Potential conflict of interest: Nothing to report. Amsterdam Molecular Therapeutics declared no commercial interest in the conclusions.
Supported by Amsterdam Molecular Therapeutics.
Adenosine triphosphate (ATP)-binding cassette (ABC) transporters are drug efflux pumps responsible for the multidrug resistance phenotype causing hepatocellular carcinoma (HCC) treatment failure. Here we studied the expression of 15 ABC transporters relevant for multidrug resistance in 19 paired HCC patient samples (16 untreated, 3 treated by chemotherapeutics). Twelve ABC transporters showed up-regulation in HCC compared with adjacent healthy liver. These include ABCA2, ABCB1, ABCB6, ABCC1, ABCC2, ABCC3, ABCC4, ABCC5, ABCC10, ABCC11, ABCC12, and ABCE1. The expression profile and function of some of these transporters have not been associated with HCC thus far. Because cellular microRNAs (miRNAs) are involved in posttranscriptional gene silencing, we hypothesized that regulation of ABC expression in HCC might be mediated by miRNAs. To study this, miRNAs were profiled and dysregulation of 90 miRNAs was shown in HCC compared with healthy liver, including up-regulation of 11 and down-regulation of 79. miRNA target sites in ABC genes were bioinformatically predicted and experimentally verified in vitro using luciferase reporter assays. In total, 13 cellular miRNAs were confirmed that target ABCA1, ABCC1, ABCC5, ABCC10, and ABCE1 genes and mediate changes in gene expression. Correlation analysis between ABC and miRNA expression in individual patients revealed an inverse relationship, providing an indication for miRNA regulation of ABC genes in HCC. Conclusion: Up-regulation of ABC transporters in HCC occurs prior to chemotherapeutic treatment and is associated with miRNA down-regulation. Up-regulation of five ABC genes appears to be mediated by 13 cellular miRNAs in HCC patient samples. miRNA-based gene therapy may be a novel and promising way to affect the ABC profile and overcome clinical multidrug resistance. (Hepatology 2012)
Hepatocellular carcinoma (HCC) is the fifth most common type of cancer worldwide. With a 5-year survival of less than 5%,1 HCC remains one of the most fatal cancers, and few treatments have proven to be effective. Major pitfalls are late diagnosis, tumor recurrence, and resistance to chemotherapeutic treatment. This is caused by a phenomenon called multidrug resistance, mediated by high expression of adenosine triphosphate (ATP)-binding cassette (ABC) transporter family members that decrease the intracellular concentration of chemotherapeutic agents.2-6 There is limited information in the literature on the expression profile of ABC genes in HCC. For example, ABCB1 (MDR1)7, 8 and ABCC3 (MRP3)9 have been shown to be up-regulated in HCC of undetermined treatment status and a high expression of ABCC1 (MRP1) has been associated with an aggressive HCC phenotype in untreated patients.10 ABCA1 and ABCG2 down-regulation and ABCC4 (MRP4) up-regulation was shown in HCC of undetermined treatment status.11 The conventional model of multidrug resistance describes a genetically altered, highly resistant subpopulation of cells selected under pressure of chemotherapeutic agents.12 Therefore, profiling HCC tissues of untreated patients is of interest, as it addresses the question of inherent multidrug resistance of HCC that has developed in the absence of chemotherapy. The regulation of ABC gene expression in HCC could be mediated by microRNAs (miRNAs), a family of small RNAs which is often dysregulated in cancer.13-15
miRNAs are ≈22 nucleotide (nt) long endogenous, single-stranded, noncoding RNAs.16 miRNAs are loaded into the RNA-induced silencing complex (RISC) where further regulation will be undertaken. If the complementarity is perfect in the “seed region” (nt 2-7 from the 5′ end of the miRNA) between the miRNA and its target in the messenger RNA (mRNA), the mRNA will be cleaved by RISC and degraded; in case of imperfect complementarity, translation will be repressed.17-20 Specific miRNAs have been shown to be involved in various biological processes, including development, cellular proliferation, apoptosis, and oncogenesis.21, 22 The finding that individual miRNAs may target several hundred genes, and that a large percentage of mRNAs may be subject to regulation by miRNAs, further underscores the emerging importance of miRNA-mediated regulation.23, 24
Because miRNAs are involved in a great number of cellular processes and pathological conditions, it is thus possible that miRNAs regulate the expression of ABC transporters. Evidence was provided by Kovalchuk et al.,25 who showed that miR-451 may regulate ABCB1 in MCF7 breast cancer cells. Additionally, both miR-451 and miR-27a regulate ABCB1 expression in multidrug-resistant A2780DX5 and KB-V1 cancer cell lines.26 Reexpression of miR-203 in vitro in the liver cancer cell lines Hep3B, HuH7, and HLF was shown to induce down-regulation of ABCE1.27 These results indicate that cellular miRNAs are implicated in mediating the regulation of the expression of at least two ABC genes, including ABCE1 in liver cancer cells.
