Mycobacterium tuberculosis and Mycobacterium marinum non-homologous end-joining proteins can function together to join DNA ends in Escherichia coli

Abstract

Mycobacterium tuberculosis and Mycobacterium smegmatis express a Ku protein and a DNA ligase D and are able to repair DNA double strand breaks (DSBs) by non-homologous end-joining (NHEJ). This pathway protects against DNA damage when bacteria are in stationary phase. Mycobacterium marinum is a member of this mycobacterium family and like M. tuberculosis is pathogenic. M. marinum lives in water, forms biofilms and infects fish and frogs. M. marinum is a biosafety level 2 (BSL2) organism as it can infect humans, although infections are limited to the skin. M. marinum is accepted as a model to study mycobacterial pathogenesis, as M. marinum and M. tuberculosis are genetically closely related and have similar mechanisms of survival and persistence inside macrophage. The aim of this study was to determine whether M. marinum could be used as a model to understand M. tuberculosis NHEJ repair. We identified and cloned the M. marinum genes encoding NHEJ proteins and generated E. coli strains that express the M. marinum Ku (Mm-Ku) and ligase D (Mm-Lig) individually or together (LHmKumLig strain) from expression vectors integrated at phage attachment sites in the genome. We demonstrated that Mm-Ku and Mm-Lig are both required to re-circularize Cla I-linearized plasmid DNA in E. coli. We compared repair of strain LHmKumLig with that of an E. coli strain (BWKuLig#2) expressing the M. tuberculosis Ku (Mt-Ku) and ligase D (Mt-Lig), and found that LHmKumLig performed 3.5 times more repair and repair was more accurate than BWKuLig#2. By expressing the Mm-Ku with the Mt-Lig, or the Mt-Ku with the Mm-Lig in E. coli, we have shown that the NHEJ proteins from M. marinum and M. tuberculosis can function together to join DNA DSBs. NHEJ repair is therefore conserved between the two species. Consequently, M. marinum is a good model to study NHEJ repair during mycobacterial pathogenesis.

Introduction

In mammalian cells, the two main pathways for repairing DNA double strand breaks (DSBs) are homologous recombination (HR) and classical non-homologous end-joining (C-NHEJ, for review see 1). HR predominates in S and G2 phase, is accurate but requires homologous sequence from the sister chromatid to complete repair. C-NHEJ can function throughout the cell cycle and joins DNA DSBs of any sequence. Repair can therefore be inaccurate as well as accurate. There is also a minor DNA DSB repair pathway called alternative non-homologous end-joining (A-NHEJ), which inaccurately joins DNA DSBs at regions of microhomology (1,2). A-NHEJ therefore results in different repair products to C-NHEJ. A-NHEJ and C-NHEJ also use different proteins for repair. Ku protein, DNA protein kinase catalytic subunit (DNAPKcs), XRCC4, a number of termini modifying proteins and DNA ligase IV are involved in C-NHEJ, while the Mre11–Rad50–NBS1 complex, Parp1, XRCC1, DNA polymerase β and DNA ligase I or III are required for A-NHEJ (1). C-NHEJ is the predominant DNA DSB repair pathway in mammalian cells and is essential not only for genome stability but also for V(D)J recombination and hence the immune system.

In prokaryotes, HR is the predominant DNA DSB repair pathway and it was only in 2001 when prokaryote NHEJ was revealed by the identification of potential NHEJ proteins by bioinformatic analyses (3–5). Prokaryote NHEJ (for review see 6,7) is similar to C-NHEJ as they are both dependent on Ku: In mammalian cells, Ku is a heterodimer protein comprising of Ku70 and Ku80, while in prokaryotes there is usually a single Ku protein (30–40kDa) that forms a homodimer. Prokaryote Ku has structural similarities to the central DNA binding domain of its mammalian counterpart (1,6) and both prokaryote NHEJ and C-NHEJ are initiated by Ku binding the DNA ends of a DSB. In C-NHEJ, Ku70/80 then recruits DNAPKcs to the DSB, DNA damage signaling is initiated, and other repair proteins bind to the damaged site and process the DNA ends to generate 3′ hydroxyl and 5′ phosphate termini (1,8). Ku70/80 is also implicated in bridging the two DNA ends and stabilizing the repair complex (9–11), allowing DNA ligase IV to join the DNA DSB. Prokaryote Ku is required for recruiting DNA ligase D to the DNA DSB (12), and DNA ligase D is a multifunctional protein with three domains: a DNA polymerase domain, a 3′ phosphoesterase/nuclease domain and an ATP-dependent DNA ligase domain (13,14). DNA ligase D provides all the activities for modifying and ligating the DNA termini (6,7). Prokaryote NHEJ is therefore much simpler than C-NHEJ as it can be performed with only two proteins (15,16). UvrD1, a DNA helicase from M. smegmatis, is the only other protein implicated in prokaryote NHEJ. Ku interacts with UvrD1 and stimulates the helicase activity (17). UvrD1 may therefore unwind the DNA termini at the DSB and so assist Ku and ligase D. A ligase D-independent NHEJ has also been described using DNA ligase C1 instead of the ligase domain of DNA ligase D (18). However, this NHEJ activity still requires the polymerase domain of ligase D (18).

Many, but not all, prokaryotes have the capability to perform NHEJ and Bacillus subtilis, Mycobacterium tuberculosis and M. smegmatis were among the first prokaryotes identified as able to repair DNA by NHEJ (3,5,15,16,19). Why specific prokaryotes are capable of NHEJ is not fully understood and the biological relevance and the functions of the Ku and ligase D may differ in different types of bacteria. Recently Ku was implicated in the protection of M. smegmatis against zinc toxicity (20) and it has been suggested that prokaryote NHEJ may play a role in genome diversification (21). The ability of a bacterial species to repair DNA DSBs using NHEJ may be related to the amount of time the bacteria exist in their natural environment in a non-replicating state. Evidence indicates that NHEJ protects the genome from DNA damage when bacteria are in stationary phase or a non-replicating state, while HR protects the genome when bacteria are replicating. B. subtilis spores carrying mutations in one or both of the Ku or ligase D genes were more sensitive to extreme dryness, UV and ionizing radiation compared to wild-type spores (22), while M. smegmatis mutants of Ku and/or ligase D were only more sensitive to ionizing radiation and desiccation during stationary phase (23). M. smegmatis deficient in recA, and hence HR, were more sensitive to damage during exponential and stationary phase with the defect being more pronounced in exponential phase (23). The type of DNA DSB may also determine the type of repair utilized by the cell. Mycobacterial NHEJ was found to be more important than HR in the repair of DNA DSBs with noncompatible DNA termini (24).

