Introduction

Health concerns pertinent to environmental pollutants, including exposures to polychlorinated biphenyls (PCBs), continue to intensify in the field of environmental health as global industrialization flourishes. PCBs are anthropogenic environmental toxicants, originally manufactured for industrial purposes and subsequently banned from production and usage when their carcinogenic properties were reported1. While one of the predominant and commonly known health effects associated with these legacy forever chemicals is cancer, other less severe but more chronic effects such as cardiometabolic diseases have also been acknowledged2,3. Notably, associations between PCB exposures and steatotic liver disease, hypertension, and diabetes have been substantiated in numerous epidemiologic and toxicological studies4,5,6,7.

Despite production being halted for decades now, PCBs still remain in our environment due to their recalcitrant nature towards both degradation and metabolism8. PCBs are still detected in the American population as documented in the National Health and Nutrition Examination Survey (NHANES) studies with heavily chlorinated PCB congeners having higher detection levels9. These polyhalogenated organic compounds are broadly classified as coplanar or non-coplanar congeners, based on the degree of chlorination on the biphenyl ring. Structure-activity relationships conferred coplanar congeners as activators of the aryl hydrocarbon receptor (AHR), similar to 2,3,7,8-Tetrachlorodibenzo-p-dioxin (TCDD) and hence also known as “dioxin-like” PCBs10. In contrast, non-coplanar PCBs are better activators of the constitutive androstane receptor (CAR) and often referred to as “non-dioxin-like” or phenobarbital-like PCBs11. Previously, our research group have demonstrated how dioxin and non-dioxin like PCBs can elicit differential effects on the liver and other metabolic parameters depending on the dose and duration of exposure12,13,14. Importantly, our studies also highlighted the significance of accounting for sex differences with PCB exposures as these toxicants implicitly act as both metabolism and endocrine-disrupting chemicals15. Indeed, female mice acutely exposed to a mixture of dioxin-like and non-dioxin-like PCBs manifested greater susceptibility to steatotic liver disease and impaired glucose uptake compared to exposed males16.

Steatotic liver disease is a spectrum of liver pathologies, initially manifested by accumulation of lipid droplets in the liver (steatosis), which may be accompanied by inflammation (steatohepatitis). If left untreated, steatohepatitis can further progress to fibrosis and cirrhosis, and ultimately hepatocellular carcinoma and fulminant liver failure. Steatotic liver disease is either a risk factor for other metabolic complications such as diabetes, obesity and cardiovascular diseases or occurs in parallel with one or more of these complications17. Multiple mechanisms with regards to PCBs’ contribution to steatotic liver disease and to an extent, glucose intolerance and diabetes, have been postulated. Primarily, PCBs have been shown to interact with hepatic xenobiotic and endobiotic receptors to elicit changes in the hepatic transcriptional machinery leading to downstream effects on hepatic lipid and carbohydrate metabolism7,18,19. Further, PCBs can also interact with nuclear receptors including thyroid and sex hormone receptors, resulting in endocrine dysfunction and disruption of overall energy homeostasis20. In addition, PCBs can induce oxidative and endoplasmic reticulum stress and trigger inflammatory responses through interactions with NF-KB, among other means4,21. PCBs can also inhibit key phosphoproteins in the liver and interrupt physiological phosphorylation processes leading to hepatic signaling disruption22,23.

More recently, other novel mechanisms identified for PCB contributions to steatotic liver disease include PCB-induced RNA modifications in the liver (hepatic epitranscriptome) resulting in changes in hepatic transcript levels for key genes regulating energy metabolism24,25. Another increasingly recognized mechanism is PCB effects on gut microbiota and intestinal function that can impact liver health and function, in part, through the gut-liver axis7,26,27. However, all these mechanistic toxicologic studies were performed majorly in male experimental models and there is a lack of understanding on how these biological events pan out in females. To address this knowledge gap, the current study aims to identify sex differences in the hepatic proteome and gut microbiome, and assess how biological sex impact PCB responses, in order to gain insight on sex-specific liver proteins and gut parameters impacted by PCB exposures. This comprehensive molecular profiling will also provide information on sex-specific pathways that potentially play a role in driving distinct PCB toxicity outcomes in males versus females. To achieve this objective, the current study adopts a holistic, multi-omic approach using liver and intestinal samples obtained from a previously characterized acute exposure mouse study16 with verified sex-specific phenotypic outcomes, in part, to ensure validity of the study purpose and objective.

Materials and methods

Animal model

The mouse exposure study was approved by the University of Louisville Institutional Animal Care and Use Committee (IACUC) under Protocol # 14077. The UofL IACUC assumes responsibility for those functions given in the National Research Council Guide for the Care and Use of Laboratory Animals, the NIH Public Health Service Policy on Humane Care and Use of Laboratory Animals, and the Animal Welfare Act. Thus, all animal experiments were performed in accordance with the National Research Council Guide for the Care and Use of Laboratory Animals. Further, the animal experiments and data reporting are in compliance with the ARRIVE guidelines. Male and female C57Bl/6 mice (7–8-week-old) were obtained from The Jackson Laboratory (Bar Harbor, ME) and housed in a temperature- and light- controlled room (12 h light; 12 h dark) with food and water ad libitum. Five mice were housed per cage in this 2-week study. All animals were fed a low-fat, synthetic diet consisting of 20.0%, 69.8%, and 10.2% of total calories from protein, carbohydrate, and fat respectively (TD06416, Envigo, Indianapolis, IN) throughout the study period. Mice were randomly divided into four groups (n=10, total=40) based on sex and exposure utilizing a 2X2 design (Male-CON, Male-PCB, Female-CON, Female-PCB), On week 0, mice were allowed to acclimate to the new facility. At the beginning of week 1, vehicle control (corn oil) or a mix of PCBs (purchased from AccuStandard, CT) solubilized in corn oil was administered to mice by oral gavage. The PCB mix comprised of the commercial PCB mixture, Aroclor 1260 (20 mg/kg) which constitutes a higher proportion of non-dioxin-like and heavily chlorinated PCB congeners, and the classic dioxin-like congener, namely PCB126 (10 μg/kg). At the beginning of week 2, a second dose of PCB126 (10 μg/kg) or vehicle control was administered to mice via oral gavage. Repeated dosing for PCB 126 was performed because, as opposed to the heavily chlorinated congeners in Aroclor 1260 which are highly resistant to metabolism, PCB 126 is less chlorinated and can be metabolized in the liver. Thus, the repeated dosing will ensure that the parent compound effect is sustained in the two-week study. At the end of week 2, mice were euthanized using ketamine/xylazine (100/20 mg/kg body weight, i.p.) and samples collected for further analysis. A schematic diagram depicting the experimental design, timeline, dosing regimen and downstream analyses is provided in Supplemental Fig. S1.

