Regulation of retinoic acid signaling in the embryonic nervous system: a master differentiation factor.

This review describes some of the properties of retinoic acid (RA) in its functions as a locally synthesized differentiation factor for the developing nervous system. The emphasis is on the characterization of the metabolic enzymes that synthesize and inactivate RA, and which determine local RA concentrations. These enzymes create regions of autocrine and paracrine RA signaling in the embryo. One mechanism by which RA can act as a differentiation agent is through the induction of growth factors and their receptors. Induction of growth factor receptors in neural progenitor cells can lead to growth factor dependency, and the consequent developmental fate of the cell will depend on the local availability of growth factors. Because RA activates the early events of cell differentiation, which then induce context-specific differentiation programs, RA may be called a master differentiation factor.

Restricted expression and retinoic acid-induced downregulation of the retinaldehyde dehydrogenase type 2 (RALDH-2) gene during mouse development.

Retinaldehyde dehydrogenase type 2 (RALDH-2) was identified as a major retinoic acid generating enzyme in the early embryo. Here we report the expression domains of the RALDH-2 gene during mouse embryogenesis, which are likely to indicate regions of endogenous retinoic acid (RA) synthesis. During early gastrulation, RALDH-2 is expressed in the mesoderm adjacent to the node and primitive streak. At the headfold stage, mesodermal expression is restricted to posterior regions up to the base of the headfolds. Later, RALDH-2 is transiently expressed in the undifferentiated somites and the optic vesicles, and more persistently along the lateral walls of the intraembryonic coelom and around the hindgut diverticulum. The RALDH-2 expression domains in differentiating limbs, which include presumptive interdigital regions, coincide with, but slightly precede, those of the RA-inducible RAR beta gene. The RALDH-2 gene is also expressed in specific regions of the developing head, including the tooth buds, inner ear, meninges and pituitary gland, and in several viscera. Administration of a teratogenic dose of RA at embryonic day 8.5 results in downregulation of RALDH-2 transcript levels in caudal regions of the embryo, and may reflect a mechanism of negative feedback regulation of RA synthesis.

Sagittal band expression of COUP-TF2 gene in the developing cerebellum.

In the developing cerebellum, the medio-lateral compartmentalization of the adult cerebellum is preceded by the transient expression of factors which divide the cortex into similar parasagittal stripes. Here we report that COUP-TF2, an orphan member of the nuclear receptor family which suppresses RA actions by forming heterodimers with RXR, shows a pattern of sagittal bands in developing mouse cerebellum. The band pattern changes according to the developmental stage. At embryonic day 13 it is expressed in the lateral half of the cerebellum, but at later stages the expression is divided into several parasagittal bands. By postnatal day 5 the COUP-TF2 expression substantially decreases to low, but detectable, levels.

A retinoic acid synthesizing enzyme in ventral retina and telencephalon of the embryonic mouse.

Most retinoic acid (RA) in the embryonic mouse is generated by three retinaldehyde dehydrogenases (RALDHs). RALDH1 (also called E1, AHD2 or ALDH1) is expressed in the dorsal retina, and RALDH2 (V2, ALDH11) generates most RA in the embryonic trunk. The third one, RALDH3 (V1), synthesizes the bulk of RA in the head of the early embryo. We show here that RALDH3 is a mouse homologue to ALDH6, an aldehyde dehydrogenase cloned from adult human salivary gland (Hsu, L.C., Chang, W.-C., Hiraoka, L., Hsien, C.-L., 1994. Molecular cloning, genomic organization, and chromosomal localization of an additional human aldehyde dehydrogenase gene, ALDH6. Genomics 24, 333-341), which was recently reported to act as a RALDH (Yoshida, A., Rzhetsky, A., Hsu, L.C., Chang, C., 1998. Human aldehyde dehydrogenase gene family. Eur. J. Biochem. 251, 549-557). RALDH3 expression begins in the surface ectoderm over the optic recess. In rapidly changing expression patterns it labels the appearance of several ectodermal structures: it marks the formation of the lens and the olfactory organ from ectodermal placodes, and it delineates the beginning eyelid field. Within the optic vesicle, RALDH3 is expressed in the ventral retina and the dorsal pigment epithelium. In the telencephalon, RALDH3 is expressed at high levels in the lateral part of the ganglionic eminence. From here it extends via the piriform cortex into the lower part of the septum. Of the three RALDHs, RALDH3 shows the strongest predilection for epithelia.

Dorsal and ventral rentinoic territories defined by retinoic acid synthesis, break-down and nuclear receptor expression.

