Biochemical, Structural and Mutational Landscapes of Base Excision Repair Enzymes and Cancer: From Atomic Resolution to Tumor Signatures

Abstract

Purpose

Base excision repair (BER) is the predominant pathway for repairing non-bulky oxidized and alkylated DNA base lesions, and its fidelity depends on the coordinated action of lesion-specific DNA glycosylases and downstream repair enzymes. This review aims to summarize recent structural, biochemical, and genomic insights into three base excision repair enzymes, MUTYH DNA glycosylase, NTHL1 DNA glycosylase, and DNA polymerase β.

Conclusion

This review outlines how MUTYH, NTHL1 and DNA polymerase β protect the genome from mutagenesis, highlights major germline variants associated with disease, and synthesizes the current knowledge on the characteristic single base substitution (SBS) mutational signatures that occur when these repair enzymes are dysfunctional.

Introduction

The base excision repair (BER) pathway is the primary mechanism for removing non-bulky oxidized and alkylated DNA lesions (David et al. 2007; Brooks et al. 2013) (Figure 1). DNA glycosylases initiate the BER pathway by scanning, recognizing, and excising DNA lesions. The eleven human DNA glycosylases are grouped into monofunctional and bifunctional glycosylases: The monofunctional DNA glycosylases, such as MUTYH, possess only glycosylase activity, leaving an abasic site after removal of the DNA lesion. Bifunctional DNA glycosylases, such as NTHL1, possess both glycosylase and lyase activity; they nick the DNA backbone either once (though β-elimination, leaving a 3’-aldehyde) or twice (via β,δ-elimination, leaving a 3’phosphate), which generates a single strand break. Either apurinic endonuclease 1 (APE1) or polynucleotide kinase (PNK) processes the resulting DNA, leaving a free 3’ -OH for DNA polymerase β (Pol β) to incorporate the correct deoxynucleotide opposite the templating base. The DNA ligase IIIα (LIGIII)- X-ray repair cross complimenting 1 (XRCC1) complex (Cappelli et al. 1997) then seals the gap. In short patch (SP) BER, a single base is removed and replaced, whereas in long patch BER two or more nucleotides are removed (Dianov et al. 1992). The following sections will examine three key enzymes of the BER pathway, MutY/MUTYH and Nth/NTHL1 DNA glycosylases, as well as DNA polymerase β, and conclude with an overview of the current understanding of mutational signatures associated with oxidative damage and/or defective repair.

Figure 1: Diagram of the short-patch base excision repair pathway.

The eleven human DNA glycosylases are grouped into monofunctional (MUTYH, MBD4, TDG, SMUG1, UNG, MPG) and bifunctional glycosylases (NTHL1, OGG1, NEIL1, NEIL2, and NEIL3). We note that NEIL3 was found to have a weak lyase activity (Liu M et al. 2010) and was described as being mainly monofunctional (Krokeide et al. 2013)

1-. MutY DNA glycosylases

MutY Structures

MutY enzymes are monofunctional adenine glycosylases that prevent mutations associated with 8-oxoguanine (OG) by removing adenines that are incorporated opposite OG by DNA polymerases during DNA replication (David and Williams 1998; Brieba et al. 2004; David et al. 2007). MutY and its human homolog MUTYH, along with the OG glycosylases, Fpg (bacteria) and OGG1 (human), and the dOGTP hydrolase NUDT1, are part to the “GO” repair pathway (Michaels et al. 1992; David et al. 2007).

The architecture of bacterial and mammalian MutY enzymes comprises an N-terminal catalytic domain and a C-terminal OG recognition domain that are linked by an interdomain connector (IDC) that traverses the major groove of DNA (Fromme et al. 2004; Lee S and Verdine 2009; Russelburg et al. 2020; Nakamura et al. 2021; Trasviña-Arenas et al. 2025). The N-terminal domain of MutY is similar in sequence and structure to other Helix-hairpin-Helix superfamily members, such as EcoNth/NTHL1 (Trasviña-Arenas et al. 2021) (see below) and harbors a catalytic pocket for the extruded adenine and conserved catalytic residues (Guan et al. 1998).The first glimpse of the structure of a full-length MutY bound to DNA was a catalytically inactive D138N Geobacillus stearothermophilus (Gs) MutY bound to an OG:A containing duplex (termed lesion-recognition complex, LRC) (Fromme et al. 2004); however, the adenine base was not fully engaged in the active site, suggesting that the structure represents an inactive conformation. Making the substrate resistant to catalysis by introducing a 2’-F in the deoxyribose sugar enabled capture of a structure with wild-type (WT) Geobacillus stearothermophilus MutY (GsMutY) bound to the OG:FA-duplex (termed the fluorinated lesion recognition structure, FLRC). (Lee S and Verdine 2009) Notably, all of these structures required tethering the enzyme to the DNA via a disulfide to obtain crystals. Several structures were solved by the David and Horvath laboratories using a pyrrolidine nucleotide (1-azaribose or 1N) that harbors a positively charged N at the position corresponding to C1’ in the transition state (TS) (Woods et al. 2016; Russelburg et al. 2020); these structures, referred to as transition-analog complexes (TSAC), have OG or G opposite the high affinity TS mimic 1N; in this case, crystals formed without the need for disulfide crosslinking. The OG-TSAC structure provided mechanistic inferences, confirmed biochemically, that MutY enzymes use a unique double displacement “retaining” mechanism (Woods et al. 2016). The G-TSAC structure revealed the role of a C-terminal “FSH” loop in MutY to detect and engage OG lesions providing a means to select mispaired As for excision over As within T:A bps (Lee AJ et al. 2020; Russelburg et al. 2020).

Mammalian MutYs have an extended IDC of ~60 amino acids. The first structure of MUTYH was of the N-terminal domain with the IDC lacking the C-terminal domain and was not in complex with DNA (Luncsford et al. 2010). This structure revealed that the extended IDC folds into a structure that would be anticipated to project away from the DNA to serve as a “landing pad” for interactions with protein partners. Indeed, the binding sites for APE1, SIRT6 and MSH6 with MUTYH are localized within the IDC (Raetz and David 2019). Mammalian MutYs were later shown to require Zn coordination by Cys residues within the IDC for optimal activity (Engstrom et al. 2014). Molecular dynamics and modeling predicted that the presence of the Zn ion would facilitate DNA substrate engagement; on this basis, this region was termed the “Zinc Linchpin Motif” (Engstrom et al. 2014; Nuñez et al. 2018). A few years later, a structure of the mouse Mutyh protein bound to an abasic site product analog (THF):OG duplex (PDB ID code 7EF9) (Nakamura et al. 2021) revealed the Zn coordination site and portions of the IDC, though much remained disordered. The first crystal structure of human MUTYH bound to an OG:1N duplex (Trasviña-Arenas et al. 2025) (PDB ID code 8FAY) revealed an allosteric network spanning from its [4Fe-4S]²⁺ cluster – a universally conserved cofactor in MutY homologs among all three domains of life – to the enzyme active site. However, this structure lacked the Zn ion, presumably due to its lability during crystallization. The David lab recently elucidated the structure of human MUTYH with both metal cofactors bound (Hashemian & David, unpublished results; PDB ID code 11IP).

