Aldehyde Dehydrogenase 6, a Cytosolic Retinaldehyde Dehydrogenase Prominently Expressed in Sensory Neuroepithelia during Development*

We have isolated the chick and mouse homologs of human aldehyde dehydrogenase 6 (ALDH6) that encode a third cytosolic retinaldehyde-specific aldehyde dehydrogenase. In both chick and mouse embryos, strong expression is observed in the sensory neuroepithelia of the head. In situ hybridization analysis in chick shows compartmentalized expression primarily in the ventral retina, olfactory epithelium, and otic vesicle; additional sites of expression include the isthmus, Rathke's pouch, posterior spinal cord interneurons, and developing limbs. Recombinant chick ALDH6 has aK 0.5 = 0.26 μm,V max = 48.4 nmol/min/mg and exhibits strong positive cooperativity (H = 1.9) toward all-trans-retinaldehyde; mouse ALDH6 has similar kinetic parameters. Expression constructs can confer 1000-fold increased sensitivity to retinoic acid receptor-dependent signaling from retinol in transient transfections experiments. The localization of ALDH6 to the developing sensory neuroepithelia of the eye, nose, and ear and discreet sites within the CNS suggests a role for RA signaling during primary neurogenesis at these sites.

Genes with promoters that contain retinoic acid response elements (RAREs) 1 are subject to regulation by the ligand-dependent nuclear transcription factors, the retinoic acid receptors RARs and RXRs. Disturbances in vitamin A signaling, either by vitamin A deficiency, through teratogenic excess of the receptor ligand retinoic acid (RA), or by retinoic acid receptor knockout studies (1)(2)(3), have shown that retinoid signals participate in vertebrate morphogenesis within specific temporal windows and target tissues. Affected tissues include the eye, craniofacial structures, heart, circulatory, urogenital, res-piratory system, limbs, and the anterior-posterior axis of the central nervous system (4 -8).
Precise control over the effective concentration of the receptor ligands, all-trans-and 9-cis-RA, is therefore a central requirement for proper receptor function and is maintained by the balance between synthesis and degradation. In retinoic acid-sensitive target tissues, ligands could be derived either by uptake from the low levels circulating in serum or through the in situ metabolism of the prohormone vitamin A (retinol) by alcohol dehydrogenases (including members of the short chain dehydrogenase/reductase and medium chain alcohol dehydrogenase families) and aldehyde dehydrogenases (reviewed in Refs. 9 and 10). Such an intracrine mechanism has the advantage that ligand synthesis and signaling can be tightly coupled through the cell-specific regulation of the respective enzymes, a property that would be desirable in various developmental models where RA is thought to act locally as a morphogenic signal. Degradation of RA proceeds primarily through further oxidative metabolism by members of the cytochrome P450 family of enzymes such as CYP26 (11).
In vertebrates, phylogenetic analysis indicates 13 families of aldehyde dehydrogenases that fall into two main clades. The "class 3" group, which consists mostly of substrate-specific dehydrogenases, and the "class 1/2" dehydrogenases that have broader substrate specificity (12). The "class 1/2" group contains members that can utilize retinaldehyde at submicromolar concentrations (13)(14)(15)(16). Recent reports confirm the essential role of ALDH1 and RALDH2 in RA signaling in vivo. For instance, premature expression of ALDH1 or RALDH2 by mRNA injection into Xenopus embryos results in the induction of premature RA synthesis and teratogenic effects (17), while the targeted disruption of RALDH2 in mice has shown its essential and specific role in RA signaling during axial rotation (body turning) and heart and limb morphogenesis (18).
Studies using retinoic acid-sensitive ␤-galactosidase reporter transgenic mice or indicator cell lines (19 -21) have demonstrated that RA signaling is restricted to specific tissues or regions during vertebrate embryogenesis and that many of these sites, but not all, co-localize to the expression pattern of ALDH1 and RALDH2. One notable discrepancy is the ventral neural retina, where evidence from in situ hybridization and zymographs of isoelectric focusing gels in both mouse and chick support the presence of an additional high activity retinaldehyde dehydrogenase that is distinct from ALDH1 or RALDH2 (22,23).
