DNA methylation and gene expression differences in children conceived in vitro or in vivo

Abstract

Epidemiological data indicate that children conceived in vitro have a greater relative risk of low birth-weight, major and minor birth defects, and rare disorders involving imprinted genes, suggesting that epigenetic changes may be associated with assisted reproduction. We examined DNA methylation at more than 700 genes (1536 CpG sites) in placenta and cord blood and measured gene expression levels of a subset of genes that differed in methylation levels between children conceived in vitro versus in vivo. Our results suggest that in vitro conception is associated with lower mean methylation at CpG sites in placenta and higher mean methylation at CpG sites in cord blood. We also find that in vitro conception-associated DNA methylation differences are associated with gene expression differences at both imprinted and non-imprinted genes. The range of inter-individual variation in gene expression of the in vitro and in vivo groups overlaps substantially but some individuals from the in vitro group differ from the in vivo group mean by more than two standard deviations. Several of the genes whose expression differs between the two groups have been implicated in chronic metabolic disorders, such as obesity and type II diabetes. These findings suggest that there may be epigenetic differences in the gametes or early embryos derived from couples undergoing treatment for infertility. Alternatively, assisted reproduction technology may have an effect on global patterns of DNA methylation and gene expression. In either case, these differences or changes may affect long-term patterns of gene expression.

Figure 1.

Hierarchical cluster analysis (heat map) of the methylation microarray data, based on all 1536 CpG loci, for 10 in vitro and 13 in vivo controls (12 for blood). Each sample was assayed in duplicate. Each duplicate clustered together (data not shown) so the mean beta values were used in this figure for ease of representation. Red indicates complete methylation, green indicates absence of methylation and intermediate levels are indicated by intermediate shades. Note that all cord blood samples cluster separately from placental samples.

Figure 2.

Beta values (fraction of CpG methylated at the indicated site) of CpG sites at which mean methylation differences were different at two sites within the same gene (two-way ANOVA, P ≤ 0.08). (A) CEBPA in cord blood, (B) COPG2 in cord blood, (C) MEST in placenta and (D) SERPINF1 in placenta. The points represent data from 12 in vivo conceptions/24 beta values per CpG site versus 10 in vitro conceptions/20 beta values per CpG site. The in vitro individuals have slightly higher average methylation at the two CEBPA and COPG2 sites in cord blood and slightly lower average methylation at MEST and SERPINF1 CpG sites in placenta.

Figure 3.

Boxplots with whiskers showing transcript levels for four genes in which mean transcript levels differ significantly (Student's t test, P ≤ 0.05) between in vitro and in vivo populations (A) CEBPA in cord blood, (B) COPG2 in cord blood, (C) MEST in placenta and (D) SERPINF1 in placenta. The plots represent data from 93, 85, 69 and 68 in vivo conceptions for CEBPA, COPG2, MEST and SERPINF1, respectively, versus 70, 73, 69 and 68 in vitro conceptions for CEBPA, COPG2, MEST and SERPINF1, respectively. Boxplot interpretation: the box contains the middle 50% of the data. The upper edge of the box indicates the 75th percentile of the data set, and the lower edge indicates the 25th percentile (the range of the middle two quartiles is known as the inter-quartile range). The mean (open square) is shown and the line in the box indicates the median value with the ends of the vertical lines or ‘whiskers’ indicating the 95th and 5th percentiles of the data set. The symbol (x) depicts the 99th and 1st percentiles of the data set, and (-) depicts the maximum data range.