MRP3, a new ATP-binding cassette protein localized to the canalicular domain of the hepatocyte

Daniel F. Ortiz, Shaohua Li, Ramachandran Iyer, Xingming Zhang, Phyllis Novikoff, Irwin M. Arias


Bile secretion in liver is driven in large part by ATP-binding cassette (ABC)-type proteins that reside in the canalicular membrane and effect ATP-dependent transport of bile acids, phospholipids, and non-bile acid organic anions. Canalicular ABC-type proteins can be classified into two subfamilies based on membrane topology and sequence identity: MDR1, MDR3, and SPGP resemble the multidrug resistance (MDR) P-glycoprotein, whereas MRP2 is similar in structure and sequence to the multidrug resistance protein MRP1 and transports similar substrates. We now report the isolation of the rMRP3 gene from rat liver, which codes for a protein 1522 amino acids in length that exhibits extensive sequence similarity with MRP1 and MRP2. Northern blot analyses indicate that rMRP3 is expressed in lung and intestine of Sprague-Dawley rats as well as in liver of Eisai hyperbilirubinemic rats and TR mutant rats, which are deficient in MRP2 expression. rMRP3 expression is also transiently induced in liver shortly after birth and during obstructive cholestasis. Antibodies raised against MRP3 recognize a polypeptide of 190–200 kDa, which is reduced in size to 155–165 kDa after treatment with endoglycosidases. Immunoblot analysis and immunoconfocal microscopy indicate that rMRP3 is present in the canalicular membrane, suggesting that it may play a role in bile formation.

  • MRP
  • bile duct ligation
  • TR rat
  • apical hepatocyte domain
  • glycoprotein

bile secretion is an essential route for excretion and recirculation of drugs, vitamins, and lipids as well as endogenous and exogenous toxins. Biliary constituents such as bile acids, phospholipids, and glutathione and glucuronide conjugates exhibit concentrations in bile that are manyfold higher than those found in cytoplasm, indicating that secretion is an active process. Various ATP-binding cassette (ABC)-type proteins, which are sorted to the apical domain of the hepatocyte, promote bile formation by mediating ATP-dependent translocation of biliary components across the canalicular membrane. SPGP effects ATP-dependent transport of taurocholate (3, 7), generating a large fraction of the osmotic gradient that drives bile flow. MDR3 mediates flipping of phosphatidylcholine in the canalicular membrane, thereby fostering its release into bile (28, 33). MRP2 (also called cMOAT and cMRP) is primarily responsible for non-bile acid-dependent bile flow, effecting biliary secretion of glutathione, sulfate, and glucuronide conjugates, which include conjugated bilirubin (2, 12, 31). MDR1 is associated with transport of a wide variety of small, hydrophobic, cationic drugs into bile (16) and flipping of short-chain phospholipids in the canalicular membrane (41).

A number of inheritable human disorders are associated with mutations in genes that code for canalicular transporters. Mutations in the SPGP gene are associated with subtype II of progressive familial intrahepatic cholestasis (PFIC) (36). PFIC III and the Dubin-Johnson syndromes have been linked with mutations in the MDR3 (5) and MRP2 (17,32, 42) genes, respectively. Single nucleotide changes in the MRP2 and MDR3 mRNAs introduce nonsense codons that interrupt the protein coding sequences and result in destabilization of the mutant transcripts. Animal models have been generated or identified for some of these diseases. For example, mdr2−/− knockout mice (35) exhibit a pattern of cholestasis and cirrhosis similar to that observed in PFIC III patients. TR mutant rats (2, 31) and Eisai hyperbilirubinemic rats (EHBR) (12) were identified as hyperbilirubinemic mutants deficient in biliary secretion of non-bile acid organic anions and were shown to harbor point mutations in the MRP2 gene that severely reduce RNA and protein levels. Lith mice, which manifest cholesterol cholelithiasis, overexpress SPGP and display altered taurocholate levels in bile (43).

