Rab1a and multiple other Rab proteins are associated with the transcytotic pathway in rat liver.

To better understand the function of Rab1a, we have immunoisolated Rab1a-associated transport vesicles from rat liver using affinity-purified anti-Rab1a-coated magnetic beads. A fraction enriched in endoplasmic reticulum (ER) to Golgi transport vesicles (CV2, ρ = 1.158) was subjected to immunoisolation, and proteins of the bound and non-bound subfractions were analyzed by Western blotting. To our surprise, we found that immunoisolated vesicles contained not only ER markers (105-kDa form of the polymeric IgA receptor (pIgAR)) but also transcytotic markers (dIgA and the 120-kDa form of pIgAR), suggesting that Rab1a is associated with transcytotic vesicles in rat liver. To investigate this possibility, we used an antibody to the cytoplasmic domain of pIgAR to immunoisolate transcytotic vesicles from a fraction (CV1, ρ = 1.146) known to be enriched in these vesicles. Rab1a was detected in the immunoadsorbed subfractions. The composition of the vesicles immunoisolated from the CV1 fraction on anti-Rab1a was similar to that of transcytotic vesicles immunoisolated from the same fraction on pIgAR. Both were enriched in transcytotic markers and depleted in ER and Golgi markers. The main difference between the two was that those isolated on anti-Rab1a appeared to be enriched in postendosomal transcytotic vesicles, whereas those isolated on pIgAR contained both pre- and postendosomal elements. Analysis of anti-Rab1a isolated vesicles using [α-32P]GTP overlay demonstrated the presence of multiple GTP-binding proteins. Some of these were identified by immunoblotting as epithelia-specific Rab17 and ubiquitous Rabs1b, −2, and −6. Taken together, these results indicate that: 1) Rab1a is associated with both ER to Golgi and postendosomal transcytotic vesicles, and 2) multiple GTP-binding proteins are associated with each class of isolated vesicle.

So far more than 30 Rab proteins have been identified, and some of them have been localized to specific organelles and transport vesicles in mammalian cells (8 -12). In addition, in vitro and in vivo studies have demonstrated that each transport step in the exocytic and endocytic pathways involves at least one Rab protein. For example, Rab1a, Rab1b, and Rab2 are located in the ER 1 to Golgi intermediate compartment and cis Golgi cisternae and are required for ER to Golgi and in-traGolgi transport (13)(14)(15); Rab4 and Rab5 are present in early endosomes (11,16) and are involved in regulation of vesicular transport between the plasma membrane (PM) and early endosomes (17)(18)(19); and Rab7 and Rab9 are associated with late endosomes and are required for transport to or from late endosomes (20 -22).
Although common transport pathways in most cells use ubiquitously expressed Rabs, highly differentiated cells with unique transport pathways may require specific Rab proteins. To date several tissue-or cell type-specific Rabs have been identified (23)(24)(25)(26). For example, Rab3a is preferentially expressed in neuroendocrine cells and neurons, where it is specifically associated with synaptic vesicles and appears to mediate neurotransmitter release (23,24). Rab3d is mainly expressed in adipocytes and pancreatic acinar cells and is thought to control regulated exocytosis of glucose transportercontaining vesicles and zymogen granules (25,27). Rab17 is specifically expressed in epithelial cells, where it is localized at the basolateral PM and in apical endosomes and is assumed to be involved in regulation of transcytosis across epithelial cells (26).
It is not yet clear whether one Rab or a set of Rabs are required for each transport step. Recent studies have localized more than one Rab in the same vesicular carriers and related intracellular compartments (28,29), suggesting that multiple Rabs may associate with vesicular carriers operating at the same relay. If multiple Rabs are present on a single class of transport vesicles, are the sets of Rabs relay-specific or is there overlap? To address these questions, isolation and characterization of distinct classes of homogeneous vesicular carriers and identification of their associated Rab proteins are required.
An excellent system for studying this problem is the hepatocyte in which the major intracellular transport pathways (exocytosis, endocytosis, and transcytosis) are well characterized (30,31). In addition, specific markers (different forms of the polymeric IgA receptor (pIgAR) and dimeric IgA (dIgA)) are available for vesicles operating along these pathways (32)(33)(34).
In this study we have immunoisolated Rab1a-associated ves-icles from rat liver fractions and, to our surprise, found that Rab1a is associated with vesicles involved in transport along both the exocytic and transcytotic pathways, leading us to postulate that transcytosis may be regulated through interactions between specific (Rab17) and ubiquitous Rabs (Rab1a and others).

EXPERIMENTAL PROCEDURES
Materials and Antibodies-Sprague-Dawley rats (Bantin & Kingman, Fremont, CA), weighing 150 -250 g, were used in these experiments. Magnetic beads (M500) were from Dynal Inc. (Lake Success, NY). S107 hybridoma cells were a gift from Dr. D. Bole (University of Michigan). Recombinant Rab1a and Rab1b proteins (15,35) and antibodies for Rab1, rabbit polyclonal p68, and mouse monoclonal m5c6b (13) were generously provided by Dr. W. Balch (Scripps Research Institute); recombinant Rab6 was kindly provided by Dr. B. Goud (Institut Pasteur). Affinity-purified anti-peptide antibodies to Rab1a and Rab6 were purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). Affinity-purified antibody to the cytoplasmic tail of pIgAR was prepared as described (36), affinity-purified anti-peptide antibody against Rab2 was generously provided by Dr. J. Larkin (Columbia University), antisera against Rab17 and Rab18 were kindly provided by Dr. M. Zerial (European Molecular Biology Laboratory), antiserum for 2,3-sialyltransferase was a gift from Drs. J. Barasch and Q. Al-Awqati (Columbia University), and monoclonal antibodies to ARF (ID9) and EGF receptor (clone 13) were gifts from Drs. R. A. Kahn (National Institutes of Health) and G. Gill (University of California, San Diego), respectively. Affinity-purified polyclonal antibodies for BiP and rat dIgA were purchased from Affinity Bioreagents Inc. (Neshanic, NJ) and the Binding Site Inc. (Birmingham, United Kingdom), respectively.
