CYP26A1 and CYP26C1 Cooperate in Degrading Retinoic Acid Within the Equatorial Retina During Later Eye Development

View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Elsevier - Publisher Connector

Developmental Biology 276 (2004) 143–157 www.elsevier.com/locate/ydbio

CYP26A1 and CYP26C1 cooperate in degrading retinoic acid within the equatorial retina during later eye development

Yasuo Sakaia, Tuanlian Luoa, Peter McCafferya, Hiroshi Hamadab, Ursula C. Dr7gera,c,*

aEunice Kennedy Shriver Center at the University of Massachusetts Medical School, Waltham, MA, United States bDevelopmental Genetics Group, Graduate School of Frontier Biosciences, Osaka University, and CREST, Japan Science and Technology Corporation (JST), 1-3 Yamada-oka, Suita, Osaka 565-0871, Japan cDepartment of Psychiatry, Harvard Medical School, Boston, MA, United States

Received for publication 9 June 2004, revised 2 August 2004, accepted 20 August 2004 Available online 16 September 2004

Abstract

In the embryonic mouse retina, retinoic acid (RA) is unevenly distributed along the dorsoventral axis: RA-rich zones in dorsal and ventral retina are separated by a horizontal RA-poor stripe that contains the RA-inactivating enzyme CYP26A1. To explore the developmental role of this arrangement, we studied formation of the retina and its projections in Cyp26a1 null-mutant mice. Expression of several dorsoventral markers was not affected, indicating that CYP26A1 is not required for establishing the dorsoventral retina axis. Analysis of the mutation on a RA-reporter mouse background confirmed, as expected, that the RA-poor stripe was missing in the retina and its projections at the time when the optic axons first grow over the diencephalon. A day later, however, a gap appeared both in retina and retinofugal projections. As explanation, we found that CYP26C1, another RA-degrading enzyme, had emerged centrally in a narrower domain within the RA-poor stripe. While RA applications increased retinal Cyp26a1 expression, they slightly reduced Cyp26c1. These observations indicate that the two enzymes function independently. The safeguard of the RA-poor stripe by two distinct enzymes during later development points to a role in maturation of a significant functional feature like an area of higher visual acuity that develops at its location. D 2004 Elsevier Inc. All rights reserved.

Keywords: Developing mouse retina; Dorsoventral axis; Temporal retina; CYP26A1; CYP26B1; CYP26C1; Retinoic acid homeostasis; Visual streak; Anisotropy; Moon illusion

Introduction in the adult (Saari, 1994). RA transduces its signaling by binding to ligand-activated nuclear transcription factors, Retinoic acid (RA), the major active derivative of vi- the RARs and RXRs (Chambon, 1996; Sucov and Evans, tamin A, plays critical roles in growth, differentiation, 1995). When RA is present, expression of its target genes morphogenesis, and homeostasis (Mark and Chambon, is regulated by different combinations of RAR/RXR 2003). The eye stands out in retinoid dependency: ocular heterodimers (Kastner et al., 1997). In addition to the malformations are the most sensitive indicators of vitamin presence of RA, also its absence can be critical for the A deficiency during development (Hale, 1937), and the regulation of gene expression by the RA receptors. For use of retinaldehyde as visual chromophore makes the instance, a repressive function of unliganded RARs is retina the organ with by far highest retinoid concentration essential for proper head formation in early embryos (Koide et al., 2001; Weston et al., 2003). The local levels of RA are determined by two types of enzymes, the RA- synthesizing retinaldehyde dehydrogenases (RALDHs) and * Corresponding author. Eunice Kennedy Shriver Center at the University of Massachusetts Medical School, 200 Trapelo Road, Waltham, RA-degrading members of the cytochrome P450-linked MA 02452. Fax: +1 781 642 0017. oxidase family, the CYP26 enzymes (Maden, 2002). In E-mail address: [email protected] (U.C. Dr7ger). humans and mice, four RALDHs and three CYP26s have

