Disruption of the Lecithin:Retinol Acyltransferase Gene Makes Mice More Susceptible to Vitamin A Deficiency*

Lecithin:retinol acyltransferase (LRAT) catalyzes the esterification of retinol (vitamin A) in the liver and in some extrahepatic tissues, including the lung. We produced an LRAT gene knock-out mouse strain and assessed whether LRAT-/- mice were more susceptible to vitamin A deficiency than wild type (WT) mice. After maintenance on a vitamin A-deficient diet for 6 weeks, the serum retinol level was 1.34 ± 0.32 μm in WT mice versus 0.13 ± 0.06 μm in LRAT-/- mice (p < 0.05). In liver, lung, eye, kidney, brain, tongue, adipose tissue, skeletal muscle, and pancreas, the retinol levels ranged from 0.05 pmol/mg (muscle and tongue) to 17.35 ± 2.66 pmol/mg (liver) in WT mice. In contrast, retinol was not detectable (<0.007 pmol/mg) in most tissues from LRAT-/- mice after maintenance on a vitamin A-deficient diet for 6 weeks. Cyp26A1 mRNA was not detected in hepatic tissue samples from LRAT-/- mice but was detected in WT mice fed the vitamin A-deficient diet. These data indicate that LRAT-/- mice are much more susceptible to vitamin A deficiency and should be an excellent animal model of vitamin A deficiency. In addition, the retinol levels in serum rapidly increased in the LRAT-/- mice upon re-addition of vitamin A to the diet, indicating that serum retinol levels in LRAT-/- mice can be conveniently modulated by the quantitative manipulation of dietary retinol.

Vitamin A and its derivatives (retinoids) regulate cell proliferation and differentiation. All-trans-retinoic acid (RA) 3 is the most active retinoid (1). At a molecular level, RA acts by binding and activating the nuclear receptors, which are transcription factors that directly regulate the transcription of their target genes (2,3). RA inhibits the proliferation of many types of cells in part by regulating the expression of several cell cycle proteins (4 -10). Data indicate that vitamin A deficiency plays a role in increasing the incidence of carcinogenesis (4,5,(11)(12)(13)(14)(15)(16)(17)(18).
Mammals acquire vitamin A from their diet. Within the intestinal lumen dietary retinyl esters are hydrolyzed to retinol (19), which is re-esterified in the enterocytes and then packaged into the nascent chylomicrons (20). The majority of the dietary vitamin A (ϳ75%) is taken up by the hepatocytes (21). Within the hepatocytes, the retinyl esters are hydrolyzed to retinol, which can either be secreted or transferred to the hepatic stellate cells and stored as retinyl esters (20 -22). The lung is one of the most important extrahepatic organs for retinol storage (23)(24)(25)(26). The esterification of retinol in the intestine, liver, and lung has been reported to be catalyzed by the enzymes lecithin:retinol acyltransferase (LRAT) and acyl-CoA:retinol acyltransferase (20,(27)(28)(29)(30). The cellular retinol-binding protein type I (CRBP-I) and cellular retinol-binding protein type II (CRBP-II) play important roles in modulating the activity of LRAT (20,30,31). Retinol bound to CRBP-II (in enterocytes) and retinol bound to CRBP-I (in liver) are the preferred substrates of LRAT, and retinol bound to CRBPs is not efficiently esterified by acyl-CoA: retinol acyltransferase (20,30,32,33). Recently, the enzyme DGAT1 was reported to catalyze acyl-CoA-dependent esterification of retinol and thus may represent the acyl-CoA:retinol acyltransferase activity measured previously by various laboratories (34). LRAT activity is increased following treatment with retinoids (35)(36)(37)(38).
Retinyl esters stored in the liver and extrahepatic tissues are responsible for the maintenance of systemic and local retinol levels during periods in which dietary vitamin A is low or absent (20,39). Mice deprived of vitamin A at 3 weeks postpartum display normal serum retinol concentrations for up to 20 weeks (39). The liver is the organ in which up to 80% of the total retinol and retinyl esters in vertebrates is stored (19,20).
We have generated an LRAT gene knock-out mouse strain in which only trace amounts of retinyl esters (less than 0.2% of the value in wild type mice) could be detected in the liver and lung. These data are consistent with a recent report (40) from another laboratory that retinyl esters are greatly reduced in the liver and lung by LRAT gene disruption. We then examined the retinol and retinyl ester levels in serum and in some other tissues from the wild type and knock-out mice at various times after maintenance of the mice on a vitamin A-deficient diet. Our results show that the LRAT Ϫ/Ϫ mice become vitamin A-deficient in a period as short as 6 weeks following removal of vitamin A from their diet. Thus, the LRAT Ϫ/Ϫ mice rapidly become vitamin A-deficient and have many advantages over WT mice in studies of vitamin A deficiency.

