Expression and Characterization of a Murine Enzyme Able to Cleave β-Carotene

Because animals are not able to synthesize retinoids de novo, ultimately they must derive them from dietary provitamin A carotenoids through a process known as carotene cleavage. The enzyme responsible for catalyzing carotene cleavage (β-carotene 15,15′-dioxygenase) has been characterized primarily in rat intestinal scrapings. Using a recently reported cDNA sequence for a carotene cleavage enzyme fromDrosophila, we identified a cDNA encoding a mouse homolog of this enzyme. When the cDNA was expressed in eitherEscherichia coli or Chinese hamster ovary cells, expression conferred upon bacterial and Chinese hamster ovary cell homogenates the ability to cleave β-carotene to retinal. Several lines of evidence obtained upon kinetic analyses of the recombinant enzyme suggested that carotene cleavage enzyme interacts with other proteins present within cell or tissue homogenates. This was confirmed by pull-down experiments upon incubation of recombinant enzyme with tissue 12,000 ×g supernatants. Matrix-assisted laser desorption ionization-mass spectrometry analysis of pulled-down proteins indicates that an atypical testis-specific isoform of lactate dehydrogenase associates with recombinant carotene cleavage enzyme. mRNA transcripts for the carotene cleavage enzyme were detected by reverse transcription-polymerase chain reaction in mouse testes, liver, kidney, and intestine. In situ hybridization studies demonstrated that carotene cleavage enzyme is expressed prominently in maternal tissue surrounding the embryo but not in embryonic tissues at 7.5 and 8.5 days postcoitus. This work offers new insights for understanding the biochemistry of carotene cleavage to retinoids.

used to screen mouse expressed sequence tag sequences deposited in GenBank TM (National Center for Biotechnology Information, Bethesda, MD) to identify potential mouse homologs of the enzyme. A cDNA sequence obtained from a cDNA library prepared from mouse kidney (accession number AW044175) sharing 48% sequence identity with the D. melanogaster was identified and obtained from Research Genetics, Inc. (Huntsville, AL). This cDNA was sequenced fully in both directions by the Columbia University Comprehensive Cancer Center Core DNA Sequencing Facility.
Expression and Purification of Carotene 15,15Ј-Dioxygenase-An open reading frame for the carotene 15,15Ј-dioxygenase cDNA described above was amplified by PCR from the original clone (AW044715) and subcloned into the mammalian expression vector pcDNA3 (CCE/pcDNA3) (Invitrogen, San Diego) and into the bacterial expression vector pGEX-3X (CCE/pGEX) (Amersham Pharmacia Biotech). Both subclones were sequenced to verify orientation and also the correct reading frame in the case of the pGEX-3X vector. This latter vector was used to express carotene 15,15Ј-dioxygenase as fusion protein with bacterial glutathione-S-transferase (GST). The recombinant fusion protein was purified by affinity chromatography on glutathione-Sepharose (Amersham Pharmacia Biotech) according to the manufacturer's instructions. Expression of a GST-containing fusion protein in Escherichia coli was confirmed by Western blot analysis as described below. CCE/pcDNA3 was transfected into CHO cells using calciumphosphate transfection (25), and carotene 15,15Ј-dioxygenase expression was verified by Northern blot analysis and in vitro enzyme activity assay.
Western Blot Analysis-Recombinant dioxygenase-GST fusion protein was subjected to SDS-PAGE on a 12% gel and blotted onto a nitrocellulose membrane at 100 V for 1 h. The blot was incubated with 10% milk blocking buffer (non-fat milk in phosphate-buffered saline (PBS; 10 mM sodium phosphate, pH 7.4, 150 mM NaCl) containing 0.3% (v/v) Tween 20 (PBST)) for 1 h followed by incubation with primary goat antibody against bacterial GST protein (1:1,000 dilution, Amersham Pharmacia Biotech) for 1 h. The membrane was washed twice with PBST for 10 min and was then incubated with rabbit anti-goat IgG conjugated with alkaline phosphatase (1:10,000) for 1 h. The blot was washed twice with PBST, as described above, and signal was detected with 10% nitro blue tetrazolium and 5% 5-bromo-4-chloro-3-indolyl phosphate in buffer containing 100 mM NaCl, 100 mM Tris (pH 9.5), 5 mM MgCl 2 .
Northern Blot Analysis-Total RNA was isolated from mouse tissues employing RNAzol B (Tel-test, Inc., Friendswood, TX) according to the manufacturer's instructions for RNA isolation from tissues. For this purpose, adult male C57BL/6J mice maintained throughout life on a control chow diet were sacrificed by cervical dislocation, and liver, testes, kidney, and intestine were removed and immediately placed in liquid nitrogen and stored at Ϫ70°C until RNA extraction. The quality of the RNA was examined by visualizing 28 S and 18 S ribosomal RNA bands on ethidium bromide-stained denaturing agarose gel. For all of the total RNA preparations employed in our studies, the RNA appeared intact with the ratio of ethidium bromide staining intensities of the 28 S and 18 S bands being ϳ2.
