Agonist-induced coordinated trafficking of functionally related transport proteins for water and ions in cholangiocytes.

We previously proposed that ductal bile formation is regulated by secretin-responsive relocation of aquaporin 1 (AQP1), a water-selective channel protein, from an intracellular vesicular compartment to the apical membrane of cholangiocytes. In this study, we immunoisolated AQP1-containing vesicles from cholangiocytes prepared from rat liver; quantitative immunoblotting revealed enrichment in these vesicles of not only AQP1 but also cystic fibrosis transmembrane regulator (CFTR) and AE2, a Cl- channel and a Cl-/HCO3- exchanger, respectively. Dual labeled immunogold electron microscopy of cultured polarized mouse cholangiocytes showed significant colocalization of AQP1, CFTR, and AE2 in an intracellular vesicular compartment; exposure of cholangiocytes to dibutyryl-cAMP (100 microm) resulted in co-redistribution of all three proteins to the apical cholangiocyte plasma membrane. After administration of secretin to rats in vivo, bile flow increased, and AQP1, CFTR, and AE2 co-redistributed to the apical cholangiocyte membrane; both events were blocked by pharmacologic disassembly of microtubules. Based on these in vitro and in vivo observations utilizing independent and complementary approaches, we propose that cholangiocytes contain an organelle that sequesters functionally related proteins that can account for ion-driven water transport, that this organelle moves to the apical cholangiocyte membrane in response to secretory agonists, and that these events account for ductal bile secretion at a molecular level.

Cholangiocytes are cells that line intrahepatic bile ducts and, like other epithelia, possess discrete, specialized apical and basolateral membranes. Each cholangiocyte membrane contains specific receptors and flux molecules (i.e. channels, exchangers, and transporters) that accomplish the vectorial movement of solutes, ions, and water across the biliary epithelial barrier (1)(2)(3), resulting in ductal bile formation by as yet unclear molecular mechanisms. Recently, we proposed a molecular model for hormone-induced bile secretion by cholangio-cytes. The key feature of this model is the agonist-induced, coordinated, exocytic insertion into and endocytic retrieval from the apical cholangiocyte plasma membrane of key flux molecules that, in the unstimulated state, are sequestered in an intracytoplasmic vesicular compartment (4). We also proposed that one of these flux proteins was AQP1, a member of the aquaporin (AQP) 1 family of water channels that mediate the bidirectional, passive movement of water molecules across epithelial cells in response to osmotic gradients established by ions and solutes (1,(5)(6)(7). In support of this model are data showing that secretin, a hormone that stimulates ductal bile secretion, can also trigger the exocytic insertion of AQP1 into the apical cholangiocyte plasma membrane (6). More recently, we also provided data in hepatocytes, the other epithelial cell in the liver involved in bile formation, indicating that recycling of AQP8 may account for agonist-induced canalicular bile secretion (8). To extend this general model, we now propose that these agonist-responsive, AQP1-containing vesicles also contain other molecules that are necessary to establish the osmotic gradients that drive passive water movement.
In the studies described here, we found that cholangiocytes contain a subpopulation of vesicles in which the water channel AQP1, the chloride channel CFTR, and the chloride/bicarbonate exchanger, AE2, are sequestered in the basal state and that these proteins move together in a microtubule-dependent manner from an intracellular location to the apical cholangiocyte plasma membrane in response to choleretic stimuli. By a variety of independent and complementary biochemical and morphologic approaches, the data support the notion that a novel organelle containing proteins that accomplish both water and ion transport exists in cholangiocytes and participates in hormone-induced bile secretion. This concept is supported by a body of evidence on the physiological implications of colocalization of functionally related molecules in the kidney, including aquaporins with ion transporters, and membrane relocalization after treatment with agonists (9 -11).

EXPERIMENTAL PROCEDURES
Preparation of Cholangiocytes-Male Fisher 344 rats after induction of selective cholangiocyte hyperplasia by bile duct ligation (BDL) were used. The BDL rat is an experimental model of selective cholangiocyte proliferation useful for the study of ductal bile secretion (12). Highly purified cholangiocytes were isolated from livers of BDL rats as previously described (13).
