Rab3D, a Small GTP–Binding Protein Implicatedin Regulated Secretion, Is Associated With the

Transcytotic Pathway in Rat Hepatocytes

JANET M. LARKIN,1,2 BONNIE WOO,1 VIJAYAN BALAN,2 DAVID L. MARKS,2 BARBARA J. OSWALD,2

NICHOLAS F. LARUSSO,2 AND MARK A. MCNIVEN2

Rab3 isotypes are expressed in regulated secretory cells.Here, we report that rab3D is also expressed in rat hepato-cytes, classic models for constitutive secretion. Using re-verse transcriptase polymerase chain reaction (RT-PCR)with primers specific for rat rab3D, we amplified a 151 basepair rab3D fragment from total RNA extracted from primarycultures of rat hepatocytes. Immunoblot analysis usingpolyclonal antibodies to peptides representing the N- andC-terminal hypervariable regions of murine rab3D recog-nized a protein of ;25 kd in hepatocyte lysates, hepaticsubcellular fractions, and tissue extracts. The distributionof rab3D was primarily cytosolic; however, only membrane-associated rab3D significantly bound guanosine triphos-phate (GTP) in overlay assays. Several lines of investigationindicate that rab3D is associated with the transcytotic path-way. First, rab3D was enriched in a crude vesicle carrierfraction (CVCF), which includes transcytotic carriers. Ve-sicular compartments immunoisolated from the CVCF onmagnetic beads coated with anti-rab3D antibody were en-riched in the transcytosed form of the polymeric IgA recep-tor (pIgA-R), but lacked not only the pIgA-R precursor formassociated with the secretory pathway, but also a Golgimarker protein. Second, indirect immunofluorescence onfrozen liver sections and in polarized cultured hepatocyteslocalized rab3D-positive sites at or near the apical plasmamembrane and to the pericanalicular cytoplasm. Finally,cholestasis induced by bile duct ligation (BDL), a manipu-lation known to slow transcytosis, caused rab3D to accumu-late in the pericanalicular cytoplasm of cholestatic hepato-cytes. Our results indicate that rab3D plays a role in theregulation of apically directed transcytosis in rat hepato-cytes. (HEPATOLOGY 2000;32:348-356.)

Rab proteins are members of the ras superfamily of lowmolecular weight guanosine triphosphate (GTP)-bindingproteins implicated in the regulation of vesicle-mediatedtransport in eukaryotic cells.1 More than 40 rab proteins havebeen identified. Based on localization and functional studies,it is thought that each rab protein controls a specific step inintracellular trafficking.2 To date, the precise mechanisms bywhich rab proteins regulate vesicle-mediated trafficking arepoorly defined. However, these proteins have been describedas “molecular switches,”3 because they cycle between activeGTP- and inactive GDP–bound states,4 during which trans-port vesicles form and are transported for fusion with appro-priate target membranes. Currently, it is hypothesized thatrab proteins regulate vectorial flow of vesicular transport bycontrolling the assembly and interactions of targeting pro-teins comprising the SNARE (SNAP receptor) complex.5 Ofthe many rab proteins identified to date, 4 highly homologousrab3 isotypes (rab3A, rab3B, rab3C, rab3D) are expressed inregulated secretory cells including neurons as well as neu-roendocrine, exocrine and endocrine, and various non-neu-ronal cell types. In these cells, rab3 isotypes are localized tosynaptic and secretory vesicles, where they presumably playimportant, yet unclear, roles in regulated secretory and/orexocytic events.6

Recent studies reveal that a rab3 isotype, rab3B, is ex-pressed in nonregulated secretory cells. This GTPase has beenlocalized near tight junctions in various polarized epithelialcells, including hepatocytes, and has been hypothesized toinfluence apical transport of proteins such as junctional com-ponents.7 Thus, it is possible that rab3 isotypes may havedistinct functions in regulated and constitutive secretorycells. Alternatively, all cells may regulate distinct secretorypathways to some extent and require the participation of rab3protein in this process. To address these issues, we investi-gated the expression and localization of an endogenous rab3isotype, rab3D, in a “constitutive” secretory epithelial cell, therat hepatocyte. Using molecular, biochemical, and morpho-logic techniques, we demonstrated that rab3D is expressed inrat hepatocytes, where it may play a role in the regulation ofapical (canalicular) membrane targeting and bile secretion.

MATERIALS AND METHODS

Animals and Cell Culture. Male Sprague-Dawley rats (150-200 g)were purchased from Harlan (Cincinnati, OH) and Zivic-Miller (Por-tersville, PA). Cholestasis was induced by ligation of the commonbile duct as described by Easter et al.8 All procedures using animalswere approved by the animal welfare committee in compliance withNational Institutes of Health guidelines. Primary cultures of rat hepa-

Abbreviations: GTP, guanosine triphosphate; ZO-1, zonula occludens 1; HRP, horse-radish peroxidase; RT-PCR, reverse transcriptase polymerase chain reaction; bp, basepair; SDS, sodium dodecyl sulfate; PAGE, polyacrylamide gel electrophoresis; CVCF,crude vesicle carrier fraction; PBS, phosphate-buffered saline; pIgA-R, polymeric immu-noglobulin A receptor; BDL, bile duct ligation.

From 1Barnard College, Department of Biological Sciences, New York, NY; and 2Cen-ter for Basic Research in Digestive Diseases, Mayo Foundation, Rochester, MN.

Received December 8, 1999; accepted May 11, 2000.Supported by the National Institutes of Health grants AA09227 and DK44650 awarded

to M. McNiven; and an AGA/Industry Scholar Award and National Institutes of Healthgrant DK52553 awarded to J. Larkin.

Address reprint requests to: Janet M. Larkin, Ph.D., Department of Biological Sciences,Barnard College/Columbia University, 3009 Broadway, New York, NY 10027. E-mail:jlarkin@barnard.columbia.edu; fax: 212-854-1950.

Copyright © 2000 by the American Association for the Study of Liver Diseases.0270-9139/00/3202-0024$3.00/0doi:10.1053/jhep.2000.9110

348

tocytes (.96% purity) were prepared according to Gores et al.9 Ly-sates of 3T3-L-1 adipocytes were a gift from Dr. Giulia Baldini(Whitehead Institute, Cambridge, MA).

