Hepatic Farnesoid X-Receptor Isoforms α2 and α4 Differentially Modulate Bile Salt and Lipoprotein Metabolism in Mice

The nuclear receptor FXR acts as an intracellular bile salt sensor that regulates synthesis and transport of bile salts within their enterohepatic circulation. In addition, FXR is involved in control of a variety of crucial metabolic pathways. Four FXR splice variants are known, i.e. FXRα1-4. Although these isoforms show differences in spatial and temporal expression patterns as well as in transcriptional activity, the physiological relevance hereof has remained elusive. We have evaluated specific roles of hepatic FXRα2 and FXRα4 by stably expressing these isoforms using liver-specific self-complementary adeno-associated viral vectors in total body FXR knock-out mice. The hepatic gene expression profile of the FXR knock-out mice was largely normalized by both isoforms. Yet, differential effects were also apparent; FXRα2 was more effective in reducing elevated HDL levels and transrepressed hepatic expression of Cyp8b1, the regulator of cholate synthesis. The latter coincided with a switch in hydrophobicity of the bile salt pool. Furthermore, FXRα2-transduction caused an increased neutral sterol excretion compared to FXRα4 without affecting intestinal cholesterol absorption. Our data show, for the first time, that hepatic FXRα2 and FXRα4 differentially modulate bile salt and lipoprotein metabolism in mice.


Introduction
Bile salts are synthesized from cholesterol by well-characterized biosynthetic pathways (see [1] for review). The neutral pathway starts with 7a-hydroxylation of cholesterol by cholesterol 7a-hydroxylase (Cyp7a1) and the acidic pathway is initiated by sterol 27-hydroxylase (Cyp27a1). Newly synthesized primary bile salts are cholate (CA) and chenodeoxycholate (CDCA) in humans, whereas in rodents CDCA is rapidly converted into the more hydrophilic muricholates (MCA). Sterol 12a-hydroxylase (Cyp8b1) is required for the biosynthesis of CA and determines the ratio CA to CDCA. The chemical diversity of bile salts is further expanded by conjugation to either taurine or glycine and by the actions of intestinal bacteria via deconjugation, oxidation of hydroxyl groups and dehydroxylation, creating a bile salt pool with specific physiochemical properties. In addition to their wellestablished functions in generation of bile formation by the liver and absorption of dietary fats and fat-soluble vitamins from the intestine, bile salts are now recognized to act as 'integrators of metabolism' [1,2].
Bile salts contribute to control of a broad range of metabolic pathways, including their own synthesis in liver and transport within the enterohepatic circulation, via activation of the Farnesoid X Receptor (FXRa/NR1H4, referred to as FXR) [3,4]. FXR, a member of the nuclear hormone receptor superfamily, has four splice variants in humans and rodents [5,6]. These isoforms arise from a single gene through alternative splicing of exon 5 and the use of distinct promoters that initiate transcription from either exon 1 or exon 3. The different promoters of the FXR gene initiate expression of either FXRa1 and FXRa2 or of FXRa3 and FXRa4 transcripts. FXR isoforms show differences in spatial and temporal expression as well as in transcriptional activity [6]. FXRa1 and FXRa3 transcripts contain four additional amino acids in the hinge domain, adjacent to the DNA-binding domain, which make these two isoforms less transcriptionally active compared to FXRa2 and FXRa4. In mice all four isoforms are abundantly expressed in liver, while FXRa3 and a4 are abundantly expressed in ileum, moderately in kidney and at low levels in stomach, duodenum, and jejunum. Furthermore, FXRa1 and a2 are moderately expressed in ileum and adrenal gland [6].
In the liver, bile salt-activated FXR suppresses, amongst others, Cyp7a1 expression, in a Short Heterodimer Partner(SHP/NR0B2)-and Liver Receptor Homolog-1 (LRH-1/NR5A2)-dependent manner [7,8]. Also Cyp8b1 and Cyp27a1 are regulated by FXR, in a SHP-and Hepatocyte Nuclear Factor-4a (HNF4A)-dependent manner [9,10]. In the past decade, FXR has emerged as an important regulator of lipid as well as glucose homeostasis and to have antiinflammatory properties [1]. Hence, FXR is considered a promising target for treatment of a number of metabolic and liver diseases. Treatment with synthetic FXR agonists showed improved insulin sensitivity in diabetic mouse models [11,12] and prevented formation of atherosclerotic plaques in atherogenic-prone mice [13]. Furthermore, administration of the semi-synthetic obeticholic acid (OCA) increased insulin sensitivity and reduced markers of liver inflammation and fibrosis in patients with type 2 diabetes and nonalcoholic fatty liver disease [14].
In light of the increasing evidence for the therapeutic potential of FXR agonists in the regulation of lipid metabolism, we addressed the specific roles of FXRa2 and FXRa4, the transcriptionally most active isoforms of FXR [5,6], on bile salt and lipid metabolism in vivo by means of a liver-specific FXRa2 or FXRa4 expressing mouse model. Our data revealed some distinct functions of FXRa2 and FXRa4 in control of plasma cholesterol levels; only hepatic FXRa2 expression could rescue the elevated plasma HDL levels of FXR KO mice. Moreover, FXRa2 and FXRa4 were found to differentially regulate Cyp8b1 expression in vivo, which coincided with compositional changes in the respective bile salt pools. Our data show, for the first time, that hepatic FXRa2 and FXRa4 differentially modulate bile salt and lipoprotein metabolism in mice.