MATERIALS AND METHODS
Construction of the LRAT Gene Targeting Vector and the Generation of LRAT Ϫ/Ϫ Mice-A genomic clone (ϳ18 kb) was isolated from a 129SVJ mouse genomic library (Stratagene, catalog number 946309) and was used to construct the gene targeting vector. The probe for the genomic library screening was a PCR fragment (552 bp), which was amplified from mouse genomic DNA with the primers 5Ј-ACGCGGG-GACCGCCCCCAGGCACACTACC-3Ј and 5Ј-GCTTGTTGGAG-ACCACCTTCTGAGTGCG-3Ј in the LRAT gene. The targeting vector contained a 3.9-kb 5Ј-arm and a 9-kb 3Ј-arm. The entire exon 1, which included the ATG start codon for the LRAT protein, was replaced with a neo gene cassette that was flanked by loxP sites. The neo/loxP fragment was inserted between the SacI and SpeI sites (Fig.  1A). An HSV-TK cassette was ligated as the upstream sequence adjacent to the 5Ј-arm. The targeting vector was introduced into 129SVJ mouse embryonic stem (ES) cells (CJ7 line, obtained from Cornell/ Sloan Kettering Transgenic Mouse Facility) by electroporation using a Bio-Rad electroporator at a voltage of 230 mV and capacitance of 500 microfarads. Positive clones were isolated with G418 selection (300 g/ml active G418) and expanded. A total of 96 cell clones were screened, and three positive clones were obtained. The clones that had undergone homologous recombination were identified by Southern blot analysis and were confirmed by flanking probes outside of the construct (P1 and P2 as indicated in Fig. 1A) from both sides (Fig. 1B). After removing the neo/loxP cassette by transient transfection with a Cre expression vector (pBS185, Invitrogen, catalog number 10347-011), the heterozygous LRAT ϩ/Ϫ mouse ES cells were injected into C57BL/6J blastocysts by the Cornell/Sloan Kettering Transgenic Mouse Facility. The LRAT gene knock-out mice were generated from the chimeric mice (5 chimeras, Ͼ90% coat color chimerism) that gave germ line transmission. The mouse genotype was determined by LRAT genespecific primers in a PCR using tail genomic DNA as the template. The upstream and downstream PCR primers (G1 and G2 as indicated in Fig.  1A) were 5Ј-TCTGGCATCTCTCCTACGCTG-3Ј and 5Ј-GTTC-CAAGTCCTTCAGTCTCTTGC-3Ј, respectively. A 1.3-kb band from the wild type allele and a 0.5-kb band from the mutant allele were amplified (Fig. 1C). The expression of the LRAT gene in mouse tissues was examined by semi-quantitative RT-PCR.
Deprivation of Dietary Vitamin A-The pups (C57BL/6J background) were weaned 3 weeks postpartum, and the genotypes were determined by PCR. At 4 weeks, wild type and LRAT Ϫ/Ϫ mice were divided randomly into two groups. One group was fed the normal vita-min A-sufficient diet (control diet, Harlan TD 88406 supplies ϳ19.8 IU (5.94 g) of vitamin A/g as retinyl palmitate), and the other group was fed a vitamin A-deficient diet (Harlan TD 88407). Serum samples were obtained from the mouse tails at 2-, 4-, and 6-week intervals. At 6 weeks, mice were sacrificed, and tissue samples were harvested in the dark and stored (without rinsing) at Ϫ70°C until the retinoids were extracted.
Analyses of Tissue Retinoids-The frozen tissue samples (ϳ45-250 mg) from the liver, lung, eye, kidney, pancreas, adipose, skeletal muscle (from quadriceps femoris muscles), brain (from cerebra), and tongue were weighed and homogenized in cold phosphate-buffered saline (PBS) (100 l of PBS per 100 mg of tissue) in glass homogenizers and then transferred into Eppendorf tubes. The volumes were adjusted to 500 l with cold PBS. The retinoids were extracted into 350 l of organic solution (acetonitrile/butanol, 50:50, v/v) in the dark as described previously (41,42). The high performance liquid chromatography (HPLC) was performed using a Waters Millenium system (Waters). Each sample (100 l of the 350 l) was loaded onto an analytical 5-m reverse phase C 18 column (Vydac, Hesperia, CA) and eluted at a flow rate of 1.5 ml/min. Two mobile phase gradient systems were used, as described previously (41,42). The amounts of retinoids were calculated from the areas under the peaks detected at the wavelength of 340 nm. The levels of retinol and retinyl esters were normalized to the tissue weight.
