Genetics-Based Approach to Identify Novel Genes Regulated by the Aryl Hydrocarbon Receptor in Mouse Liver

Abstract

The aryl hydrocarbon receptor (AHR) is a ligand-activated transcription factor in the Per-Arnt-Sim superfamily of environmental sensors that is linked to several metabolic diseases, including nonalcoholic fatty liver disease. Much remains unknown regarding the impact of genetic variation in AHR-driven disease, as past studies have focused on a small number of inbred strains. Recently, the presence of a wide range of interindividual variability amongst humans was reported in response to 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD), the prototypical ligand of the AHR. In this study, a panel of 14 diverse mouse strains was exposed to TCDD for 10 days to characterize the AHR-mediated response across genetic backgrounds. Responses to TCDD are heavily dependent on genetic background. Although mice carry 1 of 4 Ahr alleles known to impact the affinity to AHR-ligands, we observed significant intra-allelic variability suggesting the presence of novel genetic modifiers of AHR signaling. A regression-based approach was used to scan for genes regulated by the AHR and/or associated with TCDD-induced phenotypes. The approach identified 7 genes, 2 of which are novel, that are likely regulated by the AHR based on association with hepatic TCDD burden (p ≤ .05). Finally, we identified 1 gene, Dio1, which was associated with change in percent body fat across the diverse set of strains (p ≤ .05). Overall, the results in this study exemplify the power of genetics-based approaches in identifying novel genes that are putatively regulated by the AHR.

The aryl hydrocarbon receptor (AHR) is a ligand-activated transcription factor and is a member of the Per-Arnt-Sim superfamily of environmental sensing proteins. The AHR has been well-studied and found to be involved in a diverse set of physiological processes, including xenobiotic response, liver development, and immune regulation (Bunger et al., 2008; Gasiewicz et al., 2010; Lahvis et al., 2005; Moura-Alves et al., 2014; Quintana et al., 2008). The canonical AHR signaling pathway is well-defined. Prior to activation, the majority of the AHR pool is found in the cytoplasm where it is bound to a heat shock protein 90 homodimer, an AHR-interacting protein, p23, and c-Src (Carver et al., 1998; Heid et al., 2000; Meyer and Perdew, 1999). Upon ligand binding, the AHR translocates into the nucleus and heterodimerizes with the aryl hydrocarbon nuclear translocator (Abel and Haarmann-Stemmann, 2010; Sorg, 2014). This heterodimer binds to dioxin response elements on DNA leading to changes in gene expression (Dere et al., 2011; Swanson et al., 1995).

Exposure to AHR ligands, such as the persistent and pervasive environmental toxicant 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD), have been linked to many adverse health effects in humans, including chloracne, immunotoxicity, and metabolic syndrome (Dornbos et al., 2016; Leng et al., 2014; Marinković et al., 2010; Uemura et al., 2009; Warner et al., 2013). Much still remains unknown, however, regarding the mechanism behind how exposures to AHR ligands can drive such pleiotropic effects. Previous reports have identified that there is a wide range of human interindividual variability in response to AHR activation (Dornbos et al., 2016). Follow-up studies in mice have indicated the presence of similar levels of interstrain responses to TCDD (Dornbos et al., 2018). Moreover, while inbred mouse strains carry distinct Ahr alleles with known differences in sensitivity to ligand-induced effects, previous reports have indicated that interstrain differences in response to ligands can be leveraged to identify novel AHR-regulated genes (Dornbos et al., 2018, 2016; Smart and Daly, 2000).

The aim of this study was to identify novel genes that are regulated by AHR-mediated signaling in the liver. A panel of 14 mouse strains was exposed to either a vehicle control, 1 or 100 ng/kg body weight/day of TCDD for 10 consecutive days. RNA sequencing was performed to determine strain-specific differences in hepatic gene expression in the liver following TCDD exposure. Regression analysis was used to scan for genes whose expression associated with (1) hepatic TCDD accumulation, (2) TCDD-induced change in body weight, and (3) body fat percentage. Results indicate several novel genes that were associated with TCDD accumulation in the liver, including Nptx1 and Slc46a3, and a single gene, Dio1, that is associated with TCDD-induced changes in percent body fat. Such findings improve the understanding of AHR-biology and the link behind AHR activation and complex disease. Finally, the findings may, in the future, point to key genes in developing treatment for adverse AHR-mediated phenotypes.