Reagents and Antibodies. Polyclonal anti-rab3D antiserum gener-ated against a peptide representing the N-terminal hypervariable re-gion (-ASEPPASPRDAA-) of murine rab3D10 and preimmune serumwere kind gifts from Dr. Giulia Baldini (Whitehead Institute, Cam-bridge, MA). Another antiserum against this peptide was generatedby Advanced ChemTech (Louisville, KY) and gave identical results.These two antisera were designated anti–rab3D-N antibody.Polyclonal anti-rab3D antiserum against a peptide representing theC-terminal hypervariable region (-SSSPGSNGKGPALGDTPP-PQPSSCSC) of murine rab3D (designated anti–rab3D-C antibody)was generated by Babco (Richmond, CA). Purified rab1, rab2, rab4,and rab6, and their respective antisera, were obtained from Dr. BrunoGoud (Pasteur Institute, Paris, France). Polyclonal antibodies torab1 and rab2 were made against the C-terminal hypervariable re-gions of these GTPases.1 A recombinant rab3B fusion protein and anantibody specific for rab3B were obtained from Dr. Kevin Kirk (Uni-versity of Alabama, Birmingham, AL). Antibodies to rab4, rab5, andrab6 were purchased from Santa Cruz Biomedical (Santa Cruz, CA).Polyclonal antibodies to zonula occludens 1 (ZO-1) and to a man-nosidase II were obtained from Dr. James Anderson (Yale UniversitySchool of Medicine, New Haven, CT) and Dr. Marilyn Farquhar(University of California San Diego, La Jolla, CA), respectively, and amonoclonal antibody to ZO-1 was obtained from Dr. Ann Hubbard(Johns Hopkins University School of Medicine, Baltimore, MD).Rhodamine-conjugated goat anti-rabbit IgG (H & L) was purchasedfrom Tago, Inc. (Burlingame, CA). Alexa 488 and 568 secondaryantibody conjugates were from Molecular Probes (Eugene, OR), andhorseradish peroxidase (HRP)-conjugated goat anti-rabbit IgG andmonoclonal anti-rabbit IgG (g-chain specific) CLONE RG-96 werefrom Sigma (St. Louis, MO).

Reverse Transcriptase Polymerase Chain Reaction. Reverse transcrip-tase polymerase chain reaction (RT-PCR) was performed accordingto Oberhauser et al.11 Briefly, total RNA was extracted from short-term cultured hepatocytes.12 First-strand cDNA was made usingMoloney murine leukemia virus reverse transcriptase (Bethesda Re-search Laboratories, Gaithersburg, MD) and was amplified by 35cycles of PCR at 95°C for 15 seconds, 54°C for 25 seconds, and 72°Cfor 20 seconds using primers based on the published sequence of ratrab3D.11 Primer sequences, which incorporate a 151-bp region be-tween the first and second GTP-binding domains, were 59-AACT-GCTTTTGATCGGGAA-39 (upstream primer) and 59-GACAAGAG-GATCAAGCTGCA-39 (downstream primer). The resulting PCRproducts were separated by electrophoresis in agarose gels and de-tected with ethidium bromide, subcloned into a PCR-ready pCRIIplasmid (TA cloning kit, Invitrogen, San Diego, CA), and sequencedusing an automated nucleotide sequencer (Applied Biosystems 373,Foster City, CA). Control PCR reactions lacking first-strand cDNAor reverse transcriptase were appropriately negative.

Preparation of Subcellular Fractions and Tissue Extracts. Subcellularfractions were prepared from rat liver homogenate as previouslydescribed.13 To prepare tissue extracts, approximately 2 g each of ratliver and fat were homogenized on ice in buffer C (50 mmol/L Tris-HCl [pH 7.4]/0.5 mol/L NaCl/0.1% TX-100/0.1% sodium dodecylsulfate [SDS]) using a 7-mL Wheaton homogenizer and a glass pes-tle. The homogenates were microfuged for 10 minutes at 4°C toremove fat by flotation and particulates by sedimentation, and togenerate supernatants, which were assayed for rab3D by immuno-blot analysis.

Immunoprecipitation of Rab Proteins From Hepatic Cytosol and Mem-branes. Subcellular fractions were solubilized in 0.4% SDS (final con-centration), and the resulting lysates were processed through immu-noprecipitation as described previously.13 Antigen-antibodycomplexes were adsorbed onto protein A–Sepharose beads, whichwere processed for SDS–polyacrylamide gel electrophoresis (PAGE)as detailed below.

Immunoisolation of Subfractions on Antibody-Coated Magnetic Beads.Immunoisolation of rab3D-associated compartments was performedaccording to Saucan and Palade14 using magnetic Dynabeads M-500(Dynal, Lake Success, NY) coated with goat anti-rabbit IgG (Fc)(Biodesign Intl., Kennebunkport, ME), and then with affinity-puri-fied anti–rab3D-C antibody. The crude vesicle carrier fraction(CVCF), used as starting material for each immunoisolation, wasseparated into a bound subfraction (B) consisting of magnetic beadswith immunoabsorbed organelles and a nonbound subfraction (NB)using a magnet. CVCF as well as B and NB subfractions derivedtherefrom were analyzed biochemically by immunoblotting, and Bsubfractions were analyzed morphologically by transmission elec-tron microscopy.

Gel Electrophoresis. Cell and tissue samples were boiled for 2 min-utes in sample buffer (62.5 mmol/L Tris-HCl [pH 6.8]/2.7% SDS/100mmol/L b-mercaptoethanol/10% sucrose [final concentration]) be-fore being subjected to SDS-PAGE in 12% gels as described byMaizel15 at a constant current of 25 mA for ;2.5 hours. SDS-PAGEreagents were purchased from Bio-Rad (Richmond, CA).

Immunoblot Analysis. Proteins separated by SDS-PAGE were trans-ferred to nitrocellulose filters, which were then processed throughimmunoblot analysis in 10% dry milk according to Johnson et al.16

Enhanced chemiluminescence (Amersham, Arlington Heights, IL)or autoradiography at 280°C was used to detect HRP-conjugatedsecondary antibody or 125I–Protein A, respectively.