For Northern blot analysis, 20 g of total RNA was loaded onto a 1.2% denaturing agarose gel, and RNA species were separated by electrophoresis at 80 V for 4 h. After electrophoresis, the agarose gel was blotted onto a positively charged Nylon membrane (Hybond-Nϩ, Amersham Pharmacia Biotech). A 32 P-labeled ([␣-32 P]dCTP) probe was prepared from the full-length murine carotene 15,15Ј-dioxygenase cDNA using a High Prime® kit (Roche Molecular Biochemicals, Indianapolis), according to the manufacturer's instructions. Hybridization was carried out overnight at 65°C. The final wash was with 2 ϫ SSC, 0.1% SDS at 65°C. The washed blot was exposed to x-ray film for autoradiography.
To verify expression of mouse carotene 15,15Ј-dioxygenase cDNA in CHO cells, total RNA was isolated from CHO cells transfected with either empty vector (pcDNA3) or CCE/pcDNA using RNAzol B according to the manufacturer's instructions for RNA isolation from cultured cells. This total RNA was subjected to the same electrophoresis and hybridization protocols described above.
Assay of Carotene 15,15Ј-Dioxygenase-In vitro enzyme assays for carotene 15,15Ј-dioxygenase were carried out using a procedure reported previously (20). As an enzyme source, depending on the experiment, we employed either crude bacterial homogenate (CCE/pGEX), the 12,000 ϫ g supernatant obtained from the bacterial homogenate, purified GST fusion protein of carotene 15,15Ј-dioxygenase, the 10,000 ϫ g supernatant obtained from CHO cells transfected with carotene 15,15Јdioxygenase, or the 9,000 ϫ g supernatant prepared from rat intestinal mucosa scrapings (20). For our standard assay, the enzyme source was incubated with 100 mM Tricine buffer (pH 8.0) containing 15 M ␤-carotene, 0.1 mM ␣-tocopherol, 0.5 mM dithiothreitol, 4 mM sodium cholate, and 15 mM nicotinamide for 1 h at 37°C in a shaking incubator. The final reaction volume was 0.2 ml.
Protein concentrations were measured using Bio-Rad Bradford protein assay reagents according to the manufacturer's instructions employing bovine serum albumin as standard. Specific activity was defined as pmol of retinal produced/mg of protein/h. Enzyme kinetics data were analyzed using EnzFit 5.0 software (Perrella Scientific, Inc.).
Retinoid and carotenoid contents of the assay mixtures were analyzed either by reverse phase or by normal phase high performance liquid chromatography (HPLC) (see below). For reverse phase HPLC analysis, the enzymatic reaction was first terminated by adding 50 l of 37% (v/v) formaldehyde (Sigma) in water, and this mixture was allowed to incubate further at 37°C for 10 min. Retinoids were then extracted into 500 l of acetonitrile, and the extraction medium was incubated on ice for 5 min. After incubation, the assay mixture was centrifuged 10,000 ϫ g for 10 min at 4°C. An aliquot of the resulting supernatant (200 l) was injected directly onto the reverse phase HPLC column and analyzed as described below. For normal phase HPLC analysis, the reaction was stopped by adding 500 l of 100% ethanol, and retinoids and carotenoids were extracted into 2.5 ml of hexane. Small amounts of 13-cis-retinol or TMMP-ROH were added to the extraction mixtures to serve as internal standards for the HPLC analyses.
HPLC Analyses-For reverse phase HPLC analysis, retinoids were separated on a 4.6 ϫ 250-mm Ultrasphere C 18 column (Beckman, Fullerton, CA) preceded by a C 18 guard column (Supelco Inc., Bellefonte, PA), using 10% water in acetonitrile containing 0.1% ammonium acetate as the running solvent at 1 ml/min. For normal phase HPLC analysis, retinal and retinol isomers and ␤-carotene were separated on a 4.6 ϫ 150-mm Supelcosil LC-Si column (Supelco Inc.) preceded by a silica guard column (Supelco Inc.) using hexane:ethyl acetate:butanol (96.9:3:0.1, v/v) as the mobile phase flowing at a rate of 0.8 ml/min. Isomers of retinol and retinal, and ␤-carotene were detected by absorbance of 325, 365, and 450 nm, respectively, using a Waters 996 photodiode array detector (Waters Associates, Milford, MA). Retinol and retinal peaks were identified by comparing retention times and spectral data of experimental compounds with those of authentic standards. Each retinol and retinal isomer was quantitated by comparing its integrated area under the peak with those of known amounts of purified standards. The loss during extraction was accounted for by adjusting the recovery to that of the internal standards, either 13-cis-retinol or TMMP-ROH.
