Characterization of a new member of the fatty acid-binding protein family that binds all-trans-retinol.

Cellular retinol-binding protein, type I (CRBP-I) and type II (CRBP-II) are the only members of the fatty acid-binding protein (FABP) family that process intracellular retinol. Heart and skeletal muscle take up postprandial retinol but express little or no CRBP-I or CRBP-II. We have identified an intracellular retinol-binding protein in these tissues. The 134-amino acid protein is encoded by a cDNA that is expressed primarily in heart, muscle and adipose tissue. It shares 57 and 56% sequence identity with CRBP-I and CRBP-II, respectively, but less than 40% with other members of the FABP family. In situ hybridization demonstrates that the protein is expressed at least as early as day 10 in developing heart and muscle tissue of the embryonic mouse. Fluorescence titrations of purified recombinant protein with retinol isomers indicates binding to all-trans-, 13-cis-, and 9-cis-retinol, with respective K(d) values of 109, 83, and 130 nm. Retinoic acids (all-trans-, 13-cis-, and 9-cis-), retinals (all-trans-, 13-cis-, and 9-cis-), fatty acids (laurate, myristate, palmitate, oleate, linoleate, arachidonate, and docosahexanoate), or fatty alcohols (palmityl, petrosenlinyl, and ricinolenyl) fail to bind. The distinct tissue expression pattern and binding specificity suggest that we have identified a novel FABP family member, cellular retinol-binding protein, type III.

Members of the FABP family solubilize and transport their respective ligands in the cytosol and, in some instances, may regulate the metabolism of their ligands (2, 6 -10). For example, CRBP-I helps to regulate retinol hemeostasis (2,5,(13)(14)(15). Recent studies of knockout mice indicate a crucial role for CRBP-I in maintaining normal hepatic retinol storage. These studies demonstrate that CRBP-I is needed to facilitate conversion of retinol to retinyl ester, thus slowing turnover of retinol from the liver (15). Likewise, CRBP-II regulates intestinal retinoid metabolism, facilitating reduction of retinal to retinol and the subsequent esterification of retinol to retinyl ester. The retinyl ester is packaged along with other dietary lipids into nascent chylomicrons (2,5,13,14). Thus, CBRP-I and CRBP-II promote retinol uptake from the diet and its storage in the liver.
We recently demonstrated that the amount of postprandial retinol taken up by heart, skeletal muscle, and adipose tissue is markedly influenced by the level of lipoprotein lipase expression in the tissue (16). Overexpression of lipoprotein lipase in heart, muscle, or adipose tissue increases retinol uptake from chylomicrons or chylomicron remnants (16). Because neither heart nor skeletal muscle express CRBP-II (2) and very little if any CRBP-I (1-5, 17, 18), we wondered whether these tissues might express other intracellular binding proteins that facilitates retinol uptake, transport, and metabolism. We now report the identification and characterization of a new member of the FABP family that binds retinol and is highly expressed in heart, skeletal muscle, and adipose tissue.

EXPERIMENTAL PROCEDURES
Identification and Sequence Analysis of a cDNA Encoding a Novel Intracellular Retinol-binding Protein-We screened sequences depos-ited with GenBank TM (National Center for Biotechnology Information, Bethesda, MD) and identified an expressed sequence tag clone with a predicted amino acid sequence that has a high degree of homology to both mouse CRBP-I and CRBP-II. We obtained this clone (AA466092) through Research Genetics, Inc (Huntsville, AL). After subcloning the cDNA into pcDNA 3 (Invitrogen, San Diego, CA), it was sequenced in both directions through the Columbia University Comprehensive Cancer Center Core DNA Sequencing Facility. Additional sequencing was performed after the cDNA was subcloned into the baculovirus expression vector pBlueBac4.5 (Invitrogen) and into the prokaryotic expression vector pET11A (Novagen, Madison WI) to verify that the cDNA was incorporated into these vectors in the proper orientation and with the reading frame intact.