Canalicular ABC-type transporters can be classified into two major subtypes based on sequence and structure. ABC-type proteins characteristically contain one or two highly conserved nucleotide-binding domains (NBD) associated with hydrophobic regions capable of spanning the membrane multiple times. SPGP, MDR1, and MDR3 share extensive amino acid sequence similarity and exhibit a symmetrical 6+6 arrangement of transmembrane helices in the hydrophobic domains linked to the NBDs. MRP2 and its homologue MRP1 are asymmetrical in structure, display an 11+6 organization of the membrane-spanning domains (18, 19), and share little amino acid sequence similarity with the MDR subfamily outside of the conserved NBD regions. We now report identification in rat liver of another member of the MRP subfamily, named rMRP3, which shares substantial sequence and structural similarities with MRP2 and MRP1. rMRP3 expression is induced by obstructive cholestasis and in mutant rats deficient in MRP2 expression. We also show that the rMRP3 protein is present in the canalicular membrane of the hepatocyte.



The amino acid sequence of the yeast bile acid transporter BAT1p (29) was aligned with the proteins most closely resembling it in yeast and mammals by using the CLUSTALW algorithm. Sequence motifs that were conserved in MRP1, YHLOp, YK83p, and YCF1p, but absent in the canalicular ABC-type transporters MDR1 and MDR3, were identified and used to design degenerate oligonucleotide primers. The primers were used for RT-PCR of 1 μg of poly(A)+ RNA prepared from human liver and TR rat liver; the latter represents a rat mutant deficient in MRP2 mRNA accumulation. The sense primer (AAAGAATTCGGIATIGTIGGIC/AGIACIGG) represented amino acid sequence GIVGRT, and the antisense primer (AAAGAATTCAA/GICC/TG/ATGIGCIATIGT) was derived from amino acid sequence TIAHRI. An EcoR I site preceded by three random nucleotides was incorporated on the 3′ terminus of each primer. PCR parameters were 3× 95°C, 40°C, 72°C and 30× 95°C, 55°C, 72°C; all steps were for 1 min. PCR products 470–520 bp in length were purified from agarose gels and cloned into the pCR2 vector (Invitrogen). Clones were classified by restriction mapping, Southern blot analysis, and nucleotide sequencing.

cDNA inserts were isolated from human liver (Stratagene) and rat duodenum (a gift of Dr. Andrew Leiter) lambda cDNA libraries by hybridization with radiolabeled human and rat RT-PCR inserts purified from pCR2. cDNA inserts of 3.0 kb and 1.5 kb were isolated for rat and human, respectively. Multiple rounds of 5′ rapid amplification of cDNA ends (RACE)-PCR of human and TR rat liver poly(A)+ RNA were performed with a kit obtained from Life Technologies and according to the manufacturer’s instructions.

A full-length cDNA was constructed for each gene by combining as much as possible of the cDNA inserts obtained from lambda cDNA library and cDNAs generated by PCR. The 5′ portion of the rat and human cDNAs were amplified from liver poly(A)+ RNA by RT-PCR using the proofreading thermostable polymerase Pfu (Stratagene). Oligonucleotide primers were designed based on sequence information obtained from the furthermost 5′ RACE products. The human gene primers were TATGCGGCCGCTCGCCTTCCTTGCAGCC (sense) and AGATCTAGAGTGTCAAAGAAGGACTGTGG (antisense), which included a NotI and Xba I, respectively, on the 3′ terminus. For the rat cDNA, primers were CTTCTAGCTGGGGTTGAG (sense) and GTCCCTGGTCCAGAAGGAG (antisense). The nucleotide sequences of the human and rat cDNAs were obtained by automated DNA sequencing of double-stranded plasmid DNA.

Northern blot analysis.

Tissues were removed from euthanized rats and frozen immediately in liquid nitrogen. Total RNA was prepared using the Trizol reagent system (LifeTechnologies). Poly(A)+ RNA was prepared from total RNA using the FastTrac 2.0 kit (Invitrogen). Poly(A)+ RNA (5 μg/lane) was loaded and separated by electrophoresis on denaturing formaldehyde-1% agarose gels. RNA was immobilized on GeneScreen plus (DuPont) membranes after capillary transfer. DNA fragments were radioactively labeled by random primer extension (Boehringer Mannheim). The rMRP3 probe was a 1.3-kb BamH I fragment obtained from rMRP3 cDNA. The rMRP2 probe was a 1.7-kb Nco I-Xho I fragment obtained from a partial rMRP2 cDNA. Hybridization at 42°C was done in 50% formamide, 6× standard saline citrate (SSC), 10× Denhardt’s, 0.5% SDS, and 100 μg/ml denatured salmon sperm DNA in a Robbins Scientific hybridization oven. The blots were washed sequentially with 2× SSC, 0.5% SDS at 25°C, 1× SSC, 0.25% SDS at 42°C, and 0.2× SSC, 0.05% SDS at 65°C and exposed to film.

