INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Members of the nuclear receptor family regulate a broad range of developmental and physiological processes by binding to DNA response elements and regulating transcriptional activation (7). The retinoid X receptors (RXRs) are unique among the nuclear receptors in that they bind DNA as a homodimer and are required as a heterodimeric partner for a number of additional nuclear receptors to bind DNA (19). The latter receptors, termed the class II nuclear receptor subfamily, include many which are established or implicated as important regulators of gene expression in the liver (reviewed below). There are three RXR genes (18), coding for RXR, -, and -, all of which are able to heterodimerize with any of the class II receptors, although there appear to be preferences for distinct RXR subtypes by partner receptors in vivo (6).

The physiological processes that are regulated by the class II receptors in the liver are diverse. LXR is activated by oxycholesterol and promotes cholesterol metabolism (22). FXR (also known as RIP14) is part of an interrelated process, in that FXR is activated by bile acids (the end product of cholesterol metabolism) (17, 21, 34), which serve to inhibit cholesterol catabolism. CAR is involved in phenobarbital induction of the cytochrome P450 2B10 (Cyp2B10) gene (9), which encodes a drug- and xenobiotic-metabolic enzyme. Interestingly, CAR has constitutively activity, but becomes inactive in response to the steroid androstane (8). PXR is activated by a wide spectrum of steroids, and induces the Cyp3A gene (Cyp3A1 in mice; Cyp3A4 in humans) which encodes a broad-spectrum oxidase that is responsible for steroid degradation and for the catabolism of numerous pharmaceutical compounds (2, 11). PPAR is activated by leukotriene B4 and by synthetic peroxisome proliferators, such as fibrates, and controls the expression of several genes which are related to fatty acid metabolism and processing (28). Although not discussed further in the context of this study, other class II nuclear receptors, including the retinoic acid and thyroid hormone receptors, are also likely to be important regulators of liver metabolism.

In the adult liver, RXR is the most abundant of the three RXRs (18), suggesting that it might have a prominent role in hepatic functions that involve regulation by class II nuclear receptors. Conventional gene targeting approaches have only partially addressed this issue. While all three RXR genes have been individually mutated in the mouse germ line, mice lacking both RXR and RXR together (14) are viable and are not known to suffer from any impairment of liver function. In contrast, lack of RXR results in midgestation embryonic lethality caused by insufficient heart development (10, 30), thereby preventing assessment of the role of RXR in postnatal liver physiology. To address this role, in the present study, we have used cre/lox-mediated recombination to selectively mutate the RXR gene in adult hepatocytes, and thereby explore the involvement of RXR in the diverse range of physiological pathways that are regulated by class II nuclear receptors in the liver.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Mice. The derivations of the conventional loss-of-function RXR allele (identified as ko), the conditional RXR allele (identified as flox), the albumin-cre (Alb-cre) transgenic line, and the conditional lacZ reporter line ROSA26R have all been described previously (4, 25, 29, 30). Liver-specific mutation of the RXR gene was achieved by crossing the Alb-cre transgene against the conditional allele. Experimental animals used in this study carried one allele of the Alb-cre transgene and at the RXR locus were either RXR^flox/flox or RXR^flox/ko. Wild-type animals used in this study were of either completely wild-type genotype (RXR^wt/wt) or carried one RXR conditional allele (RXR^flox/wt) and did not carry the cre transgene. In the absence of cre-mediated recombination, the conditional flox allele is functionally identical to the wild-type allele. Mice were housed in standard cages under a 6 a.m.-6 p.m. light-dark cycle and under standard conditions were fed either Purina PicoLab Rodent Diet 20 or Harlan Teklad 7001 diet. The latter diet was provided to control animals in high-cholesterol-diet studies, and Harlan Teklad TD86295, containing 2% cholesterol by weight in a 7001 base, was provided to experimental mice. For analysis of Cyp7A mRNA induction by dietary cholesterol, age- and littermate-matched male mice were housed individually for at least 2 weeks before beginning the high-cholesterol diet. Two hours before the onset of the light cycle at the end of the seventh night of treatment, mice were sacrificed, and liver tissue was isolated and frozen and then processed as described below.

