Thioesterase induction by 2,3,7,8-tetrachlorodibenzo-p-dioxin results in a futile cycle that inhibits hepatic β-oxidation

2,3,7,8-Tetrachlorodibenzo-p-dioxin (TCDD), a persistent environmental contaminant, induces steatosis by increasing hepatic uptake of dietary and mobilized peripheral fats, inhibiting lipoprotein export, and repressing β-oxidation. In this study, the mechanism of β-oxidation inhibition was investigated by testing the hypothesis that TCDD dose-dependently repressed straight-chain fatty acid oxidation gene expression in mice following oral gavage every 4 days for 28 days. Untargeted metabolomic analysis revealed a dose-dependent decrease in hepatic acyl-CoA levels, while octenoyl-CoA and dicarboxylic acid levels increased. TCDD also dose-dependently repressed the hepatic gene expression associated with triacylglycerol and cholesterol ester hydrolysis, fatty acid binding proteins, fatty acid activation, and 3-ketoacyl-CoA thiolysis while inducing acyl-CoA hydrolysis. Moreover, octenoyl-CoA blocked the hydration of crotonyl-CoA suggesting short chain enoyl-CoA hydratase (ECHS1) activity was inhibited. Collectively, the integration of metabolomics and RNA-seq data suggested TCDD induced a futile cycle of fatty acid activation and acyl-CoA hydrolysis resulting in incomplete β-oxidation, and the accumulation octenoyl-CoA levels that inhibited the activity of short chain enoyl-CoA hydratase (ECHS1).


MS analysis of acyl-CoA species.
Untargeted metabolomics identified dose-dependent decreases in hepatic acyl-CoAs following oral gavage every 4 days for 28 days (Table 1). Octanoyl-, hexanoyl-, butyryland acetyl-CoA were annotated based on parent ion mass, isotope similarity, and theoretical fragmentation. Scores > 40 had features matching parent ion mass and isotope distribution and MSE fragmentation data matching in silico mass fragmentation, while most metabolite scores averaged ~ 35 based on parent ion mass and Table 1. Effect of TCDD on acyl-CoA levels. Acyl-CoA levels were assessed using untargeted liquid chromatography tandem mass spectrometry. Mice (n = 4-5) were orally gavaged every 4 days for 28 days with sesame oil vehicle or TCDD. Fold-changes were calculated for each treatment group relative to the vehicle control group. Bold font and asterisks (*) denote statistical significance (p ≤ 0.05) determined using a one-way ANOVA with a Dunnett's post-hoc analysis. Scores were determined by Progenesis with 60 being the maximum value and 0 being the minimum value. Scores ranging from 30 to 40 are based on mass error and isotope distribution similarity, while score > 40 are based on mass error, isotope distribution and fragmentation score. All annotated compounds have a score distribution averaging ~ 35.
Compound ID Description C n Score www.nature.com/scientificreports/ isotope distribution metrics. The presence of 426.1 m/z and 408.0 m/z coenzyme A fragment ions observed in the MS E fragmentation mass spectra further confirmed the acyl-CoA identifications for octanoyl-, hexanoyl-, butyrl-, and acetyl-CoA 39 . Hepatic levels of hexanoyl-, butyryl-, and acetyl-CoA were repressed 34.9-, 11.8-, and 6.3-fold at 30 μg/kg TCDD, respectively. Other β-oxidation metabolites also exhibited dose-dependent decreases including butenoyl-CoA, 3-hydroxyhexanoyl-CoA and 3-hydroxybutanoyl-CoA, while octenoyl-CoA was dose-dependently induced 138.9-fold (Table 1, Supplementary Table 1). These dose-dependent decreases in FA oxidation intermediates are consistent with β-oxidation inhibition 13,17 . TAG and CE hydrolysis. Figure 1 summarizes the dose-dependent effects of TCDD on lipases, carboxylesterases, and a deacylase associated with TAG and CE hydrolysis. Note that because gene expression can vary significantly throughout the day due to circadian regulation, all fold-changes discussed in the text were derived from the circadian gene expression dataset to highlight the TCDD effects when controlled for diurnal rhythm (unless otherwise indicated). Despite similar trends, there may be fold-change discrepancies