Expression of the multidrug resistance proteins MRP2 and MRP3 in human hepatocellular carcinoma

Treatment of hepatocellular carcinoma (HCC) by chemotherapy is often impeded by the intrinsic multidrug resistance (MDR) of this frequent primary cancer of the liver. The MDR phenotype can be caused by ATP‐dependent export of chemotherapeutic drugs across the plasma membrane being mediated by transporters of the MDR P‐glycoprotein family or of the multidrug resistance protein (MRP) family. To elucidate the role of MRP family members in HCC, we analyzed the expression and subcellular localization of MRP1 (ABCC1), MRP2 (ABCC2) and MRP3 (ABCC3); all 3 isoforms have been shown to confer resistance to chemotherapeutic drugs. Semiquantitative RT‐PCR demonstrated that MRP2 and MRP3 mRNA expression in HCC was at least 10‐fold higher than MRP1 mRNA expression. MRP2 immunostaining was observed in 87% (33/38) of HCC samples. MRP2 was localized in the plasma membrane in a polarized fashion, either in trabecular structures resembling the canalicular membrane or in the luminal membrane when cells had a pseudoglandular arrangement. MRP3 was detected in all samples examined (9/9) by RT‐PCR and by immunofluorescence microscopy. MRP3 was localized to the basolateral membrane of carcinoma cells. Double‐label immunofluorescence microscopy with antibodies specific for MRP2 or MRP3 indicated that carcinoma cells expressed both MRP isoforms simultaneously. When MRP1 was detected by immunofluorescence microscopy, it was localized on the intracellular membranes of carcinoma cells. Thus, plasma membrane expression of MRP2 and MRP3, but not of MRP1, can contribute to the MDR phenotype of HCC. © 2001 Wiley‐Liss, Inc.

Successful treatment of human hepatocellular carcinoma (HCC) with chemotherapeutic drugs is often hampered by a marked drug resistance of this primary liver cancer. Cancer cells either are insensitive to drug treatment at the onset of therapy (intrinsic drug resistance) or acquire resistance after the first treatment with a chemotherapeutic agent (acquired drug resistance). This multidrug resistance (MDR) can be caused by expression of plasma membrane transporters belonging to the MDR P-glycoprotein family (ABCB) or the multidrug resistance protein (MRP) family (ABCC). These transporters mediate the ATP-dependent efflux of chemotherapeutic drugs out of cells. MDR1 P-glycoprotein exports hydrophobic, mostly basic, cytostatic drugs out of cells in an ATP-dependent manner, thus leading to MDR. 1 Expression of MDR1 mRNA 2 and MDR1 P-glycoprotein in HCC has been demonstrated. 3 However, inhibitors of MDR1 P-glycoprotein, such as verapamil, have not improved the outcome of HCC treated by chemotherapy in combination with this modulatory drug. 4 Studies on cell lines derived from HCC indicated that the MDR phenotype is attributable not only to expression of the MDR1 P-glycoprotein gene but also to other mechanisms. 5 Members of the MRP family, which are distinct from the MDR P-glycoproteins, are integral membrane glycoproteins, some of which confer resistance to chemotherapeutic drugs. 6,7 The first cloned member of this family, MRP or MRP1 (ABCC1), was isolated from a small-cell lung-cancer cell line 8 and demonstrated to confer resistance to a number of drugs, including doxorubicin, vincristine and etoposide. 9 MRP1 is expressed in the plasma membrane of many different cell types; 10,11 however, significant amounts of MRP1 mRNA were not detected in human liver. 8,12 In hepatocytes, additional MRP isoforms have been identified. 13 MRP2 was cloned from rat and human liver 14 -16 and localized to the hepatocyte canalicular membrane. 14,17 Recombinant MRP2, expressed in polarized cells, confers resistance to cisplatin, anthracyclines and etoposide. 18 Another hepatic MRP isoform, MRP3, was cloned from human liver and localized to the basolateral membrane of hepatocytes. 19,20 MRP3 also confers resistance to chemotherapeutic agents such as etoposide and methotrexate. 20 In this study, we demonstrate by semi-quantitative RT-PCR that MRP2 and MRP3 mRNA expression in HCC was at least 10-fold higher than MRP1 mRNA expression. MRP2 and MRP3 were localized to the plasma membrane of these carcinoma cells, whereas MRP1 was expressed only on the intracellular membranes of some HCCs. Both MRP2 and MRP3 may thus contribute to the intrinsic MDR phenotype of this frequent type of primary liver cancer.

