mdr2 Encodes P-glycoprotein Expressed in the Bile Canalicular Membrane as Determined by Isoform-specific Antibodies*

We have produced antibodies specific for the three P-glycoprotein (P-gp) isoforms encoded by the mouse mdrl, mdr2, and mdr3 genes. The anti-Mdr2 and anti- Mdr3 antibodies were generated against synthetic peptides derived from the “linker” region, whereas the anti-Mdrl antibody was raised against a fusion protein containing the amino terminus of Mdrl. Western blot analysis showed that the three antibodies could dis-criminate between the three isoforms in membrane fractions from Hamster cells transfected with the corresponding full-length or chimeric mdr cDNAs. Im- munocytochemistry studies of mdr-transfected cells showed that the three antibodies specifically recog- nized each P-gp isoform expressed in whole cells. Immunoblotting of normal mouse tissues revealed that the Mdr2 isoform was expressed at very high levels in liver canalicular membrane vesicles (CMV) but not in membrane vesicles prepared from the basolateral (si-nusoidal) domain (SMV). Mdr3 was detected in intes- tinal brush border membrane vesicles and also in CMV,

to reduce intracellular drug accumulation in resistant cells. P-gps are encoded by a small family of homologous genes, termed mdr or P-gp, which include two members in humans, MDRl and MDR2 (7-9), and three members in mice, mdrl, mdr2, and mdr3 (10-13), for which full-length cDNA clones have been obtained. Predicted amino acid sequence analyses indicate that the prototype P-gp is formed by two sequence homologous halves, each composed of six transmembrane (TM) domains and a nucleotide binding (NB) fold. P-gps show a very high degree of inter-and intraspecies sequence homology (7545% identity among the three mouse proteins), the most conserved domains being the NB sites and the most divergent segments being the amino terminus and a short socalled "linker" domain joining the two homologous halves. Despite this high degree of sequence homology, striking functional differences have been detected between individual mdr genes in transfection experiments. Mouse mdrl and mdr3 and human MDRl can confer drug resistance to otherwise drugsensitive cells (10,12) but mouse mdrZ (11) and human MDR3 (MDRZ) cannot (14). Moreover, qualitative and quantitative differences in the drug resistance phenotypes encoded by mouse mdrl and mdr3 have been detected (12), and single amino acid substitutions have been shown to modulate the substrate specificity of mouse and human P-gps (15, 16).
Although the normal physiological role and substrates of P-gps remain unknown, the expression of mdr mRNA transcripts is tightly regulated in an organ-and cell-specific manner. The profile of tissue expression of mdr genes has been conserved across species, with human MDRl expression overlapping that of mouse mdrl and mdr3, and human MDR2 expression overlapping that of mouse mdr2. RNA and protein analyses have shown that human MDRl is expressed at highest levels in the adrenal gland, kidney, jejunum, colon, and endothelial cells of the blood-brain barrier (17-21), whereas MDR2 is expressed mostly in the liver (18). In normal mouse tissues, mdrl is most strongly expressed in the pregnant uterus, adrenals, placenta, and kidney, whereas the highest levels of mdr2 are found in the liver and muscle, and mdr3 is mostly expressed in intestine and lung (22). Immunohistochemical analyses indicate that with the exception of the adrenals where they are found diffusely distributed, P-gps are expressed in a polarized manner on the apical membrane of secretory epithelial cells lining luminal spaces. These include glandular epithelial cells of the endometrium in the pregnant uterus (23), the biliary canaliculi of hepatocytes, the apical surface of epithelial cells of small biliary ductules, brush border of renal proximal tubular cells, pancreatic ductule cells, and columnar epithelium cells of the intestine (19,20, 24,25). The demonstrated drug efflux function of mouse Mdrl and Mdr3 in drug-resistant cells, together with their specific tissue distribution pattern and subcellular localization, have led to the proposal that they may serve a normal detoxifying role against environmental xenobiotics or transport unidentified specific endogenous cellular products. The recent evidence that Mdrl and Mdr3 can bind (26) and interact (27,28) with progesterone is in favor of the latter proposal. The predicted function of P-gps encoded by the mouse mdr2 and human MDRP is unknown as they fail to confer drug resistance in vitro and have yet to be identified in vivo with isoform-specific antibodies.
In this study, we have generated and characterized isoformspecific polyclonal antibodies directed against each of the three mouse P-gps. Immunoblotting experiments using membrane-enriched cellular fractions from normal mouse tissues show that anti-Mdrl and anti-Mdr3 antibodies recognize unique and specific isoforms in endometrium and brush border of the intestine, respectively. Studies with the anti-Mdr2 antibody show that expression of this protein is restricted to the bile canalicular membrane where it is found present at very high amounts. Our results are discussed with respect to a possible role of Mdr2 in this tissue.

