Bacterial DNA excision repair pathways

Abstract

Bacteria inhabit virtually all environments on earth where they are exposed to numerous endogenous and exogenous DNA damaging agents. To maintain genome integrity and ensure cell survival, bacteria have evolved several DNA repair pathways to correct and repair the different types of DNA damages and non-canonical bases that occur, including strand breaks, nucleotide modifications, cross-links, mismatches, and ribonucleotide incorporations. Recent advances in genome-wide screens, the availability of tens of thousands of whole genome sequences, and advances in structural biology have enabled the rapid discovery and characterization of novel DNA repair pathways and new enzymatic activities across the domain Bacteria. In this Review, we focus on several bacterial excision repair pathways. We review recent advancements in well-known pathways, including advances in base excision repair, nucleotide excision repair, we discuss several new repair processes including the EndoMS mismatch correction pathway, and the MrfAB excision repair system. This Review highlights the impressive breadth of DNA repair capabilities harnessed within the microbial world.

Introduction

Bacteria thrive in diverse environments where they encounter a myriad of endogenous and exogenous causes of DNA damage¹. All organisms need to repair large and small base damage sites, DNA strand breaks, and correct misincorporated and noncanonical nucleotides surrounded by millions of properly paired bases¹. Maintenance of genetic information through minimizing accumulation of deleterious mutations helps ensure cell survival^1–3. DNA damage interferes with accurate DNA replication while also having the potential to block replication fork progression and RNA polymerase during transcription^4,5. A block to fork progression could be lethal or lead to genomic instability, including formation of DNA double-strand breaks^4,5. Some bacteria use photoreactivation to reverse specific types of UV damage, and direct reversal of alkylation damage, while virtually all organisms use base excision repair (BER) and nucleotide excision repair (NER) to address different types of damaged nucleobases¹. Repair of single and double strand DNA breaks is critical, and organisms do so using homologous recombination, nonhomologous end joining, and single strand annealing^1,6. Thus, bacteria have evolved a series of repair pathways to address the DNA lesions that occur in vivo.

In this Review, we consider the endogenous and exogenous sources that damage DNA as well as the types of lesions bacteria encounter. Next, we discuss conserved DNA excision repair activities. We discuss bacterial BER, including more recent work describing the BER-dependent repair of DNA crosslinks. We discuss recent advances in our understanding of the mechanism behind NER and we cover three pathways involved in the correction of DNA polymerase errors. Finally, we touch on a recently discovered excision process mediated by a helicase and exonuclease. This review provides a contemporary discussion of the excision-based mechanisms bacteria use to repair the diverse set of lesions they encounter.

Overview of DNA damage sources

Cells must repair a few basic types of lesions including base damage sites (small and bulky), apurinic/apyrimidinic (AP) sites, DNA strand breaks (single and double), cross-links (interstrand, intrastrand, and protein-DNA), ribonucleotide misincorporations and mispaired bases (mismatches)^1,7. The relative abundance of each type of lesion depends on each cell’s intrinsic metabolism, the environment occupied, and the frequency with which each type of lesion occurs. In addition, bacteria need to balance the selective pressures required for genome maintenance with the occurrence of mutations in a population that could provide a selective advantage^8,9.

Endogenous damage

Endogenous sources of DNA damage result from byproducts of normal cellular metabolism that chemically react with DNA to form lesions (Figure 1)¹⁰. Reactive oxygen and nitrogen species chemically altering DNA is the most common source of endogenous lesions¹. The primary lesions caused by reactive oxygen species are damaged bases or removal of the base resulting in an AP site¹⁰. More than 20 different lesions from oxidative damage can form on intact bases¹⁰. These lesions tend to be small and include the very common 8-oxoguanine (8-oxoG), along with base deaminations¹⁰. Oxidative DNA damage also results in formamidopyrimidines and AP sites¹⁰. AP sites are frequent and are caused by hydrolysis of the N-glycosidic linkage between the nucleobase and the deoxyribose sugar¹¹; hydrolysis occurs spontaneously, by environmental and therapeutic genotoxins, and during BER by DNA glycosylases¹². If AP sites are not repaired, the lesion can block progression of DNA replication and transcription¹³. Interstrand cross-links are a type of lesion that prevents double helix unwinding during DNA replication and transcription, requiring strand breakage or generation of two AP sites to repair^14,15.

Figure 1. Examples of endogenous and exogenous sources of DNA damage encountered by bacteria.

(A) Endogenous sources of damage or errors including base deamination (1), oxidative DNA damage (2), and DNA replication errors including base pairing errors and ribonucleotide errors (3). (B) Examples of exogenous sources of damage including UV cyclobutane pyrimidine dimers (1) and 6–4 photoproducts (1) as well as the uptake of antimicrobial compounds including mitomycin C which damage DNA forming cross-links, and a monoadduct (3). Abbreviations include base excision repair (BER), nucleotide excision repair (NER), homologous recombination (HR), non-homologous end-joining (NHEJ), single strand annealing (SSA), mismatch repair (MMR) and ribonucleotide excision repair (RER).

Other endogenous processes include base-pairing errors and sugar errors, both of which occur during DNA replication^16,17. Base-pairing errors occur when the incorrect base is paired during DNA synthesis¹⁶. Sugar errors occur when a ribonucleotide is incorporated in place of the cognate deoxyribonucleotide (i.e. AMP in lieu of dAMP)^18,19. Both mismatches and sugar errors require correction; however, a distinguishing feature is that the nucleotides themselves are not damaged, rather they are misincorporated requiring removal and correction^16,17. Base-pairing errors and sugar errors increase mutation rate^16,17. When a base-pairing error occurs, it changes the DNA sequence directly if not corrected before the next round of DNA replication¹⁶. Sugar errors can increase mutagenesis when the resynthesis step or removal mechanism is error-prone^17,19,20.

