Chronic ethanol feeding enhances miR-21 induction during liver regeneration while inhibiting proliferation in rats

Abstract

Liver regeneration is an important repair response to liver injury. Chronic ethanol consumption inhibits and delays liver regeneration in experimental animals. We studied the effects of chronic ethanol treatment on messenger RNA (mRNA) and microRNA (miRNA) expression profiles during the first 24 h after two-thirds partial hepatectomy (PHx) and found an increase in hepatic miR-21 expression in both ethanol-fed and pair-fed control rats after PHx. We demonstrate that the increase of miR-21 expression during liver regeneration is more robust in ethanol-fed rats. Peak miR-21 expression occurs at 24 h after PHx in both ethanol-fed and control rats, corresponding to the peak of hepatocyte S phase in control rats, but not in ethanol-exposed livers in which cell cycle is delayed. The induction of miR-21 24 h after PHx in control rats is not greater than the increase in expression of miR-21 due to sham surgery. However, in the ethanol-fed rat, miR-21 is induced to a greater extent by PHx than by sham surgery. To elucidate the implications of increased miR-21 expression during liver regeneration, we employed unbiased global target analysis using gene expression data compiled by our group. Our analyses suggest that miR-21 may play a greater role in regulating gene expression during regeneration in the ethanol-fed rat than in the control rat. Our analysis of potential targets of miR-21 suggests that miR-21 affects a broad range of target processes and may have a widespread regulatory role under conditions of suppressed liver regeneration in ethanol-treated animals.

the adult liver has the capacity to regenerate in response to damage caused by toxic or mechanical injury or infection, through proliferation of fully differentiated, normally quiescent cells. In the experimental model of 70% partial hepatectomy (PHx), widespread hyperplasia involving the vast majority of all hepatocytes results in recovery of the initial liver mass within ∼1–2 wk after resection, through a well-synchronized cell cycle response. In the rat, hepatocytes start entering S phase after 12 h and peak DNA synthesis occurs by 24 h after PHx, followed ∼24 h later by proliferation of nonparenchymal cells (38). During the initial 24 h following PHx an extensive program of gene expression changes is activated, as previously demonstrated by our group and others (9, 19, 23) (R. Vadigepalli, unpublished data).

Chronic ethanol consumption imposes metabolic and functional changes, which promote hepatic steatosis, a condition that predisposes the liver to more progressive disease such as steatohepatitis and cirrhosis. The regenerative capacity of the liver is greatly impaired by chronic ethanol intake. Inhibition of this repair process may contribute to the pathogenesis and progression of alcoholic liver disease. Liver regeneration after PHx is inhibited and delayed in chronically ethanol-fed rodents. DNA synthesis in the ethanol-treated regenerating liver is largely attenuated at 24 h (40) and does not peak until 48–72 h following PHx (5, 6). Liver regeneration in ethanol-fed rats is marked by significant changes in the temporal profile of gene expression compared with control livers after PHx (R. Vadigepalli, unpublished data). The molecular mechanisms by which ethanol treatment causes inhibition of liver regeneration remain largely unknown.

microRNAs (miRNAs) are small, 19–22 nucleotide, noncoding RNAs that posttranscriptionally regulate gene expression. miRNAs target mRNAs through imperfect base pairing, resulting in inhibition of translation and/or destabilization and degradation of their targets. Each miRNA can potentially regulate hundreds of targets, suggesting that most mammalian coding genes could be regulated by miRNAs (8). Recent reports of dynamic miRNA expression profiles following PHx suggest a likely role for miRNAs in liver regeneration (3, 5a, 29, 36), although specific functions targeted by miRNAs in this process remain largely unknown.

Recently, we profiled miRNA expression in rat liver during the first 24 h following PHx in ethanol-fed rats and control animals (5a). We observed an increase of miR-21 expression during the course of regeneration similar to recent reports by other investigators in both rat and mouse (3, 29, 36). These results are in apparent accordance with a proproliferative role of miR-21, which has been well characterized in cancer and cultured cells. A majority of tumors overexpress miR-21 and very high levels of miR-21 are found in most cancer cell lines (22). Among the well-characterized targets of miR-21 are tumor suppressor genes such as PTEN and PDCD4 (22).

In an effort to elucidate the role of miR-21 in liver regeneration, previous reports have largely focused on a single potential target gene (3, 29, 36). This approach fails to uncover the inherent depth of regulation possible with a single miRNA. To better understand the more widespread implications of increased miR-21 expression, we employed a global target analysis approach that utilizes the detailed temporal profiling of global gene expression changes during liver regeneration in ethanol-fed and control rats [R. Vadigepalli, unpublished data, Gene Expression Omnibus (GEO) no. GSE33785].

Our data demonstrate that upregulation of miR-21 following PHx is more robust in livers from ethanol-fed rats than in control rats, despite the inhibited cell proliferation, in apparent conflict with a proproliferative role of miR-21 in liver regeneration. This novel finding allows for a comparative analysis of potential miR-21-affected functions during regeneration between naive and ethanol-adapted livers. Our results indicate that miR-21 may have a greater impact in inhibiting regeneration in ethanol-treated livers than in enhancing hepatocyte proliferation in the normal regenerative process and potentially regulates a broad range of cellular functions in ethanol-treated livers.

MATERIALS AND METHODS

Animal protocols.

Male Sprague-Dawley (Charles River, Wilmington, MA) rats were maintained on a 12-h light and dark cycle. Animals arrived between 35–49 g as littermate pairs and were allowed access to standard chow and water until reaching 120 g. Animals were fed an ethanol-containing diet according to Lieber and DeCarli (4, 26). Briefly, animals were introduced to the control liquid diet (composed of 18% of the total calories as protein, 35% as fat, and 47% as carbohydrates; Bio-Serv, Frenchtown, NJ) for 2 days. The larger rat of each pair then received a liquid ethanol diet (containing 18% of total calories as ethanol in replacement of carbohydrates) for 2 days, followed by a standard ethanol liquid diet (containing 36% of total calories from ethanol in replacement of carbohydrates) ad libitum for 5 wk. Littermate control animals were pair fed the control liquid diet described above. Male Sprague-Dawley (Harlan, Indianapolis, IN) rats with access to standard chow and water ad libitum were also used as a dietary control. All animal protocols were approved by the Thomas Jefferson University Institutional Animal Care and Use Committee.

