Immunoadsorption of hepatic vesicles carrying newly synthesized dipeptidyl peptidase IV and polymeric IgA receptor.

Hepatocytes must transport newly synthesized apical membrane proteins from the basolateral to the apical plasma membrane. Our earlier morphological study showed that the apical proteins share a late (subapical) part of the transcytotic pathway with the well characterized polymeric immunoglobulin A receptor (Barr, V. A., and Hubbard, A. L. (1993) Gastroenterology 105, 554-571). Starting with crude microsomes from the livers of [35S]methionine-labeled rats, we sequentially immunoadsorbed first vesicles containing the endocytic asialoglycoprotein receptor and then (from the depleted supernatant) vesicles containing the polymeric IgA receptor. Biochemical characterization indicated that early basolateral and late endosomes were present in the first population but not in the second. Neither Golgi-, apical plasma membrane (PM)-, nor basolateral PM-derived vesicles were significant contaminants of either population. Both vesicle populations contained 35S-labeled receptor and 35S-labeled-dipeptidyl peptidase IV. Importantly, the elevated relative specific activity of the dipeptidyl peptidase (% of 35S-labeled/% immunoblotted) in the second population indicated that these vesicles must transport newly synthesized dipeptidyl peptidase IV. A distinct kind of vesicle was immunoadsorbed from a "carrier-vesicle fraction"; surprisingly, these vesicles contained little 35S-receptor and virtually no dipeptidyl peptidase IV. These results, together with previous kinetic data from in vivo experiments, are consistent with a computer-generated model predicting that newly synthesized dipeptidyl peptidase IV is delivered to basolateral endosomes, which also contain newly synthesized polymeric immunoglobulin A receptor. The two proteins are then transcytosed together to the subapical region.


S-labeled receptor and
S-labeled-dipeptidyl peptidase IV. Importantly, the elevated relative specific activity of the dipeptidyl peptidase (% of 35 S-labeled/% immunoblotted) in the second population indicated that these vesicles must transport newly synthesized dipeptidyl peptidase IV. A distinct kind of vesicle was immunoadsorbed from a "carrier-vesicle fraction"; surprisingly, these vesicles contained little 35 S-receptor and virtually no dipeptidyl peptidase IV. These results, together with previous kinetic data from in vivo experiments, are consistent with a computer-generated model predicting that newly synthesized dipeptidyl peptidase IV is delivered to basolateral endosomes, which also contain newly synthesized polymeric immunoglobulin A receptor. The two proteins are then transcytosed together to the subapical region.
The plasma membrane (PM) 1 of polarized epithelial cells is separated into distinct domains that have different functions and compositions (2,3). Newly synthesized integral PM proteins must be delivered to the correct domain to maintain such functional polarity. The route used to deliver newly synthe-sized apical PM proteins is particularly interesting because it varies in different epithelia (3)(4)(5). In hepatocytes, all newly synthesized PM proteins studied so far are transported from the Golgi first to the basolateral PM domain. Apical PM proteins must then be internalized, sorted, and transcytosed to the apical domain (6 -8). Caco-2 cells use this indirect route for a few proteins (9,10), while Madin-Darby canine kidney (MDCK) cells deliver their PM proteins directly from the Golgi to the correct domain (11). However, even MDCK cells can transcytose apical PM proteins if they are missorted to the basolateral surface (12,13). We want to define this transcytotic pathway; because hepatocytes rely so heavily on transcytosis, we have focused on this cell type.
Fortunately, the transcytosis of one PM protein has been studied extensively in hepatocytes (14,15); this protein is the polymeric IgA receptor (pIgA-R), which transports IgA from blood to bile. Fig. 1 shows a current view of the intracellular itinerary of newly synthesized pIgA-R. Mature pIgA-R is delivered to the basolateral PM from the Golgi. Both free and ligand-bound receptor are internalized in clathrin-coated vesicles (16 -18). Then ligand and receptor are taken to early endosomes; in uptake experiments ligand appears there very quickly (16,19,20). Later the ligand (presumably still traveling with the receptor) can be visualized in a subapical tubulovesicular compartment (16,17,21,22). Following a brief delay, pIgA is released into bile, complexed to the ectodomain of the pIgA-R (now called secretory component). Because pIgA-R is cleaved at the apical PM, most of the receptor in the cell is less than 2 h old (23,24). pIgA-R can be seen in both early endosomes and in the subapical compartment by immunofluorescence (25)(26)(27), showing that at steady state there are significant amounts of receptor in both compartments, but not at the apical PM (17,23). Work in MDCK cells transfected with pIgA-R cDNA has identified multiple signals on the 103-amino acid cytoplasmic tail of pIgA-R that guide the receptor along its complicated journey (27)(28)(29).
In contrast to the pIgA-R, many apical PM proteins either have very short cytoplasmic tails with no obvious sorting signals (e.g. dipeptidyl peptidase IV (DPP IV) and aminopeptidase N; Refs. 30 -35) or have no cytoplasmic sequences whatsoever (e.g. 5Ј nucleotidase, a glycophosphatidylinositol (GPI)-anchored protein) (36,37). Yet all of these proteins are transcytosed in hepatocytes (6,7). Do they travel with pIgA-R? In an earlier study, we found that bile duct ligation (BDL) slowed the transport of vesicles to the apical PM, and led to accumulation of newly synthesized pIgA-R and apical PM proteins in a common subapical tubulovesicular compartment (1). This result supports the idea that apical PM proteins share at least some of the pIgA-R pathway.
Our goal in this study was to determine if newly synthesized pIgA-R and apical PM proteins are found in the same transcytotic vesicles under normal conditions (38). In this study, we have immunoadsorbed vesicles from rat liver with a monoclonal antibody that recognizes the tail of pIgA-R. Advantages to the use of pIgA-R as "bait" are its abundance in hepatocytes and the availability of well characterized anti-tail antibodies (24,39). One disadvantage is that the pIgA-R has a very short half-life (t1 ⁄2 ϭ 2 h), so it is present in all of the biosynthetic compartments leading to its ultimate destination, the bile (24). Consequently, immunoadsorptions using anti-pIgA-R antibodies will bind membranes derived from several different compartments. Pulse labeling protocols help alleviate this problem by allowing us to follow a cohort of newly synthesized molecules. Even so, as time passes the cohort becomes unsynchronized, so that pulse labeled pIgA-R is found in multiple intracellular compartments (24), making it difficult to determine the origin of the membranes being immunoadsorbed from a homogenate. In this study, we describe a protocol that differentiates transcytotic vesicles from basolateral endosomes and Golgiderived vesicles, two other compartments that contain pIgA-R. Using this approach, we isolated transcytotic vesicles that contain 35 S-labeled DPP IV and 35 S-labeled pIgA-R. We also found 35 S-labeled DPP IV in vesicles that contain the asialoglycoprotein receptor (ASGP-R), a marker of basolateral endosomes. Based on these results, we believe that newly synthesized DPP IV is delivered to basolateral endosomes, which also contain newly synthesized pIgA-R; the two proteins are then transcytosed together. Preliminary results have been presented elsewhere (40).