Detection of GTP-Binding Proteins in Hepatocytes Using a[32P]GTP Over-lay. Immunoprecipitated rab proteins resolved by SDS-PAGE weretransferred to nitrocellulose filters, which were processed for bind-ing of a[32P]GTP (.3,000 Ci/mmol from ICN, Irvine, CA) accordingto Lapetina and Reep.17 GTP-binding proteins were visualized byautoradiography at 280°C.

Immunofluorescence Analysis of Liver Sections and Cultured Hepatocytes.For single-label studies, semithin sections of frozen rat liver andcultures of polarized hepatocytes were processed for immunofluo-rescence as previously described.13 For double-label immunofluo-rescence studies, rat liver was fixed in situ with paraformaldehyde/lysine/periodate.18 Frozen 0.5-mm sections were cut, mounted onpolylysine-coated glass slides, and stored in phosphate-buffered sa-line (PBS) supplemented with sodium azide at 4°C before beingprocessed for double-label imuunofluorescence as follows. Sectionswere quenched in 50 mmol/L NH4Cl/PBS, blocked in 1% bovineserum albumin/PBS, and incubated for 2 hours with polyclonal anti-rab3D and monoclonal anti–ZO-1 antibodies diluted in 1% bovineserum albumin/PBS containing 0.25% Triton X-100. After beingwashed in PBS/0.25% Triton X-100, sections were incubated withanti-rabbit and anti-mouse secondary antibodies conjugated to Alexa488 and 568 dyes diluted in 1% bovine serum albumin/PBS contain-ing 0.25% Triton X-100, washed, and mounted in 50 mmol/L Tris/150 mmol/L NaCl/2 mg/mL phenylene diamine mounting medium.

Liver sections and cells were viewed using a 1003 objective on a ZeissAxiovert 35 epifluorescence microscope (Carl Zeiss, Goettingen, Ger-many) or a 603objective on an Olympus BHS-RFC microscope equippedwith a BH2-RFC reflected light fluorescence attachment (Micro-Optics,Fresh Meadows, NY). Images were photographed on Tmax p3200(Eastman Kodak Company, Rochester, NY) with a 35-mm camera.

Electron Microscopy. Transmission electron microscopy was per-formed to visualize the intracellular compartments immunoisolatedon antibody-coated magnetic beads. Bound (B) subfraction werefixed in 3% paraformaldehyde, 1% glutaraldehyde in 0.1 mol/L Na-cacodylate buffer (pH 7.2), stained en bloc in osmium tetroxide anduranyl acetate, dehydrated in ethanol and propylene oxide, and em-bedded in Epon. Prepared thin sections were viewed on a PhilipsS208 transmission electron microscope.

RESULTS

Rab3D mRNA and Protein Are Expressed in Rat Hepatocytes. Todate, rab3D has been identified only in cells that undergoregulated secretion.10,11,19-21 To determine if this GTPase

HEPATOLOGY Vol. 32, No. 2, 2000 LARKIN ET AL. 349

plays a role in “constitutive” secretory cells, we investigatedthe expression of rab3D mRNA and protein in rat hepatocytes.Using RT-PCR with primers specific for rat rab3D, we ampli-fied a 151-bp fragment (Fig. 1A) from total RNA extractedfrom primary cultures of rat hepatocytes. This 151-bp PCRproduct, which corresponds to amino acids 24-74 of therab3D protein,22 was 100% identical to the correspondingregion of rat rab3D11 and different from 3A and 3B (Fig. 1B) asdetermined by nucleotide sequence analysis.

We next investigated the expression of rab3D protein in rathepatocytes using antibodies made against the C- and N-ter-minal hypervariable regions of rab3D. As shown in Fig. 2A,antibodies to rab3D-C (lane 1) and rab3D-N (lane 2) detecteda ;25-kd protein in rat liver, the signals for which were com-peted by preincubating antibodies with their respective im-munizing peptides (lanes 3 and 4, respectively). We also in-vestigated the expression of rab3D protein in rat hepatocytesusing an antibody made against the N-terminal hypervariableregion of rab3D that had been characterized extensively byothers in previous studies on rab3D in adipocytes.10,22 Asshown in the immunoblot in Fig. 2B, this antibody recognizeda ;25-kd protein in a hepatocyte cell lysate (lane 2) and fatand liver tissues (lanes 3 and 4, respectively). The ;25-kdhepatic protein was judged to be rab3D by its immunoreac-tivity with the anti-rab3D antibody and by its comigration inthe gel with rab3D in an adipocyte cell lysate (Fig. 2C). Theanti–rab3D-N antibody also detected the ;25-kd protein incytosol prepared from rat liver, but failed to recognize recom-binant rab1, rab2, rab4, rab6, and a rab3B fusion protein (Fig.2D), thus demonstrating its specificity. A duplicate lane con-taining the rab3B fusion protein gave a positive result whenblotted with anti-rab3B antibody (Fig. 2D) as did anti-rab4antibody when the filter containing recombinant rab4 wasreprobed (data not shown).

Distribution of Rab3D in Subcellular Fractions Prepared From RatLiver. We next investigated the subcellular localization ofrab3D using biochemical techniques. Rat liver was homoge-nized and fractionated to generate a high-speed supernatant(cytosol [C]) and total microsomal pellet (membrane [M]),which were examined by immunoblot analysis (Fig. 3).Rab3D was preferentially enriched in the cytosol (C) with a

less prominent immunoreactive protein band present in thetotal membrane fraction (M). This distribution contrastedwith those of rab1, rab2, and rab4 (Fig. 3), as well as rab5 andrab6 (data not shown), which were preferentially membrane-associated. Moreover, rab3B was also preferentially localizedto membranes, in contrast to rab3D.