MATERIALS AND METHODS
Immunogens-Peptides for immunization were selected according to the deduced amino acid sequences of cDNAs for the mouse mdrl, mdr2, and mdr3 genes (12). For Mdr2 and Mdr3, synthetic peptides were derived from the highly divergent cytoplasmic linker domain of each protein and consisted of 12 amino acid residues of sequence EEFEVELSDEKA (2080, Mdr2, position 636-647) and CKSKDEIDNLDM (2037, Mdr3, position 638-648). Peptides were synthesized, blocked at their amino terminus, and conjugated separately via benzoil benzoic acid to keyhole limpet hemocyanin (KLH) and bovine serum albumin (BSA) (Alberta Peptide Institute, Edmonton, Alberta). For Mdrl, two @-galactosidase fusion proteins were produced by subcloning discrete mdrl cDNA fragments in the bacterial expression vector PUR (30). A 484-base pair HpaII fragment (nucleotides -22 to +462), including the amino terminus and the first predicted extracellular loop, was cloned into PUR to generate fusion protein 61. A 1.6-kilobase pair BglII/EcoRI fragment (nucleotides 2775 to 3' end), including transmembrane (TM) 11 through the COOH terminus, was cloned into PUR to generate fusion protein 66. The plasmid constructs were transformed into Escherichia coli, and bacterial extracts were prepared as described previously (31) except that the extraction buffers were: A, 25 mM Tris, 1 mM EDTA, 2.5% glycerol, 250 mM NaCl, pH 7.9; B, 25 mM Tris, 1 mM EDTA, 2.5% glycerol, 2.25 M NaCl, pH 7.9; C, 25 mM Tris, 1 mM EDTA, 2.5% glycerol, 250 mM NaC1, 8 M urea, pH 7.9. The final protein fraction from extraction C was analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and the single 0-galactosidase fusion protein band cut out of the gel after visualization in 0.5 M KC1.
Production of Antibodies-Approximately 0.2-0.3 mg of the KLH or BSA-conjugated Mdr2 (2080) or Mdr3 (2037) peptides, and of the purified 61 and 66 Mdrl fusion proteins were emulsified in an equal volume of complete Freund's adjuvant (Difco) and injected intramuscularly at four sites (two gluteal and two subscapular) in adult male rabbits. At 4-week intervals, the same amount of immunogen emulsified in incomplete Freund's adjuvant was injected intramuscularly. Rabbits were bled by arterial puncture beginning 2 weeks following the third immunization, and their sera were tested for the presence of serum antibodies. Positive rabbit antisera obtained after immunization with peptides 2080 and 2037 were concentrated by ammonium sulfate precipitation and further purified by affinity chromatography using a Sepharose matrix coupled to the respective peptide, according to a protocol recommended by the supplier of the resin (Pharmacia LKB Biotechnology Inc.). Briefly, the immunoglobulin fraction obtained by ammonium sulfate precipitation was dissolved in 5 ml of phosphate-buffered saline and dialyzed extensively against the same buffer. It was then added to a 2-ml suspension of the respective Sepharose matrix and shaken gently overnight a t 4 "C. Beads were collected by centrifugation, washed three times in phosphate-buffered saline, and placed into small column supports (Bio-Rad). Bound antibody was eluted in 2 column volumes of 0.2 M glycine, pH 2.8. Fractions were immediately neutralized by addition of 1 M Tris, pH 8, re-concentrated by ammonium sulfate precipitation, and equilibrated by extensive dialysis against phosphate-buffered saline. The protein concentration of both purified antibodies was adjusted to 0.5 mg/ml. Rabbit antisera obtained after immunization with the Mdrl fusion proteins 61 and 66 were used without further purification. mdr-transfected Cell Clones-The presence of specific antibodies was tested by immunoblotting against membrane fractions of cell clones stably transfected with the corresponding mdr cDNA. Drugsensitive LR73 Chinese hamster ovary cells (32) were used as negative controls (C) in all experiments, since this cell line was used as the recipient in all transfection assays. Full-length wild type or chimeric mdr cDNAs were cloned into the mammalian expression vector pMT2 and introduced as calcium phosphate co-precipitates into LR73 cells (33). The mdr-pMT2 constructs were co-transfected with plasmid pSV2ne0, and stable populations of co-transfectants were isolated after selection in Geneticin (G418, GIBCO) at 0.5 mg/ml, according to a protocol described previously (34). Mass populations of G418' colonies were harvested 10 days after co-transfection, expanded in culture, and multidrug-resistant cell clones overexpressing individual mdr cDNA and protein were further isolated by selection in culture medium containing vinblastine a t 50 ng/ml. A cell clone transfected with wild type mdrl cDNA (clone 1-l), and a cell clone transfected with wild type mdr3 cDNA (clone 3-5), both selected and maintained in 50 ng/ml vinblastine, were used as the source of Mdrl and Mdr3 membrane proteins, respectively (15). A cell clone stably expressing high levels of a Mdrl/Mdr2 protein chimera (clone 1/2) consisting of the highly sequence-divergent linker domain of Mdr2 introduced in Mdrl was selected and maintained in medium containing Adriamycin at 100 ng/ml (35). Finally, clone EX31-108, (clone 2), which is a G418' mdr2 co-transfectant expressing low amounts of full-length Mdr2 protein, was also used in some experiments. All cell lines were grown in a-minimal essential medium supplemented with 10% fetal calf serum, glutamine (2 mM), penicillin (50 units/ml), and streptomycin (50 pg/ml).
