Genome-wide roles of Foxa2 in directing liver specification.

Dear Editor, Members of Foxa transcription factor family, namely Foxa1, Foxa2, and Foxa3, play crucial roles in guiding hepatic differentiation and hepatic homeostatic maintenance (Le Lay and Kaestner, 2010; Zaret and Carroll, 2011). They act at several critical time points to regulate hepatic differentiation as early as embryonic definitive endoderm (DE) stage (Gualdi et al., 1996). Combination of both Foxa1 and Foxa2 was required for the formation of foregut DE cells and the further liver bud differentiation during embryonic hepatic development (Lee et al., 2005). Foxa proteins can bind the condensed chromatin of Alb1 enhancer and induce local chromatin relaxation prior to the initiation of transcription (Cirillo et al., 2002). The binding of Foxa on Alb1 enhancer before transcription activation determined the developmental competence by premarking the chromatin. Thus, Foxa proteins were proposed to behave as pioneer factors in DE to prime hepatic genes prior to their actual transcription activation in nascent liver (Zaret and Carroll, 2011). Although Foxa2 binding profiles were extensively mapped in adult liver cells using ChIP-sequencing (ChIP-seq) technology (Wederell et al., 2008; Tuteja et al., 2009; Wallerman et al., 2009; Hoffman et al., 2010), binding profile of Foxa proteins at the embryonic DE stage has not been obtained yet. Functions of Foxa family have been known with significant differences between embryonic hepatic developing stage and post-developing stage (Cirillo et al., 2002; Li et al., 2011). Therefore, a systematic analysis of Foxa binding at the DE stage will help elucidate its functions at early hepatogenesis stage. We first established a protocol to efficiently differentiate ES cells to DE cells and then early hepatic cells. DE cell fate commitment was confirmed by qPCR and FACS assays and in vitro hepatic differentiation analysis (Supplementary Figure S1). The obtained DE cells were used to map Foxa2 binding sites by ChIP-seq. We obtained 4.2 million reads, and 36675 peaks were generated by Model-based Analysis of ChIP-seq (MACS) program after aligning these reads to the mouse genome and retrieving only the unique reads (Figure 1A) (Zhang et al., 2008). To verify the peaks, ChIP-qPCR assays were performed. Twenty-eight out of 30 randomly selected peak sites were positive for Foxa2 binding, while all 21 randomly selected non-peak sites were negative (Supplementary Figure S2A and B), proving the specificity of the ChIP-seq results. Motif analysis showed that all of the five top scored motifs were consistent with previously documented Foxa2 binding sites (Supplementary Figure S2C) (Tuteja et al., 2009; Hoffman et al., 2010). After mapping the enriched peaks to + 50 kb of transcription start sites of the nearest RefSeq genes, 8495 genes were identified as the candidate target genes for Foxa2 in DE. In order to gain a genome-wide view on how Foxa2 affected the expression of its target genes, samples from three time points during differentiation process were taken for microarray analysis: day 0 representing ES cells (no Foxa2 expression), day 5 representing DE cells, and day 7 representing early hepatic cells. Two comparisons were made with the microarray data sets: (i) between ES cells and DE cells; and (ii) between DE cells and early hepatic cells (Figure 1B). Furthermore, the up-regulated genes obtained from the above two comparisons were mapped to the pool of Foxa2 target genes identified by ChIP-seq (Figure 1B). Seven hundred and ninety-four Foxa2-bound genes displayed higher transcription levels at the hepatic cell stage after the DE stage, though all DNA binding by Foxa2 happened at the DE stage. Late activation of these genes was statistically significant as shown by parametric analysis of gene expression (PAGE) (Supplementary Table S1) (Kim and Volsky, 2005). Our results indicated that activation of many Foxa2 target genes indeed happened later than the Foxa2 binding and revealed a widely existing uncoupling between promoter/ enhancer occupation and immediate transcription activation during the process of hepatic differentiation, consistent with previous observations made on Alb1 gene activation (Gualdi et al., 1996). Previously, Foxa binding was suggested to correlate with histone H3 lysine 4 dimethylation (H3K4me2) in human cancer cell lines (Lupien et al., 2008). In order to investigate whether the binding of Foxa2 at the DE stage could also be related to certain histone modifications, we assessed the histone modification status on Foxa2 target genes in DE cells. Genes such as Afp, Ttr, and nine other genes, whose promoter/enhancers were occupied by Foxa2 at the DE stage, and transcription activation occurred at early hepatic cell stage, were categorized as ‘late genes’ (Supplementary Figure S3A and C); genes whose expression levels were concomitantly up-regulated with Foxa2 binding at the DE stage, such as Foxa2, Gata4, and nine other genes, were grouped as ‘early genes’ (Supplementary Figure S3B and D). Histone H3 acetylation (H3ac) status on all of these genes in DE cells was first examined. H3ac around Foxa2 binding sites could only be detected on the ‘early genes’, but not on the ‘late genes’ (Supplementary Figure S4A). Next, histone H3 methylation status on these genes was examined. High levels of both H3K4me2 and H3K27me3 were detected 420 | Journal of Molecular Cell Biology (2012), 4, 420–422 doi:10.1093/jmcb/mjs037 Published online June 28, 2012