Short Chain Dehydrogenase/Reductase Rdhe2 Is a Novel Retinol Dehydrogenase Essential for Frog Embryonic Development*

Background: Retinoic acid regulates expression of numerous genes, but the enzymes required for its biosynthesis are not fully defined. Results: Rdhe2 oxidizes retinol for retinoic acid biosynthesis, and its down-regulation is lethal for frog embryos. Conclusion: Rdhe2 is a previously unrecognized retinol dehydrogenase required for retinoic acid biosynthesis. Significance: Enzymes homologous to frog rdhe2 may carry out similar functions in other species. The enzymes responsible for the rate-limiting step in retinoic acid biosynthesis, the oxidation of retinol to retinaldehyde, during embryogenesis and in adulthood have not been fully defined. Here, we report that a novel member of the short chain dehydrogenase/reductase superfamily, frog sdr16c5, acts as a highly active retinol dehydrogenase (rdhe2) that promotes retinoic acid biosynthesis when expressed in mammalian cells. In vivo assays of rdhe2 function show that overexpression of rdhe2 in frog embryos leads to posteriorization and induction of defects resembling those caused by retinoic acid toxicity. Conversely, antisense morpholino-mediated knockdown of endogenous rdhe2 results in phenotypes consistent with retinoic acid deficiency, such as defects in anterior neural tube closure, microcephaly with small eye formation, disruption of somitogenesis, and curved body axis with bent tail. Higher doses of morpholino induce embryonic lethality. Analyses of retinoic acid levels using either endogenous retinoic acid-sensitive gene hoxd4 or retinoic acid reporter cell line both show that the levels of retinoic acid are significantly decreased in rdhe2 morphants. Taken together, these results provide strong evidence that Xenopus rdhe2 functions as a retinol dehydrogenase essential for frog embryonic development in vivo. Importantly, the retinol oxidizing activity of frog rdhe2 is conserved in its mouse homologs, suggesting that rdhe2-related enzymes may represent the previously unrecognized physiologically relevant retinol dehydrogenases that contribute to retinoic acid biosynthesis in higher vertebrates.

Both deficiency and excess of RA are harmful; therefore, the levels of RA in the cells are tightly controlled. This control is especially important during embryogenesis, when different parts of the embryo require different concentrations of RA for proper development. The correct distribution of RA is achieved by a highly coordinated spatial and temporal expression of RAsynthesizing and RA-degrading enzymes. The enzymes responsible for RA degradation have been identified and include several members of the cytochrome P450 family of proteins (CYP26A1, CYP26B1, and CYP26C1), which catalyze the conversion of RA to 4-hydroxy-, 4-oxo-, and 18-hydroxy-intermediates (for review, see ref. 13). However, the enzymology of RA biosynthesis remains incompletely understood. It has been established that RA is synthesized from local stores of retinol in two steps: first, retinol is oxidized reversibly to retinaldehyde, and then retinaldehyde is oxidized irreversibly to RA (14). At least three members of the aldehyde dehydrogenase family catalyze the second step, the conversion of retinaldehyde to RA. RALDH2 (ALDH1A2) appears to be responsible for all RA signaling activity during early embryonic development (E7.5-E8.5) (15,16), whereas RALDH1 and RALDH3 contribute to RA biosynthesis at later stages (17). However, the overall rate of RA production is determined by the enzymes that catalyze the oxidation of retinol to retinaldehyde, which is the rate-limiting step of this pathway (14). Recent studies provided strong evi-dence that retinol dehydrogenase 10 (RDH10, SDR16C4), a member of the short chain dehydrogenase/reductase (SDR) superfamily of proteins, contributes to RA biosynthesis in human cells and tissues (18) and is also critical for retinaldehyde production during early stages of mouse embryogenesis (19 -21). Interestingly, the defects observed in RDH10-null embryos are more consistent with suboptimal RA signaling rather than with complete RA deficiency. For example, Rdh10 Ϫ/Ϫ embryos retain RA activity in trunk neural tube and lateral mesoderm. At the molecular level, down-regulation of RA-responsive Hoxa1 gene is not as severe as that reported for Raldh2 Ϫ/Ϫ embryos; and RAR␤ expression, which is regulated by RA, is still detectable in branchial arches and foregut tissues of Rdh10 Ϫ/Ϫ embryos (21). These observations strongly suggest that an additional retinol dehydrogenase(s) contributes to the oxidation of retinol to retinaldehyde during embryogenesis and, possibly, in adulthood. The exact molecular identity of such enzyme is currently unknown.
Recently, we reported that a member of the SDR family of proteins named human epidermal retinal dehydrogenase 2 (RDHE2, SDR16C5), is active toward all-trans-retinol (22). RDHE2 displays an increased expression in psoriatic skin (23). Psoriasis vulgaris is a chronic autoimmune inflammatory skin disease characterized by hyperproliferation of keratinocytes (24). Hyperproliferation of keratinocytes can be induced by RA (25), suggesting that RDHE2 may be involved in the pathogenesis of psoriasis through its potential role in RA biosynthesis. RDHE2 shares the highest sequence similarity with RDH10. This observation suggests that SDR enzymes that are related to RDH10 and belong to the same branch of phylogenetic tree (26) may represent the missing retinol dehydrogenases that contribute to RA biosynthesis during embryogenesis as well as in adulthood.