Expression of Recombinant Cellular Retinol-binding Proteins-Recombinant cellular retinol-binding proteins, types I, II, and III (CRBP I, CRBP II, and CRBP III) were expressed in E. coli. For expression of CRBP II, we employed a cDNA encoding rat CRBP II (26) which was cloned in a pMON vector. The rat CRBP II cDNA was provided as a generous gift by Dr. Ellen Li of Washington University, St. Louis. The conditions we employed for induction of CRBP II expression were identical to those described previously in the literature (27). Bacteria expressing the CRBP II were pelleted by centrifugation at 6,000 ϫ g for 15 min, and the pellet was resuspended in 20 mM potassium phosphate buffer (pH 7.4) containing 1 mM EDTA, 10 mM ␤-mercaptoethanol, 15% glycerol, 0.05% sodium azide, and 0.05 mM phenylmethanesulfonyl fluoride. Bacteria was then sonicated and centrifuged again at 10,000 ϫ g for 15 min, and the supernatant was used for pull-down assays as described below. Both mouse CRBP I and CRBP III clones in pET11a vector (Novagen, Madison, WI) were provided as a generous gift by Dr. Silke Vogel of Columbia University. Both clones were induced to express the proteins as described by Vogel et al. (28). Expression of the CRBP I, II, and III was verified by SDS-PAGE followed by Coomassie staining of the gel.
Pull-down Assays-Mouse carotene 15,15Ј-dioxygenase fused to GST was purified using glutathione-Sepharose and incubated with the 9,000 ϫ g supernatant prepared from a homogenate of rat intestinal scrapings or the 10,000 ϫ g supernatant from a mouse testis homogenate for 30 min at room temperature. After incubation, the mixtures were centrifuged at 500 ϫ g for 5 min at 4°C, and unbound proteins were discarded. The testis protein(s) precipitating with the dioxygenase-GST fusion protein bound to the glutathione-Sepharose was considered the pull-down product. The glutathione-Sepharose beads were then washed three times with PBS, and proteins bound to the glutathione-Sepharose were eluted with 10 mM glutathione in 50 mM Tris (pH 8.0). The proteins eluting from the glutathione-Sepharose affinity resin were separated by 12% or 15% SDS-PAGE to identify proteins that might interact with the carotene 15,15Ј-dioxygenase fusion pro-tein. Such protein bands were subjected to both N-terminal sequence analysis by automated Edman degradation and matrix-assisted laser desorption ionization-mass spectrometry (MALDI-MS) by the Protein Chemistry Core Facility of the Howard Hughes Medical Institute at Columbia University.
Characterization of Carotene 15,15Ј-Dioxygenase Expression by RT-PCR-Total RNA (3 g) from liver, testes, intestine, and kidney (for isolation details, see above) was first treated with DNase I to remove any genomic DNA contamination, and reverse transcription (RT) was performed using a preamplification kit (Life Technologies, Inc.) and random hexamers according to the manufacturer's recommendations. The RT reaction (2 l) was subsequently used for the PCR. The primers used for carotene 15,15Ј-dioxygenase expression analysis were 5Ј-AGA-CATGGGGAGGTCTTCTAC-3Ј (forward) and 5Ј-CTCGGGCTGGCAAT-AGACA-3Ј (reverse), which should give rise to a 1,006-base pair product. PCR conditions were in a total reaction volume of 25 l with 3 pmol of each primer, 3 mM MgCl 2 , 1.5 units of Taq-polymerase (Life Technologies, Inc.). Thermal cycling parameters were 94°C for 5 min (94°C for 30 s, 55°C for 30 s, 72°C for 90 s) ϫ 35 cycles and 72°C for 5 min. As negative controls, PCR incubations containing no cDNA template or without reverse transcription were employed. As a positive control, CCE/pcDNA3 plasmid DNA was utilized. The amplicons (20 l) were loaded onto a 1.2% agarose gel, separated by electrophoresis, and visualized on a UV-transilluminator upon staining with ethidium bromide.
In Situ Hybridization-In situ hybridization studies of expression of carotene cleavage enzyme mRNA in the embryonic mouse were performed using digoxigenin-labeled riboprobes essentially as described by Mendelsohn et al. (29). Similarly, the procedures we employed to stage the mouse embryos for their fixation and the sectioning of embryos were also described by Mendelsohn et al. (29). For antisense probes, the mouse carotene cleavage enzyme cDNA in pcDNA3 was linearized with KpnI, and antisense transcripts were generated with SP6 polymerase. For preparation of sense probes, the same cDNA was linearized with XhoI, and sense transcripts were generated with T7 polymerase.