between the dose response and diurnal rhythmicity studies since the former study was not controlled for sample time collection. Although TCDD induced gene expression of adipose triglyceride lipase (ATGL, Pnpla2) 4.3-fold, its paralog, Pnpla3, was repressed 37.7-fold. The requisite ATGL co-activator, CGI-58 (Abhydrolase Domain Containing 5, Abhd5) was also repressed 2.0-fold, as was G0s2 (10.9-fold), a potent ATGL inhibitor 40 . In addition, other ATGL regulators including Faf2 (-1.6-fold), Cidec (8.7-fold), Plin2 (2.6-fold) and Plin5 (-3.1-fold) exhibited differential expression 41 . Lysosomal lipase A (Lipa, 2.3-fold), which deacylates endocytosed low-density lipoprotein TAGs and CEs, was dose-dependently repressed. Similarly, hormone sensitive lipase (Lipe, 3.3-fold), and hepatic lipase C (Lipc, 12.0-fold), which deacylate extracellular low-density lipoprotein TAGs and CEs, were also dosedependently repressed. Highly expressed carboxylesterases, which serve important roles in lipid metabolism and VLDL assembly 42 , exhibited dose-dependent repression with the exception of Ces1b. Specifically, highly expressed Ces2a, Ces2e, Ces3a, and Ces3b were repressed 56.3-, 4.0-, 1,923-, and 1,333-fold, respectively. Ces mRNA, protein and/or enzymatic activity repression by TCDD and related compounds in mice and rats has been previously reported [43][44][45] . Arylamide deacetylase (Aadac), another highly expressed gene associated with hepatic TAG metabolism 40 was dose-dependently repressed 28.6-fold. Both CESs and AADAC are localized to the extracellular region or endoplasmic reticulum. Interestingly, ATGL and monoglyceride lipase (Mgll, 2.7-fold), which are primarily located in the nucleus, cytosol, and plasma membranes, were both induced. In addition to substrate preferences based on fatty acid composition and cellular location, hydrolyases channel released FAs to specific fates such as β-oxidation, membrane formation, VLDL assembly or PPAR activation 40,42 . Some lipases and carboxylesterases also exhibit cholesterol and retinyl ester hydrolysis activity.
The differential expression of genes associated with lipid hydrolysis was more pronounced at 28 days in mice treated with 30 μg/kg TCDD. Differential expression did not occur exclusively with genes exhibiting AHR enrichment. Although TCDD disrupted the diurnal rhythmicity of most lipid hydrolysis genes, this was not the case for highly expressed carboxylesterases which did not exhibit diurnal regulation. Ces genes are localized to a tandem cluster on chromosome 8 which may explain their collective repression by TCDD except for Ces1b 42 . Ces repression may also be due to the induction of proinflammatory cytokine signaling 43 . Overall, the effect of TCDD on gene expression suggests lipid hydrolysis was repressed. However, cellular lipase mRNA levels do not always correlate with enzyme activity due to extensive post-translational regulation 46 . Accordingly, hepatic levels of TAG, CEs, and bile acids were higher, as were free FAs suggesting esterification may be saturated 13,15 . FA and acyl-CoA binding proteins. We next examined the effect of TCDD on binding proteins that are important for lipid uptake, intra-/extra-cellular trafficking, and cytoprotection ( Fig. 2) 34,35 . In addition to channeling lipids to specific metabolic pathways, fatty acid binding proteins (FABPs), acyl-CoA binding protein (ACBP; aka diazepam binding inhibitor [DBI]), and sterol carrier protein 2 (SCP2) mitigate the toxicity, hydrolysis, and signaling potential of free acyl-CoAs. FABPs, DBI, and SCP2 also bind acyl-CoAs as well as other hydrophobic ligands including peroxisome proliferators, prostaglandins, bile acids, bilirubin, heme, fatty acid, and lipid metabolites. Highly expressed Fabp1, Dbi, and Scp2 encode for the majority of acyl-CoA buffering capacity and were repressed 5.9-, 7.1-, and 3.5-fold, respectively. Repression of the highly expressed DBI was consistent with gene expression patterns (Fig. 2E). Only Dbi exhibited an oscillating expression pattern that was abolished by