Identification and characterization of a hepatoma cell-specific enhancer in the mouse multidrug resistance mdr1b promoter.

The expression of multidrug resistance/P-glycoprotein genes mdr1b(mdr1) and mdr1a(mdr3) is elevated during hepatocarcinogenesis. To investigate the regulation of mdr1b gene expression, we used transient transfection expression assays of reporter constructs containing various 5′-mdr1b flanking sequences in hepatoma and non-hepatoma cells. We found that nucleotides −233 to −116 preferentially enhanced the expression of reporter gene in mouse hepatoma cell lines in an orientation- and promoter context-independent manner. DNase I footprinting using nuclear extracts prepared from hepatoma and non-hepatoma cells identified four protein binding sites at nucleotides −205 to −186 (site A), −181 to −164 (site B), −153 to −135 (site C), and −128 to −120 (site D). Further analyses revealed that, while site B alone played a major part for the enhancer function, sites A and B combined conferred full enhancer activity. Site-directed mutagenesis results also supported these results. Gel retardation experiments using oligonucleotide competitors revealed that the site B contains a dominant binding protein. This is the first report demonstrating a cell type-specific enhancer in the mdr locus. The role of this enhancer in the activation of mdr1b gene during hepatocarcinogenesis is discussed.

The expression of multidrug resistance/P-glycoprotein genes mdr1b(mdr1) and mdr1a(mdr3) is elevated during hepatocarcinogenesis. To investigate the regulation of mdr1b gene expression, we used transient transfection expression assays of reporter constructs containing various 5-mdr1b flanking sequences in hepatoma and non-hepatoma cells. We found that nucleotides ؊233 to ؊116 preferentially enhanced the expression of reporter gene in mouse hepatoma cell lines in an orientation-and promoter context-independent manner. DNase I footprinting using nuclear extracts prepared from hepatoma and non-hepatoma cells identified four protein binding sites at nucleotides ؊205 to ؊186 (site A), ؊181 to ؊164 (site B), ؊153 to ؊135 (site C), and ؊128 to ؊120 (site D). Further analyses revealed that, while site B alone played a major part for the enhancer function, sites A and B combined conferred full enhancer activity. Site-directed mutagenesis results also supported these results. Gel retardation experiments using oligonucleotide competitors revealed that the site B contains a dominant binding protein. This is the first report demonstrating a cell type-specific enhancer in the mdr locus. The role of this enhancer in the activation of mdr1b gene during hepatocarcinogenesis is discussed.
The role of the multidrug resistance transporter P-glycoprotein (P-gp) 1 in the development of resistance to a wide variety of lipophilic compounds in animal cells has been well documented (for reviews, see Refs. [1][2][3]. P-gp is a membrane protein containing multiple transmembrane domains and two intracellularly localized nucleotide binding sites. It is generally believed that P-gp functions as an efflux pump, through which cytotoxic lipophilic compounds are expelled. P-gp is encoded by a small gene family consisting of three members in rodents (4 -6) but only two in humans (7,8). However, only two rodent genes (mdr1a or mdr3 and mdr1b or mdr1) (9, 10) and one human gene (MDR1) (11) are related to the multidrug resist-ance function in cultured cells.
Tissue-specific expression of different mdr genes has been observed in the rodents, i.e. mdr1a in intestine and blood-brain barrier, mdr1b in kidney and adrenal gland, and mdr2 in the liver. Homozygous disruption of the murine mdr1a resulted in accumulation of cytotoxic drugs in the brain and increased drug sensitivity to the animals (12). Likewise, disruption of the murine mdr2 led to an impairment of hepatic phospholipid secretion in homozygous animals (13), consistent with the idea that mdr2 product is a phospholipid transporter (14).
Overexpression of MDR1 gene has been detected in human malignant biopsies (15), and in some cases, correlation between elevated MDR1 expression and poor response to chemotherapeutic agents has been reported (16). In animals, enhanced expression of mdr1a and mdr1b mRNA has been seen during hepatocarcinogenesis (17)(18)(19)(20)(21)(22)(23)(24). However, the underlying mechanisms for the activation remain to be determined. To explore the mechanisms that control the expression of mdr gene during hepatocarcinogenesis, we have isolated a genomic DNA containing the 5Ј portion of the mouse mdr1b. Using a transient expression assay, we report here the identification of an enhancer sequence proximal to the promoter of this gene that functions preferentially in hepatoma-derived cell lines.