Identification of a cDNA Clone for Mouse Kidney Carotene
15,15Ј-Dioxygenase-By homology search of the mouse expressed sequence tag data base, we identified a clone (accession number AW044715) generated from a mouse kidney cDNA library which showed 48% amino acid sequence identity to a sequence reported previously for a putative carotene 15,15Јdioxygenase from D. melanogaster (22). This cDNA clone was obtained from Research Genetics and sequenced fully in both directions. The cDNA clone appeared to represent a full-length cDNA including polyadenylation sites. An open reading frame consisting of 566 amino acids was encoded by the cDNA. The protein deduced from this sequence has a molecular mass of ϳ64,000. At the time we sequenced the cDNA, the mammalian protein that shared the most sequence identity with the deduced protein was the rat and human retinal pigment epithelium protein, RPE65, which showed 41% identity. Since then, annotated expressed sequence tags encoding carotene 15,15Јdioxygenases from various species have also appeared in Gen-Bank TM (accession numbers AJ271386, chicken; AJ278064, mouse; AF294900, NM_017429, human). The primary sequence alignments of chicken (AJ271386), mouse, and human (NM_017429) carotene cleavage enzymes are shown in Fig. 1. The identity among these sequences is 65% and between mouse and human is 85%. We note a discrepancy between our cDNA sequence and the mouse sequence (AF294900) deposited in GenBank TM which would be manifested at the C terminus of the protein. Because of a one-nucleotide difference in the open reading frame toward the 3Ј-end of the two sequences, our deduced protein has additional 17 amino acids at the C terminus compared with the sequence deposited as AF294900. The primary sequence for mouse carotene 15,15Ј-dioxygenase provided in Fig. 1 reflects our sequencing results for the cDNA clone obtained from Research Genetics (clone 2192191).
Biochemical Characteristics of Recombinant Mouse Kidney Carotene 15,15Ј-Dioxygenase-To determine whether the mouse cDNA exhibits carotene 15,15Ј-dioxygenase activity, we subcloned the open reading frame of the cDNA into a bacterial expression vector, pGEX-3X. E. coli containing this expression plasmid were treated with isopropyl-␤-D-thiogalactopyranoside (Roche) to induce recombinant protein expression. Homogenates prepared from induced bacterial cultures by sonication were subjected to SDS-PAGE, and a prominent protein band at ϳ90 kDa was observed (data not shown). This size corresponds to the expected size of bacterial GST fused to mouse carotene 15,15Ј-dioxygenase. Immunoblot analysis using commercial antiserum against bacterial GST indicated that GST was present as part of this 90 kDa band (data not shown). To identify whether the GST fusion protein possessed the ability to cleave ␤-carotene, bacterial homogenates containing the dioxygenase-GST protein or GST protein alone were incubated for 1 h with 15 M ␤-carotene in our standard assay condition (see "Experimental Procedures"). After incubation, the reaction mixture was analyzed for retinoid and ␤-carotene content by normal phase HPLC. A representative HPLC chromatogram for these assays is provided in Fig. 2. A peak that coelutes with authentic all-trans-retinal was detected for the reaction mixture containing dioxygenase-GST fusion protein but not for the reaction mixture containing GST protein alone (Fig. 2). To confirm this result further, we repeated this analysis using reverse phase HPLC. As expected, we were able to identify a peak with a retention time that corresponded to that of authentic all-transretinal. In addition, spectral analysis of the peaks obtained for both normal and reverse phase HPLC analyses was identical to that obtained upon HPLC analysis of pure all-trans-retinal. For reaction mixtures that were terminated immediately after the addition of ␤-carotene, no all-trans-retinal was produced (Fig. 2). Aside from a small amount of 13-cis-retinal (always less than 10% of the amount of all-trans-retinal produced) that was probably formed through isomerization of all-trans-retinal, no other compounds were detected upon either normal or reverse phase HPLC analysis. Taken together, these data are consistent with the suggestion that the cDNA encodes a protein that catalyzes the cleavage of ␤-carotene, primarily at the central 15,15Ј-double bond.
Cleavage of ␤-carotene by the recombinant GST fusion protein was protein-dependent over the entire protein concentration range examined (0 -550 g of bacterial protein/assay) (Fig.  3A). Similarly, the rate of ␤-carotene cleavage was time-dependent for incubation periods ranging from 0 to 120 min (Fig.  3B). The rate of ␤-carotene cleavage to all-trans-retinal showed saturation with increasing concentrations of ␤-carotene (Fig.  3C). However, the rates of product formation were not linear with respect to either protein concentration or time, and the relationship between ␤-carotene concentration and the reaction velocity was sigmoidal in nature. Computer analysis of data relating the dependence of reaction velocity on substrate concentration best fit the Hill equation, yielding a Hill coefficient of 0.659, a K 0.5 for ␤-carotene of 0.96 M, and a V max of 368 pmol/mg/h. Taken together, these kinetic data suggest that the carotene cleavage reaction involves complex interactions between an insoluble substrate dispersed in mixed micelles with cholate and Tween 40, carotene 15,15Ј-dioxygenase, and prob- ably other proteins present within the homogenates employed for these assays.
The early literature indicates that intestinal carotene 15,15Јdioxygenase requires ferrous iron for activity and that this enzyme is extremely sensitive to the presence of metal chelators like o-phenanthroline and ␣,␣Ј-bipyridyl (13,16). To test whether the carotene cleavage enzyme studied here is also iron-dependent, we investigated whether the activity of the recombinant fusion protein is sensitive to the presence of chelating agents. To this end, the effects of exogenously added EDTA, ␣,␣Ј-bipyridyl, and o-phenanthroline on recombinant carotene cleavage activity were tested for chelator concentrations ranging between 0 and 10 mM (Fig. 4). As can be seen in Fig. 4, the addition of o-phenanthroline to the enzyme strongly inhibited enzymatic activity, exhibiting more than 90% inhibition at a concentration of 0.5 mM. ␣,␣Ј-Bipyridyl also showed inhibitory actions, albeit less potent ones than o-phenanthroline. EDTA only weakly inhibited recombinant dioxygenase activity.