Construction and production of self-complementary AAV vectors
The coding sequences of murine FXRa1, FXRa3, HNF4a, SHP and RXRa were amplified using specific primer pairs harboring a Kozak consensus ATG initiation codon and a BamH1 restriction side at the 59 end and a Not1 restriction side at the 39 end (S1 Table) [15]. As a template source, cDNA was synthesized from hepatic RNA from C57BL/6J wild-type mice. The products were cloned into pcDNA5/ FRT/TO (Invitrogen, Life Technologies, Bleiswijk, The Netherlands). The presence of the correct gene was sequence-verified. Subsequently, the MYTG insertion was deleted from FXRa1 and a3 using inverse PCR primers, to generate FXRa2 and FXRa4. Self-complementary AAV (scAAV) serotype 8 vectors with the liver specific LP1 promoter driving murine FXRa2 or murine FXRa4 expression were generated by replacing the factor IX of scAAV2-LP1-hFIXco [16,17] using Pme1, EcoR1 and Hpa1. Additionally, AAV8-pseudotyped vectors were made using the packaging plasmid pAAV8-2. Production, purification and titration of all vectors were performed as described [18,19].
Cell culture, transfections and reporter assays CV1 cells (a kind gift from Ronald M. Evans, Salk Institute, San Diego, USA) were maintained in DMEM (Gibco, Breda, The Netherlands) supplemented with 10% FCS (Sigma Aldrich Chemie BV, Zwijndrecht, The Netherlands), 100 U/ml penicillin and 100 mg/ml streptomycin (Gibco, Breda, The Netherlands). Cultures were maintained at 37˚C and 5% CO2 in a humidified incubator. Cells were transiently transfected using FuGENE 6 transfection reagent (Promega, Leiden, The Netherlands) according to manufacturer's protocol. Cell lysis and luciferase assays were performed using a dual luciferase reporter assay system (Promega, Leiden, The Netherlands) according to manufacturer's protocol. Briefly, the human PGL4-SHP (CHR1_M0312_R1, Switchgear Genomics, Menlo Park, CA, USA) and Cyp8b1 [20] (a kind gift from John Y. Chiang, Northeastern Ohio Medical University, Rootstown, USA) promoter reporters were cotransfected together with the pcDNA5/FRT/TO murine FXR isoforms and the heterodimer RXR alpha/NR2B1 for 48 hrs. Cells were treated with 50 mM CDCA 24 prior to cell lysis when stated.

