Adjunct Fenofibrate Up‐regulates Bile Acid Glucuronidation and Improves Treatment Response For Patients With Cholestasis

Abstract

Accumulation of cytotoxic bile acids (BAs) during cholestasis can result in liver failure. Glucuronidation, a phase II metabolism pathway responsible for BA detoxification, is regulated by peroxisome proliferator–activated receptor alpha (PPARα). This study investigates the efficacy of adjunct fenofibrate therapy to up‐regulate BA‐glucuronidation and reduce serum BA toxicity during cholestasis. Adult patients with primary biliary cholangitis (PBC, n = 32) and primary sclerosing cholangitis (PSC, n = 23), who experienced an incomplete response while receiving ursodiol monotherapy (13‐15 mg/kg/day), defined as serum alkaline phosphatase (ALP) ≥ 1.5 times the upper limit of normal, received additional fenofibrate (145‐160 mg/day) as standard of care. Serum BA and BA‐glucuronide concentrations were measured by liquid chromatography–mass spectrometry. Combination therapy with fenofibrate significantly decreased elevated serum ALP (−76%, P < 0.001), aspartate transaminase, alanine aminotransferase, bilirubin, total serum BAs (−54%), and increased serum BA‐glucuronides (+2.1‐fold, P < 0.01) versus ursodiol monotherapy. The major serum BA‐glucuronides that were favorably altered following adjunct fenofibrate include hyodeoxycholic acid–6G (+3.7‐fold, P < 0.01), hyocholic acid–6G (+2.6‐fold, P < 0.05), chenodeoxycholic acid (CDCA)–3G (−36%), and lithocholic acid (LCA)–3G (−42%) versus ursodiol monotherapy. Fenofibrate also up‐regulated the expression of uridine 5′‐diphospho‐glucuronosyltransferases and multidrug resistance–associated protein 3 messenger RNA in primary human hepatocytes. Pearson’s correlation coefficients identified strong associations between serum ALP and metabolic ratios of CDCA‐3G (r² = 0.62, P < 0.0001), deoxycholic acid (DCA)‐3G (r² = 0.48, P < 0.0001), and LCA‐3G (r² = 0.40, P < 0.001), in ursodiol monotherapy versus control. Receiver operating characteristic analysis identified serum BA‐glucuronides as measures of response to therapy. Conclusion: Fenofibrate favorably alters major serum BA‐glucuronides, which correlate with reduced serum ALP levels and improved outcomes. A PPARα‐mediated anti‐cholestatic mechanism is involved in detoxifying serum BAs in patients with PBC and PSC who have an incomplete response on ursodiol monotherapy and receive adjunct fenofibrate. Serum BA‐glucuronides may serve as a noninvasive measure of treatment response in PBC and PSC.

Combination therapy with fenofibrate favorably alters BA‐glucuronidation as a PPARa‐mediated mechanism to reduce elevated serum BAs in patients with PBC and PSC who are incomplete responders to Ursodiol monotherapy. This study also identifies serum BA‐glucuronides and their respective metabolic ratios as indicators of response to therapy. Serum BA‐glucuronides may serve as mechanistically relevant secondary endpoints in clinical trials for PBC and PSC with anti‐cholestatic agents.

Abbreviations

ALP

alkaline phosphatase

ALT

alanine aminotransferase

ANOVA

analysis of variance

AST

aspartate aminotransferase

AUC

area under the curve

BA

bile acid

CA

cholic acid

CDCA

chenodeoxycholic acid

Combo

combination

CYP

cytochrome P450

DCA

deoxycholic acid

HCA

hyocholic acid

HDCA

hyodeoxycholic acid

LCA

lithocholic acid

Mono

monotherapy

mRNA

messenger RNA

MRP

multidrug resistance–associated protein

PBC

primary biliary cholangitis

PC

principal component

PCA

principal component analysis

PCR

polymerase chain reaction

PPARα

peroxisome proliferator‐activated receptor

PSC

primary sclerosing cholangitis

ROC

receiver operating characteristic

UGT

uridine 5′‐diphospho‐glucuronosyltransferase

ULN

upper limit of normal

Cholestatic liver diseases, such as primary biliary cholangitis (PBC) and primary sclerosing cholangitis (PSC), result in the intracellular accumulation of toxic bile constituents, notably bile acids (BAs). Reducing BA toxicity is a primary goal in treating PBC and PSC. BA‐glucuronidation is a key step in BA detoxification and a major product of the phase II pathway of drug metabolism.^{( 1 )} BA‐glucuronidation conjugates glucuronic acid to BAs at the 3/6‐hydroxyl or 24‐carboxyl BA steroid nucleus, to produce corresponding ether/ester or acyl glucuronides, and increase their hydrophilicity for renal excretion.^{( 2 )} The human UDP‐glucuronosyltransferase (UGT) family of enzymes catalyze glucuronidation reactions,^{( 3 )} predominantly through genes belonging to the UGT1A and 2B families.^{( 4} , ⁵ , ^{6 )} Remarkably, only four UGT enzymes are responsible for hepatic BA‐glucuronidation, specifically UGT1A3, UGT1A4, UGT2B4 and UGT2B7, of which the first three are regulated by the nuclear peroxisome proliferator–activated receptor alpha (PPARα).^{( 4} , ⁵ , ^{7 )} In addition, BA‐glucuronides are substrates for the multidrug resistance–associated proteins (MRPs), including MRP3, an efflux transporter located on the sinusoidal membrane of hepatocytes and involved in the hepatic excretion of BA‐glucuronides into systemic circulation for subsequent urinary excretion.^{( 8 )}

