The NHEJ (non-homologous end-joining) pathway is one of the major mechanisms for repairing DSBs (double-strand breaks) that occur in genomic DNA. In common with eukaryotic organisms, many prokaryotes possess a conserved NHEJ apparatus that is essential for the repair of DSBs arising in the stationary phase of the cell cycle. Although the bacterial NHEJ complex is much more minimal than its eukaryotic counterpart, both pathways share a number of common mechanistic features. The relative simplicity of the prokaryotic NHEJ complex makes it a tractable model system for investigating the cellular and molecular mechanisms of DSB repair. The present review describes recent advances in our understanding of prokaryotic end-joining, focusing primarily on biochemical, structural and cellular aspects of the mycobacterial NHEJ repair pathway.
DSBs (double-strand breaks) in DNA represent a lethal form of cellular damage, but evolution has arrived at various solutions to repair such deleterious lesions, thus maintaining the integrity of genomic DNA. Breaks can arise from many sources, including endogenous metabolic processes and exogenous environmental stresses. Two major DSB-repair processes are deployed by cells: HR (homologous recombination) and NHEJ (non-homologous end-joining) [1,2]. HR relies on the pairing of one of the broken strands with a complementary region on the sister chromatid by the HR repair machinery and this ‘templating’ process faithfully guides subsequent HR-mediated break repair . In contrast, NHEJ break repair is a non-template-directed process that occurs between the non-homologous termini of a DSB and is therefore considered to be more error-prone [1,2]. HR is favoured as the major DSB repair pathway in dividing cells, where the sister chromatid is available to direct accurate repair and prevent the propagation of genetic errors. In contrast, NHEJ is the dominant break repair pathway in G1/quiescent cells, where a homologous DNA template is not available and the requirement for end-joining to prevent chromosomal instability overrides the need for accurate break repair.
Conceptually, the basic mechanism of NHEJ is relatively uncomplicated. First, the broken ends are identified, the termini are then brought together (synapsis), processed by polymerase and nucleolytic activities (if required to restore complementarity) and, finally, the remaining nicks are ligated together to restore the integrity of the DNA duplex [1,2] (Figure 1). Eukaryotic NHEJ is mediated by numerous factors with more than eight core factors involved in this DSB repair process . The heterodimeric Ku70–Ku80 complex, which is responsible for recognizing and stabilizing the DNA ends, forms a toroidal-shaped ring structure with a large central hole through which the DNA passes and binds to the break termini in a sequence-independent manner . Ku recruits a large DNA-PK (DNA-dependent protein kinase), which in turn promotes end-bridging and intermolecular ligation by the DNA ligase IV–XRCC4 (X-ray repair complementing defective repair in Chinese-hamster cells 4)–XLF (XRCC4-like factor) complex  (Figure 1A). Other accessory factors are also recruited to the break to aid in end-processing, including PNK (polynucleotide kinase; a 5′ DNA kinase and 3′ phosphatase) and Artemis (a 5′→3′ exonuclease) (Figure 1A). After alignment and processing of the break has occurred, short gaps may need to be remodelled by DNA synthesis to produce ligatable ends. This polymerase-based processing is performed by members of the Pol (polymerase) X family [Pol μ, Pol λ and TdT (terminal deoxynucleotidyltransferase)], which are recruited to the break by DNA-bound Ku [1,2].
DSB repair by the NHEJ pathway
Discovery of prokaryotic NHEJ
For many years, the process of NHEJ was considered not to exist in prokaryotes. However, recent bioinformatic studies have identified the existence of bacterial orthologues of Ku [4,5], now considered to be encoded by the majority of prokaryotic genomes. Unlike the multifactor complex deployed by eukaryotes, bacteria appear to possess only a few NHEJ-specific genes, including a gene encoding a homodimeric Ku (Figure 1B). Notably, the prokaryotic Ku genes typically reside in operons containing a conserved ATP-dependent DNA ligase, LigD (ligase D) [5,6]. This operonic association suggested that prokaryotic NHEJ involves Ku homologues recruiting a repair ligase to DSBs, which has now been proven experimentally [7,8] (Figure 1B). In addition to a ligase domain, most LigDs also contain nuclease and polymerase domains [5,6]. It has been demonstrated that Mt (Mycobacterium tuberculosis) LigD possesses all of the end-processing, gap-filling and ligation activities required to repair DSBs [7–10]. In essence, bacterial NHEJ is mediated by a two-component Ku–ligase break repair complex [7,8], although it is likely that other factors participate in this DSB repair pathway.