Two-thirds partial hepatectomy (PHx) was performed under isoflurane anesthesia by removing the left lateral and medial lobes (LLM) of the liver through surgical ligation (13). Surgeries were performed between 8:00 and 11:00 AM to avoid circadian rhythm effects. Following surgery, the rats were allowed to recover with access to their appropriate diets. Where indicated, bromodeoxyuridine (BrdU) was injected intraperitoneally 22.5 h after PHx at a dose of 100 mg/kg. The rats were euthanized 1, 6, 12, 24, and 36 h (n = 4 pairs per time point) after PHx. LLM tissue was retained at the time of surgery and used as t = 0 biological control for each rat. Sham operations were performed by laparotomy followed by ∼30 s of manual manipulation of the liver without resection and livers were harvested 24 h later. Liver tissue was clamp frozen in liquid nitrogen (31) within 5 s of resection and stored at −80°C. A small section was retained and formalin fixed for histology.

Immunohistochemistry.

Liver tissue was fixed overnight in 4% neutral buffered formalin followed by 70% ethanol, paraffin embedded, and sectioned. Sections were immunostained for BrdU by using Vectastain ABC kit from Vector Laboratories (Burlingame, CA) according to manufacturer's instructions, using diaminobenzidine (DAB) (ImmPact DAB peroxidase substrate, Vector Laboratories) as a detection agent. The primary antibody was rat anti-BrdU (Accurate Chemical and Scientific, cat. no. OBT0030) and the secondary was sheep anti-rat (AbD Serotech, cat. no. AAR10B) Slides were counterstained with Harris hematoxylin. For each sample, three random fields were selected and the total and BrdU-positive hepatocytes were quantified and the mean percentage of labeled nuclei was calculated per sample. Percentage of labeled nuclei was compared between ethanol-containing diet (EtOH) and carbohydrate control diet (CHO) rats 24 h after PHx and statistical significance was determined by Student's t-test.

RNA isolation.

Total RNA was isolated from frozen liver (50–100 mg) with TRIzol (Invitrogen, Carlsbad, CA) by following the manufacturer's recommendations. RNA concentration was determined by ND-1000 (NanoDrop, Wilmington, DE). miRNA and mRNA expression analyses were performed on the same total RNA isolations for each sample.

miR-21 expression analysis.

miR-21 expression was measured by TaqMan RT-qPCR miRNA assays (Applied Biosystems, Foster City, CA) according to manufacturer's recommendations. Briefly, 10 ng total RNA was used for reverse transcription. For quantitative polymerase chain reaction (qPCR), 1 μl of the RT product was used for each reaction and TaqMan Universal PCR Master Mix, No AmpErase UNG (Applied Biosystems) was used. All assays were performed in triplicate on an ABI Prism 7000 Sequence Detection System (Applied Biosystems). Four biological replicates were used per time point, and miR-21 expression was normalized to rat small nucleolar RNA (snoRNA) expression. Relative expression was determined with the ΔΔCt method in which the normalized miR-21 expression in the LLM (t = 0) was subtracted from the normalized miR-21 expression in the remnant (PHx) tissue from the same rat.

Gene expression analysis.

Gene expression was determined by using Affymetrix GeneChip Rat Gene 1.0 ST Arrays (Affymetrix, Santa Clara, CA) in the Thomas Jefferson Nucleic Acid Core Facility. Briefly, 10 μg of each of the total RNA samples for the 6-h and 24-h time points, as well as their biological control LLM samples, for both EtOH and CHO diets, as described above, were further purified with the Qiagen RNeasy Mini kit (Qiagen, Germantown, MD) followed by ethanol precipitation. RNA samples were analyzed with an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA) to ensure quality prior to labeling and hybridization. Data were RMA normalized with the Affymetrix Expression Console (Affymetrix). Genes were considered expressed above background if they had a log₂ normalized signal of 5 and above in at least one sample. MIAME (Minimum Information About a Microarray Experiment)-compliant microarray data were deposited in the GEO database, no. GSE33785.

Target prediction.

Gene expression differences between conditions and time points after PHx (Δs) were calculated for TargetScan (25) (v5.1)-predicted miR-21 targets from gene expression data, as described in results. Δ₁, comparing change during liver regeneration in EtOH and CHO, respectively: [(24 h PHx − 24 h LLM) − (6 h PHx − 6 h LLM)]. Δ₂, comparing the change during regeneration in EtOH to that in CHO: {[(EtOH 24 h PHx − EtOH 24 h LLM) − (EtOH 6 h PHx − EtOH 6 h LLM)] − [(CHO 24 h PHx − CHO 24 h LLM) − (CHO 6 h PHx − CHO 6 h LLM)]}. “Potential targets” of miR-21 were identified as having a significantly decreased expression in EtOH livers (Δ₁) by less than −0.2 log₂ ratio and Δ_1(EtOH) < Δ_1(CHO) by less than −0.2 log₂ ratio. Potential targets in EtOH livers were also considered potential targets in CHO livers if Δ_1(CHO) was significant and less than −0.2 log₂ ratio.

Gene set enrichment analysis.

Gene set enrichment analysis (GSEA) (37) was performed to determine the distribution of predicted miR-21 targets within the ranked lists of all genes, as described in Appendix A. The gene expression dataset from Elmen et al. (7) provided in the NCBI GEO (GSE13948) was used to validate GSEA (Appendix A and Table A1). False discovery rate q values estimated by GSEA are reported. A q value of <0.05 was considered significant.

Functional annotation.

Functional annotation terms for the miR-21 potential targets were obtained from DAVID (Database for Annotation, Visualization and Integrated Discovery) Functional Annotation Tool (16, 17). Potential targets were grouped by common gene ontology (GO) terms. For comparison, GO terms were similarly identified for predicted targets that were not considered “potential” targets in our analysis.

