Nonalcoholic steatohepatitis is associated with altered hepatic MicroRNA expression


  • Accepted for presentation at the European Association for the Study of Liver Disease in Milan, Italy, 2008.

  • Potential conflict of interest: Nothing to report.


The expression of microRNA in nonalcoholic steatohepatitis (NASH) and their role in the genesis of NASH are not known. The aims of this study were to: (1) identify differentially expressed microRNAs in human NASH, (2) tabulate their potential targets, and (3) define the effect of a specific differentially expressed microRNA, miR-122, on its targets and compare these effects with the pattern of expression of these targets in human NASH. The expression of 474 human microRNAs was compared in subjects with the metabolic syndrome and NASH versus controls with normal liver histology. Differentially expressed microRNAs were identified by the μParaflo microRNA microarray assay and validated using quantitative real-time polymerase chain reaction (PCR). The effects of a specific differentially expressed miRNA (miR-122) on its predicted targets were assessed by silencing and overexpressing miR-122 in vitro. A total of 23 microRNAs were underexpressed or overexpressed. The predicted targets of these microRNAs are known to affect cell proliferation, protein translation, apoptosis, inflammation, oxidative stress, and metabolism. The miR-122 level was significantly decreased in subjects with NASH (63% by real-time PCR, P < 0.00001). Silencing miR-122 led to an initial increase in mRNA levels of these targets (P < 0.05 for all) followed by a decrease by 48 hours. This was accompanied by an increase in protein levels of these targets (P < 0.05 for all). Overexpression of miR-122 led to a significant decrease in protein levels of these targets. Conclusions: NASH is associated with altered hepatic microRNA expression. Underexpression of miR-122 potentially contributes to altered lipid metabolism implicated in the pathogenesis of NASH. (HEPATOLOGY 2008;48:1810–1820.)

Nonalcoholic steatohepatitis (NASH) is part of a spectrum of nonalcoholic fatty liver disease (NAFLD) and can progress to cirrhosis in 15% of subjects.1 The mechanisms by which NASH develops and progresses to cirrhosis have not been fully defined.

MicroRNAs (miRNA) are non–protein-coding, small single-stranded RNA, typically 21 to 23 nucleotides long, that regulate gene expression via messenger RNA (mRNA) degradation or translational inhibition.2, 3 Currently, 873 human miRNAs have been identified (miRBase 11.0, April 2008). Their expression is both organ-specific and dependent on the stage of development.4, 5 Micro RNAs also regulate other important cellular processes such as metabolism, immune function, cell proliferation, apoptosis, and tissue development and differentiation.6–10 These may be particularly germane to the genesis of NASH, which is characterized by abnormal lipid metabolism, activation of apoptosis, cellular regenerative responses, and inflammation.11 There are currently no published data on the patterns of hepatic miRNA expression in NASH.

The aims of this study were to: (1) identify differentially expressed miRNAs in human NASH and validate these independently; (2) tabulate potential targets of differentially expressed miRNAs; and (3) evaluate and compare the effects of silencing and overexpression of a specific differentially expressed miRNA, microRNA (miR-122), which is known to target hepatic lipogenic genes, on the pattern of lipogenic gene expression in human NASH. It is hoped that these data would serve as the basis for future hypothesis generation and hypothesis-driven studies of the role of miRNAs in hepatic lipid metabolism and pathogenesis of NASH.


FAS, fatty acid synthase; HMG-CoA reductase; 3-hydroxy-3-methyl-glutaryl-coenzyme A reductase; miRNA, microRNA; mRNA, messenger RNA; NAFLD, nonalcoholic fatty liver disease; NASH, nonalcoholic steatohepatitis; RT-PCR, quantitative real time polymerase chain reaction; SAM, significance analysis of microarray; SREBP, sterol responsive element binding protein.

Materials and Methods

Study Population.

Consecutive subjects with the metabolic syndrome with or without suspected NAFLD were considered for this study, which was approved by the institutional review board. Informed consent was obtained, and each subject underwent routine clinical assessment, radiological, hematological, biochemical, and serological testing. The metabolic syndrome was diagnosed using the Adult Treatment Panel III criteria.12 Alcohol consumption was assessed clinically and considered to be significant when more than 20 g/day for women and more than 30 g/day for men. NAFLD was suspected in those with (1) either abnormal liver enzymes or radiological evidence of a fatty liver and negative study for other common causes of liver disease, and (2) absence of clinically significant alcohol consumption.