Antibodies and reagents
The EAG5 antibody 17,21 was raised in rabbits against the 12 amino acids (EAGIENVNSTKF) at the carboxy terminus of the human MRP2 sequence. 14 The FDS antibody was raised in rabbits against the 24 C-terminal amino acids (FDSPANLIAARGIFYG-MARDAGLA) and the ALL antibody against the 20 -amino acid peptide (ALLIEDTLSNHTDLTDNDPV) corresponding to amino acids 876 -895 of the human MRP3 sequence. 19 The monoclonal antibody (MAb) QCRL1 against MRP1 22 and the MAb M 2 III-6 against MRP2 23 were from Alexis (San Diego, CA).
The protease inhibitors pepstatin and PMSF and the protein standard mixture (M r 26,600 -180,000) for SDS-PAGE were from Sigma (St. Louis, MO). RNase inhibitor (RNasin), StrataScript Moloney murine leukemia virus reverse transcriptase, Taq DNA polymerase, restriction enzymes and ␤-actin primers were from Stratagene (La Jolla, CA). Lysozyme and ampicillin were from Roche (Indianapolis, IN). All other chemicals were of analytical grade and obtained either from Merck (Darmstadt, Germany) or Sigma.

Human tissue samples
Tissues of 38 HCCs were from patients undergoing partial hepatectomy or liver explantation because of histopathologically confirmed primary cancer of the liver. Four patients had received chemotherapy prior to surgery (samples 9, 12, 24 and 27 in Table  I). In some cases, corresponding noncancerous liver tissue was also obtained. Structurally normal liver tissue was obtained from noncancerous tissue of 5 livers explanted for cholangiocellular carcinoma. After resection, cancerous and noncancerous liver tissues were separated, snap-frozen immediately in liquid nitrogen and stored at -80°C before further processing. For immunohistochemic and immunofluorescence analyses, cryosections (4 -5 m) were prepared with a cryotome (Leica, Bensheim, Germany), air-dried overnight and fixed in precooled (-20°C) acetone or methanol, respectively. Tissue sections were stored at -20°C until use for immunohistochemistry or immunofluorescence microscopy.

Immunohistochemic analysis and histopathology
Immunohistochemic analysis of MRP2 expression was performed by the avidin-biotin-peroxidase technique using the polyclonal EAG5 antibody and the vectastain ABC Elite detection system (Vector, Burlingame, CA) with peroxidase as the enzyme and 3-amino-9-ethylcarbazole as the chromogen. After thawing the tissue sections, endogenous biotin was blocked. Sections were then incubated for 60 min at room temperature with the EAG5 antibody and subsequently with the secondary, biotinylated antirabbit antibody. After each antibody incubation, unbound antibodies were removed by serial washing with PBS. Sections were then incubated with the freshly prepared avidin-biotinylated enzyme complex solution. After another washing step, the chromogen solution was added. The enzyme reaction was microscopically controlled and stopped with distilled water when optimal contrast was reached, at the latest after 30 min. Nuclei were then stained with Mayer's hemalum (Merck). For diagnostic histopathology, tissue sections were stained with hematoxylin and eosin. Specimens were analyzed by a reference pathologist (WJH). Cancerous tissue was graded according to Edmondson and Steiner: 24 I, well-differentiated; II, moderately differentiated; III, poorly differentiated; IV, undifferentiated.

Immunofluorescence microscopy
Sections were rehydrated in PBS after thawing, incubated with primary antibodies for 60 min, washed 3 times for 10 min with PBS and then incubated with the combined secondary antibodies for 60 min. After washing 3 times with PBS, sections were rinsed in distilled water, air-dried and mounted in Moviol (Hoechst, Frankfurt, Germany). All antibodies were diluted with PBS containing 5% FCS at the following concentrations: EAG5, FDS and M 2 III-6 at 1:100; ALL and QCRL1 at 1:40; Cy2-conjugated antirabbit IgG and Cy3-conjugated antimouse IgG (Jackson Immunoresearch, West Grove, PA) at 1:400. Fluorescence imaging micrographs were taken with a digital video camera (Hamamatsu, Hamamatsu, Japan) on an Axiovert S100TV microscope (Carl Zeiss, Jena, Germany). Pictures were analyzed using Openlab imaging software (Improvision, Coventry, UK).