To identify the membrane-bound compartment with whichrab3D associates, hepatic subcellular fractions were examinedby immunoblot analysis (Fig. 4A-4C) and a[32P]GTP overlay(Fig. 4D). The fractions analyzed were the homogenate (H),cytosol (Cyt), total microsomes (TM), Golgi light (GL), andGolgi heavy (GH) fractions, a fraction enriched in rough en-doplasmic reticulum (ER), and the CVCF, known to be en-riched in transcytotic carriers.23,24 As partial validation of thecell fractionation protocol, the distributions of a mannosidaseII, a Golgi resident protein, and the polymeric IgA receptor(pIgA-R) were determined by immunoblot analysis (Fig. 4Aand 4B, respectively). Both of these membrane proteins wereappropriately undetectable in the cytosol (Cyt). As shown inFig. 4A, a mannosidase II was enriched in GL and GH frac-tions when compared with the TM from which these Golgifractions were prepared and was slightly detectable in theCVCF. As shown in Fig. 4B, the pIgA-R was highly enriched inthe TM over the low level detected in the H. Consistent withits trafficking along the biosynthetic and transcytotic path-ways, the pIgA-R was enriched in the GH and CVCF whencompared with the TM from which these fractions were pre-pared. Low levels of receptor were detected in the GL and ER.A duplicate set of cell fractions was then assayed for theirrab3D content as shown in Fig. 4C. Rab3D was detected in Cytand TM, consistent with its cycling on and off of membranes,was enriched in the GH and CVCF, and was barely detectablein the GL and ER fractions. Densitometric analysis of the blotin Fig. 4C showed that rab3D in GH and CVCF was enriched3- and 4.5-fold, respectively, when compared with the TMfrom which these fractions were prepared. Finally, the enrich-ment of rab3D in these fractions was confirmed by immuno-precipitation, followed by a[32P]GTP overlay. As shown inFig. 4D, anti-rab3D antiserum immunoprecipitated a GTP-binding protein with an apparent molecular weight of ;25kd. Consistent with the immunoblot in Fig. 4C, rab3D was

FIG. 1. Rab3D mRNA expressionin rat hepatocytes. (A) Detection ofrab3D PCR product from mRNA (1mg) isolated from rat hepatocytes.Using rab3D-specific primers, cDNAwas amplified by RT-PCR. The re-sulting PCR product was resolved byagarose gel electrophoresis and de-tected by ethidium bromide stain. Aspredicted by the choice of primers, a151-bp fragment (arrow) was gener-ated. (B) Comparison of the nucleo-tide sequence of the 151-bp PCRproduct with those of various rab3isotypes. Boxes designate nucleo-tides that differ among compared se-quences.

350 LARKIN ET AL. HEPATOLOGY August 2000

markedly enriched in GH and CVCF. However, in sharp con-trast to immunoblot analyses reported in Figs. 3 and 4B, cy-tosolic rab3D was less well detected than membrane-associ-ated rab3D when assayed by GTP overlay.

Immunoisolation of Rab3D-Associated Compartments From theCVCF. Our interest in hepatic transcytosis focused our stud-ies on rab3D associated with the CVCF, a membrane prepa-ration proven to be enriched in transcytotic compart-ments.23,24 To confirm that the rab3D detected in the CVCFwas membrane-bound and not adventitiously trapped cytoso-

lic GTPase, we performed immunoisolation of rab3D-associ-ated compartments using magnetic beads coated with anti-rab3D antibody. We reasoned that only those membrane-bound compartments in the CVCF associated with rab3D ontheir cytosolic aspect would bind to antibody-coated beads.After incubation with antibody-coated beads, the CVCF wasresolved into nonbound (NB) and bound (B) subfractions. Asshown in Fig. 5A, transmission electron microscopy revealedthat the immunoabsorbed rab3D-associated compartmentsconsisted predominantly of membrane-bound spheres ;80

FIG. 2. Immunoblot analysis of rab3D protein in hepatocytes. (A) Approximately 50-mg protein samples of a postnuclear supernatant prepared from ratliver were resolved by SDS-PAGE, transferred to nitrocellulose, and immunoblotted with anti–rab3D-C and anti–rab3D-N antibodies (1:100) in the absence(lanes 1 and 2) and presence (lanes 3 and 4) of their respective immunizing peptides (10 mg/mL). Incubation with primary antibodies was followed by asecondary antibody conjugated to HRP (1:5,000). Rab3D was detected by chemiluminescence. Both antibodies recognized a protein of ;25 kd. (B) Approx-imately 50-mg protein samples of cultured hepatocytes (lane 2) and detergent extracts of epididymal fat pads (lane 3) and rat liver (lane 4) were resolved bySDS-PAGE, transferred to nitrocellulose, and immunoblotted with anti–rab3D-N antibody (1:100). Rab3D was detected by chemiluminescence. Anti–rab3D-Nantibody recognized a major protein of ;25 kd (closed arrow). A faint higher molecular weight signal is nonspecific, because it is detected in lane 1 containingprestained molecular-weight markers and does not appear in blots performed with previously used antibody (lane 4). (C) Approximately 50-mg proteinsamples of cultured adipocytes, here used as a positive control, and hepatocytes were resolved by SDS-PAGE, transferred to nitrocellulose, and immunoblottedwith anti–rab3D-N antibody (1:200 dilution), followed by 125I–Protein A. Rab3D was detected by autoradiography (closed arrow). (D) To test the specificityof the anti–rab3D-N antibody used in (B and C), 10 mg of hepatic cytosol (the source of native rab3D, closed arrows) and recombinant rab1, rab2, rab4, andrab6 were processed through immunoblot analysis, and rab3D was detected by chemiluminescence. Duplicate samples of a rab3B fusion protein (40 mg/lane;open arrow) were similarly processed for immunoblot analysis with antibody either to rab3D-N (1:100) or rab3B (1:100).

HEPATOLOGY Vol. 32, No. 2, 2000 LARKIN ET AL. 351

nm in diameter, with no obvious coats or distinguishing con-tent. As shown in Fig. 5B, control magnetic beads not coatedwith anti-rab3D antibody had no immunoabsorbed mem-brane-bound compartments, demonstrating the specificity ofthe immunoisolation.