Subjects who gave informed consent underwent a percutaneous core liver biopsy with a 15-gauge Microvasive biopsy device under ultrasound or laparoscopic guidance. One liver core was fixed in formalin for histological assessment and another was snap frozen in liquid nitrogen and stored at −80°C for future studies. Each liver biopsy specimen was fixed and stained with hematoxylin-eosin and Masson's trichrome stain. All biopsy specimens were evaluated by a single pathologist and scored using the NASH clinical research network criteria.13 Steatohepatitis was defined by the presence of steatosis, cytological ballooning, and inflammation.13 Based on the laboratory, sonographic, and histological findings, subjects were selected and classified as follows: Test group: NASH; control group: normal liver histology, ultrasound, and liver enzymes. Exclusion criteria included a fatty liver alone, cirrhosis or bridging fibrosis, a liver biopsy less than 2 cm long for histological characterization, and the use of statins.

The identification of differentially expressed miRNAs in NASH was performed in two phases. The first involved a comparison of hepatic expression of all known human miRNAs at the time of these studies (based on miRBase 9.2, 2007, n = 474 targets) in a set of 15 subjects with NASH versus 15 controls using a micro-array chip assay (LC Sciences, Houston, TX). These study findings were further validated by quantitative real-time PCR (RT-PCR) in a total of 25 subjects with NASH and 25 controls, as defined previously.

RNA Preparation, miRNA Purity Assessment, and μParaflo MicroRNA Microarray Assay.

Total RNA was extracted from 10 mg liver tissue (approximately 1 cm biopsy length) using Trizol (Invitrogen, Carlsbad, CA) for both RT-PCR and microarray studies. For the microarray assay, before hybridization, RNA quality was further assessed by the ultraviolet spectrum intensity ratio (A260:A280 > 1.8) and the 28S:18S RNA ratio using the RNA 6000 Nano assay reagent kit (Agilent, Palo Alto, CA, P/N G2938-90030) and the eukaryote total RNA Nano assay.14 The presence of small RNA was detected in each sample from the corresponding peaks in the RNA electropherogram.

A challenge in miRNA studies is avoidance of contamination by other small RNA species, such as precursor miRNA and also separation of mature miRNAs from longer hairpin and pri-miRNA molecules. These were accomplished by isolation, separation, and purification of miRNAs using the Ambion flashPAGE fractionater system (Ambion, Austin, TX) according to manufacturer protocols. The absence of precursor miRNA contamination was further confirmed by adding miRNA precursor detection probes complementary to the precursor hairpin loops to the microarray chips. At assay conditions optimized for mature miRNA detection, none of the differentially expressed miRNAs reported in this study had a corresponding precursor signal above detection level.

Microarray analysis was done using miRBase 9.2 probe content and performed as previously published.15 Four to eight micrograms total RNA was size-fractionated using YM-100 Microcon centrifugal filter (Millipore, Billerica, MA). Small RNAs (<300 nt) isolated were 3′-extended with a poly(A) tail using poly(A) polymerase. An oligonucleotide tag was ligated to the poly(A) tail for fluorescent dye staining (Cy3 and Cy5 were used for the two RNA samples in dual-sample experiments). Hybridization was performed overnight on a μParaflo microfluidic chip using a micro-circulation pump (Atactic Technologies).16, 17 On the microfluidic chip, each detection probe consisted of a chemically modified nucleotide coding segment complementary to target miRNA (miRBase and or other RNA (control sequences) and a spacer segment of polyethylene glycol to extend the coding segment away from the substrate. After RNA hybridization, tag-conjugating Cy3 and Cy5 dyes were circulated through the microfluidic chip for dye staining. Fluorescence images were collected using a laser scanner (GenePix 4000B, Molecular Device) and digitized using Array-Pro image analysis software (Media Cybernetics).

RT-PCR Measurement of miRNA Expression.