RNA isolation, reverse transcription and semiquantification of mRNA expression
Total RNA was isolated from human HCC tissue (approx. 100 mg) in RNA clean solution according to the manufacturer's instructions (Hybaid, Franklin, MA). Total RNA (5 g) was reversetranscribed using 5 g oligo(dT) 18 primer as described. 19 Synthesized sscDNAs were purified using Microcon (Bedford, MA) 100 filter units.
MRP1, MRP2 and MRP3 mRNA expression in relation to ␤-actin mRNA expression was determined using the LightCycler System and the FastStart DNA Master SYBR Green I kit (both from Roche). PCRs were performed according to the manufacturer's instructions with 4 mM MgCl 2 , 0.5 M of the respective sense and 0.5 M of the respective antisense primers and 1-fold LightCycler-FastStart DNA Master SYBR Green I mix in a total volume of 20 l. Cycling conditions were as follows: 10 min denaturation at 94°C, followed by 50 cycles of 10 sec denaturation at 94°C, 15 sec primer annealing at 68°C and 45 sec of fragment elongation at 72°C. For semiquantitative PCR analysis, the following primers were used: sense primer omrp1/3ЈRT.for (5Ј-CTGACAAGCTA-GACCATGAATGT-3Ј) and antisense primer omrp1/3ЈRT.rev (5Ј-TCACACCAAGCCGGCGTCTTT-3Ј) for detection of MRP1   mRNA. For ␤-actin amplification, a commercially available ␤-actin primer pair was used (Stratagene). The amount of ␤-actin, MRP1, MRP2 and MRP3 single-stranded cDNA was determined using a serial plasmid dilution (human MRP2 cDNA in the expression vector pcDNA3.1, 18 from 1 ϫ 10 4 to 1 ϫ 10 2 fg) as amplification standard. The ␤-actin mRNA concentration, calculated in relation to this standard curve, was set to 100% and the respective MRP mRNA value is given as a percentage of ␤-actin amplification.

Immunolocalization of MRP2 and MRP3 in HCC
MRP2 was localized in human liver and in corresponding HCC by immunohistochemistry and immunofluorescence microscopy (Figs. 1,2). The EAG5 antibody gave a signal exclusively in the canalicular membrane of hepatocytes in normal and cirrhotic liver (Figs. 1a,2a). The MAb M 2 III-6 was also used for immunofluorescence microscopy. Both antibodies, EAG5 (Fig. 2a) and M 2 III-6 ( Fig. 2b), stained identical structures, i.e., the canalicular membrane. MRP2, analyzed by immunohistochemistry, was expressed in 33/38 HCC samples (87%, Table I). There was no apparent correlation between the number of cells expressing MRP2 and tumor grade. In HCC cells, MRP2 was localized in a strictly polarized fashion, either yielding "spider-like", canalicular structures in trabecular HCC (Figs. 1b, 2c,d,f) or covering the luminal membrane of tumor cells in pseudoglandular HCC (Figs. 1c,2e).
MRP3 localization in human liver and corresponding HCCs was analyzed by immunofluorescence microscopy using the polyclonal FDS antibody (Fig. 3a-d,f). In addition, the polyclonal ALL antibody, which is also directed against MRP3, 19 was used in some cases (Fig. 3e). In normal human liver, only weak expression of MRP3 was observed (Fig. 3a). Double-label immunofluorescence microscopy with the polyclonal FDS antibody and the M 2 III-6 MAb showed that MRP3 and MRP2 were simultaneously expressed in the carcinoma cells (Fig. 3b). In all 9 of the 9 HCC samples analyzed (Table I), MRP3 was detected in the basolateral membrane of HCC cells by either FDS or ALL staining.

Immunoblot analysis of MRP2 and MRP3 in HCC
In addition to immunolocalization, MRP2 and MRP3 expression in crude membranes of some HCCs was analyzed by immunoblotting (Fig. 4). The EAG5 antibody indicated expression of the 190 kDa MRP2 glycoprotein in all tissue samples. The same samples were analyzed for MRP3 expression using FDS directed against human MRP3. MRP3 was detected in all tissue samples tested. The double-band pattern for MRP3, with a 190 and a 170 kDa band, was also seen in normal human liver and may be attributable to at least 2 different splice variants. 19 Using the QCRL1 antibody for MRP1, no bands were detected in the 4 tumor samples analyzed, indicating low, if any, MRP1 expression.