MATERIALS AND METHODS
Chemicals-All chemicals were purchased from Sigma and were of reagent grade unless otherwise noted. Tran 35 S-label and ultrapure sucrose were obtained from ICN Radiochemicals (Irvine, CA). Trasylol was from FBA Pharmaceuticals (West Haven, CT). Octyl-␤-D-glucopyranoside was acquired from Boehringer Mannheim.
Animals-Young adult male Sprague-Dawley rats (200 -250 g; CD strain; Charles River Breeding Laboratories, Wilmington, MA) were housed with free access to laboratory chow and water. All animals were fasted during the dark cycle before sacrifice. Bile duct ligation was performed as described previously (1).
The protein synthesis inhibitor, cycloheximide (CHX) was given as an intraperitoneal injection of 1.0 mg/100 g of body weight (from 3 mg/ml solution in 10% EtOH-saline).
Rats were given 20 mCi of Tran 35 S-label by saphenous vein injection. No chase was used. After various times, the animals were sacrificed by decapitation and the livers were excised and perfused with ice-cold 0.9% saline in preparation for subcellular fractionation. 125 I-Asialoorosomucoid (ASOR) was prepared by neuraminidase treatment of orosomucoid, iodinated using chloramine T as described (41,42) and was administered by saphenous vein injection. The animals were sacrificed 2 min later, and the livers were excised and perfused (43). Approximately 5 min elapsed between the ASOR injection and the homogenization of the livers.
Preparation of Fractions-Livers were homogenized in 4.3 volumes of 0.25 M sucrose, 3 mM imidazole, pH 7.4 (0.25 M Suc/Im) containing protease inhibitors (100 units/ml Trasylol, 1 mM phenylmethylsulfonyl fluoride, and 5 g/ml each of antipain, leupeptin, and benzamidine) by 7 strokes in a Potter homogenizer at 4200 rpm. Microsomal fractions and carrier vesicle fractions (CVF) were prepared as described by Sztul et al. (38) with the following modifications. The filtered homogenate was centrifuged at 3500 ϫ g for 10 min (60Ti rotor, Beckman L7-55) and the supernatant was centrifuged at 180,000 ϫ g for 60 min (60Ti rotor, Beckman L7-55). The resulting pellet was resuspended in 20 ml of 0.25 M Suc/Im for use in immunoadsorptions and/or analysis. To make CVF, the microsomal pellet was resuspended in 14 ml of 1.22 M Suc/Im and 7 ml was successively overlaid with 8.5 ml of 1.15 M Suc/Im, 0.86 M Suc/Im, and 0.25 M Suc/Im. These gradients were centrifuged at 82,500 ϫ g for 3 h (SW28 rotor, Beckman L7-55), and the 1.15 M Suc/Im fraction below the white band of protein at the 1.15 M Suc/Im/0.86 M Suc/Im interface was collected.
Protein concentrations were determined with the BCA assay (Pierce).
Preparation of Immunoadsorbents-Immunoadsorbents were prepared by a modification of the method given by Sztul et al. (38). 1 ml of 50% Protein A-Sepharose bead slurry was incubated with 2 mg/ml goat anti-mouse IgG (Jackson ImmunoResearch Laboratories, West Grove, PA) overnight at 4°C and was then washed twice with 137 mM NaCl, 8.5 mM Na 2 HPO 4 , 2.6 mM KCl, 1.5 mM KH 2 PO 4 (PBS). This will be called GAM immunoadsorbent. To produce SC166 immunoadsorbent, 1 ml of GAM immunoadsorbent was incubated with 4 mg of SC166 IgG overnight (39) and then washed again in PBS. Both immunoadsorbents were cross-linked with dimethylpimelimidate as described (44) and stored as 50% slurries in PBS/NaN 3 .
Immunoadsorption of 35 S-Labeled Fractions-After an overnight fast, rats were given 20 mCi of Tran 35 S-label and sacrificed at 25 or 110 min. Then either microsomes or CVF were prepared. Similar fractions were obtained from rats whose bile ducts had been ligated 24 h before injecting 35 S-label; these rats were sacrificed at 110 min. To prevent Golgi-derived vesicles from binding to the SC166 immunoadsorbent, two rats were given 1.0 mg of CHX/100 g of body weight 15 min after injection of radioactivity and fractions were prepared 95 min later. The combined labeling and CHX chase time was 110 min.
Immunoadsorptions were performed at 4°C as shown in Fig. 2. For most experiments, 5.6 mg of microsomes or 1.4 mg of CVF diluted into 14 mls of 0.25 M sucrose, 150 mM NaCl, PBS, pH 7.4, 1% bovine serum albumin (Suc/PBS/BSA), were immunoadsorbed in 15-ml conical tubes. Generally, immunoadsorptions from microsomes used 4 times more protein than immunoadsorptions from CVF. Variations are noted in the figure legends. The fractions were first precleared by addition of fixed Staphylococcus aureus (from ATCC) (43), that had been washed three times in Suc/PBS/BSA (700 l of 22.5% SA/14 ml of microsomes or CVF). This mixture was incubated on a platform shaker at 200 oscillations/min for 1.25 h. The samples were spun at 3000 ϫ g for 15 min (3750 rpm, Beckman GS-6R centrifuge, GH-3.8 rotor). The pellet was rinsed three times by resuspension-centrifugation, first in 7 ml of Suc/ PBS/BSA, then in 7 ml of Suc/PBS (twice). All the rinses were discarded. The final pellet was designated "preclear." The precleared supernatant (from the initial centrifugation) was transferred to a new tube and incubated as described above, but with affinity-purified anti ASGP-R antibody (84 g/14 ml). Fixed S. aureus was added, and the incubation was continued for a second hour. The sample was centrifuged and rinsed as above. This washed S. aureus pellet was designated the "first ASGP-R immunoadsorption" (A1). The A1 supernatant was transferred to a fresh tube containing affinity-purified anti ASGP-R antibody, and the process was repeated. The resulting S. aureus pellet was designated the "second ASGP-R immunoadsorption" (A2). The A2 supernatant was split into two 7-ml aliquots, and 490 l of either SC166 or GAM immunoadsorbent (50% slurry) was added to an aliquot. The mixtures were incubated overnight (12-16 h) with end over end rotation, centrifuged at 300 ϫ g for 3 min and were rinsed as described above. The SC166 and GAM samples were centrifuged at the lowest After synthesis in the ER and processing in the Golgi, mature pIgA-R is transported from the trans-Golgi network to the basolateral PM. Receptors are internalized via clathrin-coated vesicles and delivered to early basolateral endosomes. Ligand (pIgA presumably bound to receptor) is next found in a subapical tubulovesicular compartment. Because this compartment is not continuous with basolateral endosomes, transport is probably mediated by vesicular transcytotic carriers. After delivery to the apical PM, pIgA-R is cleaved quickly and the ectodomain bound to pIgA is released into bile. However, there is a delay between the arrival of pIgA in the subapical compartment and release into bile, indicating that there may also be a vesicle mediated delivery step between the subapical compartment and the apical PM. These putative vesicular transport steps are indicated by question marks.