Significantly, not all prokaryotes possess a NHEJ pathway, as the Ku and LigD genes are absent from many bacterial species, including Escherichia coli [5,11]. Currently, no clear link explaining the distribution pattern of NHEJ in prokaryotes has been discovered. This raises the question as to how bacteria acquired the NHEJ machinery and in what form did they inherit it? One possibility is that NHEJ genes were acquired by bacteria via horizontal gene transfer. The evidence for such gene transfer is exemplified by the distribution of NHEJ genes in phylogenetically distant organisms, such as proteobacteria, actinomycetes, Parachlamydia and low-GC Gram-positive bacteria [5,11]. It is notable that, with one exception, NHEJ genes are not present in the third kingdom of life, archaea. Another possible explanation is that NHEJ genes arose first in prokaryotes and bacteriophages, and were subsequently lost by some bacteria over evolutionary time.
Domain structure of bacterial NHEJ DNA-repair ligases
As discussed, the prokaryotic NHEJ repair apparatus is a two-component system, comprising Ku and LigD, which together possess the majority of the activities required for break recognition, end-processing and ligation [7,8]. In many bacterial species, including Mycobacterium, LigD is a multidomain protein encoded on a single polypeptide consisting of an N-terminal polymerase domain (PolDom), a central nuclease domain (NucDom) and a C-terminal ATP-dependent ligase domain (LigDom) [5,6,12], although the order of these domains varies from species to species [5,6]. Indeed, in many bacteria (especially in thermophilic species), these domains are encoded as individual polypeptides rather than as multidomain proteins. The most studied and versatile activity on LigD resides in PolDom (discussed below), the DNA primase/polymerase domain. Significantly, PolDom exhibits the combined polymerase activities of eukaryotic Pol X family members, which have been implicated in NHEJ end-processing in eukaryotes [8,13,14]. NucDom possesses a novel 3′–5′ nucleolytic activity capable of resecting 3′-ends and appears to be required for end-processing before ligation [8,15,16]. Finally, LigD contains a ‘minimal’-sized ligase (LigDom) that is capable of sealing nicked breaks that are formed following synapsis and processing of the DSB [7,8].
Bacterial NHEJ polymerases belong to the eukaryotic DNA primase superfamily
In both sequence and structural terms, the bacterial NHEJ polymerases are members of the AEPs (archaeo-eukaryotic primases) [5,6,17], a large conserved family of polymerases that are essential components of DNA replication in all eukaryotic organisms. Primases initiate replication of new DNA strands by first synthesizing a dinucleotide product, which is subsequently elongated by up to 14 ribonucleotides . The resultant 3′-hydroxy group of this RNA primes extension by the replicative DNA polymerases, as these enzymes are unable to initiate DNA strand synthesis de novo [17,18]. Although classified as a DNA primase, PolDom possesses a remarkable and unique assortment of nucleotidyltransferase activities, including DNA-dependent RNA primase, terminal transferase and DNA-dependent RNA/DNA gap-filling polymerase activities (Figure 2) [8,13,14]. All of these polymerase activities have a requirement for bivalent metal ions and, with the exception of the priming activity, are directly relevant to NHEJ-mediated DSB-repair processing [9–11].
acterial NHEJ polymerase possesses a variety of nucleotidyltransferase activities
DNA primases are generally split into two major superfamilies that have no evolutionary relationship, despite being related functionally [17,18]. The DnaG family comprises primases from bacteria and bacteriophages, whereas the other family is made up of AEPs. The AEP family is typically characterized by the small subunit of the heterodimeric eukaryotic primases, which associates with the DNA Pol α and B subunits to form the Pol α primase complex. More robust sequence alignment methods have expanded the superfamily to include homologous primases from archaea, viruses and bacteria [5–7,17]. PolDom is a member of a subfamily of AEPs implicated in NHEJ. Functional studies have established that this divergent clade of NHEJ ‘primases’ have evolved specific structural and biochemical features required for their specific role in DSB break repair that are distinct from the basic activities required for priming DNA synthesis [8,13,19].