Duplicate aliquots of CVCF, as well as NB and B subfrac-tions generated therefrom, were partially characterized by im-munoblot analysis. The distribution and recovery of proteins

in the CVCF were followed in the NB and B subfractions bydensitometry, and the data were normalized to CVCF. Asshown in Fig. 5C, 80% of CVCF rab3D was recovered exclu-sively in the B subfraction, which lacked detectable levels of amannosidase II, consistent with rab3D-associated compart-ments being significantly depleted of secretory organelles. Totest whether the mature form of the pIgA-R, a known transcy-totic itinerant,23,24 traffics through rab3D compartments, NBand B subfractions were analyzed for their pIgA-R content.Although the 116-kd and 120-kd pIgA-R forms were not in-dividually resolved, all pIgA-R forms were detected in theCVCF and NB subfraction. Importantly, the B subfractioncontained only mature pIgA-R forms and no 105-kd precur-sor, consistent with rab3D-associated compartments beingsignificantly depleted of secretory organelles and enriched intranscytotic compartments.

Immunofluorescence Localization of Rab3D in Rat Hepatocytes.We next investigated the cellular localization of rab3D onfrozen liver sections by double-label indirect immunofluores-cence. As shown in Fig. 6, rab3D-positive sites were detectedat or near the apical plasma membrane surrounding bilecanaliculi as viewed in cross-section (arrowheads) or longitu-dinal section (arrows). The staining pattern for ZO-1, a majorprotein of tight junctions,25 was apical but did not exactlycolocalize with that for rab3D, suggesting that this GTPase isnot restricted to junctional complexes.

To show that rab3D detected in liver sections was not theresult of bile duct epithelial cells, hepatocytes were culturedunder conditions that promoted the formation of polarizedcouplets surrounding a bile canaliculus before being pro-cessed through immunofluorescence localization of variousantigens (Fig. 7). Specific rab3D staining was diffusely scat-tered throughout the cytoplasm, localized to punctate struc-tures in the pericanalicular cytoplasm, and competed by pre-incubating the antiserum with immunizing peptide (1peptide). This staining pattern for rab3D was distinct from

FIG. 3. Distribution of various rab proteins in cytosolic and membranefractions from rat hepatocytes. Approximately 40-mg protein samples of cy-tosolic and total membrane fractions prepared from whole liver were resolvedby SDS-PAGE, transferred to nitrocellulose, and immunoblotted with poly-clonal antibodies (1:100) to the indicated rab proteins. Incubation with pri-mary antibodies was followed by a secondary antibody conjugated to HRP.Rab proteins were detected by chemiluminescence. Rab3D as detected withanti–rab3D-C antibody was preferentially cytosolic, whereas the other rabproteins were primarily membrane-associated. Anti–rab3D-N antibody gaveresults identical to those obtained with anti–rab3D-C antibody (data notshown).

FIG. 4. Detection of rab3D in hepatic cellular fractions by immunoblot analysis and GTP overlay. (A-C) Approximately 40-mg protein samples of hepatichomogenate (H), cytosol (Cyt), and total microsomes (TM) and ;20-mg protein samples of Golgi light (GL), Golgi heavy (GH), CVCF, and endoplasmicreticulum (ER) prepared from rat liver were resolved by SDS-PAGE, transferred to nitrocellulose, and immunoblotted with antibodies to a mannosidase II(1:200), pIgA-R (1:100), and rab3D-N (1:100). Incubation with primary antibodies was followed by a secondary antibody conjugated to HRP. Proteins weredetected by chemiluminescence. Rab3D was enriched in the Golgi heavy fraction (GH) and the CVCF. (D) Approximately 50-mg protein samples of hepaticcellular fractions were immunoprecipitated using anti–rab3D-N antibody (1:50). Immunoprecipitated proteins were resolved by SDS-PAGE, transferred tonitrocellulose, processed through a[32P]GTP overlay, and detected by autoradiography. A GTP-binding protein of ;25 kd was enriched in the Golgi heavyfraction (GH) and the CVCF, but fell below the level of detection in the total microsomes (TM) and cytosol (Cyt).

352 LARKIN ET AL. HEPATOLOGY August 2000

those for ZO-1 and rab3B, as well as a mannosidase II andrab2, a GTPase implicated in ER to Golgi trafficking.

The apical localization of rab3D reported in Figs. 6 and 7supports our hypothesis that this GTPase may be associatedwith the transcytotic pathway in rat hepatocytes. To furtherexplore this possibility, we induced cholestasis in rats via bileduct ligation (BDL), a condition known to slow transcytosis ofpIgA-R and other apical plasma membrane proteins such thatthey accumulate in a tubulovesicular compartment locatedbeneath the apical plasma membrane.23,26 Thus, if rab3D isassociated with the transcytotic pathway in hepatocytes, BDLmight be expected to affect its localization. We tested thisprediction using indirect immunofluorescence of sham-oper-ated and BDL livers as shown in Fig. 8. In sham-operatedcontrols, rab3D-positive sites were localized at or near apicalmembranes seen in cross-sectional (arrowheads) and longitu-dinal (arrows) views. In addition, a fine, punctate fluores-cence signal was detected in the pericanalicular cytoplasm. Inresponse to BDL, rab3D-positive sites accumulated in the

pericanalicular cytoplasm around dilated bile canaliculi (ar-rowheads) or along distorted apical plasma membranes. Thepunctate cytoplasmic signal was less than in controls. As ex-pected, pIgA-R–positive sites in the pericanalicular cytoplasmat or near bile canaliculi in BDL hepatocytes were more nu-merous than in sham-operated controls. Taken together,these data suggest that rab3D is localized to the transcytoticpathway.

DISCUSSION

In this study, we used molecular, biochemical, and mor-phological technologies to demonstrate that rab3D mRNA(Fig. 1) and protein (Figs. 2-8) are expressed in a constitutivesecretory cell, the rat hepatocyte. The majority of hepaticrab3D is cytosolic rather than membrane-associated (Fig. 3),in contrast to another rab3 isoform, rab3B. Membrane-boundrab3D is associated with both a subcellular fraction enrichedin transcytotic carriers (Figs. 4 and 5) and the apical domainsof polarized hepatocytes (Figs. 6-8). Thus, rab3D may be in-volved in the regulation of apical trafficking in hepatocytes, aprocess that has consequences with respect to bile secretionand to maintaining structural and functional polarity of theplasma membrane.