Semi-quantitative RT-PCR-The tissues from the liver, lung, kidney, small intestine, and brain were dissected and stored in RNAlater (Ambion) at Ϫ70°C. Total RNA was extracted using the mini RNAeasy columns (Qiagen). Total RNA (1 g each) was used for reverse tran- Restriction sites present in the gene are as follows: E, EcoRV, B, BamHI; Bst, BstZ17I, loxP site. Note that the labeling does not represent exact scale. The probes (P1 and P2) for testing clones by Southern analysis are shown. B, identification of heterozygous LRAT ϩ/Ϫ mouse ES clones by Southern blot analysis. In the left panel genomic DNA (20 g) from each clone was digested with EcoRV and loaded on a 1% agarose gel. The Southern blot was performed with a 5Ј-flanking probe (P1). The positive clones were identified based on the 4.2-kb band. In the right panel genomic DNA was extracted from the ES clones after the neo cassette was removed. The DNA was digested with BamHI and EstZ17I. The Southern blot was performed with a 3Ј-flanking probe (P2). The positive heterozygous clones were confirmed, based on the expected 14-kb band and 10-kb band of the WT allele. C, genotyping. Genomic DNA was extracted from each mouse tail. The genotypes were assessed by PCR using G1 and G2 primers. PCR bands of 1.3 and 0.5 kb were amplified from the wild type allele and the knock-out allele, respectively. scription in a 20-l reaction using Super Script TM II Reverse Transcriptase (RT) from Invitrogen, according to the manufacturer's instructions. The cDNA thus produced was diluted to 100 l with diethyl pyrocarbonate/water, of which 1 l was used in a PCR as follows. The gene-specific primers and the number of cycles for PCR are listed as follows. For LRAT (GenBank TM accession number AF255061), the 5Јprimer 5Ј-CTGACCAATGACAAGGAACGCACTC-3Ј and 3Ј-primer 5Ј-CTAATCCCAAGACAGCCGAAGCAAGAC-3Ј were used for 40 cycles, and a 370-bp product was expected. For Cyp26A1 (GenBank TM accession number NM_007811), the 5Ј-primer 5-GCAGATGAAGCG-CAGGAAATACG-3 and 3Ј-primer 5-CCCACGAGTGCTCAATCA-GGA-3 were used for 36 cycles, and a 635-bp product was expected. For the dehydrogenase/reductase member 1 (Dhrs1, GenBank TM accession number NM_026819), the upstream primer 5-GACTTCCATCCAGG-CACGCACTC-3 and downstream primer 5-CATCTAGCCGCCCTT-TCTGTTCC-3 were used for 26 cycles, and a 388-bp product was expected. All of these primers are located in two different exons so genomic DNA contamination (if it exists) does not interfere with RNA analyses. Primers for the testis-specific mRNAs, the activator of CREM in testis (5Ј-primer, 5Ј-ATGACAAGTAGTCAATTTGATTGT-3Ј, and 3Јprimer, 5Ј-CTAAGCGTCAGTGTCTGCCC-3Ј), cAMP-responsive element modulator (5Ј-primer, 5Ј-ATGACCATGGAAACAGTTGA-ATC-3Ј, and 3Ј-primer, 5Ј-TGGTAAGTTGCCATGTCACC-3Ј), sperm fibrous sheath component (5Ј-primer, 5Ј-TATGCTTTGTGG-ATGTGTCT-3Ј, and 3Ј-primer, 5Ј-TGAGGAGCCAGTTGAGGAC-A-3Ј), mitochondrial capsule selenoprotein (5Ј-primer, 5Ј-GACTCAC-TAGACTGCTGAGGA-3Ј, and 3Ј-primer, 5Ј-CACCTGGTGGGGA-CTGTGG-3Ј), transition protein 1 (5Ј-primer, 5Ј-ATGTCGACCAG-CCGCAAGC-3Ј, and 3Ј-primer, 5Ј-CCACTCTGATAGGATCTTTG-G-3Ј), proacrosin (5Ј-primer, 5Ј-GTGTGTGCAGGGTATCCTGAA-GGCAAGATT-3Ј, and 3Ј-primer, 5Ј-GCGAAGAGAAGGAGGGTG-AAAAGAGACCAT-3Ј), protamine 1 (5Ј-primer, 5Ј-ATGGCCAGAT- Total RNA was extracted from wild type and LRAT Ϫ/Ϫ mouse tissues. Expression of LRAT mRNA in the kidney, small intestine, brain, lung, and liver was detected by RT-PCR. The 5Ј-primer (L1) for PCR was located in the exon 1 of LRAT gene, and the 3Ј-primer (L2) was located in exon 2. A 370-bp band was expected. GAPDH was used as a control mRNA for loading. Three pairs of mice were tested, and one representative result is shown. B, histological changes in testes displayed by hematoxylin/eosin staining. Testes were from 4-week-old and 3-month old LRAT ϩ/ϩ (WT) and LRAT Ϫ/Ϫ (knock-out) mice. For analyses of testis-specific mRNAs, total RNA was extracted