Preparation of Radiolabeled dIgA and Biotinylated Asialofetuin-S107 hybridoma cells, which secrete dIgA, were metabolically labeled with 100 Ci/ml Tran 35 S-label (ICN) for 20 h at which time medium was collected, and dIgA was concentrated with a Centriprep 30 concentrator (Amicon) to 0.5 ml (1.5 ϫ 10 9 cpm/ml) for use as a specific transcytotic vesicle marker.
Asialofetuin was biotinylated using an immunoprobe biotinylation kit (Sigma) according to the manufacturer's instructions. After incubation the biotinylated protein was passed though a Sephadex G-25 column to remove free biotinamidocaproate-N-hydroxysulfosuccinimide, then concentrated with a Centriprep 10 to 4 mg/ml. The level of biotinylation was determined by the avidin-HABA assay provided in the kit. The molar ratio of biotin/asialofetuin was 1.5.
In Vivo Labeling-Rat liver was labeled in vivo by injection through the portal vein. 2 mCi of Tran 35 S-label (to label ER to Golgi and intraGolgi transport vesicles) or 7.5 ϫ 10 8 cpm dIgA (to label transcytotic vesicles) were injected into the portal vein 15 or 30 min before removing the livers. Similarly, 2 mg of biotinylated asialofetuin (B-AF, to label endosomes) was injected 6 min prior to removing the livers. The livers were flushed in situ with cold 0.25 M sucrose supplemented with protease inhibitors (chymostatin, leupeptin, antipain, pepstatin, all at 1 g/ml), then removed and homogenized for subcellular fractionation.
Subcellular Fractionation-Liver homogenization and preparation of cytosol and total microsomal membranes were as described by Saucan and Palade (36). The procedure for preparation of Golgi and vesicular carrier-enriched fractions was modified from Sztul et al. (37) by adding one more layer of sucrose (1.18 M). Briefly, total microsomal membranes were adjusted to 1.24 M sucrose and loaded at the bottom of a 32-ml discontinuous sucrose gradient with 8 ml each of 1.18, 1.14, 0.86, and 0.25 M sucrose and centrifuged at 82,000 ϫ g (25,000 rpm, SW28 rotor) for 3 h. Bands at the interface between 0.25 M/0.86 M and 0.86 M/1.14 M sucrose, which are enriched in Golgi elements, were collected and designated as Golgi light and Golgi heavy fractions (36). Fractions 1.14 and 1.18 were defined as carrier vesicle fraction 1 and 2 (CV1 and CV2), respectively, and fraction 1.24 as the residual microsome fraction. The protein concentration of each fraction was determined by BCA assay (Bio-Rad).
Immunoisolation-Magnetic beads (M500) as provided by the manufacturer are smooth-surfaced beads activated with p-toluene-sulfonyl chloride to provide reactive groups for coavalent binding of antibodies containing primary amino or sulfhydryl groups. Affinity-purified goat anti-rabbit IgG (Fc) (Biodesign, Kennebunk, ME) was bound to the beads by incubation of 10 g/mg beads in 0.1 M borate buffer, pH 9.5, for 16 -24 h at room temperature. The amount of bound IgG (usually 3-6 g/mg beads) was assessed by measuring the A 280 before and after incubation. The beads were then incubated with the appropriate primary antibody (anti-Rab1a or anti-pIgAR) in 5% fetal calf serum (FCS) in phosphate buffered saline (PBS) for 16 h at 4°C and washed three times for 15 min each with 1% FCS in PBS (PBS/FCS) after each incubation.
Fraction CV1 or CV2 was incubated with primary antibody-coated magnetic beads at 50 -70 g of protein/5 mg of beads in PBS/FCS for either 3 h or overnight. Beads with the bound subfractions were collected with a magnet (38) and washed four times for 15 min each with PBS/FCS, 200 mM NaCl. Supernatants were centrifuged at 100,000 ϫ g for 1 h to pellet the membranes, which represent the non-bound subfractions. An equal aliquot (50 -70 g of protein) of starting material for immunoisolation was also centrifuged as above. All samples were resuspended in equal volumes of Laemmli sample buffer and boiled, and equal amounts were loaded onto gels. The proteins were separated by SDS-PAGE on 5-15% gradient gels, in order to determine the distribution of proteins in the bound versus the nonbound subfractions.
Immunoprecipitation-For immunoprecipitation of 35 S-pIgAR and 35 S-dIgA from liver fractions, 0.5 ml (for pIgAR) or 0.5 mg of protein (for dIgA) of each fraction was lysed in RIPA buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 0.1% SDS, 1% Triton X-100, 0.5% deoxycholate), and antibody for pIgAR (2 l) or dIgA (1 g/ml) was added. Samples were subjected to SDS-PAGE and analyzed on a PhosphorImager (Molecular Dynamics) using ImageQuant software. The amount of pIgAR and dIgA was normalized to the total volume and protein concentration, respectively.
For immunoprecipitation of Rab1a and Rab6 from immunoisolated vesicles, bound subfractions were lysed in Triton X-100 lysis buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 1% Triton X-100). Antibodies for Rab1a or Rab6 (1 g/ml) were added, and the lysates were rocked for 2 h at 4°C. Antibody-antigen complexes were collected on protein A-Sepharose. The supernatants depleted of Rab1a or Rab6 were then precipitated with 6% trichloroacetic acid at 4°C for 30 min. Samples were resuspended in equal volumes of SDS sample buffer, and proteins were separated by SDS-PAGE as above.