Exogenous damage

Exogenous sources of DNA damage are chemicals or radiation that arise from outside the cell, causing direct and indirect damage inside the cell¹. Lesions from exogenous sources are specific to the environment inhabited (Figure 1). For bacteria that are plant and animal pathogens or commensals, the exogenous sources of damage also include reactive oxygen and nitrogen species^21–23. Pathogenic bacteria taken up by macrophages or plant pathogens and symbionts invading plant tissues are faced with an oxidative burst as a common host defense mechanism^23,24. Animal gut commensals and pathogens passing through the animal digestive tract encounter nitrosative stress, oxidative stress, and lesions that can result from the acidic environment in the stomach²⁵.

Bacteria present in nature could be exposed to ultraviolet light (UV), gamma rays and many secondary metabolites produced by soil-dwelling bacteria when present in the same environment^26–28. UV exposure causes cyclobutane pyrimidine dimers (CPD) and 6–4 photoproducts, which are considered bulky adducts²⁹. Gamma rays can directly cause single and double strand DNA (ssDNA and dsDNA) breaks while also forming hydroxyl radicals that can produce both types of breaks and base damage sites^30,31.

An important group of antibiotic-producing bacteria are from the genus Streptomyces²⁸. Streptomyces produce a glycopeptide called phleomycin that causes DNA breaks, streptozotocin which is capable of alkylating DNA, and mitomycin C (MMC) and azinomycin B (AZB), which can form DNA cross-links and single bulky lesions^32–35. Below, we discuss the excision repair pathways responsible for addressing DNA lesions that arise from both endogenous and exogenous sources of DNA damage.

Overview of Base Excision Repair (BER)

Base excision repair (BER) is a process that removes and replaces damaged or non-canonical bases in DNA^14,36. The enzymes involved in BER are highly conserved throughout life, underscoring their importance³⁷. First, a DNA glycosylase recognizes a lesion containing a damaged nucleobase. Two types of DNA glycosylases exist, and both release damaged bases but generate different products³⁸. Monofunctional glycosylases cleave the N-glycosidic bond to release the damaged base, causing an AP site (Figure 2A)³⁸. After AP site creation, the DNA backbone is cleaved by an AP endonuclease, leaving a 3’-OH and a 5’ deoxyribose-phosphate (5’-dRP) which is processed by RecJ to create a substrate for DNA resynthesis³⁹. Bifunctional glycosylases also contain lyase activity³⁶. After lesion recognition and N-glycosidic bond cleavage, AP lyase activity of bifunctional glycosylases processes the DNA backbone through β−elimination resulting in a 3’ unsaturated aldehyde and 5’ phosphate³⁶. The 3’-blocking lesion is removed, DNA polymerase fills the gap and DNA ligase seals the nick³⁶.

Figure 2. Advancements in Base Excision Repair.

(A) Formation of an AP site through hydrolysis of the N-glycosidic bond (adapted from¹²⁹). (B) Guanine oxidation into 8-oxo-guanine and mismatched pairing to adenine. (C) AlkD (5kub, shown in pink) protein contacts with depurination intermediate analogue 1aR (1-aza-2’, 4’-dideoxyribose) within DNA (shown in green). Red sphere is a water nucleophile. Dashed black lines are hydrogen bonds. CH/π interactions with W109 and W187 to 1aR (adapted from⁵²). (D) Summary of the EndoV (YwqL)/ExoA repair of deaminated bases (adapted from⁶³). EndoV identifies and incises 3’ of the lesion. If the lesion is xanthine or hypoxanthine, EndoV cuts 5’ of the lesion and pol I fills in the gap. DNA ligase seals the nick to complete repair. If the lesion is uracil or an AP site, the initial endonuclease activity creates a 3’ flap containing the lesion. ExoA uses its 3’ to 5’ exonuclease activity to remove the lesion containing strand, followed by pol I and ligase. (E) Left: Proposed stepwise formation of the thiazolidine linkage between HMCES and DNA based on YedK. 1. AP site within DNA, 2. Nucleophilic attack on AP site, 3. Schiff base intermediate, 4. Thiazolidine link between HMCES and DNA. Right: YedK (6nua, shown in purple) DNA-protein crosslinks with AP site (dark yellow) within DNA (yellow). Critical hydrogen bonds between protein and DNA shown as black dashed lines (adapted from⁵⁶). Note: E105 is shown with two conformers. (F) Left: Mycobacterium tuberculosis UdgX in salmon pink (6ioa) aligned with a family 4 UDG from Thermus thermophilus in light gray (1ui0), uracil shown in orange and Fe-S cluster in black. Arrow pointing at characteristic protruding loop (adapted from⁶¹). Middle: Mycobacterium tuberculosis UdgX (6ioa) interactions with uracil (orange). H178 is the residue responsible for excision of uracil. H178, along with E52, Q53, F65, and N91 hydrogen bond to uracil. Black dashed lines are hydrogen bonds (adapted from⁶⁰). Right: Mycobacterium tuberculosis apo-UdgX (6ail) in magenta aligned with uracil-bound in salmon (6ajo). H109 changes position within the enzyme (shown with dashed black arrow) when uracil is bound and excised from the DNA backbone to form a covalent bond with the AP site (in gray). H109 serves as the nucleophile instead of water.