Fenofibrate, a Food and Drug Administration (FDA)–approved cholesterol‐lowering fibrate and PPARα agonist, offers potential therapeutic benefit for patients with PBC^{( 9} , ^{10 )} and PSC^{( 11 )} who are incomplete responders to ursodiol monotherapy. We previously demonstrated that PPARα activation by fenofibrate transcriptionally activates the hepatic transporter multidrug resistance protein (MDR)‐3.^{( 12 )} MDR3 is critically involved in biliary phosphatidylcholine secretion, which is needed to inactivate the toxic detergent action of BAs and prevent damage to epithelial cells lining the bile duct epithelia. Furthermore, the therapeutic efficacy of combination treatment with fenofibrate and ursodiol in patients with PBC and PSC was shown by normalized biochemical markers of cholestasis and reduced toxicity of serum BA and composition, particularly 7α‐hydroxy‐4‐cholesten‐3‐one (C4), the BA precursor, as well as reduced primary and secondary BAs, toward levels observed in healthy controls.^{( 13 )} Also, fenofibrate has low reported adverse events in patients with PBC.^{( 14 )}

Although glucuronidation plays less of a major role in BA metabolism under homeostatic conditions,^{( 15 )} this phase II conjugation pathway becomes an important therapeutic pathway during cholestatic conditions to reduce BA toxicity.^{( 16} , ^{17 )} Individual and total BAs have been examined for their potential as prognostic markers in acute liver injury^{( 18} , ¹⁹ , ²⁰ , ²¹ , ^{22 )}; however, BA‐glucuronides have yet to be evaluated in chronic cholestatic liver injury. This study investigates the efficacy of fenofibrate to activate PPARα‐mediated BA glucuronidation, as a therapeutic mechanism to reduce the toxicity of serum BAs during adult chronic cholestatic liver diseases. In addition, this study examines serum BA and BA‐glucuronides as secondary measures of treatment response for patients experiencing an incomplete response to ursodiol monotherapy and for those receiving combination therapy with fenofibrate.

Materials and Methods

Chemicals are listed in the Supporting Materials and Methods.

Study Population

This retrospective study examines the biochemical responses and serum BA and BA‐glucuronide concentrations of patients with PBC and PSC who were treated with either (1) ursodiol monotherapy (13‐15 mg/kg/day) for at least 6 months and continued to experience an incomplete biochemical response or (2) combination treatment with ursodiol and off‐label fenofibrate (145‐160 mg/day) after having an incomplete response while receiving ursodiol monotherapy for at least 6 months before the addition of fenofibrate. An incomplete biochemical response is defined as serum alkaline phosphatase (ALP) levels ≥ 1.5 times the upper limit of normal (ULN). Treatments were part of the patients’ routine clinical care at the Yale Autoimmune and Cholestatic Liver Diseases Program (New Haven, CT), as previously described.^{( 13 )} Patients provided written consent to have their blood samples de‐identified and stored for research purposes at the Yale Liver Center Clinical Registry (New Haven, CT). Ongoing registry sample collection has provided new samples included in the biochemical analyses (n = 5 PBC and n = 4 PSC; combination fenofibrate cohort). All samples were drawn during morning routine clinic visits, which is the only period when patients are seen, and sampling times did not vary between the patient groups. These de‐identified samples were determined to be exempt from the institutional review boards at Yale University (New Haven, CT) and the University of Rhode Island (Kingston, RI). The study protocol conforms to the ethical guidelines of the 1975 Declaration of Helsinki as reflected in a priori approval by the institutional review committee. Serum samples from healthy de‐identified adult donors (BioIVT, Westbury, NY) did not have any clinically diagnosed liver disease and serve as healthy controls.

Serum Liver Enzymes

Serum ALP, alanine aminotransferase (ALT), aspartate aminotransferase (AST), total bilirubin, and lipid panels were measured by standard clinical laboratories (Yale New Haven Hospital [New Haven, CT] or Quest Diagnostics [Secaucus, NJ]).

BA and BA‐Glucuronide Extraction and Liquid Chromatography Coupled With Tandem Mass Spectrometry

Serum BA, BA‐glucuronide extraction, and liquid chromatography–tandem mass spectrometry methodology are provided in the Supporting Materials and Methods. The total serum BAs include the sum of the following 17 compounds: cholic acid (CA), chenodeoxycholic acid (CDCA), deoxycholic acid (DCA), lithocholic acid (LCA), hyodeoxycholic acid (HDCA), hyocholic acid (HCA), ursodeoxycholic acid (UDCA), glycochenodeoxycholic acid (GCDCA), glycodeoxycholic acid (GDCA), glycoursodeoxycholic acid (GUDCA), glycolithocholic acid (GLCA), glycocholic acid (GCA), taurolithocholic acid (TLCA), taurocholic acid (TCA), taurochenodeoxycholic acid (TCDCA), taurodeoxycholic acid (TDCA), and tauroursodeoxycholic acid (TUDCA). The following BA‐glucuronides were measured: CDCA‐3G and CDCA‐24G, DCA‐3G and DCA‐24G, LCA‐3G and LCA‐24G, HDCA‐6G and HCDA‐24G, HCA‐6G and HCA‐24G, and CA‐24G. All metabolites are also listed in Supporting Table S1.

Cell Culture and Quantitative Real‐Time Polymerase Chain Reaction Analysis

Cryopreserved human hepatocytes (Sekisui Xenotech, LLC, Kansas City, KS) were maintained on collagen‐coated plates and cultured according to the manufacturer’s protocol in a 37°C, 5% CO₂ incubator. Hepatocytes were treated with DMSO (vehicle) or fenofibrate (50 µM) for 24 hours. Total RNA was extracted from hepatocytes using TRIzol (Invitrogen, Carlsbad, CA). Hepatic messenger RNA (mRNA) levels of PPARα, UGT1A1, 1A3, 1A4, 2B4, 2B7, CYP3A4, MRP2, MRP3, CYP4A11 (positive control for PPARα), and PPIA were analyzed using Taqman probes (Applied Biosystems, Foster City, CA). Quantitative real‐time polymerase chain reaction (PCR) analysis was performed using a Roche LightCycler Detection System (Basel, Switzerland).