Bile duct ligation.

Male Sprague-Dawley rats 250–300 g in weight were anesthetized by intraperitoneal injection of pentobarbital sodium (50 mg/kg body wt). The abdominal cavity was opened aseptically, and the common bile duct was ligated at two locations and cut between the ligatures. Sham-operated animals underwent the same procedure in which the common bile duct was exposed but not ligated or cut. Livers were extracted at the times indicated and processed for subcellular fractionation, RNA preparation, or immunocytochemistry. Sinusoidal membrane vesicles (SMV) and canalicular membrane vesicles (CMV) were prepared as described (11).

Antibodies and immunoblot analysis.

The “linker” domain of MRP3, which resides immediately downstream of the first NBD, exhibits little or no homology with other proteins in the database. A 216-bp DNA fragment encoding amino acids 876–948 of human MRP3 was ligated in frame with the glutathioneS-transferase (GST) coding region of pGEX5-3x (Pharmacia). The construct was transformed into Escherichia coli, and a GST-MRP3 fusion protein was overexpressed and purified by affinity chromatography on GSH-Sepharose (Pharmacia). Antisera raised against the GST fusion protein recognized an isopropyl-β-d-thiogalactopyranoside inducible peptide of relative molecular mass 39 kDa in extracts ofE. coli expressing the GST-MRP3 fusion. IgG was purified from rabbit sera by affinity chromatography on protein A-Sepharose (Pharmacia). Anti-GST antibodies were cleared from IgG by passage through a GST-Sepharose column (Pierce). GST-MRP3 was bound to Affigel-10 (Bio-Rad) following the manufacturer’s instructions and used to purify MRP3-specific IgG by affinity chromatography. Immunoblot analysis was performed as described (29). Briefly, 50 μg of protein were incubated for 10 min at 65°C in 2.5% SDS, 0.1 M Tris ⋅ HCl, pH 6.8, 1 M urea, and 5 mM dithiothreitol and separated by SDS-PAGE. Proteins were electrotransfered to nitrocellulose and incubated with antibody diluted in 0.25% (wt/vol) gelatin, 0.5% (wt/vol) BSA, 50 mM Tris ⋅ HCl, pH 8.0, 150 mM NaCl, and 0.05% (vol/vol) Tween 20. Secondary antibodies were horseradish peroxidase-conjugated goat anti-rabbit IgG and anti-mouse IgG (Bio-Rad). Antibody staining was visualized by enhanced chemiluminescence (Pierce)


Liver fragments (5 × 5 × 2 mm) were immersed in 0.25 M sucrose-PBS for 30 min at 0°C, submerged in optimum cutting temperature (OCT) compound (Tissue-Tek), and frozen on dry ice. Sections (6–8 μm) cut with a cryostat were incubated in blocking buffer [PBS containing 3% (wt/vol) BSA, 2% fetal calf serum, 0.02% (vol/vol) Triton X-100, and 0.1 M glycine] at 37°C for 1 h. Slides were washed in PBS at room temperature for 5 min and incubated for 2 h with affinity-purified anti-MRP3 IgG (50 μg/ml) at 37°C. After washing with PBS for 5 min, the slides were immersed in 3% (vol/vol) formaldehyde-PBS at 0°C for 2 min, washed with three changes of PBS for 15 min, and incubated in blocking buffer at room temperature for 15 min. The sections were stained with 10 μg/ml C219 IgG (Signet) for 2 h at room temperature, washed with PBS for 5 min, and incubated with Texas red-labeled goat anti-rabbit IgG and Cy2-labeled goat anti-mouse IgG (Jackson ImmunoResearch) at 15 μg/ml in blocking buffer for 45 min at room temperature. Slides were washed for 5 min with three changes of PBS at room temperature, mounted with Crystalmount (Biomeda), and observed at ×400 magnification by confocal microscopy.


An MRP-like protein exhibits differential expression in liver of TR rats, EHBR, and normal rats.