Target validation.

HEK293 cells were grown in DMEM/F12 media and plated in poly-l-lysine-coated 24-well plates the day prior to transfection. A dual reporter miTarget microRNA 3′ untranslated region (3′UTR) expression clone (pEZX-MT02, Genecopoeia, Rockville, MD) with a firefly luciferase reporter gene linked to the rat Crebl2 3′UTR containing the putative miR-21 target site was used. The plasmid also contained a Renilla luciferase gene under control of a separate promoter for normalization of transfection efficiency. Cotransfection was performed in 24-well plates with 150 ng plasmid and 10 pmol pre-miR-21 or pre-miR-control (Ambion, Foster City, CA) with use of Lipofectamine 2000 (Invitrogen). At 18 h following transfection, cells were replated in triplicate in 96-well plates, and 24 h later luminescence was determined on a Synergy HT Microplate Reader (Bio Tek Instruments, Winooski, VT) by use of the Luc-Pair luciferase assay kit (Genecopoeia). Firefly luciferase reporter luminescence was normalized to Renilla luciferase luminescence. Experiments were performed on three separate occasions.

Statistical analysis.

Statistical significance was determined with a two-tailed Student's t-test. Analysis of change between PHx and LLM samples were paired by animal. All other observations were unpaired. A P value <0.05 was considered significant.

Statistical significance between cumulative distributions was determined with a two-sided Kolmogorov-Smirnov test. A P value of <0.05 was considered significant.

RESULTS

miR-21 expression in liver regeneration.

In a parallel microarray profiling study of miRNA expression following PHx we found an increase in miR-21 expression in both CHO and EtOH rats (5a). These findings prompted us to investigate more closely the role of miR-21 during the inhibited regeneration in the EtOH liver. We quantified miR-21 expression dynamics during the first 36 h of liver regeneration following PHx in both CHO and EtOH rats using RT-qPCR. Our data demonstrate a significant increase in miR-21 expression at 24 h following PHx in the CHO liver compared with LLM samples from the same animals (Fig. 1A). This peak of miR-21 induction following PHx coincides with peak hepatocyte DNA synthesis in the regenerating rat liver (38).

Fig. 1.

miR-21 expression during liver regeneration in chronically ethanol-fed (EtOH) and pair-fed control (CHO) rats. A: miR-21 expression was determined by RT-quantitative PCR in remnant livers during the first 36 h following partial hepatectomy (PHx). Each replicate was normalized to its own left lateral and medial lobes (LLM). At t = 0, EtOH LLM was normalized to CHO LLM and there was no significant difference in miR-21 expression between EtOH and CHO prior to PHx. Peak miR-21 expression was at 24 h after PHx in both EtOH and CHO rats. B: miR-21 levels are increased at 24 h after sham operation compared with the relative expression in the LLM tissue. Results found in chow-fed rats were comparable to results in CHO rats. In the EtOH rat, the increase in miR-21 24 h after PHx was greater than the increase in miR-21 due to sham operation. Data are mean relative expression ± SE, n = 4 for EtOH and CHO rats and n = 3 for chow-fed rats. For all symbols denoting significance, P < 0.05. *PHx or Sham vs. LLM; ^PHx vs. Sham; #EtOH PHx vs. CHO PHx.

Hepatic miR-21 levels were similar in CHO and EtOH livers prior to surgery and the increase in miR-21 expression in remnant livers after PHx in EtOH animals followed the same temporal pattern as in CHO rats (Fig. 1A). Surprisingly, there was a more robust miR-21 increase following PHx in EtOH rats than in pair-fed CHO rats. At 24 h after PHx this difference in miR-21 induction between EtOH and CHO animals was significant. Although 24 h after PHx is the peak of hepatocyte replication in CHO and chow-fed rats (12), liver regeneration is inhibited and delayed by chronic ethanol consumption (5, 6, 40) and very few hepatocytes have entered S phase by 24 h in the EtOH rats (Fig. 2).

Fig. 2.

Hepatocyte proliferation 24 h after PHx. A: labeling of bromodeoxyuridine (BrdU)-incorporated cells demonstrates that very few hepatocytes are replicating at 24 h after PHx in the EtOH rat. B: significantly fewer hepatocytes are replicating at 24 h after PHx in the EtOH rat (7.2%) compared with the CHO rat (35.8%). Data are means ± SE, n = 4. *P < 0.05.

To account for changes due to surgical stress, we measured miR-21 expression 24 h after sham surgery, a procedure that does not induce hepatocellular proliferation (12, 24). miR-21 expression 24 h following sham surgery in CHO rats was also increased compared with LLM samples and to a level not significantly different from the increased level following PHx (Fig. 1B). Because this finding differs from previous reports on miR-21 expression after sham surgery in chow-fed rats (3), we further investigated the miR-21 response at 24 h after PHx using chow-fed rats as a dietary control for the liquid diet pair-feeding protocol. miR-21 expression was increased 24 h after both PHx and sham in chow-fed rats to a similar extent as in CHO rats (Fig. 1B). However, in the EtOH rats, the increase in miR-21 expression 24 h following PHx was significantly greater than miR-21 induction following sham (Fig. 1B). Together these results suggest that upregulation of miR-21 may not be sufficient to drive cell cycle progression in the regenerating liver, at least in the ethanol-fed rat.

miR-21 global target enrichment analysis.

Posttranscriptional regulation of gene expression by miRNAs likely contributes to the program of gene expression changes during liver regeneration. Although miRNAs can regulate gene expression through translational repression, it has previously been demonstrated that effects of miRNA regulation can be noted in gene expression microarray data (27). We utilized the genomic scale gene expression data compiled by our group to investigate gene expression changes during regeneration in livers from ethanol-fed and pair-fed control animals (R. Vadigepalli, unpublished data, GEO no. GSE33785). The same animal pairs and RNA preparations were used in both studies. To assess the impact of the miR-21 increase on gene expression changes during liver regeneration in CHO and EtOH rats, we took an unbiased approach through global analysis of predicted miR-21 targeted gene expression.