Animal experiments
FXR knock-out mice [21] backcrossed on a C57BL/6J background and C57BL/6J wild-type mice were housed individually in a temperature-and light-controlled facility with 12 hours light-dark cycling and received food and water ad libitum. Experiments were approved by the Ethical Committee for Animal Experiments of the University of Groningen. Age-matched male FXR knock-out received 1610 11 AAV vector genomes or PBS injected into the retro-orbital sinus. PBS-injection, rather than AAV-injected was used as control since GFP from our AAV-GFP induced several responses, which was already know from previous studies [22,23]. When indicated, mice received chow (RMH-B, Hope Farms, Woerden, The Netherlands) supplemented with 0.5% (w/w) cholate (Calbiochem, La Jolla, CA, USA) in the final week before termination.

Experimental procedures
Four weeks after the scAAV injections, cholesterol fluxes were measured as described previously [24]. Briefly, at day 0 mice received an intravenous dose of 0.3 mg (0.73 mmol) cholesterol-D7 (Cambridge Isotope Laboratories, Inc, Andover, MA, USA) dissolved in Intralipid (20%, Fresenius Kabi, Den Bosch, The Netherlands) and an oral dose of 0.6 mg (1.535 mmol) cholesterol-D5 (Medical Isotopes, Inc, Pelham, NH, USA) dissolved in MCT oil (Pharmacy UMCG, Groningen, The Netherlands). Blood spots were collected from the tail daily for 10 days. At the end of the experiment, mice were anesthetized and after puncturing the gallbladder and disposal of its contents, hepatic bile was collected for 20 minutes from the common bile duct via the gallbladder [25]. Tissues were excised and feces was collected from individual mice for 72 hours prior to termination.

Analytical procedures
Hepatic lipids were extracted according to Bligh & Dyer [26]. Plasma and liver triglycerides, total cholesterol and free cholesterol contents were determined using commercially available kits (Roche Diagnostics, Mannheim, Germany and DiaSys Diagnostic Systems, Holzheim, Germany). Plasma levels of alanine aminotransferase (ALAT) and aspartate aminotransferase (ASAT) were determined using commercially available kits (Spinreact, Santa Coloma, Spain). Biliary phospholipid content was determined according to Böttcher et al. [27]. Cholesterol in bile was measured according to Gamble et al. [28]. Pooled plasma samples were subjected to fast protein liquid chromatography (FPLC) gel filtration using a Superose 6 column (GE Healthcare, Little Chalfont, UK). Biliary and fecal bile salt composition were quantified using capillary gas chromatography (Hewlett-Packerd gas chromatograph; HP 6890) equipped with a FID and a CP Sil 19 capillary column; length 25 m, internal diameter 250 mm and a film thickness of 0.2 mm (Chrompack BV, Middelburg, The Netherlands). Bile salts were methylated with a mixture of methanol and acetyl chloride and trimethylsilylated with piridyne, N, O-Bis(trimethylsilyl) trifluoroacetamide and trimethylchlorosilane. The murine bile salt species include cholate (CA), deoxycholate (DCA), chenodeoxycholate (CDCA), a-muricholate (a-MCA), b-muricholate (b-MCA), v-muricholate (v-MCA), hyodeoxycholate (HDCA), lithocholate (LCA) and ursodeoxycholate (UDCA). We consider cholate and deoxycholate as CA-derived bile salts and the others as CDCA-derived bile salts. Fecal cholesterol and its derivatives (also known as neutral sterols) were trimethylsilylated with pyridine, N, O-Bis (trimethylsilyl) trifluoroacetamide and trimethylchlorosilane and measured on the GC equipped with the same column. Cholesterol enrichment was determined by capillary gas chromatography on a Agilent gas chromatograph (7890A; Amstelveen, The Netherlands) equipped with a 30 m60.25 mm column, with a film thickness of 0.25 mm (ZB-5; Bester, Amstelveen, The Netherlands) connected to a Agilent mass spectrometer (5975C). Cholesterol was derivatized with N, O-Bis (trimethylsilyl) trifluoroacetamide containing 5-10% trimethylchlorosilane. Isotope ratios were determined in the selected ion monitoring mode on m/z 458 (M0) to 465 (M7).

