Abstract

Retinoic acid (RA) affects development and function of the brain, but little is known about how much is made locally and where it is distributed. To identify RA-sensitive neural processes, we mapped the RA-synthesizing retinaldehyde dehydrogenases (RALDHs) during postnatal brain formation of the mouse. High and stable RALDH expressions mark the basal ganglia, olfactory bulbs, hippocampus and auditory afferents as major sites of RA actions in the functional brain. During the early postnatal period, transient and very high RALDH3 expressions distinguish two developmental events: (i) the colonization of the nucleus accumbens and the olfactory bulbs by neuronal precursors and (ii) the maturation of selected parts of the cerebral cortex. In the cortex, RALDH3 is transiently activated in postmigratory layer II/III neurons during formation of their dendritic arbors and it is transported in their axons across the corpus callosum. RALDH3-expressing cortical regions include most of the limbic lobe, with strongest expression in the anterior cingulate cortex, medial and lateral secondary visual cortices, auditory cortical areas, the secondary motor cortex and some association areas. The transient cortical expression points to a brief RA-critical period during differentiation of the cortical network that serves in the coordination of sensory-motor activity with emotional and recently learned information.

Introduction

Retinoic acid (RA), an active form of vitamin A, is essential for development of the vertebrate embryo. It functions as ligand for the activation of transcription factors, the nuclear RA receptors. Different classes of RA receptors are expressed in characteristic, but wide-spread and overlapping patterns throughout all locations of the embryo (Mangelsdorf et al., 1994). Because of this overlap, the receptor expression provides only limited cues as to when and where RA is active in the embryo. Consistent with the abundant expression, genetic eliminations of most single receptors cause no or only subtle developmental defects (Mark et al., 1999). A critical determinant is the restricted availability of RA itself. RA is synthesized from retinaldehyde through oxidation by retinaldehyde dehydrogenases—RALDH1, RALDH2 and RALDH3—three enzymes that were first detected biochemically in the eye of the embryonic mouse (McCaffery et al., 1992). The RALDHs are expressed much more sparsely than the RA receptors: expression sites of different RALDHs usually do not overlap and within the brain they can be separated across hundreds of micrometers by tissue totally lacking RA synthesis (Smith et al., 2001). The small lipid RA can diffuse through solid tissue, presumably forming concentration gradients that span for considerable distances from the RALDH expression sites into RALDH-free regions. The morphogenetic significance of RA gradients lies in the specific RA thresholds which are required for the transcription of particular genes (Boncinelli et al., 1991). These thresholds differ by several orders of magnitude of RA concentrations between different genes.

The early brain is completely devoid of RA synthesis, but is flanked by two regions of extremely high RALDH expression, one located rostrally in the eye and face, the other caudally in the trunk (Smith et al., 2001; Niederreither et al., 2002). Many different lines of evidence indicate that the morphogenesis of the early hindbrain depends on an RA gradient that emanates from the RALDH-rich trunk (Gavalas and Krumlauf, 2000) and RA diffusing from the eye and face influences the early fore-brain (Enwright and Grainger, 2000). As the embryonic brain grows, most neuronal tissue remains free of RA synthesis, but RALDH expression moves closer: RALDH3 begins to be expressed centrally within the embryonic telencephalon in the lateral ganglionic eminence and RALDH2 appears externally in the meninges (Smith et al., 2001). Comparisons of RALDH expressions at different locations in the developing embryo show that levels are very high at sites and during periods of neurogenesis and neuronal differentiation, but decline with maturation (McCaffery et al., 1993; McCaffery and Dräger, 1994a,b). This is consistent with a role of RA in neuronal determination and promotion of neurite outgrowth.

Recent observations indicate that retinoid-dependent processes are also essential for several functions in the adult brain: RA is required for formation of new song neurons in male finches (Denisenko-Nehrbass et al., 2000), normal motor performance is controlled by RA receptors expressed in the basal ganglia (Krezel et al., 1998) and the presence of RA receptors and RA is necessary for the cellular mechanisms that underlie learning and memory in the hippocampus (Chiang et al., 1998; Misner et al., 2001). RA receptors and retinoid-binding proteins continue to be expressed throughout the adult brain in characteristic and overlapping patterns (Krezel et al., 1999). However, reports on how much RA is made in the brain and where it is localized, are contradictory. Traditionally, the brain was not included in RA measurements, because its RA content was believed to be negligible (Napoli, 1994). More recently, traces of RA synthesis (Zetterstrom et al., 1999) or low levels were detected in the adult brain (Connor and Sidell, 1997; Zetterstrom et al., 1999). General synthesis was described to be very high, twice the rate as in the retinoid-rich liver (Dev et al., 1993), or robust synthesis was localized selectively to the hippocampus (Misner et al., 2001). The total RA content was reported to be derived largely from the circulation (Kurlandsky et al., 1995), or, under different conditions, it was found to be synthesized almost exclusively locally within the brain (Werner and Deluca, 2002). These divergent conclusions reflect the particular measurement techniques, because RA cannot be fixed and visualized directly, and because the trace lipid RA is difficult to detect by HPLC in the lipid-rich brain.

As the RALDHs are critical determinants for the sites of RA actions, their expression sites provide the best available histological criteria to estimate RA distribution. Localization of the mRNAs gives only a partial view for the brain, because neurons can transport functional RALDH enzymes axonally across long distances. Here, we used a combination of several techniques to localize the sites of RA synthesis in the postnatal brain. We will first describe the stable RALDH expression in the adult brain and then transient RALDH3 patterns during early postnatal brain maturation. The RALDH patterns we observed show intriguing spatial and temporal parallels to subtle neuroanatomical abnormalities described in schizophrenia (Lewis and Lieberman, 2000; Sawa and Snyder, 2002), an illness which has previously been connected with the retinoid system (Goodman, 1998). Schizophrenia, a severe mental disorder of unknown etiology, is commonly believed to be of neurodevelopmental origin and due both to genetic and environmental factors. The subtle neuroanatomical abnormalities in schizophrenic brains, which are most pronounced in regions involved in the integration of sensory-motor functions with cognition, drive and emotion, show signs of impaired neuronal differentiation (Selemon and Goldman-Rakic, 1999). In a separate article, we explain how the sites and developmental stages of RA actions described here fit together with well-established knowledge from the retinoid field to provide a hypothetical pathophysiological mechanism for schizophrenia (U.C. Dräger et al., manuscript under revision).