In the current study we hypothesized that ABC transporter gene expression is regulated by cellular miRNAs, resulting in a specific HCC phenotype. We show up-regulation of 12 ABC transporters in HCC, including eight which have not been previously associated with HCC. Subsequently, up-regulation of 11 cellular miRNAs and down-regulation of 79 was shown. Interestingly, 25 down-regulated miRNAs had predicted targets in six up-regulated ABC genes, of which we confirmed ABCA1, ABCC1, ABCC5, ABCC10, and ABCE1.
Samples were received as frozen tissue from the Centre de Ressources Biologiques Foie, France (courtesy of Prof. Dr. F. Degos, Hôpital Beaujon, Paris, and Prof. Dr. B. Clément, INSERM UMR991, Rennes, France). They were taken from 19 HCC patients and included paired tumor tissue (HCC) and adjacent healthy liver (AHL) tissue from each patient, frozen 15-105 minutes postsampling. RNA preparations from healthy liver (HL) samples from three patients with pancreatic cancer were obtained from Dr. F. Schaap, Academic Medical Center, Amsterdam.28 All patients provided informed consent in writing. No donor organs were obtained from executed prisoners or other institutionalized persons.
Total RNA was isolated from frozen tissue with Trizol (Invitrogen, Carlsbad, CA). RNA quality was assessed by checking for presence of 28S and 18S RNA on a 1.2% agarose gel (data not shown).
Reverse-Transcription Quantitative Polymerase Chain Reaction (RT-qPCR) for ABC Genes.
RNA was reverse-transcribed using random hexamer primers with the Dynamo kit (Finnzymes, Espoo, Finland) using 1.4 μg of total RNA. Expression of ABC genes was analyzed using custom-designed FAM-labeled 384-well Taqman Gene Expression Array (Applied Biosystems, Foster City, CA). Complementary DNA (cDNA) input per loading port (48 wells) was 1 μg. Custom array included 15 ABC genes and internal control 18S. Arrays were run in triplicate on a 7900HT RT-qPCR system (Applied Biosystems).
RT-qPCR for miRNAs.
miRNA RT reactions were performed with the TaqMan MicroRNA Reverse Transcription Kit (Applied Biosystems) according to manufacturer's instructions using 1 μg RNA. Specific miRNA targets were preamplified using Taqman PreAmp MasterMix and Megaplex PreAmp Primers (Applied Biosystems). Cellular miRNA expression from 10 HCC and 3 HL samples was analyzed using 384-well Taqman Array Human MicroRNA A cards v2.0 (Applied Biosystems) including 378 miRNAs and six controls (mammalian U6 in quadruplicate, RNU44, RNU48) as described in the manufacturer's instructions, including a preamplification step using Megaplex Primer, Human Pools Set v. 3.0 (Applied Biosystems). Arrays were run on a 7900HT RT-qPCR system (Applied Biosystems). For individual miRNA assay, RT reactions were performed with the TaqMan MicroRNA Reverse Transcription Kit (Applied Biosystems) using 10 ng RNA and 3 μL miRNA-specific RT-stem-loop primer (Applied Biosystems) according to the manufacturer's instructions. Taqman assay was done in 20 μL using 9 μL cDNA (5× diluted), 1 μL miRNA-specific primer with FAM-labeled fluorogenic probe (Applied Biosystems) and 10 μL Taqman 2× Universal PCR Master Mix (Applied Biosystems) and run in duplicate on a 7500 RT-qPCR system (Applied Biosystems).
