, indicating that this predicted GTP-binding mutant has dominant effects on the function of endogenous Rab10. It is generally believed that GTP-binding mutant forms of GTPases inhibit the function of endogenous GTPases by binding and thereby sequestering GTPase exchange factors (GEFs; Feig, 1999). These “dead-end complexes” competitively inhibit the formation of the GTP-bound, activated form of the endogenous GTPase. This mechanism has been demonstrated to underlie the dominant-negative effects observed by expression of GTP-binding mutant forms of Sec4p (Collins et al., 1997; Walch-Solimena et al., 1997), Ras (Feig and Cooper, 1988; Hwang et al., 1993; Schweighoffer et al., 1993), Rab3a (Burstein et al., 1992), Rab3d (Chen et al., 2003), and Ypt1 (Jones et al., 1995). Thus the dominant effects of Rab10-T23N may reflect competitive binding of the mutant protein to a Rab10 GEF.
Similar, although less pronounced effects on early endosome membrane transport were observed in cells expressing Rab10-Q68L. These observations contrast with those of several other GTPases, such as Ras, Rab5, and Rab7, whose function is augmented by expression of “dominant-active” GTPase hydrolysis mutants, resulting in a phenotype opposite that of cells expressing the “dominant-negative” GTP-binding mutant (Barbacid, 1987; Stenmark et al., 1994; Bucci et al., 2000). However, in this respect Rab10 is similar to the closely related Sec4p, whose function is inhibited by overexpression of either the GTP-binding or GTP-hydrolysis mutants (Walworth et al., 1989, 1992). Thus unlike Ras, the function of Rab10 and Sec4p may not depend upon the level of the GTP-bound form so much as upon a cycle of GTP binding and hydrolysis. According to this model, cycling of the endogenous protein is inhibited by sequestration of a Rab effector protein through high-affinity binding to the GTP-hydrolysis mutant.
Alternatively, Rab10-Q68L may inhibit the function of endogenous Rab10 by mislocalizing Rab10 effectors to the ARE. The function of endogenous Ras can be inhibited by expression of a constitutively active form of Ras lacking membrane association determinants, which is believed to result in sequestration of Ras effectors in the cytosol (Stacey et al., 1991).
Although Rab proteins were once considered as markers of specific transport compartments, it is increasingly appreciated that their distributions are dynamic, determined by cycles of GTP binding and hydrolysis (Ali and Seabra, 2005). In our model, Rab10 cycles between sorting endosomes and common endosomes, with GTP-binding mediating association of Rab10 with the membranes of sorting endosomes and GTP hydrolysis mediating its dissociation from common endosomes. The fact that the bulk of Rab10 localizes to common endosomes suggests that GTP hydrolysis is rate-limiting, leading to an accumulation of Rab10 in the target compartment, common endosomes. The predominant association of Rab10 with the target compartment is similar to Rab5, Rab7, and Rab11a, all of which are largely associated with their transport targets, early endosomes, late endosomes, and recycling endosomes, respectively.
At first glance, the striking relocation of Rab10 carrying the Q68L mutation to the ARE is surprising. However, because Rab10-Q68L is effectively locked in the GTP-bound state and thus locked onto membranes, its distribution may well reflect nonselective bulk membrane transport to compartments downstream of common endosomes, including the ARE, much in the same way that bulk membrane markers are transported to the peri-nuclear recycling compartment of fibroblasts (Mayor et al., 1993; Presley et al., 1993).
We find that although the majority of Rab10-T23N is cytosolic, a fraction is frequently associated with the TGN of polarized MDCK cells. This localization may not be functionally meaningful, because the corresponding mutation disrupts the association of Sec4p (the yeast homologue of Rab10) with Rab-GDI, which is necessary for appropriate membrane targeting (Collins et al., 1997). However, the distribution of Rab10-T23N may suggest that this GDP-locked Rab10 interacts with a TGN-localized GEF. To the degree that these interactions vary between cell types or species, localization to the TGN may predominate in some tissues, perhaps explaining the previous observations of Rab10 in peri-nuclear structures adjacent to and partially overlapping with markers of the Golgi complex and the TGN of fibroblasts (Chen et al., 1993), sea urchin embryonic cells (Leaf and Blum, 1998), and intestinal epithelia of C. elegans (Chen et al., 2006). In fact, we occasionally observe Rab10 in the TGN of incompletely polarized cells (our unpublished data), suggesting that its distribution and thus its function may be developmentally regulated.
These results suggest that under certain conditions, Rab10 may play a role in TGN transport. In this regard, it is interesting to note that Rab8, a protein with which Rab10 shares 66% amino acid identity (Chen et al., 1993) likewise associates with recycling endosomes, but regulates basolateral transport of newly synthesized proteins (vesicular stomatitis virus G-protein and LDL receptor), suggesting a role in TGN-to-plasma membrane transport via recycling endosomes. However, in contrast to studies of Rab8, in which expression of the GTP-hydrolysis mutant disrupted the basolateral polarity of the LDL receptor (Ang et al., 2003), expression of mutant forms of Rab10 had no effect on basolateral polarity of TfR nor on apical polarity of gp135.
Taken together, these observations suggest that Rab10 mediates transport from basolateral sorting endosomes to common endosomes, the first protein associated with this transport step. Rab10 is also the first protein marker of common endosomes and when compared with Rab11 clearly distinguishes common endosomes from the ARE. Because we find little evidence of the wild-type protein in any compartment other than common endosomes, the itinerary of Rab10 would apparently involve transient associations of Rab10 with sorting endosomes. Insofar as the GTP-hydrolysis and GTP-binding mutant forms of Rab10 stabilize interactions between Rab10 and its effectors or GEFS, they may be useful in identifying crucial Rab10-interacting proteins, thus illuminating the molecular regulation of Rab10 and basolateral membrane transport in polarized epithelial cells.
This article was published online ahead of print in MBC in Press (http://www.molbiolcell.org/cgi/doi/10.1091/mbc.E05-08-0799) on April 26, 2006.
The online version of this article contains supplemental material at MBC Online (http://www.molbiolcell.org).
apical recycling endosome
Madin- Darby canine kidney cells
polymeric Ig receptor
We thank J. Byars, P. Fonarev, P. Schweinsburg, and Barbara Sturonas-Brown for technical assistance. Microscopy studies were conducted at the Indiana Center for Biological Microscopy. This work was supported by National Institutes of Health (NIH) Grant DK51098, a grant (Indiana Genomics Initiative) from the Lilly Endowment to the Indiana University School of Medicine (K.D.), NIH Grant GM67237, MOD Grant 5-FY02, and support from the Chicago Community Trust Searle Scholars Program (B.G). C.C. was supported by the Anne B. and James B. Leathem Scholarship Fund.