Identification of HAX-1 as a protein that binds bile salt export protein and regulates its abundance in the apical membrane of Madin-Darby canine kidney cells.

ATP-binding cassette (ABC)-type proteins are essential for bile formation in vertebrate liver. BSEP, MDR1, MDR2, and MRP2 ABC transporters are targeted to the apical (canalicular) membrane of hepatocytes where they execute ATP-dependent transport of bile acids, drugs, amphipathic cations, phospholipids, and conjugated organic anions, respectively. Changes in activity and abundance of transporters in the canalicular membrane regulate bile flow; however, little is known regarding cellular proteins that bind ABC transporters and regulate their trafficking. A yeast two-hybrid screen identified HAX-1 as a binding partner for BSEP, MDR1, and MDR2. The interactions were validated biochemically by glutathione S-transferase pull-down and co-immunoprecipitation assays. BSEP and HAX-1 were over-represented in rat liver subcellular fractions enriched for canalicular membrane vesicles, microsomes, and clathrin-coated vesicles. HAX-1 was bound to BSEP, MDR1, and MDR2 in canalicular membrane vesicles and co-localized with BSEP and MDR1 in the apical membrane of Madin-Darby canine kidney (MDCK) cells. RNA interference of HAX-1 increased BSEP levels in the apical membrane of MDCK cells by 71%. Pulse-chase studies indicated that HAX-1 depletion did not affect BSEP translation, post-translational modification, delivery to the plasma membrane, or half-life. HAX-1 depletion resulted in an increased peak of metabolically labeled apical membrane BSEP at 4 h and enhanced retention at 6 and 9 h. HAX-1 also interacts with cortactin. Expression of dominant negative cortactin increased steady state levels of BSEP 2-fold in the apical membrane of MDCK cells, as did expression of dominant negative EPS15. These findings suggest that HAX-1 and cortactin participate in BSEP internalization from the apical membrane.

ABC 1 proteins in the canalicular membrane of hepatocytes generate the driving force that establishes bile flow. The ABC transporters MDR1 (ABCB1), MDR2 (ABCB4) (MDR3 in humans), and BSEP/SPGP (ABCB11) (bile salt export protein, also called sister of P-glycoprotein), which are closely related in structure and sequence, mediate ATP-dependent transport of biliary constituents. BSEP transports conjugated bile acids (1), and MDR2 is essential for phospholipid transfer into bile (2,3). MDR1 mediates excretion of hydrophobic, cationic drugs to bile (4), and MDR1 overexpression is an important cause of multidrug resistance in tumor cells (5). MRP2 (ABCC3), which has a different structure and belongs to a different ABC transporter subfamily, mediates secretion of organic anions (6,7). Mutations that affect expression or activity of these proteins are associated with liver disease. Defects in BSEP cause progressive familial intrahepatic cholestasis type II (8); mutations in MDR3 are manifested by progressive familial intrahepatic cholestasis type III (9), and MRP2 defects give rise to Dubin-Johnson syndrome (6). Disease can also result from defective trafficking of transporters to the canalicular membrane. Mutations that impair MDR3 exit from the endoplasmic reticulum are associated with cholestasis of pregnancy (10); mutations in BSEP that impair expression of the mutant proteins in the apical membrane of MDCK cells are linked to progressive familial intrahepatic cholestasis type II (11); and deletion of two amino acids in MRP2 of a Dubin-Johnson patient resulted in intracellular retention and accelerated degradation of MRP2 (12). Thus, elucidating pathways that govern trafficking of ABC transporters to and from the canalicular membrane can provide critical insight into mechanisms underlying normal biliary secretion and cholestasis (bile secretory failure).
Targeting and trafficking of integral membrane proteins depend on sequence motifs in the proteins, interactions with signaling and trafficking networks, and polypeptides that bind the membrane proteins. Few proteins have been identified that associate with ABC transporters and participate in their mobilization and targeting. Interactions with PDZ domain proteins NHERF/EBP50 and E3Karp (13,14) are essential for apical expression of cystic fibrosis transmembrane regulator (15). MRP2 retention in the apical membrane requires interaction with radixin (16) and a carboxyl-terminal moiety that binds PDZK1 (17,18). MDR1, MDR2, and BSEP do not contain obvious PDZ-interacting motifs, and apical expression of MDR proteins is not affected in the liver of radixin knockout mice (16). Thus, trafficking of MDR transporters is controlled by a different subset of proteins. By using a yeast two-hybrid screen, we identified HAX-1 as a binding partner for MDR1, MDR2, and BSEP. HAX-1 is a 34-kDa polypeptide that interacts with a heterogeneous group of proteins that include cortactin (19 -23). We present evidence indicating that HAX-1 associates with BSEP, MDR1, and MDR2 in the canalicular membrane of rat hepatocytes and that HAX-1 and cortactin regulate BSEP abundance in the apical membrane of MDCK cells.