Expression of Recombinant Protein in Sf9 Cells-The protein encoded by the cDNA described above was expressed in Spodoptera frugiperda (Sf9) cells using a baculovirus expression system (Invitrogen), according to the manufacturer's protocol. The mouse cDNA was subcloned into the pBlueBac4.5 vector and transfected together with disabled virus into actively dividing Sf9 cells. Recombinant virus containing the cDNA insert was plaque purified. The presence of the cDNA in the recombinant virus was verified by PCR. The initial low titer viral stock was used to generate a high titer viral stock. For expression of the protein, Sf9 cells were plated at a density of 10 7 cells/150-cm 2 flask and infected with recombinant virus at a multiplicity of infection of 10. Cells were harvested 72 h after infection, and the washed cell pellet was homogenized in 20 mM potassium phosphate, pH 7.4, containing 1 mM EDTA, 10 mM ␤-mercaptoethanol, 15% (v/v) glycerol, 0.05% (w/v) sodium azide, 7.5 M aprotinin, and 0.5 mM phenylmethanesulfonyl fluoride using a Dounce homogenizer. The homogenate was centrifuged at 8000 ϫ g for 20 min to remove cell debris and the nuclear fraction. Cytosolic and the crude microsomal fractions were obtained upon centrifugation at 100,000 ϫ g for 1 h at 4°C using a Beckman TC-100 ultracentrifuge (Beckman Coulter, Fullerton, CA). The presence of recombinant 15-kDa protein in the crude microsomal and cytosolic fractions was assessed on a 15% SDS-PAGE gel to verify protein expression.
Expression of Recombinant Protein in Escherichia coli-As an alternative to expressing the cDNA in Sf9 cells, we employed a prokaryotic expression vector to generate recombinant protein in E. coli. For this purpose, the cDNA was cloned into a pET11A expression vector (Novagen, Madison, WI). To clone the cDNA into the pET11A vector, a NdeI restriction site upstream of the translation start site and a BamHI restriction site downstream of the translation termination site were created through PCR amplification of the cDNA employing primers containing the specified restriction sites. The sequence of the 5Ј primer we employed was 5Ј-GGGAATTCCATATGCCAGCAGACCTCAGCGG-TAC-3Ј and that of the 3Ј primer was 5Ј-CGCGGATCCTCAGGCTCT-CTGGAAGGTTTG-3Ј. Each PCR amplification reaction contained forward and reverse primers (0.2 M for each), 0.2 mM of each dNTP, 1.5 mM MgCl 2 , 2 units of Taq DNA polymerase (Life Technologies, Inc.), and 5 l of 10ϫ PCR buffer in 50 l. The following PCR conditions were employed: initial denaturation at 95°C for 10 min, followed by 35 cycles of denaturation at 95°C for 45 s, annealing at 55°C for 45 s, and extension at 72°C for 1 min in a DNA Thermal Cycler (PerkinElmer Life Sciences). The PCR product was subcloned into pCR II according to the instructions of the supplier (Invitrogen). Subsequently, the cDNA insert was excised from pCR II and directionally subcloned into the NdeI and BamHI restrictions sites of the pET11A vector. Recombinant protein was expressed in BL23(DE3) E. coli (Novagen). E. coli containing the pET11A expression vector were grown at 30°C to an A 660 of 0.6 and expression was induced with 100 mM isopropyl-1-thio-␤-D-galactopyranoside for 3 h at 30°C. After harvesting of the induced bacterial cells by centrifugation at 10,000 ϫ g for 30 min at 4°C, the cell pellet was resuspended in 20 mM potassium phosphate, pH 7.4, containing 1 mM EDTA, 10 mM ␤-mercaptoethanol, 15% (v/v) glycerol, 0.05% (w/v) sodium azide, 7.5 M aprotinin, and 0.5 mM phenylmethanesulfonyl fluoride and sonicated with three bursts of 30 s duration at halfmaximal intensity. During sonication, the resuspended bacterial pellet was kept on ice. Following sonication, the soluble fraction was obtained upon ultracentrifugation at 100,000 ϫ g at 4°C using a Beckman TC-100 ultracentrifuge (Beckman Coulter, Fullerton, CA).
Protein Purification-Recombinant CRBP-III protein expressed either in Sf9 cells or in E. coli was purified from the cytosol fraction obtained from Sf9 cell or E. coli homogenates (as described above) using two size exclusion columns. The cytosol fraction was initially applied to a Bio-Gel P-100 polyacrylamide gel exclusion column (100 ϫ 2.5 cm) (Bio-Rad), equilibrated in homogenization buffer. Following absorbance at 280 nm monitored elution of proteins from this column. The recombinant protein eluted from the Bio-Gel P-100 column in approximately 60 -69 ml of buffer, an elution volume that should contain proteins with masses ranging from 10 to 20 kDa. The fractions thought to contain the 15-kDa protein were confirmed by analysis on 15% SDS-PAGE gels. These fractions were combined and concentrated to 2 ml using Centriplus concentrators with a molecular weight cut off of 3000 (Millipore, Bedford, MA). The concentrate of fractions containing the 15-kDa protein was then applied to a Bio-Gel P-30 sizing column (120 cm ϫ 2.5 cm) (Bio-Rad) equilibrated with homogenization buffer. Again, protein was followed by absorbance monitored at 280 nm. The presence of a 15-kDa protein in fractions eluting from the Bio-Gel P-30 column was identified by analysis on 15% SDS-PAGE gels. These fractions were combined and concentrated prior to use for the binding assays described below.