TCDD (Fig. 2F). Despite AHR enrichment at 2 h in the presence of a putative DRE (pDRE), Fabp1, Dbi, and Scp2 exhibited minimal repression within 168 h of a single bolus gavage of 30 µg/kg TCDD. At 28 days, dose-dependent repression of Fabp1, Dbi, and Scp2 was observed with 10 and 30 µg/kg TCDD. Fabp2 and 5 were also repressed 4.5-and 1.9-fold, respectively. Despite the 6.6-, 14.0-, and 120.1-fold induction of Fabp4, 7, and 12, respectively, their induction would likely be insufficient to compensate for the loss of buffering capacity provided by FABP1, which showed expression levels ~ 125-fold higher than the other FABPs. FABP1, unlike other FABPs that only bind one FA, binds two FAs and accounts for 7-11% of the cytosolic protein in normal human liver 34 . Fabp1, 5, and 12 all had multiple AHR enrichment sites but only Fabp2 and 12 exhibited differential expression at 168 h. Acyl-CoA binding domain containing proteins 4 and 5 (Acbd4 and 5), which facilitate acyl-CoA transfer to peroxisomes, were also repressed 2.7-and 3.1-fold, respectively. Consequently, decreased binding protein levels may impair FA and acyl-CoA channeling to specific pathways while increasing the potential for toxicity, signaling, and membrane disruption, as well as acyl-CoA susceptibility to hydrolysis. FA activation. Twenty-five mouse acyl-CoA synthetases (ACSs) catalyze the irreversible activation of FAs for β-oxidation or lipid biosynthesis. Each exhibits different tissue expression, subcellular locations, and substrate preferences while also channeling substrates to different pathways 47 . Twenty-two ACS genes were detected in the mouse liver including short-(Acss1-2; carbons length (C) 2-4), medium-(Acsm1-5; C6-10), long-(Acsl1, www.nature.com/scientificreports/ 3-6; C12-18), and very long-(Slc27a1-6; C12-30) members. Of the 17 abundantly expressed ACS genes, 15 were repressed including highly expressed (> 10,000 counts) Acsm1 (-12.5-fold), Acsl1 (-9.1-fold), Acsl5 (-5.6-fold), Slc27a2 (-5.9-fold), and Slc27a5 (-16.9-fold) following TCDD treatment (Fig. 3). In general, repression was dosedependently observed within 168 h of a single bolus gavage of 30 µg/kg TCDD. Most ACS loci exhibited AHR  www.nature.com/scientificreports/ enrichment in the presence of a pDRE at 2 h and exhibited disrupted diurnal expression. Although ACS mRNA levels poorly correlate with protein and enzyme activity 48 , TCDD-elicited gene repression was consistent with decreases in ACSM3 and ACSL1 protein levels (Fig. 3F,H). Collectively, these data point to ACS repression by TCDD being a direct target of AHR activation.
Mitochondrial and peroxisomal transport. While SCFAs and MCFAs passively enter mitochondria, LCFAs and VLCFAS must be transported into the mitochondria 49 . At the outer mitochondrial membrane, carnitine palmitoyltransferase I (Cpt1a; -2.7-fold) replaces CoASH in the activated FA with carnitine for transport of the acyl-carnitine (Fig. 4). Carnitine/acylcarnitine translocase (Slc25a20 aka CACT; -1.6-fold), transports acyl-carnitine species across the inner mitochondrial membrane into the matrix where carnitine palmitoyltransferase 2 (Cpt2; no change in gene expression) reactivates the acyl-carnitine into acyl-CoA in preparation for β-oxidation. Cpt1a and Slc25a20, which both possess pDREs, were also repressed within 168 h with evidence of AHR enrichment at 2 h. ATP binding cassette subfamily member D (ABCD) 1 and 2, and to a lesser extent ABCD3, transport longand very long-chain acyl-CoAs into peroxisomes 50 . ABCD1 (Abcd1), which prefers saturated and unsaturated 18-22C acyl-CoA substrates, and ABCD3 (Abcd3), with a preference for bile acid conjugated FAs 51 , were both repressed 2.2-fold (Fig. 4). Abcd1 and 3 exhibited negligible differential expression in the time course study despite AHR enrichment with pDREs. In contrast, Abcd2 showed disrupted oscillating expression in the time course and dose response studies. Although Abcd2 prefers C22:6-and C24:6-CoAs 51 , the level of very long unsaturated FAs in TAGs and CEs was increased by TCDD 13 .