Comparative genomic analysis shows that mammalian genomes contain a second gene homologous to RDHE2, RDHE2-similar (SDR16C6), located next to SDR16C5, whereas the fully annotated amphibian genome of African clawed frog Xenopus tropicalis contains a single gene in the position syntenic to mammalian SDR16C5 and SDR16C6 (Fig. 1). This, together with the phylogenetic tree topology (26), suggests that mammalian SDR16C5 and SDR16C6 may have arisen as a result of a recent duplication event. The protein product encoded by SDR16C6 has not been characterized in any species, but it is possible that SDR16C6 complements the function of SDR16C5, and together these two genes represent the functional equivalent of the single sdr16c5 (rdhe2) in amphibians.
Frog embryos are ideally suited for functional analysis of genes during development because they are easy to manipulate, the results of gain-of-function and loss-of-function assays in the whole organism can be obtained in 2-3 days, and a wide range of developmental effects can be characterized using well established markers. Therefore, in this study, we chose the easily tractable experimental model, Xenopus laevis, as the initial model to determine the physiological significance of SDR16C5/6 for RA biosynthesis during development, while avoiding the redundancy found in mice.

EXPERIMENTAL PROCEDURES
Retinoids-All-trans-retinol was purchased from Toronto Research Chemicals (Toronto, ON, Canada), and all-trans-retinal and all-trans-RA were from Sigma-Aldrich. Stock solutions were prepared in ethanol; working dilutions for cell culture experiments were prepared in ethanol and for treatment of frog embryos, in dimethyl sulfoxide.
Expression Constructs, Morpholino Oligonucleotides, and Antibody Generation-For overexpression in frog embryos and HEK293 cells, the rdhe2 coding sequence was amplified from the neurula stage cDNA using Pfx polymerase (Stratagene-Agilent Technologies) and cloned into EcoRI-XhoI sites of pCS105 vector. The coding sequence of the mouse Sdr16c6 gene was amplified from IMAGE:40129845 expressed sequence tag (EST) clone (Open Biosystems, Thermo Fisher Scientific) and cloned into BamHI-XhoI sites of pCMV-Tag4a vector (Stratagene). The coding sequence of frog raldh2 was PCR-amplified from tailbud embryo cDNA and cloned into XhoI and XbaI sites of pCS2 plasmid. A rescue rdhe2 construct containing mismatches to the endogenous transcripts was generated using site-directed mutagenesis with rdhe2 in pCS105 as a template. Sequences of all primers used in this study are provided in supplemental Table 1. Rdhe2 and rdh10 antisense morpholino oligonucleotides (MOs)(5Ј-CTA CTT GTA AAA GAA TCT CTG CGA G-3Ј and 5Ј-GGA AGA ACT CGA GCA CTA TGT GCA T-3Ј) and nontargeting control morpholino were purchased from Gene Tools (Philomath, OR).
For rdhe2 antibody production, a fragment corresponding to amino acids 67-305 was subcloned into NdeI-BamHI sites of pET19 vector in-frame with the N-terminal His 10 tag (Novagen). The recombinant protein was expressed in Escherichia coli BL21 (DE3) cells and purified using nickel-nitrilotriacetic acid affinity resin (Qiagen) according to the manufacturer's protocol. Polyclonal antiserum was raised in a rabbit by Cocalico Biologicals (Reamstown, PA). The antiserum detected 10 ng of purified recombinant protein fragment at a 1:10,000 dilution.
For Western blot analysis, embryos were homogenized in PBS with 1% Triton X-100 and centrifuged to remove debris; extracts were separated in 12% SDS-PAGE, transferred to PVDF membrane, and incubated with rdhe2 antiserum (1:5,000 dilution) or anti-FLAG antibody (F7425; Sigma-Aldrich) at 1:2,000 dilution. Expression in Cell Culture-HEK293 cells were transfected with rdhe2/pCS105 expression construct using Lipofectamine 2000 (Invitrogen). On the day after transfection, cells were treated with 5 M all-trans-retinaldehyde for 3 h or with 10 M all-trans-retinol overnight. Retinoids were extracted from cell culture medium and cells, separated by normal phase HPLC, and analyzed as described before (18) except the mobile phase was hexane/ethyl acetate/acetic acid (90:9.75:0.25, v/v/v) at a flow rate of 0.7 ml/min.