Data Analysis

Results are reported as mean ± SEM. Serum liver enzymes and lipids were analyzed using two‐tailed unpaired t‐tests with Welch’s correction, and quantitative real‐time PCR data were analyzed using two‐tailed unpaired Student t test (GraphPad Prism 9). Unless otherwise stated, serum BA and BA‐glucuronide concentrations were log₂‐transformed to obtain normal distribution before statistical analysis,^{( 23 )} then analyzed using one‐way analysis of variance (ANOVA) (R Studio, v.4.0.4). Further analyses of data within cohorts were performed using a linear mixed model with subject ID as the blocking factor (limma package v.3.38.3^{( 24 )}). Details are provided in Supporting Table S1. Strip charts were made for each metabolite in R using the ggplot function of the ggplot2 package (v. 3.3.2). Indicated P values between the groups were derived from the contrasts for the linear mixed model. Means of enzyme concentrations were compared in R using ANOVA with treatment (control, monotherapy, and combination) as the fixed effect, and multiple comparisons were conducted using the Tukey method (multcomp package in R). Metabolic ratios were analyzed by the Kruskal‐Wallis test with Dunn’s multiple comparisons. Within treatment groups, values below the limit of detection were set to the limit of detection. Pearson’s correlation coefficient (r²) was used to measure the association of serum BAs with serum ALP values. Extreme outliers defined as values ≥ 3.0 times the interquartile range either above Q3 or below Q1 were excluded from analysis. The ability of serum BAs and BA‐glucuronides to indicate the response to therapy for each patient cohort (e.g., ursodiol monotherapy nonresponders or combination therapy with fenofibrate) was measured using receiver operating characteristic (ROC) analysis curves, as implemented in the plotROC package, Rstudio.^{( 25 )} The binary variable for ROC curves was defined as 0 for the healthy control state and 1 for either the mono or combination treatment state; P values were calculated for the area under the curve (AUC) for each ROC curve using the Mann‐Whitney test in Rstudio. Principal component analysis (PCA) was conducted using R prcomp function and plotted with the ggbiplot package (v.0.55). Heatmaps were constructed using the heatmap.2 function of the gplots package (3.0.4), and values were scaled by row (metabolite).

Results

Clinical Characteristics and Serum Markers of Liver Injury

Serum was obtained from patients with PBC (n = 32; monotherapy: n = 16 and combination: n = 16) and PSC (n = 23; monotherapy: n = 11 and combination: n = 12) before and for a median of 16 months (range: 1‐64 months) after the addition of fenofibrate to ursodiol (duration of fenofibrate therapy is based on the time from initiating therapy to when sample analyses occurred). One patient with PSC had cirrhosis; no other patients had advanced fibrosis. Patient demographics are presented in Table 1A. Supporting Table S2 provides the clinical characteristics and the associated serum total BAs, BA‐glucuronides, and ALP based on fibrosis scores obtained at the time of diagnosis; these scores provide the range in severity of disease for the patients with an incomplete response and receiving ursodiol monotherapy. In the combination therapy cohorts, characteristics obtained before the addition of fenofibrate were available for a subset of patients with fibrosis scores of 1 (n = 1, PBC), 2 (n = 1, PSC), and 3 (n = 4, 2 PBC and 2 PSC), which represent a similar range in the severity of underlying disease compared with the ursodiol monotherapy cohort.

TABLE 1.

Demographics of All Patients Included in the Study (A) and Serum Liver Biochemistries of Patients With PBC and PSC (B)

	Controls (n = 33)	PBC		PSC
		Monotherapy (n = 16)	Combination (n = 16)	Monotherapy (n = 11)	Combination (n = 12)
A. Demographics
Age, median (range, years)	41 (24‐61)	54 (44‐70)	60 (48‐72)	42 (24‐66)	45 (33‐76)
Male	18 (55%)	0	0	10 (91%)	5 (42%)
Female	15 (45%)	16 (100%)	16 (100%)	1 (9%)	7 (58%)
White	10 (30%)	11 (69%)	10 (63%)	8 (73%)	10 (84%)
Black	17 (52%)	0	0	3 (27%)	1 (8%)
Hispanic	6 (18%)	3 (19%)	4 (25%)	0	0
Unknown	0	2 (12%)	2 (12%)	0	1 (8%)
B. Serum biochemistries
ALP (U/L)	52 ± 2.1	391 ± 95	97 ± 10**	621 ± 135	147 ± 32**
ALT (U/L)	6‐34#	74 ± 14	31 ± 3**	110 ± 14	51 ± 11**
AST (U/L)	11‐33#	62 ± 10	37 ± 4*	104 ± 17	51 ± 11*
Total bilirubin (mg/dL)	<1.2#	0.9 ± 0.3	0.6 ± 0.1	4 ± 1.4	0.4 ± 0*
Creatinine (mg/dL)	0.7 ± 0	0.6 ± 0.1	0.7 ± 0.1	0.6 ± 0.1	0.7 ± 0

Serum samples from healthy de‐identified adult donors (BioIVT) serve as healthy controls. ^#Reference ranges obtained from Yale New Haven Hospital Clinical laboratory. Data represent mean ± SEM, analyzed by pairwise t‐tests with Welch’s correction.

^* P < 0.05.

^** P < 0.01.

Combination therapy with fenofibrate significantly reduced serum ALP levels by 76% versus ursodiol monotherapy, and normalized levels for 82% of patients. Serum ALT and AST levels were also normalized after combination treatment with fenofibrate, decreasing values by 55% and 46% versus ursodiol monotherapy, respectively. In addition, total serum bilirubin levels decreased by 76% by fenofibrate treatment (Table 1B). These results demonstrate the efficacy of combination treatment with fenofibrate and ursodiol in reducing elevated liver biochemistries in patients with cholestasis who have a subtherapeutic response to ursodiol monotherapy.