Here we describe the cloning and characterization of a third cytosolic aldehyde dehydrogenase, ALDH6, that can synthesize RA and is expressed in the sensory neuroepithelia including the ventral retina. We also show that ALDH6 is able to specif-ically transactivate the RAR-dependent signaling pathway when transfected into cells and can sensitize cells to retinol by shifting the dose-response curve 1000-fold. Comparisons with other ALDHs suggest that ALDH6 is equivalent in efficacy to RALDH2 and 10-fold better than ALDH1 in supporting RA-dependent signaling.
In Situ Hybridization for ALDH6 Expression-cDNA fragments encoding chick ALDH6 (nucleotides 433-1458 in sequence AF152358), chick ALDH1 (nucleotides 597-1178 (27)), mouse ALDH1 (AHD-2), and ALDH6 (nucleotides 433-1458 in sequence AF152359) were used to prepare digoxygenin-labeled antisense RNA probes. Whole mount in situ hybridization was performed as described by Wilkinson (28), and section in situ hybridization was performed as described by Ishii et al. (29), except that both hybridization and posthybridization washes were carried out at 69°C and RNase treatment was omitted. Chick embryos were embedded in O.C.T compound (Sakura Finetechnical Co., Ltd., Tokyo) for frozen sections. Mouse embryos and E4 -E7 chick embryos were embedded in paraffin. The sections were cut at 6 -12-m thickness.
Aldehyde Dehydrogenase Assays-Recombinant chick ALDH6 enzyme was prepared from the expression vector pBAD-cALDH6 in DH5␣ bacterial cultures induced with 0.2% arabinose. Purification from lysates was by sequential column chromatography with a 100 ϫ 2.5-cm Sephacryl S-300 gel filtration column (Amersham Pharmacia Biotech) developed in phosphate-buffered saline. Active fractions were dialyzed against 20 mM HEPES, pH 7.5, applied to a preparative 10 ϫ 2.5-cm Macroprep High Q column (Bio-Rad), and eluted with a gradient from 0 to 1 M NaCl in 60 min. Final purification was by a TSK-Gel DEAE-5PW (Tosoh, Tokyo, Japan) anion exchange column with a gradient of 0 -0.3 M NaCl in 30 min. Aldehyde dehydrogenase activity was monitored by NADH-dependent formazan dye formation at 566 nm. Briefly, 1-10 g of protein were assayed in 200 l of enzyme buffer (50 mM Tris-Cl, pH 8.5, 200 mM KCl, 250 M NAD, 250 M nitro blue tetrazolium, 8 M phenomethyl sulfonate, and 0.26% gelatin) with varying concentrations of substrate aldehydes. Retinaldehyde dehydrogenase activity was confirmed by HPLC assay as described below. Protein concentrations were assayed with the Bio-Rad Protein Kit, and purity was determined by SDS-polyacrylamide gel electrophoresis analysis.
Kinetic parameters for aldehyde substrates and NAD were determined by HPLC quantitation of retinoic acid formation or NADH production at 340 nm. Retinaldehyde dehydrogenase activity was assayed with 0.1-0.4 g of protein and 2 M all-trans-retinaldehyde in 1 ml of enzyme buffer (50 mM HEPES, pH 8.0, 200 mM KCl, 2 mM NAD, 1 mM MgCl 2 , and 1 mM dithiothreitol) at 37°C for 10 min. Retinoids were extracted by the addition of 250 l of acetonitrile/butanol (1:1) and 200 l of saturated potassium phosphate buffer according to the method of McLean (30), separated using a 5 ϫ 4.1-mm C18 reverse phase TSK-Gel SuperODS column (Tosoh, Tokyo, Japan), and quantitated by photodiode array detection. HPLC gradient conditions were as follows: flow rate, 1.5 ml/min; 100% 50 mM ammonium acetate, pH 6.9; linear change from 0 -60% acetonitrile in 3.0 min; isocratic at 60% acetonitrile for 3.0 min; linear change from 60 to 100% acetonitrile in 1.5 min; isocratic at 100% acetonitrile for 2.5 min.