Differentially expressed miRNAs in NASH were defined as those detected with a signal change (versus controls) in any direction with a P-value < 0.05, a minimum signal intensity > 500, and a significance analysis of microarray (SAM) test with a q-value < 5%. Four miRNAs were selected for RT-PCR validation. Two were randomly selected from overexpressed miRNAs (miR-34a, miR-146b), one from underexpressed miRNAs (miR-122), and one that was not significantly altered (miR-451). Total RNA was isolated and extracted from frozen liver tissues as previously stated. Complementary DNA was synthesized using TaqMan MicroRNA Reverse Transcriptase [Applied Biosystems (ABI), Foster City, CA] according to manufacturer's instructions. RT-PCR was performed with the RT product, TaqMan Gene Expression Master Mix, primers, and probe mix (TaqMan MicroRNA Assay kit, ABI), which was incubated in a 96-well plate at 95°C for 10 minutes, followed by 40 cycles of 95°C for 15 seconds and 60°C for 10 minutes. The relative abundance of miRNA was normalized to small nuclear RNA U6 (RNU6B, ABI) and was expressed as percent change, either increase or decrease, in NASH versus controls. RT-PCR primer sequences are the exact antisense copy of the mature miRNA sequence that can be found in the miRNA registry.18

RT-PCR for mRNA Expression of Several Lipogenic Genes.

RT-PCR was performed to examine the mRNA expression of several predicted miR-122 gene targets: fatty acid synthase (FAS), 3-hydroxy-3-methyl-glutaryl-coenzyme A reductase (HMG-CoA reductase) and their transcriptional activators, the sterol-response-element-binding protein-1c and 2 (SREBP-1c, SREBP-2). After reverse transcription (ThermoScript Reverse Transcriptase kit, Invitrogen), RT-PCR was performed using SYBR Green PCR master mix (BioRad, Hercules, CA) on an ABI Prism 7300 Sequence Detection System with the following conditions: 50°C for 2 minutes and then at 95°C for 10 minutes followed by 40 cycles of amplification (95°C for 15 seconds; 60°C for 30 seconds; 80°C for 30 seconds). Glyceraldehyde-3-phosphate dehydrogenase was used as the endogenous normalizer. Primer sequences are shown in Table 1.

Table 1. Quantitative Real-Time PCR Primer Sequences
PrimerSequence (5′ to 3′)Name
  1. F, forward; R, reverse.

SREBP-1cCAGCCCCACTTCATCAAGG (F)Sterol regulatory element binding
SREBP-2AACGGTCATTCACCCAGGTC (F)Sterol regulatory element binding
HMGCRGTCATTCCAGCCAAGGTTGT (F)3-Hydroxy-3methylglutaryl-
 TCCTGTCCACAGGCAATGTA (R) coenzymeA reduclase
GAPDHCAATGACCCCTTCATTGACC (F)Glyceraldehyde-3-phosphate

Identification of Potential miRNA Gene Targets.

Predicted gene targets of all differentially expressed miRNAs were identified using several databases, including,,,, PicTar, and TargetScan. For genes that were not elucidated from these resources, a PubMed literature search was performed. All predicted targets identified in any database were then categorized into different functional classes identified using the gene ontology molecular function hierarchy ( Genes with functions that are potentially relevant to the pathogenesis of NASH, that is, lipid and glucose metabolism, inflammation, oxidative stress and cellular response, apoptosis, cell growth and differentiation, translation regulation and protein processing, were sorted for presentation in Supporting Tables 1-6. Within major gene families, gene targets that are commonly studied or are of active interest in the context of NASH, such as lipogenesis (for example, FAS) inflammation [c-JUN kinase), were summarized for inclusion in the body of the manuscript (Table 2).

Table 2. MicroRNA Gene Targets
Functional ClassGene TargetOverexpressed miRNAUnderexpressed miRNA
  1. Gene abbreviations were adopted from Ensembl and NCBI.