Raw data were analyzed with RQ Manager (Applied Biosystems). Normalization was performed with DataAssist v. 2.0 (Applied Biosystems): for ABC genes HCC were normalized to their paired AHL; for miRNA HCC were normalized to HL (screen) or AHL (validation), and the relative gene expression 2−ΔΔCt method of Livak and Schmittgen29 was used. We set a 2-fold threshold for changes in gene expression per individual patient, i.e., changes lower than 0.5-fold were considered down-regulation and higher than 2-fold were considered up-regulation. Statistical analysis was performed using a two-tailed paired t test and Wilcoxon matched-pairs test and differences were considered significant when P < 0.05. For miRNA data analysis an additional manual screen was performed in order to check that the control group had consistent Ct values (Ct obtained for all three HL samples and less than 1.5 Ct variation between all three). All miRNAs that had deviating Ct values between the three HL samples were excluded from the analysis. Statistical analysis was performed using a two-tailed t test and differences were considered significant when P < 0.05.
miRNA Target Prediction.
Softwares TargetScan24 and PicTar23 were used for ABC 3′untranslated region (UTR) target prediction of cellular miRNAs. Additionally, 3′UTR sequences were manually screened for miRNA seed-matching sequences. Predictions are presented in Supporting Tables S1, S2.
Luciferase reporters were made by cloning of ABC 3′UTR sequences (Tables S3, S4), in the renilla luciferase gene in the psiCheck-2 vector (Promega, Madison, WI). Constructs with mutated miRNA seed sequence in the ABC genes (nt 2 to 6) were synthesized by IDT (Coralville, IA). Primary miRNA (pri-miRNA) sequences were amplified (primer sequences, Table S3) from human adult normal breast tissue genomic DNA (Biochain, Hayward, CA). miRNA expression plasmids were made by cloning of the pri-miRNAs in the pcDNA6.2 vector (Invitrogen). All constructs were verified by sequencing (Macrogen, Seoul, Korea).
Cell Lines and Transfections.
Human embryonic kidney (HEK) 293T cells were cultured according to the American Tissue Culture Collection (ATCC) instructions. Cells were plated in 6-, 24-, or 96-well plates 1 day prior to transfection. Transfections were performed with Lipofectamine 2000 or LTX reagent (Invitrogen) according to the manufacturer's instructions. For luciferase assays, HEK293T cells were cotransfected with 5 ng of Luc-ABC reporter that contains both firefly and renilla luciferase genes and 150 ng of the corresponding miRNA expression constructs. Expression values when the miR-Control (miR-Ctrl) was transfected were set at 1.
Transfected cells were assayed at 72 hours posttransfection and firefly and renilla luciferase activities were measured with the Dual-Luciferase Reporter Assay System (Promega) according to the manufacturer's instructions. The relative luciferase activity was calculated as the ratio between the renilla and firefly luciferase activities.
In order to perform ABC gene and miRNA expression profiling, tissues were sampled from HCC and AHL from 19 patients. Three patients received chemotherapy (FR06, FR16, and FR17) prior to sampling, whereas 16 were untreated. Seventy-four percent of the HCCs included in the study arose in patients with alcoholic cirrhosis and 16% in patients with hepatitis C virus (HCV)-associated cirrhosis. Thirty-nine percent of the tumors had a size of 31-50 mm, 28% of 51-100 mm, and 22% of the tumors were smaller than 30 mm (Table 1).
