The biochemistry of DNA Oxidation- and repair-mediated active DNA demethylation

Cells of multicellular organisms, no matter how specialized they are, share the same genetic information, stored in their deoxyribonucleic acid (DNA) sequence. They obtain their identity during lineage commitment and differentiation, where specific gene expression patterns are established and subsequently maintained. This process does not involve the alteration of the DNA sequence itself; instead, it is achieved through mechanisms that modulate the accessibility of the DNA to the transcription machinery and thus control how the genetic code is read and applied. Faithful development and survival of complex multicellular organisms is thus not only depending on the genetic code but is also controlled by an additional layer of information called the epigenetic code. In mammals, the epigenetic information is stored mainly in two forms, posttranslational histone tail modifications and DNA methylation. DNA methylation of the fifth carbon of cytosines (C) yielding 5-methylcytosine (5mC) is predominantly found in palindromic CpG dinucleotides affecting roughly 60 - 80% of them (Bird 2002). Epigenetic memory that comprises both layers of epigenetic information is generally maintained during cell division and in particular DNA methylation poses a fundamental and heritable barrier that prevents regression into an undifferentiated state and loss of cellular identity (Messerschmidt et al. 2014; Seisenberger et al. 2013). DNA methylation is established by DNA methyltransferases (DNMTs) that, in our current understanding, either focus on the (de novo) establishment or the maintenance of DNA methylation across cell generations (Jurkowska et al. 2011; Law and Jacobsen 2010). Despite its crucial role, DNA methylation patterns are not only statically maintained but are also subject to dynamic regulation through active as well as passive mechanisms. DNA demethylation events have been observed locus-specifically in differentiated cells (Kangaspeska et al. 2008; M. S. Kim et al. 2009b; Metivier et al. 2008) as well as on a global scale during early development (Oswald et al. 2000; Seisenberger et al. 2012; Smith et al. 2014). Although global erasure of DNA methylation can be obtained efficiently through passive dilution by inhibiting the methylation maintenance machinery, a major caveat of this process is the dependence on repeated DNA replication, reducing the dynamic flexibility. In contrast, DNA demethylation also occurs in an active manner, involving enzymatic activities that can process 5mC and revert it back to unmodified C. While the catalytic mechanism of DNA methylation is well understood and established, the process of removing DNA methylation has puzzled researchers for a long time and a variety of mechanisms have been proposed (Ooi and Bestor 2008; S. C. Wu and Zhang 2010). Many of these pathways, however, have failed to find sufficient support, most often due to a lack of reproducibility or convincing biochemical as well as biological evidence.
In recent years, major advances have been made in the understanding of DNA demethylation and some promising candidate mechanisms have emerged (H. Wu and Zhang 2014). Compelling biochemical as well as biological evidence points towards an involvement of the ten eleven-translocation (TET) family of dioxygenases in the removal of DNA methylation (Pastor et al. 2013). The family consists of 3 members, TET1-3, that share a conserved catalytic core domain enabling iterative oxidation of 5mC to generate 5-hydroxymethylcytosine (5hmC), 5-formylcytosine (5fC), and 5-carboxylcytosine (5caC), which could serve as intermediates in active or passive DNA demethylation processes (He et al. 2011; Huang et al. 2014; Inoue et al. 2011; Ito et al. 2011). The thymine DNA glycosylase (TDG), originally identified as biochemical activity excising thymine (T) and uracil (U) when mispaired with guanine (G), is able to recognize and excise the TET-mediated 5mC oxidation products 5fC and 5caC (He et al. 2011; Maiti and Drohat 2011). A role for TDG in epigenetic programming and DNA demethylation has also been implicated by gene inactivation studies in animals (Cortazar et al. 2011; Cortellino et al. 2011; Saito et al. 2011) as well as in ES cells (Raiber et al. 2012; L. Shen et al. 2013; C. X. Song et al. 2013a). Together, these findings gave rise to a novel concept of active DNA demethylation (Kohli and Zhang 2013; H. Wu and Zhang 2014) involving TET-mediated 5mC oxidation followed by TDG-initiated DNA repair to release the oxidized 5mC derivative and re-establish the unmethylated state. Despite the fact that this mechanism is plausible and has been widely accepted, there is in fact little evidence supporting a direct link of TET with TDG and DNA repair and mechanistic details, coordination, regulation and targeting of this process remain to be clarified.