Transcription factorsHomeobox protein (HOX)146b, 199a*, 455, 128b, 128a, 127, 214188, 191*, 139, 145, 92b,
  99b, 181b, 27b, 100, 16, 23a, 23b, 125b 122, 26b
 Sex-determining region Y-box (SOX)224, 34a, 200a, 199a*, 128b, 214, 99b,28, 126
   214, 99b, 181b, 24, 27b, 23a, 23b, 125b188, 145, 203, 92, 361, 122, 30d, 26b
 Hepatocyte nuclear factors (HNF)146b, 214, 23b617, 375, 92b, 26b
Protein synthesis andEukaryotic translation initiation23a, 23b, 146b, 128a, 128b, 199a*, 100122, 188, 191*, 92b, 26b,
 processing factor (eIF)125b, 13b, 27b, 16, 139, 214, 181b 145, 139
 Ubiquitin conjugating enzymes (UBE)16, 23b, 146b, 27b, 199a*, 16, 128a, 128b 125b, 214122, 139, 145, 92b, 188, 191*, 26b
 X-box binding protein 1 (XBP1), EDEM34a, 125b, 200a, 214 
Oxidative stress responseCytochrome C-oxidase99b, 423, 27b, 127, 128a, 128b601, 198, 361, 563
 Biliverdin reductase99b, 100, 221 
Cell growth and cell cycleInterferon regulatory factor (IRF)16, 24, 23a, 23b, 125b, 128a, 128b, 214145, 26b
 Cyclin dependent kinases (CDK)34a, 199a*, 99b26, 122
 Phosphatidylinositol kinases (PIK)23b, 27b26b, 145
 MYCN 122
Cell differentiationFAD104, AEBP2, MMD23b, 125b, 181a, 214, 146b, 23b, 146b26b, 122, 145, 139
  16, 27b, 30d122, 375, 92b
Anti-apoptosisBCL2L (1, 2), AVEN16, 24, 214, 200a145, 92b
Pro-apoptosisBCL2L (7, 11)27b, 125b, 16, 24, 181, 199a*, 224, 24,26b, 92b
  214, 221, 222 
 PDCD4, GAS116, 21, 199a*, 200a, 34a122, 145
 PHLDA216, 24, 128a, 128b, 181b, 199a*, 423188, 375
Anti-inflammatorySuppressor of cytokine signaling16, 199a*, 214, 23a, 23b, 27b, 125b139, 188, 191*
  (SOCS), IL4128a, 128b, 181b 
Pro-inflammatoryMitogen activated protein kinases23b, 125b, 199a*, 214, 27b, 128a, 128b145, 188, 191*, 92b, 139
  (MAPK)16, 181b, 23a 
 C-jun kinase27b, 128a, 128b, 214 
 Interleukin precursors125, 214, 16, 128a, 128b, 23b122, 188, 191*
 RAS related proteins16, 23a, 23b, 27b, 214, 146b, 125b, 127149, 26b, 122, 139, 126, 145
  199a*, 128a, 128b 
Cholesterol metabolismOxysterol binding proteins128b, 146b, 214, 181b, 23a, 23b, 199a*145, 26b, 188, 191*
  200a, 221, 125b, 27b, 224 
 HMGCR, ACAT125b, 224122, 145, 188, 375
 Oxysterol receptor LXR423145, 191*, 375, 574, 92b, 139
Fatty acid metabolismPPAR, ACSL, ACC, DGAT, SREBP, ACLY, FAS128a, 128b, 146, 24, 125b, 214, 23b, 16188, 122, 26b, 145, 191*
 AMPK, SCD, FABP, LDLR, VLDLR, ELOV30d, 199a*, 23a, 181b, 200a203
Carbohydrate metabolismALDOA, GYS, glycogenin, PRKC, PYGL23b, 27b, 34a, 221, 21, 125b, 128b, 181b122, 145, 375
Insulin signalingIRS1, myotrophin, Islet123a, 16, 27b, 128a, 128b375

Functional Validation of a Differentially Expressed miRNA (miR-122).

miR-122, the most abundant miRNA in the liver, was underexpressed in NASH (P < 0.05),19 and has been shown to be involved in lipid and cholesterol metabolism,6, 20 which are at the core of fatty liver disease. This miRNA was therefore chosen for functional validation study using both HepG2 (time course) and Huh7 cells (24 hours' transfection) by both silencing and overexpressing miR-122 (anti-miR and pre-miR, Ambion) using reverse transfection as recommended by the manufacturer (catalog no. AM1540, Ambion). Control experiments were conducted simultaneously using negative control RNA duplex (Ambion) and optiMEM culture medium (Invitrogen). The mRNA and protein expression of predicted lipogenic gene targets of miR-122 including FAS, HMG-CoA reductase, and their respective transcription factors SREBP-1c and SREBP-2 were measured by RT-PCR and western blot, respectively. A common internal calibrator was used across all experiments.

Western Blot Analysis.

Cells were lysed and human liver tissue homogenized using lysis buffer (Sigma, St. Louis, MO; catalog no. C2978). Cell and tissue lysate (25 μg protein/lane for cells and 20 μg protein/lane for liver tissue lysate) was separated by 4% to 12% Bis-Tris Nu-PAGE gel (Invitrogen) and transferred using a standard protocol. The membranes were incubated with antibodies against FAS, HMG-CoA reductase, SREBP-1c, and SREBP-2 (Santa Cruz Biotechnology Inc., Santa Cruz, CA). Signals obtained were normalized to beta-actin (Santa Cruz) for both nuclear (SREBP-1c and SREBP-2) and cytoplasmic extracts as previously published and used as a loading control.21–23

Data Analysis.