Materials and Methods

The methods used for immunohistochemistry, in situ hybridization and Northern blotting, were standard and followed previously summarized protocols (Smith et al., 2001). The main experimental difficulties were with the RALDH antisera: because the sequences of the three RALDHs are rather similar, the antisera tend to show some cross-reactivity. However, because the RALDH mRNA probes are highly specific, we tested all localizations for protein and mRNA of all three RALDHs. In addition, the new RALDH3 antiserum required preabsorption with Triton-X-100-insoluble extracts prepared from paraformaldehyde-fixed brains, in order to remove cross-reactivities with intermediate filaments, in particular neurofilaments.

Results

Stable RALDH Expression Patterns in the Adult Brain

RALDH2

In the embryo, all RALDH2 expression in the forebrain is limited to the meninges, and neuronal expression is seen only in a few neurons of the hindbrain and spinal cord (Zhao et al., 1996; Niederreither et al., 1997; Smith et al., 2001). A similar pattern continues throughout postnatal development into the adult: all RALDH2 expression in the rostral postnatal brain is located in meninges and only in the medulla can a few neurons be detected (not shown). The distribution of RALDH2 protein in the rostral brain is shown on coronal sections through different levels of the hippocampus, viewed at low and higher magnifications, to illustrate a meningeal specialization (Fig. 1ae): RALDH2 is strongly expressed in the remnants of the meninges, which covered the outer telencephalic surface in the embryo and which condensed into evenly spaced channel-like structures (arrows), as the medial telencephalic rim curled inwards to form the hippocampus. In the more rostral sections (Fig. 1a,c,d), the meningeal channels are cut perpendicular and the caudal section (Fig. 1b,e) gives a tangential view.

RALDH1

Neuronal RALDH1 expression in the embryonic brain is restricted to dopaminergic cells in the ventral tegmentum, which project to the forebrain, and minor levels are detected prenatally in the meninges (McCaffery and Dräger, 1994; Smith et al., 2001). These two RALDH1 sites remain similar in the adult brain, but the expression levels increase dramatically: a rather small number of neurons in the ventral midbrain (Fig. 1h,i) send vast amounts of RALDH1 enzyme by axonal transport to the ventral telencephalon, the dorsal striatum (Fig. 1f,g) and the nucleus accumbens shell. During the first postnatal week, a few of the RALDH1-labeled axons can be seen to stray also into the dorsal telencephalon, mainly medio-rostrally into the cingulate cortex; at later stages, aberrant fibers are no longer detectable (not shown). These transient misprojections are minor and are likely an example of exuberant early projections. RALDH1 is also strongly expressed in the adult meninges, in particular in the RALDH2-positive channels that surround the hippocampus (Fig. 1h,j). In addition, the blood vessels of the entire brain appear labeled by the RALDH1 antiserum. Because tissue edges, like blood vessels and meninges, are notorious for histological artifacts, we compared dissected samples for RALDH expressions by Northern blotting (Fig. 2a): the outer cortex preparation, which consisted of upper cortical layers and meninges, contains both high RALDH1 and RALDH2 levels, and the lower cortical layers have only RALDH1. Isolated hippocampi express RALDH2 and very high RALDH1 mRNA levels. The RALDH1 mRNA comes in two transcript sizes: the blood vessels and meninges of the adult brain contain a heavier band (∼2.5 kb) and a lighter band (∼2 kb) is expressed at all other locations, including embryonic retina and brain, adult retina, substantia nigra, liver and kidney (Smith et al., 2001).

RALDH3

RALDH3 expression is overall very low in the adult forebrain: it is not detectable in adult dorsal telencephalon, either by Northern blotting (Fig. 2) or histological screens. A small number of neurons in the ventral telencephalon and the hypothalamus contain RALDH3 protein (not shown). By far the strongest forebrain expression is in the olfactory bulbs (Fig. 1k,m): a subpopulation of periglomerular cells expresses high amounts of RALDH3 enzyme throughout their cell bodies and dense processes that fill the glomeruli diffusely. Double-labeling for tyrosine hydroxylase (TH) shows that the RALDH3-containing periglomerular cells are distinct from the dopaminergic interneurons (Fig. 1mo). The small somata of the periglomerular cells can also be detected by in situ hybridizations for RALDH3 mRNA (not shown). In addition to the high RALDH3 expression, the adult olfactory bulb is surrounded by RALDH1- and RALDH2-positive meninges that follow the olfactory nerve fibers for some distance into its ventral side, as illustrated in Figure 1l for a parasagittal RALDH2-labeled section. These combined RALDH expressions make the olfactory bulbs, next to the striatum, the RA-richest parts of the adult forebrain.

Compared to the overall sparse RALDH3 expression in the adult forebrain, very high levels are detectable in the hindbrain: the cochlear nucleus appears almost solidly labeled by the RALDH3 antiserum (Fig. 3g). When we tested the same region by Northern blotting and in situ hybridization, however, practically no RALDH3 mRNA was found to be present in the cochlear nucleus itself, although mRNA could be detected nearby in large RALDH3-positive neurons of the facial nucleus (Fig. 3d and not shown). In order to determine the origin of the RALDH3 enzyme in the cochlear nucleus, we tested the inner ear; for technical reasons, we sectioned early postnatal heads, when the skull was not yet ossified. In situ hybridizations show very strong RALDH3 mRNA expression in the somata of the spiral-ganglion neuron within the cochlea (Fig. 3a). Immunohisto-chemical preparations indicate that RALDH3 enzyme is transported along the peripheral axonal branches towards the auditory hair cells (Fig. 3c) and in the centrifugal branches along the auditory nerve towards the cochlear nucleus (Figs. 3b,f).