MATERIALS AND METHODS

 

Mouse Panel Study

All animal handling was in accordance with Texas A&M University’s Institutional Animal Care and Use Committee. Fourteen mouse strains were used: (1) C57BL/6J, (2) A/J, (3) BALB/cJ, (4) FVB/NJ, (5) C3HeB/FeJ, (6) CBA/J, (7) DBA/1J, 8) NOD/ShiLtJ, (9) NZO/HlLtJ, (10) 129S1/SvlmJ, (11) BXD40, (12) BXD91, (13) BXD100, and (14) CC019/TauUNC. Animals were obtained from The Jackson Laboratory (Bar Harbor, Maine) and mated at 6–8 weeks of age. Females were checked daily for vaginal plugs. The focus of this article is on liver-related phenotypes; pregnancy-related endpoints will be reported in subsequent manuscripts. If a plug was present, mice were separated, weighed, and randomly placed into a TCDD dose group: vehicle control, 1, or 100 ng/kg/day. TCDD was administered to mice using peanut butter as the vehicle daily for 10 consecutive days. Maternal body weights and composition were recorded at gestation day 1 (D1; initial dose) and D11 (euthanasia). Body composition was performed using an EchoMRI-100H Body Composition Analyzer (EchoMRI). Both weights and body composition results were reported as mean ± SEM. On day 11 (D11), mice were anesthetized with avertin, euthanized via CO₂ asphyxiation, and tissues were snap frozen in liquid nitrogen. Pregnancy success was assessed by the presence of absorption sites or embryos within the uterus under a dissection microscope; only pregnant mice were included in downstream analysis. LabDiet 5058 chow and water were provided ad libitum during mating and throughout treatment. Mice were maintained under constant 12-h light/dark cycles, temperature (74°F ± 2°F set point), and humidity (average 51%: range 39–52%) during mating and pregnancy.

TCDD Burden Analysis

TCDD accumulation was measured in liver tissue via gas chromatography-mass spectrometry (GC/MS). TCDD was measured in the livers of all 14 strains of the mouse panel that were treated with 1 ng/kg/day (n = 3) or 100 ng/kg/day (n = 3) of TCDD. A subset of strains (n = 9) were randomly chosen to assess the level of TCDD in the vehicle control groups. Sample extraction and purification procedure were developed at the Dow Chemical Company (Midland, Michigan) based on U.S. Environmental Protection Agency method 8290. Briefly, 50 ml of 5% benzene in hexane solution, 10 ng of ¹³C-TCDD (Wellington, Lot no. MD0480912), and 30 ml of concentrated HCl (trace metal grade) were added to each sample in sequence. Lab control spike (LCS) sample, which contains 10 ng of native TCDD (Wellington Environmental, Lot no. 90STN1013) and 0.5 g of corn oil, and method blank (MB) sample (0.5 g of corn oil) were analyzed along with each set of samples. All samples were shaken for 1 h on a shaking bed, vented to release pressure, and then shaken overnight. After shaking, the organic layer was transferred to new bottle for subsequent sample purification (column clean-up). Acid/base silica column and alumina column were utilized for sample purification. The samples extracts were then purged to dryness and reconstituted with 20 µl of ¹³C-1,2,7,8-tetrachlorodibenzofuran (TCDF) (Wellington Environmental, Lot no. 020701), which was treated as the injection standard to account for the recovery of internal standard (¹³C-TCDD). All samples were quantified using either gas chromatograph/high efficiency triple-quadrupole system (GC/MS/MS, Agilent 7000 series) or low-resolution single-quadrupole GC/MS (HP 5973/6890), depending on the dosing level of TCDD. Both instruments were equipped with 30 m × 0.25 mm DB-5ms column and data were quantified via isotopic dilution approach using either chemstation (Agilent) or masshunter (Agilent) software. Results of all quality control samples (MB and LCS) have passed method criteria indicating satisfying analytical quality. The average level of TCDD in the vehicle control group was 2.8 ng/kg liver.