Isolation of Membrane Fractions-Crude membrane extracts from cultured cell clones were prepared by homogenizing cells in hypotonic buffer as described previously (12), and purified membrane fractions were further isolated by ultracentrifugation (150,000 X g for 3 h) at 4 "C on a discontinuous density gradient of 60, 45, 35, and 30% sucrose. Membrane vesicles present at the 35 and 45% interfaces were collected, washed, and stored at -80 "C in NTE (0.01 M Tris, 0.1 M NaCl, 0.01 M EDTA, pH 7.5) containing 40% glycerol. Protein concentration was determined using an amido black based commercial assay (Bio-Rad). Using methods which we have described and validated for preparation of canalicular membrane vesicles (CMV) and sinusoidal membrane vesicles (SMV) from rat liver, we prepared similar membrane fractions from the livers of 40 mice. Each liver was perfused with 0.9% saline prior to Dounce homogenization. CMV were isolated by nitrogen cavitation and calcium precipitation (36). As compared with homogenates, these preparations were selectively enriched in canalicular membrane markers; enrichment was 40-fold in leucine aminopeptidase activity. Ouabain-inhibitable Na+,K'-ATPase activity, which is localized in the basolateral plasma membrane domain of hepatocytes, was enriched less than 2-fold as compared with homogenates. SMV were isolated from homogenates by differential and sucrose-Ficoll density-gradient centrifugation (37). Ouabain-inhibitable Na',K+-ATPase activity was enriched 24-fold.
CMV and SMV were stored at -70 "C in 10 mM Hepes-Tris, pH 7.4, 0.25 M sucrose, and 0.2 mM CaC12 until used. Small intestinal brush border vesicles (BBMV) were isolated by differential centrifugation from 30 mice as described previously for comparable rat tissues (38, 39). Aminopeptidase M activity was used as a marker enzyme. When activity in BBMV was compared with that in homogenates, aminopeptidase M was 37-fold enriched. These results compare well with similar studies performed using rat small intestinal brush border preparations (39). Mouse endometrial membranes from gravid uterus were prepared by nitrogen cavitation as described by Yang et al. (29).
Western Blotting-SDS-PAGE was performed according to standard protocols (40, 41). Briefly, 20 pg of purified membrane fractions or 30 pg of purified membranes from tissues were mixed with Laemmli sample buffer for 5 min at room temperature before loading onto 7.5% SDS-polyacrylamide gels. After electrophoresis, gels were equilibrated in transfer buffer (20% methanol, 0.76 M glycine, 2.5 mM Tris, pH 8), and proteins were transferred to nitrocellulose sheets by Western blotting (0.5 A at 100 V, at 4 "C) for 1-3 h. The blots were blocked overnight at 4 "C in TBST (10 mM Tris, pH 8,0.15 M NaCl, 0.05% Tween 20) and 1% BSA, followed by incubation with the specific anti-Mdr antibodies for 1 h at room temperature and then with the appropriate goat anti-mouse or goat anti-rabbit antibody conjugated to alkaline phosphatase. Specific antigen-antibody complexes were revealed by incubation with 5-bromo-4-chloro-3-indoyl phosphatep-toluidine and nitro blue tetrazolium (Bio-Can, Montreal, Quebec). The rabbit polyclonal anti-Mdrl 61 and 66 antisera as well as the purified anti-Mdr2 (2080) and anti-Mdr3 (2037) antibodies were all used at a 1:lOO dilution, whereas the mouse anti-P-gp monoclonal antibody C219 (Centocor, Canada), which recognizes all P-gp isoforms (24, 42), was used a t a 1:300 dilution.