from the testes of 4-week-old WT and LRAT Ϫ/Ϫ mice. The sequences of primers and conditions for PCR were according to Zhang et al. (43). These experiments were repeated three times with very similar results. ACCGATGCTG-3Ј, and 3Ј-primer 5Ј-GTGGCGAGATGCTCTTGA-AG-3Ј), and protamine 2 (5Ј-primer, 5Ј-ATGGTTCGCTACCGAAT-GAGG-3Ј, and 3Ј-primer ,5Ј-TTAGTGATGGTGCCTCCTACA-3Ј), were as described previously (43). GAPDH (upstream primer, 5-ACC-ACAGTCCATGCCATCAC-3; downstream primer, 5-TCCACCAC-CCTGTTGCTGTA-3, 23 cycles, expected product size 451 bp) was used as an internal control for the semi-quantitative RT-PCR. The PCR was performed using the following conditions: 94°C for 30 s, 58°C for 30 s, and 72°C for 80 s, with a final extension at 72°C for 10 min. Taq polymerase was from Invitrogen (catalog number 18038-042). The PCR products were subjected to agarose gel electrophoresis. The gel images stained with ethidium bromide were recorded with a FluorChem 8800 system (Alpha Innotech, San Leandro, CA).
Statistical Methods-Each experiment was repeated at least three times. The means Ϯ S.E. were calculated using Graphpad Prism Program (version 4.0a). One-way analysis of variance was used to analyze differences in the mean values of retinol levels among multiple experimental groups. The unpaired t test was used to determine differences between two groups. Differences with a p value of Ͻ0.05 were considered to be statistically significant.

RESULTS
Production of the Homologous LRAT Ϫ/Ϫ Mouse Line-To address the physiological functions of the LRAT gene, a targeting vector was constructed ( Fig. 1A), and the LRAT gene was disrupted by homologous recombination. That the homologous recombination was correct was demonstrated by using both 5Ј-and 3Ј-flanking probes (Fig. 1B). The neo gene cassette was removed from two ES cell clones, and the clones without the neo cassette were used to produce the LRAT gene knockout mouse line. This is in contrast to the LRAT Ϫ/Ϫ mice produced by another laboratory, which still contain the neo R gene (40). The genotype was determined by PCR (Fig. 1C). The expression of LRAT mRNA in the kidney, small intestine, brain, liver, and lung is shown ( Fig. 2A). The LRAT transcript was expressed in all tissues tested in wild type mice. In contrast, no expression of the LRAT gene was detected in any of the tissues tested from LRAT Ϫ/Ϫ mice. The results of the RT-PCR were consistent with the Northern blot analysis using the exon 2 cDNA fragment as a probe (data not shown).
To explore the physiological consequences of the disruption of the LRAT gene, a complete necropsy was performed on 4-week-old and 3-month-old mice at the Genetically Engineered Mouse Phenotype Core Facility of Cornell/Memorial Sloan-Kettering Cancer Center. With a sufficient source of dietary vitamin A (25 IU (or 7.5 g)/g), the LRAT Ϫ/Ϫ mice were apparently healthy. However, the testes of the LRAT Ϫ/Ϫ mice at the age of 4 weeks and at 3 months showed marked hypoplasia and oligospermia (Fig. 2B). In the 4-week-old LRAT Ϫ/Ϫ mouse, diffuse testicular hypoplasia/atrophy was prominent. The LRAT Ϫ/Ϫ mice also exhibited oligospermia with the absence of sper-matogenic epithelium, dilated seminiferous tubular lumens, and occasional luminal giant cells. A similar but milder change was observed in 3-month-old male LRAT Ϫ/Ϫ mice. The results of the histological examinations are shown (Fig. 2B). The expression in testis of a panel of testisspecific genes was analyzed at the mRNA level by RT-PCR. The expression of these genes in LRAT Ϫ/Ϫ testicular tissues was similar to that seen in WT testicular tissues (Fig. 2B). These analyses indicate that there is an absence of mature sperm in the testes of LRAT Ϫ/Ϫ mice but that the process of spermatogenesis may not be completely disrupted. No other gross differences were seen between the wild type and LRAT Ϫ/Ϫ mice on histological examination with conventional staining.