Rab3D Is Detected in Rat Hepatocytes. Toward understandingits function in hepatocytes, we determined the intracellularlocalization of rab3D using biochemical techniques. Unlikeother hepatic rab proteins, including rab3B, which were pref-erentially membrane-associated, the majority of hepaticrab3D was cytosolic as determined by immunoblot analysis(Figs. 2 and 3). This result is consistent with rab3B and rab3Cin non-neuronal cells.27,28 The functional significance of thiscellular localization currently is unknown but may relate tothe phosphorylation state of the GTPases, because phosphor-ylated rab3B has been reported to be cytosolic rather thanmembrane-associated.27

In contrast to results obtained by immunoblot analysis,rab3D was poorly detected in cytosol but highly detectable ina crude vesicle carrier fraction when assayed by a[32P]GTPoverlay (Fig. 4D). One possible explanation for this result isthat cytosolic and membrane-associated rab3D possess dis-tinct posttranslational modifications (i.e., phosphorylation),which may not only influence their localization, but also theirability to bind GTP. Thus, the intensity of the signal obtainedin the crude vesicle carrier fraction by GTP overlay may be areflection of the physiologic state of rab3D as well as an en-richment of this GTPase in the fraction.

Rab3D Is Associated With the Apically-Directed TranscytoticPathway in Rat Hepatocytes. Immunofluorescence studies onfrozen liver sections (Fig. 6) and cultured cells (Fig. 7) revealthat rab3D is localized at the apical poles of hepatocytes,though not identically colocalized with ZO-1. These data sug-gest that rab3D-positive sites represent either transport vesi-cles en route to the apical plasmalemma or apical compart-ments through which transcytosed proteins traffic. Ourmorphologic findings are supported by cell fractionation (Fig.4) and immunoisolation data (Fig. 5) that indicate that rab3D-associated membrane-bound compartments immunoab-sorbed from the CVCF are significantly depleted of a Golgimarker and enriched in mature, transcytosed forms of thepIgA-R. Furthermore, detection of rab3D in the Golgi heavyfraction (GH; Fig. 4) is consistent with a report showing thattranscytotic carriers are enriched not only in the CVCF, butalso in the Golgi heavy fraction.29 Finally, BDL, a condition of

FIG. 5. Immunoisolation of rab3D-associated membrane-bound com-partments. The CVCF was prepared from rat liver and used as starting mate-rial for immunoisolation with magnetic beads coated with either linker anti-body coupled to affinity-purified anti–rab3D-C antibody or linker antibodyalone. After incubation with magnetic beads, the CVCF was resolved intononbound (NB) and bound (B) subfractions. (A, B) Aliquots of bound (B)subfractions were processed for analysis by transmission electron micros-copy. As shown in (A), beads coated with anti-rab3D antibody immunoad-sorbed spherical compartments of ;80 nm in diameter. As shown in (B),beads coated with linker antibody alone had no immunoadsorbed membrane-bound compartments. (C) Equivalent volumes of NB and B subfractionsnormalized to CVCF were subjected to SDS-PAGE to separate constituentproteins, which were transferred to nitrocellulose. Duplicate nitrocellulosemembranes were probed with antibodies to rab3D (1:100), a mannosidase II(1:100), or pIgA-R (1:100), followed by secondary antibody conjugated toHRP (1:5,000). Proteins were detected by chemiluminescence. Most of therab3D was recovered in the B subfraction, which was depleted of a manno-sidase II and significantly enriched in mature pIgA-R forms. Bars 5 0.1 mm.

HEPATOLOGY Vol. 32, No. 2, 2000 LARKIN ET AL. 353

reduced bile flow accompanied by an intracellular accumula-tion of transcytosed proteins, altered the localization ofrab3D. Current studies aimed at detailed biochemical charac-terization of immunoisolated rab3D-associated compart-ments from control and BDL rats will further clarify the role ofrab3D in transcytosis.

Detection of rab3D at apical poles of hepatocytes is consis-tent with the localization of other rab3 isotypes at apical andapical-equivalent domains of polarized cells. For instance,rab3A is localized to synaptic vesicles in presynaptic terminalsof fully polarized neurons and to axonal processes of cellslacking presynaptic specializations.30 These localizations areconsistent with the hypothesis that neuronal axons are equiv-alent to apical domains in epithelial cells.31 Furthermore,

FIG. 7. Indirect immunofluorescence localization of various rab proteinsin cultured rat hepatocytes. Rat hepatocytes were cultured under conditionsthat promoted a polarized phenotype with basolateral and apical plasmamembrane domains before being permeabilized, fixed, and incubated withpolyclonal antibodies to ZO-1 (1:250), a mannosidase II (1:100), rab3B (1:100), rab2 (1:200), rab3D-N (1:100), and rab3D-N (1:100 in the presence ofspecific rab3D-N immunizing peptide [10 mg/mL]). Primary antibodies weredetected with secondary antibody conjugated to rhodamine (1:100). ZO-1and rab3B defined the apical plasma membrane, whereas a mannosidase IIand rab2 defined the pericanalicular cytoplasm. Rab3D-positive sites werelocalized to punctate structures in the pericanalicular cytoplasm near theapical plasma membrane and diffusely scattered throughout the cytoplasm.Nucleus. Bars 5 10 mm.

4™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™FIG. 6. Indirect immunofluorescence localization of rab3D on frozen sec-

tions of rat liver. Frozen 0.5-mm sections were processed for double-labelimuunofluorescence as follows. Sections were incubated with polyclonalanti-rab3D (1:500) and monoclonal anti–ZO-1 (1:200) antibodies before beingincubated with anti-rabbit and anti-mouse secondary antibodies conjugatedto Alexa 488 and 568 dyes (1:400), respectively. Rab3D- and ZO-1–positivesites were detected near the apical plasma membrane surrounding bile cana-liculi visualized in cross and longitudinal sections (arrowheads and arrows,respectively), although their staining patterns were not identical. *Sinusoid.Bars 5 10 mm.

354 LARKIN ET AL. HEPATOLOGY August 2000

rab3D is associated with apically positioned zymogen gran-ules in rat pancreatic acinar cells.21 Similarly, rab3B is local-ized to apical domains of various epithelial cells, includinghepatocytes.7 Presumably, each of these rab3 isotypes partic-ipates in one of the final steps of secretion.