Paired analysis was performed in a subset of patients (n = 7: PBC = 5 and PSC = 2) who received ursodiol monotherapy ≥6 months and experienced an incomplete biochemical response (e.g., ALP ≥ 1.5 times the ULN [baseline]), followed by blood sampling after receiving combination therapy with fenofibrate. As given in Table 2, reductions in all serum biochemistries occurred following the addition of fenofibrate, most notably in serum ALP. Indeed, the addition of fenofibrate reduces total serum BA levels, which correspond with increased serum BA‐glucuronide concentrations.

TABLE 2.

Paired Analysis of Serum Biochemical Markers of Liver Injury, Total Serum BAs, and BA‐Glucuronides in a Subset of Patients (n = 7, PBC = 5, and PSC = 2)

No. of Patients (Treatment)	ALT (U/L)	AST (U/L)	Total Bili (mg/dL)	ALP (U/L) (% change)	Total Serum BA (µM)	Serum BA Gluc (nM) (fold change)	Feno Duration (months)
PBC
1 mono	21	32	0.8	271	185	763	—
1 combo	27	31	0.6	96 [−65%]	70	1420 [+1.9]	29
2 mono	25	26	0.26	214	9.2	280	—
2 combo	16	25	0.16	91 [−57%]	3.3	670 [+2.4]	14
2 combo	19	24	0.3	88 [−59%]	30.3	1,457 [+5.2]	49
3 mono	129	155	3.2	1,679	23.6	1,684	—
3 combo	47	78	0.8	77 [−95%]	39.1	1,606	40
3 combo	44	68	1.2	94 [−94%]	51	4,742 [+2.8]	53
4 mono	51	48	0.5	163	45.4	597	—
4 combo	24	27	0.5	73 [−55%]	6.4	1,250 [+2.1]	44
5 mono	18	25	0.5	156	74.3	309	—
5 combo	28	36	0.6	124 [−21%]	19.1	1,456 [+4.7]	7
PSC
6 mono	40	24	0.48	155	8.1	79	—
6 combo	26	17	0.51	92 [−41%]	21.5	635 [+8.1]	1
6 combo	33	25	0.5	114 [−26%]	15.8	229 [+2.9]	15
7 mono	100	58	0.61	307	11	404	—
7 combo	35	31	0.52	147 [−52%]	8	403 —	7
7 combo	30	31	0.6	117 [−62%]	7.7	732 [+1.8]	19

Abbreviations: Bili, bilirubin; combo, combination with fenofibrate; Feno, fenofibrate; Gluc, glucuronide; mono, ursodiol monotherapy.

The FDA‐approved indication of fenofibrate is to reduce elevated cholesterol. Although this was not the indication for the addition of fenofibrate in these patients, treatment with fenofibrate reduced serum lipids in a subset of patients with available lipid levels for analysis, including total cholesterol (−20%), low‐density lipoprotein (−38%, P = 0.07), triglycerides (−34%, P = 0.06), and increased high‐density lipoprotein (+23%) versus ursodiol monotherapy, respectively (Supporting Table S3), confirming the efficacy of fenofibrate in reducing serum lipids and providing further evidence of a drug effect.

mRNA Expression in Human Hepatocytes

To determine the activation of PPARα and its regulated genes involved in BA‐glucuronidation, human hepatocytes were treated with fenofibrate (50 µM), and gene expression was measured by quantitative real‐time PCR. Treatment with fenofibrate significantly up‐regulated the mRNA expression of PPARα, UGT1A3, UGT1A4, CYP3A4, MRP2, MRP3, and CYP4A11 versus DMSO (vehicle) (Fig. 1A). No effect on UGT2B7 expression was expected. These in vitro data confirm that fenofibrate/PPARα‐mediated activation of UGTs and MRPs can be involved in BA‐glucuronidation and efflux out of the liver for subsequent excretion and detoxification.

FIG. 1.

(A) Quantitative real‐time PCR analysis shows that fenofibrate (50 µM) treatment for 24 hours up‐regulates PPARα, UGT1A1, UGT1A3, UGT1A4, UGT2B4, CYP3A4, MRP2, MRP3, and CYP4A11 mRNA expression in cryopreserved hepatocytes compared with DMSO (vehicle control). Relative mRNA expression was calculated using the ∆∆Ct method, normalized to PPIA and expressed as a fold‐change over DMSO. Data are expressed as mean ± SEM (n = 3‐4 individual cases of hepatocytes), analyzed by two‐tailed unpaired Student t test. *P < 0.05, **P < 0.01, and ****P < 0.0001 versus DMSO. Combination treatment with fenofibrate decreases total serum BAs (B) and increases total serum BA‐glucuronides (C). Log₂‐transformed data were analyzed by one‐way ANOVA. **P < 0.01 and ****P < 0.0001. Total serum BAs and BA‐glucuronides are defined in the Supporting Materials and Methods.

Total Serum BA and BA‐Glucuronide Concentrations

The total serum BAs levels observed in patients on ursodiol monotherapy in this study are comparable to serum BA concentrations reported in untreated patients with PBC and PSC.^{( 26 )} Combination therapy with fenofibrate reduced total serum BA concentrations by 54% (28 ± 4 vs. 61 ± 14 µM) compared with ursodiol monotherapy and trended toward levels found in healthy controls (5 ± 0.8 µM) (Fig. 1B).

