Unique microRNA Alterations in Hepatocellular Carcinomas Arising Either Spontaneously or due to Chronic Exposure to Ginkgo biloba Extract (GBE) in B6C3F1/N Mice

Abstract

Ginkgo biloba extract (GBE) is used in traditional Chinese medicine as an herbal supplement for improving memory. Exposure of B6C3F1/N mice to GBE in a 2-year National Toxicology Program (NTP) bioassay resulted in a dose-dependent increase in hepatocellular carcinomas (HCC). To identify key microRNAs that modulate GBE-induced hepatocarcinogenesis, we compared the global miRNA expression profiles in GBE-exposed HCC (GBE-HCC) and spontaneous HCC (SPNT-HCC) with age-matched vehicle control normal livers (CNTL) from B6C3F1/N mice. The number of differentially altered miRNAs in GBE-HCC and SPNT-HCC were 74 (52 up and 22 down) and 33 (15 up and 18 down), respectively. Among the uniquely differentially altered miRNAs in GBE-HCC, miR-31 and one of its predicted targets, Cdk1 were selected for functional validation. A potential miRNA response element (MRE) in the 3’-untranslated regions (3’-UTR) of Cdk1 mRNA was revealed by in silico analysis and confirmed by luciferase assays. In mouse hepatoma cell line HEPA-1 cells, we demonstrated an inverse correlation between miR-31 and CDK1 protein levels, but no change in Cdk1 mRNA levels, suggesting a post-transcriptional effect. Additionally, a set of miRNAs (miRs-411, 300, 127, 134, 409–3p, and 433–3p) that were altered in the GBE-HCCs were also altered in non-tumor liver samples from the 90-day GBE-exposed group compared to the vehicle control group, suggesting that some of these miRNAs could serve as potential biomarkers for GBE exposure or hepatocellular carcinogenesis. These data increase our understanding of miRNA-mediated epigenetic regulation of GBE-mediated hepatocellular carcinogenesis in B6C3F1/N mice.

INTRODUCTION

Ginkgo biloba extract (GBE) is a popular herbal medicinal product in the USA and Europe (EMA/HMPC/321095/2012 2014; Saper 2014; Smith 2019) for its putative neuroprotective, cardioprotective and antioxidant effects (Mahadevan and Park 2008). GBE has been used widely for treatment of Alzheimer’s disease, cardiovascular disease, dementia, memory loss, and cerebral ischemia (Diamond and Bailey 2013; Herrschaft et al. 2012; Pereira et al. 2013). However, in the largest clinical trial for GBE, the Ginkgo evaluation of memory (GEM) study with over 3,000 volunteers of 75 years or older, twice daily doses of 120 mg GBE were not effective in reducing the incidence of dementia or Alzheimer’s disease (DeKosky et al. 2008) slowing the rate of cognitive decline (Snitz et al. 2009), reducing blood pressure or hypertension (Brinkley et al. 2010), or reducing cardiovascular mortality or events (Kuller et al. 2010). Findings of toxicity associated with Ginkgo biloba from in vitro, in vivo, and human studies are summarized in a recent review (Mei et al. 2017). The authors noted potential for drug-botanical interactions due to induction of cytochrome P-450 2C9 and 2C19 by GBE and reviewed the conflicting reports of GBE-induced bleeding events (Mei et al. 2017).

GBE was nominated to the National Toxicology Program (NTP) for a 2-year carcinogenicity bioassay for the following reasons: (1) GBE and its active ingredients, the flavonoids and ginkgolides, have biological activities, (2) GBE can be consumed in large doses over an extended period of time, (3) there are insufficient studies to evaluate for potential carcinogenicity hazard after prolonged use, and (4) some of the chemical constituents may possess mutagenic properties. NTP’s GBE rodent cancer bioassay has indicated that the chronic exposure to GBE in B6C3F1/N mouse resulted in a dose dependent increase in hepatocarcinogenicity (NTP 2013). Transcriptomic studies on GBE-exposed HCC (GBE-HCC) indicated overrepresentation of genes associated with cancer signaling, HCC development, xenobiotic metabolism, and oxidative stress pathways. In addition, increased β-catenin (Ctnnb1) mutations and alterations in Wnt/β-catenin signaling were demonstrated in GBE-exposed HCC compared to spontaneous HCC (SPNT-HCC) (Hoenerhoff et al. 2013).

It has become increasingly apparent that epigenetic mechanisms are at play in carcinogenesis in addition to various genetic alterations. Genome-wide promoter DNA methylation profiling of SPNT-HCC and GBE-HCC has identified differentially methylated promoter regions in some of the caner related genes (Kovi et al. 2019). MicroRNAs (miRNAs) epigenetically influence gene expression and have regulatory mechanisms with significant roles in regulating carcinogenesis (Lujambio and Lowe 2012). The importance of miRNAs in cancer is highlighted by the observation that nearly half of the known aberrant expression of miRNAs are located in cancer associated genomic regions (Wiklund et al. 2010). On the relationship between miRNAs and hepatocellular carcinoma (HCC) in humans, several studies have detected the aberrant expression of specific miRNAs in malignant HCC, compared to normal hepatocytes (Masaki 2009). Additionally, dysregulation of miRNAs at different stages of HCC was noticed through miRNA profiling of liver tumor tissue, serum and urine of HCC patients and healthy subjects (Dutta and Mahato 2017).

Determining the mechanisms of GBE-induced hepatocarcinogenicity in rodents may aid in assessing the health risks of human exposure. We hypothesized that genetic and epigenetic pathways dysregulated in GBE-exposed mouse HCC may reflect key pathways altered in human HCC. In this study, to identify key miRNAs that modulate GBE-induced hepatocarcinogenesis, we characterized the pattern of altered miRNAs occurring in spontaneous and GBE-exposed HCC. Based on these data, we identified the early changes of miRNA expression in non-tumor liver samples from B6C3F1/N mice exposed to GBE for 90 days (subchronic study) as well as in GBE exposed non-tumor (GBE-ENT) liver samples from the 2-year study. Additionally, using in vitro approaches, we have functionally validated a uniquely altered miR-31 in GBE-exposed HCC and one of its in silico predicted target genes, Cdk1.

MATERIALS AND METHODS

Tissue collection and miRNA extraction for miRNA array.

In the NTP bioassay, male and female B6C3F1/N mice were exposed to 0, 200, 600, and 2,000 mg/kg GBE by corn-oil gavage, 5 days a week for 2 years. Dose selection was based on the limits of gavageability and results from the 3-month studies suggesting a lack of dose-limiting toxicity at the high dose of 2,000 mg/kg (Rider et al. 2014). The GBE used for the NTP 2-year bioassay was obtained from Shanghai Xing Ling Science and Technology Pharmaceutical Company, Ltd. The key values measured in the extract included 31.2% flavonol glycosides, 15.4% terpene lactones, and 10 ppm ginkgolic acids. The NTP determined that the Shanghai Xing Ling extract and its constituents are comparable to EGb761^®, that is available in the marketplace (Catlin et al. 2018; NTP 2013). Incidences of hepatic lesions in B6C3F1/N mouse exposed to GBE are listed in Table 1. During necropsy all HCCs larger than 0.5 mm in diameter were snap frozen in liquid nitrogen. Upon examination of morphology using H & E slides prepared from the counterpart of frozen tissues, a subset of selected frozen “tumor-only” samples from GBE-HCC (n=5 from 2000 mg/kg group, 3 males and 2 females), SPNT-HCC (from vehicle control groups, n=5, 3 male and 2 female) and normal age-matched livers (CNTL, n=5, 4 male and 1 female) from vehicle control B6C3F1/N mice from the 2-year NTP bioassay were used for miRNA array analysis. Following tissue homogenization, total RNA including miRNA was extracted using the MirVANA^® miRNA Isolation Kit (Life technologies, Carlsbad, CA) following the manufacturer’s protocols. RNA integrity was measured with Bioanalyzer (Agilent Technologies, Santa Clara, CA) and all samples had RIN values greater than 7 (Catlin et al. 2018).