EXPERIMENTAL PROCEDURES
Materials and Antibodies-The Cytotrap yeast two-hybrid system and vectors were purchased from Stratagene (La Jolla, CA), and pEGFP-N1 and pEYFP-N2 vectors were from Clontech (Palo Alto, CA). The Silencer Express RNAi kit was from Ambion (Austin, TX), and Easytag NEG-772 [ 35 S]methionine and -cysteine protein labeling mix was from PerkinElmer Life Sciences. The anti-cCAM105 Ab669 antibody was a gift from S. H. Lin (24). Anti-HAX-1 and anti-calnexin antibodies were purchased from BD Biosciences; C219 antibody was from Signet (Dedham, MA); anti-mannosidase II antibody was from Covance Research (Richmond, CA); anti-clathrin heavy chain TD.1 monoclonal antibody was from Sigma; anti-Na/K-ATPase IgG was from United States Biochemical Corp.; anti-cortactin 4F11 monoclonal antibody was from Upstate Biotechnology, Inc. (Lake Placid, NY); anti-GFP antibody was from Abraxis (Warminster, PA); fluorescently labeled secondary antibodies were from Jackson ImmunoResearch (West Grove, PA); and isotype-specific goat anti-mouse IgG1 and -mouse IgG2a were from Bethyl Laboratories (Montgomery, TX). LVT90-Tu41 anti-BSEP IgG was prepared as described (25). Sulfo-NHS-LC-LCbiotin and streptavidin-agarose beads were purchased from Pierce. All other reagents were from Sigma.
Yeast Two-hybrid Screen-A PCR DNA fragment coding for amino acids 620 -698 of rat MDR2 was cloned into the pSOS vector (Stratagene) to generate bait plasmid pSL49. Other pSOS bait plasmids prepared in a similar manner are described below. A liver cDNA library was generated in the pMyr vector from rat liver poly(A) ϩ RNA using the Superscript system (Invitrogen). Library plasmid DNA was co-transformed into temperature-sensitive cdc25h yeast with pSOS bait plasmids. Yeasts grown for 24 -48 h at 25°C were replica-plated on galactose plates and incubated at the nonpermissive temperature of 37°C for 3-5 days. Individual colonies that grew at 37°C were picked and expanded in liquid media at 25°C. Washed cells were spotted on galactose or glucose solid media and incubated at 37°C.
Plasmid Constructs-The plasmids used in these studies are listed in Table I. Yeast two-hybrid bait plasmids in the pSOS vector (Stratagene) and GST protein expression plasmids in pGEX5-3x vector (Amersham Biosciences) were constructed by ligation with PCR fragments that encode the designated BSEP, MDR2, or MDR1a regions. Rat Mdr2 and Mdr1a PCR fragments were derived from plasmids provided by Dr. Jeffrey Silverman. Bsep fragments were derived from K4 plasmid (gift of Dr. Peter Meier). MDR1a-EGFP was described previously (26), and BSEP-EYFP was made by in-frame ligation of rat Bsep with enhanced yellow fluorescent protein in pEYFP-N1. 2 GFP fusion plasmids containing EPS15 were a gift of A. Benmerah. DnEPS15 (EPS15 ⌬95-295) is described in Ref. 27, and EPS15 (DIII⌬2) is described in Ref. 28.
GST Pull Down-GSH-agarose beads (Amersham Biosciences) were used to purify GST, GST-MDR2, and GST-BSEP fusion proteins from BL-21 Escherichia coli cells harboring plasmids pGEX5-3x, pSL25, and pSL27, respectively (see Table I). 5 mg of protein from a rat liver homogenate 2000 ϫ g pellet were solubilized for 30 min in Dulbecco's PBS containing 1% Triton X-100 and protease inhibitors (see the composition of SH buffer). Cleared 13,000 ϫ g supernatants were incubated for 2-3 h with 10 g of GST protein bound to GSH-Sepharose beads. The beads were washed four times with PBS-Triton, and bound proteins were eluted with SDS loading buffer for immunoblot analyses.
Culture, Transfection, and Immunofluorescence Microscopy of Mammalian Cells-MDCK II cells (gift of Dr. Enrique Rodriguez-Boulan) were grown in transwells (Corning Glass) and transfected using Lipo-fectAMINE 2000 (Invitrogen). Transwells were washed with PBS and fixed for 10 min in ice-cold methanol. Transwell filters were blocked for 1 h in 3% bovine serum albumin and 3% normal donkey serum in Dulbecco's PBS and then incubated for 1-2 h with anti-HAX-1 antibody diluted 1/25 in IF buffer (3% bovine serum albumin in Dulbecco's PBS). Membranes were washed 10 times in PBS, blocked for 30 min with IF buffer, and incubated with Texas Red-labeled donkey anti-mouse antibody for 45 min. Filters were washed 10 times with PBS and mounted on glass slides. BSEP-EYFP, MDR1-EGFP, and Texas Red-stained HAX-1 were visualized using a Leica TRS2 laser confocal microscope.
