Retinoids and vertebrate development.

In conclusion, it is obvious that many of the abnormalities in pattern formation and organ formation that result from the exogenous addition of RA during embryogenesis are related at least in part to the ability of RA to change the pattern of expression of the clusters of homeobox genes in the mammalian embryo. The majority of experimental protocols to study RA-induced changes in embryonic development have utilized treatments with excess, exogenous RA, but it has not yet been proven that endogenous RA acts as a key signaling molecule in developing embryo. Thus, we need to develop an understanding of the metabolic enzymes that control the levels of the important endogenous retinoids, including RA, in the developing embryo. The continuing study of retinoids and their receptors will provide us with significant new information about the regulatory programs that control differentiation and development. Moreover, these studies will lead to discoveries relevant to the clinical use of retinoids in the areas of cancer prevention and treatment, and dermatology.

in cultured cells or in animals. Many of these RAREs consist of a direct repeat of two motifs (5' (A/G)G(G/T)TCA) separated by a 5-base pair spacer. Such RAREs, known as DR5 elements, have been found upstream of the RAR 012 gene (13), the RAR p2 gene (14), the RAR y2 gene ( E ) , and 3' to the homeobox gene Hox 1.6 (now called Hox a-1) (16). Other RA-responsive genes have RAREs with spacers of different lengths, from 1 to 5 base pairs (for review, see Refs. 8 and 11). DR1 elements are recognized not only by RARs but also by RXRs and "orphan" receptors such as COUP-TF1, Arp-1, and ear2. Orphan receptors are members of the steroid thyroidhetinoid receptor family that have no known ligand.
While primary target genes, i.e. those directly activated by RAand its receptors, contain RAREs to which RARs and RXRs bind directly, negative regulation of gene expression by RA can occur via several different indirect mechanisms. For instance, the REX-1 gene encodes a transcription factor, and this gene is negatively regulated by RA in F9 teratocarcinoma cells. The expression of the REX-1 gene in undifferentiated F9 teratocarcinoma cells is mediated through an octamer motif (consensus sequence ATTTGCAT), and the decrease in its expression in response to RA is mediated at least in part via this same motif (17). In some other genes, the negative regulation by RA is mediated via an AP-1 binding site (e.g. Ref. 18; for review * This minireview will be reprinted in the Minireview Compendium, which will be available in December, 1994. This work was supported by grants from the National Institutes of Health. that the "receptor complex does not actually bind to the octamer or AP-1 sites but rather interferes with the activation of these genes by octamer or AP-1 transcription factors, respectively. A different mode of negative regulation by RA is exemplified by the oct-3/4 gene. The oct-3/4 gene encodes a transcription factor called "octamer 3/4," which binds to the "octamer" DNA sequence ATTTGCAT. In both teratocarcinoma cells and embryonic stem cells the expression of the oct-3 / 4 gene is inhibited upon RA treatment. The negative regulation of the oct-3/4 gene by RA does not appear to involve direct binding of the RARs. Instead, it is likely to involve both the repression of an enhancer by a mechanism not clearly defined (19) and the repression of the promoter via binding of a member of the orphan nuclear hormone receptor family (20,21). Therefore, the response of a particular cell type t o the signaling molecule all-trans-RA depends on the intracellular levels and types of RARs, RXRs, and different orphan receptors in the cells (Fig. 2).
How do the RARs interact with basal transcription factors and RNA polymerase II? While little information is available, on some promoters and in certain cell types an E1A-like protein is required as a co-activator for the TATA box binding protein TBP, which cooperates with RAR to effect RA-dependent transcriptional activation by RNA polymerase I1 (22,23). Such protein interactions may facilitate a step that is rate-limiting for the formation of a transcriptional preinitiation complex, thereby increasing the rate of transcription.
While the ligand RA has multiple receptors and all cell types assayed to date express a t least one and usually more than one of the RAR and RXR genes, vitamin A (retinol) does not bind with high affinity to either the RARs or the RXRs. Nevertheless, derivatives of retinol are present in cells and have important functions with respect to cell growth regulation. The derivative 14-hydroxy-4,14retroretinol promotes the growth of B lymphocytes in culture and the activation of T lymphocytes by antigen receptor-mediated signals (24, 25); another retinol derivative, anhydroretinol, antagonizes the effects of 14-hydroxy-4,14-retroretinol and thus exerts inhibitory effects on lymphocyte growth (26). Whether these retinol derivatives exert their effects through binding to novel receptors is a crucial question that remains t o be answered.