Immunocytochemistry-Immunocytochemical staining of cultured cells with polyclonal antisera was performed as described previously (23). Briefly, the cells were grown on slide chambers coated with 1% pig skin gelatin and fixed in 3.7% paraformaldehyde followed by immersion in cold acetone. Primary antibody (a 1:200 dilution was used for anti-Mdrl 61 and anti-Mdr3 2037, and a 1:50 dilution for anti-Mdr2 2080) was added to slides for 1 h, washed three times in TBS (0.05 M Tris, 0.15 M NaCl, pH 7.6), followed by incubation with a 1:500 dilution of a secondary goat anti-rabbit antibody conjugated to alkaline phosphatase (Promega, Bio-Tek). Slides were then stained with a solution of fast red 1 (Dakopatts, Copenhagen) dissolved in 100 ml of the following buffer: 20 mg of napthol phosphate, 2 ml of dimethyl formamide, 98 ml of 0.1 M Tris, pH 8.2, and 100 p1 of 1 M levamisole stock, counterstained with hematoxylin, and mounted with Aquamount.
Photoaffinity Labeling-Purified membrane extracts from mdrtransfected cells or CMV liver vesicles were incubated with the prazosin analogue [1251]iodoarylazidoprazosin (Du Pont-New England Nuclear, final concentration 30 nM) in a reaction buffer containing 0.1 M Tris, pH 8, and 0.05 M NaCl for 1 h at 4 "C, followed by crosslinking under UV for 5 min, as described previously (29). The unincorporated radioactive prazosin analogue was removed by centrifugation (100,000 x g, 20 min), and labeled P-gps were recovered by immunoprecipitation with mouse anti-P-gp monoclonal antibody C219 (1:lOO) or anti-Mdr3 2037 (dilution 1:lOO) or anti-Mdr2 2080 (dilution 1:50). Immunoprecipitations were carried out for 16 h at 4 " C in 0.4 ml of buffer containing 50 mM Tris, 0.2% SDS, 1% Triton X-100, and 150 mM NaCl, pH 7.5. Immune complexes were isolated by incubation for 2 h at 4 'C with protein A-Sepharose beads, followed by four washes in buffer containing 0.1% Triton X-100,0.03% SDS, 150 mM NaC1, 50 mM Tris, and 5 mg/ml BSA and two washes in buffer containing 150 mM NaCl and 50 mM Tris. The beads were then resuspended in Laemmli sample buffer and electrophoresed on a 7.5% SDS-polyacrylamide gel. Gels were fixed, dried, and exposed to Kodak X-AR films with an intensifying screen (Kronex, E. I. Du Pont de Nemours & Co.) at -70 "C for 2-4 weeks.

RESULTS
Production and Characterization of Isoform-specific Antibodies-The series of Mdr fusion proteins and synthetic peptides used as immunogens to generate the isoform-specific antibodies in rabbits are illustrated in Fig. 1. The Mdrl fusion protein 61 was derived from the highly sequence divergent amino-terminal domain, including the predicted transmembrane domain 1 (TM1) and the first extracellular loop (position 1-122). Mdrl fusion protein 66 contained the entire COOH-terminal domain (position 885-1276) including TMll-12 and the highly conserved nucleotide binding site 2 (NB2). The Mdr2 peptide 2080, EEFEVELSDEKA (position 636-647) and the Mdr3 peptide 2037, CKSKDEIDNLDM (position 638-648) were both derived from the linker domain which bears no sequence homology amongst mouse P-gps and thus would not be expected to have shared epitopes. These proteins and synthetic peptides were used to continuously immunize rabbits, followed by periodic testing of sera for the presence of anti-Mdr antibodies. The putative specificity of the rabbit antisera was tested by Western blotting, using purified membrane fractions from either drug-sensitive Chinese hamster LR73 cells (C) or drug-resistant LR73 cell clones expressing discrete P-gps bearing the individual epitopes used for immunization. In the case of Mdrl and Mdr3, multidrug-resistant cell clones (clones 1-1 and 3-5, respectively) transfected with each cDNA and expressing similar levels of the corresponding Mdr protein were isolated previously (15). Since the full-length mdr2 cDNA does not confer drug resistance, and geneticin-resistant mdr2 co-transfectants 648), were synthesized, blocked at both extremities, and conjugated express low levels of protein,* we have not been able to analyze the full-length Mdr2 protein in membrane fractions from these cells by immunoblotting. To overcome this limitation, we have introduced the Mdr2 epitope used for immunization (linker domain) into an Mdrl background (35). The resulting Mdrl/Mdr2 chimera can confer drug resistance, and a cell clone overexpressing this chimeric P-gp (clone 1/2) was isolated by selection in Adriamycin. The presence of Mdr proteins in these clones was first ascertained using the mouse anti-P-gp monoclonal antibody C219 which recognizes the epitope VQEALD near NB1 and NB2 which is common to the three mouse Mdr isoforms (24). C219 recognized proteins of apparent molecular mass 170 kDa in cell clones transfected with either wild type mdrl (clone 1-1) or chimeric mdrllmdr2 cDNA (clone 1/2), and of 150-kDa species in the mdr3 transfectant (clone 3-5), which were absent in LR73 control cells (Fig. 2 A ) . The additional 95-kDa species detected by C219 is likely to be a proteolytic fragment previously reported by Yoshimura et al. (43). The differences in electrophoretic mobilities and