M. tuberculosis is a pathogenic mycobacterium that results in tuberculosis in humans and M. tuberculosis NHEJ proteins (Mt-Ku and Mt-Lig) have been extensively studied in vitro. More recently, M. tuberculosis strains deleted in Rv0937c (Mt-Ku) and Rv0938 [M. tuberculosis ligase D (Mt-Lig)] were found to be able to survive inside macrophage if HR was functional. However, loss of NHEJ and HR decreased the ability of M. tuberculosis to survive the damage induced by nitric oxide or superoxide inside the macrophage (25). However, virulence studies in mice and guinea pigs with M. tuberculosis-deficient in both HR and NHEJ indicated that both pathways could be eliminated without compromising survival of the bacteria (26). Further work is needed to understand why survival of the mutants is impaired in macrophage, but not in an in vivo infection model.

M. smegmatis, a biosafety level 1 organism, has been used to characterize mycobacterial NHEJ repair (23,27). Apart from examining the accuracy and mechanism of repair (19,27,28), work in M. smegmatis has revealed that acetylation of Ku correlates with a decrease in NHEJ and that this point of regulation is controlled by the Sir2 deacetylase encoded by MSMEG_5175 (29,30). M. smegmatis Ku has also been shown to bind zinc and it has been proposed that Ku may bind zinc that is released during infection of macrophages, hence promoting survival of mycobacteria in an environment where zinc could be toxic (20). M. smegmatis is not a pathogenic mycobacterium and does not survive and persist inside macrophage by mechanisms similar to M. tuberculosis. Studies to examine the relevance of zinc binding or Ku acetylation to NHEJ activity during infection cannot therefore be studied in M. smegmatis. M. marinum is a waterborne pathogenic mycobacterium and is closely related to M. tuberculosis with ~85% amino acid identity for the ~3000 orthologs of the two mycobacterium species (31,32). M. marinum deploys similar mechanisms to M. tuberculosis to survive and persist inside macrophage and results in a chronic wasting disease in some fish (31). M. marinum can result in human skin infections similar to dermal M. tuberculosis infection, however human infections are usually limited to the skin or the extremities as M. marinum grows poorly at 37°C (31). The optimal growth for M. marinum in the laboratory is 30–32°C (33). While M. tuberculosis is a biosafety level 3 organism, M. marinum is a bisoafety level 2 organism and is now an accepted model of pathogenic mycobacteria (31,34,35). We therefore cloned the M. marinum Ku and ligase D genes and expressed them in E. coli to test their ability to repair DNA DSBs using a plasmid DNA recircularisation assay. We previously generated E. coli expressing Mt-Ku and Mt-Lig and determined repair was ~90% inaccurate (16). We have now compared the accuracy of repair between M. marinum and M. tuberculosis NHEJ in E. coli, and demonstrated that the M. marinum and M. tuberculosis NHEJ repair proteins can work together to perform NHEJ. This study therefore provides evidence that the NHEJ pathways in the two organisms are very similar and that M. marinum is a good model to study NHEJ during pathogenesis.

Materials and methods

Bacterial strains

A wild-type E. coli K12 strain (BW35-Hfr KL16(PO-45) thi-1 relA1 spoT1 e14^-λ^-), previously obtained from Dr. Susan S. Wallace (University of Vermont, Burlington, VT, USA), provides the background for the E. coli strains generated in this work. BWKuLig#2, BWKu containing pAH69 and BWLig containing pINT-ts were generated previously (16). These strains express the Mt-Ku and/ or Mt-Lig proteins after growth in arabinose. BW23474 was obtained from Coli Genetics Stock Center (Yale University, New Haven, CT, USA) and used to propagate the derivatives of pAH143 and pLA2. M. marinum (strain 1218R; (36)) was previously obtained from Lucia Barker (University of Minnesota) and was grown in Middlebrook 7H9 broth (Difco, Michigan) supplemented with ADS (5% BSA Fraction V, 2% Dextrose, 0.81% NaCl) to 10%, 0.5% glycerol, 100 μg/ml cycloheximide and 0.2% Tween 80. All other bacteria were grown in Luria-Bertani (LB) broth supplemented with the appropriate antibiotic to maintain plasmid replication or to select for plasmid integration.

Plasmids

pUCDUB is a derivative of pUC19 and pLA2, pAH69 and pINT-ts (37) were originally obtained from Mary Berlyn from the Coli Genetics Stock Center (Yale University). pLA2 encodes kanamycin resistance, while pUCDUB, pAH69 and pINT-ts encode carbenicillin resistance. pAHHisKu (a derivative of the CRIM vector pAH143 that integrates into the HK022 site in the E. coli genome; 16) encodes gentamicin-resistance and Mt-Ku. The two plasmids used in the end-joining assay (16) were pBestluc that expresses firefly luciferase and confers carbenicillin resistance and pACYC184 that encodes chloramphenicol resistance. To propagate replicating plasmids in bacteria, the following antibiotics were added to the medium: 15 μg/ml gentamicin, 50 μg/ml kanamycin, 34 μg/ml chloramphenicol or 100 μg/ml carbenicillin.

Oligodeoxyribonucleotides

Oligodeoxyribonucleotides (oligonucleotides) were purchased from Eurofins/Operon Technologies Inc. (Alameda, CA, USA) or the DNA facility at Iowa State University. Primers used to clone and sequence the Mm-Ku and M. marinum ligase D (Mm-Lig) genes are shown in Table 1. Two sets of PCR primers were used to independently clone each gene (Ku and Ligase D) twice from 1218R genomic DNA.

Table 1.

Oligonucleotides used to clone and sequence the Mm-Ku and Mm-Lig genes

Name	DNA sequence	Use
MM1	d(AAAAAACATATGCGCTCCATCTGGAAGGGTTC)	To clone Mm-Ku from genomic DNA. Product ligated into pLA2
MM2	d(AAAAAAGGTACCTCAGGACTTGGCAGCTTGTTTGGC)
MM6	d(AAAAAAAAGCTTATGTGAACAAGTCGGGCGC)	To clone Mm-Ku from genomic DNA. Product ligated into pUCDUB
MM7	d(AAAAAAGGATCCGACCCGGGTTCTGGGC)
MM3	d(AAAAAACATATGGGTTCGGCGCAGGCCTGG)	To clone Mm-Lig from genomic DNA. Product ligated into pAH vector
MM4	d(AAAAAAGGTACCTCATTCGCGCACCACCTCACTCG)
MM5	d(AAAAAAGGATCCTCACCGTCCGCACTTCG)	To clone Mm-Lig from genomic DNA. Product ligated into pUCDUB
MM8	d(AAAAAAAAGCTTCACCGATCACCAGCCCACG)
MM9	d(TCCAGGAAGTAGCTGCGG)	To sequence Mm-Ku gene
MM10	d(GTCATGGTGGTGCATACCC)
M13For	d(GTTGTAAAACGACGGCCAGT)	To sequence Mm-Ku and Mm-Lig in pUCDUB
M13Rev	d(TCACACAGGAAACAGCTATGA)
MM16	d(GTGTTCCTGGACTGGAGCC)	To sequence Mm-Lig gene
MM17	d(TGCTGGTCGAAGCCGACC)
MM18	d(GCTGCGGGAGCGCAGTCG)
MM19	d(TGGATCGTTGCTCATGG)
MM20	d(TGGCGCCGCTGCATACCG)

The oligonucleotides His3 and His4 contained 5′ phosphates and were used to generate the double-stranded DNA sequence encoding the N-terminal histidine tag upstream of the Mm-Ku and Mm-Lig coding sequences in the expression vectors (Table 2). The CRIM1 and CRIM 3 primers were used to sequence the Mm-Ku and Mm-Lig expression constructs (Table 2) and primers P1λ, P4λ, P1HK, P4HK, P2 and P3 (Table 2) were used for PCR to check for plasmid integration(16) into the E. coli genome.

