Trifluoroacetate mediated reduction of atherosclerosis in mice

Abstract

Trifluoroacetate (TFA) has been assumed to be an innocuous counterion in countless synthetic bioactive peptides and a few FDA-approved therapeutics. Here we show that TFA is in fact bioactive, inducing peroxisome proliferation and causing significant metabolic effects in three strains of mice (wild-type C57BL/6 J, Ldlr^−/−, and Apoe^−/−) and cultured liver cells. In a high-fat diet induced Ldlr^−/− mouse model of atherosclerosis, TFA reduced the levels of plasma LDL cholesterol, triglycerides, and the development of atherosclerotic lesions. These physiological effects were observed with TFA alone, or with TFA present as a counterion of a variety of short, unrelated synthetic peptide sequences. Multiple lines of mechanistic evidence supported that TFA acts by binding to peroxisome proliferator activated receptor (PPAR)-α to induce peroxisome proliferation. These findings indicate the need to assess the contribution of TFA to phenotypic effects observed in some studies using synthetic peptides and, if these effects extend to humans, suggest potential therapeutic applications as well as broader health implications of widespread environmental TFA exposure.

Trifluoroacetate (TFA), long assumed to be an inert peptide counterion, is a bioactive modulator of PPARα signaling that induces peroxisome proliferation, alters lipid metabolism, and attenuates atherosclerosis in mice.

Introduction

Peptides are typically prepared and employed as trifluoroacetate (TFA) salts in preclinical stages of research. TFA is a reagent during solid-phase peptide synthesis (SPPS), and is a component in buffer systems typically used for reversed-phase high-performance liquid chromatography (HPLC) purification of peptides, yielding peptides as TFA salts following purification. Ion pairing between TFA and cationic moieties in peptides (Lys, Arg, His, free N-terminus) is persistent, requiring a dedicated ion exchange step to remove TFA from the purified compounds¹. A given peptide molecule will contain varying numbers of associated molecules of TFA, depending on the number of cationic moieties present in the sequence. TFA is a counterion in at least two FDA-approved peptide drugs, bivalirudin and corticorelin¹.

Although the presence of TFA counterions is often overlooked, some previous reports have hinted that TFA can affect certain biological pathways. For instance, at relatively high concentrations (millimolar range), TFA can activate the lactate receptor hydroxycarboxylic acid receptor 1 (HCA1)², free fatty acid receptor 2 (FFA2)³, ATP-sensitive potassium channels⁴, and TFA can act as an allosteric modulator of the glycine receptor⁵. Also, modestly different antibacterial activities were reported for alkylguanidino urea antibacterial compounds when the counterion was switched from TFA to Cl^–6. Trifluoroacetate was reported to inhibit proliferation of osteoblasts and chondrocytes in cell culture⁷. Similarly, calcium trifluoroacetate was reported to prevent angiogenesis, although the proposed mechanism was not related to TFA but rather to the alteration of cytosolic calcium levels and the prevention of calcium signaling by VEGF⁸.

In the course of our ongoing efforts to develop peptides to treat atherosclerosis^9,10, we were faced with a perplexing observation– several short, unrelated peptide sequences intended as negative controls all provoked a similar phenotypic effect in vivo (reductions in plasma cholesterol levels in a mouse model of atherosclerosis). This was especially unexpected given that the peptides were administered orally, and had been expected to undergo rapid digestion in vivo, which would preclude any biological effects. Eventually, we made the discovery that the observed reductions in plasma cholesterol were not directly related to the peptides themselves, but rather stemmed from TFA that was present with the peptides as counterions.

Here, we show that TFA is a bioactive modulator of peroxisome proliferator-activated receptor (PPAR)-α signaling that induces peroxisome proliferation, alters lipid metabolism, and attenuates atherosclerosis in mice. TFA can reduce the development of atherosclerosis in the two most common mouse models of atherosclerosis, Ldlr^−/− (B6.129S7-Ldlr^tm1Her/J) and Apoe^−/− mice (B6.129P2-Apoe^tm1Unc/J), and can also provoke mechanistically-related phenotypic effects in a wild-type C57BL/6J mouse model of metabolic disease. Based on transcriptomics analysis of tissue from TFA-treated animals, along with experiments both in vivo and in cell culture, we established that the mechanism involves TFA binding to PPAR-α to induce proliferation of peroxisomes. Because most synthetic peptides in preclinical stages of research contain TFA counterions, the discovery that TFA impacts cellular metabolism in vivo and in vitro means that TFA may have played a potentially confounding role in numerous biological studies in which synthetic peptides have been employed. Moreover, the biological effects of TFA reported here may be relevant to human health, although the effects of TFA on human systems require further study. TFA is highly persistent and environmental concentrations of TFA are increasing^11–15, resulting in higher (albeit still low) levels of TFA in humans relative to other polyfluoroalkyl substances (PFAS)¹⁶. TFA is also a major metabolite of common inhaled anesthetics, leading to high plasma TFA levels in human patients for days following anesthetic administration^17–19.

Results

Discovery that TFA reduces plasma cholesterol levels in Ldlr^−/− mice

During efforts to develop peptide-based treatments for atherosclerosis, we found that a 23-residue linear peptide caused a considerable reduction in total plasma cholesterol levels and atherosclerotic plaques in high-fat diet-fed Ldlr^−/− mice, even when administered orally and despite having undetectable plasma concentrations (lack of oral bioavailability)⁹. Considering that linear peptides made up of natural L-amino acids should undergo rapid digestion in the gut, we were left with the possibility that a shorter peptide fragment could be the active species acting on targets in the gut. To test this idea, a set of four shorter peptides derived from the 23-residue peptide were synthesized (Fig. 1a). The sequences of these peptides were based on results from in vitro pepsin digestions of the parent 23-residue peptide (Supplementary Fig. 2). These fragment peptides ranged from four to eight residues in length and, together, spanned the full length of the parent peptide sequence (Fig. 1a). We also synthesized two additional previously reported tetrapeptides unrelated to the 23-residue parent peptide, one of which (FREL) was reported to prevent the development of atherosclerosis in vivo, while the other of which (KERS) was reported to be ineffective²⁰. Each of these six short peptides was individually administered orally (ad libitum in the drinking water) to high-fat diet (HFD, 1.25% cholesterol, 15.8% fat, and no cholate)-fed Ldlr^−/− mice for two weeks at a 90 mg/kg/day dose (see Supplementary Fig. 1 for dosing schedules). Unexpectedly, every one of these peptides caused a statistically significant reduction in plasma cholesterol levels, ranging from 24 to 55% lower than that of vehicle-treated animals (Fig. 1b). The basis for this effect was perplexing. The peptides were too short to fold, were not expected to self-assemble, and lacked any sequence homology. We concluded that the only commonality among the short peptide fragments in question must be the presence of TFA counterions (present following standard HPLC purification as the counterion of cationic groups on the peptides). To test if the TFA counterions were somehow causing the observed phenotypic changes, one of the short peptides (hexamer LSALEE) was converted from the TFA salt to the chloride salt²¹; ¹⁹F-NMR confirmed the complete removal of TFA from the peptide (Supplementary Fig. 3). In contrast to the TFA salt of the peptide, the LSALEE peptide administered as the chloride salt completely lost its ability to lower plasma cholesterol levels (Fig. 1b). This finding implicated TFA, albeit indirectly, as the active species causing reduced plasma cholesterol levels.

Fig. 1. Peptide-TFA salts reduced plasma cholesterol levels in HFD-fed Ldlr^−/− mice, whereas a corresponding peptide-Cl^– salt did not.

a In vitro pepsin digestions of the parent 23-residue linear apoA-I mimetic peptide yielded a number of peptide fragments (blue bars). Based on the observed fragments, four short synthetic peptides were chosen for in vivo testing (red bars). b In vivo studies with TFA or Cl^–-salts of the short linear peptides led us to the discovery that removal of the TFA counterion abolished the observed plasma cholesterol reductions in vivo. Peptides were administered to female HFD-fed Ldlr^−/− mice ad libitum in the drinking water for 2 weeks (75 mg peptide/kg/day = 28 μmol peptide/kg/day for 23-residue parent peptide; 90 mg peptide/kg/day = ~130 μmol peptide/kg/day for shorter peptides). All peptides were present as the TFA salt, except LSALEE-Cl^–. Depending on the number of cationic moieties in the particular peptide sequence, the expected amount of TFA administered as peptide counterion ranged from 18 to 53 mg TFA/kg/day. Scatter plots are shown with mean ± SD. The shaded region indicates one standard deviation above and below the mean (dotted line) for the PBS vehicle group. Each datapoint represents one biological replicate for PBS vehicle (n = 23 mice), 23-residue parent peptide (n = 8), FREL (n = 4), KERS (n = 7), CGVLESF (n = 8), KASF (n = 6), EEWTKKLQ (n = 7), LSALEE-TFA (n = 9), LSALEE-HCl (n = 7). p values were determined by one-way ANOVA comparing each experimental group to the PBS vehicle control, post hoc Sidak’s multiple comparison test; ns, not significantly different.

TFA mediates dose-dependent cholesterol-lowering effects following oral or intraperitoneal administration

A series of dosing studies (see Supplementary Fig. 1 for dosing schedules) unambiguously established that TFA was responsible for the observed reductions in plasma total cholesterol. Dose-dependent reductions in cholesterol levels were observed following daily TFA administration for two weeks in the context of either i.p. injection or oral delivery using female or male Ldlr^−/− mice (Fig. 2a). Plasma cholesterol levels were not significantly affected by TFA treatment at 3 mg/kg/day, but began trending downward at doses of 9 mg/kg/day and higher. The overall magnitude of plasma cholesterol reduction was similar following 2-week treatment with the 27 mg/kg/day dose of TFA (in the range of 25–35% compared to vehicle-treated control animals), regardless of whether the TFA was administered ad libitum in the drinking water, by once-daily i.p. injection, or by once-daily oral gavage (Fig. 2a). Notably, the 27 mg/kg/day TFA dose was chosen because this dose was in the middle of the range of TFA doses that were present in the studies as counterion with the fragment peptides (depending on the number of cationic moieties in the particular peptide sequence, the expected amount of TFA administered as peptide counterion in those studies ranged from 18 to 53 mg/kg/day). Analysis of lipoprotein profiles in plasma of the mice indicated that the reductions in plasma total cholesterol caused by TFA treatment were due to TFA dose-dependent reductions in VLDL/LDL cholesterol, whereas HDL cholesterol levels were unchanged by TFA treatment (Supplementary Fig. 4). For sodium trifluoroacetate, a 27 mg/kg/day dose is equivalent to 200 μmol/kg/day. In contrast to TFA, matched 200 μmol/kg/day doses of acetate, included as a negative control, did not affect plasma cholesterol levels, whether administered by i.p. injection or orally in the drinking water (Fig. 2a). Altogether, these data indicated that TFA treatment produces dose-dependent lipid-lowering effects in male and female HFD-fed Ldlr^−/− mice regardless of the route of administration, which begin to be observable at doses around 9 mg/kg/day.

Fig. 2. TFA exerts dose-dependent cholesterol-lowering effects, has a long in vivo half-life, and reduces the development of atherosclerosis in HFD-fed Ldlr^−/− mice.

a Compared to PBS vehicle controls, sodium trifluoroacetate (TFA) reduced plasma total cholesterol levels of HFD-fed Ldlr^−/− mice in a dose-dependent fashion following 2-week administration by daily i.p. injection, daily oral, or 2-week administration ad libitum in the drinking. In studies comparing matched 200 μmol/kg/day doses of sodium TFA (=27 mg/kg/day) and sodium acetate (=17 mg/kg/day) as a negative control, acetate did not affect plasma cholesterol levels. Sample sizes for female i.p. study: n = 4 mice for each group; male i.p. study: PBS vehicle (n = 4), acetate (n = 4), TFA (n = 5); female oral gavage: PBS vehicle (n = 17), 9 mg/kg TFA (n = 7), 27 mg/kg TFA (n = 10), 81 mg/kg TFA (n = 7); male oral drinking water: PBS vehicle (n = 8), acetate (n = 9), TFA (n = 8). p values were determined by one-way ANOVA comparing each experimental group to the corresponding PBS vehicle control, post hoc Sidak’s multiple comparison test; ns, not significantly different. b Pharmacokinetic traces of TFA in plasma following oral gavage or i.p. administration of a single 27 mg/kg or 81 mg/kg dose to Ldlr^−/− mice. For sodium trifluoroacetate, a 27 mg/kg/day dose is equivalent to 200 μmol/kg. For each timepoint, samples were taken from n = 2 animals. The curves represent a fit of the data to a pharmacokinetic equation for single-dose plasma concentration. c TFA (27 mg/kg/day, n = 8) reduced the development of atherosclerotic lesions in HFD-fed female Ldlr^−/− mice compared to PBS vehicle (n = 7) following 10-week administration ad libitum in the drinking water. p values were determined by unpaired two-tailed Student’s t test. Representative images of aortic sinus cross-section (d) and aorta (e) are shown, with lipid stained red. All scatter plots in the figure are shown as mean ± SD. Each datapoint represents one biological replicate.

TFA exhibits a relatively long in vivo half-life

To measure the plasma concentration and half-life of TFA following i.p. or oral gavage administration to mice, a ¹⁹F-NMR-based assay was developed (Supplementary Fig. 5). Mice were given a single 27 mg/kg dose of TFA, and blood was drawn from the animals at various time points. The plasma was diluted with water and scanned by ¹⁹F-NMR along with a coaxial tube insert containing trifluoromethoxybenzene as an internal concentration standard. Based on this analysis, we found that plasma concentrations of TFA were very similar for either oral or i.p. routes of administration (similar exposure AUC) (Fig. 2b). For both routes of administration, the 27 mg/kg dose of TFA resulted in a maximal plasma TFA concentration of ~450 µM around 15 min post-administration, which slowly decreased to ~150 µM at 24 h post-administration (Fig. 2b). The observed long half-life in plasma is presumably due to TFA binding/ion exchange to plasma proteins. A similar single-dose pharmacokinetic study using a higher 81 mg/kg oral dose of TFA resulted in a maximal plasma concentration of 1.3 mM around 1 h post administration, with around 500 µM remaining in plasma at 24 h post-administration (Fig. 2b). To determine plasma concentrations of TFA over the course of a two-week daily dosing regimen, a 27 mg/kg dose of TFA was administered daily by oral gavage, and plasma samples were taken one hour after TFA gavage on day 3, day 7, and day 14 to estimate maximum plasma TFA concentrations; plasma samples were also taken 30 min before TFA gavage on day 14 (23.5 h after TFA gavage on day 13) to estimate minimum plasma TFA levels. Because successive doses were administered every 24 h, before complete elimination of TFA from the plasma, multi-day dosing led to a progressive accumulation in the observed maximum and minimum plasma concentrations (average C_max = 0.4 mM, 0.6 mM, 0.9 mM, and 1.0 mM on day 1, day 3, day 7, and day 14, respectively) (Supplementary Fig. 6). By the completion of the study, the average minimum plasma concentration was 0.6 mM, indicating that the plasma TFA level would have fluctuated between 0.6 mM and 1.0 mM using 27 mg/kg doses once a steady state was reached after the first few days of the typical multi-day studies.