Table 1. Clinical Data of the 19 Patients from Which Paired HCC and Adjacent Healthy Liver (AHL) Tissue Samples Were Analyzed. Obtained from the CRB-Foie, France
Patient information (n = 19)
FR01-03, FR05-08, FR11-FR16, FR18-19
FR04, FR09-10, FR17
FR01-03, FR05-08, FR10-11, FR14-16, FR18-19
FR04, FR09, FR17
FR01-06, FR08-11, FR13-17, FR19
FR07, FR12, FR18
Portal embolization, TACE/lipiodol
FR01, FR03-05, FR07-15, FR18-19
FR02, FR09, FR12-13
FR01, FR05-07, FR11, FR17-18
FR03, FR10, FR14, FR16, FR19
FR01-04, FR07-08, FR11-14, FR16-19
FR01-03, FR05-06, FR09, FR11-13, FR17, FR19
Twelve ABC Transporters Are Up-regulated in HCC Compared with AHL.
To obtain ABC expression signatures in HCC patients, RNA was isolated from 19 paired HCC and AHL samples, and expression of 15 ABC genes was determined by custom-made ABC Taqman microfluidic array. Biological function of the 15 ABC genes is described in Table 2. Student's two-tailed paired t test and Wilcoxon matched-pairs test were performed to determine whether changes in gene expression between AHL and HCC were significant (P < 0.05). Both tests indicated that 12 ABC genes were significantly up-regulated in HCC compared with the paired AHL control samples (Fig. 1). From these genes, ABCA2, ABCC11, and ABCE1 showed a mild up-regulation of 1.6 to 1.7-fold on average. Higher changes in expression profiles were detected for ABCB1, ABCB6, ABCC1, ABCC2, ABCC3, ABCC4, ABCC5, ABCC10, and ABCC12 genes, which were on average 2.0 to 5.3-fold up-regulated (Fig. 1). Expression of ABCA1, ABCC6, and ABCG2 was not significantly changed. However, there was heterogeneity in ABC gene regulation between patients, as not 100% of the HCC samples showed an up-regulated profile (Table S5). Globally, up-regulation higher than 2-fold was observed in more than 15% of the patients for ABCC6 and ABCG2 and in more than 30% of the HCC patients for ABCA1. Interestingly, an up-regulation higher than 2-fold was observed in more than 50% of the HCC patients for ABCB6, ABCC1, ABCC4, ABCC5, ABCC10, and ABCC12. A majority of the patient samples presented an up-regulation higher than 1.5-fold change (Table S4).
Table 2. Biological Function and Cellular Localization of the 15 Profiled ABC Transporters as Described in OMIM and UniProtKB Databases
(Putative) Cellular Localization
Cholesterol efflux, anion transporter
Sterol transport, reactive oxygen species (ROS) protection
Apical plasma membrane
Multidrug resistance, transport of organic anions and cations
Outer mitochondrial membrane
Binds heme and porphyrins and functions in their ATP-dependent uptake into the mitochondria
Basolateral plasma membrane
Multidrug resistance, organic anion transport
ABCC2 (MRP2, cMOAT)
Apical plasma membrane
Multidrug resistance, organic anion transport
Basolateral plasma membrane
Multidrug resistance, organic anion transport
Basolateral plasma membrane
Multidrug resistance, organic anion transport
Multidrug resistance, transport of chemotherapeutic agents, 6MP, Platinum
Basolateral plasma membrane
Multidrug resistance, resistance against inhibitors of topoisomerase II etoposide and teniposide, and the anthracyclines doxorubicin and daunorubicin.
Putatively involved in cellular detoxification through lipophilic anion extrusion
Stimulates the ATP-dependent efflux of a range of physiological and synthetic lipophilic anions, steroid sulfates, glucuronides, the monoanionic bile acids glycocholate and taurocholate, and methotrexate
No putative transport function. Antagonizes the binding of 2-5A (5′-phosphorylated 2′,5′-linked oligoadenylates) by RNase L through direct interaction with RNase L and therefore inhibits its endoribonuclease activity, antagonizes the anti-viral effect of the interferon-regulated 2-5A/RNase L pathway
Apical plasma membrane
Transports broad spectrum of chemotherapeutic agents (mitoxantrone, daunorubicin, doxorubicin, daunorubicin)
Association of Changes in ABC Expression Profile with Clinical Parameters.