RNA isolation and measurement of mRNA levels by quantitative real-time PCR
Tissue samples for isolation of RNA were snap frozen in liquid nitrogen and stored at minus 80˚C. Samples were homogenized and total RNA using TRI-Reagent (Sigma, St. Louis, MO, USA). RNA concentration was determined using the Nanodrop spectrophotometer (NanoDrop 2000c, Thermo Scientific Inc., Walham, MA, USA). cDNA was obtained from total RNA using the RT procedure using Moloney-Murine Leukemia Virus (M-MLV) reverse transcriptase (RT) (Invitrogen, Life Technologies, Bleiswijk, The Netherlands) with random primers. Gene expression was measured with a 7900HT FAST system using FAST PCR mix, Taqman probes and MicroAmp FAST optical density 96-well plates (Applied Biosystems Europe, Nieuwekerk ad IJssel, The Netherlands). PCR results of liver and intestine were normalized to 36b4 mRNA levels. Primer and probe sequences can be found at (http://www.RTprimerDB.org).

Explorative Illumina microarray analysis
For explorative microarray analysis, total hepatic RNA was prepared from separate groups of 0.5% cholate-fed FXRa2-or FXRa4-transduced FXR knockout mice (n56 per group) using TRI-reagent (Sigma-Aldrich, St. Louis, MO, USA). RNA quality and concentration was assessed with an Biorad Experion Bioanalyzer. Starting with 200 ng of RNA, with an RNA Quality Indicator of at least 8. RNA was amplified and labeled using the Illumina TotalPrep RNA Amplification Kit (Applied Biosystems, Nieuwekerk ad IJssel, The Netherlands).
The WG-6 v2 expression arrays, containing 45281 transcripts (Illumina, San Diego, USA), were processed according to the manufactures protocol and slides were scanned immediately. Quality control, normalization (quantile), batch correction (Combat), prefiltering (fold change 1.1) and statistics (IBMT) were performed in MADMAX [29]. A list of significant changed annotated genes between FXRa2 (n56) and FXRa4 (n56), including False Discovery Rates (FDR)-corrected p-values (10%) was generated. All microarray data reported are described in accordance with MIAME guidelines and are available in the GEO database (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc5GSE51805). Identification of overrepresented functional categories among responsive genes and their grouping into functionally related clusters was performed using the DAVID Functional Annotation Clustering tool [30].

Statistics
Data are shown as Tukey's Box-and-Whiskers plots or average ¡ standard deviations. Statistical analysis was assessed using Kruskal-Wallis H test followed by Conover post-hoc comparison or students t-test using Brightstat [31].

Characterization of the mouse model
At first we established the diurnal rhythmicity and spatial expression of the four murine FXR isoforms. Bile salt synthesis shows a strong circadian rhythm [1,32], but mRNA levels of hepatic FXR isoforms did not show any diurnal variation (S1A Figure). Most prominent differences in expression between the isoforms were observed in skeletal muscle, intestine and kidney (S1B Figure), confirming previous reports [5,6], while hepatic expression of FXR isoforms was comparable. To directly compare the physiological roles of hepatic FXRa2 and FXRa4, we generated self-complementary adeno-associated virus particles (scAAV) expressing the coding regions of murine FXRa2 or FXRa4 behind the liver-specific fatty acid-binding protein 1 (LP1) promoter. Hepatic FXR mRNA expression (Fig. 1A) and protein levels (Fig. 1B) were similarly induced by scAAV particle injection. Compared to PBS-injected controls, plasma ASAT and ALAT levels were unchanged in particle-injected mice, implying that stable scAAV transduction did not negatively impact on liver cell integrity (Fig. 1C/D).