Determination of the N-terminal Amino Acid Sequence of the Purified Recombinant Protein-To confirm that we had indeed purified a protein predicted by the CRBP-III cDNA, the N-terminal sequence of the purified recombinant protein was analyzed. For this purpose the protein was electrophoresed on a 15% SDS-PAGE gel and transferred electrophoretically to a polyvinylidene difluoride transfer membrane (Millipore). The membrane was stained with Coomassie Blue, and the protein band running at approximately 15 kDa was excised from the membrane and subjected to automated N-terminal sequencing (21). N-terminal sequence analysis was performed at the Columbia University Howard Hughes Protein Chemistry Core Facility.
Total RNA Extraction and Northern Blot Analysis-Female and male adult C57Bl/6J mice were sacrificed by CO 2 asphyxia, tissues removed, immediately placed in liquid nitrogen, and stored at Ϫ70°C until RNA extraction. Total RNA from liver, kidney, heart, lung, spleen, skeletal muscle (gastronemius), small intestine, ovaries, fallopian tubes, testes, seminal vesicles, and the ovarian and epididymal fat depots was extracted using RNAzol (Tel-Test Inc., Friendswood, TX). Total RNA (25 g/lane) was resolved by gel electrophoresis in 1% agarose containing 0.98 M formaldehyde (22). For all RNA samples, the ratio of intensities of the 28 and 18 S ribosomal RNA bands after staining with ethidium bromide was approximately 2. Immediately following electrophoresis, the RNA was transferred by capillary action to a positively charged Nylon membrane (Amersham Pharmacia Biotech) and subsequently hybridized at 65°C with 32 P-labeled probes for the mouse cDNA and with control probes for rat CRBP-I (22) or glyceraldehyde-3-phosphate dehydrogenase (CLONTECH, Palo Alto, CA). The probes were labeled with 32 P using a random priming kit (Roche Molecular Biochemicals). Hybridized membranes were washed at a final stringency of 0.1ϫ SSC, 1.0% SDS at 65°C and exposed to Kodak AR-2 film at Ϫ80°C. An RNA ladder from Ambion (Austin, TX) was used to estimate transcript sizes.
In Situ Hybridization-In situ hybridization studies of expression of the cDNA in the embryonic mouse was performed using digoxigeninlabeled riboprobes essentially as described in Mendelsohn et al. (23). Similarly, the procedures we employed to stage the mouse embryos, for their fixation and the sectioning of embryos are also described in Mendelsohn et al. (23). For antisense probes, the cDNA in pcDNA3 was linearized with EcoRI, and antisense transcripts were generated with T7 polymerase. For preparation of sense probes, the same cDNA was linearized with HindIII, and sense transcripts were generated with T3 polymerase.
Isolation of Adipocytes and Stromal Vascular Cells-Adipocyte and stromal vascular cell fractions were prepared from mouse epididymal adipose tissue essentially as described by Tsutsumi et al. (24) for the rat tissue. For this purpose, epididymal fat pads from five adult C57Bl/6J mice were combined (approximately 1 g of tissue), rinsed in 0.9% NaCl at room temperature, and minced. The minced tissue was digested with 0.1% collagenase D (Roche Molecular Biochemicals) in 100 mM HEPES, pH 7.4, containing 120 mM NaCl, 50 mM KCl, 5 mM glucose, 1 mM CaCl 2 , and 1.5% bovine serum albumin for 60 min in a shaking water bath at 37°C. Undigested tissue was removed by filtering the digest through a 250-m nylon mesh. The resulting filtrate was centrifuged at 1000 ϫ g for 10 min at 25°C. The adipocytes (top layer) and the stromal vascular cells (pellet) were collected by aspiration using plastic pipettes. Total RNA was extracted from the two cell fractions as described above and used for analysis by RT-PCR.
RT-PCR Procedures-RT-PCR was used to identify the cellular sites of expression of the novel retinol-binding protein and, for comparison, CRBP-I in adipose tissue and several other tissues and primary cell isolates. For this purpose, cDNA was generated from approximately 1 g total RNA prepared from primary isolates of mouse stromal vascular cells and adipocytes using a Superscript Preamplification Kit (Life Technologies, Inc.). The target cDNAs were directly amplified by PCR using PCR beads (Amersham Pharmacia Biotech). The primers used for amplification of the novel retinol-binding protein were identical to those described above for the cloning of cDNA into the pET11A prokaryotic expression vector. For CRBP-I, the primers were designed based on the published mouse cDNA sequence (GenBank TM accession number X60367). As the 5Ј primer we employed 5Ј-GGGAATTCCATATGCCT-GTGGACTTCAACGGGTA-3Ј, and as the 3Ј primer we used 5Ј-CGCG-GATCCCAGTGTACTTTCTTGAACACT-3Ј. The same PCR conditions employed above also were used to amplify CRBP-I cDNA. The amplified cDNA was electrophoresed on a 1% agarose gel.