Concentrations of TFA were also measured in liver tissue following two-week daily oral gavage of 27 mg/kg or 81 mg/kg doses of TFA, with livers being harvested between 25 and 29 h after the final administration of TFA. The observed concentrations of TFA in liver tissue were 0.5 mM and 0.8 mM for the 27 mg/kg and 81 mg/kg doses, respectively (Supplementary Fig. 6). Thus, for similar timepoints following multi-day 27 mg/kg TFA administration, the TFA concentration measured in liver (0.5 mM at ~25 h post-administration) were similar to that measured in plasma (0.6 mM at 23.5 h post administration), indicating that TFA does not accumulate in liver under the conditions studied.

TFA reduces the development of atherosclerosis in HFD-fed Ldlr^−/− mice

Given the key contributing role that high plasma LDL cholesterol levels play in the development of atherosclerosis, the observation of TFA-mediated reductions in plasma LDL cholesterol suggested that TFA might have the potential to reduce the development of atherosclerotic lesions in HFD-fed Ldlr^−/− mice. To assess this possibility, we carried out a 10-week atheroprotection study. Female mice were used for this study to allow for cage swapping (which is not possible with male mice) to facilitate balancing ages/weights and to ensure a more uniform initial gut microbiome between groups. The mice were fed a standard chow diet until ∼10 weeks of age and were then switched to the HFD along with commencement of daily TFA treatment (27 mg/kg/day) for 10 weeks (see Supplementary Fig. 1 for dosing schedules). As had been observed in the 2-week studies, plasma cholesterol levels were reduced by two weeks of TFA treatment (29% reduction; p = 0.00005, Supplementary Fig. 7), and remained lower in the TFA-treated mice at the 6-week (34% reduction, p = 0.0015, Supplementary Fig. 7) and 10-week timepoints (28% reduction, p = 0.0010, Supplementary Fig. 7). Following the 10-week study, we observed a reduction in the volumes of atherosclerotic lesions in the hearts of the mice (41% reduction) (Fig. 2c, 2d, Supplementary Fig. 8). Likewise, a reduced plaque area was observed in en face aortic sections (46% reduction) (Fig. 2c, 2e, Supplementary Fig. 9). Consistent with the reduced development of atherosclerosis, TFA treatment reduced the circulating levels of serum amyloid A (SAA), a marker of systemic inflammation (Supplementary Fig. 10). In a follow-up 8-week atheroprotection study, a lower dose of TFA (1.4 mg/kg/day) did not cause changes in plasma cholesterol or triglyceride levels over the course of the study (consistent with the results from the dosing studies), nor did it reduce the development of aortic lesions (Supplementary Fig. 11).

Overall, including various studies described herein to characterize TFA dose dependency, route of administration, mechanism of action, and TFA analogs (vide infra), we have replicated the cholesterol-lowering effect promoted by TFA in Ldlr^−/− mice in eight independent two-week vehicle-controlled trials using an oral route of administration, and three independent two-week vehicle-controlled trial using i.p. administration. These findings convincingly support that TFA, at concentrations that could be present with experimental peptides as counterion, can reduce plasma lipid levels and the development of atherosclerosis in a leading mouse model of the disease (Ldlr^−/− mice).

TFA upregulates genes involved in fatty acid metabolism and acyl-CoA biochemistry

To explore possible mechanistic underpinnings of the observed effects, RNA-seq analysis of tissues from the TFA-treated Ldlr^−/− mice was carried out. Few gene expression changes were observed in tissue from intestine (ileum) or fat pad, whereas analysis of liver tissue revealed more numerous dose-dependent changes in gene expression. Principal component analysis of the liver data indicated that TFA doses of 9 mg/kg, 27 mg/kg, and 81 mg/kg each had distinct gene expression profiles compared to the vehicle control and 3 mg/kg TFA groups (Fig. 3a). The changes in gene expression were time- and dose-dependent, with the highest 81 mg/kg TFA dose showing a more pronounced effect than the lower doses, and a 27 mg/kg dose showing a greater effect after 14-day treatment than a 3-day treatment (Fig. 3a). At the highest dose of TFA (81 mg/kg), gene expression was altered for 686 genes with an adjusted p-value < 0.05 (Supplementary Data 1). Pathway enrichment analysis of the gene expression data indicated that the pathways most upregulated by TFA treatment were related to the metabolism of fatty acids and proliferation of peroxisomes, which are organelles that carry out lipid and fatty acid metabolism²² (Fig. 3a, Supplementary Data 1). Many of the upregulated genes (e.g., enoyl-CoA hydratase and 3-hydroxyacyl CoA dehydrogenase, EHHADH; acyl-CoA oxidase 1, ACOX1) encode enzymes specific to peroxisomal beta-oxidation of fatty acids, while others encode for enzymes outside of the peroxisome (e.g., acyl-CoA thioesterase 1, ACOT1; acyl-CoA thioesterase 2, ACOT2) that are nonetheless target genes of the transcription factor peroxisome proliferator-activated receptor (PPAR)-α and are upregulated by peroxisome proliferating agents²². Numerous upregulated enzymes were involved in processing of various acyl-coenzyme A (CoA) species, which would be consistent with increased conversion of fatty acids to acyl-CoA upon peroxisome proliferation. However, we also considered that expression changes related to CoA biochemistry could also reflect the incorporation/participation of TFA as a structural analog of acetate in metabolic pathways.

Fig. 3. Gene and protein expression changes indicate that TFA stimulates peroxisome proliferation in Ldlr^−/− mice.

a Left, heatmap showing biological pathways with fold-enrichment >5 as determined by pathway enrichment analysis of gene expression data from liver tissue of Ldlr^−/− mice following three day or 2-week TFA treatment (i.p.). “Score” represents the agglomerated z-score, which reflects the standardized enrichment signal of a given pathway in all samples, incorporating directionality (activation or repression). A positive score indicates activation of the pathway, while a negative score indicates repression. Right, principal component analysis of the gene expression data. b Confocal microscopy at 60× magnification of liver cells from the TFA- or vehicle-treated Ldlr^−/− animals, stained magenta for PMP70, a marker of peroxisome proliferation. Magenta, AF568 stain for PMP70; blue, DAPI stain for nuclei; green, autofluorescence. The experiment was repeated with five biological replicates per group. c Western blot for PMP70 in liver tissue from biological replicates of treated Ldlr^−/− mice. Similar results were obtained for seven biological replicates per group.

Based on the biological pathways and enzymes implicated by the transcriptomics analysis, we experimentally tested two main mechanistic possibilities for the action of TFA. First, we examined if TFA treatment induced peroxisome proliferation, leading to shifts in lipid metabolism including enhanced fatty acid oxidation and decreased lipogenesis. Second, we explored if TFA impacted metabolism by generating trifluoroacetylated metabolites and/or proteins, possibly through the intermediacy of trifluoroacetyl-CoA (analog of acetyl-CoA).

TFA induces peroxisome proliferation in Ldlr^−/− mice

To experimentally establish if TFA was causing peroxisome proliferation in vivo as suggested by the RNA-seq analysis, we carried out a targeted study to look for known biomarkers of peroxisome proliferation. TFA was administered to Ldlr^−/− mice daily for a 2-week period, at concentrations ranging from 9 mg/kg/day to 81 mg/kg/day. The study included one arm involving mice fed HFD, as well as another arm involving mice fed a CHOW diet. Because our earlier studies had indicated that both male and female Ldlr^−/− mice were susceptible to the phenotypic effects caused by TFA, female animals were used for the HFD arm and male animals for the CHOW arm. As a positive control in the HFD arm, we used the drug fenofibrate (100 mg/kg), a clinically used fibrate drug that induces peroxisome proliferation by activating PPAR-α. Consistent with the expected effects of peroxisome proliferation, we observed increased levels of peroxisome membrane protein 70 (PMP70), a fatty acid transporter protein that is a marker for peroxisomes, using confocal microscopy (Fig. 3b, Supplementary Fig. 12) and Western blotting (Fig. 3c) of liver tissue from the treated animals. In both arms of the study, TFA caused dose-dependent increases in liver palmitoyl-CoA oxidase activity (Fig. 4a), which is associated with peroxisomal beta oxidation of fatty acids. We also observed increases in liver weight (Fig. 4c), consistent with the effects of peroxisome proliferation in rodents²³. TFA treatment led to reductions in plasma triglyceride levels of the mice fed HFD (Fig. 4b), whereas no changes were observed in plasma free fatty acid levels (Supplementary Fig. 13). In the mice fed a CHOW diet, no changes were observed in triglyceride or cholesterol levels (Supplementary Fig. 13). Targeted metabolomics analysis of bile acid levels in the feces and plasma of TFA-treated Ldlr^−/− mice revealed that bile acid levels were generally lower in the TFA-treated animals, although the changes did not reach the level of statistical significance in most cases (Supplementary Figs. 14, 15). The observation of reduced levels of bile acids in TFA-treated mice are consistent with previous reports that peroxisome proliferation causes altered/suppressed bile acid synthesis^24,25. In general, the phenotypic effects provoked by TFA were similar to those caused by fenofibrate, the positive control agent for induction of peroxisome proliferation.

Fig. 4. TFA induces palmitoyl-CoA oxidation, lowers plasma triglycerides, and exerts other phenotypic effects in mice.

a Liver tissue palmitoyl-CoA oxidase activity, b plasma triglyceride levels, and c liver weight in Ldlr^−/− mice following two weeks of daily oral gavage of TFA. Female animals were used for the HFD arm and male animals for the CHOW arm. Fenofibrate (100 mg/kg) in 0.5% methylcellulose (MC) vehicle was used as a positive control; this fibrate drug induces peroxisome proliferation via activation of PPAR-α. In panel (a), sample sizes for HFD treatments: PBS vehicle (n = 7 mice), 9 mg/kg TFA (n = 7), 27 mg/kg TFA (n = 10), 81 mg/kg TFA (n = 7), MC vehicle (n = 7), fenofibrate (n = 7). In panels (b) and (c), sample sizes for HFD treatments: PBS vehicle (n = 16), 9 mg/kg TFA (n = 7), 27 mg/kg TFA (n = 10), 81 mg/kg TFA (n = 7), MC vehicle (n = 7), fenofibrate (n = 7). In panels (a) and (c), sample sizes for CHOW treatments: n = 8 for all groups. d In Apoe^−/− mice, daily administration of TFA (27 mg/kg) in the drinking water for 10 weeks led to changes in liver tissue palmitoyl-CoA oxidase activity in the context of either a CHOW or HFD diet. In panel (d), sample sizes were: female/vehicle/CHOW (n = 7), female/TFA/CHOW (n = 9), female/vehicle/HFD and female/TFA/HFD (n = 10), male/vehicle/CHOW and male/TFA/CHOW (n = 8), male/vehicle/HFD (n = 9), male/TFA/HFD (n = 10). e TFA treatment led to reductions in atherosclerotic lesions in the aorta of male Apoe^−/− mice. Sample sizes were: female/PBS vehicle and female/TFA (n = 10), male/PBS vehicle (n = 9), male/TFA (n = 10). f In the male Apoe^−/− mice, TFA reduced levels of inflammation markers SAA and CRP. Sample sizes were: vehicle/CHOW/SAA (n = 5), vehicle/HFD/SAA (n = 6), TFA/HFD/SAA (n = 6), vehicle/CHOW/CRP (n = 6), vehicle/HFD/CRP (n = 7), TFA/HFD/CRP (n = 7). g In wild-type C57BL/6J male mice, daily oral gavage of 18 mg/kg TFA for five weeks led to changes in liver tissue palmitoyl-CoA oxidase activity (n = 10 mice for each group). h Reductions in plasma triglyceride and cholesterol levels (although not statistically significant in the case of cholesterol). Sample sizes for triglycerides and cholesterol were: PBS vehicle/CHOW (n = 10), PBS vehicle/HFD (n = 14), TFA/HFD (n = 10). i Altered gene expression in the liver for genes associated with peroxisome proliferation, namely ACOX1 and ACADS. Sample sizes were: HFD/PBS vehicle (n = 14), HFD/TFA (n = 10). In all scatter plots, each point represents one biological replicate and the data are shown as mean ± SD. In panels (a–h), p values were determined by one-way ANOVA comparing the experimental group to its corresponding vehicle group (demarcated by dotted lines), post hoc Sidak’s multiple comparison test; ns not significantly different. In panel (i), p values were determined by unpaired two-tailed Student’s t test. The TFA dose was 27 mg/kg/day unless otherwise noted, and animals were fed a HFD unless a CHOW diet is noted.

TFA treatment induces peroxisome proliferation in Apoe^−/− and wild-type C57BL/6J mice

Having verified that TFA-induced peroxisome proliferation in Ldlr^−/− mice, we carried out focused studies in two additional strains of mice, Apoe^−/− mice and wild-type C57BL/6J mice, to establish if TFA treatment would cause similar effects in these strains. Apoe^−/− mice have been widely used for studying the anti-atherogenic effects of drugs, including antihypertensives, LXR agonists, PPAR agonists, statins, etc.²⁶. Over the course of 10 weeks, Apoe^−/−mice were treated with a daily 27 mg/kg oral dose of TFA (see Supplementary Fig. 1 for dosing schedules). The study included arms involving male and female mice, fed a standard CHOW diet or a HFD that commenced at the beginning of the TFA treatment. We observed gene expression changes indicative of peroxisome proliferation in liver tissue from both male and female Apoe^−/− mice, after either three days or two weeks of TFA treatment (Supplementary Data 1). The expression changes in PPAR-α target genes²² that were similar to those observed in Ldlr^−/− mice (Supplementary Data 1), with pathway enrichment analysis indicating upregulation in metabolism of fatty acids and peroxisome proliferation. We also observed increased palmitoyl-CoA oxidation in liver tissue in both male and female Apoe^−/− mice (Fig. 4d). Unlike the Ldlr^−/− mice, we did not observe TFA-induced reductions of plasma cholesterol or triglyceride levels in Apoe^−/− animals (Supplementary Fig. 16). This was not surprising, because previous studies have reported that plasma lipid levels are often not reduced in Apoe^−/− mice by peroxisome proliferator fibrate drugs (or other agents such as apoA-I mimetic peptides or antibodies), even in cases where the treatment leads to increased or decreased development of atherosclerotic plaques^20,27–32. Indeed, analysis of atherosclerotic lesions in the Apoe^−/− animals indicated that TFA treatment caused a 26% reduction in the of atherosclerotic lesions volumes in the hearts (not statistically significant, Supplementary Figs. 17, 18) and a 30% reduction in plaque area in en face aortic sections (p = 0.02, Fig. 4e, Supplementary Fig. 19) in male mice. In the female Apoe^−/− mice, atherosclerotic lesions were not significantly impacted by TFA treatment (Fig. 4e, Supplementary Figs. 17, 20, 21). Consistent with the reduced development of atherosclerosis in the male Apoe^−/− mice, TFA reduced the circulating levels of SAA and C-reactive protein (CRP), markers of inflammation that correlate with atherosclerosis (Fig. 4f).