FXR isoforms differentially modulate lipid metabolism
The potency of hepatic FXRa2 and FXRa4 to normalize the characteristic plasma lipid profile of FXR knock-out mice [21,33] was evaluated first. Both isoforms reduced the elevated total plasma cholesterol levels (table 1) seen in PBS-injected FXR KO, FXRa2 being more effective than FXRa4 in this respect, but neither FXRa2 nor FXRa4 completely normalized cholesterol levels to those seen in wildtype mice. FPLC-profiling revealed reduced cholesterol levels in the VLDL-, LDLand particularly HDL-sized fractions of FXRa2-and FXRa4-expressing mice compared to PBS-injected controls and wild-type mice (Fig. 1E), with FXRa2 showing the most pronounced reduction of HDL-cholesterol. Both FXRa2 and FXRa4 caused a lowering of plasma triglyceride (TG) levels, yet, wild-type levels were not reached in either case.

FXR isoforms differentially modulate bile salt composition
Next, we tested whether the physiochemical properties of the bile salt pool were differentially affected by FXRa2 and FXRa4. The total amount of biliary and fecal bile salt excretion was, surprisingly, not different between groups ( Fig. 2A and C). Yet, the bile salt profile in FXRa2-expressing mice showed a shift towards CDCAderived muricholates relative to CA-derived bile salts that were more prominent in FXRa4-expressing mice ( Fig. 2B and D). Specifically, biliary CA and fecal DCA were reduced in the first, while the relative and absolute abundances of a-MCA, CDCA and HDCA were increased (S2 Table). These data imply a shift in bile salt synthesis towards the production of relatively hydrophilic muricholates in FXRa2-expressing mice. Based on similar biliary and fecal bile salt excretion rates, the latter reflecting hepatic bile salt synthesis under steady sate, it can be assumed that the sizes of circulating bile salt pools remained unaffected upon expression of either FXRa2 or FXRa4.

Non-biliary cholesterol excretion is enhanced in FXRa2transduced mice
Since muricholates are more hydrophilic than CA is, we tested whether the differences in the hydrophobicity of the bile salt pool between FXRa2-and FXRa4-expressing mice affect intestinal cholesterol absorption. Fecal neutral sterol output was increased in FXRa2 mice compared to PBS controls and FXRa4 mice (Fig. 3). This increase was not due to diminished fractional cholesterol absorption or increased biliary cholesterol secretion, since these were similar between groups. Hence, the differences in fecal sterols must come directly from the intestine, i.e., reflect the transintestinal cholesterol excretion pathway (TICE) [34,35]. Hepatic FXRa2 expression increased this pathway by 100% on average. Thus, FXRa2-expressing mice showed induction in fecal neutral sterol output and TICE without altering intestinal cholesterol absorption efficiency.

FXR isoforms differently regulate primary bile salt biosynthesis genes
To gain insight in the modes of action by which hepatic FXRa2 and FXRa4 act on metabolic processes, an explorative microarray was performed on the livers of 0.5% cholate-fed FXR KO mice transduced with scAAV-FXRa2 or scAAV-FXRa4. Cholate-feeding was employed to obtain supraphysiological activation of the FXR      Table). Broad scale KEGG pathway analysis [30] revealed, as expected, that the primary bile salt biosynthetic pathway pops up within the top 5 of most relevant pathways. Compared to expression levels seen in the livers of FXRa4-transduced mice, FXRa2-transduced mice showed lower expression levels of Cyp7a1 (19% reduction) and Cyp8b1 (17% reduction), indicating that FXRa2 transrepressed both Cyp7a1 and Cyp8b1 more strongly than FXRa4. Since Cyp8b1 activity is the major determinant of CA biosynthesis in the neutral branch of bile salt synthesis, the data support the notion that hepatic FXRa2 particularly inhibits the CA production. On the other hand, FXRa4-transduced mice showed lower expression of Cyp7b1 (60% reduction). Since Cyp7b1 is involved in CDCA biosynthesis via the so called acidic or alternative branch of bile salt synthesis, these data indeed indicate that hepatic FXRa4 particularly inhibits CDCA/MCA production, as expected on the basis of analysis of biliary and fecal bile salt composition (S2 Table). Also Cyp17a1, a C21 steroidogenic enzyme, was differentially regulated by the FXR isoforms (62% reduction in FXRa2 versus FXRa4). Some of the differentially expressed genes as well as FXR itself and its targets Shp and Abcb11/Bsep were confirmed using quantitative PCR together with some intestinal FXR targets, in which the differences were even more pronounced (table 2). Consistent with the shift towards CDCA-derived muricholates relative to CA-derived bile salts in FXRa2-expressing mice, hepatic Cyp8b1 mRNA expression was reduced compared to FXRa4-expressing mice (reduction of 33%) and PBS-injected FXR KO (reduction of 60%). CYP7B1, the enzyme of the alternative bile salt biosynthetic pathway, was only down-regulated in the liver of FXRa4-expressing mice on 0.5% cholic acid-feeding. The expression of Abcb11/ Bsep also showed differential regulation by the two FXR isoforms on mRNA level. FXRa2-expressing mice showed higher Abcb11/Bsep mRNA expression compared to FXRa4 in both conditions. Bile flow in these animals was increased compared to FXRa4-expressing and control mice (table 1), yet, secretion of endogenous bile salts was unaffected implicating stimulation of bile salt-independent bile flow. Surprisingly, the intestinal FXR target genes Fgf15 and Fabp6/Ibabp were affected in the FXRa2-and FXR4a-transduced mice on both diets (table 2), indicating a cross-talk between liver and intestine.