Fluorescence Binding Assays-All fluorescence measurements were carried out on an Aminco luminescence spectrometer (Spectronic Unicam, Rochester, NY) equipped with a magnetic stirring unit using a 2-ml fluorescence cuvette. To assess potential binding of all-trans-retinol, 13-cis-retinol, and 9-cis-retinol to the purified recombinant protein, we measured the enhancement of retinol fluorescence observed upon binding. For this purpose, the protein solution was excited at 330 nm (bandpass of 4 nm), and retinol emission was measured at 480 nm (bandpass of 4 nm). To assess the potential binding of 9-cis-, 13-cis-, and all-trans-retinoic acid, the excitation wavelength was set at 360 nm (4 nm bandpass), and emission was monitored at 470 nm (4 nm bandpass) (25). To assess potential binding of fatty acids, fatty alcohols, sphingosine, and ceramide, potential changes in tryptophan fluorescence were monitored upon excitation at 290 nm and emission at 340 nm (bandpasses of 4 nm) (25). Concentrated retinoid stocks were prepared by dissolving each retinoid in absolute ethanol. Stock concentrations were calculated based on the absorbance of each retinoid at its respective absorbance maximium (26). Diluted ethanolic solutions prepared from these stocks were used for binding assays. Fatty acid, fatty alcohol, sphingosine, and ceramide stocks diluted in ethanol were used to assess potential binding through changes in tryptophan fluorescence.
For these titrations, purified recombinant protein at a final concentration of 2.5 M was dissolved in 150 mM sodium phosphate, pH 7.4, containing 5 mM KCl and 10 mM HEPES. Retinoids or other lipids were added in 2 l (100 pmol/l) aliquots to an initial assay volume of 2 ml. To estimate changes in fluorescence arising from unbound retinoids, identical titration curves were performed for buffer alone. Calculations were done essentially as described by Cogan et al. (25).
HPLC Analyses of Retinol Isomers Following Fluorescence Binding Studies-Because all-trans-, 13-cis-, and 9-cis-retinol are light sensitive and easily isomerize and degrade upon exposure to intense light like that employed in the fluorescent titrations, we determined whether the all-trans-, 13-cis-, or 9-cis-retinol remained intact during/following the titrations. Thus, following titration with these retinol isomers, the solutions were extracted and subjected to analysis by normal phase HPLC. For this purpose, the solution was transferred to a 15-ml conical tube; 2 ml of absolute ethanol was added, and the mixture was vortexed well. Retinoids were then extracted into 3 ml of hexane. After one backwash against 0.5 ml of distilled water, the hexane extract was evaporated to dryness under a gentle stream of N 2 . The dried extract was immediately reconstituted in 120 l of hexane and injected to a normal phase HPLC system (see below). To establish retention times and to assess purity of these stocks, the diluted retinoid stocks used for titration were extracted simultaneously with the protein solutions generated in the fluorescence titrations.
Retinol isomers were separated on 4.6 ϫ 150 mm Supelcosil silica column (Supelco, Belefonte, PA), preceded by a silica guard column (Supleco) using hexane/ethyl acetate/n-butanol (96.9:3:0.1 v/v/v) as the mobile phase at a flow rate of 0.8 ml/min. Elution of retinol isomers were monitored at 325 nm. The identities of each isomer was based on the absorbance spectrum obtained from an inline Waters 996 Photodiodearray detector (Waters, Milford, MA). Concentrations of all-trans-, 13-cis-, and 9-cis-retinol in the extracts were calculated based on standard curves relating integrated peak area with known amounts of authentic all-trans-retinol, 13-cis-retinol, or 9-cis-retinol.
Western Blot Analysis-Recombinant protein expressed in E. coli and recombinant rat CRBP-I (27) were analyzed on a 15% SDS-PAGE gradient gel (Bio-Rad), and the protein was transferred to a polyvinylidene difluoride transfer membrane (Millipore) by electroblotting. The membrane was blocked in 5% nonfat dry milk in Tris-buffered saline, pH 7.4, containing 0.1% Tween-20 (TTBS) overnight. The blocked membrane was washed the next morning three times with TTBS and incubated with rabbit anti-CRBP-I antibody (28) for 1 h. The blot was visualized using ECL Western blot reagents (Amersham Pharmacia Biotech) according to the manufacturer's instructions.