Phylogenetic Analysis

Genomic and amino acid sequences were predicted using the automated Mouse Gene and Protein Sequence predictor (Dornbos et al., 2020). Multiple sequence alignments were performed using Multiple Alignment using Fast Fourier Transform (MAFFT) algorithm (Katoh et al., 2005). Clustal-formatted alignment and phylogenetic tree outputs were created based on the output from MAFFT software. Phylogenetic trees were visualized with FigTree v1.4.2.

Heritability Analysis

Heritability was assessed as previously described (Dornbos et al., 2018). Briefly, a regression model was fit to estimate the proportion of variance in the dependent variable (ie, hepatic TCDD burden) that is not attributed to variance across independent variable replication (ie, intrastrain variability). The 95% CIs of the multiple R² value was calculated via bootstrapping (n = 1000) using the bias corrected and accelerated method implemented in boot library in R (Canty and Ripley, 2017).

RNA Isolation

Frozen liver (approximately 50 mg) was homogenized in 1 ml of TriZOL with chrome-steel beads using a Mixer Mill 300 (Life Sciences, Carlsbad, California) for 4 min. Following, total RNA was extracted per the manufacturer’s instructions with an additional 5:1 phenol:chloroform extraction step (Sigma Aldrich, St Louis, Missouri). RNA purity (260/280 ratio) and concentration (ng/µl) were assessed with a NanoDrop 1000 spectrophotometer. RNA quality was determined using an Agilent 2100 Bioanalyzer. All RNA samples had RNA integrity numbers ≥7.

RNA Sequencing

Sequencing

RNA samples from the 14 strains of mice treated with either vehicle control (n = 3) or 100 ng/kg/day (n = 3) were sequenced by Novogene (Sacramento, California). RNA quality was verified on an Agilent 2100 Bioanalyzer (Santa Clara, California) prior to sequencing. RNA was fragmented via sonication with a BioRuptor (Diagenode, Denville, New Jersey). Library preparation was performed using a NEBNext Ultra RNA kit (New England Biolabs, Ipswich, Massachusetts). An additional quality control step was performed to assess concentration, molarity and fragment size prior to sequencing. The samples were sequenced using the PE150 sequencing strategy with reagents from Illumina (San Diego, California).

Processing

Read pair processing and analysis was performed as previously described (Green et al., 2017). Read quality was assessed using FastQC v0.11.5 (www.bioinformatics.babraham.ac.uk/projects/fastqc/; last accessed March 18, 2021) and adapter trimming was performed using Trimmomatic v0.38 (Bolger et al., 2014). Strain-specific reference pseudogenomes (Build 37) and corresponding MOD files were downloaded from http://www.csbio.unc.edu/CCstatus/index.py?run=Pseudo (last accessed March 18, 2021) and used for alignment using Bowtie2 v2.3.2 (Langmead and Salzberg, 2012). Alignment of reads for BXD strains (BXD30, BXD91, and BXD100) whose pseudogenomes were not available was performed using the pseudogenomes of parental strains C57BL/6J and DBA/2J. Coordinates of aligned reads were converted to common mm9 reference genome coordinates using the Lapels package (https://github.com/shunping/lapels; last accessed March 18, 2021). Alignments made to 2 reference genomes (eg, BXD mice) were merged using the Suspenders package (https://github.com/holtjma/suspenders; last accessed March 18, 2021). Gene counts were determined using HTSeq-count v0.6.1 (Anders, 2010). Differential expression analysis was performed between treatment groups within strains using DESeq2 v3.8 (Love et al., 2014). Genes were considered differentially expressed when |fold-change| ≥ 1.5 and adjusted p-value (false discovery rate [FDR]) ≤ .05. Sequencing data is deposited in the gene expression omnibus (GEO; GSE167328).

DAVID Analysis

Functional annotation clustering of differentially expressed genes (DEGs) from individual mouse strains was assessed using the gene functional classification tool on the Database for Annotation, Visualization and Integrated Discovery (DAVID, v. 6.8; Huang et al., 2009a,b). Enrichment scores higher than 1.3 were considered significant.