Levels of Retinol and Retinyl Esters in the Liver and Lung-We then examined the retinol and retinyl ester levels in two major organs for retinol storage, liver and lung, in 10-week-old wild type and LRAT Ϫ/Ϫ mice. These mice were fed a normal vitamin A-sufficient diet for the 6 weeks after weaning; the analytical results are shown (Fig. 3). The levels of retinyl esters detected in the LRAT Ϫ/Ϫ livers and lungs were 3.41 Ϯ 0.41 and 1.29 Ϯ 0.43 pmol/mg, respectively, which were 0.15 and 0.08% of the retinyl esters present in the livers (2227.46 Ϯ 436.40 pmol/mg) and lungs (1619.33 Ϯ 373.80 pmol/mg) of the wild type mice (TABLE  TWO).
Disruption of the LRAT gene also resulted in a decrease in retinol levels in the liver and lung. The retinol levels were 2.20 Ϯ 0.39 pmol/mg in the livers and 1.45 Ϯ 0.30 pmol/mg in the lungs of the LRAT Ϫ/Ϫ mice, which were 4.27 and 4.89%, respectively, of the levels seen in the livers (51.54 Ϯ 11.91 pmol/mg) and lungs (29.68 Ϯ 5.92 pmol/mg) of WT mice (TABLE TWO). These data show that both the retinyl esters and retinol levels in these tissues are drastically reduced in the LRAT Ϫ/Ϫ mice even on a normal vitamin A-sufficient diet.
Serum Retinol Levels in Mice Fed a Vitamin A-deficient Diet-To test whether LRAT Ϫ/Ϫ mice were more susceptible to vitamin A deficiency, the wild type and LRAT Ϫ/Ϫ mice were placed either on a control, vitamin A-sufficient diet or a vitamin A-deficient diet at 4 weeks of age. The concentration of retinol in the serum was measured at 2, 4, and 6 weeks after initiation of the vitamin A-deficient diet. The concentration of serum retinol ranged from 1.41 Ϯ 0.29 (mean Ϯ S.E.) to 1.68 Ϯ 0.18 M in wild type mice fed the control diet, whereas in the LRAT Ϫ/Ϫ mice fed the control diet it ranged from 1.64 Ϯ 0.16 to 1.78 Ϯ 0.06 M. There were no significant differences in the concentrations of serum retinol between WT and LRAT Ϫ/Ϫ mice fed the vitamin A-sufficient (control) diet (p Ͼ 0.05, one-way analysis of variance).
In wild type mice fed the vitamin A-deficient diet, the serum retinol was 1.   DECEMBER 2, 2005 • VOLUME 280 • NUMBER 48 serum retinol concentration was significantly lower than that seen in WT mice on the vitamin A-deficient diet (p Ͻ 0.05, unpaired t test). The quantitative results of the serum retinol determinations are tabulated (TABLE ONE). These results show that depletion of the serum retinol in the LRAT Ϫ/Ϫ mice was significantly faster than that seen in wild type mice after initiation of the vitamin A-deficient diet.

Analysis of Tissue Retinoids in WT and LRAT
Ϫ/Ϫ Mice-After a period of 6 weeks on the vitamin A-deficient diet, retinoids were extracted from the liver, lung, eye, kidney, pancreas, adipose tissue, skeletal muscle, brain, and tongue of wild type and LRAT Ϫ/Ϫ mice and subjected to HPLC analyses. The levels of tissue retinol and retinyl esters are shown in TABLE TWO. The levels of retinyl esters and retinol in tissues from wild type and LRAT Ϫ/Ϫ mice, fed either the vitamin A-sufficient (control) or vitamin A-deficient diet, are as follows.
Tissue Retinyl Esters from Mice Fed the Control, Vitamin A-sufficient Diet for 6 Weeks-The levels of retinyl esters in most other extrahepatic tissues of LRAT Ϫ/Ϫ mice, including the eye, kidney, pancreas, muscle, brain, and tongue, were lower than the levels in the same tissues of WT mice even when the LRAT Ϫ/Ϫ mice were maintained on a control, vitamin A-sufficient diet. The adipose tissue from LRAT Ϫ/Ϫ mice was the only tissue type that showed a higher level of retinyl esters than that of WT mice (3.79 Ϯ 0.85 pmol/mg in LRAT Ϫ/Ϫ versus 1.49 Ϯ 0.55 pmol/mg in WT) (TABLE TWO). These data show that in the LRAT Ϫ/Ϫ mice the aforementioned tissues contained lower amounts of retinyl esters.