Table 2.

Oligonucleotides used to generate the expression vectors, check for integration of the vectors in the E. coli genome and to analyze the repair products from the DNA repair assay.

Name	DNA sequence	Use
His3	d(TATGGGCCATCATCATCATCATCATCATCATCATCACAGCAGCGGCCATATCGAAGGTCGTCA)	Primers annealed and inserted into the expression vectors to add the N-terminal histidine tag
His4	d(TATGACGACCTTCGATATGGCCGCTGCTGTGATGATGATGATGATGATGATGATGATGGCCCA)
CRIM1	d(ATCCATAAGATTAGCGG)	Primers used to sequence the Mm-Ku and Mm-Lig expression vectors
CRIM3	d(CGGATATTATCGTGAGG)
P1λ	d(GGCATCACGGCAATATAC)	Primers for PCR to check for plasmid integration at the λ or HK022 phage attachment sites in the E. coli genome
P4λ	d(TCTGGTCTGGTAGCAATG)
P1HK	d(GGAATCAATGCCTGAGTG)
P4HK	d(GGCATCAACAGCACATTC)
P2	d(ACTTAACGGCTGACATGG)
P3	d(ACGAGTATCGAGATGGCA)
Luc1	d(TGGATGGCTACATTCTG)	Primers used to analyze the repair products from the plasmid DNA end-joining assay
Luc5	d(GCCTGGTATCTTTATAG)
Hind3	d(GAACGTGACGGACGTAAC)
Luc3	d(ATGTGGATTTCGAGTCGTCT)
R	d(TCATCGTCTTTCCGTGCT)

Primers Luc1, Luc5, Luc3, Hind3 and R (Table 2) were used to sequence or PCR the repair products to determine the size of deletions in the repair products in the DNA end-joining assay.

Preparation of M. marinum genomic DNA

M. marinum was grown to late log-phase at 30°C and glycine added to a final concentration of 1%. The culture (10ml) was grown for a further 24h before harvesting the bacteria by centrifugation. The bacteria were resuspended in 1ml GTE (50mM glucose, 25mM Tris-HCl pH 8, 10mM EDTA), subjected to centrifugation and then resuspened in 450 μl GTE and 50 μl lysozyme solution (10mg/ml lysozyme, 25mM Tris–HCl pH 8.5). The sample was incubated overnight at 37°C, prior to addition of 100 μl SDS and 50 μl 10mg/ml proteinase K and incubation at 55°C for 40min. Sodium chloride was added to a final concentration of 1.2M prior to addition of 65°C CTAB solution (0.3M Cetrimide, 0.78M NaCl) and incubation at 65°C for 10min. The sample was then extracted twice with an equal volume of chloroform-isoamyl alcohol (24:1) and the DNA was precipitated from the aqueous layer with 0.7 volumes of isopropanol. The DNA was collected by centrifugation, washed with 70% ethanol and resuspended in TE (10mM Tris–HCl pH 8, 1mM EDTA).

PCR amplification of the M. marinum Ku and Ligase genes and production of bacterial expression plasmids.

Coding regions for Mm-Ku and Mm-Lig were amplified using M. marinum 1218R genomic DNA, sequence specific primers (Table 1) and Phusion™ DNA polymerase according to maufacturer’s recommendations using annealing temperatures of 67°C for MM1 and MM2, 68°C for MM3 and MM4, 66°C for MM5 and MM8, and 61°C for MM6 and MM7. PCR products generated using MM1 and MM2 or MM3 and MM4 were digested with Nde I and Kpn I, purified following electrophoresis through an agarose gel and ligated into the Nde I-Kpn I sites in either pLA2 or pAHHisKu generating pLmKu and pAHmLig, respectively. Ligation products were transformed into a pir⁺E. coli strain (BW23474) for selection and propagation. Cloned Mm-Ku and Mm-Lig inserts were sequenced by the DNA facility at Iowa State University. pLmKu was then digested with Nde I and a double-stranded oligonucleotide encoding a N-terminal histidine tag was ligated upstream of the Mm-Ku coding region. The double-stranded oligonucleotide was generated by annealing His3 and His4 (16). The plasmid generated (pLHismKu) was used to create E. coli strains (Table 3) that expressed histidine-tagged Mm-Ku when bacteria were grown in 0.2% L-arabinose.

Table 3.

E. coli strains created for this study

Strain created	Starting strain	Plasmid integrated	NHEJ protein(s) expressed when grown in arabinose
LHmKu	BW35	pLHismKu	M. marinum Ku
LHmLig	BW35	pAHHismLig	M. marinum Ligase
LHmKumLig	LHmLig	pLHismKu	M. marinum Ku M. marinum Ligase
LHmKutLig	BWLig (ref. 16)	pLHismKu	M. marinum Ku M. tuberculosis Ligase
LHtKumLig	BWKu (ref. 16)	pAHHismLig	M. tuberculosis Ku M. marinum Ligase

PCR products generated using MM5 and MM8 or MM6 and MM7 were digested with BamHI and HindIII, purified following electrophoresis through an agarose gel and ligated into the BamHI and HindIII sites of pUCDUB generating pDUBmLig and pDUBmKu, respectively. Ligation products were transformed into E. coli TOP10 for selection and propagation. The cloned Mm-Ku and Mm-Lig genes were sequenced by the Iowa State University DNA Facility. pDUBmLig was used to generate pAHHismLig. Briefly a blunt-end cut at the 3′ end of the Mm-Lig gene was generated by digestion with BamHI and treatment with Klenow fragment and the Mm-Lig insert removed following further digestion with NdeI; there is a NdeI site (CATATG) at the translation start site of the Mm-Lig gene. To generate NdeI-blunt end linear DNA from the pAHHisKu plasmid, the plasmid was digested with KpnI and treated with Klenow fragment and the insert encoding histidine-tagged Mt-Ku removed after further digestion with NdeI. The NdeI-blunt Mm-Lig insert was then ligated into the pAH linear DNA producing pAHmLig. The sequence of the Mm-Lig coding region was re-checked prior to digestion with NdeI and insertion of a double-stranded oligonucleotide encoding a histidine tag (annealed His3 and His4) at the 5′ end of the Mm-Lig gene. This plasmid (pAHHismLig) was sequenced again before being used to generate E. coli strains expressing histidine –tagged Mm-Lig from the arabinose-inducible promoter (Table 3).

