Mycobacterium smegmatis PafBC is involved in regulation of DNA damage response

Two genes, pafB and pafC, are organized in an operon with the Pup-ligase gene pafA, which is part of the Pup-proteasome system (PPS) present in mycobacteria and other actinobacteria. The PPS is crucial for Mycobacterium tuberculosis resistance towards reactive nitrogen intermediates (RNI). However, pafB and pafC apparently play only a minor role in RNI resistance. To characterize their function, we generated a pafBC deletion in Mycobacterium smegmatis (Msm). Proteome analysis of the mutant strain revealed decreased cellular levels of various proteins involved in DNA damage repair, including recombinase A (RecA). In agreement with this finding, Msm ΔpafBC displayed increased sensitivity to DNA damaging agents. In mycobacteria two pathways regulate DNA repair genes: the LexA/RecA-dependent SOS response and a predominant pathway that controls gene expression via a LexA/RecA-independent promoter, termed P1. PafB and PafC feature winged helix-turn-helix DNA binding motifs and we demonstrate that together they form a stable heterodimer in vitro, implying a function as a heterodimeric transcriptional regulator. Indeed, P1-driven transcription of recA was decreased in Msm ΔpafBC under standard conditions and induction of recA expression upon DNA damage was strongly impaired. Taken together, our data indicate an important regulatory function of PafBC in the mycobacterial DNA damage response.

a consensus sequence in the upstream region of these genes (tTGTCRgtg -8 nt -TannnT) that resembles the recA P1 promoter and is responsible for the observed LexA-independent regulation 14 . A recent study in Msm showed that the clp gene regulator (ClgR; MSMEG_2694/Rv2745c) recognizes the P1 promoter region 15 . Moreover, deletion of clgR in Msm results in impaired induction of various genes related to DNA repair mechanisms under DNA damaging conditions. This indicates a regulatory function of ClgR in the LexA/RecA-independent DNA damage repair pathway in mycobacteria.
A post-translational protein modification pathway akin to ubiquitination contributes to the survival of Mtb in host macrophages. In this modification pathway termed pupylation, mycobacteria (and other actinobacteria) make use of the small ubiquitin-like protein Pup to target proteins for degradation by the bacterial proteasome 16,17 . Pup is covalently attached via its C-terminal glutamate to lysine side chains of proteasomal substrates by action of the Pup ligase PafA 18,19 . Before ligation can occur, Pup, which in all mycobacteria is encoded with a C-terminal glutamine, must be rendered competent for ligation through deamidation of its C-terminal glutamine residue by the enzyme Dop (deamidase of Pup) 19,20 . Toward pupylated substrates Dop can act as a depupylase 21,22 , thereby counteracting the ligase activity. Pupylated proteins are recognized by the proteasomal ATPase ring, termed Mpa (mycobacterial proteasomal ATPase) in mycobacteria, which unfolds the substrate and translocates it into the proteolytic chamber of the proteasome complex [23][24][25] . The genes encoding the core proteins involved in pupylation and proteasomal degradation are clustered together in the so-called Pup-proteasome gene locus. In all mycobacterial species the gene pafA encoding the Pup ligase is organized in an operon together with two other genes, pafB and pafC 26 . Mtb PafB and PafC were shown to copurify, suggesting that they might form a complex 26 . Deletion of the coding genes does not affect pupylation in Mtb and therefore has no influence on the degradation of proteasomal substrates. Unlike an Mtb pafA deletion mutant, a strain lacking pafBC displays only a slight decrease in resistance against nitrosative stress in vitro 26,27 . However, PafB and PafC were linked to conjugal DNA transfer in Msm 28 . Deletion of the corresponding genes does not impair the donor strain's ability to transfer DNA. In contrast, the corresponding recipient strain is strongly compromised in DNA uptake. A recent study indicates that PafC contributes to the intrinsic resistance of mycobacteria to fluoroquinolone antibiotics 29 . However, the underlying mechanism for any of the observed phenotypes remained elusive.