After separation the gel was removed and equilibrated in 62.5 mM Tris-HCl, pH 6.8, 2.3% SDS, 5% 2-mercaptoethanol, and 10% glycerol for 15-30 min. The gel strips were then cut and applied on the top of 10% SDS gels and the proteins separated by SDS-PAGE (second dimension) on a Bio-Rad minigel apparatus.
Immunoblotting-After electrophoresis on either two-dimensional or one-dimensional gels, the separated proteins were transferred to PVDF membranes (Millipore). The membranes were incubated with primary and then secondary antibodies in 1% FCS/PBS, 0.2% Triton X-100 for 1 h each, washed three times for 15 min each between each incubation, and detected by enhanced chemiluminescence (ECL, Amersham) according to manufacturer's instructions. Radiolabeled dIgA was detected and quantitated on a PhosphorImager using ImageQuant software.
[␣-32 P]GTP Overlay-[␣-32 P]GTP overlay was modified from the procedure detailed by Nickel et al. (40). Briefly, PVDF membranes were soaked in 50 mM phosphate buffer, pH 7.6, containing 10 mM MgCl 2 , 2 mM dithiothreitol, 0.3% Tween 20, 100 mM ATP twice for 15 min each, followed by incubation with [␣-32 P]GTP (DuPont NEN) in the same buffer at 1 Ci/ml for 2 h. The membranes were washed with the same buffer six times for 2 min each and air-dried. The bound [␣-32 P]GTP was detected by autoradiography (2 h to overnight exposure) on Kodak X-Omat film.
Quantitation of Rabs and Total Protein on Isolated Vesicles-For quantitation of the amount of Rab1a or Rab6 associated with immunoisolated vesicles, standard curves of recombinant Rab1a or Rab6 with 5-or 10-ng increments were made over a 10 -100-ng range. Recombinant proteins and immunoisolation samples were loaded onto the same gels, subjected to SDS-PAGE, and transferred to PVDF membranes. The PVDF membranes were blotted with anti-Rab1 or anti-Rab6 followed by 50 Ci of 125 I-protein A (DuPont NEN). The bound 125 I-protein A was quantitated on a PhosphorImager, and a standard curve was plotted for each protein. The amount of Rab1a or Rab6 associated with the isolated subfractions was then determined from the standard curve. The numbers shown represent the averages of four gels for Rab1a and two for Rab6.
For quantitation of the total protein bound to immunoadsorbents, starting material and non-bound subfractions were spun at 100,000 ϫ g to remove the serum proteins added (as blocking agents) during immunoisolation. The membrane pellets were resuspended in PBS, and the protein concentration was determined by BCA assay. The total protein bound to the beads was determined by subtracting non-bound from the starting material. The numbers were normalized to g of protein/mg of beads (mean Ϯ S.D., n ϭ 3).
Electron Microscopy-Vesicles immunoisolated on magnetic beads were fixed, stained and sectioned as described previously (36,41). In brief, beads were fixed for 45 min in 4% paraformaldehyde, 0.1% glutaraldehyde in 0.1 M cacodylate-HCl buffer, pH 7.2, followed by 1% OsO 4 in the same buffer for 1 h. Samples were stained in block with 2% uranyl acetate, dehydrated, and embedded in Epon. Thin sections were cut, stained with 2% uranyl acetate and lead citrate, and examined in a Philips CM-10 electron microscope.

Characterization of Rat Liver
Fractions-Subcellular fractions obtained from rat liver total microsomes by flotation in a discontinuous sucrose gradient have been characterized previously (36). To obtain more enriched ER to Golgi and transcytotic vesicle fractions, we modified the fractionation procedure and separated the crude vesicular fraction into two carrier vesicle fractions, CV1 and CV2.
Characterization of these fractions was carried out using pIgAR and dIgA as markers. dIgA is present only in transcytotic vesicles. pIgAR has three forms with different mobilities and locations (32-34) (see Fig. 1): a 105-kDa ER form (endo H-sensitive), a 116-kDa fully glycosylated Golgi form (endo H-resistant), and a 120-kDa phosphorylated form, which is a marker for the transcytotic pathway (34). To determine the forms of pIgAR present in CV1 and CV2, Tran 35 S-label was injected into the portal vein, and 15 or 30 min later the liver was removed and processed for fractionation. After 15 min, ϳ35% of the newly synthesized 105-kDa form of the pIgAR was found in the CV2 fraction and another 45% was associated with the residual microsomes, indicating that CV2 is enriched in ER to Golgi transport vesicles ( Fig. 1, top panel). CV1 contains little (Ͻ8%) of the 105-or 116-kDa forms, suggesting that this fraction contains relatively few ER to Golgi and intraGolgi transport vesicles. The 120-kDa mature form is not seen at 15 min after injection of the label (37). However, 30 min after injection Ͼ36% of the 120-kDa form of the receptor was found in CV1 (data not shown). Similarly, when 35 S-dIgA was injected into the portal vein to label transcytotic vesicles of hepatocytes, most (ϳ54%) of the dIgA was recovered in the CV1 fraction at 30 min after injection and only 11% in CV2 (Fig. 1,  middle panel). These results indicate that whereas CV2 is enriched in ER to Golgi transport vesicles, CV1 contains very few of these vesicles and is enriched in transcytotic vesicles.
Contamination of CV1 and CV2 with other organelles has also been examined using protein disulfide isomerase (ER), ␣-mannosidase II (Golgi), and B-AF (endosomes) as markers. CV1 contains a small amount of the total protein disulfide isomerase and ␣-mannosidase II (1.5 and 11%, respectively) in the total microsomes and ϳ30% of the total B-AF. By contrast, the percentages of the total of these three markers in CV2 are 40, 30, and 10%, respectively (data not shown). Therefore, CV1 and CV2 represent fractions highly enriched in transcytotic and ER to Golgi vesicular carriers, respectively, and because of this are appropriate as starting material for immunoisolation of the respective carriers.