Transient RALDH3 Expression in the Early Postnatal Brain

Ventral Telencephalon

During embryonic forebrain development, RALDH3 is strongly expressed in the ventral telencephalon and it represents a major local RA source at early stages. In the later embryo, RALDH3 mRNA levels decline, as is apparent on entire telencephala of different ages tested by Northern blotting (Smith et al., 2001). When such assays are extended to postnatal ages (Fig. 2b), the total amounts of telencephalic RALDH3 mRNA can be seen to rise again transiently during the first postnatal week and then to decline slowly to undetectable adult levels. In order to find out where this transient RALDH3 expression is located, we screened postnatal brains of different ages by in situ hybridization and immunohistochemistry. For the ventral telencephalon, Figure 4c,e gives the localization of RALDH3 mRNA at postnatal day 1.5 (P1.5), and Figure 4d,f shows RALDH3 protein in adjacent sections. As in the late embryo (Fig. 4a,b), RALDH3 mRNA in the early postnatal brain is located mostly in the rostral subventricular zone (SVZ) of the lateral ventricles, but only small amounts remain. The neighboring antiserum-labeled sections (Fig. 4d,f) indicate that high RALDH3-protein expression persists in small cells that appear to move. Figure 4gm illustrates RALDH3-labeled cells in continuity with the SVZ during early postnatal days: some seem to disperse along the SVZ (Fig. 4g), some head towards the septum and the nucleus accumbens (Fig. 4h,i) and others head towards the olfactory bulb via the rostral migratory stream (Fig. 4km). While RALDH3-labeled cells in the septum and in the olfactory tubercle (Fig. 4j) exhibit already differentiated neuronal features, most of the streaming cells are tiny, consisting mainly of a stubby process with large growth cone and tight perisomal region. This is shown in Figure 4m for a cell that seems to have made a right-angle turn from the rostral migratory stream, located centrally in the olfactory bulb, towards the glomerular layer. Such RALDH3-labeled cells are most common during the first postnatal week, but a few much more weakly labeled ones can be found in the adult olfactory bulb.

Dorsal Telencephalon

While RALDH3 mRNA declines in the ventral telencephalon, it increases dorsally in parts of the cerebral cortex. The locations and the rising phase of the cortical expression are shown on low-power coronal brain sections for different ages reacted by in situ hybridization (Fig. 5). In the embryonic dorsal telencephalon, no RALDH3 can be detected (Fig. 4a,b), but all expression seems to begin after birth. At P1.5, the cortical expression is still weak; it increases over the following few days up to a maximum around P6; by P10 levels have declined again significantly and by P13 only remnants remain visible. The labeled regions form one continuous band that extends along the medial cortex and wraps around the occipital pole; a second continuous region is labeled in the lateral part of the caudal telencephalon, adjoining the rhinal fissure (arrows). Both the lateral and particularly the medial parts are much larger than single cytoarchitectonic areas. As judged by their locations, the labeled regions include both limbic and neocortical areas. A tentative list of the medial areas, using a simplified version of Paxinos and Franklin’s (Paxinos and Franklin, 2001) nomenclature for the adult mouse, includes the following limbic areas: piriform cortex, medial orbital cortex, prelimbic cortex, cingulate cortical areas, retrosplenial cortices and presubiculum; as well as the following neocortical areas: parietal association cortices, secondary motor area and medial secondary visual area. The caudo-temporal areas include the lateral secondary visual area, the secondary and primary auditory areas, the temporal association cortex and the ectorhinal cortex. Neither by in situ hybridization nor by Northern blotting did we detect RALDH3 mRNA in the hippocampus. The labeling intensity changes over time, both absolutely and relatively. Along the medial band, intensity steps are apparent and become more abrupt with time. Although these appear to demarcate future cytoarchitectonic boundaries, the labeling intensity does not distinguish limbic from neocortical areas: it is strongest in the anterior cingulate cortex (ACC) and more rostral limbic areas, as well as in the parietal association cortex (PtA) and the medial secondary visual area (V2M).

Comparisons of the in situ preparations with neighboring antiserum-labeled sections allow the identification of RALDH3-positive cells (Figs 6 and 7). The enzyme is synthesized in immature neurons that are located in the most superficial stratum of the cortical plate, i.e. in the future layer II or II/III. These are the last neurons to be generated during the neuronogenetic period, which ends by embryonic day 17 in the mouse (Caviness et al., 2000). In the anterior cingulate cortex (ACC) at P1.5 (Fig. 7c), the labeled cells still have mostly the bipolar appearance of early postmigratory neurons. Over the following days they assume the appearance of pyramidal cells: first, they sprout densely branched processes from their apical dendrites which occupy the inner half of the marginal layer I (Fig. 7f,g). These apical dendritic tufts account for the bistratified labeling that becomes apparent in some cortical regions of the antiserum-treated preparations but not in the in situ hybridizations (Figs 6ce and 7b,e). In the later preparations the labeled pyramidal cells show signs of basal dendrites, as illustrated for the parietal association cortex (Fig. 7h), but cellular details are difficult to judge, since most cells in the densely packed layer II seem to contain RALDH3.