Dehydrogenation of activated acyl-CoA species. Despite the repression of lipid hydrolases and
decreased acyl-CoA binding capacity, Table 1 indicates ongoing β-oxidation in the presence of TCDD as evidenced by the presence of short-and medium-chain acyl-CoAs. The first step involves acyl-CoA dehydrogenation between C2 and C3 to produce trans-2-enoyl-CoA. In the mitochondria, this oxidation is catalyzed by acyl-CoA dehydrogenases (ACAD) which exhibit different tissue expression, subcellular locations, and substrate preferences. Highly expressed Acad11, which preferentially oxidizes saturated C22-CoA 52 , was repressed in the time course and dose response (3.8-fold at 30 µg/kg TCDD) studies with AHR enrichment at 2 h in the presence of a pDRE (Fig. 5). However, highly expressed Acadvl (induced 1.6-fold), which has overlapping substrate preferences, likely offsets the decreased expression of Acad11 while the effects on Acadm, which preferentially oxidizes medium-chain acyl-CoAs, were negligible. Overall, the effects of TCDD on mitochondrial acyl-CoA dehydrogenase gene expression were modest. In contrast, highly expressed acyl-CoA oxidase 1 (ACOX1), the rate-limiting step in peroxisomal β-oxidation, was repressed in the time course, dose response and circadian (3.9-fold at 30 µg/kg TCDD) studies in the presence of AHR enrichment at 2 h and pDREs (Fig. 5). Accordingly, ACOX1 protein levels were also repressed (Fig. 5F). The results suggests peroxisomal β-oxidation is impeded by TCDD, consistent with the accumulation of free and esterified very long-and long-chain FAs within TAGs, CEs, and phospholipids 13,15 . Hydration of trans-2-enoyl-CoA. The hydration of trans-2-enoyl-CoA to 3-hydroxyacyl-CoA is catalyzed by enoyl-CoA hydratases depending on the number of carbons in the substrate (Fig. 5). In the mitochondria, long chain enoyl-CoAs are hydrated by the HADHA subunit of the mitochondrial trifunctional protein (MTP). HADHA exhibits the highest activity for C12-16 enoyl-CoAs with virtually no activity towards C4 enoyl-CoAs 53 . The expression of Hadha was not affected by TCDD. In contrast, ECHS1, the short chain enoyl-CoA hydratase, prefers C4 enoyl-CoAs with diminishing activity towards C10 enoyl-CoAs 54 . Although Echs1 transcript levels was repressed 2.0-fold in the presence of AHR enrichment and pDREs, protein levels were induced (Fig. 5H). Furthermore, Fig. 5I shows that octenoyl-CoA inhibited the hydration of crotonyl-CoA, the preferred C4-enoyl substrate of ECHS1. A tenfold higher concentration of octenoyl-CoA inhibited the hydration of crotonyl-CoA > 90% providing compelling evidence that octenoyl-CoA can inhibit ECHS1 activity. TCDD does not directly inhibit ECHS1 enzymatic activity at concentrations previously reported to repress HepG2 cell proliferation and colony formation ( Supplementary Fig. S2) 55 .
In peroxisomes, Ehhadh, which encodes for the enoyl-CoA hydratase subunit of the liver bifunctional enzyme (L-BPE), was induced 7.7-fold in the absence of AHR enrichment and pDREs. After several cycles, the resulting peroxisomal medium-chain acyl-CoAs are transported to the mitochondria for further β-oxidation. Despite the expression of multiple acyl-CoA dehydrogenases that produce medium-chain enoyl-CoAs, there is only one enoyl-CoA hydratase, ECHS1, with a preference for medium and short-chain acyl-CoAs. Consequently, Echs1 repression may contribute to the dose-dependent accumulation of octenoyl-CoA (Table 1, Fig. 5).
Eci1 and 2, and Decr1 and 2 isomerize mono-and poly-unsaturated acyl-CoA cis double bonds to the 2-trans configuration, and the reduction of 2,4-dienoyl-CoAs to trans-3-enoyl CoA, respectively (Supplementary Fig. 1). Eci1, Eci2 and peroxisomal Decr2 were repressed by TCDD (2.1-, 2.2-, and 2.9-fold, respectfully) while highly expressed mitochondrial Decr1 was unaffected by TCDD. All four genes exhibited AHR enrichment in the presence of a pDRE. Without appropriate standards, it was not possible to determine the presence of cis-versus trans-enoyl-CoA or 2-versus 3-enoyl-CoA or the existence of polyunsaturated enoyl-CoA species. Collectively, the data suggests Δ 2,3 -octenoyl-CoA is present although other octenoyl-CoA species may also exist.