Rab1a Is Tightly Membrane-associated in Rat Liver-Rab1a has been localized by immunogold labeling to the membranes of clusters of vesicles located between the ER and the Golgi region of NRK cells (14,42) and has been shown to be required for ER to Golgi and intraGolgi transport (14,15,35). Since Rab1a is known to be associated with ER to Golgi transport vesicles, we decided to use it to isolate this population of vesicles.
Rab1a is anchored to the membrane via geranylgeranyl moieties covalently attached at cysteine residues located at the carboxyl terminus (43)(44)(45). According to current models of Rab function (1,5), Rab proteins could be found either in the cytosol or associated with membranes. The distribution of Rab1a in liver fractions was determined by immunoblotting. As reported for cultured cells (43,46), in rat liver most of the Rab1a is associated with membranes at steady state (Fig. 1, bottom panel, compare Cyt and TM), and it has a broad distribution across the gradient. The association of Rab1a with membranes is resistant to treatment with either high pH (100 mM Na 2 CO 3 , pH 11) or high salt (1 M KCl) (data not shown). Therefore, it behaves as an integral membrane protein. The tight association of Rab1a with membranes permitted us to use it to immunoisolate Rab1a-associated vesicles using an affinity-purified, anti-Rab1a antibody.
Characterization of Rab1a-specific IgG-The antibody used for immunoisolation was affinity-purified IgG raised against a 19-amino acid peptide near the carboxyl terminus of Rab1a (amino acids 181-199; AGGAEKSNVKIQSTPVKQS). Sequence analysis of the 32 known Rabs showed that there is no homology of the peptide to any other Rab. Rab1a and Rab1b are the closest isoforms, which share 92% identity (47) and differ in only 14 amino acids, 9 of which are restricted to amino acids 180 -199. Therefore this antibody was expected to be specific for Rab1a. Western blotting on purified recombinant Rab proteins confirmed that it detected Rab1a but not Rab1b ( Fig. 2A), indicating this antibody can discriminate between the two proteins. In total microsomes the antibody reacted with a single protein, which co-migrated with recombinant Rab1a and could be detected by p68, a well characterized polyclonal antibody that recognizes both Rab1a and Rab1b (13).
To rule out the possibility that anti-Rab1a antibody may cross-react with other Rab proteins that have the same mobility as Rab1a, total microsomes were subjected to two-dimensional electrophoresis, followed by immunoblotting with anti-Rab1a. A single spot with the expected pI (pH 6.0) and mobility of Rab1a (48) was detected (Fig. 2B). The same spot can be detected by p68 (data not shown). Since the likelihood that two different proteins could have the same pI and mobility is very low, we concluded that the anti-Rab1a is specific for Rab1a.
Immunoisolation of Rab1a-associated Vesicles-The CV2 fraction (d ϭ 1.158) was chosen as the starting material for immunoisolation because it is enriched in ER to Golgi transport vesicles. Anti-Rab1a-coated magnetic beads were incubated FIG. 1. CV2 and CV1 are enriched in ER to Golgi transport vesicles and transcytotic vesicles, respectively. Tran 35 S-label or 35 S-dIgA were injected into the portal vein, and 15 or 30 min later the liver was removed, homogenized, and fractionated as described under "Experimental Procedures." pIgAR or dIgA were immunoprecipitated from each fraction, separated by SDS-PAGE, and analyzed on a Phos-phorImager. ϳ80% of the 105-kDa form of the pIgAR is found in CV2 and residual microsomes (RM), and Ͼ90% of the 116-kDa form is in Golgi light (GL) and Golgi heavy (GH) (top panel). CV1 contains little (Ͻ8% and 4%, respectively) of either form. Most of the dIgA is found in CV1 (54%), and less in Golgi heavy (18%) and CV2 (11%) (middle panel). Rab1a determined by immunoblotting has a broad distribution across the gradient (bottom panel).
with this fraction, collected with a magnet, and washed. Distribution of Rab1a in non-bound and bound subfractions after immunoisolation was determined. Immunoblotting with affinity-purified anti-Rab1a IgG showed that the immunoisolation was highly efficient; Ͼ95% of the Rab1a was detected in the bound subfraction (Fig. 3, lane 3), and little, if any, was left in the non-bound subfraction (Fig. 3, lane 2). In contrast, no Rab1a was detected on the beads coated with control IgG (Fig. 3, lane 6).
To identify the sources of the immunoisolated vesicles, we examined the form of pIgAR they contain. We found, as expected, the 105-kDa (ER) form of pIgAR in the bound subfraction after immunoisolation with anti-Rab1a (Fig. 3, top panel,  lane 3). However, to our surprise, a considerable amount of the mature, 120-kDa form was also present in the same subfraction. Since the mature form of pIgAR is associated only with transcytotic carriers (33), this finding suggested that the population of vesicles immunoisolated from the CV2 fraction with anti-Rab1a includes not only vesicles involved in ER to Golgi transport, but also those involved in transcytosis.
Rab1a Is Associated with Transcytotic Vesicular Carriers-To further investigate this possibility, we immunoisolated transcytotic vesicles from CV1 (d ϭ 1.146), the fraction enriched in transcytotic vesicular carriers, using an antibody against the cytoplasmic domain of pIgAR (37) and assayed for the presence of Rab1a. The 120-kDa form of pIgAR and 35 S-dIgA were used as markers for the transcytotic pathway (37). As expected, vesicles immunoisolated from CV1 with beads coated with anti-pIgAR contained the transcytotic content marker dIgA and the mature (120-kDa) form of pIgAR (Fig. 4, lanes 4 -6) (37). By immunoblotting Rab1a was detected only in the bound subfraction (Fig. 4, lane 6). To verify that this band is indeed Rab1a, high resolution two-dimensional PAGE followed by immunoblotting was performed. A single spot at the expected pI and mobility of Rab1a was detected in the bound fraction (data not shown). These results provide further evidence that Rab1a is associated with transcytotic vesicles.