We subsequently determined whether the changes in expression of ABC genes correlated with clinical parameters, e.g., treatment, patient gender, differentiation state, and size of the tumor. First we determined the effect of HCC treatment, as we included 16 untreated and three treated patients in this study. When excluding the three treated patients (FR06, FR16, and FR17) from the analysis, up-regulation was significant for 10 ABC genes (ABCA2, ABCB1, ABCB6, ABCC2, ABCC3, ABCC4, ABCC5, ABCC10, ABCC11, and ABCE1). Expression of ABCA1, ABCC1, ABCC6, ABCC12, and ABCG2 was not significantly changed in untreated patients. Out of the 19 patients, only one, FR06, had a stable or down-regulated expression for the entire gene set, with 0.3 to 1.3-fold changes for all genes. Patient FR06 had a well-differentiated 50-mm initial tumor and was previously treated with TACE.
We then determined if there was an association between ABC up-regulation and gender. Expression of ABCC6 was significantly higher in females than in males (female, n = 4, AFC = 2.7; male, n = 15, AFC = 1.2; P = 0.0341), whereas for the expression of other ABC genes there was no gender difference. We then correlated the ABC expression data with the differentiation state and found that up-regulation of ABCC10 was significantly lower in well-differentiated HCC (well-differentiated, n = 4, AFC = 1.40; intermediate differentiation, n = 15, AFC = 3.25; P = 0.0081). Finally, we determined a possible association of ABC expression with tumor size. Up-regulation of ABCB6 and ABCC2 was significantly higher in patients with tumors <30 mm than in patients with tumors >31 mm (<30 mm, n = 4; >31 mm, n = 15), with AFC values of respectively 4.6 and 2.3 for ABCB6 (P = 0.0144) and 4.2 and 1.5 for ABCC2 (P = 0.0022).
Total miRNA Profiling Reveals That 79 miRNAs Are Down-regulated and 11 Are Up-regulated in HCC Patient Samples.
We hypothesized that ABC gene expression might be regulated by cellular miRNAs, i.e., ABC genes up-regulation in HCC would be the consequence of the down-regulation of cellular miRNAs. In order to obtain miRNA expression signatures, RNA was isolated from 10 HCC and three HL samples. To minimize variation in the miRNA profile, only 10 HCC with alcohol etiology were selected from the 19 available (FR01, FR03, FR05, FR06, FR07, FR08, FR10, FR11, FR14, and FR18). miRNA expression was determined by Taqman 384-well microfluidic array including 378 cellular miRNAs and six control wells and data were normalized to mammalian U6 RNA. In total, 361 out of 378 miRNAs were detectable. Changes in miRNA expression between 2-and 40-fold were considered up-regulation and changes between 0- and 0.5-fold were considered down-regulation. Average miRNA expression was compared in HCC and HL groups by two-tailed t test. miRNA expression in HCC compared with HL control was significantly higher for 11 cellular miRNAs and lower for 79 miRNAs, which accounted for respectively 3% and 22% of the detectable miRNAs (Fig. 2; Fig. S2). Analysis of the conservation of the 90 dysregulated miRNAs revealed that 87 were conserved up to the mouse, and 25 up to the chicken (Table S6). Next a subset of miRNAs quantified with the microfluidic array was cross-examined on all samples: 19 paired HCC and AHL, and three HL. Six miRNAs were selected: miR-135b, miR-145, miR-199a-3p/a/b, and miR-296 because they were consistently down-regulated in the 10 HCC patient samples (low standard deviation). Expression of these six miRNAs was quantified using single miRNA Taqman assays (Fig. 3). First, miRNA expression in HL from pancreatic cancer patients and AHL from liver cancer patients was similar (Fig. 3). This indicated that the miRNA profile is not affected in HL tissues despite the different background of these samples. Second, differences in miRNA expression observed between paired AHL and HCC samples were significant for miR-145, miR-199a-3p, miR-199a-5p, and miR-199b, hence confirming the miRNA signature in HCC from the microfluidic array. These differences were also significant for miR-135b when the two patients presenting the highest variation in each group are excluded (Fig. 3; miR-135b; FR01; FR12, FR13, and FR19; P = 0.0070). miR-296 presented a down-regulated profile but the differences were not statistically significant. When comparing all six miRNA profiles there was a trend for FR13 to recurrently present the highest variation among the AHL group, for miR-135b, miR-145, miR-199a-3p/a/b. Patient FR13 was an untreated HCV-infected male with a history of alcohol abuse.