In order to understand the biological function of PafB and PafC, we generated a pafBC deletion mutant of Msm SMR5 and employed a comparative proteomic approach using iTRAQ (isobaric Tags for Relative and Absolute Quantitation in mass spectrometry). Notably, a large part of the proteins that displayed decreased cellular levels in the deletion mutant are associated with DNA repair, including RecA. Accordingly, deletion of pafBC resulted in increased sensitivity towards DNA damaging agents, and induction of recA transcription in Msm ΔpafBC was strongly impaired under DNA damaging growth conditions. Strikingly, whereas the LexA/RecA-dependent recA promoter P2 remained inducible to wild type levels, induction of P1-driven transcription was completely blocked. Our findings establish pafBC as a new regulatory element in the LexA/RecA-independent DNA damage response and provide new insights into RecA regulation as well as the control of DNA repair mechanisms in mycobacteria.

M. smegmatis
PafB and PafC form a stable heterodimer and feature DNA binding motifs. The genes encoding PafB and PafC are located downstream of the Pup ligase gene pafA. Like in its pathogenic relative Mtb 26 , these three genes are co-transcribed in M. smegmatis as evidenced by the presence of a single mRNA encoding both PafC and PafB in the Msm wild type strain (Fig. 1A). Sequence analyses of PafB and PafC showed that they exhibit similarities to a multitude of bacterial transcription factors featuring a so-called winged helixturn-helix (wHTH) DNA-binding motif in their N-terminal region (Supplementray Fig. S1). The wHTH motif is typically comprised of three helices forming a compact helical bundle, which is capped by an adjacent hairpin ("wing") (Supplementray Fig. S1) 30 . Binding to DNA is usually mediated by insertion of helix α3 into the major groove (Supplementray Fig. S1). Furthermore, PafB and PafC contain a C-terminal WYL domain. This domain is often found in association with wHTH motifs and is possibly involved in protein dimerization and/or the binding of ligands, which thereby trigger a specific transcriptional response 31 . Indeed, our in vitro analysis of the assembly state of Msm PafB and PafC demonstrates that the two proteins form a heterodimer (Fig. 1B). Analytical gel filtration profiles of PafB and PafC expressed and purified separately (Fig. 1B, upper panel) indicate that PafB (36 kDa) with an apparent molecular weight of 40 kDa runs as a monomer, while PafC (34 kDa) runs at an apparent molecular weight of 84 kD, possibly due to homodimer formation in absence of PafB. However, when PafB and PafC are co-expressed and co-purified (Fig. 1B, lower panel), a single main elution peak shifted away from either of the positions of the individual proteins is detected at an apparent molecular weight in agreement with a PafBC heterodimer. Consistent with this interpretation, analysis of the elution peak in SDS-PAGE reveals that it contains equal amounts of PafB and PafC. Taken together, bioinformatic analysis along with the observed stable heterodimer formation between PafB and PafC imply that the two proteins might act as a heterodimeric transcriptional regulator in mycobacteria.
Generation of a M. smegmatis pafBC deletion strain. As an initial step towards understanding the role of PafBC in mycobacteria we generated a markerless deletion of the pafBC coding genes in the model organism Mycobacterium smegmatis (Msm). The mutant was obtained by transformation of Msm SMR5, the streptomycin resistant parent strain carrying a point mutation in the ribosomal protein S12 coding gene (rpsL) with the suicide plasmid pMCSΔpafBC-hyg-rpsL carrying the rpsL + gene and subsequent application of the rpsL counterselection strategy ( Fig. 2A) 32,33 . Disruption of pafBC was confirmed by Southern blot analysis, and the parent strain is hereafter referred to as Msm wild type, the deletion strain as Msm ΔpafBC. The used probe hybridized to a 2.6-kbp-fragment in the wild type strain and to a 6.5-kbp-fragment in the pafBC deletion mutant (Fig. 2B). A complemented strain was generated by transforming Msm ΔpafBC with the integrative plasmid pMV361-hyg-pafBC carrying a fragment consisting of the presumable promoter region of the pafABC operon fused to Msm pafBC. The complementation vector re-established cellular levels of PafBC similar to wild type (Fig. 2C). In vitro growth of the pafBC deletion mutant did not differ from Msm wild type when grown under standard conditions at 37 °C (Supplementray Fig. S2).