Quantitative Real Time Polymerase Chain Reaction (qRT-PCR)

Total RNA (2 µg for liver homogenate) was converted to cDNA using oligo(dT) primers and GoScript reverse transcriptase (Promega, Madison, Wisconsin). SYBR Green Master Mix (Life Technologies) was used to analyze relative gene expression. Gene expression was normalized to the geometric mean of 3 housekeeping genes: Hprt, Actb, and Gusb. Primer sequences are listed in Supplementary Table 1. All PCR was performed using a QuantStudio 3 RT-PCR System (Thermo Fisher, Waltham, Massachusetts). The 2^−ΔΔct method was used to calculate fold changes and all values are relative to the mean of their respective control mice for each strain. In all cases, n = 3 for each treatment group per strain.

Linear Regression

Regression was used to analyze DEGs across the 14 strains (n = 932), meaning (1) at least 1 strain displayed |fold change| ≥1.5 and (2) at least 1 strain displayed significant change in expression of the gene as compared with vehicle (p < .05). Linear, least-squares regression was performed to scan for genes where expression associated with TCDD burden, body fat percentage, or weight. In all cases, Python version 2.7.10 was used during regression analysis (Van Rossum and Drake, 1995). p-values were adjusted for a 5% FDR using the Benjamini-Hochberg correction (Benjamini et al., 2009).

Statistical Analyses

With the exception of the regression analyses, all statistical analyses were performed using R version 3.0.2 (R Development Core Team, 2015). Histograms and q-q plots were used to assess distributions prior to statistical analyses. Outliers within dose-groups were assessed with a Grubbs’ test; significant outliers (p < .05) were removed prior to downstream analysis. Potential significant differences across dose groups and strains were calculated with a t test or analysis of variance (ANOVA) where appropriate.

RESULTS

 

AHR-Mediated Gene Expression Across Strains

To identify novel AHR-regulated genes or modifiers of AHR signaling, 14 strains of mice were dosed with TCDD (0, 1, or 100 ng/kg body weight), a potent activator the AHR-pathway. RNA-sequencing was used to assess global changes in gene expression in the liver of each strain that received the highest dose (ie, vehicle control vs. 100 ng of TCDD/kg of body weight/day for 10 days). The results indicated that, across the 14 strains, there were a total 1024 DEGs with fold changes ≥1.5 and adjusted p ≤ .05 as compared with the controls (Figure 1). The results were verified for several known AHR battery genes, such as Cyp1a1, Cyp1a2, and Cyp1b1, and there were not significant differences in expression when compared with qRT-PCR (Supplementary Table 2).

Figure 1.

Differentially expressed genes in each strain of mice across diverse genetic backgrounds of mice. The correlation matrix indicates the total and shared number of differentially expressed genes across 14 genetically diverse strains of mice. Genes are considered differentially expressed if |fold change| ≥ to 1.5 and Benjamini-Hochberg-corrected p ≤ .05. Darker red indicates greater degree of correlation.

Mice carry 1 of 4 alleles of the Ahr. These alleles (i.e. Ahr^b1, Ahr^b2, Ahr^b3, and Ahr^d) impact susceptibility to ligand-induced AHR-mediated signaling and toxicity (Poland and Glover, 1990; Poland et al., 1994). To determine if these different alleles could, in part, explain the variation observed between mouse strains, we sought to establish which alleles the mice carry and correlate that to expression patterns. We used an automated method to establish which Ahr allele was carried by each inbred strain (Dornbos et al., 2020). We identified 3 strains which carry the Ahr^b1 allele, 5 strains that carry Ahr^b2 allele, and 6 strains that carry the Ahr^d allele (Figure 2A). In comparing the average number of shared DEGs across Ahr alleles, 7 DEGs are found across Ahr^b1 mice, 3 DEGs across Ahr^b2 mice, and 1 DEG across Ahr^d mice. In focusing on AHR battery genes, 8 of the 14 strains had significant induction of Cyp1a1 and Cyp1a2 expression upon TCDD treatment, all of which carry the high affinity Ahr^b1 or Ahr^b2 alleles. Strains carrying Ahr^b1 alleles, on average, had the greatest induction of Cyp1a1 and Cyp1a2 as compared with controls (Supplementary Table 2).

Figure 2.

