Multiomics analysis of the impact of polychlorinated biphenyls on environmental liver disease in a mouse model

Abstract

Exposure to high fat diet (HFD) and persistent organic pollutants including polychlorinated biphenyls (PCBs) is associated with liver injury in human populations and non-alcoholic fatty liver disease (NAFLD) and steatohepatitis (NASH) in animal models. Previously, exposure of HFD-fed male mice to the non-dioxin-like (NDL) PCB mixture Aroclor1260, dioxin-like (DL) PCB126, or Aroclor1260 + PCB126 co-exposure caused toxicant-associated steatohepatitis (TASH) and differentially altered the liver proteome. Here unbiased mRNA and miRNA sequencing (mRNA- and miRNA- seq) was used to identify biological pathways altered in these liver samples. Fewer transcripts and miRs were up- or down- regulated by PCB126 or Aroclor1260 compared to the combination, suggesting that crosstalk between the receptors activated by these PCBs amplifies changes in the transcriptome. Pathway enrichment analysis identified “positive regulation of Wnt/β-catenin signaling” and “role of miRNAs in cell migration, survival, and angiogenesis” for differentially expressed mRNAs and miRNAs, respectively. We evaluated the five miRNAs increased in human plasma with PCB exposure and suspected TASH and found that miR-192–5p was increased with PCB exposure in mouse liver. Although we observed little overlap between differentially expressed mRNA transcripts and proteins, biological pathway-relevant PCB-induced miRNA-mRNA and miRNA-protein inverse relationships were identified that may explain protein changes. These results provide novel insights into miRNA and mRNA transcriptome changes playing direct and indirect roles in the functional protein pathways in PCB-related hepatic lipid accumulation, inflammation, and fibrosis in a mouse model of TASH and its relevance to human liver disease in exposed populations.

1. Introduction

Polychlorinated biphenyls (PCBs) are persistent organic pollutants (POPs) which have been implicated in a range of adverse human health effects, including type 2 diabetes and cancer (Simhadri et al., 2020; van den Berg et al., 2017). Although banned from manufacture in the U.S. in 1979, PCBs still leach into the environment by thermal processes, e.g., from materials in buildings (Shen et al., 2021) and exposure to PCBs in the U.S. population is considered “ubiquitous” (Curtis et al., 2021). PCBs accumulate in soil, aquatic sediments, and in species who consume these sources, e.g., fish, cows (including dairy), and humans (Shipley et al., 2017). There are 209 theoretical PCB congeners with one or more chlorine atom substituents (EPA). PCB toxicity varies according to the degree and position of the chlorine substitution for hydrogen on two linked benzene rings. The biological effects of PCBs are historically attributed to receptor-based modes of action. Coplanar, or “dioxin-like” (DL) PCBs are more toxic and activate the aryl hydrocarbon receptor (AHR), while the non-coplanar, or “non-dioxin-like” (NDL) PCBs activate nuclear receptors, e.g., constitutive androstane receptor (CAR) and pregnane X receptor (PXR) (Safe et al., 1985). In addition, we reported that Aroclor1260 (a NDL PCB mixture) attenuated phosphorylation of epidermal growth factor (EGFR) and Hepatocyte Nuclear Factor 4 Alpha (HNF4) in the livers of male C57Bl/6 J mice, possibly contributing to the transition from steatosis to steatohepatitis (Hardesty et al., 2019b).

PCBs contribute to toxicant-associated steatohepatitis (TASH), a form of nonalcoholic fatty liver disease (NAFLD) (Heindel et al., 2017; Wahlang et al., 2019) and are associated with hepatocellular carcinoma (HCC), altered liver enzymes and mortality in exposed populations (Cave et al., 2010; Clair et al., 2018). Fatty liver diseases may progress to steatohepatitis, cirrhosis, HCC, and death or a need for liver transplantation (Cave et al., 2016). We recently reported that Aroclor1260 and PCB126 (a DL PCB) altered the murine hepatic proteome, attenuating the liver’s protective responses against nutritional stress from a high fat diet (HFD) (Hardesty et al., 2019b; Jin et al., 2020). We have also reported significant changes in post-transcriptional RNA chemical modifications in the livers of male C57Bl/J6 mice on a HFD and exposed to PCBs, suggesting that PCB exposure alters the global epitranscriptome (Klinge et al., 2021). However, little is known regarding how PCBs alter microRNA (miRNA) profiles to reprogram the liver and how this may contribute to TASH.

The role of miRNA epigenetic regulation of RNA translation in hepatic pathologies is well studied. miRNAs are an evolutionarily conserved class of short non-coding RNAs (ncRNAs) that function as post-transcriptional regulators of gene expression and interact with a broad range of signaling pathways playing pivotal roles in biological processes (He and Hannon, 2004). Several miRNAs have been implicated as regulating key targets associated with the development of NAFLD, e.g., miR-9 (Ao et al., 2016), miR-21 (Braza-Boïls et al., 2016), miR-103a (Soronen et al., 2016), and miR-122 (Yamada et al., 2013). The expression of miRNAs in liver tissue is regulated by many factors, including alcohol (Heo et al., 2019), diet (Que et al., 2020), and environmental exposures (Ferragut Cardoso et al., 2020). Exposure to a single combined dose of both Aroclor1260 + PCB126 was associated with proteome changes associated with gene ontology (GO) processes related to epigenetic mechanisms (Jin et al., 2020). Nonetheless, the impact of PCB exposure on the transcriptome and miRNA expression in a HFD-induced model of steatohepatitis has not been evaluated.

This study examined the miRNA and mRNA transcriptome in response to HFD + /− exposure to Aroclor1260 or PCB126 alone or in combination in an unbiased approach using mRNA sequencing (RNA-seq) and miRNA sequencing (miRNA-seq) to identify pathways and mechanisms that may account for the alterations in the liver proteome, e.g., liver injury markers and inflammation-associated enzymes, reported previously (Jin et al., 2020). Here we tested the hypothesis that HFD in combination with PCB exposure alters hepatic miRNA expression with selected miRs regulating the expression of proteins identified in our previous study that contribute to liver injury in TASH. We identified NDL and DL PCB-exposure unique and overlapping changes in mRNA and miRNA abundance that are associated with pathways known to be involved in liver injury as well as steatosis. Integration of the miRNA-mRNA transcriptome sets was used to identify potential mRNA targets of PCB-altered miRNAs. Interestingly, some of the differentially expressed hepatic miRNAs in exposed mouse liver were also differentially expressed in serum from a cohort of PCB-exposed subjects with a high prevalence of suspected TASH (Cave et al., 2022). Integration of the miRNA and mRNA transcriptomes with the proteome in each exposure group identified previously (Jin et al., 2020) showed little overlap, suggesting that additional mechanism of regulation are involved in protein changes in TASH.

2. Materials and methods

2.1. Study design

The experimental design is modeled in Fig. S1. This study used frozen livers that were collected from male C57Bl/6 J mice in a previous University of Louisville Institutional Animal Care and Use Committee-approved study (Jin et al., 2020). All mice were fed a HFD (42 % kCal from fat) and exposed to a single oral gavage of vehicle control (corn oil), Aroclor1260 (20 mg/kg), PCB126 (20 μg/kg), or the combination of the doses of Aroclor1260 + PCB126 (Jin et al., 2020). Thus, there were four exposure groups: HFD diet: Vehicle control (corn oil), Aroclor1260, PCB126, and Aroclor1260 + PCB126. The time of HFD feeding + PCB exposure was 12 wks. As reported (Jin et al., 2020): HFD feeding resulted in variable degrees of hepatic steatosis, as assessed by H&E staining, and neutrophil infiltration, as assessed by CAE and F4/80 staining of liver sections. The Aroclor1260 exposure group showed worsened hepatic inflammation with an increased number of inflammatory foci (CAE staining), macrophage infiltrates (F4/80 staining), and elevated plasma ALT levels, compared to the other two PCB groups. In fact, the PCB126 group showed decreased hepatic inflammation compared to the Aroclor 1260 group and decreased pro-fibrotic gene expression compared to all other groups. PCB126 also attenuated Aroclor1260-mediated inflammation in the co-exposure group. Despite the attenuated inflammation, the Aroclor1260 + PCB126 co-exposure group had increased hepatic levels of free fatty acids.