0012-1606/$ - see front matter D 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.ydbio.2004.08.032 144 Y. Sakai et al. / Developmental Biology 276 (2004) 143–157 been identified, RALDH1–4 (Grqn et al., 2000; Lee et al., the embryo, but the coordinates for its morphogenesis 1991; Li et al., 2000; Lin et al., 2003; Suzuki et al., 2000; predict the light distribution falling onto the retina in the Zhao et al., 1996) and CYP26A1, B1, and C1 (Abu-Abed functional eye (Wagner et al., 2000). In contrast to the et al., 2002; Fujii et al., 1997; MacLean et al., 2001; Ray ancient mechanisms for dorsoventral eye formation, the et al., 1997; Tahayato et al., 2003; Taimi et al., 2004; use of RA as transcriptional regulator emerged only in White et al., 1996). As RA can diffuse freely across solid chordates/vertebrates, and it often functions as a new layer tissues, the local expression sites of the metabolic of regulation superimposed over older processes (Escriva enzymes represent potential spatial patterning tools: the et al., 2000). RALDHs form the origins of RA diffusion gradients and This discrepancy in evolutionary age between dorso- the CYP26s generate abrupt steps in RA levels. ventral eye formation and transcriptional use of RA is RA synthesis in the future eye region of the mouse probably related to variations between vertebrates in the begins with neurulation at embryonic day 8 (E8), when extent to which RA is able to influence formation of the Raldh2 is transiently expressed at the anterior neural ridge, dorsoventral retina. In zebrafish, perturbations in RA levels an organizing center for both eye and forebrain formation have the most pronounced effect: RA deficiency abolishes (Wagner et al., 2000; Wilson and Rubenstein, 2000). At the ventral retina (Marsh-Armstrong et al., 1994), and RA E8.5, Raldh3 expression starts in the surface epidermis over excess duplicates it at the expense of the dorsal retina the optic recess, and a day later, it appears in the ventral (Hyatt et al., 1996). In mice, the effects are milder: genetic neural retina, the dorsal retinal pigment epithelium, and the null-mutants of the dorsal Raldh1 enzyme have no lens (Gru¨n et al., 2000; Li et al., 2000; Suzuki et al., 2000). detectable eye defect (Fan et al., 2003), and null-mutants At E9, Raldh1 and Cyp26a1 begin to be expressed in of the ventral Raldh3 only cause reductions in the size of overlapping regions of the dorsal eye vesicle (Wagner et al., the ventral retina (Dupe et al., 2003). A similar reduction 2000). Raldh1 expression then remains in the dorsal retina, in ventral-retina size is also one of the effects of vitamin A whereas Cyp26a1 moves downwards to form a horizontal deficiency in rodents (Warkany and Schraffenberger, stripe across the neural retina underneath RALDH1. By 1946), and it can be caused by genetic disruptions of E11.5, the embryonic retina is subdivided into three zones several RA receptors (Mark and Chambon, 2003). In the of different RA levels along the dorsoventral axis: mode- chick, expression of the metabolic enzymes in the eye rately high RA levels generated dorsally by RALDH1 and region begins later than in the mouse, and the shape of its very high RA levels made ventrally by RALDH3 are eye is relatively impervious to experimental RA perturba- separated by a horizontal RA-poor stripe that is created by tions (Mey et al., 2001). CYP26A1 activity (McCaffery et al., 1999). These three Here, we used Cyp26a1 null-mutants to ask whether the compartments of different RA levels are directly visible in RA-poor stripe generated by CYP26A1 is required for RA-reporter mice that are transgenic for a construct morphogenesis of the embryonic retina and its projections containing lacZ driven by sensitive RA response elements to the brain in mice. While CYP26A1 was previously the (Rossant et al., 1991): in these mice, the ganglion-cell axons only RA-degrading enzyme identified in the neural retina from the dorsal and ventral lacZ-positive retina regions are of the mouse, we found that its lack abolished the RA-poor intensely labeled all along their projections to the optic stripe only transiently during early formation of the targets in the brain (Wagner et al., 2000). retinofugal projections. A little later, the stripe became The zones of RA synthesis in the eye of embryonic partially visible, but it was now generated by CYP26C1, vertebrates match with other molecular markers of the another RA-degrading enzyme. Thus, both CYP26A1 and dorsoventral retinal axis; for example, the dorsal tran- CYP26C1 cooperate to establish a RA-poor stripe during scription factor Tbx5 overlaps with Raldh1 expression and later retina development, which makes it likely that it is the ventral transcription factor Vax2 with Raldh3 (Peters, necessary for formation of a functionally important feature, 2002). Whereas the partition of the developing eye into such as a retinal region specialized for higher acuity dorsal and ventral compartments is very ancient, however, vision. an involvement of RA in this pattern is much more recent: although the genome of Drosophila does not contain RA receptors, the eye disc in embryonic flies is subdivided Materials and methods into dorsal, equatorial, and ventral zones by multiple patterning factors (Brodsky and Steller, 1996). In adult Mice flies, this embryonic pattern gives rise to the morphologically distinct dorsal and ventral divisions of the compound The production of the Cyp26a1- and Cyp26b1-null mice eye, which are separated by a horizontally oriented border and the polymerase chain reaction (PCR) strategies for region. Similarly in mice, the coordinates of the embryonic detection of their genotypes were described previously eye are already the same as those used for functional (Sakai et al., 2001; Yashiro et al., 2004). The breeding vision; that is, the eye is not formed in the segmental colony of the RARE-hsplacZ reporter mice was a gift from J. scheme in which the rest of the nervous system is set up in Rossant (Rossant et al., 1991). The embryos were staged Y. Sakai et al. / Developmental Biology 276 (2004) 143–157 145 following Theiler’s (1972) criteria. All experiments were Results conducted following procedures approved by the recom- mendations of the Panel on Euthanasia of the American Dorsoventral markers Veterinary Medical Association. In the normal embryonic retina, Cyp26a1 occupies a In situ hybridization horizontal stripe region that abuts the dorsal and ventral RALDH territories (McCaffery et al., 1999). To test whether In situ hybridizations on whole mounts and sections were this arrangement signals a flexible state that is responsive to performed according to previously published protocols local RA levels, we examined the expression patterns of (Henrique et al., 1995; McCaffery et al., 1999). The probe several dorsoventral marker genes in Cyp26a1 null-mutants. for Vax2 (Schulte et al., 1999) was provided by T. Furukawa The transcription factor T-box5 (Tbx5) is an early marker for and the one for Tbx5 (Koshiba-Takeuchi et al., 2000)byT. the dorsal retina and the ventral anterior homeobox gene Ogura. Probes for Raldh1, Raldh3, Cyp26a1, and Cyp26b1 (Vax2) for the ventral retina in mice and chicken (Koshiba- were prepared as described (Fujii et al., 1997; Li et al., Takeuchi et al., 2000; Schulte et al., 1999). In situ 2000; Yashiro et al., 2004). A probe for Cyp26c1 was hybridizations on whole mounts and sections of mutant generated to the nucleotide residues 360–855 of the mouse and wild-type embryos revealed no differences in the Cyp26c1 cDNA. expression pattern of either Tbx5 or Vax2 at E9.5 to E14.5, the ages tested. In the double-labeled preparations RA administration shown here (Fig. 1A), the two markers are separated by a gap that is the same in normal and mutant embryos. For the RA applications, aliquots from a stock solution of Similarly, the expressions of Raldh1 and Raldh3 are not 0.1 M all-trans RA (Sigma) in dimethyl sulfoxide (DMSO) affected in the Cyp26a1À/À retina (Fig. 1B). These results were emulsified in tissue culture medium containing 10% indicate that generation of the RA-poor stripe is not essential fetal calf serum and injected intraperitoneally (Luo et al., for patterning the retina along its dorsoventral axis. 2004). RA doses were adjusted to the weight of the mouse À/À at 30 mg/kg for pups, a dose that is close to the LD50 at this Altered expression of a RA-reporter in Cyp26a1 mice: age (Luo et al., in preparation), and 60 mg/kg in pregnant local and systemic effects mice, a high yet not toxic dose for adults. RA reporter animals, including the RARE-hsplacZ mice b-Galactosidase staining made by Rossant et al. (1991), are widely used as indicators of endogenous RA signaling. Their lacZ expression patterns For whole-mount staining of RA-reporter tissues, the reflect combinations of local RA levels and of propensities to dissected samples were fixed in 0.2% glutaraldehyde and respond to RA. For instance, in the RALDH-rich retina, the processed following Rossant et al.’s (1991) protocol. For lacZ patterns signal mainly the RA distribution, whereas in sections, the embryos were briefly perfused transcardially brain tissue that receives most of its RA by diffusion the lacZ with 2% glutaraldehyde. The heads or brains were rapidly labeling marks primarily RA-responsive sites (Luo et al., embedded in 4% agarose, sliced at 150 Am on a vibratome, 2004). To visualize directly the effects of eliminating and the slices were processed with 5-bromo-4-chloro-3- CYP26A1 activity, we crossed the RARE-hsplacZ transgene indolyl-h-d-galactopyranoside (X-gal) for lacZ visualiza- (Rossant et al., 1991) onto the Cyp26a1 mutant stock. The tion (Luo et al., 2004). heads or brains of perfusion-fixed embryos were serially sliced on a vibratome at 150 Am and reacted with X-gal for DiI labeling lacZ expression. The coronal section (Fig. 2A) through the eyes of an E15 homozygous Cyp26a1À/À mutant embryo The optic projections were traced in E15.5–18.5 shows that the gap between dorsal and ventral retina is embryos with DiI (1,1-dioctadecyl-3,3,3V,3V-tetramethylin- missing, while it is readily visible in the slice through eyes of docarbocyanine perchlorate; Molecular Probes), following a normal embryo (Fig. 2B). a protocol provided by Pierre Godement (Godement et In addition to this local effect in the retina where al., 1987). All embryos in litters from heterozygous CYP26A1 is normally present, the Cyp26a1À/À mutation Cyp26a1+/À matings were perfused transcardially with exerted a widespread intensification of lacZ staining at 4% paraformaldehyde in phosphate-buffered saline (PBS) many sites, at most of which CYP26A1 is never expressed; and stored for one to several days in the same fixative. this was immediately obvious when all slices of mutant Then one eye was carefully packed with homogenized and normal heads were viewed side by side. In the DiI crystals, and the embryos were stored in PBS at illustrated rostral sections (Figs. 2A, B), the cornea, room temperature. After 6–8 weeks, the optic pathways conjunctiva, and frontonasal region are abnormally heavily were dissected and photographed with epifluorescence labeled in the mutant compared to the normal embryo. In optics on a Zeiss Axioplan. the brain, the general intensification was most pronounced 146 Y. Sakai et al. / Developmental Biology 276 (2004) 143–157