Determination of the dorso-ventral dimension of the vertebrate retina is known to involve retinoic acid (RA), in that high RA activates expression of a ventral retinaldehyde dehydrogenase and low RA of a dorsal dehydrogenase. Here we show that in the early eye vesicle of the mouse embryo, expression of the dorsal dehydrogenase is preceded by, and transiently overlaps with, the RA-degrading oxidase creating a trough between very high ventral and moderately high dorsal RA levels. Most of the RA receptors are expressed uniformly throughout the retina except for the RA-sensitive RARbeta, which is down-regulated in the CYP26 stripe. The orphan receptor COUP-TFII, which modulates RA responses, colocalizes with the dorsal dehydrogenase. The organization of the embryonic vertebrate retina into dorsal ventral territories divided by a horizontal boundary has parallels to the division of the Drosophila eye disc into dorsal, equatorial and ventral zones, indicating that the similarities in eye morphogenesis extend beyond single molecules to topographical patterns.

Dorsal and ventral retinal territories defined by retinoic acid synthesis, break-down and nuclear receptor expression.

Determination of the dorso-ventral dimension of the vertebrate retina is known to involve retinoic acid (RA), in that high RA activates expression of a ventral retinaldehyde dehydrogenase and low RA of a dorsal dehydrogenase. Here we show that in the early eye vesicle of the mouse embryo, expression of the dorsal dehydrogenase is preceded by, and transiently overlaps with, the RA-degrading oxidase CYP26. Subsequently in the embryonic retina, CYP26 forms a narrow horizontal boundary between the dorsal and ventral dehydrogenases, creating a trough between very high ventral and moderately high dorsal RA levels. Most of the RA receptors are expressed uniformly throughout the retina except for the RA-sensitive RARbeta, which is down-regulated in the CYP26 stripe. The orphan receptor COUP-TFII, which modulates RA responses, colocalizes with the dorsal dehydrogenase. The organization of the embryonic vertebrate retina into dorsal and ventral territories divided by a horizontal boundary has parallels to the division of the Drosophila eye disc into dorsal, equatorial and ventral zones, indicating that the similarities in eye morphogenesis extend beyond single molecules to topographical patterns.

Receptive fields of single cells and topography in mouse visual cortex.

The visual cortex was studied in the mouse (C57 Black/6J strain) be recording from single units, and a topographic map of the visual field was constructed. Forty-five percent of the neurons in striate cortex responded best to oriented line stimuli moving over their receptive fields; they were classified as simple (17%), complex (25%) and hypercomplex (3%). Of all preferred orientations horizontal was most common. Fifty-five percent of recpetive fields were circularly symmetric: these were on-center (25%), off-center (7%) and homogeneous on-off in type (23%). Optimal stimulus velocities were much higher than those reported in the cat, mostly varying between 20 degrees and 300 degrees/sec. The field of vision common to the two eyes projected to more than one-third of the striate cortex. Although the contralateral eye provided the dominating influence on cells in this binocular area, more than two-thirds of cells could also be driven through the ipsilateral eye. The topography of area 17 was similar to that found in other mammals: the upper visual field projected posteriorly, the most nasal part mapped onto the lateral border. Here the projection did not end at the vertical meridian passing through the animal's long axis, but proceeded for at least 10 degrees into the ipsilateral hemifield of vision, so that at least 20 degrees of visual field were represented in both hemispheres. The magnification in area 17 was rather uniform throughout the visual field. In an area lateral to area 17 (18a) the fields were projected in condensed mirror image fashion with respect to the arrangement of area 17. Medial to area 17 a third visual area (area 18) was again related to 17 as a condensed mirror image.

Studies of visual function and its decay in mice with hereditary retinal degeneration.

Functional implications of mouse hereditary retinal degeneration have been studied at the level of the superior colliculus and visual cortex in the C57BL/6J-le rd strain. On autoradiography at a light-microscopic level, following eye injection with radioactive compounds, central visual structures appeared normal. A slight reduction in ipsilateral retinal projection was probably related to reduced retinal pigmentation associated with the light ear (le) mutation. In recordings from visual cortex and tectum in rd mice older than five months the cells discharged with highly rhythmic maintained activity. This ongoing activity depended on retinal input, since temporary asphyxia of the eye stopped it immediately. The frequency of the rhythm was influenced by the anesthesia. In these older mice no visual receptive fields could be mapped, but in a few tectal recordings it was possible to suppress the maintained activity by diffuse, very intense illumination. As in normal mice, no auditory or somatosensory responses were observed in the visual cortex or upper tectal layers. In recordings from tectum before the age of three weeks retinotopic topography and receptive fields were normal. By day 24 no receptive fields could be recorded from parts of the tectum representing the central 90--100 degrees of the visual field, whereas within a peripheral ring responses were still roughly normal under photopic conditions. Over the following four months these peripheral responses faded away slowly. Incremental thresholds, especially in the scotopic range, were elevated, rising slowly to unmeasurable values. Similarly during dark adaptation the thresholds fell to values several log units above those reached in normal mice; these values of dark adapted thresholds in rd mice rose with age. This is consistent with morphological changes known to occur in the retina as a consequence, of the rd mutation the rods degenerating before the cones.

