Analysis of HSD3B7 knockout mice reveals that a 3α-hydroxyl stereochemistry is required for bile acid function

Construction of Hsd3b7 Knockout Mice.

We took advantage of the small size of the mouse Hsd3b7 (2.7 kb, six exons) to assemble a targeting vector that would allow complete replacement of the gene with a cassette encoding neomycin resistance after homologous recombination in ES cells (Fig. 1A). To ensure that the resulting phenotype was due to loss of Hsd3b7 and not due to transcriptional interference from the neomycin resistance gene, the cassette was flanked by LoxP sites and contained a gene specifying Cre recombinase linked to a testis-specific promoter (24). Activation of the Cre recombinase gene in germ cells of chimeric male mice was predicted to result in cassette excision, leaving behind a 34-bp LoxP site in place of Hsd3b7 (Fig. 1A).

Homologous recombination in ES cells occurred at a high frequency (40%) after introduction of the targeting vector, and six independently derived ES cell clones produced chimeric animals that transmitted the altered Hsd3b7 through the germline. The mutant allele was inherited in Mendelian fashion (Fig. 1B), and expected numbers of wild-type, heterozygous, and homozygous offspring were born. Equal numbers of male and female pups were obtained in crosses between heterozygous carriers of the mutant gene. Fertility and fecundity were normal in mice heterozygous and homozygous for the introduced mutation.

The effects of the mutation on the expression of HSD3B7 mRNA and protein in the liver were determined by real time RT-PCR and immunoblotting. Abundant HSD3B7 mRNA was detected in the livers of wild-type mice; levels of this mRNA were reduced by half in heterozygous mice and to undetectable levels in homozygous mice (data not shown). As judged by SDS/PAGE and immunoblotting (Fig. 1C), the normal HSD3B7 protein migrated with an apparent molecular weight of ≈33,000. The amount of this protein was reduced by half in mice heterozygous for the disrupted gene and was undetectable in animals homozygous for the mutant allele.

Phenotype of Hsd3b7^−/− Mice.

Mice lacking HSD3B7 were outwardly normal at birth, but a majority failed to grow at a normal rate and died within the first 18 days of postnatal life when maintained on standard laboratory chow and water (Fig. 2A). Death occurred in two waves, with ≈45% of homozygous male and female animals dying in the first 4 days of life and another ≈45% dying between days 8 and 18. Only a few animals survived beyond weaning (postnatal day 21). Addition of a pan-vitamin supplement to the drinking water and a bile acid (0.1% cholic acid) to the diet during late gestation and continuing through postnatal day 28 resulted in an approximate doubling of the survival frequency. When the level of dietary bile acid was increased to 0.5% and vitamin supplementation was maintained, survival of the mutant mice approached that of wild-type mice (Fig. 2A), The fur of Hsd3b7^−/− suckling mice and nursing mothers maintained on unsupplemented chow and water was oily (Fig. 2B), and the skin was scaly (Fig. 2C). These traits were not observed in wild-type mice or those heterozygous for the introduced mutation. Vitamin and bile acid supplementation prevented the appearance of these features in homozygous mice.

Adult Hsd3b7^−/− animals maintained on vitamin and bile acid supplements through postnatal day 28 and thereafter switched to normal laboratory chow attained body weights similar to those of wild-type controls (Table 1). There was no apparent organomegaly in tissues of the enterohepatic circulation, and plasma cholesterol levels were normal. Kidney and liver cholesterol levels were not significantly different. A 32% decrease in plasma triglycerides was measured together with a 30% increase in hepatic triglyceride levels (Table 1). Liver histology of knockout animals up to 1.5 years of age was normal, and in agreement with this finding, the liver enzymes aspartate aminotransferase and alanine aminotransferase were not elevated in the plasma.

Sterol and Bile Acid Metabolism in Hsd3b7^−/− Mice.

Bile acid pool size was determined by HPLC and found to be decreased by ≈33% in HSD3b7^−/− vs. control mice, a difference that was not statistically significant (Fig. 3A). Rates of fecal bile acid excretion determined by HPLC analysis of bile acids in the stool were increased ≈2-fold in the knockout mice (Fig. 3B). Fecal neutral sterols also were excreted at a rate that was 2.8-fold higher in knockout mice than in wild-type littermates (Fig. 3C), and, consistent with this increase in sterol loss, intestinal cholesterol absorption was decreased to <9% of normal in Hsd3b7^−/− mice (Fig. 3D).