Overexpression in Frog Embryos-Embryos were obtained and maintained as described previously (27). All experiments involving X. laevis were approved by University of Alabama at Birmingham Institutional Animal Care and Use Committee. To synthesize sense mRNAs for injections, the rdhe2, rescue rdhe2, and raldh2 constructs were linearized with AscI and transcribed using SP6 polymerase and mMessage mMachine kit (Ambion). For phenotypic observation, purified transcripts (1-4 ng/embryo, as indicated in the figures) were injected into two dorsal blastomeres at the four-cell stage, and the embryos were fixed at early tadpole stages. For marker examination, sense mRNAs were injected into one blastomere of two-cellstage embryos together with a lineage tracer that encoded nuclear ␤-galactosidase (0.2 ng of mRNA/embryo). Retinol was added to the medium at stage 8 to the final concentration of 5 or 8 M. Embryos were fixed at stage 20 and processed for in situ hybridization.
Silencing of Endogenous Rdhe2 Expression-Rdhe2 and rdh10 antisense MOs as well as nontargeting morpholino were injected into two dorsal or two ventral blastomeres (30 -100 ng/embryo) or into one dorsal blastomere (15-25 ng/injection) at the four-cell stage. Where specified, MOs were mixed with a lineage tracer (0.2 ng/embryo). Morphants were fixed at stages 11-32 for marker examination or were allowed to develop until stage 42 for phenotypic observations. For rescue experiments, RA was added in dimethyl sulfoxide to culture medium at stage 8 to the final concentration of 0.01 or 0.1 M.
Whole Mount in Situ Hybridization-Rdhe2 construct in pCS105 and constructs encoding marker genes in pBlue-scriptSK(Ϫ) were linearized with EcoRI and transcribed with T3 RNA polymerase (Promega) in the presence of digoxigenin-UTP. In situ hybridization was performed as described previously (28). For histological analyses, stained embryos were embedded in paraffin and sectioned at 50 m.
RA Reporter Activation Assay-RA reporter cells F9-RARE-lacZ (Sil 15) expressing ␤-galactosidase under the control of the RA-responsive element (RARE) of human RAR␤ gene (29) was used to compare the levels of RA in tissue explants from control and rdhe2 knock-down embryos. The previously published procedure was modified as follows: early passage Sil15 cells were seeded into 12-well culture plate and grown to 80% confluence in DMEM medium with 10% FBS and G418 selection. Equal numbers of control embryos and rdhe2 morphants were dissected at stage 28 -30. Tissues were cut into smaller pieces and placed into cell culture inserts with PET membrane. The cell culture inserts were then submerged into the medium over the monolayer of Sil15 cells and incubated overnight. Cells were fixed and ␤-galactosidase activity was visualized as described previously (29).
Phylogenetic Analysis-Xenopus rdhe2 and rdh10 protein sequences were used as queries to search the following versions of invertebrate genomes available through the Ensembl data base: Strongylocentrotus purpuratus Spur2.5, Caenorhabditis elegans WS220, Nematostella vectensis (version 1.0), and Ciona intestinalis JGI2). Nonredundant and EST data bases at NCBI were used to reconstruct the predicted sequences that appeared incomplete. Protein sequences were aligned using ClustalW (30); a phylogenetic tree was obtained using Fitch-Margoliash algorithm in PHYLIP package, version 3.68 (31); Dendroscope 3 (32) was used for graphic representation of the tree.

Xenopus Rdhe2 Exhibits High Retinol Dehydrogenase Activity-
To design sequence-specific primers for isolation of frog rdhe2 cDNA, we searched the X. laevis EST data base. This search revealed the existence of two transcripts derived from rdhe2 gene. The transcripts differed in their 5Ј-untranslated regions but encoded the same 305-amino acid protein ( Fig. 2 and supplemental Fig. 1). The two transcripts may represent pseudoalleles or alternatively spliced forms.
The cDNA encoding X. laevis rdhe2 was obtained by RT-PCR of mRNA isolated from the neurula stage embryos and cloned into pCS105 expression vector as described under "Experimental Procedures." To determine whether X. laevis rdhe2 is enzymatically active and recognizes retinoids as substrates, the rdhe2 expression construct was transiently transfected into HEK293 cells, and the cells were incubated with all-trans-retinol or all-trans-retinaldehyde (Fig. 3, A and B). As shown in Fig. 3C, rdhe2-transfected cells produced 10-fold more RA and retinaldehyde from retinol than mock-transfected cells. At the same time, the amount of retinyl esters formed in these cells was significantly reduced. Thus, rdhe2 dramatically increased the ability of cells to convert all-transretinol to RA and, furthermore, was able to compete effectively for retinol with endogenous retinol-esterifying enzymes (Fig.  3A). Treatment of the cells with retinaldehyde did not reverse the direction of the reaction (Fig. 3D), indicating that rdhe2 is a strictly oxidative enzyme, similar to RDH10 (18).