To investigate the role of the UDP‐glucuronyl transferase metabolic pathway in reducing the toxicity of serum BAs, we next quantified patient serum BA‐glucuronide concentrations. Combination therapy with fenofibrate increased total serum BA‐glucuronides in PBC and PSC by about 6.0‐fold (1,064 ± 166 vs. 176 ± 26 nM, P < 0.0001) and 2.1‐fold (1,064 ± 166 vs. 505 ± 77 nM, P < 0.01) versus healthy controls and ursodiol monotherapy, respectively (Fig. 1C). Serum BA‐glucuronide concentrations were also elevated by 2.8‐fold (505 ± 77 vs. 176 ± 26 nM, P < 0.0001) in patients on ursodiol monotherapy versus healthy controls, albeit to a much lower degree than by the addition of fenofibrate. Despite this increase in BA‐glucuronide levels in patients on ursodiol monotherapy, their total serum BA concentrations remained significantly elevated compared with controls (61 ± 14 µM vs. 5 ± 0.8 µM, P < 0.0001), suggesting that the ursodiol‐mediated increases in BA‐glucuronides were not sufficient to reduce the toxicity of the total serum BA pool. In contrast, the reduction of total serum BAs along with PPARα activation demonstrate that the addition of fenofibrate reduces the toxicity of serum BAs through PPARα‐mediated BA‐glucuronidation.

Therapeutic Impact of Fenofibrate on Serum BA‐Glucuronide Concentrations and Composition

We next assessed the impact of fenofibrate on the individual BA‐glucuronide profiles. Of the 11 metabolites studied, we found that 99% of the serum BA‐glucuronides in the healthy controls consisted of five BA‐glucuronides: HCA‐6G (47%), CDCA‐3G (36%), DCA‐3G (6%), HDCA‐6G (10%), and LCA‐3G (1%). In PBC and PSC, the addition of fenofibrate increased serum HCA‐6G and HDCA‐6G concentrations by 2.6‐fold and 3.7‐fold versus ursodiol alone (P < 0.01 and P < 0.05), respectively (Fig. 2A,B). Fenofibrate therapy also increased the contribution of HCA‐6G in the serum glucuronide pool from 47% in healthy controls to 55% and 66% in patients with PBC and PSC, respectively (Fig. 3). Similarly, fenofibrate treatment increased the contribution of HDCA‐6G from 10% of the serum BA‐glucuronides in controls to 35% in patients with PBC and 20% in patients with PSC (Fig. 3). Finally, a corresponding decrease in serum HCA‐24G (−76%) was observed, whereas HDCA‐24G remained unchanged (Supporting Fig. S1A,B). These findings confirm that glucuronidation is preferred at the 6‐hydroxyl^{( 17 )} over the 24‐hydroxyl position.

FIG. 2.

Combination treatment with fenofibrate alters serum BA‐glucuronides in patients with PBC and PSC. Scatter plots show individual patient serum concentrations of HCA‐6G (A), HDCA‐6G (B) (B), CDCA‐3G (C), LCA‐3G (D), and DCA‐3G (E). Log₂‐transformed data were analyzed by one‐way ANOVA. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Abbreviations: Combo, combination; and Mono, monotherapy.

FIG. 3.

Combination treatment with fenofibrate shifts the percentage and composition of total 3G and 6G‐glucuronides (CDCA‐3G, DCA‐3G, LCA‐3G, HCA‐6G, and HDCA‐6G) in patients with PBC and PSC, closer to values observed in control subjects: healthy controls (A), PBC monotherapy (B), PBC combined (C), PSC monotherapy (D), and PSC combined (E).

When HepG2 cells are treated with CDCA‐3G, DCA‐3G, and LCA‐3G, they exhibit apoptosis and necrosis,^{( 27 )} confirming the therapeutic importance of glucuronidation of these particular compounds. In this study, ursodiol monotherapy significantly increased serum concentrations of CDCA‐3G (+2.2‐fold, P < 0.05) and LCA‐3G (+6.2‐fold, P < 0.001) levels versus healthy controls, which were markedly reduced by the addition of fenofibrate (−36% and −42%, respectively) versus ursodiol monotherapy (Fig. 2C,D). Notable reductions in the contribution of CDCA‐3G and LCA‐3G to the serum BA‐glucuronide pool were also found, particularly in PSC, decreasing from 52% to 13% and 5% to 0.3%, respectively, in the fenofibrate‐treated cohort relative to ursodiol monotherapy (Fig. 3). Fenofibrate did not significantly change DCA‐3G serum concentrations, relative to that of ursodiol monotherapy (Fig. 2E). CDCA, DCA, and LCA are glucuronidated at the C3 position by UGT2B7 during homeostatic conditions, and CDCA undergoes glucuronidation through PPARα‐induced UGT1A4 when exposed to fenofibrate.^{( 4 )} Collectively, the shift away from these three C3 glucuronides and toward C24 glucuronidation (Supporting Fig. S1) likely reduces the toxic impact of these metabolites. Moreover, this latter finding indicates that reducing elevated serum C3 glucuronides in PBC and PSC by fenofibrate is hepatoprotective.

Metabolic Ratios of Serum BA‐Glucuronides and Correlations With Serum ALP

To further investigate whether changes in the serum BA‐glucuronides reflect the amount of BA substrate available for glucuronidation (e.g., parent BA) or a change in metabolism and transport, we calculated the metabolic ratio for each serum BA‐glucuronide compared with their unconjugated parent BA (BA‐glucuronide/unconjugated parent BA). The most profound changes occurred among the 3‐carboxyl BA‐glucuronides. Significant reductions in serum CDCA‐3G decreased the metabolic ratio of CDCA‐3G/CDCA by 87% in fenofibrate‐treated patients with PBC and PSC versus those receiving ursodiol alone (Fig. 4A). Also, fenofibrate combination treatment reduced serum levels of DCA, resulting in a 77% reduction in the metabolic ratio of DCA‐3G/DCA and decreased LCA‐3G serum concentrations, which reduced the metabolic ratio of LCA‐3G/LCA by 81%, relative to ursodiol monotherapy, and nearly restoring levels for all BA‐3Gs to those found in healthy controls (Fig. 3B,C). Additionally, increased metabolic ratios of serum HCA‐6G and HDCA‐6G support the notion of enhanced BA‐glucuronide efflux out of the liver and into circulation through up‐regulated MRP3 by fenofibrate (Fig. 4D,E). The increases in these metabolic ratios were due to increases in both the parent BA compound and BA‐glucuronide compound after the addition of fenofibrate treatment. Corresponding shifts in the metabolic ratios of BA‐24Gs were also detected (Supporting Fig. S2).