possible g force to avoid nonspecific pelleting of membranes. The S. aureus immunoadsorbents required stronger centrifugation because of the difference in size. (S. aureus has an average diameter of 1 m, while Sepharose beads range from 45 to 160 m in diameter.) The washed pellet from each sample: the preclear, A1, A2, SC166 immunoadsorbent, and GAM immunoadsorbent, was divided into portions while resuspended in the last rinse. One portion was then analyzed for DPP IV content, while another portion of the same sample was analyzed for pIgA-R. After centrifugation, the final pellets were resuspended in 1 ml of 20 mM octyl-␤-D-glucopyranoside, 0.5% (w/v) Triton X-100, 300 mM NaCl, 25 mM NaPO 4 , pH 7.4, and 0.02% NaN 3 (solubilization buffer) plus protease inhibitors. After incubating on ice for 30 min, the solubilized material was saved for immunoprecipitation or analysis. Because the solubilization buffer did not release the pIgA-R attached to SC166 immunoadsorbent, the immunoadsorbent beads were added to the immunoprecipitation mixture to insure recovery of all the pIgA-R.
Immunoprecipitations-Immunoprecipitations were performed as described previously (7,45), except for pIgA-R. All steps were performed at 4°C. The solubilized samples were added to 70 l of anti-DPP IV IgG-Sepharose or anti-HA321 IgG-Sepharose and incubated overnight. pIgA-R was immunoprecipitated by incubating the solubilized sample and the SC166 or GAM immunoadsorbent with 10 l of anti-secretory component polyclonal rabbit serum for 4 h, followed by incubating 6 -15 h with 70 l of Protein A-Sepharose beads. At the same time pIgA-R, DPP IV, and HA321 were immunoprecipitated from various amounts of homogenate (2-900 g), microsomes (60 -240 g), or CVF (25-240 g). Finally all the immunoprecipitated samples were washed and made into reduced gel samples (7,45), except for HA321, which was not reduced. To quantify the amount of 35 S-labeled protein, 80% of each sample was separated on a 7.5% SDS-PAGE gel, the gel was treated with 2,5-diphenyloxazole dried, and exposed to x-ray film for 5-6 weeks. The remaining 20% of the immunoadsorbed SC166 and GAM samples were brought to 100 l in gel buffer and then run on another 7.5% gel. Smaller fractions of the remaining preclear, A1, and A2 samples were brought to 100 l in gel buffer and applied to gels. After electrophoresis, polypeptides were transferred to nitrocellulose by standard methods (46), incubated with antibodies and the resulting bands visualized using the ECL detection system (Amersham Corp.). Band intensities were quantified on a Microcomputer Imaging Device (MCID) (Imaging Research Inc., St. Catharines, Ontario, Canada). The relative amount of protein in a fraction (% of homogenate) was determined by comparing the intensity of the band of interest to a standard curve of homogenate consisting of 3 or 4 points.
Electron Microscopy-Immunoadsorbed samples and membrane fractions were fixed for 1 h in 1.5% glutaraldehyde in 0.1 M sodium cacodylate (pH 7.4) at 4°C. SC166 and GAM samples were then embedded in agarose (47), while the other samples were pelleted by centrifugation at 10,000 ϫ g for 10 min (SW 50.1 rotor, Beckman L7-55). The samples were postfixed in OsO 4 and uranyl acetate, dehydrated, and embedded in epon according to standard procedures (43,48). Membrane contrast was enhanced by staining with lead citrate for 5 min (49).
Enzyme Assays-␤-Glucuronidase activity was assayed by monitoring the release of free phenolphthalein from phenolphthalein mono-␤glucuronic acid, using a diagnostic kit (Sigma). Cathepsin D activity was determined from the release of trichloroacetic acid-soluble material from hemoglobin at pH 3, following the quality control test procedure from Sigma.
Computer Modeling-Computer models were constructed with the Stella II program (High Performance Systems, Inc. Hanover, NH). Models consisted of reservoir "stocks" connected by first order "flows" that transferred material from one stock to the next. The ER compartment also contained a conveyor function to model a required residence time. A pulse of protein synthesis was modeled as a graphical time input. This input was 0 at 0 min, rose linearly to 1 at 3 min, remained at 1 until 10 min, and then decreased linearly to 0 at 15 min. Data from previous studies (6, 7) were used to determine the kinetic parameters for movement through the biosynthetic pathway. We used the numbers of molecules of DPP IV determined in a previous study to calculate the number of molecules in each compartment. This study showed that there are 5 ϫ 10 6 molecules of DPP IV/hepatocyte and that 1.2 ϫ 10 4 molecules would be synthesized in a 15-min pulse (24).

Characterization of the Immunoadsorbed Samples-We used
in vivo metabolic labeling combined with vesicle immunoadsorption to test the hypothesis that apical PM proteins are transcytosed with pIgA-R. Our starting material, the microsomal fraction, contained Golgi-derived vesicles (indicated by the presence of sialyl transferase (ST)), basolateral endosomes (indicated by the presence of ASGP-R, transferrin receptor (Tf-R), and 125 I-ASOR internalized for 5 min) and transcytotic vesicles (indicated by the presence of 125 I-pIgA internalized for 60 min). Because pIgA-R, our target for immunoadsorption, is most likely present in all of these compartments (50, 51), we designed a protocol that specifically isolated transcytotic vesicles (Fig. 2). The preclear step removed vesicles that stuck to S. aureus nonspecifically. The next two immunoadsorption steps removed membranes that contained ASGP-R. ASGP-R is abundant in early endosomes, but it is also at the basolateral PM (52) and in late endosomes (41). Thus, vesicles derived from these compartments would be found in the A1 and A2 samples. Any remaining vesicles that contained pIgA-R but lacked ASGP-R would then be bound by the SC166 immunoadsorbent. The GAM immunoadsorbent served as a control for nonspecific binding. Comparison of the amounts of a particular protein in each of the immunoadsorbed samples gave us a quantitative picture of its distribution in the biosynthetic pathway. S. aureus was used for the first immunoadsorptions because we knew from previous work that we could remove all the endosomes in this manner. However, we did not use S. aureus for selecting pIgA-R vesicles, because it shows higher nonspecific binding than does Sepharose.
