SUMOning the base excision repair machinery for differentiation

Abstract

Besides critical and well‐studied roles in maintaining genome stability, DNA repair pathways including base excision repair (BER) are also employed by mammalian cells for targeted events such as the generation of diverse antibody repertoires and the differentiation of pluripotent cells. In this issue of The EMBO Journal, Steinacher et al (2019) provide novel mechanistic insights into the complex role of SUMOylation of key BER proteins during active DNA demethylation—a role they demonstrate to be important for cellular differentiation and that has broader implications for BER as a whole.

Groundbreaking work by Nobel laureate Tomas Lindahl measured the rate of base loss from duplex DNA in an aqueous environment, revealing the inherent instability of the molecule that stores our genetic information. This notion of spontaneous DNA damage was expanded to include DNA damage caused by reactive chemicals that are normally present in the cellular milieu, such as reactive oxygen species and alkylating agents. The vast majority of the DNA adducts introduced by these spontaneous mechanisms, which usually range from relatively minor changes in the nitrogenous bases to complete loss of the base, do not significantly distort the helical structure of duplex DNA and are repaired by BER (for an excellent review of BER mechanisms, see Krokan & Bjørås, 2013). In the case of a modified base, the repair pathway is initiated by a DNA glycosylase that cleaves the glycosidic bond, generating an abasic site. The abasic site is then cleaved by an endonuclease followed by processing of the break termini to generate 3′ hydroxyl and 5′ phosphate ends that can be extended and then joined by a DNA polymerase and a DNA ligase, respectively. A notable feature of the BER reaction mechanism is that it generates cytotoxic reaction intermediates from the relatively benign modified base, including DNA single‐strand breaks (SSBs) that, when present in excess, can hyperactivate poly(ADP‐ribose) polymerase‐1 (PARP1) and cause cellular stress and death (Calvo et al, 2013). This has prompted significant interest in understanding the mechanisms by which BER is coordinated.

While the primary role of DNA repair pathways is to protect the informational content and ensure the functional integrity of the genome, there are a number of instances in which DNA repair pathways have been co‐opted to participate in processes that are critical in complex multicellular organisms. For example, the generation of a diverse repertoire of antibodies by V(D)J recombination involves proteins from the non‐homologous end‐joining pathway that repair DNA double‐strand breaks. This bears some risk, as failed attempts at V(D)J recombination can give rise to lymphomagenic chromosomal translocations. The AID cytidine deaminase, which converts cytosine to uracil, initiates mutagenic BER that is selectively targeted to optimize antibody affinity after initial antigen recognition. However, the identification of mutational signatures associated with the aberrant expression of AID and other cytidine deaminases indicates that dysregulated mutagenic BER can contribute to the genomic instability that is a hallmark of cancer (Alexandrov et al, 2013). Thus, the utilization of DNA repair pathways for cell‐type‐specific events in metazoan organisms is subject to regulation in order to minimize the risks to genome integrity.

In addition to being the substrate of cytidine deaminases, cytosine residues in CpG sequences are methylated at carbon 5 of the cytosine base by DNA methyltransferases, forming 5‐methylcytosine (5mC). While DNA methylation is a cornerstone epigenetic strategy employed by higher eukaryotes to regulate transcription and chromatin structure, this strategy comes at a cost in terms of genome stability as methylation of cytosine to 5mC reduces the stability of the glycosidic bond, resulting in increased spontaneous base loss and a mutational signature that is evident in tumor genomes (Alexandrov et al, 2013).

Another challenge associated with transcription regulation by 5mC is the need to alter the epigenome during cell differentiation, a process that requires the removal of a large number of 5mC bases in order to establish new transcriptional programs (Fig 1A). With the groundbreaking discovery that ten‐eleven translocation (TET) proteins oxidize 5mC to 5‐formylcytosine (5fC) and 5‐carboxylcytosine (5caC) in mammalian cells, it was realized that active demethylation of 5mCs could be accomplished by BER of TET‐oxidized 5mC derivatives (Fig 1A) and replacement with unmodified cytosines (Kohili & Zhang, 2013). Indeed, TDG, which is the sole DNA glycosylase known to engage oxidized 5mC derivatives (Fig 1A), is also the only DNA glycosylase essential for mouse embryonic development, supporting a critical role for TDG‐initiated BER of TET‐oxidized 5mCs in mammalian development (Cortázar et al, 2011; Cortellino et al, 2011).