BER of Oxidative DNA damage

Guanine is highly susceptible to oxidative damage and is the most common oxidatively damaged base in cells⁴⁰. The resulting 8-oxo-guanine (8-oxoG) can pair with dC in the anti-conformation or dA when rotation around the N-glycosidic linkage occurs in the syn-conformation (Figure 2B)⁴¹. When 8-oxo-dGTP is incorporated into DNA it is highly mutagenic causing G:C to T:A transversions⁴². Before 8-oxo-dGTP is incorporated into DNA, it can be sanitized by MutT, an 8-oxo-dGTP diphosphatase⁴³. MutT converts 8-oxo-dGTP and 8-oxo-GTP to the corresponding nucleoside monophosphates preventing their incorporation into DNA and RNA^43,44. If the 8-oxo-dGTP is incorporated into DNA before removal from the dNTP pools, it becomes a substrate for BER⁴⁵. The bifunctional formamidopyrimidine-DNA-glycosylase Fgp/MutM recognizes 8-oxo-dG:C base pairs and cleaves the N-glycosidic bond to create an AP site^46,47. Then using its lyase activity, MutM cleaves the DNA backbone and heals the gap (by δ elimination) creating 3’ and 5’ phosphates⁴¹. Because DNA polymerase requires a 3’ OH, an AP endonuclease such as Exo III or Endo IV is required to remove the remaining 3’ phosphate, allowing for DNA polymerase I to fill the gap⁴¹. If mispairing in an 8-oxo-dG:A base pair occurs, monofunctional adenine DNA glycosylase MutY removes the undamaged, mispaired adenine across from the preexisting 8-oxo-dG⁴⁸. This removal creates an AP site which is then processed to allow for resynthesis generating a substrate for MutM⁴⁷.

Because 8-oxoG occurs in anti- and syn- conformations, which differ only by orientation of the base attached to the ribose moiety, it can mispair with both cytosine and adenine, respectively⁴¹. It was previously unclear whether both anti- or syn-conformations of 8-oxoG could bind to the (MutM/Fpg) required for base excision⁴⁹. Base-specific excision is the canonical mechanism by which syn-bound 8-oxoG is repaired⁴⁷. Recently a DNA base-independent excision mechanism was described representing an alternative mechanism for BER⁴⁹. Importantly, this work suggests that there is no discrimination between DNA lesions and DNA nucleotides within the active site, explaining how 8-oxoG in the anti-conformation could be repaired by MutM/Fgp⁴⁹.

Non-base flipping DNA glycosylases

Most canonical DNA glycosylases use a base-flipping mechanism for recognition and excision of damaged bases⁵⁰. In B. cereus, AlkD was found to be the first DNA glycosylase to catalyze BER without flipping the damaged nucleotide from the DNA⁵⁰. Glycosylases that use a non-base flipping mechanism can provide considerable plasticity in lesion recognition as opposed to canonical base flipping mechanisms where the damaged base is placed in the active site providing a mechanism for lesion specific interactions⁵¹. AlkD is specific for cationic lesions and interacts with the sugar-phosphate backbone using CH/π interactions with Trp109 and Trp187 (Figure 2C)⁵². Asp113 was shown to be critical for orienting the water molecule needed in nucleophilic attack, with mutation dramatically reducing excision (Figure 2C). The interaction between AlkD and a lesion is indirect, and therefore the DNA bends modestly upon lesion binding^50,52. For this reason, AlkD can accommodate different positions and sizes of alkyl substituents, including bulky alkylpurine adducts^50,52.

AlkD recognizes DNA modifications produced by yatakemycin, a highly toxic DNA alkylating agent within the spirocyclopropylcyclohexadienone family of natural products that produces lesions which do not distort or destabilize DNA⁵³. The stabilization of DNA from yatakemycin lesions occurs through CH-π interactions⁵³. Upon yatakemycin treatment, AlkD widens the minor groove around the lesion by separating the lesion and the modified strand⁵³. AlkD cleaves the N-glycosidic bond connecting the sugar and the base, creating an AP site and leaving the base attached to the YTMA adduct⁵³. AlkC is another non base flipping DNA glycosylase that recognizes 3mG and 3mA substrates⁵⁴. An Ig-like domain was described in AlkC, which was shown to cause a sharp bend in the DNA and disrupt the base stacking interactions⁵⁴. As a result, the base is exposed to make direct contact with AlkC, unlike what has been shown in AlkD^52,54. Together, AlkD and AlkC provide examples of non-canonical methods for lesion recognition that are likely to be widespread among enzymes that fall into these groups.

Protection of AP sites

AP sites not only occur as a BER intermediate, but they occur spontaneously and are considered a highly abundant lesion in vivo¹. Human HMCES and YedK from E. coli possess a SOS-response associated peptidase (SRAP) domain for AP site recognition to block the AP site from use as a template by mutagenic translesion DNA polymerases^55,56. These DNA-protein crosslinks are thought to protect DNA from hydrolysis and access by endonucleases and DNA polymerases and are therefore critical for the transient maintenance of genome stability. Additionally, the DNA-protein crosslinks are critical for AP site repair but need to be resolved for replication and transcription to proceed^55,57. It was shown previously that covalent HMCES-DNA crosslinks were resolved via ubiquitin-mediated proteolysis⁵⁵. E. coli YedK has been shown to form a stable thiazolidine link between the ring-opened AP site and the amino-terminal cysteine (Figure 2E, left)⁵⁶. First, the α-NH₂ of Cys2 performs a nucleophilic attack on the AP aldehyde C1’ carbon, resulting in a Schiff base intermediate⁵⁶. Next, the Cys2 sulfhydryl group attacks C1’ to form the favored thiazolidine linkage (Figure 2E, left). This linkage is resistant to strand cleavage and proteolysis, which differs from the typically unstable, transient cross-links in BER⁵⁶. Arg77 and Arg162 are essential for DNA binding while His160 forms a hydrogen bond to stabilize the AP site (Figure 2E, right)⁵⁶. Glu105 and Asn75 are conserved and form hydrogen bonds to the AP site (Figure 2E, right); substitution of these residues results in reduced crosslinking⁵⁶. In E. coli, the most common model for DNA-protein crosslink repair is through the combined action of NER and homologous recombination,⁵⁷ with more recent evidence showing that Lon protease helps remove DNA protein-crosslinks indicating a protease dependent mechanism⁵⁸.