Since pafB and pafC form an operon with the Pup-ligase encoding gene pafA (Fig. 1A) we investigated whether pupylation is affected by the absence of PafBC. Anti-Pup immunoblot analysis of cell-free lysates deriving from Msm wild type or Msm ΔpafBC grown under standard conditions revealed no significant differences in the band pattern or band intensities of the pupylated proteome (pupylome), suggesting that PafA pupylation activity remains unaffected (Fig. 3A). In accordance with this, we found that deletion of pafB and pafC did not alter the relative cellular amount of pafA mRNA (Fig. 3B). Similar findings were made for other proteins involved in the Pup-proteasome pathway, namely Dop, Mpa, PrcA and PrcB, which all displayed unchanged cellular protein levels in the Msm ΔpafBC mutant strain (Supplementray Fig. S3).
Deletion of pafBC affects the cellular levels of proteins involved in DNA repair pathways, transcription and recombination. In order to gain insights into the cellular function of PafBC, a quantitative proteomic analysis of the Msm ΔpafBC mutant in comparison to the wild type strain was performed using iTRAQ. We identified a total of 2554 proteins (corresponding to 38.2% of all annotated Msm proteins) in three biological replicates deriving from Msm wild type and ΔpafBC cultures grown to late exponential phase. In Msm ΔpafBC 54 proteins were found to be down-regulated and 17 proteins to be up-regulated at least 1.5-fold compared to Msm wild type (Table 1). Among the affected proteins, 11 can be assigned to DNA-related functions, The pafABC operon. Co-transcription of pafABC in Msm was confirmed by isolating mRNA from the wild type strain and a subsequent reverse transcriptase reaction. cDNA was analysed by PCR using oligonucleotides that resulted in the amplification of products spanning either pafA and pafB (AB) or pafB and pafC (BC; as indicated by black bars). PCR reactions using Msm genomic DNA (gD) and mRNA that was not subjected to reverse transcription (−) served as positive and negative control, respectively. (B) Analytical gel filtration of PafB and PafC separately (upper panel) and co-expressed and co-purified PafBC complex (lower panel). For each trace, 30 µM protein was incubated 20 min at 37 °C before running on a Superdex 200 10/300GL column at 1 ml/min at room temperature. Peak fractions analyzed by SDS-PAGE are shown as gel slices above their respective peaks.
namely DNA repair, recombination and transcription. The strongest decrease was observed for RecA, with a relative reduction of about 2.5 fold in the pafBC deletion mutant. Accordingly, when cell-free lysates deriving from cells grown under the conditions of the iTRAQ experiment were analysed by an anti-RecA immunoblot, the signal for RecA, in the ΔpafBC background was significantly reduced compared to either Msm wild type or the complemented deletion strain (Fig. 4). Further analysis showed that RecA levels are controlled by concomitant action of PafB and PafC. Expression of either pafB or pafC alone in Msm ΔpafBC did not affect RecA expression as compared to the mutant strain, while co-expression of pafB and pafC resulted in a significant increase of cellular RecA levels (Supplementray Fig. S4).