apparent molecular masses detected between Mdrl and Mdr3 have been documented previously by us (12,15) and others (28). Of the two rabbit antisera generated against Mdrl fusion proteins 66 and 61 (Fig. l), only antiserum 61 proved to be isoform-specific, reacting only with membranes from mdrl transfectants (Fig. ZB, leftpanel). The antiserum raised against fusion protein 66 reacted with both the Mdr3 and Mdrl proteins and most likely recognizes epitope(s) associated with the COOH terminus NB2 site and * E. Buschman and P. Gros, unpublished data. conserved in both proteins (Fig. 2B, right panel). Results shown in Fig. 2C demonstrate that the affinity-purified rabbit antiserum directed against Mdr2 peptide 2080 specifically recognized the linker domain of Mdr2 included in the Mdrl/ Mdr2 chimeric protein (clone 1/2), but failed to react with either wild type Mdrl or Mdr3. Finally, the afffity-purified anti-Mdr3 antibody directed against peptide 2037 reacted only with membranes from cell clones expressing the wild type Mdr3 protein, but did not react with either the wild type Mdrl or MdrllMdr2 chimera (Fig. 20). These results establish that three of the four antisera generated here are isoformspecific and can distinguish individual mouse P-gp isoforms in immunoblotting experiments. A comparison of the intensity of signals generated by C219 (Fig. ?A), anti-Mdrl (61, Fig. 2B), anti-Mdr2 (2080, Fig. 2C), and anti-Mdr3 antibodies (2037, Fig. W ) suggests that the anti-Mdr3 is the isoformspecific antibody showing the highest titer and/or greatest affinity for its epitope.
Immunocytochemical Analysis of mdr-transfected Cell Lines-To determine the usefulness of these antibodies for the immunochemical detection of individual mouse P-gps in normal tissues, their specificities were tested in fixed cells by a technique that employs fast red to visualize specific antibody binding (23). Each antibody was tested for reactivity against drug-sensitive LR73 cells, and cell clones expressing Mdrl (clone 1-l), Mdr2 (clone EX31-108, termed 2), Mdr3 (clone 3-5), the Mdrl/Mdr2 chimera (clone 1/2), and results are shown in Fig. 3. The anti-Mdrl antiserum 61 (Fig. 3, left panel) reacted with Mdrl expressing cells (clone  as well as cells expressing or the 1/2 chimera (clone 1/2), as expected, since both clones express the amino-terminal Mdrl epitope used to generate the antibody. Antiserum 61 showed no reactivity against either LR73 control cells or clones expressing Mdr2 (clone 2) or Mdr3 (clone 3-5). The anti-Mdr2 antiserum 2080 also proved to react in an isoform-specific fashion in this assay (Fig. 3, centerpanel), it reactedpositively with cells expressing either the wild type Mdr2 (clone 2) or the Mdrl/ Mdr2 chimera (clone 1/21 and failed to react with either LR73 cells or clones expressing either Mdrl (clone  or Mdr3 (clone [3][4][5]. Although the level of Mdr2 polypeptide detected in EX31-108 (clone 2) by Western or immunoprecipitation is very low: strong staining of this clone was obtained by the current protocol. Finally, the anti-Mdr3 antiserum 2037 reacted very strongly with Mdr3-expressing cells but did not react with either LR73 control cells or the other Mdr expressing cells (Fig. 3, right panel).
The immunocytochemical staining pattern for drug-resistant transfectants expressing either wild type Mdrl or Mdr3 or the Mdrl/MdrZ chimera indicated that over 90% of the cells were strongly positive and showed homogeneous staining from cell to cell with either antiserum 61 or 2037. For Mdr2expressing cells, the staining obtained by 2080 was more heterogenous, suggesting a greater variability in the amount of Mdr2 protein produced in these transfectants. A similar heterogenous staining was recently reported for the human homolog MDR2 overexpressed in BRO transfected cells (14). As noted previously in Western blots (Fig. 2), the anti-Mdr3 antiserum 2037 produced a stronger staining at an equal or lower dilution than that produced by the two other antisera. The staining profiles obtained with the three iso&psp&&   or the mdrl/mdr2 chimeric cDNA containing the mdr2 linker region inserted in mdrl (clone 1/2), mdr2 (clone 2), or mdr3 (clone [3][4][5]. Cells were fixed in paraformaldehyde and then acetone and exposed to the anti-Mdr antisera. Specific immune complexes were revealed by a goat anti-rabbit alkaline phosphatase-conjugated antibody (MOO dilution), followed by fast red substrate and counterstaining with hematoxylin. antibodies in drug-resistant transfectants (clones 1-1, 1/2, and 3-5) clearly showed an intense ring-like staining at the peripheral edge of the cells, with a more diffuse and less intense intracellular staining, suggesting that in these transfectants the Mdr isoforms are concentrated at the cell membrane. On the other hand, staining of drug-sensitive mdr2 transfectants (clone 2) by anti-Mdr2 2080 antiserum appeared to show relative homogeneous staining over the cell body with a less intense staining at the peripheral edge. Taken together, results of immunocytochemistry and immunoblotting clearly show that the three antisera are isoform-specific.