BER of Uracil

Uracil occurs in DNA when cellular dNTP pools are depleted as a result of ribonucleotide reductase inhibition, or from cytosine deamination¹. Uracil incorporated into DNA is repaired using BER and is recognized and excised by uracil DNA glycosylases (UDGs) which exist within six families¹. These families of UDGs are characterized by motif A and B sequences for substrate catalysis and stabilization of the uracil-DNA with enzyme complex¹. Recently, several studies characterized the structures and activity of UdgX from Mycobacterium smegmatis which has 60% sequence identity with family 4 but possesses a unique, positively charged protruding loop (Figure 2F, left)⁵⁹. M. smegmatis UdgX is the first uracil DNA glycosylase to form a tight complex with uracil-DNA and several structures of M. smegmatis UdgX were solved in free, uracil-bound, DNA-bound, and ssDNA-bound forms^60,61. His178 forms a hydrogen bond to uracil to facilitate N-glycosidic cleavage (Figure 2F, middle) and alanine substitution resulted in a loss of the covalent complex^60,61. His109 acts as a nucleophile instead of the typical water molecule creating a covalent linkage to the C1’ of the deoxyribose sugar upon cleavage of the N-glycosidic bond^60,61. This linkage is aided by a long-distance hydrogen bond to Gln53^60,61 (Figure 2F, right).

The overall model for repair is that MsmUdgX-DNA complex stalls the replication fork and presumably forms gaps that require RecA-mediated repair⁵⁹. The in vivo relevance of UdgX is unclear. Requiring HR to repair UdgX-DNA seems like an inefficient method to address uracil. More work will be necessary to determine if other processes in addition to HR are used to repair the UdgX-DNA complex.

Alternative repair of deamination

Adenine, guanine, and cytosine are deaminated by hydrolysis or nitric oxide and hydroxylamine⁶² to produce hypoxanthine, xanthine, and uracil, respectively⁶³. B. subtilis EndoV (YwqL), ExoA and DNA polymerase I constitute an alternative excision repair pathway (Figure 2D)⁶³. EndoV hydrolyzes the second phosphodiester bond 3’ to a lesion with activity on a range of DNA substrates, including hypoxanthine, xanthine, uracil, and AP sites⁶³. EndoV showed preference for hypoxanthine and uracil relative to AP sites, and xanthine, respectively⁶³. DNA polymerase I (polA) and EndoV are epistatic, suggesting that they cooperate although DNA polymerase I is unable to process the product from EndoV. Current work shows that after EndoV cuts at the damaged site, ExoA uses its 3’ to 5’ exonuclease activity to extend the site, followed by DNA polymerase I gap synthesis after lesion excision⁶³. The EndoV, ExoA and Pol I pathway demonstrates the need for bacteria to contain repair pathways that have overlapping functions to address common forming lesions including sites of base deamination.

BER of interstrand cross-links

Interstrand cross-links (ICLs) form a covalent bond between two DNA strands, preventing their separation during replication and transcription⁶⁴. The most common model for ICL repair involves the combined action of nucleotide excision repair (NER), DNA polymerase I, and homologous recombination¹⁵, although other models exist⁶⁵. Therefore, most current models for ICL repair in bacteria require the participation and coordination of multiple processes^15,66.

The availability of genome sequences and functional studies has uncovered a large group of widely conserved DNA glycosylases that repair ICLs¹⁴. Glycosylase-dependent ICL repair was identified in Streptomyces sahachiroi, which produces the alkylating agent azinomycin B (AZB) (Figure 3A)^67–69. These species contain the gene cluster azi that is responsible for the biosynthesis of AZB and within that gene cluster resides a DNA glycosylase alkZ⁷⁰. AlkZ is a member of a large family of glycosylases with orthologs present in several thousand diverse bacteria, which also includes human pathogens and environmental species⁶⁹. The two most well-characterized members from this family are AlkZ from S. sahachiroi and YcaQ from E. coli^68,69,71. Like mitomycin C (MMC), AZB is a bifunctional DNA alkylating agent that forms ICLs (Figure 3B)³³. Mitomycin C (MMC) forms adducts at the N6 position of adenine or the N2 position of guanine in the GC sequence⁶⁴, while AZB forms adducts at the N7 position of appropriately placed purines³³. Both agents can form a monoadduct where one strand is modified or a diadduct where both strands are modified^33,64.

Figure 3. Azinomycin B and nitrogen mustard interstrand crosslink repair.

(A) Structure of azinomycin B (AZB) (adapted from⁷¹). (B) Structure of interstrand crosslink formed by azinomycin B (adapted from⁷¹). (C) Top: AZB asymmetric crosslinked substrate and possible unhooking products. AlkZ was shown to unhook crosslinks from both sides with preference toward the GGC side (product 1). YcaQ was shown to unhook the crosslink from both sides with slight preference toward unhooking the CGG side (product 2). Bottom: Nitrogen mustard symmetric crosslinked substrate and possible unhooking products. YcaQ was able to unhook both sides of the crosslink, while AlkZ showed no activity⁷¹.