We next measured rates of cholesterol synthesis in the tissues of wild-type and knockout mice, using a [³H]H₂O incorporation assay (25). As indicated by the data in Fig. 4A, elimination of Hsd3b7 caused a 3.5-fold increase in de novo cholesterol synthesis in the liver and smaller but significant increases in the duodenal and jejunal segments of the small intestine. In contrast to these results, cholesterol synthesis rates were not changed in the terminal segment (ileum) of the small intestine or in the spleen. A modest but significant decrease in cholesterol synthesis was detected in the kidney. These findings suggested that Hsd3b7^−/− mice compensated for the loss of dietary cholesterol (Fig. 3D) by increasing rates of de novo cholesterol synthesis in the liver and initial segments of the gut (Fig. 4A). Rates of fatty acid synthesis increased ≈20% in the livers of knockout mice but were unchanged in the other tissues examined (Fig. 4B).

Fast Atom Bombardment (FAB)-MS Analyses of Bile Acids in Hsd3b7^−/− Mice.

The presence of an insignificantly different bile acid pool size in the mutant mice together with an increased rate of bile acid and fecal neutral sterol excretion and decreased cholesterol absorption suggested that the composition of the bile acid pool was different in these animals. To test this hypothesis, we first used FAB-MS, which is routinely used to diagnose the presence of HSD3B7 deficiency in human subjects (26). As shown in Fig. 5, the negative ion FAB-MS mass spectra obtained from bile pooled from wild-type and knockout mice were virtually identical and contained major ions at m/z 514 and 498 representing the taurine conjugates of typical mouse dihydroxy- and trihydroxy-cholanoic acids. These data were consistent with the presence of primary bile acids in mice of both Hsd3b7 genotypes and with the fact that the mouse is an obligate taurine conjugator. No bile acids were detected by FAB-MS analyses of urine samples from wild-type and null mice (data not shown).

For comparison, we acquired the spectrum of a urine sample from a human patient with HSD3B7 deficiency (Fig. 5). As expected, two specific pairs of ions were observed at m/z 469 and 485, and m/z 526 and 542, which represented the sulfated and glyco-sulfated metabolites of 3β,7α-dihydroxy-5-cholen-24-oic and 3β,7α,12α-trihydroxy-5-cholen-24-oic acids, respectively, that accumulate in subjects with this genetic disease. The absence of these or related 3β-hydroxy-Δ⁵ bile acids in bile and urine of adult HSD3B7 knockout mice suggested the presence of a compensatory metabolic pathway that prevented their accumulation.

GC-MS Analyses of Bile Acids in Hsd3b7^−/− Mice.

We next carried out GC-MS analyses of bile acids present in gallbladder bile of wild-type and knockout animals. As expected, the bile of wild-type mice contained three major bile acids corresponding to cholic acid (3α,7α,12α-trihydroxy-5β-cholanoic acid), β-muricholic acid (3α,6β,7β-trihydroxy-5β-cholanoic acid), and ω-muricholic acid (3α,6α,7β-trihydroxy-5β-cholanoic acid) (Fig. 6A). The bile of knockout mice contained these three bile acids and two additional major species that eluted from the GC column on both sides of cholic acid, and several minor species (Fig. 6A). The mass spectrum of the bile acid eluting just before cholic acid suggested the presence of a trihydroxy-cholanoate structure (Fig. 6B, scan 265). The molecular ion, of low intensity, was at m/z 638, and the ion sequence of three losses of trimethylsilanol (TMS), yielding m/z 548 (M-90), m/z 458 (M-[2 × 90]), and m/z 368 (M-[3 × 90]), together with the loss of the side chain (−115 Da) to give rise to the ABCD-ring fragment at m/z 253, established a trihydroxylated C-24 bile acid. The relative intensities of these ions differentiated two possible stereoisomers of a 3,7,12-trihydroxy bile acid; the intense ion at m/z 343 representing the loss of two TMS groups and the side chain was more abundant in the 3β,7α,12α-trihydroxy-5β-cholanoic acid stereoisomer (Fig. 6B, scan 265, arrow) than it was in the 3α,7α,12α-trihydroxy-5β-cholanoic acid stereoisomer (Fig. 6B, scan 278).