2.2. RNA isolation and sequencing

RNA was isolated from five mouse livers/experimental exposure using RNeasy kit (Qiagen, Germantown, MD) (Fig. S1). RNA quality and concentration were evaluated using an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA). RNA-sequencing was performed as previously described (Muluhngwi et al., 2017; Schultz et al., 2018) with minor modifications. Briefly, libraries were prepared from 1 μg of mouse liver RNA using the TruSeq stranded mRNA Library Prep Kit (Illumina, San Diego, CA), along with TruSeq RNA CD Index Plate. Following preparation, the library was validated for size, purity, and semi quantitation on an Agilent Bioanalyzer (Agilent Technologies, Santa Clara, CA). Sequencing library quantitation was done by performing a MiSeq Nano (Illumina) test run. Samples were denatured and diluted to a final concentration of 1.8pM prior to loading in to the cartridge. Two sequencing runs were performed on the Illumina NextSeq 500 using the NextSeq 500/550 High Output Kit v2.5 (75 Cycles). Data as fastq Files were downloaded from Illumina’s BaseSpace onto the KBRIN server for analysis. Quality control (QC) of the raw sequence data was performed using FastQC (version 0.10.1). For each sequencing run the sequences were considered to be high quality, and no sequence trimming was necessary. The sequences were aligned to the mm10 reference genome assembly (GRCm38.p6.fa) using STAR (version 2.6) generating alignment files in bam format (Dobin et al., 2013). Differential expression was performed using DESeq2 that uses Benjamini-Hochberg for the FDR calculation (Love et al., 2014). Raw counts were obtained from the STAR aligned bam format files using HTSeq version 0.10.0 (Anders et al., 2015). The raw counts were normalized using the Relative Log Expression (RLE) method and then filtered to exclude genes with less than 10 counts across samples. Entrez gene identifiers for significantly expressed genes had a q-value cutoff of 0.05. The raw data of the RNA-seq are available at Gene Expression Omnibus (GEO) database: RNA accession number GSE173271.

2.3. miRNA Sequencing

miRNA was isolated from the same five mouse livers/experimental exposure as used for RNA seq using Qiagen miRNA kits (Fig. S1). miRNA-sequencing was performed as previously described (Muluhngwi et al., 2017). Briefly, libraries were prepared from 1 μg of mouse liver RNA the QIAseq miRNA Library Kit (Qiagen). The library was validated for size, purity, and semi quantitation on an Agilent Bioanalyzer (Agilent Technologies, Santa Clara, CA). Sequencing library quantitation was done by performing a MiSeq Nano (Illumina) test run. Samples were denatured and diluted to a final concentration of 10pM prior to loading in to the cartridge. Two sequencing runs were performed on the Illumina NextSeq 500 using the NextSeq 500/550 High Output Kit v2.5 (75 cycles). Data as fastq Files were downloaded from Illumina’s BaseSpace onto the KBRIN server for analysis. Quality control (QC) of the raw sequence data was performed using FastQC (version 0.10.1) for each sequencing run. Preliminary adapter trimming (Trimmomatic v0.33) (Bolger et al., 2014) was performed on each of the samples to remove the Qiagen 3′ Adapter sequence. The trimmed sequences were directly aligned to the mouse mm10 reference genome assembly (mm10.fa) using the mapper.pl wrapper of the mirdeep2 package (v 0.0.7) (Friedlander et al., 2012). This script uses bowtie (version 1.1.1), generating alignment files in arf format (Langmead and Salzberg, 2012). The aligned sequences were used as inputs into the mirdeep2 package and the in-house script quantifier.pl. In addition, this script uses the mirBase release 21 (Griffiths-Jones et al., 2006) mature miRNA and miRNA hairpin sequences downloaded from ftp://mirbase.org/pub/mirbase/CURRENT/. The result is a file containing the number of reads mapping to each of the 2061 mouse (mmu) miRs for the specific sample. Differentially expressed miRs were determined using edgeR (Robinson et al., 2010). The q value cut off was ≤ 1. The raw data of the miRNA-seq is available at Gene Expression Omnibus (GEO) database: GSE195829.

2.4. MetaCore enrichment analysis

Network analysis for differentially expressed genes (DEG) in the mRNA transcriptome, miRNAs in the miRNA transcriptome, and the comparative analysis of DEGs and miRNAs from each of the exposure groups was evaluated using the web-based software MetaCore version 22.1 (Cortellis, Philadelphia, PA). Mapping procedures within MetaCore involves calculating statistical relevance of the matches found including public ontologies such as Gene Ontology (GO, www.geneontology.org). Ontologies are represented by canonical Pathway Maps, Process Networks, and GO Processes.

2.5. RNA extraction and quantitative real-time PCR (qRT-PCR)

RNA was extracted using the RNeasy Mini Kit (Qiagen, Gaithersburg, MD, USA). miRNA was isolated using the miRNeasy Mini Kit RNA (Qiagen). RNA concentration and quality were assessed using a Nano-Drop spectrophotometer (Thermo Scientific, Rockford, IL, USA). The TaqMan^® MicroRNA Reverse Transcription Kit and the High Capacity cDNA Reverse Transcription Kit for RNA (both from ThermoFisher) were used to make cDNA from miRNA and mRNA, respectively. Quantitative real-time PCR (qRT-PCR) for miRNAs (assay IDs listed in Table S27) was performed using TaqMan assays (ThermoFisher). qRT-PCR for mRNA used SYBR Green primers (Integrated DNA Technologies (IDT), Coralville, IA, USA) (Table S28). qRT-PCR was performed using an ABI Viia 7 Real-Time PCR system (LifeTechnologies) with each reaction run in triplicate. A minimum of three biological replicate experiments was performed.

2.6. Dual luciferase miRNA reporters and assays

The putative miRNA response elements in the 3′ UTRs of the PCB-regulated mouse liver genes Nol3, Smarca4, Stmn1, and Notch3 for miR-712–5p, miR-466 f-3p, miR-9–5p, miR-206–3p, respectively, were identified in miRTarget: miRNA-target interactions https://rdrr.io/bioc/miRSM/man/miRTarget.html (Ritchie et al., 2015). Oligomers containing the miRNA recognition element (MRE) and flanking region from the 3′ UTRs of each gene were purchased from IDT (Coralville, IA). Sequences for gene target/miRNAs-MREs are (f = forward, r = reverse): Nol3/miR-712–5p: f: 5′-CTAGCGGCCGCTGAAAGAAAAGAAGGAAGGAAGGAAGGAT-3′ and r: 5′- CTAGATCCTTCCTTCCTTCCTTCTTTTCTTTCAGCGGCCGCTAGAGCT-3′; Smarca4/ miR-466 f-3p f: 5′- CTAGCGGCCGC CACATATGAATGTGTATATATACATACT-3′ and r: 5′-CTAGAGTATGTATATATACACATTCATATGTGGCGGCCGCTAGAGC-T-3′; Stmn1/ miR-9–5p: f:5′- CTAGCGGCCGC TCCTAAATATCCAAAGACTGTACTGGCT-3′ and r: 5′- CTAGAGCCAGTACAGTCTTTGGATATTTAGGAGCGGCCGCTAGAGCT-3′; Notch3/ miR-206–3p: f: 5′- CTAGCGGCCGC CACTTCACTGCATTCCAGATAAGACGTT-3′ and f: 5′- CTAGAACGTCTTATCTGGAATGCAGTGAAGTGGCGGCCGCTAGAGC-T-3′. Oligomers were annealed and cloned into the Sac1 and Xba1 (New England Biolabs, Ipswich, MA) sites of the pmirGLO Dual-luciferase miRNA Target Expression Vector (Promega, Madison, WI) 3′ to the Firefly luciferase gene, transformed into competent JM109 E. coli (Promega, Madison, WI). Plasmids were purified using the DNA mini kit (Qiagen, Germantown, MD). Diagnostic Not1 (New England Biolabs) restriction enzyme digestions were performed on the purified plasmids. A 1.5 % agarose gel was run to confirm insert(s) of appropriate size(s) prior to DNA sequencing. Inserted gene 3′ UTR miRNA target sequences were confirmed by DNA sequencing and analysis using SeqManPro (DNASTAR, Madison, WI).