Specificity of cre-mediated recombination. For Southern blotting, liver DNA was isolated from dissected embryonic or postnatal stage animals by standard procedures. Genomic DNA was restriction digested with BamHI and BglII and then separated on an agarose gel and transferred to nylon membranes. The probe used in hybridizations was a HindIII-SmaI fragment of the mouse genomic RXR locus previously described as probe C (4), which hybridizes to a region outside the two loxP sites and therefore recognizes the wild-type RXR allele, the unrecombined conditional RXR allele, and the conditional RXR allele after recombination. For PCR analysis of specificity, genomic DNA was isolated from various tissues and assayed by PCR amplification with primers P1 and P3 as described previously (4). The sensitivity of this assay was determined by amplifying samples with known ratios of recombined liver genomic DNA and nonrecombined tail genomic DNA. For histochemical staining in animals carrying the conditional ROSA26R lacZ reporter gene, adult tissue was isolated and in some cases bisected manually to improve stain penetration and then fixed and stained with X-Gal (5-bromo-4-chloro-3-indolyl--D-galactopyranoside) by standard procedures. For RXR RNA analysis, total liver RNA was converted into cDNA by random priming and then amplified for 32 cycles with primers corresponding to nucleotides 552 to 574 and 952 to 931 of X66223.1 in GenBank.

Northern blotting analysis. All mice used as a source of RNA were male mice 2 to 4 months of age. Total liver RNA was isolated by guanidinium-phenol extraction. Twenty micrograms of total RNA per lane was resolved by electrophoresis on 1.2% agarose gels containing 2.2 M formaldehyde and then transferred to nylon membranes by capillary blotting. Probe cDNA fragments were labeled by random priming and hybridized to membranes in 7% (wt/vol) sodium dodecyl sulfate, 0.5 M sodium phosphate (pH 6.5), 1 mM EDTA, and 1 mg of bovine serum albumin per ml at 68°C overnight. The membranes were washed twice in 1% sodium dodecyl sulfate, 50 mM NaCl, and 1 mM EDTA at 68°C for 15 min each and autoradiographed with intensifying screens. The amount of mRNA expressed in individual samples was quantitated by densitometry and then normalized with the level of 18S rRNA; the mean and standard deviation for eight samples were calculated to validate the statistical significance of observed changes. Gene probes used were ApoAI and CIII (provided by J. Auwerx), Cyp4A1 (provided by F. Gonzalez), liver fatty acid-binding protein (LFABP; provided by J. Gordon), acylcoenzyme A (acyl-CoA) oxidase (provided by T. Osumi), catalase (purchased from American Type Culture Collection) Cyp2B10 (provided by M. Negishi), Cyp3A1 (provided by F. Gonzalez), and Cyp7A (provided by L. Chan). The RXR heterodimeric partner genes studied were mouse retinoic acid receptors (RARs) and RXRs (provided by R. Evans), PPAR (provided by S. Green), LXR (provided by D. Mangelsdorf), CAR (provided by B. Forman), PXR (provided by B. Blumberg), FXR/RIP14 (provided by D. Moore), and HNF4 (provided by F. Sladek).

