A large number of whole-animal and in vitro studies have characterized the relationships between structure and developmental toxicity of retinoids. As for any chemical, the in vivo potency results from a combination of pharmacokinetic and pharmacodynamic properties. These studies suggest that the teratogenicity of retinoids is receptor-mediated, but there are other possibilities (see below). The structural requirements for teratogenicity have been reviewed (Willhite et al. 1989; Collins and Mao 1999) and show a wide diversity in the polyene side-chain and β-cyclogeranylidene ring modifications that still retain activity, although an acidic polar terminus appears indispensable. Some aromatic retinoids (arotinoids), such as TTNPB, are 1,000-fold more potent in vivo teratogens than RA. This potency appears to be due predominantly to slower elimination and reduced affinity for CRABPs (Pignatello et al. 1999). Another aromatic retinoid, etretinate, has a very long half-life in humans after multiple exposures, with measurable concentrations in serum 2 years after cessation of intake, probably because of storage and slow release from adipose tissue (Eisenhardt and Bickel 1994). Experimental studies show that the critical pharmacokinetic characteristic for retinoid teratogenicity is the area under the curve (AUC) (Tzimas et al. 1997), rather than a transient high dose.
The receptors for retinoids are of the nuclear hormone ligand-dependent transcription-factor superfamily (Nuclear Receptors Committee 1999). They are of two types: RARs (subclass NR1B) and RXRs (subclass NR2B) (see Chambon 1996). For each type, there are three receptors, α, β, and γ (NR1B1, -2, and -3 and NR2B1, -2, and -3), each encoded by a separate gene. For all these genes, with the exception of RXR, multiple isoforms have been detected (e.g., NR1B2a, -b, -c, -d), generated by differential promoter usage and alternative splicing. Most of the embryonic effects of retinoids seem to be mediated by RAR-RXR heterodimers, but RXRs can form homodimers and can also form heterodimers with a number of other nuclear receptors, the most important being those for thyroid hormones and for peroxisome proliferators (Mangelsdorf and Evans 1995). Several isomers of RA are agonists for RARs, including all-trans-RA, 9-cis-RA, 4-oxo-RA, and 3,4-didehydro-RA, and 9-cis-RA seems to be the predominant RXR agonist (Collins and Mao 1999).
Each receptor, and in some cases each isoform, has been knocked out in mice to test for its function in development. Many combinations of knockouts have also been generated. Loss of RARβ (all isoforms), RAR 1, or RAR 2 has no phenotypic effect (Li et al. 1993; Lohnes et al. 1993; Luo et al. 1995). In contrast, disruption of all isoforms of RAR or RAR causes many of the effects of vitamin A deficiency, including growth deficiencies and male sterility (Lohnes et al. 1993; Lufkin et al. 1993). Compound RAR null mice display all the malformations induced by vitamin A deficiency, including defects of the eyes, limbs, and heart and the craniofacial, urogenital, and reproductive systems (Lohnes et al. 1994; Mendelsohn et al. 1994). An interesting recent example is the compound RAR-RARβ null mouse, which causes syndactyly and demonstrates a role of RA