Genotype-dependent differences in accumulation of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) in the liver. AHR amino acid sequence alignment was used to identify the allele carried by each strain: (1) Ahr^b1 (gray), (2) Ahr^b2 (blue), and (3) Ahr^d (green) (A). Levels of TCDD in the liver for each strain were used to calculate the mean level of TCDD found to accumulate within each Ahr allelic category. Results were used to compare mean TCDD levels of TCDD accumulation across Ahr alleles (B); asterisk (*) indicates p < .05. Mean TCDD levels in each strain were also compared within Ahr allelic categories (C). Letters over bars indicate significant differences (p < .05) with the mean of: ^aNOD/ShiLtJ, ^bNZO/HlLtJ, ^cCC019, ^dA/J, ^eFVB/NJ; ^fBALB/cJ, and ^gC3HeB/FeJ (p < .05). In all cases, gray indicates an Ahr^b1 allele (very high affinity), blue indicates an Ahr^b2 allele (high affinity), and green indicates an Ahr^d allele (low affinity). Bars indicate means and error bars indicate SE.

A key observation was that differential expression was not concordant across strains carrying the same Ahr allele. For example, 2 strains, CBA/J and C3HeB/FeJ, which carry Ahr^b2 alleles, had respective 1031.12- and 680.29-fold changes in Cyp1a1 expression as compared with an average of 20.97 for the remaining 3 Ahr^b2 mice (Supplementary Table 2). Beyond the AHR battery genes, we observe a similar trend as differential expression was not concordant across strains carrying the same Ahr allele (Supplementary Tables 3–5). Furthermore, the number of DEGs was also variable across strains carrying the same allele. For Ahr^b1 mice, the BXD strains averaged 19 DEGs, whereas C57Bl/6J had 161 DEGs. Interestingly, these strains averaged more DEGs than their other parental strain, DBA/1J, which only had 4 DEGs. A similar trend was seen across strains carrying Ahr^b2 alleles, such as A/J, which have 85 DEGs, whereas FVB/nJ and Balb/cj each have <10 DEGs. Although the majority of strains carrying Ahr^d alleles have ≤8 DEGs, NOD/ShiLtJ had 595 DEGs, which, notably, is more than observed in the other strains carrying the Ahr^b1 allele. These results suggest that, while specific Ahr alleles impact susceptibility of a strain to AHR-mediated toxicity, there are likely other genetic factors that modulate the response as well.

The observed difference in the number of DEGs led us to hypothesize that there are differences in the physiological response to AHR ligands across differing genetic backgrounds. To test this, we assessed enrichment of genes across an array of phenotypes using DAVID. We observe that, across all 14 strains, 8 strains were found with clusters with significant enrichment. Of these enrichments, all were annotated with involvement in differing physiologic responses. Although several clusters were common across the differing Ahr alleles (eg, oxidation-reduction processes), other strains had unique clusters, including extracellular matrices for NOD/ShiLtJ (Ahr^d allele; score: 3.93) and ion channels to A/J (Ahr^b2 allele; score: 3.19; Supplementary Table 6). As such, we hypothesize that the differing genetic backgrounds also impact the overall physiological response to AHR ligands.

TCDD Accumulation in the Liver

Given that ligand-activated AHR-mediated gene expression is dose-dependent, we next sought to establish the burden of TCDD that is present in the livers of the differing strains. Using GC-MS, we observed profound variability in the level of hepatic TCDD accumulation across strains. Within the 1 ng/kg/day dose group, the mean level of TCDD was 22.7 ng/kg of liver with 95% CIs ranging from 6.0 to 39.4 ng/kg of liver (Figure 3). There were > 40-fold differences across strains in the 1 ng/kg/day group with the lowest levels in NZO/HlLtJ (2.8 ng/kg liver) and highest in DBA/1J (119.5 ng/kg liver; Figure 3). Within the 100 ng/kg/day dose group, the population mean level of TCDD was 1909.7 ng/kg of liver with the 95% CI ranging from 955.1 to 2864.2 ng/kg of liver (Figure 3). As compared with the 1 ng/kg/day mice, the mean level of TCDD accumulation is > 84-fold higher in mice that received 100 ng/kg/day, but with comparable intradose levels of variability across strains with > 30-fold differences in the mean hepatic TCDD levels. The strain with lowest level of TCDD was the NOD/ShiLtJ (159.2 ng/kg liver), whereas the highest levels were found in BXD91 (5286.7 ng/kg liver; Figure 3). The mean level of TCDD in the vehicle control group (ie, background exposure) was found to be 2.8 ng/kg of liver. In comparison with serum lipid adjusted toxic equivalent factors for TCDD and dioxin-like compounds previously reported, the levels of hepatic TCDD measured in this study are consistent with concentrations found in humans following exposure over a lifetime (Nault et al., 2016).