Tissue Distribution of Carotene 15,15Ј-Dioxygenase in the Mouse-The expression pattern of carotene 15,15Ј-dioxygenase mRNA in different mouse tissues was investigated by Northern blot analysis using a full-length 32 P-labeled cDNA as probe. For this purpose, total RNA was prepared from mouse small intestine, kidney, liver, heart, muscle, and testes. Surprisingly, Northern blot analysis revealed that carotene 15,15Ј-dioxygenase mRNA is expressed most highly in the testis followed by the liver and kidney. By Northern blot of total tissue RNA, we were unable to detect expression in the small intestine. The testis expressed two transcripts (2.6 and 1.7 kilobases). The smaller transcripts were detected more strongly when the 3Јportion of the open reading frame of cDNA was used as a probe as compared with when the 5Ј-portion of the open reading frame served as a probe. It is possible that the smaller transcript might arise as a splice variant that does not contain an early exon(s). Because carotene 15,15Ј-dioxygenase was first identified in homogenates prepared from intestinal scrapings, we were concerned that we could not demonstrate expression of our cDNA clone in total RNA prepared from the entire small intestine. Hence, we performed RT-PCR analysis using total RNA prepared from mouse testis, small intestine, kidney, and liver (Fig. 5A). This analysis revealed that all four tissues express carotene 15,15Ј-dioxygenase mRNA. Moreover, the pattern of expression observed agrees with the Northern blot analysis, showing the highest level of expression in the testis followed by liver, kidney, and with a much lower level in small intestine. Also consistent with our Northern blot analysis, the testes showed two differently sized amplification products when primers spanning the 5Ј-portion of the open reading frame were employed for the RT-PCR analysis (data not shown), adding support to the notion that the smaller sized transcript is missing a portion of 5Ј-region of the cDNA.
We also investigated the expression pattern of carotene 15,15Ј-dioxygenase during early stages of mouse embryogenesis. For this purpose, we examined by in situ hybridization the tissue sites of expression of mRNA for carotene 15,15Ј-dioxygenase in mouse embryos in utero at embryonic days 7.5 (E7.5) and 8.5 (E8.5). As seen in Fig. 5B, for both E7.5 and E8.5, mRNA for this enzyme is highly expressed in maternal tissue surrounding the embryo but is not present at detectable levels in embryonic tissues.
Partial Purification and Characterization of Carotene 15,15Ј-Dioxygenase-Because bacterial homogenates containing dioxygenase-GST fusion protein possessed substantial amounts of ␤-carotene cleavage activity that was not present in homogenates prepared from bacteria transformed with the empty vector, we attempted to purify the mouse carotene 15,15Ј-dioxygenase to homogeneity from this bacterial source. Previous attempts to purify intestinal carotene 15,15Ј-dioxygenase from several mammalian species using classic column chromatography techniques have been reported in the literature (13,16). These reports indicate that purification of the enzyme to homogeneity could not be achieved. Because the recombinant dioxygenase-GST protein could be purified rapidly through its interaction with glutathione-Sepharose and the GST portion of the protein subsequently cleaved by factor X a potentially giving rise to a purified preparation of mouse carotene 15,15Ј-dioxygenase, we undertook such an affinity purification. When purified protein preparation was examined on SDS-PAGE followed by Coomassie staining, only the 90-kDa dioxygenase-GST fusion protein and one other 60-kDa protein were detected. However, the specific activity of this purified dioxygenase-GST fusion protein was only 30% of that observed in the unfractionated bacterial homogenate employed as the source of the fusion protein. This suggested that to maintain the activity of the dioxygenase-GST fusion protein another component(s) present in the crude homogenate is required for maintaining its optimal activity. Upon testing this possibility, we observed that the activity of the purified dioxygenase-GST fusion protein could be reconstituted through addition of either the 12,000 ϫ g supernatant prepared from homogenates of E. coli transformed with empty vector pGEX-3X or addition of the 10,000 ϫ g supernatant from homogenate of CHO cells transfected with empty vector pcDNA3. Because neither the bacterial supernatant nor the CHO cell supernatant possessed any detectable carotene 15,15Ј-dioxygenase activity, this indicated that the purified recombinant dioxygenase-GST fusion protein must be missing some essential component(s) that is needed either for catalysis of carotene cleavage or to maintain the stability of the enzyme.
