Retinoid Metabolism: New Insights

Abstract

Vitamin A (retinol) is a critical micronutrient required for the control of stem cell functions, cell differentiation, and cell metabolism in many different cell types, both during embryogenesis and in the adult organism. However, we must obtain vitamin A from food sources. Thus, the uptake and metabolism of vitamin A by intestinal epithelial cells, the storage of vitamin A in the liver, and the metabolism of vitamin A in target cells to more biologically active metabolites, such as retinoic acid (RA) and 4-oxo-RA, must be precisely regulated. Here, I will discuss the enzymes that metabolize vitamin A to RA and the cytochrome P450 Cyp26 family of enzymes that further oxidize RA. Because much progress has been made in understanding the regulation of ALDH1a2 (RALDH2) actions in the intestine, one focus of this review is on the metabolism of vitamin A in intestinal epithelial cells and dendritic cells. Another focus is on recent data that 4-oxo-RA is a ligand required for the maintenance of hematopoietic stem cell dormancy and the important role of RARβ (RARB) in these stem cells. Despite this progress, many questions remain in this research area, which links vitamin A metabolism to nutrition, immune functions, developmental biology, and nuclear receptor pharmacology.

A. Introduction and Focus

Vitamin A (all-trans retinol, ROL), a lipid-soluble micronutrient, is absolutely required for mammalian life. Without vitamin A mammals develop many clinical pathologies, from xerophthalmia (dryness and inflammation of the conjunctiva and cornea of the eye) to a greatly increased susceptibility to infections (1–4). Mammals must obtain vitamin A from their diet; mammals can’t synthesize vitamin A (5). Vitamin A is found in meat and dairy products, and pro-vitamin A sources, especially beta-carotene, are found in vegetables (6).

Presumably, because of the essential roles of vitamin A and because periods of nutrient limitation, including famines, have occurred throughout the course of history, ingested vitamin A is delivered, via chylomicrons (small particles of triglycerides, phospholipids, proteins, and cholesterol) to the liver for longer-term storage (7, 8). In the liver vitamin A is converted to retinyl esters (REs) by the enzyme lecithin:retinol acyl transferase (LRAT), primarily in the stellate cells, after vitamin A is transferred to these stellate cells from hepatocytes (9–13). These REs are stored for future use so that seasonal and other variations in dietary supplies don’t prevent the target cell types that require vitamin A (most cell types in the body) from obtaining the vitamin A needed to carry out appropriate molecular activities. The term ‘retinoids’ is used to describe vitamin A and its various metabolites, including retinoic acid (RA).

Homeostasis is defined as a process by which an organism adjusts to external, sometimes hostile stimuli, to maintain stable signaling internally. Homeostasis of retinoids requires multiple metabolic processes, including the storage and release of vitamin A from the liver retinyl ester (RE) stores, the transport of vitamin A into various cell types in the body (7), the metabolism of vitamin A to all-trans retinoic acid (RA), and the further oxidative metabolism of RA, to allow for the proper levels of biologically active retinoids to be achieved in specific cell types in the body. Since RA regulates stem cell differentiation and is active at nanomolar concentrations in cells, levels of this signaling molecule must be precisely regulated in a spatiotemporal manner during embryogenesis and throughout life (14). Of course, it follows that disruption of vitamin A homeostasis can likely result in pathologies. For example, overnutrition seen in obesity can negatively affect vitamin A metabolism and signaling (15), leading to diminished immune functions, increased susceptibility to respiratory viral infections, and poor responses to vaccinations (16).

We know, in both animal models and in humans, that levels of RA that are too high or too low can cause major pathological changes both during development (17–19) and in the adult (20–23). This review focuses on how retinol is metabolized to more bioactive retinoids, such as all-trans retinoic acid (RA), in specific cell types, at the correct times and in the correct amounts. The role of the bio-active retinoid, 11-cis-retinaldehye, in vision and the metabolism of vitamin A to 11-cis-retinaldehyde are not covered here, but an excellent review covers this topic (24). This review will focus on selected, recent literature on the topic of vitamin A (retinol) metabolism and the roles of this metabolism in maintaining cell function and regulating differentiation.

