Neural tube defect genes and maternal diabetes during pregnancy

Abstract

Background: Maternal diabetes during pregnancy is a well-known teratogen that increases the risk for birth defects, such as neural tube defects (NTDs). We have previously shown that maternal diabetes profoundly affects gene expression in the developing embryo, in particular a suite of known NTD genes. In rodent experimental systems, NTDs present as phenotypes of incomplete penetrance in diabetic pregnancies. This property is difficult to reconcile with observations of consistently altered gene expression in exposed embryos. We here show that maternal diabetes increases the overall variability of gene expression levels in embryos.

Results: Altered gene expression and increased variability of gene expression together may constitute the molecular correlates for incomplete phenotype penetrance.

Discussion: Based on this model, we suggest that maternal diabetes reduces the precision of gene regulation in exposed individuals. Loss of precision in embryonic gene regulation may include changes to the epigenome via deregulated expression of chromatin-modifying factors. Unraveling the mechanisms underlying such epigenetic modifications in diabetic pregnancies will help to understand how teratogenic insults compromise embryonic development and possibly provide avenues for therapeutic intervention.

Figure 1

Maternal diabetes increases the variability of gene expression levels in the developing embryo. (A) Standard deviations from the mean expression level were calculated for each gene probe on the microarray that showed a “Present” signal in all samples. Standard deviations were then plotted for control as well as diabetic samples; each data point represents a single gene probe and its standard deviations. The cloud of data points deviates from the diagonal (dotted black line) that depicts the hypothetical regression line for equal variability under both metabolic conditions. The blue solid line for the experimental results displays a shift toward the y-axis (diabetic condition), indicating that standard deviations are generally higher (on average by 1.72-fold) in embryo samples from diabetic dams. (B) Although the SD plot shows absolute values regardless of expression level of each gene, the coefficient of variation shows variability of gene expression normalized for the respective expression levels. After logarithmic transformation of the coefficient of variation for each gene probe, a histogram was obtained, which shows that the curve representing the results for embryos from diabetic dams is shifted to the right toward higher values. Again, this visualization indicates that the variability of gene expression levels is higher in embryos obtained from diabetic dams compared to embryos from control dams. These results were obtained from microarrays in which each sample consisted of pools of three embryos that each came from an independent diabetic or control pregnancy, respectively. (C) Variability of gene expression levels between individual embryos from the same control or the same diabetic pregnancy, respectively, following the same analysis and transformations as for panel B.

Figure 2

Diabetes increases the variability of gene expression levels in other experimental paradigms. (A) Microarray data from gene expression profiling in placentas from normal compared to diabetic pregnancies (Salbaum and Kappen, unpublished data) were processed as shown in Figure 1B: the coefficient of variation was determined for each gene probe, and a histogram was obtained after logarithmic transformation. The curve representing the diabetic placenta samples was shifted to higher values, similar to the results obtained in embryos from diabetic pregnancies. (B) Publicly available microarray data from diabetic versus normal human kidney (GEO record GSE1009) were treated in the same fashion as described for embryonic or placental gene expression data. Similar to our own datasets, the curve representing the coefficients of variation for the diabetic samples is shifted toward higher values, again implying that the variability of gene expression levels is higher in diabetic samples compared to control samples.