Characterization of regulators of CpG island activity in mouse stem cells

CpG island (CGI) promoters control the expression of over two thirds of all mammalian genes, many which are highly active. Nevertheless, we are still lacking an understanding how their hallmark feature, the high frequency of unmethylated CpGs, contributes to promoter activity. A recent study showed that CpG density is necessary for transcriptional activity in mouse embryonic stem cells (mESCs), independent of transcription factors (TFs) that bind complex DNA motifs (Hartl et al., 2019). This thesis explores the contribution of proteins that bind unmethylated CpGs at CGIs and characterizes a distinct case in which binding of the TF BANP (BTG3 Associated Nuclear Protein) depends on promoter CpG density.
In the first part of this thesis, we aim to identify binders of unmethylated CpGs in mESCs. To that end, we employed oligo-affinity purification coupled with quantitative mass spectrometry (oligo-AP MS) utilizing three different promoters each with three different CpG densities. We identify common and specific hits (often representing specific protein complexes) that enrich with higher promoter CpG density. For selected targets we validate increased binding with higher CpG density in vivo (based on published ChIP-seq data), including the ZF-CxxC containing proteins KMT2B (Lysine Methyltransferase 2B) and KDM2B (Lysine Demethylase 2B) as well as ZBTB2 (Zinc Finger And BTB Domain Containing 2) and SAMD1 (Sterile Alpha Motif Domain Containing 1). Motif analysis, at least in the case of ZBTB2 and SAMD1, suggests that additional motif preferences beyond “CG” exist, offering an explanation for differential binding between the factors.
In order to explore how selected CpG binders contribute to promoter activity, we tested how transcription was affected upon their deletion in mESCs, particularly in the context of CpG density and binding. Upon genetic deletion we detect small trends at bound, CpG-rich sites towards a decrease of promoter activity for Kmt2b-/- and to a minor degree Zbtb2-/- cells as well as an increase for Kdm2b-/- and Samd1-/- cells. For Zbtb2-/- and Samd1-/- cells, we tested how stably expressed luciferase reporters with different promoter CpG densities behave in the mutant backgrounds compared to wildtype (WT), and the results align with the genome-wide trends. As the effect sizes at bound and CpG-rich promoters are relatively small, we generated combinatorial protein deletions to test for additional effects. Overall, we observe that most of the transcriptional changes upon combinatorial deletions can potentially be explained as additions of the single deletions. Furthermore, transcriptional changes at commonly bound sites appear to be a mixture of what is expected based on the single deletions. Finally, we explored functional links based on bound proteins at CGIs close to genes affected by our targets of interest; we identify common patterns as well as target-specific hits.
The last part of this thesis explores the observation that binding of the TF BANP is dependent on CpG density. BANP is a strong activator in mESCs, allowing the generation of a promoter-luciferase reporter that is highly dependent on BANP (Grand et al., 2021). We generated BANP promoter constructs with five different CpG densities and determined their transcriptional activity, as well as the footprints they generate when stably expressed in cells with and without DNA methylation. We find that the BANP reporters strongly depend on CpG density with a strong decrease in activity at 50 % non-motif CpG density that is partially rescued in cells lacking DNA methylation. Single molecule footprinting (SMF) suggests that CpG density as well as DNA methylation contribute to DNA accessibility and BANP binding. Finally, we observe a disconnect between loss of reporter activity and loss of BANP binding, which requires further validation but speculatively suggests the contribution of additional factors to BANP promoter activity.
Taken together, this work identifies proteins that preferentially bind CGI promoters and show trends to either activate or inhibit transcription at their bound sites. Our study on CpG-binding proteins gives some insight into the regulation of their bound CGI promoters, offering a good framework to discover additional CGI-regulating factors and gain a better understanding of CGI regulation including complex associations and combinatorial binding. Similarly, the BANP reporter constructs provide a promising tool to disentangle the effects of CpG density, BANP binding, DNA accessibility, DNA methylation, and additional binders on promoter activity.
Understanding how CpG density contributes to (CGI) promoter activity constitutes one major step in ultimately predicting gene expression in development and disease.