Preparation of Membrane Fractions, Immunoblot Analysis, and Immunoprecipitation-Subcellular fractions were prepared from rat liver as indicated. Enrichment of marker proteins was determined by immunoblot analyses, except for lysosomes and mitochondria, which were measured by biochemical enzyme assays. Briefly, rat livers were perfused with SH buffer (250 mM sucrose, 50 mM HEPES-Tris, pH 7.4, 0.1 mM CaCl 2 , 1 g/ml pepstatin, 1 g/ml leupeptin, 0.5 mM phenylmethylsulfonyl fluoride, 5 g/ml benzamidine, 0.1 l/ml aprotinin) and homogenized in 50 ml of ice-cold SH buffer, and 2000 ϫ g supernatants and pellets were obtained by centrifugation. Canalicular membrane vesicles (CMV) were prepared from the 2000 ϫ g pellet as described (29). Immunoblot analyses indicated CMV were enriched 40 -50-fold in MDR1 and MDR2 (C219 antibody), BSEP (LVT90), and cCAM105 (Ab669). The 2000 ϫ g supernatant was used to prepare sinusoidal membrane vesicles (30) that were enriched 20 -25-fold in Na/K-ATPase, endoplasmic reticulum microsomes (31) that were enriched 15-20-fold in calnexin, and Golgi membrane vesicles (32) that were enriched 20 -25-fold in mannosidase II. Clathrin-coated vesicles prepared from whole liver homogenates by density centrifugation (33) were enriched 40 -45-fold in clathrin heavy chain. Rat liver lysosomes (34) and mitochondria (35) (gift from Dr. Ana Maria Cuervo) were enriched 25-30fold in ␤-hexosaminidase (36) and 30 -40-fold in succinate dehydrogenase, respectively. Immunoprecipitations were performed as described (25). C219 immunoprecipitates prepared in this manner predominantly contained 2 Y. Wakabayashi, personal communication. overexpression in E. coli, and expression in mammalian cells Vectors include the following: pGEX5x-3 (Amersham Biosciences) for expression of GST fusion proteins in E. coli; pSOS (Stratagene) for expression of bait proteins in S. cerevisiae; pMYR (Stratagene) for expression of prey cDNAs in S. cerevisiae; pBK-CMV (Stratagene) for expression of cDNAs in mammalian cells; pEYFP-N1 and pEGFP-C1 (Clontech) for expression of proteins as EYFP or EGFP, respectively, in mammalian cells; pQE60 for expression of His 6 -tagged proteins in E. coli; and pSec for expression of RNAi duplexes driven by the human U6 promoter. AA refers to the amino acid residues of the designated gene product that are encoded by the DNA fragment in each plasmid; NA indicates not applicable. MDR1 and MDR2 P-glycoproteins. A small amount of BSEP was also present but was less than 10% that obtained in equivalent immunoprecipitations with LVT90 (our observations and see Ref. 25).
Biotinylation of Cell Surface Proteins-Biotinylation was performed essentially as described (37). MDCK cells grown in 12-mm transwell inserts (polyester 0.4-m pore) were transfected using LipofectAMINE 2000, and media were changed every 1-2 days. 2-3 days after confluency, transwells were washed once with ice-cold DMEM without serum and twice with cold Dulbecco's PBS containing calcium and magnesium. Cells were incubated twice for 25 min with 0.3 ml of freshly made biotinylation solution (1.5 mg/ml sulfo-NHS-LC-LC-biotin, 10 mM ethanolamine, 2 mM CaCl 2 , 250 mM NaCl, pH 7.5) in either the upper or lower chamber. Transwells were washed twice with Dulbecco's PBS containing 100 mM glycine, incubated in a third wash for 20 min, and washed twice with Dulbecco's PBS. Transwell filters were incubated for 1 h in lysis buffer (1% Triton X-100, 150 mM NaCl, 5 mM EDTA, 50 mM Tris-HCl, pH 7.5, 1/1000 (v/v) aprotinin, 2 M leupeptin, 4 M pepstatin A, 1 mM phenylmethylsulfonyl fluoride), and cells were scraped off. Lysates were centrifuged at 13,000 ϫ g for 10 min, and 50 l of packed streptavidin-agarose beads were added to the supernatants, which were incubated overnight with rotation. The beads were washed three times with lysis buffer, twice with high salt buffer (lysis buffer with 500 mM NaCl), and once with 10 mM Tris, pH 7.5. Proteins were eluted from the beads with 40 l of SDS-urea loading buffer (4% SDS, 250 mM Tris-HCl, pH 6.8, 10 mM dithiothreitol, 8 M urea), separated by SDS-PAGE, and immunoblotted.
RNA Interference-A cDNA containing the canine Hax-1 coding region was generated from RNA-isolated MDCK cells by RT-PCR. The forward and reverse primers used were GAATGAGCGTCTTT-GATCTTTTCCG and GTCCAGGATGGAAAAGGGATCATC, respectively. The siRNA target finder algorithm in the Ambion web site (www.ambion.com) was used to select 19-nucleotide oligomers in canine Hax-1 to be tested for RNA interference. A control 19-nucleotide sequence was generated that had the same G-C content as the selected oligomers but did not display sequence identity with Hax-1. BLAST analysis ensured that sequence identity between oligonucleotides and mammalian cDNAs in EST data bases was 15 nucleotides or less. The sequences chosen are as follows: 365, GCTTAAGTATCCAGATAGT; 427, GAAGTGAATCCTCCAAACC; 623, GATCACTAAGCCAGATGGG; 756, GCCCTGGATGATGCCTATT; and con, GTGTACAGCGATGTT-GTCG. Double-stranded DNA fragments, which contained the selected RNAi sequences positioned downstream of the human U6 RNA polymerase III promoter, were generated by PCR using the Ambion Silencer Express kit and according to the manufacturer's instructions. The DNA fragments were cloned into the pSec vector (Ambion) and sequenced. MDCK II cells in 24-mm transwells were co-transfected with pEGFP and pSec derivatives (p365-756 and pCon). Two days later, cells were detached by trypsinization and passed 10 times through a 25-gauge needle. EGFP fluorescent cells and nontransfected cells were sorted, counted, and collected using a MOFLO (Dako-Cytomation) fluorescentactivated cell sorter. Equal numbers of cells were pelleted by centrifugation at 5000 ϫ g and lysed in SDS loading buffer. HAX-1 expression levels were determined by immunoblots using the BD Biosciences anti-HAX-1 monoclonal antibody.