EXPERIMENTAL PROCEDURES
Isolation of DNA Fragment Containing a 5Ј-mdr1b Sequence and Construction of mdr1b-CAT Chimeric Recombinants-A DNA fragment containing the 5Ј portion of mdr1b cDNA was synthesized by the reverse transcriptase-polymerase chain reaction (PCR) using oligonucleotide 1 as the left side primer and oligonucleotide 2 as the right side primer and total RNA prepared from mouse L1210/Dox 0.5 cells (25) as template. Table I shows the sequences of oligonucleotides 1 and 2 and those used in the subsequent experiments. The PCR product was nick-translated with [ 32 P]dCTP and used as a probe to screen a cosmid genomic DNA library prepared from L1210/Dox 0.5 cells (25). A positive clone designated cosmuDR11 was isolated. The restriction enzyme sites in the insert were mapped, and a 1.5-kilobase PstI-SmaI fragment containing the 5Ј-end of the mdr1b sequence was subcloned in pGEM3Z (Promega).
A fragment containing nucleotides (nt) Ϫ1069 to ϩ97 of mdr1b gene was synthesized by PCR DNA using the subcloned DNA as template, SP6 promoter sequence (in the vector) as left side primer, and oligonucleotide 3 as the right side primer. The PCR product was cloned into the PstI-HincII sites of pGEM3Z, released by XbaI/HindIII digestion, and recloned into the XbaI-SmaI sites of a vector containing the chloramphenicol acetyltransferase (CAT) gene (26), generating Ϫ1069M1CAT. A KpnI-PstI fragment was excised from Ϫ1069M1CAT and subcloned into the KpnI-PstI sites of the CAT vector to generate the first 5Јdeletion mutant, Ϫ586M1CAT. Three additional 5Ј-deletion recombinant DNA were constructed by PCR using Ϫ1069Ml3Z as template and oligonucleotide primers 4 -6 (all contain a KpnI site), each paired with SP6 promoter sequence (in the vector) as left side primer. The PCR products were digested with KpnI/PstI and inserted into the CAT vector (26), generating recombinants Ϫ233M1CAT, Ϫ158M1CAT, and Ϫ116M1CAT, respectively.
Construction of mdr1 Enhancer Subsequences in the Homologous and Heterologous Promoters-A DNA fragment containing nt Ϫ233 to Ϫ116 was synthesized using oligonucleotides 4 and 10 primers and Ϫ1069M13Z DNA as template. The fragment was cloned into pCRII TM vector (Invitrogen), released by HindIII digestion, and recloned into pBLCAT2 vector (27) in both forward and reverse orientations to generate Ϫ233/116TKCAT and Ϫ116/Ϫ233TKCAT, respectively. The same fragment with reverse orientation was recloned into the KpnI site of Ϫ116M1CAT to generate Ϫ116/Ϫ233M1CAT. Similarly, a DNA fragment containing nt Ϫ210 and Ϫ155 was synthesized by PCR using primer pair oligonucleotide 11 and oligonucleotide 12, inserted into pCRII TM intermediate, excised by BamHI, and recloned into the BamHI site of pBLCAT2 vector in both forward and reverse orientations, yielding recombinants Ϫ210/Ϫ155TKCAT and Ϫ155/Ϫ210TKCAT, respectively. To generate recombinant Ϫ162/116TKCAT, DNA fragment containing nt Ϫ162 to Ϫ116 was synthesized by PCR using oligonucleotides 15 and 10 primers, cloned into pCRII TM vector, excised by HindIII digestion, and recloned into the HindIII site of pBLCAT2 in both forward and reverse orientations.
Site-directed Mutagenesis-A total of 11 site-specifically mutagenized constructs of Ϫ233M1CAT were prepared by replacing the wild-type sequence with mutated sequences generated by PCR. Eight left side primers (oligonucleotides [22][23][24][25][26][27][28][29] and three right side primers (oligonucleotides 30 -32) were used. All of these primers contain a KpnI site (underscored) and site-specifically altered nucleotides (boldface letters) (Table I). PCR products were synthesized using Ϫ133M1CAT as a template (or Ϫ1069ML3Z for left side mutant primers), and these oligonucleotides were paired with their opposite side primers. The PCR products were digested with KpnI/PstI and cloned into the CAT vector to generate the first set of eight mutants or with KpnI and cloned into Ϫ116 MlCAT to generate the second set of three mutants. All the constructs described here were confirmed by DNA sequencing.