FIG. 4.

Combination treatment with fenofibrate alters metabolic ratios (BA‐glucuronide/parent BA) in patients with PBC and PSC versus ursodiol monotherapy: CDCA‐3G/CDCA (A), DCA‐3G/DCA (B), LCA‐3G/LCA (C), HCA‐6G/HCA (D), and HDCA‐6G/HDCA (E). Data were analyzed by the Kruskal‐Wallis test with Dunn’s multiple comparisons. *P < 0.05, **P < 0.01, and ***P < 0.001.

Correlation of Serum ALP and BA‐Glucuronide Metabolic Ratios

Pearson’s correlation coefficients (r²) between serum ALP and the concentrations of individual BAs, BA‐glucuronides, and their respective metabolic ratios were analyzed. The strongest correlations associated with elevated serum ALP levels were found in the patients receiving ursodiol monotherapy, particularly for serum CDCA‐3G/CDCA (r² = 0.62, P < 0.0001), DCA‐3G/DCA (r² = 0.48, P < 0.0001), and LCA‐3G/LCA (r² = 0.40, P < 0.001) (Fig. 5A). These results highlight the strong association of CDCA‐3G, DCA‐3G, and LCA‐3G metabolic ratios with an incomplete response to ursodiol monotherapy (i.e., serum ALP values > 1.5 times the ULN). Conversely, correlation coefficients obtained from the analysis of combined fenofibrate treatment revealed therapeutic and significant reductions in both ALP and BA‐glucuronide metabolic ratios by fenofibrate, similar to levels found in healthy control patients (Fig. 5A).

FIG. 5.

Pearson’s correlation coefficient of BA‐glucuronide metabolic ratios with ALP levels: CDCA‐3G/CDCA, DCA‐3G/DCA, and LCA‐3G/LCA (A). ROC analysis in ursodiol monotherapy versus control: CDCA‐3G and LCA‐3G (B), CDCA‐3G/CDCA and DCA‐3G/DCA ratios (C). ROC curve analyses in combined fenofibrate treatment versus control: HDCA‐6G and HCA‐6G (D) and HDCA‐6G/HDCA and HCA‐6G/HCA ratios (E). Data was log₂‐transformed before analysis.

Serum BA‐Glucuronides Indicate Treatment Response

We next analyzed the relationship between serum BA‐glucuronides and treatment response for the ursodiol monotherapy (incomplete responders) versus healthy controls by ROC curves. The AUC reveals that CDCA‐3G (AUC = 0.69, P = 0.01), LCA‐3G (AUC = 0.8, P < 0.0001), as well as the metabolic ratios of CDCA‐3G/CDCA (AUC = 0.69, P = 0.01) and DCA‐3G/DCA (AUC = 0.76, P < 0.001), respectively, are strong indicators of the incomplete response to ursodiol monotherapy (Fig. 5B,C). Further ROC curve analyses were performed on the ursodiol monotherapy cohort based on their fibrosis score (Supporting Fig. S3). Those with a fibrosis score of 1, 2, and 3 resulted with AUC values and significance proportional to the degree of their liver injury, including CDCA‐3G (AUC = 0.6, P = 0.63; AUC = 0.63, P = 0.35; and AUC = 0.72, P = 0.06; respectively) and LCA‐3G (AUC = 0.8, P = 0.9; AUC = 0.93, P < 0.001; and AUC = 0.82, P = 0.01; respectively). Similarly, the AUCs for the metabolic ratios of CDCA‐3G/CDCA (AUC = 0.47, P = 0.89; AUC = 0.7, P = 0.12; and AUC = 0.78, P = 0.02; respectively) and DCA‐3G/DCA (AUC = 0.39, P = 0.55; AUC = 0.85, P = 0.01; and AUC = 0.8, P = 0.01; respectively) demonstrate that these metabolites are strong indicators of the incomplete response to ursodiol monotherapy.

To further support the hypothesis that fenofibrate treatment improves BA‐glucuronidation and reduces the toxicity of serum BAs, ROC analysis was performed with the combined fenofibrate cohort versus healthy controls. HDCA‐6G (AUC = 0.79, P = 0.0001) and HCA‐6G (AUC = 0.93, P < 0.0001) and their metabolic ratios (AUC = 0.77, P < 0.001 and AUC = 0.67, P < 0.05, respectively) were identified as an indicator of the therapeutic response to combined fenofibrate treatment (Fig. 5D,E). Overall, these results support the use of serum BA‐6Gs as measures of the therapeutic efficacy of combined treatment with fenofibrate. Additional analyses of serum BA‐glucuronides in measuring the treatment response were assessed by ROC curves (Supporting Table S4).