Tissue Retinyl Esters from Mice Fed the Vitamin A-deficient Diet for 6 Weeks-As compared with the WT mice on a vitamin A-sufficient, control diet, the levels of retinyl esters in the livers and lungs of WT mice decreased to 58.68 and 23.76% of control levels, respectively, after being fed the vitamin A-deficient diet for 6 weeks. Thus, the internally stored retinyl esters are mobilized in WT mice in the absence of dietary retinol. The levels of retinyl esters in other tissues, including the eye, kidney, pancreas, adipose tissue, muscle, and brain, were also decreased in the WT mice on the vitamin A-deficient diet (TABLE TWO). In the LRAT Ϫ/Ϫ mice no significant mobilization of retinyl esters was found in the absence of dietary vitamin A (TABLE TWO). Of course, the retinyl ester levels are extremely low in the LRAT Ϫ/Ϫ animals on both the vitamin A-sufficient and vitamin A-deficient diets (TABLE TWO).
Tissue Retinol Levels in Mice Fed the Control Diet for 6 Weeks-As described above, the levels of retinol in the livers and lungs of WT mice were higher than those seen in LRAT Ϫ/Ϫ mice. The levels of retinol in the eyes of WT mice were also higher than those in LRAT Ϫ/Ϫ mice. In contrast, the levels of retinol in most other tissues, including the kidney, adipose tissue, muscle, brain, and tongues of the WT mice, were lower than the levels in the tissues of LRAT Ϫ/Ϫ mice (TABLE TWO). These data indicate that LRAT activity is involved in the regulation of the availability of retinol to various tissues when mice are fed a vitamin A-sufficient diet.
Tissue Retinol Levels in Mice Fed the Vitamin A-deficient Diet for 6 Weeks-In the wild type mice retinol was detectable in all tissues, ranging from 0.05 (muscle and tongue) to 17.35 Ϯ 2.66 pmol/mg (liver). In contrast, in the LRAT Ϫ/Ϫ mice, retinol was only detected at trace levels in one hepatic sample (0.11 pmol/mg) and two kidney samples (0.07 pmol/mg and 0.08 pmol/mg), whereas retinol was not detectable in all other tissue samples (TABLE TWO). These results show that the LRAT Ϫ/Ϫ mice are more susceptible to vitamin A deficiency and become vitamin A-deficient more rapidly. An almost complete absence of retinol in tissues was observed in the LRAT Ϫ/Ϫ mice after 6 weeks on the vitamin A-deficient diet (TABLE TWO).

Retinol and retinyl esters in tissues
The levels of retinol and retinyl esters were measured by HPLC. Control (Ctrl) diet, vitamin A-sufficient control diet; VAϪ diet, vitamin A-deficient diet. Data are presented as mean Ϯ S.E. There were three samples in each group (n ϭ 3).

Gene Expression in the Wild Type and LRAT Ϫ/Ϫ Mice-The
Cyp26A1 is a cytochrome P450, which catalyzes the oxidation of RA to polar retinoids (44). The expression of Cyp26A1 gene has been correlated with the level of tissue retinol (45). We examined the expression of the Cyp26A1 gene in hepatic tissue in WT and LRAT Ϫ/Ϫ mice fed various diets. Our data show that Cyp26A1 mRNA was highly expressed in the liver of LRAT Ϫ/Ϫ mice on a vitamin A-sufficient diet, but expression was undetectable by RT-PCR after the knock-out mice were fed the vitamin A-deficient diet for 6 weeks (Fig. 7A). In contrast, expression of the Cyp26A1 transcript was observed in hepatic tissue from wild type mice fed either the control or the vitamin A-deficient diet for 6 weeks (Fig. 7A).
LRAT mRNA was not detectable in the hepatic samples from LRAT Ϫ/Ϫ mice (Fig. 7A). LRAT mRNA levels in the livers of WT mice were not affected by the diet (Fig. 7A). The expression of another gene involved in retinol metabolism, Dhrs1 (dehydrogenase/reductase member 1) (46, 47), was not influenced by either the LRAT gene knock-out or the type of diet (Fig. 7A).