The signal value for each probe was considered to be detectable if it met two conditions: (1) signal intensity greater than 3× background standard deviation, and (2) spot CV less than 0.5 where CV = standard deviation/signal intensity. For repeating probes, transcript was listed as detectable only if 50% or more of the repeating probes were above detection level. After subtraction of background signal, data for individual probes were normalized to remove system-related variations. Normalization was performed using cyclic LOWESS (locally-weighted scatter plot smoothing) method as previously described.24 Additional data adjustment included data filtering, log2 transformation, and gene centering. Data filtering was used to remove miRNAs with normalized intensity values less than 32 across all samples.25 The intensities of individual signals were converted to a log2 scale and transformed as follows: value = ([value] − mean(gene)/(standard deviation[gene]).26 The relative expression of differentially expressed miRNAs (P < 0.05) was presented as a log2ratio of (mean signal intensity of NASH versus control). T-tests were performed between “control” and “test” groups, and T-values were calculated for each miRNA. P-values were computed from the theoretical t-distribution.26 Differentially expressed miRNAs with significant P-values (<0.05) were selected for cluster analysis using a hierarchical method with average linkage and Euclidean distance metric.27 Clustering plots were generated using TIGR MeV (Multiple Experimental Viewer) software from the Institute for Genomic Research. SAM were also employed to assess “significant” by chance and used to calculate a score for each miRNA on the basis of the change in expression relative to the standard deviation of all measurements. The score was presented as a q-value and expressed in % (q-value < 5% by SAM test corresponds to P < 0.05 by T-test).28


Patient Characteristics.

Two groups of subjects were screened (Table 3): (1) those with metabolic syndrome with abnormal liver enzymes or hepatic imaging suggestive of NAFLD and (2) those with metabolic syndrome and clinical factors that led them to be referred for evaluation for NAFLD, such as hepatomegaly, body mass index greater than 35, or diabetes. Of the first group, none of the 23 subjects screened refused a biopsy. Of the latter, 32 subjects were screened and 27 underwent a biopsy. All five subjects who did not get a biopsy refused the procedure. Test and control subjects were matched with respect to sex, race, body mass index, and features of the metabolic syndrome. As expected, subjects with NASH had higher aspartate aminotransferase and alanine aminotransferase levels than control subjects. NASH subjects were older than matched controls (52.5 years versus 40.3 years, P < 0.02), and had significantly higher low-density lipoprotein and triglyceride levels compared with the control group.

Table 3. Clinical and Demographic Data of Subjects with NASH and Matched Controls with Normal Liver Histology
 NASH (n = 25) Mean ± SDNormal Controls (n = 25) Mean ± SD
  • Hyperlipidemia is present if total cholesterol is greater than 200 mg/dL based on the Adult Treatment Plan III criteria and the American Heart Association guidelines.

  • *

    P < 0.02.

  • P < 0.0005.

  • P < 0.05.

Mean age (years)52.5*40.3
Body mass index (kg/m2)3539.54
 African American3/254/25
Metabolic risks  
 Diabetes mellitus type II11/257/25
Laboratory values  
 Aspartate aminotransferase (U/L)69.730.3
 Alanine aminotransferase (U/L)106.335.9
 Alkaline phosphatase (U/L)10480.4
 Total bilirubin (mg/dL)0.550.43
 Albumin (mg/dL)4.434.44
 Fasting blood glucose (mg/dL)115119
 Total cholesterol (mg/dL)214176
 Low density lipoprotein (mg/dL)14094
 High density lipoprotein (mg/dL)5751
 Triglyceride (mg/dL)177*67
Liver histology  
 Steatosis2.5 ± 0.30
 Cytological ballooning1.8 ± 0.30
 Mallory hyaline0.9 ± 0.40
 Lobular inflammation1.8 ± 0.40
 Pericellular fibrosis0.9 ± 0.20

miRNA Expression Profile in Subjects with NASH.

Of the 474 miRNAs that were probed, 46 were differentially expressed in subjects with NASH (Table 4). These are shown visually in the cluster analysis (Fig. 1). Twenty-three of 46 were overexpressed (log2 [NASH/control] range: 0.18-1.35), whereas 23 of 46 were underexpressed (log2 [NASH/normal] range: −0.21 to −2.06) (Table 4). All 46 differentially expressed miRNAs had a q-value < 5% (SAM test), corroborating that the observed changes were indeed significantly different compared with controls. Thus, NASH was associated with differential expression of 46 miRNAs that met significance both by T-test and SAM test.