The immunohistochemical preparations indicate that the differences in labeling intensity between cortical regions must reflect differences in predominantly labeled neuronal type. In the most heavily labeled cortical areas with abundant apical dendritic arborizations, as in the anterior cingulate cortex (ACC), secondary motor area (M2), parietal association cortex (PtA) and medial secondary visual area (V2M), RALDH3 seems mostly contained in pyramidal cells. In the more weakly labeled areas, including the posterior cingulate cortex (PCC) and the retrosplenial cortical areas (RS), the enzyme is expressed in smaller neuronal somata at the upper edge of the cortical plate which have symmetrically arranged dendrites. RALDH3 expression in the lateral cortical areas, which include the lateral secondary visual area (V2L), auditory areas, temporal association cortex (TeA) and ectorhinal cortex (Ect), is overall weaker. Throughout the medial cortex the labeled cells project across the corpus callosum, where they form large superficial and thinner deep bundles. Figure 6f shows the callosal projections originating from the caudal telencephalon in four coronal views from caudal to rostral: axons from the medial visual association area, the retrosplenial cortex and the presubiculum first head rostrally within the white matter to reach the callosum, where they turn sharply and cross. We saw callosal projections only between the same regions on both sides, i.e. we could not determine where else the axons project to; similarly, we did not detect axonal projections from the lateral cortical areas. Both of these negative observations probably reflect mainly technical limitations due to poor antiserum quality and they do not preclude additional labeled axonal projections.

Discussion

We began the reported studies with the aim of addressing the controversy on RA synthesis in the brain described in the Introduction, but we were not aiming for anything relevant to schizophrenia. This association emerged only subsequently when we tried to make sense of the puzzling RALDH3 expression patterns. For this reason, we will here discuss the RALDHs in a neurobiological and developmental context. In a separate article we will explain how the presented results combine with a wealth of other information on retinoids and on the known vulnerability factors for schizophrenia into a comprehensive pathophysiological disease hypothesis (U.C. Dräger et al., manuscript under revision).

We have described here the localizations of three major RA-generating dehydrogenases in the postnatal brain. Of the three, RALDH2 and RALDH3 are highly efficient in RA production, whereas RALDH1 is much less so per amount of protein. RALDH1 occurs at very high protein concentrations, however, and it has been described to function in addition as binding protein for small lipophilic compounds, including thyroid (Yamauchi and Tata, 1994) and steroid hormones (Pereira et al., 1991). The three RALDHs are likely to represent the main local sources of RA for the brain, as we did not detect any other major enzymes in previous biochemical assays (McCaffery and Dräger, 1995). By such assays several aldehyde/xanthine-oxidase activities are observed at different locations throughout the body, but their contributions to total RA synthesis appear to be relatively minor (Manthey et al., 1990). We do not have evidence for the expression of major novel RALDHs, as were recently postulated for the early embryo (Mic et al., 2002). As in the embryonic brain (Smith et al., 2001), we find only rather sparse RALDH expression postnatally. Our observations are consistent with the measurements by Kurlandski et al. (Kurlandsky et al., 1995) on the origins of tissue RA pools in the adult rat: they reported that almost 90% of total brain RA is not synthesized locally but is taken up from the plasma RA pool. Similarly, Le Doze et al. (Le Doze et al., 2000) found that intravenously injected all-trans-RA permeates into the brain white matter to concentrations six or seven times higher than those of the serum. It seems thus plausible that the RA concentrations in large parts of the brain represent a partition equilibrium between lipophilic pockets on serum proteins and the lipid-rich brain. In selected locations close to RALDHs, however, local RA levels must be significantly higher than average brain levels.

In a study on vitamin-A-deprived rats, Werner and Deluca (Werner and Deluca, 2002) observed that a large fraction of a systemically injected retinol dose is converted locally within the brain into RA and that RA is not preferentially imported from the plasma; they conclude that the brain synthesizes RA more efficiently than other tissues. One interpretation of this result, which differs from the observations in vitamin-A-sufficient rats, is that the brain might be a privileged site of retinol uptake from the circulation under conditions of severe vitamin-A deficiency. By electrophysiological analyses, Chiang, Misner and colleagues established that RA actions are absolutely required for the mechanisms of learning and memory in the adult hippocampus (Chiang et al., 1998; Misner et al., 2001); their observations characterize the hippocampus as a major site of RA usage in the brain. To account for local RA synthesis, we found expression of both RALDH1 and RALDH2 in the meninges and particularly the meningeal channels surrounding the hippocampus; RALDH1 is expressed in addition in the blood vessels (Fig. 1). In previous biochemical assays on RA-synthesizing enzymes in the adult hippocampus (McCaffery and Dräger, 1994a), we detected only RALDH2 activity. However, for technical reasons additional RA synthesis by RALDH1 cannot be excluded, because RALDH1 function tends to become undetectable in dilute samples due to its high oxidation vulnerability. Alternatively, it is possible that the enzyme translated from the heavier RALDH1 mRNA transcript, which is expressed selectively in adult meninges and blood vessels, differs from the more common product of the lower mRNA band, on which all previous functional RALDH1 assays were done (McCaffery et al., 1992; McCaffery and Dräger, 1994). For instance, it might be that the lipophilic binding properties mentioned above (Pereira et al., 1991; Yamauchi and Tata, 1994) are relatively more pronounced compared to RA synthesis in the heavier than the lower-band product, and that RALDH1 contributes to accumulation of retinol from the circulation into the brain.

Although most of our analyses were focused on the rostral brain, we illustrate in Figure 3 the axonal transport of RALDH3 protein from the inner ear to the brain stem. This situation is a parallel to the nigro-striatal RALDH1 projection, where all RALDH1 mRNA is confined to the ventral tegmentum and practically all of the functional enzyme is transported axonally to the ventral telencephalon (McCaffery and Dräger, 1994; Smith et al., 2001). The function of RALDH3 transport from the spiral ganglion might be a trophic influence of the inner ear on the cochlear nucleus. The tegmental projection onto the striatum is well known to exert a strong trophic influence. This influence is commonly ascribed to dopamine, but RA synthesis might be a significant additional component; the RALDH1 inhibitor disulfiram (Antabuse) is known to cause acute Parkinson’s disease in rare instances (Laplane et al., 1992).