Figure 3.

Mean hepatic 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) levels in 14 genetically diverse mouse strains. Gas chromatography-mass spectrometry was used to assess the hepatic TCDD burden in mice (n = 3) treated with 1 or 100 ng/kg of TCDD for 10 consecutive days. Levels are reported as ng of TCDD per kg of liver (ng/kg). Bars indicate mean level of TCDD; error bars indicate standard error. Orange and blue boxes indicate the 95% CIs of population-level mean TCDD levels for the 1 and 100 ng/kg/day dose group, respectively. White-dotted lines within colored boxes indicate the population-level means for the respective dose group.

The large variability in TCDD burden across strains led us to investigate the degree to which TCDD accumulation across strains is due to difference in genetic background. Using an ANOVA, we observe that the hepatic accumulation of TCDD within the 100 ng/kg/day dose group results a multiple R² value of 0.94 with 95% CI spanning from 0.90 to 0.96 (Supplementary Table 7). These results suggest that, in a controlled environment, the level of TCDD accumulation is quite heritable as the differences in genetic background appear to drive an estimated 15 times more of the observed variance as compared with environmental factors.

On average, the level of TCDD sequestered in the liver is impacted by the Ahr allele carried by the mouse strain (Figure 2B). The mean level of TCDD in Ahr^b1 mice is significantly higher than the mean levels found in Ahr^b2 and Ahr^d mice (p < .05). Similarly, Ahr^b2 mice sequester more TCDD than Ahr^d mice (p < .05). Notably, allelic differences in TCDD burden were only present in mice treated with 100 ng/kg/day TCDD and not at the lower dose of TCDD. Although significant differences in the mean levels of TCDD across Ahr alleles in the 100 ng/kg/day dose group are present, we also observe intra-allelic variability (Figure 2C). DBA/1J and BXD40 were found to have significantly higher TCDD burdens than other Ahr^d mice such as NOD/ShiLtJ, NZO/HlLtJ, and CC019 (p < .05). Similarly, CBA/J had significantly higher TCDD burden than all the other Ahr^b2 mice (p < .05). C3HeB/FeJ were found to accumulate significantly higher levels of TCDD as compared with A/J (p < .05). Notably, in comparing individual strains across allelic categories, several Ahr^d allele mice, such as the 129S1/SvlmJ, DBA/1J, and BXD40, were found to accumulate higher levels of TCDD than Ahr^b2 mice, such as A/J, FVB/NJ, and BALB/cJ. These results suggest that, while the Ahr allele affects the sequestration of TCDD, other genomic factors likely also impact accumulation.

Pharmacodynamics of TCDD-Induced Hepatic Gene Expression Across Strains

The interstrain variability in gene expression and hepatic accumulation across strains led us to believe that there may be additional genes that could be regulated by AHR-mediated signaling. Linear regression was used to search for gene expression that is associated with TCDD burden. Results indicated that there were 7 genes with expression that is associated with TCDD burden in the liver (Figure 4). Of note were several “positive control” genes found to be significantly associated with TCDD levels that are amongst the AHR battery, such as Cyp1a1 (beta = 0.0022, adj-p = .0099) and Cyp1a2 (beta = 0.0006, adj-p = .0429), that are associated with hepatic TCDD burden (Stevens et al., 2009). Other genes associated with TCDD accumulation that have previous links to TCDD are Nfe2l2 which encodes NRF2 (beta = 0.0002, adj-p = .0223) and Pmm1 which encodes phosphomannomutase 1 (beta = 0.0003, adj-p = .0151; Boutros et al., 2008; Nault et al., 2016, 2018; Stevens et al., 2009). Of particular interest are 2 novel genes, Nptx1 and Slc46a3, which were also found to be associated with hepatic TCDD burden.