HEK-293 T cells (ATCC, Manassas, VA) were grown in DMEM medium (Corning, Corning, NY) supplemented with 10 % Fetal Bovine Serum (FBC, R&D Systems, Minneapolis, MN) and 1 % Penicillin Streptomycin Solution (Corning, Corning, NY). Cells were seeded in 96-well plates at 0.5 × 10⁴, 1 × 10⁴, and 2 × 10⁴ cells per well and co-transfected with 0.05 μg pmirGLO plasmids + /− pre-miRNAs (pre-miR-712–5p, pre-miR-466 f-3p, pre-miR-9–5p, and pre-mIR-206–3p) specific to the plasmid or the pre-miR negative control (ThermoFisher Scientific, Waltham, MA) using FuGENE HD Transfection Reagent (Promega, Madison, WI) following the manufacturer’s protocol. 48 h post transfection, dual-luciferase assays were performed using the Dual-Glo Luciferase Assay System (Promega, Madison, WI) and fluorescence was read on a Synergy2 spectrophotometer (BioTek, Winooski, VT). Five independent, biological replicate transient transfection experiments were performed.

2.7. Statistical analysis

The Ct values of miRNAs in RT-qPCR were normalized to sno202 or u6 snRNA and variable thresholds, denote as ΔCt. Then Quantile normalization (QN) to ΔCt was used for all samples within group (Srivastava et al., 2019). The descriptive statistics for QN of ΔCt values were produced by Group (Aroclor1260, PCB126 and Aroclor1260 + PCB126) using one-way ANOVA. The mean with 95 % confidence interval, standard deviation (SD), median, minimum, and maximum are presented with p-value (Steel and Torrie, 1980). For data description, there is only one factor (Group or Replicate). Hence, one-way ANOVA is used for data comparison among 4 different groups (Vehicle Control, Aroclor1260, PCB126, and Aroclor1260 + PCB126), or 5 different replicates. A two-way ANOVA was performed to analyze the effect of Aroclor1260 and PCB126 on miRNA expression. Significance is reported as p-value < 0.05. Differentially expressed genes (DEG) from mRNA seq, DEG (p ≤ 0.05; q ≤ 0.05; log2FC ≥ 0.5, FPKM ≥ 1 in ≥ 3 of the 5) samples/treatment group were identified in the DESeq2. ANOVA based analyses were performed using SAS statistical software (The SAS System V9. Cary, NC, SAS Institute Inc; 2003) and gene expression analysis were performed using R (R Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria, 2020).

3. Results

3.1. Effects of HFD and PCB exposure on the hepatic mRNA transcriptome

To identify hepatic transcriptome changes mediated by the combination of HFD diet and PCB exposure, male C57Bl/J6 mice were exposed to 12 wks. of HFD and a single oral dose of vehicle control (corn oil), Aroclor1260, PCB126, or combined Aroclor1260 + PCB126 (Fig. S1). Differentially expressed genes (DEGs) were identified for three pairwise comparisons (vehicle control vs. Aroclor1260; PCB126; of Aroclor1260 + PCB126) using Cuffdiff2 (Trapnell et al., 2013). Table 1 shows the number of DEGs in each comparison. Only transcripts with a log₂ fold-change > 0.5 (< −0.5 for repressed transcripts) were included in the analyses. Co-exposure to Aroclor1260 + PCB126 resulted in the largest number of DEGs compared to HFD alone (Fig. 1A). PCB126 exposure regulated more transcripts than Aroclor1260 exposure (Fig. 1A). Similar trends in the number of differentially expressed proteins were observed in the proteome of these samples (Jin et al., 2020). The top 20 DEGs, up- or down-regulated, for each exposure group are listed in Tables S1–S6.

Table 1.

Differentially expressed genes (DEG) from mRNA seq of C57Bl/6 J mouse liver samples after 12 wks. of HFD and the indicated PCB exposure. The number of DEGs in each PCB exposure group was compared to vehicle control (veh). Statistical significance determined by DESeq2 analysis: p ≤ 0.05; q ≤ 0.05; log2FC ≥ 0.5, FPKM ≥ 1 in ≥ 3 of the 5 samples/treatment group.

Exposure Group	Total DEG	# genes upregulated	# genes downregulated
Aroclor1260 vs. veh.	124	89	35
PCB126 vs. veh.	278	181	97
Aroclor1260 + PCB126 vs. veh	5190	2367	3823

Fig. 1.

Differentially expressed genes (DEGs) with PCB exposure in the liver of HFD-fed mice. Fig. S1 shows the experimental design. Male C57BL6/J mice were exposed to 12 wks. of a HFD and either vehicle control or a single oral dose of the indicated PCB. Unbiased mRNA-seq identified the transcriptome. Values are an average of 5 mice/exposure group. A) Comparison of the number of DEGs after exposure to Aroclor1260, PCB126, or Aroclor1260 + PCB126 co-exposure, compared to vehicle control. B) The Venn diagram shows the overlap and differences between the number of DEGs in each PCB-exposure group compared to vehicle control. Breakout boxes identify the top 10 genes uniquely up- (left column, red up-arrow) and down- (right column, green down-arrow) regulated DEGs in that exposure group. Cut-off values are 0.5-log2 fold change and a q-value < 0.05. C) The log2fold change values of the top 10 up- and down-regulated DEGs in the indicated PCB exposure group are shown. Additional DEGs for each PCB exposure group are listed in Tables S1–S6.

Unique and common DEGs for the three PCB exposure groups, compared to vehicle control, are summarized in Fig. 1B. Twenty-five commonly regulated DEGs were identified with all three PCB exposures. This included increased abundance of Sp4 (a transcription factor that regulates VEGF), Fzd4 (a frizzled membrane receptor in the Wnt pathway), and Tfrc (a transferrin receptor involved in iron metabolism). Interestingly, 25 common genes were regulated in the three PCB exposure groups. 42 common genes regulated by Aroclor1260 and Aroclor1260 + PCB126 and 112 genes in common between PCB126 and Aroclor1260 + PCB126 exposure groups were identified (Fig. 1B). PCB126 and Aroclor1260 + PCB126 increased the abundance of cytochrome P450 family A member 2 (Cyp1a2), an AHR target gene upregulated by HFD + Aroclor1260 exposure in male mouse livers (Wahlang et al., 2014b). The top 10 DEGs for each PCB exposure are identified in Fig. 1C and their relevance to hepatic function, NAFLD, and NASH are described in Tables S7 and S8.

3.2. Pathway analysis of DEGs in PCB and HFD exposed mouse liver

MetaCore comparative pathway analysis of the DEGs in the three PCB exposure groups were compared to vehicle control and identified unique and common pathways (Fig. 2). For this analysis, all significant DEGs (p ≤ 0.05) were included. Common pathways included “HSP70- and HSP40- dependent folding” which likely reflects the induction of ER stress and the unfolded protein response (UPR) in HFD-induced NAFLD (Tsai et al., 2020), “positive regulation of the WNT/β -catenin signaling pathway” which is involved in the activation of profibrogenic hepatic stellate cells (HSCs) (Perugorria et al., 2019), and “signal transduction via AKT signaling” which regulates key inflammatory cytokines to activate immune cells in NAFLD (Matsuda et al., 2013). Pathways unique to Aroclor1260, PCB126, or combined Aroclor1260 + PCB126 exposure include those involved in the pathogenesis of TASH, e.g., immune responses (An et al., 2020), “chemokines in inflammation in adipose tissue and liver in obesity” (du Plessis et al., 2015), and “transcription of HIF-1α targets” (Kaminsky-Kolesnikov et al., 2020). Pathways associated with DEGs in the Aroclor1260 + PCB126 co-exposure group included “regulation of glucose and lipid metabolism via FXR”. Aroclor1260 exposure reduced FXR protein abundance in HFD-fed mouse liver (Hardesty et al., 2019b).