Liver function, composition, and morphology. Liver function, including blood triglyceride and cholesterol levels, was evaluated by automated analysis of major serum components. Blood samples were obtained by retroorbital extraction. Hepatic cholesterol and triglyceride were measured as previously described (22). For hematoxylin-eosin histological analysis, pieces of tissue from equivalent regions of the liver were dissected, formalin fixed, paraffin embedded, sectioned at a 5-µm thickness, and stained by standard procedures. For neutral lipid staining, oil red O was dissolved in isopropanol and mixed with water immediately prior to use. Tissue was frozen in O.C.T. (Tissue-Tek) and cryostat sectioned at a 5-µm thickness, rinsed in water, stained, rinsed in isopropanol and then water, and then stained with hematoxylin and mounted with Fisher aqueous mountant. A polarizing filter was used to visualize cholesterol crystals.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Generation of mice lacking RXR in hepatocytes. To selectively disrupt RXR gene function in postnatal hepatocytes, we have utilized a previously described (4) conditional ("floxed") allele of RXR. This allele was constructed such that small loxP sequences were inserted into introns flanking the fourth exon of the gene; this exon encodes the majority of the DNA binding domain of the RXR protein, which is essential for RXR function. This allele is fully functional and effectively wild type in the absence of cre-mediated recombination, although upon recombination, the fourth exon is specifically deleted. To mutate the RXR gene in hepatocytes, we crossed the floxed RXR allele against a transgenic line in which cre recombinase is expressed under the control of the albumin promoter (25). The albumin gene is expressed exclusively in hepatocytes, beginning in midgestation and continuing for the life of the animal, and the specificity of the albumin promoter is very well defined in transgenic studies (24). Because the albumin promoter is continually expressed in hepatocytes at all stages, cre-mediated mutation of the RXR gene is expected to be a unidirectional and cumulative process.

View larger version (48K): [in this window] [in a new window]   FIG. 1.   Efficiency of recombination of the conditional RXR allele by Alb-cre. (A) Diagrammatic representation of the chromosomal organization of the RXR alleles described in this study (adapted from reference 4). The upper line represents the wild-type (wt) allele, the middle line represents the conditional (floxed) allele, and the bottom line represents the conditional allele after cre-mediated recombination (floxed out). Exons 3 and 4 are indicated as boxes (e3 and e4, respectively), and loxP sites are indicated by solid triangles. Brackets underneath each line indicate the sizes of restriction fragments after BamHI (B) and BglII (Bg) double digestion, as illustrated in Southern blots (panel B). The cross-hatched block indicates the location of the probe used in Southern blots. Small arrows and brackets above each line indicate PCR primers and sizes of amplification products, as illustrated in panel C. Sizes are in kilobases. The diagram approximates but is not drawn to scale. (B) Time course of recombination. Genomic DNA from liver tissue from Alb-cre transgenic, RXRflox/wt animals at the indicated embryonic (E; measured from day of conception) or postnatal (measured from birth) time points was isolated and analyzed by Southern blotting. The wild-type (wt) band, which is not affected by recombination, serves as an internal standard; the positions of the nonrecombined (floxed) and recombined (floxed out) conditional RXR fragments are indicated. (C) Specificity of recombination. Genomic DNA from the indicated tissues of a 7-wk-old Alb-cre transgenic, RXRflox/wt animal was analyzed by PCR amplification. The primers used amplify to different product lengths the wild-type, nonrecombined, and recombined RXR alleles. In samples from RXRflox/wt animals in which the conditional allele is not recombined, there is an additional band intermediate in size between the wild-type and floxed products, which represents incomplete extension in one cycle from the conditional allele and subsequent extension on the wild-type template in a later amplification cycle. To the left are size standards from genomic tail DNA from unrelated animals. sm. int., small intestine; lg. int., large intestine. (D) RNA analysis of the conditional allele. To the left is a diagrammatic representation of a portion of the transcripts derived from the conditional allele (above; note this transcript is identical to a wild-type transcript) and from the conditional allele lacking exon 4 after recombination (below). The arrows and brackets indicate primer locations and sizes (in base pairs) of amplification products. To the right is the result of reverse transcription-PCR analysis of RXR mRNA using total liver RNA derived from 3-month-old mice which were RXRflox/flox and either did not () or did (+) carry the Alb-cre transgene. In cre-transgenic liver tissue, the predominant product is deleted for exon 4 sequences, but a trace amount of normal-size product is also seen, presumably representing expression of wild-type RXR mRNA in nonhepatocyte cell types.

Absence of RXR results in altered hepatic gene expression and physiology. We first addressed with mice fed a standard rodent diet and not subjected to physiological stress whether absence of RXR altered the basal expression pattern of any of the class II nuclear receptor genes that are expressed in the liver (Fig. 2). The expression of RXR and RXR was unchanged in mutant tissue relative to the wild type, indicating that a compensatory up-regulation of these RXR genes does not occur in the absence of RXR. The expression of RXR, LXR, FXR/RIP14, CAR, and HNF4 was also unchanged. The hepatic level of PXR ranged up to twofold higher in mutant livers relative to levels in wild-type animals, but this increase was not statistically significant. However, twofold increases in the levels of RAR, RAR, and PPAR were observed. It is currently uncertain to what extent these increases are physiologically relevant.