Verification of ABC miRNA Target Prediction.
miRNAs can induce posttranscriptional down-regulation of target genes, i.e., genes which have in their 3′UTR sequences complementary to an miRNA seed sequence. We hypothesized that some down-regulated miRNAs are regulating ABC expression and that this is associated with HCC. Because most of the ABC genes were up-regulated in HCC samples we concentrated on the down-regulated miRNAs as potentially influencing ABC expression. Interestingly, from the 79 down-regulated miRNAs, 25 had predicted targets in up-regulated ABC genes. We therefore determined in silico miRNA target sequences in the 3′UTRs of the up-regulated ABC genes and cross-analyzed these data with the down-regulated miRNAs (Tables S1, S2). Twenty-four cellular miRNAs were cloned in the expression vector pcDNA6.2 and Luc-ABC reporters where the 3′UTR of the six ABC genes was cloned into the dual luciferase vector psiCheck-2 were made for six ABC genes. Because the 3′UTR of ABCA1 is 3.3 kb, we made three Luc-ABCA1 variants: Luc-ABCA1-5′ contains the 5′ end of the 3′UTR, ABCA1-3′ the 3′ end, and ABCA1 is a composite containing 246 nt from the 5′ end and 303 nt from the 3′ end of the 3′UTR (Fig. S2). We subsequently validated the in silico miRNA target predictions in vitro using a luciferase reporter system. Based on the in silico predictions we cotransfected HEK293T cells with the Luc-ABC reporter plasmids that contained the miRNA predicted targets and the respective miRNAs and measured luciferase expression. Knock-down of luciferase expression would indicate that the ABC transporter is a possible target for the specific miRNA. To determine which miRNAs were efficiently down-regulating the corresponding Luc-ABC reporters, we considered all miRNAs that inhibited luciferase expression below a set threshold of 0.5 as possibly targeting the ABC gene. With this method we were able to experimentally confirm 15 ABC miRNA targets out of the 51 associations that were bioinformatically predicted (Fig. 4). miR-101 and miR-135b down-regulated Luc-ABCA1-5′; however, they had no effect on Luc-ABCA1-3′ expression. Interestingly, the composite Luc-ABCA1 was down-regulated by miR-101 and miR-135b by respectively 67% and 77%, indicating an additive effect of the targets at 5′ and 3′ ends of the 3′UTR. Several other miRNA targets were verified with the Luc-ABC assay: Luc-ABCC1 was down-regulated by miR-199a/b and miR-296, Luc-ABCC4 was down-regulated by miR-125a/b, Luc-ABCC5 was down-regulated by miR-101, miR-125a and let-7a, Luc-ABCC10 was down-regulated by let-7a/e, and Luc-ABCE1 was down-regulated by miR-26a, miR-135b, and miR-145 (Fig. 4). To experimentally verify the miRNA targets, we next mutated the predicted miRNA targets by changing the seed sequences of miRNAs in the 6 Luc-ABC reporters. We were able to confirm all Luc-ABC as bona fide miRNA targets, except for Luc-ABCC4, which was not clearly affected when the miR-125a/b sites were mutated (Fig. 5).