Fig. 2.

Enrichment analysis of DEGs identified in HFD-fed mouse liver with PCB exposure. The DEGs (q-value < 0.05, up- and down- regulated, Table 2, 3, and S1–6) in each PCB exposure group were analyzed using the Compare Experiments tool in MetaCore. MetaCore identified 731 common genes in the three data sets and 2038 genes that were identical in two of the three PCB exposure data sets. The number of genes uniquely regulated in each exposure group is indicated. The color palette is preserved in the Venn diagram (also Fig. 1B). The top five common and unique pathways for the exposure groups is shown in the order provided by MetaCore analysis.

Gene Ontology (GO) biological processes and KEGG pathways were identified for the DEGs in the three PCB exposure groups (Fig. S2 and Fig. S3). GO processes identified for DEGs in Aroclor1260 exposure group were concordant with liver pathology in TASH including inflammation, e.g., “receptor-mediated endocytosis” (Kim et al., 2017), “endothelial cell differentiation” (Tamburini et al., 2019), and “endosome transport” (Pol and Enrich, 1997). GO processes identified in the DEGs in the PCB126 exposure group included biosynthetic processes responsible for clearing toxins from the liver, e.g., “organophosphate biosynthetic processes” (Abdou et al., 2020) and “cofactor biosynthetic processes” (Rowland et al., 2013). Aroclor1260 + PCB126 co-exposure was associated with GO processes in catabolism, e.g., “proteasomal protein”, “aromatic compound”, and “cellular nitrogen compound catabolic processes”. These GO processes are concordant with pathologies reported previously with HFD and PCB co-exposure, i.e., hepatic inflammation in Aroclor1260-exposed mice (Wahlang et al., 2014b), transcription factor regulation in PCB126 exposed mice (Hardesty et al., 2019b), and cytoskeletal remodeling in Aroclor1260 + PCB126 co-exposed mice (Jin et al., 2020). Novel GO processes identified here, e.g., multivesicular sorting, vesicular transport, and autophagy, suggest a role for PCB-induced DEGs in autophagic processes during ER stress that contribute to hepatic cell death, fibrosis, and cirrhosis (Kim et al., 2021; Zhao et al., 2021).

3.3. PCBs alter miRNA expression

The impact of HFD and PCB co-exposure on mouse hepatic miRNA expression is unknown. Changes in hepatic miRNA abundance in response to PCB exposure, relative to vehicle control, are shown in Volcano plots (Fig. 3) and Heatmaps (Fig. S4). Consistent with the mRNA transcriptome (Fig. 1B), Aroclor1260 + PCB126 co-exposure resulted in the largest number of differentially expressed hepatic miRNAs (DEMs) (Table 2, Fig. 4). The top 10 up- and down-regulated DEMs are listed in Tables S9 and S10 with additional DEMs listed in Tables S11–S16. Many of the miRNAs upregulated with PCB-exposure have no known role in liver pathology, but some are implicated in pathways associated with inflammation, apoptosis, and cancer (Tables S9 and S10). Comparison of the top ten up- or down-regulated miRNAs for each exposure group is shown in Fig. 4. Five common DEMs were identified in all three PCB exposure groups (Fig. 4, Table S9 and S10).

Fig. 3.

Volcano Plots compare the impact of each of the three PCB exposures on mRNA and miRNA expression in HFD-fed mouse liver compared to vehicle control. The x-axis represents the log2 fold changes of FPKM (fragments per kilobase of transcript per million mapped reads) of mRNAs or miRs between two groups. Red indicates that the gene or miR is up- or down-regulated with a p-value less than 0.05. Pink in the mRNA plots indicates that the transcript is up or down-regulated with a q-value less than 0.05.

Table 2.

Summary of differentially expressed miRNAs (DEMs) in male C57Bl/6 J mouse liver after HFD + PCB exposure. The number of DEMs in each PCB exposure group was compared to vehicle control (veh.). Statistical significance determined by DESeq2 analysis: p ≤ 0.05; q ≤ 1; log2FC ≥ 0.

Exposure Group	Total DEMs	Upregulated miRNAs	Downregulated miRNAs
Aroclor1260 vs. veh.	69	17	52
PCB126 vs. veh.	50	28	22
Aroclor1260 + PCB126 vs. veh Control	97	51	46

Fig. 4.

Differentially expressed miRNA (DEMs) with PCB exposure in liver of HFD-fed mice. A Venn diagram illustrates common and unique differentially expressed miRNAs (DEMs) in the three PCB exposure groups compared to vehicle control. In the boxed lists of miRs, the miRs in the left and right columns are the top 10 increased and decreased miRs, respectively. Cutoff values are 1-fold change and a p-value < 0.05.

DEMs downregulated by PCB exposure have been reported to play a role in hepatic processes, liver injury, and fibrosis (Table S10). Notably, three of the downregulated miRNAs common to all three PCB exposure groups (miR-208b-2p, miR-499–5p, and miR-34a-3p) (Fig. 4) may play critical roles in hepatic function. miR-499 targets Fnip (Liu et al., 2016) that interacts with AMPK in the development of hepatic natural killer T (NKT) cells which are increased in the livers of patients with hepatic steatosis (Wolf et al., 2014). Thus, the PCB-mediated reduction in miR-499 may be associated with increased immune infiltration reported previously (Jin et al., 2020). The PCB-regulated reduction in miR-34a-3p expression reported here may be associated with FXR activation and the induction of ER stress due to abnormal lipid accumulation, inflammation, and mitochondrial dysfunction in this model (Hardesty et al., 2019a) and also detected in human NAFLD/NASH (Meex and Blaak, 2021). Taken together with the upregulated DEMs (Table S9), these results suggest a role for dysregulation of miRNAs in the progression of hepatic injury in this mouse model of PCB-induced TASH.

3.4. Pathway analysis of DEMs

MetaCore enrichment analysis of the DEMs identified in pathways in cancer (Fig. 5). Similar to the mRNA analyses, all significant DEMs (p ≤ 0.05) were included. A limitation of MetaCore analysis is that it is based on published literature; thus, many of the named pathways are associated with cancers. However, key descriptors suggest miR involvement in mechanisms associated with hepatic reprogramming, i.e., “epigenetic regulation”, “cell migration, survival, and angiogenesis”, consistent with our previous findings in this TASH model (Jin et al., 2020). Broadly, the pathways associated with DEMs included cell survival/ death, cell proliferation/differentiation including EMT and Notch signaling, metabolism, and inflammation.

Fig. 5.

Enrichment analysis of DEMs identified in HFD-fed mouse liver with PCB exposure. The DEMs, both up- and down-regulated compared to vehicle control, identified in each of the three PCB exposure groups were analyzed using the Compare Experiments tool in MetaCore. MetaCore identified 18 common DEMs in the three PCB exposure groups and 65 DEMs that were identical in two of the three data sets. The number of DEMs uniquely regulated in each PCB exposure group is indicated. The color palette is preserved in the Venn diagram in Fig. 4. The top common and unique pathways for the PCB exposure groups is shown in the order provided by MetaCore analysis.

3.5. Validation of altered mRNA and miRNA expression by qRT-PCR

Based on the highest fold-change for DEGs and DEMs and their relevance to liver-specific pathology, 10 mRNAs (Olfm3, Bnajb, Cyp1a2, Cyp1a1, Tfrc, Nr1d1, Dmd, Ppargc1a, and Prox1) and eight miRNAs (miR-205–5p, miR-301b-3p, miR-124–3p, miR-712–5p, miR-133a-5p, miR-297a-5p, miR-466 f, and miR-7b-5p) dysregulated by PCB exposure were selected for validation by qRT-PCR. As expected, mRNA qRT-PCR validated the increased abundance of Cyp1a1 and Cyp1a2 in the livers of mice exposed to PCB126 (Table S17, Fig. S4). The down-regulation of Ccl2 in the PCB126 exposure group was also confirmed by qRT-PCR (Table S17). Transcriptome results for Dnajb1, Nr1d1, and Tfrc were also confirmed (Table S17). qRT-PCR confirmed four of the eight selected miRNAs to be consistent with the DEMs identified (Table S18). Together, these data support the DEGs and DEMs identified by mRNA-seq and miRNA-seq.