Depth segregation of retinal ganglion cells projecting to mouse superior colliculus.

The retinotectal projections in the mouse were analyzed with injections of horseradish peroxidase into the superior colliculus and of radioactive amino acids into the eye. At least 70% of the ganglion cells, and possibly all of them, were found to project to the superior colliculus, including ganglion cells of all sizes. Small injections revealed that ganglion cells of different sizes terminate at different levels in the superior colliculus. The small ganglion cells that form the vast majority of all cells project predominantly to the upper stratum griseum superficiale. A small population of mainly medium-sized and large ganglion cells project to the deep stratum griseum superficiale and to the stratum opticum. The ipsilateral projection is restricted to the deep stratum griseum superficiale and stratum opticum and consists predominantly of medium-sized and large ganglion cells.

Origins of crossed and uncrossed retinal projections in pigmented and albino mice.

The extent of the binocular cortical field in albino mice, as revealed by recording from single cells, was almost normal; although the input from the ipsilateral eye was weaker than normal, most cells were driven from both eyes. By backfilling retinal ganglion cells from one optic tract with horseradish peroxidase we examined the origins of the retinofugal projections. Filled cells ipsilateral to the injected tract were concentrated in a crescent-shaped area bordering the inferior temperal retina. In black mice this area constituted 20% of the total retinal area, in albinos 17%. In black mice we counted nearly 1,000 labeled cells in the ipsilateral retina, or 2.6% of all cells filled in both eyes. Albinos had about one-third fewer filled cells ipsilaterally than black mice. Four percent of all ipsilaterally filled cells in black mice and 8% in albinos were scattered outside of the crescent region. The density of ipsilaterally projecting cells was uniform throughout the crescent region in black mice, but decreased toward the central retina in albinos. In retinas contralateral to the injection up to 39,000 cells were filled-about two-thirds of the cells in the ganglion-cell layer whose cytoplasm contained conspicuous Nissl substance. Depending on classification of unfilled cells as ganglion cells or interneurons, we estimated a total of 48,000 to 65,000 ganglion cells to exist in the retina. The size distribution of ipsilaterally projecting ganglion cells was similar in albinos and normals. Ipsilaterally projecting ganglion cells were on average 1.8-3 times larger in volume than contralaterally projecting ones in both types of mice. Displaced ganglion cells were relatively more common in ipsilateral retinofugal projections: 21% of all ipsilateral ganglion cells were displaced versus less than 1% of all the contralateral ganglion cells in black mice. In albinos only 13% of the ganglion cells in the ipsilateral retina were displaced. The overall reduction in ipsilaterally projecting cells in albinos was reflected twice as much in displaced ganglion cells as in normally placed ones.

Thy-1 antigen: a ganglion cell specific marker in rodent retina.

A monoclonal antibody, 2G12 , has been produced against a rat cerebral cortex glycoprotein fraction. It interacts with Thy-1 based on both the tissue distribution of its reactivity and the blocking of its binding by pretreatment with a rabbit anti-Thy-1 serum. In rat retina this antibody labels only cell bodies in the ganglion cell layer, optic nerve fibres and the inner plexiform layer. That the cell-body labelling was confined to ganglion cells was confirmed by double-labelling experiments. Ganglion cells were distinguished by retrograde transport of fluorescent markers injected into the superior colliculi. The retinas were dissociated into single cells and the cell suspension was labelled with 2G12 . There was almost complete coincidence of the two labels. A monoclonal antibody against mouse Thy-1.2 gave an essentially identical pattern of labelling. In both rats and mice Thy-1 was also found on the vitreal surface of the inner limiting membrane in a pattern reminiscent of that formed by the Müller cell endfeet, although these cells do not express Thy-1.

A ventrodorsal GABA gradient in the embryonic retina prior to expression of glutamate decarboxylase.