Acyl-CoA deactivation. Excess FAs can overload β-oxidation causing mitochondrial stress that reduces flux and depletes free CoASH needed for other metabolic pathways including the TCA cycle and FA metabolism. In response, acyl-CoA thioesterases (ACOTs) hydrolyze acyl-CoAs releasing the CoASH from activated FAs (Fig. 6) 56 . Therefore, we examined the effect of TCDD on cytosolic and mitochondrial thioesterase expression. Acot1 and Acot2, which exhibit a preference for long-chain acyl-CoAs, were induced 8.9-and 21.2-fold, respectively. ACOT2 is the primary mitochondrial thioesterase with little activity for < 10C acyl-CoAs. In contrast, Acot7 and Acot13, are enzymatically inhibited by CoASH, and were repressed 2.4-and 2.8-fold, respectively. Peroxisomal thioesterase Acot4, which prefers very long-and long-chain acyl-CoAs was induced 3.7-fold, while CoASH sensitive Acot8 that exhibits broad substrate preferences (C2-20 acyl-CoAs), was repressed 1.5-fold. Unlike the mitochondria, which catalyzes complete FA oxidation, peroxisomes only complete 2-5 β-oxidative cycles producing medium-chain acyl-CoAs that are hydrolyzed by ACOT3 (Acot3; induced 12.5-fold) and read- www.nature.com/scientificreports/ ily taken up by the mitochondria. Alternatively, they are transported to the mitochondria following conversion to medium-chain acylcarnitine by carnitine O-octanoyltransferase (Crot) (Fig. 7). Mitochondrial Acot9, which preferentially hydrolyzes medium-and short-chain acyl-CoAs, and is inhibited by CoASH, was also induced 6.0-fold (Fig. 6). Accumulating mitochondrial CoASH could then either be used to reactivate free FAs in the matrix or be exported to the cytosol via PMP34 (Slc25a17, no expression change) initiating a futile cycle (Fig. 7) 56 . Nudt7 which preferentially cleaves medium-chain acyl-CoAs to acyl-phosphopantetheines and 3' ,5'-ADP, was repressed 8.0-fold 57 . This futile cycle may be exacerbated by the 23.3-fold induction of Ucp2 which would not only facilitate the mitochondrial export of liberated FAs and FA peroxides, but dissipate the proton gradient and uncouple oxidative phosphorylation 58,59 . In summary, TCDD-elicited differential gene expression of thioesterases favors the hydrolysis of very long-and long-chain acyl-CoAs freeing CoASH required for other reactions while leaving medium-chain acyl-CoAs intact for peroxisomal and mitochondrial β-oxidation.

Discussion
In this study, TCDD is used as a prototypical AHR ligand and represents the cumulative burden of all AHR agonists. The dose range and regimen approached steady state levels while inducing full dose response curves for many genes in the absence of (i) necrosis or apoptosis, (ii) significant serum ALT increases, (iii) changes in food consumption and (iv) body weight loss > 15% 8,9 . To provide perspective, 30 µg/kg TCDD resulted in mouse hepatic tissue levels comparable to serum levels reported in Viktor Yushchenko following intentional poisoning, while 0.03-0.1 μg/kg resulted in serum levels comparable to the Seveso cohort of women following the 1976 chemical release accident. At 0.01 µg/kg, hepatic levels were comparable to control hepatic levels, and to dioxin-like compound levels in US, German, Spanish and British serum samples 8,64,65 . Consequently, the dose range and regimen are relevant to human exposures, and the elicited effects cannot be attributed to overt toxicity. Conservation of the AHR, and similarities in AHR-mediated dyslipidemia and metabolic disease between rodents and humans, suggest a common mechanism that may identify novel therapeutic interventions for NAFLD which currently has limited treatment options 66 .
TCDD dose-dependently induced steatosis in mice with marked increases in TAGs, CEs, phospholipids, and free FAs in the absence of acute toxicity 13,15 . This steatosis is attributed, in part, to the accumulation of dietary lipids and the inhibition of VLDL export 13,67 . Previous metabolomic analysis suggested TCDD disruptedd β-oxidation based increased palmitoylcarnitine 15 , with more targeted functional assays demonstrating repression of hepatic β-oxidation 13,17 , which was confirmed by dose-dependent decreases in shorter chain acyl-CoAs. This is consistent with reported decreases of acyl-CoAs by 2,3,7,8-tetrachlorodibenzo-p-furan (TCDF) in mice and in human liver Huh-7 cells following TCDD treatment 68,69 . We further investigated this inhibition by integrating metabolomics, ChIP-seq, and RNA-seq datasets to determine the effects of TCDD on FA oxidation, focusing on differential gene expression associated with lipid hydrolysis, FA activation, binding proteins, β-oxidation, and acyl-CoA hydrolysis.