Kinetic data were fitted using the nonlinear regression analysis program Prism (GraphPad Software, Inc., San Diego, CA).

Screening of Aldehyde Dehydrogenase PCR Amplicons from
Mouse and Chick Tissues-In the developing chick and mouse retina, ALDH1 is restricted to the dorsal neural retina, while RALDH2 is expressed in the pigmented retinal epithelium (31). However, an additional retinaldehyde dehydrogenase activity, named V1 in mouse and C-V in chick, that is biochemically distinct from ALDH1 and RALDH2 has been detected in the ventral neural retina (22,23). Since the small tissue sample size makes it impractical for classical protein purification, we isolated novel ALDH cDNA clones by designing degenerate oligonucleotide PCR primers to regions of high sequence conservation in an alignment of published human, mouse, and chick ALDH sequences for the nonallelic ALDH1/2/5/6 and RALDH2 genes. In addition to detecting the known aldehyde dehydrogenases, 4 of 26 clones from mouse and 5 of 120 clones from chick represented a new ALDH that by BLAST search of GenBank TM data base sequences shared closest homology with human ALDH6 (32) and the recently described murine RALDH3 cloned from the lateral ganglionic eminence (33). On the basis of nucleotide and amino acid similarity, we assign this new ALDH as the mouse and chick homolog of human ALDH6. Fig. 1, A and B, shows the mouse and chick nucleotide sequences and deduced open reading frame of the longest cDNAs obtained by PCR as described under "Experimental Procedures." The phylogenetic relationship between various aldehyde dehydrogenases is shown in Fig. 2 and Table I. Human, mouse, and chick ALDH6 sequences share 85-94% amino acid conservation, equivalent to that seen between species for sequences of ALDH1 (83%) or RALDH2 (96%). ALDH6 is only 69% con-served between either ALDH1 or RALDH2. The murine ALDH6 is essentially identical (99%) to RALDH-3 (Ref. 33; GenBank TM accession number AF253409). The ALDH6 open reading frame contains the NAD binding motif (GXGXXXG) at residues 235-241 and the conserved cysteine (Cys 314 ), which acts as the active site nucleophile (34). In addition, residues that are found in all catalytically active ALDHs (12) are present in equivalent positions in an optimal alignment, suggesting that this enzyme is functional.
ALDH6 Is Expressed in Sensory Neuroepithelia during Development-We performed in situ hybridization of chick and mouse embryos with antisense RNA probes of ALDH6. Figs. 3 and 4 show the expression pattern of ALDH6 in whole mount and sections in chick from stage 10 to 18. No signal was detected with sense probes. ALDH6 is first detected at stage 10 in the surface epithelium anterior to the optic vesicle but is absent from ectoderm directly overlying the lens vesicle (Figs. 3A and 4F). By contrast, the expression of ALDH1 is first detectable as the optic vesicle makes contact with the overlying ectoderm and the lens placode begins to thicken at stage 12. ALDH1 expression is initiated in the proximal layer of the vesicle directly opposite the lens placode (Fig. 4A). This region of strong ALDH1 expression involutes during optic cup formation and expands to form the dorsal neural retina (Fig. 4, A-E). In contrast, ALDH6 expression extends to both the proximal and distal ventral sides of the vesicle that are destined to become, respectively, the ventral halves of the neural and pigmented retina (Fig. 4, F-J). By embryonic day 4, ALDH1 and ALDH6 exhibit complementary nonoverlapping domains of expression in the dorsal and ventral neural retina (Fig. 4, E and J). Separating these two expression domains is a region that transects the dorso-ventral axis of the retina where neither ALDH1 nor ALDH6 is expressed. Interestingly, this stripe represents the expression domain of CYP26, a cytochrome P450 enzyme that is involved in RA breakdown (11,35,36).