BA‐Glucuronide Patterns and Variability

Finally, to identify which metabolites are responsible for the variation in BA‐glucuronide concentrations between patient cohorts, we performed principal component (PC) and heatmap analyses. Approximately 68% of the variability in BA‐glucuronide levels between patient cohorts are accounted for by three PCs. The first PC accounts for about 36% of the variability among all patient cohorts, which is driven by the healthy control groups having significantly lower concentrations of all BA‐glucuronides (Fig. 6A). The second PC accounts for 20% of the variability, in which the major decreases in LCA‐3G and CDCA‐3G separated the combination treatment cohorts away from the ursodiol monotherapy cohorts, creating a clear distinction between the treatment groups (Fig. 6A). In the third PC, 12% of the variability is accounted for by the result of DCA‐3G, which did not show significant differences between monotherapy and combination cohorts, (+1.8‐fold vs. +1.7‐fold, respectively), compared with healthy controls (Fig. 6A). When plotted on the heatmap, the variability of the metabolites results partially from the higher concentrations of the BA‐glucuronides in both the PBC and PSC combination fenofibrate therapy groups in relation to ursodiol mono, thereby creating separation between those receiving ursodiol monotherapy and patients receiving combination treatment with fenofibrate (Fig. 6B).

FIG. 6.

(A) Two PCA score plots of individual BA‐glucuronides from the serum of healthy controls, ursodiol monotherapy, and combination fenofibrate‐treated patients to account for approximately 68% of variability. Data was log₂‐transformed before PCA analysis and grouped as Control, PBC Mono, PBC Combo, PSC Mono, and PSC Combo. (B) Heatmap analysis of individual glucuronides from the serum of healthy subjects, ursodiol monotherapy, and combination fenofibrate‐treated patients. Data were log₂‐transformed before constructing the heatmap. Rows represent BA‐glucuronides. Columns represent individual samples and are clustered based on cohort (Control, PBC Mono, PBC Combo, PSC Mono, and PSC Combo).

Discussion

In this study, we hypothesized that fenofibrate exerts its therapeutic role in cholestasis in part through PPARα‐mediated UGT up‐regulation to promote BA glucuronidation in patients with PBC and PSC who have remained inadequately responsive to ursodiol monotherapy. We also examined whether BA‐glucuronides have the potential to serve as measures of efficacy to mono or combination therapy. During chronic cholestatic liver diseases, BAs accumulate within the liver and result in inflammation, fibrosis, and disease progression.^{( 28 )} This pathophysiologic process emphasizes the importance of reducing serum BAs as a therapeutic target for patients with PBC and PSC, especially those with an incomplete response to ursodiol monotherapy. In the past decade, clinical studies have demonstrated the efficacy of fibrates, particularly fenofibrate^{( 9} , ^{11 )} and bezafibrate,^{( 29 )} as therapeutic alternatives for patients with PBC or PSC who have an incomplete response to ursodiol. We^{( 13 )} and others^{( 14 )} have shown fenofibrate to be safe and well‐tolerated in PBC. although much of these data demonstrate that fibrates lower elevated serum liver enzymes during cholestasis, very little data exist on their effect on BA metabolism, specifically BA‐glucuronidation. Here, our data show that combination therapy with fenofibrate significantly increased the serum concentrations of BA‐glucuronides in patients with PBC and PSC compared with ursodiol monotherapy and healthy controls, in addition to normalizing serum ALP in 94% of patients with PBC and 67% in patients with PSC versus ursodiol monotherapy. We also found that BA‐glucuronides comprise 4% of the serum BAs in healthy controls, whereas after monotherapy with ursodiol, serum BA‐glucuronides comprise less than 1% of the total serum BA pool, likely due to the significant increases in total BA concentrations and progressive liver disease in patients with an incomplete response to ursodiol monotherapy. In stark contrast, combination treatment with fenofibrate enhanced BA glucuronidation and restored serum BA‐glucuronide levels to normal levels, comprising approximately 4% of the total BA pool in PBC and PSC. The PCA and heatmap plots show a clear separation between the healthy and treated patients along PC1, which is driven by the general increase in BA‐glucuronide concentrations in the treated patients. Additionally, the major serum BA‐glucuronides were favorably altered by combination treatment with fenofibrate, including elevated serum HDCA‐6G (P < 0.05) and HCA‐6G (P < 0.01) levels, which represent most of the BA‐glucuronide pool, compared with ursodiol monotherapy. Both of these BA‐glucuronides undergo glucuronidation by UGT2B4, which is transcriptionally regulated by PPARα.^{( 5 )} The data presented herein reveal a mechanism in which fenofibrate increases serum BA‐glucuronidation and reduces BA toxicity during cholestatic liver diseases in patients who have an incomplete response to ursodiol monotherapy.

Among secondary BAs, DCA and LCA are the most toxic. Elevated concentrations of LCA induce cholestatic liver damage^{( 30 )} and hepatotoxicity,^{( 31 )} highlighting the importance of reducing LCA serum concentrations. However, LCA‐3G exhibits cholestatic^{( 32 )} and apoptotic properties.^{( 27 )} Indeed, decreasing LCA‐3G levels and promoting LCA detoxification through additional pathways, such as sulfation, is associated with improved markers of disease for patients with PBC and PSC who are incomplete responders to ursodiol monotherapy.^{( 13 )} Hydroxylation is also an alternative route for LCA detoxification,^{( 33 )} where it serves as a substrate for cytochrome P450 (CYP) 3A4–mediated metabolism to HDCA.^{( 34 )} Bezafibrate, a pan‐PPAR agonist, transcriptionally activates PXR and subsequently increases CYP3A4 mRNA expression in HepaRG cells.^{( 35 )} However, whether bezafibrate also increases LCA hydroxylation remains to be determined. In agreement with our current data showing that fenofibrate up‐regulates the CYP3A4 mRNA expression in human hepatocytes (Fig. 1A), CYP3A4 mRNA expression increased in primary human hepatocytes treated with the PPARα agonist Wy‐14,643,^{( 36 )} which was significantly decreased in response to silencing PPARα expression.^{( 37 )} Herein, the increase in serum HDCA parent concentrations in the combination fenofibrate cohorts allow for subsequent HDCA 6‐glucuronidation by UGT2B4 and UGT2B7,^{( 4} , ^{5 )} thereby enhancing its urinary excretion. These findings shed light on an additional mechanism of LCA detoxification regulated, in part, by PPARα.