We next immunoisolated vesicles from the CV1 fraction on anti-Rab1a and compared their composition with those isolated from the same fraction on anti-pIgAR. The results were similar; dIgA was found only in the bound fraction (Fig. 4, lane 3), and the majority of the pIgAR associated with the bound fraction was the mature, 120-kDa form. There was also a smaller amount of the 116-kDa form, but the 105-kDa form was not detected (Fig. 4, lane 3). Since transcytotic vesicles are known to contain primarily the 120-kDa and a small amount of the 116-kDa form of pIgAR (33), this observation suggests that vesicles immunoisolated from the CV1 fraction on either anti-Rab1a or anti-pIgAR are associated with the transcytotic pathway, rather than with ER to Golgi transport. This conclusion is supported by data obtained on newly synthesized albumin. The bound subfraction immunoisolated on anti-Rab1a contains Ͻ1% of the marker, whereas the corresponding bound fraction from CV2 accounts for ϳ36%.
For quantitation of the Rab1a associated with the isolated vesicles, recombinant Rab1a was titrated to make a standard curve, and the amount of Rab1a present on the vesicles was determined using 125 I-protein A. The amount of Rab1a associated with vesicles immunoisolated via the two antibodies from CV1 was comparable; under conditions in which Ͼ95% of Rab1a was recovered in the bound fraction (e.g. Fig. 4), 1.249 Ϯ 0.158 and 1.253 Ϯ 0.350 ng of Rab1a/g of starting material were associated with vesicles isolated on anti-Rab1a and anti-pIgAR, respectively.
To validate that Rab1a is truly a component of transcytotic vesicles, we compared the density of Rab1a present on these vesicles isolated from CV1, which is enriched in transcytotic vesicles, with those isolated from CV2, which is enriched in ER-to-Golgi transport vesicles. The amount of total protein and Rab1a bound to the immunoadsorbents after incubation with CV1 or CV2 was determined. As shown in Table I, vesicles isolated from CV2 contained 9 -12-fold more protein than vesicles isolated from CV1. This undoubtedly reflects the high levels of newly synthesized content proteins (such as albumin) in ER-to-Golgi transport vesicles. On the other hand, similar amounts of Rab1a were present in the two populations of vesicles (Table I). Since vesicles immunoisolated from CV1 and CV2 via anti-Rab1a are similar in size and the efficiency of immunoisolation is the same (see below and Fig. 6), the beads would be expected to bind a comparable amount of vesicles from the two fractions. We therefore estimated that the density of Rab1a on transcytotic vesicles is similar to that on ER to Golgi vesicles.
Vesicles Immunoisolated on Anti-Rab1a Are Depleted of ER, Golgi, Endosome, and Basolateral PM Markers-As a further check for the purity of the immunoisolated transcytotic vesicles, we examined the distribution of markers of specific cell compartments as follows: BiP and calreticulin (ER), ␣2,3-sialyltransferase (Golgi), B-AF (endosomes), and EGF receptor (basolateral PM) (Fig. 5). Neither ER nor Golgi markers were detected in the bound subfractions after immunoisolation on either anti-Rab1a (lane 3) or anti-pIgAR (lane 6), suggesting little, if any, ER or Golgi membranes in the isolated vesicles.
B-AF, a well studied ligand for the asialoglycoprotein receptor (ASGPR), is a marker for pre-endosomal vesicles and early endosomes. It is taken up by the ASGPR at the basolateral (sinusoidal) PM, transported to early endosomes where it dissociates from the receptor, the ligand is delivered to lysosomes, and the receptor recycles back to the basolateral cell surface (49 -51). B-AF was injected into the portal vein, the livers were removed 6 min later (to allow most of the injected B-AF to reach endosomes (51), and immunoisolation was carried out on the CV1 fraction. As shown in Fig. 5, the distribution of B-AF and EGF receptor in the subfractions immunoisolated on anti-Rab1a differ from their distribution after immunoisolation on anti-pIgAR. Both markers can be detected in the non-bound (lane 5) and bound (lane 6) subfractions after immunoisolation with anti-pIgAR. However, neither B-AF nor EGF receptor were found in the bound subfraction immunoisolated with anti-Rab1a (lane 3), suggesting that it does not contain endosomes or basolateral PM. ASGPR, another early endosomal marker, had the same distribution as B-AF in the two fractions (data not shown). The finding of different distributions of endosome and basolateral PM markers in the vesicles immunoisolated on anti-Rab1a and anti-pIgAR indicates that the two populations of vesicles have different content proteins.
Transcytosis starts at the basolateral PM where the pIgAR and its ligand are internalized, after which it is sorted in early endosomes and transported to the apical (bile canalicular) domain of the PM (30,52,53). Since we detected pIgAR and dIgA (transcytotic markers) but not B-AF or EGF receptor (endosomal and basolateral PM markers) in vesicles isolated on anti-Rab1a, our data suggest that these vesicles may represent post-endosomal transcytotic vesicles, whereas vesicles isolated on anti-pIgAR are derived from the entire transcytotic pathway from basolateral to apical PM, including pre-and post-endosomal vesicles, and the subapical compartment (54).
Morphology of Vesicles Immunoisolated with Anti-Rab1a or Anti-pIgAR-The morphology of the vesicles immunoisolated on anti-Rab1a and anti-pIgAR was examined by electron microscopy. Vesicles immunoisolated on anti-Rab1a from CV1 (Fig. 6D) and CV2 (Fig. 6, A and C) or those immunoisolated on anti-pIgAR from CV1 (Fig. 6E) were relatively homogeneous  6). B-AF was injected into the portal vein, and 6 min later the liver was removed and processed for immunoisolation as described under "Experimental Procedures". B-AF was detected using streptavidin-conjugated to horseradish peroxidase followed by ECL. SialylT, ␣2,3-sialyltransferase; EGFR, EGF receptor; B-AF, biotinylated asialofetuin.