Table 4. Log2 (NASH/Normal) Ratio of Differentially Expressed miRNAs That Reached Statistical Significance by T-test (P < 0.05) and Further Confirmed by SAM Test (q < 5%)
Underexpressed MicroRNAs (n = 23)Overexpressed MicroRNAs (n = 23)
miRNAsLog2 (NASH/Normal) Ratioq value (%)miRNAsLog2 (NASH/Normal) Ratioq Value (%)
  • A q-value of <5% by SAM test corresponds to a P-value of < 0.05 by T-test.

  • *

    Expression of miR-199a excised from the 5' arm was validated in zebrafish, and the ends mapped by cloning. miR-199a* (excised from the 3' arm) (miRBase).

Figure 1.

Cluster analysis of differentially expressed miRNAs (P < 0.05). MicroRNA signature in liver from patients with NASH and normal controls. Signal intensity was expressed as log2 ratio between NASH and controls. Bright green, underexpression; black, no change; bright red, overexpression.

Validation of miRNA CHIP Data.

RT-PCR was used to validate the microarray analysis findings. The relative expressions, expressed as percent change from controls of the four chosen miRNAs (miR-34a, miR-122, miR-146b, and miR-451), are shown in Figure 2. These data confirmed that miR-34a and miR-146b were overexpressed in NASH whereas miR-122 expression was decreased (P < 0.02 for all). The expression of miR-451, which was not significantly different from controls, was also confirmed not to be significantly different between the two groups. The expression levels of these miRNAs did not correlate with the severity of individual histological features of NASH.

Figure 2.

Quantitative RT-PCR validation of miRNA expression. The relative abundance of four specific miRNAs (25 NASH versus 25 controls) was measured using RT-PCR and expressed as mean percent change ± standard error of the mean (SEM) after normalization to glyceraldehyde 3-phosphate dehydrogenase. MiR-34a and miR-146b were confirmed to be overexpressed by 99% and 80%, respectively, whereas miR-122 was confirmed to be underexpressed by 63% (P < 0.02 for all). The expression of miR-451, which was not differentially expressed by microarray, was confirmed not to be significantly different between the two groups.

Multiplicity and Cooperativity in miRNA–Target Interactions.

One miRNA can target multiple genes (multiplicity) and multiple miRNAs may target a single gene (cooperativity). These were evaluated by plotting (1) log2 rank of all 46 differentially expressed miRNAs (23 overexpressed, 23 underexpressed) versus their corresponding number of selected gene targets as presented in Supporting Tables 1-6 (multiplicity), and (2) log2 rank of these gene targets versus the number of differentially miRNAs directed at them (cooperativity) (Fig. 3). Two overexpressed miRNAs (miR-23b, miR-16) and one underexpressed miRNA (miR-122) had more than 40 selected targets (Fig. 3A,B, respectively). Eleven of 23 overexpressed and 17 of 23 underexpressed miRNAs had fewer than 10 selected targets. Of the selected genes whose products could be potentially affected by altered miRNA expression in NASH, most were targeted by fewer than four differentially expressed miRNAs (Fig. 3C,D).

Figure 3.

Multiplicity and cooperativity of miRNA–target interactions of differentially expressed miRNAs (P < 0.05). One miRNA can target more than one gene (multiplicity), and one gene can be regulated by more than one miRNA (cooperativity). The distributions for multiplicity and cooperativity were based on (1) log2 rank of differentially expressed miRNAs versus number of gene targets (based on targets presented in Supporting Tables 1-6), and (2) log2rank of gene targets versus number of miRNAs directed at them, respectively. Multiplicity and cooperativity of differentially overexpressed miRNAs were shown in (A) and (C), respectively, and differentially underexpressed miRNAs were shown in (B) and (D), respectively. Distributions decayed exponentially. Very few overexpressed and underexpressed miRNAs regulated more than 40 predicted targets (top left A, B); however, most miRNAs controlled only a few genes (bottom right A, B). A few targets appeared to be under highly cooperative miRNA regulation (top left C, D); however, most genes were targeted by fewer than four differentially expressed miRNAs (bottom right C, D).

Potential Gene Targets of Differentially Expressed miRNAs.