3.6. Examination of human miRNAs associated with PCB exposure in the mouse liver samples

We recently published a study evaluating the dysregulated circulating miRs in a human cohort (the Anniston Community Health Survey, ACHS) with increased PCB exposures and a high prevalence of suspected TASH (Cave et al., 2022). In this cohort, nine plasma miRs were significantly associated with liver necrosis. Of these, miR-99–5p, miR-122–5p, miR-192–5p, and miR-320a were increased in plasma. Other miRs, such as miR-130a-3p were associated with other liver disease biomarkers. Although none of these miRs were identified as significantly different from the vehicle control in the miRome, we measured the hepatic expression of these five miRs in the mouse TASH model. Four of the five miRs tested were differentially expressed in at least one PCB exposure group (Table 3), although the direction of the change in mouse liver was not always the same as that observed in human serum. This is not necessarily surprising given known differences in miRNA liver expression and release into circulation under conditions of hepatotoxicity and cell death. miR-192–5p was differentially expressed vs. vehicle control for all PCB treatments tested. Taking into account the respective roles of NDL Aroclor1260 and DL PCB126 on miRNA expression, a two-way ANOVA was performed to analyze the effect of Aroclor1260 and PCB126 on miRNA expression (Table 4). The results revealed that there was a statistically significant interaction between the effects of Aroclor1260 and PCB126 for miR192–5p (p < 0.0011) and for miR-320–3p (p = 0.008). Further, the analysis showed that Aroclor1260 had a statistically significant main effect on miR-192–5p (p = 0.004), and that PCB126 had a statistically significant main effect on miR-192–5p (p < 0.001) and miR-320 (p = 0.018). The interaction between Aroclor1260 and PCB126 for miR-192–5p and miR-320–3p indicates that the effect of these two classes of PCBs on hepatic miR-192–5p and miR-320–3p expression was dependent upon the specific PCB (Aroclor1260 or PCB126) exposure of the mice. Taken together, these qRT-PCR results validate key results from present study, and extend the ACHS results into an animal model system.

Table 3.

qRT-PCR analysis of miRNAs in mouse liver that were reported to be dysregulated in human serum from the PCB-exposed ACHS cohort (Cave et al., 2022). The table summarizes qRT-PCR examination of miRNA expression in livers from C57Bl/6 J male mice on HFD (vehicle control) or with exposure to Aroclor1260, PCB126, or Aroclor1260 + PCB126. Values are the mean of 10 separate animals/exposure group. The mean with 95 % confidence interval are presented with p-value The means with 95 % confidence interval are presented with p-value for the main effect (comparison to vehicle control). Statistical significance was determined on data with quantile normalization (QN) using one-way ANOVA. Values in red denote significance (p < 0.05). All calculations were performed using SAS statistical software.

miRNA	vehicle Control Mean (95 %CI)	Aroclor1260 Relative to Control Mean (95 %CI)	P-Value	PCB126 Relative to Control Mean (95 %CI)	P-Value	Aroclor1260 +PCB126 Relative to Control Mean (95 %CI)	P-Value
miR-99a-3p	9.414 (9.315–9.514)	9.515 (9.448–9.581)	0.106	9.452 (9.362–9.543)	0.582	9.630 (9.500–9.761)	0.012
miR-192–5p	−0.904 (−0.972 to −0.836)	−1.054 (−1.113 to −0.995)	0.002	−1.135 (−1.209 to −1.061)	< .001	−1.019 (−1.099 to −0.938)	0.038
miR-320–3p	4.457 (4.372–4.542)	4.377 (4.282–4.472)	0.226	4.299 (4.203–4.395)	0.019	4.474 (4.383–4.565)	0.786
miR-122–5p	−3.771 (−3.882 to −3.660)	−3.826 (−3.912 to −3.741)	0.439	−3.695 (−3.847 to −3.544)	0.435	−3.624 (−3.724 to −3.524)	0.059
miR-130a-3p	3.415 (3.309–3.522)	3.374 (3.318–3.430)	0.498	3.510 (3.418–3.602)	0.192	3.652 (3.539–3.765)	0.004

Table 4.

Two way ANOVA analysis of the interaction between Aroclor1260 and PCB126 on miRNA expression. Two-way ANOVA was performed examining the interaction between Aroclor1260 and PCB126 exposure for miRNA expression in mouse liver. The p-values were reported up to three significant digits and p < 0.05 was considered statistically significant for the main factors and interaction. The four within group comparisons using 2-sample t-test method were considered statistically significant for p ≤ 0.05. Shown are 3 significant main effects and 2 significant interaction effects. Values in red denote significance (p < 0.05). Supplementary Table 29 shows the two-way ANOVA data.

	P values
	Aroclor1260	PCB126	Interaction	Aroclor1260 =No (PCB: No vs Yes)	Aroclor1260 =Yes (PCB: No vs Yes)	PCB126 =No (Aroclor1260: No vs Yes)	PCB126 =Yes (Aroclor1260: No vs Yes)
miR-99a-3p	0.165	0.597	0.444	0.582	0.126	0.106	0.032
miR-192–5p	0.004	< 0.001	< 0.001	< 0.001	0.486	0.002	0.042
miR-320–3p	0.240	0.018	0.008	0.019	0.155	0.226	0.012
miR-122–5p	0.502	0.364	0.280	0.435	0.004	0.439	0.445
miR-130a-3p	0.540	0.167	0.059	0.192	< 0.001	0.498	0.061

3.7. Integrated miRNA-mRNA analysis

Next, we identified experimentally validated mRNA targets for DEMs (Tables S11–S16) in miRTarBase (release 6.1) that were altered in the DEGs of the same PCB exposure group (Table S19). For example, we identified 13 reciprocal relationships in which a downregulated DEM was predicted to target an upregulated DEG in the livers of mice exposed to Aroclor1260 (Table S19). The largest number of reciprocal miRNA-mRNA pairs was identified in the Aroclor1260 + PCB126 co-exposure group (Table S19).

Table 5 lists upregulated DEGs in common in the three PCB exposure groups that are the predicted targets of identified downregulated miRNAs. We used a log2FC > 0.5 for DEGs and a logFC > 0.5 for DEMs to identify biologically relevant miRNA-mRNA relationships. For example, miR-466 f-3p was downregulated in the PCB126 and Aroclor1260 + PCB126 exposure groups and targets Dmd and Ppargc1a which were upregulated in the same exposure groups (Table 5). Table 6 summarizes the downregulated DEGs with the inversely upregulated DEMs. Only one DEG, Ddit3, was downregulated in all three PCB exposure groups and is a predicted target of miR-205–5p that was upregulated in the PCB126 exposure group. Tables S20 and S21 list miRNA-mRNA pairs unique to each exposure group.

Table 5.

Upregulated genes predicted to be targeted by downregulated miRNA. The indicated DEGs (log₂ FC > 0.5) and DEMs (log FC < −0.5) were identified as significantly different from the vehicle control in the indicated PCB exposure group (Tables S1–S12) The upregulated DEGs common to all PCB exposure groups and their log2FC are listed with downregulated miRNAs predicted to target the DEGs.