Cell Culture, Transfection, and CAT Assay-NIH3T3, HeLa cells, mouse hepatoma cell lines Hepalclc and Hepalclc-BprC1 (hereafter referred to as BprC1) (28), and Hepal-6 (29) were maintained in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) supplemented with 10% fetal bovine serum (Life Technologies, Inc.). Cells were transfected with CsCl-purified (2ϫ) plasmid DNA (5 g) by the polybrene method (34) using the cotransfected pH2B-␤gal plasmid DNA as an internal reference to normalize transfection efficiency for CAT expression according to the procedure described (25). The pH2B-␤gal contains a lacZ gene linked to a basal promoter from Xenopus laevis histone H2B that promotes constitutive expression in all cells. In some experiments, pCMV-␤gal plasmid DNA was used as internal control for transfection calibration. No significant difference was found between the use of these two control plasmids.
Preparation of Nuclear Extracts, Gel Mobility Shift Assays, and DNase I Footprinting Analyses-Nuclear extracts were prepared cultured cells and mouse HCC according to the procedure described by The mobility shift assays were performed essentially as described previously (32). DNase I footprinting was carried out using the protocol described by Ohisson and Ediund (33).

RESULTS
Cell Type-specific Activity of cis-Acting Regulatory Elements in the Mouse mdr1b Promoter-Various lengths of the mdr1b 5Ј-flanking sequence were inserted into a vector containing a promoterless CAT gene (Fig. 1A). These constructs were transfected into mouse hepatoma cells (Hepalclc, BprC1, and Hepa1-6) and nonhepatoma cells (NIH3T3 and HeLa). A previous study has demonstrated that these hepatoma-derived cell lines exhibited high levels of steady-state mdr1b mRNA (25), whereas NIH3T3 cells contained very low levels of mdr1b mRNA (19). The levels of mdr1b mRNA in these hepatoma cells are comparable to those in mouse kidneys, one of the organs that contain abundant mdr1b mRNA (25). Levels of mdr1b mRNA in NIH3T3 cells are comparable to those in the normal livers (19). The murine mdr1b homolog is missing in HeLa cells, which display very low levels of MDR1 and MDR2 expression. Recombinant Ϫ1069M1CAT (Fig. 1, construct a) was transfected into these cells. 2 days after transfection, the CAT activity in the transfected cells was determined. Levels of CAT expression in BprC1 cells were comparable to those in Hepalclc and Hepal-6 cells (not shown) but were about 2-and 20-fold higher than those in NIH3T3 cells and HeLa cells (not shown), respectively, using the cotransfected pH2B-␤gal as an internal control (Fig. 1B). Similar results were obtained with pCMV-␤gal, in which the expression of lacZ gene is controlled by cytomegalovirus enhancer/promoter (not shown). These results suggest that a sequence within nt Ϫ1069 can direct reporter gene expression preferentially in hepatoma-derived cells. Cell type-specific activity of mdr1b upstream sequences have also been demonstrated previously by transient transfection assay (35,36).
The levels of CAT expression from Ϫ1069M1CAT, Ϫ586M1CAT, and Ϫ233M1CAT recombinants (Fig. 1B, constructs a-c) in the transfected BprC1 cells were not signifi-cantly different. However, deleting nt Ϫ233 to Ϫ158 and further downstream (Fig. 1B, constructs d and e) resulted in a reduction of more than 90% of CAT activity. These results suggest that sequences downstream from nt Ϫ233 are important for the expression of the mdr1b gene in mouse hepatoma cells. Similar results were seen in Hepal-6 and Hepalclc hepatoma cells (data not shown), in which high levels of mdr1b expression were also seen (25).
When the same set of recombinant plasmids were transfected into NIH3T3 cells and HeLa cells, different expression patterns emerged. In the transfected NIH3T3 cells (Fig. 1B), the level of CAT activity progressively decreased as the lengths of 5Ј-mdr1b region decreased. Furthermore, only less than 20% reduction in the levels of CAT expression was found between Ϫ233M1CAT (construct c) and Ϫ158M1CAT (construct d). In the transfected HeLa cells, there was no significant difference in the CAT activity between Ϫ233M1CAT and Ϫ158M1CAT recombinants (not shown). These results suggest that positive cis-regulatory element(s) located downstream from Ϫ233 nt of the mdr1b gene functions preferentially in hepatoma cell lines.