In addition to rab3 isotypes, rab11 is associated with apicaldomains of various epithelial cells. For example, rab11 trans-locates with the H1-K1-ATPase from a subplasmalemmal tu-bulovesicular compartment to the apical plasma membranedomain during gastric parietal cell stimulation.32 Further-more, hepatic rab11 is associated with vesicles in the peri-canalicular cytoplasm of hepatocytes.33 Additional studies areneeded to determine whether these various apical rab proteinsserve redundant or distinct functions in regulating vesicle-mediated transport to the canalicular pole of the hepatocyte.

Do Hepatocytes Have a Regulated Secretory Pathway? Transportpathways that terminate at apical poles of hepatocytes havetraditionally been viewed as being constitutive. The identifi-cation of rab3B7 and rab3D (current study) in hepatocytesgenerates the testable hypothesis that apical transport andsecretion in these cells may be regulated via signal transduc-tion processes. Several independent reports support this pos-sibility. First, glucagon stimulates bile secretion from guineapig liver apparently by influencing the transcytotic pathway inhepatocytes.34 Second, cyclic adenosine monophosphate aswell as a choleretic bile acid stimulate transcytosis and secre-tion of horseradish peroxidase into bile.35,36 Third, bile acid–induced choleresis has been correlated with an increase in thesurface area of the canalicular membrane in rat hepatocytes, aresult consistent with fusion of transport vesicles with thisapical plasmalemmal domain.37 Finally, hepatocytes, likeother transporting epithelial cells, may regulate ion transportvia vesicle-mediated insertion of various transporters into theapical plasma membrane.38,39 Future studies are needed todefine the function of hepatic rab3D and to explore the exis-tence of a regulated secretory pathway in the hepatocyte.

Acknowledgment: The authors thank Drs. G. Baldini (Co-lumbia University) for anti-rab3D antibody and adipocytes; B.Goud (Pasteur Institute) for purified rab1, -2, -4, -6, and an-tibodies to these proteins; J. Anderson (Yale University Schoolof Medicine) for polyclonal anti–ZO-1 antibody; M. Farquhar(UCSD) for anti–a mannosidase II antibody; and K. Kirk(University of Alabama at Birmingham) for rab3B fusion pro-tein and anti-rab3B antibody. Special thanks go to Dr. AnnHubbard (Johns Hopkins University School of Medicine) forproviding paraformaldehyde-fixed liver tissue and monoclo-nal anti-ZO-1 antibody, as well as critically reading the manu-script. S. Bronk and S. Kuntz provided excellent technicalassistance in culturing isolated hepatocytes and cutting someof the 2-mm-thick frozen liver sections, respectively. Dr. Lar-kin thanks the Center for Basic Research in Digestive Diseasesat the Mayo Clinic for funding and supporting the initialstages of this project, the Department of Biological Sciences atBarnard College for continued support, and the Departmentof Chemistry at Barnard College for their support of BW andher biochemistry honors thesis.

REFERENCES

1. Zahraoui A, Touchat N, Chardin P, Tavitian A. The human Rab genesencode a family of GTP-binding proteins related to yeast YPT1 and SEC4products involved in secretion. J Biol Chem 1989;264:12394-12401.

2. Olkkonen VM, Stenmark H. Role of Rab GTPases in membrane traffic.Intl Rev Cyt 1997:176:1-85.

3. Bourne HR, Sanders DA, McCormick F. The GTPase superfamily: con-served structure and molecular mechanisms. Nature 1991;349:117-127.

4. Boguski MS, McCormick F. Proteins regulating Ras and its relatives.Nature 1993;366:643-654.

5. Schimmoller F, Simon I, Pfeffer SR. Rab GTPases, directors of vesicledocking. J Biol Chem 1998;273:22161-22164.

6. Fischer von Mollard G, Stahl B, Li C, Sudhof TC, Jahn R. Rab proteins inregulated exocytosis. TIBS 1994;19:164-168.

7. Weber E, Berta G, Tousson A, St John P, Green MW, Gopalokrishnan U,Jilling T, et al. Expression and polarized targeting of a rab3 isoform inepithelial cells. J Cell Biol 1994;125:583-594.

FIG. 8. Indirect immunofluores-cence localization of rab3D andpIgA-R on frozen sections of liverfrom rats subjected to either a shamoperation or BDL. Two-micrometer-thick frozen sections of sham-oper-ated or ligated liver were acetone-fixed, allowed to air-dry, andincubated with antibodies either torab3D-N (1:100) or the pIgA-R (1:100), followed by a secondary anti-body conjugated to rhodamine(1:100). The number of rab3D- andpIgA-R–positive sites in the peri-canalicular cytoplasm at or near theapical plasma membrane viewed inlongitudinal and cross sections (ar-rows and arrowheads, respectively)was increased in BDL hepatocyteswhen compared with sham-operatedcontrols. Bars 5 10 mm.

HEPATOLOGY Vol. 32, No. 2, 2000 LARKIN ET AL. 355

8. Easter DW, Wade JB, Boyer JL. Structural integrity of hepatocyte tightjunctions. J Cell Biol 1983;96:745-749.

9. Gores GJ, Neiminen A-L, Wray BE, Herman B, Lemasters JJ. IntracellularpH during “chemical hypoxia” in cultured rat hepatocytes. Protection byintracellular acidosis against the onset of cell death. J Clin Invest 1989;83:386-396.

10. Baldini G, Scherer PE, Lodish HF. Nonneuronal expression of Rab3A:induction during adipogenesis and association with different intracellu-lar membranes than Rab3D. Proc Natl Acad Sci U S A 1995;92:4284-4288.

11. Oberhauser AF, Balan V, Fernandez-Badilla CL, Fernandez JM. RT-PCRcloning of rab3 isoforms expressed in peritoneal mast cells. FEBS Lett1994;339:171-174.

12. Chomczynski P, Sacci N. Single-step method of RNA isolation by acidguanidinium thiocyanate-phenol-chloroform extraction. Anal Biochem1987;162:156-159.