The crystal structures of a number of PolDoms have been elucidated, both as apo and nucleotide-bound forms (Figure 3A), which confirmed that these primases share significant structural homology with the replicative AEPs [13,19]. Notably, not only are the catalytic cores of the NHEJ AEPs conserved, but also so are many of the peripheral helices, which strongly suggests a common evolutionary ancestry with the more divergent members of the AEP family [13,17,19]. The catalytic triad of metal-binding aspartate residues, conserved in the NHEJ AEPs, maintain the same geometry as the equivalent residues in the active sites of other AEPs [3,19]. The co-crystal structures of PolDoms in complex with various nucleotides has identified the mode of binding of NTPs (Figure 3A) and dNTPs in the active sites of the prokaryotic NHEJ polymerase/primases and identified many of the conserved residues that are important for both nucleotide recognition and subsequent nucleotidyltransferase activities [13,19]. In addition, these structures established the molecular basis for the preferential binding of NTPs over dNTPs. In the Mt-PolDom–GTP structure, a specific interaction of the 2′-hydroxy group with conserved threonine and histidine residues in the active site stabilizes NTP binding and also orientates the sugar moiety to enhance the stacking interaction of the base with a conserved phenylalanine residue . Lack of a 2′-hydroxy group results in a less well occupied nucleotide-binding site, as observed in the co-crystal structure of Mt-PolDom bound to dGTP where the binding of dNTP is only stabilized by the triphosphate tail and base-stacking interactions. The biological significance of this preferential binding of NTPs by these polymerases is discussed below.
Co-crystal structures of the mycobacterial NHEJ polymerase
On the basis of various primary sequence and structural alignments, AEPs have been shown to have a catalytic core convergent with the Pol X family of DNA polymerases [17,20,21]. There appears to be no topological convergence between these two families, with the catalytic residues sitting on a parallel β-sheet in the Pol Xs, whereas the equivalent residues are arranged on antiparallel structural elements in the AEPs. Despite these differences, the spatial relationship of the three catalytic carboxylates is highly convergent between members of the two polymerase families, indicating that, not only do the AEPs bind the NTPs in a similar way as Pol Xs, but also they probably share a similar catalytic pathway for nucleotidyl transfer .
Biochemical characteristics of the bacterial NHEJ polymerases
As discussed, NHEJ polymerases possess a unique variety of DNA-extension activities, presumably reflecting the requirements for an assortment of extension activities during NHEJ-mediated end processing (Figure 2). PolDom can extend dsDNA (double-stranded DNA) (with 5′-overhangs) and fill in gapped dsDNA substrates in a template-dependent manner [8,12–14]. PolDom also displays template-independent terminal transferase activity on ssDNA (single-stranded DNA) and blunt-ended dsDNA substrates, although this activity is restricted to the addition of a few nucleotides [8,13]. DNA-dependent RNA primase activity has also been demonstrated for PolDom [8,13,14], confirming that these atypical AEP family members retain some primordial primase activity. Mutation of the conserved aspartate residues, involved in metal co-ordination, abolishes all of these polymerase activities, suggesting that one active site catalyses all of the distinct extension activities [8,12,14]. PolDom also possesses additional activities that are relevant for processing of DSBs, including nucleotide insertion opposite 8-oxoguanine and lesion bypass via template dislocation and realignment (Figure 2) .
PolDoms display a marked preference for the insertion of NTPs over dNTPs in vitro, with a ∼20–70-fold difference in its preference for incorporation of NTPs over dNTPs into DNA, depending on the templating base [8,13]. In stationary cells, intracellular pools of NTPs are much higher than dNTPs (>10:1), suggesting that preferential incorporation of ribonucleotides is likely to occur in vivo, although this has not been demonstrated. This preference for incorporating NTPs into DNA may be one reason a primase fold has been adopted as the NHEJ polymerase by the prokarya. If this process occurs in vivo, what is the cellular fate of the short stretches of RNA introduced into the sites of repaired DSBs? The presence of patches of RNA in DNA would severely compromise genome stability and therefore a pathway that ‘repairs’ DNA–RNA hybrids would be essential. A candidate enzyme capable of recognizing and excising short RNA patches from DNA is type II RNase H, which can process short tracts of RNA introduced into newly synthesized DNA .
DNA-break recognition by the prokaryotic NHEJ polymerases
In addition to a wide diversity of extension activities, PolDoms possess other attributes that enable it to participate in atypical polymerase activities associated with DNA end-joining. The major determinant for specific binding of PolDom to DNA is the presence of a 5′-phosphate , located on the recessed end of the break site, and NHEJ AEPs have little affinity for DNA ends in the absence of a 5′-phosphate moiety. Phosphate interaction significantly enhances nucleotide selection and polymerization, as well as promoting end remodelling . Although structurally unrelated, the eukaryotic NHEJ polymerases (Pol μ and Pol λ) possess a related ability to bind 5′-phosphated ends via a helix–hairpin–helix domain [23,24]. PolDom lacks this phosphate-binding domain, but utilizes a novel structural feature to interact with a 5′-phosphate (Figure 3B) .