MUTYH-associated polyposis (MAP)

MUTYH-associated polyposis (MAP) was first described in a British Family, “Family N”, in 2002 (Al-Tassan et al. 2002; David et al. 2007). The phenotypic presentation in Family N resembled that of familial adenomatous polyposis (FAP) with colonic polyposis and adenomas; surprisingly, MAP was not associated with inherited mutations in the adenomatous polyposis coli (APC) gene. Early clues to the genetic origin of MAP were the predominance of G:C to T:A transversion mutations in APC observed in tumors from afflicted family members, implicating defective repair of OG. Remarkably, this approach of using mutational patterns to identify the genetic origin of the polyposis phenotype happened well before the widespread adoption of mutational signature analysis. Subsequent DNA sequencing revealed that afflicted members of Family N were compound heterozygotes for two missense variants in MUTYH – Y179C and G396D – that are now considered the founder mutations of MAP(Al-Tassan et al. 2002).

The David laboratory played a key role in the discovery of MAP by demonstrating that disease-associated amino acid variants markedly reduce OG:A lesion recognition and adenine glycosylase activity of MUTYH (Al-Tassan et al. 2002; David et al. 2007). At the time – and in the 24 years since –the inability to overexpress reasonable amounts of highly active MUTYH protein bottlenecked efforts to characterize its in vitro enzyme activity. Recent improvements in MUTYH expression are due in part to co-expression of the [4Fe-4S]²⁺ cluster machinery and chaperone proteins to promote insertion of metal cofactors and facilitate proper folding. In 2002, due to the high conservation of Tyr179 and Gly396 positions among MutY homologs, functional profiling of the corresponding variants in the bacterial enzyme were used to reveal functional defects; both were found to be catalytically compromised, supporting the hypothesis that family members with the variants would have reduced OG:A repair and increased frequencies for genomic C>A mutations (Al-Tassan et al. 2002; Chmiel et al. 2003; Livingston et al. 2005). Notably, the findings using the bacterial enzymes with these two variants have been recapitulated in activity analyses with the mouse (Pope et al. 2005) and human MUTYH (Trasviña-Arenas et al. 2025) recombinant proteins.

Since the identification of Family N, the correlation of MUTYH variants with colorectal polyposis and adenomas, and high likelihood for colorectal cancer has been well-established, leading to the designation of MUTYH as a colorectal cancer susceptibility gene and the recognition of the associated predisposition syndrome as MUTYH-associated polyposis (MAP) (David et al. 2007; Banda et al. 2017; Raetz et al. 2020). Beyond the two founder variants, there have been a large number (>3,000) of germline and somatic mutations identified in the MUTYH gene with a large swath being missense variants. In addition, MUTYH variants have been found to be associated with other cancers (Poulsen and Bisgaard 2008). Many of these MUTYH variants are designated as “variants of uncertain significance” or VUS. These variants are localized throughout the protein sequence, in domains that are known to be involved in OG recognition and adenine excision, but also in regions that play indirect roles in enzyme function or are involved in mediating protein-protein interactions (Figure 2).

Figure 2: Structural overlay of germline MUTYH missense variants reported by ClinVar.

At the time of analysis, ClinVar reports 2,386 unique MUTYH missense variants, represented by a total of 5,643 independent submissions. Among the unique variants, 73% are classified as variants of uncertain significance. Notably, 21% of all ClinVar submissions localize to residues within or proximal to the [4Fe-4S]²⁺ cluster and interdomain connector (IDC), the region that harbors the Zn cofactor. To visualize variant distribution, ClinVar submissions were binned by residue position from amino acids 1–549 according to the MANE Select transcript NM_001128425.2, and a two-dimensional heatmap was generated in which submission frequency per residue was encoded using a color gradient of 0 (purple) to 30 (gold) submissions, with >30 submissions shown in red. Catalytic residues Glu134 and Asp236, as well as known founder mutation loci Tyr179 and Gly396, are highlighted. Gln338 is additionally annotated as a residue proposed to represent a polymorphism-prone locus based on population and clinical variant data. The heatmap was subsequently mapped onto a crystal structure of human MUTYH bound to duplex DNA containing an OG:THF product analog (PDB accession pending) to illustrate the three-dimensional localization of variant submissions. A magnified view highlighting the spatial relationship between the [4Fe-4S]²⁺ cluster and the active site is shown on the left.

MAP MAPs: Visualizing MUTYH variants on the MUTYH structure

Structure-guided “MAP maps,” in which MUTYH variants are visualized on the MUTYH structure, have proven valuable for predicting and rationalizing the functional impact of disease-associated substitutions. Notably, however, there is substantial variability in agreement among legacy (e.g., PolyPhen-2, SIFT), clinical (e.g., ClinVar, LOVD), and more recently developed AI-based tools (e.g., AlphaMissense) as predictors of relevant disease-associated substitutions (Cheng et al. 2023). Among these, AlphaMissense provides a simple assessment of pathogenicity (benign, ambiguous, pathogenic), which generally agrees with our experimental data on MUTYH variants, with limitations in unstructured regions of the protein and regions associated with protein-protein interactions. To contextualize these predictions within the broader landscape of clinically-observed variation, we annotated germline missense variant submissions from ClinVar, which provides the largest aggregated dataset of clinically reported variants and enables assessment of variant burden across the MUTYH sequence (Cheng et al. 2023). Aggregation of germline missense variants submitted to ClinVar is typically performed using the NCBI MANE Select or MANE Plus Clinical transcript to ensure consistent clinical annotation, as these transcripts serve as community-agreed reference standards for clinical reporting. For MUTYH, the NCBI MANE Plus Clinical transcript encodes a 549 amino-acid protein that is consistent with the most recent crystal structures of human MUTYH bound to DNA (PDB ID: 8FAY), enabling structure-based visualization of germline missense variants and revealing patterns of clinical burden across the protein (Figure 2)(Trasviña-Arenas et al. 2025).

A structural basis for dysfunction of the two founder MAP variants, Y179C and G396D, and other variants localized in the N-terminal catalytic domain and C-terminal OG recognition domain can be gleaned using the bacterial GsMutY structures. Indeed, the defective activity of Y179C and G396D, and the more hampered activity of the former, makes sense upon inspection of the GsMutY FLRC structure (Lee S and Verdine 2009). The Tyr in the FLRC structure corresponding to Y179 intercalates 5’ of the OG of the OG:A mismatch to disrupt base-stacking interactions, facilitate adenine base-flipping in the active site pocket and stabilize MUTYH on the mismatch (Livingston et al. 2005; Lee S and Verdine 2009). Gly396 is located in a tight turn region within the CTD that closely interacts with the DNA backbone near the OG phosphodiester linkage (Lee S and Verdine 2009) such that any amino acid substitution at this position would be destabilizing (Livingston et al. 2005).

MUTYH variants are also localized at or near the catalytic residues. The variants D236N, E134V and E134G represent alterations of the two key catalytic residues in human MUTYH: the Asp plays the role of the nucleophile in the base excision step and makes the transient covalent acetal intermediate, while the Glu is involved in protonation of the adenine that enhances the leaving group potential and also participates in deprotonating the water molecules that hydrolyze the enzyme-acetal intermediate to form the abasic site (Woods et al. 2016). Additionally, the Asn associated with the human variant N238S, is localized in MutY/MUTYH structures within H-bonding distance to the Asp residue, D236, suggesting a role in modulating the pK_a of the Asp nucleophile and aiding in its proper positioning (Demir et al. 2023; Trasviña-Arenas et al. 2025). Notably, the corresponding catalytic Asp in NTHL1 is a locus for SNP D239Y (rs3087468); the purified variant has been shown to be inactive as a glycosylase and its presence in cells promotes cellular transformation and genomic instability (Galick et al. 2013).