The CRABPs
In addition to the RARs and RXRs that possess both RA and DNA binding domains, a second type of high affinity RA binding protein exists in many different cell types; two such proteins, called CRABP-I and CRABP-11, have been purified and cloned to date (see Ref. 27 for review). Although the functions of the CRABPs are not entirely understood, biochemical and genetic data have provided evidence that the higher the level of the CRABP-I in the cytoplasm, the less sensitive the cell is to a given external concentration of all-trans-RA (28). How is this accomplished? Since the C W P -I protein can enhance the enzymatic conversion of RA to more polar, oxidized derivatives (29, 301, it appears that the CRABP-I protein reduces the amount of RA available to regulate gene expression in the nucleus. In addition, CRABPs may direct metabolism of retinoids to derivatives that have important biological activities. The distribution of the C W P -I and -11 proteins is somewhat more limited than that of the RARs and FtXRa (see Ref. 27 for review), but these cytoplasmic RA binding proteins are present in a variety of different cell types and must be taken into account when responses of a given cell type to RA are assessed.

Regulation of Cell Difirentiation by Retinoic Acid
The regulation of the process of differentiation in various types of cells by RA involves the ability of this signaling molecule to alter the expression of a wide variety of different types of genes. For example, in the murine teratocarcinoma stem cell differentiation model system RA addition results in profound changes in the levels of transcripts of several different types of transcription factors within a few hours to 1-2 days. Many of these effects are primary in that they are mediated directly by receptor binding to FtAFW. s that control gene expression in response to RA. At later times, RA influences the expression of genes that encode glycoproteins involved in the production of the extracellular matrix, proteases such as tissue plasminogen activator, intermediate filament proteins such as keratins, and the genes for a number of different growth factors and their receptors (for review, see Ref. 6). Many of these changes are secondary and do not involve the direct binding of an RA-receptor complex to a RARE. It is very likely that the rapid changes in the expression of a number of transcription factor genes brought about by RA are critical for differentiation to proceed. One such family of transcription factors regulated by RA is the homeobox genes.

Retinoic Acid and Regulation of Eomeobaz: Genes
Homeobox genes are transcription factors, first identified in Drosophila, that bind DNA through a helix-turn-helix motif called the homeodomain or homeobox. That homeobox genes are involved in both invertebrate and vertebrate morphogenesis has been shown Each box re resents a with a uestion mark indicates that no gene has been discovered in vertebrates. each gene in the cluster is transcribed 6' to 3', but the entire cluster is activated in a 3' to 5' direction during mammalian development. by many laboratories (for review, see Ref. 31). Four clusters of homeobox genes have been identified in the mouse and human genomes, each cluster on a different chromosome (Fig. 3). "his laboratory has demonstrated that the homeobox gene at the 3' end of the A cluster on mouse chromosome 6, Hox a-1 (formerly Hox 1.61, is induced very rapidly in a protein synthesis-independent fashion in response to RA treatment of F9 embryonic teratocarcinoma cells (32). In addition, Simeone et al. (33,34) have demonstrated that human teratocarcinoma cells (NTP-Dl) cultured in RA concentrations ranging from lo4 to lod M differentidy activate genes of the human Hox B homeobox gene cluster. Genes located in the 3'-half of the cluster were induced at peak levels by lo4 M R A , whereas a concentration of M was required to activate genes near the 5' end of the cluster. This observation suggests that the concentration of RA to which a cell is exposed will greatly influence which homeobox genes are activated. Remarkably, these researchers also demonstrated that genes in each one of the four human Hox loci are activated by RA in a sequential order temporally, colinear with their 3' to 5' arrangement in the cluster; 3' Hox genes were activated more rapidly after RA addition, whereas genes at the 5' end of the cluster responded at progressively later times after RA addition (33, 34). Similarly, Hox genes at the 3' ends of the clusters are expressed earlier in embryonic development and in more anterior regions, while Hox genes closer to the 5' ends of the clusters are expressed at later times in development and in posterior regions of the embryo (Fig. 3).
Thus, one of the key features in the regulation of vertebrate development by RA may be RA' s ability to activate specific homeobox genes. We have shown that approximately 5 kb 3' of the Hox a-1 gene transcriptional start site there is a RA-inducible enhancer which contains an RARE of the D M type that specifically binds RARs (16). This enhancer is necessary and sufllcient for the RA activation of the Hox a-1 gene, the 3'-most gene in the A cluster.
Thus, we have shown that there is a direct molecular link between RA, RARs, and the activation of a specific homeobox gene. How other Hox genes located in more 5' positions in the cluster are regulated by RA is not known, but the enhancer we have identified may function to regulate other Hox A cluster genes in addition to the Hox a-1 gene.
A pressing question at the current time concerns the nature of the target genes of these homeobox proteins (for review see Ref. 351, a question for which we have very few answers. We do have some preliminary evidence that genes involved in cell shape determination are targets of the Hox a-1 homeobox protein (36). The identification of these Hox target genes is a top priority.