FXR isoforms differentially induce SHP in vitro
In attempt to explain the differential effects of FXRa2 and FXRa4 on Cyp8b1 expression, we have screened the promoter region of the Cyp8b1 gene for FXREs using Mathinspector [36]. This screening revealed an IR1 FXRE in the promoter region of murine Cyp8b1 starting at 99 base pairs from the transcription start site. Also in the human CYP8B1 promoter region an IR1 FXRE was found, at .2500 base pairs from the transcription start site. Unlike expected, we were unable to show direct FXR-mediated regulation of Cyp8b1 expression in promoter studies, since FXRa2 could not transrepress the HNF-4a-activated Cyp8b1 reporter (Fig. 4A). However, we did find that FXR isoforms differentially induce Shp (Fig. 4B). Shp is a well-established FXR target gene whose product represses nuclear hormone receptor-mediated transactivation. Hence, this finding might indicate FXR-Shp-mediated Cyp8b1 transrepression, possibly explaining the changes in bile salt composition observed in the in vivo situation. FXR isoforms showed differences in their transcriptional activity (FXRa2.a4.a1.a3)  independent of ligand activation, which was already shown in previous studies for Abcb11/Bsep but not for Shp [5,6]. Compared to other isoforms FXRa2 showed an up to 50% increase in transcriptional activity upon activation using CDCA as ligand. However, the differences in hepatic Shp expression in vivo between chowfed FXRa2-and FXRa4-expressing mice failed to reach statistical significance and therefore the physiological relevance, if any, of isoform-specific Shp regulation awaits further studies.