Figure 1. Role of SUMOylation in active demethylation and the control of BER protein levels.

(A) Model of SUMO dynamics during active DNA demethylation. Under differentiating conditions, TET enzymes recognize and oxidize 5mC to 5fC or 5caC. TDG recognizes these derivatives and cleaves their glycosidic bond but remains associated with the abasic site. SUMOylated XRCC1 along with its interacting proteins POLβ and LIGIII interacts with the SIM2 motif of TDG forming a stable DNA–protein complex. SUMO is covalently transferred from XRCC1 to TDG via a SUMO protease and SUMO activating (E1), conjugating (E2), and ligating (E3) enzymes. TDG SUMOylation facilitates its dissociation from the abasic site, allowing APE1 to generate a single‐strand break that is extended and processed by POLβ and then ligated by LIGIII. It is unclear if the 5fc/5caC on the complementary strand is immediately engaged by TDG to symmetrically demethylate CpG loci. As a result of TET‐initiated BER, select genes, such as Pax6, that were previously repressed by DNA methylation in stem cells are now expressed in the differentiated cells. (B) Maintaining the relative stoichiometry of BER proteins is critical to ensure efficient and non‐mutagenic reactions. Ubiquitination in the nucleoplasm targets the scaffold protein XRCC1 for proteasomal degradation in the absence of DNA damage, while damage induction increases the levels of chromatin‐bound, non‐ubiquitinated XRCC1. Given that both TET activation during differentiation and oxidative stress induce SUMOylation of XRCC1, damage‐induced SUMOylation may target XRCC1 to DNA damage sites in chromatin, protecting it from degradation and enhancing BER capacity.

It is well recognized that a variety of post‐translational modifications (PTMs) of DNA repair proteins regulate their stability, localization, interactions, and activity. While there is convincing evidence that ubiquitin‐dependent proteolysis controls the steady‐state levels of key BER factors (Fig 1B; Parsons et al, 2008), involvement of the ubiquitin‐related protein SUMO (small ubiquitin‐like modifier) is less well characterized. Proteomics data have identified SUMOylation sites in multiple BER proteins, including TDG1 and XRCC1 (Hendriks & Vertegaal, 2016). Combined with recent evidence demonstrating that SUMOylation of the Saccharomyces cerevisiae Ntg1 glycosylase is important for cell survival in response to oxidative stress and DNA‐methylating agents (Swartzlander et al, 2016), it is clear that SUMOylation plays critical but yet undefined roles in BER. Interestingly, Steinacher et al (2019) demonstrate that both TDG1 and XRCC1 are progressively SUMOylated in mouse embryonic stem cells (mESCs) during differentiation. This suggests that SUMO may not only regulate the repair of DNA damage by BER, but also the TDG‐initiated BER processing of TET‐oxidized 5mC derivatives (Fig 1A).

One of the key questions regarding the mechanism of BER has been whether the pathway is coordinated by sequential handover of reaction intermediates via interactions with relatively stable enzyme–product complexes, or by the assembly of BER proteins into large repairosomes that engage damaged bases en masse. Utilizing an innovative pull‐down assay mimicking TET‐oxidized 5mC, Steinacher et al provide evidence that TDG‐initiated BER occurs by a combination of both mechanisms. First, they demonstrate that TDG but none of the other BER proteins, XRCC1, POLβ, APE1, and LIGIII, stably associates with 5caC‐containing DNA. However, upon SUMOylation of XRCC1, XRCC1‐SUMO is able to interact with DNA‐bound TDG via the SUMO‐interacting motif 2 (SIM2) of TDG to form a stable DNA–protein complex along with the XRCC1 interacting proteins POLβ and LIGIII (Fig 1A). Thus, TDG stably bound to abasic DNA in the enzyme–product complex is engaged by SUMOylated XRCC1 to assemble a multi‐protein “BERosome” at the site of base damage.

Next, Steinacher et al show that the complete repair of a 5caC‐containing substrate not only requires the expected proteins TDG, SUMO‐XRCC1, POLβ, LIGIII, and APE1, but also the SUMOylation proteins SAE1/2, UBC9, and the SUMO protease SENP2. In an insightful set of experiments, they uncover that the SUMO moiety covalently linked to XRCC1 is transferred to TDG during the BER reaction, forming TDG‐SUMO. Given that it was recently demonstrated that TDG‐SUMO has weakened substrate affinity and impaired catalytic activity (Coey & Drohat, 2018), and a previous crystallographic study demonstrated that SUMOylation alters the structure of TDG in way that can discourage DNA binding (Baba et al, 2005), these combined observations strongly suggest that SUMO is transferred from XRCC1 to TDG after TDG has excised the oxidized base in order to promote the release of TDG from the abasic site and allow access by APE1 (Fig 1A).