Fig. 1. (A) In situ hybridizations on whole heads of Cyp26a1À/À embryos and normal littermates labeled with two probes simultaneously, one to the dorsal- retina marker Tbx5 and the other to the ventral marker Vax2. (B) Paraffin-sectioned heads double-labeled for dorsal Raldh1 and ventral Raldh3 mRNA. Here, as in all following figures, dorsal is up and ventral down. Note that the spatial layout of the dorsoventral retina markers in the mutants is similar to the normal patterns. E10.5: embryonic day 10.5. in the occipital cortex, although CYP26A1 is not normally prepared whole mounts of the brains and corresponding expressed here or anywhere else in the forebrain. The retinas separately for each embryo in litters from hetero- comparison of slices through mutant (Fig. 2C) and normal zygous Cyp26a1+/À matings, and we determined the brains (Figs. 2E, F) shows an increase in overall lacZ individual genotypes by PCR. From the brains, the labeling, which is conspicuous even over the normally telencephala and hindbrains were removed, and the somewhat variable lacZ labeling of the cortex (Luo et al., remaining di- and mesencephala were flattened by a cut 2004). The lacZ intensification in the Cyp26a1À/À mutants along the ventral midline (Luo et al., 2004). In the resembles delayed effects in the cortex to earlier RA preparations illustrated (Figs. 3A–C), the optic tract is injections (Fig. 2D): when embryos are exposed to RA visible as it projects in form of a tight bundle from the excess early in telencephalon development, the RA- optic chiasm (ox, left in the figures) across the hypothala- reporter expression in the cerebral cortex can undergo a mic surface of the diencephalon. Beginning at the rostral lasting change that appears like a sensitization to endog- border of the lateral geniculate nucleus (lgn), the optic enous RA levels and that is similar to changes observed in axons then splay out to cover the central retinotopic maps, the di- and mesencephalon after specific injection times in which the ventral retina (V) is represented dorsomedially (Luo et al., 2004). Although we do not know the functional in the geniculate and the dorsal retina (D) ventrolaterally. significance of this phenomenon, we have observed it in In the wild-type embryos (+/+, Fig. 3), the axon the cortex only following RA injections around E10–11 but bundles from dorsal and ventral retina are separated by not after earlier or later applications. Here, we take this an unlabeled gap that we assume contains unlabeled observation as indicator for generalized and chronic RA axons from the CYP26A1 stripe in the retina. In the poisoning of the Cyp26a1À/À mutants in addition to the preparations of the E15 litter shown, absence of all local consequences of the mutation. Although Cyp26a1 is CYP26A1 activity in the (À/À) mutants has completely expressed only at restricted locations in the embryo, eliminated the gap in the retinofugal projections (Fig. including the retina, hindbrain and tail, elimination of the 3A), as well as the RA-free stripe in the retinas (Fig. gene thus also influences the circulating RA levels and the 3D); in the heterozygous E15 Cyp26a1+/À mutants, the general retinoid homeostasis. RA-free stripe in the retinas is partially obliterated, and the gap in the diencephalic optic projection is partly The RA reporter indicates CYP26A1 dosage and a residual filled in. Surprisingly, in the homozygous Cyp26a1À/À gap mutant embryo of the E17.5 litter (Fig. 3B), the optic projections are separated by a partial gap in the To evaluate the effects of altered CYP26A1 levels on diencephalon (see red arrow), which is less pronounced the retinas and their central projections in entirety, we than in the heterozygous projections but still clearly Y. Sakai et al. / Developmental Biology 276 (2004) 143–157 147

Fig. 2. X-gal reacted slices through embryonic heads to illustrate the effects of the Cyp26a1À/À mutation on RA-reporter expression, as described in the text. Note that the lacZ labeling of the mutant retinas (A) is continuous, whereas in normal retinas it is interrupted by a gap (B). In addition to the effects on the neural retina, the Cyp26a1À/À mutation results in an intensification of lacZ expression at selected locations including the cornea (corn.) and conjunctiva (conj.) of the eye, the frontonasal region (A), and the occipital cortex (C). Systemic RA injections cause acutely a similar increase in lacZ expression in the retina (Luo et al., 2004) and chronically in the cortex (compare D with E, F). The embryo in D had been exposed to 30 mg/kg all-trans RA at E10.5, and it was perfused at E16.5. discernible. The same phenomenon is apparent in the Comparisons of Cyp26a1 expression by in situ hybridization E18.5 litter (Fig. 3C), for which also several retinas are (Fig. 3F) reveal lower mRNA amounts in heterozygous illustrated (Fig. 3E): in the homozygous Cyp26a1À/À Cyp26a1+/À mutant than in wild-type retinas, an observation mutant embryos, a region of reduced lacZ expression is that explains the heterozygous effect on the RA-reporter visible both in the retinas and optic projections (red arrows), expression. In the homozygous Cyp26a1À/À mutant retinas, and in the heterozygous (+/À) mutants, the retina stripe and no Cyp26a1 mRNA was detectable (not shown), leaving the central gap are narrower compared to the wild-type embryo. faint gap in lacZ expression unexplained. 148 Y. Sakai et al. / Developmental Biology 276 (2004) 143–157

Fig. 3. Flat-mounted di- and mesencephala (A–C) as well as retina cups (D–E) to illustrate the effects of homozygous (À/À) and heterozygous (+/À) Cyp26a1 mutations on RA-reporter expression in the primary optic projections at three ages. The Cyp26a1 genotypes of all embryos were determined by PCR. The in situ hybridizations for Cyp26a1 on heterozygous (+/À) and wild-type (+/+) retinas (F) show a reduction in Cyp26a1 dosage in the heterozygotes. This dosage effect is directly visible as differences in lacZ expression between heterozygous and wild-type retinas and brains. Note that, at the two older ages, a partial gap appears in the retinas and retinofugal projections of homozygous Cyp26a1À/À mutants (red arrows). Abbreviations: ox, optic chiasm; ot, optic tract; lgn, lateral geniculate nucleus; V ret, from ventral retina; D ret, from dorsal retina; di, diencephalon; mes, mesencephalon.