AlkZ recognizes the ICL and AZB monoadduct^68,69. During monoadduct repair, AlkZ binds and cleaves the N-glycosidic bond producing an AP site^68,69. During AZB ICL repair, AlkZ unhooks one cross-link generating an AP site followed by unhooking of the second site (Figure 3C, top)⁷¹. As with monoadduct repair, the resulting AP sites are processed to complete repair. E. coli YcaQ also unhooks the AZB-generated monoadduct and ICL yielding an AP site that is sensitive to cleavage by an AP endonuclease (Endo IV) (Figure 3C, top)⁷¹. As discussed above, some BER glycosylases are highly specific while other glycosylases have more plasticity in lesion recognition. Streptomyces AlkZ is very specific for the monoadduct and ICL produced by AZB⁷¹. Perhaps this is not surprising given that the alkZ gene is located within the AZB biosynthesis gene cluster providing a mechanism of survival during AZB production. E. coli YcaQ has more breadth in the type of lesion recognized and excised⁷¹. YcaQ was efficient in unhooking ICLs produced by nitrogen mustards, whereas AlkZ was completely ineffective even though ICLs derived from nitrogen mustards react with guanine at the N7 position (Figure 3C, bottom)⁷¹. Given these observations, it will be important to learn the lesion specificity among the thousands of other AlkZ family members that remain uncharacterized.

Overview of Nucleotide Excision Repair (NER)

Nucleotide excision repair occurs across all domains of life⁷². NER repairs bulky DNA adducts, such as CPDs from UV exposure, and interstrand cross-links¹. In addition, it can remove non-bulky DNA lesions and ribonucleotides^73,74. The breadth of structurally unrelated adducts recognized by NER is an important feature of this pathway and it is still unclear how NER can recognize such a diverse array of nucleobases⁷⁵. The main steps of NER are the same between prokaryotes and eukaryotes, except bacteria use UvrA, UvrB, and UvrC, while eukaryotes use different proteins that are unrelated to their bacterial counterparts⁷⁵. In brief, UvrA binds ATP and searches for lesions. UvrA then recruits UvrB ATPase which forms a tight complex on DNA; the interaction timing of UvrA and UvrB is still unclear^76,77. Next, UvrC binds and makes two incisions, one 4–5 phosphodiester bonds 3’ to the damage and another 8 phosphodiester bonds 5’ to the damage^78,79. Finally, the lesion containing DNA is released by UvrD helicase and DNA polymerase fills the gap, followed by ligation (Figure 4A).

Figure 4. Overview of nucleotide excision repair.

(A) Canonical nucleotide excision repair where UvrA exists as monomers and dimerizes in solution binding ATP in the distal sites. When UvrA₂ binds undamaged DNA, it should form a “closed groove” conformation. When UvrA₂ encounters a lesion in DNA, ATP is hydrolyzed at proximal sites. The “open tray” should accommodate lesion-bearing DNA. UvrB binds each UvrA monomer and ATP is hydrolyzed at the distal sites. UvrA₂ leaves the lesion, and UvrB uses its 5’ to 3’ helicase activity to find the lesion. One UvrB encounters the lesion and clamps down, releasing the second UvrB. UvrC, a dual endonuclease (grey), binds and cleaves the DNA upstream and downstream from the lesion. UvrC and B are displaced and UvrD helicase (or another helicase in organisms lacking UvrD) removes the fragment. Lastly, DNA polymerase and ligase complete repair. See the following review for a detailed mechanistic and structural view of bacterial NER⁷⁵ (adapted from^77,81). (B) Summary of NER-induced mutagenesis (NERiM). When DNA contains opposing lesions, canonical NER performs a 12–13 nucleotide excision and removal of a lesion on one strand. The gap is filled by pol IV or pol II, and subsequent translesion synthesis by pol V occurs. The result is a repaired strand containing a mismatch across from the unrepaired lesion (adapted from⁹⁷).

Preincision complex

Binding of UvrA to ATP has been shown to promote UvrA₂ dimerization and interaction with UvrB^80,81. Recent investigations have shown the importance of the UvrA₂ dimer ATPase activity and how hydrolysis allows UvrA₂ to recognize damaged DNA⁷⁶. Each monomer of UvrA₂ has two ATPase sites, one proximal and one distal to the DNA binding groove^76,82. The distal site has high affinity for ATP and performs hydrolysis more quickly with a slow release of ADP⁷⁶. There appears to be allosteric communication between the distal and proximal sites due to the observation that ATP binding by proximal sites is necessary for ATP hydrolysis, ADP release, and catalytic turnover by distal sites⁷⁶. This distal hydrolysis creates “mixed” UvrA₂ dimers bound to ATP, ADP, or no nucleotide⁷⁶. From the kinetics, affinity, and stoichiometry experiments, it was shown that the mixed ATP/ADP/apo UvrA₂ population is necessary to interact with DNA and initiate NER. Further, ATP hydrolysis by UvrA helps form the UvrB-DNA complex, possibly via the proximal ATPase site⁸². Recent work found that ATP hydrolysis by UvrB is required to form a UvrB-DNA complex but not required for UvrA to dissociate from the UvrAB-DNA complex⁸².