Immunoisolation of AQP1-, CFTR-, and AE2-containing Vesicles-Purified cholangiocytes were sonicated, and a postnuclear supernatant was prepared. The postnuclear supernatant was further purified on a linear sucrose gradient (8.6 -40%), and the resulting microsomal fraction sedimented at 200,000 ϫ g for 60 min to remove any remaining cytosol. This microsome-rich fraction is referred to as "starting material." Immunoisolation was performed by a modification of the method of Howell et al. (14). Briefly, tosylated superparamagnetic beads (Dynabeads M-500, Dynal Inc., Oslo, Norway) were incubated overnight with a goat anti-rabbit IgG (Fc) linker secondary antibody (Biodesign International, Kennebunk, ME) at a concentration of 10 g of antibody/mg of beads in borate buffer (100 mM H 3 BO 3 , pH 9.5). In this and all subsequent steps, beads were collected with a magnetic device (MPC; Dynal, Inc.). The amount of linker antibody that bound to the beads was reproducibly ϳ5 g/mg of beads. Coated Dynabeads were rotated for 12 h with a primary polyclonal antibody to AQP1 (10 g/mg of beads) and 500 g of purified microsome starting material. Control experiments were performed using Dynabeads coated with secondary antibody alone. Beads were collected and rinsed with incubation buffer, and this immunoisolated fraction then represented the AQP1-containing vesicles specifically bound to the beads. Supernatant that remained after separation from the immunoisolated bound fraction was centrifuged at 16,000 ϫ g for 10 min, and the pellet was resuspended in 0.25 M sucrose, representing the unbound fraction. For simplicity, we refer to the microsome-rich fraction as "starting material," the fraction containing the immunoisolated AQP1 vesicles attached to the beads as "bound," and the supernatant containing non-AQP1-containing vesicles as "unbound." These three fractions were subsequently analyzed biochemically and by immunoblotting.
In a separate set of experiments aimed at determining the reversibility and specificity of the vesicle isolation technique, we used polyclonal antibodies to either CFTR or AE2 as the linking antibody to the coated immunomagnetic beads and thereby immunoisolated a population of CFTR-containing vesicles or AE2-containing vesicles, respectively, which were then analyzed biochemically and by immunoblotting.
Characterization of Isolated Vesicles-Biochemical characterization using specific organelle assays was performed to determine the purity of the starting material prior to the immunoisolation protocol. To determine the specificity of the immunoisolation technique, the same assays were performed on both the bound and the unbound fractions. N-acetyl-␤-glucosaminidase (lysosomes) activity (15), malate dehydrogenase (mitochondria) activity (16), lactate dehydrogenase (cytosol) activity (17), and alkaline phosphodiesterase 1 (plasma membrane) were assayed as previously reported (18). Microsome activity was determined by two methods: microsomal esterase, assessed by the method of Beaufay and Berthet (18), and glucose-6-phosphatase using a commercially available spectrophotometric kit from Sigma. Golgi activity was determined fluorometrically with the substrate, 4-methylumbelliferyl-␣-D-mannopyranoside, by the method of Farquhar (19). Protein concentration was determined by the fluorescamine method using bovine serum albumin as a standard (20).
As additional evidence of vesicle purity, solubilized fractions from the immunoisolation protocol were immunoblotted with specific antibodies to (i) calnexin (Affinity Bioreagents, Inc., Golden, CO), an integral membrane protein of the endoplasmic reticulum; (ii) TGN-38, a trans-Golgi network protein; and (iii) AQP4 (Alpha Diagnostics International, Inc., San Antonio, TX), a protein found in cholangiocytes exclusively in the basolateral plasma membrane.
Vesicle suspensions were prepared for transmission electron microscopy using a primary fixative of 2.5% glutaraldehyde, 0.25 M sucrose for 1 h at 4°C. A second and poststaining fixative consisting of 1% OsO 4 and 1% uranyl acetate, respectively, was followed by dehydration in alcohol and embedding in Quetol as previously described (21). Electron micrographs were generated on a Phillips 201 electron microscope.