	mRNA upregulated				miRNA downregulated
	Gene	Aroclor1260 (log2 FC)	PCB126 (log2 FC)	Combined (log2 FC)	Predicted miR	PCB Exposure Group	FC
All	Dmd	0.671	1.003	0.732	miR-466 f-3p	PCB126	−3.10
	Prox1	0.644	0.652	0.990	miR-124–3p	Aroclor1260	−2.81
	Ppargc1a	0.594	0.567	0.880	miR-466 f-3p	Aro + PCB	−3.13
Aroclor1260 and Combined	Cpeb4	0.719		0.974	miR-301b-3p	Aro + PCB	−1.08
					miR-297a-5p	Aroclor1260	−2.10
	Arap2	0.619		0.691	miR-301b-3p	Aro + PCB	−1.08
	Sh3bgrl	0.596		0.727	miR-1a-3p	Aro + PCB	−6.81
					miR-1a-3p	Aroclor1260	−4.79
	Fcho2	0.591		0.852	miR-301b-3p	Aro + PCB	−1.08
					miR-124–3p	Aroclor1260	−2.81
	Thrb	0.515		0.737	miR-712–5p	Aroclor1260	−1.93
	Adam10	0.544		0.713	miR-301b-3p	Aro + PCB	−1.08
					miR-297a-5p	Aroclor1260	−2.10
PCB126 and Combined	Eif4e3		0.888	0.795	miR-206–3p	Aro + PCB	−5.84
	Lonrf1		0.826	0.652	miR-466 f-3p	Aro + PCB	−3.13
	Arl15		0.693	0.614	miR-466b-3p	Aro + PCB	−0.69
	Gpcpd1		0.690	0.872	miR-301b-3p	Aro + PCB	−1.08
	Papss2		0.626	0.787	miR-301b-3p	Aro + PCB	−1.08
	Rufy2		0.606	0.686	miR-466b-3p	Aro + PCB	−0.69
	Fzd6		0.603	0.646	miR-301b-3p	Aro + PCB	−1.08
	Pdxk		0.587	0.503	miR-466b-3p	Aro + PCB	−0.69
	Tgfbr2		0.540	0.694	miR-196a-5p	Aro + PCB	−2.84
	Prnp		0.530	0.706	miR-301b-3p	Aro + PCB	−1.08
	Crebrf		0.522	0.846	miR-466 f-3p	Aro + PCB	−3.13
	Tmem241		0.504	0.540	miR-466e-3p	Aro + PCB	−0.69

Table 6. Downregulated genes predicted to be targeted by upregulated miRNA.

The indicated DEGs (log₂ FC < −0.5) and DEMs (log FC > 0.5) were identified as significantly different from the vehicle control in the indicated PCB exposure group (Tables S1–S12). The downregulated DEGs common to all PCB exposure groups and their log2FC are listed with upregulated miRNAs predicted to target the DEGs.

	mRNA downregulated				miRNA upregulated
	Gene	Aroclor1260 (log2 FC)	PCB126 (log2 FC)	Combined (log2 FC)	Predicted miR	Exposure Group	FC
All	Ddit3	−0.783	−1.022	−1.002	miR-205–5p	PCB126	1.95
PCB126 and Combined	Bag3		−1.027	−1.033	miR-7b-5p	PCB126	1.13
	Mpzl1		−0.687	−0.697	miR-7b-5p	PCB126	1.13
	Camk1d		−0.539	−0.621	miR-466i-5p	Aro + PCB	0.71
	Ndrg1		−0.733		miR-691	PCB126	2.76

3.8. Integrated miRNA-mRNA-Protein interaction analysis

We previously identified hepatic proteome changes in the HFD-fed mice with exposure to Aroclor1260, PCB126, and Aroclor1260 + PCB126 co-exposure (Jin et al., 2020). Only a small overlap (0.8–1.8 %) between differentially expressed proteins (DEPs) and DEGs identified here was detected (Fig. 6, Table S22). This observation suggests that transcript abundance is not a good proxy for protein abundance in this model and that regulation of transcript stability, translation, and protein stability regulate the proteome, most notably with Aroclor1260 + PCB126 co-exposure. Unlike mRNA expression, which showed the most transcript changes with Aroclor1260 + PCB126 co-exposure, more protein changes were observed with PCB126 exposure.

Fig. 6.

Integrated analysis of DEGs and differentially expressed proteins (DEPs) in HFD-fed mouse liver with PCB exposure. The number of up- and down-regulated DEGs (A) and differentially expressed proteins (DEPs) (B) (Jin et al., 2020) are displayed for each PCB exposure group (relative to the vehicle control). C) Venn diagrams illustrate the comparison of the transcriptome (DEG) and proteome (DEP) in the three PCB exposure groups. DEGs are log2FC > 0.5, q-value < 0.05, DEPs are FC > 0.5, q-value < 0.05. Additional DEGs in the 56 common DEGs for the Aroclor1260 + PCB126 co-exposure are listed in Table S15.

To identify potential miRNA targets in the in the PCB-altered hepatic proteome (Jin et al., 2020), we searched DEPs against miRTarBase and the resulting miRNAs against the DEMs for that PCB exposure group. We identified 36 reciprocal relationships in which a DEM was predicted to target a DEP within the three PCB exposure groups, i.e., (DEP up, DEM down; or DEP down, DEM up) (Table S23), including two proteins targeted by two or more DEMs and eight DEMs that target multiple DEPs. Most reciprocal DEM-DEPs pairs were identified in the PCB126 exposure group, and the Aroclor1260 + PCB126 co-exposure group (Table S23). To identify pathways regulated by PCB-induced changes in protein, miRNA, and mRNA abundance, an integrated bioinformatics analysis was applied to the miRNA and mRNA transcriptomic and proteomic data (Jin et al., 2020) (Tables S24–S26). These integrated analyses confirm that PCB exposures induced protein, miRNA, and mRNA changes associated with pathways identified to play a role in the development of steatohepatitis, e.g., “immune response signaling”, “signal transduction via JAK/STAT pathway”, “NOTCH signaling”, and “p53 signaling”.

3.9. Examination of putative miRNA regulated DEGs

To determine if proteins altered by PCB exposures in mouse liver are targeted by reciprocally expressed miRNAs in the same PCB exposure group, we identified four protein-miRNA ‘reciprocal pairs’ of interest: three proteins with increased expression and downregulated miRNAs: Nol3-miR-712–5p; Smarca4-miR-466 f-3p, Notch3-miR-206–3p, and one protein (Stmn9) that was downregulated and its verified miR-9–5p (Delaloy et al., 2010) upregulated with PCB126 exposure Stmn1-miR-9–5p (Table 6). Thus, miR-9–5p served as a positive control for our experiments. The putative MREs in the 3′ UTR, with flanking sequences for the target gene were cloned into the pmirGLO Dual-luciferase reporter vector 3′ to Firefly luciferase and confirmed by DNA sequencing. HEK-293 T cells were transiently transfected with the pmirGLO vector, the pmirGLO-MRE containing vectors, pre-miR control, or pre-miRs for each of the four miRNAs. Forty-eight h later Firefly and Renilla luciferase activities were assayed. We observed that miR-712–5p, miR-466 f-3p, miR-206–3p, and miR-9–5p inhibited luciferase activity from the Nol3, Smarca4, Notch3, and Stmn1 3′ UTR reporters (Fig. 7), respectively, but did not affect luciferase activity from the pmirGLO vector lacking the MREs (Fig. S5). These data support the reciprocal relationship of these miRNAs and their target proteins in the PCB exposure groups and confirm the previous report that miR-9 targets Stmn1 (Delaloy et al., 2010). Further experiments in tractable mouse hepatocytes will be required to further examine the miRNA inhibition of the translation of these proteins.

Fig. 7.

Validation of reciprocal gene-miRNA targets regulated by PCB exposure in HFD-fed mouse liver. HEK-293 T cells were transfected with pmirGLO dual luciferase reporters into which the 3′-UTR miRNA targets of the indicated mouse genes were cloned. As indicated cells were co-transfected with pre-miR-Control or the indicated pre-miR. Each transfection ‘condition’ was done in quadruplicate within a 96-well plate. Dual luciferase assays were performed 48 h after transfection. These data are individual points from five biological replicate experiments. Two-way ANOVA identified a significant interaction (p = 0.0085) between the gene target reporters and the transfection. The gene target p = 0.0083 and the transfection p = < 0.0001. Tukey’s multiple comparison test as a post hoc analysis **** p < 0.0001 vs. the reporter and the miR Control transfection.