In later stages of eye development, neural retina differentiates into a well organized layered structure in which various types of cells reside at specific positions. At E10, both ALDH1 and ALDH6 were found to be expressed in the outer side of the inner nuclear layer, where amacrine cells are differentiating (Figs. 5, A, B, E, and F) (37). Both transcripts were also found in the ganglion cell layer, albeit faintly. By E16, expression of both ALDH1 and ALDH6 have become faint (Fig. 5, C, D, G,  and H). Additional sites in the developing head include a band of expression in the isthmus joining the midbrain and hind brain and Rathke's pouch (Fig. 6, C and D). Although we could not detect expression during early limb bud formation (stages 16 -21), ALDH6 was observed later at E4 as diffuse staining in mesenchymal limb tissue and at E7 in the hind limb and fore limb interdigital zones and the perichondrial membranes of wing and leg buds (Fig. 6, A, B, and G). In the posterior trunk, ALDH6 staining was found in somites and a subset of spinal cord interneurons opposite the hind limb field (Fig. 6, G and H). This spinal cord expression co-localized with En-1, a specific marker for interneurons (38).
The localization of ALDH6 expression to the sensory neuroepithelia is conserved between chick and mouse (Fig. 7A), although some minor differences were observed. In mouse embryos, ALDH6 was not detected before E8.5 but became clearly visible by E9.5 in the dorsal and ventral margins of the optic vesicle and in a large area of overlying surface head ectoderm (Fig. 7B). The dynamic nature of the expression pattern is identical to that reported for RALDH3 (33). By E10.5, ALDH6 expression in surface ectoderm is rapidly reduced (Fig. 7, A and C) but is reexpressed at E11.5, where the ectoderm invaginates at the dorsal and ventral margins of the optic cup (Fig. 7D). During this period, expression in the ventral neural retina becomes more pronounced (Fig. 7, A, C, and D), and in contrast to the chick, some staining is also transiently seen in the dorsal neural retina through E11.5 (Fig. 7D). However, by E12.5 ALDH1 and ALDH6 are separated into distinct dorsal and ventral domains in the neural retina (Fig. 7, E and F). Expression of ALDH6 in retina weakens slightly by E15.5 (Fig. 7G).
The conservation of expression between mouse and chick ALDH6 extended to the olfactory placode and otic vesicle, which give rise to the sensory neuroepithelia of the nose and ear. Whole mount and section in situ hybridization of mouse embryos showed initially strong uniform expression in the developing olfactory epithelium at E10.5 (Fig. 7, A and C). Starting at E11.5, ALDH6 expression became progressively more restricted to the dorso-lateral neuroepithelium (Fig. 7D). By E15.5, transverse sections of olfactory structures exhibited a punctated cellular expression pattern in the sensory epithelium and the mesenchymal stroma directly underlying it that was restricted to discreet zones within the developing turbinates (Fig. 7H). Expression in the otic vesicle was transient and not detected after E10.5. In coronal sections of brain at E15.5, we also confirmed expression in the lateral ganglionic eminence as described in detail by Li et al. (33).
Mouse and Chick ALDH6 Can Utilize Retinaldehyde and Are Inhibited by Citral-In order to confirm that ALDH6 is indeed the retinaldehyde-specific dehydrogenase of ventral retina, we constructed inducible bacterial expression vectors containing the open reading frame. Conversion of all-trans-retinal to retinoic acid was only detected in lysates from bacterial cultures containing pBAD-ALDH6 and was dependent on induction of the protein (Fig. 8). Purified chick and mouse ALDH6 enzymes were used for determination of kinetic parameters and substrate specificity (Table II). Activity was found to be strongly stimulated with increasing pH between 7 and 8 but was constant above pH 8.0. Enzyme assays were therefore carried out at pH 8.0. ALDH6, like other dehydrogenases in the "class 1/2" clade, can utilize a variety of aldehyde substrates preferring aliphatic longer chain aldehydes; K m values for acetaldehyde, benzaldehyde, and octanal were 3.    8 (n ϭ 3). The submicromolar affinity for retinaldehyde and substantial V max support the conclusion that ALDH6 can function as a retinaldehyde-specific dehydrogenase in vivo.