View larger version (69K): [in this window] [in a new window]   FIG. 2.   Expression of nuclear receptor genes in RXR-deficient liver tissue. (A) Retinoid receptor genes. (B) Additional class II heterodimeric partners of RXR. Blots were prepared with RNA isolated from wild-type (+/+) and deficient (/) liver tissue and sequentially hybridized with the probes indicated. RNA was isolated from eight wild-type and eight mutant livers to control for variations in diet, hormonal status, and/or other physiological parameters that might influence gene expression, although these and other panels in this study present the results from three samples per genotype. RIP14 is also known as FXR.

View larger version (566775K): [in this window] [in a new window]   FIG. 3.   Target gene expression in RXR-deficient liver tissue. (A) PPAR target genes. (B) Apolipoprotein genes. (C) Target genes for LXR, FXR, CAR, and PXR. Experimental details are as described for Fig. 2.

View larger version (29K): [in this window] [in a new window]   FIG. 4.   Lipid parameters in normal and RXR-deficient mice fed a standard or high-cholesterol diet. (A) Liver/body mass ratio. (B) Hepatic cholesterol. (C) Hepatic triglyceride. (D) Serum cholesterol. (E) Serum triglyceride. The data shown were obtained from age-matched female mice of the indicated genotypes maintained for 3 months either with a low-cholesterol or high-cholesterol diet; where investigated, results for male mice were generally very similar to those for females. The results shown are means ± standard errors.

Failure in cholesterol homeostasis in RXR-deficient mice. In order to define the extent to which the absence of RXR compromises not only the basal status of hepatic class II nuclear receptor pathways, but also the response to physiological stress, we challenged our mice with a high-cholesterol (2% by weight) diet. Excess dietary cholesterol is accommodated in normal mice by increasing bile acid production and consequent fecal elimination, a process which requires LXR (22). Our approach was similar to the previous evaluation of the LXR^/ mouse, and in a manner which is very similar to the LXR knockout phenotype, the livers of mice lacking RXR specifically in hepatocytes and fed the high-cholesterol diet became progressively fatty and enlarged. After 1 week of treatment, the livers of mutant mice were already obviously more pale than those of wild-type mice or mutant mice provided with a low-cholesterol diet (data not shown), and by 1 month of treatment, the livers of these mice were almost completely white (Fig. 5A and B). The liver/body weight ratio in liver-specific RXR mutant mice maintained on the high-cholesterol diet for 1 month was twice that of normal controls fed the same diet (data not shown), and after 3 months (the longest period measured in this study), it was triple that of control mice (Fig. 4A). Even with such pronounced hepatomegaly, mutant mice on the 2% cholesterol diet were normal in externally apparent parameters and were not moribund, lethargic, or sick.

View larger version (100K): [in this window] [in a new window]   FIG. 5.   Morphology and histology of mouse livers. Mice were fed either a low-cholesterol diet (A, C, and E) or a 2% cholesterol diet (B, D, F, G, and H) for 1 (A and B) or 2 (C to H) months. (A and B) Morphology of whole livers from hepatocyte-specific RXR mutant mice. The normal appearance of livers from mutant mice on a standard diet (A) is identical to that of wild-type mice fed either diet (data not shown). (C and D) Histology of wild-type mouse livers. Challenge with a high-cholesterol diet does not alter histology. (E and F) Histology of mutant mouse livers. Although the untreated mutant tissue is of normal appearance (E), there is a substantial increase in hepatocyte cell size and vacuole material in high-cholesterol-fed mice (F). Panels C to F are hematoxylin-and-eosin stained and are at the same magnification; the scale bar in panel F is 50 µm. (G and H) Oil red O staining of liver tissue. When viewed under normal light (G), hepatic triglyceride (darker red staining) is preferentially located close to blood vessels, whereas hepatic cholesterol (lighter staining) is accumulated more distally. Panel H is the same section as panel G, but viewed with polarized light to illuminate cholesterol crystals (which appear white). The inset is a close-up view of a superimposition of panels G and H. Arrows in panels G and H and the inset are provided for register and point to the same two small blood vessels. The scale bar in panel G is 100 µm.