Integration of pAHHismLig or pLHismKu into the E. coli chromosome

The plasmids were integrated into the E. coli chromosome as previously described (16,37). Bacteria contained pAH69 for integration of pAH143-derived plasmids, or pINT-ts for integration of pLA2-derived plasmids and were grown at 30°C on solid medium or in liquid culture containing 100 μg/ml carbenicillin. Electrocompetent bacteria were prepared according to Seidman et al. (38) and bacteria transformed with 100ng of the integrating plasmid. Bacteria were then grown at 37°C for 1h, followed by 42°C for 30min. Replication of pAH69 and pINT-ts is inhibited at these temperatures, while the integrases expressed by pAH69 and pINT-ts are activated. The integrases are required for plasmid insertion into the E. coli chromosome at the λ phage attachment site (pLA2-derived vectors) or the HK022 phage site (pAH143-derived plasmid). Bacteria were grown on solid medium containing 5 μg/ml gentamicin (pAH143-derived plasmid) or 10 μg/ml kanamycin (pLA2-derived plasmids). After incubation at 37°C overnight, colonies were grown on solid LB medium without antibiotic before regrowth on the solid medium containing antibiotic. Each strain was tested for carbenicillin resistance to make sure that pAH69 or pINT-ts were no longer present in the new bacterial strains.

Single copy plasmid integration was confirmed by PCR using primers P1(HK or λ), P2, P3, and P4 (HK or λ)as previously described (16).

Western analysis

Cell-free extracts were prepared from bacteria grown in the absence (LB) or presence (Ara+) of 0.2% arabinose for 1h according to Wright et al. (39) Protein (50 µg) was subjected to electrophoresis through a tris-glycine 4–20% gradient SDS polyacrylamide gel and transferred to 0.2 µm nitrocellulose membrane. Western analysis was performed using an anti-His-tag monoclonal antibody (final concentration 0.2 µg/ml; Novagen, Madison, WI, USA), a secondary horse radish peroxidase antibody (diluted 1:3000; Amersham, Piscataway, NJ, USA) and chemiluminescent substrate (ECL-plus substrate, Amersham, Piscataway, NJ, USA; 39).

Plasmid end-joining assay

The Cla I recognition sequence in pBestluc is situated in the coding region of firefly luciferase and accurate repair of Cla I-linearized DNA is required to result in expression of active firefly luciferase in bacteria. The end-joining assay was performed using Cla I-linearized pBestluc according to Malyarchuk et al. (16) Briefly, bacteria were grown without (LB) and with (Ara+) the addition of L-arabinose (0.2%) and prepared for electroporation (38). Bacteria were co-transformed with linearized pBestluc (50ng) and 0.1ng pACYC184 using the Bio-Rad Gene Pulser Xcell Electroporation System (Hercules, CA, USA) at 2.5 KV, 200 Ω and 25 μF, transferred to 1.5ml LB and grown for 1.5h at 37°C and 250rpm. A portion of the culture was grown overnight at 37°C on triplicate plates containing solid medium supplemented with 100 µg/ml carbenicillin to identify bacteria containing re-circularized pBestluc, or 34 µg/ml chloramphenicol to identify bacteria containing pACYC184. The colonies were counted and a ratio of the carbenicillin resistant (Carb^R) colonies/chloramphenicol resistant (Cm^R) colonies was calculated for each transformation. This ratio corresponds to the total repair of pBestluc. By normalizing the Carb^R colonies to the Cm^R colonies, each sample was normalized for bacterial transformation efficiency. To identify colonies containing accurate or inaccurately repaired pBestluc, Carb^R colonies were transferred to nylon membranes and sprayed with firefly luciferase substrate. Colonies with accurately repaired pBestluc emit light when treated with this substrate due to expression of active luciferase. The number of colonies that expressed active (Luc⁺) or inactive (Luc⁻) firefly luciferase were counted and the ratio of Luc⁺Carb^R/ Cm^R (accurate repair) and Luc^- Carb^R/Cm^R (inaccurate repair) was calculated for each transformation. The percentage of accurate or inaccurate repair for each transformation was calculated:

$$\%\text{ Accurate}:\frac{{\text{Luc}}^{+}{\text{Carb}}^{\text{R}}/{\text{Cm}}^{\text{R}}}{{\text{Carb}}^{\text{R}}/{\text{Cm}}^{\text{R}}\left(\text{Total Repair}\right)}\text{\hspace{0.17em}X\hspace{0.17em}1}00$$

$$\%\text{ Inaccurate}:\frac{{\text{Luc}}^{-}{\text{Carb}}^{\text{R}}/{\text{Cm}}^{\text{R}}}{{\text{Carb}}^{\text{R}}/{\text{Cm}}^{\text{R}}\left(\text{Total Repair}\right)}\text{\hspace{0.17em}X\hspace{0.17em}1}00$$

PCR was used (16,39) to determine the presence of deletions or insertions in pBestluc in Luc^- Carb^R colonies. The sequence across the junction was determined by isolating plasmid from the bacteria using the Wizard Plus Miniprep DNA purification system (Promega) and sequencing was performed by the DNA Facility at Iowa State University (Ames, IA, USA).

Statistical analysis

The Carb^R/Cm^R ratios from the end-joining assays were compared for each bacterial strain grown in LB or Ara+ using Student–Newman–Keuls Test. The amount of total repair (Carb^R/Cm^R ratio), accurate repair (Luc⁺Carb^R/Cm^R) or inaccurate repair (Luc⁻ Carb^R/Cm^R) for all the strains grown in either LB or Ara⁺ were compared using the least squares means test with Tukey–Kramer adjustment for multiple comparisons.

To determine whether there was an equal distribution of deletion sizes for the size categories (< 50bp, 51–500bp, 501–1000bp, >1001bp), samples were tested to determine whether each size category contained 25% of the total population by using a Chi-Square test.

All analyses were performed using the SAS 9.4 program (SAS Inc, Gary, NC, USA). All P < 0.05 were considered statistically significant.

Results

Cloning the M. marinum Ku and ATP-dependent DNA ligase D genes from strain 1218R

The coding sequence for Mt-Ku (Rv0937c; genbank accession number AL123456, strain H37Rv) or Mt-Lig (Rv0938; genbank accession number AL123456, strain H37Rv) was used with the NCBI nucleotide blast program (http://blast.ncbi.nlm.nih.gov/Blast.cgi) to identify the M. marinum orthologs of these genes in M. marinum strain M (accession number CP000854.1). The Mt-Ku sequence identified gene MMAR_4575 with 86% identity to the Rv0937c sequence, while the Mt-Lig sequence identified MMAR_4573 with 80% identity to the Rv0938 sequence. The Ku and ligase D gene arrangements on the M. marinum and M. tuberculosis chromosomes are shown in Figure 1. While there are only 115 base pairs (bp) between the Ku and ligase D genes in M. tuberculosis, there are 1105bp between the two genes in M. marinum. In fact, there is a gene (MMAR_4574) encoding fructokinase situated between the Ku and ligase D genes on the M. marinum chromosome. There are no known orthologs in M. tuberculosis for this gene (http://mycobrowser.epfl.ch/marinolist.html) and no gene was identified in M. tuberculosis when the NCBI nucleotide blast program was run with the sequence for MMAR_4574.