Our experimental design provides the advantage of comparing two Δs, differences in gene expression between two conditions that differ in the level of miR-21, to identify predicted targets with expression profiles consistent with miR-21-dependent regulation. Δ₁ is the temporal differential in miR-21 expression between 6 and 24 h after PHx; Δ₂ derives from the differential between miR-21 responses in EtOH livers compared with CHO control livers. For Δ₁, a negative correlation is predicted between the changes in the expression of miR-21 and the expression of miR-21 targets during liver regeneration. Thus we expect a decrease in the differential expression of miR-21 target genes from 6 to 24 h after PHx. For Δ₂, the predicted decrease in miR-21 target expression during the time frame from 6 to 24 h after PHx would be greater in EtOH livers than in the CHO control livers, reflecting the more pronounced increase of miR-21 during that response period in EtOH livers.

To determine whether miR-21 expression correlates with altered gene expression during liver regeneration we used GSEA (37). GSEA calculates the enrichment of a gene set at either the bottom or top end of a list of genes rank ordered between two conditions, or phenotypes. If increased miR-21 expression affects gene expression during liver regeneration, we would expect enrichment of miR-21 targets among genes with higher differential expression at 6 h after PHx than at 24 h after PHx, i.e., genes that decrease expression from 6 to 24 h after PHx would be enriched for miR-21-predicted targets. For validation of GSEA to investigate miRNA effects on gene expression, we used the gene expression dataset of miR-122 inhibition in vivo from Elmen et al. (7) (see Appendix A). Interestingly, of the target prediction algorithms tested in this control dataset, only TargetScan predictions gave the expected enrichment results (see Appendix A).

TargetScan was used to generate a list of predicted miR-21 target genes. The gene expression dataset of all genes expressed in liver was ranked on the basis of the ratio of differential expressions of 24 h PHx (PHx to LLM) to 6 h PHx (PHx to LLM), and the gene list of miR-21-predicted targets was plotted on the ranked dataset. Notably, there was no enrichment of miR-21-predicted targets over this time frame in CHO livers, suggesting that the modest increase of miR-21 during liver regeneration in CHO animals is insufficient to elicit a decrease in target gene expression that is significantly enriched against the background of gene expression changes during this time frame (Fig. 3A).

Fig. 3.

Gene set enrichment analysis (GSEA) of miR-21-predicted targets after PHx. GSEA was performed for differentially expressed TargetScan-predicted miR-21 targets from 6 to 24 h after PHx in CHO (A) and EtOH (B). C: triple log₂ ratio (Δ2) of EtOH vs. CHO for differential expression from 6 to 24 h to demonstrate greater decrease in predicted targets over that time frame in EtOH than in CHO livers. GSEA-estimated FDR q values are reported; q < 0.05 is considered significant. D–F: corresponding cumulative distribution plots of differential expression of predicted miR-21 targets compared with differential expression of nontargets. P values were determined by Kolmogorov-Smirnov test. P < 0.05 is considered significant.

However, as expected for miR-21-dependent regulation of gene expression, miR-21-predicted targets are significantly enriched among genes that decrease during regeneration in the EtOH liver (Fig. 3B). This is further visualized by the significant shift to the left in the cumulative distribution of differential expression from 6 to 24 h after PHx of miR-21-predicted targets compared with nonpredicted targets (Fig. 3E). Comparing the differential expression during liver regeneration between the two diets, there is an enrichment of predicted targets (Fig. 3C) among genes with decreasing expression in EtOH livers compared with CHO livers. This outcome could also have resulted if targets were enriched among genes with a higher differential expression at 24 h after PHx than at 6 h in CHO. However, as demonstrated in Fig. 2A there is an even distribution of the predicted targets in the CHO livers. These results suggest that potential miR-21 targets are affected more significantly during regeneration in EtOH livers than in CHO control livers.

We performed the same miR-21 enrichment analysis described above with predictions from several other target prediction algorithms, including DIANA-microT (20), MicroCosm (11), microRNA.org (2), and PicTar (21) (Appendix A and Table A2). The reported performance of different miRNA target prediction algorithms varies greatly, frequently showing little overlap between targets predicted by various algorithms (34). Therefore, we separately tested the combined and overlapping predictions from multiple algorithms. None of the algorithms tested showed an enrichment of predicted targets negatively correlating with miR-21 expression in CHO liver regeneration. However, as with our GSEA validation results, only TargetScan-predicted targets had significant enrichment in the EtOH livers. Thus the TargetScan prediction set best identified the possible miR-21 targets in our system. On the basis of these findings, all further target analyses were performed with the TargetScan predictions.

Biological annotation analysis of potential miR-21 targets.

Our target analyses suggest that miR-21 has a greater effect on gene expression following PHx in the EtOH liver than in the control liver. To identify genes potentially regulated by miR-21 during regeneration in the EtOH livers we utilized the two miR-21 expression Δs described above. We identified as potential targets those TargetScan-predicted target mRNAs that decrease between 6 and 24 h after PHx in EtOH livers (Δ_1(EtOH)) and that decrease to a greater extent in the EtOH livers than in CHO livers over this period (Δ₂) (Appendix B Table A3). The subset of potential targets that also have a negative Δ_1(CHO) was further identified as possible targets of miR-21 during regeneration in control livers. Three miR-21-predicted targets identified by this analysis, Spry1, Spry2, and Timp3, had been previously identified in other systems (10, 33, 39) (Fig. 4, Appendix B Table A3). A fourth potential target selected by our analysis, Peli1, was identified by Marquez and colleagues (29) as a target of miR-21 in mouse liver regeneration (Appendix B Table A3). All four genes had been previously validated as genuine miR-21 targets in vitro (10, 29, 33, 39).

Fig. 4.