/

/

View larger version (40K): [in this window] [in a new window]   FIG. 6.   Attenuated induction of Cyp7A by dietary cholesterol in RXR-deficient mice. Hepatic Cyp7A message levels in RXR-deficient mice provided with a low (control)- or high-cholesterol diet for 7 days are shown. There is a two- to threefold induction of Cyp7A between the first two pairs of animals and no induction in the second two pairs. The mice used in these assays are derived from a mixed-strain background, which might account for some of the variability in the level and inducibility of Cyp7A expression.

/

/

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The RXRs were initially identified on the basis of their ability to promote DNA binding of certain nuclear receptors in cell-free assays (33). Numerous biochemical and transfection assays performed since these initial observations have supported the model in which RXRs are obligate heterodimeric partners of the class II receptors and in fact have served to define the class II subfamily as requiring heterodimerization with RXR. Nonetheless, the in vivo evidence for the role of heterodimerization has so far only been established for RXR-RAR, based primarily on the analysis of RXR and RAR compound mutant mice (10, 20). In this study, we extend this principle beyond the more closely related retinoid receptor subfamily (i.e., RXRs and RARs) to include LXR, PPAR, CAR, PXR, and FXR, with the demonstration that physiological pathways regulated by each of these receptors are compromised in the absence of RXR.

In addition to confirming the previously suspected role of RXR heterodimeric partners in regulating hepatic physiology, our results also provide insight into which of the nuclear receptor pathways are most critical in the regulation of certain pathways in vivo. For example, Cyp7A expression is upregulated by LXR and downregulated by FXR; the substantial increase in expression we observed after RXR mutation suggests that in normal mice on a standard diet, the dominant regulation of the Cyp7A gene is negative regulation via the RXR-FXR heterodimer. Similarly, of the multiple nuclear receptor complexes which have been defined by transfection assays to regulate the ApoAI and ApoCIII genes (PPAR, HNF4, ARP, EAR, etc.), our results demonstrate the importance of the RXR-PPAR complex, although they do not necessarily discount the involvement of these additional receptors in apolipoprotein expression under basal or stressed conditions. In some cases, some ambiguity in defining the involvement of potentially competing receptor pathways remains. For example, the fourfold decrease in Cyp3A1 expression we observed as a consequence of RXR mutation could result from reduction of an RXR-CAR or RXR-PXR complex, because both heterodimers have been implicated in Cyp3A1 regulation. It is also possible, perhaps even likely, that both receptors simultaneously regulate the basal level of this target gene.

Some of the target genes we explored gave surprising results. The gene coding for acyl-CoA oxidase is a prototype PPAR target gene, being one of the first of the fatty acid oxidation pathway genes to be defined as fibrate inducible and with an element in its promoter that was among the first to define the PPAR response element (12). In vivo, however, acyl-CoA oxidase expression is not altered by RXR mutation. Similarly, the ApoCIII promoter is inhibited by fibrate (similar to the response of ApoAI) and contains a well-defined PPAR response element. The increase in ApoCIII expression we observed in the absence of RXR is consistent with removal of an inhibitory regulatory agent, which might be the RXR-PPAR heterodimer. However, in PPAR^/ mice, the ApoCIII gene is normally expressed (23). While we cannot account for these results at this time, one possibility is that these genes are controlled by other regulatory pathways in vivo, either independent of or in addition to the known pathways. There may also be compensatory pathways that become active in the absence of RXR and influence gene expression in a secondary manner. It is also possible, at least for acyl-CoA oxidase, that the presence of RXR or RXR allows normal expression of the gene coding for this product to occur, despite the alteration in expression of other PPAR target genes that occurs in the absence of RXR.