The elucidation of the crystal structure of the Mt-PolDom in complex with dsDNA (containing a recessed 5′-phosphate and a 3′-overhang) has provided significant insights into DNA recognition by this family of NHEJ DNA-repair polymerases (Figures 3B and 3C) . The structure revealed that the 5′-phosphate of the duplex strand is bound by a conserved positively charged pocket on the surface of the protein (Figure 3B). PolDom induces a bending of the template strand (by over ∼90°) via specific contacts with the surface of the enzyme. As discussed above, PolDom possesses many of the combined NHEJ activities present in the eukaryotic NHEJ polymerases (Pol λ, Pol μ and TdT), some of which contain a surface-exposed structural loop (loop 1) that promotes connectivity between the two discontinuous ends [24,26,27]. The PolDom–DNA co-crystal structure revealed that the bacterial NHEJ polymerases also possess a functionally analogous structural element, and this surface loop plays an important role in facilitating end-synapsis  (Figures 3B and 3C). In this structure, a homodimeric arrangement of the polymerases promotes the direct interaction of the 3′-overhanging termini of two DNA ‘ends’ (Figure 3C). Loop 1 acts as a molecular ‘match-maker’ promoting base-pairing, base-flipping, base-stacking and other interactions to occur between the 3′-termini, which results in stable synaptic complex formation between the opposing ends of a DNA break. This complex has also been shown to exist in solution and to be functionally relevant and important for templated extension of the termini of DSBs. It has been proposed that this arrangement may be required to mediate the processing of ‘difficult’ breaks, such as 3′-overhangs . Mutation of loop 1 ablated the formation of the synaptic intermediates on termini with 3′-overhanging ends , implicating PolDom directly in the end-synapsis process. Further structural and functional studies are required to understand how LigD and Ku co-operate to promote break-synapsis and choreograph end-processing and ligation.
Processing and repair of non-homologous DNA breaks
Higher eukaryotes require a large number of factors to repair DSBs by NHEJ (Figure 1A) . In contrast, the prokaryotic two-component NHEJ apparatus appears to possess most of the processing activities required for remodelling and repair of DSBs , and therefore the simplicity of this break repair apparatus makes it an ideal model system for studying the molecular mechanisms of the end-joining process. As discussed, the bacterial NHEJ complex is capable of bringing two discontinuous DNA ends together, processing the termini and ligating the resulting complementary ends to restore genome integrity (Figure 1B) [7–10]. Examination of the sequence of repaired break junctions provides insights into the molecular processes that drive NHEJ. During eukaryotic NHEJ, short stretches of DNA sequence homology (microhomologies) are often exploited to align the termini of non-homologous breaks [1,2]. Similarly, bacterial NHEJ preferentially exploits DNA sequence microhomologies, internal to the overhangs of incompatible ends, to promote end-synapsis, resection, resynthesis and ligation of DSBs [8,25,28]. Co-ordination of the nucleotidyltransferase and ligase activities of LigD is highlighted by its preference for filling-in of DNA duplexes containing gaps adjacent to regions of microhomology, before ligation of these intermediates [8,28]. As discussed, the polymerase domain of LigD plays a direct role in promoting the connection between break termini , especially 3′-ends, and it is likely that it plays a general role in promoting microhomology-mediated end-joining.
Terminal transferase activity has also been implicated in NHEJ repair, especially when at least one of the DSB termini is blunt-ended. When blunt-ended DNA is joined to a termini containing a 3′-overhang [8,25,28], the polymerase activity of LigD often performs a non-templated base addition to the 3′-terminus of the blunt-end as it cannot extend 5′-recessed ends, as DNA polymerases operate in a 5′–3′ direction. This de novo synthesis activity is even more crucial when both ends are blunt-ended. During in vivo repair assays, where both ends of a DSB are blunt-ended, ∼50% of the DSB repair events resulted in a mutagenic outcome where a single frameshift mutation was observed . This frameshift is the result of one blunt end undergoing non-templated base addition at the 3′-terminus and templated addition occurring at the other blunt end before ligation.