Gene ontology (GO) grouping of potential miR-21 targets identified by gene expression analysis by negative correlation. Identified genes have expression profiles that negatively correlate with miR-21 expression and are thus candidate targets for miR-21 in liver regeneration of ethanol-fed animals. Bar graphs represent differential expression at 6 and 24 h following PHx compared with LLM. All bar graphs are on a −1 to 1 log₂ scale with the following exceptions: Chd7, Dusp8, Nfib, PDCD4 are on a −2 to 2 log₂ scale and Btg2 and PPARα are on a −3.5 to 3.5 log₂ scale. *Potential regulation by miR-21 in CHO liver regeneration. §Genes that have previously been demonstrated to be miR-21 targets. †Pten is not predicted by TargetScan to be a target of miR-21.

To elucidate the functional impact of miR-21 regulation of gene expression after PHx in EtOH livers, we investigated the functionality of the ensemble of potential targets using the DAVID Functional Annotation Tool (16, 17) (Fig. 4). Four main functional groups were identified comprising 15 of the 24 potential targets in regenerating EtOH livers. Of these, regulation of transcription was the largest functional grouping, with a subgrouping of the more specific identifier “transcriptional factor activity.” Other functional groups identified were RNA splicing, telomere maintenance, and negative regulation of MAP kinase activity. The remaining nine predicted targets that were not grouped by functional annotation include genes involved in nucleocytoplasmic transport, inhibition of matrix metalloproteinases, NF-κB activation, ubiquitin conjugation, and deubiquitination. The subset of genes identified to also be possible targets in CHO liver regeneration was distributed among the functional annotation terms (Fig. 4, *) and ungrouped genes (Appendix B Table A3).

We also investigated the functionality of TargetScan-predicted targets for miR-21 not identified by our analysis as a potential target of miR-21 in liver regeneration. Some of these predicted targets have been previously validated as miR-21 targets (Fig. 4, §) and others may be miR-21 targets in other contexts. We refer to these as “nonresponsive targets.” The functional annotation and expression profiles for some of these nonresponsive targets are given as examples in Fig. 4. These genes can be classified similarly to the potential miR-21 targets identified by our criteria (data not shown). Interestingly, in the nonresponsive target gene set, several genes are grouped as negative regulators of cell proliferation, although this functional annotation was not assigned to any of the responsive targets. Given the results of this analysis, miR-21 may impact many processes during regeneration in the ethanol-treated liver, but its functions may not positively affect proliferation the way it appears to do in the context of cancer (22).

miR-21 target validation.

Our analysis identified Crebl2 as a previously unrecognized potential miR-21 target gene in both EtOH and CHO liver regeneration. Crebl2 is a transcription factor of the CREB/ATF family and in humans this gene is located on the short arm of chromosome 12 at a locus commonly deleted in malignancies (14), making it a candidate tumor suppressor gene. However, to the best of our knowledge, no functional studies have demonstrated this function for Crebl2. Recently, a requirement for Crebl2 in adipogenesis and in the regulation of adipocyte glucose uptake and lipogenesis was identified (28), indicating a role for Crebl2 in metabolism. A potential metabolic target of miR-21 in liver regeneration is interesting given the metabolic demand on hepatocytes during regeneration and the altered metabolism in ethanol-adapted livers.

We performed an in vitro reporter assay for this previously unidentified miR-21 target Crebl2. In vitro reporter assays demonstrate miRNA-mediated effects on target expression. Although these assays do not provide definite information as to whether the miRNA affects expression of the target under the physiological conditions of an in vivo context, such as liver regeneration, they do provide information as to whether a given miRNA can affect expression of a given target. Since miRNA-mediated effects can be both context and cell type specific, we used an easily transfectable cell line, HEK293, to demonstrate miR-21-mediated effects on the expression of a Crebl2 reporter. HEK293 cells were cotransfected with a dual reporter luciferase vector containing the 3′UTR of Crebl2 and either the miR-21 precursor (pre-miR-21) or a scrambled pre-miR-control oligo (Fig. 5). Overexpression of miR-21 decreased Crebl2 reporter levels by nearly 40%, indicating that miR-21 can directly regulate Crebl2 expression and may target the Crebl2 transcript during liver regeneration.

Fig. 5.

miR-21 regulation of Crebl2, a predicted target identified by global target expression analysis. A: TargetScan-predicted miR-21–Crebl2 interaction. B: transfection of HEK293 cells with pre-miR-21 causes inhibition of a luciferase reporter gene linked to the 3′ untranslated region of Crebl2. Data are means ± SE, n = 3 experimental replicates of transfections with 3 technical triplicates per replicate. *P < 0.05.

DISCUSSION

In this study, we present the novel finding that the increased miR-21 expression during liver regeneration is more robust in livers from chronically ethanol-treated animals than in livers from pair-fed control animals. The induction of miR-21 in control rats due to PHx was comparable to the induction of miR-21 due to sham operation. However, in the ethanol-fed rats, induction of miR-21 due to PHx was significantly greater than induction following sham operation. Global gene expression analysis revealed that this greater response of miR-21 during the suppressed regeneration process in the livers of ethanol-treated rats correlated with a greater effect on predicted target expression than in the normal regeneration process. Functional analyses identified potential widespread roles of miR-21 in regeneration of the ethanol-exposed liver, highlighting possible miR-21 involvement in diverse cellular functions in this context.

Although several reports on miRNAs in liver regeneration agree that there is an increase of miR-21 during liver regeneration following PHx, there are discrepancies in the details of miR-21 expression. Our data demonstrate a smaller peak miR-21 expression than reported by others. In CHO livers, the peak miR-21 expression at 24 h after PHx was ∼1.4-fold higher than in naive livers, whereas others report peak changes of 2-fold to nearly 4-fold (3, 29, 36). Remarkably, we found a similar increase of miR-21 24 h after palpation of the liver without resection in a sham surgery, in both EtOH and CHO livers. This result was reproduced in a chow-fed model as a control for the dietary conditions imposed by liquid diet pair feeding. Although neither of the cited studies performed in mice report miR-21 expression in sham operated animals, Castro and colleagues (3) did considerable work in a chow-fed rat model to profile expression following sham surgeries and found the increase in miR-21 expression to be significantly greater after PHx than after sham surgery. Reasons for this discrepancy are not clear but may include possible differences in sham surgical procedures.

