Genetic dissection of retinoid esterification and accumulation in the liver and adipose tissue.

Approximately 80–90% of all retinoids in the body are stored as retinyl esters (REs) in the liver. Adipose tissue also contributes significantly to RE storage. The present studies, employing genetic and nutritional interventions, explored factors that are responsible for regulating RE accumulation in the liver and adipose tissue and how these influence levels of retinoic acid (RA) and RA-responsive gene expression. Our data establish that acyl-CoA:retinol acyltransferase (ARAT) activity is not involved in RE synthesis in the liver, even when mice are nutritionally stressed by feeding a 25-fold excess retinol diet or upon ablation of cellular retinol-binding protein type I (CRBPI), which is proposed to limit retinol availability to ARATs. Unlike the liver, where lecithin:retinol acyltransferase (LRAT) is responsible for all RE synthesis, this is not true for adipose tissue where Lrat-deficient mice display significantly elevated RE concentrations. However, when CrbpI is also absent, RE levels resemble wild-type levels, suggesting a role for CrbpI in RE accumulation in adipose tissue. Although expression of several RA-responsive genes is elevated in Lrat-deficient liver, employing a sensitive liquid chromatography tandem mass spectrometry protocol and contrary to what has been assumed for many years, we did not detect elevated concentrations of all-trans-RA. The elevated RA-responsive gene expression was associated with elevated hepatic triglyceride levels and decreased expression of Pparδ and its downstream Pdk4 target, suggesting a role for RA in these processes in vivo.

literature ( 16,(34)(35)(36). Routinely, animals were allowed ad libitum access to water and a standard nutritionally complete rodent chow diet (W. F. Fisher and Sons, Inc.). All mice were maintained on a 12 h dark-light cycle, with the period of darkness between 7:00 PM and 7:00 AM, in a conventional barrier facility. For our studies, male and/or female mice at 3 months of age were employed and routinely sacrifi ced in the morning between 10:00 and 11:00 AM. The animal experiments described in this report were conducted in accordance with the National Research Council's Guide for the Care and Use of Laboratory Animals ( 37 ) and were approved by the Columbia University Institutional Animal Care and Use Committee.

Nutritional manipulations
For nutritional studies, mice were maintained on a chow diet (25 IU of retinol/g of diet) until they reached approximately 3 months of age. At 3 months of age, mice were randomized onto one of two different diets: 1 ) a retinoid-suffi cient (basal) diet providing 22 IU of retinol/g diet (Test Diet, 5755, St. Louis, MO); or 2 ) a retinoid-excess diet containing 25 times the amount of retinol in the basal diet or 550 IU of retinol/g diet. These purifi ed diets were formulated otherwise according to the AIN-93M formulation ( 38 ). Each diet was nutritionally complete for all other nutrients and only differed in their retinol content. Depending on the specifi c experiment, mice were maintained continuously on one of these diets for up to 12 weeks prior to sacrifi ce. For these nutritional studies, diet and water were provided to all animals on an ad libitum basis until the time of sacrifi ce.

HPLC analysis of retinol and REs
Tissue and serum retinol and RE levels were determined by HPLC protocols described previously ( 24 ). Briefl y, serum, liver, and epididymal adipose tissue were fl ash-frozen in liquid N 2 after dissection and stored immediately at Ϫ 80°C prior to analysis. Tissues were homogenized in 10 vol of PBS [10 mM sodium phosphate (pH 7.2), 150 mM sodium chloride] using a Polytron homogenizer (Brinkmann Instruments, Westbury, NY) set at halfmaximal speed for 10 s. An aliquot of serum, the total adipose tissue, or a 200 l aliquot of the liver homogenate was then treated with an equal volume of absolute ethanol containing a known amount of retinyl acetate as an internal standard. The retinoids present in the homogenates were extracted into hexane. After one backwash against doubly distilled water, the hexane extract was evaporated to dryness under a gentle stream of N 2 . Immediately upon reaching dryness, the retinoid-containing fi lm was redissolved in 40 l of benzene for injection onto the HPLC column. The extracted retinoids were separated on a 4.6 × 250 mm Ultrasphere C 18 column (Beckmann, Fullerton, CA) preceded by a C 18 guard column (Supelco, Bellefonte, PA) using 70% acetonitrile-15% methanol-15% methylene chloride as the running solvent fl owing at 1.8 ml/min. Retinol and REs (retinyl palmitate, oleate, linoleate, and stearate) were detected at 325 nm and identifi ed by comparing the retention times and spectral data of experimental compounds with those of authentic standards. Concentrations of retinol and REs in the tissues were quantitated by comparing integrated peak areas of each retinoid against those of known amounts of purifi ed standards. Loss during extraction was accounted for by adjusting for the recovery of internal standard added immediately after homogenization of the samples.