Table 1.

Incidences of hepatic lesions in B6C3F1/N mice exposed to Ginkgo biloba leaf extract (GBE) by gavage in subchronic (90-day) and chronic (2-year) National Toxicology Program bioassays (NTP, 2013).

A. 90-day subchronic study
	Males						Females
Dosage (mg/kg)a	0	125	250	500	1,000	2,000	0	125	250	500	1,000	2,000
Hepatocellular hypertrophy	0	0	10** [1.4]b (100%)	10** [1.7] (100%)	10** [2.7] (100%)	10** [2.7] (100%)	0	0	4* [1.0] (40%)	10** [1.2] (100%)	9** [1.6] (90%)	10** [1.9] (100%)
Necrosis, focal	0	0	1 [1.0] (10%)	0	5* [1.0] (50%)	9** [1.0] (90%)	0	0	0	0	0	1 [1.0] (10%)

B. 2-year chronic study
	Males				Females
Dosage (mg/kg)c	0	200	600	2000	0	200	600	2000
Hepatocellular hypertrophy	3 [1.7] (6%)	19** [2.6] (38%)	35** [3.0] (70%)	23** [3.2] (46%)	0	18** [2.2] (36%)	37** [2.1] (74%)	37** [2.9] (74%)
Necrosis	9 [1.9] (18%)	15 [2.1] (30%)	17* [1.9] (35%)	19* [2.3] (38%)	4 [2.3] (8%)	2 [2.0] (4%)	6 [1.5] (10%)	11 [2.0] (20%)
Hepatocellular adenoma, includes multiple	31 (62%)	46** (92%)	33 (66%)	33 (66%)	17 (35%)	37** (74%)	41** (82%)	48** (96%)
Hepatocellular carcinoma, includes multiple	22 (44%)	31* (62%)	41** (82%)	47** (94%)	9 (18%)	10 (20%)	15 (30%)	44** (88%)

^a10 male and female B6C3F1/N mice were exposed to 0, 125, 250, 500, 1,000, and 2,000 mg/kg GBE by gavage, once daily, 5 days per week for 90 days.

^bSeverity grade based on 0–4 grading scale (0 = no significant lesion, 1 = minimal, 2 = mild, 3 = moderate, 4 = severe).

^c50 male and female B6C3F1/N mice were exposed to 0, 200, 600, and 2,000mg/kg GBE by gavage, once daily, 5 days per week for two years.

Significantly different from controls

^*p < 0.05

^**p < 0.01 by the poly-3 test.

miRNA array hybridization and data analysis.

miRNA expression analysis was conducted using Affymetrix GeneChip^® miRNA 3.0 Array (Affymetrix, Santa Clara, CA) following the Affymetrix hybridization protocols. According to the manufacturer’s protocol, the Affymetrix FlashTag™ Biotin HSR labeling (Affymetrix, Santa Clara, CA) was used to label 200 ng of total RNA. Each labeled sample was hybridized for 16 hours at 48°C in a rotating hybridization oven using the Affymetrix Eukaryotic Target Hybridization Controls (Affymetrix, Santa Clara, CA) and protocol. Array slides were stained with streptavidin/phycoerythrin utilizing a double-antibody staining procedure and then washed for antibody amplification according to the GeneChip Hybridization, Wash and Stain Kit user manual. Arrays were scanned in an Affymetrix Scanner 3000 and data was obtained using the GeneChip^® Command Console Software (AGCC; Version 3.2) and Expression Console (Version 1.2). The raw miRNA data were normalized using the Robust Multichip Analysis (RMA) approach (Bolstad et al. 2003). The mouse miRNAs were retained for the analysis of differential expression. To identify differentially expressed miRNAs, the linear modeling approach and empirical Bayes statistics implemented in the limma package were applied (Ritchie et al. 2015). The pairwise analyses were performed by comparing GBE-HCCs and SPNT-HCCs with CNTL livers from B6C3F1/N mice. The p-values were adjusted by the Benjamini-Hochberg method (Benjamini and Hochberg 1995). miRNAs with absolute fold change over 2 and adjusted p-value less than 0.05 were considered as significantly differentially expressed. The in silico predicted target genes of the differentially altered miRNAs in this study were validated using the mRNA array data previously published from this study (Hoenerhoff et al. 2013; Kovi et al. 2019). The miRNA data from this study and the corresponding mRNA data are deposited in NCBI GEO database at GSE139252 and GSE29813, respectively. Partek Genomics Suite was used to perform principal component analysis (PCA) on the normalized data and to generate heat maps to compare samples.

miRNA array data validation by quantitative real-time PCR (qRT-PCR) assays.

miRNA isolated from the same frozen tissues of 5 GBE-HCC, 5 SPNT-HCC and 5 age-matched liver samples from vehicle controls were used for validation by qRT-PCR. In addition, miRNA was also isolated from two 20μm sections from formalin-fixed, paraffin-embedded (FFPE) exposed non-tumor (ENT) liver from high dose GBE-exposed group (n=9) and age-matched control normal liver (n=10) from 2-year carcinogenicity bioassay as well as FFPE sections from high dose group (2000 mg/kg) from 90-day study (n=6, 3 male and 3 female) and the age-matched vehicle controls (n=6, 3 male and 3 female) using MirVANA miRNA Isolation Kit (Life technologies, Carlsbad, CA). Quantitative RT-PCR (qRT-PCR) was used to validate miRNA array results and selected miRNAs. qRT-PCR was performed using TaqMan^® miRNA Assay (Life technologies, Carlsbad, CA) on ABI PRISM 7900HT Sequence Detection System (Applied Biosystems, Foster City, CA). snoRNA202 (Life technologies, Carlsbad, CA) was used as the endogenous control for normalization of miRNA levels. snoRNA202 was recommended as an endogenous control due to its high abundance and least variability across the normal tissues and cell lines that were tested by the manufacturer and also previously reported (Bouhaddioui et al. 2014). Relative quantities were normalized to endogenous control values and fold changes were calculated by using the 2^-ΔΔCt method and the p-values were determined as previously described (Schmittgen and Livak 2008; Yuan et al. 2006).

in silico analysis.

miR-31 was selected for the further functional analysis since it was uniquely down-regulated in GBE-HCC among the differentially expressed miRNAs. Depending on the tumor type, miR-31 has either a tumor suppressor or an oncogenic activity and is among the most frequently altered microRNAs in a large variety of human tumor types (Cottonham et al. 2010; Kim et al. 2015; Nagy et al. 2018; Siow et al. 2014; Valastyan et al. 2009; Wszolek et al. 2011; Xu et al. 2018). Potential miRNA response elements (MREs) in the 3’-untranslated region (3’-UTR) of mRNA transcripts for mouse miR-31 were investigated by performing an online search with the bioinformatics tools, TargetScan (http://www.targetscan.org), miRBase (http://www.mirbase.org) and microRNA.org (http://www.microrna.org) (Griffiths-Jones 2004).

Cell culture.