Rdhe2 Exhibits Tissue-specific Expression Pattern during Embryogenesis-Quantitative PCR analysis of frog embryos at different stages of development using primers specific for each of the two transcripts revealed that both transcripts were present in unfertilized eggs, and their levels increased at the onset of FIGURE 2. Alignment of 5 regions of ESTs corresponding to two major rdhe2 transcripts. The first ATG codon is in bold. Different 5Ј-untranslated regions and a silent one-nucleotide difference in the coding sequence are boxed. Antisense morpholino sequence corresponds to positions Ϫ1 to ϩ24. The mismatches between the morpholino and the partial sequence of rdhe2 rescue construct are marked by asterisks. MARCH 16, 2012 • VOLUME 287 • NUMBER 12 neurulation (data not shown). In agreement with the quantitative PCR data, Western blot analysis of embryos using rdhe2specific rabbit polyclonal antiserum (Fig. 4A) demonstrated that rdhe2 is a maternal protein present in the egg and throughout early embryonic development. In gastrula embryos, rdhe2 was equally distributed between the animal and vegetal fractions and between the dorso-anterior and ventro-posterior halves ( Fig. 4B). At early tadpole stages, rdhe2 expression was detected in notochord, eyes, head, tail, back, and belly (composed mainly of the endoderm) (Fig. 4C). In contrast to gastrula stages, the levels of rdhe2 protein varied among different segments, with the highest amount of rdhe2 detected in the belly and notochord when normalized per total protein content (Fig.  4C). This observation implied that rdhe2 protein underwent redistribution as development proceeded from gastrula to tadpole stages.

Frog Retinol Dehydrogenase Essential for Development
The spatial distribution of rdhe2 was examined further by whole mount in situ hybridization (Fig. 4, D-G). The rdhe2 mRNA was below detection limit in embryos prior to neurula, consistent with the low levels of transcripts as indicated by quantitative PCR. However, the rdhe2 signal was detected in notochord starting at tailbud stages (Fig. 4D) and then in brain, otic vesicle, eye, and olfactory bulb at early tadpole stages (stages 36 -39) (Fig. 4, D and E). Sectioning of embryos in transverse direction revealed that the posterior neural tube did not express high levels of rdhe2 (Fig. 4F), but the staining for rdhe2 intensified in hindbrain region and extended into forebrain, including olfactory bulb (Fig. 4, E and G-I). A strong staining was also observed in the eyes and otic vesicle (Fig. 4, G and H). This broad expression pattern of rdhe2 suggested that it may contribute to RA biosynthesis and regulation of embryonic patterning at multiple sites during development and most notably in the hindbrain and in the notochord.
Overexpression of Frog Rdhe2 Results in Embryonic Defects Resembling RA Toxicity-If rdhe2 promotes the conversion of retinol to RA in vivo, then overexpression of rdhe2 should induce phenotypes that mimic those with elevated levels of RA. To test this hypothesis, we injected the mRNA encoding rdhe2 into two dorsal blastomeres of four-cell stage frog embryos. Western blot analysis confirmed that rdhe2 protein level was . The amount of products extracted from the cells and medium was summarized, normalized per total cellular protein, and expressed as mean Ϯ S.D. (error bars), n ϭ 4. Note that there is essentially no difference in the amount of RA produced by rdhe2transfected cells compared with mock-transfected cells when all-trans-retinaldehyde is provided as substrate (D). This is because the rate of RA biosynthesis in this case is determined by the exogenously provided retinaldehyde (5 M), which masks the difference in the amount of retinaldehyde produced from endogenous retinol by rdhe2-expressing versus control cells. indeed increased (Fig. 4A). Similar to overexpression of frog rdh10 (33), overexpression of rdhe2 induced mostly mild defects. This could be due to the limited availability of retinol in early frog embryos because the main embryonic vitamin A storage form in frogs is a protein-bound retinaldehyde (34). To increase the availability of retinol, we supplemented the medium with 5 or 8 M retinol. At these concentrations, most of the uninjected embryos developed normally, with only a small proportion displaying mild defects in head formation (Fig. 5, A and B). In comparison, embryos injected with rdhe2 RNA exhibited a higher percentage of head malformations. Furthermore, the defects in rdhe2 mRNA-injected embryos were more severe than in control embryos and included reduced head structures, smaller or absent eyes, anteriorly shifted or cyclopic eyes, and shortened body axis (Fig. 5A); all of these defects were previously shown to be associated with overproduction of RA (35,36). This phenotype could be mimicked in control embryos by treatments with higher (Ͼ10 M) retinol concentrations, suggesting that overexpression of rdhe2 poten-tiated the toxicity of retinol treatment by increasing the conversion of retinol to RA. Co-injection of rdhe2 mRNA with raldh2 mRNA resulted in a further increase in the incidence and severity of the phenotypes compared with raldh2 injections alone (Fig. 5B), suggesting that the two enzymes work in concert to convert retinol to RA. Interestingly, increasing the concentration of retinol from 5 to 8 M enhanced the toxicity in embryos injected with rdhe2, but not in embryos injected with raldh2 alone. This indicated that rdhe2 acted in a concentration-dependent manner, catalyzing the rate-limiting step, the oxidation of retinol to retinaldehyde, in RA biosynthesis.