Recovery of pIgA-R- Fig. 3A shows an immunoblot of the pIgA-R that was found in samples which were analyzed after the initial sedimentation (i.e. no washing). pIgA-R was present in each immunoadsorbed sample, but the amount varied dramatically (Fig. 3B). The preclear and A1 samples contained Ͼ65% of the pIgA-R. Of the total microsomal pIgA-R, the SC166 sample consistently contained Ͼ10%, which was 6-fold more than in the GAM control. ϳ7% of the pIgA-R remained unbound. Overall, we could account for 80 -100% of the pIgA-R through the four steps of the immunoadsorption protocol. However, only 30% of the microsomal pIgA-R was recovered in comparable samples that had been washed prior to analysis (A1 ϩ A2 ϩ SC166 ϳ 23%; unbound ϳ 7%). The difference between washed and unwashed samples was predominantly in the A1 and A2 samples, suggesting to us that immunoadsorbed vesicles may have broken during the S. au- The protocol for immunoadsorption of vesicles. Vesicular fractions were sequentially incubated once with fixed S. aureus, twice with anti ASGP-R antibodies and fixed S. aureus, and then with SC166 or GAM immunoadsorbent. After each incubation, the samples were centrifuged and the pellets were washed as described under "Materials and Methods." The preclear contains vesicles that adhere to S. aureus nonspecifically, A1 and A2 contain vesicles with ASGP-R, and SC166 contains any vesicles with pIgA-R but without ASGP-R. GAM serves as a control for vesicles that adhere to the linker antibody or Sepharose beads.
reus rinsing procedure, with consequent pinching off and loss of unattached membrane. We used washed samples in most experiments despite this loss, because we wanted to avoid contamination by nonspecific sticking of membranes, particularly contamination with PM.
The Material Immunoadsorbed by SC166 Is Not from Endo-somes-To be sure that early endosomes were substantially depleted from the SC166 sample, we examined microsomes from animals that had been given 125 I-ASOR 5 min before sacrifice. At this time, 94% of the labeled ASOR had been taken up by the liver (data not shown). According to previous work, the ligand is still bound to ASGP-R and is found in early endosomes (43,53). 77% of the 125 I-ASOR present in homogenate was found in the starting microsomes, and of this, 90% was sedimentable (100,000 ϫ g for 1 h) (data not shown). When we used these microsomes in our immunoadsorption protocol, more than 98% of the sedimentable ASOR was in unwashed preclear, A1, and A2 samples (Fig. 3C), confirming our earlier studies (43). Essentially no ASOR was found in the immunoadsorbed SC166 sample (i.e. 0.7% of the ASOR was in each sample (SC166, GAM, and unbound)). ASGP-R appeared to be substantially depleted from the unbound sedimentable material, but we could not quantify the amount in our immunoadsorbed samples, because IgG heavy chains (40 -50 kDa) migrated at the same molecular weight range as the receptor and interfered with the immunoblot analysis (data not shown).
We also determined the distributions of DPP IV and Tf-R throughout the immunoadsorption. The SC166 sample contained ϳ0.1% of the homogenate DPP IV, which was ϳ2.5 times more DPP IV than was found in the GAM sample (range 2-3.3-fold). In contrast, only 0.03% of the homogenate Tf-R was found in either the SC166 or GAM sample. We found that Tf-R was predominantly in the A1 sample (data not shown), suggesting that this recycling receptor was present in the vesicles that contained the ASGP-R. Neither cathepsin D nor ␤-glucuronidase activity, markers of late endosomes and lysosomes (41), was present in the immunoadsorbed SC166 sample. However, both activities were found in the immunoadsorbed A1 and A2 samples. These results indicate that the SC166 samples did not include either early or late endosomes, in contrast to the A1 and A2 samples, which contained both.
Electron Microscopy of the Immunoadsorbed Samples-Examination of the immunoadsorbed samples by electron microscopy revealed the presence of small vesicles bound to SC166 (Fig. 4A); very few vesicles were seen in the GAM control (Fig.  4B). The SC166 vesicles ranged from 60 to 150 nm in diameter and were smaller and more uniform than the tubules and vesicles present in the A1 and A2 samples (43; data not shown). These samples also contained a few vesicles with attached ribosomes. However, both the endosomal (A1 and A2) and transcytotic (SC166) vesicles were minor components of total microsomes, because there was no obvious difference between the starting and unbound material (data not shown).
Newly Synthesized DPP IV Is in Vesicles That Contain Newly Synthesized pIgA-R-Next, we analyzed the distributions of  Fig. 2, except that gel samples were made directly from unwashed pellets at each step. 50% of the preclear and A1 and 25% of the unbound sedimentable material are shown on this immunoblot. B, recovery of pIgA-R in the various samples. Bands from immunoblots like the one shown in Fig. 2A were quantified by video densitometry. This graph shows the average of three experiments. 25% of the pIgA-R from microsomes was in the preclear, 36% was in the A1 sample, 12% in the SC166 sample, but only 2% was in the GAM sample. 6% of the pIgA-R remained unbound; most of this was the immature ER form. We could account for 80 -100% of the starting pIgA-R in unwashed samples; however, washing the vesicles after each immunoadsorption resulted in the loss of as much as 50% of the pIgA-R, indicating the adsorbed vesicles probably break and lose membranes during the washes. C, 125 I-ASOR was removed by immunoadsorption with S. aureus and anti ASGP-R antibodies. Animals were given 125 I-ASOR for 2 min before sacrifice; the labeled ASOR should be in early endosomes at this time. Microsomes were prepared and used in the immunoisolation protocol described in A. The graph shows the distribution of the sedimentable ASOR in unwashed immunoadsorbed samples (100% ϭ the sedimentable 125 I-ASOR found in microsomes, 90% of the total ASOR). Essentially no ASOR is left after A2, showing that early endosomes have been efficiently removed. The pattern seen here, preclear Ͻ A1 Ͼ Ͼ A2 binding, is characteristic of the binding of membranes containing ASGP-R. As seen in B, 12% of the microsomal pIgA-R was found in immunoadsorbed SC166 samples, demonstrating the presence of vesicles containing pIgA-R but not ASGP-R. DPP IV and pIgA-R in samples immunoadsorbed from liver microsomes of rats labeled with Tran 35 S-label for 110 min. The choice of labeling time was based on results from our earlier metabolic studies, which indicated that most of the newly synthesized apical PM proteins should have moved beyond the basolateral PM and would be en route to the apical surface at this time (6,7). The difficulties in following a "pulse"-labeled cohort in vivo have been discussed (7,24). The results presented in Fig. 5 (A and C) show that newly synthesized pIgA-R and DPP IV were both present in vesicles bound by the SC166 immunoadsorbent. (Only the upper band, which represents mature DPP IV, was quantified.) The dramatic difference in the signals between the SC166 and GAM samples shows that pIgA-R containing vesicles specifically adhered to the SC166 immunoadsorbent and that these vesicles contained 35 S-labeled DPP IV. Although the presence of immature proteins indicated that some ER-derived membranes were also bound to the SC166 immunoadsorbent (ER does contain pIgA-R), we analyzed only the mature DPP IV, which we know is derived from post-ER compartments (7). Furthermore, the electron micrographs indicate that most of the immunoadsorbed vesicles were free of bound ribosomes and, conversely, most of the vesicles with attached ribosomes remained in the unbound material.