Down-regulation of RecA in
Msm ΔpafBC occurs at the transcriptional level. The changes in cellular levels of various proteins in Msm ΔpafBC implied a transcriptional regulation of the corresponding coding genes. It should be noted, however, that some of the down-regulated proteins found in our iTRAQ analysis were previously identified as possible or, like RecA, genuine targets of the pupylation pathway. In addition, we find that RecA accumulates in a proteasome-deficient strain of Msm (Fig. 5A, bottom blot), also indicating that RecA is degraded via the Pup-proteasome pathway in Msm. However, the significantly reduced levels of RecA in the Msm ΔpafBC mutant strain (Fig. 4, second lane) cannot be due to increased degradation of the protein, since Msm wild type and Msm ΔpafBC exhibit similar turn-over rates for a C-terminally His 6 -tagged RecA variant expressed in either of the strains (Fig. 5A, upper and middle blot). This result strongly supports the notion that transcription of recA in Msm is either directly or indirectly regulated by PafBC. Carrying out relative quantification by qRT-PCR, we found recA transcript levels decreased by about 36% in the Msm ΔpafBC mutant strain as compared to the parental strain (Fig. 5B, upper panel). In a concomitant control, we checked the transcriptional levels of another player involved in DNA damage repair, namely uvrB 34 , whose corresponding gene product was not covered by the analysed iTRAQ samples. In analogy to recA, expression uvrB decreased by 37% in Msm ΔpafBC compared to wild type (Fig. 5B, lower panel).
Next, we examined whether PafBC influences transcription of recA by binding to its promoter region. To this end, an electrophoretic mobility shift assay (EMSA) was performed using purified Msm PafBC and a DNA fragment comprising the 300 bp upstream of the recA gene. Indeed, PafBC was found to bind to the recA promoter region (Supplementray Fig. S5). However, PafBC displays a rather promiscuous binding behaviour in vitro and comparable interaction with different promoter regions used as negative controls in the assay was observed (Supplementray Fig. S5). Various attempts to increase specificity by testing different buffer conditions were not successful. Therefore, the EMSA did not allow for a final conclusion concerning the binding of PafBC to the recA promoter region.
Deletion of pafBC renders M. smegmatis more susceptible to DNA damaging conditions. Like in other prokaryotes, RecA in mycobacteria holds a key role in homologous recombination and was shown to be a crucial element in regulating the SOS response upon DNA damage 10,35,36 . Thus, the observed decrease of RecA prompted us to perform a phenotypical analysis of the Msm ΔpafBC mutant strain under DNA-damaging conditions.
In the growth experiments DNA stress was emulated by exposing the bacterial cultures to the DNA interstrand crosslinking agent mitomycin C. In both, Mtb and Msm, recA was shown to be upregulated upon exposure to mitomycin C 5 , 37,38 . In an initial approach we determined the minimal inhibitory concentration (MIC) of mitomycin C for Msm wild type to be 160 ng/ml and 4 ng/ml for the pafBC deletion strain. We subsequently performed a survival assay using 50% of the MIC for Msm wild type. The Msm wild type strain, Msm ΔpafBC and the complemented strain were grown in shaking cultures and mitomycin C was added at an OD of 0.1. After 8 h, serial dilutions of the respective cultures were spotted onto LB agar plates without mitomycin C. Msm wild type and the complemented strain were only mildly affected by mitomycin C, when compared to the controls grown in absence of the agent (Fig. 6A). In contrast, the number of viable cells in cultures of the pafBC deletion mutant decreased by about two orders of magnitude upon exposure to mitomycin C. Similar observations were made when Msm ΔpafBC was exposed to UV irradiation. After a UV dose of 25 mJ/cm 2 viability of the deletion strain decreased by approximately two orders of magnitude as compared to the Msm SMR5 parental strain (Fig. 6B).