Determination of Tissue Retinol Concentrations-Heart, epididymal fat, and skeletal muscle (gastronemius) were dissected from adult male C57Bl/6J mice under yellow light and immediately frozen in liquid N 2 and stored at Ϫ70°C until analysis. On the day of analysis, each tissue sample was weighed, thawed on ice, and homogenized with a Polytron homogenizer (Brinkmann Instruments, Westbury, NY) in 2 ml of 150 mM sodium phosphate, pH 7.4, containing 150 mM NaCl. The synthetic retinoid, all-trans-9-(4-methoxy-2,3,6-trimethylphenyl)-3,7-di-methyl-2,4,6,8-nonatetraen-1-ol (a gift from Dr. Christian Eckhoff, Hoffman La Roche Inc., Nutley, NJ) was added as internal standard to monitor recovery of the retinoids during extraction and HPLC analysis. The tissue homogenates were denatured with equal volume of 100% ethanol, extracted into 3 ml of hexane and backwashed once with 0.5 ml of distilled water. After evaporation of the hexane under a gentle stream of N 2 to dryness, the samples were reconstituted in hexane and analyzed by normal phase HPLC essentially as described above (19).

Identification of a cDNA Encoding a Novel Protein with
Sequence Identity to CRBP-I and CRBP-II-We sought to isolate intracellular proteins that bind retinol with high affinity from tissues that contain little or no CRBP-I or CRBP-II. Searching expressed sequence tags deposited in GenBank TM , we located a full-length expressed sequence tag, originally cloned from mouse mammary tissue. The full sequence for this clone is shown in Fig. 1A. 2 The cDNA encodes a protein consisting of 134 amino acids with a theoretical isoelectric point of 6.13 and a calculated molar extinction coefficient of 25,920 (29). The cDNA carries a putative polyadenylation site 166 base pairs downstream of the stop codon.
Based on the deduced amino acid sequence, the cDNA encodes a previously unidentified member of the intracellular FABP family (Fig. 1B). The alignment of the primary amino acid sequence for the encoded protein with several other members of the FABP family, CRBP-I, CRBP-II, ALBP, and hearttype fatty acid-binding protein is shown in Fig. 1B. Also shown are elements of the secondary structure characteristic of FABP family members. These elements, 10 stretches of ␤-sheet structure and 2 ␣-helix domains, are also present in the encoded protein sequence (2, 6 -10, 30). The amino acid sequence of this protein shares 57 and 56% identity with murine CRBP-I and CRBP-II, respectively, 35% identity with heart FABP, and 38% with ALBP. The protein has a 2-amino acid insertion at amino acid positions 75 and 76 in a turn site between the ␤E and ␤F strands (Fig. 1B). Of all FABP proteins only the two cellular retinol-binding proteins, CRBP-I and CRBP-II, show this two amino acid insertion. Thus, the protein encoded by this cDNA appears to be most closely related to CRBP-I and CRBP-II. On the basis of this similarity and additional properties described below, we term the protein cellular retinol-binding protein, type III (CRBP-III).
The three-dimensional structures of several members of the FABP protein family have been determined (2, 6 -10, 30, 31). Several amino acid residues participate in ligand binding. As shown in Table I, amino acid residues 4, 40, 108, 126, and 128 (based on the numbering of ALBP) interact with ligand and help confer specificity (6 -10, 30 -36). At three of these five positions, CRBP-I, CRBP-II, and CRBP-III contain the same amino acids, lysine 40, glutamine 128, and phenylalanine 130 (Table I). The three proteins differ at residue 4. At position 108, CRBP-III carries histidine instead of the glutamine found in CRBP-I and CRBP-II. This difference between CRBP-I, CRBP-II, and CRBP-III might affect ligand specificity or binding affinity.
Expression of CRBP-III cDNA in Adult Mouse Tissues-We examined the pattern of tissue expression for CRBP-III mRNA in adult mice by Northern blot analysis. Fig. 2A provides a representative Northern blot that gives the tissue pattern of CRBP-III expression. For comparison, tissue CRBP-I mRNA levels are also shown in Fig. 2A. A single CRBP-III transcript with an approximate size of 0.7 kilobases was detected in several tissues. The size of the CRBP-III transcript is similar to that of rat CRBP-I mRNA, although the tissue expression pattern is quite distinct (Fig. 2A). As seen in Fig. 2A, CRBP-III mRNA was observed in heart, epididymal fat, and skeletal muscle but not in liver, kidney, brain, lung, or eyes. In other Northern blots, CRBP-III expression was not detected in spleen, testes, ovaries, fallopian tubes, or seminal vesicles (data not shown). In contrast, CRBP-I ( Fig. 2A) is highly expressed in kidney, liver, lung, testis, epididymal fat, and eyes and, to a lesser extent, in heart. Interestingly, CRBP-III is highly expressed in two adult tissues, heart and skeletal muscle, that express little or no CRBP-I (2, 17). Both CRBP-III and CRBP-I are expressed in adipose tissue (2,24). Low levels of CRBP-III expression were also detected in small intestine, an organ that also expresses CRBP-II (data not shown) (2,24).