Mouse hepatoma cell line HEPA-1 cells (Hepa 1–6, ATCC CRL-1830) were cultured in Dulbecco’s modified Eagle’s medium (DMEM, Gibco, Carlsbad, CA) supplemented with 10% fetal bovine serum (FBS, Gibco, Carlsbad, CA) and 1% sodium pyruvate (Gibco, Carlsbad, CA) and incubated in 5% CO₂ at 37°C in a water-saturated atmosphere. The pMIR-pREPORT miRNA expression reported vector system was obtained from Ambion (Austin, TX). Precursors for miR-31 and the negative control preparations were obtained from Ambion (Austin, TX). These constructs were transiently transfected using Lipofectamine 2000^® (Invitrogen, Carlsbad, CA). Antisense oligonucleotides (AsOs) for mmu-miR-31 and the negative control were obtained from Exiqon (Woburn, MA). All oligonucleotides were commercially synthesized at Integrated DNA Technologies (Coralville, IA).

Western blot analysis.

Ten or 50 pmol of precursors or AsOs for miR-31 or their respective controls were transiently transfected into HEPA-1 cells by using Lipofectamine 2000^® (Invitrogen, Carlsbad, CA). After 72 hours, the cells were harvested and homogenized in RIPA buffer (ThermoFisher Scientific, Rockford, IL) to prepare whole cell homogenates. Samples were separated through 10% Tris/glycine/SDS gel electrophoresis and were transferred to a nitrocellulose membrane (ThermoFisher Scientific, Rockford, IL). The membrane was blocked with 5% nonfat milk in blocking buffer and was blotted with 1:3000 anti-CDK1 antibody (Abcam, Cambridge, MA) or 1:5000 anti-GAPDH antibody (Cell Signaling Technology, Danvers, MA). Goat anti-rabbit horseradish peroxidase (GE Healthcare Life Sciences, Pittsburgh, PA) was used as a secondary antibody. The protein bands on the membranes were developed by using a SuperSignal West Femto kit (ThermoFisher Scientific, Rockford, IL). The blots were scanned, and labeling intensities of the bands were quantitated by using ImageJ (National Institutes of Health, USA, http://imagej.nih.gov/ij/). The relative CDK1 protein levels were analyzed through densitometric image analysis and expressed relative to GAPDH levels with the mean ± S.E. of triplicate observations and p < 0.05 based on Student t-test.

Reporter plasmid constructs.

To construct luciferase reporter plasmids, target oligonucleotides were inserted between the MluI and HindIII sites downstream from the luciferase gene in the pMIR-REPORT miRNA expression reporter vector (ThermoFisher Scientific, Rockford, IL). The sequence from position 1393 to position 1413 (370~390 bp downstream from the stop codon) in the mouse Cdk1 mRNA (5’-TTCCTGGTCAGTTTCTTGCCA-3’), containing a putative mmu-miR-31 recognition element, was termed Cdk1MRE. Three copies of the Cdk1MRE (5’-CGCGTTTCCTGGTCAGTTTCTTGCCANNNNNNNTTCCTGGTCAGTTT CTTGCCANNNNNNNTTCCTGGTCAGTTTCTTGCCAA −3’; the MRE is underlined) were cloned into the pMIR-REPORT miRNA expression reporter vector (termed pMIR/3XMRE). The complementary sequence of the three copies of Cdk1MRE was cloned into the pMIR-REPORT miRNA expression reporter vector as a negative control (termed pMIR/3XMRE-Rev). Oligonucleotides containing the perfect matching sequence of mature miR-31 (5’- CGCGTCAGCTATGCCAGCATCTTGCCTA - 3’; the perfect matching sequence is underlined) was cloned into the pMIR-REPORT miRNA expression reporter vector as a positive control (termed pMIR/c-31). All constructs were verified through Sanger sequencing.

Luciferase assays.

A series of luciferase reporter plasmids or empty vector controls were transiently transfected, with the internal control pMIR-REPORT β-gal control plasmid, into HEPA-1 cells by using Lipofectamine 2000^® (Invitrogen, Carlsbad, CA). miR-31 precursors (50 pmol; Applied Biosystems, Foster City, CA), precursor controls, miRCURY AsOs for miR-31 (50 pmol; Exiqon, Woburn, MA), or AsOs controls were transiently transfected into cells. After 72 h, the cells were resuspended in 1x Passive lysis buffer (Promega, Madison, WI) or lysis solution (Applied Biosystems, Foster City, CA). Luciferase and β-galactosidase assays were performed by using a Dual Luciferase Reporter Assay System (Promega, Madison, WI) and Dual-Light System (Applied Biosystems, Foster City, CA). Luciferase activities were normalized with β-galactosidase activities.

Pathway/functional analysis.

Enrichment analysis was performed using Ingenuity Pathway Analysis software (IPA, Ingenuity Systems Inc., CA). The unique or common significantly differentially expressed miRNAs in Spontaneous or GBE-induced HCC samples as compared to the normal controls were uploaded to IPA. The IPA microRNA Target Filter module was then applied to identify the messenger RNA target genes for these microRNAs. This microRNA Target Filter module derives miRNA targets from TargetScan, TarBase, miRecords and the Ingenuity Knowledge base. The identified miRNA-targeted mRNA genes were then associated to the corresponding mRNA expression data from previous studies (Hoenerhoff et al. 2013; Kovi et al. 2019). miRNA-targeted genes with adjusted p-value < 0.05 and an absolute fold change of 2 were considered as statistically significant and used as input datasets for IPA analysis to further investigate enriched canonical signaling pathways. The Benjamini and Hochberg corrected p-value < 0.05 was applied to define the significantly enriched pathways (Benjamini and Hochberg 1995).

Statistical analyses.

Data are expressed as the mean ± S.E. of triplicate determinations or of three independent experiments, and Student’s t-test was performed on data from the western blot analysis and luciferase assays. The differences were considered significant at P < 0.05. For microarray data, error bars were not included as all statistical calculations were performed on the log2 transformed scale. The difference on the log2 scale represents the log2(fold change). The fold change is a measure of the average difference between two groups (SPNT-HCC vs. CNTL or GBE-HCC vs. CNTL).

RESULTS

Unique miRNAs were altered in HCCs arising spontaneously or due to chronic GBE exposure in B6C3F1/N mice

Principal component analysis (PCA) of global mouse miRNA expression profiles demonstrated nearly distinct clustering of age-matched vehicle control normal liver (CNTL), spontaneous HCC and GBE-exposed HCCs (Figure 1A). Significantly differentially expressed miRNAs were visualized using hierarchical cluster analysis in order to determine distinct miRNA expression patterns between the groups. There were discernible differences in miRNA expression between the groups (Figure 1B). Using a false discovery rate (FDR) of adjusted p-value of 0.05 and fold change of + or −2, the number of differentially expressed miRNAs in comparing GBE-HCC or SPNT-HCC to CNTL were 74 (52 up and 22 down) and 33 (15 up and 18 down), respectively. Unique differentially expressed miRNAs in GBE-HCC and SPNT-HCC were 59 (45 up and 14 down) and 18 (8 up and 10 down), respectively (Figure 2A, Table 2–5). There are 15 (7 up and 8 down) miRNAs common to both SPNT-HCC and GBE-HCCs. Differential gene expression profile of SPNT-HCC and that of GBE-HCC compared to age-matched normal control is depicted in Figure 2B and detailed analysis including functional pathway analysis was previously reported (Hoenerhoff et al. 2013). The list of top ten uniquely downregulated and upregulated miRNAs in GBE-HCCs includes mmu-miR-31, −3472, -miR-31*, −1893, −763, −122, 3470a, 871–3p, −470, −592, −1902, and mmu-miR-127, −379, −541, −411, −134, −376b, −409–3p, −382, −337–5p, −300, respectively (Table 2). Whereas in SPNT-HCCs, uniquely downregulated and upregulated miRNAs include mmu-miR-193, −5132, −203, −10a, −22*, −378*, −341*, −802, and mmu-miR-146b, −193b, −182, −210, −99b, −574–3p, −674, −320, respectively (Table 3). There are 15 miRNAs that are conserved in both SPNT-HCC and GBE-HCCs; downregulated and upregulated miRNAs include, mmu-miR-1948, −5115, −455*, −30e*, −184, −743a, 5133, −878–3p and mmu-miR-486, −3107, −27a, −23a, −181a, −106b*, and −339–5p, respectively (Table 4). Of the 59 and 18 unique differentially expressed miRNAs in GBE-exposed and SPNT-HCCs in mouse, 40 miRNAs are predicted to target 576 genes (383 upregulated and 193 downregulated) in GBE-HCC and 15 miRNAs with known targets are predicted to regulate 332 (214 upregulated and 118 downregulated) target genes in SPNT-HCC (Table 5).

