TDG-mediated Active DNA Demethylation in the Genome-wide Control of Transcriptional Initiation and Elongation

The transcription of DNA into RNA is key to convert the genetic code - a linear sequence of the four “letters” of the genetic alphabet - into three-dimensional biological structures with functions that are central to create and maintain life. Differential regulation of transcription allows the use of a single genic locus in multiple ways (gene-variants), allowing cells to respond to different circumstances such as the surrounding environment, the stage in development of an organism or even the circadian clock. The possibility of this adaptive use of a gene vastly increases the functional spectrum contained in a single coding region, but also requires a complex regulatory system to read and convey the right information at the right time. One way of how this regulation is achieved, is being studied in the field of epigenetics. This research addresses covalent but reversible modifications to DNA or DNA-packaging proteins that are able to influence transcriptional levels. As these modifications can persist throughout multiple cell cycles and are actively passed on to daughter cells, epigenetic marks are being considered as heritable without actually altering the genetic code itself.
Study of the DNA repair enzyme Thymine DNA glycosylase (TDG) revealed that DNA excision repair contributes to an epigenetic mechanism of transcriptional regulation, as TDG is able to indirectly remove methylated bases from the DNA (Jacobs and Schär, 2012; Schuermann et al., 2016). Beside G•T and other non-canonical DNA mispairings and base modifications, TDG can recognise and excise 5-formylcytosine (5fC) and 5-carboxylcytosine (5caC) from DNA, both of which are products of 5-methylcytosine (5mC) oxidation by ten-eleven translocation proteins (TETs, He et al., 2011; Ito et al., 2011). Methylation marks on DNA, placed on a C followed by a G (CpG), have been shown to affect transcription in many ways. The most direct and intuitive way is by the inhibition of transcription factor binding at gene promoters. Methylated cytosines thereby interfere with transcriptional initiation and the final expression level of a gene. TET and TDG-mediated active DNA demethylation can, in this case, remove these 5mCs and render the promoter more accessible to the transcription machinery and restore expression levels. This seemingly rather clear mechanistic concept of active DNA demethylation through oxidation and base excision could be shown on the biochemical level five years ago (Weber et al., 2016), but in vivo data supporting the whole process including TDG and the DNA repair part is rather scarce and/or convoluted. It is most likely attributable to the intrinsic dynamicity of this multistep process and the vast number of interactions of the active DNA demethylation machinery with other epigenetic regulators that remain to be untangled.
The Schär group and others reported that TDG depletion, unlike any other DNA glycosylase, leads to the death of mouse embryos around day E10.5 (Cortázar et al., 2011; Cortellino et al., 2011). This phenotype was not just accompanied by alterations in DNA methylation, but (even more so) by altered histone modifications, and consequential changes of RNA expression during differentiation. Other groups then described intermediates of active DNA demethylation being present outside of promoters and all over the genome of embryonic stem cells (ESCs, Raiber et al., 2012; Shen et al., 2013), thereby increasing the complexity of the epigenetic landscape and its regulation by TDG (Kolendowski et al., 2018). A central but until now least-understood concept, before TDG can be fully established as an omni-present tool for transcriptional regulation, is the following: oxidation of 5mC through TET enzymes, followed by TDG-mediated excision and BER, provides an unmodified C that can again be methylated (Weber et al., 2016). This proposes the possibility that 5mC species can undergo multiple rounds of methylation and demethylation in the same cell, rather than being a stable mark (reviewed in Parry et al., 2020). This brings two further challenges into the research of TDG and active DNA demethylation: First, capturing of the effective methylation state is difficult/not meaningful as the dynamics of the (de-)methylation of a CpG have to be measured. And second, not just the 5mC species need to be investigated in their interaction with DNA-binding factors but also all of the constantly present DNA demethylation factors, too – as well as their interactors. A very recent study addressed the first aspect and measured methylation and oxidation rates of CpGs and could indeed show that active 5mC oxidation exceeds passive demethylation by far and that CpGs with different rates of oxidation, i.e. active DNA demethylation, associate with different genomic locations, transcriptional levels and histone modifications (Ginno et al., 2020). While this nicely showed the flexibility of TET-mediated 5mC oxidation, the further modulation of/by processing via TDG and BER to control transcription remains to be addressed.
Although TDG was described to be beneficial for transcriptional activation of certain genes, it was only rarely shown for multiple targets and often lacked the direct link to the actual presence and activity of TDG (Hassan et al., 2017, 2020; Hu et al., 2014; Léger et al., 2014). In order to contribute filling that lack of information, I studied TDG and its effect on transcriptional regulation in murine ESCs, which are a very dynamic experimental model with high activity of active DNA demethylation, and applied locus-specific as well as genome-wide experimental procedures.