LC/MS/MS analysis of RA
Serum and tissue levels of all-trans -RA were determined by ultra high-performance liquid chromatography tandem mass spectrometry (LC/MS/MS) using a Waters Xevo TQ MS ACQUITY UPLC system (Waters, Milford, MA). For this analysis, we only employed and other tissues are able to catalyze RE formation; one involves an acyl-CoA-dependent mechanism and the other an acyl-CoA-independent process (11)(12)(13)(14)(15). The acyl-Co-Adependent mechanism is attributed to an acyl-CoA:retinol acyltransferase (ARAT) activity (9)(10)(11)(12). The acyl-CoA-independent synthesis of REs was shown early on to involve the transesterifi cation of a fatty acyl group from the sn-1 position of a membrane phosphatidylcholine to retinol, and the enzyme responsible for this transesterifi cation was termed lecithin:retinol acyltransferase (LRAT) (13)(14)(15). Studies of Lrat -defi cient ( Lrat Ϫ / Ϫ ) mice have established that LRAT is responsible for most, but not all, REs synthesized in the body (16)(17)(18). Very few REs can be detected in the liver, eyes, or lungs of Lrat Ϫ / Ϫ mice fed a control chow diet ( 17 ). The sole tissue where substantial REs accumulate in Lrat Ϫ / Ϫ mice is adipose tissue, where concentrations of REs are elevated by 2-to 3-fold over those measured in matched wild-type (WT) mice ( 17,18 ).
Several groups have reported in vitro studies demonstrating that recombinant diacylglycerol acyltransferase 1 (DGAT1), an enzyme that catalyzes the formation of triglyceride from diglycerides and fatty acyl-CoAs (19)(20)(21), also catalyzes the acyl-CoA-dependent esterifi cation of retinol ( 17,22,23 ). Thus, DGAT1 possesses ARAT activity. Subsequently, through investigations of Dgat1 -defi cient ( Dgat1 Ϫ / Ϫ ), Lrat Ϫ / Ϫ , and Lrat Ϫ / Ϫ /Dgat1 Ϫ / Ϫ mice, we established that DGAT1 can act as a physiologically important ARAT in mouse intestine ( 24 ). Shih et al. ( 25 ), through study of Dgat1 Ϫ / Ϫ mice, demonstrated that DGAT1 acts physiologically in catalyzing RE synthesis in mouse skin. Dgat1 is highly expressed in the liver, where it has been shown to have a role in hepatic triglyceride synthesis ( 19,20,26 ). Dgat1 is also highly expressed in adipose tissue ( 26 ). The literature suggests that ARAT activities may become active only in times of excessive retinol availability (27)(28)(29). The literature also proposes that cellular retinol-binding proteins (CRBPs) prevent retinol from being delivered to ARAT activities ( 13-15, 28, 29 ). The liver is very much enriched in CRBP type I (CRBPI), and adipose tissue expresses both CRBPI and CRBPIII ( 28,30,31 ). Moreover, the literature proposes that the ability to esterify retinol contributes to regulation of retinol conversion to RA ( 32,33 ). The present studies are aimed at gaining understanding of the roles of DGAT1 and its ARAT activity and CRBPI in hepatic and adipose RE synthesis, and of how RE formation may infl uence RA availability and RA-responsive gene expression.