Some potential targets of differentially expressed miRNAs are specifically pertinent to the pathogenesis of NASH. These included a number of transcription factors, key elements of protein translation and processing, apoptosis, cellular growth and differentiation, inflammation, and metabolism (Table 4, Supporting Tables 1-6). Key metabolic targets included cytochrome oxidase, biliverdin reductase, and several hepatic lipogenic enzymes. Sleep apnea and intermittent hypoxia has been postulated as a cofactor for liver injury in NASH.29, 30 Interestingly, all of the known hypoxia-inducible miRNAs were overexpressed in NASH, with the exception of miR-26b.31 Cyclin-dependent kinases and DNA repair enzymes were predicted targets of several differentially expressed miRNAs. Stress response enzymes such as mitogen-activated protein kinase (MAPK) and c-jun-n-terminal kinase were also predicted targets of several differentially expressed miRNAs (miR-128a, 128b, 23a, 23b, 125b, 27b).

Effects of Silencing and Overexpressing miR-122 on Targets Involved in Lipid Synthesis and Comparison with Pattern of Expression of the Targets in NASH.

There is a paucity of data on the pattern of expression of lipogenic genes in human NASH. The hepatic mRNA and protein expression of FAS, SREBP-1c, HMG-CoA reductase, and SREBP-2 was therefore first measured to provide reference data for studies of miR-122. The mRNA levels of SREBP-1c, FAS, and HMG-CoA reductase were significantly increased in subjects with NASH (Fig. 4; P < 0.05 for all). Their protein levels were also increased. Whereas SREBP-2 mRNA was not increased, the levels of mature protein were increased in subjects with NASH. On silencing of miR-122 in cell culture, there was a significant increase in the mRNA levels of FAS, HMG-CoA reductase, SREBP-1c, and SREBP-2 by 24 hours, which dropped below baseline values by 48 hours (P < 0.02 for all) (Fig. 5A-D). This was followed by a significant increase in protein levels, which peaked at 48 hours (P < 0.02 for all) (Fig. 5E, F). Conversely, after overexpression of miR-122, the mRNA levels of these target genes did not change initially but decreased by 48 hours (P < 0.03 for all) (Fig. 6A-D). This was accompanied by a corresponding decrease in levels of mature SREBP-1c and SREBP-2 and their downstream targets FAS and HMG-CoA reductase, respectively (P < 0.003 for all) (Fig. 6E, F). These findings were further validated by measuring the effects of overexpression and silencing miR-122 on HMG-CoA reductase in Huh7 cells (Fig. 7).

Figure 4.

Hepatic mRNA and protein expression of lipogenic genes in human subjects with NASH and matched controls. Messenger RNA levels of hepatic lipogenic genes were expressed as relative fold change from a single human liver sample used as the internal calibrator across all RT-PCR experiments. Overall, the hepatic mRNA levels of FAS, SREBP-1c, and HMG-CoA reductase were increased significantly in subjects with NASH compared with controls (P < 0.05), whereas the hepatic mRNA level of SREBP-2 was not different between the two groups. As shown in B and C, SREBP-2, HMG-CoA reductase, SREBP-1c, and FAS protein were increased in NASH subjects (P < 0.02 for SREBP-2, P = not significant for the rest).

Figure 5.

Silencing of miR-122 in HepG2 cells. Using RT-PCR, mRNA expression of miR-122 targets involved in lipogenesis in subjects with NASH versus controls was evaluated after silencing miR-122 in cell culture. As shown in (A) through (D), mRNA levels of FAS, HMG-CoA reductase, SREBP-1c, and SREBP-2 increased significantly at 24 hours (P < 0.05 for all) after silencing of miR-122. The mRNA levels of these lipogenic genes then dropped significantly below baseline values by 48 hours (P < 0.02 for all). This was accompanied by a significant increase in protein expression of HMG-CoA reductase, FAS, and mature SREBP1c and SREBP2 (P < 0.02 for all) at 48 hours as shown in (E) and (F) (None: no treatment; scrambled: transfection with RNA duplex containing a scrambled sequence without known gene targets; anti: transfection with miR-122 silencing RNA).

Figure 6.

Overexpression of miR-122 in HepG2 cells. Using RT-PCR, mRNA expression of miR-122 targets involved in lipogenesis was evaluated after overexpression of miR-122 in vitro. As shown in (A) through (D), mRNA levels of FAS, HMG-CoA reductase, SREBP-1c, and SREBP-2 remained unchanged at 24 hours after overexpression. This was followed by a significant decrease in mRNA expression by 48 hours (P < 0.03 for all). Protein expression of HMG-CoA reductase, FAS, and mature SREBP1c and SREBP2 decreased significantly (P < 0.003 for all) at 48 hours after miR-122 overexpression, as shown in (E) and (F) (None: no treatment; scrambled: transfection with RNA duplex containing a scrambled sequence without known gene targets; pre: transfection with miR-122 RNA).