RNA-seq analysis identified seven of nine lipid hydrolases repressed by TCDD with only Pnpla2 and Lpl showing induction. Yet, TCDD increased hepatic free FA levels which could serve as substrate for β-oxidation. The increase in FA levels is attributed not only to increased hepatic uptake of dietary FAs and mobilized peripheral lipids following CD36 induction, but also reduced ACS levels 13,15,17 . ACSs catalyze a two-step, ATP-dependent, reaction producing activated FAs required for β-oxidation to proceed. In addition, genes encoding the predominant FA binding proteins FABP1, DBI, and SCP2, which also bind acyl-CoAs, heme, bile acids, retinoids, and other hydrophobic ligands, were repressed by TCDD, therefore reducing cytosolic, peroxisomal, www.nature.com/scientificreports/ and mitochondrial acyl-CoA binding capacity and trafficking. Furthermore, FABPs, DBI, and SCP2 buffer the toxicity, block the signaling potential of free ligands, and protect acyl-CoAs from hydrolysis 34,70 . For example, long-chain acyl-CoAs are powerful disruptors of membrane bilayers, inhibit diverse enzyme activities, and alter ion channel function 71 . Consequently, decreases in ACS and binding proteins would impede FA metabolism by reducing activated FA levels, and increasing acyl-CoA susceptibility to hydrolysis. CoASH is an obligate cofactor for > 100 different metabolic reactions. It does not passively diffuse across membranes, and therefore distinct cytosolic, peroxisomal, and mitochondrial CoASH and acyl-CoA pools exist. Hereditary or acquired conditions involving CoA ester accumulation, often referred to as CoASH sequestration, toxicity, or redistribution (CASTOR) disease, lead to the accumulation of one or more acyl-CoA species that reduce free CoASH levels causing adverse effects 72 . Deficiencies in short-, medium-, long-, and very long-chain acyl-CoA dehydrogenase, trifunctional protein, and carnitine shuttle activities are associated with CASTOR disease 72 . This phenomena is also associated with the toxicity of valproic, salicylic, pivalic, phenylbutyric, and benzoic acids and does not solely arise from free CoASH depletion 72,73 .
In response to CoASH sequestration, thioesterases prevent acyl-CoA levels from reaching toxic levels 38 . TCDD dose-dependently induced cytosolic, peroxisomal, and mitochondrial thioesterases including Acot9 which prefers medium-chain acyl-CoAs, but is strongly inhibited by CoASH favoring octanoyl-CoA accumulation. Furthermore, Nudt7 which hydrolyzes free CoASH and acyl-CoAs to (acyl)phosphopantetheines and 3' ,5'-ADP, was repressed 57 . Nudt7 repression and thioesterase induction would increase free CoASH levels and de-activate FAs in an ATP-consuming futile cycle that inhibits β-oxidation. Collectively, (i) ACS repression and thioesterase induction would reduce activated FA levels required for β-oxidation, while (ii) decreased FABPs, DBI, and SCP2 binding capacity would increase acyl-CoA susceptibility to hydrolysis and compromise trafficking substrates to metabolic pathways. Moreover, Ucp2 induction would export liberated FAs from the matrix, uncoupling oxidative phosphorylation and further compromise energy production. The accumulation of TAGs and CEs containing long-chain FAs 13 , the dose-dependent decrease in acyl-CoA levels (Table 1), and the detection of palmitoylcarnitine in serum, all suggest TCDD impaired FA oxidation 15 . Increased palmitoyl (C16)-, tetradecanoyl (C14)-, and decanoyl (C10)-carnitine levels are also reported in factory workers with 29.49-765.35 pg TEQs/g lipid levels 74 . Elevated serum acylcarnitine levels are associated with NAFLD 75 , suggesting the effects of TCDD in mice may have relevance in humans.