RT-PCR of poly(A)+ RNA from TR liver identified a number of MRP-like sequences that differed from MRP2, the canalicular multiple organic anion transporter. The primers were degenerate oligonucleotides capable of hybridizing to the NBD2 coding domains of MRP-type subfamily members but not to the MDR-type subfamily. Northern blot analysis indicated that probes derived from one of the PCR fragments hybridized to a 5500-nt transcript present in lung and intestine of normal Sprague-Dawley rats and liver of mutant EHBR. The mRNA was also elevated in liver of TR mutant rats but exhibited very low levels in normal Sprague-Dawley rat liver (Fig. 1). A full-length cDNA for the gene was assembled from lambda phage inserts and an RT-PCR fragment generated by a proofreading thermostable polymerase. Nucleotide sequence analysis indicated that the cDNA contained a single long open reading frame (ORF) capable of encoding a polypeptide 1522 amino acids in length that bears the hallmarks of an ABC-type protein (Fig.2). The ORF contains two hydrophobic polytopic membrane-spanning regions and two nucleotide-binding domains (NBD1 and NBD2) harboring Walker A and B nucleotide-binding motifs and the C moiety characteristic of ABC-type proteins.

Fig. 1.

Expression of rMRP3 gene in rat tissues. Northern blot of RNA prepared from Sprague-Dawley (normal) male rats and liver from an Eisai hyperbilirubinemic rat (EHBR) mutant. Each lane contains 5 μg of poly(A)+ RNA. The blot was hybridized to a radioactively labeled 480-bp DNA fragment exhibiting homology to MRP subfamily proteins. The probe recognizes a 5500-nt transcript in intestine, lung, and EHBR liver and faint signals in other tissues. The blot was stripped and rehybridized to a glyceraldehyde-3-phosphate dehydrogenase (GAPDH) probe to show comparable loading of RNA in each lane.

Fig. 2.

Amino acid sequence deduced from a full-length rMRP3 cDNA. Analysis of the nucleotide sequence obtained from a 5103-bp cDNA indicated the presence of a 1522-amino acid open reading frame, which initiates with a methionine codon that resides within a Kozak consensus sequence. Two highly conserved nucleotide-binding domains (NDB) characteristic of ATP-binding cassette (ABC)-type proteins can be identified that contain consensus Walker A and B motifs as well as the ABC C signature moieties (shown in boldface in sequence and labeled as A, B, and C). Analysis of the amino acid sequence with the TMAP, TmPred, and PHD transmembrane prediction algorithms identified 10–12 transmembrane domains in the amino-terminal region and 4–6 membrane-spanning segments in the carboxy region, which have been indicated in the sequence with a stippled underline. PROSITE analysis identifies 9 possible asparagine-linked glycosylation sites. Six of the sites reside within the NBD1 and NDB2 domains and are probably not decorated with sugars. The remaining putative receptor asparagine residues have been circled.

A blast search of the GenBank databases indicated the partial human expressed sequence tag (EST) sequence MRP3 (21), so named for its similarity to MRP1 and MRP2, is 80% identical to the rat nucleotide and protein sequences. This level of identity is equivalent to that observed between rat and human homologues of MRP2 and MDR3. Accordingly, the new gene has been named rMRP3. Outside of hMRP3, the protein most similar to rMRP3 is MRP1 (55% identity). Other proteins exhibiting significant similarity are the canalicular multiple organic anion transporter MRP2 (45%), EBCR (42%), Caenorhabditis elegans MRP1 and MRP2 (39%), the sulfonylurea receptor SUR1 (31%), and Saccharomyces cerevisiae YCF1 (37%) and BAT1 (31%). After we had completed this analysis, the full-length sequences for human MRP3 (20) and the rat MLP2 (9), which is identical to rMRP3, were published.

Like other members of the MRP subfamily of ABC-type proteins, rMRP3 is asymmetric and has an amino-terminal polytopic transmembrane domain that is significantly larger than the corresponding region downstream of the NBD. Computer-aided analysis of the deduced amino acid sequence by Kyte-Doolittle hydropathy plot and algorithms that predict transmembrane segments (TmPred, TMAP, and PHD) suggested that rMRP3 contains 10–12 putative membrane-spanning helices in the amino terminus and 4–6 transmembrane segments in the carboxy region. The hydropathy plots of the rMRP3 and MRP1 are almost identical, and structural studies reveal that MRP1 exhibits a 6+5 arrangement of membrane-spanning helices upstream of the first NBD and six transmembrane helices between the two NBDs (18). The amino terminus of MRP1 is extracellular and glycosylated (8); PROSITE analysis of the rMRP3 sequence indicated that it contains a potential N-linked glycosylation site in the amino terminus and a second site downstream of the first NBD that match the position of glycosylation sites in MRP1. Based on the amino acid sequence analysis and similarity with MRP1, a putative membrane topology for rMRP3 is diagrammed in Fig.3.