Approximately 21% of all clinically reported germline variants in MUTYH localize proximal to its two metal cofactors – the [4Fe-4S]²⁺ cluster and the Zn ion – both of which are distal to the catalytic active site (Figure 2). This enrichment suggests that variants can be signposts for functionally important regions in BER glycosylases like MUTYH. The first realization that mammalian MutY enzymes may harbor a second metal site beyond the [4Fe-4S]²⁺ cluster emerged during inspection of variant clustering in this region along with multiple sequence alignments of mammalian MutY homologs (Engstrom et al. 2014). The presence of three conserved Cys residues in the IDC suggested a metal binding site and subsequent inductively coupled plasma – mass spectrometry (ICP-MS) based metal analysis revealed that this was a single Zn ion. Although the fourth Zn ligand was initially proposed to be a Cys residue in the N-terminal domain (Nuñez et al. 2018), more recent structural analysis of mouse Mutyh instead identified a His residue as the fourth ligand (Nakamura et al. 2021).

Insight into functional roles for the [4Fe-4S]²⁺ cluster originated from inspection of locations of cancer-associated variants (CAVs) near the [4Fe-4S]²⁺ cluster within the MUTYH TSAC.(Trasviña-Arenas et al. 2025) This analysis revealed a hydrogen-bond network made up of CAVs that spans the 20 Å between the [4Fe-4S]²⁺ and the active site (Figure 2). The connection involves four highly conserved residues starting with the [4Fe-4S]²⁺ cluster ligand Cys290 that H-bonds with Arg241, which in turn interacts with Asn238, which ultimately hydrogen-bonds to the catalytic Asp236. Notably, severing the Arg-Asn bond by the CAV mutations R241Q or N238S ablated catalytic activity while preserving OG:A DNA binding. Activity assays and molecular dynamics indicated that severing the bridge altered the position and protonation state of the catalytic Asp, suggesting that DNA binding at the [4Fe-4S]²⁺ cluster allosterically regulates adenine excision at the active site. A similar H-bonding network can be found in all of the [4Fe-4S]²⁺ cluster containing HhH-GPD superfamily of BER glycosylases (Trasviña-Arenas et al. 2025), suggesting a similar allosteric mechanism is operative in these enzymes to fine-tune repair or disable repair in the face of oxidative stress.

2-. Nth DNA glycosylases

Endonuclease III/Endo III or Nth was reported in 1976 as an enzyme capable of nicking DNA (hence the name endonuclease)(Radman 1976). It appears to be the same E. coli enzyme previously known as X-ray endonuclease that was shown to specifically nick X-irradiated (Strniste and Wallace 1975) and OsO4-treated DNA (Armel et al. 1977). Nth was later found to be conserved in bacteria, yeast, and mammals. Further studies showed that Nth is a bifunctional DNA glycosylase possessing both N-glycosylase and abasic site (AP) lyase activities (Brooks et al. 2013; Das et al. 2020).

Nth was originally shown to excise thymine glycol (Tg) (Demple and Linn 1980; Katcher and Wallace 1983). Using gas chromatography /mass spectrometry (GC/MS), Miral Dizdaroglu and colleagues found that Nth excises a number of oxidized pyrimidines, in addition to Tg: 5,6-dihydrothymine (DHT), 6-dihydrouracil (DHU), uracil glycol (Ug), 5-hydroxy-5-methylhydantoin, 5-hydroxy-6-hydrothymine, 5-hydroxy-6-hydrouracil, and 5-alloxan (Dizdaroglu et al. 1993). A later study revealed additional oxidized pyrimidines such as 5,6-dihydroxyuracil (5-OHU), and 5,6-dihydroxycytosine (5-OHC) and a purine-derived substrate, 6-diamino-5-formamidopyrimidine (Dizdaroglu et al. 2000). These findings show that even though Nth/NTHL1 belong to the HhH/GPD superfamily of DNA glycosylases like MutY/MUTYH and share structural elements such as an iron sulfur [4Fe-4S]²⁺ cluster and helix-hairpin-helix (HhH) motif, these glycosylases vary substantially in terms of substrate specificity (Brooks et al. 2013; Das et al. 2020). Unlike MUTYH, which specifically recognizes and excises A opposite OG, Nth/NTHL1 has a broad substrate specificity (Cadet et al. 2000), some of which is recapitulated in single base substitution (SBS) signature 30 (see below).

The structure of E. coli Nth solved in 1992 by the Tainer lab (Kuo et al. 1992) revealed a two-domain architecture with an N-terminal domain harboring the iron-sulfur [4Fe-4S]²⁺ cluster and a C-terminal domain comprising the HhH DNA binding motif. Later a crystal structure of Geobacillus stearothermophilus Nth crosslinked to DNA via borohydride trapping provided a first glimpse at a protein-DNA complex (Fromme and Verdine 2003). Nth glycosylases harbor two charged residues important for catalysis, an aspartate and a lysine (Dodson et al. 1994): The aspartate in the N-terminal [4Fe-4S]²⁺ cluster domain deprotonates the lysine side chain located in the six-helical barrel C-terminal domain, a key catalytic nucleophile that forms a transient covalent Schiff-base intermediate with the abasic site (Ikeda et al. 1998). The Nth crystal structures showed that the DNA binds along a cleft between the two domains. The structures further revealed that the [4Fe-4S] domain of Nth makes contacts to the DNA backbone and provided a structural basis for why guanine is preferred as an opposite base over adenine (Fromme and Verdine 2003). They did not, however, reveal why DNA containing a pyrimidine as an opposite base is weakly bound.

The structure of an active human NTHL1 protein construct (aa 64–312) was solved in 2021 and revealed an unprecedented open conformation of the enzyme (Carroll et al. 2021) (Figure 3), with a distance of about 23 Å between the catalytic lysine and aspartate residues. This is in contrast to the bacterial Nth where the unliganded and DNA-bound structures are both in the “closed” conformation and nearly superimposable (Kuznetsov et al. 2015) (Figure 3). The observed open form of NTHL1 led to the hypothesis that the enzyme must close upon binding a DNA lesion in order to assemble the active site. After processing, NTHL1 would be predicted to revert back to the open conformation and resume scanning the DNA for lesions. In human NTHL1, the two domains are connected by two linkers: linker 1 (aa 104–125) and linker 2 (aa 230–240). A chimera protein engineered by replacing the longer and less conserved of the two linkers in human NTHL1 (linker 1) with the corresponding shorter linker from bacterial Nth yielded an enzyme in the closed conformation, suggesting that the extended linker is necessary for the interdomain rearrangement in the human enzyme. A movie made with PyMOL illustrating the open-closed transition in NTHL1 is shown as supplemental information (Supplemental movie 1, made with PDB files 7RDS and 7RDT). Molecular dynamics simulations of human NTHL1 described three distinct conformations: open, closed, and “bundle”, with only two of the conformations (open and bundle) being described as stable (Odstrcil et al. 2024). Interestingly, when given the NTHL1 DNA sequence, AlphaFold predicts the closed conformation (Supplemental Figure 1)(Jumper et al. 2021).

Figure 3. Molecular structures of human NTHL1 and bacterial Nth.