Induction of recA under DNA-damaging conditions is impaired in Msm
ΔpafBC. The increased sensitivity of the Msm ΔpafBC mutant towards mitomycin C and UV radiation strongly implied that the induction of recA, which is usually observed under DNA damaging conditions 5 , is hampered in the Msm ΔpafBC mutant. Indeed, when we performed an anti-RecA immunoblot analysis, levels of RecA increased significantly in Msm wild type after the exposure to mitomycin C (Fig. 7). In contrast, the pafBC deletion mutant exhibited significantly impaired upregulation of RecA levels under the same DNA damaging growth conditions (Fig. 7A), which could be rescued by complementation with pafBC. In addition, we addressed the question whether induction of recA under DNA damaging conditions is accompanied by upregulation of PafBC. Therefore, Msm wild type grown in the presence of mitomycin C was analysed by an anti-PafBC Western Blot (Fig. 7B). However, no alterations in PafBC levels were observed compared to cells grown under standard conditions. PafBC affects transcriptional control of recA via the LexA/RecA-independent promoter. In contrast to other bacteria, recA is controlled by two promoters in mycobacteria: a LexA/RecA-dependent promoter (P2), which is activated as part of the classical SOS response, and a LexA/RecA-independent promoter (P1). Transcriptional activity of both, the P1 and the P2 promoter, was previously shown to be inducible by mitomycin C 6,7 . As we observed a decrease of RecA levels in Msm ΔpafBC under both standard and DNA damaging conditions, we addressed the question through which of the two promoters PafBC affects transcription of the corresponding gene. We constructed reporter strains of Msm wild type and Msm ΔpafBC that allowed us to monitor transcriptional activities of P1 and P2 independently. A DNA fragment encompassing the entire recA promoter region and the first 11 codons of recA was fused to GFP (Fig. 8A). Either P1 or P2 was subsequently mutated in order to determine the individual contributions of the two promoter sites with respect to the presence of PafBC. The constructs were introduced into the genome of the respective strain by an integrative vector.
In the Msm wild type strain reporting on the activity of the entire P1/P2 promoter region, GFP fluorescence was observed to be about 6-fold higher in presence of mitomycin C than in the absence of the agent (Fig. 8B). In the ΔpafBC strain the mitomycin C-induced response was markedly reduced and resulted only in an approximately 3-fold increase reaching less than half of the wild type fluorescence value. When P2 was inactivated, the response to mitomycin C stress could only be observed for the wild type strain. In contrast, mutation of P1 retained the response to mitomycin C in both strains, although the ΔpafBC strain displayed only about three-fourths of the wild type level. The weaker P2 response of the ΔpafBC strain may be explained by the fact that P2 activity is dependent on RecA 6 , whose level was shown is significantly reduced in the ΔpafBC strain (Figs 4A and 7A). In absence of mitomycin C, GFP fluorescence in the ΔpafBC strain was approximately half of the wild type level for the reporter constructs exhibiting an intact P1 promoter, while GFP fluorescence reached the same level under standard conditions when P1 was mutated. Together, these findings clearly demonstrate that P1-driven transcription of recA is absent in the pafBC deletion mutant while P2-driven transcription remains largely unaffected.

Discussion
The response to DNA damage in mycobacteria is regulated by two independent mechanisms. Besides the ubiquitous LexA/RecA-dependent SOS response, most DNA damage-inducible genes in mycobacterial species are controlled in a LexA/RecA-independent manner via the P1 promoter [5][6][7] . Notably, some genes, like e.g. recA, are regulated by both mechanisms 8,14 . The presented study broadens the scope of P1-driven transcription of DNA repair genes by adding PafBC as a novel regulatory player to the DNA damage response in mycobacteria. The goal of our study was to gain insights into the physiological function of pafB and pafC in mycobacteria. Since these genes form an operon with the Pup-ligase coding gene pafA (Fig. 1A) 26 in mycobacteria, they were generally considered to be associated with the Pup-proteasome gene locus. The formation of a heterodimeric complex (Fig. 1B) and the presence of a winged helix-turn-helix DNA-binding motif at the N-termini of PafB and PafC (Fig. S1) indicated a regulatory function at the transcriptional level. However, under the conditions tested we found no influence of PafBC on any of the components known to be involved in pupylation or proteasomal degradation (Fig. 3, Supplementray Fig. S3). Rather, quantitative proteome analyses revealed that levels of various proteins related to DNA damage repair were altered in a PafBC-dependent manner (Table 1). Accordingly, Msm ΔpafBC displayed an increased susceptibility to DNA damaging agents (Fig. 6). Finally, using the example of RecA, we demonstrated that PafBC exerts its influence on transcription of DNA-damage inducible genes at the level of the LexA/RecA-independent P1-promoter (Figs 5B and 8). To our knowledge, this is the first study showing a role of PafBC in DNA damage response.