Pulse-Chase Labeling-MDCK cells cultured in 24-mm transwell filters were transfected as described above. Two days after transfection, cells were washed twice and incubated for 30 min with methionine-and cysteine-free DMEM containing 5% dialyzed fetal bovine serum. Transwells were placed on 0.15 ml of pulse media consisting of DMEM containing 1 mCi/ml [ 35 S]cysteine/methionine (Amersham Biosciences Easytag NEG772) for 30 min. Cells were washed twice and transferred to wells containing chase media (DMEM, 10% fetal bovine serum and cysteine and methionine at 180 mg/liter each). At specified time points, transwells were washed twice with ice-cold DMEM, and apical surface proteins were biotinylated as described above. Following PBS/glycine washes, membranes were excised from transwells, and cells were lysed by shaking for 1 h in 1.5 ml of 2ϫ PBS, 1% Triton X-100, 40 mM octyl glucoside. Lysates were diluted to 10 ml with 2ϫ PBS, 0.5% Triton X-100, and cell debris was removed by centrifugation at 10,000 ϫ g for 20 min. Cell extracts were incubated for 1 h with 50 l of anti-rabbit IgG-Sepharose beads (Sigma). Beads were removed by centrifugation, and BSEP-EYFP was immunoprecipitated from pre-cleared lysates using an anti-GFP polyclonal antibody (Abraxis) and anti-rabbit IgG-Sepharose beads. Beads were washed three times with 2ϫ PBS, 0.5% Triton X-100. Immunoprecipitated BSEP-EYFP was released from anti-GFP antibody by boiling for 10 min in 75 l of 1.5% SDS and diluted to 1.5 ml with lysis buffer. Biotinylated BSEP-EYFP was extracted by overnight incubation with streptavidin-agarose beads, which were har-vested by centrifugation at 1000 ϫ g and washed three times with lysis buffer. Proteins were eluted with 40 l of SDS-urea loading buffer and separated by SDS-PAGE. Gels were fixed in 40% methanol, 10% acetic acid and exposed to phosphor screens for 2-7 days.
Real Time RT-PCR-Replica 12-mm transwells containing MDCK cells were transfected on the same day and in the same manner as those used for biotinylation. Cells were washed twice with ice-cold PBS, and RNA was isolated using 250 l of TRIzol reagent according the manufacturer's instructions. RNA from each sample was treated with RNasefree DNase (Invitrogen) for 20 min at 37°C. DNase was inactivated by a 20-min incubation with EDTA at 65°C. First strand cDNA was generated in a 20-l reaction containing 0.5 g of DNA-free RNA and using oligo(dT) primers, Superscript II reverse transcriptase, and RNase H. Real time PCR was done by using 1 l of the reverse transcriptase (RT) reaction and Hot Start SYBR green system (Qiagen) according to the manufacturer's instructions. Amplification of PCR products was monitored in real time for 40 cycles by using Stratagene Mx4000 multiplex quantitative PCR instrument. Bsep primers were sense CGATGACCTCAGAAATAATCCTGG and antisense GCCG-CAATGATGTTAGTGAAGG. Canine ␤-actin was used as a control with sense primer CGATGACCTCAGAAATAATCCTGG and antisense primer ATCCACGAGACCACCTTCAACTCC.  38). A yeast two-hybrid screen, which used the rat MDR2 linker domain as bait, resulted in isolation of full-length Hax-1 cDNAs on three separate occasions. The rat Hax-1 nucleotide sequence exhibited 93 and 85% identity with mouse (GenBank TM accession number NM_011826) and human (GenBank TM accession number NM_006118) Hax-1 genes, respectively. The deduced rat HAX-1 protein (rHAX-1) sequence was 94 and 79% identical to mouse and human HAX-1 amino acid sequences, respectively.
HAX-1 Binds Amino Acid Motifs Proximal to the Nucleotide Binding Domains-Two-hybrid analyses of pSL47 and pSL49 deletion derivatives revealed that rHAX-1 interacted in yeast with portions of the MDR2 and BSEP linkers that are proximal to the amino-terminal nucleotide binding domains (Fig. 1, C  and D). The HAX-1 binding region was circumscribed to 20 amino acids in BSEP and to 16 residues in MDR2. These are the only regions of the BSEP and MDR2 linker domains that share significant amino acid sequence identity (see Fig. 1E). Moreover, these linker sequence fragments, particularly the MDR2 linker, exhibit significant similarity with the HAX-1 binding region of the herpesvirus membrane protein K15P (23) (Fig. 1E).
GST-BSEP and GST-MDR2 Fusion Proteins Bind HAX-1 in Rat Liver Homogenates-Rat liver homogenates were incubated with GSH-agarose beads adsorbed with chimeric GST proteins containing 91 amino acids of the BSEP linker domain or 79 amino acids of the MDR2 linker. GST-BSEP and GST-MDR2 beads effectively extracted HAX-1 from liver homogenates, whereas GST alone did not (Fig. 2). This experiment indicates that BSEP and MDR2 linker domains efficiently bind HAX-1 in the presence of the myriad proteins of a rat liver extract.