Dose rationale

The PCB doses utilized in the current study were designed to mimic current PCB bioaccumulation patterns in humans and routinely used by our group in PCB exposure mouse models16,18. The congener composition in Aroclor 1260 also strongly resemble human PCB bioaccumulation patterns11. Circulating levels of PCBs (lipid-adjusted) reported in measurements from NHANES participants and PCB-exposed populations such as the Anniston Community Health Survey (ACHS) range from around 70 to 170 ng/g9,28,29. The National Toxicology Program (NTP) has also performed extensive studies on PCB tissue distribution in rodents in 2-year gavage studies and demonstrated that PCB levels in the liver were approximately 10 times higher than lipid-adjusted serum levels, irrespective of the dose administered or exposure duration30. Based on extrapolations from these NTP studies, a 20 mg/kg cumulative dose of Aroclor 1260 would result in serum PCB levels at approximately 170 ng/g and liver PCB tissue levels at around 3000 ng/g. Thus, the Aroclor 1260 dose is expected to yield serum PCB levels similar to those measured in the ACHS participants. With regards to PCB 126, exposures by total PCB mass to dioxin-like PCBs including PCB 126 is >0.1% of the total PCB exposures in NHANES participants9, reflective of the cummulative dose of the 20 μg/kg (0.1% of Arcolor 1260) used in the study.

Proteomics

Protein samples (n=5) were extracted from mouse livers in 1% SDS modified RIPA buffer using bead homogenization. To avoid cage effects, mouse livers were selected randomly for each group but ensuring that at least two liver samples of a group came from animals housed in a different cage than the other three samples of that same group. Protein concentrations were measured using Pierce™ BCA Protein Assay Kit (Thermo Fisher Scientific, Waltham, MA) according to the manufacturer’s instructions. Protein lysates (200 μg) were trypsinized using the modified filter-aided sample preparation method. Firstly, protein samples were reduced and denatured using dithiothreitol and 8 mol/L urea, respectively, and alkylated using iodoacetamide. This was followed by centrifugation through a high molecular weight cutoff centrifugal filter (Millipore Sigma, 10k MWCO) and subsequent overnight digestion with sequencing grade trypsin (Promega, Madison, WI, USA) at 37 °C. The digested peptides were collected and cleaned with a C18 Proto™ 300 Å ultra microspin column. Digested peptide samples (50 μg) were then labeled with tandem mass tag (TMT) TMT10plex™ isobaric label reagent set (Thermo Fisher Scientific). Labeled samples were concentrated and desalted with Oasis HLB extraction cartridges (Waters Corporation, Milford, MA) using a modified protocol for extraction of the digested peptides. Next, samples were then separated by high pH reversed phase separation with fraction concatenation on a Beckman System Gold LC system supplemented with 126 solvent module and 166 UV–Vis detector in tandem with a BioRad Model 2110 Fraction Collector. Liquid chromatography/mass spectrometry was used to measure TMT-labeled peptides. Briefly, every high pH reversed phase fraction was dissolved in 50 μL solution of the combination of 2% v/v acetonitrile with 0.1% v/v formic acid. 1 μL of each fraction was analyzed on EASY-nLC 1000 UHPLC system (Thermo Fisher Scientific) and an Orbitrap Elite–ETD mass spectrometer (Thermo Fisher Scientific). The raw spectral data obtained from the mass spectrometer were analyzed using Proteome Discoverer (2.5.0.400 Settings).

Proteomic analysis

Statistical analyses of the proteomic data were performed using R software. Proteins exhibiting 40% or more missing values were excluded from the analysis, which resulted in 3837 proteins out of 4533 identified. All 3837 proteins showed complete observations in all 20 samples and no imputation on missing values was needed. Quantile normalization method was then conducted on the complete dataset across different groups using the function “normalize.quantiles” in R package “preprocessCore”31. Subsequently, the data were analyzed using linear models, with sex and exposure as the main effects, and the response variable being the log-transformed expression data (base 2). Contrasts using moderated t-tests were generated to explore exposure effect (PCB vs. control) for each sex, sex difference for each exposure group, as well as the interaction between exposure and sex using “Limma” package32. The results were reported as mean expression levels, log fold-change (FC), p-value, and p-value adjusted by the Benjamini-Hochberg (BH) method to control the false discovery rate (FDR)33. As this was a discovery-based, high throughput analysis, the FDR was set at <0.1, and the unadjusted p-value were also provided. Computational analyses on significantly altered proteins for each comparison group (up-regulated proteins at log2 FC ≥0.5, down-regulated proteins at log2 FC ≤−0.5, along with adjusted p-value <0.10) were performed using MetaCore (Clarivate Analytics). Briefly, the protein lists for each comparison group which included both upregulated and downregulated proteins along with the respective fold change were uploaded to MetaCore. Analyses including “GO Process”, “Interaction by Protein Function”, “Enrichment by Diseases (by Biomarkers)” and “Enrichment by Pathway Maps” were performed for each comparison group using “One-click Analysis” with parameters set at p-value <0.05 and FDR <0.05.