TABLE I Comparison of the amount of Rab1a and total protein immunoisolated from the CV1 and CV2 fractions
Vesicles were isolated from either CV1 or CV2 via anti-Rab1a or anti-pIgAR. The amount of total protein bound to the beads was calculated by subtracting the nonbound from the starting material as described under "Experimental Procedures." Recombinant Rab1a over a range of 10 -100 ng was subjected to SDS-PAGE, transferred to PVDF membrane, and immunoblotted. Rab1a was detected by 125 I-protein A, and a standard curve was plotted. Samples of immunoisolations were processed in the same gels to determine the amount of Rab1a associated with isolated vesicles. The values are expressed as the mean Ϯ S.D. from three experiments. with a diameter of 60 -100 nm. There is no major difference in the size and morphology of vesicles immunoisolated on Rab1a from the two fractions. Control beads coated with normal IgG lacked any associated vesicular elements (Fig. 6B).
Multiple Small GTP-binding Proteins Are Found in Vesicles Immunoisolated via Anti-Rab1a or Anti-pIgAR-The experiments described above demonstrated that a reasonably homogeneous subpopulation of transcytotic vesicles can be immunoisolated from the CV1 fraction using anti-Rab1a. We next set out to determine how many small GTP-binding proteins are associated with this class of vesicles. After immunoisolation, samples were analyzed by [␣-32 P]GTP overlay. We found at least three major bands were selectively associated with the bound subfraction, whereas most of the more slowly migrating bands and small amounts of the other bands remained in the non-bound subfraction (data not shown). We also found multiple small GTP-binding proteins associated with the bound subfraction immunoisolated on anti-pIgAR, with the pattern being quite similar to that obtained after immunoisolation on anti-Rab1a. These data indicate that multiple specific small GTPbinding proteins are associated with vesicles immunoisolated via either anti-Rab1a or anti-pIgAR.
Identification of the Small GTP-binding Proteins Associated with Transcytotic Vesicles-To date, over 30 Rabs have been identified (3,6), but up to now no Rab protein has been linked to the transcytotic pathway. Most Rabs are ubiquitously expressed in all cell types, but some are cell type-specific. To determine which Rabs associate with immunoisolated transcytotic vesicles, we first examined two epithelia-specific Rabs: Rab17 and Rab18 (26,29). Most of the Rab17 was present in the bound fractions of vesicles immunoisolated on anti-Rab1a (Fig. 7, lane 3) or anti-pIgAR (Fig. 7, lane 6). In contrast, Rab18 was detected only in the non-bound subfractions (lanes 2 and 5). These data suggest that epithelia-specific Rab17 is associated with transcytotic vesicles, whereas Rab18 is associated with some other type(s) of vesicular carrier, which remains to be identified.
In addition to the epithelia-specific Rabs, we investigated the presence of ubiquitously expressed Rabs (Rabs1b, -2, and -6) and ARF in vesicles immunoisolated on anti-Rab1a. Although Rab1b has the same distribution as Rab1a in NRK cells (15,42), we found that Rab1b has a broader distribution than Rab1a in rat liver. Rab1b was found in comparable amounts in both the non-bound (lanes 2 and 5) and bound (lanes 3 and 6) subfractions by immunoblotting, whereas Rab1a was detected only in the bound subfractions (lanes 3 and 6). Rab2 and Rab6 were also distributed between the nonbound and bound frac-tions (lanes 3 and 6), suggesting that these Rabs are also associated with elements of the transcytotic pathway. In contrast, ARF was found only in the non-bound subfractions (lanes 2 and 5). The smaller amounts of Rab2 and Rab6 remaining in the non-bound subfractions (lanes 2 and 5) are presumably associated with other populations of vesicular carriers.
To rule out the possibility that the presence of multiple Rabs in the bound subfractions was due to nonspecific binding of trace quantities of Rabs to the matrix of the immunoadsorbents, we carried out additional immunoprecipitation experiments with Rab6 and quantitated the amount of Rab6 associated with the bound vesicles. Rab6 was used as a representative Rab because antibody suitable for immunoprecipitation was available. After immunoisolation on anti-Rab1a or anti-pIgAR, the isolated vesicles were lysed in Triton X-100 lysis buffer, and immunoprecipitation with anti-Rab6 was performed on the solubilized proteins. A single ϳ24-kDa band was precipitated (Fig. 8, lanes 5 and 10). The same band was depleted from the supernatants (Fig. 8, lanes 4 and 9). When the amount of Rab6 associated with isolated transcytotic vesicles FIG. 6. EM morphology of vesicles immunoisolated on anti-Rab1a and anti-pIgAR. Vesicles immunoisolated on magnetic beads were fixed, stained in block with uranyl acetate, and processed for routine electron microscopy as described under "Experimental Procedures." A, low magnification view showing that the beads coated with anti-Rab1a isolated from CV2 are uniformly covered with small (60 -100 nm) vesicles, which are relatively homogeneous in size and morphology. B, control sample in which nonspecific IgG was coupled to the beads shows minimal binding of vesicular structures to the beads. Vesicles isolated from either the CV1 fraction (D) or the CV2 fraction (C) on anti-Rab1a or from the CV1 fraction on anti-pIgAR (E) are similar in size and morphology. Note that tubular structures (arrows) are occasionally seen (E). Bar, 0.  1-3) or anti-pIgAR (lanes 4 -6) as described in Fig. 4, and the presence of ARF and various Rab proteins (indicated on the right) was determined by immunoblotting. Rab17 and most of the Rab2 and Rab6 are associated with vesicles immunoisolated on anti-Rab1a (lane 3) or anti-pIgAR (lane 6). Rab1b is evenly distributed between the bound (B, lanes 3 and 6) and non-bound (NB, lanes 2 and 5) subfractions, whereas Rab18 and ARF are found primarily in the non-bound subfractions. SM, starting material. was quantitated by immunoblotting using 125 I-protein A for detection, ϳ1.292 Ϯ 0.44 and 1.111 Ϯ 0.06 ng of Rab6/g of starting material was detected in vesicles isolated on anti-Rab1a and anti-pIgAR, respectively. These results indicate that Rab6 is associated with immunoisolated transcytotic vesicles at a level comparable to that of Rab1a (see "Rab1a Is Associated with Ttanscytotic Vesicular Carriers").