In a set of experiments designed to assess the specificity of the antibody for CFTR used for immunoblotting, we performed immunoblots for CFTR and AQP1 on cell lysates from cultured mouse cells obtained from a line of CFTR (Ϫ/Ϫ) knockout mice. In these immunoblots, the cell lysate was positive for AQP1 and negative for CFTR, demonstrating the specificity of the CFTR antibody (data not shown).
Dual Labeled Immunogold for AQP1, CFTR, and AE2-Immortalized, nonmalignant mouse cholangiocytes (23) were grown on type 1 collagen (Biocoat; Becton Dickinson, Bedford, MA) in culture media of minimal essential media (Invitrogen) with 10% fetal bovine serum and 1ϫ penicillin/streptomycin. Confluent, polarized cholangiocytes were treated with media alone (basal) or media with the choleretic agonist, dibutyryl-cyclic AMP (100 M), forskolin (3 M), and isobutylmethylxanthine (100 M) (to prevent the degradation of cAMP) for 30 min at 37°C. The cells were then processed for fixation and immunogold labeling according to the method of McCaffery and Farquhar (24) and then exposed to primary antibodies to AQP1 (rabbit polyclonal) and CFTR (mouse monoclonal). Grids were then incubated with a mixture of donkey anti-rabbit IgG linked to 6-nm gold (AQP1) and donkey antimouse 12-nm gold (CFTR) and examined with a Joel 1200 electron microscope at 80 kV.
Electron micrographs from both basal and agonist groups were printed at a final magnification of ϫ30,000, and pictures were randomized for counting. Vesicles were quantified in four groups: (i) vesicles containing only 6-nm gold particles, representing specific labeling for AQP1; (ii) vesicles containing only 12-nm gold particles, labeled for CFTR; (iii) vesicles labeling for both AQP1 and CFTR; and (iv) vesicles without any labeling. In a parallel experiment, vesicles were exposed to 5-nm gold particles labeled for AE2 and 10-nm gold particles labeled for CFTR. The same criteria used to quantitate the associated gold particles for AQP1 and CFTR was used as described for AE2 and CFTR. The micrographs represent photographs from six grids from each of three different vesicle preparations. Micrographs were counted randomly in each group and counted three times to obtain means Ϯ S.E. Data are expressed as the number of gold particles in each group as a percentage of the total.
Animal Model for in Vivo Experiments-In vivo experiments were performed using rats in which the common bile duct was ligated for 1 FIG. 1. A, electron micrographs indicating that vesicles containing AQP1 (arrowheads) specifically attach to the surface of a magnetic bead coated with AQP1 antibody (top panel). In the lower panel, no vesiclelike structures are seen attached to control beads coated only with the secondary, non-AQP1 antibody (magnification, ϫ30,000). B, immunoblotting for calnexin, TGN-38, AQP4, and AQP1 was performed using 10 g of positive control and 25 g each of starting material, vesicles that associated with the magnetic beads (bound) and unbound fractions.
In separate experiments, BDL rats were injected intravenously with 0.5 mol/100 g of body weight colchicine or ␤-lumicolchicine in phosphate-buffered saline 2 h prior to administration of secretin. Bile was collected, and cholangiocytes were isolated immediately after secretin administration as previously described (6).
For immunoblotting experiments, apical and basolateral plasma membranes were prepared from isolated cholangiocytes derived from whole liver as previously described by us (21).