Detection of Mdr Isoforms in Normal Mouse Tissues-The three Mdr-specific antibodies were used to detect expression of the three isoforms in membrane fractions from normal mouse tissues. Tissues known to express either mdr mRNAs or proteins as detected by gene-specific hybridization probes (22) or generic anti-P-gp antibodies (42) were used. These included liver canalicular and sinusoidal membrane vesicles (CMV and SMV, respectively), intestinal brush border membranes (BBMV), and membrane fractions from endometrium of pregnant uterus (29). Immunoblots of membrane fractions from normal tissues (30 pg) or mdr-transfected cells (20 pg) were individually probed with generic (C219) or isoformspecific antibodies (Fig. 4). The monoclonal antibody C219 ( species of lesser abundance. These two latter species were not consistently detected in independent CMV preparations and may represent degradation products of the abundant 140-150-kDa band. No major C219-reactive species was detected in SMV. A 150-kDa isoform was identified in the intestinal BBMV fraction and a 170-kDa isoform in endometrium membrane fractions (data not shown). The anti-Mdrl antiserum 61 recognized Mdrl in control transfected cell clones (1-1, 1/ 2), failed to detect the C219-reactive species present in BBMV or CMV (Fig. 4B), but recognized a 170-kDa species in the endometrium of gravid uterus (Fig. 4C) as the major Mdr isoform in this tissue which is in agreement with previous studies using gene-specific hybridization probes (23,25). In other experiments, neither anti-Mdr2 nor Mdr3 antisera reacted with the Mdr isoform in the gravid uterus (data not shown). The isoform-specific anti-Mdr2 antiserum 2080 identified Mdr2 in control transfectants and also recognized the 150-kDa isoform present in liver CMV, but did not react with either SMV or intestinal BBMV preparations (Fig. 40). The gel in Fig. 4 0 was run longer in order to clearly distinguish the difference in electrophoretic mobility between the Mdr2 P-gp isoforms present in clone 112 and in liver CMV. Finally, the anti-Mdr3 antiserum 2037 identified the 145-kDa intestinal BBMV isoform as being encoded by mdr3 and also detected a 140-kDa isoform in CMV (Fig. 4E), indicating that both Mdr2 and Mdr3 are expressed in bile canaliculi. The observations that (i) P-gp staining by C219 in Mdrl/Mdr2 (clone 1/2) and Mdr3 (clone 3-5) transfectants and in CMV is similar, (ii) Mdr staining by anti-Mdr2 antiserum in Mdrl/ Mdr2 transfectants and in CMV is similar, and (iii) Mdr staining by the high titer/or affinity anti-Mdr3 antiserum is much stronger in Mdr3 transfectants than in CMV suggest that Mdr2 is the major P-gp isoform expressed in bile canaliculi and that Mdr3 is expressed a t a much lower level.
Photolabeling of Liver Canalicular Membranes with r"I] Zodoarylazidoprazosin-Rat CMV prepared according to the same protocol used to isolate mouse liver CMV used here can sustain ATP-dependent daunomycin transport (44), suggesting the presence of a functionally active Mdr isoform in bile canaliculi. Since in transfection assays, cDNA for either mouse mdr2 or human MDR2 fail to confer drug resistance, we wished to determine whether either or both Mdr2 or Mdr3 isoforms could bind the photoactivatable drug analogue ['*'I] iodoarylazidoprazosin (IAAP) in cross-linking experiments. IAAP has been shown to bind to P-gp (45), and the binding site of IAAP has been tentatively mapped to the same 6-kDa tryptic peptide fragment as that labeled by azidopine, another photoactivatable P-gp substrate (46). Membrane fractions from mdr transfectants or CMV were incubated with IAAP, cross-linked under-UV, followed by immunoprecipitation with either C219, anti-Mdr2, or anti-Mdr3 antibodies (Fig. 5). The (2219 antibody (Fig. 5A) could immunoprecipitate photolabeled P-gps from liver CMV (-150 kDa) and membranes from cell clones expressing Mdr3 (150 kDa) and the Mdrl/Mdr2 chimera (170 kDa). The anti-Mdr3 antiserum 2037 (Fig. 5B) could immunoprecipitate photolabeled Mdr3 from Mdr3 expressing cell clones but also a species of similar electrophoretic mobility (-150 kDa) in CMV preparations. Although the specific signal detected on the autoradiogram was weak, the specific photolabeled Mdr3 protein was consistently detected in independent experiments (data not shown). The small amount of Mdr3 detected by IAAP binding is consistent with the relatively small amount of Mdr3 detected in CMV by immunoblotting with anti-Mdr3 antiserum 2037 (Fig. 4E).