Do the RARs Have Distinct Roles in Gem Regulation?
The question of the specificity of the different RARa with respect to the regulation of gene expression is a significant one. All three receptors bind all-trans-RA with similar &ties, and most cell types express more than one RAR gene. Does each receptor type have specific functions? The use of homologous recombination has permitted the isolation of mice in which both alleles of a particular receptor type have been disrupted. However, disruption of the RAR a or the RAR y gene did not result in major morphological abnormalities or embryonic lethality, suggesting that some degree of functional redundancy exists among the different types of RARs (37-39). The effects of the disruption of RAR genes by homologous recombination were also assayed in a less complex cell culture system. In F9 cells the disruption of the RAR y gene leads to the loss of the appropriate activation of the Hox a-1 gene in response to RA, as well as to the loss of the RA induction of the laminin B1 and collagen type N genes, two genes that are normally induced by RA in wild type F9 cells at late times (approximately 2 days) after RA addition (40). Disruption of both RAR a alleles leads to the loss of the RA-associated activation of both the Hox b-1 (formerly called Hox 2.9) gene and the C W P -I 1 gene, which is normally seen in F9 wild type.' These results are most easily interpreted as indicating that the three RAR types do exhibit different affinities for the RAREs andlor for other regulatory proteins bound near the RAREs on enhancers in different primary RA target genes. Additional gene disruption experiments involving other genes in the RA differentiation pathway in cell culture model systems should lead to a better understanding of the functions of the RARs, FXRs, and their primary transcription factor gene targets in cell differentiation.