Statistical Analysis-Densitometric scanning was performed using the software Molecular Analyst (Bio-Rad). All values are expressed as mean Ϯ S.E. Significance was determined using Student's t test; p Ͻ 0.05 was considered statistically significant. Fig. 1A are electron micrographs confirming that the vesicles immunoisolated with an antibody to AQP1 specifically attach to magnetic beads coated with this antibody. In the top panel, numerous vesicle-like structures can be seen attached to the surface of a magnetic bead coated with AQP1 antibody. In the lower panel, no such structures are seen when control beads coated only with the secondary, non-AQP1 antibody were used. Fig. 1B displays results of immunoblots for proteins of the endoplasmic reticulum (calnexin), Golgi (TGN-38), plasma membrane (AQP4), and AQP1. Calnexin, TGN-38, and AQP4 were detected at their expected molecular weights and in their respective positive controls but were absent from the bound fraction. AQP1 was present at 28 kDa in the positive control of rat RBC plasma membrane and in the microsome and bound fractions. Table I shows a biochemical analysis of the microsomal starting material, the fraction bound to the immunomagnetic beads, and the unbound fraction. As expected, the microsomal derived starting material showed an enrichment in activity of the microsomal markers, microsomal esterase, and glucose-6-phosphatase; enzyme activities for lysosomes, mitochondria, and plasma membranes were undetectable, demonstrating that the starting material did not contain these organelles. The amounts of measurable lactate dehydrogenase (cytosolic marker) and ␣-D-mannopyranoside (Golgi marker) were negligible.

Characterization of AQP1-containing Vesicles-Shown in
Immunoblotting of Isolated Subcellular Fractions-The immunoblot shown in Fig. 2A illustrates that AQP1, CFTR, and AE2 were all significantly (p Ͻ 0.05) enriched in the beadassociated AQP1-immunoisolated vesicles compared with the microsomal preparation: AQP1, 2.4-fold; CFTR, 1.7-fold; AE2, 2.8-fold. There was virtually no detectable AQP1, CFTR, or AE2 in the lanes loaded with an equal amount of protein from the unbound fraction. Fig. 2, B and C, shows the results of immunoblotting for AQP1, CFTR, and AE2 on fractions from the reverse immunoisolations using antibodies to CFTR to isolate CFTR-containing vesicles (B) or antibodies to AE2 to isolate AE2-containing vesicles (C). In vesicles isolated with the antibody to CFTR (B), the bound fractions were significantly (p Ͻ 0.05) enriched 3.2fold over the starting material for CFTR, 2.5-fold for AQP1, and 2.7-fold for AE2. In vesicles isolated with the antibody to AE2 (C), the bound fractions were significantly (p Ͻ 0.05) enriched 3.4-fold over the starting material for AE2, 2.6-fold for AQP1, and 2.5-fold for CFTR. Fig. 3, A and B, illustrates colocalization in two sets of dual labeled immunogold experimental combinations (AQP1 with CFTR (top) and AE2 with CFTR (bottom)). Our observations FIG. 2. A, immunoblotting for AQP1, CFTR, and AE2 on subcellular fractions prepared from microsomes (starting material), vesicles specifically associated with the magnetic beads (bound), and vesicles that did not bind to the beads (unbound) using AQP1 antibody. B, reverse vesicle immunoisolation using CFTR antibody; C, reverse vesicle immunoisolation using AE2 antibody. A 10% polyacrylamide gel was loaded with fractions for AQP1 (50 g of total protein/lane), CFTR (100 g of total protein/lane), and AE2 (50 g of total protein/lane) . Data are expressed in arbitrary densitometry units as means Ϯ S.E. *, p Ͻ 0.05 (Student's t test). revealed populations of small beads alone (6-nm gold for AQP1 and AE2), large beads alone (12-nm gold for CFTR), and clusters of colocalized small and large beads together shown in the enlarged insets. Quantitation revealed that AQP1 and CFTR and AE2 and CFTR colocalized in 25 and 34%, respectively, of the total vesicles counted (Fig 3, A1 and B1). Following agonist stimulation (Fig. 4), we observed relocation of the gold particles to the apical plasma membrane (Fig. 4, A1 and B1) compared with the basal state (Fig 4, A and B), consistent with our previous biochemical and morphologic data. Since the normal mouse cholangiocytes were grown in a polarized confluent manner prior to fixation, we concluded that the agonist-induced migration is domain-specific. Quantitation of the gold particles (Fig. 4, C and D) revealed a significant 2-fold or greater increase in the apical membrane-associated redistribution. In addition, the apical membrane-associated labeling contains numerous clusters of small and large gold particles, implying that the transporters may remain associated after agonist-induced redistribution to the apical membrane. Statistical analysis was performed on the quantitation of gold particles in the basal state according to the Kullback-Liebler hypothesized model of randomness (26,27) to assess "degree of association" between AQP1 and CFTR and between AE2 and CFTR in an effort to minimize the possibility that the apparent colocalization of the immunogold beads was a random event.