In a collaborative effort, I could show for the first time that TDG (and TET) is essential in the mediation of cytotoxicity upon a stress signal that is not caused by base-analogues as substrates for TDG (Kunz et al., 2009), but by the interference of the BER pathway downstream of TDG (see 4.1/appendix I). We observed that treatment of naïve ESCs, cultivated in 2i-medium, are extremely sensitive to an inhibitor of the DNA repair protein Poly-[ADP-ribose] polymerase 1 (PARP1) called Talazoparib (Tal). PARP-inhibitors are usually used in treatment of cancer cells that are defective in homologous repair (HR). There, inhibition of repair leads to an accumulation of SSBs, which, for example by collapse of replication forks, result in an overload of toxic DSBs that kill the HR-deficient cells. It should therefore not affect healthy (somatic) cells that are HR-proficient. The toxic effect of Tal in ESCs, which are highly proficient in HR, was therefore surprising and demanded further examination. We could show that the Tal-mediated cell death is not based on the increased generation of DSBs, but on the transmission of the damage signal via the damage sensor and transcription factor, (tumour) protein p53. Depletion of TDG did not reduce the level of DSBs, but drastically reduced expression of all examined p53 target genes that were upregulated upon the treatment with Tal. Among these genes were classical mediators of apoptosis, like Fas (Fas Cell Surface Death Receptor), Alox5 (Arachidonate 5-Lipoxygenase) or Tap1 (Tocopherol-associated protein 1), indicating that TET/TDG-mediated active DNA demethylation is necessary to transmit a systemic stress response that includes at least 1500 genes. Furthermore, did we observe an increased transcription of several thousand repetitive elements upon the treatment with Tal that was also drastically reduced after the depletion of TET, TDG or p53. The expression of these elements, notably caused an upregulation of genes involved in necroptosis and an interferon-like response, very likely contributing to the observed cell death. Reduction of these transcriptional responses in absence of TDG additionally correlated with a reduced amount of generated SSBs, hinting towards a link of TDG-mediated excision of 5fC and 5caC and expression.
Together, I could clearly underline the essential role of TDG activity as a general factor in the toolbox of transcriptional regulation, by showing that the Tal-mediated activation of these p53-targets is depending on the presence of TDG and correlates with increased SSBs in TDG-proficient cells.
In the Tal-mediated stress response, we saw that the recruitment of p53 to its targets genes is slightly impaired upon the deletion of TDG but that the chromatin accessibility at these promoters, which is usually well correlated with transcriptional initiation, is not altered. Based on this and the fact that oxmCs as well as TETs and TDG are frequently found after the transcription start site (TSS) as well as throughout the whole gene body, I aimed to address transcriptional elongation in dependence of active DNA demethylation. Performing a precision run-on sequencing (PRO-seq), I observed that transcriptional initiation in TDG depleted ESCs is not significantly changed, but the pause release of RNA polymerase II (RNAP2) is clearly reduced (see 4.3/appendix III). Interestingly, this was true for the stress induction by Tal but also in unchallenged TDG KO ESCs and indicates that TDG and active DNA demethylation are involved in another step of transcriptional control, in addition to promoter demethylation. To figure out what mechanism leads from active DNA demethylation to pause release of RNAP2, I analysed the histone variant H2A.Z. H2A.Z was described as an important factor in facilitating transcriptional initiation by its presence upstream of the TSS, but at the same time hindering productive transcriptional elongation after the TSS (Giaimo et al., 2019; Mylonas et al., 2020). By analysing published datasets of (acetylated) H2A.Z, I could correlate the abundance of this histone variant with sites of active DNA demethylation, SSBs and the release of paused RNAP2. In combination with a dataset that was previously generated in our laboratory, I could show that H2A.Z levels are strongly increased in ESCs depleted of TDG and that genes with an increased level of H2A.Z at their TSS also display a reduced release of paused RNAP2. This implies a role of TDG in the eviction of H2A.Z during RNAP2 pause release. Furthermore, did I observe that not only transcriptional elongation of genes is hindered in TDG null ESCs, but also that transcription at enhancers is impaired. Although this was recently published (Kolendowski et al., 2018), neither the connection of transcriptional elongation and enhancer activity, nor the crucial abundance of H2A.Z at enhancers (Brunelle et al., 2015), was investigated in dependence of TDG. These observations, however, need further verification but are well supported by current literature and provide a solid basis on the description of how TDG-mediated active DNA demethylation affects transcriptional elongation via H2A.Z (Schaukowitch et al., 2014; Williams et al., 2015).
Last but not least, I could contribute to the description of a novel step in the processing of DNA lesions by BER. Combination of the genome-wide data set of SSBs in murine ESCs, with newly obtained biochemical data in our laboratory, allowed to show that modification of BER-enzymes by PARP1 facilitates the initial generation of SSBs by TDG and the abasic site endonuclease APE1, by increasing the dissociation of BER factors after processing of their substrate. This adds a new level of PARP1-mediated regulation of BER in addition to its function in general SSB repair, as was previously assumed (see 4.2/appendix II, Fisher et al., 2007; Hanzlikova et al., 2016). Focussing the analysis on genomic locations that are not responding to the stress signalling evoked by Tal, I could show that PARP-inhibition leads to a reduction of SSBs mediated by active DNA demethylation. These sites that undergo targeted DNA demethylation are contrasted by sites with increased genomic instability (e.g., simple repeats), which gain SSBs upon inhibition by Tal because of PARP1 failing to repair spontaneously occurring damage. This data contributes to the understanding of random vs. instructed DNA damage in the genome and thereby again fortifies the role of active DNA demethylation as a targeted process in ESCs that is constantly ongoing.
Taken together I could underline that TDG is not just an accessory protein to TETs for the excision of oxmCs but that on top of this role, it possesses its own essential functions necessary for transcriptional regulation and contributes to the epigenetic landscape in a fundamental way.