4. Discussion

PCBs are metabolism-disrupting chemicals (MDCs) that are associated with NAFLD in human populations (Cave et al., 2022). The precise mechanisms by which PCBs in combination with HFD induce NAFLD progression to steatohepatitis in the ‘two-hit’ model of TASH remain to be determined. Our previous study demonstrated that co-exposure to HFD and a single dose of the NDL PCB mixture Aroclor1260 increased HFD-induced steatosis and worsened inflammation and fibrosis in a mouse model of NAFLD progression to TASH by inhibiting hepatic EGFR signaling pathways (Hardesty et al., 2018, 2017, 2019a). One pathway identified with PCB exposure was a decrease in proteins associated with “epigenetic regulation of gene expression” (Jin et al., 2020). Epigenetic regulation includes miRNA post-transcriptional regulation of the proteome and serum miRNAs are new markers of liver injury in PCB-exposed human populations (Cave et al., 2022). The aim of this study was to identify the regulatory transcriptional-protein networks altered in PCB-exposure-induced TASH. Unbiased mRNA- and miRNA-seq analyses of livers from HFD-fed mice exposed to Aroclor1260, PCB126, and the combination of Aroclor1260 + PCB126 has uncovered deeper insight into TASH. Our multi-omics approach is the first to integrate the mRNA transcriptome, miRome, and proteome in this TASH model.

There was little overlap between DEGs and DEPs within PCB exposure groups. Although we detected a low percentage of coordinate reciprocal relationships between DEPs and DEGs putatively targeted by DEMs, we identified and validated four reciprocal (inversely) associated protein-miRNA ‘pairs’ that were regulated by the PCB exposures. We demonstrated that miR-712–5p targets the MRE in Nol3 (nucleolar protein 3), which was increased by Aroclor1260 exposure whereas miR-712–5p was reduced. Nol3 is a key apoptotic regulator that was increased in female low-density lipoprotein receptor (LDLR) knockout mice fed a western diet (Garcia-Jaramillo et al., 2019). Smarca4 (SWI/SNF Related, Matrix Associated, Actin Dependent Regulator Of Chromatin, Subfamily A, Member 4) protein expression was increased in PCB126 exposed liver whereas miR-466 f-3p was reduced and we demonstrated that miR-466 f-3p targets the MRE in the 3′ UTR of Smarca4. Overexpression of this SWI/SNF ATPase involved in chromatin remodeling increases the transcription of inflammatory cytokines inducing steatohepatitis in male C57Bl/J6 mice fed a HFD (Tian et al., 2013). We demonstrated that miR-206–3p targets the MRE in the 3′ UTR of Notch3 (Notch Receptor 3). Notch3 protein was increased whereas miR-206–3p was downregulated with Aroclor1260 + PCB126 exposure. Notch3 was upregulated in iPSC-derived hepatocytes from patients with NAFLD (Duwaerts et al., 2021). Notch3 was highly expressed in the liver tissue of patients with chronic hepatitis (Chen et al., 2012) and it has also previously been associated with liver fibrosis and hepatocellular carcinoma (Auguet et al., 2020; Schwabe et al., 2020). Stmn1 (Stathmin 1) protein was downregulated and miR-9–5p was increased by PCB126 exposure. Our data are in agreement with a report that Stathmin 1 protein and transcript are low in normal mouse liver, but were increased after oxidant injury from CCl₄ exposure (Zhao et al., 2020). Further experiments will be necessary to validate our results demonstrating miRNA-3′ UTR MRE targeting of these liver genes in liver cells.

The lack of overlap between DEGs and DEPs suggests that exposure to the NDL-PCB mixture in Aroclor1260, the DL-PCB126, and the co-exposure to Aroclor1260 and PCB126 elicit complex, direct and indirect systemic receptor-mediated effects resulting in the observed transcriptomic and proteomic profiles. Further, these observations suggest that transcript abundance is not a good proxy for protein abundance in this model and that regulation of transcript stability, translation, and protein stability regulate the proteome. Likewise, few miRNA-mRNA targets were concordantly regulated in opposition, i.e., miRNA up and mRNA target down or miRNA down and mRNA target up, including those protein-miRNA ‘pairs’ that we validated here. We suggest that additional mechanisms of regulation are involved in protein changes in TASH. Notably, despite the lack of direct overlap or linkage between mRNAs, miRNAs, and proteins, integrated MetaCore enrichment analysis identified pathways associated with hepatic injury in NAFLD and NASH. However, the sensitivity of the sequencing techniques utilized probably exceeded that of the proteomic techniques, especially for proteins with low abundance.

Many more RNAs and miRNAs were altered by co-exposure of HFD-fed mice to Aroclor1260 + PCB126, suggesting that the combination of the DL- and NDL- PCBs activate or inhibit cell receptors that initiate crosstalk involving novel cell signaling responses. Co-exposure to Aroclor1260 + PCB126 altered the abundance of hepatic genes enriched in pathways associated with the accumulation of lipids, hepatocyte damage, and the development of liver fibrosis, all changes that we reported in TASH (Jin et al., 2020). For example, we identified the dysregulation of hepatic genes that play a role in lysophosphatidic acid (LPA) signaling via GPCR2, e.g., Cd36, Pik3cb, Rock1, and Foxo1, with Aroclor1260 + PCB126 co-exposure. LPA causes ER stress, inflammation, and necrosis in hepatocytes (Neuschwander-Tetri, 2010) and is elevated with liver injury (Watanabe et al., 2007). Aroclor1260 + PCB126 co-exposure also altered “regulation of actin cytoskeleton organization by kinase effectors of RhoGTPases”. These data agree with the liver changes in cytoskeletal-associated proteins, i.e., troponin C, actin α, myosin light chain 3, and type II keratins, reported previously in this co-exposure model of TASH (Jin et al., 2020).

Regulation of metabolism, including “bile acids regulation of glucose and lipid metabolism via FXR”, was another pathway altered by the co-exposure to Aroclor1260 + PCB126. While the NDL-PCB congeners in Aroclor1260 do not bind FXR or LXR directly (Wahlang et al., 2014a), the combined effects of Aroclor1260 + PCB126 altered the transcript expression of a number of hepatic transcription factors, i.e., Fxr, Ppara, Foxo1, Foxa 2,3, Rxra, Nr1h3, and Hnf1a, which regulate glucose and lipid metabolism. However, no change was detected in protein expression of Rxra, Nr1h3, and Hnf1a in this co-exposure model (Jin et al., 2020). This suggests factors controlling transcript stability or translation may be altered in this co-exposure group. For example, Ppargc1 is upregulated with Aroclor1260 + PCB126 co-exposure and is a putative target of miR-466 f-3p that was downregulated in this exposure group. In addition, miR-206 was downregulated with Aroclor1260 + PCB126 co-exposure. miR-206 represses LXRα which regulates liver lipid metabolism (Zhong et al., 2013); thus, the decrease in miR-206 and miR-466 f-3p may be involved in the liver fat accumulation in this model. Future studies will examine these possibilities.

We reported that the interaction between HFD and Aroclor1260 attenuated the expression of receptors and transcription factors protective against NASH, i.e., peroxisome proliferator-activated receptors (PPARα/δ/γ), NRF2 (Nuclear Factor, Erythroid 2 Like 2), and cAMP-responsive element binding protein (CREB-1) (Hardesty et al., 2019b). Here, we found no change, or in the case of Ppara and Pparg, a small increase in mRNA transcript levels for these transcription factors with Aroclor1260 exposure. The differences between proteomic and transcriptomic results are likely due to post-transcriptional regulation and stability of transcripts compared to proteins in the frozen liver samples. We did not identify any changes in known or putative Ppara or Pparg regulating miRNAs with Aroclor1260 exposure. Pathway analysis of the DEGs in the Aroclor1260 exposure group revealed “enrichment of immune response pathway via major histocompatibility complex (MHC) class I molecules”, consistent with previous reports of elevated pro-inflammatory cytokines and increased macrophage infiltration this model (Wahlang et al., 2014b). The increased transcript abundance of Toll-like receptors 3 and 7 (Tlr3, Tlr7) also suggest activation of hepatic immune responses with Aroclor1260 exposure. Transcript levels of heat shock proteins involved in the MHC class 1 immune response signaling, e.g., Hsph1 (Hsp105) and Hspa1a (Hsp70), were reduced. Aroclor1260 exposure downregulated Hsp70 which is an anti-inflammatory protein with decreased hepatic expression in obese patients and inversely correlated with disease progression in NASH (Di Naso et al., 2015). Conversely, Aroclor1260 exposure upregulated the abundance of transcripts of chemokine receptors, i.e., Ccr2, Ccr5, and Cd36, associated with liver inflammation in obesity, type 2 diabetes, and metabolic syndrome and with activation of Kupffer cells, resident liver macrophages involved in fibrosis (Kolios et al., 2006). These findings are consistent with our report that Aroclor1260 exposure exacerbated neutrophil and macrophage infiltration of the livers of HFD-fed mice (Jin et al., 2020) and with transformation from steatosis to steatohepatitis, and hepatocyte necrosis (Wahlang et al., 2014b) previously reported.