There was a surprisingly large amount of both 35 S-and total DPP IV in the A1 sample. Furthermore, much more DPP IV and pIgA-R were found in it than in the A2 sample, which was suggestive of specific binding of both proteins in the A1 immunoadsorption step.
As a positive control, we also performed immunoadsorptions using microsomes obtained from the livers of BDL animals, since increased amounts of both DPP IV and pIgA-R are found in the subapical compartment after BDL (1). More 35 S-labeled DPP IV was in the SC166 samples from ligated animals 2 (data not shown), which is consistent with the idea that we immu-noadsorbed vesicles from the subapical compartment. The increase was modest, as we expected from our previous immunoelectron microscopic quantification (1), and was within the variation in our measurements.
The Enrichment in Specific Activity of DPP IV in the SC166 Sample Is Not Due to Golgi Contamination-Analysis of the SC166 sample showed that ST was present, indicating that Golgi-derived vesicles containing pIgA-R were being immunoadsorbed (Fig. 6). Furthermore, in studies of animals labeled for 25 min, when most of the newly synthesized DPP IV should still be in the Golgi complex (7), 35 S-labeled DPP IV was easily detected in the SC166 sample (data not shown). Because some DPP IV remains in the Golgi complex even at 110 min after labeling, we were concerned that the newly synthesized DPP IV detected in the immunoadsorbed SC166 samples was in Golgi-derived vesicles, not in vesicles from compartments involved in transport between the basolateral and apical PM. Therefore, after a 15-min pulse, we treated rats with CHX for the next 95 min. This reduced the Golgi contamination in two ways. First, the amount of pIgA-R left in the Golgi under these conditions should have been very small (24), which would decrease Golgi contamination of the SC166 immunoadsorbent. Second, there should have been less reincorporation of the 35 S label, which would reduce the amount of labeled DPP IV in the Golgi complex. Fig. 6 shows that the amount of ST found in the SC166 sample was reduced to background levels by CHX treatment, indicating that Golgi-derived vesicles were not bound to the SC166 immunoadsorbent.
When we examined the distribution of DPP IV immunoadsorbed from CHX-treated microsomes (Fig. 7), we found that the amount of mature 35 S-labeled DPP IV in the SC166 sample was not changed by CHX treatment. Thus the newly synthesized DPP IV in the SC166 samples was not from Golgi vesicles.
We reasoned that a bona fide transcytotic compartment should not contain newly synthesized basolateral PM proteins, while earlier compartments (including the Golgi complex) might contain newly synthesized proteins en route to both membrane domains. So we looked for a newly synthesized basolateral PM protein (HA321) in the samples immunoadsorbed from CHX-treated microsomes (Fig. 7). 35 5. Immunoadsorption of DPP IV (A and B) and pIgA-R (C and D) from liver microsomes of animals given Tran 35 S-label for 110 min in vivo. After 110 min of labeling, microsomes were prepared and 2600 g were used in the immunoisolation protocol as described under "Materials and Methods" and in Fig. 2. The vesicles bound to the immunoadsorbents were solubilized, divided into portions, either DPP IV or pIgA-R was immunoprecipitated and gel samples were made from the immunoprecipitates. The preclear, A1, and A2 samples were run on 1 gel, while the SC166 and GAM samples were run on another. Each gel contained a standard curve of immunoprecipitated homogenate and 2 dilutions of microsomes. A, fluorogram showing 35 S-labeled DPP IV. Various amounts of each immunoadsorbed sample (indicated by Ad) were loaded to obtain bands of similar densities; the caption shows how much sample was loaded on the gel (equivalent g of starting microsomes used in the protocol). Lanes showing DPP IV directly immunoprecipitated from microsomes are indicated by Ad:None. Newly synthesized DPP IV is seen in all samples; much less is found in the GAM and A2 samples. The lower band is immature DPP IV, which is still present at this time. B, immunoblot showing the total amount of DPP IV in the various samples. There was a surprisingly large amount of total DPP IV found in the A1 sample. The substantial decrease in A2 suggests that this DPP IV was in vesicles that also contain ASGP-R. C, fluorogram showing 35 S-labeled pIgA-R. The immunoadsorbed A1 and SC166 samples contained the most newly synthesized pIgA-R, although a small amount was present in the A2 and GAM samples. D, immunoblot showing the distribution of pIgA-R in the various immunoadsorbed samples. A substantial amount of pIgA-R is found in the SC166 sample, while very little is seen on GAM immunoadsorbent. Presumably the pIgA-R found in the A1 and A2 samples is in early basolateral endosomes, a compartment known to be involved in the transcytosis of pIgA-R.
HA321 was never detected in the SC166 sample. (A small amount of HA321 (0.05% of homogenate) was detected in the immunoblot shown in Fig. 7, but in another sample there was no immunoreactive protein. ) We did not use CHX routinely because it altered the subcellular distribution of pIgA-R and changed the morphology of many hepatocytes (data not shown).
The SC166 Samples Are Enriched in Newly Synthesized DPP IV-We quantified the amounts of 35 S-labeled and immunoreactive DPP IV in the samples from the sequential immunoadsorptions (Fig. 8). Since we would predict that transcytotic vesicles would contain predominantly newly synthesized mol-ecules, the ratio of newly synthesized to total protein should be higher in these vesicles than in homogenate. This ratio will be higher for a protein with a long half-life and low synthetic rate, like DPP IV; most of the DPP IV protein is at the apical PM and very few newly synthesized molecules are in transit to that surface (24). In contrast, older pIgA-R is removed once it reaches the apical PM (22,54,55). Thus, most of the pIgA-R within the hepatocyte is newly synthesized; this means that the specific activity of pIgA-R within transport vesicles is never much greater than the specific activity of homogenate. The ratio ( 35 S-labeled protein, % of homogenate/immunoblotted protein, % of homogenate), which we called the specific activity, indicates the relative enrichment of newly synthesized molecules in a sample in comparison to homogenate.
Microsomes contained about half the 35 S-labeled DPP IV from homogenate, but showed little increase in specific activity (Fig. 8A). Although the early endosomal samples (A1ϩA2) contained 10% of 35 S-labeled DPP IV, the specific activity did not increase (Fig. 8B). In contrast, the SC166 samples had less 35 S-labeled DPP IV (3%), but this DPP IV had a specific activity 25 times higher than homogenate (Fig. 8C, note the change of scale on the specific activity axis). Such a high specific activity indicated that we had immunoadsorbed vesicles involved in the transport of newly synthesized DPP IV 110 min after synthesis. It is formally possible that these vesicles contained older unlabeled pIgA-R, while other vesicles in the SC166 samples contained the newly synthesized pIgA-R. However, the short halflife of pIgA-R in hepatocytes predicts that no membranes would contain primarily unlabeled pIgA-R.