Figure 1.

Differential miRNA expression profiling of spontaneous and GBE-exposed hepatocellular carcinomas (HCCs) in B6C3F1/N mice. A. Principal component analysis (PCA) of global mouse miRNA expression profiles demonstrated nearly distinct clustering of age-matched normal control liver (blue, CNTL), spontaneous HCC (green, SPNT-HCC) and GBE-exposed HCC (red, GBE-HCC) samples. B. Hierarchical cluster analysis (heatmap) illustrates significant differences in global miRNA expression between normal control liver, SPNT-HCC and GBE-HCC with a foldchange of 2 (up or down) and false discovery rate of ≤ 0.05. The number of samples for miRNA expression study is 5 per each group.

Figure 2.

Differential miRNA and mRNA expression profiling of spontaneous and GBE-exposed hepatocellular carcinomas (HCCs) in B6C3F1/N mice. For each miRNA and mRNA expression fold change of 2 and false discovery rate of 5% were used for generating the list of differentially expressed miRNA and mRNAs in SPNT-HCC and GBE-HCCs.

Table 2.

Unique differentially altered miRNAs in GBE exposed Hepatocellular carcinomas compared to age-matched control liver.

Transcript ID	Symbol	Fold change	Adjusted p-value
Unique downregulated miRNA (n=14)
mmu-miR-31	miR-31–5p (and other miRNAs w/seed GGCAAGA)	−26.99	0.008301
mmu-miR-3472	miR-3472 (miRNAs w/seed AAUAGCC)	−16.41	0.008762
mmu-miR-31-star	miR-31–3p (and other miRNAs w/seed GCUAUGC)	−12.68	0.008297
mmu-miR-1893	miR-1893 (and other miRNAs w/seed GCGCGGG)	−5.40	0.023396
mmu-miR-763	miR-1207–3p (and other miRNAs w/seed CAGCUGG)	−5.26	0.045606
mmu-miR-122-star	miR-122–3p (miRNAs w/seed AACGCCA)	−4.81	0.016169
mmu-miR-3470a	miR-3470a (miRNAs w/seed CACUUUG)	−4.77	0.042241
mmu-miR-871–3p	miR-871–3p (miRNAs w/seed GACUGGC)	−3.07	0.018054
mmu-miR-470	miR-470–5p (miRNAs w/seed UCUUGGA)	−2.78	0.045823
mmu-miR-592	miR-592 (and other miRNAs w/seed UUGUGUC)	−2.63	0.021041
mmu-miR-1902	miR-1902 (miRNAs w/seed GAGGUGC)	−2.54	0.008646
mmu-miR-26b-star	miR-26b-3p (and other miRNAs w/seed CUGUUCU)	−2.23	0.023396
mmu-miR-145	miR-145–5p (and other miRNAs w/seed UCCAGUU)	−2.14	0.032758
mmu-miR-26b	miR-26a-5p (and other miRNAs w/seed UCAAGUA)	−2.06	0.008646
Unique upregulated miRNA (n=45)
mmu-miR-127	miR-127–3p (miRNAs w/seed CGGAUCC)	141.55	0.008301
mmu-miR-379	miR-379–5p (and other miRNAs w/seed GGUAGAC)	108.43	0.008301
mmu-miR-541	miR-541–5p (miRNAs w/seed AGGGAUU)	99.62	0.008301
mmu-miR-411	miR-411–5p (and other miRNAs w/seed AGUAGAC)	73.04	0.008301
mmu-miR-134	miR-3118 (and other miRNAs w/seed GUGACUG)	54.95	0.008301
mmu-miR-376b	miR-376a-3p (and other miRNAs w/seed UCAUAGA)	53.94	0.008646
mmu-miR-409–3p	miR-409–3p (miRNAs w/seed AAUGUUG)	53.14	0.008301
mmu-miR-382	miR-382–5p (miRNAs w/seed AAGUUGU)	48.07	0.008762
mmu-miR-337–5p	miR-337–5p (miRNAs w/seed GGCGUCA)	46.20	0.008301
mmu-miR-300	miR-300–3p (miRNAs w/seed AUGCAAG)	44.70	0.008301
mmu-miR-431	miR-431–5p (and other miRNAs w/seed GUCUUGC)	35.72	0.00913
mmu-miR-409–5p	miR-409–5p (and other miRNAs w/seed GGUUACC)	24.54	0.008646
mmu-miR-434–3p	miR-434–3p (miRNAs w/seed UUGAACC)	19.75	0.00903
mmu-miR-433–3p	miR-433–3p (miRNAs w/seed UCAUGAU)	18.92	0.030024
mmu-miR-154	miR-154–5p (miRNAs w/seed AGGUUAU)	17.88	0.01523
mmu-miR-299-star	miR-299a-5p (and other miRNAs w/seed GGUUUAC)	17.86	0.00903
mmu-miR-299	miR-299a-3p (and other miRNAs w/seed AUGUGGG)	17.10	0.019207
mmu-miR-540–3p	miR-540–3p (and other miRNAs w/seed GGUCAGA)	15.05	0.013657
mmu-miR-411-star	miR-411–3p (and other miRNAs w/seed AUGUAAC)	13.41	0.008301
mmu-miR-495	miR-495–3p (and other miRNAs w/seed AACAAAC)	13.08	0.008646
mmu-miR-434–5p	miR-434–5p (miRNAs w/seed CUCGACU)	13.04	0.010328
mmu-miR-487b	miR-487b-3p (miRNAs w/seed AUCGUAC)	11.85	0.013657
mmu-miR-665	miR-665 (and other miRNAs w/seed CCAGGAG)	11.64	0.042241
mmu-miR-543	miR-543–3p (and other miRNAs w/seed AACAUUC)	10.97	0.008646
mmu-miR-485	miR-485–5p (and other miRNAs w/seed GAGGCUG)	10.94	0.01523
mmu-miR-329	miR-329–3p (and other miRNAs w/seed ACACACC)	9.45	0.012802
mmu-miR-673–5p	miR-673–5p (and other miRNAs w/seed UCACAGC)	8.86	0.010232
mmu-miR-3096–3p		6.56	0.020519
mmu-miR-381	miR-381–3p (and other miRNAs w/seed AUACAAG)	6.53	0.034568
mmu-miR-376a	miR-376a-3p (miRNAs w/seed UCGUAGA)	6.01	0.034568
mmu-miR-127-star	miR-127–5p (miRNAs w/seed UGAAGCU)	5.69	0.026278
mmu-miR-485-star	miR-485–3p (miRNAs w/seed GUCAUAC)	5.04	0.026278
mmu-miR-155	miR-155–5p (miRNAs w/seed UAAUGCU)	4.79	0.041353
mmu-miR-337–3p	miR-337–3p (and other miRNAs w/seed CAGCUCC)	4.62	0.012022
mmu-miR-410	miR-344d-3p (and other miRNAs w/seed AUAUAAC)	4.15	0.008301
mmu-miR-342–3p	miR-342–3p (miRNAs w/seed CUCACAC)	3.79	0.008301
mmu-miR-505–5p	miR-505–5p (and other miRNAs w/seed GGAGCCA)	3.18	0.023396
mmu-miR-3096b-5p		3.04	0.021256
mmu-miR-132	miR-132–3p (and other miRNAs w/seed AACAGUC)	2.79	0.041353
mmu-miR-532–3p	miR-532–3p (miRNAs w/seed CUCCCAC)	2.65	0.01476
mmu-miR-5099	miR-5099 (miRNAs w/seed UAGAUCG)	2.54	0.008914
mmu-miR-1981	miR-1971 (and other miRNAs w/seed UAAAGGC)	2.41	0.008646
mmu-miR-425	miR-425–5p (and other miRNAs w/seed AUGACAC)	2.34	0.008301
mmu-miR-500	miR-501–3p (and other miRNAs w/seed AUGCACC)	2.22	0.008301
mmu-miR-222	miR-221–3p (and other miRNAs w/seed GCUACAU)	2.06	0.012046