Tracings of retinofugal projections with DiI merely a consequence of missing reporter expression rather than a true gap between the projections from dorsal and As we began to doubt our previous assumption that the ventral retina, we decided to test this question directly. All gap in the retinofugal projections of RA-reporter mice is embryos in litters of heterozygous matings were perfused Y. Sakai et al. / Developmental Biology 276 (2004) 143–157 149 transcardially for optimal fixation, and one eye was packed appear grossly altered by the lack in CYP26A1 activity, with the dye DiI, which equilibrates by diffusion within as far as is visible by simple inspections of the whole- continuous lipophilic compartments such as axonal mem- mount preparations. For later embryonic ages, we tested branes, if given enough time (Godement et al., 1987; head preparations as well as isolated retinas (Figs. 5C– Lukas et al., 1998). At all ages tested up to E18.5, no gap H). For Cyp26b1, which we had expected to find, no was detectable within the retinofugal projections both in trace was detectable at any age in the neural retinas of normal and mutant embryos, as illustrated here (Fig. 4) for normal embryos or of Cyp26a1À/À mutants; it is, a normal E15 and a mutant E16.5 embryo. When scruti- however, expressed in the retinal pigment epithelium nized at higher magnifications, the detailed optic-fiber (Abu-Abed et al., 2002). As lack of Cyp26b1 is difficult trajectories in the mutants were indistinguishable from to illustrate convincingly in isolated retinas, we show normal littermates. This makes it likely that the central (Fig. 5C) the unlabeled eyes in head-whole mounts next visual maps in the mutant do not have any gross to positive controls, the emerging whisker follicles. In abnormality, as the optic axons travel largely parallel and screens of Cyp26b1À/À null mutants with the same in a topographically ordered arrangement. techniques as described here for Cyp26a1À/À embryos, no changes were apparent in the staining patterns of the Search for another RA-degrading enzyme in the retina RARE-hsplacZ transgene (data not shown). Similar to Cyp26b1,noCyp26c1 was detectable in the The DiI labelings indicated unambiguously that the eyes at E11.5 (Fig. 5D). Isolated older retinas, however, gap in lacZ expression in the older homozygous revealed traces beginning around E14. By E15, Cyp26c1 Cyp26a1À/À mutants must be due to RA degradation in expression had become sufficiently strong to trace a the retina by a mechanism different from CYP26A1 horizontal line across the retina at a similar position as activity. Two additional RA-degrading enzymes are the wider Cyp26a1 stripe (Fig. 5E). In the E16 retinal known, CYP26B1, which was originally identified as whole mounts that are shown both from back and front an expressed sequence tag (EST) in an adult human (Fig. 5F), Cyp26c1 forms a substantial but narrower stripe. retina library (White et al., 2000), and CYP26C1, which In newborn retinas double-labeled for the RA reporter and had not been detected in the eye (Tahayato et al., 2003). the RA-degrading enzymes, Cyp26c1 occupies the center We asked whether expression of these two enzymes is of the lacZ-free stripe (Fig. 5G), and Cyp26a1 more or less affected in the homozygous Cyp26a1À/À mutant embryos, fills it (Fig. 5H). and we searched for expression in the normal eye region throughout development. The whole embryos at E9.5– Expression of Cyp26a1 and Cyp26c1 in postnatal retinas 10.5 (Figs. 5A,B) show no Cyp26b1 or Cyp26c1 expression in the eye regions of the Cyp26a1À/À mutants, The retinal whole mounts showed that expression of like in the wild-type embryos. Outside of the eye, the both Cyp26a1 and Cyp26c1 continues for several days normal expression patterns of the two enzymes do not postnatally, but detection became increasingly difficult in

Fig. 4. DiI-labeled optic projections in normal embryos and a Cyp26a1À/À mutant. The telencephala were removed, and rostral is to the left; the general layout of the labeled axons in the brain is shown in the low-magnification insert of a normal E16.5 brain. Only the thick fiber bundle consists of optic axons; the two weaker bundles marked by asterisk are retrogradely labeled motor axons projecting to the eye. Abbreviations: ox, optic chiasm; med, medulla. 150 Y. Sakai et al. / Developmental Biology 276 (2004) 143–157

Fig. 5. Different preparations to illustrate the absence of Cyp26b1 expression in embryonic retinas (A, C), the expression of Cyp26c1 during later developmental stages (B, D–G), and comparisons of Cyp26c1 and Cyp26a1 expression (E, G, H). All preparations are in situ hybridizations which, in G and H, were done following X-gal reactions for the RA-reporter transgene. In F–H, the same retina cups are shown both from back and front. P0.5: postnatal day 0.5. the thick preparations. For this reason, we analyzed enzymes decrease gradually over the first two postnatal section of retinas from normal RA-reporter mice at weeks (Fig. 6A). The Cyp26a1 signal diminished evenly successive ages. The retinas were briefly reacted with across the nasotemporal dimension, whereas Cyp26c1 X-gal, aligned in the same dorsoventral and nasotemporal disappeared first in the nasal retina and persisted longest orientation into one block (Fig. 6B), sectioned on a in the peripheral temporal retina (not illustrated). The cryostat, and reacted in parallel to assure optimal labeled cell bodies become confined to the inner nuclear comparability of the in situ hybridization signals. Such layer, as shown for postnatal day 6 (P6) (Fig. 6C). After preparations showed that the signal intensities of both about P14, both enzymes became undetectable. Y. Sakai et al. / Developmental Biology 276 (2004) 143–157 151

Fig. 6. Sections of normal retinas from RA-reporter mice of different postnatal ages, cut on a cryostat. The retina cups were aligned with help of the lacZ patterns, stacked together into two blocks, and processed in parallel to assure that the in situ hybridization signals are comparable. The arrows (B) point to the expression sites of the two Cyp26 enzymes at low magnifications. (A) Enlarged retina segments shown in the same temporal sequence; in the Nomarski view (C), the retinal layers are indicated. Note that expression of both Cyp26a1 and Cyp26c1 decreases with age and condenses in cell bodies in the inner nuclear layer (inl); onl, outer nuclear layer; opl, outer plexiform layer; ipl, inner plexiform layer; gcl, ganglion-cell layer; P14, postnatal day 14.

Responses of Cyp26a1 and Cyp26c1 to RA Cyp26c1, we first compared the normal retinas in heterozygous litters with those of Cyp26a1 mutants, in which the As the patterns of Cyp26a1 and Cyp26c1 are similar RA levels are increased. In the homozygous mutant (À/À) in the retina and as the two genes are located very close retina of the E18.5 litter (Fig. 7A), Cyp26c1 expression is together on chromosome 19 (see Discussion), we won- weaker, as compared to the wild-type (+/+) and hetero- dered whether expression of the two enzymes is regulated in zygous (+/À) retinas. A second pregnant Cyp26a1+/À a similar way. Cyp26a1 was originally identified as a gene mouse was injected with a very high RA dose (60 mg/kg) whose transcription is increased by RA (Ray et al., 1997; 6 h before collection of her embryos at E18.5, and the White et al., 1996). To test whether this applies also to retinas were processed in parallel with those from the not- 152 Y. Sakai et al. / Developmental Biology 276 (2004) 143–157