Human NTHL1 is expected to undergo a large conformational change upon binding DNA, unlike bacterial Nth. Human NTHL1 (PDB 7RDS; 7RDT; Carroll et al. 2021)(open conformation: light blue; closed conformation: dark blue; linkers 1 and 2: green; catalytic residues: yellow); Geobacillus stearothermophilus Nth (GsNth) (shown in beige with DNA in light blue; PDB 1ORN; Fromme et al. 2003); Escherichia coli Nth (EcoNth) (shown in pink; PDB 2ABK; Kuo et al. 1992).

NTHL1-associated polyposis (NAP)

Similarly to MUTYH, NTHL1 has been linked to familial inheritance colorectal cancer (CRC), and adenomatous polyposis (Weren et al. 2015). Multiple studies have found that NTHL1-associated polyposis (NAP) has a biallelic germline nonsense NTHL1 mutation, p.Q82* (previously named p.Q90*; the discrepancy arises from isoforms being produced by alternative initiation), which renders the DNA glycosylase inactive. Biallelic mutations have been found in multiple tumor types, notably colorectal, breast and endometrial cancers (Grolleman et al. 2019). The prevalence of NAP is estimated to be 1:114,770 in individuals of European decent (Weren et al. 2018), which is rarer than MUTYH-associated polyposis (MAP) (estimated incidence 1:5,000–40,000 in Europe). The NAP tumors revealed a strong C>T transition pattern (Weren et al. 2015), suggesting that they likely stem from failure to repair an oxidized lesion other than OG (see below). Recent studies have reported NTHL1 as a recessive multi-tumor susceptibility gene (Nurmi et al. 2023).

NTHL1 variants

In addition to p.Q82*, three other nonsense NTHL1 variants, p.Y130*, p.R153*, and p.Q287*, were assayed and all were found to be defective in repairing 5-OHU lesions, unlike the missense variants R19Q, V179I, V217F, and G286S (Shinmura et al. 2019). In contrast, the same G286S variant was found to have impaired glycosylase and lyase activity on both Tg:A and DHU:G, and reduced lyase activity on AP:G (Doublié lab, unpublished data).

SNP rs3087468, found in 6.2% of the population, is a single amino acid substitution where D239 is replaced by a tyrosine, D239Y. Not surprisingly, because this aspartate is a catalytic residue, the resulting glycosylase variant is inactive on Tg- and DHU-containing DNA (Galick et al. 2013). In collaboration with Miral Dizdaroglu, the Sweasy lab showed that D239Y is inactive for removal of a variety of DNA base lesions beyond Tg and DHU, namely 5-OHC, 5-OH-5-MeHyd, FapyAde, 8-OH-Ade and FapyGua (Kant et al. 2022). Further, expression of this variant in MCF10-A non-transformed mammary epithelial cells led to genomic instability and cellular transformation (Galick et al. 2013). A knockin mouse model carrying the mutation corresponding to D239Y, Nthl1 D227Y, was used to isolate murine embryonic fibroblasts (MEFs) and to investigate the cellular effects of the mutation. MEFs derived from the D227Y mouse had decreased glycosylase activity on Tg and exhibited an increase in double-strand breaks, genomic instability, and replication stress (Marsden et al. 2022). In Nthl1^−/− mice, it was shown that a back-up DNA glycosylase like Neil1 can repair oxidized base damage and thus compensate for the absence of functional Nthl1 (Takao et al. 2002). However, in the case of NTHL1D239Y/Nthl1D227Y, the malfunctioning glycosylase is thought to recognize, bind and stay on the DNA lesion, which would block access to other glycosylases and prevent repair (Marsden et al. 2022).

Another rare SNP (rs2302172), the missense mutation R33K, is located in the flexible N-terminal extension of NTHL1 (aa 1–63) that contains both a mitochondrial transit peptide and nuclear localization signal and is dispensable for activity (up to residue ~80) (Liu X and Roy 2002; Carroll et al. 2021). The glycosylase and lyase activity of the variant were shown to be similar to WT using Tg-containing oligonucleotides. Further analysis in collaboration with Miral Dizdaroglu using irradiated calf thymus DNA showed that R33K excises similar levels of nine DNA base lesions, as gauged by GC-MS/MS techniques (Marsden et al. 2020). Expression of NTHL1 R33K in MCF10-A cells induces cellular transformation, as measured by increased proliferation and anchorage-independent growth (Marsden et al. 2020). Since the enzymatic activity of the variant is similar to WT, one possible explanation for the effect of the mutation in cells is that the flexible N-terminal domain in NTHL1 serves as a “landing pad” for other protein factors and that the mutation might affect protein-protein interactions. Known interactors of NTHL1 include Y-box protein (Marenstein et al. 2001), XPG (Bessho 1999), PCNA and P53 (Oyama et al. 2004).

In addition to having multiple germline variants, NTHL1 is reported to be upregulated in some cancers, such as non-small cell lung cancer (NSCLC) (Limpose et al. 2018). Overexpression of NTHL1 in non-transformed human epithelial cells leads to replication stress and cellular transformation (Limpose et al. 2018). As mentioned above, XPG is one of the proteins known to interact with NTHL1. It is possible that the overexpression of NTHL1 sequesters XPG from its important function in homologous recombination and leads to the use of more error-prone processes for double-strand break repair.

3-. DNA Polymerase β

In the short-patch base excision repair (SP BER) pathway, DNA Polymerase β (Pol β) is responsible for selecting and inserting the correct nucleotide following recognition and removal of lesions by a DNA glycosylase (Figure 1). The role of Pol β in DNA repair is critical for genomic stability, and Pol β cancer variants have been shown to induce increased mutation frequency and cellular transformation (Lang et al. 2004; Yamtich et al. 2012). We note that knocking out Pol β in mice is embryonic lethal, unlike MUTYH and NTHL1 (Sugo et al. 2000; Rognlien et al. 2018). Pol β is therefore essential for embryonic development, possibly because other repair polymerases cannot fully compensate for its loss during early embryogenesis, or because it participates in other essential cellular processes, such as BER-dependent DNA demethylation. DNA Pol β is the smallest human DNA polymerase and does not have a proofreading exonuclease domain, resulting in a modest fidelity of one mutation every 3,000 – 7,000 single base insertions (Kunkel and Loeb 1981; Kunkel 1985). Because of its small size and importance in genome integrity, Pol β has served as a model for DNA synthesis and there is a wealth of structural and biophysical information on Pol β and its variants. Still, one of the remaining complex questions is how polymerases accurately select and incorporate nucleotides. The insight gleaned by structural and kinetic probing of Pol β cancer variants on the mechanism of faithful dNTP selection by Pol β will be the focus of this section.

Pol β structure and reaction pathway

Pol β is composed of an 8 kDa lyase domain (aa 1–90) and a 31 kDa polymerase domain which is further subdivided into the thumb (aa 91–150), palm (aa 151–261) and fingers (aa 262 – 335) subdomains (Figure 4A). The lyase domain is responsible for removal of the 5’ deoxyribose phosphate (dRP) group that may be present after removal of the damaged base. Pol β also contains a helix-hairpin-helix (HhH) motif that is responsible for binding downstream of the abasic site. This, coupled with another HhH motif found in the thumb domain that binds upstream, helps orient DNA in the binary complex. The palm domain contains catalytically active residues which coordinate the divalent metal cations that assist in DNA synthesis and in concert with the fingers domain, make up the dNTP binding site in the ternary complex. Transition from binary (open) to ternary (closed) is characterized by fingers subdomain movement (Figure 4B).