Restoration of Vitamin A in Animals Fed the Vitamin A-deficient Diet-To determine whether the hepatic expression of the Cyp26A1 transcript would be restored upon the re-addition of vitamin A to the diet, a vitamin A-supplemented diet (Harlan, TD 96008, 250 IU (75 g) of vitamin A/g) was provided to the wild type and LRAT Ϫ/Ϫ mice that had been fed the vitamin A-deficient diet for 6 weeks. The expression of the Cyp26A1 transcript in the liver was examined by RT-PCR, and the concentration of serum retinol was measured by HPLC 48 h after the re-addition of vitamin A to the diet. The RT-PCR data showed that the Cyp26A1 gene was highly expressed in the hepatic tissues from both wild type and LRAT Ϫ/Ϫ mice upon readdition of vitamin A to the diet (Fig. 7B). Similarly, the serum retinol concentrations were increased in both WT and LRAT Ϫ/Ϫ mice following the re-addition of vitamin A to the diet (Fig. 7C). These results indicate that the hepatic expression level of the   . All-trans-retinol and retinyl esters in extrahepatic tissues from wild type and LRAT ؊/؊ knock-out mice. Eye, kidney, pancreas, adipose tissue, skeletal muscle, brain, and tongue were obtained from mice fed a vitamin A-sufficient control diet or a vitamin A-deficient diet for 6 weeks. The values correspond to mean Ϯ S.E. (n ϭ 3). RE, retinyl esters; Vit, vitamin; Ctrl, control. DECEMBER 2, 2005 • VOLUME 280 • NUMBER 48

JOURNAL OF BIOLOGICAL CHEMISTRY 40231
Cyp26A1 transcript is correlated with the tissue retinol level in the LRAT Ϫ/Ϫ mice.

DISCUSSION
We generated LRAT knock-out mice in order to analyze the effects of LRAT on the process of carcinogenesis, because expression of LRAT is low or absent in many types of human tumors (18,42,48,49). In comparison to the LRAT gene knock-out mouse strain reported recently by another laboratory (40), our LRAT knock-out mice have the neo R gene removed. This neo excision allowed us to avoid any possible deleterious effects resulting from the presence of the neo promoter (50 -53). We wanted to determine whether the LRAT Ϫ/Ϫ mice are more readily made vitamin A-deficient when placed on a vitamin A-deficient diet in order to use this model for future studies of the process of carcinogenesis. Retinyl esters, stored in the liver, are sufficient for sustaining life for up to 75 weeks in wild type mice (54). This makes it extremely difficult to generate vitamin A deficiency in WT mice. Our results show that disruption of the LRAT gene causes a loss of more than 99.5% of hepatic retinyl esters even when mice are on a control, vitamin A-sufficient diet (TABLE TWO). A great reduction in retinyl esters is also seen in the lung tissues of LRAT Ϫ/Ϫ mice maintained on a control, vitamin A-sufficient diet (TABLE TWO). This significant decrease in retinyl esters is the biological rationale for using these LRAT Ϫ/Ϫ mice as a model for vitamin A deficiency studies.
In other reports of mouse models of vitamin A deficiency using WT mice, the dietary retinol must be removed either during gestation or at parturition. Postnatal mice are generally not considered suitable for use in vitamin A deficiency studies because it is difficult to deplete vitamin A in wild type mice (54 -60). Two genetically modified mouse strains, the CRBP-I gene knock-out mice and the serum retinol-binding protein (RBP4) gene knock-out mice, have been used previously for studies of vitamin A deficiency (33,61,62). The CRBP-I protein has been reported to play a role in retinol metabolism. CRBP-I is involved in the oxidation and esterification of retinol, as well as in the hydrolysis of retinyl esters (63)(64)(65). Disruption of the CRBP-I gene in mice results in a 50% decrease in hepatic storage of retinyl esters (33). In CRBP-I Ϫ/Ϫ mice fed a vitamin A-deficient diet, it takes 23 weeks for serum retinol levels to decrease to a level of 0.05 g/ml (ϳ0.17 M) (33), which is much longer than the time required to observe such a decrease in LRAT Ϫ/Ϫ mice. The RBP4 Ϫ/Ϫ mouse is another knock-out line that has been used for vitamin A deficiency analyses (62). RBP4 is primarily synthesized in the liver and transports retinol from the liver to target tissues (62). The fetal offspring from RBP4 Ϫ/Ϫ dams display a phenotype of vitamin A deficiency upon maternal dietary vitamin A deprivation, and this is an excellent mouse model of embryonic vitamin A deficiency (66). Adult RBP4 Ϫ/Ϫ mice exhibit a low serum retinol level. One study showed that the serum retinol levels in RBP4 Ϫ/Ϫ mice fed a control, vitamin A-sufficient diet are about 6.8% of the levels seen in wild type mice (62). After 1 week on a vitamin A-deficient diet, RBP4 Ϫ/Ϫ mice showed further decreases in serum retinol levels (62). However, RBP4 Ϫ/Ϫ mice on a control, vitamin A-sufficient diet accumulate higher levels of retinol in the liver than wild type mice, and this high hepatic retinol level is not decreased when the RBP4 Ϫ/Ϫ mice are maintained on a vitamin A-deficient diet (62). Therefore, vitamin A deficiency may not occur in all tissues of the RBP4 Ϫ/Ϫ mice.