Distribution of HAX-1 in Rat Liver Subcellular Fractions-HAX-1 was primarily associated with 100,000 ϫ g precipitable material (named 100 P) from rat liver homogenates. HAX-1 association with the 100 P fraction was unaffected by high pH or high salt but was dissociated by 1 M urea or 1% Triton (Fig.  3A), indicating strong affinity of HAX-1 for cellular membranes. rHAX-1 was enriched in CMV, clathrin-coated vesicles, and endoplasmic reticulum-enriched microsomal fractions (Fig.   3B). Only trace amounts of rHAX-1 were present in Golgi, lysosomes, mitochondria, or basolateral membrane vesicles. BSEP, MDR1, and MDR2 were also enriched in CMV and clathrin-coated vesicle fractions.
To determine whether HAX-1 is associated with the transporters in the canalicular membrane, BSEP was immunoprecipitated from CMV using antibody LVT90. Immunoblot analysis indicated that, in addition to BSEP, the immunoprecipitates contained HAX-1 (Fig. 4C). LVT90 did not immunoprecipitate HAX-1 from liver fractions that contained HAX-1 but were devoid of BSEP. Likewise, HAX-1 accompanied MDR1 and MDR2 immunoprecipitated from CMV by antibody C219 (Fig. 4D). As with LVT90, HAX-1 was not detected in C219 immunoprecipitates of liver fractions that did not contain MDR1 or MDR2. Thus, HAX-1 is present in protein complexes that contain BSEP, MDR1, or MDR2 in the canalicular membrane of hepatocytes.
HAX-1 Co-localizes with BSEP-EYFP in the Apical Domain of Polarized MDCK II Cells-MDCK II cultures were co-transfected with constructs expressing rHAX-1 and BSEP-EYFP. The cells were fixed, immunostained, and visualized by laser confocal fluorescence microscopy. Subcellular co-localization of BSEP-EYFP and HAX-1 varied with cell polarization. In nonpolarized MDCK II cells, HAX-1 and BSEP-EYFP were predominantly intracellular, and co-localization was sparse (data not shown). One to 2 days after confluency, cells showed differentiation of apical and lateral membranes. HAX-1 partially co-localized with BSEP-EYFP in the apical membrane; however, most HAX-1 was located immediately below the apical domain, where it partially localized with BSEP-EYFP (Fig. 5,  upper panel). In polarized columnar cells, 4 -5 days after confluency, BSEP-EYFP was mostly restricted to the apical membrane where it co-localized with HAX-1 (Fig. 5, lower panel).
Reduction of HAX-1 Expression by RNA Interference Increases Apical Membrane BSEP in MDCK Cells-RNAi plasmids were generated that express short stem-loop RNAs homologous to 19-bp sequences in the canine Hax-1 mRNA. A matched control that displayed no homology with Hax-1 was also made. MDCK cells co-transfected with pEGFP and RNAi plasmids were purified by fluorescence-activated cell sorting. Transfection efficiency of MDCK cultures varied from 30 to 60%. However, cells transfected with two plasmids displayed greater than 85% co-transfection efficiency (as determined by fluorescence microscopy of cells co-transfected with pEGFP and pECFP plasmids, not shown). Therefore, 85% of cells sorted by EGFP expression were presumed to contain RNAi plasmids. Immunoblots revealed that HAX-1 levels were reduced by 7-75% in EGFP-positive cells expressing different RNAi transcripts (Fig. 6A). Nonfluorescent cells, or cells co-transfected with the control plasmid, displayed HAX-1 levels indistinguishable from nontransfected controls. Transfection with plasmid p365 consistently reduced HAX-1 expression by greater than 70% relative to the pCon control plasmid.