We next determined if TFA treatment caused peroxisome proliferation in wild-type C57BL/6J mice, using a diet-induced model of metabolic disease. Male mice were used for these studies because males are more susceptible to metabolic disease^33,34. The mice were fed a high-fat diet for six weeks, at which point daily oral TFA treatment (18 mg/kg/day) was commenced; TFA treatment and HFD feeding continued for five weeks, after which animals were sacrificed (see Supplementary Fig. 1 for dosing schedules). Like Ldlr^−/− and Apoe^−/− mice, TFA-treated wild-type C57BL/6J mice exhibited increased palmitoyl-CoA oxidation in liver tissue (Fig. 4g). TFA treatment also lowered plasma SAA (Supplementary Fig. 10), triglycerides (Fig. 4h), and cholesterol levels in the C57BL/6J mice (Fig. 4h), although the change was not statistically significant in the case of cholesterol. Expression of peroxisome proliferation-associated genes (e.g., ACOX1, ACADS) was increased by TFA treatment, as determined by qPCR (Fig. 4i).

Food intake did not significantly differ between vehicle control animals vs. TFA-treated animals in any of the studies for Ldlr^−/−, Apoe^−/−, or wild-type C57BL/6J mice (Supplementary Fig. 22), ruling out differences in cholesterol or fat uptake as a source of TFA-mediated phenotypic differences. Likewise, TFA treatment did not cause any differences in weight gain in Ldlr^−/−, Apoe^−/−, or wild-type C57BL/6J mice (Supplementary Fig. 23), nor did TFA cause changes in the plasma levels of liver enzymes alanine transaminase (ALT) or aspartate aminotransferase (AST) (Supplementary Fig. 23), which are standard markers of liver damage or disease. Taken together, these studies indicate that TFA induces peroxisome proliferation across several mouse strains (Ldlr^−/−, Apoe^−/−, and wild-type C57BL/6J mice), although the resulting phenotypic effects were somewhat strain dependent.

TFA induces peroxisome proliferation by binding to PPAR-α

To gain further insight into the mechanism by which TFA was inducing peroxisome proliferation, we established in vitro cell culture models. In rodent liver FAO cells, TFA treatment increased the expression of several peroxisome proliferation-associated genes (e.g., EHHADH, ACOX1, and ACAA1), as confirmed by qPCR (Fig. 5a). Western blotting of FAO cell lysates also confirmed that TFA treatment led to higher levels of EHHADH protein in the cells (Supplementary Fig. 24). Consistently, TFA treatment also increased palmitoyl-CoA oxidation activity, a functional assay for peroxisome proliferation (Supplementary Fig. 25). In contrast, neither gene expression changes by qPCR nor changes in palmitoyl-CoA oxidation activity were observed in TFA-treated human liver HepG2 cells, in line with previous findings that rodent liver cells are more sensitive to induction of peroxisome proliferation than human cells^31,32.

Fig. 5. TFA acts by inducing peroxisome proliferation, rather than generating trifluoroacetylated metabolites/proteins or altering the gut microbiome.

a Expression of several PPAR target genes in rat FAO liver cells was increased by TFA treatment, as indicated by qPCR. The effect was lost in PPAR-α knockout FAO cells. The data are shown as mean ± SD for technical triplicates. p values were determined by one-way ANOVA comparing each experimental group to the no treatment control, post hoc Dunnett’s test. The experiment was repeated twice independently with similar results. b Reporter assay in human HepG2 liver cells transfected with different mouse PPAR isoforms, in which luciferase expression is driven by activation of the PPAR response element. A different positive control was used for each PPAR isoform: 0.1 mM ciprofibrate for PPAR-α, 1 µM GW501516 for PPAR-δ/β, and 10 µM rosiglitazone for PPAR-γ. The data are shown as mean ± SD for technical triplicates. p values were determined by one-way ANOVA comparing each experimental group to the no treatment control, post hoc Dunnett’s test. The experiment was repeated twice independently with similar results. c A coactivator recruitment assay based on biolayer interferometry (BLI) was used to determine if TFA promotes binding of purified PPAR-α to a coactivator peptide. On the left are shown BLI association/dissociation traces for different putative ligands of PPAR-α. On the right is a plot of binding responses (maximum shift in the association phase) for palmitic acid and TFA, which had EC₅₀ values of 0.2 and of 4.9 mM, respectively. Acetate did not promote binding at the highest concentration tested of 10 mM. Data are shown as mean ± SD for technical triplicates. The experiment was repeated twice independently with similar results. d Data including RNA-seq, confocal microscopy, western blotting, metabolomics, BLI binding assays, and biochemical assays indicated that TFA induces peroxisome proliferation by activating peroxisome proliferator-activated receptor (PPAR)-α. e While TFA is structurally similar to acetate, we did not observe trifluoroacetylated metabolites or proteins, nor trifluoroacetyl-CoA, using a combination of metabolomics, proteomics, and biochemical assays. f TFA treatment did not alter the gut microbiome composition, as determined by 16S sequencing of fecal microbiome samples from PBS vehicle- or TFA-treated wild-type C57BL/6J mice (4-week treatment, 27 mg/kg/day TFA, n = 10–15/group). On the left is shown a Bray-Curtis principal component analysis of beta diversity between samples. On the right, plots of Chao1 index alpha diversity are shown as mean ± SD. Each datapoint represents one biological replicate; sample sizes were CHOW/PBS vehicle (n = 10), HFD/PBS vehicle (n = 15), HFD/TFA (n = 10). p values were determined by one-way ANOVA comparing the experimental group to PBS vehicle group, post hoc Sidak’s multiple comparison test; ns not significantly different.

We suspected that the actions of TFA were mediated by the transcription factor peroxisome proliferator-activated receptor (PPAR)-α, because most of the upregulated genes observed by RNA-seq were targets of PPAR-α and because TFA had caused similar phenotypic effects in mice as fenofibrate, which is a PPAR-α agonist (Fig. 4). To test this hypothesis, we used a reporter assay in human liver HepG2 cells transiently transfected with a PPAR response element-driven luciferase gene and a plasmid containing one of the three PPAR isoforms (PPAR-α, PPAR-δ/β, and PPAR-γ)^35,36. In the case of human (Supplementary Fig. 26) or mouse (Fig. 5b) PPAR-α, TFA concentrations of 1 mM or 10 mM significantly activated luciferase expression. Luciferase expression was activated by TFA treatment following 6-h, 24-h, or 48-h treatments (Supplementary Fig. 26). On the other hand, luciferase was not activated by TFA with human or mouse PPAR-δ/β or PPAR-γ isoforms (Fig. 5b, Supplementary Fig. 26), implicating PPAR-α as the isoform responsible for TFA-mediated peroxisome proliferation. To confirm this further, we knocked out the PPAR-α gene in FAO cells and found that TFA completely lost the ability to induce the expression of peroxisome proliferation-associated genes, such as EHHADH (Fig. 5a), further supporting that the observed effects of TFA were mediated by PPAR-α.

TFA did not cause significant changes in the expression level of PPAR-α as indicated by the RNA-seq data, ruling out changes in PPAR-α protein levels as the driver of TFA-induced peroxisome proliferation. To establish if TFA was causing peroxisome proliferation by directly binding to PPAR-α as a ligand, we carried out a PPAR-α coactivator recruitment assay using biolayer interferometry (BLI) (Fig. 5c, Supplementary Fig. 27). Binding of ligands to PPAR-α causes a conformational change that promotes association of the PPAR-α/ligand complex to various nuclear receptor coactivator proteins at the LXXLL motif present in the coactivator protein. Accordingly, a biotinylated LXXLL motif peptide corresponding to amino acids 135–153 of PPAR coactivator 1α (PGC-1α) was synthesized. The peptide was adhered to a streptavidin-coated BLI tip so that BLI could be used to assess whether the presence of putative ligands would promote the binding of purified PPAR-α in solution to the LXXLL motif on the tip. As positive controls, ciprofibrate and palmitic acid were found to promote PPAR-α binding with EC₅₀ values of 5.4 μM and 190 μM, respectively (Fig. 5c, Supplementary Fig. 27), close to previously reported values for these molecules in a coactivator recruitment assay³⁷. TFA clearly promoted binding of PPAR-α to the LXXLL peptide, albeit with a weaker EC₅₀ value of 4.9 mM (Fig. 5c). Trichloroacetate (TCA) also promoted binding, with an EC₅₀ value of 0.9 mM (Supplementary Fig. 27). On the other hand, sodium acetate did not promote any observable binding at the highest concentration tested (10 mM) (Fig. 5c). These results establish that TFA activates PPAR-α by directly binding it as a ligand to induce peroxisome proliferation (Fig. 5d), as opposed to some other indirect mechanism of activation.

Altogether, these experiments support that TFA activates PPAR-α in liver cells as a ligand, leading to changes in gene expression and enzyme activity associated with peroxisome proliferation.

No trifluoroacetylated metabolites or proteins were detected

Based on the observed gene expression changes in numerous enzymes involved in CoA biochemistry, and the structural similarity of TFA to acetate, we wondered if TFA might be impacting metabolism by generating trifluoroacetylated metabolites and/or proteins (analogs of acetylated proteins), possibly through the intermediacy of trifluoroacetyl-CoA (analog of acetyl-CoA). As a first step to test this hypothesis, we carried out an untargeted metabolomics analysis to identify metabolites that were significantly changed in concentration following TFA treatment. For the analysis, we compared plasma from female Ldlr^−/− mice treated by daily oral gavage for two weeks with a 27 mg/kg TFA daily dose or vehicle. The TFA treatment altered the concentrations of a few hundred plasma metabolites with a fold-change >1.5, p-value < 0.01, and maximum peak intensity >5000 (Supplementary Data 2). We were able to identify about 15 of these metabolites, most of which are phospholipids and fatty acids (Supplementary Data 2). As would be expected, TFA was observed in the plasma of treated mice, but not in the plasma of vehicle control animals. Interestingly, one of the identified metabolites was pantothenic acid, a component of coenzyme A (CoA), which increased about 2-fold in plasma concentration upon TFA treatment. We were unable to identify any metabolites consistent with the presence of a covalently attached trifluoroacetyl moiety, despite searching the mass fragmentation spectra for the presence of trifluoroacetyl fragment ions, and searching for pairs of parent ion masses that differed by the mass of a trifluoroacetyl group (Fig. 5e).

Biochemical assays were used to explore the potential formation of TFA-CoA in vitro. These studies generally followed previously established procedures, in which an organic acid of interest is incubated with rat liver microsomes or liver cell lysate in the presence of ATP and CoA to determine if the acid is converted into the corresponding acyl-CoA derivative^38,39. Whereas we clearly observed the formation of the acyl-CoA derivatives of acetate and ciprofibrate by LCMS in this assay, we did not observe any evidence of TFA-CoA formation (Supplementary Fig. 28). Likewise, incubation of purified acetyl CoA synthase protein from yeast with acetate and ATP led to the formation of acetyl-CoA, whereas no TFA-CoA was observed when acetate was replaced with TFA. Finally, coincubation of TFA along with acetate or ciprofibrate in the above assays did not affect the formation of either acetyl-CoA or ciprofibrate-CoA (Supplementary Fig. 28), suggesting that the presence of TFA did not inhibit or compete with the other acids as a substrate. Thus, there is no experimental support for the formation of TFA-CoA (Fig. 5e), although we cannot strictly rule out that TFA-CoA was generated and rapidly hydrolyzed.

Protein acylation, especially with short-chain acids such as acetate, can alter cellular phenotypes by modifying the epigenetic and cell signaling landscapes. To investigate if TFA treatment led to the formation of trifluoroacetylated proteins in vivo, analogous to protein acetylation, we carried out a quantitative proteomics study using liver tissue from mice that had been treated with TFA (27 mg/kg/day) or vehicle for two weeks. The liver samples were homogenized, labeled using isobaric tandem mass tags (TMT)⁴⁰, and analyzed by LCMS to identify any potential differences between the treatment conditions. We carried out both an open search of the data using MSFragger⁴¹ to identify unspecified mass differences between predicted and experimentally observed peptide spectra, and a closed search of the data using Proteome Discoverer with trifluoroacetylation of lysine as a potential modification. We observed no evidence of trifluoroacetylated proteins using either of these approaches (Fig. 5e), though other expected protein modifications were identified (Supplementary Data 3).

TFA treatment did not directly affect the gut microbiome in vitro or in vivo

A number of studies indicated that TFA was not operating through a mechanism involving direct effects on the gut microbiota. Using in vitro gut microbiota cultures of cecal material taken from HFD-fed Ldlr^−/− mice, we established that overnight treatment with TFA did not cause changes to the microbiome composition (Supplementary Fig. 29). In contrast, two positive control gut microbiome remodeling cyclic peptides⁴² shifted the in vitro microbiota community into distinct compositions as expected, regardless of whether the peptides were administered as the TFA or the Cl^– salt (Supplementary Fig. 29). In a separate in vitro study, the gut microbiota meta-transcriptomics profile of fresh gut microbiota cultures derived from HFD-fed Ldlr^−/− mice was unchanged following a 6-hour in vitro treatment with TFA, both for transcription activity at the organism or transcript function level (Supplementary Fig. 29). To test the potential effects of TFA on gut microbiota in vivo, fecal samples from TFA- or vehicle-treated Ldlr^−/− mice or wild-type C57BL/6J mice were analyzed using 16S sequencing. As expected, the mouse diet (CHOW vs HFD) led to clear differences in gut microbiome composition (Fig. 5f, Supplementary Fig. 30). On the other hand, TFA did not cause observable changes in alpha diversity (Chao1 or Shannon index) or beta diversity (Bray-Curtis dissimilarity) in either strain of mouse compared to the vehicle control group (Fig. 5f, Supplementary Fig. 30). We conclude that TFA does not directly impact gut microbiota. This conclusion is consistent with the observation that oral and i.p. routes of TFA administration both provoked similar phenotypic outcomes, even though the initial concentrations of TFA in the gut would likely be considerably higher following oral administration.