Fig. 3.

Hypothetical membrane topology of rMRP3. A: Kyte-Doolittle hydropathy plots of rMRP3 and MRP1 highlight the marked similarity in structure between the 2 proteins. B: diagram indicating location of putative structural features on the rMRP3 amino acid sequence. Transmembrane segments are indicated by dark bars and nucleotide-binding domains are indicated as shaded boxes labeled NBD1 and NBD2. C: a model suggesting a hypothetical membrane topology for rMRP3 that is based on the structural analysis performed on MRP1. These studies indicated an extracellular amino terminus, 6+5 transmembrane segments in the amino-terminal polytopic membrane-spanning domain, and 6 membrane helices in the carboxy terminus. Putative polysaccharide decorations on consensus asparagine glycosylation sites exposed to the extracellular milieu are indicated.

rMRP3 mRNA levels change upon birth and during obstructive cholestasis.

Little or no rMRP3 transcript was detected in livers of 18- or 20-day-old rat fetuses. Levels of rMRP3 RNA increased postpartum, peaked at 5–7 days after birth, and then decreased slowly to levels observed in adults (Fig.4 A). The MRP2 RNA species was also low in fetuses and increased postpartum. However, the rate of increase in MRP2 mRNA, when compared with levels present in adults, was markedly slower than was observed for rMRP3.

Fig. 4.

rMRP3 mRNA levels increase in liver early in development and after bile duct ligation. Autoradiograph of Northern blots containing 5 μg of poly(A)+ RNA prepared from 20-day-old rat fetuses and rat pups at different times after birth (A) and RNA derived from livers and intestine of rats that have undergone ligation of the common bile duct (BDL) for 24 h and 72 h and from sham-operated animals. (B) Equivalent amounts of RNA from livers of control and TR rats were also included. Blots were hybridized with a radioactively labeled rMRP3 probe, stripped, and rehybridized to an MRP2 probe. The different mRNA species that hybridize to the MRP2 probe are generated by alternative processing of the 3′-untranslated region of the primary transcript (13). Hybridization to a GAPDH probe was used to show comparable loading of RNA. These results are representative of several experiments.

rMRP3 expression is markedly increased in liver of MRP2-deficient TR rats and EHBR, suggesting that accumulation in the hepatocyte of a compound normally secreted into bile may be responsible for the increase in rMRP3 mRNA levels. An initial test of this hypothesis is the response of rMRP3 to ligation of the common bile duct. Northern blot analysis revealed that there was a severalfold increase in rMRP3 mRNA 24 h after bile duct ligation (Fig.4 B). The high levels were maintained after 72 h, 5 days, or 7 days of bile duct ligation (not shown). This effect appears to be specific to the liver; no change in rMRP3 was observed in intestine or lung of bile duct-ligated rats, the two other rat tissues that display significant expression of rMRP3.

rMRP3 is present in the canalicular membrane.

Antibodies were raised against residues 876–948 of rMRP3, which reside immediately downstream of the first NBD and represent the variable linker region of rMRP3. This region is 62% identical in rat and human MRP3 but exhibits little similarity with the MRP1-,MRP2-, SPGP-, or MDR-encoded P-glycoproteins. The antisera recognized a polypeptide with a relative molecular mass of 190–200 kDa in immunoblots of CMV proteins purified from rat liver. No detectable cross-reactive material was detected in SMV preparations (Fig.5 A). Antibody staining of the 200-kDa protein was blocked by the GST-MRP3 fusion peptide but not by GST alone. The c219 monoclonal antibody, which cross-reacts withMDR-encoded P-glycoproteins, stains a 170-kDa band in CMV proteins that is not detected in SMV. c219 staining was not affected by GST-MRP3 or GST. The molecular mass of 169 kDa predicted for rMRP3 from the deduced amino acid sequence is smaller than the apparent molecular mass of 200 kDa observed in SDS-PAGE. Treatment of CMV proteins with peptide N-glycosidase F (PNGaseF), which cleaves glycoprotein asparagine-linked oligosaccharides (39), decreased the apparent molecular mass of rMRP3 to a size more in agreement with that deduced for the cDNA sequence (Fig. 5 B). The 190- to 200-kDA signal was also detected in CMV purified from TR rats and rats that had undergone bile duct ligation (Fig. 5 C). Rat liver subcellular fractions greatly enriched in lysosomal, mitochondrial, endoplasmic reticulum, or cytosolic proteins exhibited no detectable immunostaining with the MRP3 antibodies (not shown).