Endogenous Retinoids in Embryos: Are Retinoid Gradients
Involved in Pattern Formation? It has been hypothesized that during development a small number of cells in specific regions of the embryo can become specialized such that these cells become a source of a signaling molecule that can then spread to surrounding tissues. This results in a concentration gradient of the signaling molecule. Such a hypothetical signaling molecule whose concentration is "read" by cells to determine their spatial position in the embryo relative to the source is termed a "morphogen." There has been much speculation about the possibility that concentration gradients of biologically active retinoids exist along the embryonic anteroposterior axis during the critical periods in embryogenesis during which major organs such as limbs and heart are forming. If exogenous RA can activate homeobox genes and other genes encoding transcription factors involved in cell differentiation, can RA function as a "morphogen?" To address this question it is certainly important to assess whether endogenous retinoids are present in the developing embryo and how they are distributed. It is therefore somewhat surprising that relatively few direct measurements of endogenous RA in the developing embryo have been performed. Thaller and Eichele (41) directly measured all-trans-RA in the chick limb bud by HPLC. Embryonic tissues have also been assessed for their levels of biologically active retinoids using more indirect assays. Using these assays, several groups have demonstrated that sources of RA (regions where RA is produced) are somewhat restricted in the developing embryo, For instance, Hensen's node, a region in the chick embryo that is similar to the Spemann organizer of amphibians and the primitive streak in the mouse in that it can induce a second body axis when grafted into a host embryo, is one source of RA or other active retinoids (42,43). Other sources of RA or a biologically active derivative are the floor plate of the neural tube (44) and the alar regions of the cervical and lumbar spinal cord (45). Chen et al. (46) reported that active retinoids were present at higher concentrations in the posterior rather than anterior positions of neurula stage Xenopus embryos.
In summary, there is evidence that biologically active retinoids are present in developing embryos. Either a spatial or temporal RA gradient would allow cells along the anteroposterior axis of the embryo to respond to different concentrations of active retinoids. However, as retinol, RA, and their metabolites are relatively unstable, easily oxidized during their extraction from tissues and difficult to separate by HPLC, much caution must be exercised with respect to the identification and quantitation of endogenous retinoids in embryonic cells and tissues. It is also important to note that the vast majority of experiments that have been performed involve the application of exogenous RA to developing embryos and tissues rather than the measurement or manipulation of endogenous RA levels.

Effects of Exogenous RA on Vertebrate Development: Limb, Hindbrain, and Craniofacial Abnormalities
The effects of RA added exogenously during embryogenesis are extraordinary, and only a small number of these effects can be discussed here. RAcan exert profound effects on pattern formation, i.e. the development of the embryo in a spatially organized fashion. Thus, RA is considered to be a potent teratogen. However, the fact that RA can alter the development of so many different structures in the embryo also points to the fact that retinoids are likely to be key regulatory molecules. A recurring theme in this section is that RA alters Hox gene expression in the organs and tissues that develop abnormally in response to RA.
One of the most thoroughly studied structures is the developing limb. A tremendous amount of interest was generated by the observation a decade ago that local application of all-trans-RA to the anterior portion of the chick wing bud could induce a mirror image duplication of digits (47). How are these extra digits formed? It is known that genes of the Hox D (formerly Hox 4) cluster are involved in limb pattern formation and that several of these genes (d9, d10, d l l , d12, and d13) are expressed primarily in the posterior portion of the developing vertebrate limb under normal circumstances. However, when RAis locally applied to the anterior margin of the developing limb bud, the expression of Hox D cluster genes expands to include expression in the anterior half of the wing bud as well as the posterior half (48). These experiments indicate that RA can activate genes of the Hox D cluster and that the activation of these Hox D cluster genes in the anterior portion of the developing limb bud, a portion of the bud in which they would not normally be expressed, is part of the mechanism by which RA application to the anterior portion of the limb bud leads to the formation of extra digits. The recently described vertebrate "segment polarity" gene called sonic hedgehog is also normally expressed only in a small region of cells at the posterior margin of the limb bud, but local application of RA to the anterior portion of the limb bud induces a high level of sonic hedgehog expression in the anterior limb bud (49). Thus, both the anterior expression of the sonic hedgehog gene and the genes in the Hox D cluster are thought to be important in the conversion of the RA signal into extra digits generated from cells in the anterior portion of the limb bud. Remarkably, the Hox D genes appear to be directly involved in specifying digit identity, as it has been reported that the ectopic overexpression of the Hox d-11 gene (in the absence of RA) can change a digit 1 into a digit 2 (50).
Excess exogenous RA can cause craniofacial malformations and malformations in the hindbrain region (51-53). The ability of RA to alter the segmental organization of the developing hindbrain results in part from its ability to alter the pattern of expression of Hox genes, and in particular the Hox b-1 gene (54). As the alteration in the expression of the Hox b-1 gene in response to RA is quite rapid and the Hox b-1 gene encodes a transcription factor, it is likely that the aberrant expression of Hox b-1 results in some of the morphological defects in the hindbrain observed in the embryos treated with RA (54, 55).
Treatment of mice with RA on day 8 of gestation also leads to a high incidence of spina bifida (failure of neural tube closure). This condition most likely results from excessive cell death induced by RA in the tissues that underlie the posterior neural plate, such as mesenchymal tissues in the region of the primitive streak, rather than from direct effects of RA on the neuroepithelium (56). Thus, some of the teratogenic effects of RA may be caused by excessive cell death induced by RA in regions in which some programmed cell death would normally occur.