Structural analogs of TFA reduce plasma lipid levels in Ldlr^−/− mice

Intrigued by the observed lipid-lowering effects of TFA in vivo, we carried out a small exploration of the structure-activity space by investigating a number of TFA analogs in HFD-fed Ldlr^−/− mice (Fig. 6). The molecules were each administered orally at a 200 µmol/kg daily dose (equivalent dose to 27 mg/kg/day of sodium TFA) for two weeks, and included the small organic acids acetate, trichloroacetate (TCA), difluoroacetate, 2,2-difluoropropionate, and 2,2-difluorobutyrate. We also examined the 2,2,2-trifluoroacetamide and N-ethylacetamide derivatives of TFA, trifluoromethane-sulfonamide, and the carboxylic acid isosteres⁴³ 5-trifluoromethytetrazole and 3-(trifluoromethyl)-1,2,4-oxadiazol-5-ol. The sulfonamide caused acute toxicity in the mice and was discontinued before completion of the study; the remaining molecules were all well-tolerated. Neither TCA nor acetate affected plasma cholesterol levels in the mice (Fig. 6). Likewise, neither of the carboxylic acid isosteres 5-trifluoromethytetrazole or 3-(trifluoromethyl)-1,2,4-oxadiazol-5-ol reduced plasma cholesterol or triglyceride levels (Fig. 6, Supplementary Fig. 31). On the other hand, difluoroacetate and 2,2-difluorobutyrate brought about statistically significant reductions in cholesterol levels. Both of the TFA acetamides also reduced plasma cholesterol and triglyceride levels; however, follow-up studies revealed that these molecules were relatively quickly converted to TFA in vivo (half-life <1 h) upon oral administration (Supplementary Fig. 32). ¹⁹F-NMR analysis of plasma drawn one hour following oral gavage of 2,2,2-trifluoroacetamide showed a 1.8:1 ratio of TFA to trifluoroacetamide present in plasma, due to conversion of the administered trifluoroacetamide to TFA in vivo (Supplementary Fig. 32). Similarly, at the one-hour timepoint following oral gavage of N-ethyl-trifluoroacetamide, we observed a 3.8:1 ratio of TFA to 2,2,2-trifluoroacetamide present in plasma (none of the administered N-ethyl-trifluoroacetamide was observed), indicating that N-ethyl-trifluoroacetamide is also converted to TFA, possibly through the intermediacy of 2,2,2-trifluoroacetamide (Supplementary Fig. 33). Thus, the reductions in plasma lipid levels observed for the TFA acetamides are likely due to TFA itself. In contrast, ¹⁹F-NMR analyses of plasma from mice treated with difluoroacetate or 2,2-difluorobutyrate indicated that each of these analogs was unchanged and observable in plasma at the one-hour timepoint following oral gavage (Supplementary Fig. 34), suggesting that difluoroacetate and 2,2-difluorobutyrate directly cause cholesterol reductions in the mice.

Fig. 6. Several structural analogs of TFA reduce plasma cholesterol levels in HFD-fed Ldlr^−/− mice.

Each analog was administered orally for 2 weeks at 200 μmol/kg/day (equivalent dose to 27 mg/kg/day of sodium TFA). The reductions in plasma cholesterol levels observed for N-ethyl-trifluoroacetamide and 2,2,2-trifluoroacetamide are likely due to the effects of TFA itself, based on the observation of rapid conversion of these agents into TFA in vivo following oral administration. N-ethyl-trifluoroacetamide, 2,2,2-trifluoroacetamide, 5-trifluoromethyltetrazole, 3-(trifluoromethyl)-1,2,4-oxadiazol-5-ol, and trifluoromethane sulfonamide were administered by oral gavage, whereas the remainder of the TFA analogs were administered in the drinking water. The data for PBS vehicle and trifluoroacetate include groups treated by oral gavage and groups treated via drinking water. The data represent a combination of eight independent vehicle-controlled studies for TFA, two independent studies for 2,2,2-trifluoroacetamide, and a single study for the other chemicals. The data are shown as mean ± SD. Each datapoint represents one biological replicate. Sample sizes were: PBS vehicle (n = 80 mice), TFA (n = 68), N-ethyl-trifluoroacetamide (n = 9), 2,2,2-trifluoroacetamide (n = 15), difluoroacetate (n = 9), 2,2-difluorobutyrate (n = 9), 2,2-difluoropropionate (n = 8), trichloroacetate (n = 8), acetate (n = 9), 5-trifluoromethyltetrazole (n = 8), 3-(trifluoromethyl)−1,2,4-oxadiazol-5-ol (n = 8), trifluoromethane sulfonamide (n = 8). p values were determined by one-way ANOVA comparing the experimental group to PBS vehicle group, post hoc Dunnett’s test; ns not significantly different. The dotted line and gray shaded area correspond to the mean and ±one SD for PBS vehicle, respectively.

Discussion

Here we show that TFA impacts fatty acid and lipid metabolism in multiple biological contexts (Ldlr^−/−, Apoe^−/−, and wild-type C57BL/6J mice and cultured human HepG2 and rat FAO liver cells) through a mechanism involving promotion of peroxisome proliferation mediated by binding of PPAR-α. Given that TFA is a counterion in many peptides employed in early research and development settings, these findings raise the possibility that TFA may be confounding or contributing to phenotypic changes observed in some studies involving peptides. Indeed, it was in the context of control studies using short peptides that we first discovered that TFA reduced the levels of plasma cholesterol and triglycerides in high-fat-fed Ldlr^−/− mice. One issue raised by these findings is that the biological effects of TFA would not be removed altogether when using scrambled peptides as negative control compounds if the scrambled peptides also contained TFA counterions; it would be preferable to employ peptides and controls that do not contain TFA counterions.

Altogether, our data suggest that TFA-mediated induction of peroxisome proliferation is rather general across multiple mouse strains and cell lines tested, but the resulting phenotypic changes are context dependent. The observed effects induced by TFA differed somewhat among the three strains of mice examined, as well as between sexes. In all three strains of mice (as well as cultured rat liver cells), TFA treatment caused upregulation of PPAR-α response genes and increased palmitoyl-CoA oxidase activity, supporting the induction of peroxisome proliferation. Likewise, plasma SAA levels were reduced by TFA treatment in all three strains of mice; this biomarker of chronic inflammation is known to be repressed by activation of PPAR-α⁴⁴. The most dramatic effects on plasma lipid levels were observed in HFD-fed Ldlr^−/− mice, where significantly reduced plasma cholesterol and triglyceride levels were found in both male and female mice. TFA treatment also reduced the development of atherosclerotic lesions in female Ldlr^−/− mice (male Ldlr^−/− mice were not included in the atherosclerosis study). Triglyceride levels were significantly reduced in wild-type male C57BL/6J mice, whereas cholesterol levels were unchanged (only male C57BL/6J mice were studied in this model because females are less susceptible to metabolic disease^33,34). In contrast, plasma lipid levels were unchanged in HFD-fed male or female Apoe^−/− mice, but TFA treatment nonetheless reduced the development of atherosclerotic lesions in male Apoe^−/− mice. Sex-based differences are frequently observed in mouse atherosclerosis studies⁴⁵. Notably, sexual dimorphism has previously been observed with PPAR-γ agonists in mice, where the agents were protective against atherosclerosis in male animals, but not females (similar to the results reported here for Apoe^−/− mice)⁴⁶. Importantly, the observed sexual dimorphism and context-dependent differences in phenotypic effects could be especially problematic regarding the potential for TFA to cause artefactual effects in biological studies, because such effects could be easier to overlook in control studies.

While TFA-induced peroxisome proliferation and fatty acid oxidation pathways in both Ldlr^−/− and Apoe^−/− mice, its effects on circulating lipoproteins differed substantially between these models. This distinction is not unexpected, because the two models develop hyperlipidemia through fundamentally different mechanisms. Ldlr^−/− mice retain apoE-mediated remnant clearance pathways but have impaired LDL receptor-dependent uptake of apoB-containing lipoproteins, whereas Apoe−/− mice lack apoE entirely and therefore have more severe defects in remnant lipoprotein clearance. Therefore, interventions that decrease hepatic triglyceride synthesis, VLDL secretion, or fatty acid availability may produce larger reductions in circulating lipoproteins in Ldlr^−/− mice than in Apoe^−/− mice. Importantly, the observation that TFA activates peroxisomal pathways and reduces inflammatory markers in both strains supports the conclusion that its molecular mechanism is conserved, while the magnitude of the plasma lipid response depends on the underlying genetic defect governing lipoprotein metabolism.

There are conflicting data in the literature regarding the interplay of ApoE, peroxisome proliferation, lipid levels, and atherosclerosis. In one study, PPAR-α agonists (e.g., ciprofibrate and fenofibrate) were reported to aggravate hypercholesterolemia and accelerate development of atherosclerosis in Apoe^−/− mice by interfering with chylomicron remnant removal from circulation³¹. In a different study, fenofibrate reduced the development of atherosclerosis in male Apoe^−/− mice even though it did not reduce plasma lipid levels³². These latter findings are similar to those reported here, where plasma lipid levels were unchanged in Apoe^−/− mice, but TFA treatment reduced the development of atherosclerotic lesions in the male mice. In addition to effects on lipid levels, PPAR-α agonists exert known anti-inflammatory effects in vascular tissues^32,47. Thus, it is possible that the anti-atherosclerotic effects of TFA in Apoe^−/− mice (where lipid levels were not reduced) were due to anti-inflammatory properties stemming from peroxisome proliferation; the observed reductions in SAA and CRP mediated by TFA in male Apoe^−/− mice support this possibility. The effects of TFA on immune cell populations and inflammatory responses, including plaque macrophages and immune cells in blood and liver, remain to be investigated and may help explain strain-specific phenotypic differences.

Because TFA is present in peptide samples as a counterion of basic residues (Lys, Arg, His, or free N-terminus), the amount of TFA contained with a given peptide depends on the number of basic sites in the peptide sequence. Under the conditions of our studies, TFA doses of 27 mg/kg/day reproducibly reduced plasma cholesterol levels compared to vehicle in HFD-fed Ldlr^−/− mice, while doses of 9 mg/kg/day caused noticeable changes in gene expression and caused reductions in lipid levels that did not reach the degree of statistical significance. These TFA doses represent levels that could be present as peptide counterions, especially for longer or more cationic peptide sequences. Further studies are warranted to determine if the presence of TFA may be unexpectedly contributing to the outcomes of certain biological studies involving peptides. The phenotypic effects brought about by TFA may be especially likely to play a confounding or synergistic role in areas of research involving agents being developed to cause lipid lowering, anti-inflammatory effects, induction of peroxisome proliferation, or treatment of atherosclerosis.

To our knowledge, the only previous indication that TFA can reduce lipid levels in vivo is to be found in a 2014 patent⁴⁸, which claimed that orally administered calcium trifluoroacetate lowered plasma cholesterol levels in a male Gottingen minipig (n = 1, daily administration for four weeks), as did intravenously administered calcium trifluoroacetate in male or female beagle dogs (n = 2 or 4 animals/sex/group, daily administration for 2–4 weeks). The TFA doses reported in the patent ranged from 52 mg/kg/day up to 1020 mg//kg/day, which are higher than those used in the studies described here (typically, 27 mg/kg/day). The patent does not describe the diets of the treated animals, nor does it give any discussion of the relevant mechanism(s) of action. The data reported in the patent do not appear to have been published in a scientific journal, nor has the patent been previously cited. Nevertheless, this patent provides some support that TFA exerts effects on metabolism in vivo beyond rodent species, which were the focus of the studies here.

Earlier studies, generally carried out in cell culture, have indicated that longer chain perfluorinated acids (PFAS), such as perfluorobutyrate and perfluorooctanoate^36,49–51, and some halogenated acetates³⁵, can stimulate peroxisome proliferation. However, these previous studies, as well as toxicological studies of TFA^12,52, indicated that TFA only weakly stimulated peroxisome proliferation, if at all. In contrast, multiple lines of evidence from our studies (including gene expression changes, confocal microscopy, western blotting, metabolomics, biochemical assays, and PPAR-α binding studies) support that TFA can induce peroxisome proliferation by activating PPAR-α. This discrepancy could be due to a number of factors, such as differences in TFA dose, duration of treatment, or different biological systems under study. Also, the animals under study here typically consumed a high-fat diet, which may make the effects of TFA more pronounced because of the higher plasma lipid levels present.

The peak TFA plasma level following a single 27 mg/kg dose was ~0.4 mM, although the peak TFA level grew progressively with multi-day dosing to 0.6 mM, 0.9 mM, and 1 mM on day 3, day 7, and day 14, respectively. Our data indicate that plasma TFA levels would have fluctuated between 0.6 mM and 1.0 mM TFA for the 27 mg/kg dose, once a steady state was reached after the first few days of the typical multi-day studies. Higher steady-state levels would be expected for the 81 mg/kg dose; with a single 81 mg/kg TFA dose, the peak plasma concentration was 1.3 mM, with 0.5 mM TFA remaining in plasma at 24 h. We also measured TFA levels in liver, which is the organ most responsible for and most responsive to peroxisome proliferation. Liver TFA concentrations were similar to plasma TFA concentration following two week daily oral gavage, indicating that TFA does not accumulate in liver under the conditions studied, but (like plasma) liver concentrations of TFA would have fluctuated between around 0.6 mM and 1.0 mM TFA for the 27 mg/kg dose once a steady state TFA level was reached. Several lines of evidence support that these plasma concentrations of TFA are sufficient to activate PPAR-α in vivo, albeit sub-maximally, given the observed EC₅₀ of 4.9 mM for TFA in the coactivator recruitment assay. First, the BLI curves in Fig. 5c indicate that TFA concentrations as low as 0.5 mM TFA activated PPAR-α in the coactivator recruitment assay relative to the background observed for no ligand or acetate negative controls. Second, the cell culture studies shown in Fig. 5 and Supplementary Fig. 24 indicate that TFA treatments of 1 mM significantly increased gene expression of PPAR response genes, activated the PPAR-α luciferase reporter assay, and increased protein levels of EHHADH. Third, the gene expression data shown in Fig. 3a revealed that the 27 mg/kg TFA dose had almost no effects at the earlier day 3 timepoint, whereas this dose produced a modest but clear effect on gene expression at day 14, consistent with the progressive increase in plasma TFA levels observed in the multi-day PK study for this TFA dose.

As a strong organic acid, TFA exists almost completely in the anionic form at physiological pH values, raising the question of how TFA would cross cellular membranes to gain entry to the nucleus to activate PPAR-α. Previous studies in erythrocytes have found that TFA is actively transported across cellular membranes by both monocarboxylate transporters and anion transporters, with intracellular TFA concentrations equal to or exceeding extracellular TFA concentrations within around 10 min^53,54. While simple diffusion of TFA across bilayers was originally proposed as an additional pathway into cells⁵³, a later study reported that simple diffusion of TFA was negligible⁵⁴.

While our findings implicate a mechanism involving peroxisome proliferation for TFA-induced lipid perturbations, it should be noted that other biological pathways may also play a role. For example, GPR81 and GPR109a are G-protein coupled receptors for lactate and butyrate, respectively, but they also bind other small organic salts, including TFA^2,55,56. TFA binds to GPR81 with similar affinity as lactate (~6 mM EC₅₀), albeit with only ~60% of the maximum response stimulated by lactate²; TFA can also bind to GPR109a at high (>10 mM) concentrations². Activation of these receptors inhibits the breakdown of fatty acids in adipose tissue, resulting in reduced transport of fatty acids to the liver⁵⁷. The beneficial effects of the drug niacin, including increased HDL levels, lowered LDL, and lowered triglycerides are due to activation of GPR109a⁵⁶. Additionally, GPR81 is involved in regulating immune tolerance and suppressing inflammation⁵⁸. Thus, the known phenotypic effects associated with activation of GPR81 and/or GPR109a would be consistent with some of those we observed for TFA-treated mice.

