Retinoic Acid Signaling in the Functioning Brain

See allHide authors and affiliations

Science's STKE  28 Feb 2006:
Vol. 2006, Issue 324, pp. pe10
DOI: 10.1126/stke.3242006pe10


Retinoic acid, an active form of vitamin A, regulates gene expression throughout the body, and many components of the signaling system through which it acts are present in the brain. Very little is known, however, about how retinoic acid functions in neurobiological systems. Several studies have provided evidence that retinoic acid plays a role in sleep, learning, and memory, but the precise mechanisms through which it influences these processes remain unclear. All of these processes involve local or long-range inhibition and synchronized neuronal activity between separate locations in the brain. A critical component in the generation of the synchronized firing of cortical neurons (cortical synchrony) is a network of inhibitory interneurons containing parvalbumin, a cell population affected by retinoid perturbations, such as exposure to a vitamin A overdose. An understanding of the role of retinoids in normal brain function would provide clues to the long-standing question of whether abnormalities in retinoic acid signaling contribute to the pathogenesis of some brain diseases with uncertain etiologies that involve both genetic and environmental factors.

Some of the most intriguing papers in science are those that demonstrate indisputably a fact that cannot be explained with existing information. This sort of unexplainable fact appears in the meticulous study by Maret et al. on retinoic acid (RA) signaling and sleep (1), and in general, such open questions characterize the state of retinoid research on the functioning brain (2). Although most components of the retinoid-signaling pathway are present in the mature brain, very little is known about where in the brain RA acts, what proteins it influences, and how RA signaling affects brain function. Moreover, these questions are barely being investigated.

Vitamin A acts through one of the most powerful and pervasive biological systems that exist for the regulation of gene expression (3). Many genetically encoded proteins are required for retinoid uptake, metabolic conversion, storage, transport, intracellular binding, and transcriptional regulation (Fig. 1). RA binds to 14 different RA-receptor proteins that regulate transcription either as heterodimers with each other or with other members of the nuclear receptor gene family, including the receptors for thyroid hormone, vitamin D, and peroxisome proliferator–activated receptor (PPAR). Many of the retinoid-linked nuclear receptors are widely expressed throughout the brain in an overlapping fashion. Because of the large number of receptor isoforms and the enormous number of different heterodimeric combinations, RA-dependent processes can be influenced by a vast range of combinatorial receptor constellations. The expression of many brain proteins is known from in vitro studies to be potentially regulated by RA (2). Because specific RA actions in vivo depend on the cellular and developmental context, however, one cannot infer from the in vitro data which of the many possible sequelae of RA regulation take place at a particular site and developmental stage in the brain (2).

Fig. 1.

Very simplified schematic of the many molecules involved in retinoid biosynthesis and handling, and of the transcriptional processes influenced by binding of RA to its nuclear receptors. These include the various retinoids, the enzymes implicated in their biosynthesis, the retinoid-binding proteins that help regulate their location and activity, and the nuclear receptors and cofactors through which the retinoids act to regulate transcription.

The effects of perturbations on the elaborate retinoid system underline its importance. For instance, vitamin A deficiency during early development leads to severe malformations of different organs, which tend to phenocopy genetically caused or sporadically occurring deformities, such as inborn heart, lung, kidney, skeletal, and eye defects (4). This indicates that RA exerts a critical regulatory role on many of the genes that establish the basic body plan in the early embryo. One of the oldest questions in retinoid research concerns what determines the sites of RA actions and the locations of the defects caused by vitamin A deficiency. Most components of the retinoid-signaling pathway were believed at some time to be the critical spatial determinants of RA function, until experimental elimination of their genes revealed inconsistencies in this interpretation of their role. For the most part, mice lacking a single one of the retinoid genes (RA receptors, retinoid-binding proteins, or retinoid-converting enzymes) appear more or less normal (3). When different knockout mice are bred together, however, the compound mutants exhibit all the symptoms and malformations found in vitamin A–deprived embryos and known collectively as the fetal vitamin A–deficiency syndrome (3, 4). The effects of different retinoid mutations are additive and are exacerbated by nutritional vitamin A deprivation. The retinoid genes whose elimination causes the most severe vitamin A–deficiency syndrome are the retinaldehyde dehydrogenases (RALDHs), the enzymes that create high RA concentrations locally (5, 6). In contrast to the widespread expression of the RA receptors, expression of the RALDHs is restricted to a few sites, with most intervening tissue being totally free of any RA synthetic enzymes (5, 79).

The observation that all genetic and environmental disturbances converge on the same pleiotropic pathological syndrome demonstrates a broadly distributed systemic quality to vitamin A’s actions. Although the different retinoid proteins each serve specific localized functions, they also exhibit functional redundancies with each other, so that each helps to protect the entire organism from vitamin A deficiency or from malfunction in any of the other retinoid genes. These functional redundancies make it difficult to infer the crucial sites of action of a particular retinoid gene from its normal expression patterns. The poor correlation between normal gene expression and anatomical or functional abnormalities in knockout mice is most pronounced for the RA receptors, whereas RALDH expression sites represent the relatively best indicators of the locations of RA signaling. The difficulty in localizing RA actions represents a conceptual hurdle for understanding RA function in the context of those aspects of brain research that depend on the precise topographic mapping of neuronal functions.