The Vesicles Containing Newly Synthesized DPP IV Are Not Present in a "Carrier" Vesicle Fraction-We also examined immunoadsorptions from the CVF, because previous studies indicated that it was enriched in transcytotic vesicles (38,56). By immunoblot analysis, CVF also contains early endosomes, Golgi derived vesicles and a small amount of PM (data not shown). Approximately 50% of the pIgA-R from the CVF was found in A1ϩA2 samples and another 20% bound to SC166 immunoadsorbent, accounting for the majority of the pIgA-R in the starting fraction (data not shown). Fig. 9 shows the distribution of DPP IV and pIgA-R immunoadsorbed from CVF. The immunoadsorbed A1ϩA2 samples were similar to those adsorbed from microsomes. However, we found very little newly synthesized DPP IV in the SC166 sample.
We also performed immunoadsorptions using CVF from BDL rats. There was no increase in the amount of newly synthesized DPP IV immunoadsorbed by SC166, showing that the subapical compartment was not in this fraction. In addition, the specific activity of the DPP IV was low in all of the samples that were immunoadsorbed from CVF. (The quantification of DPP IV in the SC166 samples from CVF is given in Fig. 10.) We also examined immunoadsorptions from animals labeled for 90 min, to look for transcytotic vesicles that might have already passed through a compartment found in the 110-min CVF. However, the results matched those of the 110-min labeling experiments (data not shown).
The SC166 samples from CVF consistently contained 1-2% of both the 35 S-labeled and total pIgA-R. Moreover, there was less 35 S-labeled pIgA-R in immunoadsorptions of CVF from rats labeled for 25 min than in similar immunoadsorptions from rats labeled for 90 or 110 min. Finally, SC166 samples from CVF did not contain HA321 (data not shown). These characteristics indicated that the SC166 samples from CVF contained vesicles involved in the transport of pIgA-R after it had passed through the basolateral PM; however, these vesicles lacked newly synthesized DPP IV.  35 S-labeled DPP IV. After CHX treatment, the immunoadsorption of Golgi-derived vesicles was greatly reduced, but the amount of labeled DPP IV in the immunoadsorbed SC166 sample did not decrease. The distribution of 35 S-labeled DPP IV in the immunoadsorbed samples was not affected by CHX treatment. B, immunoblot showing the distribution of DPP IV. The distribution of total DPP IV in the immunoadsorption was unchanged by CHX treatment. Because the amount of labeled DPP IV in vesicles bound to SC166 was not decreased by CHX treatment, the newly synthesized DPP IV in this sample was not from Golgi-derived vesicles. C, fluorogram showing newly synthesized HA321. All of these samples were run on a single gel. Newly synthesized HA321 was found in the A1 sample, not in the SC166 sample. D, immunoblot showing the distribution of HA321, these samples were also run on a single gel. There was no 35 S-labeled and little immunoreactive HA321 in the SC166 sample (in fact, an examination of the blot shows mainly background staining). This suggests that the vesicles containing DPP IV and pIgA-R that bind to SC166 immunoadsorbent arise from a compartment involved in transcytosis of apical proteins, after separation from basolateral proteins.

The Characteristics of the Immunoadsorbed Samples Depend on the Fractions Used, Not on the Treatment of the Animal-
The amounts of pIgA-R and DPP IV found in SC166 samples immunoadsorbed from the same fraction appears to be similar in all our treatment groups. Because the variation between animals given different treatments was as great as the variation between groups, we combined the data from all the groups. In immunoadsorptions of microsomes from control, BDL-, or CHX-treated animals labeled for 110 min, 3% of the 35 S-labeled DPP IV was found in the SC166 samples with a specific activity of 25-times homogenate ( Table I) (average of data from 6 animals). These vesicles also contained an average of 13% of the 35 S-labeled pIgA-R. In contrast, in immunoadsorptions of CVF from control or BDL animals labeled for 110 or 90 min, only 0.07% of the 35 S-labeled DPP IV was found in the SC166 samples with a specific activity of 1.4-times homogenate ( Table  II)  were quantified by video densitometry (data from BDL animals are also shown). The darkness of each band was compared to a standard curve of homogenate to obtain the percentage of the starting homogenate found in the immunoadsorbed samples. 35 S values were calculated from fluorograms, and total amounts were calculated from immunoblots. The specific activity was then calculated by taking the ratio of 35 S-labeled protein to total protein for each sample. Each bar shows data from a different animal. A, DPP IV in the starting microsomes. About 40% of the 35 S-labeled DPP IV from homogenate was in the microsomal fraction; slightly less was found in microsomes from CHX-treated animals. The specific activity of the DPP IV in microsomes was about the same as homogenate (1.2 times homogenate overall), so the microsomal fraction was not enriched in newly synthesized DPP IV. B, DPP IV in the A1 and A2 samples. Membranes that contained ASGP-R also contained about 10% of the newly synthesized DPP IV (A1ϩA2, 80% contributed by A1); however, the specific activity of these samples was low, about 1.8 ϫ homogenate. C, DPP IV in the SC166 samples. 3% of the newly synthesized DPP IV from homogenate was found in the immunoadsorbed SC166 samples. But the specific activity of this DPP IV was 25 times that of homogenate (note the change of scale on the right y axis), indicating these samples do contain a compartment(s) transporting newly synthesized DPP IV. Moreover, neither the amount nor the specific activity was decreased by CHX treatment, indicating that this DPP IV was in a post-Golgi compartment.  Fig. 9 were quantified by video densitometry (data from BDL animals are also shown) and analyzed as described in Fig. 8. Each bar shows data from a different animal. Very little newly synthesized DPP IV was found in the SC166 samples; only 0.1% of the homogenate DPP IV. Furthermore the specific activity of this DPP IV was low (1.4 times homogenate). Thus, we did not find newly synthesized DPP IV being transported with pIgA-R in vesicles found in the CVF.
2% of the 35 S-labeled pIgA-R, much less than was found in the samples from microsomes.

DISCUSSION
Three Kinds of Vesicles Transport pIgA-R from Basolateral to Apical PM in Hepatocytes; Only Two of These Carry DPP IV-We were able to immunoadsorb three biochemically distinct vesicles that all appeared to be involved in the transport of newly synthesized pIgA-R between the basolateral and apical PM. First, we found vesicles that contained ASGP-R, a large amount of newly synthesized pIgA-R (18% of homogenate), and a substantial amount of newly synthesized DPP IV (10% of homogenate). Next we found vesicles that lacked ASGP-R but contained a substantial amount of newly synthesized pIgA-R (13% of homogenate) and some newly synthesized DPP IV (3% of homogenate). Finally, there were vesicles in the CVF that contained a small amount of newly synthesized pIgA-R (2% of homogenate), but lacked both ASGP-R and DPP IV. Only the middle group, the vesicles with pIgA-R and DPP IV, contained DPP IV with a high specific activity, a feature we expected in vesicles transporting newly synthesized apical PM proteins. What subcellular compartments involved in the transcytosis of pIgA-R give rise to these three kinds of vesicles?