The cellular sites of CRBP-III and CRBP-I expression in adipose tissue were compared. CRBP-III was expressed in both primary adipocytes and stromal vascular cell preparations (Fig.  2B). This pattern of expression is different from that of CRBP-I, which is almost entirely restricted to stromal vascular cells (Ref. 24 and Fig. 2B). CRBP-III expression was unaffected by inactivation of ob, which increases adipose tissue formation. Northern blot analysis showed that the concentration of CRBP-III mRNA was equivalent in ob/ob and wild type mice (data not shown).
Because Northern analysis failed to reveal CRBP-III expression in liver, kidney and testes, we checked for CRBP-III expression in these three tissues by the more sensitive RT-PCR analysis. Nevertheless, even by RT-PCR analysis, we were unable to detect CRBP-III expression in these tissues (Fig. 2B).
Expression of CRBP-III cDNA in the Embryonic Mouse-The prominent expression of CRBP-III cDNA in heart and skeletal muscle of adult mice led us to examine its tissue expression pattern during organogenesis. Fig. 3 shows the results of in situ hybridization analysis of muscle tissue in E10 mouse embryos with digoxigenin-labeled antisense riboprobes. At E10 and at later stages, expression of the CRBP-III cDNA was robust in the walls of the dorsal aorta, the outflow tract, heart, and somites (Fig. 3, A and C). In heart, expression was highest in the atrial myocardium and in the ventricular trabeculae (Fig.  3B). At embryonic day 14, expression was localized in the developing muscle in heart, dorsal aorta, trachea, and bronchi of the lungs, as well as in skeletal muscle (data not shown). Overall, these data indicate that CRBP-III is expressed most strongly in developing heart and in muscle tissue destined to become part of the cardiovascular and arterial system.
Expression and Purification of Recombinant CRBP-III-To characterize the biochemical properties of CRBP-III, we trans-  1. A, the nucleotide sequence for the cDNA clone encoding a new member of the FABP protein family. The start and the termination codons are underlined. B, the amino acid sequence deduced from this cDNA. For comparison, the primary sequence encoded by the cDNA is aligned with those of four other members of the fatty acid binding protein family (2,6,8): CRBP-I, CRBP-II, ALBP, and heart FABP. The boxed amino acids designated ␤A-␤J represent 10 stretches of ␤-sheet structure, and those designated ␣I and ␣II designate stretches of ␣-helix.
fected CRBP-III cDNA into Sf9 cells. CRBP-III was recovered in the cytosolic fraction. The protein migrated on a 15% SDS-PAGE gel with an apparent molecular mass of ϳ15 kDa, in good agreement with the deduced molecular mass of 15.4 kDa. Purification from the Sf9 cytosol was achieved by fractionation through two successive size exclusion columns. The first fractionation on Bio-Gel P-100 yielded a preparation that was approximately 70% homogeneous based on SDS-PAGE gels (data not shown). After further fractionation on a Biogel P-30 column, a protein that migrated as a single band on a 4 -20% SDS-PAGE gradient gel at the same position as recombinant rat CRBP-I was detected (Fig. 4A). The sequence of the 30 N-terminal amino acids of the purified recombinant protein was identical to that predicted from the cDNA sequence: PADLSGTWNLLSSDNFEGYMLALGIDFATR. Despite the high degree of sequence identity between CRBP-III and CRBP-I, there was no cross-reactivity with rabbit antibodies directed against rat testis CRBP-I (Fig. 4B and Ref. 28).
Because 9-cis-retinol can readily isomerize to the all-transisomer, we were concerned that 9-cis-retinol isomerized to alltrans-retinol prior to binding to CRBP-III. To eliminate this possibility, we measured the rate of isomerization of all-transretinol and 9-cis-retinol in assay buffer. We found that both all-trans-retinol and 9-cis-retinol retained their isomeric configurations during the titration reaction. At the end of the titration, 99 and 95% of the added all-trans-retinol and 9-cisretinol were recovered as all-trans-and 9-cis-retinol, respectively. About 5% of the 9-cis-retinol was isomerized to all-transretinol. No isomerization of 13-cis-retinol to either the alltrans-retinol or 9-cis-retinol isomer was detected.