GABA is known to function as a neurotransmitter in the mature nervous system, and in immature neurons it has been linked to neurotrophic actions. While most GABA is generated by glutamate decarboxylase (GAD), an alternative synthetic pathway is known to originate from putrescine, which is converted via gamma-aminobutyraldehyde in an aldehyde-dehydrogenase-requiring step to GABA. In a search for the role of two aldehyde dehydrogenases expressed in segregated compartments along the dorsoventral axis of the developing retina, we assayed dorsal and ventral retina fractions of the mouse for GABA by high performance liquid chromatography. We found GABA to be present in the embryonic retina, long before expression of GAD, and ventral GABA levels exceeded dorsal levels by more than three-fold. Postnatally, when GAD became detectable, overall GABA levels increased, and the ventrodorsal concentration difference disappeared. Our observations indicate that prior to the formation of synapses the embryonic retina contains a ventrodorsal GABA gradient generated by an alternate synthetic pathway.

Ganglion cell distribution in the retina of the mouse.

The distribution of ganglion cells in the mouse retina was studied with the use of Nissl criteria for distinguishing cell types in the ganglion cell layer. Retrograde filling with horseradish peroxidase (HRP) from the optic fiber tract helped to validate Nissl criteria and served to identify displaced ganglion cells. We estimated a total of 117,000 nonvascular cells in the ganglion cell layer; of these, 70,000 were probably ganglion cells, and 47,000 could not be classified. The density of the presumed ganglion cells was highest-more than 8000 cells/mm2-just temporal to the optic disk, and lowest-less than 2000 cells/mm2-in the most dorsal retina. The retinal region with highest ganglion cell density was slightly elongated in a nasotemporal direction. About 2% of all HRP-filled ganglion cells had their cell bodies in the inner nuclear layer. These displaced cells differed in topographical distribution from the normally positioned ganglion cells: although occurring throughout the retina, they were more common along the retinal periphery. Measurements of ganglion cell areas showed a tendency toward larger size with eccentricity. At no retinal location did cell-size histograms reveal clearly separate size classes.

Early posttranslational modifications of the three neurofilament subunits in mouse retinal ganglion cells: neuronal sites and time course in relation to subunit polymerization and axonal transport.

We have characterized stages in the posttranslational processing of the three neurofilament subunits, High (NF-H), Middle (NF-M), and Low (NF-L), in retinal ganglion cells in vivo during the interval between synthesis in cell bodies within the retina and appearance of these polypeptides in axons at the level of the optic nerve (optic axons). Neurofilament proteins pulse-labeled by injecting mice intravitreally with [35S]methionine or [32P]orthophosphate, were isolated from Triton-soluble and Triton-insoluble fractions of the retina or optic axons by immunoprecipitation or immunoaffinity chromatography. Within 2 h after [35S]methionine injection, the retina contained neurofilament-immunoreactive radiolabeled proteins with apparent molecular weights of 160, 139, and 70 kDa, which co-migrated with subunits of axonal neurofilaments that were dephosphorylated in vitro with alkaline phosphatase. The two larger polypeptides were not labeled with [32P]orthophosphate, indicating that they were relatively unmodified forms of NF-H and NF-M. About 75% of the subunits were Triton-insoluble by 2 h after isotope injection, and this percentage increased to 98% by 6 h. Labeled neurofilament polypeptides appeared in optic axons as early as 2 h after injection. These subunits exhibited apparent molecular weights of 160, 139, and 70 kDa and were Triton-insoluble. The time of appearance of fully modified polypeptide forms differed for each subunit (2 h for NF-L, 6-18 h for NF-M, 18-24 h for NF-H) and was preceded by the transient appearance of intermediate forms. The modified radiolabeled subunits in optic axons 3 days after synthesis were heavily labeled with [32P]orthophosphate and exhibited the same apparent molecular weights as subunits of axonal neurofilaments (70 kDa, 145 and 140 kDa, and 195-210 kDa, respectively). Whole mounts of retina immunostained with monoclonal antibodies against NF-H in different states of phosphorylation demonstrated a transition from non-phosphorylated neurofilaments to predominantly phosphorylated ones within a region of the axon between 200 and 1000 microns downstream from the cell body. These experiments demonstrate that the addition of most phosphate groups to NF-M and NF-H takes place within a proximal region of the axon. The rapid appearance of modified forms of NF-L after synthesis may imply that processing of this subunit occurs at least partly in the cell body. The presence of a substantial pool of Triton-insoluble, unmodified subunits early after synthesis indicates that the heaviest incorporation of phosphate occurs after neurofilament proteins are polymerized.(ABSTRACT TRUNCATED AT 400 WORDS)

Influence of the choroid plexus on cerebellar development: analysis of retinoic acid synthesis.