The head defects induced by excessive RA reflect the posteriorization of frog embryos (35,37). To determine whether overexpression of rdhe2 similarly caudalized the embryos, we examined the expression of several regional neural markers. In these experiments, rdhe2 RNA was co-injected with a lineage tracer, the RNA encoding nuclear ␤-galactosidase, into the animal region of one blastomer of two-cell stage embryos. The embryos were collected at the neurula stages and stained with the ␤-galactosidase substrate Red-Gal to label the injected side. Expression patterns of the forebrain marker otx, the retina marker rx2A, the midbrain marker en2, and the hindbrain marker krox20 (for rhombomeres 3/5) were assayed by in situ hybridization. Both krox20 and en2 displayed an anterior shift on the injected side, and some embryos also displayed a reduction of otx expression. In addition, staining with the retina homeobox gene rx2A revealed a reduction of the eye field and a decreased intensity of staining on the injected side of the embryo (Fig. 5C). Injection of tracer mRNA alone did not cause any noticeable changes in marker expression. The observed anterior shift of the neural markers was similar to that caused by overexpression of rdh10 (33) and was consistent with an increased conversion of retinol to RA by the ectopically expressed rdhe2, leading to posteriorization of the embryos.
Endogenous Rdhe2 Is Required for Early Frog Embryogenesis-The above gain-of-function studies demonstrated that rdhe2 can promote RA production to control embryonic development when ectopically expressed. To determine whether the endogenous rdhe2 plays a role in RA-dependent developmental processes, we employed a loss-of-function approach, using the translational inhibitory antisense MO (Fig. 2). The MO recognizes the mRNA sequence from Ϫ1 to ϩ24 relative to the translational start site, a region identical in the two alternative transcripts reported for rdhe2, except for the single-nucleotide difference at ϩ6. The MO is thus expected to knock down translation of rdhe2 from both transcripts. Co-injection of rdhe2 mRNA with rdhe2 MO resulted in a reduction of ectopically expressed rdhe2 at the neurula stages (Fig. 6A), confirming the effectiveness of the MO. The level of endogenous rdhe2 was not affected at this stage (Fig. 6A), which may reflect the existence of a stable pool of maternal rdhe2 protein in neurula. At later stages, rdhe2 MO did decrease the levels of endogenous rdhe2 protein (Fig. 6B).
As expected from the expression pattern of rdhe2, injection of rdhe2 MO in different embryonic regions induced defects in different structures (Fig. 6C). Ventral injections of rdhe2 MO resulted predominantly in defects in tail elongation and the bent tail phenotype, whereas dorsal injections of rdhe2 MO  MARCH 16, 2012 • VOLUME 287 • NUMBER 12 induced a range of defects depending on the doses of MO. At the high dose of 100 ng, about half of the embryos displayed gastrulation defects, and all perished before the tailbud stages. In contrast, parallel injections of 100 ng of control MO did not cause any noticeable abnormalities in embryos observed until early tadpole stages (stage 42). When the doses of rdhe2 MO were reduced to 25-50 ng, the majority of embryos were able to go through gastrulation and neurulation, but displayed an array of defects. The most noticeable defect in embryos injected dorsally was the failure to complete neural tube closure in the head region (Fig. 6C, severe phenotype with an arrow pointing to the open neural tube). This often led to lysis of embryos before tailbud stages. Other defects included the reduction of head and eyes, enlargement of ventro-posterior structures, delay in tail elongation, bent tails, and shortening of the body axis (Fig. 6C). The spectrum of phenotypes observed in rdhe2 morphants resembled defects caused by RA deficiency in Xenopus and mouse embryos (33, 38 -41) and was consistent with the expression pattern of the endogenous enzyme. Manifestation of the phenotypes was dose-dependent, with the higher percentage of embryos displaying more severe defects at 50-ng dose of rdhe2 MO compared with 30-ng dose (Fig. 6D).

Frog Retinol Dehydrogenase Essential for Development
To confirm that the embryonic defects were caused by the down-regulation of rdhe2 and did not represent off-target effects, we generated rdhe2 rescue construct in which six nucleotides were replaced with silent mutations that prevented hybridization of the rescue mRNA with the MO (Fig. 2). Co-injection of rdhe2 MO with the mutated mRNA led to a partial rescue of the MO phenotype, reducing the proportion of embryos with severe defects and increasing the number of embryos with mild phenotypes (Fig. 6E). This result confirmed that the observed morphant phenotypes were caused specifically by the decreased expression of rdhe2 protein.
Interestingly, the phenotypes resulting from the down-regulation of rdhe2 were more severe than those in rdh10 MO-injected embryos (Fig. 6C). Co-injection of rdhe2 and rdh10 MOs resulted in further increase in the number of severely affected embryos (Fig. 6F). The similarity between rdhe2 and rdh10 morphant phenotypes suggested that, like rdh10, rdhe2 is essential for the maintenance of RA levels during development. The fact that the embryo depends on the supply of retinaldehyde produced by retinol dehydrogenases suggests that the retinaldehyde stored in embryos in the protein-bound form may not be readily available for direct oxidation to RA.