Transient RALDH expression events are likely to mark critical periods of RA actions, as defined by Shenefelt (Shenefelt, 1972). This is the case for the earliest expression in the prospective forebrain: RALDH2 appears for a few hours during the start of neurulation at the rostral edge of the neural plate, where the eyes and telencephalon will develop (Wagner et al., 2000; Smith et al., 2001). This event marks the most teratogenic period for the eyes and brain, because an excess of RA just prior to and during the start of RALDH2 expression causes the most severe malformations (Shenefelt, 1972). The transient RALDH3 expression in migrating precursors of the newborn ventral telencephalon undoubtedly represents the postnatal tail of the neuronal migration from the remnants of the embryonic lateral ganglionic eminence. It seems plausible that the tiny migrating cells turn down RALDH3 transcription for economic reasons, while on route. The origin, path and final phenotype of precursors migrating out from the ganglionic eminence have been beautifully illustrated by grafting labeled cells into the early embryo (Wichterle et al., 2001): lateral-ganglionic-eminence-derived cells migrate antero-ventrally to transform into distinct neuronal populations in the striatum, nucleus accumbens, olfactory tubercle and olfactory bulb.

The transient expression of RALDH3 in an odd assortment of cortical areas clearly does not demarcate similar cytoarchitectonic areas nor neoversus allocortex, but it appears to distinguish a network of cortical circuits that connects peristriate visual areas with other cortical regions (Vogt and Miller, 1983). RALDH3 appears in early postmigratory neurons and it accompanies their major period of dendritic arborization. Although neurogenesis is complete by E17 in the mouse, the medial cortex is last to mature (Caviness et al., 2000), which makes it likely that new neurons continue to arrive through the first part of RALDH3 expression. The RALDH3-labeled cells are the youngest subpopulation of the entire callosal projection, which originates in addition to layers II/III from layer V and the subplate (DeAzevedo et al., 1997). The labeled cortical areas represent a system involved in the integration of sensorimotor circuits with emotional and remembered information (Vogt and Miller, 1983). In particular, the anterior cingulate cortex is now considered as the cortical center where perception is linked with motor control to provide a network for motivation, cognition, intelligent behavior and emotional self-control in humans (Paus, 2001). It is intriguing to find that RALDH3, which is the most important RA source for the embryonic retina, also marks the differentiation stage of the cortical system for visual working memory and cognitive functions. This cortical network is evolved to much higher sophistication in primates, whose visual capacity also dramatically exceeds that of mice. It may be that higher mental functions evolved as a corollary of higher visual capacity, since logical reasoning, in the absence of any visual input, was found to activate an occipitoparietal–frontal network in humans (Knauff et al., 2002), which includes the areas homologous to the RALDH3-labeled cortex in the mouse.

The sites of RALDH3 expression in the mouse represent the cortical regions that correspond to areas in the human brain which are disproportionately affected in schizophrenia and mood disorders, and the corresponding developmental stage is during the second trimester, which is also a vulnerable period for the acquisition of a schizophrenia predisposition (DeAzevedo et al., 1997; Lewis and Lieberman, 2000; Pearlson, 2000; Sawa and Snyder, 2002). This lead will be pursued ielsewhere (U.C. Dräger et al., manuscript under revision).

E.W. and T.L. contributed equally to this work.

We thank F. Grün for his contributions to the generation of the new RALDH3 antiserum, R. Lindahl for the RALDH1 antiserum, J. Crandall and other members of the E.K. Shriver Center for help and comments, and J. Adams, F. Benes, W. Blaner and U. Hämmerling for discussions. This work was supported by NIH grant EY01938.

Figure 1.

Stable expression sites of the three RALDHs in the adult brain, detected by antisera. (ae) Coronal sections showing RALDH2 in meninges; arrows point to the meningeal channels around the hippocampus. (fj) RALDH1 expression in tegmental neurons projecting to the striatum, and in meninges and blood vessels: (f) low-power view of horizontal section through the olfactory bulbs and the dorsal striatum; (g) parasagittal section showing expression in the striatum, blood vessels and meninges; (hj) coronal sections showing RALDH1-positive neurons in the ventral tegmentum, as well as meningeal channels around the hippocampus (arrows) and blood vessels. (ko) Olfactory bulb: (k) low-power view of horizontal section showing RALDH3 in glomeruli; (l) parasagittal section labeled for RALDH2 to show expression in meninges and nerve sheaths around the ventrally entering olfactory axons; (mo) higher magnification of the glomeruli double-labeled for RALDH3 and tyrosine hydroxylase (TH). Scale bars: a, b = 2 mm; f = 1 mm; mo = 100 μm; all others = 500 μm.

Figure 1.

Stable expression sites of the three RALDHs in the adult brain, detected by antisera. (ae) Coronal sections showing RALDH2 in meninges; arrows point to the meningeal channels around the hippocampus. (fj) RALDH1 expression in tegmental neurons projecting to the striatum, and in meninges and blood vessels: (f) low-power view of horizontal section through the olfactory bulbs and the dorsal striatum; (g) parasagittal section showing expression in the striatum, blood vessels and meninges; (hj) coronal sections showing RALDH1-positive neurons in the ventral tegmentum, as well as meningeal channels around the hippocampus (arrows) and blood vessels. (ko) Olfactory bulb: (k) low-power view of horizontal section showing RALDH3 in glomeruli; (l) parasagittal section labeled for RALDH2 to show expression in meninges and nerve sheaths around the ventrally entering olfactory axons; (mo) higher magnification of the glomeruli double-labeled for RALDH3 and tyrosine hydroxylase (TH). Scale bars: a, b = 2 mm; f = 1 mm; mo = 100 μm; all others = 500 μm.