16S rRNA sequencing

Microbial genomic DNA was extracted from mouse cecal samples (n=10) using the DNeasy PowerSoil Pro Kit (Qiagen, Germantown, MD) according to the manufacturer’s instruction. Illumina MiSeq technology targeting the variable V3 and V4 regions of 16S ribosomal RNA was used to determine cecal bacterial composition. Libraries were prepared using Illumina’s 16S library preparation guide and Illumina’s Nextera Index Kit (FC-121–1012). Briefly, quantitation of microbial genomic DNA was performed using Qubit Broad Range (BR) assay (Thermo Fisher Scientific) in a Qubit 2.0 Fluorometer. Amplicon PCR was carried out using primers that were complementary upstream and downstream of the region of interest with overhang adapters. The 16S variable region was amplified using 12.5 ng of microbial genomic DNA and Amplicon PCR Clean‐Up carried out using AMPure XP beads. Index PCR was then performed to attach dual indices and Illumina sequencing adapters using the Nextera XT Kit (FC-121–1012). Following the Index PCR Clean-Up using SpriSelect beads, sequencing libraries were aliquoted and mixed to make pooled libraries and concentrations determined on the Bioanalyzer. Next, after normalization, the pooled library was denatured and mixed with the PhiX control library. Sequencing was done on a Nano-300 cycle test chip (MS-103–1001) to confirm sample concentration followed by Illumina MiSeq Reagents kit v3 (600 cycles) (MS-102–3003) at 10 pM and 30% PhIX.

16S rRNA metagenomics analysis

The microbial diversity analysis and taxonomic profiling were conducted using Quantitative Insights into the Microbial Ecology (QIIME) version 2–2022.2.234. The pipeline utilized for bioinformatics analysis is described as follows: i) Quality control of the 40 raw sequences FASTQ files was performed using FastQC version 0.10.135, ii) Demultiplexed sequence reads were imported to QIIME2-2022.2, and trimming was performed to remove low-quality data using the parameters p-trunc-len-f 270 and p-trunc-len-r 220. These trimming thresholds were selected after examining the interactive quality plots (Supplemental Fig. S2) to ensure that high quality sequence regions were retained while still maintaining sufficient overlap between paired-end reads, iii) Denoising of reads into amplicon sequence variants (ASVs) was performed using DADA2 QIIME2-2022.2 plugin36, iv) Taxonomic classification of ASVs was done using VSEARCH qiime2-2022.2 plugin at similarity cutoff value at 97% followed by mapping of ASVs using Greengenes 13_8_99-nb classifier37, v) Phylogenetic tree was constructed using QIIME phylogeny align-to-tree-mafft-fasttree algorithm. For data rarefaction, it was determined that the smallest and largest number of read counts are 12,698 and 47,091, respectively (Supplemental Table S1). Based on the alpha rarefaction curves (Supplemental Fig. S3), the sequence sampling depth of 23, 192 was utilized when generating all alpha and beta diversity core metrics results. The Faith’s PD curve for M-CON-4 (12,698 reads) did not reach saturation at this depth, whereas rarefaction curves for all other samples plateaued above 20,000 reads. To avoid compromising diversity and taxonomic assessment, this sample was excluded from further analysis. In addition, this sample showed low reverse strand sequencing quality. After removal, the control group retained nine samples, which was sufficient for statistical power, and therefore all subsequent analyses were conducted without this sample. Beta diversity PERMANOVA analyses were performed using the QIIME2 pipeline, and principal component analysis plots were generated using Partek Flow 11.0.24.0414. To account for any potential cage effect, an additional PERMANOVA analysis using adonis2 function in R, which allows inclusion of “cage” as a covariate, was performed. Alpha diversity output obtained from QIIME2 was visualized and analyzed using the Kruskal Wallis test with Dunn multiple correction in GraphPad PRISM Version 10.4.2. Taxonomic data was represented as bar plots using GraphPad PRISM version 10.4.2, heatmaps using Partek genomics suite version 7.21.1119, and analyzed using the Kruskal Wallis test with Bayesian q-value correction38. Linear discriminant analysis Effect Size (LEfSe) was used to identify the enrichment profile of taxa characterizing the differences between the experimental groups39. Analysis was performed on family and genus relative abundance data using Huttenhower Lab Galaxy Server 2.0. Significant enrichment of taxa was calculated via a series of Kruskal Wallis non-parametric tests to determine the presence of significant alterations in taxonomic abundance between groups (adjusted p <0.05). This was followed up using linear discriminant analysis to estimate the effect size of any significant alterations set at linear discriminant analysis (LDA) cutoff score >2 and p <0.05.

RT-PCR and western blot analysis

Total RNA was extracted from mouse ileal samples (n=10) with RNA-STAT 60 (Tel-Test, Austin, TX) using bead homogenization. RNA purity and quantity was assessed with the NanoDrop One (Thermo Fisher Scientific) and cDNA was synthesized from total RNA using qScript cDNA Synthesis Kit (Quantabio, Beverly, MA). RT-PCR was performed on the CFX384 TM Real-Time System (BioRad, Hercules, CA) using iTaq Universal Probes Supermix and Taqman probes purchased from Thermo Fisher Scientific (Supplemental Table S2). Four housekeeping genes were tested and stability analyzed using RefFinder40, a web-based comprehensive tool for screening reference genes by integrating four computational programs (geNorm, NormFinder, BestKeeper, and the comparative ΔCt method). For the current study, Ppia was identified as the most stable housekeeping gene from the four tested (Supplemental Fig. S4). Gene expression levels were calculated according to the 2−ΔΔCt method. The levels of mRNA were normalized relative to the levels of the housekeeping gene (Ppia) and mean expression levels in unexposed, male mice were set at ∼1.

Liver tissue was homogenized in RIPA Buffer, supplemented with protease, and phosphatase inhibitors (Sigma-Aldrich, St. Louis, MO) to obtain protein samples for western blot analysis (n=6). Protein concentration was determined by Pierce™ BCA Protein Assay Kit (Thermo Fisher Scientific). Protein (30 μg) was separated on a 4–15% gradient SDS Gel (BioRad, Hercules, CA), transferred, blocked, and incubated with primary antibody according to the manufacturer’s instructions. The membranes were incubated in Pierce™ ECL agents (Thermo Fisher Scientific) and luminescent signals were captured with BioRad Chemidoc Imaging System. Western blot bands were quantified using BioRad Image Software. Primary antibody for CYP1A2 (D2V75) was obtained from Cell Signaling Technology (Danvers, MA, USA), CD36 (ab133625) and flavin monooxygenase 3 (FMO3, ab126711) from Abcam (Cambridge, MA, USA), and Actin (A5441) from Millipore Sigma (St. Louis, MO).