Figure 1.

Operons for the non-homologous end-joining proteins. The Ku and ligase D open reading frames are transcribed in the opposite direction to each other on the chromosome in M. marinum (A) and M. tuberculosis (B). The chromosome nucleotide numbers are from Strain M for M. marinum (Refseq number NC 010612) and from H37Rv for M. tuberculosis (Refseq number NC 000963).

To clone the Mm-Ku and Mm-Lig genes from the 1218R strain of M. marinum, genomic DNA was isolated from the bacteria and PCR primers were designed using the sequence from strain M for MMAR_4575 or MMAR_4573. Since PCR can introduce errors during amplification a high fidelity DNA polymerase (Phusion™) was used for the PCR. Each gene was also independently cloned twice using two different PCR strategies and the gene sequence from both cloning strategies had to be identical to be considered as accurate gene sequence.

The Mm-Ku gene from 1218R genomic DNA was amplified with primers MM1 and MM2 or MM6 and MM7 (Table 1) and cloned into the NdeI-KpnI sites of pLA2 and the BamHI–HindIII sites of pUCDUB, respectively. The Mm-Ku coding sequence (876bp) from 3 clones of the resulting pLmKu and 1 clone of the pDUBmKu were sequenced on both DNA strands. The sequence was compared to the known MMAR_4575 sequence from M. marinum strain M and 4 point changes were identified and confirmed on all four clones. This was a 0.46% change in DNA sequence that resulted in two amino acid changes compared to the strain M gene (Figure 2). The two amino acid changes were a histidine at amino acid 86 instead of an arginine, and an arginine at amino acid 277 instead of a lysine.

Figure 2.

Alignment of the Ku amino acid sequences. The sequences for M. marinum (Strain M) and M. tuberculosis (H37Rv) were obtained from the database (Refseq number NC 010612 and NC 000963, respectively). The database sequences were aligned with the amino acid sequence translated from the Ku open reading frame cloned from M. marinum genomic DNA strain 1218R (this study). Alignments were generated using Blastp (http://blast.ncbi.nlm.nih.gov/Blast.cgi?PROGRAM=blastp&PAGE_TYPE=BlastSearch&LINK_LOC=blasthome). Line 1 marks the end of the Ku core domain and the beginning of the minimal C terminus and line 2 marks the beginning of the extended C terminus as defined by McGovern et al. (45) The single letters that are bold and underlined are the amino acids that are different compared to M. marinum strain 1218R.

The Mm-Lig gene from 1218R genomic DNA was amplified with primers MM3 and MM4 or MM5 and MM8 and cloned into the NdeI-KpnI sites of pAHHisKu and the BamHI-HindIII sites of pUCDUB, respectively. As for the Mm-Ku, multiple clones of the Mm-Lig coding sequence were sequenced (two in the pAH vector and four in the pUCDUB vector) and the DNA sequence compared to MMAR_4573 from M. marinum strain M. Eleven point changes out of 2313 nucleotides were identified on the multiple clones and this was a 0.48% change in DNA sequence compared to strain M. This resulted in two amino acid changes (Figure 3): a proline instead of alanine at amino acid 104 and a valine instead of alanine at amino acid 463.

Figure 3.

Alignment of the Ligase D amino acid sequences. The sequences for M. marinum (Strain M) and M. tuberculosis (H37Rv) were obtained from the database (Refseq number NC 010612 and NC 000963, respectively). The database sequences were aligned with the amino acid sequence translated from the ligase open reading frame cloned from M. marinum genomic DNA strain 1218R (this study). Alignments were generated using Blastp (http://blast.ncbi.nlm.nih.gov/Blast.cgi?PROGRAM=blastp&PAGE_TYPE=BlastSearch&LINK_LOC=blasthome). The single letters that are bold and underlined are the amino acids that are different compared to M. marinum strain 1218R.

Strain M was the first M. marinum strain to be completely sequenced and is frequently used as a model organism. More recently the complete sequence of strain E11 was submitted to genbank (accession number HG917972.2) and this E11 strain is almost identical in DNA sequence to strain 1218R for the Mm-Ku and Mm-Lig genes. Only one nucleotide difference was identified in each of these genes. The amino acid sequence was identical for the E11 and 1218R strains for Mm-Lig and only one amino acid change was identified between the two strains for Mm-Ku (arginine at amino acid 277 instead of a lysine). This amino acid change is conserved as both lysine and arginine have basic side chains and are positively charged at neutral pH. The changes in amino acid sequence between strain M, E11 and 1218R are likely normal strain variations. Previously, partial sequencing of seven protein coding sequences in 22 different M. marinum strains identified point changes in the sequence between the strains (40) and five different M. marinum genotypes were identified. Examination of a total of 26 M. marinum strains demonstrated less than 3% nucleotide variation (32). Therefore the less than 0.5% nucleotide variation in the Ku and ligase D genes between strain M and 1218R is within the range of variation seen between M. marinum strains from around the world.

The M. tuberculosis and M. marinum Ku protein sequences are 80% identical, but the Mt-Ku is 18 amino acids shorter than the Mm-Ku and the gaps in the protein are situated near the C-terminus of the protein (Figure 2). For ligase D, the M. tuberculosis and M. marinum protein sequences are also 80% identical and the Mt-Lig is 11 amino acids shorter than the Mm-Lig. Eight of the amino acids absent from the M. tuberculosis protein occur very close to the N-terminus of the protein (Figure 3).

Generation of E. coli that express M. marinum and M. tuberculosis Ku and Ligase proteins

Previously we generated an E. coli strain expressing the Mt-Ku and Mt-Lig (BWKuLig#2, Figure 4; 16) using the CRIM vector system developed by Haldiman and Wanner (37). In order to test the activity of the M. marinum Ku and ligase D, we generated two CRIM expression vectors using the cloned Mm-Ku and Mm-Lig genes: pLHismKu and pAHHismLig. These plasmids can express N terminal histidine-tagged Mm-Ku or Mm-Lig from the arabinose-inducible promoter (P_araB; 37). The advantage of the CRIM vector system is that these plasmids integrate into the E. coli chromosome and multiple vectors can integrate into the same genome: pLHismKu integrates into the lambda attachment site, while pAHHismLig integrates into the HK022 attachment site. pLHismKu and pAHHismLig were individually integrated into the KL16 E. coli strain (BW35) generating LHmKu and LHmLig, respectively (Table 3). This KL16 E. coli strain was previously used to create BWKuLig#2 (16). To generate a strain (LHmKumLig) that could be induced to express both the Mm-Ku and Mm-Lig, the pLHismKu plasmid was integrated into the genome of the LHmLig strain (Table 3). To determine whether the M. tuberculosis NHEJ proteins could function to repair DNA with the M. marinum NHEJ proteins, strains were created that could be induced to express the Mm-Ku and Mt-Lig (LHmKutLig) or the Mt-Ku and Mm-Lig (LHtKumLig; see Table 3). All strains were checked using PCR for single copy integration of the plasmids (16) and by western analyses with cell-free extracts from cultures grown with or without 0.2% arabinose (Figure 4) to confirm protein expression. The Ku and ligase proteins were detected using an antihistidine tag antibody. Proteins were only expressed when grown in the presence of arabinose and the size difference between the Mt-Ku and Mm-Ku is evident from the western analyses. Mm-Ku was also expressed at a higher level than Mt-Ku in E. coli (Figure 4), while expression of the Mt-Lig and Mm-Lig were similar.