Concentrations of All-trans-retinol in Murine Skeletal
Muscle, Heart, and Epidydimal Fat-To determine whether the retinol binding constants for CRBP-III were compatible with tissue retinol levels, we measured the concentrations of alltrans-retinol in these tissues by normal phase HPLC. We are not aware of previous reports in the literature that provide heart and muscle levels of retinol or retinyl ester, as measured by modern HPLC techniques. Mean concentrations of all-transretinol for mouse heart, skeletal muscle, and epidydimal fat were 0.37, 0.12, and 2.76 nmol/g, respectively (Table III). No retinyl esters were detected in heart or skeletal muscle; however, as previously reported (24), retinyl esters were detected in adipose tissue (data not shown). Our low limit of detection for retinyl ester measurement was less than 0.003 nmol/g. Thus, CRBP-III is expressed in tissues that contain significant alltrans-retinol concentrations but in two tissues that contain little or no retinyl ester. DISCUSSION In this report, we describe a new member of the intracellular FABP protein family, CRBP-III. CRBP-III is highly expressed in adult tissues that express little or no CRBP-I and CRBP-II, viz heart and muscle. In adipose tissue it is expressed along with CRBP-I. In contrast, CRBP-III is not expressed in liver, lungs, testis, or kidney, which express relatively high levels of CRBP-I. CRBP-III is expressed, albeit it at relatively low levels, in small intestine along with CRBP-II. Of the FABP proteins, CRBP-III shares the most sequence identity with CRBP-I and CRBP-II, 57 and 56% identity, respectively. Like CRBP-I and CRBP-II it is Յ40% identical with other members of the family (1)(2)(3)(4)(5). It shares predicted structural motifs with the CRBP-I and CRBP-II proteins. Like CRBP-I and CRBP-II, CRBP-III carries a 2-amino acid insertion between ␤E and ␤F sheets and a lysine residue at position 40 (Fig. 1B) that is not found in other members of the FABP family (6 -10, 30 -37).
CRBP-III binds all-trans-, 13-cis-, and 9-cis-retinol with relatively high affinity but does not bind retinoic acid or retinaldehyde isomers, long chain, short chain, or branched chain fatty acids or fatty alcohols. The amino acid sequence of CRBP-III is consistent with its binding specificity. The positively charged lysine 40 helps stabilize binding of retinol to CRBP-I and CRBP-II through interactions with the polyene chain of the retinoid (6 -10, 30 -37). At positions 128 and 130, which also contact ligand, CRBP-III, like CRBP-I and CRBP-II, carries glutamine and phenylalanine. FABP family members that primarily bind fatty acids contain arginine and tyrosine residues at positions 128 and 130, respectively. At position 108, a site critical in determining ligand specificity, CRBP-III contains a histidine residue; CRBP-I and CRBP-II carry glutamine, and FABP proteins that bind fatty acids carry arginine (6 -10, 30 -37). For CRBP-I and CRBP-II, the glutamine residue at position 108 is proposed to stabilize retinol binding through hydrogen bond formation with the hydroxyl group of the retinol molecule. Substitution of the glutamine at position of CRBP-I with an arginine residue decreases the affinity of the mutant CRBP-I for retinol. Moreover, the same substitution in CRBP-II increases the affinity of the mutant CRBP-II for both retinoic acid and fatty acids (30 -33). Similarly for intestinal FABP, substitution of the arginine at position 108 with a glutamine increases binding affinity of the mutant intestinal FABP for all-trans-retinol (35). We propose that histidine 108 in CRBP-III acts both as a hydrogen bond donor and hydrogen bond acceptor. Histidine 108 thus could stabilize retinol binding by participating as a donor in hydrogen bond formation between the hydroxyl group of the retinol molecule in a manner similar to but less effective than the interaction between retinol and glutamine 108 of CRBP-I.
CRBP-III binds all-trans-retinol with an apparent dissociation constant of 109 nM, compared with CRBP-I and CRBP-II, which bind with apparent dissociation constants in the range of 10 -16 nM (1)(2)(3)(4)(5). The weaker binding affinity of CRBP-III for retinols is consistent with the idea that CRBP-III may have a different physiological role(s) than CRBP-I or CRBP-II. CRBP-I and CRBP-II are proposed to play important roles in retinol metabolism, especially in facilitating retinol esterification (2,14,15) and retinol oxidation to retinoic acid (2,5). CRBP-I knockout mice are unable to esterify retinol efficiently and have difficulty in maintaining hepatic retinoid stores (15). On this basis, it has been suggested that CRBP-I and CRBP-II deliver retinol to the enzyme lecithin:retinol acyltransferase, which catalyzes retinyl ester formation (2,14,15). CRBP-III cannot play this role in heart or skeletal muscle, because these tissues lack detectable levels of retinyl ester (Table III).