The choroid plexus of the fourth ventricle is conspicuous both in location and size: it protrudes over the outer hindbrain, closely apposed to the caudal external surface of the cerebellum, and it is disproportionately large early on. While the developing cerebellum is known to respond to retinoic acid (RA), it does not express significant levels of RA synthesizing enzyme. Retinaldehyde dehydrogenase levels in the choroid plexus, however, are very high, with maxima during the pre- and postnatal periods of cerebellar morphogenesis. Explants assays demonstrate release of a neurite-outgrowth promoting activity from the choroid plexus, whose levels parallel the levels of RA synthesizing enzyme here, and which can be mimicked by RA. These observations characterize the choroid plexus as a paracrine, growth-promoting organ for the developing cerebellum, with the effects mediated through temporally regulated RA production.

A novel assay for retinoic acid catabolic enzymes shows high expression in the developing hindbrain.

We have employed a novel technique that determines the relative capacity of tissues to catabolize all-trans retinoic acid (RA) to a metabolite incapable of activating a RA reporter cell line. This assay uses the microsomal fraction of tissues from the developing mouse and detects a pathway which requires NADPH and is inhibitable by ketoconazole, suggesting that a cytochrome P450-dependent enzyme may be required. High catabolic activity was detected transiently in the developing cerebellum which peaked at postnatal day 2. The medulla oblongata was the only other CNS region with high catabolic capacity, its earlier expression peak, between embryonic days 16 and 17, likely reflecting its earlier maturation. In the CNS, the hindbrain is exceptional in its high expression of RA catabolic enzymes, suggesting a unique function for regulated RA levels in this region.

Murine BC1 RNA in dendritic fields of the retinal inner plexiform layer.

Rodent BC1 RNA is a non-messenger RNA polymerase III transcript that is almost exclusively expressed in nerve cells. BC1 RNA has been localised in somatic and dendritic domains of neurons, and its location has been interpreted to indicate a functional role in extrasomatic postsynaptic protein synthesis. In previous in situ hybridisation experiments, it has been demonstrated that in the retina most of the BC1 labelling signal was confined to the ganglion cell layer and the inner plexiform layer. Dendritic processes of several types of neurons form the neuritic plexus of the inner plexiform layer, and in order to determine the contribution of ganglion cells to the BC1 labelling signal, we eliminated this cell type by transecting the optic nerve unilaterally in newborn mice. Deletion of the ganglion cells resulted in a significant reduction although not a complete elimination of the BC1 signal in the inner plexiform layer. These data indicate that dendritic processes of both ganglion cells and amacrine cells contain BC1 RNA.

A sensitive bioassay for enzymes that synthesize retinoic acid.

Retinoic acid (RA) is a potent regulator of gene transcription and it plays a pivotal role in neural development. As the compound is active at nanomolar concentrations, standard RA detection methods based on high-pressure liquid chromatography (HPLC) are poorly suited for neurodevelopmental questions and single RA measurements require pooling of tissues from very large numbers of embryos. An alternative approach is to determine the potential for RA synthesis by assaying for the enzymes that catalyze the last step of RA synthesis, the irreversible oxidation of retinaldehyde to RA. In a quantitative comparison of retinaldehyde dehydrogenase levels with endogenous RA levels, we found a good concordance between the two parameters. For the detection of retinaldehyde dehydrogenases we have developed an assay which involves analyses of protein fractions, separated by isoelectric focusing (IEF), with a RA responsive cell line; operationally, this technique is a zymography bioassay. This method is exquisitely sensitive: tissue volumes as small as a sugar grain can be assayed, and it can be used on all species. Separation by IEF allows in a single run the identification and relative quantitation of several enzyme isoforms.

Postnatal effects of retinoic acid on cerebellar development.

Retinoic acid(RA) is a potent teratogen to which the early CNS is known to be highly sensitive. However, very little is know about the postnatal effects of RA. The cerebellum is a candidate for postnatal RA toxicity, as it develops late and exhibits temporal patterns of RA synthesis that are synchronized with developmental stages. Northern blotting shows that the cerebellum expresses receptors for RA, predominantly of the RXR group activated by 9-cis RA. To determine whether development can be disrupted by RA excess, newborn rats were injected with RA at a time when endogenous RA levels are normally low. Histological examinations at postnatal day 14 revealed loss of a subpopulation of proliferating granule cells, and adult cerebella showed clusters of ectopic granule cells located in the molecular layer. Thus, although the granule cells may normally be regulated by endogenous RA, they are sensitive to abnormally high RA concentrations.