To explore this observation in more detail, we investigated first whether the factor(s) present in the 12,000 ϫ g CHO cell supernatant was dialyzable through a membrane with a molecular weight cutoff of 3,500. Even after exhaustive dialysis against PBS, addition of the dialyzed bacterial supernatant to the purified recombinant dioxygenase-GST fusion protein was still able to restore carotene cleavage activity to the purified fusion protein. Thus, it would appear that the component(s) responsible for restoring the dioxygenase activity is not a small dialyzable molecule. We next asked whether the restoration of carotene cleavage activity came about through a nonspecific process that was dependent solely on protein concentration. We did not observe any restoration of enzymatic activity when the purified fusion protein was incubated with bovine serum albumin at a protein concentration that was similar to those of the 12,000 ϫ g supernatant used to restore activity. This suggested that a specific interaction between a component(s) in the bacterial supernatant and the CHO cell homogenates is needed to restore ␤-carotene cleavage activity to the recombinant dioxygenase-GST fusion protein. We asked further whether the factor(s) responsible for reactivation was heat-sensitive. When either boiled (5 min) or native CHO cell supernatant was added to the purified recombinant dioxygenase-GST preparation, only the undenatured supernatant restored enzymatic activity. Moreover, this reactivation of the fusion protein by CHO cell supernatant was protein-dependent (Fig. 6).
The recombinant dioxygenase-GST fusion protein was stable when stored at Ϫ20°C for over a month if kept as bacterial homogenate and in the absence of repeated freezing and thawing. Dioxygenase activity, however, decreased rapidly when 1% Triton X-100 was added to bacterial homogenates to improve the lysis of bacterial wall. For Triton X-100-containing homogenates, most of the dioxygenase activity was lost within 2-3 days even when the homogenates were kept in Ϫ20°C.
Interactions of Recombinant Mouse Carotene 15,15Ј-Dioxygenase with Other Mammalian Proteins-Because both our kinetic studies reported above and our studies on the stability of carotene cleavage enzyme suggested that proteins present in cells and tissues are likely important for mediating and/or maintaining carotene 15,15Ј-dioxygenase activity, we undertook pull-down experiments to identify proteins that interact physiologically with the enzyme. For this purpose, dioxygenase-GST fusion protein bound to glutathione-Sepharose was incubated with supernatant fractions prepared either from mouse testis homogenate or rat intestinal scraping homogenate; after repeated washes against PBS, materials that interacted with the dioxygenase-GST fusion protein were analyzed by 12% SDS-PAGE. As can be seen in Fig. 7, a unique testis protein specifically coprecipitated with the glutathione-Sepharose-bound dioxygenase-GST fusion protein. The identity of this pulled-down protein was determined by MALDI-MS analysis. The sizes of the tryptic fragments obtained upon digestion (Fig. 8) of the pulled-down protein were identical to those predicted by data base scan (ProFound at prowl.rockefeller.edu/cgi-bin/ProFound and MS-Fit at prospector.ucsf. edu) for tryptic digests of mouse L-lactate dehydrogenase C (LDH-C). Two prominent protein bands were pulled down from the rat intestinal scraping 9,000 ϫ g supernatant preparation.
However, when these protein bands were taken for N-terminal sequence analysis by automated Edman degradation, the Nterminal sequences were found to be blocked. Upon MALDI-MS analysis, tryptic fragments obtained from each of the bands did not match those of any protein sequence reported in data bases (ProFound at prowl.rockefeller.edu/cgi-bin/ProFound and MS-Fit at prospector.ucsf.edu). Thus, although we were able to detect rat intestinal proteins that apparently interact with mouse carotene 15,15Ј-dioxygenase, we were unable to identify these proteins. Because CRBP II is proposed to have a role in facilitating ␤-carotene cleavage to retinal and in facilitating the subsequent reduction of retinal to retinol (30, 31), we examined whether recombinant CRBP II expressed in bacteria will un-FIG. 6. The activity of purified mouse carotene 15,15-dioxygenase is restored in a protein-dependent manner upon addition of protein fraction from sham transfected CHO cells. Gluthathione-Sepharose affinity-purified mouse carotene 15,15Ј-dioxygenase-GST fusion protein (ϳ 40 g of total protein) was incubated with increasing amounts of the 12,000 ϫ g supernatant protein prepared from sham transfected CHO cell homogenates, and the resulting mixture was assayed in the presence of 15 M ␤-carotene for 1 h at 37°C. The production of all-trans-retinal was monitored by normal phase HPLC at 365 nm. The 12,000 ϫ g supernatant was added to the reaction mixture either as native protein or after 5 min of denaturation at 100°C. All-trans-retinal production was increased upon addition of increasing amounts of native CHO cell homogenate. This effect saturated when ϳ280 g (40 l) of protein was added to the affinity-purified carotene 15,15Ј-dioxygenase. q, native CHO cell homogenate; E, boiled CHO cell homogenate. dergo protein-protein interactions with the purified dioxygenase-GST fusion protein. Thus, we incubated the 12,000 ϫ g supernatant of bacteria homogenates containing recombinant rat CRBP II with purified dioxygenase-GST fusion protein bound to glutathione-Sepharose. No significant binding was detected between the two recombinant proteins upon SDS-PAGE analysis of the glutathione-Sepharose pull-down product. Similarly, we were unable to pull down either mouse CRBP I or mouse CRBP III upon incubation of these recombinant proteins with the purified dioxygenase-GST fusion protein. DISCUSSION Because animals are incapable of the de novo synthesis of retinoids, the first biochemical event needed to facilitate retinoid actions within the body is the enzymatic conversion of provitamin A carotenoids, like ␤-carotene, to retinal. Through studies carried out in the 1960s, it was established that this conversion, termed carotene cleavage, is catalyzed by the enzyme carotene 15,15Ј-dioxygenase (11,32,33). This early work demonstrated convincingly that this enzyme is a soluble ferrous iron-dependent protein that utilizes molecular O 2 as a substrate and incorporates both O 2 atoms into the product retinal (11,13,16,32,33). Carotene cleavage activity has been demonstrated in cytosol preparations from small intestine, liver, lung, and kidney for a variety of mammalian species including man, rats, hogs, guinea pigs, and rabbits (10, 14 -21). Because carotene 15,15Ј-dioxygenase lost enzymatic activity when attempts were made to purify it from cytosol fractions, details regarding the biochemical properties of this enzyme(s) have remained elusive. The identification and cloning of a carotene cleavage enzyme from maize mutants (34) several years ago facilitated the recent cloning of a cDNA from D. melanogaster that, upon expression in bacteria, catalyzed ␤-carotene cleavage to retinal (22). By sequence homology analysis employing the published sequence for this Drosophila cDNA, we have identified a cDNA that is expressed in adult mouse intestine, liver, kidney, and testis and maternal tissues during embryogenesis and which encodes a 64-kDa protein possessing carotene 15,15Ј-dioxygenase activity.