The mass spectrum of the bile acid eluting just after cholic acid suggested the presence of a dihydroxy-cholanoate structure (Fig. 6B, scan 286). The molecular ion was at m/z 550 consistent with a dihydroxy C-24 bile acid. Sequential losses of two TMS groups gave rise to ions at m/z 460 and 370, followed by the loss of the side chain (−115 Da) to account for the ABCD-ring fragment at m/z 255. The ion at m/z 405 in this spectrum, which is accounted for by loss of carbon atoms C-1 to C-4 of the A-ring, was diagnostic of a 3,6-dihydroxy-bile acid vs. a 3,7- or a 3,12-dihydroxylated bile acid. The ion at m/z 323 is also prominent in 3,6-dihydroxy bile acids, being formed by cleavage across the B- and C-rings. From these data and the retention index of the peak, we deduced that this species represented the methyl ester-TMS ether derivative of 3β,6α-dihydroxy-5β-cholanoic acid. For comparison, the mass spectrum of the major trihydroxylated 6β-hydroxy bile acid, β-muricholic acid, is shown in Fig. 6B, scan 337.

Supporting information (SI) Table 2 summarizes the individual bile acids detected and their concentrations in pooled gallbladder bile samples from wild-type and knockout mice, which were determined as described in SI Materials and Methods. Together, these MS analyses indicated that Hsd3b7^−/− mice contained reduced amounts of normal bile acids and two additional atypical bile acids, 3β,7α,12α-trihydroxy-5β-cholanoic acid and 3β,6α-dihydroxy-5β-cholanoic acid.

Gene Expression in Hsd3b7^−/− Mice.

To determine the regulatory effects of the altered bile acid pool composition in Hsd3b7^−/− mice, we used real-time RT-PCR to assess mRNA levels in tissues of the enterohepatic circulation. As shown by the data in Fig. 7A, the amounts of cholesterol 7α-hydroxylase and sterol 12α-hydroxylase mRNAs, which are normally repressed by bile acids acting through FXR, were elevated 4-fold compared with those of wild-type mice. As a control, the sterol 27-hydroxylase mRNA level, which does not respond to bile acids, was unchanged in the knockout mice. Unexpectedly, the levels of two other normally constitutively expressed mRNAs, encoding the bile acid biosynthetic enzyme steroid 5β-reductase and racemase, were increased ≈3-fold in the mutant mice. SHP mRNA levels, which are normally induced by FXR and bile acids, were decreased ≈2-fold in Hsd3b7^−/− livers. The increased rate of cholesterol synthesis reported in Fig. 4A, was reflected by an elevation in the level of mRNA encoding the rate-limiting enzyme in this pathway, hydroxymethylglutaryl CoA reductase, whereas increases in mRNAs encoding acetyl-CoA carboxylase 1, fatty acid synthase, and stearoyl-CoA desaturase 1 mirrored the enhanced rates of fatty acid synthesis measured in these animals. The changes observed in these mRNA levels were confirmed by microarray hybridization (data not shown) and, in the case of cholesterol 7α-hydroxylase, at the protein level by immunoblotting (Fig. 7A Inset).

In the gut, the alteration in bile acid pool composition led to a decrease in the levels of two known FXR target mRNAs, those specifying SHP and fibroblast growth factor 15 (Fig. 7B). The increased rate of cholesterol synthesis was paralleled by an elevation in the amount of hydroxymethylglutaryl CoA reductase mRNA.

Construction of Targeting Vector.

A plasmid vector designed to delete the six exons of Hsd3b7 by homologous recombination in mouse ES cells was constructed by using standard methods (40), the vector pPolIIshort-neobPA-HSVTK (41), and the “ACN cassette” (24).

ES Cell Culture.

Mouse I1C ES cells derived from the 129SvEv strain were used in homologous recombination experiments as described in ref. 20. The frequency of ES cell clones that underwent homologous recombination was 40% (132/384 screened). Injection of four of these clones into C57BL/6J blastocysts produced 24 chimeric males that had contributions from the ES cells ranging from 50% to 80% based on the amount of agouti coat color. Of these mice, six transmitted the disrupted gene through the germline. Experiments reported here were performed in mice of mixed strain (C57BL/6J-129SvEv) descendants (F₂ and subsequent generations) of these animals. No phenotypic differences were observed between knockout mice derived from two independent ES cell clones.

PCR Genotyping.

Routine genotyping of mice was performed on genomic DNA extracted from tails (DirectPCR lysis reagents; Viagen Biotech, Los Angeles, CA), using three oligonucleotide primers in a multiplex PCR. The oligonucleotides used were: wild-type allele 5′ primer, GCCACAGTGGGTGAAGCAGG; targeted allele 5′ primer, AAGGACAGAACCTGCCCATTTGGT; and common 3′ primer, GGTAGCTGAAGTAAGCAGACCAGC.

Animals and Diets.