Effects of RA on Embryonic Skin Differentiation
Retinoids can alter the fate of embryonic epithelia. For instance, in embryonic chicken skin explant cultures, RA addition to the culture medium leads to a decrease in the size of the feather buds or the transformation of the buds into scalelike structures (57). If RA is added locally to the cultured skin via an RA-soaked anion exchange bead, a circular zone of inhibition of feather formation is observed, with a rim of disoriented feather buds around this zone of inhibition. How does RA effect this change in feather bud patterning? Amazingly, in the developing skin there are anteroposterior gradients of expression of homeobox proteins in the individual developing feather buds with the highest expression of Hox proteins on the anterior side of each bud. These gradients are "microgradients" in the sense that there is a Hox expression microgradient in each developing feather bud, and there are approximately 20,000 feathers per bird. After growth in the presence of RA, the Hox protein staining of the feather buds is much more diffuse (57), like that in normal scales. Thus, excess, exogenous RA disrupts the normal Hox codes of skin appendages in the developing chick.
In the mouse, all of the genes of the Hox B locus except Hox b-1 are expressed in the skin of embryos at days 16-18 of gestation, a time when the epidermis is differentiating and undergoing stratification. Additionally, many Hox genes in the other clusters (A, C, and D) are expressed at this time in development, as well as in adult skin (58). The data strongly suggest that homeoboxgenes have regulatory functions both in the early differentiation of embryonic skin and in the differentiation of epidermal cells in the adult.

Effects of RA on Inner Ear Development
Exposure of the cochlea, an organ of the mammalian inner ear that develops from the otic vesicle, to RA in vitro at mouse embryonic day 14 resulted in a large increase in the number of cells that develop as hair cells (59,60). Rather than just a single row of inner hair cells, the vertebrate sensory epithelia that send neural signals to the central nervous system, two rows of inner hair cells developed. Moreover, in the presence of RA up to 11 rows of outer hair cells developed rather than the three to five rows of outer hair cells normally observed; outer hair cells do not connect to afferent nerve fibers but provide mechanical amplification of acoustic signals. These results indicate that retinoids have an important function in the development of the cochlea and may be involved in the induction of differentiation of presensory cells to become hair cells and supporting cells. Strikingly, RA can stimulate the regeneration in vitro of mammalian auditory hair cells in cochleas that have been explanted after treatment with an ototoxin that destroyed the hair cells (61). It was previously thought that these sensory hair cells of the cochlea did not regenerate and that most hearing loss resulting from the death of sensory hair cells was irreversible. The mechanisms involved in these effects of RA on sensory epithelial cell regeneration are not understood.

Summary
In conclusion, it is obvious that many of the abnormalities in pattern formation and organ formation that result from the exogenous addition of RA during embryogenesis are related at least in part to the ability of RA to change the pattern of expression of the clusters of homeobox genes in the mammalian embryo. The majority of experimental protocols to study "induced changes in embryonic development have utilized treatments with excess, exogenous RA, but it has not yet been proven that endogenous RA acts as a key signaling molecule in developing embryo. Thus, we need to develop an understanding of the metabolic enzymes that control the levels of the important endogenous retinoids, including RA, in the developing embryo. The continuing study of retinoids and their receptors will provide us with significant new information about the regulatory programs that control differentiation and development. Moreover, these studies will lead to discoveries relevant to the clinical use of retinoids in the areas of cancer prevention and treatment, and dermatology. a n d Alex Langston for critically reading this manuscript prior to publication,