An unexpected result was the dose-dependent increase in octenoyl-CoA. In the mitochondria, long-chain acyl-CoAs are oxidized by multiple dehydrogenases that were minimally affected by TCDD. The trifunctional protein (MTP), encoded by Hadha and Hadhb, carries out the enoyl-CoA hydratase, hydroxyacyl-CoA dehydrogenase, and 3-ketothiolase activities. Hadha expression was not affected by TCDD, Hadhb was modestly induced, and thiolases (Acaa1b, Acaa2) were repressed. Enoyl-, 3-hydroxy-and 3-keto-acyl-CoA intermediates are not substrates for thioesterases. Very long-and long-chain acyl-CoAs that escape thioesterase hydrolysis and complete several β-oxidation spirals would result in the production of medium-chain acyl-CoAs. TCDD induced thioesterases with a preference for very long-and long-chain acyl-CoAs. Acot9 which prefers medium-chain acyl-CoAs was also induced but is strongly inhibited by CoASH 38 . Peroxisomal MCFAs produced by peroxisomal β-oxidation would be ferried into the mitochondrial matrix via carnitine O-octanoyltransferase (Crot) which prefers medium-chain acyl-CoAs 36,57 , MCFA trafficking is not dependent on binding proteins and tend to be metabolized by hepatic mitochondrial β-oxidation following activation as opposed to being incorporated into TAGs, CEs or phospholipids 49 . We posit that accumulating peroxisomal and mitochondrial octanoyl-CoAs are oxidized to octenoyl-CoA by ACADM which was largely unaffected by TCDD. Accumulating octenoyl-CoA . (E) Circadian regulated genes are denoted with a "Y". An orange 'X' indicates abolished diurnal rhythm following oral gavage with 30 μg/kg TCDD every 4 days for 28 days. ZT indicates statistically significant (P1(t) > 0.8) time of maximum gene induction/repression. Counts represent the maximum number of raw aligned reads for any treatment group. Low counts (< 500 reads) are denoted in yellow with high counts (> 10,000) in pink. Differential expression with a posterior probability (P1(t)) > 0.80 is indicated with a black triangle in the top right tile corner. Protein subcellular locations were obtained from COMPARTMENTS and abbreviated as: cytosol (C), mitochondrion (M), mitochondrial outer membrane (OMM), mitochondrial inner membrane (IMM), nucleus (N), peroxisome (P), and plasma membrane (PM). (F/H) Capillary electrophoresis was used to assess ACOX1 and ECHS1 protein levels in total lysate prepared from liver samples harvested between ZT0-3 (n = 3). Bar graphs denote the mean ± SEM. Statistical significance (*p ≤ 0.05) was determined using a one-way ANOVA followed by Dunnett's post-hoc analysis. (G) Diurnal expression of Acox1 was assessed by RNA-seq (n = 3). Asterisks denotes a posterior probability (P1(t) > 0.8) within the same timepoint comparing vehicle to TCDD. Diurnal rhythmicity ( ‡) was determined using JTK_CYCLE for each treatment. Circadian data are plotted twice along the x-axis to better visualize the gene expression rhythmicity. (I) ECHS1 activity was assessed by monitoring the depletion of crotonyl-CoA which has an absorbance at 263 nm. Statistical significance (*p ≤ 0.05) was determined using a one-way ANOVA followed by Dunnett's post-hoc analysis. The heatmap was created using R (v4.0.4). Plots were created using GraphPad Prism (v8.4.3). The biochemical reaction was created using Adobe Illustrator (v25.2).  54,76 . This is somewhat analogous to ACADM deficient mice where octanoyl-CoA oxidation is not compensated by other acyl-CoA dehydrogenases 77 . Moreover, in addition to being a poor substrate for hydration, we showed octenoyl-CoA inhibited the hydration of crotonyl-CoA, the preferred substrate of ECHS1. When β-oxidation is hindered, FAs can undergo ω-hydroxylation, primarily by CYP4A10 and CYP4A14, with subsequent oxidation by poorly defined cytosolic alcohol and aldehyde dehydrogenases to produce DCAs 78 . In humans, ω-oxidation is a secondary pathway responsible for 5-10% of FA metabolism under normal conditions. During fasting, starvation, or when β-oxidation is impeded due to an inborn error of metabolism, ω-oxidation   www.nature.com/scientificreports/ is considered a rescue pathway producing DCAs that undergo peroxisomal β-oxidation to produce TCA cycle and gluconeogenesis intermediates 79 . Conversely, DCAs are also associated with oxidative stress, lipotoxicity and PPAR activation that induces β-oxidation 80 . However, TCDD repressed Ppara expression and inhibits PPARαmediated gene expression 81 , thus compromising β-oxidation induction by DCAs. Long-chain DCAs can be reactivated to the corresponding CoA ester and undergo chain shortening by peroxisomal β-oxidation to produce DCAs of varying length following hydrolysis by thioesterases 82 . For example, dodecanedioic acid (12C) can be oxidized to sebacic (C10) and suberic acids (C8) while undecanedioic acid (11C) is rarely detected since it is partially oxidized in peroxisomes before further mitochondrial metabolism. Interestingly, TCDD decreased serum azelaic (C9) acid levels due to the repression of CES3 that hydrolyzes monoesters to release azelaic acid 43 . By extrapolation, we surmise that decreasing DCA levels ( Table 2) at higher TCDD doses is inversely correlated with the repression of Ces genes that exhibit different DCA monoester preferences as reported for other carboxylesterase substrates 43,83 .