Additional statistical analyses

Graphs were plotted using GraphPad Prism 10.4.2 (GraphPad Software Inc., La Jolla, CA, USA) unless specified otherwise. Western blot and gene expression data were analyzed by two-way ANOVA, for two factors, namely “Sex” and “PCBs” and their interactions in GraphPad Prism 10.4.2. This was followed by multiple comparison tests with Bonferroni correction for sub-group analyses (adjusted p <0.05 was considered statistically significant). In the event of high data variability in a group, an outlier test was performed using the ROUT method with false discovery rate (FDR) set at 0.1 %. In terms of reporting the findings, based on the research objective and study design, basal sex differences were reported first, followed by PCB effects.

Results

Effects of biological sex and PCB exposure on the hepatic proteome

A total of 3837 hepatic proteins were analyzed after adjustment for missing values with the exposed male versus female group showing the greatest number of differentially expressed (DE) proteins among all comparison groups (Fig. 1A, B & C). A principal component analysis of the normalized dataset was also performed to visualize the distinct proteomic profiles between groups (Fig. 1D). While biological sex was a major driver for the DE proteins observed (44 upregulated and 41 downregulated in MCON vs. FCON), PCB exposure further increased the number of DE proteins between males and females (61 upregulated and 82 downregulated in MPCB vs. FPCB). PCB exposure within each sex however resulted in relatively modest changes (8 upregulated in MCON vs. MPCB and 3 upregulated in FCON vs. FPCB). A list of the top ten DE proteins for each comparison group is shown in Table 1 and a list of all DE proteins for each comparison group is presented in Supplemental ExcelTable 1. Notably, there were no similar DE proteins identified between all 4 comparison groups (Table 1 and Fig. 1E). Nonetheless, 66 common DE proteins were identified in both the MCON vs. FCON and MPCB vs. FPCB groups including flavin-containing monooxygenase 3 (FMO3) and cytochrome P450 enzymes such as CYP2A4. Only 2 common DE proteins were identified in both MCON vs. MPCB and FCON vs. FPCB groups, namely CYP1A1 and CYP1A2. Unique DE proteins identified in the MCON vs. MPCB group included CYP2B19, CYP4A10, CYP2A5, CYP2C29, CYP5A and retinol-binding protein 4 (RBP4). Only one unique DE protein was observed in the FCON vs. FPCB group, namely HIV-1 tat interactive protein 2 (HTATIP2), an oxidoreductase.

Fig. 1
figure 1

Effect of sex and PCBs on the hepatic proteome. (A) A summary table of the overall results is presented. Proteins exhibiting 40% or more missing values were excluded from the analysis, which resulted in 3837 proteins being analyzed out of 4533 identified. The number of differentially expressed (DE) proteins was obtained for each comparison group. p-Not Adj is p <0.05 and p-Adj is p <0.1, after correcting for false discovery rate. FC - fold change. Volcano plots were generated to graphically depict alterations in hepatic proteins for (B) MCON vs. FCON and (C) MPCB vs. FPCB. (D) Principal component (PCA) analysis of the normalized raw proteomic data was performed and a PCA plot generated. (E) A Venn diagram showing the number of shared or unique DE proteins based on the four comparison groups was constructed.

Full size image
Table 1 The top ten differentially expressed (DE) proteins based on ascending adjusted p-values for each comparison are presented.
Full size table

To validate the proteomic findings, western blot analysis of selected DE proteins including FMO3 (upregulated in both female groups), CYP1A2 (upregulated in both PCB groups) and CD36 (upregulated only in the exposed female group vs. the exposed male group) was performed. Importantly, CD36 is involved in hepatic lipid uptake and its upregulation is implicated in steatotic liver development41. In concordance with the proteomic results, protein expression of FMO3 was increased in both the female groups with and without exposure (Fig. 2A and Supplemental Fig. S5) with two-way ANOVA reporting a significant sex effect. Exposed females also exhibited a trend towards significance for higher FMO3 expression vs. their sex-matched controls (p=0.051), potentially reaffirming the proteomic findings wherein the MPCB vs. FPCB group displayed higher FMO3 fold induction than MCON vs. FCON. In contrast to FMO3, CYP1A2 expression was upregulated only in PCB-exposed groups, regardless of sex (Fig. 2B and Supplemental Fig. S5) with two-way ANOVA reporting a significant PCB effect. Lastly, CD36 protein expression was largely driven by sex, and this difference was more pronounced in the exposed male vs. exposed female groups (Fig. 2C and Supplemental Fig. S5). Nonetheless, exposed females also showed a trend towards significance for increased CD36 protein expression levels (p=0.054), compared to their sex-matched controls. Overall, the data from the western blot analysis aligned with the proteomic output results.

Fig. 2
figure 2

Validation of selected proteins using western blot analysis. Selected hepatic proteins that were significantly altered were further validated by Western blot analysis, namely (A) FMO3, (B) CYP1A2, and (C) CD36. Representative blots are shown. The densitometry ratios of the target protein to the corresponding control protein, Actin was calculated. Note: FMO3 and CD36 share the same Actin control. A list of p-values from the two-way ANOVA analysis is also provided. Values are mean ± SD, *p <0.05.