Figure 4. Pol β structures and kinetic scheme.

(A) DNA polymerase β architecture (PDB ID: 1BPY, ternary correct) with highlighted lyase (green) domain and thumb (cyan), palm (blue) and fingers (purple) subdomains. (B) Alignment of binary (blue, 1BPX) and ternary (red, 1BPY) Pol β complexes. α-helix N (N-helix) in the fingers subdomain (marked by the black dashed box) rearranges to form the closed ternary enzyme. (C) Reaction scheme of Pol β in short-patch BER, with PDB codes given for each structure (1BPD, 1BPX, 1BPY, 1BPZ (Sawaya et al. 1994) and (Freudenthal et al. 2013). Figure adapted from (Liptak et al. 2018).

Pol β shares a similar reaction pathway with the other DNA polymerases (summarized in Figure 4C) (Raper et al. 2018). Briefly, apo Pol β binds single-nucleotide gapped DNA (sgDNA) to form the binary complex, followed by dNTP binding. This triggers rapid closing of the fingers subdomain to surround the dNTP in the binding pocket, resulting in the ternary complex. Other conformational rearrangements then likely occur. For Pol β a pre-chemistry, non-covalent step (NCS) has been detected using a stop-flow Fӧrster resonance energy transfer (FRET) assay to probe the rates of conformational steps in the reaction pathway (Towle-Weicksel et al. 2014). Although still unclear, this step likely plays a role in orienting the binding pocket for catalysis (Schlick et al. 2012; Towle-Weicksel et al. 2014). Finally, chemistry occurs with phosphodiester bond formation, followed by fingers re-opening, pyrophosphate release and DNA dissociation.

Use of dNTP analogues to study the effect of leaving group basicity on k_pol supports that chemistry is the rate determining step for both correct and incorrect nucleotide insertion by Pol β (Sucato et al. 2008; Oertell et al. 2014). Intriguingly, coupling this technique with stop-flow FRET, which can determine rates for the pre-chemistry conformational steps, revealed that there is also a linear free-energy relationship with the fingers closing step during correct nucleotide incorporation (Alnajjar et al. 2024). Further probing revealed that the orientation of the dNTP phosphate group in the binding pocket, which is important for polymerization, is impactful to the rate of fingers closing and suggests that conformation rearrangements can influence chemistry (Alnajjar et al. 2024).

The role and importance of conformational rearrangements in DNA polymerase selectivity has been highly debated. In 1958, Koshland introduced the induced fit model for enzymes which hypothesized that conformational movements caused by substrate binding can result in an arrangement of the binding pocket that promotes catalysis (Koshland 1958). Later it was argued that for enzymes in which catalysis is the rate determining step, conformational changes prior to chemistry cannot affect selectivity (Fersht 1974). This conclusion is based upon the assumption that binding of correct and incorrect substrates result in the same conformation changes, a point highlighted by Post and Ray who posited that selectivity can be achieved by conformational changes if they result in different geometries of the binding pocket for correct and incorrect binding (Post and Ray 1995). Since then, a wealth of structural, kinetic and modelling data have supported the idea that DNA polymerases achieve selectivity in some part from the conformational movements prior to chemistry, although a specific model remains elusive (Tsai and Johnson 2006; Raper et al. 2018; Das et al. 2020).

Like all DNA polymerases, Pol β must accurately select the correct nucleotide out of a pool of similarly structured molecules and change that preference in successive rounds of insertion. Polymerase fidelity (Equation 1) is a measure of how well the enzyme can select (dNTP binding affinity, K_d) and insert (insertion rate, k_pol) competing nucleotides (correct vs incorrect). For Pol β these parameters are calculated from single-turnover kinetic assays (Joyce 2010). Studying Pol β variants with altered fidelities is an invaluable means to identifying residues which have unique function in selection and incorporation of correct vs incorrect dNTPs. Specifically, cancer variants will be discussed grouped by how their altered fidelity is driven by a lack of discrimination at the level of binding or polymerization. For an in depth review of Pol β cancer variants see recent review (Sawyer and Sweasy 2022).

Equation 1:

Fidelity calculation

$$fidelity=\frac{[{(\frac{k_{\mathit{pol}}}{K_{d,\mathit{app}}})}_{\mathit{correct}}+{(\frac{k_{\mathit{pol}}}{K_{d,\mathit{app}}})}_{\mathit{incorrect}}]}{{(\frac{k_{\mathit{pol}}}{K_{d,\mathit{app}}})}_{\mathit{incorrect}}}$$

K_d determinant – dNTP binding and fingers subdomain movement

Ground state binding of dNTPs by the Pol β-sgDNA binary complex is measured by the dissociation constant (K_d), a thermodynamic parameter that reflects the binding equilibria that precedes the rate determining step (Joyce 2010). DNA polymerases use a template base to select the incoming nucleotide, but early studies indicated that hydrogen bonding from Watson-Crick base pairing is not sufficient to explain fidelity (Goodman 1997). It has been hypothesized that binding pocket geometry and base stacking are also important contributors to binding affinity (Goodman 1997; Moran et al. 1997). This was confirmed by structures of DNA polymerases in complex with DNA and nucleotide (Doublié et al. 1999).

Isoleucine 260 is a hydrophobic residue buried in the hinge region, a track of amino acids that connect the fingers and palm subdomain. Mutation of this residue to methionine was identified in prostate cancer and expression of this variant in mouse cells results in cellular transformation (Dalal et al. 2005; Donigan, Tuck, et al. 2012). In vitro, I260M has an altered fidelity due to lower binding of the correct dNTP, while binding of the incorrect dNTPs remain unchanged compared to wild type (WT) (Dalal et al. 2005; Fijen et al. 2024). Interestingly, mutation of I260 to glutamine also changes Pol β fidelity, but primarily by binding incorrect nucleotides with higher affinity (Starcevic et al. 2005; Klvana et al. 2012).

Follow-up work with I260Q combining NMR, X-ray crystallography and FRET assays revealed that I260Q has an altered dynamic between the binary and ternary state, namely the binary complex is in a pre-closed conformation, mimicking the WT ternary state (Liptak et al. 2018). It was hypothesized that the increased affinity that I260Q has for incorrect nucleotides could be because it is already poised in a closed conformation for their binding. I260M, which binds correct dNTP, with lower affinity than WT Pol β, was also subject to FRET analysis, but fingers closing was not detectible for correct insertion indicating that either the fingers do not close appreciably for I260M or that too small of a population undergoes movements to be detectable (Fijen et al. 2024). Together, these mutations to this key residue in the hinge region illustrate how fingers movement (or the lack of) is critical for selection of nucleotides at the level of binding affinity.

Outside of the hinge region, other variants have also been shown to modulate fidelity at the level of K_d through modulation of fingers movement. Glutamic acid 288 is a solvent accessible residue located at the end of the N-helix within the fingers subdomain which is responsible for contacting the DNA in the closed ternary complex. The E288K mutant was identified in human colorectal and breast cancer (Donigan, Sun, et al. 2012; Ciriello et al. 2015). In cells, expression of E288K leads to an increase in mutation frequency at A:T base pairs and in vitro work showed that it is a sequence specific mutator opposite template A (Murphy et al. 2012). This and later work confirmed that the fidelity differences of E288K are due to a worse binding affinity than WT for dTTP opposite template A coupled with an increased affinity for incorrect dNTPs (Klvana et al. 2012; Murphy et al. 2012; Mahmoud et al. 2017). Kinetic analysis by stop-flow FRET revealed that E288K has a faster rate of fingers closing and a slower rate of reopening (dNTP release) compared to WT leading to the conclusion that E288K stabilizes the ternary complex (Mahmoud et al. 2017). This is supported by previous molecular modelling data of E288K that showed the mutated lysine residue is able to positively interact with the negative phosphate backbone of the DNA in the ternary complex (Murphy et al. 2012).