Retinoids are needed to maintain the health of the heart and cardiovascular system in the adult. Experimentally induced myocardial infarction induces mobilization of hepatic retinol and its delivery to the damaged heart (38). We suggest that CRBP-III mediates retinol uptake from circulating retinol-RBP and/or helps facilitate retinol oxidation to retinoic acid within the heart. The precise biochemical mechanisms responsible for these roles must still be elucidated.
At days 9.5-10.5 during mouse embryogenesis, expression of CRBP-III is most pronounced in the developing heart and cardiovascular system (Fig. 3). CRBP-I is much more widely distributed at this embryonic stage than CRBP-III and is expressed highly in liver, gut, tongue, and spinal cord and other components of the central nervous system (39,40). At embryonic day 12.5, CRBP-I is expressed in the epicardial layer of the developing heart (40). It is well established that retinoids are critically needed during embryonic heart and cardiovascular system development. Both vitamin A deficiency and vitamin A excess during pregnancy cause a wide spectrum of defects including embryonic heart defects (41). Knockout mice that lack retinaldehyde dehydrogenase type 2 and the ability to synthesize retinoic acid from retinol die at mid-gestation showing severe heart defects (42). Similarly, knockout mice lacking functional retinoid X receptor ␣ develop severe cardiovascular values for all-trans-and 13-cis-retinol are given as the means Ϯ S.D. for four independent determinations, and that for 9-cis-retinol is given as the mean of duplicate independent determinations. No binding was evidenced for alltrans-, 13-cis-, or 9-cis-retinoic acids, all-trans-, 13-cis-, or 9-cis-retinaldehydes, phytanic acid, octanoic acid, lauric acid, myristic acid, palmitic acid, oleic acid, linolenic acid, arachadonic acid, docosahexanoic acid, palmityl alcohol, petrosenlinyl alcohol, ricinolenyl alcohol, sphingosine, or lignoceric ceramide.  III All-trans-retinol concentrations in mouse heart, skeletal muscle, and epidydimal fat Values are given as the means Ϯ S.D. for n ϭ 4. The values represent only tissue levels of all-trans-retinol. The presence of all-trans-retinyl esters was detected in adipose tissue but not in heart or muscle. Our low limits of detection for both retinol and retinyl ester are each less than 0.003 nmol/g of tissue. defects and die by embryonic day 16 (43). A major cause of the embryonic lethality associated with the retinoid X receptor ␣ knockout has been attributed to impaired cardiovascular system development (43). Interestingly, CRBP-III, retinaldehyde dehydrogenase and retinoid X receptor ␣ are colocalized in the myocardium ventricules of the developing mouse heart at embryonic days 9.5-10.5 (42,43). Considering both the temporal and spatial pattern of expression of CRBP-III and these other proteins in the embryonic heart, we speculate that CRBP-III plays a role in maintaining normal retinoic acid homeostasis in the developing heart and cardiovascular system. Both CRBP-III and CRBP-I are expressed in adipose tissue (Ref. 24 and Fig. 2B). Adipose tissue consists of adipocytes and stromal vascular cells, which include adipocyte precursors, macrophages, fibroblasts, red blood cells and other blood cells (44). Stromal vascular cells account for over 50% of the protein and DNA content of rat white adipose tissue (44 -46). CRBP-III is expressed in both adipocytes and stromal vascular cells (Fig.  2B). The expression of CRBP-I in adipose tissue is almost entirely limited to stromal vascular cells (Ref. 24 and Fig. 2B). The distinct cellular expression patterns of CRBP-III and CRBP-I implies that the two proteins have different physiological roles within adipose tissue.
Adipose tissue also contains ALBP, a FABP family member that shares 38% amino acid identity with mouse CRBP-III and CRBP-I. ALBP, which binds unsaturated and saturated fatty acids and retinoic acid but not retinol, is expressed more highly in adipocytes compared with stromal vascular cells (11,47,48).
One aim of this work was to elucidate the mechanism of tissue retinol uptake and intracellular processing. We earlier showed a correlation between lipoprotein lipase expression in heart, skeletal muscle, and adipose tissue and the amount of retinol taken up by these tissues from chylomicron-bound retinyl ester (16,49). We proposed that lipoprotein lipase facilitates uptake of postprandial retinol by hydrolyzing retinyl esters (16,49). In tissues other than heart and skeletal muscle, intracellular processing of the retinol is thought to be promoted by CRBP-I (2,14,17,50,51) or CRBP-II (2,14,52). We now report the identification and characterization of a new intracellular retinol binding protein, CRBP-III, which may substitute for CRBP-I or CRBP-II in these tissues. However, definitive evidence for this or other physiologic actions of CRBP-III remains to be gathered, and this will be the direction of our future studies.