A Computer Model of the Transport of Newly Synthesized DPP IV-We used the Stella modeling program to explore the movement of DPP IV through the biosynthetic pathway (Fig. 11). The model is based on simple assumptions. We assumed that exit from a compartment is random and proportional to the amount in the compartment, so we have used first order rate constants to model movement between compartments. However, synthesis in the ER required a discrete residence time, because most proteins cannot leave the ER until they have attained the proper conformation (57,58). The kinetics of the maturation and movement of DPP IV determined in earlier in vivo studies from this laboratory were used to define the rate constants used in the model (6, 7).
We determined the steady state distribution of DPP IV and the distribution of a pulse 110 min after synthesis in a model that included the known biosynthetic compartments and only one transcytotic compartment between the basolateral and apical PM (Fig. 11A). This initial choice was based on our identification of the SC166-immunoadsorbed vesicles as the only obvious compartment that was involved in the transcytosis of newly synthesized DPP IV. The movement of DPP IV through the biosynthetic pathway predicted by this model is slightly different from that found in the experimental studies (6, 7) (Fig. 11B). In particular, DPP IV arrives at the apical PM more quickly than was found experimentally, and the peak at the basolateral PM is smaller than the data indicates, although it occurs at the correct time. It is important to remember that the experimental data are from in vivo studies that involved single point determinations from a series of animals; this means that the exact distribution of 35 S-labeled DPP IV is not well defined experimentally. In addition, the amounts in the various compartments are often estimates based on the recoveries in the fractions. Given the uncertainty in the experimental data, the model does a good job of predicting the rate of DPP IV maturation and arrival at the PM. The amount of newly synthesized DPP IV in the transcytotic compartment and the specific activity calculated from the model are shown in Fig. 11C. The model predicts that 5.6% of the DPP IV made in a 15-min pulse should be in this compartment 110 min after synthesis. We found an average of 3% of the labeled DPP IV from homogenate in the SC166 sample from microsomes. Since we may have lost bound membrane during the washes, the amount we found is in good agreement with the model. Although we did not find much 35 S-labeled DPP IV, it is not an unreasonably small amount. The specific activity in the transcytotic compartment in the model is 100 times that of homogenate, while the specific activity in the SC166 samples was 25 times homogenate, indicating that the experimental sample has more unlabeled DPP IV than would be predicted, based on our simple model. However, if we had immunoad-

TABLE II
Amounts of newly synthesized and total DPP IV and pIgA-R immunoadsorbed from CVF Bands from immunoblots and fluorograms like those shown in Fig. 9 were quantified by video densitometry and analyzed like Table I. Values are the average of measurements from 5 animals Ϯ the standard deviation (2, 110-min labeled control; 2, 110-min labeled BDL; 1, 90-min labeled control). While some DPP IV was found in samples adsorbed from CVF, there was very little newly synthesized DPP IV in the immunoadsorbed SC166 sample (Im SC166). The specific activity (SA) of DPP IV in the SC166 sample immunoadsorbed from CVF was much lower than the specific activity of DPP IV in SC166 samples immunoadsorbed from microsomes.

TABLE I Amounts of newly synthesized and total DPP IV and pIgA-R immunoadsorbed from microsomes
Bands from immunoblots and fluorograms like those shown in Figs. 5 and 7 were quantified by video densitometry. The darkness of each band was compared to a standard curve of homogenate to obtain the percentage of the total homogenate found in the immunoadsorbed samples. 35 S values were calculated from fluorograms and total amounts were calculated from immunoblots. The specific activity (SA) is the ratio of 35 S to total amounts. If the variation between animals in a treatment group was larger than the variation between groups, the data from the different groups was combined. Values for DPP IV are the average of measurements from control, BDL-, and CHX-treated animals Ϯ the standard deviation (n ϭ 6) (range for Im SC166 ϭ 1.0 -7.8%, SA ϭ 10 -25 times homogenate). Values for pIgA-R are the average of measurements from the control and BDL animals (n ϭ 4); samples from CHX-treated animals contained less pIgA-R and were not used (range for Im SC166 ϭ 8 -21%, SA ϭ 0.8 -2-fold). sorbed even a small amount of apical PM (as little as 0.05%), it would lower the experimental specific activity dramatically. The generation (during homogenization) of a few vesicles derived from the apical PM with the tail of pIgA-R in the correct orientation would result in the immunoadsorption of some apical PM. There are several reasons to think that the transcytotic compartment is the subapical compartment described in the transcytosis of pIgA-R. First, the SC166 samples contain a significant amount of the homogenate pIgA-R, both immuno-reactive and 35 S-labeled. The two compartments that contain observable pIgA-R in hepatocytes are basolateral endosomes and the subapical compartment (23)(24)(25). Since the SC166 samples lack basolateral endosomal markers, the most likely source is the subapical compartment. The modest increase in newly synthesized DPP IV immunoadsorbed from BDL animals is consistent with this identification, as the two molecules are together in the subapical compartment after BDL (1).
DPP IV Also Passes through Basolateral Endosomes-The characteristics of the DPP IV found in the immunoadsorbed A1 FIG. 11. Analysis of the transcytosis of DPP IV using a Stella model. A, a schematic drawing of the model. The boxes represent intracellular compartments, and the arrows represent the movement of DPP IV from one compartment to the next. The cloud at the beginning signifies the input of newly synthesized protein, while the cloud at the end signifies degradation. The ER conveyor is a mathematical construct that retains newly synthesized DPP IV for a set period (4 min) before allowing movement to the Golgi. All other transport steps are modeled as first order processes. There is only one long lived transcytotic compartment. B, comparison of the model to the in vivo kinetics of DPP IV biosynthesis. These graphs show data from earlier studies on the maturation and movement of newly synthesized DPP IV from the ER to the apical PM in comparison to the kinetics calculated from the model. The left graph shows movement from the ER through the Golgi, while the right side shows movement through the basolateral PM to the apical PM. The model matches the kinetics of DPP IV biosynthesis fairly well, although the model brings DPP IV to the apical membrane a little quickly. The model also predicts a smaller amount of newly synthesized DPP IV at the basolateral membrane at 70 min. However, the data show only that the peak of basolateral DPP IV occurs around 70 min; we obtained the peak amount by subtracting the amounts in the ER and Golgi at this time from 100%. If a substantial amount of newly synthesized DPP IV is basolateral endosomes, as is indicated by the immunoadsorption data, then the amount at the basolateral PM decreases. Adding basolateral endosomes to the kinetic model also delays the arrival of newly synthesized DPP IV at the apical PM, so the revised model fits the data better. C, characteristics of DPP IV at 110 min after synthesis. The model predicts that 6% of the newly synthesized DPP IV should be in the subapical compartment at this time. We found about 3% of the 35 S-labeled DPP IV in the immunoadsorbed SC166 samples, so the model indicates that this small amount is not unreasonable. The model predicts that DPP IV in all the biosynthetic compartments should have a high specific activity and that in the transcytotic compartment the specific activity should be even higher than the 25-fold increase we found in the SC166 samples. However, even a very small amount of apical PM in our immunoadsorbed samples would decrease the specific activity dramatically because of the large amount of unlabeled DPP IV at the apical surface. and A2 samples (10% of the 35 S-labeled DPP IV and 5% of the total DPP IV, with a relative specific activity of 2 times homogenate) cannot be fit into our current kinetic model. The model predicts that the relative specific activity of all the biosynthetic compartments, including the ER, should be about 30 times that of homogenate. Moreover, putting a significant amount of the total DPP IV (more than 2%) into any biosynthetic compartment at steady state prevents the pulse of newly synthesized DPP IV from reaching the apical PM for 6 h. Since we know that ASGP-R is present in many intracellular compartments, it is likely that the A1 and A2 samples contain a mixture of vesicles and that the DPP IV in these samples is present in two kinds of vesicles, some that contain mainly labeled DPP IV and some that contain older, unlabeled DPP IV.