Another variant that selectively modifies dNTP binding affinity is E295K. Identified in human gastric cancer, expression of this variant in mouse cells led to cellular transformation (Lang et al. 2007). Like E288K, E295K is a sequence specific mutator opposite template A, but is driven by a much more dramatic loss in binding to correct dTTP (Eckenroth et al. 2017). Interestingly, similar results for E295A were also seen indicating that this could be due to a loss in interaction of the glutamic acid (Kraynov et al. 2000). Structural investigation of the E295K mutant with correct or incorrect dNTP bound in the ternary complex showed that it formed a closed state for incorrect (dCTP) but not correct. This again highlights the importance of finger movements for fidelity as reflected in dNTP binding affinity.

k_pol determinant – Chemistry and binding pocket alignment

Formation of the phosphodiester bond (catalysis or chemistry step) is rate determining for Pol β when inserting both correct and incorrect nucleotides. For WT, structural data indicates that the slower rate of catalysis for incorrect vs correct is due to altered geometry of the binding pocket due to displacement of the DNA and the incoming dNTP (Freudenthal et al. 2013). Many Pol β variants show decreased catalysis that equally affects incorrect and correct incorporation so that there is no impact to their fidelity. Therefore, Pol β variants that are k_pol determinant have mutated residues that differentially affect the rate of incorrect or correct insertion. One such residue is tyrosine 265, located in the hinge region. The Pol β Y265C variant was shown to have a mutator phenotype in vivo and in vitro (Opresko et al. 1998; Clairmont et al. 1999). When expressed in mice, they developed an autoimmune pathology strongly resembling systematic lupus erythematous (Senejani et al. 2014; Rahim et al. 2022). Kinetic investigation of Pol β with the residue mutated to histidine revealed an ~100 fold decrease in k_pol for correct, while k_pol for incorrect remained the same (Shah et al. 2001).

Another mutant, K289M, shows similar trends in k_pol differences. This variant in the N-helix region of the fingers subdomain was identified in human colorectal cancers and in mouse cells was shown to be a sequence specific mutator at G:C base pairs within a track of adenines, notably in the APC sequence, known to be highly mutated in colorectal cancers (Lang et al. 2004). Further characterization showed that expression of K289M resulted in cellular transformation in mouse cells (Sweasy et al. 2005). dGTP and dCTP analogues with template C were used to probe leaving group pK_a on k_pol rates and had the interesting discovery that while chemistry remained rate determining for incorrect insertion, it was not for correct insertion by K289M (Alnajjar et al. 2017). Follow-up studies using stop-flow FRET assays confirmed that for correct insertion of dGTP opposite template dC the non-covalent step becomes rate limiting for K289M (Alnajjar et al. 2019).

Although still unclear, the role of the NCS could be important for active site alignment after fingers closing. Initial probing of the NCS revealed that it is still present with calcium cations (which do not allow for chemistry to occur) but is not detectible if there is no 3’OH group on the primer (Towle-Weicksel et al. 2014). This indicates that this step may be involved in orienting the nucleophilic 3’OH for chemistry, but further investigation of this step in correct vs incorrect incorporation is critical to make stronger conclusions about the contribution of the NCS to WT Pol β fidelity.

4-. Mutational signatures

Mutation Signatures Associated with Oxidative Damage and Repair-Deficiency

As mentioned above, MUTYH, NTHL1, and Pol β repair a chemically diverse set of DNA lesions, including oxidized purines and pyrimidines and abasic sites, generated by oxidative stress. Several mutational signatures of unrepaired oxidative DNA damage are now well defined through large-scale sequencing and experimental validation (Figure 5). The COSMIC mutational signature SBS18 (Alexandrov et al. 2020), dominated by C>A substitutions at NpCpG contexts (when normalized to target size), is the canonical signature of oxidative stress and is often referred to as the “ROS signature.” Cancers arising in tissues chronically exposed to these conditions, such as the lung, gastrointestinal tract, and kidney, frequently display SBS18. Its burden in these cancers is typically low, ranging from 0.1 to 10 mutations per Mb, yet this signature is present in roughly 30% sequenced tumors. However, SBS18 reflects OG mispairing with adenine during replication (Shibutani et al. 1991), resulting in G:C>T:A substitutions that are conventionally displayed as the complementary pyrimidine substitution (i.e. C>A) in signature analysis (Alexandrov, Nik-Zainal, Wedge, Campbell, et al. 2013). Therefore, SBS18 is more precisely the OG signature.

Figure 5:

Oxidative damage mutation signatures. (A) COSMIC signatures linked to oxidation damage. SBS18 is caused by OG DNA. ID5 correlates with SBS18 in human cancers and a similar signature is induced by oxidants in cell culture (Cordero et al. 2024), although the causative lesion in unidentified. (B) Signatures in tumors associated with glycosylases-deficiency. Biallelic mutation of NTHL1, MUTYH, and OGG1 have been linked to SBS30, SBS36, and SBS108 (Drost et al. 2017; Grolleman et al. 2019; Degasperi et al. 2022; Robinson et al. 2022), respectively.

Experimental validation of SBS18’s origination from OG lesions was based initially on the similarity of the signature with mutations induced by chemicals that create OG lesions, such as KBrO₃. KBrO₃ treatment of human cell lines primarily produces C>A substitutions, with similar sequence preferences as SBS18 (Kucab et al. 2019; Cordero et al. 2024). However, some differences, such as higher mutagenicity of CC dinucleotides in KBrO₃-treated cells indicates that the OG-inducing oxidant likely influences lesion formation and the subsequently derived mutation spectrum (Cordero et al. 2024). Consequently, mutation signatures associated with deficiency of OG repair enzymes provide stronger evidence for the etiology of SBS18. OGG1 recognizes and excises OG when paired with cytosine (Radicella et al. 1997), while a different glycosylase, MUTYH, can remove adenines mispaired with OG (Ohtsubo et al. 2000) to limit the mutations associated with this lesion. Human OGG1 contains the two α-helical domains common to all HhH–GPD family members such as NTHL1 and MUTYH, and a third antiparallel β-sheet domain similar to AlkA (Bruner et al. 2000). Like most other DNA glycosylases, OGG1 flips the lesion out of the DNA double helix. Interestingly, hOGG1 and other OG-DNA glycosylases like OGG2 and AGOG specifically recognize the protonated N7 position of OG, rather than the oxygen at C8 (Faucher et al. 2012). OGG1 knockout cells display mutation spectra similar to SBS18 (van den Boogaard et al. 2021; Zou et al. 2021). The frequent accumulation of SBS18 mutations in cultured wild-type cells makes identifying an OGG1 signature difficult as loss of the glycosylase merely emphasizes an existing signature. More recently tumors containing biallelic disruption of OGG1 carry SBS108 (Degasperi et al. 2022), which is highly similar to SBS18. MUTYH-deficiency produces its own distinct, but related signature, SBS36 (Pilati et al. 2017). SBS36, like SBS18 is dominated by C>A transversions arising from the fixation of OG:A mismatches. However, slight differences in trinucleotide sequence preferences between SBS18 and SBS36 (Robinson et al. 2022) indicate that repair specificities also influence the resulting mutation spectra of the same lesion. SBS36 is observed in MUTYH knockout cells (van den Boogaard et al. 2021) but was first identified in tumors with biallelic loss of MUTYH function, arising somatically or from individuals suffering from MAP (Pilati et al. 2017). The strong mechanistic correspondence between lesion chemistry, repair deficiency, and mutational outcome establishes oxidative mutagenesis as one of the most experimentally validated endogenous processes shaping the cancer genome.