We also considered the possibility that the effects of TFA were due to the formation of a transient trifluoroacetyl-CoA intermediate that perturbed metabolic pathways in vivo. This hypothesis was based on the structural similarity of TFA to acetate, the previous observations that some exogenous organic acids (including fibrate drugs) are converted to the corresponding acyl-CoA derivatives in rat microsomes^38,39, and the changes we observed in gene expression for numerous enzymes involved in CoA processing (Supplementary Data 1). In the case of TFA, our experiments do not support the formation of TFA-CoA species or other covalent trifluoroacetylated metabolites or proteins. We could not detect evidence of small molecule metabolites containing trifluoroacetyl groups using untargeted metabolomics, nor did we detect any signs of trifluoroacetylated proteins using a proteomics approach, nor did we detect the formation of TFA-CoA with liver microsomes or purified acetyl-CoA synthetase. Thus, while we cannot rule out the possibility that TFA-CoA species are generated in vivo and hydrolyzed/metabolized too rapidly to detect, it appears more likely that the altered gene expression observed for numerous enzymes involved in CoA biochemistry is a reflection of their involvement during peroxisome proliferation in processing acyl-CoA intermediates in fatty acid metabolism (Fig. 5)^50,59. Trifluoroacetylated proteins are observed in patients following treatment with certain inhaled anesthetics such as halothane, although these proteins are believed to arise not from TFA but rather trifluoroacetyl chloride produced via oxidative metabolism of the anesthetic^17,60.

In addition to TFA, we found that difluoracetate and 2,2-difluorobutyrate also significantly reduced plasma cholesterol levels in Ldlr^−/− mice, while several other TFA analogs did not. Trichloroacetate (TCA) did not affect cholesterol levels in the mice, which was somewhat unexpected because TCA is reported to more effectively induce peroxisome proliferation than TFA in cell culture³⁵ and TCA was more effective that TFA in promoting PPAR-α activation in our coactivator recruitment assay. The lack of in vivo efficacy for TCA could be due to poorer adsorption or distribution compared to TFA, or faster metabolism or excretion in vivo. We observed that 2,2,2-trifluoroacetamide and N-ethyltrifluoroacetamide underwent relatively rapid conversion to TFA (half-life <1 h) following oral administration to the mice. N-ethyltrifluoroacetamide gave rise to both TFA and trifluoroacetamide. N-deethylation of N-ethyl acetamides to the corresponding acetamide (CONHEt → CONH₂) has been observed previously in the microsomal metabolism of N,N-diethyl-toluamide (DEET)⁶¹. In other molecular contexts, trifluoracetamide groups are more stable against metabolism, as exemplified by Paxlovid (nirmatrelvir) in which only ~4% of the dose administered to humans is eventually converted to the des-trifluoroacetylated metabolite⁶².

If the findings here in mice extend to human systems, TFA or its analogues might have therapeutic applications, in analogy to existing peroxisome proliferator fibrate drugs prescribed as lipid-modifying agents to reduce triglyceride and LDL cholesterol levels and increase HDL levels^63,64. Further, the findings reported here may have other implications for human health. TFA is found at higher levels in humans relative to other PFAS¹⁶, although still at plasma levels much lower than were present in our studies. Based on the limited studies available, TFA contamination seems to be ubiquitous in the environment^65,66, and environmental TFA levels are predicted to increase in coming decades due to newer PFAS alternatives^11,14. Another potential source of human exposure to TFA is treatment with the inhalation anesthetics such as halothane and isoflurane¹⁷, which are among the most commonly used agents for induction and maintenance of general anesthesia for surgery. These anesthetics are metabolized in the liver to TFA as a major metabolite^17–19. In one study, peak plasma concentrations of TFA in patients following treatment with halothane were reported to be 580 µM at 24 h following the start of anesthesia, and remained higher than 200 µM for at least 96 h¹⁸. Whereas newer anesthetics such as isoflurane (~2% metabolized to TFA) and desflurane (~0.2% metabolized to TFA) have become the standard of care in developed nations, halothane (~50% metabolized to TFA) is still widely used in the developing world^17,60.

Our study raises several questions that merit future research. Most importantly, further studies are warranted to better characterize the response of human tissues to TFA exposures. We observed that TFA treatment activated human PPAR-α in a reporter assay in human liver HepG2 cells, but neither gene expression changes nor changes in palmitoyl-CoA oxidation activity were observed in the human liver cells. This is consistent with previous findings that rodent liver cells are more sensitive to induction of peroxisome proliferation than human cells^23,67. Also, it would be useful to better characterize the differences in phenotypic effects of TFA across mouse strains and sexes. In this regard, more detailed characterizations of tissues from TFA-treated animals, especially atherosclerotic plaques and liver phenotypes, would be helpful, as would a more detailed analysis of TFA effects on lipoprotein synthesis, secretion, and clearance. Studies using PPAR-α^−/− mice could be potentially informative, especially for linking the effects of TFA to observed changes in lipid levels. Last, our findings suggest that TFA or its analogs may exert beneficial effects in metabolic diseases other than atherosclerosis, such as diabetes, but this needs to be experimentally verified.

Methods

Materials

Amino acids, Rink amide MBHA resin, and N-hydroxybenzotriazole (HOBt) were purchased from Advanced ChemTech, CombiBlocks, or NovaBiochem. Sodium trifluoroacetate was from Alfa-Aesar. N-ethyl-trifluoroacetamide, 2,2-difluorobutyric acid, and 2,2-difluoropropionic acid were from CombiBlocks. 2,2,2-trifluoroacetamide, 5-trifluoromethyltetrazole, and trifluoromethane sulfonamide were from Oakwood. Difluoroacetatic acid and sodium trichloroacetate were from Acros Organics. 3-(trifluoromethyl)-1,2,4-oxadiazol-5-ol was from 1ClickChemistry. Trifluoroacetic acid was obtained from Halocarbon Products (Hackensack, NJ). Antibodies were from Invitrogen or Life Technologies. Other chemicals were purchased from Sigma-Aldrich or Fisher Scientific or as noted below. Oligonucleotides were ordered from Integrated DNA Technologies (IDT) unless otherwise noted.

Solid-phase peptide synthesis (SPPS)

Peptides were synthesized using standard Fmoc chemistry with an Advanced Chemtech Apex 396 peptide synthesizer. A typical synthesis was performed on 0.09 mmol scale using 0.6 mmol/g Rink amide MBHA resin (Novabiochem) to prepare a C-terminal amide. Standard side chain protecting groups included Cys(Trt), Gln(Trt), Lys(Boc), Thr(tBu), Trp(Boc), Glu(OtBu), Arg(Pbf), Ser(tBu). ‘[PEG]’ refers to amino)ethoxy]ethoxy]acetic acid, incorporated using Fmoc-AEEA (CAS # 166108-71-0). The biotin moiety was incorporated using 4 eq. D-biotin (Combi-Blocks), 4 eq. PyOxim (ethyl cyano(hydroxyimino)acetato-O2] tri-1-pyrrolidinylphosphonium hexafluorophosphate), and 8 eq. N, N-diisopropylethylamine (DIPEA) with NMP as solvent. Chain elongations were achieved using DIC and HOBt in NMP (N-methylpyrrolidinone) with 90-min couplings. Fmoc deprotection was achieved using 25% piperidine in NMP. Peptides were cleaved from the resin with concomitant side chain deprotection by agitation in a solution of 94:2.5:2.5:1 TFA/ethanedithiol/TIS/water for 3 h. The peptide was precipitated with ether, centrifuged, and washed three additional times with ether. The crude peptides were purified by preparative reverse-phase (RP)-HPLC on a Vydac or Phenomenex reversed-phase C18 column. Purity was confirmed by analytical RP-HPLC. Purified peptides were characterized by analytical HPLC-MS and/or MALDI-TOF mass spectrometry. Analytical reverse-phase HPLC was performed using a Zorbax 300-SB C-18 column connected to a Hitachi D-7000 HPLC system. Binary gradients of solvent A (99% H₂O, 0.9% acetonitrile, 0.1% TFA) and solvent B (90% acetonitrile, 9.9% H₂O, 0.07% TFA) were employed for HPLC.

Parent 23-residue peptide

The sequence of parent peptide is H-Cys(CH₂CONH₂)-Gly-Val-Leu-Glu-Ser-Phe-Lys-Ala-Ser-Phe-Leu-Ser-Ala-Leu-Glu-Glu-Trp-Thr-Lys-Lys-Leu-Gln-CONH₂. The N-terminal Cys residue was alkylated with iodoacetamide⁹, resulting in a residue mass of 161.04 Da. MW = 2671.9 g/mol.

FREL

Peptide sequence H-Phe-Arg-Glu-Leu-COOH. MW = 563.6 g/mol.

KERS

Peptide sequence H-Lys-Glu-Arg-Ser-COOH. MW = 518.6 g/mol.

CGVLESF

Peptide sequence H-Cys(CH₂CONH₂)-Gly-Val-Leu-Glu-Ser-Phe-COOH. The N-terminal Cys residue was alkylated with iodoacetamide⁹, resulting in a residue mass of 161.04 Da. MW = 811.7 g/mol.

KASF

Peptide sequence H-Lys-Ala-Ser-Phe-COOH. MW = 451.5 g/mol.

EEWTKKLQ

Peptide sequence H-Glu-Glu-Trp-Thr-Lys-Lys-Leu-Gln-COOH. MW = 1061.2 g/mol.

LSALEE

Peptide sequence H-Leu-Ser-Ala-Leu-Glu-Glu-COOH. MW = 660.7 g/mol.

Biotin-PEG-EAEEPSLLKKLLLAPANTQ-NH₂

Peptide sequence biotin-PEG-Glu-Ala-Glu-Glu-Pro-Ser-Leu-Leu-Lys-Lys-Leu-Leu-Leu-Ala-Pro-Ala-Asn-Thr-Gln-CONH₂. MW = 2435.9 g/mol.

In vitro pepsin digestions of parent 23-residue peptide

To determine what peptide fragments would be produced by pepsin digestion, the parent peptide (0.5 mg/mL peptide, prepared in 10% acetic acid) was incubated with pepsin (P-6887 from Sigma) at 37 °C in 10% acetic acid (pH 2.1). Three different incubations were carried out to obtain less extensive and more extensive digests of the peptides: 0.5 U pepsin for 160 min, 1500 U pepsin for 3 min, and 1500 U pepsin for 160 min. Aliquots of the reactions were quenched at the desired times by addition of pepstatin (10 μg/mL) and dilution with acetonitrile. Samples were analyzed by HPLC-MS and product fragments identified manually based on UV absorbance and mass.

Peptide counterion exchange

The TFA counterion in LSALEE•TFA was exchanged with HCl by repeated lyophilization from aqueous HCl solutions²¹. Peptide LSALEE•TFA (340 mg) was dissolved in 30 mL of 100 mM HCl solution and lyophilized overnight. The process was repeated twice. Complete removal of TFA from the peptide was confirmed using ¹⁹F-NMR analysis.

19F-NMR

Spectra were recorded on a Bruker AV Neo 399 spectrometer equipped with a 5 mm BBFO smart probe. A coaxial tube insert containing 500 µM trifluoromethoxybenzene (CAS 456-55-3) dissolved in d6-DMSO was used to provide a lock signal and internal concentration standard. For each spectrum, the pulse signal was modified to be centered between the outermost ¹⁹F peaks (for example, for spectra containing TFA at −78 ppm and trifluoromethoxybenzene at −58 ppm, the pulse was set to −68 ppm). The d1 relaxation time was increased to 8 s. Spectra were typically recorded with 64–90 scans.

Animals

The TSRI Institutional Animal Care and Use Committee approved all experimental protocols involving live animals. The group size was typically n = 8, which was chosen based on prior experience in atherosclerosis efficacy studies indicating that this number is sufficient to detect significant differences of greater than 20% between groups in plasma cholesterol levels and atherosclerotic lesions. In studies involving female mice, groups were randomized and balanced by swapping individual mice between cages based on mouse weight and age prior to the beginning of treatment; in studies involving male mice (because males fight if introduced to new cage mates), groups were randomized and balanced by choosing cages providing the best balance of mouse weight and age. Studies were not blinded. No data points were omitted from analysis.

LDL receptor-deficient (Ldlr^−/−) mice on a C57BL/6J/6J background (B6.129S7-Ldlr^tm1Her/J, Jackson strain # 002207) were initially purchased from Jackson Laboratories (Bar Harbor, ME) and were bred in house. The mice were weaned at 4 weeks of age and were fed ad libitum a standard chow diet (LabDiet 5053). Mouse cohorts selected for study were between 10 and 12 weeks old when studies were initiated. The high-fat diet (HFD) used for Ldlr^−/− mice contained 1.25% cholesterol, 37.3% kcal from fat, and no cholate (Inotiv TD.94059).

Apoe deficient (Apoe^−/−) mice on a C57BL/6J/6J background (B6.129P2-Apoe^tm1Unc/J, Jackson strain # 002052) were initially purchased from Jackson Laboratories (Bar Harbor, ME) and were bred in house. The mice were weaned at 4 weeks of age and were fed ad libitum a standard chow diet (LabDiet 5053). Mouse cohorts selected for study were between 6 and 7 weeks old when studies were initiated. The high-fat diet (HFD) used for Apoe^−/− mice contained 0.2% cholesterol, 42% kcal from fat, and no cholate (Inotiv TD.88137).

10-week-old C57BL/6J/6J DIO mice (Jackson stock # 380050) or C57BL/6J/6J DIO control mice (stock # 380056) were purchased from Jackson Laboratories (Bar Harbor, ME). These mice are fed a HFD (Research Diets # D12492i, 60% kcal from fat) or matched control diet (Research Diets # D12450Ji, 10% kcal from fat) starting at 6 weeks of age at Jackson Laboratories. The mice were maintained on the same diets at Scripps Research. The mice were kept at Scripps Research for two weeks prior to start of treatment to allow the animals to acclimate.