Fig. 7. Late embryonic and newborn retina cups to illustrate the reactions of Cyp26a1 and Cyp26c1 to altered RA levels. Each vertical panel represents one experiment; that is, the single retinas in each block are representative of several retinas per batch in a series of different conditions, which were reacted in parallel to assure perfect comparability of the in situ hybridization signals; a quantitative comparison of signal strength is only possible along the same vertical block but not between blocks. (A) Retinas from two E18.5 litters of Cyp26a+/À matings. One of the pregnant mice was injected with 60 mg/kg all-trans RA 6 h before collecting the embryos. The Cyp26a1 genotypes of all embryos were determined by PCR. (B) Retinas of normal newborn mice collected at different times following injections of 30 mg/kg all-trans RA. Note that RA excess causes a dramatic increase in Cyp26a1 expression and a minor reduction in Cyp26c1 levels. Cyp26c1 expression is stronger in the temporal (T) than nasal retinas (N). injected heterozygous litter. The line-up of retinas from both down-regulated in a slower time course, first visible here litters (Fig. 7A) shows that the excessive RA dose causes a in the 4-h sample. The alignment of all retinas from one very slight further down-regulation in Cyp26c1 expression side illustrates that the Cyp26c1 intensity is higher in the compared to the genetic reduction in CYP26A1. The temporal (T) than in nasal (N) retinas. This asymmetry is intensity of Cyp26a1, by comparison, is drastically in- consistent with the mentioned longer persistence of creased in the wild-type retina in response to the RA excess, Cyp26c1 in the temporal than nasal retina during the and in the heterozygous retina, the increase is still striking postnatal signal decline. but proportionately smaller. To analyze the time course of the RA effects, we injected normal newborn pups with 30 mg/kg RA, Discussion collected their retinas at successive times, and reacted all left retinas for one and the right retinas for the other In this study on Cyp26a1 null-mutants, we show that enzyme. These experiments show that RA rapidly and CYP26A1 and CYP26C1 act in parallel during normal strongly increases Cyp26a1 expression with a maximum development to generate a RA-poor stripe in the later after6–9h(Fig. 7B). The RA-induced Cyp26a1 embryonic and early postnatal retina. All Cyp26a1À/À intensification spilled over the normal stripe territory into mutants are embryonic or perinatal-lethal, with half of the the adjoining retina. Cyp26c1, in contrast, is slightly embryos already dead by E10.5, and only a few Y. Sakai et al. / Developmental Biology 276 (2004) 143–157 153 homozygous mutants surviving up to late embryonic Also in Cyp26b1À/À mutants did we not detect any consistent stages; the heterozygous Cyp26a1+/À mutants are viable abnormalities in the retina and its projections (not illustrated and appear indistinguishable from wild-type mice (Abu- here). Elimination of the Raldh2 gene causes symptoms of Abed et al., 2001; Sakai et al., 2001). The homozygous most severe vitamin A deficiency and is lethal by about E9.5; mutants exhibit severe caudal truncations, defects of the morphology of the eye region at the time of death, hindbrain patterning and homeotic transformation of including the expressions of Raldh1 and Raldh3 in it, appears vertebrae, a range of malformations that recapitulate some normal (Mic et al., 2002; Niederreither et al., 1999). The of the effects of exposing early embryos to RA excess Raldh1 null-mutants show no apparent defect in the eye or at (Morriss-Kay and Ward, 1999; Shenefelt, 1972). As in the any other site, but they are slightly more sensitive to retinol previous studies on the Cyp26a1À/À mutants no abnor- toxicity (Fan et al., 2003; Molotkov and Duester, 2003). malities were detected in the retina that is a major site of Raldh3 null-mutants exhibit only mild abnormalities, includ- CYP26A1 expression in normal mice, we did now a more ing shortening of the ventral retina with lens rotation, detailed search for defects in the mutant eye and its central persistence of the primary vitreous body as retrolenticular projections with help of a RA reporter, the RARE-hsplacZ membrane, and absence of Harderian glands (Dupe et al., transgene (Rossant et al., 1991). In the homozygous null- 2003). In Raldh1-, Raldh2-, Raldh3-, and Cyp26a1-null mutants, the RA reporter showed that normally up to mice, the RARE-hsplacZ reporter shows abnormal RA about E15, CYP26A1 is alone responsible for dividing the distributions in the retina. Dorsoventral retina markers, retina into three RA-concentration zones, and later its RA- however, such as Tbx5 and Vax2, are not affected in any of inactivating function is reinforced by CYP26C1. Cyp26c1 the mutants (Dupe et al., 2003; Fan et al., 2003). It can thus be mRNA becomes faintly visible around E14, and over the concluded that the dorsoventral identity of the mouse retina is following days its expression intensifies. In the mutants determined independently of RA synthesis and that neither lacking CYP26A1, the remaining CYP26C1 activity RA itself nor the correct subdivision of the developing retina creates by E16.5 a narrower RA-poor stripe that is visible into three zones of different RA levels is critical for early eye on the RA reporter background. In the morphologically formation. normal-appearing heterozygous Cyp26a1+/À mutants, the Why then is the developing retina subdivided into three sensitive RA-reporter tool did reveal an effect of the zones of different RA levels? It is highly unlikely that this reduced CYP26A1 amount: the lacZ-free stripe was pattern is completely dispensable for normal eye develop- narrower compared to the wild type; the Cyp26a1 ment, as it is conserved between vertebrates and as some expression stripe, by comparison, was only weaker but of it persists throughout life (McCaffery et al., 1991; Mey not thinner. This indicates that the width of the lacZ-free et al., 2001). The question about the functional signifi- stripe reflects an equilibrium between RA diffusing cance of the RA compartments in the retina is part of the tangentially from the dorsal and ventral RALDH-contain- long-standing problem of what determines the location and ing compartments and the catabolic efficiency of the timing of RA actions. While initially the expression CYP26 enzymes. Postnatally, the expression of both patterns of the RA receptors were taken as the natural Cyp26a1 and Cyp26c1 continues within the inner nuclear candidates for creating the spatial specificity, mouse null- layer for some time but disappears around P14. mutants for almost every single receptor turned out to be relatively normal (Mark and Chambon, 2003). The Role of RA in eye formation observations that the Raldh2 and Cyp26a1 knockout mice are much more severely affected than any of the single Normal development of the eye is known to depend on receptor null-mutants suggested that the variations in local properly regulated RA levels: microphthalmia was the first RA concentration must be the primary cues for the sites of example of congenital malformation found to be induced RA actions. This explanation, however, is undermined by by vitamin A deficiency (Hale, 1937), and RA excess can the observations that the malformations in RALDH cause exophthalmia or microphthalmia (Shenefelt, 1972). knockout embryos can be prevented by feeding RA to The detailed effects of RA perturbations vary to some the pregnant dams (Dupe et al., 2003; Mic et al., 2002). extent between different vertebrates, as described above in The same dietary regimen can even thwart the malforma- the Introduction. In all species, however, where the retinal tions caused by severe vitamin A deprivation in early RALDH and RA distributions have been analyzed, RA embryos (White et al., 1998). How can a universal levels in ventral retina are considerably higher than dorsally flooding of the embryo with RA compensate for the (Gru¨n et al., 2000; Marsh-Armstrong et al., 1994; McCaff- intricate localization patterns of the RALDHs? An ery et al., 1993; Mey et al., 1997, 2001). explanation for all these paradoxical observations cannot Knockout mutants have been created for three of the four be that most of the factors in retinoid actions are Raldh genes and two of the three Cyp26 genes in mice. The dispensable, but it must be the opposite: RA-mediated only noticeable gross ocular abnormality in the Cyp26a1À/À transcriptional regulation apparently matters so much for mutants is a little smaller eye size at early stages compared to survival and normal function that each step is assured by wild type, but this difference becomes less obvious later on. multiple potentially redundant mechanisms. Neither the 154 Y. Sakai et al. / Developmental Biology 276 (2004) 143–157 null-mutants nor the RA-rescued mice are completely in the body, where the ligand-dependent activation normal, and the phenotypes of the knockout mice reveal function of the RXRa—and thus possibly 9-cis RA—is only parts of the normal functions of particular genes indispensable (Mark and Chambon, 2003). (Mark and Chambon, 2003). With respect to normal eye formation and function, the very early event of dorsoven- What function could justify the reinforced maintenance of a tral-axis determination is followed by many different RA-poor zone in the retina? processes, all of which can include RA-dependent mechanisms. The high complexity of the retinoid system We know of no molecular marker whose expression in and the observations on the null-mutant mice suggest that the embryonic retina coincides with the CYP26 enzymes, mutations in its many different components or down- comparable to the colocalizations of Raldh1 with Tbx5 or stream targets may be involved in multifactorial and Raldh3 with Vax2. The only feature that colocalizes with polygenic human diseases. the RA-poor stripe is a horizontally elongated area of higher visual acuity in the adult mouse (Dra¨ger and Hubel, Relationship of CYP26A1 and CYP26C1 1976). A similar area occurs in all mammals, and in its extreme form, it is called visual streak, as in the rabbit. It is In both mouse and human, the Cyp26a1 and Cyp26c1 one of two forms of such visual specializations found in genes are located close together and in the same mammals; the other is the area centralis or fovea in the transcriptional direction, on chromosome 19 and separated temporal retina, which serves focused visual attention. The by a 4-kb interval in mice and on chromosome 10 spaced visual streak and its equivalents function as specializations 13 kb apart in humans (Taimi et al., 2004 and to monitor visual space along the horizon. They are unpublished data). This suggests that one of the genes characterized by anisotropy in their central representations, has duplicated from the template of the other. A such that the vertical dimension of visual space is vastly comparison of the expression patterns of the two genes enlarged compared to the horizontal dimension in the brain makes it likely that Cyp26c1 arose subsequent to (Dow et al., 1985; Hughes, 1971; Rosa et al., 1999). In Cyp26a1. It does not seem plausible that the two genes rabbits, the central anisotropy is very large (approximately share a common regulatory system for their expression in 7Â); in mice and primates, it is comparatively more subtle the retina, as all observations indicate that CYP26A1 and (approximately 2Â). CYP26C1 act independently here. First, there is a The visual field of the mouse is illustrated in Fig. 8 discrepancy in expression regions and stages between (adapted from (Dra¨ger and Hubel, 1976; Dra¨ger and Olsen, Cyp26c1 and Cyp26a1. Second, in contrast to Cyp26a1, 1980)); its geometry reflects the positions of the eyes Cyp26c1 mRNA is not increased in response to RA but is placed laterally and on top of the mouse’s head. Fig. 8A slightly down-regulated. Third, the human CYP26C1 shows the extent of the monocular and binocular fields enzyme was shown to be more efficient in catabolizing projected onto a globe; Fig. 8B shows the most common 9-cis RA than the other two CYP26s, suggesting a light distribution, and Fig. 8C shows the area of highest different physiological role (Taimi et al., 2004). This visual acuity, which contains also the projection of the area relatively higher affinity for 9-cis RA is reminiscent of centralis in the binocular field, which is only very poorly the conclusion drawn from many different RA-receptor developed in mice (Dra¨ger and Olsen, 1981). The entire knockout mice that ocular morphogenesis is the only site region of higher visual acuity extends parallel and above