Full size image

Pathway analysis for differentially expressed proteins and identification of sex-specific enriched pathways

To identify key biological processes and pathways impacted by changes in hepatic proteins for each comparison group, downstream computational enrichment analyses were performed using MetaCore. With regards to “GO Processes”, similar biological processes were affected in the MCON vs. FCON and MPCB vs. FPCB groups including metabolic and catabolic processes, xenobiotic responses, lipid synthesis and breakdown, and carbohydrate and protein processes (Fig. 3). These results highlighted basal sex differences in energy metabolism particularly for fatty acid synthesis and oxidation. Processes enriched only in the PCB-exposed male and female groups included “omega-hydroxylase P450 pathway”, “dibenzo-p-dioxin metabolic process”, “hepatocyte differentiation” and metal processes. Processes identified in all 4 groups included hormone processes such as regulation of hormone levels, implicating that both sex and PCBs may modulate these endpoints. Additional analysis, namely “Enrichment by Pathway Maps” was performed for the PCB-exposed groups and the top 10 enriched pathways shown in Table 2. Pathway enrichment analysis demonstrated that while in males, “aryl hydrocarbon receptor signaling pathway” was the top enriched pathway, “estradiol metabolism” was the topmost enriched pathway in females. PCB exposure in males and females however enriched similar pathways related to “dioxin-like” chemicals and primarily driven by CYP1A1 and CYP1A2 upregulation.

Fig. 3
figure 3

Sex and PCB effects on gene ontology (GO) processes. A heat map showing different processes that were altered by sex and/or PCB exposures according to the -log(p-value) for each comparison. The first experimental group mentioned in each comparison serves as the reference group. The processes were obtained by GO analysis enrichment of the different proteins altered by sex and/or PCB exposures using MetaCore. A - MCON vs. FCON, B - MPCB vs. FPCB, C - MCON vs. MPCB, D - FCON vs. FPCB.

Full size image
Table 2 The top ten enriched pathways along with the generated p-values and FDR values obtained from MetaCore using the “Enrichment by Pathway Maps” analysis for selected comparison groups.
Full size table

Interaction by protein function analysis and disease enrichment

To better assess protein activity and function based on DE proteins for each group, “Interaction by Protein Function” analysis was performed in MetaCore (Fig. 4A). Transcription factors were the major output from this analysis, while enzymes, co-activators and to an extent, micro-RNAs, were also identified. In terms of xenobiotic receptors, AHR showed the highest activity (by z-score), and this was observed only in the MPCB vs. FPCB and FCON vs. FPCB comparison groups, suggesting that AHR activation was predominant in the exposed female group. Surprisingly, CAR and pregnane-xenobiotic receptor (PXR) activity were observed only in the MCON vs. FCON and MPCB vs. FPCB groups. For endobiotic receptors, liver-X-receptor (LXR)-alpha activity was observed only in the MCON vs. MPCB and FCON vs. FPCB with the latter group showing a higher z-score, while peroxisome proliferator-activated receptor alpha (PPAR alpha) activity was observed in all comparison groups except in the FCON vs. FPCB group. Furthermore, for most hormone receptors, protein activity was only observed in the MCON vs. FCON and MPCB vs. FPCB groups. Likewise, for other transcription factors including hepatocyte nuclear factors, protein activity was only observed in the MCON vs. FCON and MPCB vs. FPCB groups. Significantly, APOB activity was observed only in the MPCB vs. FPCB group. However, with regards to micro-RNAs, activity for multiple microRNAs, including the liver specific miR-122-5p, was observed only in the FCON vs. FPCB group.

Fig. 4
figure 4

Interaction by Protein Function (IPF) and Disease Enrichment analyses. (A) An IPF analysis was performed on the DE proteins obtained for each comparison group using MetaCore. A heat map of selected objects obtained from the IPF analysis, along with their corresponding z-scores, was constructed. Increased z-score denotes increased activity. (B) A disease enrichment analysis was also performed on the DE proteins for each comparison group using MetaCore and a heatmap of selected enriched diseases based on the -log(p-value) was constructed. A - MCON vs. FCON, B - MPCB vs. FPCB, C - MCON vs. MPCB, D - FCON vs. FPCB.

Full size image

In addition to protein function, disease enrichment analysis was also performed to identify if differences in the protein landscape can modulate disease endpoints. Selected diseases including liver, intestinal, cardio-metabolic, and hormone-related diseases were examined (Fig. 4B). In general, “liver diseases” was enriched in all 4 comparison groups while only the MCON vs. MPCB and FCON vs. FPCB showed enrichment for “chemical and drug-induced liver injury”. Further, only the MPCB vs. FPCB group showed enrichment for “fatty liver”. “Intestinal diseases” were enriched only in the MCON vs. FCON and MPCB vs. FPCB groups while “endocrine system diseases” were enriched only in the MCON vs. MPCB and FCON vs. FPCB.

Effects of biological sex and PCB exposure on gut bacterial diversity

The 16S rRNA gene sequencing was performed to determine the sex-specific PCB effects on the gut microbiome. Differences in microbial diversity within individual samples were evaluated using alpha diversity measures, and differences in diversity between samples were accessed using beta diversity measures. Multiple diversity metrics were included to capture distinct aspects of microbial community structure (e.g., richness, evenness, phylogeny, and abundance), as is standard in exploratory microbiome studies. These different metrics offer unique and biologically relevant insights, in addition to providing complementary information and supporting a transparent and robust interpretation of the results.

Alpha diversity indices namely Observed Species which measures species richness within a group, and Shannon Index which measures species richness and evenness within a group, indicated a significant sex difference, with females showing lower alpha diversity than their male counterparts, irrespective of exposure (Fig. 5A). In contrast, the Pielou Index which measures evenness within a group, and Faith Phylogenetic Diversity (Faith_pd) which measures species evolutionary relatedness within a group, indicated no differences between the groups (Supplemental Fig. S6). Beta diversity indices namely Jaccard which measures compositional similarity between groups, and Bray Curtis which measures compositional dissimilarity between groups, revealed a significant compositional difference between all groups (Fig. 5B). Notably, these differences in composition between all groups remained significant after adjusting for cage effects (Supplemental Table S3). In comparison to Jaccard and Bray Curtis, UniFrac tests measures compositional relatedness between groups. Unweighted UniFrac which measures absence and presence of taxa, and Weighted which measures absence, presence, and abundance of taxa, both indicated a significant compositional difference between males and females irrespective of exposure (Supplemental Fig. S6). Overall, the diversity analyses reveal that alterations in microbial diversity were mainly due to sex differences. However, both exposure and sex had a significant effect on the composition of microbial communities that are distinctive between all groups.