Figure 4: Gene expression associated with TCDD liver burden. The plots indicate genes with expression that had a significant association to TCDD liver burden for the mouse panel in the linear regression analysis (p<0.05). The beta, R2 and adjusted P-value determined by the regression analysis are listed with each gene: (A) Cyp1a1, (B), Nptx1, (C) Htatip2, (D) Slc46a3, (E) Pmm1, (F) Nfe2l2 and (G) Cyp1a2.

AHR-Mediated Changes in Body Measurements Across Strains

Although TCDD has been well-established to induced wasting in rodents, we observed that TCDD-induced changes in weight and body fat percentage were quite variable across strains over the dosing scheme of this study (Seefeld et al., 1984). The average change in weight percentage ranged from +1% to −6% (Figure 5A). Although 4 strains that carry the Ahr^b2 allele ranged from about 1 to −1% weight change, the FVB/nJ, which carries an Ahr^b2 allele, lost 6% of body weight. Of particular interest is that the change in weight for the FVB/nJ was greater than all strains carrying the Ahr^b1 allele (Figure 5A). The change in body fat percentage was also variable across strains, ranging from −7% to +5% (Figure 5B). Although the Ahr^b1 and Ahr^d allele mice ranged from +0.2 to −1.4 and +1.6 to −1.2, respectively, the change in percent body fat for Ahr^b2 allele mice was less concordant. For example, CBA/J mice gained 4.75% body fat, whereas C3HeB/FeJ mice lost 6.61% body fat, which is 9-fold higher as compared with the Ahr^b1 mice. Of note, the C3HeB/FeJ mice, while losing body fat, were observed to gain weight. It should be noted that though these changes are TCDD-induced, the results might be confounded by the pregnancy of the mice.

Figure 5.

Aryl hydrocarbon receptor-mediated changes in physiological parameters. The plot indicates changes in (A) body fat and (B) body weight from the start to the end of the experiment of the 2,3,7,8-tetrachlorodibenzo-p-dioxin mice compared with vehicle control are indicated for each strain. In all cases, n ≥ 3 mice.

The variability observed across strains led us to wonder whether we may identify novel genes associated with change in weight and/or fat percentage. Regression results indicated that the expression of the Dio1 gene, which encodes iodothyronine deiodinase 1, was associated with change in fat percentage (beta = −0.1723, adj-p = .0457; Figure 6). This gene is of particular interest given the expression profile in humans where it is primarily found in thyroid and liver, both tissues where Dio1 has been shown to be dysregulated in mice following exposure to TCDD (Boutros et al., 2009; GTEx Portal). Furthermore, Dio1 catalyzes the conversion of T4 to T3 in humans, which may suggest a mechanism behind the TCDD-induced changes in body composition (Agnihothri et al., 2014; Liu et al., 2017; Panicker et al., 2008). Finally, we did not identify gene expression that was associated with change in body weight.

Figure 6.

Association of Dio1 expression and change in percent body fat. The plot indicates negative association of Dio1 expression with the change in fat percentage in mice treated with 2,3,7,8-tetrachlorodibenzo-p-dioxin. The beta, R², and adjusted p-value are indicated on the plot.

DISCUSSION

Previous reports have suggested that interstrain differences in mice in response to AHR ligands can point to novel genes that may be regulated by the AHR (Chapman and Schiller, 1985; Dornbos et al., 2018). Discovering these genes may improve the understanding of the AHR’s role in complex disease and potentially lead to treatment options for adverse AHR-mediated phenotypes. Here, a genetically diverse panel of 14 strains was treated with TCDD to characterize novel genes regulated by AHR-mediated signaling in the liver, an important target organ of AHR-mediated adverse health outcomes.