Disruption of pafB or pafC in Msm was previously shown to hamper a recipient strain's ability to take up and/or integrate DNA via homologous recombination (HR) into its genome during conjugation 28 . RecA is a key element in HR, as it binds to the DNA strands and catalyzes strand exchange, which eventually results in recombination. In the iTRAQ analysis RecA displayed the highest decrease in cellular levels of those proteins that could be identified in the Msm pafBC deletion strain (Table 1). Thus, against the background of this observation the impact of disrupted pafB or pafC on conjugation can be attributed to impaired RecA-dependent HR rather than impaired DNA uptake.
Very recently, a M. smegmatis mutant that carries a transposon insertion in pafC was shown to be hypersensitive to various fluoroquinolones, like moxifloxacin or ciprofloxacin 29 . The data presented here provide an explanation for the increased sensitivity. Quinolones target topoisomerase IV and DNA gyrase and their action on these enzymes eventually results, amongst others, in multiple DNA strand breaks 39 . Thus, while Msm wild type is able to at least partially compensate the action of quinolones by a functional DNA damage repair, the impaired response of ΔpafBC to DNA damaging conditions (Figs 6 and 7) leads to increased sensitivity to these drugs.
At least under in vitro conditions PafBC binding lacks specificity as reflected by the results of our EMSA approach (Supplementray Fig. S4). It is possible that additional mechanisms might exist in vivo to bring about specificity. Many transcription factors are targeted by post-translational modifications like phosphorylation or acetylation that modulate their binding affinity 40,41 . For example, VirS, the transcription factor that regulates the mycobacterial monooxygenase (mymA) operon, is phosphorylated by the kinase PknK 42 . Phosphorylation increases the affinity of VirS for the mymA promoter. Furthermore, PafB and PafC feature a WYL domain in their C-terminal half. WYL domains were previously postulated to act in ligand sensing and binding 31 . It is therefore conceivable that PafBC relies on an additional interaction partner in order to specifically bind its genuine target, which is absent in the in vitro EMSA experiment. It could be argued that PafBC does not act as a specific transcriptional regulator but might function analogous to members of the Dps (DNA binding protein from starved cells) protein family 43   shown to bind DNA without apparent sequence specificity, thereby shielding the DNA molecule from damaging agents [44][45][46] . The fact that PafBC did not exhibit a change in cellular levels under DNA damaging conditions, makes such a mode of function very unlikely (Fig. 7B). Furthermore, most factors with wHTH motifs interact with specific DNA sequences, while DNA binding of Dps is mediated simply by the positive charge of N-terminal lysine residues 47,48 .
The involvement of PafBC in regulating the DNA damage response in Msm is clearly supported by our data. However, it remains elusive how the signal of DNA damage is eventually translated into PafBC-dependent transcriptional control. Regulation of DNA repair-related genes by PafBC could occur directly by binding to the P1 promoter, or indirectly by e.g. affecting the transcription of other regulators. Recently, the clp gene regulator ClgR was shown to recognize the LexA/RecA-independent P1 promoter 15 . Analogous to our observation with the Msm ΔpafBC mutant, upregulation of recA and other genes involved in DNA repair is impaired in an Msm ΔclgR mutant upon exposure to DNA damaging conditions 15 . The coding genes of 15 proteins that displayed altered levels in Msm upon deletion of pafBC either harbour the P1 consensus motif recognized by ClgR in the promoter region or are part of operons which bear the motif in their promoter region (Table S1). Based on these observations different scenarios regarding P1-driven transcription are conceivable: i) PafBC and ClgR act independently on the P1 promoter, but share certain targets, like e.g. recA, which are regulated by one or the other protein, depending on distinct cellular signals and/or additional transcriptional factors ii) transcription is regulated by concerted binding to the P1 promoter by both, PafBC and ClgR iii) the impact of PafBC could be indirect by affecting transcriptional activity of clgR.