In summary (Fig. 9), we propose that β-oxidation inhibition by TCDD is due to the differential expression of genes peripheral to the spiral itself. More specifically, TCDD represses lipid hydrolysis, FA activation, and binding protein expression, while inducing thioesterases. However, most of the effects reported here are limited to gene expression and selected proteins, and do not consider post-translational modifications that can also regulate enzyme activity. Nevertheless, the data suggest TCDD induces a futile cycle of FA activation by ACSs and de-activation by thioesterases that inhibits complete oxidation of FAs, uncouples oxidative phosphorylation, and depletes ATP levels. Consequently, FAs underwent ω-oxidation producing PPAR ligands in an attempt to induce β-oxidation. However, Ppara and Fabp1, the encoded proteins, of which, deliver ligands to PPARα, were repressed thwarting efforts to induce β-oxidation. This would not only further deplete energy levels but also increase free FA and DCA levels that likely exacerbate TCDD toxicity. The inhibition of ECHS1 activity and accumulation of mitochondrial octenoyl-CoA may also contribute to toxicity by precipitating metabolic decompensation due to the exhaustion of free CoASH required by enzymes in other pathways. Moreover, octenoyl-CoA may itself be toxic by inappropriately inducing signaling, and/or possess lytic (detergent-like) properties that disrupt membranes. Although many of these effects are consistent with the development and progression of NAFLD, as well as AHR-mediated hepatotoxicity following activation by TCDD and related compounds, their relevance in human models warrants further investigation. This should include complementary metabolomic analyses of other fractions and compartments such as serum and urine for metabolites associated with impaired β-oxidation and NAFLD.

Methods
Animal treatment. Postnatal day 25 (PND25) male C57BL/6 mice weighing within 10% of each other were obtained from Charles River Laboratories (Kingston, NY). Mice were housed in Innovive Innocages (San Diego, CA) containing ALPHA-dri bedding (Shepherd Specialty Papers, Chicago, IL) in a 23 °C environment with 30-40% humidity and a 12 h/12 h light/dark cycle. Aquavive water (Innovive) and Harlan Teklad 22/5 Rodent Diet 8940 (Madison, WI) was provided ad libitum. TCDD (AccuStandard, New Haven, CT) was dissolved in acetone and diluted in sesame oil to a working stock. The acetone from the working stock was then evaporated off using nitrogen gas. On PND28, mice were orally gavaged at the start of the light cycle (zeitgeber [ZT] 0) with 0.1 ml sesame oil vehicle (Sigma-Aldrich, St. Louis, MO) or 0.01, 0.03, 0.1, 0.3, 1, 3, 10, and 30 μg/kg body weight TCDD every 4 days for 28 days for a total of 7 treatments. The first gavage was administered on day 0, while the final gavage was on day 24 of the 28-day study. This dosing regimen was selected to approach steady state levels given the 8-12 day half-life of TCDD in mice 84 . Comparable treatment has been used in previous studies 8,9,13,16,17,43 . Mice were fasted six hours prior to the end of the study. On day 28, vehicle-and TCDD-treated mice were weighed and euthanized. Liver samples were immediately flash frozen in liquid nitrogen and stored at − 80 °C until analysis. This study was conducted in accordance with relevant guidelines and regulations. All animal procedures were approved by the Michigan State University (MSU) Institutional Animal Care and Use Committee (IACUC; PROTO201800043) and meet the ARRIVE guidelines.
Liquid chromatography tandem mass spectrometry. Samples were extracted using methanol:water:chloroform as previously described with slight modifications 9 . Briefly, untargeted extractions were reconstituted with 400 µl tributylamine with no dilutions for analysis. Analysis was performed using a Xevo G2-XS QTOF attached to a Waters UPLC (Waters, Milford, MA) with negative-mode electrospray ionization run in MS E continuum mode. LC phases, gradient rates, and columns were used as previously published 9 . For untargeted acyl-CoA analysis, MS E continuum data was processed with Progenesis QI (Waters) to align features, peaks, deconvolute, and annotate metabolite peaks. Metabolite annotations were scored based on a mass error < 12 ppm to Human metabolome Database entries 85 , isotopic distribution similarity, and theoretical fragmentation comparisons to MS E high-energy mass spectra using the MetFrag option with each metric contributing a max of 20 points towards a max score of 60. Raw signals for each compound abundances were normalized to a correction factor calculated using a median and mean absolute deviation approach by Progenesis QI. Significance was determined by a one-way ANOVA adjusted for multiple comparisons with a Dunnett's post-hoc test.