B. Retinoids Act Primarily, But Not Exclusively, By Binding To Retinoic Acid Receptors on DNA

Retinoic acid (RA), a metabolite of vitamin A, is an endogenous agonist for the retinoic acid receptors α, β, and γ (RARα, β, and γ (RARA, RARB, RARG); also named NR1B1, NR1B2, and NR1B), which are members of the nuclear receptor protein family. Depending on the context, these nuclear receptors can act as either repressors or activators of gene transcription (Fig. 1). They possess both a ligand binding domain and a DNA binding domain. The endogenously produced ligand, RA, induces a ‘molecular switch’ that involves changing the regulatory proteins bound to the RARs, often turning the RARs, which are dimerized with retinoid X receptors (RXRs) α, β, and γ, from repressors to activators of transcription (25). Transcriptional activation takes place via binding of the RA-bound RARs to enhancers containing RAREs (retinoic acid response elements) that exist in promoters and/or introns of primary RA target genes (those genes to which the RARs directly bind) such as Hoxa1, Hoxb1, and RARβ. (26–31).

Figure 1. Metabolism of Vitamin A (Retinol) in Cells and Major Actions of Retinoic Acid in Cells.

All-trans retinol is oxidized to retinaldehyde and to retinoic acid; further oxidation by CYP26a1, b1, c1 can also take place. Retinoic acid moves to the nucleus where it acts as an endogenous agonist for the retinoic acid receptors (RARs) α, β, and γ when these receptors are in a heterodimer with the RXRs α, β, and γ. Then, transcriptional activation or repression of primary target genes, those genes with RAREs (retinoic acid response elements/enhancers to which the RARs bind) takes place. RARα can also function in the cytoplasm, where in the absence of RA it inhibits the translation of target RNAs in neurons; in this case the ligand, RA, blocks this effect. 4-Hydroxy-RA and 4-oxo-RA both can act as endogenous agonists for the RARs in model systems, and recent data indicate a role in animal models in hematopoietic stem cells.

Although most studies of RARs have focused on their actions in the nucleus, recently RARα has also been shown to act in the cytoplasm of cells in the hippocampus of the brain as an ‘RA-induced postsynaptic regulator of protein synthesis’ (32). Additionally, RA can activate extracellular signal regulated kinase 1/2 (ERK1/2) via an RAR independent mechanism (33).

RARs α, β, and γ all bind RA, but with different affinities (34), and synthetic retinoids that selectively bind one type of RAR have also been synthesized for use in the treatment of a number of different diseases (35). 4-Oxo-retinol is also an agonist for the RARs (36). Pertinent to this discussion, the RARs α, β, and γ exert different functions in different cell types and tissues, indicating some unique functions of the three RARs in animals, despite the fact that all three RARs bind RA and are expressed in most cell types (37–42). Furthermore, RA can signal within the same cell or can be secreted to influence cells nearby. In fact, RA ‘morphogen’ gradients assists in patterning of organs during development (43). I will focus on how retinoids are metabolized to produce the biologically active derivatives that are required for the development or function of specific cell types.

C. Metabolism of Vitamin A to Retinaldehyde Reveals a Sexually Dimorphic Impact on Energy Metabolism and the Key Roles of Retinoids in Development

The short-chain dehydrogenases/reductases (SDR) enzyme family (44–46), consists of many members, some of which can metabolize retinol to retinaldehyde (retinal, RAL) (Fig. 1). One member of this SDR family, the NAD-dependent enzyme retinol dehydrogenase (RDH10), carries out the oxidation of retinol to retinaldehyde. RDH10 is required for survival during embryogenesis, and without RDH10, there is a major deficiency in RA, resulting in the failure of many developmental processes (47–52). Genetic studies in mice demonstrate that RDH10 regulates energy expenditure (53). Notably, heterozygous RDH10+/− male mice show relatively small decreases in tissue RA (<25%), but, when placed on a high-fat diet, males decrease fatty-acid oxidation and develop insulin resistance In contrast, female RDH10+/− mice do not show these changes, though both males and females exhibit increased adiposity on a high-fat diet compared to wild-type mice. Additionally, male running endurance decreases, while for females, running endurance increases. These sexually dimorphic effects are in part the result of different amounts of reduction in RA levels in these RDH10+/− male vs. female mice, since low-dose RA, given chronically, can correct these metabolic pathologies (54, 55).