Animals were housed in cages containing wood shavings in a temperature-controlled room (22°C) with 12-hr light/dark cycling. All experiments were performed toward the end of the 12-hr dark phase of the cycle, and all mice were in the fed state at the time of study unless otherwise stated. Experiments were approved by the University of Texas Southwestern Institutional Animal Care and Research Advisory Committee.

Survival studies used a cereal-based rodent diet, Teklad 7002 (Harlan Teklad, Madison, WI), which contained 6% (wt/wt) fat, 24% (wt/wt) protein, and 5% (wt/wt) fiber. To increase survival of Hsd3b7^−/− pups, pregnant and nursing mothers were provided water supplemented with vitamins (CritterVites; Mardel, Glendale Heights, IL) and fed Teklad 7002 supplemented with either 0.1% (wt/wt) or 0.5% cholic acid (catalog nos. TD03626 and TD04046, respectively; Harlan Teklad). These supplements were begun on gestation day 12 and continued until the fourth postnatal week, at which time the animals were fed Teklad 7001 (Harlan Teklad), which contained 4% (wt/wt) fat, 24% (wt/wt) protein, and 5% (wt/wt) fiber. To ensure complete purging of cholic acid provided as a supplement to newborn and juvenile Hsd3b7^−/− mice, animals in which bile acid pool size and composition were determined were fed the unsupplemented Teklad 7001 diet a minimum of 2 months before study.

Intestinal Cholesterol Absorption.

Cholesterol absorption was measured by a fecal dual-isotope ratio method (42).

Bile Acid Composition and Pool Size.

Bile acids were isolated from the gallbladder, surrounding liver tissue, and small intestine of individual mice by extraction into ethanol with [24-¹⁴C]taurocholic acid (New England Nuclear, Waltham, MA) added as an internal standard. Individual bile acids were resolved and quantitated after HPLC on a Hewlett Packard (Palo Alto, CA) series 1100 system with refractive index detector (38).

Fecal Bile Acid and Neutral Sterol Excretion.

Mice were singly housed in fresh cages for 3 days, and their feces were collected and extracted as described in refs. 20 and 38.

Cholesterol and Fatty Acid Synthesis.

Rates of de novo sterol and fatty acid synthesis were measured by using a [³H]H₂O incorporation assay (25).

Plasma and Tissue Lipid Levels.

Mice were exsanguinated from the vena cava under isoflurane anesthesia. Cholesterol and triglyceride levels were determined in plasma and 100-mg aliquots of frozen liver from knockout and wild-type mice, using commercially available kits (Chol 1127771; Roche Diagnostics, Nutley, NJ; and Infinity; Thermo Electron, Melbourne, Australia) (43).

RNA and Protein Blotting.

Total RNA was extracted from liver and small intestine, using RNA Stat-60 kits (Tel-Test B, Friendswood, TX). For real time RT-PCR, equal amounts of RNA isolated from individual mice were pooled, and cDNA derived from 20 ng of total RNA was used in each reaction.

Protein extracts were prepared by homogenizing freshly harvested livers in 0.25 M sucrose buffer [0.25 M sucrose/20 mM Tris-acetate, pH 7.4/1 mM EDTA and one tablet of an EDTA-free protease inhibitor mixture tablet per 7 ml of buffer (Roche Diagnostics, Mannheim, Germany)], using a polytron, followed by centrifugation at 600 × g for 10 min. Membranes were prepared by subjecting the supernatant to an additional centrifugation at 100,000 × g for 30 min at 4°C. The resulting pellets were resuspended in 50 mM Tris·acetate, pH 7.4/1 mM EDTA/20% (vol/vol) glycerol, and protease inhibitor as described above. A rabbit polyclonal antibody that recognizes amino acids 340–354 of the mouse HSD3B7 was used to detect protein expression (22). Cholesterol 7α-hydroxylase and sterol 27-hydroxylase proteins were identified by using rabbit polyclonal antisera (44).

FAB Mass Spectrometry (FAB-MS).

Bile was collected from the gallbladders of mice fasted for 4 h. Bile acids were analyzed directly with no prior work-up of the sample. Negative ion FAB-MS spectra were recorded after placing the equivalent of 1 μl of bile onto a small drop of glycerol spotted on the target probe. The probe was introduced into the ion source of an Autospec Q magnetic sector mass spectrometer, and a beam of fast atoms of cesium was fired at the target. Negative ion spectra (scan time 1.5 sec) were recorded over the mass range (m/z) 50–800 Da and compared with those of known bile acid conjugates and biosynthetic intermediates (45, 46).