Fig. 8. (A) Visual fields visible to the left and right eyes of a mouse plotted onto a globe (Dra¨ger and Olsen, 1980); the monocular fields are drawn as stippled yellow and red and the binocular field as solid orange. The asterisk indicates the projection of the optic disc (=blind spot). (B) Light levels are commonly much higher from above than below. (C) Approximate domain of maximal visual resolution including the region seen by the area centralis within the binocular visual field (Dra¨ger and Hubel, 1976; Dra¨ger and Olsen, 1981). Y. Sakai et al. / Developmental Biology 276 (2004) 143–157 155 the horizon, as mice look slightly upwards towards a Note added in proof stimulus of interest. This depiction illustrates the similarity of the visual field to the RA distribution in the embryonic While this paper was in press, we learned that Ribes, retina: the RALDH1-expressing region develops into the Petkovich and Dolle´ observed abnormalities in RALDH dorsal retina onto which the ground projects, the RALDH3 expressions in retinas of another Cyp26a1 null-mutant containing region into the retina seeing the sky, and the (personal communication). CYP26 expression gives rise to the area of highest acuity or maximal visual attention. The observation described above that the RA-poor stripe is backed up by two Acknowledgments independent enzymes indicates that development of a functionally important retinal specialization at its location We thank J. Rossant for the RA reporter mice, T. depends on very low RA levels such as on the conditions Furukawa for the Vax2 probe, and T. Ogura for the Tbx5 required for the repressive action of unliganded RA probe. P. Godement sent us a protocol for the eye injections receptors (Weston et al., 2003). with DiI. We also thank K. Yashiro and M. Uehara for An anisotropy of visual space similar to the mouse is helpful discussions. This work was supported by NIH apparent psychophysically in humans as optic illusions. For Grants EY01938 and EY13272. instance, when an observer is asked to generate histograms for the orientations of all elongated contours in the visual surrounding as they project onto his eyes, he finds that all References vertical lines form a tight cluster around the direction of gravity, whereas the projections of horizontal lines scatter Abu-Abed, S., Dolle, P., Metzger, D., Beckett, B., Chambon, P., Petkovich, over a wide range of angles; this phenomenon represents the M., 2001. The retinoic acid-metabolizing enzyme, CYP26A1, is perspective view. As any normal environment contains, of essential for normal hindbrain patterning, vertebral identity, and development of posterior structures. Genes Dev. 15, 226–240. course, a similar number of horizontal and vertical contours, Abu-Abed, S., MacLean, G., Fraulob, V., Chambon, P., Petkovich, M., the histogram distribution seems initially puzzling. It Dolle, P., 2002. Differential expression of the retinoic acid-metabolizing reflects a subconscious distortion of where the visual enzymes CYP26A1 and CYP26B1 during murine organogenesis. environment is located. Although we think of the visual Mech. Dev. 110, 173–177. world as a sphere or hemisphere, it is closer to a horizontal Brodsky, M.H., Steller, H., 1996. Positional information along the dorsal– ventral axis of the Drosophila eye: graded expression of the four- slab of space: most important features are localized close to jointed gene. Dev. Biol. 173, 428–446. the horizon, and the awareness of the large space towards Chambon, P., 1996. A decade of molecular biology of retinoic acid the ceiling or sky is compressed. The same distortion receptors. FASEB J. 10, 940–954. accounts probably also for the moon illusion, which is the Dow, B.M., Vautin, R.G., Bauer, R., 1985. The mapping of visual space powerful impression that the moon is larger close to the onto foveal striate cortex in the macaque monkey. J. Neurosci. 5, 890–902. horizon than up in the sky, despite its identical angular Dr7ger, U.C., Hubel, D.H., 1976. Topography of visual and somatosen- dimensions. An explanation for this illusion is an anisotropy sory projections to mouse superior colliculus. J. Neurophysiol. 39, in visual distance perception (Kaufman and Kaufman, 91–101. 2000). Dr7ger, U.C., Olsen, J.F., 1980. Origins of crossed and uncrossed retinal The reason for bringing up human psychophysics in projections in pigmented and albino mice. J. Comp. Neurol. 191, 383–412. discussing experimental results on mice is to advocate that Dr7ger, U.C., Olsen, J.F., 1981. Ganglion-cell distribution in the retina of the functional significance of the observations described the mouse. Invest. Ophthal. Visual Sci. 20, 285–293. here ought to be followed up in a species with highly Dupe, V., Matt, N., Garnier, J.M., Chambon, P., Mark, M., Ghyselinck, elaborate vision. While it will be interesting to eventually N.B., 2003. A newborn lethal defect due to inactivation of retinalde- examine the retinal phenotype in conditional double null- hyde dehydrogenase type 3 is prevented by maternal retinoic acid treatment. Proc. Natl. Acad. Sci. U. S. A. 100, 410–421. mutants for the two enzymes, visual function in mice is far Escriva, H., Delaunay, F., Laudet, V., 2000. Ligand binding and nuclear less specialized and much more difficult to study than in receptor evolution. Bioessays 22, 717–727. humans. If our hypothesis is correct that the development of Fan, X., Molotkov, A., Manabe, S., Donmoyer, C., Deltour, L., Foglio, M., regions of higher visual acuity depends on locally very low Cuenca, A., Blaner, W., Lipton, S., Duester, G., 2003. Targeted RA levels in the retina, one wonders whether this might also disruption of aldh1a1 (raldh1) provides evidence for a complex mechanism of retinoic acid synthesis in the developing retina. Mol. apply to the human fovea. It will be interesting to explore Cell. Biol. 23, 4637–4648. where and when the corresponding CYP26 enzymes are Fujii, H., Sato, T., Kaneko, S., Gotoh, O., Fujii-Kuriyama, Y., Osawa, K., expressed in the developing human eye. Because the proper Kato, S., Hamada, H., 1997. Metabolic inactivation of retinoic acid by a function of the retinoid system depends both on genetic and novel P450 differentially expressed in developing mouse embryo. environmental factors, disturbances in vitamin-A levels and EMBO J. 16, 4163–4173. Godement, P., Vanselow, J., Thanos, S., Bonhoeffer, F., 1987. A study in RA-dependent genes may play a role not only in early eye developing visual systems with a new method of staining neurones and malformations (Hornby et al., 2003), but also in some later their processes in fixed tissue. Development 101, 697–713. arising ophthalmologic problems. Grqn, F., Hirose, Y., Kawauchi, S., Ogura, T., Umesono, K., 2000. ALDH6, 156 Y. Sakai et al. / Developmental Biology 276 (2004) 143–157