Hypothetically, the proportions of vesicles with small gold beads only, large beads only, and both small and large beads should all be equal to 1 ⁄3. To measure the departure from randomness for each reading from a given micrograph, the Kullback-Liebler distance was computed. A calculated value of Ͻ0.05 indicates a significant departure from randomness. The measured value of the colocalized AQP1/CFTR and AE2/ CFTR was 0.031 and 0.030, respectively, supporting our inter-pretation that the observed colocalization of small and large beads was not a random event.
Apical Versus Basolateral Insertion of AQP1, CFTR, and AE2 in Secretin-treated BDL Rats-Shown in Fig. 5A are data from in vivo experiments demonstrating the secretin-induced relocation of AQP1, CFTR, and AE2 to the apical plasma membrane of rat cholangiocytes.
In the unstimulated state, any AQP1 in the plasma membrane is present principally in the apical domain. Following intravenous infusion of 10 Ϫ7 M secretin into BDL rats, which causes a 2-4-fold increase in bile secretion (Fig. 5C), the total amount of AQP1 is increased in both the band representing total plasma membrane and in the apical domain by 2.4-and 2.2-fold, respectively, relative to membranes isolated from rats infused with vehicle alone. Immunoblots for CFTR and AE2 were performed on the same fractions of total plasma membranes as well as fractions enriched in the apical and basolateral plasma membrane domains. In the absence of an intravenous infusion of secretin (i.e. vehicle infusion), total plasma membrane CFTR and AE2 are predominantly apical in their location. Following IV infusion of secretin, the amount of CFTR and AE2 in plasma membranes, like AQP1, also increases 3.8-fold in the total plasma membrane and 2.4-fold in the apical plasma membrane.
We next employed the microtubule blocker colchicine, which we had previously shown inhibits secretin-induced bile flow (6). These data are shown again in Fig. 5C for the purpose of correlating secretin-induced changes in bile flow with the blocked increase in movement to the plasma membrane of AQP1, CFTR, and AE2. As shown in Fig. 5B, pretreatment of rats with colchicine, followed by administration of secretin, nearly abolished the increase of AQP1, CFTR, and AE2 in cholangiocyte plasma membranes. The inactive analog, ␤-lumi- colchicine, did not affect the action of secretin on AQP1, CFTR, and AE2 membrane distribution or bile flow compared with rats receiving no treatment. DISCUSSION This is the first report to our knowledge suggesting that cholangiocytes contain agonist-responsive specialized organelles in which functionally related proteins involved in water and ion transport are sequestered. Using independent and complementary biochemical and morphologic techniques, we found that (i) in the unstimulated or basal state, AQP1, CFTR, and AE2 colocalize to an intracellular vesicular compartment in cholangiocytes, and (ii) after exposure of cholangiocytes to dibutyryl-cAMP in vitro or after secretin infusion in vivo, AQP1, CFTR, and AE2 redistribute together to the apical cholangiocyte plasma membrane, an event that is blocked by the microtubule inhibitor, colchicine. The data are consistent with the notion that secretin, by interacting with specific cAMP-coupled receptors on the basolateral domain of cholangiocytes, induces microtubule-dependent movement of vesicles containing AQP1, CFTR, and AE2 and exocytic insertion of these functionally related flux proteins into the apical plasma membrane domain. Increased CFTR trafficked to the secretory pole of cholangiocytes would facilitate active secretion of Cl Ϫ , driving HCO 3 Ϫ secretion into the ductal lumen by AE2, result-ing in passive movement of water through AQP1 and net ductal bile secretion. We used immunoisolation to prepare a subpopulation of vesicles from cholangiocytes that were enriched in AQP1, CFTR, and AE2. The degree of enrichment we observed in the immunoisolated fraction seems reasonable given that our starting material is a purified microsomal fraction rather than a postnuclear supernatant. Moreover, using three different antibodies, we achieved the same relative degree of enrichment of the three proteins in the immunoisolated fraction. In comparably designed studies to immunoisolate endothelial plasmalemmal vesicles, the caveolin fraction that bound to the magnetic beads was enriched 4 -5-fold compared with a more crude starting material of whole rat lung (25). In addition, the antibodies and proteins in the isolation and wash buffer (required for the immunopurification) adhere to the beads and cause an important underestimation of actual enrichment.