Expression of Developmental Marker Genes Is Altered in Rdhe2
Morphants-Because the RA pathway regulates neural patterning and our overexpression assays confirmed that ectopic rdhe2 can shift several regional neural markers anteriorly, we investigated whether these neural markers were also affected by rdhe2 knockdown. This experiment was performed as described above (Fig. 5C) except that rdhe2 MO was injected instead of rdhe2 RNA. In situ hybridization analysis showed that the expression of otx, en2, and krox20 was weakened on the injected side. A posterior shift of the otx and krox20 domains was detected in some embryos (Fig. 7A), in agreement with the known anteriorizing effect of RA deficiency on neural development (33). Staining of en2 was often asymmetrical (Fig. 7B), which might reflect the defects in neural tube closure. In addition, hoxd4, a gene expressed in the posterior hindbrain, showed a consistent reduction and/or posterior shift of its expression domain (Fig. 7C). Collectively, these results suggested that RA signaling was compromised in the brain when the level of rdhe2 protein was reduced.
A recent study in frogs has convincingly shown that the blocking of RA signaling from mesoderm to neuroectoderm suppressed the expression of mid-axial hox genes in neurula embryos and led to distortion of segmental identity and reduction of posterior hindbrain (42). This prompted us to investigate whether the observed reduction of hoxd4 and the posterior hindbrain in rdhe2 morphants reflected a localized decrease in RA signaling. Human, murine, and zebrafish HoxD4 genes were all shown to contain the RARE in their 3Ј-untranslated region; actin (top) antibodies shows a decrease in endogenous rdhe2 levels. C, array of phenotypes displayed by rdhe2 and rdh10 morphants compared with control morphants. Phenotypes were divided into severe, strong, moderate, and mild groups with the representative embryos shown for each group. An arrow shows an incompletely closed neural tube typical for severely affected rdhe2 morphants. Dramatic phenotypic changes in morphants compared with the moderate decrease in protein levels in B may reflect a strong localized effect of MO at sites where newly synthesized rdhe2 protein is critical for embryogenesis. D, proportion of different phenotypic groups in rdhe2 morphants compared with control morphants and wild-type embryos. E, co-injection of rdhe2 rescue mRNA (4 ng) and rdhe2 MO (30 ng) decreasing the frequency and severity of defects displayed by morphants. F, proportion of different phenotypic groups in embryos injected with 25 ng of rdhe2 MO or rdh10 MO, or a mixture of both (50 ng total). Control MO (25 ng) was added to individual rdhe2-and rdh10-MO injections to equalize the total oligonucleotide amount.
however, this has not been reported in frogs. To determine whether the frog hoxd4 expression is regulated by RA, we treated frog embryos with 0.1 M exogenous RA. Whole mount in situ hybridization with hoxd4 probe revealed that the RA treatment resulted in an increased staining intensity and anterior expansion of hoxd4 domain (Fig. 7D). This experiment demonstrated that hoxd4 expression can serve as a sensitive reporter of RA levels in frog embryos. Next, we tested whether hoxd4 expression in rdhe2 morphants could be rescued by exogenous RA supplementation. As shown in Fig. 7D, treatment of rdhe2 morphants with RA restored the expression of hoxd4 in a dose-dependent manner. Supplementation of morphants with a high concentration of RA (0.1 M) expanded hoxd4 expression beyond its normal domain. However, this expansion was not as strong as that in control morphants treated with the same concentration of RA (Fig. 7D) because of the reduced base-line levels of RA in rdhe2 morphants. Treat-ment with 0.01 M exogenous RA restored hoxd4 expression to approximately wild-type levels and boundaries. Thus, our finding that the RA-regulated hoxd4 was severely reduced in neurula stage rdhe2 morphants provided strong evidence that RA levels were significantly decreased in developing hindbrain, consistent with the function of rdhe2 in RA biosynthesis.
RA deficiency was reported to cause a failure in the anterior neural fold closure, leading to the exencephaly phenotype in mouse embryos (38,41). This is similar to the neural tube closure defects we observed in rdhe2 morphants. To assess the process of neural induction and neurulation in more detail, we examined the expression of sox3, a pan-neural marker, at the neural plate stages (stage 14). Induction of sox3 was normal in rdhe2 morphants, with a strong expression of sox3 detected in the neural plate. However, the domain of sox3 expression was widened compared with that in embryos injected with control MO (Fig. 7E), indicating that neurulation was delayed. It is interesting to note that in Raldh2 knock-out mice specification of neuroectoderm occurred normally in the absence of RA, but the later step of neurulation was affected by RA deficiency, leading to a seemingly widened pattern of Sox2 staining in the neural plate and a delay in the neural tube closure (16), reminiscent of the neurulation defects in rdhe2 morphants.
The presence of rdhe2 transcript in the notochord suggested that it may play a role in development of this tissue. To test whether this was the case, we analyzed the expression pattern of sonic hedgehog (shh) as the marker gene for the notochord. In situ hybridization with shh probe at the neurula stages revealed that shh was expressed in a shortened and less defined region in the notochord of rdhe2 morphants (Fig. 7F), demonstrating that rdhe2 indeed regulates the notochord development. The question of whether shh is an RA target gene is controversial because up-regulation of SHH expression in chick was observed under high pharmacological doses of exogenous RA (43), and its expression during limb patterning appeared to be independent of RA (44), whereas a recent report suggested that Shh is a direct target of RA signaling during genital tubercle formation in mice (45). However, SHH is thought to regulate rapid expansion of forebrain and midbrain vesicles, which collapse upon local ablation of SHH signaling in chick embryos (46). Thus, regardless of whether shh is directly regulated by RA, the decreased shh expression in rdhe2 morphants may contribute to the observed defects in anterior neural development.