Figure 4.

Expression of the NHEJ proteins in E. coli. Bacteria were grown with or without 0.2% L-arabinose (Ara) and cell-free extracts prepared and subjected to electrophoresis through a tris-glycine 4–20% gradient SDS polyacrylamide gel. Proteins were transferred to nitrocellulose and Mm-Ku, Mm-Lig, Mt-Ku and Mt-Lig proteins were detected with an anti-histidine tag antibody.

DNA end-joining activity of mycobacterium NHEJ proteins in E. coli

To test end-joining activity, Cla I-linearized pBestluc was co-transformed with pACYC184 into the E. coli strains. Re-circularisation of pBestluc results in carbenicillin resistant colonies (Carb^R), while pACYC184 encodes the chloramphenicol resistance (Cm^R) gene and is used to normalize for transformation efficiency: the greater the Carb^R/Cm^R ratio, the higher the level of total repair of pBestluc. Since the mycobacterium NHEJ proteins are only expressed in the presence of arabinose (Ara+), a comparison of the Carb^R/Cm^R ratios when the bacteria are grown in Ara+ or LB reveals the amount of total repair performed by the NHEJ proteins. Dividing the Carb^R/Cm^R from the Ara+ samples by the Carb^R/Cm^R for the LB samples provides a measure of how much total repair occurred above background. Ara+/LB ≤ 1 indicates that no plasmid re-circularisation occurred in the bacteria due to expression of the Ku and/or ligase. There was no statistical difference between the Carb^R/Cm^R for the LB and Ara+ samples for strains LHmKu and LHmLig (Figure 5). The Ara+/LB for LHmKu and LHmLig strains were 0.84 and 0.96, respectively, indicating repair did not occur when only one of the NHEJ proteins was expressed. However, when bacterial strains had two vectors integrated into their genome to express a Ku and a ligase the Ara+ Carb^R/Cm^R was significantly greater than the LB Carb^R/Cm^R (P < 0.05). This indicates repair occurred when both proteins were expressed in the same cell even when the Ku and ligase coding sequences were from different mycobacterium species (Figure 5).

Figure 5.

Recircularisation of plasmid DNA by the NHEJ proteins in E. coli. Bacteria were grown in LB or LB supplemented with 0.2% L-arabinose (Ara+) prior to the preparation of electrocompetent bacteria. Bacteria were then electroporated with Cla I-linearized pBestluc (50ng) and 0.1ng pACYC184 and the bacteria cultured on solid medium containing carbenicillin or chloramphenicol. The ratio of Carb^R to Cm^R colonies was calculated to compare the level of repair by the different E. coli strains expressing the Ku and/or Ligase proteins from M. marinum and M. tuberculosis. The average and standard error calculated from at least six transformations are shown graphically. For Ara+ samples, *P < 0.05 compared to LHmKumLig Ara+.

Expression of the Mm-Ku and Mm-Lig at the same time dramatically increased re-cirularisation of pBestLuc (Figure 5). In fact total repair was significantly greater in the E. coli strain expressing Mm-Ku and Mm-Lig (LHmKumLig), than in BWKuLig#2, which expresses the Mt-Ku and Mt-Lig (P < 0.0001). The Ara+/LB ratio was approximately 3.5 times higher for LHmKumLig (Ara+/LB = 169) than BWKuLig#2 (Ara+/LB = 49).

To determine whether the M. marinum and M. tuberculosis NHEJ proteins are interchangeable, repair was compared in strains expressing the Mm-Ku and the Mt-Lig (LHmKutLig) and the Mt-Ku and Mm-Lig (LHtKumLig). As can be seen from the Ara+ samples Figure 5, both of these strains were able to re-circularize pBestluc, but LHtKumLig repaired a significantly greater amount of plasmid than LHmKutLig (P < 0.0001). Interestingly, bacterial strains expressing the same ligase performed equivalent amounts of total repair: there was no significant difference for LHmKumLig (Ara+Carb^R/Cm^R = 0.87±0.11) compared to LHtKumLig (Ara+Carb^R/Cm^R = 0.88±0.05), and also for LHmKutLig (Ara+Carb^R/Cm^R = 0.089±0.02) compared to BWKuLig#2 (Ara+Carb^R/Cm^R = 0.15±0.01). Data also indicates that bacterial strains expressing a Ku and the Mm-Lig performed significantly greater amounts of total repair compared to those expressing a Ku and the Mt-Lig (Figure 5).

Cla I linearisation of pBestluc results in a DSB with a 2bp 5′ GC overhang in the 3′ end of the firefly luciferase coding region. If the sequence at the Cla I site is altered by deletions or insertions during DNA recircularisation then the firefly luciferase produced by the bacteria has no activity (16,39,41). To test for firefly luciferase activity, the colonies were transferred to nylon, sprayed with luciferase substrate and the filter placed against autoradiographic film. Colonies expressing active luciferase were identified by emission of light. The Luc⁺Carb^R/Cm^R (Figure 6A) and the Luc^-Carb^R/Cm^R (Figure 6B) ratios were calculated for the Ara+ bacteria and the bacteria grown in LB. There was no significant difference between any of the strains for Luc⁺Carb^R/Cm^R or Luc^-Carb^R/Cm^R when the bacteria were grown in LB. The background amount of circular plasmid expressing active or inactive luciferase was therefore identical in all strains when the NHEJ proteins were not expressed. When considering the Ara+ samples for Luc⁺Carb^R/Cm^R or Luc^-Carb^R/Cm^R (Figure 6A and B), the strains that expressed different ligases (Mt-Lig or Mm-Lig) performed significantly different amounts of accurate and inaccurate repair: LHmKumLig and LHtKumLig were comparable, but significantly different from BWKuLig#2 and LHmKutLig, and BWKuLig#2 results were similar to those from LHmKutLig.

Figure 6.