a cytosolic retinaldehyde dehydrogenase prominently expressed in Changing patterns of the retinoic acid system in the developing retina. sensory neuroepithelia during development. J. Biol. Chem. 275, Dev. Biol. 158, 390–399. 41210–41218. McCaffery, P., Wagner, E., O’Neil, J., Petkovich, M., Dr7ger, U.C., 1999. Hale, F., 1937. The relation of maternal vitamin A deficiency to Dorsal and ventral retinal territories defined by retinoic acid synthesis, microphthalmia in pigs. Texas State J. Med. 33, 228–232. break-down and nuclear receptor expression. Mech. Dev. 82, 119–130. Henrique, D., Adam, J., Myat, A., Chitnis, A., Lewis, J., Ish-Horowicz, D., (Corrections 85, 203–214). 1995. Expression of a Delta homologue in prospective neurons in the Mey, J., McCaffery, P., Dr7ger, U.C., 1997. Retinoic acid synthesis in the chick. Nature 375, 787–790. developing chick retina. J. Neurosci. 17, 7339–7441. Hornby, S.J., Ward, S.J., Gilbert, C.E., 2003. Eye birth defects in humans Mey, J., McCaffery, P., Klemeit, M., 2001. Sources and sink of retinoic acid may be caused by a recessively-inherited genetic predisposition to the in the embryonic chick retina: distribution of aldehyde dehydrogenase effects of maternal vitamin A deficiency during pregnancy. Med. Sci. activities, CRABP-I, and sites of retinoic acid inactivation. Brain Res. Monit. 9, HY23–HY26. Dev. Brain Res 127, 135–148. Hughes, A., 1971. Topograhical relationships between the anatomy and Mic, F.A., Haselbeck, R.J., Cuenca, A.E., Duester, G., 2002. Novel retinoic physiology of the rabbit visual system. Doc. Ophthalmol. 30, 33–160. acid generating activities in the neural tube and heart identified by Hyatt, G., Schmitt, E.A., Marsh-Armstrong, N., McCaffery, P., Dr7ger, conditional rescue of Raldh2 null mutant mice. Development 129, U.C., Dowling, J.E., 1996. Retinoic acid establishes ventral retinal 2271–2282. characteristics. Development 122, 195–204. Molotkov, A., Duester, G., 2003. Genetic evidence that retinaldehyde Kastner, P., Mark, M., Ghyselinck, N., Krezel, W., Dupe´, V., Grondona, dehydrogenase Raldh1 (Aldh1a1) functions downstream of alcohol J.M., Chambon, P., 1997. Genetic evidence that the retinoid signal is dehydrogenase Adh1 in metabolism of retinol to retinoic acid. J. Biol. transduced by heterodimeric RXR/RAR functional units during mouse Chem. 278, 36085–36090. development. Development 124, 313–326. Morriss-Kay, G.M., Ward, S.J., 1999. Retinoids and mammalian develop- Kaufman, L., Kaufman, J.H., 2000. Explaining the moon illusion. Proc. ment. Int. Rev. Cytol. 188, 73–131. Natl. Acad. Sci. U. S. A. 97, 1281–1289. Niederreither, K., Subbarayan, V., Dolle, P., Chambon, P., 1999. Embryonic Koide, T., Downes, M., Chandraratna, R.A., Blumberg, B., Umesono, K., retinoic acid synthesis is essential for early mouse post-implantation 2001. Active repression of RAR signaling is required for head development. Nat. Genet. 21, 444–448. formation. Genes Dev. 15, 2111–2121. Peters, M.A., 2002. Patterning the neural retina. Curr. Opin. Neurobiol. 12, Koshiba-Takeuchi, K., Takeuchi, J.K., Matsumoto, K., Momose, T., Uno, 43–48. K., Hoepker, V., Ogura, K., Takahashi, N., Nakamura, H., Yasuda, K., Ray, W.J., Bain, G., Yao, M., Gottlieb, D.I., 1997. CYP26, a novel Ogura, T., 2000. Tbx5 and the retinotectum projection. Science 287, mammalian cytochrome P450, is induced by retinoic acid and defines a 134–137. new family. J. Biol. Chem. 272, 18702–18708. Lee, M.-O., Manthey, C.L., Sladek, N.E., 1991. Identification of mouse Rosa, M.G., Krubitzer, L.A., Molnar, Z., Nelson, J.E., 1999. Organization liver aldehyde dehydrogenases that catalyze the oxidation of retinalde- of visual cortex in the northern quoll, Dasyurus hallucatus: evidence for hyde to retinoic acid. Biochem. Pharmacol. 42, 1279–1285. a homologue of the second visual area in marsupials. Eur. J. Neurosci. Li, H., Wagner, E., McCaffery, P., Smith, D., Andreadis, A., Dr7ger, 11, 907–915. U.C., 2000. A retinoic acid synthesizing enzyme in ventral retina Rossant, J., Zirngibl, R., Cado, D., Shago, M., Gigue`re, V., 1991. and telencephalon of the embryonic mouse. Mech. Dev. 95, Expression of a retinoic acid response element-hsplacZ transgene 283–289. defines specific domains of transcriptional activity during mouse Lin, M., Zhang, M., Abraham, M., Smith, S.M., Napoli, J.L., 2003. embryogenesis. Genes Dev. 5, 1333–1344. Mouse RALDH4: molecular cloning, cellular expression, and activity Saari, J.C., 1994. Retinoids in photosensitive systems. In: Sporn, M.B., in 9-cis-retinoic acid biosynthesis in intact cells. J. Biol. Chem. 278, Roberts, A.B., Goodman, D.S. (Eds.), The Retinoids: Biology, 9856–9861. Chemistry, and Medicine. Raven Press, New York, pp. 351–385. Lukas, J.R., Aigner, M., Denk, M., Heinzl, H., Burian, M., Mayr, R., 1998. Sakai, Y., Meno, C., Fujii, H., Nishino, J., Shiratori, H., Saijoh, Y., Rossant, Carbocyanine postmortem neuronal tracing. Influence of different J., Hamada, H., 2001. The retinoic acid-inactivating enzyme CYP26 is parameters on tracing distance and combination with immunocyto- essential for establishing an uneven distribution of retinoic acid along chemistry. J. Histochem. Cytochem. 46, 901–910. the anterio-posterior axis within the mouse embryo. Genes Dev. 15, Luo, T., Wagner, E., Grqn, F., Dr7ger, U.C., 2004. Retinoic acid signaling 213–225. in the brain marks formation of optic projections, maturation of the Schulte, D., Furukawa, T., Peters, M.A., Kozak, C.A., Cepko, C.L., 1999. dorsal telencephalon, and function of limbic sites. J. Comp. Neurol. Misexpression of the Emx-related homeobox genes cVax and mVax2 470, 297–316. ventralizes the retina and perturbs the retinotectal map. Neuron 24, MacLean, G., Abu-Abed, S., Dolle, P., Tahayato, A., Chambon, P., 541–553. Petkovich, M., 2001. Cloning of a novel retinoic-acid metabolizing Shenefelt, R.E., 1972. Morphogenesis of malformations in hamsters caused cytochrome P450, Cyp26B1, and comparative expression analysis by retinoic acid. Teratology 5, 403–418. with Cyp26A1 during early murine development. Mech. Dev. 107, Sucov, H.M., Evans, R.M., 1995. Retinoic acid and retinoic acid receptors 195–201. in development. Mol. Neurobiol. 10, 169–184. Maden, M., 2002. Retinoid signalling in the development of the central Suzuki, R., Shintani, T., Sakuta, H., Kato, A., Ohkawara, T., Osumi, N., nervous system. Nat. Rev., Neurosci. 3, 843–853. Noda, M., 2000. Identification of RALDH-3, a novel retinaldehyde Mark, M., Chambon, P., 2003. Functions of RARs and RXRs in vivo: dehydrogenase, expressed in the ventral region of the retina. Mech. Dev. genetic dissection of the retinoid signaling pathway. Pure Appl. Chem. 98, 37–50. 75, 1709–1732. Tahayato, A., Dolle, P., Petkovich, M., 2003. Cyp26C1 encodes a novel Marsh-Armstrong, N., McCaffery, P., Dowling, J.E., Gilbert, W., retinoic acid-metabolizing enzyme expressed in the hindbrain, inner ear, Dr7ger, U.C., 1994. Retinoic acid is necessary for development of first branchial arch and tooth buds during murine development. Gene the ventral retina in zebrafish. Proc. Natl. Acad. Sci. U. S. A. 91, Expression Patterns 3, 449–454. 7286–7290. Taimi, M., Helvig, C., Wisniewski, J., Ramshaw, H., White, J., Amad, M., McCaffery, P., Tempst, P., Lara, G., Dr7ger, U.C., 1991. Aldehyde Korczak, B., Petkovich, M.A., 2004. Novel human cytochrome P450, dehydro-genase is a positional marker in the retina. Development 112, CYP26C1, involved in metabolism of 9-cis and all-trans isomers of 693–702. retinoic acid. J. Biol. Chem. 279, 77–85. McCaffery, P., Posch, K.C., Napoli, J.L., Gudas, L., Dr7ger, U.C., 1993. Theiler, K., 1972. The House Mouse. Development and Normal Stages Y. Sakai et al. / Developmental Biology 276 (2004) 143–157 157

