THE MULTIFACETED MISMATCH REPAIR
During the past two decades, my group has been primarily interested in studying the biochemistry and biology of the postreplicative mismatch repair (MMR) system in human cells. As mutations in MMR genes are associated with hereditary non-polyposis colon cancer (HNPCC, also known as Lynch Syndrome), one of the most common inherited cancer predisposition syndromes, we have been trying to understand how MMR functions and how its malfunction leads to malignant transformation. However, evidence emerging from several different directions implicated MMR proteins also in other pathways of DNA metabolism and we are now changing direction and concentrating our efforts on some of these processes, mostly in human systems. We have also begun to explore the potential of other systems, specifically the DT40 chicken bursal B cells.
We are also revisiting the field of DNA demethylation, with a specific focus on the molecular mechanism of the process and its effects on different pathways of DNA metabolism.
Biochemistry of mismatch repair
To improve replication fidelity, MMR must detect non-Watson-Crick base pairs and direct their repair to the nascent DNA strand. Eukaryotic MMR in vitro requires pre-existing strand discontinuities for initiation; consequently, it has been postulated that MMR in vivo initiates at Okazaki fragment termini in the lagging strand, and at nicks in the leading strand generated by the mismatch-activated MLH1/PMS2 endonuclease. We were able to show that a single ribonucleotide in the vicinity of a mismatch can act as an initiation site for MMR in human cell extracts and that MMR activation in this system is dependent on RNase H2 (Figure 1). As loss of RNase H2 in S. cerevisiae resulted in a mild MMR defect that was reflected in increased mutagenesis, MMR in vivo might also initiate at RNase H2 generated nicks. We therefore proposed that ribonucleotides misincoporated during DNA replication serve as physiological markers of the nascent DNA strand (Ghodgaonkar et al., 2013).
The minimal MMR system could be reconstituted from purified proteins some time ago (Constantin et al., 2005; Zhang et al., 2005), but our genetic screens (Cejka and Jiricny, 2008) and proteomic analysis of the MMR interactome (Cannavo et al., 2007) identified several polypeptides that strongly associate with MLH1 and PMS2, yet are apparently not required for minimal MMR.
In order to learn whether the identified interactors play accessory role(s) in MMR, we set up the reconstituted system in our laboratory, in addition to the in vitro assay that makes use of nuclear extracts of human cells.
We have also set out to study the involvement of nucleases in human MMR. The rationale for this work is the finding that the mutator phenotype of MSH2- and MLH1-deficient cells is considerably stronger than that of cells lacking EXO1, the only exonuclease implicated in MMR to date. This suggests that additional nucleases compensate (at least partially) for the lack of EXO1. One of the candidate nucleases we have been studying is the proofreading activity of polymerase-δ. Using a novel approach developed in our laboratory, we have been able to stably replace the endogenous large subunit of pol-δ with a variant that is error-prone, a variant lacking the 3’→5’ proofreading activity and a third variant affected in both these functions (Figure 2). Phenotypic analysis of these cells revealed that all three cell lines had substantially elevated mutation frequencies, which implies either that their MMR capacity was saturated, or that MMR does not address errors that escape the proofreading exonuclease.
Figure 1. Schematic representation of ribonucleotide-directed mismatch repair. Ribonucleotides incorporated into the nascent strand during replication are removed by RNasH2. Should a mismatch be generated in the vicinity, the MMR system can hijack the strand break arising during the ribonucleotide removal as an initiation site for EXO1-catalysed degradation of the error-containing nascent strand. The single-stranded gasp is stabilised by RPA, until it is filled-in by the replicative polymerase.
Figure 2. Biochemical characterization of polymerase-δ variants expressed in human cells. The enzymes were isolated by affinity chromatography and tested for their ability to extend a radiolabelled 17-mer primer annealed to a 31-mer template. All enzyme variants, wild type (WT), error-prone (EP), proofreading-deficient (PD) or the double mutant (DM) could extend the primer in the presence of all four dNTPs, albeit with varying efficiencies. The exonuclease defect in the PD and DM variants is clearly evident from the lack of degradation products of the 17-mer. All variants were able to incorporate dCMP (but no other nucleotide) opposite the G at position 18 of the template. The figure represents an autoradiograph of a 10% polyacrylamide gel.