Finally, for each paired HCC patient sample we tested the correlation between a specific ABC expression profile and a corresponding validated miRNA (Fig. 6; Fig. S3). We expected an inverse ABC-miRNA correlation; therefore, tumors with a high ABC expression should simultaneously present a low validated miRNA levels and vice versa. As anticipated, our positive control, the previously published ABCE1/miR-203 pair, presented a good qualitative correlation with 9 out of 10 tumors having high ABCE1 and low miR-203 levels (Fig. S3). However, the correlation coefficient R2 = 0.6433 indicating that the samples do not fit a linear regression, likely due to the low number of samples (n = 10) and the absence of samples displaying down-regulated ABCE1 expression in the sample set. We therefore discarded R2 as a quantitative readout and determined only a qualitative response, i.e., if for each ABC/miRNA pair a majority of tumors present a high ABC expression and a low validated miRNA level. ABCC5/miR-101 pair presented a good correlation with 9 out of 10 HCC samples being high for ABCC5 and low for miR-101 (Fig. 6). ABC/miRNA pairs ABCC5/let-7a, ABCC5/mir-125a, and ABCC5/miR-125a showed similar results (Fig. 6). The other verified ABC/miRNA pairs also showed inverse correlations in expression profile in each individual patient tumor (Fig. S3). This negative correlation would require validation on a larger sample set but provides indication of a miRNA regulation of ABC genes in HCC.
In the current study we quantified the expression of 15 ABC transporters in 19 paired HCC and AHL patient samples. The majority had not received chemotherapy prior to sampling (16/19 untreated patients) and in most (14/19) the etiology was alcoholic cirrhosis. We showed that 12 ABC genes were up-regulated in HCC. In several patients the ABC genes were up-regulated up to 2-fold and the physiological relevance of such a mild regulation needs additional attention. We speculate that in the context of chemotherapy, even changes of 1.5-fold may tip the toxic concentration of the drug due to changes in efflux activity of the ABC genes in the tumor cells, therefore resulting in a significant physiological effect. Up-regulation of some of these transporters has been described previously, e.g., ABCB17, 8 and ABCC39 were up-regulated in HCC. The expression of three ABC genes, ABCA1, ABCC6, and ABCG2, was not significantly changed in this study. Interestingly, ABCA1 and ABCG2 down-regulation was shown in HCC compared with adjacent HL in patients of unknown treatment status,11 and the two genes were respectively 14.6 and 9.3-fold up-regulated in TACE-treated samples.30 These mixed results may indicate a high variability in the expression of ABCA1 and ABCG2 in HCC patients, possibly linked to treatment status. In our study, only one patient presented a stable or down-regulated expression of all 15 transporters. This patient, FR06, had received transarterial chemoembolization (TACE) treatment, but as the other two TACE-treated patients did not present such a profile we cannot make any conclusion on the effect of TACE on ABC expression. Second, we were able to demonstrate that in the 16 untreated patients the expression of 10 ABC transporters was significantly up-regulated in HCC compared with paired AHL samples. In untreated patients, only up-regulation of ABCC1 has been shown to be associated with a more aggressive HCC phenotype.10
To our knowledge this is the first publication reporting up-regulation of a broad range of ABC transporters in untreated HCC patients. This includes eight ABC genes: ABCA2, ABCB6, ABCC2, ABCC3, ABCC5, ABCC10, ABCC11, and ABCE1, whose association with HCC has not been reported so far. Due to the size of some subpopulations, the described associations between ABC profile and clinical parameters will have to be confirmed on a larger patient population. Nevertheless, the results of the ABC profiling raise the question of the possible regulation pathways implicated in the phenomenon of ABC genes up-regulation.