Our data suggest that miR-21 induction is not sufficient to regulate cell cycle progression during liver regeneration, particularly in the ethanol-fed rat. We obtained the same temporal expression pattern for miR-21 in both the ethanol-exposed and control livers. Both our data and those of Castro and colleagues (3) demonstrate a peak of miR-21 expression at 24 h after PHx in the rat, coinciding with the peak of DNA replication in hepatocytes during normal regeneration. However, regeneration in EtOH livers is delayed and the peak of S phase does not occur until 48–72 h after PHx (5, 6). Interestingly, the peak of miR-21 expression after PHx is reported to occur at the same time and even earlier in mice, at 12–24 h (29, 30, 36), although cell cycle progression occurs later in mice than in rats. In the mouse, hepatocyte DNA synthesis peaks between 36 and 42 h following PHx, depending on the strain, much later than the reported peaks of miR-21 expression in the mouse model. Given this temporal profile, miR-21 expression during liver regeneration does not appear to be tightly linked to cell cycle progression.

During the preparation of this manuscript a report by Ng and colleagues (30) demonstrated delayed entry into S phase following PHx in mice treated with a miR-21 inhibitor. The findings of this report suggest that induction of miR-21 is necessary for promoting cyclin D1 protein expression through regulation of an upstream regulator of the translational machinery. Cyclin D1 expression is considered to be a reliable marker for G₁ to S phase transition in the regenerating liver and it is induced at an earlier time following PHx in the rat than in the mouse (1). Likewise, cyclin D1 expression is delayed in the ethanol-fed rat (15). Given the temporal expression pattern of miR-21 following PHx as outlined above, the mechanism of miR-21 regulation of cell cycle progression in mouse liver regeneration may not be directly transferable to the regenerative process in the rat. Additionally, and as the results of the present study suggest, the role of miR-21 during inhibited regeneration in the ethanol-fed rat may differ from that in the normal regeneration process. Further studies examining the effect of miR-21 inhibition following PHx in the ethanol-fed rat will aid in the elucidation of a role for miR-21 induction under conditions where regeneration is inhibited.

Two validated targets of miR-21, Btg2 and Peli1, were previously identified as possibly regulated by miR-21 in the context of liver regeneration (29, 36). As we demonstrate in Fig. 4, the criteria for our analysis do not support Btg2 (36), a cell cycle inhibitor, as being regulated by miR-21 after PHx in the ethanol-treated liver. However, our analysis does identify Peli1 (29) as a potential target, both in ethanol-treated and in control livers. Peli1 is required for NF-κB activation, an early event required for transcriptional activities important for cell cycle progression and inhibition of apoptosis (18, 32). The levels of Peli1 are increased at 6 h post-PHx in animals on both diets, but to a lesser extent in EtOH rats. This correlates with the attenuated increase in binding-activity of NF-κB in the ethanol-exposed liver compared with the control liver following PHx (41), although activation of NF-κB is reportedly earlier than 6 h. The decrease in Peli1 levels by 24 h indicates a potential role for miR-21 in attenuating the early proproliferative signals of liver regeneration. It is not clear, however, how further reducing these levels in the EtOH liver may affect regeneration. Marquez and colleagues (29) suggest miR-21 is part of a negative feedback loop with NF-κB, in which NF-κB activates miR-21 transcription in liver regeneration. However, transcriptional regulation of miR-21 in liver regeneration has not yet been characterized.

In addition to Peli1, three of the other potential targets of miR-21 during liver regeneration in the ethanol-fed rat as identified by our analysis had been previously validated by others as miR-21 targets: Spry1, Spry2, and Timp3 (10, 33, 39). We additionally validated Crebl2 as a novel target of miR-21. The recent connection of Crebl2 to adipogenesis and lipogenesis (28) makes it an intriguing potential target of miR-21 in the context of liver regeneration. The regenerative process is marked by hepatic lipid accumulation in association with the upregulation of genes involved in adipogenesis (35). Suppression of lipid accumulation inhibits regeneration following PHx (35); thus the downregulation of Crebl2, potentially by miR-21, following PHx in the ethanol-fed rat may be an inhibitory mechanism in the regenerative process. By contrast, it has also been suggested that Crebl2 may be a tumor suppressor gene (14) and thus its downregulation during regeneration in the ethanol-fed rat may be a compensatory mechanism. Further investigation is needed to elucidate the role of Crebl2 in the impairment of the regenerative process in the ethanol-fed rat.

The promiscuous nature of miRNAs makes it highly unlikely that miR-21 regulates a single factor or affects a sole process during liver regeneration. Indeed, the ensemble of potential targets identified by our analysis is involved in many different cellular processes. Multiple regulatory components of some processes appear to be affected, including genes involved with regulation of transcription, RNA splicing, telomere maintenance, and MAP kinase activity. The subset of targets predicted to also be regulated by miR-21 in normal regeneration is distributed among the functional groups identified for the EtOH liver. This indicates that some of the same processes may be affected by miR-21 during regeneration in both EtOH and control livers. However, because our data and analysis suggest miR-21 does not have a major impact on the gene expression during liver regeneration in the control animal, we did not specifically address miR-21 function in the CHO liver. The mechanisms of the regeneration process, and how it is inhibited, are less well characterized in the EtOH liver. Our findings suggest that miR-21 has many potential downstream and secondary effects, making it difficult to assess the functional contribution of miR-21 to the inhibition of regeneration caused by chronic ethanol feeding. We also cannot exclude the possibility that the potentiation of the miR-21 response in EtOH livers may be a compensatory response to the inhibition of cell cycle progression.

Our study examined the possible functional role of miR-21 in regeneration of the ethanol-exposed liver by focusing on potential targets predicted by a single algorithm and affected at the transcript level. Although the effects of miRNA regulation can be noted in gene expression data (27), miRNAs can also regulate their targets through inhibition of translation without affecting transcript levels. In addition, different target prediction algorithms identify largely distinct sets of targets. These factors imply a vast potential for miR-21 regulation through what are likely many more possible targets in the context of ethanol-inhibited liver regeneration. Our global approach to identifying the role of miR-21 in liver regeneration of ethanol-treated rats demonstrates how extensive the reach of miRNA regulation can be.