To determine the 3Ј boundary of the cis-regulatory element, we carried out similar analysis using Ϫ233MlCAT constructs with progressive deletions from Ϫ116 to Ϫ172 (Fig. 1A, recombinants f, g, and h). Removing sequences between Ϫ116 and Ϫ130 or between Ϫ116 to Ϫ155 resulted in reduction of no more than 30% CAT activity in BprC1 cells. Removal of additional sequences, i.e. Ϫ155 to Ϫ172, resulted in reduction of 45% of activity. These results, together with those from the 5Ј-deletion assay, suggest that the sequence located between Ϫ233 and Ϫ172 plays a major role for the expression of the reporter constructs in BprC1 cells, whereas the sequence between Ϫ172 and Ϫ155 may also be contributory but comparatively less substantial.
The cis-Regulatory Element Can Function in Both an Orientation-and Promoter Context-independent Manner-To further characterize the hepatoma-related cis-regulatory element located downstream from nt Ϫ233 of the mdr1b promoter, we inserted nt Ϫ233 to Ϫ116 into Ϫ116M1CAT vector in the reverse orientation ( Fig. 2A, construct b). Transient transfection assay showed that the resultant construct exhibited about 6-fold enhancement in CAT activity, in comparison with the 9-fold increase seen in the same construct with forward orientation ( Fig. 2A, construct a). This result indicated that the cis-regulatory element located Ϫ233 to Ϫ116 can function in both forward and reverse orientations. All the constructs, regardless of either orientation, yielded very poor enhancement activity in NIH3T3 cells ( Fig. 2A), supporting the cell type specificity of the regulatory element.
To investigate whether nt Ϫ233 to Ϫ116 can enhance reporter gene expression from a heterologous promoter, we inserted this sequence in both forward and reverse orientations into pBLCAT2 (Fig. 2B, constructs d-f). This vector contains a CAT reporter gene driven by the basal promoter from the thymidine kinase (TK) gene. As shown in Fig. 2A, the forward version of construct (construct d) displayed a 10-fold increase in the CAT activity in BprC1 cells, whereas it displayed about a 2-fold increase for the reverse version (construct e). Again, all of these constructs exhibited a very low enhancement of CAT activity in NIH3T3 cells. These results indicate that nt Ϫ233 to Ϫ116 can enhance the reporter gene expression in an orientation-and promoter context-independent manner. Furthermore, this sequence showed hepatoma specificity in directing the expression of the reporter gene even in a heterologous promoter. Therefore, this sequence can be considered as an enhancer preferentially active in the hepatoma cells.
Multiple Protein Binding Sites in the Enhancer-To deter- mine whether there were specific protein binding sites in the enhancer, we carried out DNase I footprinting analysis. A DNA fragment containing nt Ϫ233 to Ϫ116 was synthesized by PCR using labeled primers at either end. The labeled DNA was incubated with nuclear extracts from BprC1 cells and digested with nuclease. Remarkably reproducible footprints (four experiments) were observed in the crude nuclear extracts, suggesting that the trans-acting factors are either in high abundance or have high affinity for their target sequences. Two DNase I protection sites, located approximately nt Ϫ205 to Ϫ186 (site A) and Ϫ181 to Ϫ164 (site B), were found in both the coding and noncoding strands (Fig. 3A). Some of these footprints were characterized by the flanking DNase I hypersensitive sites in the presence of nuclear extracts, e.g. site B in the coding strand and site A in the non-coding strand. Due to technical limitations, the borders of these sites could only be assigned with approximation. Furthermore, binding sites located on one strand were not always exactly superimposed on another strand, and the footprints detected in one strand were not equally discernible on the other.