LigD also possesses 3′-ribonuclease and 3′-phosphatase activities, exemplified by the removal of 3′-phosphates and nucleotide flaps, the latter intermediates are often a product of microhomology-mediated synapsis events [8,15,16]. The nucleolytic activities of LigD reside in the NucDom, which has a preference for cleaving recessed 3′-ends and is relatively poor at cleaving 3′-overhangs . The probable role for the 3′-phosphatase activity in DSB repair is to process 3′-phosphated termini that have been generated during DNA breakage. Removal of the 3′-phosphate is essential to restore the 3′-hydroxy group that acts as the essential nucleophile for nucleotide addition and ligation reactions. In contrast, 3′-resection activity in DSB repair appears to be less obvious, although it has been observed in various in vitro and in vivo LigD-dependent NHEJ repair assays [8,28]. As discussed, PolDom can promote the bridging of 3′-protruding ends leading to the formation of DNA synaptic intermediates , which contain unpaired 3′-ends. These termini would require 3′-resectioning to enable gap-filling and ligation to complete the break-repair process. Stringent regulation of this nuclease activity would be required to prevent excessive resectioning that could lead to the loss of genetic information, therefore it is likely that this activity is not deployed unless absolutely required.
In summary, if the ends of DNA breaks are precise then only Ku and the ligase activity are required to repair them. However, if the DSBs are imprecise breaks, Ku and the polymerase activity are needed to promote the synapsis of the non-homologous ends, before end-processing and ligation. When available, bacterial NHEJ utilizes microhomologies on the opposing termini to direct this synapsis process. Realignment of the ends via microhomology-mediated end-bridging is certainly the preferred option, thus reducing the loss of excessive genetic information. In the absence of microhomology (e.g. blunt-ended DSBs) LigD employs other back-up activities, such as terminal transferase activity, to create de novo connections between the termini. After formation of the synaptic complex, the termini are ready to be processed in a sequential manner by the polymerase and nuclease activities before ligation.
Physiological importance of bacterial NHEJ
NHEJ break repair is not conserved in all bacterial species, suggesting that this repair pathway may not be essential for cell survival . If this is the case, then why has NHEJ been retained by the vast majority of prokaryotic organisms? Recent reports suggest that NHEJ plays an essential role in the survival of both sporulating and non-sporulating bacteria. Although NHEJ mutant strains (ligD and ku) of the non-sporulating mycobacterium Mycobacterium smegmatis are not sensitive to DSBs induced by IR (ionizing radiation) during exponential growth phase, stationary-phase cells are much more sensitive to such breaks , establishing that NHEJ is much more important for the survival of non-dividing cells. Although M. smegmatis strains deficient in NHEJ are sensitive to IR-induced damage, terrestrial environments do not produce high doses of such radiation. Significantly, IR-resistant bacteria can be isolated from natural sources by selecting for strains that are resistant to desiccation, suggesting that IR-resistance is simply a result of the adaptation to a common physiological stress, namely desiccation  and, notably, NHEJ-deficient M. smegmatis strains are very sensitive to desiccation .
During transmission within airborne droplets, desiccation may be a real threat to the survival of many bacterial species, including pathogens such as M. tuberculosis. NHEJ may confer resistance to desiccation and is likely to be essential for the survival of such organisms. In addition, M. tuberculosis can survive for many years in a dormant state inside macrophages, while retaining the ability to initiate infection when the host immunity is compromised. Bacilli residing within macrophages are continually exposed to endogenous genotoxic stresses, such as oxidative damage. It is likely that the primordial NHEJ repair pathway evolved to protect the genomes of quiescent bacteria against the genotoxic damage induced by extreme environmental stresses, where HR-mediated repair is not a viable option .
Under conditions of limited nutrients, many prokaryotic species (e.g. Bacillus) undergo a programmed cell differentiation process known as sporulation . The resulting spores protect the dormant haploid cell from a wide range of environmental stresses, including UV radiation, freezing, drying and heating. It is essential that genome stability is maintained under conditions of prolonged dormancy so that viable spores can germinate and actively divide once food becomes available. Bacillus subtilis spores deficient in NHEJ are highly sensitive to stresses that induce DSBs [7,32,33], therefore a functional NHEJ repair pathway is essential for bacterial spore viability under conditions that give rise to DSBs. Indeed, the expression of NHEJ genes in B. subtilis is specifically up-regulated upon induction of sporulation , suggesting that NHEJ plays a specific role in maintaining genome stability in spores.
This work was funded by the Biotechnology and Biological Science Research Council [grant numbers 8/C17246, BB/D522746 and BB/F013795/1].
DNA Damage: from Causes to Cures: Biochemical Society Annual Symposium No. 76 held at Robinson College, Cambridge, U.K., 15–17 December 2008. Organized and Edited by Richard Bowater (University of East Anglia, U.K.), Rhona Borts (Leicester, U.K.) and Malcolm White (St. Andrews, U.K.).
DNA-dependent protein kinase
ligase domain of LigD
nuclease domain of LigD
polymerase domain of LigD
X-ray repair complementing defective repair in Chinese-hamster cells 4