These results indicate that the Mdr3 isoform, present in transfected cell clones (clone 3-5) and in liver CMV, can bind IAAP. However, upon immunoprecipitation with the anti-Mdr2 antiserum 2080 (Fig. 5C) isoform was visible in cells expressing the MdrllMdr2 chimera, but no Mdr2-photolabeled peptide was detected in CMV, even after prolonged exposure of the autoradiogram. Although the possibility exists that the antibody titer of 2080 was not high enough to detect photolabeled Mdr2 in CMV, this is unlikely since the amount of photolabeled Mdr protein detected by generic anti-P-gp antibody C219 in CMV is similar to that detected in the chimera 112 (Fig. 5A), whereas the anti-Mdr2 antibody readily detected the photolabeled 112 chimeric P-gp but gives no signal in CMV (Fig. 5C). Therefore, the data suggest that Mdr2 in CMV binds IAAP poorly, if at all.

DISCUSSION
We have used fusion proteins and synthetic oligopeptide immunogens derived from Mdr protein segments which are not conserved among the three members of the mouse family t o raise isoform-specific anti-Mdr antibodies. The specificity of these antisera was ascertained both by immunoblotting of membrane fractions and by immunocytochemistry on whole cells stably transfected and expressing wild type or chimeric Mdr proteins harboring the specific epitopes used as immunogens. These antisera should prove useful for the immunohistochemical identification of the three mouse isoforms in normal tissues at the cellular and subcellular levels. Although the three mouse mdr cDNAs are predicted to be composed of a n identical number of 1276 residues, differences in the apparent molecular mass of the three isoforms were noted by Western blotting in membrane fractions of transfectants and normal tissues. The Mdrl protein detected in mdrl-transfectants and normal endometrium (Fig. 4C) appeared to be approximately 170 kDa in size, and the Mdr2 and Mdr3 proteins showed faster electrophoretic mobilities of 140-150 kDa (Fig. 4, A, D , and E ) . In addition, the Mdr3 protein detected in liver CMV appeared smaller (140 kDa) than that detected in BBMV (145 kDa, Fig. 4E). The reasons for these differences in apparent molecular mass are not fully understood but could reflect different post-translational modifications (47) of the distinct isoforms in transfected cells (for example, Mdrl uersus Mdr3) or the same isoform in different tissues (Mdr3, CMV uersus BBMV). Differences in the extent of glycosylation or phosphorylation could be implicated; indeed, Mdr3 shows three and Mdrl four Asn-X-Ser (Thr) consensus sites for N-linked glycosylation (12) in their first predicted extracellular loops, and two-dimensional analysis of Mdrl and Mdr3 phosphopeptides reveals distinct profiles? Also, variations were noted in the immunocytochemical stain-A. Veillette and P. Gros, unpublished data. ing profiles of transfected cells by the three isoform-specific antibodies (Fig. 3). All mdrl and mdr3 drug-resistant transfectant cells stained homogeneously positive, however, individual mdr2-transfected cells showed a more heterogenous staining pattern varying from strong to weak intensity. These differences are most likely caused by the fact that continuous expression of Mdrl and Mdr3 is required for survival of transfected cells in drug-containing medium, whereas no such selective pressure can be applied on Mdr2. Expression of Mdr2 in these cells is only maintained by the co-transfected ne0 marker. The staining profile of mouse Mdr2 detected here is similar to that observed in BRO cells transfected and overexpressing the human MDRB homolog and stained with (2219 antibody (14).