The function of PafBC is apparently not restricted to DNA damage. This is clearly indicated by the large number of proteins (about 80% of the proteins identified in our iTRAQ analysis) whose expression is influenced Values are shown as mean ± SD and represent three biological replicates measured in triplicates. An unpaired two-tailed Student's t-test was applied for statistical analysis; ****p value < 0.001.
by PafBC independent of the P1 promoter and that are not related to DNA repair (Table 1).This points towards a more global impact of PafBC in the cell's regulatory networks. In accordance with an earlier study we did not find any influence of PafBC on pupylation or proteasomal degradation at least under the conditions tested ( Fig. 3 and Supplementray Fig. S3) 26 . Nevertheless, the organization of pafB and pafC in an operon with pafA implies functional coupling and therefore a role for PafBC in the Pup-proteasome system (PPS). Interestingly, conditional expression of ClgR in Mtb results in the induction of prcB and prcA 49 . Given the possible link of PafBC and ClgR indicated above, it is therefore tempting to speculate that these two regulators together might  modulate levels of the PPS upon a particular cellular signal. In summary, our study assigns the first known function to the genes pafB and pafC in Msm. The elucidation of the function of PafBC as a heterodimeric transcriptional regulator involved in the LexA/RecA-independent mycobacterial DNA damage response adds a new facet to the intricate system of DNA damage control in mycobacteria. Yet, the co-occurrence of pafBC with the Pup-proteasome gene locus remains elusive. This association could have evolved since both systems -DNA repair and Pup-proteasome-dependent protein degradation -might play a role under specific stress conditions and are therefore regulated by overlapping regulatory mechanisms. Whether PafBC and the Pup-proteasome system indeed share also a more direct functional link remains to be further scrutinized.

Methods
Bacterial strains, oligonucleotides, media and growth conditions. The oligonucleotides used in this study are summarized in table S2. Strains used were Mycobacterium smegmatis (Msm) SMR5, Msm ΔpafBC and the complemented strain Msm ΔpafBC-pafBC (see following section). Bacteria were grown in LB medium supplemented with 0.05% (v/v) TWEEN-80 (LB-T) or 7H9 medium (Difco Laboratories) without OADC supplement. When appropriate, hygromycin B was added to a final concentration of 100 µg ml −1 . Growth curves of Msm SMR5, Msm ΔpafBC and Msm ΔpafBC-pafBC, were measured in 24-well plates in a total volume of 0.5 ml. OD 600 measurements were carried out in a Synergy H1 Hybrid Multi-Mode Microplate Reader (Biotek). For proteomic analyses, strains were grown in 7H9 medium without OADC. Cells were routinely grown at 37 °C in a shaking incubator at 140 rpm.

Disruption of pafBC in M. smegmatis.
To generate a markerless deletion of pafBC in Msm the knock-out plasmid pMCS-hyg-rpsL-ΔpafBC was generated as follows: a 1.5 kbp NcoI/HindIII fragment encompassing the 5' pafBC flanking sequence and a 1.5 kbp HindIII/NdeI fragment encompassing the 3' flanking sequence were amplified from Msm chromosomal DNA. The PCR products were cut with the respective restriction enzymes and were cloned into the vector pMCS-hyg-rpsL. Competent cells of streptomycin-resistant Msm mc 2 155 SMR5 were transformed with the knock-out plasmid and the rpsL counterselection strategy 32,33 was applied to delete a 1.9 kbp fragment comprising pafB and pafC. Deletion of the genes was confirmed by PCR and Southern blot analysis, using a pafA specific 0.5 kbp probe. Msm ΔpafBC was complemented with the integrative plasmid pMV361-hyg-amp containing the genes encoding wild type PafB and PafC fused to 300 bp upstream sequence of the pafABC operon, encompassing the native promoter.