Fig. 5.

Subcellular localization of rMRP3 in plasma membrane fractions purified from rat liver. Immunoblots of canalicular membrane vesicle (CMV) and sinusoidal membrane vesicle (SMV) proteins separated by SDS-PAGE. A: affinity purified anti-MRP3 IgG (30 ng/ml) recognizes a 190- to 200-kDa protein in CMV but not SMV (left). Preincubation of the antibody with 0.5 μg/ml of purified glutathioneS-transferase (GST)-MRP3 essentially abrogates binding of the antibody to the 200-kDa peptide (middle). Substitution of GST alone for the fusion peptide has no effect on antibody binding (not shown). The blot was stripped and incubated with the c219 monoclonal antibody that recognizes MDR-encoded P-glycoproteins (right). B: immunoblot stained with anti-MRP3 IgG showing that incubation of SDS-solubilized CMV proteins with peptide N-glycosidase F for 4 h (+PNGaseF) results in a 30- to 40-kDa reduction in the apparent molecular mass of rMRP3 relative to untreated (control) or mock treated (−PNGaseF) CMV. C: immunoblot stained with anti-MRP3 IgG showing CMV and SMV proteins derived from TR mutant rats, bile duct-ligated, and sham-operated animals. In all cases, the 200-kDa protein is detected in CMV, but not SMV, fractions. In all blots, CMV and SMV lanes contained 50 μg of protein.

Fluorescence immunomicroscopy of rat liver cryosections indicated that rMRP3 colocalizes with canalicular ABC-type proteins recognized by the c219 antibody (see Fig. 6). Strong staining of rMRP3 was observed in the canaliculus of TR and bile duct-ligated rats; less intense staining was detectable in liver sections derived from sham-operated animals. Sparse lateral staining was observed with anti-MRP3 antibodies, particularly in tissues derived from bile duct-ligated animals; however, the strongest signal originated from the canaliculus. No canalicular staining was observed in liver sections incubated with preimmune IgG or when anti-MRP3 IgG was preincubated with the GST-MRP3 fusion protein. Preincubation with GST alone had no effect on immunostaining, and c219 staining was unaffected by GST-MRP3 or GST.

Fig. 6.

rMRP3 resides primarily in the canalicular membrane of hepatocytes. Immunofluorescence confocal microscopy images of frozen rat liver sections prepared from TR rats, animals that had undergone bile duct ligation for 72 h (BDL), or animals that had been sham operated (control). Sections were stained with affinity-purified anti-MRP3 IgG and c219 monoclonal antibody, which recognizes MDR P-glycoproteins. Bound antibodies were labeled with goat anti-rabbit IgG-Texas red (rMRP3; left) and goat anti-mouse IgG-Cy2 (c219; middle) and were visualized by fluorescence confocal microscopy. Superimposition of the 2 images (rMRP3 + c219;right) indicates colocalization of rMRP3 and P-glycoproteins in yellow.


The rMRP3 gene encodes a protein that exhibits significant amino acid sequence similarity with the glutathione conjugate pumps MRP1 (4) and MRP2 (2, 12, 31) of mammals and YCF1 (37) and Atmrp1 (25) of yeast and plants, respectively. Analysis of the rMRP3 primary sequence outlines a putative membrane topology similar to that determined empirically for MRP1, suggesting that rMRP3 has an extracellular amino terminus, 11 membrane-spanning helices in the amino-terminal region, and 6 transmembrane domains between the two NBDs. rMRP3 contains several consensus N-linked glycosylation sites that match the location of sites empirically identified in MRP1 (8). Treatment of membrane extracts with endoglycosidase reduces the apparent molecular mass of the protein that reacts with anti-MRP3 antibodies, suggesting that rMRP3 undergoes a pattern of decoration with oligosaccharides similar to that of MRP1.