Apical or basolateral surface proteins of MDCK cells transfected with Bsep-EYFP were labeled with a membrane-impermeable biotinylation reagent and purified by binding to streptavidin-agarose. Immunoblots of biotinylated proteins indicated that BSEP-EYFP was predominantly localized to the FIG. 3. HAX-1 distribution in rat liver fractions. A, HAX-1 is associated with membrane fractions. A rat liver 2000 ϫ g pellet enriched in HAX-1 was homogenized in 2% Triton X-100 (2% Trit), 1% Triton X-100 (1% Trit), 1 M urea, Na 2 CO 3 buffer, pH 11, or 0.5 M NaCl. Con, control. Immunoblots of 100,000 ϫ g supernatant (S) or pellets (P) of these homogenates indicated that HAX-1 was removed from precipitable material by detergent or chaotropic agents. B, immunoblot analysis of purified liver subcellular fractions. Equal amounts of protein (15 g) from each subcellular fraction were immunoblotted and stained for HAX-1 (HAX-1 monoclonal antibody), MDR1 and MDR2 (C219 monoclonal antibody), or BSEP (LVT90 antisera). Subcellular fractions tested are as follows: SMV, sinusoidal membrane vesicles; 2 k, the 2000 ϫ g pellet that contains crude canalicular membranes; gol, Golgi vesicles; mic, ER microsomal fractions; mit, mitochondria; lys, lysosomes; CCV, clathrin-coated vesicles. . Immunoblots of LVT90 immunoprecipitates indicated that HAX-1 co-immunoprecipitated with BSEP. HEK-293T cells express human HAX-1, which has a slightly smaller M r than the rat ortholog. Thus, two bands were seen in immunoprecipitates from cells expressing human and rat HAX-1. Nonimmune serum does not immunoprecipitate BSEP or HAX-1, and HAX-1 is not immunoprecipitated by LVT90 in the absence of BSEP. B, MDR1 and HAX-1 are co-immunoprecipitated by the anti-MDR1 C219 antibody. Extracts were prepared from HEK293 cells transfected with the pMDR1a-EGFP plasmid (ϩ) or the pEGFP-N1 empty vector (Ϫ). Immunoblots of proteins immunoprecipitated (I.P.) by the C219 monoclonal antibody (C219) or normal mouse IgG 2a (N.I.) revealed that HAX-1 specifically co-immunoprecipitates with MDR1a-EGFP. C, immunoblot of proteins immunoprecipitated (Immunopcpt.) from CMV by the anti-BSEP LVT90 antibody. LVT 90 (LVT) or normal rabbit (nl) IgG were incubated with 0.4 mg of rat liver CMV (cmv) or 1 mg of a 2000 ϫ g supernatant (2 k S), which contains HAX-1 but does not contain detectable BSEP or MDR2. Immunoblot analysis revealed that LVT90 immunoprecipitated BSEP from CMV but not from the 2000 ϫ g supernatant fraction (top panel). HAX-1 was present in LVT90 immunoprecipitates derived from CMV, but not from 2000 ϫ g supernatant (bottom panel), indicating that LVT90 precipitates HAX-1 only when BSEP is present. 20 g of CMV and 50 g of 2000 ϫ g supernatant protein were loaded in the gel to show relative levels of BSEP and HAX-1. LVT90 was visualized using protein A-horseradish peroxidase conjugates, and the HAX-1 IgG 1 monoclonal antibody was visualized with goat anti-mouse IgG. D, immunoblots of C219 immunoprecipitates. MDR1 and MDR2 were immunoprecipitated from CMV by the C219 IgG 2a . Immunoblot analysis of immunoprecipitates indicated that HAX-1 co-immunoprecipitated with MDR1 and MDR2 from CMV. Lanes containing normal mouse IgG 2a or C129 immunoprecipitates from the 2000 ϫ g supernatant fraction did not contain HAX-1, indicating that HAX-1 is immunoprecipitated by C219 only in the presence of MDR1 and MDR2. C219 IgG 2a was visualized using protein-A-horseradish peroxidase conjugates and HAX-1 IgG 1 was visualized using isotype-specific goat anti-mouse IgG 1 conjugated to horseradish peroxidase. apical membrane (Fig. 7A). Cells co-transfected with Bsep-EYFP and p365 RNAi plasmid contained 71 Ϯ 7% more BSEP-EYFP in the apical membrane relative to controls transfected with pCon (Fig. 7, B and C). HAX-1 depletion by p365 did not induce mistargeting of BSEP-EYFP to the basolateral membrane (not shown). RNA analyses by real time RT-PCR indicated that Bsep transcript levels were not significantly different in cells that contained p365 or pCon (Fig. 7D).

HAX-1 Associates with Cortactin and Expression of Dominant Negative Cortactin Increases BSEP Levels in the Apical
Membrane-HAX-1 and cortactin co-localize in fibroblast lamellipodia and interact in vitro (19). Analysis of subcellular fractions from rat liver revealed that cortactin is highly enriched in CMV (Fig. 9A). Immunoprecipitation of HAX-1 from CMV also precipitated cortactin (Fig. 9B) indicating that HAX-1 interacts with cortactin in the canalicular membrane.
To determine whether cortactin participates in transporter trafficking, MDCK cells were transfected with plasmids that express BSEP and a dominant negative form of cortactin, which lacks the carboxyl-terminal SH3 domain (39). Cell surface biotinylation indicated that co-expression of dominant negative cortactin was associated with a greater than 2-fold increase in apical membrane BSEP levels (Fig. 10, A and B). Real time RT-PCR of RNA derived from replica transwells revealed no significant differences in Bsep transcript levels in cells expressing GFP or GFP fused to dominant negative cortactin (Fig. 10C).
Expression of Dominant Negative EPS15 Increases Apical Membrane BSEP-Cortactin has been implicated in clathrin-

FIG. 5. HAX-1 and BSEP-EYFP colocalize in the apical domain of MDCK II cells. MDCK II cells were
transfected with plasmids expressing BSEP-EYFP and rHAX-1. Cells were fixed in methanol, and HAX-1 was visualized with the monoclonal antibody and goat anti-mouse IgG-Texas Red. Localization of HAX-1 at the apical membrane and co-localization with BSEP-EYFP increased with time in culture and polarization of MDCK II cells. The upper panels show cells 2 days after transfection, whereas cells in the lower panels were fixed 5 days after transfection. XZ sections of stained cells were obtained with a Leica TRS2 confocal microscope. Sequential scanning with argon (514 nm) and krypton (568 nm) lasers was used to avoid bleed through of BSEP-EYFP fluorescence into the Texas Red detection channel.

FIG. 6. Depletion of HAX-1 expression in MDCK cells by RNA interference.