It is well established that PCB126, a DL PCB, activates AHR and induces hepatic expression of Cytochrome P450 drug-metabolizing enzymes (Hardesty et al., 2017; Wahlang et al., 2017a). The increase in Cyp1a2 and Cyp1a1 transcript abundance in the PCB126 exposure groups is consistent with the increased protein expression with these exposures levels (Jin et al., 2020; Wahlang et al., 2017b). In contrast to the reduction in CCl2 in the PCB126 exposure group, Ccl2 was previously found to be upregulated in the livers of mice with methionine-choline deficient (MCD) and PCB126 exposure- induced NAFLD (Wahlang et al., 2017a). These data suggest that HFD differentially contributes to the PCB126 regulation of CCl2. Integrated analysis of the mRNA-seq, miRNA-seq, and the proteome data from the HFD + PCB126 exposure group identified upregulation of the Wnt/β-catenin signaling pathway mediated by increases in Fzd4, Wnt11, Wnt5b, and Prox1 abundance. While the precise role of non-canonical Wnt signaling in NASH and NAFLD is unclear, the pathway was activated in HSCs during hepatic fibrosis (Myung et al., 2007) and β-catenin upregulated lipid metabolism in a HFD-fed mouse model (Behari et al., 2014). HSCs are activated following liver injury (Puche et al., 2013). Interestingly, PCB126 exposure downregulated miR-466 f which has been validated to target and decrease protein expression of Wnt, β-catenin, and p-P65 (Li and Yan, 2021). Integrated pathway analysis for the PCB126 exposure group also included “hypoxia-inducible factor 1-α (HIF-1α)-mediated transcription”. HIF-1α plays a key role in HSC activation in liver fibrogenesis (Moon et al., 2009). The increase of Tfrc transcript abundance in PCB126 exposure suggests dysregulation in iron metabolism that may be involved in the pathways “sequestering of iron ion” and “sequestering of intracellular iron ion” identified in our proteome analysis for this exposure (Jin et al., 2020). While we have previously shown that PCB126 exposure induced steatosis, but not fibrosis, in NAFLD (Jin et al., 2020), the alteration of Wnt and HIF1-α signaling pathways suggest a possible mechanism for PCB126-induced HSC activation. Further studies will be necessary to examine this possibility.

To address the human health relevance of the changes in hepatic miRNA expression identified here we examined the expression of several miRs that were recently associated with PCB-associated TASH in the ACHS cohort (Cave et al., 2022). Significant changes were observed in mouse liver levels for each of the miRs examined in at least one treatment group. While the direction of the changes varied in some cases between mouse liver and human serum, this is not necessarily surprising due to the different compartments and species analyzed. Furthermore, the PCB exposures (actual vs. assessed) were somewhat different. For example, PCB 126 exposures were only assessed in a subset of ACHS-I participants (Pavuk et al. 2014). Therefore, potential associations between miRs and PCB 126 have not been determined in ACHS-I. However, we were able to demonstrate that several of the miRs that were associated with PCBs and liver disease in ACHS-I were differentially regulated by PCBs in this fatty liver disease animal model. The most consistent finding across studies was for miR-192–5p. While PCBs did not regulate expression of hepatic miR-122–5p levels in mice, it is important to note that circulating miR levels were not measured in the animal models. Moreover, liver and circulating levels of MiR-122–5p levels may even be discordant in NAFLD (Gjorgjieva et al., 2019; Lopez-Sanchez et al., 2021; Pirola et al., 2015; Zhang et al., 2021). In the follow-up ACHS II cohort, additional dioxin-like PCBs have been measured, including PCB 126 (Birnbaum et al., 2016; Pittman et al., 2020). Our ongoing analyses looking at associations between PCB congeners and miRNAs in ACHS II will shed more light into how dioxin-like molecules regulate miRs in the context of human environmental exposures.

5. Conclusions

This study is the first to employ a multi-omics approach to address the regulatory miRNA-mRNA-Protein network dysregulated by PCB-exposure in a HFD-fed model of TASH. The data presented here demonstrate that HFD and DL-, NDL-, and co-exposure to DL and NDL-PCBs alter hepatic mRNA and miRNA expression with minor concordance between mRNA transcripts and protein changes in this model (Jin et al., 2020). Similarly, only a few miRNA changes are directionally associated with changes in their mRNA or protein targets. The poor correlation between expression levels of mRNA and protein is not uncommon and can be attributed to multiple layers of post-transcriptional and post-translational modification (de Sousa Abreu et al., 2009; Vogel and Marcotte, 2012). These observations suggest that the regulation of the hepatic proteome in the HFD-fed mice exposed to PCBs is likely mediated by additional mechanisms controlling transcript stability, processing, and translation and by factors regulating protein abundance.

Integrated analysis revealed that the PCB-disrupted transcriptome changes identified here are related to pathways associated with NASH and NAFLD. In our previous studies, we thought that PCB-mediated hepatic receptor activation was the major mode of action of hepatic toxicity (Hardesty et al., 2019a). However, when we used the receptor knockout mouse models (Car, Pxr, Ahr), we still observed PCB toxicity implicating off-receptor targets for PCB action in TASH (Wahlang et al., 2021). This study addressed one of those targets, i.e., miRNAs. However, since only ~ 10 % of the observed DEGs and ~4 % of proteins are associated with miRNAs, additional mechanisms regulating gene expression require further research, e.g., epitranscriptomic regulation of individual transcripts (Alfonzo et al., 2021; Klinge et al., 2021).

Among the weaknesses of our study is that the mRNA and miRNA were isolated from frozen liver samples which may affect mRNA transcript abundance. Thus, it will be important to confirm these findings with coordinated analysis on freshly prepared samples. While we have described differences in the hepatic mRNAome, miRome, and proteome with HFD and PCB exposure, we have not yet elucidated the specific mechanisms connecting these observations. Additionally, we did not examine the impact of sex in this model. Lastly, the study includes a single, acute PCB exposure rather than chronic long-term exposure mimicking human exposure. These important parameters will be examined in future work.

Abbreviations:

ACHS

Anniston Community Health Survey

AHR

aryl hydrocarbon receptor

ALT

Alanine transaminase

AR

androgen receptor

AST

aspartate transaminase

CAR

constitutive androstane receptor

CREB-1

cyclic AMP-responsive element binding protein

CYP

cytochrome P450

DEG

differentially expressed genes

DEM

differentially expressed miRNA

DL

dioxin-like

EGFR

epidermal growth factor receptor

EMT

epithelial-mesenchymal transition

ER

endoplasmic reticulum

FXR

farnesoid X receptor

GO

gene ontology

HCC

hepatocellular cancer

HFD

high fat diet

HNF4

Hepatocyte Nuclear Factor 4 Alpha

HSCs

hepatic stellate cells

KEGG

Kyoto Encyclopedia of Genes and Genomes (https://www.kegg.jp/)

mRNA seq

mRNA sequencing

miRNA seq

miRNA sequencing

NAFLD

non-alcoholic fatty liver disease

NASH

non-alcoholic steatohepatitis

NDL

non-dioxin-like

NRF2

Nuclear Factor, Erythroid 2 Like 2

PCB

polychlorinated biphenyl

PPAR

peroxisome proliferator-activated receptor

PXR

pregnane X receptor

QN

Quantile normalization

qRT-PCR

quantitative real-time PCR

TASH

toxicant-associated steatohepatitis

TCDD

tetrachlorodibenzo-p-dioxin

Appendix A. Supporting information

Supplementary data associated with this article can be found in the online version at doi:10.1016/j.etap.2022.103928.