Animals, animal husbandry, and diets
The mutant mouse lines we employed have all been described in the literature and include Lrat Ϫ / Ϫ ( 16,17 ), The Lrat Ϫ / Ϫ and CrbpI Ϫ / Ϫ mice originally described for a mixed C57Bl/6J/129sv genetic background were employed in our studies. Dgat1 Ϫ / Ϫ mice were obtained from Jackson Labs in the C57Bl/6J genetic background. Using conventional breeding protocols we also generated Lrat Genotypes of the mice were determined by protocols already described in the ARAT activities can contribute to RE synthesis when retinol is present in excess of normal amounts (27)(28)(29). We investigated these possibilities in matched male WT, Lrat Ϫ / Ϫ , Dgat1 Ϫ / Ϫ , and Lrat Ϫ / Ϫ mice fed a diet containing a 25-fold excess of retinol compared with standard dietary levels for 4 weeks. However, we were unable to detect substantial RE concentrations in the livers of Lrat Ϫ / Ϫ or ( Table 1 ). This is contrary to what has been reported in the literature by Yamaguchi et al., who proposed, based on cell culture studies, that DGAT1 is the major contributor to the ARAT activity contributing to RE formation in hepatic stellate cells ( 44 ), the cellular site for RE storage in the liver ( 7,8,10 ). These investigators also reported that ablation of Dgat1 expression in cultured cells using antisense oligonucleotides results in increased expression of Lrat ( 44 ). We were unable to confi rm this published fi nding in our studies of Dgat1 Ϫ / Ϫ mice. Lrat mRNA levels assessed by qPCR for matched WT and Dgat1 Ϫ / Ϫ livers were identical ( Fig. 1A ). Similarly, Dgat1 mRNA levels were not different for WT and Lrat Ϫ / Ϫ livers ( Fig. 1B ). We also attempted to confi rm the published studies of Yamaguchi et al. ( 44 ) in vivo, using adenovirus constructs to rescue RE synthesis in Lrat Ϫ / Ϫ or Lrat Ϫ / Ϫ mice. However, adenovirus rescue vectors injected into the circulation of these mice were cleared predominantly by hepatocytes with very little being taken up by hepatic stellate cells, the cellular site of retinoid storage in the liver. Consequently, it was not possible to use this standard approach for rescuing hepatic Lrat expression to further validate our fi ndings from nutritional and genetic studies. The literature indicates that DGAT1 contributes to triglyceride-rich lipoprotein (VLDL) secretion from hepatocytes ( 45,46 ). Because REs are present in VLDLs, we asked whether DGAT1 might act to facilitate RE incorporation into VLDLs. Figure 2 provides evidence that LRAT is responsible for the synthesis of most REs that are incorporated into VLDLs and secreted from the liver. When RE concentrations were normalized for VLDL triglyceride levels, these concentrations were not different for WT or Dgat1 Ϫ / Ϫ mice. Very little RE was detected in VLDLs obtained from Lrat Ϫ / Ϫ mice. Thus, LRAT-catalyzed RE formation appears to be primarily responsible for most of the LC/MS grade acetonitrile and LC/MS grade water purchased from Thermo Fisher (Pittsburgh, PA). All-trans -and 9-cis -RA were purchased from Sigma-Aldrich. Penta-deuterated all-trans -RA was employed as an internal standard and was purchased from Toronto Research Chemicals (North York, Ontario, Canada). Retinoid concentrations were verifi ed spectrophotometrically using published values ( 39 ). Tissue homogenates were extracted using the two-step acid-base extraction described by Kane et al. ( 40 ). All-trans -RA was detected and quantifi ed using the multiple reaction monitoring mode employing the following transitions: all-trans -RA, m/z 301.16 → 123.00; penta-deuterated all-trans -RA, m/z 306.15 → 127.03; and 9-cis -RA, m/z 301.16 → 123.00.