Although the functional role of P-gps in normal tissues is still unknown, the expression of specific mdr mRNAs is known to be tightly regulated in a tissue-specific fashion. Moreover, this pattern of expression is conserved across species with human MDRl expression overlapping that of rodent mdrl and mdr3, whereas human MDRB and mouse mdr2 are expressed in the same tissues (17-20, 22, 23). The primary goal of our study was to generate immunological reagents to identify specific Mdr isoform expression at the protein level in normal tissues known to express high levels of mdr mRNA transcripts. We detected the highest levels of Mdrl, Mdr2, and Mdr3 protein expression in membrane fractions from endometrium, bile canaliculus, and intestine, respectively. In the case of mouse (Mdrl and Mdr3) and human (MDR1) isoforms known to confer drug resistance in transfected cells, two possible physiological roles have been proposed. These isoforms could function as drug efflux detoxifying mechanisms to protect the intracellular milieu against environmental xenobiotics or could transport as yet unidentified cellular products. In support of the former hypothesis is the known drug efflux function of these isoforms in MDR cells and their distribution in certain normal tissues such as the bile canaliculus, intestinal brush border, the kidney proximal tubule, endothelial cells of the blood brain barrier, and pluripotent precursor cells of the hemopoietic system (48). On the other hand, the expression of these isoforms in other anatomical sites such as pancreatic ductules, endometrial glands of the pregnant uterus, and the adrenal cortex does not support a detoxifying role but rather suggests the transport of other unknown substrates. The finding that mouse Mdrl (Mdrlb) and Mdr3 (Mdrla) can bind (26) and interact (27-29) with progesterone is also in favor of this proposal. The proposed role of mouse Mdr2 and its human MDR2 homolog is even more mysterious, since neither confer drug resistance to transfected cells, and since the absence of isoform-specific antibodies has precluded identification of these isoforms in normal tissues at the cellular and subcellular levels. In the present study, we have identified the bile canaliculus as a major site of normal Mdr2 expression, confirming and extending earlier studies with gene-specific hybridization probes that identified liver as the major site of mdr2 RNA expression (22). We have determined that Mdr2 expression is restricted to CMV and is completely absent from SMV. Moreover, the level of Mdr2 expression in CMV was found to be very high and comparable with the high degrees of protein expression attained in mdrl and mdr3 transfectants (Fig. 2, A and D ) . Both the very high amounts and the striking compartmentalization of protein expression suggest that Mdr2 plays an important function in CMV. At present, the nature of this physiological role can only be speculated upon. ATP-dependent daunomycin transport has been detected previously in CMV (44), suggesting that Mdr2 could function as a drug efflux pump at this site. In order to investigate the possibility of drug binding to Mdr2 in CMV, we chose iodinated IAAP for its high specific activity rather than tritiated azidopine which has lower specific activity. Our findings that; (i) low levels of Mdr3 are co-expressed with Mdr2 in CMV, (ii) photolabeling by the drug analog IAAP could be demonstrated only for Mdr3 but not Mdr2, and (iii) mdr3 but not mdr2 can confer drug resistance to transfected cells, together suggest that Mdr3 may be responsible for the daunomycin transport detected previously in CMV (44). The finding that the Mdr3 isoform in CMV was only weakly photolabeled by IAAP was most likely due to the relatively small amount of Mdr3 detected in CMV.
What then could be likely candidate substrates for MdrS? Although most substrates for Mdrl and Mdr3 are small organic cations of relative hydrophobicity having molecular weights of approximately 400 to 1500, substrates for Mdr2 need not be affected by these physical or chemical properties. However, our demonstration that NB1 and NB2 of Mdr2 can functionally complement in chimeric proteins the drug efflux function of Mdrl (35) suggests that the mechanism of substrate transport by Mdr2 is likely to be ATP-dependent. Two distinct ATP-dependent transport systems have recently been described in inside-out CMV from rat liver (49-53).4 One is specific for bile acids (49, 50)4 and the other for nonbile acid organic anions, such as bilirubin glucuronides, GSH conjugates, carboxyfluorescein diacetate, and other similar ligands (51)(52)(53). A candidate ATP-dependent bile acid transporter of apparent molecular mass 110 kDa has been identified and reconstituted in proteoliposomes (50). The size of the transporter is incompatible with the apparent electrophoretic mobility and predicted molecular mass of Mdr2, estimated at greater than 140 kDa. The ATP-dependent nonbile acid organic anion transporter has not yet been isolated but is known to be functionally absent in CMV from a mutant rat strain (TR-) showing an autosomal recessive inherited defect (51)(52)(53). The transport defect in TR-rats is limited to the liver, whereas comparable transport by the kidney remains intact (54). We are currently testing the candidacy of these and other molecules as Mdr2 substrates in cell clones stably transfected and expressing the mdr2 gene.
In conclusion, we have generated isoform-specific antibodies that can recognize individual mouse P-gps by immunoblotting, immunocytochemical staining, and immunoprecipitation. In particular, we have generated an antibody against the enigmatic member of the mdr family (mouse mdr2 and human MDR2) that does not encode multidrug resistance. We have used this antibody to demonstrate that Mdr2 protein expression is compartmentalized within the liver, is highly polarized, and is restricted to the canalicular but not basolat-era1 membrane of hepatocytes. Our findings also show that in CMV, only Mdr3, and not Mdr2, appears capable of combining the photoactivatable drug analogue IAAP. The observation that drug molecules bind poorly if at all to Mdr2 may explain its inability to confer drug resistance.