All of the BA‐24Gs that we studied are formed by UGT1A3,^{( 4 )} in which the most significant changes following fenofibrate therapy occurred with LCA‐24G and DCA‐24G. Our findings agree with reports that PPARα activation increased UGT1A3 expression and activity, including the production of LCA‐24G.^{( 1} , ^{38 )} In fact, LCA‐24G is more actively produced compared with LCA‐3G in healthy human liver microsomes.^{( 17 )} In support of this action, a PPARα response element has been identified in human UGT1A3, ^{( 7 )} which is supported by our UGT1A3 mRNA data. Up‐regulation of UGT1A3 is also important for LCA 24‐glucuronidation to reduce the hydrophobic and cytotoxic properties of LCA. However, many acyl glucuronides, including serum CA‐24G (Supporting Fig. S2F), were found at low concentrations, likely because the carboxyl group at C24 remains intact until glycine or taurine conjugation occurs.^{( 15 )}

Activation of PPARα also increases the mRNA expression of MRP3,^{( 39} , ^{40 )} an efflux transporter responsible for secreting BA‐glucuronides into the systemic circulation for urinary excretion, which is likely to be involved in the BA detoxification process in the current study. In support of this notion, urinary levels of BA‐glucuronides were increased in patients with cholestasis during biliary stenting, suggesting that this detoxification pathway may serve as a protective measure to reduce toxic BAs during cholestasis.^{( 17 )} Although urinary BA‐glucuronides were not available for study, the up‐regulated MRP3 mRNA expression following fenofibrate treatment in human hepatocytes supports our hypothesis that PPARα therapy enhances the urinary excretion of BA‐glucuronides, ultimately to reduce BA toxicity during cholestasis.

In addition, combination treatment with fenofibrate decreased total bilirubin concentrations in serum by 35% and 89% in PBC and PSC, respectively. Bilirubin is conjugated to bilirubin diglucuronide by UGT1A1 and 1A4 to facilitate its biliary excretion through MRP2.^{( 34} , ⁴¹ , ⁴² , ^{43 )} In support of this pathway, our data show that fenofibrate also up‐regulates human UGT1A1, 1A4, and MRP2 mRNA expression, suggesting that activation of PPARα by fenofibrate also promotes bilirubin excretion through conjugation to glucuronic acid and excretion into bile through MRP2, thereby reducing elevated serum levels during cholestasis.

Serum and urinary total BAs and individual BAs have been evaluated as potential mechanistic biomarkers of liver injury.^{( 20} , ^{21 )} Recently, He et al. showed that CA and CDCA were predictive diagnostic markers in severe alcohol‐associated hepatitis.^{( 22 )} However, the role of serum BA‐glucuronides as secondary measures of treatment response in cholestatic liver injury remains unknown. In this study, Pearson correlation coefficients demonstrate that metabolic ratios of serum CDCA‐3G and LCA‐3G concentrations are strongly associated with elevated serum ALP levels in patients experiencing an incomplete response to ursodiol monotherapy. Similarly, PCA plots identified a strong separation between treatment groups as a result of the decreases found in LCA‐3G and CDCA‐3G after combined fenofibrate treatment in comparison to ursodiol monotherapy (Fig. 3A). Together with the ROC curve results, these findings support the possible role of serum BA and serum BA‐glucuronides as mechanistically relevant secondary endpoints in clinical trials for PBC and PSC with anti‐cholestatic agents.

Furthermore, the ROC curves generated from serum HCA‐6G and HDCA‐6G concentrations in combination fenofibrate group versus healthy controls (AUCs > 0.79, P < 0.0001) are consistent with the heatmap in which increases for HDCA‐6G and HCA‐6G are found in both PBC and PSC fenofibrate combination groups in comparison to ursodiol monotherapy and healthy controls. Importantly, HCDA‐6G and HCA‐6G are not cytotoxic in vitro.^{( 27 )} Together, these data demonstrate the importance of serum BA‐glucuronides, in addition to reduced serum ALP levels, as measures of response to therapy (e.g., incomplete response to ursodiol monotherapy versus therapeutic efficacy of adjunct fenofibrate treatment).

Toxic BA accumulation is a major concern for both patients with PBC and PSC; thus, both cholestatic patient cohorts have been included in our study. Data for each cohort are presented to demonstrate how combination fenofibrate therapy affects BA glucuronidation in each disease. Human BA‐glucuronides have been shown to have higher concentrations in urine than in serum.^{( 4} , ^{27 )} Nevertheless, favorable differences in reduced serum liver enzymes and increased serum BA‐glucuronide concentrations following the addition of fenofibrate remain evident, supporting the role of serum BA‐glucuronides as secondary markers of response to therapy for cholestatic liver injury. Our study provides compelling evidence that fenofibrate‐mediated BA‐glucuronidation is a mechanism to detoxify the toxic primary (e.g., CDCA, CA) and secondary BAs (e.g., DCA, LCA) during cholestasis. In particular, the serum concentrations of CDCA‐3G and LCA‐3G greatly decreased after combination treatment with fenofibrate (Fig. 2C,D) compared with ursodiol monotherapy and approached healthy control levels, thereby reducing their toxicity. These trends in concentrations changes are also supported by the correlation data and ROC curves.

In summary, this is a study of BA‐glucuronidation in patients with PBC and PSC receiving combination therapy with fenofibrate and ursodiol. The findings shed light on therapeutic mechanisms for future treatment options for patients with chronic cholestatic liver diseases who remain subtherapeutic while taking ursodiol monotherapy.

Data Availability

The data that support the findings of this study are available from the corresponding author upon reasonable request. Some data may not be made available because of privacy or ethical restrictions.

Supporting information

Fig S1‐S3

Table S1‐4

Supplementary Material