Triglyceride analysis
Triglyceride concentrations were determined enzymatically using a commercial colorimetric triglyceride kit (Wako), according to the manufacturer's instructions.

RNA isolation, reverse transcription, and qualitative real-time PCR
Total RNA from the liver was isolated using the RNA-Bee (Tel-Test) reagent according to the manufacturer's instructions. Potential contaminating genomic DNA present in the liver RNA isolates was removed by DNase treatment and chromatography on RNeasy columns (Qiagen). Reverse transcription was performed using random hexamer primers to generate cDNAs according to the supplier's instructions (Invitrogen). Quantitative polymerase chain reaction (qPCR) was performed for 40 cycles for 15 s at 95°C and 60 s at 60°C using an ABI 7000 sequence detection system (Applied Biosystems). TaqMan probes and primers for Ppar ␣ , , cellular-retinoic acid-binding protein type I ( CrabpI ), CrabpII , and 18S transcripts were designed by and obtained from ABI (Applied Biosystems). Quantifi cation of mRNA levels was performed by comparing the C t value of each sample to a standard curve generated by serial dilution of the appropriate tissue cDNA. For each of these standard curves, the correlation coeffi cients were 0.99 or greater. Values are normalized to 18S rRNA levels.

Hepatic VLDL production and triglyceride analysis
To assess the roles or effects of LRAT, DGAT1, and RBP4 in facilitating RE incorporation into nascent VLDLs, mice were fasted for 4 h and then injected with the total lipase inhibitor P-407, at 1 mg/g body weight by ip injection ( 41,42 ). Immediately prior to injection (0 h) and 6 h after injection (a time previously shown to assure a linear rate of triglyceride accumulation in P-407-treated mice ( 43 ), serum was obtained and processed for retinoid analysis by HPLC and triglyceride analysis as described above.

Statistical analyses
All data were analyzed for statistically signifi cant differences using standard procedures consisting of an unpaired t -test for comparisons of two groups or an ANOVA followed by post hoc analysis if more than two groups of mice were being compared.