Table 5.

Differentially expressed miRNAs and corresponding differentially expressed targeted genes in GBE-HCC and SPNT-HCC. Statistical analysis identified significant correlations at the adjusted p-value of 0.05 and fold change of 2 for both miRNA and mRNA expression levels. DmiRs: Differentially expressed miRNAs, FDR: False discovery rate, FC: Fold change

	GBE-HCC			SPNT-HCC			Common
	Total DmiRs-Targets	Downregulated miRNA-Targets	Upregulated miRNA-Targets	Total DmiRs-Targets	Downregulated miRs-Targets	Upregulated miRs-targets	Total DmiRs-Targets	Downregulated miRs-Targets	Upregulated miRs-targets
miRNA	59	14	45	18	10	8	15	8	7
miRNA with targets	40	7	33	15	8	7	9	3	6
FDR 0.05, FC 2	576	202	508	332	211	204	383	116	296
Up regulation in mRNA	383	138	337	214	130	134
Downregulation in mRNA	193	64	171	118	81	70

Table 3.

Unique differentially altered miRNAs in spontaneous Hepatocellular carcinomas compared to age-matched control liver

Transcript ID	Symbol	Fold change	Adjusted p-value
Unique downregulated miRNA (n=10)
mmu-miR-193	miR-193a-3p (and other miRNAs w/seed ACUGGCC)	−5.98	0.00136324
mmu-miR-5132	miR-5132–5p (miRNAs w/seed CGUGGGG)	−5.03	0.0468464
mmu-miR-203	miR-203a-3p (and other miRNAs w/seed UGAAAUG)	−4.97	0.04142236
mmu-miR-10a	miR-10a-5p (and other miRNAs w/seed ACCCUGU)	−4.54	0.03940415
mmu-miR-22-star	miR-22–5p (and other miRNAs w/seed GUUCUUC)	−4.18	0.00993676
mmu-miR-193-star	miR-193a-5p (miRNAs w/seed GGGUCUU)	−2.63	0.00700403
mmu-miR-378-star	miR-378a-5p (miRNAs w/seed UCCUGAC)	−2.45	0.01446302
mmu-miR-378b	miR-378b (and other miRNAs w/seed UGGACUU)	−2.35	0.01085118
mmu-miR-341-star	miR-341–5p (miRNAs w/seed GGUCGGC)	−2.17	0.03763051
mmu-miR-802	miR-802–5p (miRNAs w/seed CAGUAAC)	−2.06	0.0127478
Unique upregulated miRNA (n=8)
mmu-miR-146b	miR-146a-5p (and other miRNAs w/seed GAGAACU)	92.32	4.45E-06
mmu-miR-193b	miR-193a-3p (and other miRNAs w/seed ACUGGCC)	9.24	0.01446302
mmu-miR-182	miR-182–5p (and other miRNAs w/seed UUGGCAA)	8.84	0.03020976
mmu-miR-210	miR-210–3p (miRNAs w/seed UGUGCGU)	4.88	0.03940415
mmu-miR-99b	miR-100–5p (and other miRNAs w/seed ACCCGUA)	3.56	0.00909049
mmu-miR-574–3p	miR-574–3p (miRNAs w/seed ACGCUCA)	2.96	0.0212642
mmu-miR-674	miR-674–5p (and other miRNAs w/seed CACUGAG)	2.27	0.01377644
mmu-miR-320	miR-320b (and other miRNAs w/seed AAAGCUG)	2.07	0.00909049

Table 4.

Common miRNA altered in both spontaneous HCCs and GBE-exposed HCCs in B6C3F1/N mouse.

Transcript ID	Symbol	SPNT-HCC		GBE-HCC
		Foldchange	Adjusted p-value	Foldchange	Adjusted p-value
Unique downregulated miRNA (n=8)
mmu-miR-1948	miR-1948–3p (miRNAs w/seed UUAGGCA)	−7.45	0.02977	−5.51	0.035085
mmu-miR-5115		−6.11	0.037631	−4.85	0.043805
mmu-miR-455-star	miR-455–5p (and other miRNAs w/seed AUGUGCC)	−5.40	0.010851	−3.42	0.026278
mmu-miR-30e-star	miR-30a-3p (and other miRNAs w/seed UUUCAGU)	−4.09	0.037631	−3.47	0.0401
mmu-miR-184	miR-184 (and other miRNAs w/seed GGACGGA)	−3.64	0.037631	−3.12	0.041654
mmu-miR-743a	miR-743b-3p (and other miRNAs w/seed AAAGACA)	−3.57	0.03021	−3.13	0.028495
mmu-miR-5133	miR-5133 (miRNAs w/seed CUGGAGC)	−2.67	0.038528	−2.49	0.034568
mmu-miR-878–3p	miR-878 (and other miRNAs w/seed CAUGACA)	−2.50	0.037631	−2.50	0.023242
Unique upregulated miRNA (n=7)
mmu-miR-486	miR-486–5p (and other miRNAs w/seed CCUGUAC)	3.87	0.024559	3.30	0.02441
mmu-miR-3107	miR-486–5p (and other miRNAs w/seed CCUGUAC)	3.47	0.02949	3.11	0.023795
mmu-miR-27a	miR-27a-3p (and other miRNAs w/seed UCACAGU)	3.14	0.001363	2.15	0.00913
mmu-miR-23a	miR-23a-3p (and other miRNAs w/seed UCACAUU)	2.72	0.001363	2.12	0.008301
mmu-miR-181a	miR-181a-5p (and other miRNAs w/seed ACAUUCA)	2.62	0.037631	2.60	0.023396
mmu-miR-106b-star	miR-106b-3p (miRNAs w/seed CGCACUG)	2.53	0.02247	2.82	0.008646
mmu-miR-339–5p	miR-339–5p (and other miRNAs w/seed CCCUGUC)	2.42	0.021264	2.38	0.012802

The miRNA array data was validated using the 10 most significant miRNAs by quantitative RT-PCR. The selected miRNAs include, mmu-miR-146b, −27a, −411, −300, −134, −409–3p, 433–3p, −31, and −122). The qRT-PCR results corroborated the expression patterns observed with the miRNA array (Figure 3A). Among differentially expressed miRNAs in GBE-exposed HCC compared to normal livers, mmu-miR-31 was strongly downregulated (−27-fold change in array data and ~ −129-fold by qRT-PCR) with no significant change in spontaneous HCCs (Figure 3A). Based on the target gene predications from miRNA and mRNA expression data in this study, 82 putative target genes were identified for mmu-miR-31 (Supplemental Table S1). Among differentially expressed miRNAs in spontaneous HCC compared to normal livers, miR-146b was strongly upregulated (93-fold change in array data and 9-fold change by qRT-PCR) with no significant change in GBE-exposed HCC (Figure 3A). For mmu-miR-146b, there are 37 predicted target genes (Supplemental Table S2).