Only early endosomes are present in sufficient quantity in the A1ϩA2 samples to account for 10% of the newly synthesized DPP IV. We speculate that the unlabeled DPP IV in A1ϩA2 is from late endosomes. Internalized material from the basolateral and apical surfaces meet at the level of late endosomes in MDCK cells (59), so this compartment could contain proteins from both PM domains in hepatocytes. Thus, the unlabeled DPP IV found in A1 and A2 could be older DPP IV internalized from the apical PM and on its way to degradation (23).
If the model is adjusted to show transit of newly synthesized DPP IV through basolateral endosomes, newly synthesized DPP IV leaves the basolateral PM earlier and arrives at the apical PM later. Thus, the revised model fits the experimental kinetic data better. It also predicts the movement of pIgA-R from the Golgi to the apical PM fairly accurately (data not shown).
There Is a Transport Step Where pIgA-R and DPP IV Are Separate-In this study we also found vesicles that contained labeled pIgA-R, little labeled DPP IV, and yet had many characteristics of vesicles involved in the late transport of pIgA-R. This is consistent with earlier studies showing that vesicles immunoadsorbed from CVF were enriched in mature forms of pIgA-R, contained dimeric IgA and therefore were transcytotic vesicles (38). Antibodies generated against components of samples immunoadsorbed from CVF recognize a 108-kDa protein that has been implicated in vesicle fusion throughout the biosynthetic pathway (38,56,60); thus, these vesicles could be carriers of newly synthesized pIgA-R. However, the low recovery of labeled pIgA-R from a similar starting fraction (CVF) in our study (2% of homogenate), together with our failure to find labeled DPP IV in immunoadsorptions of CVF from BDL ani-mals, suggests that these vesicles are not part of the subapical compartment. They could be vesicles involved in internalization of pIgA-R from the basolateral PM, in delivery of pIgA-R to the apical PM from the subapical compartment, or in retrieval of pIgA-R from the apical PM. It is possible that pIgA-R and DPP IV are internalized from the basolateral PM separately, because DPP IV lacks the cytoplasmic sequences associated with rapid uptake into clathrin-coated vesicles (61)(62)(63). At present, we favor the notion that SC166 samples from CVF contain pIgA-R internalized from the basolateral PM in vesicles without sufficient ASGP-R to be bound in the A1 or A2 steps.
DPP IV and pIgA-R Share the Same Transcytotic Pathway-Our analysis of the immunoadsorbed samples, combined with kinetic modeling, suggests that DPP IV uses the same transcytotic pathway that carries pIgA-R (Fig. 12). DPP IV is internalized from the basolateral PM and delivered to early endosomes where it is sorted into transcytotic carriers. Then it arrives at the subapical compartment on the way to the apical PM. After several days at the apical PM, DPP IV is retrieved and sent to late endosomes where a significant amount accumulates before being degraded. We believe these are all "permanent" compartments; the nature of the transport intermediates is still unknown. While it seems likely that all transport intermediates are vesicles, this has not been proved. Mostov (64) has suggested that in MDCK cells the connection between basolateral and apical endosomes is tubular, based on the presence of basolateral receptors in apical endosomes. However, other researchers have reported contradictory findings (65), and we have no evidence of basolateral PM proteins in the subapical compartment of hepatocytes.
Our data suggest that a single transcytotic pathway could carry all proteins from basolateral endosomes to the apical PM in hepatocytes. Different proteins may be internalized by different mechanisms at the basolateral PM: clathrin-mediated uptake for pIgA-R and ASGP-R (16,17,66,67), non-clathrinmediated uptake for DPP IV (68,69); and perhaps caveolaemediated uptake for GPI-anchored proteins like 5Ј nucleotidase (70,71). However, they all arrive at early endosomes (72); presumably they can then be sorted into transcytotic vesicles. A comparison of our kinetic models for transcytosis of 125 I-pIgA ligand and 35 S-DPP IV indicates that the transcytosis of both 3 The HA321 data indicated that only a small amount of the basolateral PM in microsomes was immunoadsorbed. This appears to contradict our results with Tf-R and ASGP-R, since these proteins were removed from the starting microsomes and they are certainly present in basolateral PM as well as early endosomes. However, the explanation lies in the relative distributions of the various proteins in the two compartments. If 20% of the Tf-R is at the PM (73, 74) and 25% of the PM was recovered in microsomes, then about 5% of the total receptor present in microsomes was from basolateral PM (30 -40% of the total Tf-R is in microsomes) presumably the rest was from early endosomes. If we assume that all of the HA321 in microsomes represents basolateral PM and only 5% was bound to our immunoadsorbent, then 90 -95% of Tf-R from basolateral PM or about 5% of the total receptor should remain unbound. We found only 1% of the Tf-R left unbound or about 1/5 of the amount we expected. One possible explanation is that the places in the basolateral PM where recycling receptors like ASGP-R and Tf-R are clustered were more efficiently immunoadsorbed, resulting in an almost complete removal of these proteins, while the rest of the basolateral PM remained unbound.
FIG. 12. Transcytosis of DPP IV and pIgA-R. Compartments containing DPP IV are shown in black, those with pIgA-R are shown in white, and compartments where the two proteins are together are shown in gray. Since DPP IV lacks a signal for inclusion in clathrincoated vesicles, it is probably internalized from the basolateral surface by vesicles that do not contain pIgA-R. DPP IV is then transported to early endosomes, which do contain pIgA-R, and both proteins are transcytosed across the cell to the subapical compartment. Finally, they are delivered to the apical PM; we do not know if the same delivery vesicles take both proteins to the apical PM.
proteins proceeds with the same kinetics. This suggests that DPP IV is both internalized and sorted into the transcytotic pathway efficiently. In contrast, 5Ј nucleotidase is transcytosed very slowly, as if internalization or sorting is difficult for GPIanchored proteins. Thus, while the same transcytotic pathway may be used by all proteins, the interactions with the machinery may be quite different. Finally, if DPP IV is internalized separately from pIgA-R, there should be another class of endocytic vesicles that lack both pIgA-R and ASGP-R. Clearly, more study is needed to determine the exact pathway used by the multiple types of apical membrane proteins going from the basolateral to the apical PM.