AUTHOR CONTRIBUTIONS

R.P.D and J.B.H conceived and designed research. R.P.D. performed experiments and analyzed data. R.P.D, R.V., and G.E.G. designed modes of analysis. R.P.D. and J.B.H. interpreted results of experiments and analysis. R.P.D. drafted manuscript. R.P.D., R.V., G.E.G, and J.B.H edited and revised manuscript.

GRANTS

This work was supported by the following grants: AA008714, AA014986, AA018873, AA017261, AA16919. R. P. Dippold was supported by National Institute on Alcohol Abuse and Alcoholism training grant T32 AA007467.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

APPENDIX A

Validation of GSEA.

For validation of GSEA to investigate miRNA effects on gene expression, we used the gene expression dataset of miR-122 inhibition in vivo from Elmen et al. (7). GSEA calculates an enrichment score based on the overrepresentation of a gene set at the two ends of a ranked list of genes (37). Enrichment scores are calculated within the GSEA program by an increasing running sum when a predicted target is encountered in the ranked list of genes. Statistical significance of the observed enrichment score is estimated by GSEA relative to a null distribution of enrichment scores obtained through permutations of the gene set and is presented as false discovery rate (FDR) q values.

Briefly, in this study Elmen et al. inhibited miR-122 in vivo in mice using an anti-miR-122 LNA oligonucleotide and demonstrated extensive reduction of available miR-122 lasting over 1 wk (7). As a result of miR-122 inhibition, increases in several putative miR-122 targets were observed. Levels of let-7 were demonstrated to be unchanged with miR-122 inhibition (7), and thus let-7 was used as a negative control in our enrichment analysis. For our GSEA validation analysis, predicted targets for miR-122 and let-7 were obtained from TargetScan (25), DIANA-microT (20), and MicroCosm (11) and used as gene sets. Gene expression data from miR-122-inhibited liver at day 3 and day 9 were compared with data from control liver to obtain rank-ordered gene lists, and GSEA (37) was used to determine enrichment of miR-122 or let-7 within these lists (Table A1). Interestingly, only TargetScan-predicted targets gave the expected result of significant enrichment of miR-122 targets among genes with increased expression due to miR-122 inhibition in both time points investigated as well as no significant enrichment of let-7-predicted targets for these comparisons. These results indicate that GSEA is an appropriate tool for investigating miRNA effect on gene expression.

Table A1.

Validation of GSEA for miRNA predicted target sets with gene expression data

	Anti-miR-122 day 3 vs. Control (Elmen et al.)	FDR q Value	Anti-miR-122 day 9 vs. Control (Elmen et al.)	FDR q Value
miR-122
TargetScan	LNA-antimiR-122 day 3	0.024	LNA-antimiR-122 day 9	0.007
DIANA-microT	LNA-antimiR-122 day 3	0.222	control	0.741
Microcosm	LNA-antimiR-122 day 3	0.058	LNA-antimiR-122 day 9	0.218
let-7
TargetScan	LNA-antimiR-122 day 3	0.328	LNA-antimiR-122 day 9	0.188
DIANA-microT	LNA-antimiR-122 day 3	0.736	LNA-antimiR-122 day 9	0.028
Microcosm	LNA-antimiR-122 day 3	0.298	LNA-antimiR-122 day 9	0.609

Publicly available (GEO no. GSE13948) gene expression data from Elmen et al. (7) were used. Gene set enrichment analysis (GSEA) demonstrates significant enrichment of TargetScan predicted miR-122 targets, but not predicted let-7 targets,[1] among genes with higher expression in the miR-122-inhibited liver compared with control. GSEA-estimated false discovery rate (FDR) q values for enrichment are reported.

miR-21-predicted target GSEA.

GSEA (37) was performed to determine the distribution of predicted miR-21 targets within the ranked lists of all genes. In our analyses, the gene list is comprised of all genes from the gene expression data and is ranked from low to high by the differential expression in each temporal comparison [e.g., (EtOH 24 h PHx − EtOH 24 h LLM) vs. (EtOH 6 h PHx − EtOH 6 h LLM)] and for the differential response between EtOH and CHO {[(EtOH 24 h PHx − EtOH 24 h LLM) − (EtOH 6 h PHx − EtOH 6 h LLM)] vs. [(CHO 24 h PHx − CHO 24 h LLM) − (CHO 6 h PHx − CHO 6 h LLM)]}. The gene set is the list of predicted targets from the indicated target prediction algorithm or combinations of these algorithms (Table A2). Only TargetScan-predicted targets demonstrate enrichment in genes with a greater differential expression at 6 h than at 24 h, or a decrease from 6 to 24 h, after PHx in EtOH (Table A2, Fig. 2B). Target prediction algorithms used to generate gene sets were TargetScan (25), DIANA-microT (20), MicroCosm (11), microRNA.org (2), and PicTar (21). False discovery rate q values estimated by GSEA are reported. A q value of <0.05 was considered significant.

Table A2.