Figure 2.

Northern blots of RALDH mRNA in postnatal tissues. (a) Adult and P3 liver, adult retina, outer and inner cortex, hippocampus and occipital cortex, probed for RALDH expressions; the signals in the three blots were aligned for easier comparison. (b) Variations in RALDH3 expression in entire cerebral hemispheres, tested in two separately collected age series.

Figure 2.

Northern blots of RALDH mRNA in postnatal tissues. (a) Adult and P3 liver, adult retina, outer and inner cortex, hippocampus and occipital cortex, probed for RALDH expressions; the signals in the three blots were aligned for easier comparison. (b) Variations in RALDH3 expression in entire cerebral hemispheres, tested in two separately collected age series.

Figure 3.

(Top) RALDH3 expression in the cochlea and hindbrain. (a,b) Low-power views of parasagittal sections through P5 cochlea reacted for RALDH3 mRNA (a) and RALDH3 protein (b), to show mRNA in the spiral ganglion (sg) and enzyme transport along the auditory nerve fibers; the labeling over the basilar membranes represents auto-fluorescence; neuronal somata in the vestibular ganglion (vg) are not labeled. (c) Higher magnification of coronal section showing the spiral-ganglion somata and their peripheral axons towards the auditory hair cells that are not visible. (d) Large RALDH3-positive neurons in the facial nucleus (fn) and their axonal projections. (eg) Low to higher magnification views of the hindbrain to show labeling of the cochlear nucleus (cn) at P5 and in the adult mouse; the round profiles in the adult do not represent cell bodies themselves but are large terminals that surround the octopus cells. Scale bars: a,b,e,f = 500 μm; c,g = 50 μm; d = 100 μm.

Figure 3.

(Top) RALDH3 expression in the cochlea and hindbrain. (a,b) Low-power views of parasagittal sections through P5 cochlea reacted for RALDH3 mRNA (a) and RALDH3 protein (b), to show mRNA in the spiral ganglion (sg) and enzyme transport along the auditory nerve fibers; the labeling over the basilar membranes represents auto-fluorescence; neuronal somata in the vestibular ganglion (vg) are not labeled. (c) Higher magnification of coronal section showing the spiral-ganglion somata and their peripheral axons towards the auditory hair cells that are not visible. (d) Large RALDH3-positive neurons in the facial nucleus (fn) and their axonal projections. (eg) Low to higher magnification views of the hindbrain to show labeling of the cochlear nucleus (cn) at P5 and in the adult mouse; the round profiles in the adult do not represent cell bodies themselves but are large terminals that surround the octopus cells. Scale bars: a,b,e,f = 500 μm; c,g = 50 μm; d = 100 μm.

Figure 4.

(Bottom) Transient RALDH3 expression in the ventral telencephalon. (af) Comparisons of RALDH3 mRNA and protein at two late embryonic ages (a,b) and in neighboring sections at P1.5 (cf): note that the mRNA is confined to the region surrounding the lateral ventricle (lv) and that the protein-containing cells appear to move outwards; na, nucleus accumbens, s, septum. (gm) Coronal sections showing RALDH3 protein at P1.5–3; lateral is to the right; note the small cells dispersing along the subventricular zone (SVZ), towards the nucleus accumbens and towards the olfactory bulb along the rostral migratory stream (rms); RALDH3-positive neurons in the olfactory tubercle (ot) have begun differentiation; the arrow in (m) points to the olfactory bulb periphery, where the growth cone (gc) of the small cell is heading for. Scale bars: af = 500 μm; g,h,i,l = 200 μm; j,k = 100 μm; m ;= 50 μm.

Figure 4.

(Bottom) Transient RALDH3 expression in the ventral telencephalon. (af) Comparisons of RALDH3 mRNA and protein at two late embryonic ages (a,b) and in neighboring sections at P1.5 (cf): note that the mRNA is confined to the region surrounding the lateral ventricle (lv) and that the protein-containing cells appear to move outwards; na, nucleus accumbens, s, septum. (gm) Coronal sections showing RALDH3 protein at P1.5–3; lateral is to the right; note the small cells dispersing along the subventricular zone (SVZ), towards the nucleus accumbens and towards the olfactory bulb along the rostral migratory stream (rms); RALDH3-positive neurons in the olfactory tubercle (ot) have begun differentiation; the arrow in (m) points to the olfactory bulb periphery, where the growth cone (gc) of the small cell is heading for. Scale bars: af = 500 μm; g,h,i,l = 200 μm; j,k = 100 μm; m ;= 50 μm.

Figure 5.

(Top) Topography of transient RALDH3 mRNA expression sites in the cerebral cortex. Coronal sections through P1.5, P3, P5 and P6 brains treated by in situ hybridization; randomly selected sections are arranged in each vertical row from caudal to rostral. The designations of cortical areas is only based on topography, but not any additional criteria: V2M, medial secondary visual area; RS, retrosplenial cortical areas; PtA, parietal association cortex; M2, secondary motor area; ACC, anterior cingulate cortex. Scale bar = 2 mm.

Figure 5.

(Top) Topography of transient RALDH3 mRNA expression sites in the cerebral cortex. Coronal sections through P1.5, P3, P5 and P6 brains treated by in situ hybridization; randomly selected sections are arranged in each vertical row from caudal to rostral. The designations of cortical areas is only based on topography, but not any additional criteria: V2M, medial secondary visual area; RS, retrosplenial cortical areas; PtA, parietal association cortex; M2, secondary motor area; ACC, anterior cingulate cortex. Scale bar = 2 mm.

Figure 6.