Immunoblots of extracts from MDCK cells co-transfected with pEGFP and four pSec plasmid derivatives that express small interfering RNAs homologous to canine Hax-1. Cells were detached from transwell filters, and EGFP-positive cells (ϩEGFP) were sorted from nonfluorescent cells (ϪEGFP) by fluorescent-activated cell sorter. Con, control. Protein extracts from equal numbers of cells were separated by SDS-PAGE and blotted with anti-HAX-1 or anti-Na ϩ /K ϩ -ATPase antibodies. Quantification of the immunoblots indicated that HAX-1 expression in ϩEGFP-transfected cells was significantly reduced by p427 (47 Ϯ 5%) and p365 (76 Ϯ 7%) (n ϭ 3, ** indicates p Ͻ 0.01; Student's t test). No differences in HAX-1 expression were observed in extracts from nontransfected ϪEGFP cells mediated receptor endocytosis (40,41). To determine whether clathrin participates in BSEP internalization from the apical membrane, MDCK cells were co-transfected with a plasmid expressing dominant negative EPS15. EPS15 specifically interacts with epsin (42) and the AP-2 adaptor (43). The dominant negative EPS15-EH21 mutant, which lacks the amino-terminal EH (EPS15 homology) domains, inhibits clathrin-coated pit membrane proteins were isolated with streptavidin-agarose and immunoblotted with the anti-BSEP LVT90 antibody. B, immunoblots of apically biotinylated proteins (apical) and total extracts (total) derived from cells co-transfected with Bsep-EYFP and with the p365 Hax-1 RNAi plasmid, or the pCon control plasmid. HAX-1 abundance was reduced by 50% in p365 total extracts, which contain proteins from transfected and nontransfected cells, suggesting that HAX-1 expression in transfected cells may be reduced by greater than 70% (see Fig. 6). HAX-1 depletion was accompanied by a 26 Ϯ 5% (*, p Ͻ 0.05) increase in total BSEP levels and by a 71 Ϯ 12% (**, p Ͻ 0.01) increase in apical membrane levels of BSEP. C, (mean of three experiments in triplicate, and one experiment in quadruplicate, Ϯ S.E., n ϭ 13). D, real time reverse transcription PCR analysis of Bsep RNA in transfected cells. Total RNA was isolated from MDCK cells transfected with plasmids expressing BSEP-EYFP, p365, and pCon. Residual genomic and plasmid DNA was removed by treatment with RNase-free DNase. Real time RT-PCR analysis using two pairs of Bsep-specific primers revealed that Bsep RNA levels were not significantly different in cells containing p365 or pCon (two experiments in triplicate, n ϭ 6). Negative controls included RNA isolated from cells that did not express Bsep and RNA which had been treated with DNase but had not undergone reverse transcription (R.T.). In all cases, real time PCR results were normalized to expression of endogenous canine ␤-actin RNA. formation at the plasma membrane and receptor endocytosis (27). Expression of EH-21 EPS15 in MDCK cells resulted in a 94 Ϯ 17% increase in apical membrane BSEP (Fig. 11), which indicates that clathrin participates in BSEP internalization from the apical membrane. DISCUSSION Trafficking of transporters in hepatocytes is rigorously controlled. Newly synthesized MDR1 and MDR2 traffic directly from the trans-Golgi network to the canalicular membrane (25,26). BSEP travels from Golgi to intracellular compartment(s) from which it cycles to and from the apical domain (25). Recruitment and removal of canalicular membrane MDR trans-porters responds to cAMP, taurocholate (25,44,45), and changes in osmolarity (46). The molecular mechanisms that control trafficking of MDR-type proteins in hepatocytes are unknown. HAX-1 is the first binding partner to be described for the canalicular ABC proteins MDR1, MDR2, and BSEP.
Association of HAX-1 with MDR1, MDR2, and BSEP was initially detected in a yeast two-hybrid assay. The interactions in yeast were corroborated by three lines of evidence. GST fusion proteins containing MDR2 or BSEP linker domains extracted HAX-1 from rat liver homogenates. HAX-1 co-localized with BSEP and MDR1 in mammalian cells in culture and co-purified with the transporters in liver subcellular fractions. Finally, HAX-1 was specifically co-immunoprecipitated with BSEP, MDR1, and MDR2 from cell extracts and liver subcel- A, MDCK II cells were co-transfected with plasmids expressing BSEP-EYFP or a fusion protein (GFP-⌬cort) consisting of GFP and dominant negative cortactin. Immunoblots of apically biotinylated proteins indicated apical membrane BSEP (apical) was 210 Ϯ 13% (**, p Ͻ 0.01) higher in cells expressing GFP-⌬cort compared with controls (B) (three experiments in triplicate n ϭ 9). Expression of BSEP in the whole cell extracts (total) was increased by 28 Ϯ 14% (total). C, real time RT-PCR analysis indicated that Bsep RNA levels were not significantly different in cells expressing GFP or GFP-⌬cort (two experiments in triplicate, n ϭ 6). Negative controls included RNA isolated from cells that did not express Bsep and RNA that had been treated with DNase but not had not undergone reverse transcription (R.T.). As before, real time PCR results were normalized to expression of the endogenous canine ␤-actin RNA.