Figure 3.

Comparison of expression profiles of selected miRNAs from miRNA array data and with that of qRT-PCR validation of selected miRNAs in spontaneous and GBE-exposed hepatocellular carcinomas (HCCs) in B6C3F1/N mice (A) and exposed non-tumor liver from high dose GBE-exposed group from 2-year time point (B), normalized to age matched normal control livers. Student’s t-test was performed on the qRT-PCR data and the differences were considered significant at P < 0.05 (*)

Potential biomarkers for GBE exposure or hepatocellular carcinogenesis

In order to determine if there are any miRNAs that could potentially serve as biomarkers for GBE-exposure or GBE-induced hepatocellular carcinogenesis, we have analyzed the miRNA expression in non-tumor livers from B6C3F1/N mice treated with 2000 mg/kg GBE for 90 days and for 2 years with no significant microscopic lesions (exposed non-tumor, GBE-ENT group). Three miRNAs, mmu-miR-146b, −127 and −433–3p are significantly overexpressed in exposed non-tumor (ENT) liver from high dose group from the 2-year time point (Figure 3B). Interestingly, miRs-411, 300, 127, 134, 409–3p, and 433–3p were overexpressed in exposed liver from the high dose group from the 90-day time point compared to vehicle control group (Figure 4).

Figure 4.

Quantitative RT-PCR of miRNAs in livers from B6C3F1/N mice treated with 2000 mg/kg GBE for 90 days. The miRNA expression in GBE livers was normalized to normal livers of age matched vehicle control. snoRNA202 expression level was used as the endogenous control for normalization of miRNA levels. Relative expression levels were calculated by using the 2^-ΔΔCt method. Student’s t-test was performed on the qRT-PCR data and the differences were considered significant at P < 0.05 (*).

Identification of miR-31 target site within the 3’-UTR of Cdk1

In the in-silico analysis using the bioinformatics tools, TargetScan, miRBase and microRNA.org, mouse miR-31 was predicted to target 3’-UTR of Abcc4, Mmp12, Sp5, Cd48, Lgr5, Epst1, Glu1, Rnf43, Hmox1, Cdk1, E2f1/2 etc. Some of the previously reported targets of miR-31 in hepatocellular carcinomas include CDK2, HDAC2 (Kim et al. 2015), Smad4 (Ruoming et al. 2015), SP1 (Zhao et al. 2017), E2F2 (Li et al. 2015b), etc. Cdk1 mRNA was one of the predicted targets of miR-31 among cancer-related genes. The predicted interaction between miR-31 and targeting site within the 3’-UTR of Cdk1 (mirSVR score; − 0.9277) is illustrated in Figure 5A. The functional validation of Cdk1 and miR-31 was undertaken since the miR-31 binding sequence in the Cdk1 mRNA is highly conserved across species and Cdk1 was altered in the corresponding mRNA dataset.

Figure 5.

Regulation of mouse CDK1 protein levels through ectopic overexpression or silencing of miR-31 in HEPA-1 cells. A. Predicted target sequence for miR-31 in mouse Cdk1 mRNA. The numbering refers to the start codon ATG in translation starting with A as 1; the coding region continues to position 891, with a stop codon at positions 892 to 894. The sequence of the putative Cdk1MRE (red box) is located 370 to 390 bp downstream from the stop codon in the 3’-UTR of mouse Cdk1 mRNA. CDK1 and GAPDH levels were determined through immunoblot analysis. HEPA-1 cells were transfected with 10 or 50 pmol of precursors (B) or 10 or 50 pmol of AsOs (C). Lanes represent results for one of triplicate observations. The relative CDK1 protein levels were analyzed through densitometric image analysis and expressed relative to GAPDH levels. Values are the mean ± S.E. of triplicate observations. *p < 0.05; Significantly different from precursor or AsOs control.

Effect of miR-31 on mouse CDK1 protein expression in mouse hepatoma cell line

The correlation between miR-31 and Cdk1 was examined by evaluating CDK1 expression in HEPA-1 cells after overexpression or silencing of miR-31. In these experiments, overexpression was achieved by transfecting cells with the precursor of mmu-miR-31. miR-31 silencing was achieved by transfecting cells with antisense oligonucleotides (AsOs) of miR-31 designed to specifically target mature miR-31. Ectopic expression of miR-31 at 50 pmol level significantly decreased CDK1 protein level, as measured through immunoblot analysis with an antibody specific to CDK1, compared with the GAPDH control (Figure 5B). At 10 pmol of miR-31 precursor, CDK1 protein level tended to decrease. Conversely, transfection with 10 and 50 pmol of AsOs for miR-31 significantly increased endogenous CDK1 protein levels, compared with GAPDH (Figure 5C). These results confirmed that miR-31 regulated mouse CDK1 protein level in HEPA-1 cells.

Regulation of CDK1 by MRE in 3’-UTR of CDK1 in mouse hepatoma cell line

To determine whether the putative mouse miR-31 recognition element in the 3’-UTR of Cdk1 mRNA (Cdk1MRE) was functionally recognized by miR-31, luciferase assays were performed in HEPA-1 cells. The luciferase activity of the pMIR/3XMRE construct, which contained three copies of the putative miR-31 Cdk1 recognition site downstream from the luciferase gene, was significantly decreased when pMIR/3XMRE was cotransfected with 50 pmol of the precursor for miR-31 (p < 0.001, Figure 6A). The negative control pMIR/3XMRE-Rev construct, which contained three copies of the inverted putative recognition site, had no effect when precursor for miR-31 was cotransfected (Figure 6A). The luciferase activity of the positive control pMIR/c-31 construct, which contained the perfect matching sequence of mature miR-31, was significantly decreased when precursor for miR-31 was cotransfected (p < 0.01, Figure 6A). Conversely, the luciferase activity of pMIR/3XMRE construct was significantly increased when pMIR/3XMRE was cotransfected with 50 pmol of the AsOs for miR-31 (p < 0.01, Figure 6B). The negative control pMIR/3XMRE-Rev construct had no effect when AsOs for miR-31 was cotransfected (Figure 6B). The luciferase activity of the positive control pMIR/c-31 construct was significantly increased when AsOs for miR-31 was cotransfected (p < 0.001, Figure 6B). Taken together, these results indicate that miR-31 functionally recognizes Cdk1MRE to negatively regulate the expression of CDK1.

Figure 6.

A series of luciferase reporter plasmids or empty vector controls were transiently transfected, with the internal control pMIR-REPORT β-gal control plasmid, into HEPA-1 cells with 50 pmol of the precursors for miR-31 or control precursors (A) or with 50 pmol of AsOs for miR-31 or AsO controls (B). Luciferase activities were normalized with β-galactosidase activities. Each column represents the mean ± S.E. of triplicate determinations. **p < 0.01, ***p < 0.001; Significantly different from precursor or AsO control.