Semiquantitative analysis of MRP1, MRP2 and MRP3 mRNA expression in human liver and HCC
MRP mRNA expression was analyzed by semiquantitative RT-PCR standardized for expression of ␤-actin mRNA (Table II, Fig.  5). Therefore, a standard curve was created by amplifying fragments on known template concentrations. The amount of ␤-actin mRNA in the RT reaction was calculated in relation to this standard curve and the amount of MRP mRNA is given as a percentage of ␤-actin mRNA. The amplification products had the expected size of 350 bp for MRP1, 284 bp for MRP2 and 452 bp for MRP3 (Fig. 5). As an internal control and for standardization, a ␤-actin fragment of 661 bp was amplified. cDNA fragments were subcloned and sequenced. Semiquantitative analysis showed expression of MRP3 mRNA in 7/7 HCCs examined and in the corresponding liver samples (Table II). MRP2 mRNA was also detected in cancerous as well as benign samples. Whereas the relative MRP2 to MRP3 mRNA expression ratio was about 16, the  b,d,f), which was raised against the 202-amino acid C terminal part of human MRP2. 23 The same tissue section (cirrhotic, noncancerous tissue corresponding to sample 1, Table I) was double-stained with the EAG5 (a) and M 2 III-6 (b) antibodies. Both antibodies stained identical structures, i.e., the canalicular (apical) membranes. Scale bars ϭ 20 m.  (Table I) showed fluorescent basolateral signals with the FDS and ALL antibodies (c-f). Scale bars ϭ 20 m.
relative MRP1 mRNA expression ratio was about 0.6, indicating that there are about 25 times more transcripts for MRP2 and MRP3 than for MRP1.