Fig. 5
figure 5

Gut microbial composition analysis. (A) Alpha diversity indices namely observed species (OTUs - observed taxonomic units) and Shannon Index were calculated to measure differences in microbial diversity within individual samples. Values are mean ± SD, *Adjusted p <0.05. (B) Differences in microbial diversity between samples were accessed using beta diversity measures including Bray Curtis and Jaccard algorithms. PCA plots for these algorithms along with the respective Permanova-pairwise comparisons are presented.

Full size image

Sex and PCB exposure modulate gut bacterial abundance at higher taxa levels

Bacterial relative abundance at varying levels were further analyzed to identify the differences in microbial composition observed in the beta diversity analysis. At the phyla level (Fig. 6A and Supplemental Table S4), distinct sex differences were observed with females generally having lower abundance for Bacteroidetes, Proteobacteria and Tenericutes, and higher abundance for Actinobacteria, OD1, and Verrucomicrobia compared to males. However, in PCB-exposed females, there was only a significantly lower abundance for Tenericutes, and higher abundance observed only for Actinobacteria and OD1 (Parcubacteria) compared to exposed males. No significant alterations at the phyla level were observed in either males or females with PCB exposures. Additionally, there was no change in dysbiosis index (Firmicutes/Bacteroidetes) between samples, indicating that neither sex nor exposure led to significant microbial dysbiosis.

Fig. 6
figure 6

Effects of sex and PCB exposure at different taxonomic levels. Heatmaps depicting bacterial abundance at the (A) phyla and (B) family levels were generated. (C) Linear discriminant analysis (LefSe) was performed for relative bacterial abundance at the genus level and the results for each comparison group is presented.

Full size image

Similarly, at the family level (Fig. 6B and Supplemental Table S5), 36 bacteria were identified with 17 significant alterations in abundance observed in the unexposed female versus male groups while 12 significant changes were observed in the PCB-exposed female versus male groups. Prominent observations include decreased abundance for Bacteroidaceae and Staphylococcaceae and increased abundance for Clostridiaceae and Coriobacteriaceae in unexposed females versus their male counterparts. However, these effects disappeared except for Coriobacteriaceae in PCB-exposed females versus their male counterparts. Notably, in PCB-exposed females, there was an increase in Bifidobacteriaceae and decreased Ruminococcaceae abundance compared to PCB-exposed males. Further, PCB effects at the family level were only observed in PCB-exposed males wherein PCB exposure resulted in an increased abundance for Dehalobacteriaceae and decreased abundance for Bifidobacteriaceae and Leuconostocaceae compared to their sex-matched controls.

Effects of sex and PCB exposure on gut bacterial abundance at lower taxa levels

At the genus level, 57 bacteria were identified, and 23 significantly altered bacterial abundance were observed in the unexposed female versus male group while only 15 were observed in the PCB-exposed female versus male group (Supplemental Table S6). These observations indicate that differences in microbial composition were mainly observed due to changes in sex as compared to PCB exposure.

To gain further insight on the effect size and distinguishing profiles between experimental groups at the genus level, LEfSe analysis was performed and LDA scores generated (Fig. 6C). The genus Bacteroides showed the largest difference in the unexposed mice, with males having higher abundance for Bacteroidetes, in addition to Staphylococcus, Oscillospira, Anaerotruncus, Anaeroplasma, Solibacillus and Dorea. Although some of these observations were retained in PCB-exposed males versus females, other distinctive changes at the genus level include increased abundance for Adlercruetzia and Dehalobacterium in males, while PCB-exposed females showed higher abundance for Solibacillus and Bifidobacterium. Interestingly, PCB-exposed males also showed increased Ruminococcus, as well as Adlercruetzia and Dehalobacterium when compared to their sex-matched controls while PCB-exposed females showed increased abundance for Solibacillus and Sporosarcina in comparison to their sex-matched controls. Overall, LEfSe results showed distinguished bacterial enrichment profiles not only due to basal sex differences but also due to PCB exposure and warrant further investigations into the pathogenic and beneficial role of these enriched bacteria.

Sex-specific PCB-mediated alterations on ileal gene expression

To further explore and identify potential sex-specific PCB effects on gut properties, and how these effects could impact gut microbiome and vice versa, mRNA levels for genes encoding proteins involved in gut permeability and function were measured. Being an immunologically responsive component of the gut, in addition to its function in gut barrier maintenance, the ileum was chosen as a representative tissue for evaluation. Initial assessment of genes encoding gut barrier proteins including tight junction protein 1 (Tjp1), claudins (Cldn1,2) and occludin (Ocln) were carried out. While no significant sub-group differences were observed in Tjp1, Cldn1 and Cldn2 mRNA levels (Supplemental Fig. S7), significant PCB effects were noted for some genes. PCB-exposed males showed lower mRNA levels for Ocln (Fig. 7A) compared to their sex-matched controls. Likewise, PCB-exposed males also showed decreased expression for Reg3g, the gene encoding regenerating islet-derived protein 3 gamma, an anti-microbial peptide (Fig. 7B). Only a sex effect was observed for the mucin-encoding gene, Muc1 (Fig. 7C) while no differences were observed for Muc2 (Supplemental Fig. S7). Additionally, gene expression of healthy and functional mucosal markers, namely the cathelicidin antimicrobial peptide (Camp), trefoil factor 3 (Tff3) and fibroblast growth factor 15 (Fgf15) were assessed. While no sex or PCB effects were observed for Camp, a sex effect was observed for Tff3 (Supplemental Fig. S7). Further, unexposed females also exhibited higher Fgf15 mRNA levels vs. unexposed males (Fig. 7D). Finally, to evaluate xenobiotic receptor expression in the gut, ileal mRNA levels of Cyp1a1, an AHR target gene and PCB target was measured. As expected, PCB exposure resulted in Cyp1a1 gene induction, indicative of AHR activation (Fig. 7E). Surprisingly, sex differences were observed in ileal Cyp1a1 induction levels with PCB-exposed males demonstrating more robust induction than exposed females despite no differences in ileal gene expression levels for the AHR (Fig. 7F).