Despite the focus on SBS18 as the ROS signature (Alexandrov, Nik-Zainal, Wedge, Aparicio, et al. 2013; Alexandrov et al. 2020), other mutation signatures are likely associated with oxidative DNA damage, while the cognate lesions underlying them remain elusive. Treatment of immortalized human retinal pigment epithelial cells (RPE-1-hTERT) with KBrO₃ causes a dramatic increase in C>A substitutions consistent with OG being the primary lesion the oxidant forms. However, this treatment also increases single base T-deletions preferentially outside of homopolymer tracks, constituting an indel signature resembling COSMIC ID5 (Cordero et al. 2024). These T-deletions also lack any specific enrichment of flanking C:G bases that would be suggestive of the mutations occurring through an alternative error-prone bypass process of OG. Consequently, the simplest explanation for the occurrence of KBrO₃-induced T-deletions is that the oxidant also induces a T-based lesion (potentially thymine glycol) that preferentially results in mutagenic deletion of the damaged base. Oxidation potentially causing small insertion and deletion (ID) signature 5 mutations is also supported by co-occurrence of ID5 and SBS18 in primary cells within normal colon crypts (Lee-Six et al. 2019), as well as correlation of ID5 mutation abundance with the number of SBS18 mutations in human cancers (Cordero et al. 2024). Definitive assignment of ID5 as a second ROS signature however is complicated by the correlation of ID5 with multiple other substitution signatures caused by base damage occurring by mechanisms distinct from oxidation (Cordero et al. 2024). The generally low complexity of sequence contexts available for classifying indel signatures (Koh et al. 2025) suggests that ID5 may represent a composite signature of multiple different types of damage, including oxidative lesions.

SBS30 is highly likely to originate from an oxidation lesion other than OG. This signature is composed of C>T substitutions at CA and CC dinucleotides and associates with cancers containing inactivating human Nth-like 1 (NTHL1) mutations (Alexandrov et al. 2020). Drost et al. identified NTHL1 deficiency as the root of SBS30, using human intestinal organoids (Drost et al. 2017). SBS30 was also reported in breast cancers, and retrospectively the breast tumor in which SBS30 was first identified was determined to be NTHL1 deficient (Nik-Zainal et al. 2016). Grolleman et al. later identified four more breast tumors where SBS30 accounts for 80% of the mutations, suggesting that NTHL1 deficiency has driven the formation of these tumors (Grolleman et al. 2019). Transcriptome sequencing of a pancreatic neuroendocrine tumor revealed SBS30, and NTHL1 loss, implicating NTHL1 loss as a driver of another tumor type (Wong et al. 2018). These experimental and cancer analyses support that NTHL1 enzymatic activity specifically protects against an endogenous DNA lesion that preferentially results in C>T substitutions. While the identity of the lesion that causes SBS30 is currently unknown, 5-hydroxycytosine (5-OHC) or 5-hydroxyuracil (5-OHU) are strong candidates. NTHL1 displays enzymatic activity towards the removal of both of these oxidation lesions in vitro (Dizdaroglu et al. 1999). Formation of 5-OHU through cytosine glycol oxidation intermediates would ultimately template for dATP incorporation across from what was a cytosine base prior to damage and result in C>T substitution. Moreover, RB69 DNA polymerase, which has significant homology to human replicative polymerases, efficiently inserts dATP opposite 5-OHC, only trailing insertion of the non-mutagenic dGTP in order of preference (Zahn et al. 2011). Consequently, 5-OHC also primarily results in C>T substitutions in H₂O₂-induced cell free systems (Sugiyama and Sanyal 2024) or in E. coli vectors (Kreutzer and Essigmann 1998). Similar C>T substitutions are observed upon upregulation of cellular ROS in yeast containing persistent single stranded (ss) DNA (Degtyareva et al. 2019). These substitutions occur primarily at similar sequence contexts at those in NTHL1 mutant tumors suggesting they may originate from the same lesion and providing experimental evidence for ROS-involvement in the formation of the signature. Additionally, hydrogen peroxide induced C>T substitutions in yeast only occurred in genomic regions predicted to be single stranded suggesting that either the causative lesion forms only in ssDNA or it is normally efficiently removed by BER in dsDNA but persists in ssDNA where BER cannot function. Efficient removal limiting ROS-induced C>T substitution is consistent with the lesser prominence of SBS30 compared to SBS18 in human cancers (Alexandrov et al. 2020), the former of which requires loss of NTHL1 to become more prominent. In rare tumors that display SBS30 in the presence of functional NTHL1, polymerase-based incorporation of 5-(hydroxymethyl)-2’-deoxyuridine (HMdU) nucleotides at template G bases could serve as an alternative mechanism for establishment of the signature (Mellac et al. 1993).

While the two most prominent ROS-signatures are emphasized by deficiencies in glycosylases that serve primary roles in global removal of ROS-induced base damage, whether deficiencies in other steps of BER either modifies existing or generates unique signatures is less unclear. Previous works have clearly highlighted associations of DNA Pol β and DNA ligase III-deficiency on carcinogenesis [reviewed in (Wallace et al. 2012)]. The current working hypothesis for this association is that these deficiencies result in genome instability and increase the frequency of cancer driving mutations. However, insufficient occurrence of these deficiencies has been observed within large consortia-based sequencing efforts to determine whether they also produce the expected increase in mutation rate and subsequent mutation signatures. Targeted experimental sequencing of model systems are therefore required to determine if these variants are truly mutagenic and assess risk for individuals who carry these germline variants.

Topography of oxidation-induced mutations: Footprints of BER activity.

Despite historic treatment of mutation as a random process, DNA lesion formation and repair processes are impacted by underlying sequence context and chromatin structure which results in variable mutation density across the genome (Schuster-Bockler and Lehner 2012; Haradhvala et al. 2016; Otlu et al. 2023). The topological features that are enriched for specific mutation types provides additional support for the origins of a mutational signature as well as provides information on the in vivo function of DNA repair enzymes during tumor development. Chromatin structure, transcription factor binding sites, replication timing, and transcriptional and replication strand biases are common features that differentially govern where in the genome specific mutation signatures occur (De and Michor 2011; Schuster-Bockler and Lehner 2012; Haradhvala et al. 2016; Brown et al. 2018; Mao et al. 2018; Pich et al. 2018; Frigola et al. 2021; Liu M et al. 2021; Sivapragasam et al. 2021; Otlu et al. 2023). For SBS18 mutations, the primary feature that influences mutation density is chromatin structure, with higher mutation abundance occurring in more heterochromatic regions of the genome (Figure 6). This effect appears to be due to tightly packed nucleosomes within these regions limiting repair of OG. OG mapping technologies indicate the lesion forms similarly in hetero- and euchromatic regions of the genome and structures of OGG1 interacting with OG within nucleosomal DNA suggest the enzyme would only be capable of excising solvent exposed lesions (Maher et al. 2013; Cordero et al. 2024). OG-induced mutations, unlike UV-light induced mutations display no enrichment for transcription factor binding or strand biases in repair proficient cells, however transcriptional strand bias can be seen upon loss of OGG1 or DNA pol eta, which protects against OG-induced mutations in human cells (Cordero et al. 2024).

