A function of thymine DNA glycosylase-initiated DNA repair in maintaining epigenome stability

The Thymine DNA Glycosylase (TDG) was initially discovered by its ability to excise the deamination
products of cytosine and 5-methylcytosine (5-mC), and therefore thought to initiate base excision
repair (BER) of the resulting G•U and G•T mismatches. Later, TDG was also found to act in concert
with transcription factors in the regulation of gene expression.
With its apparently two-sided nature, TDG has riddled researchers for many years and the stimuli
and interactions that control TDG function are still under investigation. The aim of my thesis was to
dissect the role of TDG in DNA repair with a focus on its regulation by post-translational modification,
and to investigate how TDG-initiated BER contributes to epigenetic stability at CpG islands (CGIs)
during cell differentiation.
Both described functions of TDG, in DNA repair and in the regulation of gene expression, require its
post-translational modification and non-covalent interaction with the small ubiquitin-like modifiers,
SUMO1 and SUMO2/3. Extensive biochemical studies by our laboratory have shown that
SUMOylation of TDG may induce its dissociation from the abasic (AP-) site after base excision.
However, in vivo evidence corroborating an involvement of SUMOylation in TDG-dependent BER has
been pending and the function of non-covalent SUMO-binding has remained elusive. I thus
generated a Fluorescence Resonance Energy Transfer (FRET) system to monitor the interaction
between TDG and SUMO1 or SUMO3 in cells. I was able to confirm a modulation of the SUMO1-TDG
interaction dynamics in response to DNA damage, whereas the interaction with SUMO3 remained
unaffected, suggesting that modification by SUMO3 might regulate TDG function in a context other
than DNA repair.
To investigate the biological function of TDG genetically, we generated a Tdg knockout mouse. In
contrast to any other known DNA glycosylase, deletion of Tdg caused embryonic lethality. Further
characterization of MEFs isolated from TDG-proficient and -deficient embryos revealed no evidence
for a DNA repair defect, but a significant number of misregulated genes in differentiated Tdg-/- cells,
as well as a loss of active histone marks, gain of repressive histone modifications and an
accumulation of 5-mC at CGI promoters. A phenotype we did not observe in embryonic stem cells.
From these data, we proposed a dual function of TDG in maintaining active chromatin states at
promoters in differentiating cells, first by structurally coordinating histone modifying enzymes and
second by counteracting errors of the DNA methylation machinery by initiating repair of aberrantly
methylated cytosines in CGIs. Consistent with a TDG-dependent engagement of DNA repair at such
sites, we found BER factors to associate with these promoters and DNA repair intermediates to
accumulate in differentiating cells in a TDG dependent manner.
To investigate further how TDG is involved in DNA methylation control, we mapped DNA methylation
in the genomes of TDG-proficient and -deficient mouse embryonic stem cells (ESCs), neuronal
progenitor cells (NPs) and MEFs and found differential methylation to arise only with differentiation.
Further characterization of the resulting differentially methylated regions (DMRs) revealed that those
overlapping with a CGI were almost exclusively hypomethylated in TDG-deficient compared to -proficient cells,
reflecting a failure to establish methylation at these CGIs during differentiation. In
search of the reason for this failure in a 24 h differentiation timecourse, we found global 5-mC levels
to rise with differentiation in cells lacking TDG activity, in parallel to the generation of the final
products of TET-protein catalyzed 5-mC oxidation, 5-formylcytosine (5-fC) and 5-carboxylcytosine (5-
caC), the latter two of which are proposed intermediates of active DNA demethylation and
substrates for TDG. Differentiation thus appeared to induce methylation but also the intermediates
of active demethylation. We therefore analyzed 5-mC and 5-caC levels at the CGI DMRs and found
both to rise with differentiation in wildtype cells, suggesting that the loss of pluripotency induces a
cycle of DNA methylation and demethylation specific CGIs. In Tdg knockout cells, though, this
induction appeared to fail whereas in cells expressing a catalytically dead mutant TDG (TDG-cat), the
cycle of methylation and demethylation was induced but blocked by the inability of TDG-cat to excise
5-caC.
Taken together, in collaboration with colleagues from different laboratories I was able to show that
differentiation triggers a state of high epigenetic plasticity at these CGIs and that catalytically active
TDG is required to maintain an equilibrium of DNA methylation and demethylation. The imbalance of
epigenetic marks resulting from knockout of TDG disrupts gene expression programs and the
accumulation of aberrations eventually leads to loss of viability on the cellular and on the organismic
level.