RA signaling has been also shown to influence the organization of somites in Xenopus embryos (47). To determine whether somitic tissue was affected by the knockdown of rdhe2, we examined the expression of the myogenic protein myod. In rdhe2 morphants, myod was expressed at the levels similar to those in control embryos, but rdhe2 morphants clearly displayed partially irregular somites and disrupted somitic boundaries (Fig. 7G). Frog embryos treated with RAR antagonist or expressing a dominant negative form of RAR were shown to have smaller somites and sometimes unidentifiable somitic boundaries (47). Studies in Raldh2-, Rdh10-, and Rere-deficient mouse embryos have also implicated RA in the control of regularity and bilateral symmetry of somitogenesis (21, 48 -52).
Endogenous Rdhe2 Contributes to RA Biosynthesis in Vivo-Our gain-and loss-of-function assays and marker expression analyses all support the hypothesis that rdhe2 acts to promote RA production. To obtain further proof of rdhe2 function in RA biosynthesis, we employed a highly sensitive RA detection assay based on the activation of ␤-galactosidase reporter driven by a RARE promoter derived from human RAR␤ gene expressed in F9-RARE-lacZ cells (29). Our preliminary studies demonstrated that Sil15 cells grown in a monolayer responded to treatment with 0.5-50 nM RA by developing blue color in a dose-dependent manner. Importantly, treatment of Sil15 cells with retinaldehyde also resulted in blue staining, indicating that these cells were capable of oxidizing retinaldehyde, which would be produced by retinol dehydrogenases, to RA. Thus, Sil15 RA reporter cells represent a valid model for analyzing the activity of retinol dehydrogenases.
To compare the levels of retinaldehyde and RA in control and rdhe2 morphant embryos, equal numbers of stage 30 -32 embryos from both groups were dissected into head, trunk (dorsal and ventral), and tail segments. These tissues were incubated in cell culture inserts submerged into the medium over Sil15 cells monolayer. As can be seen in Fig. 7H, tissues from the head and dorsal trunk region of rdhe2 morphants produced a much weaker staining in Sil15 cell monolayer than equivalent tissues from control morphants. The observed results were reproduced in a series of three independent experiments. Interestingly, endodermal tissues removed from the ventral belly region did not activate the RA reporter, indicating that the stores of retinaldehyde present in the embryos were not released efficiently into the medium. Thus, both assays of RA levels, i.e. using the reporter cells and the endogenous RA sensitive hoxd4 gene, demonstrated that the knockdown of rdhe2 expression resulted in significantly reduced retinoid signaling, providing further evidence for the role of rdhe2 as a retinol dehydrogenase.
Frog Rdhe2 Has Two Catalytically Active Homologs in Mammals-As discussed in the Introduction, mammals have two genes in the position syntenic to frog rdhe2. Our previous studies have shown that the enzyme encoded by one of these genes, SDR16C5, recognizes all-trans-retinol as a substrate (22). To determine whether the product of the second homolog, SDR16C6 (RDHE2S), exhibits a retinol dehydrogenase activity, we obtained the corresponding mouse EST clone from Open Biosystems and subcloned the coding region of the cDNA into pCMV-Tag4a vector. The RDHE2S expression vector was transiently transfected into HEK293 cells, and the cells were incubated with all-trans-retinol. As shown in Fig. 8, the cells overexpressing RDHE2S produced significantly more RA than mock-transfected cells. This experiment demonstrated for the first time that the mouse RDHE2S can oxidize retinol to retinaldehyde in living cells. Thus, the ability to contribute to RA biosynthesis is conserved in both mammalian homologs of frog rdhe2, suggesting that, together, these two enzymes carry out the function of a single rdhe2 enzyme in frogs.
Despite sharing a relatively low (ϳ43%) protein sequence identity with rdh10, frog rdhe2 and mammalian rdhe2-related enzymes are the most similar proteins to rdh10 encoded by their respective genomes. Phylogenetic analysis places them into the same clade within SDR16C family, separately from other members (26, 53), thus confirming their common origin (supplemental Fig. 2 and supplemental Table 2). It is important to point out that, in terms of sequence identity, proteins from nonchordate taxa located at the basal positions of RDH10/ RDHE2-related clade of the tree, including those from sea urchin S. purpuratus and sea anemone N. vectensis, are more similar to RDHE2 than to RDH10. This indicates that RDHE2related enzymes are more likely to represent the ancestral, whereas RDH10, a later derived form. Further studies of RDHE2-related enzymes in various species may help to elucidate the origins of the RA biosynthetic pathway.