RESULTS
The literature has long indicated that an acyl-CoAdependent enzymatic activity, an ARAT, present in liver homogenates, can catalyze synthesis of REs (9)(10)(11)(12). DGAT1, which is expressed in the liver, has been shown to be a physiologically signifi cant ARAT in the intestine and skin ( 24,25 ). It also has been proposed in the literature that Mice were maintained for 4 weeks on a diet providing 25 times more retinol than a standard vitamin A-suffi cient basal diet. Prior to being placed on the excess-retinol diet, all mice were maintained from weaning on a standard vitamin A-suffi cient chow diet. All values are given as mean ± SD. a P < 0.01 different from WT mice. b P < 0.05 different from CrbpI Ϫ / Ϫ mice. c P < 0.05 different from Lrat Ϫ / Ϫ mice.
for age-and diet-matched male and female WT mice, the hepatic retinol levels were much greater, by approximately 50-fold, than those of Lrat Ϫ / Ϫ mice; 81.5 ± 46.7 nmol/g for males and 49.3 ± 14.4 nmol/g for females. We examined both male and female mice because of known gender differences in hepatic total retinol accumulation ( 17 ). Liu and Gudas ( 18 ) reported that the expression level of Cyp26A1 , an enzyme that is induced by RA and that catalyzes retinoid degradation, is upregulated in Lrat Ϫ / Ϫ mice. We were able to confi rm this fi nding and able to extend it to CrbpI Ϫ / Ϫ and Lrat Ϫ / Ϫ / CrbpI Ϫ / Ϫ mice, which also showed elevated levels of Cyp26A1 mRNA ( Fig. 4A ). In addition to elevated expression of Cyp26A1 , we observed statistically signifi cant elevations in hepatic expression of another RA-inducible transcript, Rar ␤ 2 , for Lrat Ϫ / Ϫ and Lrat ( Fig. 4B ). However, we did not detect differences in hepatic mRNA expression levels of CrabpI or CrabpII . Thus, expression levels for a number of RA-inducible genes are likely elevated in the livers of these mutant mice. It is generally assumed that elevated expression levels of Cyp26A1 and Rar ␤ 2 refl ect elevated cellular all-trans -RA REs that are incorporated into VLDLs. Interestingly, mice totally lacking expression of Rbp4 , and hence unable to mobilize hepatic retinol ( 36 ), are able to mobilize REs from the liver bound to VLDL at levels that are identical to those of WT mice ( Fig. 2 ).
Cellular retinol-binding proteins, like CRBPI, which is highly expressed in the liver, have been proposed to sequester retinol and prevent it from being acted upon by ARAT activities (27)(28)(29). To address whether this might account for our inability to demonstrate the existence of a hepatic ARAT in vivo, we conventionally bred Lrat Ϫ / Ϫ with CrbpI Ϫ / Ϫ mice to generate mice defi cient in both genes,  ( Table 1 ), retinol is still present in these livers. Interestingly, as seen in Fig. 3 , hepatic retinol concentrations for male and female Lrat Ϫ / Ϫ / CrbpI Ϫ / Ϫ mice fed a control diet were markedly diminished, by 10-to 20-fold, compared with matched Lrat Ϫ / Ϫ mice. Moreover,  Ϫ / Ϫ mice that had been fasted 4 h prior to P-407 administration by ip injection. All values are given as means ± SD for six mice per group. Statistical signifi cance: a, P < 0.01 compared with WT, Dgat1 ( Fig. 5A ).
Because CrbpI is expressed in adipose tissue, in a separate study we asked whether the absence of CrbpI affects adipose retinol levels as it does in the liver. Indeed, adipose tissue total retinol levels, which are elevated by approximately 3-fold for Lrat Ϫ / Ϫ compared with WT mice, were diminished in adipose tissue from matched Lrat Ϫ / Ϫ / CrbpI Ϫ / Ϫ mice to levels identical to WT mice ( Fig. 5B ). We also undertook studies to identify whether there might be differences in expression of known RA-responsive genes in adipose tissue obtained from these mice. However, unlike the liver, we did not detect statistically signifi cant differences in mRNA expression levels for Rar ␤ 2 , Cyp26A1 , or Cyp26B1 for the different mouse lines (data not shown). We also did not observe differences in Rbp4 , CrabpI , or CrabpII mRNA levels between the different lines. While studying the Lrat Ϫ / Ϫ / CrbpI Ϫ / Ϫ mice, we observed visually that these mice seemed to accumulate more hepatic fat than WT mice. We assessed this possibility in age-and diet-matched male WT, Lrat Ϫ / Ϫ , CrbpI Ϫ / Ϫ , and mice showed a statistically signifi cant elevation in fasting triglyceride levels compared with WT mice ( Fig. 6A ) Fig. 4C-E ). Our LC/MS/MS methods allowed for a very clean separation of all-trans -RA in tissue extracts. We did not encounter any problems that might be associated with matrix effects for either the LC separations ( Fig. 4D ) or the fragmentation as assessed from the daughter ion spectrum of the endogeneous all-trans -RA ( Fig. 4E ). Surprisingly, and contrary to what has been inferred based on gene expression data, serum and hepatic steady-state concentrations of all-trans -RA were not elevated for Lrat Ϫ / Ϫ compared with WT mice ( Fig. 4C ). These levels were actually signifi cantly reduced in the serum and livers of the mutant mice. This was also the case for hepatic all-trans -RA levels for CrbpI Ϫ / Ϫ and Lrat Ϫ / Ϫ / CrbpI Ϫ / Ϫ mice as well (data not shown). We take this to indicate that elevated expression of CYP26A1 results in increased catabolism and lower hepatic all-trans -RA concentrations. We were also interested in measuring 9-cis -RA concentrations in addition to all-trans -RA by LC/MS/MS. However, 9-cis -RA was not present in the livers at a level that we felt we could accurately measure. This can be seen in the LC/MS/ MS profi le provided as Fig. 4D . The peak for all-trans -RA is very substantial for this liver extract, and for all other liver extracts we analyzed. There is a small peak with a retention time of approximately 8.15 min, which is the retention time at which authentic 9-cis -RA elutes. Given the size of this peak, it is possible that the small amount of 9-cis -RA present may have been formed as an artifact during extraction and processing, because it is well known that all-trans -RA can undergo some isomerization to its cis -isomers.
To understand whether DGAT1 is responsible for the REs present in Lrat -defi cient adipose tissue, we measured total retinol levels (retinol + REs) for epididymal fat pads obtained from mice lacking both Lrat Ϫ / Ϫ and Dgat1 Ϫ / Ϫ mice. These levels were not statistically Scd1 , and Acc ) and fatty acid oxidation ( Cpt1 ) but observed no signifi cant differences (data not shown). As shown in Fig. 6C , we observed a marked downregulation in expression of the key regulatory enzyme Pdk4 , which is a known target gene for Ppar ␦ transcriptional regulation ( 47 ).