We also determined the kinetic parameters for citral, which acts as a high affinity slow turnover substrate of aldehyde dehydrogenases that has been used as a selective competitive inhibitor of RA synthesis. By inference, tissues disturbed by citral treatment are considered to be sites of RA synthesis and signaling. The ventral retina appears to be particularly sensitive to disruption by citral (39). Inhibition curves with citral at constant retinal concentrations were determined and used to calculate K i values. Citral effectively inhibits RA synthesis of chick ALDH6 with a K i of 98 nM and V max of 4.2 nmol/mg/min (Table II). By comparison, rat RalDH(I) and RalDH(II) have reported K m values of 1 and 12 M (14), while human ALDH1 has a K m and V max of 4.2 M and 73.2 nmol/min/mg, respectively (40). Although RA synthesis from all three aldehyde dehydrogenases (ALDH1, RALDH2, and ALDH6) can be inhibited by citral, the lower affinity and higher turnover of citral by ALDH1 (40) suggest that it is less sensitive to disruption. The selective loss of the ventral retina by citral may therefore be a consequence of the intrinsic kinetic properties of ALDH1 and ALDH6.
Comparison of ALDHs for Activation of RA Signaling-Since aldehyde dehydrogenases exhibit broad substrate specificities, in vitro activity toward retinal may not necessarily imply functional significance in vivo. One consideration is the effective concentration of substrate encountered within the cell; the high affinity of retinal for CRBPs (41) suggests that free retinal is present only in nanomolar concentrations, although the high turnover of retinaldehydes during the visual cycle within the mature retina may represent a special case. We therefore tested chick ALDH6 for its ability to activate RA-dependent signaling under more physiologically relevant conditions. GAL-NR fusion constructs can be used to quantitate specific ligand-dependent signaling in transient transfections. Various GAL-NR constructs were therefore transfected either with or without chick ALDH6 expression plasmid into the cell line JEG-3. JEG-3 is a human choriocarcinoma cell line that does not synthesize RA from 10% fetal bovine serum or 10 nM retinol in serum-free medium and that, by Northern hybridization and PCR, does not express ALDH1, RALDH2, or ALDH6 (data not shown). RA-dependent transactivation through GAL-RAR could be detected either by exogenous addition of RA or in the presence of ALDH6 and a source of retinol (Fig. 10). The ability to support transactivation from retinol with ALDH6 indicates that aldehyde dehydrogenase activity is limiting in JEG-3 cells. Activation was specific to the GAL-RAR pathway, since no specific activation with ALDH6 was observed with other GAL-NR constructs: Gal-RXR, -VDR, -GR, or -PPAR. Identical results were observed with mouse ALDH6 (data not shown).
We also directly compared the efficacy of the various ALDHs to transactivate RA signaling under conditions of uniform promoter strength (Fig. 11). In transient transfections, the doseresponse curve for retinol was shifted 100-fold in the presence of chick ALDH1 and 1000-fold with chick RALDH2 or ALDH6. We also noted some increased GAL-RAR activity in ITLB medium (retinol-free) with ALDH6 or RALDH2 versus vector-only transfected cells (Figs. 10 and 11). This residual activity is likely to represent the mobilization and conversion into RA of the endogenous cellular pools of retinoids (i.e. retinyl esters or CRBP-bound retinol) that could not be effectively washed out during removal of serum-containing medium following transfection. Either longer incubation in retinol-free medium or an additional wash at a later time point could reduce this effect. Hence, in JEG-3 cells, RA signaling is dependent on the expression of a retinaldehyde dehydrogenase activity, and furthermore, the specific ALDH expressed can determine the level of RA signaling from an equivalent retinol source. DISCUSSION In this paper, we have described the chick and murine homologs of human ALDH6, which correspond to the ventral retina chick C-V and murine V1 (also recently cloned as RALDH3 (33) 43)). Although a direct comparison is difficult without further evaluation of chick ALDH1 and mouse ALDH1, our kinetic data for ALDH6 in general support the model that the ventral retina ALDH6 is likely to be the more efficient retinaldehyde dehydrogenase.