Based on these experiments, we conclude that 1) multiple Rab proteins, including specialized and ubiquitous Rabs, apparently associate with transcytotic carrier vesicles; and 2) Rab1a and Rab6 are present in transcytotic vesicles with comparable stoichiometry. DISCUSSION In the present work we have used subcellular fractionation of rat liver to obtain fractions enriched in either ER to Golgi transport vesicles (CV2) or transcytotic vesicles (CV1) and, in the case of CV1, de-enriched in other cell components. A specific subset of transport vesicles was further purified from these fractions by immunoisolation on anti-Rab1a antibody. In characterizing the immunoisolated vesicles, we found that: 1) Rab1a is not only associated with ER to Golgi transport vesicles, but also is present on transcytotic vesicles; 2) multiple Rabs, including epithelia-specific Rab17 and ubiquitous Rabs1b, -2, and -6, are found on the membranes of transcytotic vesicles; and 3) ARF is not found in association with transcytotic vesicles.
Rab1a Is Associated with Transcytotic Vesicles as Well as ER to Golgi Transport Vesicles-The initial goal of this study was to isolate ER to Golgi transport vesicles using Rab1a as the target. For this purpose we utilized an antibody that is Rab1aspecific and does not cross-react with any other Rabs, including the closely-related protein, Rab1b. The specificity of anti-Rab1a for Rab1a was established experimentally by immunoblotting on recombinant Rab1a and -1b proteins as well as on total microsomes after separating the proteins by high resolution two-dimensional PAGE.
To obtain a highly purified preparation of ER-to-Golgi transport vesicles, we prepared a fraction, CV2, from rat liver enriched in ER to Golgi carrier vesicles and subjected this fraction to immunoisolation on anti-Rab1a-coated magnetic beads. Using pIgAR as a marker, to our surprise, we detected in the bound subfraction not only the 105-kDa/ER form of pIgAR, but also the 120-kDa form of pIgAR, which is found primarily in transcytotic vesicles. This suggested that, in addition to being associated with ER to Golgi transport vesicles, Rab1a is also present on vesicular elements associated with the transcytotic pathway in rat liver.
To further investigate this possibility, we carried out immunoisolation on a fraction, CV1, known to be enriched in transcytotic vesicles. Once again we found that vesicles immunoisolated on anti-Rab1a contained the two best transcytotic markers: dIgA and its receptor, the 120-kDa form of pIgAR. Moreover, Rab1a was detected by immunoblotting in transcytotic vesicles immunoisolated on anti-pIgAR-coated magnetic beads, and its presence was confirmed by high resolution twodimensional PAGE. This method has been used to analyze the complexity of the large superfamily of small GTP-binding proteins. At least 28 small GTP-binding proteins (20)(21)(22)(23)(24)(25), including Rab1a, have been mapped by their relative pI and mobility (55). Our experiments suggest that the density of Rab1a in transcytotic vesicles is comparable to that on ER to Golgi vesicles, indicating that Rab1a is truly associated with transcytotic vesicles.
The association of Rab1a with transcytotic vesicles was unexpected, because Rab1a has been well characterized in cultured NRK cells, where it was localized to pre-Golgi and Golgi elements by immunogold labeling and shown to be required for ER to Golgi and intraGolgi transport (14,35,42). The finding that Rab1a is associated with transcytotic vesicles in hepatocytes, which are highly polarized epithelial cells, suggests that the distribution of Rab proteins may vary in different cell types.
Vesicles Isolated on Anti-Rab1a Are Post-endosomal Transcytotic Vesicles-Sztul et al. (37) have previously immunoisolated transcytotic vesicles from a carrier vesicle-enriched fraction using an antibody against the cytoplasmic domain of pIgAR. We have modified the fractionation procedure to obtain a more enriched fraction of transcytotic vesicles, CV1, which contains ϳ54% of the transcytotic marker (dIgA) but Ͻ8% of the ER and Golgi markers (105-kDa and 116-kDa pIgAR, respectively). Immunoisolation carried out with either anti-Rab1a-or anti-pIgAR-coated magnetic beads yielded a reasonably homogeneous population of transcytotic vesicles based on the presence of transcytotic markers (dIgA and the 120-kDa pIgAR) and the absence of ER (calreticulin, 105-kDa pIgAR) and Golgi (sialyltransferase) markers. Further analysis of the two populations of immunoisolated vesicles suggests that those isolated on anti-pIgAR are involved in both pre-endosomal and post-endosomal steps in transcytosis, whereas those isolated on Rab1a consist mainly of post-endosomal vesicles involved in transport from endosomes to the apical (bile canalicular) PM.