The timing of UvrA and B interaction is still not fully resolved. After UvrA dimerization, it is unclear if UvrA recognizes lesions alone or forms a heterotetramer with UvrB prior to lesion recognition^83,84. Work by Pakotiparapha et al. showed UvrA₂-UvrB₂ stoichiometry during lesion recognition in vitro^77,85. They also found that the ATPase activity of UvrA is required for recruitment of UvrB⁷⁷. More recent in vivo single molecule studies suggest that UvrA₂ searches in the absence of UvrB, followed by recruitment of UvrB⁸¹. Structural characterization of UvrA shows that it forms the “closed groove” or “open tray” configurations (Figure 4A)⁷⁷. The idea is that UvrA can change conformations between the “open tray” and “closed groove” during the search for damaged DNA⁷⁷. The “closed groove” appears to only accommodate undamaged DNA while the “open tray” can accommodate lesion-bearing DNA^75,77. One possibility is that UvrA toggles between the two conformations as it searches for lesions⁷⁵. Upon damage recognition, UvrA converts to an “open tray” perhaps stimulating the DNA damage recognition pathway^75,77. After formation of the “open tray” UvrA hydrolyzes ATP and leaves the lesion⁸². Helicase activity of UvrB occurs in the 5’ to 3’ direction on DNA toward the lesion^86,87. UvrB kinks the DNA after recognizing the damage to form a stable UvrB-DNA complex, called the pre-incision complex^86,88,89. Repair is complete after incision by UvrC and liberation of the lesion-containing segment by UvrD, followed by resynthesis and ligation¹.

Additional NER factors

Other proteins have been shown to participate in NER in addition to UvrABC and UvrD^90,91. When O⁶-alkylguanine adducts mispair with thymine they are mutagenic⁹¹. Alkyltransferases typically repair these adducts back to guanine⁹². Alkyltransferase-like (ATL) proteins have high sequence similarity to alkyltransferases but lack the catalytic cysteine residue⁹³. As a result, ATLs can bind O⁶-alkylguanine adducts but do not transfer alkyl groups, like canonical alkyltransferases⁹⁴. E. coli YbaZ is in the ATL protein family and works with NER to repair large O⁶-alkylguanine adducts and other lesions⁹¹. Direct interaction between YbaZ and UvrA has been measured, suggesting that YbaZ acts as a “molecular flag” to enhance the efficiency of NER on these lesions⁹¹. This provides an important example of a previously uncharacterized gene that enhances the efficiency of the NER pathway for repair of certain lesions. There could be many more protein factors to be discovered that aid in NER repair of specific lesions.

Mutagenesis during NER

An important consideration for NER is that the resynthesis step (gap filling) could be error-prone, leading to lower fidelity repair. In B. subtilis, hexavalent chromium generates reactive oxygen stress, eliciting the SOS response⁹⁵. At the same time, DNA-protein cross-links were abundant and processed by RecA and NER⁹⁵. In addition to being involved in processing of UV-induced photodimers⁹⁶, the Y-family DNA polymerases YqjH and YqjW can act after NER to facilitate low-fidelity DNA synthesis⁹⁵. This work suggests that NER, homologous recombination, and Y-family DNA polymerases can act together to maintain viability while also resulting in mutagenesis during the resynthesis step⁹⁵.

In another example, a subpathway for UV-induced lesion repair was identified in E. coli⁹⁷. This pathway was termed NER-induced mutagenesis (NERiM) for repair of two adjacent lesions on opposing strands (Figure 4B)⁹⁷. Following UV damage, UvrA and UvrC work to complete canonical NER on lesion-bearing DNA. Next, DNA polymerases Pol II and Pol IV perform gap-filling⁹⁷. Finally, Pol V is involved in translesion synthesis of the template strand across the opposing lesion resulting in mutagenesis⁹⁷. Lower fidelity resynthesis allows for DNA replication to bypass a lesion on the template strand. However, overuse of the error-prone pathway has the potential to increase detrimental mutations and therefore decrease cell fitness.

Overview of Mismatch Repair (MMR)

DNA replication errors escape the 3’ to 5’ proofreading activity of replicative polymerases in all organisms. For bacteria, replication errors (mismatches) occur once every 15.5 rounds of replication or approximately one error every 59 million bases replicated¹⁶. Although this error rate may seem infrequent, mismatch repair increases the overall fidelity of genome replication to one error per ~909 rounds of replication¹⁶. MMR recognizes two major classes of errors, single base mismatches, and insertion and deletion loops (INDELs)¹. Replicative DNA polymerases have intrinsic error bias and as a result are more prone to form G-T mismatches instead of unstable C-C mismatches in vivo⁹⁸. Consequently, the affinity of the mismatch sensing protein MutS for G-T errors is much higher than the affinity of MutS for mismatches that occur less frequently to the point where MutS fails to even bind a C-C mispair⁹⁹.This suggests that MutS has adapted to recognize the most common errors found in vivo.

Methyl-independent MMR

In bacteria, MMR is either methyl-directed or methylation-independent^100,101. The methyl-directed pathway was identified in E. coli and has been well characterized¹⁰¹. DNA adenine methylase (Dam) and hemi-methylation specific endonuclease, MutH, are components of the methylation-dependent mismatch repair pathway present in E. coli and a few closely related bacteria¹⁰². In this system, MutH nicks the unmethylated DNA strand representing the newly synthesized strand¹⁰¹.

The vast majority of bacteria lack MutH and Dam methyltransferase¹⁰². Considerable evidence in bacteria lacking MutH/Dam has shown that mismatch repair is coupled to the site of DNA replication^{100,103–106}. In bacteria, the replication sliding clamp (β-clamp or DnaN) aids in the recruitment of DNA repair proteins to sites of replication^107,108. MutS contains a replication sliding clamp (DnaN) binding motif (QL[SD]LF) that is important for mismatch repair and recruitment of MutS to the site of DNA synthesis serving to confine the search for mismatches to nascent DNA increasing the overall efficiency of mismatch recognition¹⁰⁹. Single molecule tracking of MutS in B. subtilis has shown that MutS moves rapidly around the cell but MutS molecules slow and dwell to search the site of DNA synthesis after encountering the replisome¹⁰⁶. MutS variants that prevent mismatch recognition have no effect on recruitment to the replisome^106,109. However, MutS variants that prevent interaction with the replication sliding clamp reduce or abolish the ability of MutS to dwell at the site of replication impairing the mismatch repair pathway and increasing mutagenesis^106,109. Recently, it was shown in the distantly related alpha-proteobacterium Caulobacter crescentus that MutS and MutL localize to the replisome when ectopically expressed in live cells¹¹⁰. Therefore, targeting MMR proteins to the site of DNA replication through interaction with the replication sliding clamp has emerged as a conserved feature in bacteria lacking a methyl-directed signal. One possibility is that targeting of MutS to the replisome is critical to aid in distinguishing the nascent strand from the template in the absence of a methylation signal.