13. Larkin JM, Oswald B, McNiven MA. Ethanol-induced retention of nas-cent proteins in rat hepatocytes is accompanied by altered distribution ofthe small GTP-binding protein rab2. J Clin Invest 1996;98:2146-2157.

14. Saucan L, Palade GE. Membrane and secretory proteins are transportedfrom the Golgi complex to the sinusoidal plasmalemma of hepatocytes bydistinct vesicular carriers. J Cell Biol 1994;125:733-741.

15. Maizel JV. Polyacrylamide gel electrophoresis of viral proteins. In: Mar-amorosch K, Koprowski H, eds. Methods in Virology. Volume 5. NewYork: Academic Press, 1971:179-246.

16. Johnson DA, Gautsch JW, Sportsman JR, Elder JH. Improved techniqueutilizing nonfat dry milk for analysis of proteins and nucleic acids trans-ferred to nitrocellulose. Gene Anal Techn 1984;1:3-8.

17. Lapetina EG, Reep BR. Specific binding of [a32P] GTP to cytosolic andmembrane-bound proteins of human platelets correlates with the activa-tion of phospholipase C. Proc Natl Acad Sci U S A 1987;84:2261-2265.

18. Bartles JR, Hubbard AL. Preservation of hepatocyte plasma membranedomains during cell division in situ in regenerating rat liver. Dev Biol1986:118;286-295.

19. Ohnishi H, Ernst SA, Wys N, McNiven M, Williams JA. Rab3D localizesto zymogen gransules in rat pancreatic acini and other exocrine glands.Am J Physiol 1996;271:G531-G538.

20. Weber E, Jilling T, Kirk KL. Distinct functional properties of Rab3A andRab3B in PC12 neuroendocrine cells. J Biol Chem 1996:271:6963-6971.

21. Valentijn JA, Sengupta D, Gumkowski FD, Tang LH, Konieczko EM,Jamieson JD. Rab3D localizes to secretory granules in rat pancreaticacinar cells. Eur J Cell Biol 1996;70:33-41.

22. Baldini G, Hohl T, Lin HY, Lodish HF. Cloning of a rab3 isotype predom-inately expressed in adipocytes. Proc Natl Acad Sci U S A 1992;89:5049-5052.

23. Larkin JM, Palade GE. Transcytotic vesicular carriers for polymeric IgAreceptors accumulate in rat hepatocytes after bile duct ligation. J Cell Sci1991;98:205-216.

24. Sztul E, Kaplin A, Saucan L, Palade G. Protein traffic between distinctplasma membrane domains: isolation and characterization of vesicularcarriers involved in transcytosis. Cell 1991;64:81-89.

25. Anderson JM, Glade JL, Stevenson BR, Boyer JL, Mooseker MS. Hepaticimmunohistochemical localization of the tight junction protein ZO-1 inrat models of cholestasis. Am J Pathol 1989;134:1055-1062.

26. Barr VA, Hubbard AL. Newly synthesized hepatocyte plasma membraneproteins are transported in transcytotic vesicles in the bile duct-ligatedrat. Gastroenterology 1993;105:554-571.

27. Karniguian A, Zahraoui A, Tavitian A. Identification of small GTP-bind-ing rab proteins in human platelets: thrombin-induced phosphorylationof rab3B, rab6, and rab8 proteins. Proc Natl Acad Sci U S A 1993;90:7647-7651.

28. Cormont M, Tanti J-F, Zahraoui A, Obberghen V, Tavitian A, LeMarch-and-Brustel Y. Insulin and okadaic acid induce Rab4 redistribution inadipocytes. J Biol Chem 1993;268:19491-19497.

29. Sztul ES, Howell KE, Palade GE. Biogenesis of the polymeric IgA receptorin rat hepatocytes. II. Localization of its intracellular forms by cell frac-tionation studies. J Cell Biol 1985;100:1255-1261.

30. Matteoli M, Takei K, Cameron R, Hurlbut P, Johnston PA, Sudhof TC,Jahn R, et al. Association of rab3A with synaptic vesicles at late stages ofthe secretory pathway. J Cell Biol 1991;115:625-633.

31. Ikonen E, Parton RG, Hunziker W, Simons K, Dotti CG. Transcytosis ofthe polymeric immunoglobulin receptor in cultured hippocampal neu-rons. Curr Biol 1993;3:635-644.

32. Goldenring JR, Soroka CJ, Shen KR, Tang LH, Rodriguez W, VaughanHD, Stoch SA, et al. Enrichment of rab11 a small GTP-binding protein ingastric parietal cells. Am J Physiol 1994;267:G187-G194.

33. Goldenring JR, Smith J, Vaughan HD, Cameron P, Hawkins W, NavarreJ. Rab11 is an apically located small GTP-binding protein in epithelialtissues. Am J Physiol 1996;270:G515-G525.

34. Lenzen R, Hruby VJ, Tavoloni N. Mechanism of glucagon-inducedcholeresis in guinea pigs. Am J Physiol 1990;259:G736-G744.

35. Hayakawa T, Bruick R, Ng OC, Boyer JL. DBcAMP stimulates vesicletransport and HRP excretion in isolated perfused rat liver. Am J Physiol1990;259:G727-G735.

36. Hayakawa T, Ng OC, A M, Boyer JL. Taurocholate stimulates transcytoticvesicular pathways labelled by horseradish peroxidase in the isolatedperfused rat liver. Gastroenterology 1990;99:216-228.

37. Gautam A, Ng OC, Strazzabosco M, Boyer JL. Quantitative assessment ofcanalicular bile formation in isolated hepatocyte couplets using micro-scopic optical planimetry. J Clin Invest 1989;83:565-573.

38. Benedetti A, Strazzabosco M, Ng OC, Boyer JL. Regulation of activity andapical targeting of the Cl2/HCO3

2 exchanger in rat hepatocytes. ProcNatl Acad Sci U S A 1994:91:792-796.

39. Bradbury NA, Bridges RJ. Role of membrane trafficking in plasma mem-brane solute transport. Am J Physiol 1994;267:C1-C24.

356 LARKIN ET AL. HEPATOLOGY August 2000