(Bottom) Transient RALDH3 mRNA and enzyme expression in the occipital cortex. In situ hybridization (b) and immunohistochemical detection (all others) of RALDH3 at P3 (a) and P5 (all others). Note that the labeled neurons send axons towards the white matter (wm) and the corpus callosum (cc). The medial (V2M) and lateral (V2L) secondary visual areas are labeled, but not the primary visual cortex in between; RS, retrosplenial cortical areas; the designations of V2L, the temporal association cortex (TeA) and the ectorhinal cortex (Ect) are only approximate, because no borders are apparent. All scale bars = 500 μm.

Figure 6.

(Bottom) Transient RALDH3 mRNA and enzyme expression in the occipital cortex. In situ hybridization (b) and immunohistochemical detection (all others) of RALDH3 at P3 (a) and P5 (all others). Note that the labeled neurons send axons towards the white matter (wm) and the corpus callosum (cc). The medial (V2M) and lateral (V2L) secondary visual areas are labeled, but not the primary visual cortex in between; RS, retrosplenial cortical areas; the designations of V2L, the temporal association cortex (TeA) and the ectorhinal cortex (Ect) are only approximate, because no borders are apparent. All scale bars = 500 μm.

Figure 7.

Transient RALDH3 mRNA and enzyme expression in the fronto-parietal cortex. In situ hybridization (a) and immunohistochemical detection (all others) to illustrate the maturation of the RALDH3-positive neurons; PtA, parietal association cortex; M2, secondary motor area; ACC and PCC, anterior and posterior cingulate cortices; the brackets in (g) indicate the upper cortical layers. Scale bars: ae = 500 μm; fh = 50 μm.

Figure 7.

Transient RALDH3 mRNA and enzyme expression in the fronto-parietal cortex. In situ hybridization (a) and immunohistochemical detection (all others) to illustrate the maturation of the RALDH3-positive neurons; PtA, parietal association cortex; M2, secondary motor area; ACC and PCC, anterior and posterior cingulate cortices; the brackets in (g) indicate the upper cortical layers. Scale bars: ae = 500 μm; fh = 50 μm.

References

Boncinelli E, Simeone F, Acampora D, Mavilio F (
1991
) Hox gene activation by retinoic acid.
Trends Genet
 
7
:
329
–334.
Caviness VS, Takahashi T, Nowakowski RS (
2000
) Neocortical malformation as consequence of nonadaptive regulation of neuronogenetic sequence.
MRDD Res Rev
 
6
:
22
–33.
Chiang MY, Misner D, Kempermann G, Schikorski T, Giguere V, Sucov HM, Gage FH, Stevens CF, Evans RM (
1998
) An essential role for retinoid receptors RARbeta and RXRgamma in long-term potentiation and depression.
Neuron
 
21
:
1353
–1361.
Connor MJ, Sidell N (
1997
) Retinoic acid synthesis in normal and Alzheimer diseased brain and human neural cells.
Mol Chem Neuropathol
 
30
:
239
–252.
DeAzevedo LC, Hedin-Peirera C, Lent R (
1997
) Callosal neurons in the cingulate cortical plate and subplate of human fetuses.
J Comp Neurol
 
386
:
60
–70.
Denisenko-Nehrbass NI, Jarvis E, Scharff C, Nottebohm F, Mello CV (
2000
) Site-specific retinoic acid production in the brains of adult song birds.
Neuron
 
27
:
359
–370.
Dev S, Adler A, Edwards R (
1993
) Adult rabbit brain synthesizes retinoic acid.
Brain Res
 
632
:
1
–4.
Enwright JF, Grainger RM (
2000
) Altered retinoid signaling in the heads of Small eye mutant mouse embryos.
Dev Biol
 
221
:
1
–22.
Gavalas A, Krumlauf R (
2000
) Retinoid signaling and hindbrain patterning.
Curr Opin Genet Dev
 
10
:
380
–386.
Goodman AB (
1998
) Three independent lines of evidence suggest retinoids as causal to schizophrenia.
Proc Natl Acad Sci USA
 
95
:
7240
–7244.
Knauff M, Mulack T, Kassubek J, Salih HR, Greenlee MW (
2002
) Spatial imagery in deductive reasoning: a functional MRI study.
Brain Res Cogn Brain Res
 
13
:
203
–212.
Krezel W, Ghyselinck N, Samad TA, Dupe V, Kastner P, Borrell IE, Chambon P (
1998
) Impaired locomotion and dopamine signaling in retinoid receptor mutant mice.
Science
 
279
:
863
–867.
Krezel W, Kastner P, Chambon P (
1999
) Differential expression of retinoid receptors in the adult mouse central nervous system.
Neuroscience
 
89
:
1291
–1300.
Kurlandsky SB, Gamble MV, Ramakrishnan R, Blaner WS (
1995
) Plasma delivery of retinoic acid to tissues in the rat.
J Biol Chem
 
270
:
17850
–17857.
Laplane D, Attal N, Sauron B, de Billy A, Dubois B (
1992
) Lesions of basal ganglia due to disulfiram neurotoxicity.
J Neurol Neurosurg Psychiatry
 
55
:
925
–929.
Le Doze F, Debruyne D, Albessard F, Barre L, Defer GL (
2000
) Pharmacokinetics of all-trans retinoic acid, 13-cis retinoic acid, and fenretinide in plasma and brain of rat.
Drug Metab Dispos
 
28
:
205
–208.
Lewis DA, Lieberman JA (
2000
) Catching up on schizophrenia: natural history and neurobiology.
Neuron
 
28
:
325
–334.
McCaffery P, Dräger UC (
1994
) High levels of a retinoic-acid generating dehydrogenase in the meso-telencephalic dopamine system.
Proc Natl Acad Sci USA
 