In comparison to the mouse models discussed above, LRAT Ϫ/Ϫ mice have some distinct advantages. Vitamin A deficiency can be rapidly achieved in adult mice by removal of dietary vitamin A. As shown by our data, retinol levels are not only greatly reduced in the serum but are also virtually absent in most tissues in LRAT Ϫ/Ϫ mice after 6 weeks on a FIGURE 7. The expression of the Cyp26A1 transcript in liver. Gene expression was examined by semi-quantitative RT-PCR. A, expression of Cyp26A1, LRAT, and Dhrs1 mRNAs in hepatic tissues. The experiment was performed three times on three identical (in terms of age and diet) groups of mice (groups A-C), and the results are shown. There were four mice in each group (two WT and two LRAT Ϫ/Ϫ ), and they were maintained either on a vitamin A-sufficient, control diet (Ctrl Diet) or on a vitamin A-deficient diet (Vit AϪ Diet) for 6 weeks. Note the changes in cycle numbers for PCR (Cyp26A1, 36 cycles; LRAT, 33 cycles; Dhrs1, 26 cycles; and GAPDH, 23 cycles). To show that the PCR was in linear range and not saturated, a serial dilution of one RT sample (group A, Wt, Ctrl Diet) was performed, and PCR with LRAT and GAPDH primers was repeated. Wt, wild type LRAT ϩ/ϩ mice; KO, mutant LRAT Ϫ/Ϫ mice. B, expression of the Cyp26A1 transcript in hepatic tissues after re-addition of a vitamin A-sufficient diet. The experiment was performed twice on two identical (in terms of age and diet) groups of mice (groups D and E). Four mice (two WT and two LRAT Ϫ/Ϫ , 4 weeks old) in each group were first fed the vitamin A-deficient diet for 6 weeks. Then two mice (one WT and one LRAT Ϫ/Ϫ ) from each group were changed to a vitamin A-supplemented diet (labeled as Vit Aϩϩ Diet), whereas the vitamin A-deficient diet was continued for the other mice (labeled as Vit AϪ Diet). After 48 h, the serum and hepatic samples were harvested from all mice. Analytical results for Cyp26A1 mRNA expression are shown. C, serum retinol after the restoration of the vitamin A-sufficient, control diet for 48 h. HPLC results from one experimental group (group D) are shown. The retinol peak is indicated by the arrow. ROL, all-trans-retinol.
vitamin A-deficient diet (Figs. 4 -6 and TABLE ONE and TWO). In addition, the retinol levels in serum rapidly increase in the LRAT Ϫ/Ϫ mice upon re-addition of vitamin A to the diet (Fig. 7C), indicating that serum retinol levels in LRAT Ϫ/Ϫ mice can be conveniently modulated by the quantitative manipulation of dietary retinol. These advantages make the LRAT Ϫ/Ϫ mice an ideal mouse model of vitamin A deficiency for a broad range of studies.
We used Cyp26A1 in this study as a molecular marker to monitor the retinol status in the liver. The expression of the Cyp26A1 gene is precisely regulated by the level of retinoids (27,44,45,67,68). Previous studies showed that the hepatic expression of the Cyp26A1 gene was induced either by an acute loading of RA or long term exposure to dietary vitamin A, whereas its expression was down-regulated upon administration of a vitamin A-deficient diet (45,69). In our study we detected Cyp26A1 expression in wild type mice but not in the LRAT Ϫ/Ϫ mice fed a vitamin A-deficient diet (Fig. 7A). These data provide additional evidence, at molecular level, that these LRAT Ϫ/Ϫ mice are vitamin A-deficient. The expression of the Cyp26A1 gene was rapidly induced upon re-administration of vitamin A in the diet, further indicating that Cyp26A1 is a sensitive marker to monitor tissue retinol levels (Fig. 7C). The hepatic expression of Cyp26A1 mRNA is higher in the LRAT Ϫ/Ϫ mice than in WT mice fed a control, vitamin A-sufficient diet, suggesting that the expression of Cyp26A1 is up-regulated in the absence of LRAT activity. Increased expression of Cyp26A1 mRNA in the lungs and testes of LRAT Ϫ/Ϫ mice fed a vitamin A-sufficient diet is also observed (data not shown). Both the LRAT and Cyp26 genes have been shown to be responsive to the change of local retinol levels (27). The induction of the enzyme CYP26A1 in the LRAT Ϫ/Ϫ mice suggests that LRAT is not only responsible for the retinol storage but also functions as a regulator of local retinoid concentrations and signaling. Similar levels of all-trans-retinoic acid are seen by HPLC in the liver and the extrahepatic tissues (lung and testis) of wild type and LRAT Ϫ/Ϫ mice (not shown). It is likely that in the absence of LRAT, CYP26A1 is activated to metabolize retinoic acid in a compensatory mechanism.