GSEA of miR-21 target gene sets from multiple prediction algorithms

	Dataset Count	CHO 24 hPHx.LLM 6 hPHx.LLM	FDR q Value	EtOH 24 hPHx.LLM 6 hPHx.LLM	FDR q Value	EvC 24 hPHxLLM 6 hPHxLLM	FDR q Value
Individual algorithms
TS*	121	6 h	0.36	6 h	0.015	CHO	0.074
PT	81	6 h	0.376	6 h	0.395	CHO	0.297
DmT	54	24 h	0.326	6 h	0.411	CHO	0.218
mr	134	24 h	0.443	6 h	1	CHO	0.02
mc	218	24 h	0.233	24 h	1	CHO	0.248
Combination of algorithms
DmT+PT+TS+mr+mc	382	24 h	0.11	6 h	0.39	CHO	0.0053
Overlap of algorithms
>1 algorithm	124	24 h	0.557	6 h	0.163	CHO	0.042
>2 algorithm	42	24 h	0.728	6 h	0.202	CHO	0.059
Overlap of individual algorithms
DmT/PT	23	6 h	0.579	6 h	0.75	CHO	0.67
DmT/TS	33	6 h	0.34	6 h	0.774	CHO	0.137
DmT/mc	7	24 h	0.338	24 h	0.9	EtOH	1
DmT/mr	8	24 h	0.392	24 h	1	CHO	0.24
PT/TS	52	6 h	0.521	6 h	0.3	EtOH	0.64
PT/mc	13	24 h	0.264	24 h	1	EtOH	0.93
PT/mr	13	24 h	0.73	6 h	0.389	CHO	0.77
TS/mc	17	24 h	0.64	24 h	1	CHO	0.92
TS/mr	20	6 h	0.52	6 h	0.23	CHO	0.631
mc/mr	51	24 h	0.35	6 h	0.77	CHO	0.27
Dmt/PT/TS	14	6 h	0.26	6 h	0.29	CHO	0.39
Dmt/PT/mc	2
DmT/PT/mr	1
DmT/TS/mc	1
DmT/TS/mr	3
DmT/mc/mr	1
PT/TS/mc	4
PT/TS/mr	4
TS/mc/mr	5

Targets predicted by individual algorithms as well as overlapping targets predicted by multiple algorithms were tested. The numbers of predicted targets from each algorithm, or combination of predicted targets from different algorithms, present in the dataset are noted in the second column. Only TargetScan (TS, *)-predicted targets demonstrate enrichment in genes with a greater differential expression at 6 h than at 24 h, or a decrease from 6 h to 24 h, after partial hepatectomy (PHx) in ethanol-containing diet (EtOH). GSEA-estimated FDR q values for enrichment are reported.

CHO, carbohydrate control diet; PT, PicTar; DmT, DIANA-microT; mr, microRNA.org; mc, microcosm.

APPENDIX B

Identification of potential miR-21 targets.

TargetScan-predicted miR-21 target genes were identified as “potential targets” of miR-21 during liver regeneration in the chronically ethanol-fed rat if their expression decreased between 6 and 24 h after PHx in EtOH livers and had a greater decrease in expression over this period in the EtOH livers than in the CHO livers (Table A3). A subset of these potential targets was also identified as possible targets of miR-21 during regeneration in the CHO livers (Table A3).

Table A3.

Genes identified as potential targets of miR-21 in regeneration of the chronic ethanol-treated liver

Gene Symbol	Entrez Gene ID	Ensembl Gene ID	Ensembl Transcript ID	EtOH	CHO	EtOH vs. CHO
Acvr2a	29263	ENSRNOG00000005334	ENSRNOT00000007404	−0.524	0.266	−0.790
Atpaf1	313510	ENSRNOG00000010169	ENSRNOT00000013533	−0.688	0.556	−1.244
Brwd1	304061			−0.532	0.033	−0.565
Chd7	312974			−1.493	−0.191	−1.302
Crebl2	362453	ENSRNOG00000007115	ENSRNOT00000009324	−0.470	−0.172	−0.298
Dusp8*	117280	ENSRNOG00000033790	ENSRNOT00000044477	−1.499	−1.123	−0.375
Hnrnpu	362317	ENSRNOG00000007937	ENSRNOT00000032505	−0.471	0.048	−0.518
Krit1	306254	ENSRNOG00000028227	ENSRNOT00000016581	−0.709	−0.116	−0.593
Pbrm1	305549	ENSRNOG00000006329	ENSRNOT00000009319	−0.552	−0.049	−0.502
Pcbp2	291078	ENSRNOG00000016705	ENSRNOT00000022902	−0.536	−0.120	−0.416
Peli1§	498407	ENSRNOG00000007042	ENSRNOT00000009214	−0.769	−0.198	−0.571
Prpf4b	501560	ENSRNOG00000006632	ENSRNOT00000008938	−0.659	−0.378	−0.282
Purb	361814	ENSRNOG00000000520	ENSRNOT00000000627	−0.420	0.024	−0.444
Rps6ka3	294981	ENSRNOG00000025371	ENSRNOT00000067034	−0.670	−0.457	−0.213
Sfrs3	306141	ENSRNOG00000010058	ENSRNOT00000013342	−0.763	0.196	−0.960
Spry1§	361403	ENSRNOG00000020435	ENSRNOT00000027695	−0.548	−0.214	−0.334
Spry2§	25358	ENSRNOG00000004303	ENSRNOT00000005746	−0.395	−0.088	−0.308
Terf2	290794	ENSRNOG00000011625	ENSRNOT00000015813	−1.043	−0.144	−0.899
Timp3*§	309126	ENSRNOG00000014999	ENSRNOT00000020543	−2.171	−1.443	−0.728
Tnks	641452	ENSRNOG00000013741	ENSRNOT00000051335	−0.397	−0.195	−0.202
Tnpo1	81920	ENSRNOG00000013741	ENSRNOT00000051335	−0.494	−0.008	−0.486
Ube2d3	363014	ENSRNOG00000005933	ENSRNOT00000061818	−0.410	−0.025	−0.386
Yap1*	363014	ENSRNOG00000005933	ENSRNOT00000061822	−0.659	−0.303	−0.356
Yod1	363982	ENSRNOG00000025704	ENSRNOT00000030476	−0.777	−0.557	−0.219

Potential targets are significantly decreased from 6 h to 24 h after PHx in EtOH as determined by a 2-tailed, unpaired Student's t-test, P < 0.05, and are more substantially decreased in the EtOH liver than in the CHO liver by >0.2 log2 ratio.

^*Genes that are also significantly decreased from 6 h to 24 h after PHx in CHO liver. The log2 double ratio of (24 h PHx − 24 h LLM) − (6 h PHx − 6 h LLM) expression is given for EtOH and CHO. EtOH vs. CHO is a comparison of those log2 double ratios. Genes marked with

^§have been previously published as targets of miR-21.

LLM, left lateral and medial liver lobes.