At the molecular level, RDH10 transcription is increased by the transcription factor FoxO1 in cultured liver cells when serum is removed (56). In cultured embryonic stem cells oxidation of retinol by RDH10 is required for ethanol to induce primary RA-target genes such as Hoxa1, Cyp26a1, and Stra6. Intriguingly, ethanol also increases RDH10 mRNA levels by a mechanism not yet clear (57).

The SDR enzyme dehydrogenase/reductase SDR member 3 (DHRS3) is an NADPH-dependent enzyme which, conversely, reduces retinaldehyde to retinol and causes embryonic lethality because of excess RA (58–61). An association of RDH10 with DHR3 exists and this association facilitates the activities of each of these proteins (62). Various other SDR family members have been reported to exhibit oxidoreductase activity using retinoid substrates in numerous tissues. For instance, Dhrs9 plays a critical role in the hair follicle cycle and in inhibiting squamous cell carcinomas (63).

D. Oxidation of Retinaldehyde (RAL) to Retinoic Acid Regulates Cell Development and Function

Following the conversion of retinol to retinaldehyde, RAL is irreversibly oxidized to RA by members of the ALDH1A family of enzymes, ALDH1a1, ALDH1a2, and ALDH1a3 (formerly RALDH1,2,3) (64, 65). These enzymes ALDH1a1, 2, and 3 also catalyze the oxidation of aliphatic aldehydes, such as hexanal and 4-hydroxy-2-nonenal. Enzymes in this family are expressed at different levels in various cell types and most likely play tissue-specific roles in the production of RA from RAL (66–70). The key role of ALDH1a2 in development is underscored by the report of humans with two different missense variants in ALDH1a2 that result in a severe, multiple congenital anomaly syndrome with neonatal lethality (71). Other members of this ALDH1A family include ALDH1a1, which contributes to the synthesis of RA in adipose tissue (72–74), and ALDH1a3, which is active in cardiac atrial appendage progenitor cells (75). As might be expected, higher RA levels generally reduce the expression of ALDH1a2 in RA-target cells (58). A vitamin A deficient diet causes an increase in ALDH1a2, a decrease in ALDH1a1, and no significant change in ALDH1a3 transcript levels in murine kidneys (76). These changes in ALDH1a2 indicate that the synthesis of RA can be modulated in response to intracellular levels of RA; when RA levels are low, ALDH1a2 transcripts are increased in some organs, such as the liver and kidney.

Since the reaction catalyzed by ALDH1a2 is such a critical step in the synthesis of RA from RAL (Fig. 1), it is not surprising that the ALDH1a2 gene is transcriptionally controlled by several different transcription factors. Information is available about how ALDH1a2 transcription is regulated in some tissues, both during development and in the activation of dendritic cells, key cells that initiate and modulate immune responses (77). For instance, the T-box transcription factor, TBX5, controls ALDH1A2 transcription in the foregut lateral plate mesoderm during development, ultimately generating RA and activating endodermal Shh (Sonic hedgehog) expression (78). ALDH1a2 levels are also increased by estrogen in human endometrium (79).

E. The Metabolism of Retinaldehyde to Retinoic Acid by ALDH1a2 in Subtypes of Dendritic Cells Modulates Immunity

Dendritic cells (DCs) capture and present antigens to naïve T-cells, generating antigen-specific immune responses. Here, we focus on the DCs in the intestine, where many advances in our understanding of the actions of ALDH1a2 on DC function have been elucidated. The intestinal immune system exhibits some differences from the systemic immune system. Cells in the intestinal tract are exposed to large numbers of antigens, many of which are from harmless substances, and Treg cells, a type of T-cell, are primarily responsible for immunosuppression in the intestine (80). During homeostasis a subtype of DCs, CD103⁺ DCs, assists in maintaining intestinal tolerance by migrating to mesenteric lymph nodes, the lymph nodes between layers of the mesentery, a membrane that attaches the intestine to the abdominal wall. There these intestinal DCs a) induce the gut-homing receptors CCR9 and α4β7 on T and B-cells, b) promote development of Foxp3⁺ regulatory T-cells, and c) promote IgA class switching in naïve B-cells (81, 82) (Fig. 2). Notably, RA, produced in CD103⁺ DCs by transcriptional activation of ALDH1a2, plays a major role in regulating these processes (83–87). Furthermore, in the presence of RA, naïve T-cells can no longer properly differentiate into T-helper 17 (Th17 cells). Additionally, T-cell production of interleukin-4 and interferon-γ, cytokines that block T-reg differentiation, is reduced by RA (88–90). In aggregate, these effects of DCs that produce RA in the intestine influence the balance between Treg and Th17 cells, favoring Treg cells. Importantly, in contrast to this CD103⁺ subtype of DCs in the intestine, DCs in the spleen and lymph nodes draining the skin generally don’t exhibit high ALDH1a2 expression (91, 92).