Discussion
The most important finding of this study is it the demonstration, for the first time, of physiologically relevant differences in in vivo activity between the two transcriptionally most active FXR isoforms, i.e., FXRa2 and FXRa4. Our data demonstrate that FXRa2 and FXRa4 differentially affect transcriptional control of bile salt and lipoprotein metabolism in mice.
One of the phenotypic hallmarks of FXR knock-out mice concerns the plasma lipid profile with high levels of HDL-cholesterol and concomitantly elevated VLDL-cholesterol and triglyceride levels [11]. In our model, a first biological readout was the (partial) normalization of this profile. Both FXR isoforms reduced HDL-and VLDL-cholesterol levels. However, total plasma cholesterol levels in FXRa2-expressing mice were markedly reduced compared to FXR KO values, while FXRa4 was less effective. In both situations, cholesterol levels found in wildtype mice were not reached, yet, the results underscore the role of hepatic FXR in control of plasma HDL levels. The existence of relationships between bile salt and lipid metabolism has been recognized for many years. The production of triglyceride-rich VLDL-particles, as well as their clearance from the circulation have been proposed to be influenced by FXR activity through modulation of hepatic apoc2/apoc3 expression [37,38] as well as VLDLR expression in cells [39]. On the other hand, bile salt-activated FXR lowers HDL-cholesterol, which has been attributed to suppression of Apoa1 expression [40] and, very recently, to suppression of CETP expression which is, however, not expressed in wild-type mice [41]. The explorative microarray data of 0.5% cholic acid-fed FXRa2-and FXRa4-expressing mice revealed small but significant differences in hepatic Apoc2 (20% upregulated in FXRa2) and Vldlr (30% upregulated in FXRa4) expression (S3 Table). Although relatively small, these differences indicate differential regulation of VLDL production as well as clearance from the circulation. Recently, the triglyceride-lowering capacity of FXR was linked to synthesis of 12ahydroxylated bile salts via Foxo1, a key regulator of glucose metabolism [42]. Murine Foxo1 ablation caused elevated liver and plasma triglyceride levels, two signature lipid abnormalities of diabetes and the metabolic syndrome. These changes were associated with a relative deficiency of 12a-hydroxylated bile salts and in expression of Cyp8b1. In diabetes, the impaired insulin response led to dysfunctional Foxo1, which drives bile salt synthesis towards 12a-hydroxylated bile salts, leading to an abnormally hydrophilic bile salt pool affecting cholesterol and triglyceride metabolism. These data coincide with those observed in our model: FXRa2-expressing mice showed a stronger transrepression of Cyp8b1 compared to FXRa4, thereby generating less 12a-hydroxylated bile salts and their plasma triglyceride levels tended to be higher than in FXRa4-expressing mice (table 1). Furthermore, FPLC data revealed slightly lower HDL-cholesterol in FXRa2-expressing mice compared to those expressing FXRa4. Thus, FXRa2 was found to target primarily the neutral branch of the bile salt biosynthesis pathway by transrepressing Cyp7a1 and Cyp8b1, thereby inhibiting CA production. FXRa4, on the other hand, was found to particularly suppress the acidic branch of the bile salt biosynthesis pathway Cyp7b1, targeting primarily CDCA and MCA synthesis.
The changes in expression levels were consistent with compositional changes in the bile salt species in bile as well as in feces. The changes in bile salt composition alter the physiochemical properties of the bile salt pool. FXRa2 mice generate a more hydrophilic bile salt pool compared to FXRa4. In the cholate-fed situation, i.e., a state of FXR activation, also Cyp17a1 was found to be differentially expressed by FXRa2 and FXRa4. FXRa2-expressing mice transrepressed Cyp17a1 62%more compared to FXRa4. Recently, increased production of 17-hydroxy steroid metabolites, the product of Cyp17a1, have been associated with juvenile onset cholestasis [43]. We did not observe large differences in bile flow or the grade of liver damage between the FXR isoforms that might indicate onset of cholestasis.
In the bile salt pool of Cyp7a1-as well as Cyp8b1-knock-out mice, in which CA was largely replaced by MCA, fractional cholesterol absorption from the intestine was decreased [44] while the intestinal cholesterol synthesis was induced [45] as a rescue mechanism. Since we observed similar alterations in bile salt pool composition in the FXRa2 mice, we analyzed cholesterol kinetics [24] to determine whole body cholesterol balance. Our data show increased fecal neutral sterol output in FXRa2-expressing mice, without concomitant reduction of fractional cholesterol absorption. Since there was no difference in biliary cholesterol secretion between FXRa2 and a4 -expressing mice, increased TICE accounted for the difference in neutral sterol excretion. The question arises how FXRa2 induces this effect. Our