Solutions for administration to mice

For the majority of studies involving Ldlr^−/− or Apoe^−/− mice, PBS was used as vehicle. The pH was adjusted to 7 using 5 N NaOH in cases where the test agent (TFA analog or short peptide) caused a drop in pH of the vehicle. Solutions containing TFA were prepared using sodium trifluoroacetate, not trifluoroacetic acid. Acetate buffer (10 mM) at pH 4 was used as vehicle for the study involving wild-type C57BL/6J mice. Stock solutions of test agent in vehicle were prepared fresh every one or two weeks, stored at −80 °C, thawed, and filtered through a 0.22 µm sterile syringe filter before administration. Test agents were administered by daily i.p. injection (0.2 mL) or by daily oral gavage (0.2–0.25 mL), or in the drinking water for two or ten weeks. The concentration of TFA for i.p. or gavage varied from 1 to 60 mM to achieve the corresponding 1.4–81 mg/kg dose. In the drinking water, a 1 mM TFA solution was used to achieve a dose of ~27 mg/kg/day, because mice drank ~4.5 mL per day. The prepared drinking water solution was added to 50 mL graduated polypropylene tubes fitted with standard rubber stoppers and stainless-steel sippers of a type used routinely in mouse water bottles. The solution was changed daily. There was no significant difference in water or food consumption between groups. Fenofibrate was prepared fresh daily as a suspension, 20 mg in 2 mL of 0.5% methylcellulose (MC) and administered by oral gavage (0.2 mL) for a dose of 100 mg/kg. Feeding and treatment schedules for the in vivo efficacy studies are shown in Supplementary Fig. 1.

Analysis of mouse plasma lipids and biomarkers

Blood (200 µL) was collected after a 6 h fast by retro-orbital puncture into a heparinized capillary tube and transferred to K₂EDTA collection tube. Plasma was separated immediately by centrifugation of the blood samples at 2000 × g for 10 min at 4 °C and stored at −80 °C. Plasma total cholesterol (TC) was measured using an enzymatic colorimetric method kit (Amplex® red cholesterol assay kit, No. A12216, Life Technologies) according to the manufacturer’s instructions. Plasma triglyceride, free fatty acids, SAA, and CRP levels were measured by using a Triglyceride Colorimetric Assay Kit (Cayman, No. 10010303), Free Fatty Acid Assay Kit (Sigma-Aldrich MAK466), Mouse SAA Mouse ELISA Kit (Life Technologies, No. KMA0021), and Mouse CRP ELISA kit (Invitrogen, No. EM20RB), respectively, following the manufacturer’s instructions. qPCR for ACOX1 and ACADS genes in C57BL/6J mice was carried out using TaqMan™ Gene Expression Assay (Catalog # 4331182), assay ID Mm01246834_m1 for ACOX1 and assay ID Mm00431617_m1 for ACADS.

Lipoprotein profile analysis

Plasma was pooled from three mice chosen randomly from each group (15 μL from each mouse, 45 μL total pool) following a two-week daily oral gavage treatment in female Ldlr^−/− mice. Lipoproteins were separated by using a Biozen 3 μm dSEC-2 column (200 Å, 300 ×7.8 mm, flow rate = 1 ml/min, Phenomenex part # 00H-4788-K0). The plasma was centrifuged at 10,000 × g for 10 min at 4 °C to remove particulates (floating material was gently mixed into the liquid before removing the supernatant for SEC injection). 30 μL of the pooled plasma was injected on the system eluted with 10 mM tris-HCl buffer, pH 7.4, containing 1 mM EDTA and 150 mM NaCl; a fraction size of 0.25 mL was used. Total cholesterol in the FPLC fractions was measured using the Amplex® red cholesterol assay kit (No. A12216, Life Technologies) according to the manufacturer’s instructions.

Evaluation of atherosclerosis

Atherosclerotic lesion severity was assessed in the aortae as previously described⁶⁸ with minor modifications. At euthanasia, animals were perfused with PBS, followed by 4% formaldehyde (10% UltraPure EM Grade from Polysciences diluted in PBS, pH 7.2). For en face analysis, the entire mouse aorta was dissected from the proximal ascending aorta to the bifurcation of the iliac artery by using a dissecting microscope. Adventitial fat was removed, and the aorta was opened longitudinally, pinned flat onto black dissecting wax, stained with Sudan IV, and photographed at a fixed magnification. The photographs were digitized, and total aortic areas and lesion areas were calculated by using Adobe Photoshop version CS4, Chromatica V, and NIH Scion Image software (http://rsb.info.nih.gov/nih-image/Default.html). The results were reported as a percentage of the total aortic area that contained lesions.

As a second assessment of atherosclerosis, lesions of the aortic root (aortic sinus) were analyzed by a modification of current methods of valve lesion analysis⁶⁸. Utilizing stereological principles, lesion volume was estimated across a fixed distance of the aortic sinus. After 10 min fixation in 4% paraformaldehyde during perfusion, hearts were cut at an angle perpendicular to the atria of the heart and embedded in OCT (Tissue-Tek). Frozen hearts were sectioned on a Leica cryostat, with 10-μm sections collected from the beginning of the aortic sinus (defined as when a valve leaflet became visible) to 500 μm below the beginning of the sinus. For hearts cut at an angle that resulted in valve leaflets not appearing in the same section (due to poor section angle), the lagging leaflet was used to determine the 500-μm distance. Sections were collected in duplicate at 50-μm intervals. Sections were stained with oil red O, counterstained with Gill hematoxylin 1 (Fischer Scientific International), photographed, and digitized for lesion analysis. Scoring of valve lesion areas was done for each of the 3 valve cusps individually. Lesion areas found only within the valve cusp were measured. Lesion volume estimation was determined from a 1 in 10 sampling rate; hence, valve cusps spaced at 140 μm were used to determine the lesion volume for a total of 4 sections analyzed per valve cusp. Lesion volume was calculated from an integration of the measured cross-sectional areas. Prediction of the coefficient of error (CE) in approximating lesion volume was computed using the Cavalieri estimator derived from a covariogram analysis of an ordered set of estimates of cross-sectional areas. This yielded CE values of less than 10% that were acceptable for a stereological computation of lesion volume. Supplementary Data 4 contains cusp measurements used in the analysis, along with the plaque volume calculations.

Harvesting of liver samples

Euthanized mice were perfused with sterile 10 mL PBS (left ventricle to right atrium) to remove blood from the liver. The liver was excised and weight recorded. Sections of liver were cut and processed according to the various downstream processing needs. A small amount, ~30 mg, was placed into RNAlater (Invitrogen/Thermo Fisher Scientific # AM7021, Lithuania) and frozen at −80 °C for RNA extraction. One of the liver lobes was cut into ~5 mm sections and placed into 3% paraformaldehyde in PBS. Sections of ~30 µg size were placed into lysing buffer (male mice) or flash frozen in liquid nitrogen (female mice) for extraction of PMP70 for western blotting. Sections of ~100 µg size were flash frozen in liquid nitrogen for analysis of palmitoyl-CoA oxidase activity.

Confocal immunofluorescence imaging of liver sections

Liver tissue was fixed in 3.5 mL freshly prepared 3% paraformaldehyde in PBS overnight at 4 °C. Following fixation, the tissue was transferred to 3.5 mL 15% sucrose in PBS. When the tissue sank to the bottom of the tube, it was transferred to 3.5 mL 30% sucrose in PBS. Tissue samples were stored overnight at 4 °C, providing ample time for the samples to sink to the bottom of the tube. Tissue samples were blotted on a kimwipe and transferred to a mold containing OCT (Tissue-Tek # 4583, Sakura Finetek USA, Torrance, CA) and frozen on a block of dry ice. Frozen samples were transferred to a −80 °C freezer for storage. Samples were sectioned onto slides at 10-micron thickness using a Leica 1800 cryostat set at −20 °C, 2 sections per slide. Sectioned slides were stored at −20 °C. Slides were washed once in 0.3 M glycine in PBS for 5 min and three times in PBS for 5 min at room temperature. Slides were permeabilized in 0.1% Triton X-100 in PBS for 5 min on ice. Secondary antibody binding sites were blocked in 5% normal goat serum +0.1% Triton X-100 for one hour at room temperature. Slides were incubated with primary antibody (polyclonal rabbit anti-PMP70 # PA1650, Life Technologies, Carlsbad, CA) at final concentration of 2 µg/mL in PBS containing 5% normal goat serum. Primary antibody was applied at 100 μL per section, overlaid with a coverslip, and placed in a humidity chamber for three hours at room temperature. Coverslips were floated off the sections in PBS and slides were washed three times in PBS for 5 min. Secondary antibody (goat anti-rabbit AF 568 # A11036, Life Technologies, Carlsbad, CA) was applied at final concentration of 4 µg/ml in PBS containing 5% normal goat serum. Secondary antibody was also applied at 100 μL per section, overlaid with a coverslip, and placed in a humidity chamber for one hour at room temperature. Coverslips were floated off the sections in PBS and slides were washed three times in PBS for 5 min. Slides were washed once in 0.1% Triton X-100 in PBS for 5 min and three more times in PBS for 5 min then stained for 2 min with DAPI (Biolegend # 422801, San Diego, CA) at final concentration of 0.5 μg/mL. Slides were washed a final three times in PBS, mounted with Vectashield (Vector Laboratories # H-1400, Burlingame, CA) at 30 μL per section, covered with a coverslip (22 × 50 × 1.5 mm) and sealed with clear nail varnish. Sealed slides were stored at 4 °C, covered from light. Images were collected on a Zeiss LSM 710 confocal microscope, using a 60× Plan-Apochromat NA 1.4 objective (oil). Single 60× Z-planes were imaged sequentially with the following settings: pixel dwell time of 2.33 microseconds, and pixel dimensions of 0.098 microns (1380 × 1380 pixels); DAPI- 2.0% laser power, emission filter 410–585 nm, master gain 696, digital offset 7.0, pinhole 58 µm; green autofluorescence- 77.3% laser power, emission filter 493–574 nm, master gain 790, digital offset 0.00, pinhole 51 microns; red fluorescence- 7.9% laser power, emission filter 568–712 nm, master gain 790, digital offset 8.0, pinhole 51 microns.

TFA pharmacokinetics in vivo

Ldlr^−/− mice (~20–25 g) were maintained on a chow diet. Sodium trifluoroacetate was formulated as a 25 mM solution in pH 7 PBS, and sterile filtered through a 0.2 µm filter. For the 24-h PK studies, groups of two mice received a 27 mg/kg or 81 mg/kg dose of TFA via intraperitoneal injection (0.2 mL) or oral gavage (0.2 mL). Three such groups were used so that no animal had more than two survival blood draws. Blood was drawn from the retro-orbital sinus into a heparinized capillary tube at different intervals beginning 15 min after dosing. Final blood collections (~500 µL) were done by cardiac puncture. For the two-week PK studies in plasma and liver, groups of three mice received a 27 mg/kg daily dose of TFA via oral gavage (0.2 mL) for 14 days. Blood was drawn from animals (retroorbital) on day three at 1 h post gavage, on day seven at 1 h post gavage, on day 14 at 1 h pre-gavage, and on day 14 at 1 h post gavage. Plasma was isolated immediately from the whole blood by centrifugation at 2000 × g for 10 min at 4 °C. Immediately after blood collection and plasma separation, plasma samples were flash frozen in liquid nitrogen and stored at −80 °C until analysis. Livers were harvested on day 15 (between 25 and 29 h following the final administration of TFA on day 14), flash frozen in liquid nitrogen, and stored at −80 °C until analysis. Liver tissue (~80 mg) was weighed into a tared vial and homogenized in 0.6 mL RIPA lysis buffer (Pierce) using a pellet pestles cordless motor. Samples were incubated at 4 °C for 30 min, sonicated in a bath sonicator for 2 min, and vortexed for 30 s. Samples were then centrifuged at 16,000 × g for 15 min at 4 °C to pellet tissue debris. The lysate was removed and kept at 4 °C until ¹⁹F-NMR analysis the same day. TFA concentrations in the samples were determined by using ¹⁹F-NMR, with a lower limit of quantification of 10 µM. Plasma samples (~150 µL) were thawed and diluted with deionized water to a final volume of 0.5 mL; liver lysate samples were used without dilution for NMR analysis. Spectra were recorded using a coaxial tube insert containing 500 µM trifluoromethoxybenzene (CAS 456-55-3) dissolved in d6-DMSO to provide a lock signal and internal concentration standard. A standard curve for TFA in plasma was generated by spiking known concentrations of TFA into 150 µL plasma from untreated animals and diluting with 350 µL deionized water. TFA concentrations were calculated by interpolation from the standard curve of the NMR integral for TFA relative to that of the internal concentration standard. Spiking of known TFA amounts into liver homogenate from animals that had not received TFA indicated that the TFA standard curve generated for plasma samples could also be used to give accurate concentrations of TFA for liver lysate. To back-calculate the concentration of TFA in the original liver tissues, we assumed that the liver tissue was composed entirely of water (no solid tissue). Because liver is ~70–75% water in reality, the actual TFA concentration within liver cells would be somewhat higher than the TFA concentrations reported.

Stability of TFA analogs in vivo

Male Ldlr^−/− mice (~25 g) were maintained on a chow diet. The TFA analog (trifluoroacetamide, N-ethyl-trifluoroacetamide, 2,2-difluorobutyric acid, or difluoroacetic acid) was formulated as a 25 mM solution in PBS, with the pH adjusted to ~7 using 5 M NaOH if necessary, and sterile filtered through a 0.2 µm filter. Groups of four mice received a 200 µmol/kg dose (equivalent dose to 27 mg/kg of sodium TFA) of the analog via oral gavage (0.2 mL). One hour following administration, the mice were sacrificed and blood collected by cardiac puncture (~500 µL). Plasma was isolated immediately from the whole blood by centrifugation at 2000 × g for 10 min at 4 °C. Immediately after blood collection and plasma separation, plasma samples were flash frozen in liquid nitrogen and stored at −80 °C until analysis by using ¹⁹F-NMR. Plasma samples were thawed and diluted with deionized water to a final volume of 0.5 mL. Spectra were recorded using a coaxial tube insert containing 500 µM trifluoromethoxybenzene (CAS 456-55-3) dissolved in d6-DMSO to provide a lock signal and internal concentration standard. Spiking of the plasma with TFA or the TFA analogs was used to confirm the identity of the observed peaks.

RNA-seq sample preparation and sequencing

Mouse liver tissue that had been stored in RNAlater at −80 °C after harvest (around 20–30 mg) was used for RNA extraction with an RNeasy Mini Kit (Qiagen) following the manufacturer’s instructions. RNA quality was assessed by Agilent’s Bioanalyzer or Tapestation (Agilent Technologies). RNA quantity was measured by Qubit RNA BR/HS kit (Invitrogen). The rRNA was removed from total RNA with Ribo-Zero Plus rRNA Depletion Kit (Illumina). RNA-seq libraries were constructed using NEBNext Ultra II RNA Library Prep Kit for Illumina (NEB). Samples were purified with NEBNext® Sample Purification Beads (NEB) and quantified by Qubit dsDNA HS Assay Kit (Invitrogen). NEBNext Multiplex Oligos for Illumina (Dual Index Primer Set1) (NEB) was used to barcode libraries. Libraries with different barcodes were pooled, and paired-end sequencing (150 bp) was performed on an Illumina NextSeq platform at the Scripps Research Institute next-generation sequencing core.