Slow-wave sleep (SWS), which is distinguished by large-amplitude, low-frequency fluctuations in recordings of the electroencephalogram (EEG), is one of the two components of normal sleep; the other is rapid eye movement (REM) sleep, which is characterized by a low-voltage, high-frequency EEG (10). SWS emerges only slowly during postnatal life; in human babies, sleep is dominated by a form of REM sleep, but in the adult, SWS occupies about 80% of total sleep time (10). The EEG during SWS is characterized by slow waves mostly in the delta frequency range (1 to 4 Hz); sleep deprivation studies show that these delta oscillations are a reliable indicator of sleep depth and reflect accumulated sleep need (1). Maret et al. (1) studied a genetic difference in the EEG patterns of SWS between two inbred mouse strains: DBA mice show relatively less delta frequency content in their SWS EEG power-density spectrum (quantified through fast Fourier transform) than do C57BL mice. The authors traced this difference to a polymorphism in the gene that encodes the retinoic acid receptor β (RARβ), one of the nuclear receptors through which RA regulates gene transcription (3). The functional consequences of the RARβ polymorphism are reduced power in the delta frequency range of the EEG in DBA mice accompanied by a reduction in sleep depth (1). The authors confirmed the dependence of the sleep phenotype on the RARβ gene by crossing the different inbred strains with RARβ null mutants, whose SWS EEG appears to be similar to the C57BL pattern (1). Thus, increased expression of RARβ leads to a reduction in the EEG delta power density.

The different frequencies of the oscillations detectable in the EEG are due to differences in the synchronized firing rates of cortical neurons (11), rhythms that are generated within corticothalamic feedback loops or directly in the cortex (12, 13). The ontogenetic emergence of synchronized firing depends critically on a late-developing network of parvalbumin-containing inhibitory interneurons that are coupled synaptically and by gap junctions (11). The unexplainable point in the Maret study (1) concerns the mechanism by which RARβ, which is clearly not required for the generation of cortical synchrony, influences delta oscillations in the cortex: RARβ is expressed in neither the cortex nor the thalamus, at least not at detectable levels (14, 15). A similar conundrum applies to an electrophysiological analysis of the hippocampus in RARβ null mutants (16): Recordings from hippocampal slices reveal a striking absence of long-term potentiation and long-term depression, even though RARβ has not been detected in the normal hippocampus (14, 15). RARβ is highly expressed in the corpus striatum of the basal ganglia, but this structure is not known to play a prominent role in sleep or learning and memory. One possible explanation for these localization discrepancies is that RARβ is expressed in low abundance at all the missing sites and has thus not been detected there yet, but is still present in amounts that are sufficiently high to explain the strain differences (1) and the symptoms in knockout mice (16); however, other possibilities exist.

Although RA is widely recognized as playing a critical role in very early development before the existence of a functional brain, it has not been clear what role such a broadly distributed system of transcriptional control might play in the mature brain. A major gap in neuroscience knowledge concerns the basis for a range of complex human brain diseases, many of which lead to psychiatric problems. Here, a mechanism is lacking that can explain why widely distributed brain regions are vulnerable to similar environmental and different genetic perturbations (17, 18). Because normal vitamin A function depends both on a sufficient nutritional supply of the vitamin and on many different genes, disturbances in RA signaling have been suggested to function as a common denominator in the pathophysiology of some of these diseases (2, 19, 20). These suggestions have not had much influence on psychiatric research so far, however, because little specific information is available about the normal role of RA in brain maturation and function.

The Maret study (1) leaves no doubt that retinoids play a role in synchronizing cortical activity. Impaired cortical synchrony is a conspicuous feature of several poorly understood brain disorders with etiologies that involve both genetic and environmental components, such as autism and schizophrenia (18, 21). Parvalbumin-containing interneurons, which play a critical role in cortical synchronization (11), are abnormal in several psychiatric diseases (22), and they are also a major target of experimental retinoid perturbations during postnatal brain development in mice (23). Parvalbumin-positive interneurons have been implicated in cortical plasticity, for instance, in the plastic changes that occur in response to monocular visual deprivation of young animals (24); plasticity in juvenile brains is a specialized form of learning and memory. In the adult brain, retinoid perturbations are known to result in defective learning and memory (2, 25). All of these intriguing observations underscore the need to clarify how RA signaling is integrated in normal brain maturation and function.

The null mutants that lack different RA receptors demonstrate that the expression patterns of these receptors are a relatively poor indicator of the location of RA actions in the embryo, which probably reflects the widespread, overlapping expressions and functional redundancies of the different RA receptors (3). The poor correlation between the sites of abundant RARβ expression in the postnatal brain and the symptoms of the different inbred strains (1) and knockout mice (16) can be seen as just another manifestation of the same phenomenon. As already mentioned, the retinoid genes whose expression sites correlate best with the locations of RA signaling are the RALDHs (59). The only RALDH expressed within neurons of the postnatal cerebral cortex is RALDH3, which appears around birth in a band along the medial cortex and creates tangential differences in RA levels across the cortex (26). The RALDH3 territory designates the pathway for visuospatial integration and the attentional and executive networks as preferential sites of RA signaling in the postnatal cortex (26). These sites are where synchronous activity during sleep is most pronounced (27), and they are also cortical regions that are known for late maturation, extensive plasticity, and vulnerability to genetic and environmental insults (18, 2830).

Because most components of the retinoid system continue to be expressed in the postnatal brain and, often, in novel constellations compared with the embryo (2), the clinically relevant vitamin A–linked malformations are probably not the monstrous early ones (which tend to be lethal anyhow) (4, 31), but functional brain defects that arise later. All the piecemeal, accumulating evidence points to an important neurobiological role for RA in the postnatal brain, but the precise targets and mechanisms of RA signaling in brain function remain a mystery (Fig. 2).

Fig. 2.

Vitamin A, as symbolized by the carrot, is critical for brain development and function. Retinoid signaling has recently been demonstrated to modulate cortical synchrony during SWS in adult mice through mechanisms and targets that remain undefined (1).


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
View Abstract

Navigate This Article