from Fertilization to 4 Weeks of Age. Springer-Verlag, Berlin, levels of all-trans-retinoic acid. Proc. Natl. Acad. Sci. U. S. A. 95, Heidelberg, New York. 13459–13464. Wagner, E., McCaffery, P., Dr7ger, U.C., 2000. Retinoic acid in the White, J.A., Ramshaw, H., Taimi, M., Stangle, W., Zhang, A., Everingham, formation of the dorso-ventral retina and its central projections. Dev. S., Creighton, S., Tam, S.P., Jones, G., Petkovich, M., 2000. Biol. 222, 460–470. Identification of the human cytochrome P450, P450RAI-2, which is Warkany, J., Schraffenberger, E., 1946. Congenital malformations induced predominantly expressed in the adult cerebellum and is responsible for in rats by maternal vitamin A deficiency. I. Defects of the eye. Arch. all-trans-retinoic acid metabolism. Proc. Natl. Acad. Sci. U. S. A. 97, Ophthalmol. 35, 150–169. 6403–6408. Weston, A.D., Blumberg, B., Underhill, T.M., 2003. Active repression by Wilson, S.W., Rubenstein, J.L.R., 2000. Induction and dorsoventral unliganded retinoid receptors in development: less is sometimes more. patterning of the telencephalon. Neuron 28, 641–651. J. Cell Biol. 161, 223–228. Yashiro, K., Zhao, X., Uehara, M., Yamashita, K., Nishijima, M., White, J.A., Guo, Y.D., Baetz, K., Beckett-Jones, B., Bonasoro, J., Hsu, Nishino, J., Saijoh, Y., Sakai, Y., Hamada, H., 2004. Regulation K.E., Dilworth, F.J., Jones, G., Petkovich, M., 1996. Identification of of retinoic acid distribution is required for proximo-distal pattern- the retinoic acid-inducible all-trans-retinoic acid 4-hydroxylase. J. Biol. ing and outgrowth of the developing mouse limb. Dev. Cell 6, Chem. 271, 29922–29927. 411–422. White, J.C., Shankar, V.N., Highland, M., Epstein, M.L., DeLuca, Zhao, D., McCaffery, P., Ivins, K.J., Neve, R.L., Hogan, P., Chin, W.W., H.F., Clagett-Dame, M., 1998. Defects in embryonic hindbrain Dr7ger, U.C., 1996. Molecular identification of a major retinoic-acid development and fetal resorption resulting from vitamin A synthesizing enzyme: a retinaldehyde-specific dehydrogenase. Eur. J. deficiency in the rat are prevented by feeding pharmacological Biochem. 240, 15–22.