Immunolocalization of MRP1 in HCC
Because low expression of MRP1 mRNA was detected by RT-PCR (Table II), MRP1 expression was analyzed by immunofluorescence microscopy using the QCRL1 antibody. In benign liver tissue, no staining of hepatocytes was observed (Fig. 6a). Similarly, in HCC samples 1, 7 and 28, no MRP1 expression was detected in carcinoma cells. In sample 9 (Fig. 6c), several carcinoma cells showed intracellular staining, indicating expression of MRP1 on intracellular membranes. A similar staining pattern was observed in sample 27 (data not shown). In sample 13, almost 90% of carcinoma cells showed expression of MRP1 localized to intracellular membranes rather than to the plasma membrane (Fig. 6e). DISCUSSION HCC is one of the most common cancers in the world. 26 Chemotherapy, either with a single agent or with a multidrug regimen, does not prolong life in most cases 27 and is thus not a promising option in the treatment of this primary cancer of the liver. 28 At first, expression of MDR1 P-glycoprotein was suggested to cause MDR in HCC. However, evidence has not been obtained to prove that inhibitors of MDR1 P-glycoprotein can overcome MDR in HCC. 4 This suggested additional mechanisms of conferring drug resistance. The MRP family of ATP-dependent drug-export pumps has been elucidated and some of its members shown to confer drug resistance. 7,29 In the present study, we describe that the isoforms MRP2 and MRP3 were expressed in the plasma membrane of HCC cells, possibly contributing to the intrinsic MDR phenotype of HCC.
The founding member of the MRP family, MRP1, was cloned from the drug-resistant small-cell lung-cancer cell line H69AR 8 and shown to confer resistance to doxorubicin, vinblastine and etoposide. 9 MRP1 is overexpressed in a number of cancer types and contributes to MDR in these malignancies. 6 In HCC, however, MRP1 is most likely not involved in the MDR phenotype. MRP1 mRNA expression, detected by RNase protection assay, is very low in liver. 8,12 This result was confirmed in the present study by quantitative RT-PCR (Table II). MRP1 mRNA expression was at least 10-fold lower than MRP2 or MRP3 mRNA expression. MRP1 was not detectable in hepatocytes by immunofluorescence (Fig. 6). Other cells in the liver, e.g., mast cells, 30 are therefore a likely source of MRP1 mRNA in human liver. In some HCC samples, MRP1 expression was detected on intracellular membranes (Fig. 6); however, plasma membrane staining of these carcinoma cells was not observed. Whether MRP1 is functional in transiently sequestering drugs into the lumen of these intracellular vesicles remains to be seen.
Because of the similar substrate specificities of MRP1 and MRP2, 7 it was suggested that MRP2 also confers drug resistance. This suggestion was supported by transfection of MRP2 antisense  (Table I). With the MRP1-specific QCRL1, no signal was detected, indicating low, if any, MRP1 expression in these HCC samples. Values represent expression of the respective MRP isoform as a percentage of ␤-actin mRNA expression. Noncancerous, benign tissues of samples 1, 13, 14, 27 and 28 were diagnosed as cirrhotic.  18 primer yielding single-stranded cDNAs. For PCR analysis, 3 different primer pairs were used, which were specific for human MRP1 mRNA (350 bp fragment), MRP2 mRNA (284 bp fragment) and MRP3 mRNA (452 bp fragment). A 661 bp fragment for ␤-actin generated with specific primers for the human sequence demonstrated integrity of the isolated RNA and was used for standardization of MRP mRNA expression (Table II). RNA into HepG2 cells, which decreased MRP2 protein and increased sensitivity to cisplatin. 31 MDR conferred by MRP2 was directly demonstrated in cells expressing recombinant MRP2. 18 Additional results confirm the transport of chemotherapeutic agents such as methotrexate, cisplatin and camptothecin by MRP2. [32][33][34][35][36] Resistance to etoposide and methotrexate conferred by MRP3 was shown after expression of recombinant MRP3 in polarized MDCK 20 and HEK 37 cells.
Our findings of high levels of MRP2 and MRP3 mRNA expression in relation to MRP1 mRNA expression (Table II) and our immunofluorescence data (Table I) indicate that MRP2 and MRP3 are the MRP isoforms that contribute to the resistance of HCC to cytotoxic agents. MRP2 expression was restricted to the apical membrane in normal liver 17 as well as in HCC (Figs. 1,2). MRP2 was localized in the apical membrane of these carcinoma cells, yielding either a spider-like expression pattern in trabecular HCC or a circular expression pattern in pseudoglandular HCC, with MRP2 localized to the cell membrane facing the lumen of the pseudoglandular structures sometimes filled with bile (Figs. 1,2). The site of MRP2 expression in HCC cells thus corresponds to the expression site in normal hepatocytes where the pump exerts its normal function. This suggests a similar function of MRP2 in HCC cells, i.e., extrusion of substrates including chemotherapeutic drugs out of the cell. In comparison, renal clear-cell carcinomas lacked a distinct apical-to-basolateral tumor cell polarity and MRP2 appeared in part on intracellular membranes. 38 In some poorly differentiated HCCs, no trabecular or luminal structures were observed; and these tumors lacked immunohistochemically detectable MRP2 (Fig. 1d). In carcinoma cells, MRP3 was restricted to the basolateral membrane, as was MRP3 in hepatocytes of normal human liver. 19,20 The site of MRP3 expression in carcinoma cells thus corresponds to its normal localization. Therefore, in HCC cells, an MRP3 function similar to that in normal cells, i.e., export of substances into blood, is suggested.
Cancer cells are often resistant to chemotherapeutic drugs at the onset of therapy. MRP2 may account, at least in part, for this intrinsic drug resistance of HCC because it is expressed in normal liver 12,17 as well as in benign, cirrhotic tissue (Table II). Other normal human tissues in which MRP2 is expressed include duodenum and kidney. 12,38 The apical isoform MRP2 functions there as an export pump for anionic conjugates. 39 In contrast to MRP2, MRP3 expression is low in normal human hepatocytes. 19 However, cholestasis, which may also occur in HCC, can cause upregulation of MRP3 in rats 40 and probably in humans. 19 In conclusion, MRP2 and MRP3 are expressed in the apical (luminal) and basolateral plasma membrane domains of HCC cells, respectively. We propose that both MRP isoforms contribute to the intrinsic resistance of HCC to a wide variety of cytotoxic, chemotherapeutic agents. The development of inhibitors directed against MRP1, 41 which is hardly expressed in HCC, is far advanced. When substances become available for the inhibition of MRP2 and MRP3, they will provide improved efficacy in the chemotherapy of this frequent malignant tumor. -Immunofluorescence microscopy of MRP1 in benign, cirrhotic tissue (a) and in HCC (c,e) using the QCRL1 MAb. 22 As a control, MRP2 was visualized with the polyclonal EAG5 antibody (b,d,f). Whereas hepatocytes do not express MRP1, intracellular MRP1 staining was observed in samples 9 (c) and 13 (e). Arrowheads point to cells that simultaneously express MRP1 and MRP2. Scale bars ϭ 20 m.