DISCUSSION
All-trans-retinol can be oxidized to all-trans-retinaldehyde by several members of the cytosolic medium-chain alcohol dehydrogenase family and by the endoplasmic reticulum membrane-associated members of the SDR family of proteins (54). Gene knock-out studies in mice have shown that some of these enzymes (alcohol dehydrogenases) play a role in detoxification of excessive dietary retinol (55) whereas others (RDH1) may be important for RA biosynthesis in certain adult tissues under vitamin A deficiency (56), but most of these candidate enzymes are not essential for RA biosynthesis during embryonic development under conditions of normal vitamin A supply. In contrast, RDH10 is the first retinol dehydrogenase found to be essential for the early stages of development in vitamin A-sufficient mice. Although RDH10 is clearly important, two independent studies have now shown that embryos that lack a functional RDH10 gene still retain RA signaling at several sites (19,21). The results of our study suggest that rdhe2-related SDRs might represent the missing RDHs responsible for this RA signaling. This conclusion is supported by several lines of evidence. First, X. laevis rdhe2, which represents the genomic equivalent of mammalian SDR16C5 and SDR16C6, exhibits an excellent activity toward all-trans-retinol, comparable with that of RDH10 when expressed in the same cell culture system (18). Thus, enzymatically, rdhe2 is a potent retinol dehydrogenase. Second, our gain-of-function experiments show that rdhe2 is capable of oxidizing retinol in vivo, conveying the phenotype consistent with overproduction of RA. Conversely, loss-offunction studies using MO-mediated knockdown of the endogenous rdhe2 enzyme reveal phenotypic defects and changes in the expression of marker genes consistent with the decrease in retinoid signaling in morphant tissues. Finally, two different assays of RA levels, i.e. using the reporter cells and the endogenous RA-sensitive hoxd4 gene, demonstrate that the knockdown of rdhe2 expression results in significantly reduced retinoid signaling. Together, these results establish that rdhe2 functions as a retinol dehydrogenase essential for embryonic development in frogs.
Expression pattern of rdhe2 overlaps with some of the reported domains of rdh10 expression (33), namely, forebrain, midbrain, otic vesicle, and eye. However, rdhe2 expression in the forebrain appears to be stronger than that of rdh10, especially in the olfactory bulb. In the eye, rdhe2 expression domains are broader than those of rdh10 and include both the neural retina and the lens. Our results also show that although frog rdhe2 is highly expressed in the brain at late tailbud and tadpole stages, it is absent from the developing head earlier. This dynamic expression pattern is consistent with the previous reports that the proper development of anterior structures in vertebrates requires exclusion of RA from the head and active repression by unliganded RARs at early stages (57,58), whereas later, RA synthesis occurs at several sites of the developing brain (59,60). Although rdh10 expression was reported in hindbrain-midbrain boundary, rdhe2 is expressed at high levels in the hindbrain. The requirement of RA signaling to establish segmental identity of the hindbrain, and observed changes in the RA-responsive posterior hindbrain marker hoxd4 in rdhe2 morphants suggest that rdhe2 activity may be important for hindbrain patterning. The most striking difference with rdh10 expression pattern is the high expression of rdhe2 throughout the notochord. Thus, in some areas of the developing embryo, such as the notochord and hindbrain, rdhe2 could be largely responsible for generation of retinaldehyde, except for the most posterior notochord where rdh10 also appears to be present.
Despite the well established role of RA in the patterning of posterior central nervous system (CNS) (for review, see Ref. 61), enzymes of the RA biosynthesis pathway appear to be largely absent from CNS (62). In mice, RALDH2 was not detected in the posterior neural tube, whereas RDH10 was detected only in the floor plate (19,63,64). These findings led to the paracrine model of RA action (65), which implies that RA is synthesized in the trunk mesoderm and diffuses from mesoderm to exert transcriptional control in CNS. The expression pattern of rdhe2 suggests that at least the first step of RA biosynthesis for CNS, the oxidation of retinol, may also occur in the notochord, which underlies the developing neural tube, serving as the source of retinaldehyde for a yet unidentified retinaldehyde dehydrogenase.
Frog rdhe2 has two homologs in mammals, one of which, RDHE2 (SDR16C5), is known to recognize all-trans-retinol as a substrate with NAD ϩ as the preferred cofactor, which suggests that it functions in the oxidative direction in the cells. As shown in this study, the second mammalian homolog, RDHE2S (SDR16C6), also exhibits an oxidative all-trans-retinol dehydrogenase activity when expressed in living cells. Compared with rdhe2, the mammalian homologs appear to have lower activities, but this could be due to the specific assay conditions that have not yet been perfected to reveal the full enzymatic potential of these enzymes. The fact that SDR16C5 is up-regulated in psoriasis suggests that, functionally, SDR16C5 bears enough physiological significance to be regulated at the transcriptional level. Future studies will reveal whether the two mammalian enzymes contribute to RA biosynthesis in overlapping or complementary pattern during development and in adult tissues and whether the level of their contribution is altered under certain pathophysiological conditions.