Further evidence in support of this conclusion is the differential ability of ALDH expression constructs to confer increased sensitivity to retinol for RAR-dependent signaling in transient transfections of JEG-3 cells. Both mouse and chick ALDH6 constructs were able to shift the dose-response curve of the RAR-signaling pathway 1000-fold with retinol and are equally effective as RALDH2; ALDH1 is 10-fold less effective than ALDH6 or RALDH2 but nevertheless able to elicit a robust signaling response within the physiological range of retinol concentrations. These data are consistent with the efficacy of ALDH1 and RALDH2 to synthesize moderate and high amounts of RA when mRNA is injected into Xenopus oocytes (17).
These data now raise the question of whether different ALDH enzymes are functionally redundant within the RA sig-naling pathway or whether they impart additional signaling information such as setting boundaries, thresholds, and/or gradients of RA. These concepts are most clearly visualized in the neural retina, where three distinct domains can be distinguished along the dorso-ventral axis on the basis of the expression pattern of ALDH1, CYP26, and ALDH6 that parallel functional RA-dependent signaling as observed in activation of RARE-␤-galactosidase reporter cells and animals (36,44). The juxtapositioning of opposing synthetic and degradative enzyme activities suggests that steep gradients in RA concentrations could be generated over short distances that would radically alter the transcriptional regulation of RA-sensitive target genes, essentially establishing sharp boundaries, but that within the dorsal and ventral ALDH domains the response is largely uniform.
Nevertheless, significant differences in kinetic properties of retinaldehyde dehydrogenases would also provide an intrinsically robust mechanism to quantitatively control RA synthesis and thereby regulate unique transcriptional responses encoded by threshold levels of RA. Measurement of RA synthetic ability in whole retina explants detects differences between the dorsal and ventral halves (36), indicating that although the dorsal and ventral halves of the retina are competent to synthesize the same active ligand, quantitative differences may define their positional identity; moderate levels equate with dorsal, and high levels represent ventral. In this manner, dorsal and ventral phenotypes can be discriminated and regulated within the same signaling pathway. Consistent with such a model are observations reported with zebrafish embryos, where elevated  (45). Whether in fact alternate retinaldehyde dehydrogenases elicit different RA signaling outputs in tissues, such as the neural retina, which are largely homogenous in structure and function, has not yet been experimentally addressed but provides a plausible mechanism of how to encode positional information or direct subtype cellular specialization within RA-sensitive tissues.
It is therefore intriguing to speculate if the prominent expression of ALDH6 in the developing neuroepithelia of the eye, nose, and ear indicates a potential selective requirement for early sensory neurogenesis pathways. In observations similar to those described above for the retina, a failure of localized RA synthesis and signaling in the invaginating olfactory placode and dorso-lateral olfactory epithelium observed in citraltreated and homozygous Pax6 Sey mutant mice coincides with severe broad disruption of the olfactory pathway (46). Later   9. Kinetics of RA synthesis and inhibition by citral. A, RA synthesis from all-trans-retinaldehyde by chick ALDH6. B, inhibition of RA synthesis by citral at 4 M retinaldehyde. Activity was assayed using 0.1-0.3 g of recombinant protein at 37°C, pH 8.0, as described under "Experimental Procedures." Data are representative of assays performed with 8 -12 substrate data points and quantitated by peak integration of HPLC chromatograms. Run-to-run quantitation errors were determined to be less than Ϯ3%.
during olfactory epithelium development, retinoid signaling becomes restricted and defines a subset of olfactory receptor neurons interspersed with nonresponsive cells (47). Reminiscent of these patterns are the changes seen in ALDH6 expression. Shifts in the anterior-posterior axis by either RA deficiency or excess also severely affect otic development (6,48,49).