RNA-seq analysis

RNA-Seq data were processed and aligned using the NF-core RNAseq v3.5 pipeline⁶⁹. Reads were trimmed using Trim Galore (v0.6.7) and aligned using Star (v2.6.1d) against the GRCm38 reference genome and annotation. Transcript quantification was performed using Salmon (v1.5.2). Differential analysis was performed using DESeq2 (v1.28.0) and the R statistical environment (v4.0.2). Pathway analysis was performed using pathfindR (v1.6.3)⁷⁰ inside the R statistical environment (v4.0.2). The mouse KEGG gene set was used to determine the active subnetworks from the protein-protein interaction dataset. The false discovery rate (FDR) was used as the p-value correction methodology (p < 0.05). An adjusted p < 0.05 was used for statistical analysis and pathways with fold-enrichment >5 are shown in Fig. 3. The “score” displayed in the heatmap represents the agglomerated z-score, calculated using the score_terms() function in pathfindR package. This score reflects the standardized enrichment signal of a given pathway in all samples, incorporating directionality (activation or repression). A positive score indicates activation of the pathway. A negative score indicates repression. The RNA-seq data is deposited in Bioproject at accession #: PRJNA917696.

Meta-transcriptomics sequencing and analysis

In an anaerobic chamber, cecal material from HFD-fed Ldlr^−/− mice was resuspended in Chopped Meat Carbohydrate Broth (BD, 297307) at a ratio of 1.5 ml per 100 mg of cecal content. The mixture was vortexed for 5 min and then allowed to settle and strained through a 100 µm filter. The resulting supernatant was treated with 1 mM TFA or vehicle control and then incubated anaerobically at 37 °C for 6 h. The bacteria were collected by centrifugation and stored in PowerProtect DNA/RNA solution (Qiagen, Cat. 14800) at −80 °C. Total RNA was extracted using the RNeasy PowerFecal Pro Kit (Qiagen, Cat. 78404) according to the manufacturer’s instructions. The RNA quality was evaluated using the TapeStation (Agilent Technologies), and the quantity was determined with the Qubit RNA BR Kit (Invitrogen). Library preparation and sequencing were subsequently carried out by Novogene (Sacramento, CA). The SAMSA2 metatranscriptome analysis pipeline (https://github.com/transcript/samsa2) was used for organism and gene function identification.

16S rRNA gene sequencing and analysis

16S rRNA sequencing was conducted following a previously established method⁴² with minor modifications. In short, in an anaerobic chamber, cecal contents from mice fed HFD or CHOW diets were resuspended in PBS (1.5 ml per 100 mg of cecal content). The mixture was vortexed for 5 min, allowed to settle, and strained through a 100 µm filter. Supernatant was diluted 12.5-fold with Chopped Meat Carbohydrate Broth (Anaerobe Systems). TFA or a control gut microbiome remodeling cyclic peptide was added at the desired concentration, and the mixture was incubated anaerobically at 37 °C overnight. Bacterial cells were then collected, and DNA was extracted using the DNeasy UltraClean Microbial Kit (Qiagen) according to the manufacturer’s instructions. The V3–V4 region of the 16S rRNA gene was amplified using KAPA HiFi HotStart ReadyMix (Kapa Biosystems) with the following primers: forward primer, (5′-AATGATACGGCGACCACCGAGATCTACACNNNNNNNNACACTCTTTCCCTACACGACGCTCTTCCGATCTACTCCTACGGGAGGCAGCAG-3′) and reverse primer, (5′-CAAGCAGAAGACGGCATACGAGATNNNNNNNNGTGACTGGAGTTCAGACGTGTGCTCTTCCGATCTnnnnnnnnnnGGACTACHVGGGTWTCTAAT-3′). Each sample was tagged with a unique 8-base barcode (N) for identification and a 10-base unique molecular identifier (n) to correct for PCR duplication artifacts during library preparation. Amplicons were quantified using the Qubit dsDNA Assay Kit (Invitrogen), and equimolar amounts of DNA from each sample were pooled and purified with Agencourt AMPure XP beads (Beckman Coulter). The final library was sequenced on an Illumina NextSeq platform. Sequencing data were demultiplexed and analyzed for taxonomic identification using QIIME2 pipeline (https://qiime2.org).

Cell culture

Rat hepatoma FAO cells (obtained from Sigma-Aldrich, cat. # 89042701, RRID: CVCL_0269) were cultured in Ham’s F-12K (Kaighn’s) medium (Gibco) containing 10% FBS (HyClone), 100 U/ml penicillin, and 100 µg/ml streptomycin. The HepG2 cell line (obtained from ATCC, cat. # HB-8065, RRID: CVCL_0027) was cultured in DMEM medium (Gibco) containing 10% FBS (HyClone), 100 U/ml penicillin, and 100 µg/ml streptomycin. Cells were maintained at 37 °C in a humidified atmosphere of 95% air and 5% CO₂.

Luciferase reporter assay

We followed a previously reported procedure³⁵ with some modifications. Plasmid PPRE X3-TK-luc, which encodes a firefly luciferase gene under the control of peroxisome proliferator response element (PPRE), was used to monitor the effect of TFA on different PPAR isoforms. Beta-galactosidase was used to normalize the luciferase signal by co-transfection with pSV-β-galactosidase control vector (Promega). All three PPAR isoform cDNAs were cloned into the HindIII site of pBApo-CMV plasmid (Takara) by NEBuilder® HiFi DNA Assembly kit (NEB). The primers used are listed in Supplementary Table 1. All constructed plasmids were verified by Sanger sequencing (Eton bioscience). To carry out the luciferase reporter assay, HepG2 cells were seeded into a 24-well plate at ~2.0 × 10⁵ cells per well in 0.5 mL DMEM medium (Gibco) containing 10% FBS (HyClone), 100 U/ml penicillin, and 100 µg/ml streptomycin and grown overnight. Next, 400 ng PPRE X3-TK-luc plasmid, 100 ng pSV-β-galactosidase control vector, and 20 ng pBApo-CMV containing one of the PPAR isoforms were added to the culture media for transfection with lipofectamine 3000 (Invitrogen), according to manufacturer’s protocol. After 24 h, the medium was replaced by fresh DMEM containing 10% FBS with different concentrations of TFA or positive control compound and incubated for the desired time. The cells were washed with PBS, and then were scratched and lysed into 150 µl reporter lysis buffer (Promega). Luciferase activity in the lysate was measured using the Bright-Glo™ Luciferase Assay System (Promega) and normalized by β-galactosidase activity using β-Galactosidase Enzyme Assay System (Promega). All luciferase experiments were replicated at least twice.

FAO cell treatment and RT-qPCR analysis

Rat hepatoma FAO cells were seeded in a 24-well plate at ~2.0 × 10⁵ cells per well in Ham’s F-12K (Kaighn’s) medium (Gibco) containing 0.5% FBS (HyClone), 100U/ml penicillin, and 100 µg/ml streptomycin and grown overnight. The media was then replaced with fresh media containing 1 mM TFA, 10 mM TFA, or 0.1 mM ciprofibrate. The treated cells were cultured for 24 h before RNA isolation. For total RNA isolation, the supernatant was discarded, and the cells were washed with PBS. Total RNA was isolated using Trizol reagent (Invitrogen) following the manufacturer’s protocol. Reverse transcription was carried out using QuantiTech Reverse Transcription Kit (Qiagen). The cDNA was diluted 10-fold and used for quantitative PCR with PowerUp™ SYBR™ Green Master Mix (Applied Biosystems) on a Bio-Rad CFX384 Touch Real-Time PCR Detection System, using primers specific for each gene (Supplementary Table 1). The ΔΔCt method was used for quantification. Data were normalized to housekeeping gene glyceraldehyde-3-phosphate dehydrogenase (GAPDH). All RT-qPCR experiments were replicated at least twice.

CRISPR-Cas9 gene knockout of PPAR-α

To easily screen mutants, a fragment deletion strategy was used instead of indel mutation. To construct the FAO PPAR-α^−/− cell line, two gRNAs were designed using CRISPOR website⁷¹ and inserted into the pSpCas9(BB)-2A-Puro (PX459) plasmid by Golden gate method. The constructed plasmids were verified by Sanger sequencing. The gene knockout protocol was modified from the previously described method⁷². Briefly, FAO cells were seeded into a 24-well plate at ~5.0 × 10⁴ cells per well and transfected with the constructed plasmid by lipofectamine 3000 (Invitrogen). After 6 h, the transfection reagents were removed by replacing the media with fresh medium. The transfected cells were incubated for 48 h to express the puromycin resistant protein. Cells were harvested and resuspend in fresh medium containing 20 µg/ml puromycin (Invitrogen) and seeded into a 60 mm petri dish. After 3 days, the medium was replaced with fresh medium containing no puromycin. A dilution-based method was used to isolate a single clonal cell line that had the designed mutation. The expected deletion mutant of the PPAR-α gene was confirmed by PCR.

Western-blot analysis

Protein fractions from fresh or flash-frozen tissue were extracted using T-PER tissue protein extraction reagent (Thermo Fisher Scientific) according to the manufacturer’s protocol. Around 25 mg of liver tissue was homogenized in 500 µL extraction reagent containing 1× Halt™ protease and phosphatase inhibitor cocktail (Thermo Fisher Scientific) using a pellet pestles cordless motor. After homogenization, samples were centrifuged at 10,000 × g for 5 min at 4 °C to pellet tissue debris. The supernatant was filtered using a 0.22 µm filter, and the total protein concentration was measured by using a BCA protein assay kit (Thermo Fisher Scientific). SDS-PAGE was carried out using a Bolt 4–12% gradient Tris-Bis gel following the manufacturer’s protocol. 25–50 µg total protein was loaded on each well and run at 100 V for 100 min. PVDF membrane (Thermo Fisher Scientific) was used for transfer at 30 V for 2 h by wet transfer. After transfer, the membrane was blocked in 5% BSA for 1 h at room temperature. Immunoblotting was carried out using Pierce Fast Western Blot Kit and SuperSignal West Pico Substrate (Thermo Fisher Scientific), following the manufacturer’s protocol. Briefly, the rabbit anti-PMP70 antibody (Thermo Fisher Scientific cat. PA1-650, RRID: AB_2219912) and anti-GAPDH antibody (Thermo Fisher Scientific cat.PA1-987, RRID: AB_2107311) were diluted 1:1000 in the diluent buffer and incubated with the PVDF blot at room temperature for 1 h. After washing, the PVDF blot was incubated with HRP-conjugated antibody at RT for 10 min. Excess secondary antibody was removed using three washed with washing buffer. SuperSignal West Pico Substrate was used to detect the HRP-conjugated secondary antibody. Imaging was conducted via Bio-Rad ChemiDoc XRS+ Imaging system.

Palmitoyl-CoA oxidation assay

Around 150 mg liver tissue that had been perfused with PBS was homogenized in 750 µL lysis buffer containing 50 mM Tris-HCl (pH 8.0), 152 mM KCl, 0.1% triton X-100 and 1× Halt™ protease and phosphatase inhibitor cocktail (Thermo Fisher Scientific). Homogenization was performed using a pellet pestles cordless motor on ice. Homogenates were centrifuged at 2500 × g for 15 min at 4 °C. The supernatant was collected, and protein concentration was determined using a BCA protein assay kit (Thermo Fisher Scientific). Supernatant was stored at −80 °C until analysis. Palmitoyl-CoA oxidation activity was determined using slight modifications of a previously described method³⁵. Briefly, around 100 µg protein was preincubated in 1 ml reaction mixture containing 50 mM Tris-HCl (pH 8.0), 0.2 mM NAD⁺, 0.01 mM FAD, 0.1 mM coenzyme A, 6.25 mM dithiothreitol, 0.15 mg/ml bovine serum albumin, 1.5 mM KCN and 0.01% (w/v) triton X-100 for 5 min in a cuvette at 37 °C in a Cary 100 UV–Vis spectrophotometer. Palmitoyl-CoA oxidation was initiated by adding 0.05 mM palmitoyl-CoA substrate (Sigma Aldrich, cat # P9716). The absorbance at 340 nm was recorded for 15 min at 37 °C. The rate of NADH formation was calculated using the initial linear region of the curve (first 5 min) and converted to NADH concentration units using an extinction coefficient of 6.22 mM⁻¹ cm⁻¹.

Mouse PPARα protein expression and purification

The full-length mouse PPARα gene was amplified from the plasmid pBApo-CMV-mPPARα (as described in the Luciferase reporter assay section) using the primer pair 28a-mPPAR-α-F/28a-mPPAR-α-R. The PCR product was purified and cloned into the BamHI/XhoI sites of pET28a-sumo vector using the NEBuilder HiFi DNA Assembly Kit (NEB) to get pET28a-sumo-mPPARα. To obtain a tag-free PPAR-α protein, a WelQut protease recognition site was incorporated at the N-terminus of the PPAR-α gene via primer design. The final plasmid was confirmed by Sanger sequencing (Eton Bioscience). Full-length mouse PPAR-α was expressed in a soluble form as an N-terminal His₆-SUMO-tagged protein in E. coli Rosetta (DE3) cells (Novagen) and purified using Ni-NTA affinity chromatography followed by size-exclusion chromatography. The transformed Rosetta (DE3) harboring pET28a-sumo-mPPAR-α was cultured overnight at 37 °C in LB medium supplemented with 50 μg/mL kanamycin and 17 μg/mL chloramphenicol. The overnight culture was diluted 1:100 into fresh LB medium containing the same antibiotics and grown at 37 °C until the optical density at 600 nm (OD₆₀₀) reached 0.4–0.6. Protein expression was induced by adding 0.5 mM isopropyl β-D-1-thiogalactopyranoside (IPTG), and cells were further cultured at 18 °C for 20 h. Cells were harvested by centrifugation and resuspended in lysis buffer containing 20 mM Tris-HCl (pH 7.5), 500 mM NaCl, 1 mM TCEP, 10 % glycerol, then lysed by sonication. The lysate was clarified by centrifugation at 12,000 × g for 30 min at 4 °C. The supernatant was applied to a HisTrap HP column (GE Healthcare), washed with washing buffer containing 20 mM Tris-HCl (pH 7.5), 500 mM NaCl, 60 mM imidazole, 1 mM TCEP, and 10 % glycerol, and eluted with buffer containing 250 mM imidazole. The eluted protein was concentrated using 10-kDa Amicon Ultra centrifugal filters (Millipore) and incubated with WelQut protease (Thermo Scientific) at 4 °C overnight to cleavage the His₆-SUMO tag. The digestion mixture was diluted 10-fold with lysis buffer and passed through a HisTrap HP column to remove WelQut protease and the cleaved tags. The flow-through containing tag-free PPAR-α protein was concentrated and further purified by size-exclusion chromatography on a Superdex 75 column (Pharmacia Biotech) with a running buffer containing 20 mM HEPES-NaOH (pH 7.5), 150 mM NaCl. Protein concentration was determined using BCA assay kit (Thermo Scientific). The protein was resolved in SDS-PAGE and stained with Coomassie blue.