Figure 6:

Topography of OG mutagenesis in human cells. Mutations caused by OG are enriched in histone-occluded DNA, likely due to inhibited BER. This impact is observed at multiple levels of chromatin compaction, being highest in heterochromatin, but also being observed in individual nucleosomes. These regional mutation associations can be used to study BER within tumor samples. Images modified from (Cordero et al. 2024).

SBS30 induced mutations appear to follow similar topological associations as SBS18, being enriched in compact chromatin regions, showing nucleosome periodicity, but lacking transcriptional or replication strand biases (Otlu et al. 2023). While the distribution of lesions causing SBS30 is unknown, the similarity of SBS30 and SBS18 topology likely derives from common impacts of chromatin structure on NTHL1 and OGG1 glycosylases. The topological characteristics are likely to be shared among most BER-associated mutation signatures due to commonalities of lesion processing within nucleosomal DNA. Position-dependent inhibition of many BER enzymes have been reported: OGG1, UNG, TDG (Barbosa and Delaney 2025; Ferrara and Delaney 2025) and NTHL1 (Odell et al. 2011) glycosylases, and AP endonuclease (APE1) are less inhibited at solvent-facing lesions (i.e., deoxyuridine or thymine glycol) compared to histone-obscured positions (Hinz et al. 2010). BER-associated gap-filling and ligation reactions are more strongly altered within nucleosome structure, with long-patch synthesis being completely inhibited (Meas and Smerdon 2016). Pol β-mediated filling is limited to 1 base of synthesis (Weaver et al. 2025) and most limited in the central portion of the nucleosome, whereas DNA Ligase III activity is strongly inhibited throughout nucleosomal DNA (Sutton et al. 2024). The latter inhibitory effects on BER likely require nucleosome displacement for repair to be completed, either mediated through DNA Ligase III/XRCC1 (Odell et al. 2011) or chromatin remodeling complexes like SWI/SNF (Sutton et al. 2024). Nucleosome disruption, however, is unlikely to significantly alter the topology of mutational signatures produced by BER-repaired DNA lesions as the efficiency of lesion recognition is the most prominent factor that influences the likelihood a lesion will eventually result in a mutation.

Variable Impacts of BER variants on mutation signatures

Currently, the characterization of BER-deficiency mutation signatures has relied primarily on analysis of cell lines or tumors with full inactivation of either OGG1, MUTYH, or NTHL1. Cell line experiments have exclusively utilized CRISPR/Cas9 deletion of the respective genes (Drost et al. 2017; van den Boogaard et al. 2021; Zou et al. 2021) while biallelic variants in tumors displaying SBS36 or SBS30 often have inactivating MUTYH Y179C (Pilati et al. 2017; Degasperi et al. 2022; Georgeson et al. 2022; Robinson et al. 2022) or NTHL1 G82* nonsense alleles (Belhadj et al. 2019; Grolleman et al. 2019). However, other missense variants exist that could result in partial loss of function and unique mutation signatures or topologies. MUTYH G396D is one such variant that has been suggested to confer an altered SBS36 signature (Robinson et al. 2022). As discussed above, G396D impacts a tight turn region of MUTYH decreasing DNA binding and activity, but not to the extent of Y179C. In most MAP tumors, G396D can contribute to compound heterozygous deficiency with Y179C. Moreover, a recently described tumor with the Y179C/G396D biallelic inactivation displayed a significantly different mutation signature than Y179C homozygous tumors, suggesting the G396D mutation could modify the signature (Robinson et al. 2022). While still a possibility, the same tumor also contains OGG1 R46Q and G308E germline variants, indicating the difference in signature may result from a superposition of the OGG1- and MUTYH-deficiency signatures (SBS108 and SBS36, respectively). Currently, the underlying structural and enzymatic reasons G308E inactivates OGG1 have yet to be reported. G308E is the most common germline polymorphism associating with SBS108 (Degasperi et al. 2022), which itself has subtle differences from experimentally derived signatures caused by OGG1 knockout (Zou et al. 2021), indicating G308E may also be a partial loss of function allele with a unique signature. The mutagenic impacts of other germline missense polymorphisms associated with cancer risk, including NTHL1 D239Y, R33K, and Pol β variants K289M, P242R are largely unknown. Pol β K289M increases mutation frequency of a reporter in mouse cells ~2-fold (Lang et al. 2004), however, no whole genome sequencing compatible with mutation signature analysis has been conducted on any of these variants to determine how truly mutagenic they are as well as their potential towards promoting carcinogenesis through mutation-based mechanisms. Establishment of this connection will be of significant importance in further characterizing the pathogenicity of the large number of variants of uncertain significance that occur in BER genes (Georgeson et al. 2022).

5-. Concluding Remarks

Base excision repair protects cells from oxidative DNA damage. As highlighted in this review, structural and biochemical analyses of MUTYH, NTHL1, and DNA polymerase β revealed that BER fidelity depends on factors such as coordinated domain motions, allosteric mechanisms, and precise control of catalytic chemistry. At the genome scale, BER failure leaves discernible imprints in tumor DNA. Mutational signatures such as SBS18, SBS36, and SBS30 provide strong evidence linking specific oxidative lesions and repair defects to cancer development, while their chromatin-dependent topologies underscore the importance of genome organization in shaping mutational outcomes. These findings establish BER deficiency as a link between DNA damage and cancer mutagenesis.

The integration of biochemical and structural insights with genomics and tumor sequencing will be essential for interpreting variants of uncertain significance and for a better understanding of how BER dysfunction contributes to cancer initiation and progression.

Biographical Note

SH is a postdoctoral fellow in the laboratory of Joann Sweasy at the Eppley Institute in Cancer Research.

CC is a graduate student in the Cellular, Molecular, and Biomedical Sciences program at the University of Vermont. His thesis research focuses on mutational signatures of oxidative and UV damage.

MH is a graduate student in the Chemistry and Chemical Biology graduate program at the University of California, Davis. His thesis research is focused on kinetics and structural analysis of MUTYH cancer associated variants.

MBN is a Chemistry undergraduate at the University of Vermont. His work centers on the NTHL1 DNA glycosylase, combining X-ray crystallography, biophysical methods and molecular dynamics.

SSD is a Professor of Chemistry at the University of California, Davis. Her research has focused on using chemical biology approaches to study base excision repair glycosylases.

SAR is a Professor of Microbiology and Molecular Genetics at the University of Vermont. His research investigates mechanisms of mutation that contribute to tumor evolution, with particular emphasis on elevated DNA damage and dysfunctional DNA repair.

JBS is the Director of the Fred and Pamela Buffett Cancer Center and Eppley Institute for Research in Cancer. She is the Robert F. and Myrna L. Krohn Chair in Cancer Research and holds the Fred and Pamela Buffett Presidential Chair.

SD is a Green and Gold Professor of Microbiology and Molecular Genetics at the University of Vermont. Her research focuses on the structural biology of enzymes involved in base excision and double strand break repair.