ARAT activities are not involved in RE synthesis in the liver
The literature indicates that ARATs are involved in the synthesis of hepatic REs ( 9-12, 28, 29 ). We have reported that DGAT1 can act as a physiologically signifi cant ARAT in the mouse intestine ( 24 ) and Shih et al. ( 25 ) established that DGAT1 acts physiologically as an ARAT in mouse skin. It is well established that DGAT1 acts to facilitate triglyceride storage/metabolism and lipid droplet formation in the liver (19)(20)(21). Because DGAT1 is highly expressed in the liver, this raises a question as to whether DGAT1 might also act as an ARAT in the liver. Moreover, DGAT1 is expressed both in hepatocytes and in hepatic stellate cells ( 44 ), the cellular site in the liver where REs are stored and where LRAT is primarily expressed ( 48 ). Even though our earlier studies of Lrat Ϫ / Ϫ mice established that these mutant mice have very low levels of hepatic REs (<0.1% of matched WT levels) suggesting that LRAT is responsible for the preponderance of hepatic RE synthesis when mice are maintained on a standard chow diet ( 17 ), the literature suggests a role for an ARAT in hepatic RE formation. This extensive literature maintains that tissue ARAT activities may only become active when high levels of retinol are available and/or when the capacities of CRBPs like CRBPI and CRBPII to bind retinol and channel it to LRAT have been exceeded (27)(28)(29)49 ). Indeed, our earlier work, which established DGAT1 as a physiologically relevant ARAT in the intestine, also established that one of the actions of CRBPII in the intestine was to channel retinol to LRAT for esterifi cation ( 23 ). To directly address these possibilities, we employed a nutritional approach, feeding a 25fold excess retinol diet for 4 weeks, coupled with a genetic approach, in an attempt to demonstrate LRAT-independent RE formation. Our data do not support the idea that an acyl-CoA-dependent ARAT enzyme(s) contributes to hepatic RE formation in vivo. Our data are consistent with Ϫ / Ϫ mice were not signifi cantly different nor were the expression levels of Ppar ␥ in adipose tissue obtained from these different genotypes (data not shown). We also examined possible changes in expression for genes involved in hepatic lipogenesis ( Fas ,  Statistical signifi cance: a, P < 0.01 compared with WT mice or CrbpI Ϫ / Ϫ mice; b, P < 0.01 compared with Lrat Ϫ / Ϫ mice.
suggests coordinated gene regulation as proposed by these authors.