To determine whether additional protein binding sites located downstream from Ϫ164 that could not be adequately resolved under the gel electrophoretic conditions were favorable for detection of sites A and B, we carried out footprinting analyses using end-labeled fragments spanning nt Ϫ50 to Ϫ185 for the coding strand and Ϫ210 to Ϫ80 for the non-coding strand (Fig. 3B). At least two additional protein binding sites, located approximately at Ϫ153 to Ϫ135 (site C) and Ϫ128 to Ϫ119 (site D), were detected. For the reason as mentioned above, site C could be better detected in the non-coding strand (by the characterized flanking hypersensitive sites) than in the coding strand. The reason for this is not clear but could be due to the unfavorable cleavage bias of A and C residues by the nuclease (see lanes 1 and 2 in Fig. 3B). These results demonstrated multiple protein binding sites in the mdr1b enhancer region. Fig. 3B also shows one footprint (indicated by bars) downstream from Ϫ120 nt. The functional aspect of this footprint was not further characterized.
Sites A and B Are Important for the Hepatoma-specific and Orientation-independent Enhancer Functions-To investigate whether these protein binding sites are important for the en-  1-3) or the noncoding strand (lanes 4 -6). The labeled DNA was mixed with (lanes 2, 3, 5, and 6) or without (lanes 1 and 4) BprC1 nuclear extract in the presence of poly dI⅐dC competitors and digested by DNase I. Lanes G and T contain molecular size markers generated by dideoxynucleotide sequencing ladders. The two footprints (sites A and B) in both coding and noncoding strands are shown by brackets. For detecting footprints downstream from Ϫ160, end-labeled oligonucleotides with sequence Ϫ185 to Ϫ50 (coding strand,  lanes 1-4)) and Ϫ210 to Ϫ80 (noncoding strand, lanes 5-7) were used as probes. Two footprints, C and D, are shown by brackets (panel B). Another footprint downstream from Ϫ120 is shown by bars.
hancer function and to dissect the essential domain in the enhancer, oligonucleotide sequences spanning these binding sites, either as an individual or in combination, were inserted into pBLCAT2 vector in both forward and reverse orientations. Fig. 4 shows that BprC1 cells transfected with recombinant Ϫ210/Ϫ155TKCAT (construct b), which contains both sites A and B, exhibited about 9-fold increase in CAT activity, as compared with those transfected with vector alone. This level of increase is comparable to that of the same cells transfected with Ϫ233/Ϫ116 recombinant (construct a). Even the reverse version of Ϫ210/155TKCAT (construct b) still showed 3-fold increase of CAT activity. NIH3T3 cells transfected with these recombinants showed no more than 2-fold increase in either orientations. This result suggests that sites A and B contain most, if not the entire, enhancer function seen in Ϫ233/ Ϫ116TKCAT. Recombinant Ϫ210/Ϫ185TKCAT (construct c), which contains site A only, exhibited no enhancing activity in both BprC1 and NIH3T3 cells, whereas recombinants Ϫ185/ Ϫ155TKCAT (construct d, containing site B) in both orientations displayed 7-fold increased activity in BprC1 cells lines. No enhancer activities were found with recombinants Ϫ162/ Ϫ116TKCAT, Ϫ153/Ϫ135TKCAT, and Ϫ127/Ϫ120TKCAT (constructs e through g) containing sites C and D, combined or in individual, respectively. We conclude that the entire enhancer function is mostly located in sites A and B, with site B as a major player. These results are in general agreement with those using deletion mutants as shown in Fig. 1.
To further substantiate this finding, we carried out sitespecific mutagenesis of nucleotides downstream from Ϫ233 in Ϫ233M1CAT. A total of 11 mutants were prepared (Fig. 5A). These constructs were transfected into BprC1 cells. The CAT activities in the transfected cells were determined (Fig. 5B). Mutants b through e, which contain specific mutations between nt Ϫ226 and Ϫ200, had CAT activities comparable to that of the wild-type construct (construct a). Mutants f through i, which contain mutations spanning from the 3Ј-half of site A to the 5Ј-half of site B, however, exhibited a 90% reduction of CAT activity in BprC1 cells but no reduction in NIH3T3 cells. Mutations in sites C and D (mutants j-l) exhibited minor reduction in CAT activities in BprC1 cells in relative to those found in mutant e, and no significant reduction in NIH3T3 cells. These results collaborate the notion that sites A and B are important for the enhancer function as mentioned above.