Another hallmark of the methylation-independent mismatch repair pathway is found in MutL homologs that contain a metal-dependent endonuclease active site with the conserved sequence DQHA(X)₂E(X)₄E^111,112. The endonuclease active site was originally identified in the human MutL homolog PMS2 and was shown to cleave DNA using manganese¹¹¹. The C-terminal structure for B. subtilis MutL with the endonuclease active site revealed a replication sliding clamp (DnaN) binding site located in an extended region of the C-terminal domain¹¹². Mutation of the endonuclease active site or the DnaN binding motif in MutL abolished mismatch repair in vivo demonstrating the importance of MutL-dependent DNA incision and interaction with the sliding clamp¹¹². CryoEM structures of the B. subtilis MutL C-terminal domain and DnaN along with enzymatic assays demonstrated that interaction between MutL and the replication sliding clamp stimulated MutL incision of DNA¹¹³. Therefore, in organisms with a methylation-independent pathway, coupling of MutS and MutL to replication through DnaN is critical for both mismatch recognition and stimulating DNA incision.

Endonuclease Mismatch-Specific Repair

The MutS/MutL MMR pathway is well conserved across all three domains of life,^114,115 yet some organisms including Actinobacteria and some archaea lack mutS and mutL¹¹⁶. Instead, these organisms contain a recently described endonuclease mismatch-specific (EndoMS) protein also known as NucS^117,118. EndoMS/NucS has been shown to combine with DnaN to form a noncanonical mismatch correction system (Figure 5A)^{118,117,119,120}. EndoMS contains a mismatch-specific RecB-like nuclease domain (Figure 5B) and conserved catalytic domain (Figure 5C)¹²¹. Tyr41 and Trp77 exhibit planar stacking with the mismatched base to create the recognition site (Figure 5B). Asn76 forms hydrogen bonds with the base and aids in base discrimination at this site (Figure 5B)¹²¹. It was shown that EndoMS prefers G or T mismatches over A or C mismatches. Arg44, Gln78, and Glu103 form critical hydrogen bonds for dsDNA interaction (Figure 5B)¹²¹. Within the conserved catalytic domain that resembles type II restriction systems, a critical residue for catalysis Asp165 was mutated to alanine to prevent DNA cleavage (Figure 5C)¹²¹. C. glutamicum EndoMS binds and cleaves G/T, G/G and T/T mismatches in vitro¹²⁰. The absence of EndoMS function in vivo correlates with transition mutations, suggesting a failure to correct DNA replication errors^118,120. Further, deletion of the endoMS gene leads to an increase in mutation rate and the mutation rate is synergistic when combined with a replicative DNA polymerase mutant¹²⁰. EndoMS from several organisms contains the QL[SD]LF DnaN binding motif^119,120. DnaN interaction stimulates EndoMS activity on mismatched substrates in vitro, demonstrating a functional interaction. Further, in cells the EndoMS-DnaN interaction is required for mismatch correction along with EndoMS nuclease activity^119,120. while over expression of an EndoMS variant that is unable to bind DnaN is still defective in mismatch correction. Together, these results indicate that DnaN targets EndoMS to mismatches while also regulating mismatch-dependent endonuclease activation^119,120. Although MutS and EndoMS evolved independently, both require interaction with the DnaN to target their search for mismatches to nascent DNA further underscoring the importance of mismatch recognition near the replisome^109,120. One aspect that is unclear is how EndoMS cleavage would be directed to the nascent strand. In vitro reactions show cleavage of both strands 5’ to the mismatch¹²⁰, which would not occur in vivo. EndoMS/NucS nucleases have also been shown to cleave several other substrates in addition to mismatched bases, including flapped DNA structures and deaminated bases¹²². Continued investigation of EndoMS/NucS in DNA repair and mismatch correction will be important to fully understand the different ways this family of proteins contributes to genome maintenance.

Figure 5. Structural insights of EndoMS restriction endonuclease activity.

(A) G-T mismatch within duplex DNA (top). EndoMS (green) binds β-clamp (grey) and recognizes the mismatch (middle). In vitro, EndoMS cleaves the third phosphodiester bond 5’ of the mismatch on both sides (bottom)¹²⁰. In vivo only one 5’ nick should be present on the newly replicated strand. The mechanism of strand discrimination is unknown. (B) Base discrimination site within EndoMS (5gke). dT shown in green and dG shown in dark pink. Critical hydrogen bonds to atoms within dT and dG for base discrimination are shown as black and gray dashed lines, respectively. Light gray dashed line represents a critical hydrogen bond between S47 and N76 (adapted from¹²¹). (C) Catalytic site of EndoMS (5gke). Green sphere represents Mg²⁺. D165 in the structure was substituted to alanine to prevent reaction catalysis. Black dashed lines represent hydrogen bonds (adapted from¹²¹).

Genome-wide screens for novel DNA repair activities

Advances in genome-wide screens and homology searches have rapidly enhanced the discovery and characterization of new DNA repair pathways and novel enzymatic activities¹²³. The identification of the large family of ICL repair glycosylases described above highlights the discovery potential of homology searches across sequenced genomes^68,69. Recently, the use of random transposon insertions in B. subtilis followed by deep-sequencing (Tn-seq) identified several previously characterized genes including MrfA and MrfB^123–125.