Biolayer interferometry assay

The binding of trifluoroacetate and other ligands to purified mouse PPAR-α was assessed via a coactivator recruitment assay using biolayer interferometry (BLI), which detects the physical interaction between PPAR-α and a biotin-labeled peptide containing the LXXLL motif. In this case, the peptide used was derived from amino acids 135–153 of the PPAR coactivator 1α (PGC-1α). The sequence of the biotinylated PEG-PGC1α peptide was: biotin-PEG-EAEEPSLLKKLLLAPANTQ-CONH₂. BLI binding assays were performed on an Octet R8 system (Sartorius) using streptavidin (SA) biosensors (Sartorius). All steps were conducted at 25 °C with shaking at 1000 rpm in black 96-well plates (Greiner), with each well containing 200 μL of solution. A reaction buffer containing 20 mM HEPES-NaOH (pH 7.4), 150 mM NaCl, 1 mM EDTA, 1 mM DTT, 0.005 % Tween-20, and 0.1 % fatty acid-free bovine serum albumin (BSA) was used throughout the study, including for dilution of biotin-PEG-PGC1α peptide and protein. After equilibration in reaction buffer for 30 min, SA biosensors were loaded with 20 nM biotin-PEG-PGC1α peptide for 600 s, followed by a 300 s wash in reaction buffer. The loaded sensors were then immersed in 150 nM mouse PPAR-α protein solutions containing varying concentrations of sodium trifluoroacetate or other ligand for the association step (1800 s), followed by a dissociation step (1800 s) in reaction buffer. Binding responses (maximum shift in the association phase) were plotted and fitted using nonlinear regression with a variable slope model to determine EC₅₀ values, using GraphPad Prism 10. Binding experiments involving TFA were performed in triplicate.

Biochemical assays for formation of acetyl-CoA and trifluoroacetyl-CoA

Reactions were carried out in 50 µL buffer containing 100 mM Tris-HCl (pH 8.0), 5 mM ATP, 5 mM MgCl₂, 2 mM CoA, 1 mM DTT, and 0.15% Triton X-100. Sodium acetate and sodium trifluoroacetate were both added at a final concentration of 10 mM. Acetyl-CoA synthetase (Sigma Aldrich, A1765) was used at 5 µM. For liver tissue lysate reactions, 70 µg of total protein was added to the 50 µL mixture. To generate ciprofibroyl-CoA, 2 mM ciprofibrate (TCI America) and 50 µg of rat microsomes (Gibco) were included. After incubating at 37 °C for 1 h, the reactions were quenched by adding 200 µL of chloroform/methanol (2:1, v/v) followed by vigorous mixing. After centrifugation, the upper phase was directly used for LC-MS analysis. Each sample (10 μL) was injected onto an Agilent Poreshell 120 EC-C8 (4.6 ×50 mm, 2.7 µm) analytical column using the Agilent InfinityLab LC/MSD system (Agilent Technologies). Buffer A (0.1% formic acid in water) and buffer B (0.1% formic acid in MeCN) were used in a gradient (flow rate 0.5 mL/min) of 5 % buffer B for 1 min, followed by a linear increase to 100 % buffer B over 9 min, 100 % buffer B for 2 min, followed by descent to 5% buffer B.

Targeted bile acid metabolomics

Bile acids were extracted from plasma or feces samples according to published procedures⁷³. Briefly, fecal samples were lyophilized and homogenized. Five milligrams of fecal powder were extracted with 250 µL of methanol containing heavy internal standards. Fifty microliters of plasma were extracted with 150 µL of methanol containing heavy internal standards. After vortexing for 10 min and centrifuging (16,000 × g, 4 °C, 10 min), supernatants were transferred to glass vials for injection and analysis by LCMS. Bile acids were analyzed on a Dionex Ultimate 3000 LC system (Thermo) coupled to a TSQ Quantiva mass spectrometer (Thermo) fitted with a Kinetex C18 reversed phase column (2.6 µm, 150 ×2.1 mm i.d., Phenomenex). The following LC solvents were used: solution A, 0.1% formic acid and 20 mM ammonium acetate in water, solution B, acetonitrile/methanol (3/1, v/v) containing 0.1% formic acid and 20 mM ammonium acetate. The following reversed phase gradient was utilized: at a flow rate of 0.2 mL/min with a gradient consisting of 25–29% B in 1 min, 29–33% B in 14 min, 33–70% B in 15 min, up to 100% B in 1 min, 100% B for 9 min and re-equilibrated to 25% B for 10 min, for a total run time of 50 min. The injection volume for all samples was 10 µL, the column oven temperature was set to 50 °C and the autosampler kept at 4 °C. MS analyses were performed using electrospray ionization in positive and negative ion modes, with spray voltages of 3.5 and −3 kV, respectively, ion transfer tube temperature of 325 °C, and vaporizer temperature of 275 °C. Multiple reaction monitoring (MRM) was performed by using mass transitions between specific parent ions into corresponding fragment ions for each analyte.

Untargeted metabolomics

The untargeted metabolomics study was carried out by the Scripps Research Center for Metabolomics and Mass Spectrometry core. Plasma samples (100 µL) from female Ldlr^−/− mice treated by daily oral gavage for two weeks with a 27 mg/kg TFA daily dose (n = 10) or PBS vehicle (n = 10) were extracted with cold MeOH (400 µL) and analyzed individually by reversed phase LCMS using water/acetonitrile/formic acid buffer system (column: Waters ACQUITY UPLC BEH C18 Column, 130 Å, 1.7 µm, 1 mm × 100 mm, flow rate 0.15 mL/min), in both positive and negative mode (Bruker Impact II Q-TOF instrument). MS parameters were: End Plate Offset (500 V), Capillary (4000 V), Nebulizer (29.0 psi), Dry gas (8.0 l/min), Dry temp (200 °C), mass range (40–1000 m/z), Rolling average (On 2×), Spectra rate (2.00 Hz). MS data was analyzed (peak alignment, integration, annotation, statistics) using XCMSOnline (www.xcmsonline.scripps.edu) to run statistical comparison between the sample groups. The following cutoff thresholds were used for analysis: a fold-change >1.5, p-value < 0.01, and maximum peak intensity >5000. The XCMSOnline output for positive and negative mode analyses is provided in Supplementary Data 2. The mass fragmentation spectra were manually searched for the presence of potential trifluoroacetyl fragment ions (97.0 Da for F₃CCO and 69.0 Da for F₃C fragments). The lists of metabolite M/Z from XCMSOnline were also searched manually for pairs of parent ion masses that differed by the mass of a trifluoroacetyl modification (96.0 Da for +F₃CCO-H). The raw metabolomics data for this study is available for download from the National Metabolomics Data Repository with the Study ID ST003595 at https://doi.org/10.21228/M88R7M.

Preparation of samples for proteomics

The experiment included samples from mice treated with PBS vehicle (n = 3) or TFA (n = 3). Samples of mouse liver homogenate (100 µl, 2 mg protein/ml) were combined with urea (48 mg) in Protein LoBind Tubes (2.0 ml, Eppendorf 022431102). To each sample was added a solution of tris(2-carboxyethyl)phosphine hydrochloride and potassium carbonate (50 µl, 100 mM and 300 mM respectively, in DPBS), and the samples were incubated at 37 °C for 30 min with shaking. A solution of iodoacetamide (70 µl, 400 mM in DPBS) was then added and the samples were incubated at ambient temperature for a further 30 min in the absence of light. Cold chloroform (220 µl), methanol (880 µl), and water (440 µl) were then added to each sample, and the mixtures were briefly vortexed followed by centrifugation (10,000 × g, 10 min). The upper liquid phases were carefully removed, and cold methanol (600 µl) was added to each sample. The samples were homogenized in a water bath sonicator, centrifuged (10,000 × g, 10 min), and the supernatants were removed. To each sample was added triethylammonium bicarbonate solution (TEAB, 160 µl, 100 mM, pH 8.5) and the protein residue was resuspended by probe sonication (1 s, 7 pulses, 15 % amplitude, 15 ms on, 40 ms off). A solution of LysC (20 µl, 50 µg / ml in TEAB, New England BioLabs P8109S) was added to each sample. Samples were then incubated for 2 h at 37 °C with shaking. Solutions of trypsin (20 µl, 100 µg/ml in TEAB, Promega V5111), Protease Max (2 µl, 1 % w/v in TEAB, Promega V2072), and calcium chloride (2 µl, 100 mM) were added, and the mixtures were incubated overnight at 37 °C with shaking. The samples were centrifuged (12,000 × g, 10 min) and peptide concentrations were estimated using a Pierce Quantitative Fluorometric Peptide Assay according to the manufacturer’s instructions (Thermo 23290). Volumes corresponding to 25 µg digested protein were transferred to new Protein LoBind Tubes (1.5 ml, Eppendorf 022431081), and each sample was made up to 35 µl with TEAB followed by the addition of acetonitrile (12.5 µl). The peptides were labeled with a TMT10plex™ Isobaric Label Reagent Set (Thermo 90406) according to the manufacturer’s instructions. The reactions were quenched through the addition of hydroxylamine solution (3 µl, 5% w/w in water) and incubated for 15 min before the addition of formic acid (5 µl). Solvents were then removed from the samples by vacuum centrifugation. Each residue was dissolved in trifluoroacetic acid solution (40 µl, 0.1% v/v), and the solutions were combined together. An additional volume of trifluoroacetic acid solution (200 µl, 0.1% v/v) was used to wash any residual peptide solutions into the combined solution (final volume 600 µl). The combined peptide solution was eluted into eighteen fractions using a Pierce High pH Reversed-Phase Peptide Fractionation Kit (Thermo 84868) according to the manufacturer’s instructions. The first fraction was combined with the tenth, the second with the eleventh, etc. to yield nine solutions which were then dried by vacuum centrifugation. The residues were redissolved in formic acid solution (65 µl, 0.1% in water), centrifuged (10,000 × g, 10 min), and the supernatants transferred to LC-MS vials for analysis.

Mass spectrometric proteomic analysis

Proteomic samples were analyzed as described previously⁷⁴. Briefly, from each sample 10 µl was loaded on to a precolumn (Acclaim PepMap 100, 75 µm × 2 mm) and separated using an Acclaim PepMap RSLC analytical column (75 µm × 15 cm) on an UltiMate 3000 RSLCnano system (Thermo Fisher Scientific). The following gradient was applied at a flow rate of 300 µl/min and column temperature of 35 °C: 2% buffer B for 10 min, rising to 30% buffer B over 192 min, followed by 60% buffer B for 5 min, 60 % rising to 95% buffer B over 1 min, 95% buffer B for 5 min, falling to 2% buffer B over 1 min, and held at 2% buffer B for 6 min (total 220 min, buffer A is 0.1% v/v formic acid in water, buffer B is 0.1% v/v formic acid in acetonitrile). The eluted peptides were analyzed with a Thermo Fisher Scientific Orbitrap Fusion Lumos mass spectrometer. The spectrometer cycle time was 3 s and a nano-LC electrospray ionization source was used with an applied voltage of 2000 V. The data were collected using Xcalibur (v.4.1.50). The MS1 scan range was set to 375 to 1500 m/z, with a maximum injection time of 50 ms (dynamic exclusion enabled, repeat count 1, duration 20 s). The resolution was set to 120,000 with an automatic gain control value of 1 × 10⁶ions. The peptides selected for MS2 analysis were isolated with the quadrupole (isolation window 1.6 m/z) and fragmented using collision-induced dissociation (30% collision energy), and the resultant fragments were detected using the ion trap (automatic gain control 1.8 × 10⁴, maximum injection time 120 ms). The MS3 spectra were generated through high-energy collision-induced dissociation (65% collision energy). Synchronous precursor selection was used to isolate up to 10 MS2 ions for MS3-based quantitation.

Proteomic data analysis

Proteomic analysis was performed with using the SEQUEST HT algorithm in the Proteome Discoverer 3.0 software package (Thermo Fisher Scientific). Fragment tolerances were set to 0.6 Da, and precursor mass tolerances set to 10 ppm with one missed cleavage site allowed. Carbamidomethyl (C, +57.02146) and TMT-tag (K and N-terminal, +229.1629) were specified as static modifications while oxidation (M, +15.994915) and optionally trifluoroacetate (K, +95.9823) were specified as variable modifications. Spectra were searched against the Mus musculus protein database (UniProt, 17,158 sequences) using a false discovery rate of 1% (Percolator). MS3 peptide quantitation was performed with a mass tolerance of 20 ppm. Protein abundances were normalized to the median abundance of each channel. Identified proteins were required to have at least two unique peptides. TMT ratios obtained by Proteome Discoverer were log₂-transformed, and p-values were calculated via Student’s two-tailed t tests with three biological replicates. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository⁷⁵ with the dataset identifier PXD057035. Open search was performed using the MSFragger (4.0) and Philosopher (5.1.0) software packages running under FragPipe (20.0), using the ‘Open’ workflow with mass range of −150 to 500 Da^76–79.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Conceptualization: L.J.L. and M.R.G. Methodology: W.T., A.S.B., R.M., K.M., J.A.R.G., J.J.Z., C.G.P., A.S., L.J.L., and M.R.G. Investigation: W.T., A.S.B., R.M., K.M., J.A.R.G., J.J.Z., L.C., A.F.M.P., and L.J.L. Visualization: W.T., A.S.B., J.A.R.G., J.J.Z., and L.J.L. Funding acquisition: A.S. and M.R.G. Project administration: C.G.P., A.S., L.J.L., and M.R.G. Supervision: C.G.P., A.S., L.J.L., and M.R.G. Writing – original draft: W.T., L.J.L., and M.R.G. Writing – review & editing: all authors.

Peer review

Peer review information

Nature Communications thanks Anton Gisterå, Susanna Hofmann, Xiong Su and the other anonymous, reviewer for their contribution to the peer review of this work. A peer review file is available.

Funding

National Institutes of Health grant R01HL118114 (MRG) National Institutes of Health grant R01GM102491 (AS) The Skaggs Institute of Chemical Biology (MRG) Ferring Foundation (AS).

Data availability

The data generated in this study are available in the Supplementary Information/Source Data file. Mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD057035. Untargeted metabolomics data for this study is available for download from the National Metabolomics Data Repository with the Study ID ST003595 at https://doi.org/10.21228/M88R7M. The RNA-seq data for this study is available for download from Bioproject at accession # PRJNA917696, and 16S rRNA sequencing reads have been deposited in MG-RAST under accession MGP93528, MGP93529 and MGP93586. Source data are provided with this paper.

Competing interests

Scripps Research has filed a provisional patent covering novel therapeutic uses of TFA and structural analogs (application no. 63/763,336), with M.R.G., L.J.L., W.T., A.S.B., J.A.R.G., and J.J.Z. listed as inventors. The remaining authors declare that they have no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77341-2.