The concept that transporters for water and ions reside in the same vesicle in cholangiocytes is teleologically appealing, since water moves passively in response to osmotic gradients established by active Cl Ϫ -driven HCO 3 Ϫ secretion, the currently accepted paradigm of ductal bile formation. These transient gradients are thought to be established by the transport of bicarbonate via the Cl Ϫ /HCO 3 Ϫ exchanger located in the apical cholangiocyte membrane and operating in parallel with the CFTR chloride channel. Recently published observations, (3,5,6,28) and those reported here support the notion that a secretin-induced increase in cholangiocyte cAMP both triggers the exocytic insertion of AQP1 into the cholangiocyte apical membrane and activates apical CFTR chloride channels. Previously published data from our laboratory have demonstrated that exposure of highly purified isolated rat cholangiocytes to secretin caused significant, dose-dependent increases in osmotic membrane water permeability (P f ) (e.g. increased by 60% with 10 Ϫ7 M secretin), which was reversibly inhibited by the water channel blocker, mercuric chloride (5).
Several lines of evidence exist that support the colocalization of water with ion transporters or other regulators of exocytosis, based on data generated in the tubule cells of the kidney using techniques of immunblotting and immunostaining. These include SNAP-25-associated Hrs-2 colocalizing with AQP2 (11), AQP2 with VAMP2 (10), and AQP2 with CFTR and the amiloride-sensitive epithelial Na ϩ channel (EnaC) (9). Evidence for the cAMP-mediated stimulation of apical targeting of vesicles containing flux molecules has been shown not only by us for AQP1 in cholangiocytes but also in a number of other epithelia, including in T84 colonocytes for CFTR and in kidney tubule epithelia for AQP2 (28 -32). Moreover, it has been reported that secretin stimulates Cl Ϫ /HCO 3 Ϫ exchange in isolated cholangiocytes (33) and increases CFTR in the apical membrane of pancreatic duct cells via cAMP (34). The last three amino acids in the COOH terminus of CFTR (TRL) comprise a PDZ-interacting domain that appears to be required for the targeting of CFTR to the apical plasma membrane in human airway and kidney epithelial cells (35). In addition, it has been proposed that CFTR and its PDZ proteins in concert with ezrin affect other epithelial ion channels in their movement to a subapical compartment (35). Since AQP1 does not contain a known apical targeting signal, colocalization of AQP1 in a vesicle with CFTR would provide a plausible mechanism explaining how AQP1 can redistribute to the apical cholangiocyte plasma membrane in response to a secretory agonist. In kidney collecting tubule cells, AQP2 colocalizes with VAMP1 and -2, and these proteins are all trafficked to the apical plasma membrane in response to vasopressin (36). Thus, the model of  A1 and B1). Agonist stimulation caused a redistribution of dual labeled immunogold and demonstrates a consistent and significant movement of vesicles containing colocalized AQP1 with CFTR and AE2 with CFTR to the apical plasma membrane (arrowheads). Quantitation of the apical membrane-associated labeling (C and D) revealed a significant 2-fold or greater increase in the apical membrane-associated redistribution. Bar, 200 nm at a magnification of ϫ 30,000. *, p Ͻ 0.05 (Student's t test).
agonist-driven, polarized targeting and insertion of flux proteins into specific plasma membrane domains of secretory epithelia has precedent.