We chose to further investigate the possibility of miRNA-mediated ABC gene regulation in association with HCC. So far only miR-203 has been reported to specifically target ABCE127 but its implication in HCC still needs to be determined. Our miRNA screen in 10 HCC patients identified significant changes in expression of 90 miRNAs, including 11 up-regulated miRNAs and 79 down-regulated miRNAs, of which 25 had predicted ABC targets. Interestingly, 97% of the dysregulated miRNAs were highly conserved in mammals (up to the mouse), indicating their possible association with HCC, as has been shown for several diseases.31 Seventy-nine of 90 dysregulated miRNAs are down-regulated and this is in agreement with the evidence that miRNAs are generally down-regulated in cancer.32 Many miRNAs identified during the current screen confirmed findings obtained with larger sample sets or sample sets coming from patients with a different cancer etiology. Similar to other publications, we observed down-regulation of miR-101,33 miR-122,34-37 miR-125a,37, 38 miR-130a,34, 39 miR-145,39-41 miR-199a,33, 36, 38, 39 miR-199b,38, 40 and up-regulation of miR-21.33-37, 40-42 Seventy-four percent of our patients had alcoholic cirrhosis, and ethanol-treatment has been linked in the literature to the miR-199 family.43 However, no correlation between the miRNA profiles and viral HCC etiology could be determined, probably due to the small patient size (14 alcohol, three HCV, one HBV, and one alcohol/HCV). Our screen also identified 12 miRNAs with predicted ABC targets that have not been previously associated with HCC. Down-regulated miRNAs were reported to be repressed by oncogenes: miR-145 by Ras,44 and miR-26a and miR-195 by Myc45 as well as let-7.45, 46 The most down-regulated miRNAs, miR-383 and miR-654-3p, had no putative targets in any of the 15 ABC genes. These two miRNAs, together with the two most up-regulated miRNAs, miR-96 and miR-182, require further study for understanding their relevance in HCC. Down-regulated miRNAs are of interest because they can act as tumor suppressors.39 Cellular miRNAs can also act as oncogenes,47 and their up-regulation in cancer will cause down-regulation of their tumor-suppressive targets. In general, these miRNAs are potentially relevant for HCC therapy: tumor suppressor miRNAs can be introduced back in a cancer cell, thereby repressing tumorigenesis, and oncogenic miRNAs can be inhibited by using synthetic miRNA antagonists or virally-delivered sponge-like sequences.48, 49 This brings exciting possibilities for the use of miRNAs as therapeutics.
In the current study we experimentally verified 13 predicted miRNAs targets in five ABC genes using luciferase reporter assays. We were able to prove that the miRNA effect was sequence-specific by mutating the targets in the reporters and by cotransfecting miRNAs not having targets in ABC genes (data not shown). Except for ABCC4, our mutational analysis revealed some new miRNA targets in ABC genes. Strikingly, we were able to show that for several miRNA-ABC pairs, a very high proportion of the analyzed tumors have an increased ABC gene expression level together with a reduced level of miRNA. Thus far the only evidence of miRNA-mediated regulation of ABC gene expression in HCC has been provided by Furuta et al.,27 who showed that miR-203 regulates the expression of ABCE1, which is involved in translation initiation, but has not been linked to multidrug resistance. With this perspective, we are currently working on in vitro validation of the miRNA-mediated regulation of endogenous ABC gene expression with a special focus on the ABCC subfamily. Future research will concentrate on delivery of these miRNAs as gene therapy, either in miRNA-replacement therapy for HCC or as a novel indirect strategy to induce down-regulation of ABC transporters instead of direct short hairpin RNA (shRNA)- or artificial miRNA-mediated gene therapy approaches.50 The focus should be on ABCB6, ABCC1, ABCC4, ABCC5, ABCC10, and ABCC12 as these genes were up-regulated in more than 50% of the patients.
The authors thank Françoise Degos, Bruno Clément, Bruno Turlin, and the Centre de Ressources Biologiques Foie (France) for providing clinical samples and data, and Cees B.M. Oudejans (Department of Clinical Chemistry, VU University Medical Center, Amsterdam, The Netherlands) for kindly providing access to the ABI 7900HT.