When this mouse cDNA was expressed in either bacteria or CHO cells that lack endogenous carotene cleavage activity, expression of the cDNA conferred the activity on the bacterial and CHO cell homogenates (Fig. 2). All-trans-retinal along with some 13-cis-retinal (always Ͻ10% of all-trans-retinal) were the sole products that could be detected upon incubation of alltrans-␤-carotene with the recombinant protein. The 13-cis-retinal likely arises as an isomerization product of all-trans-retinal produced upon all-trans-␤-carotene cleavage because the concentration of 13-cis-retinal detected was directly dependent on the amount of all-trans-retinal produced upon ␤-carotene cleavage. HPLC peaks suggesting the presence of apocarotenals in extracts from incubation mixtures were not observed. These data indicate that cleavage of the ␤-carotene occurs primarily at the central 15,15Ј-carbon-carbon double bond. Unexpectedly, the dependence of reaction velocity for the recombinant dioxygenase showed a sigmoidal relationship with ␤-carotene concentration. Our kinetic data best fit the Hill equation. Analysis of the data gave rise to a K 0.5 for ␤-carotene of 0.95 M, a V max of 368 pmol/mg/h, and a Hill coefficient of 0.659. As far as we are aware, the early literature on the kinetic properties of carotene 15,15Ј-dioxygenase indicates that the enzyme displays a hyperbolic relationship between reaction velocity and substrate concentration, with reported apparent K m values ranging from 0.52 to 9.5 M and reported V max values ranging from 23.8 to 1,300 pmol/mg protein/h (13,(35)(36)(37)(38)(39). We do not understand the basis for this discrepancy between our kinetic data obtained using recombinant enzyme and data obtained by early investigators employing partially purified enzyme preparations. However, in keeping with this early literature, the recombinant mouse dioxygenase was markedly inhibited by the chelating agents o-phenanthroline and ␣,␣Јbipyridyl, but to a much lesser degree by EDTA (Fig. 4). This observation is consistent with the reported properties of partially purified intestinal carotene cleavage enzyme obtained from hog, rabbit, and guinea pig (13,39). In this regard, the mammalian carotene 15,15Ј-dioxygenase may be different from the Drosophila enzyme, which was strongly inhibited by 10 mM EDTA (22). Taken together, these date support the conclusion that we have identified a murine cDNA that encodes an enzyme that possesses catalytic properties that are very similar to those originally reported for mammalian carotene 15,15Ј-dioxygenase.
Our attempts to purify catalytically active recombinant mouse carotene 15,15Ј-dioxygenase-GST fusion protein to homogeneity from bacterial homogenates were unsuccessful. As proved to be the case for early attempts to purify carotene 15,15Ј-dioxygenase from tissue sources, the purified dioxygenase rapidly lost catalytic activity. However, the activity of the purified recombinant dioxygenase was quickly restored upon addition of either the 12,000 ϫ g supernatant from homogenates of bacteria transformed with empty pGEX-3X vector or the 12,000 ϫ g supernatant from sham transfected CHO cell homogenates that do not express carotene 15,15Ј-dioxygenase mRNA. Thus, it appears that a protein(s) present in these supernatants is responsible for this effect and that the activating factor(s) was both nondialyzable and heat-labile. It is possible that such interactions are significant for maintaining or facilitating the physiologic actions of carotene 15,15Ј-dioxygenase, but this remains to be established conclusively by further investigations.