Furthermore, retinoid sensitivity is seen in sensory pathways in the spinal cord. Elevated RAR/RXR expression and RA treatment increases (in contrast to citral treatment, which decreases) the number of Isl1ϩ primary sensory neurons in Xenopus tadpoles (50), and a sonic hedgehog independent retinoid-mediated pathway appears to control ventral spinal cord progenitor identity and promote interneuron diversity (including En1-and En2-positive neurons) in chick (51). Since ALDH6 localizes to a subset of ventral spinal cord interneurons but is distinct from the adjacent expression of RALDH2 present in the paraxial mesoderm and motor neurons arising from the most ventral lateral horn of the spinal cord (31), it may contribute to the synthesis of this local RA source and these specific neuronal subsets.
Of note also is the localization of chick ALDH6 to the isthmus and Rathke's pouch. The isthmus is a source of graded diffusible factors with organizer activity that specify the boundary of the mesencephalon from the segmented patterning of the rhombencephalon (52). While RA acts as a posteriorizing factor in determination of the anterior-posterior axis and misregulation in signaling alters specification of rhombomeres within the hind brain (8, 49, 53), RA synthesis from RALDH2 and liganddependent signaling extends only to the level of the first somite in developing chick and mouse embryos (54,55). The expression of chick ALDH6 in a sharp, narrow line of demarcation across the anterior margin of the isthmus suggests that RA signaling may also influence a boundary rostrally of the hind brain. Chick ALDH6 is also a specific marker of Rathke's pouch and defines the oral ectoderm in the region where it contacts the neural ectoderm of the diencephalon that together are destined to fuse and form, respectively, the glandular and neural portions of the pituitary. Localized RA synthesis by ALDH6 is likely to affect aspects of pituitary development and has the potential to impact the hypothalamic-pituitary-gonadal axis, through the synergistic action between retinoid receptor signaling and pituitary-specific transcription factors such as Pit-1 (56,57).
In conclusion, we have shown that ALDH6 is strongly expressed in the ventral retina, is a functional NAD-dependent aldehyde dehydrogenase, has high activity toward all-transretinaldehyde, and is sensitive to citral inhibition. Hence, the expression pattern and biochemical characteristics fit well with the mouse V1 (RALDH3) and chick C-V activities. The discreet localization of ALDH6 highlights additional domains of RA synthesis during embryogenesis that are spatially distinct from sites of RALDH2 and ALDH1 expression and suggests a role in the development of a restricted set of neural tissues.  11. RA-dependent transactivation by chick aldehyde dehydrogenases. JEG-3 cells were plated at 20 -40% confluency, transfected with 600 ng of DNA/well RAREϫ2-tk-Luc plus pCMX-␤Gal plus pCAGGS vector or chick ALDH constructs (1:1:1) for 8 h, washed twice with phosphate-buffered saline, and incubated for 24 h in ITLB/Dulbecco's modified Eagle's medium with the indicated concentrations of retinoids. Dose-response curves are as follows. Open squares, pCAGGS vector with retinol (ROL); closed squares, pCAGGS vector with alltrans-RA; closed circles, ALDH1 with retinol; open circles, RALDH2 with retinol; closed triangles, ALDH6 with retinol. Dose-response curves for RA with ALDHs are omitted for clarity but were essentially the same as the response observed with pCAGGS vector. Data points represent the means of triplicates; the S.E. was less than 15% for all points. R.L.U., relative luciferase units.