In order to gain further insight into TET and TDG-mediated active DNA demethylation, I set out to address some of the imminent mechanistic questions by in vitro reconstitution of oxidative DNA demethylation along the TET-TDG axis in combination with base excision repair (BER). I first showed that TDG by itself has no detectable enzymatic activity on 5mC and 5hmC but efficiently recognizes and processes 5caC, a modification that does not affect regular Watson-Crick basepairing (Supplementary results 4.4.1) (He et al. 2011; Maiti and Drohat 2011). The proposed oxidative DNA demethylation mechanism implies a coupled action of TET and TDG to facilitate an efficient but coordinated removal of 5mC. To investigate the coupling mechanism, I tested a potential physical interaction of the two enzymes, through multiple experimental approaches. I could demonstrate that TET1 and TDG physically interact through domains located in the N-terminus as well as the catalytic domain (TET1CD) of TET1. Recombinant TET1CD/TDG complex, purified from Escherichia coli (E.coli) cells co-expressing both proteins, turned out to act as ‘demethylase’ by combining both enzymatic activities to remove 5mC and 5hmC from synthetic DNA oligonucleotides. After successful reconstitution of 5mC base release with purified recombinant proteins, I combined this activity with the BER machinery and showed complete reconstitution of active DNA demethylation via oxidized intermediates in vitro, providing the first experimental evidence that this process is functional in the proposed manner. Moreover, investigation of the process operating at symmetrically modified CpGs suggested that symmetric DNA demethylation is obtained through a processive mechanism that is highly coordinated and acts sequentially on both strands to protect the DNA from the formation of DNA double strand breaks (DSBs). However, the sequential and coordinated repair of two nearby substrates on opposite strands, beneficial in terms of avoiding the formation of DNA DSBs, could have an impact on mutagenesis of CpG dinucleotides. I could show that at fully methylated CpG sites, where spontaneous hydrolytic deamination may occur coincident with oxidative DNA demethylation, the repair of the resulting G/T mismatch is highly disfavored in presence of a G•5caC base pair. The preferential repair of 5caC can then occasionally create a C to T mutation and, hence, lead to the loss of the CpG dinucleotide (Appendix I). Additional experiments revealed that neither TET nor TDG activity is restricted to double-stranded DNA or a CpG context, suggesting that TET-TDG-mediated DNA demethylation might also occur in other biologically relevant contexts including non-CpG methylation, single-stranded DNA or R-loops (Supplementary results 4.4.1).
The process of active DNA demethylation by TET-TDG-BER has to occur in a tightly regulated and highly coordinated manner to ensure accuracy and genome integrity. TDG was previously described to be regulated by posttranslational modification and non-covalent interaction with the small ubiquitin-like modifiers (SUMO), SUMO1 and SUMO2/3 (Hardeland et al. 2002; Mohan et al. 2007; Steinacher and Schar 2005) and other BER factors are also amongst the increasing number of reported SUMO targets (Cremona et al. 2012; Weber et al. 2014). The biochemical investigation of the functional consequences of SUMO modification, however, has been lagging behind due to the difficulty to generate appreciable amounts of recombinant SUMOylated proteins. Therefore, I, in collaboration with David Schürmann, established a recombinant SUMOylation system, coupling efficient SUMO-conjugation with affinity purification of modified target proteins, and present tools and strategies to generate SUMOylated proteins using versatile binary expression vector systems in protease-deficient E.coli. We successfully modified the BER factors TDG and XRCC1 and could show that purified SUMO-modified TDG had retained the expected biochemical properties (Appendix II). I was then also able to modify the N-terminus of TET1 with SUMO1 as well as SUMO3 using the recombinant SUMOylation system and identified SUMO interaction motifs (SIMs) in the TET1 sequence by in silico prediction. This indicates that SUMOylation might also be prominently involved in the coordination and regulation of TET-TDG mediated DNA demethylation processes (Supplementary results 4.4.2).
We reasoned that other factors previously proposed to contribute to DNA demethylation might be involved in the concerted action of the TET and TDG enzymatic activities, exerting a regulatory or structural function. The growth arrest and DNA-damage-inducible protein 45 (Gadd45) family of proteins has previously been implicated in active DNA demethylation through Xeroderma pigmentosum group G (XPG)-dependent DNA repair (Barreto et al. 2007; Schmitz et al. 2009) or BER of activation-induced deaminase (AID)-based deamination products (Cortellino et al. 2011; Rai et al. 2008). In collaboration with Zheng Li and Guoliang Xu at the Chinese Academy of Science in Shanghai, I set out to re-investigate a potential role of Gadd45a in the context of oxidative DNA demethylation and provide several lines of evidence that Gadd45a serves as a regulator in the TET-TDG-mediated DNA demethylation pathway. Together, we showed that Gadd45a synergizes with TET and TDG to activate a methylated reporter gene in transfected cells. Moreover, Gadd45a physically interacted with TDG and potentiated TDG glycosylase activity to remove 5fC and 5caC from genomic DNA of transfected HEK293T cells. Finally, deletion of Gadd45a/b in mouse ES cells led to hypermethylation at specific genomic loci, which also gained increased DNA methylation levels and are enriched in 5fC in TDG-deficient cells. Despite the diverse molecular functions that have been attributed to Gadd45a, we were able to connect Gadd45 proteins with DNA demethylation along the TET-TDG axis and propose a regulatory function. My specific contribution to this work was the biochemical examination of Gadd45a on the enzymatic function of TET and TDG (Appendix III).
Taken together, the work presented in my PhD thesis advances our understanding of TET-TDG-mediated active DNA demethylation and the underlying mechanisms. I was able to show that TET associates with BER by physically interacting with TDG and to provide proof of concept that DNA demethylation can be achieved through the coordinated action of an intricate network of enzymes consisting of TET, TDG and the core components of the DNA BER system. Without question, additional factors and regulatory mechanisms, like Gadd45a and SUMO modification, for which I was able to provide preliminary evidence, will turn out to contribute to coordination, targeting and regulation of this active DNA demethylation process. Additional findings that 5mC oxidation and repair by TET1 and TDG is neither limited to CpG dinucleotides nor to double-stranded DNA suggest that this pathway of DNA demethylation could operate in as yet unidentified biological contexts.