Our immunogold data provide additional evidence for the colocalization of AQP1, AE2, and CFTR and are reminiscent of the work of Nielsen et al. (36), who demonstrated dual labeling and colocalization of AQP2 with VAMP2. The apparent increase in membrane-associated vesicles containing only AQP1, which was greater than the individual increases of CFTR alone and AE2 alone, may indicate the presence of specific signal or target proteins on the plasma membrane that are selective to AQP1 and not CFTR or AE2. Alternatively, the vesicles con-  's t test). Data for AQP1 were published previously (6). The same samples were then used to generate the data for CFTR and AE2. C, effect of colchicine on secretin-induced bile flow. One-week BDL rats were injected with colchicine or lumicolchicine as described for B. After 2 h, secretin (1 ml, 10 Ϫ7 M) or vehicle was administered intravenously for 20 min, and bile was simultaneously collected. Data are means Ϯ S.E. for 4 -6 animals in each experimental group. **, p Ͻ 0.01 compared with corresponding (Ϫ) secretin values. *, p Ͻ 0.05 significant decrease in bile flow relative to secretin plus lumicolchicine. Data were published previously (6). taining both flux proteins may fuse with the plasma membrane where the proteins are subsequently sorted independently, causing a difference in the net number of gold particles remaining on the membrane at the time of fixation. Whether these functionally related flux proteins are also retrieved together upon withdrawal of the secretory stimulus or in response to a cholestatic stimulus (e.g. somatostatin) remains to be elucidated. Data generated using the model of isolated bile duct units from AQP1-ko mice were interpreted as suggesting that AQPs may not be rate-limiting for transcellular water movement in biliary epithelia; interpretation of these data should also include the possibility that up-regulation of other AQPs that we have shown are present in cholangiocytes may compensate for the absence of AQP1 in cholangiocytes (37).
Thus, although recycling as a regulatory mechanism for epithelial cell transport is an accepted paradigm, our work here substantially extends this concept by introducing the novel notion that a single intracellular hormone-responsive vesicular compartment might contain several proteins that are linked both anatomically and functionally in an integrated fashion to accomplish a cellular process, in this case coupled water and electrolyte secretion.
Our in vivo observation that administration of the microtubule depolymerizing agent, colchicine, inhibited the secretininduced increase of AQP1, AE2, and CFTR in the cholangiocyte plasma membrane as well as secretin-induced ductal bile secretion is consistent with studies in kidney-proximal tubule cells, showing that microtubules are required for insertion of AQP1 into the plasma membrane (38). Moreover, the extension of our observations to this in vivo situation provides additional evidence in a more physiologically relevant model that the molecular events we describe are important in ductal bile secretion.
It seems plausible that defects or alterations in the coordinated trafficking of related proteins could be responsible for a variety of diseases. For example, it has been reported that both message and protein expression of the Cl Ϫ /HCO 3 Ϫ exchanger is diminished in patients with primary biliary cirrhosis, a disease in which the cholangiocyte is the target cell (39). The secretion of bicarbonate into bile is strongly linked to increased efflux of chloride into the lumen and the conductance of apical CFTR (40). Moreover, the aquaporin family of water channel proteins also include mutations and abnormalities in transport that may result in nephrogenic diabetes insipidis (AQP2) (41), cataracts (AQP0) (42), Sjögren's syndrome (AQP5) (43,44), or an impaired urinary concentration ability (AQP1) (45).
In conclusion, the data presented in this study, together with our previous observations (5,6), support our proposed molecular model of ductal bile secretion. The model is now extended by the proposed existence of a biochemically and morphologically characterized novel transporting organelle in cholangiocytes that appears to contain proteins for water and chloride transport, flux proteins that play integral roles in hormone-induced ductal bile secretion. Clearly, additional studies will be necessary to fully elucidate the quantitative relevance of our observations to total bile formation.