FIG. 11. Dominant negative EPS15 increases BSEP abundance in the apical membrane of MDCK cells. A, MDCK cells were cotransfected with plasmids expressing BSEP-EYFP and fusion proteins containing GFP and dominant negative EPS15 (dnEPS15), GFP, and a noninhibitory EPS15 derivative (EPS15), or GFP alone (GFP). Immunoblots of biotinylated proteins revealed that cells expressing dnEPS15 contained 94 Ϯ 17% (**, p Ͻ 0.01) more BSEP in the apical membrane relative to controls expressing GFP alone (n ϭ 9, three experiments in triplicate) (B). Cells expressing EPS15 exhibited 21 Ϯ 12% less apical membrane BSEP, which was not significantly different to controls. C, real time RT-PCR analysis of Bsep RNA. Total RNA isolated from MDCK cells expressing BSEP-EYFP and GFP, GFP-dnEPS15, or GFP-EPS15 was analyzed as described in Fig. 7. No significant differences in Bsep RNA levels were detected between cells expressing GFP, GFP-dnEPS15, or GFP-EPS15 (two experiments in triplicate, n ϭ 6). Negative controls included RNA isolated from cells that did not express Bsep and RNA that had been treated with DNase but not had not undergone reverse transcription (R.T.). Real time PCR results were normalized to expression of the endogenous canine ␤-actin RNA. lular fractions, which confirms that HAX-1 binds MDR proteins in mammalian cells.
HAX-1 specifically interacted with the linker domains of MDR1, MDR2, and BSEP and did not bind other intracellular loops. The linker domains were previously thought to represent unstructured connecting loops that span the homologous halves of MDR proteins; however, these regions also have regulatory functions. Linker domain ubiquination of Saccharomyces cerevisiae MDR-type transporter, STE6p, regulates its endocytosis and intracellular trafficking (47). Phosphorylation of the MDR1a linker mediates its interaction with a volume-gated chloride channel (48) and may regulate drug transport activity (49). Additionally, the MDR1b linker has been proposed to function as a dimerization domain (50). Interaction of HAX-1 with the linker domains of MDR-type transporters may also have regulatory function.
The Hax-1 gene codes for a 34-kDa protein that is ubiquitously expressed in mammalian tissues but appears to have no orthologs or paralogs in Drosophila, Caenorhabditis elegans, or yeast. HAX-1 was first identified in a yeast two-hybrid screen as a binding partner for HS-1 (20), a kinase substrate, and a regulator of actin polymerization. HAX-1 also binds the HS1 paralog cortactin/EMS1 (19) as well as other proteins (21-23, 51, 52).
HAX-1 is closely associated with cellular membranes. Immunofluorescence microscopy of cultured cells revealed that HAX-1 resides predominantly in endoplasmic reticulum, mitochondria, or plasma membrane (19,20,23). 3 Analyses of liver subcellular fractions indicated that HAX-1 is predominantly associated with the 100,000 ϫ g sediment. The HAX-1 primary sequence has been proposed to contain a transmembrane domain in the carboxyl terminus (23); however, this region of 10 -12 amino acids is considerably shorter than most membrane-spanning helices, and HAX-1 does not contain an identifiable signal sequence. Thus, HAX-1 association with cellular membranes may be mediated by its interaction ABC transporters and other integral membrane proteins (19,23).
The subcellular localization of HAX-1 is variable and depends on cell type. HAX-1 is associated with the endoplasmic reticulum of fibroblasts (19) and with mitochondria of B lymphoma cells (20) but not mitochondria purified from rat liver. In nonpolarized MDCK II cells, HAX-1 resides primarily in intracellular membranes and did not co-localize with BSEP or MDR1. With increasing cell polarity, HAX-1 shifted toward the apical membrane, and co-localization with BSEP increased. Immunoblots of rat liver subcellular fractions revealed that HAX-1 was most highly enriched in CMV, ER microsomes, and clathrin-coated vesicles. HAX-1 co-immunoprecipitated with BSEP, MDR1, and MDR2 from CMV. Thus, HAX-1 associates with MDR transporters in the apical membrane of hepatocytes and polarized cells in culture.
To explore its role in BSEP trafficking, Hax-1 expression was ablated by RNAi. HAX-1 depletion was accompanied by a 71% increase in steady state apical membrane BSEP. Pulse-chase studies revealed no effect of Hax-1 RNAi on BSEP half-life or rates of translation, post-translational modification, or delivery to the apical membrane. However, apical membrane levels of metabolically labeled BSEP continued to rise in HAX-1-depleted cells after they declined in control cells. Apical membrane-labeled BSEP abundance remained higher than controls 4 -5 h thereafter, suggesting that HAX-1 depletion retarded BSEP internalization (Fig. 8). Further support that the likely mechanism involves clathrin-mediated endocytosis is as fol-lows: 1) BSEP and HAX-1 are present in clathrin-coated vesicles; 2) expression of dominant negative EPS15, which selectively blocks clathrin mediated endocytosis (27), doubled the apical membrane concentration of BSEP; and 3) expression of dominant negative cortactin also doubled the amount of BSEP in the apical membrane. Cortactin, an actin and HAX-1-binding protein, participates in clathrin endocytosis (41,53).
Our studies suggest that HAX-1 participates in previously unrecognized clathrin-mediated endocytosis of BSEP, and possibly other ABC transporters, from the apical plasma membrane. We are currently exploring HAX-1 interaction with specific components of the endocytic machinery.