91
:
7772
–7776.
McCaffery P, Dräger UC (
1994
) Hotspots of retinoic acid synthesis in the developing spinal cord.
Proc Natl Acad Sci USA
 
91
:
7194
–7197.
McCaffery P, Dräger UC (
1995
) Retinoic acid synthesizing enzymes in the embryonic and adult vertebrate.
Adv Exp Med Biol
 
372
:
173
–183.
McCaffery P, Lee M-O, Wagner MA, Sladek NE, Dräger UC (
1992
) Asymmetrical retinoic acid synthesis in the dorso-ventral axis of the retina.
Development
 
115
:
371
–382.
McCaffery P, Posch KC, Napoli JL, Gudas L, Dräger UC (
1993
) Changing patterns of the retinoic acid system in the developing retina.
Dev Biol
 
158
:
390
–399.
Mangelsdorf DJ, Umesono K, Evans RM (1994) The retinoid receptors. In: The retinoids: biology, chemistry, and medicine (Sporn MB, Roberts AB, Goodman DS, eds), pp. 319–349. New York: Raven Press.
Manthey CL, Landkamer GJ, Sladek NE (
1990
) Identification of the mouse aldehyde dehydrogenases important in aldophosphamide detoxification.
Cancer Res
 
50
:
4991
–5002.
Mark M, Ghyselinck NB, Wendling O, Dupe V, Mascrez B, Kastner P, Chambon P (
1999
) A genetic dissection of the retinoid signalling pathway in the mouse.
Proc Nutr Soc
 
58
:
609
–613.
Mic FA, Haselbeck RJ, Cuenca AE, Duester G (
2002
) Novel retinoic acid generating activities in the neural tube and heart identified by conditional rescue of Raldh2 null mutant mice.
Development
 
129
:
2271
–2282.
Misner DL, Jacobs S, Shimizu Y, de Urquiza AM, Solomin L, Perlman T, De Luca LM, Stevens CF, Evans RM (
2001
) Vitamin A deprivation results in reversible loss of hippocampal long-term synaptic plasticity.
Proc Natl Acad Sci USA
 
98
:
11714
–11719.
Napoli JL (1994) Retinoic acid homeostasis. Prospective roles of betacarotene, retinol, CRBP, and CRABP. In: Vitamin A in health and disease (Blomhoff R, ed.), pp. 135–188. New York: Marcel Dekker.
Niederreither K, Fraulob V, Garnier J-M, Chambon P, Dollé P (
2002
) Differential expression of retinoic acid-synthesizing (RALDH) enzymes during fetal development and organ differentiation in the mouse.
Mech Dev
 
110
:
165
–171.
Niederreither K, McCaffery P, Dräger UC, Chambon P, Dollé P (
1997
) Restricted expression and retinoic acid-induced downregulation of the retinaldehyde dehydrogenase type 2 (RALDH-2) gene during mouse development.
Mech Dev
 
62
:
67
–78.
Paus T (
2001
) Primate anterior cingulate cortex: where motor control, drive and cognition interface.
Nature Reviews
 
2
:
417
–424.
Paxinos G, Franklin KBJ (2001) The mouse brain in stereotaxic coordinates, 2nd edn. New York: Academic Press.
Pearlson GD (
2000
) Neurobiology of schizophrenia.
Ann Neurol
 
48
:
556
–566.
Pereira F, Rosenmann E, Nylen E, Kaufman M, Pinsky L, Wrogemann K (
1991
) The 56 kDa androgen binding protein is an aldehyde dehydrogenase.
Biochem Biophys Res Commun
 
175
:
831
–838.
Sawa A, Snyder SH (
2002
). Schizophrenia: diverse approaches to a complex disease.
Science
 
296
:
692
–695.
Selemon LD, Goldman-Rakic PS (
1999
) The reduced neuropil hypothesis: a circuit based model of schizophrenia.
Biol Psychiatry
 
45
:
17
–25.
Shenefelt RE (
1972
) Morphogenesis of malformations in hamsters caused by retinoic acid.
Teratology
 
5
:
403
–418.
Smith D, Wagner E, Koul O, McCaffery P, Dräger UC (
2001
) Retinoic acid synthesis for the developing telencephalon.
Cereb Cortex
 
11
:
894
–905.
Vogt BA, Miller MW (
1983
) Cortical connections between rat cingulate cortex and visual, motor and postsubicular cortices.
J Comp Neurol
 
216
:
192
–210.
Wagner E, McCaffery P, Dräger UC (
2000
) Retinoic acid in the formation of the dorso-ventral retina and its central projections.
Dev Biol
 
222
:
460
–470.
Werner EA, Deluca HF (
2002
) Retinoic acid is detected at relatively high levels in the CNS of adult rats.
Am J Physiol Endocrinol Metab
 
282
:
E672
–E678.
Wichterle H, Turnbull DH, Nery S, Fishell G, Alvarez-Buylla A (
2001
) In utero fate mapping reveals distinct migratory pathways and fates of neurons born in the mammalian basal forebrain.
Development
 
128
:
3759
–3771.
Yamauchi K, Tata JR (
1994
) Purification and characterization of a cytosolic thyroid-hormone-binding protein (CTBP) in Xenopus liver.
Eur J Biochem
 
225
:
1105
–1112.
Zetterstrom RH, Lindquist E, Mata de Urquiza A, Tomac A, Eriksson U, Perlmann T, Olson L (
1999
) Role of retinoids in the CNS: differential expression of retinoid binding proteins and receptors and evidence for presence of retinoic acid.
Eur J Neurosci
 
11
:
407
–416.
Zhao D, McCaffery P, Ivins KJ, Neve RL, Hogan P, Chin WW, Dräger UC (
1996
) Molecular identification of a major retinoic-acid synthesizing enzyme: a retinaldehyde-specific dehydrogenase.
Eur J Biochem
 
240
:
15
–22.