Pathways that are altered by miRNA-mediated regulation of gene expression in mouse HCC

Analysis revealed significant correlations at the fold change of 2 and an adjusted p-value of 0.05 level for 576 and 332 differentially expressed miRNA target genes in GBE-HCC and SPNT-HCC compared to control, respectively. The number of positive correlations was generally higher than negative correlations in GBE-HCCs (Table 5). Since there were significant differences in differentially expressed miRNAs and mRNAs between SPNT-HCC and GBE-HCC, we performed a core and comparison analysis with GBE-HCC or SPNT-HCC in relation to CNTL samples using Ingenuity Pathway Analysis (IPA) software. Several pathways had enrichment of miRNA targeted genes including LPS/IL-1 mediated inhibition of RXR function, AHR receptor signaling, and NRF2-mediated oxidative stress response in GBE-HCC and some of the cancer related pathways such as cyclins and cell cycle regulation, apoptosis signaling, and molecular mechanisms of cancer in both SPNT-HCC and GBE-HCCs (Figure 7). In addition to several pathways that are relevant to cell growth and proliferation in general, pathways involved in immune response, xenobiotic metabolism, and NRF2-mediated oxidative stress response were most likely to be influenced by the differentially regulated miRNAs in SPNT-HCC and GBE-HCCs.

Figure 7.

miRNA target prediction and functional analysis: A. Predicted target genes of differentially expressed miRNAs in SPNT-HCC and GBE-HCC are analyzed for significant enriched functional categories (Fisher’s exact test p values). Significantly enriched categories including cancer related pathways are shown with -log [p-value]. Blue bar= spontaneous HCC and orange bar = GBE-HCC.

DISCUSSION

In the NTP 2-year GBE rodent cancer bioassay, there was a dose dependent increase in hepatocellular carcinoma incidence (NTP 2013). A significant hepatocarcinogenic response was observed in mice at the lowest dose tested of 200 mg/kg GBE, corresponding to a human equivalent dose of 16 mg/kg, which is approximately 4 times higher than the recommended dose of 240 mg GBE in a 60 kg human (4 mg/kg). This estimation is consistent with toxicokinetic work comparing systemic exposures of GBE in NTP rat studies with data from a rat study using a different GBE test material and a human pharmacokinetic study (Waidyanatha et al. 2019).

Aberrant regulation of non-coding RNAs (ncRNAs) has been reported to have an important role in tumor development and progression (Iorio and Croce 2012). The ncRNAs can be divided into two major categories, long non-coding RNAs and short non-coding RNAs (miRNAs) (Lujambio and Lowe 2012). miRNAs are highly conserved small non-coding RNA molecules regulating the expression of approximately 30% of all human genes at the transcriptional and post-transcriptional levels (Fu et al. 2012). Numerous miRNAs are reported to be differentially expressed in human hepatocellular carcinomas (Xu et al. 2018). In this study, we have compared the miRNA expression profiles of mouse hepatocellular carcinomas arising either spontaneously or due to chronic exposure to GBE. Further, we have correlated the miRNA-gene interactions by comparing the differential expression profiles of miRNA and mRNA of SPNT-HCCs and GBE-HCCs. By comparing miRNA profiles of spontaneous HCCs, GBE-exposed HCCs, GBE-exposed non-tumor liver from the 2-year timepoint, and GBE-exposed liver from the 90-day study, we hypothesized that a subset of these unique differentially altered miRNAs likely play an important role in hepatocellular carcinogenesis in SPNT-HCC or GBE-exposed HCC, by targeting genes involved in cancer signaling pathways.

Using the Affymetrix GeneChip^® miRNA 3.0 Array platform, when compared to normal livers, there were 18 and 59 unique differentially altered mouse miRNAs in HCCs arising either spontaneously or due to chronic GBE exposure in B6C3F1/N mice, respectively, at FDR ≤ 0.05 and fold change 2. There are 15 miRNAs common to both spontaneous and GBE-exposed mouse hepatocellular carcinomas. Among these miRNAs, miR-21, miR-221, miR-226, miR-224, miR-122 and miR-199a were reported to be consistently altered in human HCCs (Meng et al. 2007; Wang et al. 2008; Wang and Lee 2011). These altered miRNAs in HCC were reported to dysregulate proliferation, and/or apoptosis, angiogenesis, metastasis through targeting various molecules including PTEN, CDKN1B/p27, SMAD4, and CDKN1C/p57 (Fornari et al. 2008; Meng et al. 2007; Wang et al. 2013). Down-regulated miRNAs (mir-122, mir-199a) often function as tumor suppressors (Gramantieri et al. 2007; Hou et al. 2011).

Of the 59 and 18 unique differentially expressed miRNAs in GBE-exposed and SPNT-HCCs in mouse, 40 miRNAs are predicted to target 576 genes (383 upregulated and 193 downregulated) in GBE-HCC and 14 miRNAs with known targets have been predicted to regulate 332 (214 up- and 118 down-regulate) target genes in SPNT-HCC. However, there are no known or predicted target genes for the subset of differentially expressed miRNAs in both GBE-HCC and SPNT-HCC; these miRNAs may have unique and novel target genes. Several miRNAs including miR-31, miR-181a, miR-122, miR-31–5p, miR-221, miR-155, and miR-146b are reported to be altered in human HCCs and are also altered in mouse HCCs (Xu et al. 2018). Two miRNAs, miR-221 (upregulated) and miR-122 (down regulated) in GBE-HCCs were reported to be consistently altered in human HCCs (Liu et al. 2018; Simerzin et al. 2016). Six unique miRNAs predicted and validated as prognostic markers in human HCC from various independent microarray and RNA-Seq based data sets in TCGA are miR-149, miR-139, miR-3677, miR-146b, miR-584 and miR-31 (Nagy et al. 2018). Of these six unique human HCC biomarkers, mmu-miR-31 and mmu-miR-146b are uniquely differentially expressed in mouse GBE-HCC and SPNT-HCC, respectively, from this study.

One of the miRNAs uniquely altered in GBE-exposed HCC, mmu-miR-31 was downregulated in GBE-exposed HCC with no significant change in spontaneous HCC compared to normal livers. miR-31 is described as both a tumor suppressive and an oncogenic miRNA in a variety tumor types (Laurila and Kallioniemi 2013). For instance, miR-31 was significantly overexpressed in colon and oral cancer (Cottonham et al. 2010; Siow et al. 2014), whereas, in HCC and bladder cancer (Wszolek et al. 2011), miR-31 was markedly downregulated. In HCC, miR-31 was reported to act as a tumor suppressor by regulating the cell cycle and epithelial mesenchymal transition, and the low expression of miR-31 was associated with a poor prognosis in HCC patients (Kim et al. 2015; Li et al. 2015b; Liu et al. 2010; Rasheed et al. 2015; Ruoming et al. 2015; Sossey-Alaoui et al. 2011; Zhao et al. 2017). Although some target genes of miR-31 in human HCC have been reported, including HDAC2, CDK2 (Kim et al., 2015), NDRG-3 (Du et al. 2017), and SP1 (Zhao et al. 2017), very little is known about the functional role of miR-31 in rodent HCCs resulting from chronic chemical exposure. Among the predicted target genes of miR-31 include genes involved in cell cycle (CDK1, MCM2), Wnt pathway (PLCE1, SMO), hepatic cholestasis (IL1RAP, IL1RN, SLC10A1), LPS/IL1-mediated inhibition of RXR function (FMO4, IL1RAP, IL1RN, SLC10A1), role of macrophages and fibroblasts (CXCL12, FCGR1A, PLCE1, SMO).