Transcytosis starts at the basolateral (sinusoidal) PM, where the pIgAR and its ligand are internalized into cells via coated vesicles and transported to early endosomes where they are sorted from other membrane and content proteins. Transcytotic vesicles then bud off from endosomes and transport pIgAR with bound pIgA to the apical (bile canalicular) domain of the PM (30,52,53). Although the early stages of transcytosis from internalization at the basolateral PM to sorting at the early endosome have been well characterized in hepatocytes, the late steps of the pathway in the apical region are less clear. In Madin-Darby canine kidney cells, it has been shown that dIgA is transcytosed and delivered to an apical endosome before reaching the apical PM (56,57). In hepatocytes it has been shown that pIgAR is transported to a subapical compartment, where it colocalizes with several apical membrane proteins (58). By immunodepletion of endosomes (using ASGPR as the target) followed by immunoadsorbtion for transcytotic vesicles (using pIgAR), Barr et al. (54) have recently isolated a population of vesicles containing transcytosed dipeptidyl peptidase IV and pIgAR. Kinetic data and computer-generated models suggest that the isolated fraction constitutes the subapical compartment. We have also detected dipeptidyl peptidase IV in FIG. 8. Immunoprecipitation of Rab6 from vesicles isolated on anti-Rab1a or anti-pIgAR. Immunoisolation was carried out on anti-Rab1a (lanes 1-5) or anti-pIgAR (lanes 6 -10) as described in Fig. 4. After immunoisolation, the bound membranes were solubilized in Triton X-100 lysis buffer, divided into equal portions, followed by immunoprecipitation (IP) with either anti-Rab6 (lanes 4, 5, 9, and 10) or anti- Rab1a (lanes 2, 3, 7, and 8)  vesicles isolated on both anti-Rab1a and anti-pIgAR. We do not know whether our immunoisolated vesicles are postendosomal transport vesicles or represent vesicles derived from the subapical compartment or a mixture of the two.
Contamination or Multiplicity of Rabs per Relay?-In principle the presence of a Rab protein (Rab1a) in different vesicular carriers could reflect intercontamination. This can be ruled out in the case of Rab1a, since its ratio to exocytic and transcytotic markers is clearly different. To test the possibility that there are more than one Rab per relay of transport carriers, we examined the association of small GTP-binding proteins with the isolated vesicles by [ 32 P]GTP overlay and by immunoblotting with anti-Rab1a and antibodies to other small GTPbinding proteins. We found that multiple small GTP-binding proteins were present on transcytotic vesicles, including Rab-1b, -2, -6, and -17 as well as Rab1a. Several pieces of evidence indicate that the presence of multiple Rabs on the transcytotic vesicles was not due to nonspecific binding; (i) none of these Rabs were detected on magnetic beads coated with nonspecific IgG (see Fig. 3), (ii) Rab6, a representative of these Rab proteins, can be immunoprecipitated from isolated vesicles (see Fig. 8), and (iii) the amount of Rab6 associated with the transcytotic vesicles is comparable with that of Rab1a.
Little is known about the molecular mechanisms by which Rabs regulate membrane traffic. Over 30 Rabs have been identified in mammalian cells. As there are less than 30 relays of transport carriers, multiple Rab proteins may be associated with the same carrier vesicles and function together in vesicular transport.
Supporting this idea, multiple Rabs have also been found in other transport pathways. For example, at least seven small GTP-binding proteins were found to be associated with zymogen granule membranes (59 -61), and Rab3b, -18, -20, and -22, as well as Rab4 and Rab5 are associated with early endosomes (11,16,29,62). The function of most of these Rabs is unknown.
Rab Proteins Are Associated with More than One Population of Vesicular Carriers-We have identified at least four Rabs (Rab1b, -2, -6, and -17) on vesicles isolated from CV1 on anti-Rab1a. Rab17 is expressed only in epithelial cells in the kidney, liver, and intestine and has been localized to the basolateral PM and apical endosome region by electron microscopy (26). Our finding that Rab17 associates with transcytotic vesicles suggests that it may be involved in transcytosis. In contrast, Rab18, another Rab protein highly expressed in epithelia, was found only in the non-bound subfractions.
Additional Rabs found on the isolated transcytotic vesicles, i.e. Rab1b, -2, and -6, are ubiquitously expressed in all cell types and have been localized to pre-Golgi and Golgi membranes, intermediate compartment, and medial and trans Golgi and trans Golgi network, respectively (14,24). Our data suggest that similar to Rab1a, Rab1b, -2, and -6 are associated with more than one population of transport vesicles operating in both the exocytic and transcytotic pathways.
It is not clear why these ubiquitous Rab proteins are associated with the transcytotic pathway. It is possible that each transport step requires one "major" Rab (specific Rab) and a set of "helper" Rabs (common Rabs). Each of these ubiquitous Rabs may serve as a "major" Rab in a particular step of the exocytic pathway and as a "helper" Rab in the transcytotic pathway, and interactions among multiple Rabs may control the specificity and directionality of vesicular transport. Another possible explanation is that a given Rab controls (as a switch or kinetic proof reader) the entry of a factor into the fusion/targeting complex. Since these complexes have common components in addition to specific components, the same Rab (e.g. Rab1a) could be found at multiple relays.
In support of this hypothesis, some Rab proteins have been found in two transport pathways. For example, in addition to its function in the endocytic pathway, Rab4 is associated with glucose transporter (Glut4)-containing vesicles in adipocytes, and its distribution is modified by insulin. Upon stimulation with insulin, Rab4 is released from the membrane of Glut4containing vesicles into the cytosol and Glut4 is transported to the cell surface, suggesting that Rab4 may be involved in insulin stimulated translocation of Glut4-containing vesicles (63).
Rab5 has been found to be associated with both apical and basolateral endosomes in polarized Madin-Darby canine kidney cells (64), to be involved in endocytosis at both the apical and basolateral PM, and to be required for homotypic fusion of apical or basolateral endosomes in vitro (64,65). Yet fusion between the two sets of endosomes or mixing of the respective content does not occur (66 -68). This indicates that Rab5 alone would not be sufficient to target vesicles to either apical or basolateral compartments. Targeting may be controlled through interaction between Rab5 and other regulatory proteins, which could be other Rabs located in the same vesicular carriers.
Our data have implicated a network of Rab proteins in the transcytotic pathway in rat liver. Important questions to be addressed in future studies are what role they play in the transcytotic pathway and how they function together. It may be possible to address these questions experimentally by using mutant Rab proteins and determining their effect on transcytotic transport in polarized cultured epithelial cells.