Multiple Proteins Bind to the mdr1b Enhancer-To investigate the protein(s) that recognize the enhancer sequence, we used the DNA fragment containing nt Ϫ233 to Ϫ116 (sites A through D) as a probe in gel mobility retardation assays. Nuclear extracts from BprC1 cells were mixed with 32 P-labeled probe and nonspecific competitor, poly(dI)⅐poly(dC), followed by various amounts of specific competitors containing sequences in sites A-D. DNA-protein complexes were resolved by polyacrylamide gel electrophoresis under neutral conditions and visualized by autoradiography. In the absence of competitors, a major retarded band (Fig. 6A, arrow) and several minor bands were observed from the autoradiograms. Under the conditions used, this major band could be effectively competed by a sequence containing site B but not by those containing A, C, and D. This result suggests that the major retarded band must arise by protein binding to site B. This was also evident when nt Ϫ210 to Ϫ155 (containing sites A and B) was used as probe (Fig. 6B). These results suggest that either protein recognizing site B sequence is much more abundant than those recognizing other sites or that the affinity of the protein bound to site B supersedes those to other sites. When individual sequences containing sites A-D were used as competitors to probe protein binding to nt Ϫ162/Ϫ116, which contains sites C and D, only site C sequence could compete efficiently, suggesting that protein binding to site C dominates site D (Fig. 6C). Similar results were observed when nuclear extracts from NIH3T3 (Fig. 6,   FIG. 4. CAT expression analysis of various portions of the mdr1b enhancer. CAT expression from various recombinant plasmid DNA (a-g) in BprC1 and NIH3T3 cells are shown. In all cases, F and R refer to forward and reverse versions, respectively, of the mdr constructs that were used. The structures of the plasmid DNA are schematically shown. CAT expression was normalized to the ␤-galactosidase activity using the cotransfected pH2B-␤gal plasmid DNA. mdr1b sequences in these constructs are as follows: a, Ϫ233 to Ϫ116; b, Ϫ210 to Ϫ155, c, Ϫ210 to Ϫ185; d, Ϫ185 to Ϫ155; e, Ϫ162 to Ϫ116; f, Ϫ153 to Ϫ135; and g, Ϫ127 to Ϫ120. D-F), normal mouse livers, and mouse HCC (not shown) were used. These results, taken together, suggest that multiple proteins in hepatoma-and nonhepatoma-derived cell extracts recognize the enhancer with different capacities. The finding that site B contains a dominant binding protein is consistent with the functional assay that distinguishes site B from other binding sites (Fig. 4).
The precise reasons that oligonucleotides containing site A and site D sequences failed to show competitions in gel shift assays are not clear but could be due to the following possibilities: (i) protein binding to these sites, as detected by DNase footprinting assays (Fig. 3), may require neighboring sequences, since the probes used in these two assay systems were not entirely the same; (ii) protein bindings to site A and site D may require protein occupancies to site B and site C for stabilization; (iii) site A and site D may not be genuine protein binding sites; instead, they may be created by protein-protein interactions, using sites B or C as anchoring points, thereby masking nuclease accessibilities of the neighboring sequences. Likewise, the failure of detecting enhancing activities for site C and site D in CAT assay (Fig. 4) suggests that proteins recognizing these sites may serve only a structural role. Further studies are needed to clarify these possibilities. DISCUSSION Recent studies from several laboratories have demonstrated that several putative positive and negative cis-regulatory elements are present in the promoter regions of the rodent and human mdr genes (35)(36)(37)(38)(39)(40)(41). In addition, several protein binding sites have been located at the promoter region of the murine mdr1b gene (42,43). The study presented here identified an enhancer located between nt Ϫ233 and Ϫ116 in the mouse mdr1b gene in which four DNase I footprints are located.
Further analyses showed that site A and site B together possess full enhancer function, but site B plays a predominant role. Strikingly, these sequences can promote preferentially in hepatoma-derived cell lines. This is the first cell type-specific enhancer found in the mdr locus (a cell type-specific enhancer upstream from the human MDR1 gene was reported (44), but association to the MDR1 could be due to cloning artifact (45)).