The RAR gene is regulated through a response element that is activated by RXR heterodimers with RAR (31). In the absence of RXR, expression of the RAR gene is increased (Fig. 2A), which is consistent with the removal of an inhibitory agent. Possibly, this inhibitor is RXR-RAR, which like certain other nuclear receptor heterodimers has been demonstrated to have silencing activity in the absence of ligand via recruitment of transcriptional corepressors (5). In this context, it will be particularly interesting to explore thyroid hormone-regulated gene expression in our mice, because the thyroid receptor has a very strong transcriptional repression domain in the absence of ligand (13).

We chose to examine the cholesterol pathway in detail because the LXR knockout phenotype has been characterized and is particularly striking. Thus, challenge with a high-cholesterol diet offers the most direct way to define RXR as the in vivo partner of one of its class II partners, not only under standard conditions, but also under physiological stress. In fact, the liver phenotypes of both types of animals in response to a high-cholesterol diet are quite similar, being manifested in an accumulation of hepatic cholesterol and an elevated serum cholesterol level. It should be noted that there are differences between the two animal models as well (in basal hepatic and serum lipid content, in the extent of hepatic cholesterol accumulation after the high-cholesterol diet, and in Cyp7A expression). We cannot at the moment explain these differences, although we suggest that the absence of RXR alters multiple physiological pathways in addition to LXR that collectively account for the differences seen. Nonetheless, our results indicate that RXR is required for at least a large component of the normal cholesterol responsiveness through LXR. There may be pharmaceutical relevance to this observation, because the RXR-LXR heterodimer is activated independently by retinoids and by oxycholesterols (35). It may be possible, for example, to stimulate the elimination of cholesterol with RXR-specific agonists.

The results of this study also indicate that the primary site of LXR action in the process of cholesterol metabolism is in hepatocytes. Bile acid homeostasis is a highly dynamic process involving synthesis in hepatocytes, storage and secretion in the gall bladder, and reabsorption in the intestine. In principle, cholesterol homeostasis might primarily require LXR action in the gall bladder or intestine (where LXR is abundantly expressed), with hepatic events being a secondary or minor aspect of cholesterol physiology. The cholesterol phenotype seen in LXR mutant mice does not address this issue, because the LXR gene is mutated in all tissues of the animal. However, by specifically disrupting RXR, the LXR partner receptor, in hepatocytes, the critical role of LXR in the liver is established.

Many of the class II nuclear receptor genes have been mutated in the mouse germ line, providing the opportunity to compare the hepatic phenotypes of these knockouts with the RXR liver-specific mutant phenotype, as we have done here for LXR. We have found that certain components of the fibrate-inducible PPAR pathway are also compromised in the absence of RXR (Y. Wan et al., unpublished observations), consistent with the alterations we observed in the expression profile of PPAR target genes in mice maintained on a standard diet and without physiological challenge. Perhaps more valuable will be the opportunity to address with our mice the hepatic role of class II receptors for which germ line mutations have not yet been established, including the FXR, CAR, and PXR receptors and their cognate ligands.

The central role of RXR as a heterodimeric partner for at least 10 different vertebrate class II nuclear receptor gene subfamilies (1) regulating very diverse developmental and physiological processes has been somewhat enigmatic. This relationship is evolutionarily broad, in that the Drosophila ultraspiracle locus encodes an RXR homolog which is required as the heterodimeric partner of the ecdysone receptor (36). One of the most intriguing aspects of our study is the realization that multiple metabolic pathways in the mouse liver are coordinately dependent upon RXR. Fatty acid and cholesterol metabolism in particular are ancient processes required in all metazoan organisms that must regulate lipid absorption, storage, and utilization. The heterodimeric dependence of class II nuclear receptors on RXR may have initially evolved specifically to coordinately regulate these fundamental metabolic pathways. Once in place, the class II subfamily may have expanded and evolved to accommodate new functions and ligands (such as retinoic acid, which is unique to vertebrates, and ecdysteroids, which are found in insects), yet preserved the requirement for RXR heterodimerization.