Fig. 7
figure 7

Sex and PCB effects on ileal gene expression. Ileal mRNA levels were measured for genes encoding permeability, gut mucosal, and other markers, namely (A) Occludin (Ocln), (B) regenerating islet-derived protein 3 gamma (Reg3g), (C) mucin 1 (Muc1), (D) fibroblast growth factor 15 (Fgf15), (E) cytochrome P450 1A1 (Cyp1a1) and (F) the aryl hydrocarbon receptor (Ahr). A list of p-values from the two-way ANOVA analysis is also provided. Values are mean ± SD, *p <0.05.

Full size image

Discussion

Emerging toxicological studies have documented sex differences in liver and metabolic phenotype with PCB exposures15. Epidemiological studies on historic cohorts such as the “Yucheng” population with accidental, high-dose exposures to PCBs, demonstrated that the female sex may be more susceptible to metabolic disease outcomes including diabetes42. Prenatal PCB exposures were also associated with significantly more obesogenic effects in female adolescents vs. males43. In contrast, exposure studies to dioxin-like PCBs using rodent models demonstrated that male rats and mice manifested greater susceptibility to disruption in glucose homeostasis while female rats were more predisposed to PCB126-mediated disruption in fatty acid metabolism and lipid homeostasis15,44. Overall, sex-specific outcomes with PCB exposures, while highly dependent on the dose, duration, and nature of exposure, reiterate the ability of these chemicals to interfere with normal metabolic and endocrine processes and interact with hepatic xenobiotic receptors as part of their mechanistic contributions to metabolic and liver toxicity.

Phenotypic observations previously reported from this study, which represent effects of both dioxin-like and non-dioxin-like PCBs, included enhanced steatosis, glucose intolerance, and inflammation in the female sex compared to their exposed male counterparts16 (Supplemental Fig. S8). While exhibiting decreased hepatic HNF4 alpha expression, exposed female mice also showed decreased hepatic mRNA and protein levels for phosphorylated epidermal growth factor receptor (EGFR) which was interpreted as one of the plausible mechanisms for the sex-specific hepatic and metabolic toxicity16. It has been hypothesized that inhibition of EGFR phosphorylation by PCBs can induce hepatic signaling disruption leading to alterations in normal energy metabolism22. Another attributable factor for the observed sex-specific effects was the pro-androgenic and anti-estrogenic effects of this particular PCB mixture in females that fueled their sensitivity to metabolic disruption16 given that estrogens and hepatic estrogen receptor (ER) alpha expression play critical roles in maintenance of insulin sensitivity45, lipid homeostasis46 and oxidative stress responses47. While these phenotyping findings shed light on the sex-specific toxicity profile elicited by this combination of PCBs, they also suggest other off-hepatic PCB effects including potential multi-organ interactions that could have ramifications on toxicant-associated steatotic liver disease (TASLD) development.

Findings from the current study which focused on how short-term exposures to the non-dioxin-like Aroclor 1260 plus dioxin-like PCB126 affect the hepatic protein landscape in male and female mice revealed that the massive differences in protein expression were largely driven by biological sex. Higher basal protein levels for FMOs including FMO3 and multiple P450 enzymes were observed in these young adult C57BL/6 female mice compared to their male counterparts. While higher basal mRNA levels for most P450 enzymes including Cyp2b10 and Cyp3a11 have been reported in this mouse strain48, the current data affirm those findings at the protein levels as well. Similarly, sex differences in hepatic expression of FMOs have also been reported49,50, and the current observations are in concordance with those reports. In terms of sex-specific PCB effects, distinct upregulated proteins identified in the male exposed group included xenobiotic detoxification enzymes, primarily CYP2C29, a CAR target51, CYP4A10, a PPAR alpha target52, and CYP2A5, a downstream target of NRF2 activation and the AHR53. Strikingly, CYP2A5 (CYP2A6 in humans) is also upregulated with alcohol consumption and recognized as a mediator in alcohol-associated liver disease54. Additionally, other proteins involved in processes other than xenobiotic detoxication were CYB5A, postulated to be involved in sterol biosynthesis55, and RBP4, a vitamin A carrier in hepatocytes. Increased circulating RBP4 levels have been associated with MASLD, but reports remain controversial56. In contrast, HTATIP2, an oxidoreductase, was the only upregulated protein in PCB-exposed females. While numerous cancer studies have reported HTATIP2 as a tumor suppressor, hepatic HTATIP2 has been implicated in lipid metabolism, particularly lipid droplet formation57,58. These novel findings shed light on alternative key proteins that may be impacted by PCBs, besides the well-studied xenobiotic receptor targets.

Although the short-term PCB exposure resulted in modest changes in hepatic protein expression, it also contributed to distinct sex-specific biological processes as a consequence of the altered hepatic proteome, albeit both sexes showing AHR signaling (CYP1A1, CYP1A2 upregulation) as an enriched pathway. Indeed, IPF analysis suggested that AHR and LXR activation were more pronounced responses in the female sex and calls to attention the likelihood of sex-specific AHR activation by exogenous ligands which may partially account for differences in toxicity outcomes, and susceptibility of female livers to changes in lipid metabolism with toxicant exposures. Furthermore, when compared to exposed males, PCB-exposed females also exhibited increased protein expression for CD36, accompanied by disease enrichment for “fatty liver” in this group. CD36 is a fatty acid translocating protein and shared AHR target gene. Upregulated CD36 expression by the classic AHR ligand, TCDD, have been partially credited for increased steatosis with high-fat dietary intake59,60. It should be noted however, that biological sex largely drove the differences in lipid processes as observed in the GO Process findings, while the PCB effects were modest. Further, “estradiol metabolism” was identified as the most enriched pathway in the PCB-exposed female group, possibly as a consequence of the augmented AHR activation. AHR-ER interactions have been confirmed through in vitro studies, and increased AHR activation is often associated with dampened ER signaling and reduced estrogen levels61. Although the current study did not record a significant decrease in estradiol (E2) circulatory levels in the PCB-exposed group, a nonsignificant trend was observed16.