Figure 2. Vitamin A Actions in the Intestine Influence Immunity.

Dendritic cells (DCs) from the bone marrow produce some ALDH1A2 and migrate to the intestine. There, Vitamin A from the diet is metabolized by intestinal epithelial cells (enterocytes) into retinoic acid (RA). By association with the epithelial cells that produce RA, these intestinal DCs, characterized by the expression of CD103, produce more ALDH1a2, as RA can activate ALDH1A2 transcription. These intestinal DCs then generate RA for IgM+ B-cells present in the lamina propria of the intestine, allowing them to undergo class switching to IgA. In the mesenteric lymph nodes these DCs, after migration from the lamina propria, provide RA to B cells. B-cells recirculate from the mesenteric lymph nodes via the bloodstream and enter the intestine using α4β7, becoming IgA secreting plasma cells. These high ALDH1A2-expressing DCs also drive T-cells to differentiate into T-Reg cells rather than TH17 T-cells. Panel on the right: In the DCs RA from the enterocytes binds to the RAR:RXR heterodimer and, along with GM-CSF, drives transcription of the ALDH1a2 gene, which leads to more RA synthesis from retinaldehyde. The protein tristetraprolin is involved in degrading ALDH1A2 mRNA, so in the absence of tristetraprolin, an RNA-binding protein, more RA is present since the ALDH1A2 transcript is more stable in the DCs.

How is the ALDH1a2 gene transcriptionally activated in CD103⁺ DCs to produce the RA that influences intestinal immunity? First, bone marrow-derived DCs express some ALDH1a2 that produces RA, and RA then promotes a mucosal differentiated DC phenotype (93, 94). In bone marrow derived DCs a heterodimer of the transcription factors PU.1 and IRF4 bind at −1961/−1952 of the ALDH1a2 promoter to activate transcription (95). These mucosal DCs continuously arrive in the intestine from the circulation (96). In the intestine, epithelial cells (enterocytes) absorb and metabolize vitamin A from the diet. The RA produced by this enterocyte metabolism induces further differentiation of these DCs in the intestine and initiates greater induction of ALDH1a2 transcription in these intestinal DCs (91, 97–99) (Fig. 2). There is evidence that the induction of ALDH1a2 transcription occurs via RARβ binding to two potential RAREs located within 1 kb 5’ of the ALDH1a2 transcription start site (100). GM-CSF (granulocyte macrophage-colony stimulating factor) and IL-4 plus RA can greatly increase ALDH1a2 transcripts in DC cells (101). Cooperative binding of the transcription factor Sp1 and the RARα/RXR heterodimer, binding at a RARE half site in the ALDH1a2 promoter just upstream of the transcription start site, is required for RA and GM-CSF to enhance ALDH1a2 transcription (102) (Fig. 2). Notch signaling is also required for the transcriptional activation of ALDH1a2 in DCs (103). Thus, RA is both a product of ALDH1a2 and a transcriptional activator of ALDH1a2 in these DCs. Furthermore, ALDH1a2 activity in these DCs declines with age, suggesting that the increased inflammation associated with aging results in part from a reduction in antigen specific Treg cell production (104). A key question that remains unanswered is how vitamin A and/or RA is transferred from the enterocytes into DCs to stimulate their differentiation in the intestine and to initiate RA production from ALDH1a2. Similarly, we don’t understand how the RA produced by DCs is transferred to and transported into T-cells to stimulate their differentiation into Treg cells. B and T-cells in the immune system express both RARα and RARγ constitutively so they can respond to the RA produced by DCs (105). RARβ actions in enterocytes are essential for both the generation of gut homing CD4+ T-cells and IgA-producing B-cells, but not for production of Treg cells in the intestine (106).