Based on the observations regarding the need of the recombinant carotene cleavage enzyme for other protein factors to maintain its activity and on the kinetic data that suggest complex interactions between the ␤-carotene and the protein species that catalyze its cleavage to retinal, we investigated by pull-down assay the possibility that recombinant dioxygenase-GST fusion protein interacts with other proteins present in tissues. SDS-PAGE analysis of pull-down assays employing the 12,000 ϫ g supernatant from mouse testis, the tissue with the highest level of carotene 15,15Ј-dioxygenase expression (see Fig. 5A), showed a major protein species that copurified with the dioxygenase-GST fusion protein bound to glutathione-Sepharose. MALDI-MS analysis of the protein excised from the SDS-PAGE gel identified the protein as an atypical testis-specific LDH-C. LDH-C was originally identified in the 1960s and demonstrated to catalyze the reversible oxidation of lactate and other secondary alcohols (45,46). However, the physiologic role of this enzyme has often been questioned because the other widely distributed lactate dehydrogenase species (LDH-A and LDH-B) are both highly expressed in the testis (45,46). We are unable to find in the literature any information as to whether LDH-C will catalyze either the oxidation or the reduction of aldehydes like retinal, the product of ␤-carotene cleavage. If LDH-C can catalyze either retinal reduction to retinol or retinal oxidation to retinoic acid, this would suggest that carotene 15,15Ј-dioxygenase acts physiologically in concert with other proteins involved in retinoid metabolism. We are presently investigating this possibility.
The tissue pattern of carotene 15,15Ј-dioxygenase mRNA expression in the mouse is somewhat surprising (Fig. 5). This expression pattern does not overlap fully with carotene 15,15Јdioxygenase activity levels reported in early studies, which indicate that the highest specific activity level of carotene 15,15Ј-dioxygenase is present in small intestine followed by the liver, brain, lung, and kidney (20). As seen in Fig. 5, dioxygenase mRNA expression is relatively low in the mouse small intestine compared with testis, liver, and kidney. This may be attributed to species differences between the mouse and the rat (and other mammalian species) because the published studies of tissue activity levels have not employed mouse small intestine as a source of enzyme activity. Moreover, the literature indicates that carotene 15,15Ј-dioxygenase activity is limited to only the proximal portion of the small intestine, including the duodenum and part of jejunum (8,15,40) and is present only in mature functional enterocytes within these anatomic regions (18). Because we were concerned about the possible rapid degradation of intestinal RNA, as a source of intestinal total RNA we employed the entire small intestine without first separating the mucosal lining. This may also have contributed to the apparent relatively low levels of dioxygenase mRNA in the mouse small intestine having the effect of diluting the dioxygenase mRNA in the total RNA pool. Alternatively, the enzyme encoded by the cDNA isolated from a kidney library may be  Fig. 7, lane 4, was submitted to tryptic digestion followed by MALDI-MS analysis. Based on the mass (m/z) values for the 21 numbered tryptic fragments, the excised protein was identified as the testis-specific isoform, LDH-C. The peaks labeled S represent internal standard, and those labeled T reflect known peaks arising from trypsin fragments. Panel B, predicted mass (m/z) values for each of the numbered lactate dehydrogenase C fragments shown in panel A. These predicted mass values for LDH-C fragments were obtained from protein sequence data bases (ProFound at prowl.rockefeller.edu/cgibin/ProFound and MS-Fit at prospector.ucsf.edu). The measured mass values for these tryptic fragments differed for each fragment by less than 0.5 mass (m/z) unit from those predicted by the data base scan. distinct from the one studied earlier in intestinal homogenates. However, this possibility seems unlikely because the biochemical characteristics of the recombinant enzyme strongly resemble those reported for carotene 15,15Ј-dioxygenase preparations from intestine. The relatively high level of dioxygenase mRNA expression in testes was also unexpected. There are reports in the literature that ␤-carotene accumulates in the testes in humans and ferrets (41,42). It is possible that this testicular ␤-carotene may serve as a previously unrecognized source of retinoid for maintaining testis function.
The presence of carotene 15,15Ј-dioxygenase mRNA, at days 7.5 and 8.5 of mouse embryogenesis, in maternal tissue at the site of embryo implantation suggests that this enzyme may be acting to provide needed retinoid to the embryo. The anatomic site of dioxygenase expression appears to coincide with the region of the uterus which may be important in nutrient exchange. Thus, the presence of carotene cleavage activity in this maternal tissue would be consistent with the notion that maternal carotene cleavage to retinoid is an important mechanism through which the developing embryo can acquire retinoid that is needed for directing gene expression during developmental processes (43).
In summary, we have identified a mouse kidney cDNA that encodes a 64-kDa protein that catalyzes the cleavage of ␤-carotene to retinal in vitro. The recombinant enzyme possesses most of the biochemical properties that have been reported for mammalian intestinal carotene 15,15Ј-dioxygenases. Based on our characterizations, we suggest that the protein encoded by this cDNA is indeed the same enzyme that was first described in the 1960s as carotene 15,15Ј-dioxygenase. This identification raises new possibilities for extending our understanding of both the biochemistry of this enzyme and the biochemical processes that are important for carotenoid processing and retinoid formation in vivo.