Accuracy of plasmid DNA repair by the NHEJ proteins in E. coli. The Carb^R colonies obtained from the plasmid DNA recircularisation experiments (Figure 5) were transferred to nylon and sprayed with luciferase substrate. Colonies containing accurately repaired plasmid express active luciferase and were identified by light emission. The ratio of Carb^R to Cm^R colonies was calculated for the colonies expressing active luciferase (Luc⁺) and inactive luciferase (Luc^-) and the percentage of accurate and inaccurate repair was calculated for each transformation. The average and standard error, calculated from at least six transformations, is shown graphically for the Carb^RLuc⁺/Cm^R (A), Carb^RLuc⁻/Cm^R (B) and for the percentage of accurate or inaccurate repair (C). For the Ara+ samples for Carb^RLuc⁺/Cm^R (A) and Carb^RLuc⁻/Cm^R (B), *P < 0.05 compared to LHmKumLig Ara+. For the % accurate repair (C), ^#P < 0.05 compared to the % of accurate repair for LHmKumLig.

Since strains expressing different types of ligases repaired significantly different amounts of plasmid, the percentage of accuracy of repair was examined (Figure 6C). The most accurate repair was performed by Mm-Ku and Mm-Lig in LHmKumLig (48% ± 1.8) and the least accurate was by Mt-Ku and Mt-Lig in BWKuLig#2 (8.7% ± 0.6%). The accuracy of repair by Mm-Ku and Mt-Lig (in LHmKutLig) and Mt-Ku and Mm-Lig (in LHtKumLig) were 39% ± 1.2 and 33% ± 4.7, respectively. Unlike the amount of repair achieved, the accuracy was not dictated only by the type of ligase expressed in the bacteria.

The products of inaccurate repair

Previously, we determined that Mt-Ku and Mt-Lig (BWKuLig#2) introduced deletions at the Cla I site during inaccurate repair that ranged in size from < 50bp to > 1000bp and that deletions occurred at regions of microhomology, with one of the repeat sequences deleted at the site of ligation (39). PCR was therefore used to determine the size of the deletion for the inaccurate repair products from LHmKumLig, LHmKutLig and LHtKumLig (Table 4).

Table 4.

PCR analysis of repair products

Strain	Number of colonies analyzed	Approximate deletion sizes in bp in Luc- colonies
		≤50	51–500	501–1000	>1000
LHmKumLig	67	5 (7.5%)	9 (13.4%)	10 (14.9%)	43 (64.2%)
LHtKumLig*	83	23 (27.7%)	23 (27.7%)	9 (10.8%)	28 (33.7%)
LHmKutLig*	50	13 (26%)	6 (12%)	8 (16%)	23 (46%)

Cla I-linearized DNA was transformed into the indicated strains and the Carb^R colonies were tested for luciferase activity. The deletions in pBestluc in the Luc^- colonies were determined using PCR followed by electrophoresis of PCR products through an agarose gel.

*P < 0.05 compared to LHmKumLig.

Inaccurate repair by all strains resulted in products with a wide range of deletion sizes. Statistically, no strain had an equal distribution (25%) of deleted products among the size groups (Table 4); indicating deletion sizes were not introduced randomly. The distribution of the number of deleted products in the size groups was significantly different between the bacterial strains, except for LHmKutLig and LHtKumLig, and only LHtKumLig had a similar distribution to the previously published results for BWKuLig#2 (39). The size of deletions introduced could therefore not be attributed to expression of one particular protein (Ku or ligase) from a particular mycobacterium species. LHmKumLig did have a higher percentage (64%) of deletions greater than 1000bp compared to that previously found for BWKuLig#2 (26%; 39) and compared to LHtKumLig (33.7%) and LHmKutLig (46%). Therefore even though the M. marinum Ku and ligase repaired a higher percentage of the Cla I-linearized pBestluc accurately (Figure 6C), larger deletions were introduced in the inaccurately repaired plasmids.

Inaccurately repaired plasmid was isolated from LHmKumLig and sequenced across the Cla I restriction region and examples are shown in Table 5. As previously found for BWKuLig#2, deletions occurred at regions of microhomology and one copy of the repeat sequence was deleted during the joining of the DNA ends.

Table 5.

Sequence of repair products from LHmKumLig

Deletion size	Sequence before repair	Sequence after repair	Microhomology region
12 bp	GAATTG−ATTGTTA	GAATTGTTA	ATTG
60 bp	CCGCTTG−GTTACAA	CCGCTTGTTACAA	G
134 bp	GTTGACCGC−CCGCCGCC	GTTGACCGCCGCC	CCGC
661 bp	GGTTGCAAA−CAAATTG	GGTTGCAAATTG	CAAA

LHmKumLig bacteria were transformed with Cla I-linearized DNA and the Luc^- Carb^R colonies were tested for deletions in pBestluc. Colonies were selected that contained a range in size of deletions and the pBestLuc was isolated from these bacteria and sequenced. The junctions of the deletion were identified and are shown above. The dash (−) in the ‘sequence before repair’ relates to the sequence between the microhomology regions in pBestluc. The microhomology regions are shown in bold.

Discussion

Studying NHEJ in E. coli has the advantage that there is little or no background activity in E. coli that will re-circularize transformed linear plasmid DNA. It is therefore easy to determine which ‘foreign’ proteins expressed in E. coli are taking part in NHEJ. As previously found for M. tuberculosis (15,16) and M. smegmatis (19), both the Ku and ligase D from M. marinum are required to repair DNA DSBs by NHEJ. This work also clearly demonstrates that the Mt-Ku can operate with Mm-Lig, and the Mm-Ku can function with Mt-Lig to repair the linear plasmid DNA. Mt-Ku does not co-ordinate repair with all DNA ligases. Mt-Ku could not stimulate DNA re-joining by T4 DNA ligase in vitro (15) and was unable to stimulate NHEJ in yeast when yeast NHEJ DNA ligase (Dnl4) was present (15). Therefore the ability of the M. marinum and M. tuberculosis Ku and ligase D proteins to be interchangeable is an indication of the high homology between the repair proteins and between the NHEJ repair pathways of M. marinum and M. tuberculosis.

Work in M. smegmatis using linearized plasmid has determined that DNA DSBs with 5′ overhangs or blunt end termini are the preferred substrates for mycobacterium NHEJ (19,28). For a DNA DSB with a 4bp 5′ overhang, repair was 43% accurate in wild-type M. smegmatis and repair of the plasmid DNA was reduced 125–200 times with deletion of the Ku or ligase D genes (28). Repair by M. smegmatis NHEJ was therefore predominantly inaccurate and repair products contained insertions, insertions with unidirectional deletions and a small number of bidirectional deletions. Using linearized plasmid DNA with a 2bp 5′ overhang, we have determined from this work and our previous study(16) that repair by M. tuberculosis or M. marinum NHEJ proteins in E. coli is also highly mutagenic (~91% and 52% inaccurate repair, respectively). As previously found for Mt-Ku and Mt-Lig (16), M. marinum NHEJ in our E. coli model results in inaccurate repair due to deletions. The lack of insertions may indicate that another mycobacterium protein is required in vivo to stimulate the function of the DNA polymerase domain of ligase D. The DNA end-joining also occurred at sites of microhomology with 1 copy of the repeat sequence deleted upon ligation. This repair product is similar to that generated by microhomology mediated end-joining in mammalian cells (42). Mt-Ku and Mt-Lig in vitro also use regions of microhomology as well as resection and gap-filling when repairing partially complementary DNA ends (15). The use of microhomology regions was not, however, evident in NHEJ products of repaired plasmid in M. smegmatis (19,28). Conversely, when the DNA DSB was introduced by I-SceI into the M. smegmatis genome (27) the NHEJ repair products predominantly contained deletions with joining at sequences of microhomology. M. marinum and M. tuberculosis NHEJ repair products in the E. coli model therefore resemble M. smegmatis NHEJ repair products of I-SceI cleaved genomic DNA. Even though the products of inaccurate repair are different in the different models, all the studies are in agreement that NHEJ is mutagenic in mycobacterium. Recently, prokaryote NHEJ repair was implicated in stationary-phase mutagenesis in Pseudomonas putida (21). NHEJ mutagenic repair in mycobacterium may therefore play a role in the generation of population diversity as well as maintaining an intact genome during phases of dormancy.