MrfA is a member of the highly conserved Superfamily 2 helicases characterized by a C-terminal domain of unknown function (DUF1998)¹²⁴. This group of helicases are present in a wide range of organisms including Mycobacterium SftH and S. cerevisiae Hrq1^126,127. MrfB is a DnaQ-like metal-dependent nuclease active on linear and nicked plasmid DNA¹²⁴. B. subtilis strains with a mrfA and/or mrfB transposon insertions resulted in a growth sensitivity to MMC, but not phleomycin or MMS^123,124. Further, the action of MrfAB was shown to be specific for MMC since mrfA and mrfB deletion mutants failed to sensitize B. subtilis to growth in the presence of another cross-linking agent including psoralen or the bulky guanine monoadduct producing agent 4- nitroquinoline 1 oxide (4-NQO)¹²⁴. Bacterial two-hybrid interaction data showed that MrfA and MrfB interact and isolation of genomic DNA from MMC treated cells showed that MrfA and MrfB did not repair the ICL, suggesting that these proteins are specific to the MMC monoadduct¹²⁴. MrfA and B exist together in select species of Proteobacteria and Firmicutes, including B. subtilis and B. cereus¹²⁴. Overexpression of mrfB from B. cereus and S. pneumoniae in the B. subtilis mrfB deletion background restores growth on MMC to varying degrees¹²⁴. mrfB from P. aeruginosa, a member of the γ-Proteobacteria phylum, does not restore growth on MMC in the B. subtilis ΔmrfB background, indicating its enzymatic activity differs from that of the B. subtilis protein. It is possible that even in species with MrfA and B functioning together, the MrfB activity has evolved to function differently across phyla¹²⁴.

The crystal structure for B. subtilis MrfA was solved including the domain of unknown function¹²⁸ providing novel insight into this group of helicases (Figure 6A). Within the basic groove, kinked DNA is stabilized through hydrogen bonding to residues Asn157-Ser163 and at the 3’ end of DNA with His168 and Lys171¹²⁸. Residues in the RecA1 and RecA2 domains (Arg148, Arg144, Gly137, Lys108, Arg333, Gly326, and Arg298) bind DNA via the sugar-phosphate backbone, and these interactions are well conserved in MrfA homologs¹²⁸. F483 forms π-π stacking with the DNA backbone, while Ser486 and Ser489 stabilize the backbone with hydrogen bonds (Figure 6A)¹²⁸. These interactions are not highly conserved in other homologs, leading authors to suggest that they could be important for helicase activity but are likely not required¹²⁸. MrfA contains a dual RecA domain with Zn²⁺ binding in the conserved DUF¹²⁸. Structural and biochemical characterization of MrfA revealed a novel translocation mechanism (Figure 6B). MrfA is proposed to use a “skipping rope” like mechanism where a molecular “rope” is attached between the winged helix (WH) and oligonucleotide/oligosaccharide binding (OB) domains. In this model the molecular “rope” skips ahead moving along ssDNA, while also preventing backsliding of the helicase¹²⁸. The Zn²⁺ binding domain in the DUF contributes to MrfA action by causing a large conformational change induced upon DNA binding¹²⁸. Given the conservation of MrfA-type helicases involved in repair of MMC-induced lesions, it will be interesting to learn how these helicases load onto DNA and if the translocation and conformational changes observed in MrfA are also conserved in other related helicases.

Figure 6. MrfA involvement in DNA repair.

(A) MrfA (6znq) contacts with DNA in the loading strand adapted from¹²⁸. Note: K108¹²⁸ is K109. Black dashed lines represent hydrogen bonds. Yellow dashed line represents π-π stacking. A and B are adapted from the following work¹²⁸. (B) It is unknown how MrfA loads onto DNA or how it identifies a mitomycin C-induced lesion. MrfA contains seven domains: a N-terminal region (NTR, not shown), RecA1 and RecA2 (core domains, grey), a winged helix domain (WHD, green), an oligonucleotide/oligosaccharide binding domain (OB, blue), a connector element (CON, maroon), and a C-terminal DUF1998 that contains a unique MrfA zinc binding domain (MZB, light green)¹²⁸. DNA shown in bright green with a DNA lesion in red. Upon loading DNA, the OB domain shifts toward RecA2 domain¹²⁸. The OB domain binds RecA2. As a result, the rope (purple) connecting the OB domain and winged-helix domain undergoes a conformational change. The tightening of this rope across the ssDNA prevents the loading strand from slipping backward¹²⁸. ATP is bound, causing DNA translocation through an “inchworm”-like mechanism. The mechanism for coupling DNA translocation and unwinding is unknown¹²⁸. As new ATP is hydrolyzed, DNA is pumped through the protein and products are released.

Conclusions

Because bacteria thrive in such a wide range of habitats, their DNA repair pathways display a broad and unique collection of mechanisms. In this Review we have discussed non-base flipping DNA glycosylases and glycosylases that repair specific types of naturally occurring cross-linking damage. An important future direction will be to study the numerous uncharacterized AlkZ family members to determine their substrate specificity. We have also discussed the MrfAB excision system. Another important future direction will be to determine the mechanism of lesion recognition and to determine if the unstudied enzymes that are part of this family contribute to repair of other types of lesions in the diverse bacteria that use this system. Bacteria have served to drive discovery of novel DNA repair processes and provided new mechanistic insight into established ones; we expect that many new DNA repair strategies will be uncovered in the millions of largely unstudied bacterial species. We look forward to the insights that will be gained in the coming years.