This study showcases the impact of genetics on physiological response. Although previous reports have established that differing Ahr alleles have differing affinities for TCDD, we observe distinct intra-allelic differences across strains. An example is found amongst BXD strains, which are a cross of C57Bl/6J (Ahr^b1) and DBA/2J (Ahr^d) strains (Taylor et al., 1999). We observed that BXD100 and BXD91 strains, both of which inherited the Ahr^b1 allele from the C57Bl/6J, have TCDD burden levels similar to C57Bl/6J (approximately 4000 ng/kg; Figure 1C), but have far less DEGs (n = 14 and 23, respectively) as compared with C57BL/6J (n = 161, Figure 1). On the other hand, the BXD40 which inherited the Ahr^d allele from the DBA/2J have TCDD burden levels similar to a DBA/1J, a close genetic cousin of the DBA/2J (approximately 1600 ng/kg; Figure 2C), but are found to have more DEGs (n = 78) as compared with the DBA/1J (n = 4) as well as the Ahr^b1 carrying BXD100 and BXD91 (Table 1). Such results suggest that, while the Ahr allele is known to impact sensitivity to AHR ligands with C57Bl/6J and DBA/2J used as the flagship examples, other genetic factors impact the response (Greig et al., 1984; Poland and Glover, 1990).

Linear regression analysis was used to identify genes that were associated with TCDD liver burden. In total, 7 genes were identified: Cyp1a1, Cyp1a2, Nptx1, Slc46a3, Nfe2l2, Htatip2, and Pmm1. Although these genes have different functions, previous reports have suggested that they have been previously linked to liver disease (Nault et al., 2016, 2018; Wu et al., 2007; Zhao et al., 2019a,b). Of note are the 2 novel genes reported in this article: Nptx1 and Slc46a3. Both genes are reported to play a role in the progression of liver injury (Wu et al., 2007; Zhao et al., 2019a,b). In human hepatocellular carcinoma (HCC), expression of NPTX1 has been reported to be negatively associated with tumor size and metastasis (Zhao, et al., 2019b). NPTX1 has also been linked to the Ak strain transforming (AKT) pathway in HCC cells, which is impaired in cells that are AHR deficient (Wu et al., 2007; Zhao et al., 2019b). Given that the AHR activation is associated with the progression of liver injury and can lead to HCC, these results suggest a mechanism where an AHR antagonist may be a potential therapy for HCC. Similarly, expression of Slc46a3, which encodes a lysosomal transporter, can impact the severity of HCC tumor progression. Humans with low levels of SLC46A3 are reported to have more aggressive phenotypes and cells with overexpressed SLC46A3 inhibited the levels of migration and invasion compared with control cells (Zhao et al., 2019a). As such, the results in this study suggest that Slc46a3 may be a link to how AHR-mediated toxicity may ultimately lead to HCC. Further studies are required to characterize the role of Nptx1 and Slc46a3 in AHR-mediated toxicity in the liver.

Regression analysis also suggests that Dio1 is associated with the TCDD-induced change in fat percentage. Although this finding may provide some insight on the potential mechanism of TCDD-induced wasting, it may also impact our understanding of AHR-mediated liver injury. One study indicated that the expression of DIO1 in both rodents and humans with chronic liver injury and fibrosis was significantly dysregulated (Bohinc et al., 2014). Furthermore, hypothyroidism has been associated with nonalcoholic fatty liver disease due to thyroid hormone’s regulation of hepatic lipid metabolism (Sinha et al., 2018). As such, the findings in this study may indicate an AHR-mediated link between Dio1 expression and TCDD-induced liver injury.

In conclusion, this study showcases the power in using mouse genetic based approaches in toxicology (Dornbos and LaPres, 2018; Harrill and McAllister, 2017). The results indicate a wide-range in AHR-mediated responses across diverse genetic backgrounds. Furthermore, this study pointed out novel genes that are likely regulated by the AHR. This study provides insight into the impact of AHR-mediated signaling in liver damage and may suggest new therapeutic targets for liver-related disease.

SUPPLEMENTARY DATA

Supplementary data are available at Toxicological Sciences online.

FUNDING

National Institute of Environmental Health Sciences Superfund Basic Research Program (NIEHS SBRP P42ES4911). AgBioResearch at Michigan State University (to J.J.L. and T.Z.). National Institute of Environmental Health Sciences Training Grant at Michigan State University (T32 ES007255 to A.J. and P. D.). Financial assistance for GC/MS analysis was provided by the Dow Chemical Company.

DECLARATION OF CONFLICTING INTERESTS

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. Dr. Hui Lin works for Dow chemical Company and performed the quantitative analysis of hepatic TCDD levels but did not partipcipate in other aspects of the research or the conclusions drawn here.