MMR and interstrand cross-link repair
In 2007, we identified KIAA1018 as a strong interactor of the MMR protein MLH1. Because the protein is recruited to chromatin by mono-ubiquitylated Fanconi protein FANCD2, it has been renamed FANCD2-associated nuclease 1, FAN1. We could show that FAN1 is an exo/endonuclease, which preferentially cleaves 5’ flaps and D-loops in vitro. We were able to show that human (Kratz et al., 2010) and chicken DT40 (Yoshikiyo et al., 2010) cells lacking FAN1 were hypersensitive to agents that induce interstrand cross-links (ICLs), and that FAN1 deficiency also lowered recombination efficiency and double-strand break repair. Interestingly, although hypersensitivity to ICL-inducing agents is one of the key hallmarks of Fanconi anemia (FA), the FAN1 gene does not appear to be mutated in FA patients. We are now trying to understand the biological relevance and/or importance of the binding of FAN1 to MLH1 and PMS2, as well as identify its molecular role in the processing of ICLs.
The role of MMR proteins in antibody diversification
The generation of our vast antibody repertoire involves three processes: VDJ recombination, somatic hypermutation (SHM) and class switch recombination (CSR). All of these processes irreversibly alter the genome of B cells. Whereas the random recombination of the variable (V), diversity (D) and join (J) regions of the immunoglobulin (Ig) genes takes place in unstimulated B-cells already in the bone marrow, the latter processes are initiated upon antigen stimulation of the cells in germinal centers. SHM/CSR is triggered by activation-induced cytidine deaminase (AID), which is induced in antigen-stimulated B cells and which converts cytosines in certain sequence contexts to uracils. Although uracil processing by base excision repair is generally error-free, in stimulated B cells it gives rise to mutations. Surprisingly, evidence obtained from knock out mouse models and, more recently, also from patients, showed that a subset of these mutations is dependent on MMR. Thus, while MMR is a high-fidelity process, MMR proteins appear to act as mutators during SHM/CSR. Using defined uracil-containing substrates, we could show that base excision repair and MMR compete for the AID-generated U/G mispairs. This interference gives rise to long tracts of single-stranded DNA, which are not efficiently filled-in by the replicative polymerases due to low enzyme concentrations and depleted nucleotide pools outside of S phase. This appears to trigger mono-ubiquitylation of PCNA and recruitment to DNA of translesion polymerases such as polymerase-η. We postulate that the deployment of these error-prone polymerases in the repair of MMR-generated gaps leads to mutations during SHM/CSR (Figure 3).
We are currently attempting to obtain mechanistic insights into the CSR process, using an in vitro system capable of mimicking the AID-triggered in vivo recombination events.
Figure 3. Putative scheme of MMR function during G1- and S-phases of the cell cycle. Lesions bound by MMR proteins outside of S-phase activate non-canonical MMR, during which the endonuclease activity of MutLα introduces nicks into either DNA strand. These might be used for loading of EXO1, which would result in the generation of long single-stranded gaps. Due to low nucleotide pool concentrations and low levels of replicative polymerases, the gaps might persist for some time, which could trigger PCNA ubiquitylation and recruitment of error-prone polymerase(s) such as pol-η. In contrast, lesions generated during S-phase would be repaired with high fidelity, due to the existence of free termini that direct MMR to the nascent strand, the ready availability of dNTPs and higher concentrations of replicative polymerases.
Repair of O6-methylguanine in Xenopus laevis egg extracts
The MMR system has been shown to be involved in the processing of DNA damage other than base/base mismatches and IDLs. It is largely responsible for the cytotoxicity of the mutgenic O6-methylguanine (MeG), such that MMR-deficient cells are up to 100-fold more resistant to killing by methylating agents of the SN1 type than their MMR-proficient counterparts. In order to understand the molecular basis of this resistance, we asked whether DNA substrates carrying defined base modifications are addressed by the MMR system in vitro. We devised a method of preparing such substrates, using a combination of primer extension reactions on single-stranded substrates and a “nickase” – an enzyme capable of incising specifically only a single DNA strand of its recognition sequence. We succeeded in incorporating MeG into our substrates, and were able to show that it is addressed by the MMR system. However, our in vitro MMR assay does not faithfully mirror the process in which a mispair is addressed immediately after it was generated by the polymerase, i.e. in the context of DNA replication. In order to gain insights into the mechanism of postreplicative MMR, we are attempting to make use of MeG present in the template that is undergoing replication in nucleoplasmic extracts of Xenopus laevis eggs. In this system, we can follow replication, repair, DNA damage signalling and possibly also recombination in the same assay.