To provide evidence for a physiological function of the mechanism uncovered by the in vitro experiments, Steinacher et al demonstrate that mESCs with a mutant TDG SIM2 domain fail to efficiently differentiate into neurites when subjected to all‐trans retinoic acid differentiating conditions. Notably, the TDG SIM2 mutant cells fail to upregulate genes necessary for neurogenesis when placed into differentiating conditions, supporting the hypothesis that the interaction of TDG with SUMOylated proteins is necessary for epigenetic reprogramming of stem cells. Additionally, XRCC1 null mESCs also fail to differentiate and do not appropriately alter their transcriptional programming. These cellular data convincingly support roles for both XRCC1 and SUMOylation in TET‐initiated BER during differentiation.

Unanswered questions remain about the cleavage of the abasic site generated by TDG and how this process is coordinated with the transfer of SUMO from XRCC1 to TDG and the subsequent reactions catalyzed by POLβ and LIGIII (Fig 1A). Although Steinacher et al did not detect APE1 in the DNA‐BER protein complex nucleated by SUMO‐XRCC1, it is possible that APE1 transiently interacts with the TDG–abasic DNA complex and/or XRCC1 to gain access to the abasic site, and that the interaction of the XRCC1 complex with the BER intermediates facilitates the subsequent processing of the cleaved abasic site by the XRCC1 interacting proteins POLβ and LIGIII. Understanding the regulation of APE1 activity during TDG‐BER is important because the simultaneous processing of 5mCs by TDG‐BER in gene regulatory regions with clustered CpG sites could, in theory, lead to DNA double‐strand breaks if APE1 acts at closely spaced abasic sites on opposite strands. Thus, further work is needed to delineate how the later steps of active demethylation by TDG‐initiated BER are regulated to safeguard genome stability.

Mutations in the gene encoding XRCC1 and in the genes encoding a number of XRCC1‐associated proteins have been identified as the causative factors in several rare neurodegenerative syndromes (Hoch et al, 2017). While it evident that these mutations cause defects in the XRCC1‐dependent repair of base adducts and single‐strand breaks, resulting in hyperactivation of PARP1 and reduced cell viability, it is not clear why neurodegeneration occurs in distinct neuronal cell types in the different syndromes. The results reported by Steinacher et al raise the intriguing possibility that defects in the differentiation of neuronal stem cells may contribute to the neurodegeneration that characterizes these syndromes and may underlie the cell‐type specificity.

Finally, the findings by Steinacher et al (2019) may be applicable to the regulation of BER as a whole. While it is evident that maintaining the relative stoichiometry of BER proteins is important for genome stability (Calvo et al, 2013), the mechanisms involved in this regulation are not well characterized. As XRCC1 is a scaffold protein that mediates the stability of the key BER proteins LIGIII and POLβ (Parsons et al, 2008), and since XRCC1 mediates the assembly of the multi‐protein “BERosome” at the site of base damage, XRCC1 is a likely key regulatory target for BER activity. Taken together with previous work demonstrating that multiple BER enzymes, including XRCC1, have increased stability in response to DNA damage as a consequence of forming chromatin‐bound repair complexes that are protected from ubiquitin‐dependent degradation (Parsons et al, 2008), the DNA damage‐induced SUMOylation of XRCC1 suggests that BER capacity is fine‐tuned to match DNA damage levels by the opposing effects of ubiquitination and SUMOylation on XRCC1 (Fig 1B). It will be interesting to determine the mechanisms by which DNA damage enhances SUMOylation of XRCC1, and whether SUMO‐XRCC1 interacts with other DNA glycosylases in addition to TDG. We anticipate that many more important roles for SUMOylation will be uncovered in the various BER repair subpathways that safeguard genome stability over the coming months and years. As such, the work by Steinacher et al represents an important step forward in understanding how a critical BER reaction is regulated to facilitate the differentiation of metazoan cells and raises intriguing future directions for the study of BER and human disease.

The EMBO Journal (2019) 38: e101147