The protein tristetraprolin, encoded by the Zfp36 gene, has many activities, including reducing the expression of IL-23 and IL-6, mediators of inflammation (107). Zfp36 null mice, therefore, exhibit a multiorgan inflammatory syndrome similar to spondyloarthritis (108). Recently it was shown that ALDH1a2 is one direct target of tristetraprolin, but that the increase in ALDH1a2 mRNA level observed in the absence of tristetraprolin (in Zfp36-deficient DCs) occurs via a non-transcriptional mechanism. Tristetraprolin is an AU-rich element (ARE) RNA binding protein, and tristetraprolin reduces the stability of ALDH1a2 mRNA in DC cells by binding to AREs in its 3’UTR and targeting ALDH1a2 mRNA for degradation by recruitment of deadenylases (109). When tristetraprolin is absent, ALDH1a2 mRNA is more stable, resulting in increases in T-reg homeostasis because of greater RA production by the DCs (110). Thus, ALDH1a2 is regulated both at the level of transcriptional activation and mRNA stability in intestinal DCs to control immune tolerance (Fig. 2).

It is likely that ALDH1a2 is activated in other cell types by inflammatory signals and cellular stress and/or damage, as is the case in the kidney (111) and in stellate cells of the liver, which also increase ALDH1a2 transcripts and, in the presence of DCs and TGFβ, induce FoxP3+ Treg cells (112). ALDH1a2 is also induced and required for zebrafish fin and heart regeneration (113, 114). More research very likely will reveal the regulation of ALDH1a2 in other cell types.

F. Oxidation of Retinoic Acid by Cytochrome P-450 Enzymes and the Regulation of These Enzymes in Numerous Cell Types.

Further oxidation of the RA ionone ring via hydroxylation at the C-4 position is catalyzed primarily by the CYP26 family of enzymes, CYP26A1, B1, and C1 (115–121) (Fig. 1). Evidence indicates that retinoids are the only substrates for this CYP26 family of enzymes (118, 119). Oxidation of 4-hydroxy- to 4-keto-RA is also mediated by these enzymes (122, 123). Oxidized metabolites of RA are also conjugated by glucuronosyltransferases in Phase II metabolism for clearance (124).

The Cyp26 enzymes are key proteins for regulating the internal levels of RA in cells. CYP26A1 is transcriptionally activated by RA via RAR/RXR binding to two RAREs in the CYP26A1 promoter (125, 126). The CYP26A1 gene can also be transcriptionally activated by RA metabolites (127), suggesting a role for CYP26A1 in RA elimination. Moreover, the transcription of CYP26A1 is controlled by epigenetic regulators, such as histone deacetylases, in addition to RA (128–130).

Knockouts of CYP26A1 and CYP26B1 are embryonic lethal (131), but if these enzymes are knocked out postnatally, dermatitis, splenomegaly, and inflammation ensue (132). These postnatal knockout mice exhibited increased RA levels in the liver, serum, skin, and intestines, again underlining the importance of these enzymes in regulating the levels of RA in the body (132). In humans, some individuals with missense mutations in CYP26B1 have survived, but with complex developmental anomalies (133, 134). Inhibitors that are selective for CYP26B1 have been generated and may be of use for treatment of keratinization disorders in the skin, eg. Darier disease (135).

During development, Cyp26a1 and Cyp26b1 play major roles in establishing gradients of RA and, thus, in regulating the differentiation of various stem cells (136–140). For instance, Cyp26a1 is required for normal hindbrain patterning in embryos (141). The absence of Cyp26a1 results in altered spinal motor neuron subtype identity when embryonic stem cells are differentiated into neuronal cells, indicating a key role for retinoids in directing neural differentiation (142). Normal testis development requires CYP26B1 since without CYP26B1 activity, the reproductive tract is feminized (143, 144). Additionally, during lung development the lack of Cyp26b1 activity results in fewer alveolar type 1 cells, failure of alveolar inflation, and postnatal lethality in rodents (145). Spleen organogenesis and growth are impaired in mice that lack Cyp26b1 (146). Cyp26b1 plays a key role in heart valve morphogenesis (147). In the developing prefrontal cortex of the brain a frontal to temporal gradient of RA exists, and RA signaling is limited to the pre-frontal cortex by CYP26B1, which is highly expressed in the prospective motor cortex. RARβ and CYP26B1 work together to produce proper patterning of the prefrontal and motor areas and pre-frontal cortex-mediodorsal thalamus connections (39, 148). In summary, the regulation of CYP26B1 and the ability of CYP26B1 to metabolize RA is probably critical for many aspects of development, especially in terms of establishing transient RA concentration gradients.