The rMRP3 transcript is present at low levels in normal adult liver but is elevated in TR rats and EHBR, suggesting that accumulation of an MRP2 substrate, or an event secondary to this increase, regulates rMRP3 expression. Supporting this hypothesis is the observation that rMRP3 mRNA levels increase rapidly during obstructive cholestasis. Many changes occur in hepatocytes after bile duct ligation that may induce the rMRP3 RNA increase. However, bile duct ligation is accompanied by accumulation in the hepatocyte of MRP2 substrates, a reduction of MRP2 RNA levels in liver, and almost complete disappearance of MRP2 from the canalicular membrane (40). The connection between the developmental patterns of MRP2 and rMRP3 expression in newborn liver is less clear but still consistent with the hypothesis. rMRP3 RNA accumulates rapidly after birth and then declines to adult levels by day 15. The MRP2 transcript undergoes a slow, steady increase in accumulation that has not reached adult levels by day 15. However, the increase in MRP2 function by the seventh day after birth may be sufficient to clear inducing substrates from the hepatocyte and lead to decreased rMRP3 expression.

Immunofluorescence confocal microscopy and immunoblot analyses of subcellular liver membrane fractions indicate that rMRP3 is present in the canalicular membrane, suggesting that it may play a role in bile formation. However, the function of rMRP3 in liver is unknown. Eukaryotes harbor multiple homologues of MRP-like proteins. A complete inventory of S. cerevisiae ABC-type proteins identifies six different genes (6), Arabidopsis thaliana contains at least three (34), and mammals express eight or more (MRP1 throughMRP6, SUR1, and SUR2). Some of the MRP subfamily members exhibit broad substrate specificities and transport a wide variety of amphiphylic organic anions; MRP1 (22) and MRP2 (15, 26) transport many glutathione and glucuronide conjugates, as do the yeast YCF1 (23) and plant Atmrp1 (25) and Atmrp2 (24). However, the homologues do not behave identically; MRP1 and MRP2 display different affinities for specific substrates (15), as do the plant homologues Atmrp1 and Atmrp2 (24). MRP2-deficient TRrats and EHBR secrete compounds into bile that are related to MRP2 substrates and include unconjugated bilirubin (30), sulfate-conjugated drugs (38), and others (14). rMRP3 may transport one or more of these compounds.

Not all MRP subfamily members transport glutathione and glucuronide conjugates. The yeast MRP-like protein BAT1p mediates ATP-dependent transport of bile salts that include taurocholate, glycocholate, and taurochenodeoxycholate (29). Bile salt secretion is the principal force driving bile flow, and there is evidence suggesting that the canalicular membrane contains a bile salt transporter in addition to the SPGP taurocholate transporter. Isolated CMV display robust ATP-dependent transport of glycocholate; however, insect cells heterologously expressing SPGP exhibit no detectable glycocholate transport (7). Moreover, the polarized WIF-B hepatoma-fibroblast fusion cell line (10) exhibits vectorial transport and accumulation of fluorescently labeled glycocholate in the canalicular vacuoles (1) that form between adjacent cells and that contain canalicular transport ATPases (27). However, WIF-B cells contain no detectable SPGP mRNA but exhibit high levels of rMRP3 mRNA (data to be published elsewhere). rMRP3 is similar in sequence and structure to the yeast bile acid transporter BAT1p and may mediate transport of a subset of bile acids that are poor substrates for SPGP.


We thank Fernado Agarraberes for help and suggestions in staining and visualization of immunoconfocal fluorescence images.


While this study was in press, Konig et al. (Hepatology 29: 1156–1163, 1999) reported localization of human MRP3 to the basolateral hepatocyte domain. Further studies are required to investigate possible species differences and specificity of the immunologic observation.


  • Address for reprint requests and other correspondence: D. F. Ortiz, Dept. of Physiology, Tufts School of Medicine, 136 Harrison Ave., Boston, MA 02111 (E-mail: dortiz{at}

  • This work was supported by the American Liver Foundation Bobby Banks Scholar Award and National Institutes of Health Grants DK-51005 and DK-35652.


View Abstract