We discovered a putative miRNA response element (MRE) in the 3’-UTR of cyclin dependent kinase Cdk1 mRNA and hypothesized that miR-31 expression levels could alter the protein levels of CDK1. We tested this hypothesis by ectopic expression of synthetic precursor and AsOs and demonstrated that Cdk1 is negatively regulated post-transcriptionally by miR-31 and can potentially play a similar role in GBE-exposed HCC in B6C3F1/N mice. CDK1, a key protein in cell cycle regulation, forms complexes with its cyclin partners that phosphorylate a variety of target substrates. Phosphorylation of these proteins leads to cell cycle progression (Enserink and Kolodner 2010). In HCC, CDK1 protein overexpression is directly related to advanced stage, portal invasion, intrahepatic metastasis, poor differentiation, large tumor size and poor prognosis in a large study of HCC patients (Ito et al. 2000). CDK1 protein overexpression may play a significant role in GBE-induced hepatocarcinogenesis in mice. In addition to the regulation of global gene expression via RNA-polymerase II, Cdk/cyclin complexes have been implicated in cell cycle-independent deregulation of Wnt/β-catenin signaling pathway by potentiating β-catenin-mediated transcription within the nucleus (Firestein et al. 2008) and phosphorylation of Lrp6, a Wnt co-receptor at the plasma membrane that contributes to stabilization of β-catenin (Davidson and Niehrs 2010). Selective accumulation of point mutations and short deletions in Ctnnb1 gene in GBE-exposed mouse HCCs has been reported by our group (Hoenerhoff et al. 2013). The interplay between the genetic alterations in the β-catenin gene and epigenetic regulation of members of the Wnt/β-catenin pathway either by miRNA (Peng et al. 2017) or by promoter DNA methylation (Kovi et al. 2019), may synergistically contribute to hepatocarcinogenesis in GBE-exposed mice.

It is a well-known fact that a single miRNA can target multiple genes, while multiple miRNAs can target a single gene. Recently, multiple mRNA targets of miR-31 other than CDK1 were reported, including LATS2, PPP2R2A (Liu et al. 2010), WAVE3 (Sossey-Alaoui et al. 2011), HDAC2, CDK2 (Kim et al. 2015), GNA13 (Rasheed et al. 2015), SMAD4 (Ruoming et al. 2015), E2F2 (Li et al. 2015b), and SP1 (Zhao et al. 2017). In addition to these genes, additional putative targets may also play important roles in carcinogenesis (Supplemental Table S1). Therefore, future studies should investigate the potential roles played by miR-31 and its multiple targets especially in chemical induced hepatocellular carcinogenesis in rodents.

To the best of our knowledge, this is the first study to evaluate the global miRNA expression profiles in mouse hepatocellular carcinomas resulting from chemical exposures. Among the miRNAs altered in spontaneous HCC, downregulated miR-146b and miR-193b are known to be dysregulated in human HCC (Li et al. 2017; Xu et al. 2010). Target genes of miR-146b include TRAF6, NFKB1, KIT, CDKN1A, AKT3, HSPA1B, and MMP16 (Hsu et al. 2011). Further studies for the functional analysis of these miRNAs may help to provide a better understanding of rodent and human hepatocellular carcinogenesis. Some of the miRNAs which are dysregulated both in SPNT-HCC and GBE-HCC are also reported to be differentially altered in human HCC, and these include miR-27a (Li et al. 2015a), miR-23a (Mohamed et al. 2017), miR-27a and miR-23a (Huang et al. 2008).

The differentially altered miRNAs in GBE-exposed HCC in mice were comparable to those altered in human HCC, such as, miR-31, miR-122*, miR-145, miR-222, miR-300, miR-409, and miR-411. Especially, down-regulation of miR-145 (Ding et al. 2016; Law et al. 2012; Liang et al. 2018) and up-regulation of miR-222 (Liu et al. 2018; Wong et al. 2010; Zhang et al. 2015) are well known to be related to hepatocellular carcinogenesis. However, some miRNAs that are not commonly altered (e.g., miR-229*, and miR-1949) or altered in the opposite direction (e.g., miR-134, miR-127, miR-329 and miR-431), miR-134 (Zha et al. 2014), miR-127 (Huan et al. 2016), miR-329 (Zhou et al. 2016), miR-431 (Pan et al. 2015) in human HCC, suggests that GBE-exposed HCC may have a distinct miRNA signature when compared to human HCCs.

To determine if there are any miRNAs that could potentially serve as a biomarker for GBE exposure and/or early biomarkers of hepatocellular carcinogenesis, we have analyzed miRNA expression in GBE exposed non-tumor (GBE-ENT) mouse livers from the 2-year study and from the 90-day study (there were no tumors or significant morphological changes at the 90-day time point for the selected samples). The expression of miRs-411, 300, 127, 134, 409–3p and 433–3p were significantly increased in the GBE-treated group from the 90-day study compared to age-matched vehicle control group, and the expression of miRs-146b, 127 and 433–3p were significantly increased in GBE-ENT liver tissue from 2-year time point compared to the age-matched vehicle control group. In the 90-day GBE study, although hepatocellular hypertrophy and focal necrosis were found in the livers, there were no preneoplastic hepatic foci (NTP 2013) (Table 1). Based on these results, we could speculate that the miRNAs that are altered in GBE-HCC may indicate biomarkers of exposure or biomarkers of hepatocellular carcinoma, miRNAs that are altered in GBE-ENT livers at the 2-year timepoint may indicate biomarkers of GBE exposure but not for HCC and unique miRNAs that are altered in GBE-exposed livers at 90-day may indicate early biomarkers of GBE exposure or early HCC even though there are no histological lesions suggestive of a preneoplastic process. Using this approach, of the selected 10 miRNAs, it may be suggested that miR-433–3p may be a biomarker of GBE exposure as it was altered in all the three groups (HCC, liver from 90 days and GBE-ENT liver from 2 years); miR-300 and 409–3p may be an early biomarker of GBE-exposed HCC since it was present in GBE-HCC and GBE-exposed liver at 90 days but not significantly altered in GBE-ENT at 2 years, and also it was not present in the miRNAs that are common to both GBE-HCC and SPNT-HCC. Although there is some evidence linking hepatocellular carcinogenesis and miR-300 (Zhang et al. 2016), validation of additional miRNAs from HCCs resulting from various chemical-exposures or from livers after short term exposure are necessary to validate these findings. In addition, although this qRT-PCR analysis was performed using miRNA extracted from FFPE liver sections, all the miRNAs analyzed were amplified with relative ease, indicating that archival FFPE tissues can be leveraged for miRNA-based biomarker identification.

In conclusion, this work shows that miRNAs are uniquely altered in GBE-exposed HCC and spontaneous HCC in mice, and down-regulated miR-31 may play a significant role in GBE-induced hepatocellular carcinogenesis by targeting Cdk1 mRNA and in turn, potentially synergistically contributing to β-catenin/Wnt signaling pathway. In addition, putative biomarkers of exposure and/or mouse hepatocellular carcinogenesis may be discovered following the approach outlined in this manuscript.

Abbreviations

AsOs

Antisense oligonucleotides

CNTL

Control

DEGs

Differentially expressed genes

DmiRs

Differentially expressed miRNAs

FC

Fold change

FDR

False discovery rate

FFPE

Formalin-fixed paraffin embedded

GBE

Ginkgo biloba extract

H&E

Hematoxylin and eosin

HCC

Hepatocellular carcinoma

miRNA

MicroRNA

mmu-miR

Mus musculus microRNA

MRE

miRNA response element

NTP

National Toxicology Program

RMA

Robust multiarray normalization

SPNT

Spontaneous