Figure 7.

Functional validation of miR-122 in Huh7 cells. Silencing of miR-122 resulted in a significant increase in HMG-CoA transductase CR protein (P = 0.03), whereas overexpression of miR-122 led to decreased HMGCR protein expression.


Micro RNAs are important regulators of gene expression and affect mRNA stability and function.2, 3 The current study provides evidence that NASH is associated with altered hepatic miRNA expression. The potential targets of differentially expressed miRNAs are known to play a role in lipid metabolism, cell growth and differentiation, apoptosis, and inflammation; all these are key processes involved in the development and progression of NASH.11 Although our findings do not provide direct proof of the involvement of these differentially expressed miRNAs in the pathogenesis of NASH, they serve as a broad blueprint and resource for future hypothesis generation and hypothesis-driven studies of the role of miRNAs in the development and progression of NASH. This was indeed a principal objective of this study.

It should be noted that the field of miRNA research, although relatively young, is growing rapidly. Since the time of performance of the microarray studies approximately 12 months ago, several hundred new miRNA species have been identified (miRbase 11.0, April 2008, total 873 human miRs). The functions of most of these remain to be experimentally verified. In the current study, approximately 10% (46/474) of the tested miRNAs were differentially expressed in NASH. Most of the differentially expressed miRNAs were either overexpressed or underexpressed by at least 50%. Furthermore, most differentially expressed miRNAs had a very limited number of targets (based on pertinent targets selected for presentation in Supporting Tables). Conversely, the great majority of targets were affected by less than four differentially expressed miRNAs. Although these suggest that the observed changes are likely to have a biological effect, this possibility needs to be confirmed by specific hypothesis-driven studies.

The pattern of protein expression of key lipogenic genes seen in human NASH can be reproduced by silencing miR-122 in cell culture. These data are in line with and further extend previously published observations in mice.6, 20 The mRNA expression of FAS, SREBP-1c, HMG-CoA reductase, and SREBP-2 increased initially after silencing miR-122 and then declined. It is possible that miR-122 degrades the mRNA of its target genes and the initial increase in mRNA levels reflected silencing of this effect. Conversely, the decrease in mRNA levels of these genes after overexpression may reflect increased mRNA degradation by miR-122. FAS, HMG-CoA reductase, SREBP-1c, and SREBP-2 protein levels increased over a similar time course and peaked at 48 hours after miR-122 silencing. Although precursor SREBP1c and SREBP2 protein levels did not change significantly after miR-122 silencing, the mature forms showed a significant increase by 48 hours posttransfection, suggesting a possible role of miR-122 in the posttranslational regulation of SREBP maturation. Also, the progressive increase in protein levels of its targets after silencing miR-122 suggests the possibility that this was due to de-repression of the inhibitory effects of miR-122 on protein translation. These possibilities now need to be explored in future studies designed to clarify the mechanisms by which miR-122 regulates its targets.

The pattern of miRNA expression could be affected by the development of cirrhosis. This possibility was minimized by excluding those with bridging fibrosis or cirrhosis from this study. The focus of this study was to determine, in the background of the metabolic syndrome, whether NASH was associated with altered hepatic miRNA expression. Therefore, only subjects with the metabolic syndrome with either a normal liver (defined by enzymes, ultrasound, and liver biopsy) or NASH were studied. Consequently, these data do not provide any information on hepatic miRNA expression in lean normal individuals without the metabolic syndrome or those with isolated hepatic steatosis.

In summary, NASH is associated with changes in the hepatic expression of miRNAs. The potential consequences of these changes can affect insulin signaling, lipid metabolism, cellular responses to stress and apoptosis, inflammation, and response to tissue injury. Silencing miR-122 in vitro reproduced a pattern of key hepatic lipogenic gene mRNA and protein expression that is similar to that seen in human NASH. The mechanisms by which miR-122 affects lipid homeostasis need to be defined further. It is hoped that these data will lead to specific hypothesis-generated studies, which will define the role of miRNAs in the genesis of fatty liver and progression to steatohepatitis and further add to the knowledge regarding the pathogenesis of this condition.