CRBPI acts to prevent catabolism and loss of hepatic retinol
It has been proposed that CRBPI prevents retinol from being converted to REs by ARAT activities or exposure to nonspecifi c enzymes that may catalyze retinol oxidation ( 27-29, 34, 49, 50 ). Our data do not support the notion that CRBPI acts to prevent hepatic or adipose ARAT activities, like DGAT1, from catalyzing RE synthesis. Rather, our data are convincing that CRBPI prevents elimination or loss of retinol from the liver, and from adipose tissue as well (see Fig. 3 ). The absence of CRBPI from Lrat Ϫ / Ϫ livers (in Lrat Ϫ / Ϫ / CrbpI Ϫ / Ϫ mice), which possess no REs and hence hepatic retinol levels and metabolism can be very cleanly assessed, results in an 8-to 20-fold reduction in the level of hepatic retinol. Molotkov et al. ( 50 ) have proposed that hepatic CRBPI limits nonspecifi c oxidation of retinol by alcohol dehydrogenase 1 and proposed that this increases the ability of hepatic "esterifying enzymes" to produce REs for storage. Because retinol cannot be esterifi ed in the livers of Lrat Ϫ / Ϫ / CrbpI Ϫ / Ϫ mice, our data establishes directly that hepatic CRBPI prevents loss of retinol from the liver. Interestingly, although the simple absence of CRBPI from adipose tissue does not affect the total retinol (retinol + REs) level found in adipose tissue ( Fig. 5B ), the absence of CRBPI from Lrat Ϫ / Ϫ mice results in a signifi cant reduction of adipose total retinol. Total retinol levels present in Lrat Ϫ / Ϫ adipose tissue are approximately 2-or 3-fold elevated over those of age-, gender-, and diet-matched WT mice ( 17 ) ( Fig. 5B ). The absence of CRBPI from Lrat -defi cient adipose tissue results in adipose tissue total retinol levels which are similar to those of matched WT mice. There are two possible bases for this observation. It is possible, that like in the liver, CRBPI prevents oxidation and loss of adipose retinol. However, because adipose total retinol levels are similar for WT and CrbpI Ϫ / Ϫ mice, we believe that this is unlikely. Alternatively, because the molecular identity of the enzyme(s) responsible for RE formation in Lrat Ϫ / Ϫ / Dgat1 Ϫ / Ϫ adipose tissue is not known, possibly there is a previously unsuspected CRBPI-dependent retinol esterifying activity present in adipose tissue. This possibility needs to be explored in future research.
Elevated hepatic mRNA levels for known RA-responsive genes should not be taken to indicate that hepatic steady-state RA concentrations are elevated Liu and Gudas ( 18 ) have demonstrated that Cyp26A1 mRNA expression is elevated in the livers of Lrat Ϫ / Ϫ mice. Earlier studies showed Cyp26A1 mRNA expression is induced either by acute loading with RA or long-term exposure to dietary retinoids, whereas expression was downregulated upon administration of a retinoid-defi cient diet ( 51,52 ). We have confi rmed the published observation of Liu and Gudas ( 18 ) that Cyp26A1 expression is elevated in the livers of chow-fed Lrat Ϫ / Ϫ mice and have established further that expression of the retinoid-responsive transcription factor RAR ␤ 2 is also elevated in the livers of chow-fed the conclusion that LRAT is solely responsible for hepatic RE synthesis. This includes both RE storage in hepatic stellate cells and RE incorporation into nascent VLDLs in hepatocytes. Although DGAT1 is a physiologically relevant ARAT in the intestine and skin ( 24,25 ), we were unable to obtain any evidence that it has this role in the liver. Moreover, contrary to what has been proposed by Yamaguchi et al. ( 44 ) from cell culture studies, we observed no relationship between Lrat and Dgat1 gene expression in the liver that were fasted in the morning for 4 h after diet was removed from their housing prior to sacrifi ce. Statistical signifi cance: a, P < 0.05 compared with WT mice. Livers (n = 6 per strain) were taken for RNA isolation and assessment of Ppar ␦ (B) and Pdk4 (C) mRNA levels by qPCR. All values are given as means ± SD. Statistical significance: a, P < 0.01 compared with WT mice. effects on hepatic endocannabinoid signaling ( 62 ), and human observational studies ( 63,64 ). However, the specifi c mechanisms underlying these observations are not well-established. Given the focus of our research, we undertook only a limited survey to identify possible general molecular pathways that might be responsible for the observation. To this end, we examined by qPCR expression levels of a number of key regulators of hepatic lipid metabolism. We did not detect signifi cant differences between matched mutant and WT mice in hepatic expression of regulatory genes commonly associated with hepatic steatosis, specifi cally Ppar ␣ and Ppar ␥ . Strikingly though, Ppar ␦ mRNA levels were downregulated by more than 75% and levels of the Ppar ␦ target gene Pdk4 ( 47 ) were similarly downregulated in the livers of CrbpI Ϫ / Ϫ , Lrat Ϫ / Ϫ , and Although it is generally believed that PPAR ␦ exerts its effects on lipid metabolism primarily through actions in skeletal muscle ( 65 ), there is evidence that PPAR ␦ also controls hepatic energy substrate homeostasis through coordinated regulation of glucose and fatty acid metabolism ( 66 ). Interestingly, all-trans -RA has been proposed to be a PPAR ␦ agonist ( 4,5 ). We believe that the observations of elevated hepatic triglyceride accumulation in CrbpI Ϫ / Ϫ and Lrat Ϫ / Ϫ / CrbpI Ϫ / Ϫ mice and elevated RA-responsive gene expression in these livers are directly related. However, further investigations will be needed before this possibility can be conclusively established.