Site A contains the imperfect inverted repeat sequence 5Ј-ACTTACCTGAACACGTAAAG (underscored). Mutations in the first half repeat (Fig. 5, mutant e) failed to abolish the function of the enhancer, whereas the second half did (Fig. 5, f  and g), suggesting that the 5Ј boundary of the enhancer may begin at the second half of site A. We searched DNA sequences recognized by transcription factors in the GenBank and found that the second half repeat of site A contains a subsequence resembling the sequence motif CGT(A/C)A that is critical for binding of a group of cellular transcription factors, i.e. ATP, CREB, E4F1, or E4TF3, and EivF. Although these transcription factors are capable of activating E1a-and cyclic AMP (cAMP)-inducible promoters (46 -49), different promoters respond very differently to these inducers. Thus, whether mdr1b enhancer can be modulated by E1a or by cAMP remains to be determined. However, it may be relevant to note that cAMPdependent protein kinase regulated sensitivity of mammalian cells to multiple drugs has been reported (50). Site B contains two tetranucleotide direct repeats 5Ј-GTATGTAAATGTCT-GAGG (Fig. 5) and a potential HNF-1 binding site ((G/A)TTA-ATN(A/T)T(T/C)AG) (51). In addition, one copy of p53 binding site ( Ϫ188 AAGACAAGTCT Ϫ178 ) (52) is located between sites A and B of the mdr1b enhancer. However, it has been reported that a single copy of the recognition sequence was insufficient for transcriptional activation by p53 (52,53). In this context, FIG. 6. Gel mobility retardation assay of protein-DNA complexes. End-labeled probe nt Ϫ233 to Ϫ116 containing sites A through D (panels A and D), Ϫ210 to Ϫ155 (sites A and B, panels B and E), and Ϫ162 to Ϫ116 (sites C and D, panels C and F) were mixed with nuclear extract prepared from BprC1 cells (panels A-C) or from NIH3T3 cells (panels D-F) in the presence of poly dI⅐dC and various amounts of unlabeled, double-stranded enhancer DNA competitors, whose nucleotide sequences span A (Ϫ210 to Ϫ185), B (Ϫ185 to Ϫ155), C (p153 to Ϫ135), and D (Ϫ127 to Ϫ120) elements as indicated.
mdr1b enhancer may not be sensitive to the function of p53, although previous studies have demonstrated that the promoters of human MDR1 (38) and Chinese hamster Pgp1 (55) (both are homologs of murine mdr1a) are modulated differentially by wild-type and mutant p53. In any event, the identities of interacting protein factors that confer cell type-specific enhancer function remain to be demonstrated.
DNase I footprinting and gel mobility retardation assays apparently suggested that protein factors recognizing specific binding sites in the mdr1b enhancer are present in both hepatoma and non-hepatoma cells. This does not preclude the possibility that these proteins have a role in the regulation of mdr1b expression, specifically in hepatoma cells. For example, ATBF1, a multiple homeodomain zinc finger protein that selectively down-regulates hepatoma cell-specific enhancer of human ␣-fetoprotein gene, is present in both hepatoma and nonhepatoma cells (56,57). At this time, we cannot exclude the possibility of differential posttranscriptional modification of these proteins (e.g. phosphorylation, poly(ADP-ribosylation), etc.) resulting in transcriptional activation of mdr1b gene in liver tumors. Alternatively, these enhancer binding proteins may interact with other non-DNA binding factors to confer tissue specificity in the similar manner as the recently identified novel B cell-derived coactivator (OCA-B), which potentiates the activation of immunoglobulin promoters by octamerbinding transcription factors (58). Several such coactivators have also been implicated in the transcriptional regulation of liver-specific gene expression (59,60). Molecular cloning of genes encoding these mdr1b enhancer binding proteins should facilitate understanding of the complex control mechanism of mdr1b gene expression.
Levels of mdr1b mRNA are elevated during hepatocarcinogenesis. Studies of mdr1b expression in this system are hampered by the lack of culture cell systems that mimic the in vivo situation. Culturing primary hepatocytes in vitro for several hours shows spontaneous activation of mdr1b expression (61)(62)(63). Whether the enhancer described here is involved in the mdr1b expression in vivo remains to be determined. Investigations using transgenic animals and/or targeting gene delivery to HCC (54, 64) may allow us to address these issues. These experiments are currently under way in our laboratory.
In summary, we have characterized a hepatoma-specific enhancer in cultured cells. The identification of this enhancer may serve as a molecular basis for future studies of the regulation of mdr1b expression during hepatocarcinogenesis. These studies may eventually increase our understanding of how mdr gene expression in HCC is controlled and hence facilitate the development of approaches to control intrinsic drug resistance in this disease.