How is the transcription of Cyp26b1 regulated? Given the complex RA gradients that Cyp26b1 is involved in generating, the expectation is that the regulation of Cyp26b1 transcription is complex, and more data are needed in this area of research. For instance, during gonad development the transcription factor Steroidogenic Factor 1 (SF1) and Sex-Determining Region Y-Box 9 (Sox9) positively regulate Cyp26b1 transcription, allowing for RA degradation and thus blocking germ cell differentiation in response to RA. Conversely, the Notch target Hairy/Enhancer-of-Split Related with YRPW Motif (Hey1), a repressor, binds to the Cyp26b1 promoter in Sertoli cells, blocking Cyp26b1 expression (149). In most cell types the regulation of Cyp26b1 transcription is not well understood, but it is likely that it is complicated and cell type specific.

G. Biological Activities of 4-OH-RA and 4-Oxo-RA and the Maintenance of Hematopoietic Stem Cells

An open question in this field is whether the other retinoid metabolites, e.g., 4-OH-RA and 4-oxo-RA, are required for any cellular processes. These RA metabolites have biological activity (150–152), but it is not clear that they play any essential roles in terms of the regulation of gene expression in mammals. In general, these more oxidized metabolites of RA are thought to be degradation products on the way to the elimination of RA from the body (153). There is evidence that these oxidized metabolites of RA are not involved in mouse development (131), but, in this report, Cyp26b1 was intact, making it difficult to rigorously determine if 4-oxo-RA generated by Cyp26b1 has any developmental functions.

Recent studies on the effects of dietary vitamin A on dormancy in hematopoietic stem cells (HSCs) have identified some intriguing roles of RA and its oxidized metabolites. RA signaling can maintain quiescence in hematopoietic stem cells by reducing the rate of protein translation and lowering levels of reactive oxygen species (ROS), and this is associated with specific expression of RARβ, (and not RARα and RARγ,) in these dormant HSCs (41). Moreover, these dormant HSCs can metabolize ROL to RA in a cell autonomous fashion (41). Recently, 4-oxo-RA, produced by Cyp26b1, was shown to be essential for HSC maintenance, and again, RARβ was also shown to be required for this HSC maintenance (154) (Fig. 3). 4-Oxo-RA was previously shown to be a potent agonist specifically for RARβ (155, 156).

Figure 3. Vitamin A is Metabolized to Retinoic Acid and then to 4-OH and 4-oxoRA.

4-Oxo-RA can act as an agonist with highest affinity for RAR β to mediate hematopoietic stem cell self-renewal and reduce cell differentiation and cell cycle transit (modified from ref. 154).

It is possible that in other cell types, Cyp26a1, rather than Cyp26b1, is required to generate 4-oxo-RA and that 4-oxo-RA is a key ligand required for transcriptional activation. Using a differentiation protocol that results in parietal endoderm, a type of epithelial cell, we showed that Cyp26a1 null embryonic stem cells displayed greatly increased intracellular RA levels and high levels of Hoxa1 mRNA, as might be expected, but surprisingly, also exhibited more resistance to RA-induced cell proliferation arrest and a large reduction in the parietal endoderm differentiation markers expressed later in the differentiation process, compared to wild type embryonic stem cells (157). These data from embryonic stem cells suggest that RA metabolite, such as 4-oxo-RA, generated by Cyp26a1, are indeed required for the complete differentiation of embryonic stem cells into parietal endoderm cells (Fig. 3). Thus, oxidized metabolites of RA may be essential signaling molecules that can act as agonists for the RARs, indicating that the regulation of signaling in various cell types by vitamin A is more complex than previously thought. Much more interrogation is needed to determine how widespread this requirement for signaling by 4-oxo-RA is both during embryonic development and postnatally in mammals, as much of the data now in the literature indicate that Cyp26a1 and Cyp26b1 are primarily involved in reducing RA levels by inactivating the RA signal.

H. Summary

The complexity of the metabolism of vitamin A to RA and 4-oxo-RA reflects the importance of these signaling molecules in diverse processes in numerous cell types and the observations that endogenous concentrations of RA that are either too high or too low can cause abnormalities during development. There is much more to learn in this fascinating area of research.