M. marinum NHEJ in the E. coli model system was more accurate than M. tuberculosis NHEJ (48 and ~9% accurate repair, respectively), and is similar to the level of accuracy of M. smegmatis NHEJ repair (43% accurate repair; 28). When examining the different strains expressing the combinations of Ku and ligase D the most accurate repair was in the order of Mm-Ku and Mm-Lig > Mm-Ku and Mt-Lig > Mt-Ku and MmLig >> Mt-Ku and Mt-Lig. This indicates that both the Mm-Ku and the Mm-Lig contribute to the increased accuracy of repair.

Sequence analysis predicts that the Mm-Lig has DNA polymerase, phosphoesterase/nuclease as well as a DNA ligase domain. Decreased activity of the DNA polymerase domain and potentially the phosphoesterase/nuclease domain compared to Mt-Ku could increase M. marinum NHEJ fidelity. However, there was no evidence of insertions in the repair products generated by LHmKumLig or BWKuLig#2 strains (16), so an altered DNA polymerase activity could not have changed the accuracy of repair in the E. coli system. The ligase activity could modify repair fidelity; slower ligation could result in the Ku–DNA complex destabilizing before ligase D completes repair. The Mm-Lig was able to achieve a greater amount of repair than Mt-Lig in the same amount of time no matter which Ku initiated repair. This suggests that in E. coli Mm-Lig repaired Cla I-linearized DNA faster or more efficiently than Mt-Lig. Although beyond the scope of this study, in vitro biochemical studies with pure proteins are needed to compare the ligase, polymerase and phosphoesterase activities of Mt-Lig and Mm-Lig to understand how the ligases differ in specific domain repair activities.

The higher level of Mm-Ku expression in E. coli compared to Mt-Ku could contribute to the greater fidelity of M. marinum NHEJ repair. A higher level of Ku in the cell could protect a greater number of DNA ends from degradation by E. coli nucleases. We previously demonstrated that total repair as well as accuracy was increased when Mt-Ku and Mt-Lig were expressed in a recB mutant E. coli strain (16). Loss of RecB likely prevented degradation of the linear plasmid DNA by the RecBCD nuclease, allowing the Mt-Ku more time to bind the DNA DSB before the termini were modified by E. coli nucleases. Protection of the DNA end by Ku will also be influenced by the binding affinity of Ku for the Cla I-induced DNA DSB, as well as the stability of the Ku–DNA complex. The increased accuracy of repair by M. marinum NHEJ could therefore be due to different DNA binding affinities of the Mm-Ku and Mt-Ku. Recently, work has highlighted the importance of the structure of the C-terminus of Ku in binding DNA (43–45). The prokaryote Ku structure is now defined in terms of the core DNA binding domain (from the N terminus to [1], Figure 2), the minimal C-terminus (between [1] and [2] on Figure 2) and the C terminal extension ([2] to the end of the protein in Figure 2; 43–45). The interaction of Ku and ligase D and Ku stimulation of ligase D activity requires the minimal C terminus (45), and the C terminal extension is implicated in altering the types of DNA that can be bound by Ku, the movement of the Ku along the DNA and the bridging of the two DNA ends by Ku (43–45). Mm-Ku and Mt-Ku have the same size core DNA binding domain and minimal C terminus (Figure 2), as well as high sequence homology in these regions. This explains why Mt-Ku and Mm-Ku can both bind Mt-Lig or Mm-Lig and why the type of Ku does not alter the total amount of repair achieved by each ligase D. However, Mm-Ku has a C-terminal extension that consists of 32 amino acids, 11 of which are basic (Figure 2), but Mt-Ku has only 14 amino acids beyond the region of the minimal C-terminus, only 3 of which are basic, and so does not have the C terminal extension (43,45). The basic residues may form electrostatic interactions with the DNA backbone (43,45). Mt-Ku requires DNA with free ends to bind DNA, while M. smegmatis Ku, which also has a C terminal extension, can bind supercoiled as well as linear DNA (44). Therefore, M. marinum Ku would be predicted to bind DNA that does not contain free ends, as well as DNA with a DSB. The C terminal extension is not required to protect DNA from degradation by exonucleases (45) and so lack of end protection by Mt-Ku does not explain why M. tuberculosis NHEJ is less accurate than M. marinum NHEJ. The M. smegmatis Ku C terminal extension has been found to promote DNA end-joining (43), and the B. subtilis Ku C terminal extension is involved in the formation of Ku–DNA networks, bringing DNA ends together, and limiting the internal movement of Ku on the DNA away from the DNA ends (45). It has been proposed that Ku proteins with the C terminal extension can be positioned on DNA before damage, then quickly move to the DNA ends when the DSB is induced and accumulate at the termini to promote DNA end bridging. The minimal C terminal section of Ku then promotes ligase D binding and ligation (45). The C terminal extension of Mm-Ku may therefore promote accurate repair by increasing the amount of Mm-Ku at the DNA termini, enhancing bridging of the DSB DNA termini and stabilizing the alignment of the cohesive DNA overhangs for ligation. The C terminal extension of M. marinum Ku and M. smegmatis Ku may explain why these two mycobacterium species have similar fidelity of NHEJ repair, and the lack of the C terminal extension on Mt-Ku may be why M. tuberculosis NHEJ is more mutagenic. Detailed biochemical studies will be required to confirm the function of the Mm-Ku C terminal extension in fidelity of NHEJ.

In summary, this work has shown that although there are differences between the accuracy and the amount of repair that can be achieved by M. marinum NHEJ and M. tuberculosis NHEJ proteins in E. coli, the NHEJ proteins from these two mycobacterium species are interchangeable and the pathways are very similar. Studying NHEJ in M. marinum under different stress conditions will therefore provide information about the importance of M. tuberculosis NHEJ for survival of this pathogenic mycobacterium.

Funding

This work was supported by the National Cancer Institute at the National Institutes of Health (CA 85693).

Genbank accession number for M. marinum Ku strain 1218R: KX258459.

Genbank accession number for M. marinum Ligase D strain 1218R: KX264330.