Lrat
Ϫ / Ϫ mice. Both the Cyp26A and Rar ␤ 2 genes are known to contain RA response elements, which render the genes responsive to all-trans -RA availability (1)(2)(3)53 ). This is usually taken to suggest that steady-state RA levels are elevated in tissues/cells expressing elevated levels of RA-responsive genes. However, as seen in Fig. 4C, D , serum and liver levels of all-trans -RA, assessed using very specifi c and sensitive LC/ MS/MS protocols, were actually signifi cantly lower in Lrat Ϫ / Ϫ compared with matched WT mice. This likely arises from the increased expression and catabolic actions of Cyp26A1 , and possibly other RA metabolizing cytochromes. These data could be taken to suggest that the genes for Cyp26A1 and Rar ␤ 2 , and possibly other RA-responsive genes, may respond to increased fl uxes of all-trans -RA (i.e., increased synthesis as well as increased degradation) rather than simply increased steady-state levels of this retinoid. However, there are other molecular processes that may explain increases in Cyp26A1 and Rar ␤ 2 gene expression. These include possible differences in the half-lives of these mRNA species, effects of other transcription factors that modulate the expression of these genes, or possible differences in how all-trans -RA may be partitioned between the cytosol and nucleus within the cell. This later possibility though seems unlikely because we saw no differences in hepatic expression of CrabpI or CrabpII mRNA for the different mouse lines. The LC/MS/MS protocols we used to measure all-trans -RA concentrations in liver extracts also allowed for separation and detection of purifi ed 9-cis -RA. However, we did not detect much 9-cis -RA in any of our liver extracts, although we could readily detect 9-cis -RA when it was exogenously added to liver homogenates. Thus, we conclude that 9-cis RA is simply not present at substantial levels endogenously in the liver. Kane et al. ( 40 ) have reported a similar inability to detect 9-cis -RA in mouse tissues, also using state-of-the-art LC/MS/MS instrumentation. When 9-cis -RA originally was reported to be the natural ligand for the retinoid X receptors , its level in mouse liver was reported to be 4 ng/g wet weight ( 54 ). This reported level is clearly incorrect and likely too large by at least two orders of magnitude. Based on our data, we believe that relative to all-trans -RA there is very little 9-cis -RA in the mouse liver.

Increased hepatic RA-responsive gene expression is associated with elevated triglyceride levels
While undertaking our investigations aimed at understanding hepatic retinoid storage, we observed that the fasting livers of CrbpI Ϫ / Ϫ and Lrat Ϫ / Ϫ