In conformity with the proteomic results, metagenomic analyses also exemplified how biological sex majorly drives changes in microbial diversity. Our findings demonstrated that Actinobacteria and Tenericutes were more abundant in male mice while female mice have lower Bacteroidetes abundance. While these findings reaffirmed previous reports on sex differences in gut microbiota in experimental models62,63, they also supported evidence of sex differences in human gut microbiome studies wherein female participants displayed lower Bacteroidetes abundance64. PCB exposure also impacted microbial composition differently in males and females. One intriguing observation was the sex-specific impact of PCB exposure on bacterial abundance noted by the greater abundance of Bifidobacteriaceae, a key bacterial family involved in bile acid metabolism, in PCB-exposed females but not males. Basal sex differences in secondary bile acid levels have been reported with females exhibiting higher secondary bile acids than males65, and this may subsequently affect the abundance of bile acid metabolizing bacteria. Although bile acids are yet to be measured in the current study, our findings suggest that PCB exposure may potentially modify bile acid composition through gut microbiota changes specifically in females. Importantly, bile acid-mediated effects on gut microbiota have been postulated as a pathway in regulation of lipid metabolism genes and overall metabolic homeostasis65. Increased levels of Bifidobacterium, a bacterial genus strongly correlated with carbohydrate metabolism, butyrate production and host health benefits66, was also observed in PCB-exposed females but not males and appear contradictory when considering the observed phenotypic findings in the female sex. Even though basal sex differences may act as a driver for the increased Bifidobacterium in PCB-exposed females, PCB-mediated effects in carbohydrate metabolism observed in this group may also act as a probable driver stimulating growth and possibly metabolic activity of this particular genus67. Nonetheless, the current study only accounted for acute effects with exposure and longer-term exposures may not necessarily reflect these patterns. Another compelling sex-specific observation in the current study was the increased abundance of Dehalobacterium only in PCB-exposed males. This is fascinating given that the Dehalobacter genus constitutes of bacterial populations implicated exclusively in degradation of organohalides such as dichloromethane68,69. As organohalides, PCBs, especially higher molecular weight congeners, are also subjected to microbial degradation through de-chlorination under anaerobic conditions. In fact, utilization of PCB dechlorinating bacteria including Dehalococcoides sp. has been successfully embraced as a strategy in PCB remediation efforts on contaminated soil and water70. Thus, it can be postulated that the increased abundance in Dehalobacterium with PCB exposure may, in part, serve as a bacterial response to the presence of parent organohalides and/or their metabolites in the gut. One plausible explanation for the observed Dehalobacterium abundance only in males could be differences in PCB gut distribution and bioaccumulation between males and females. This is also supported by the observed sex-specific induction of ileal Cyp1a1 in males but not females, although this observation also suggests that AHR activation by PCBs may be tissue-specific contingent on sex. However, a better understanding as to why this microbial compositional change is observed only in males with PCB exposure remains to be elucidated.

The hepatic proteome-gut microbiome data, coupled with our previously reported phenotyping observations, offered a more nuanced outlook on PCBs’ contribution to metabolic dysfunction and TASLD/TASH outcomes. The findings suggest that, in addition to biological sex, organ-organ interactions should be considered when predicting toxicity outcomes with environmental exposures especially for persistent compounds such as PCBs that can impact multiple organs simultaneously yet have tissue-specific effects. Particularly, these toxicity effects may not be concerted and appear somewhat dissociative. Regardless, the current findings underscore major sex-specific mechanistic pathways implicated in these toxicity phenotypes including sensitivity of female mice to hepatic AHR activation and repercussions thereof on hepatic estrogen metabolism and overall energy homeostasis.

Despite the current study’s strength in identifying mechanistic pathways with toxicant exposures using omics platforms, it is not without limitations. Firstly, the use of a short-term study design may not necessarily capture human exposure paradigms which are often chronic and occur over time. The minimal protein alterations observed with PCBs could also be due to the short period of exposure, with some computational analyses driven by a modest number of DE proteins, specifically the FCON vs. FPCB group. Secondly, the study did not reckon in its analyses, how phases of the estrus cycle in female mice may have influenced the measured endpoints for this sex. Further, while investigating gut microbiota changes, the study did not extend towards functional analyses and bacterial metabolite profile. Significantly, the study only considered environmental factors albeit TASLD often requiring more than one “hit” to develop and progress4. The unequal sample sizes for the proteomics vs.16S RNA sequencing experiments also limited the ability to perform integrated bioinformatic analyses specific to each mouse. Lastly, PCB body burdens were not assessed in the current study. In reconciliation with these current study limitations, future studies will include performing longer-term exposures in the presence of additional factors such as hypercaloric diets and other relevant lifestyle factors to better translate our research findings to exposed and affected populations. Use of whole genome sequencing and expanding the analysis on gut microbiota profile to functional metagenomics will also provide an in-depth picture on PCB effects on the gut and consequent impact of those changes on metabolic homeostasis. Further, future multi-omic platforms will also incorporate uniform sample sizes for integrated multi-omic analyses.

Conclusions

In closing, the present study provided valuable information on sex differences in the hepatic proteome and gut microbiome in the routinely used C57BL/6 mouse strain. Importantly, the current findings highlighted how acute PCB exposures transform the liver protein landscape in female mice, which is currently unknown knowledge, and provided insights on prospective mechanisms steering the sex-specific TASLD endpoints observed. In particular, only modest observations were made with regards to sex-specific DE proteins with PCB exposures. Nonetheless, the current study presented a perspective on how off-liver PCB effects may also promote sex-specific TASLD/metabolic outcomes. Furthermore, the omics data also aid in hypothesis generation and fueling additional research questions such as the consequential impact of liver vs. intestinal AHR activation with PCB exposures and underlying sex differences; impact of PCBs on bile acid metabolism; sex differences in PCB distribution and bioaccumulation and how that consequentially modulates gut-liver toxicity; and lastly, how long-term PCB exposure can affect intestinal permeability and function.