In this issue, Ali-Khan et al. (pages 139–152) have used gene expression analysis to elucidate signaling pathways potentially involved in the mechanism of the developmental toxicity of excess vitamin A (retinol acetate). The ability to examine simultaneously the expression of thousands of genes in a tissue of interest has emerged relatively recently. While some have characterized the use of this technology to elucidate mechanisms of toxicity as “fishing trips,” the approach can be powerful for generating hypotheses. Global gene array analysis has been used to some success in identifying developmental pathways affected by the human and murine teratogen, valproic acid (Kultima et al., 2004; Okada and Fujiwara, 2006), and has also been used for expression profiling across normal developmental stages (e.g., Bonner et al., 2003; Gheorghe et al., 2006; Wang et al., 2005). Here, the experimental model is the murine limb exposed to retinol while developing in vitro. The premise is straightforward; by examining changes in the embryonal transcriptome in response to developmental insult, one can formulate hypotheses as to the developmental pathways that are critically affected.

In practice, it is far more complex. An embryo undergoing organogenesis is a rapidly changing organism with heterogeneous tissues and cells. Highly transcriptionally active (e.g., the authors report that 40% of all genes on the arrays had detectable transcripts in control limbs), it is likely, though not well documented, that the transcriptome of the organogenesis-stage embryo is changing in a time frame of hours or even minutes. Superimpose on this moving targets the effects of a toxicant and one can quickly get into difficulty in interpreting the results of gene array experiments. To reduce these difficulties, Ali-Khan and Hales have chosen to analyze gene expression in developing limbs at a relatively early time point, after 3 h of culture with retinol. Seeking to detect the earliest transcriptional responses to insult, this approach also minimizes transcriptional differences due to diverging rates of development between treated and control limbs. The choice of the 3-h time point is reasonable—the authors report a total of 81 genes whose expression is upregulated by retinol exposure. Yet, it is still a single snapshot in time. Additional later time points might have allowed elucidation of pathogenetic processes through downstream events, and provided context for the early changes reported here.

Another complexity is the choice of exposure concentrations. Based on previous work (Ali-Khan and Hales, 2006), the authors have chosen two concentrations of retinol, one which is highly dysmorphogenic (affecting multiple structures in the limb as well as limb growth) and a lower dose known to produce significant but milder limb dysmorphology. None of the limbs cultured in the present study were allowed to develop to a stage where they could be evaluated for morphogenesis. Lack of a contemporary set of cultured and treated limbs to confirm morphological effects would have strengthened the study. Mechanisms of action for most toxicants probably change to some extent across dose, and the authors acknowledge that here. Not surprisingly, many more genes were significantly upregulated at the higher, more teratogenic concentration. Indeed, only a single gene was significantly upregulated at the lower concentration, and that was eyes absent homolog 2 (Eya2). This gene is known to be retinol responsive and to mediate apoptosis in the eye (Matt et al., 2005). Although not a part of the present work, the eye is well known to be a target of excess vitamin A and other retinoids, as well as vitamin A deficiency. In contrast, 50 genes were significantly upregulated at the higher concentration. The use of two concentrations of retinol greatly enhances the value of this study. Using K-means cluster analysis, four distinct expression dose-response patterns were identified. One might surmise that those genes whose expression exhibited a monotonic dose response, particularly those most affected at the lower dose, might be more likely to be direct targets of vitamin A. While the authors present this idea, they also acknowledge that all four expression pattern groups of genes contained members from the same functional gene families: transcription factors and regulators, growth factors/cytokines and their receptors, intracellular signaling, and cell-cycle–related genes. Thus, while establishing dose-response patterns is important, these findings illustrate the difficult interpretation that remains.

In addition to Eya2, the single gene significantly upregulated at the lower retinol dose, Ali-Khan and Hales chose three others, Id3, Hes1, and Snai1, for confirmation by qRT-PCR of vitamin A induced expression changes. All four genes were confirmed by qRT-PCR to be upregulated by the higher dose of vitamin A. In a minor divergence from the array results, Eya2 expression was nominally but not statistically elevated at the low dose of retinol compared to control.