data show a differential effect of the FXR isoforms on plasma lipoprotein and bile salt pool distribution. Regulatory effects of plasma lipoproteins on the rate of TICE have not been described and the (lipoprotein) donor of TICE has still not been identified [46,47]. The influence on the hydrophobicity of the bile salt pool may be more important. In vitro studies have shown that the presence of bile salts in the intestinal lumen, particular when combined with phospholipids, strongly stimulate this pathway but a clear influence of bile salt hydrophobicity was not demonstrated. A clear association between bile salt hydrophobicity and cholesterol absorption has been observed [48]. However, apparently in the present study the changes in the hydrophobicity of the bile salt pool were too subtle to affect cholesterol absorption. Hence it is not inhibition of cholesterol absorption that induces the increase in TICE but another as yet unidentified factor must be involved. We do speculate, however, that the change in the hydrophobicity of the bile salt pool drives the effect on TICE.
FXR regulates bile salt hydrophobicity via modulation of Cyp7a1 and Cyp8b1 expression through a SHP/LRH1-and HNF-4a-dependent pathway. To explain our results we speculated that FXR would directly transrepress Cyp8b1 without interference of SHP. Promoter region analysis using Mathinspector [36] revealed a FXRE (agggcaggaacct) in the promoter region of both murine and human CYP8B1. Reporter studies, however, did not confirm direct FXR-mediated regulation of Cyp8b1 expression using our murine FXR isoforms. Although hepatic gene expression of SHP did not significantly differ between the FXR isoform-transduced mice in vivo, our in vitro studies did show differential induction of SHP by both FXR isoforms. SHP is an orphan nuclear receptor that is regulated by many other nuclear receptors and transcription factors involved in metabolism and cancer [49] and also has been shown to be involved in control of Cyp8b1 expression [50]. SHP can repress transcription factor-mediated transactivation by inhibiting DNA binding, competing for binding of coactivators or recruitment of corepressors. Hepatic SHP mRNA expression shows circadian rhythmicity, regulated by the clock gene Nr1d1(Rev-erba) [51]. Because of the interaction with many other transcription factors and the circadian rhythmicity, the in vivo regulation of SHP is complex. Our laboratory very recently demonstrated that human FXR regulates SHP expression through direct binding to an LRH-1 binding site, independent of the known IR-1 and LRH-1 sites [52], which underscores the fact that the exact mechanisms by which SHP is regulated in vivo remains enigmatic. The differences in transcriptional activation of SHP, i.e., FXRa2 being more effective than FXRa4, may contribute to the differences observed in hepatic Cyp8b1 expression, resulting in the concomitant differences in the bile salt pools between FXRa2-and FXRa4-expressing mice. Yet, more studies are required to substantiate this suggestion.
The intriguing question remains why FXR isoforms exist and what their specific physiological roles in different tissues actually are. FXRa2 and FXRa4 differ within their N-terminal parts, i.e., a 37 amino acid extension of FXRa4 compared to FXRa2. This difference can cause conformational changes in the FXR protein and, thereby, influence the transcriptional activity and/or transactivation/ repression capacity on target genes. A dozen transcriptional cofactors have been described to influence FXR transactivation [53], but actions specific for the specific FXR isoforms have not been described. The relative abundance of hepatic FXRa3,4 was increased in children suffering from progressive familial cholestasis type I [54]. Also in cholangiocarcinomas [55] and colorectal adenocarcinoma [56] the relative abundance of hepatic and intestinal FXRa3,4 was increased. Apparently, disease states can be associated with a switch in FXR isoform expression in the liver and intestine which may influence outcome of disease progression. Recently, Vaquero et al. [57] reported that activation of human FXR depends on the pattern of FXR isoform expression and bile salt composition. They showed that cell-specific pattern of FXR isoforms determine the overall tissue sensitivity to FXR agonists and may be involved in the differential response of FXR target genes to FXR activation. The present study shows that the FXR isoforms influence physiology differentially and although murine and human bile salt and lipoprotein metabolism differ from each other the differential influence of the isoforms on the bile salt synthesis pathways will probably be similar.
In conclusion, we show for the first time physiological evidence for differential roles of FXRa2 and FXRa4 in control of bile salt and lipoprotein metabolism in mice. This might, in light of the increasing evidence for therapeutic potential of FXR agonists, be of importance for design of treatment strategies for metabolic diseases.