Preparation of cell-free lysates and immunoblot analyses. Cells were harvested by centrifugation and were washed once with 1 Vol. PBS supplemented with 0.05% (v/v) TWEEN-80. Pellets were then resuspended in lysis buffer containing 100 mM Tris-HCl pH 8, 50 mM NaCl, 1 mM EDTA, 1 mM DTT, 1 mM MgCl 2 , 1 mM PMSF, supplemented with 1x protease inhibitor cocktail (Sigma-Aldrich), DNase and RNase. Cells were lysed using a FastPrep-24 Bead Beater (MP Biomedicals; 2 × 30 s at 6 m/s, 0.15mm zirconia beads). Cell debris and insoluble material was removed by centrifugation (16,000 × g, 5 min, 4 °C). Protein concentration of cell-free lysates was determined by the Bradford assay. Lysates were routinely separated by SDS-PAGE and were subjected to immunoblot analyses, using polyclonal anti-PafBC antiserum (1:1,000) raised against purified Msm PafBC (Eurogentec), anti-Pup antiserum (1:10,000) 19 or anti-His 6 -tag antibody (1:10,000; Bethyl Laboratory, Montgomery, TX, USA). For detection a goat anti-rabbit IgG alkaline phosphatase antibody-horseradish peroxidase (HRP) (BioRad) or a rabbit anti-mouse IgG alkaline phosphatase antibody-HRP (Promega) was used together with Amersham ECL Prime Western Blotting Detection Reagent (1:1 ratio; GE Healthcare Life Sciences) as a substrate for the horseradish peroxidase. To control for equal loading, membranes were stripped and control immunoblots were performed using a monoclonal antibody that was raised against the RNA polymerase subunit β (RpoB) of E. coli (1:7,500; GeneTex). For immunoblotting of RecA, cells were harvested by centrifugation and were washed once with 700 µl PBS. Pellets were then resuspended in lysis buffer (PBS supplemented with 1 mM PMSF, 1x c0mplete (Roche) protease inhibitor), bead beaten and centrifuged as above. Lysates were separated by SDS-PAGE and were subjected to immunoblot analyses, using commercially available anti-RecA antibody (1:1,000, MBL International, clone ARM414) raised in mouse and commercially available anti-RpoB antibody (1:1,000, BioLegend, clone 8RB13) raised in mouse. For detection, HRP-conjugated anti-mouse IgG antibody (1:250,000, abcam #ab6789) together with Amersham ECL Select Western Blotting Detection Reagent (GE Healthcare Life Sciences) was used.
Determination of the minimal inhibitory concentration (MIC). MICs were determined as described 50 . Briefly, cultures of Msm SMR5 and Msm ΔpafBC were grown to mid-log phase in LB-T. Cultures were diluted to an OD 600 of 0.01 and aliquots of 50 µl were transferred to 96-well plates containing two-fold serial dilutions of mitomycin C, resulting in a final concentration ranging from 500 to 1 ng ml −1 . Each strain was tested in triplicates in two independent experiments. Plates were incubated at 37 °C. The MIC was defined as the mitomycin C concentration at which no visible growth was observed after three days of incubation.
Assessing survival after mitomycin C and UV exposure. Msm SMR5, Msm ΔpafBC (both carrying the empty plasmid pMV361-hyg-amp) and the complemented strain Msm ΔpafBC-pafBC were grown in LB-T to OD 600 of 1.0 and were subsequently diluted to an OD 600 of 0.1. Cultures were supplemented with 80 ng ml −1 mitomycin C. Controls were grown in the absence of mitomycin C. After 8 h of incubation, serial dilutions of the cultures were spotted onto LB agar containing 100 µg ml −1 hygromycin B. Plates were incubated at 37 °C for 3 days and survival was assessed by comparing control samples to the samples deriving from mitomycin C-treated cultures.
For the UV irradiation assay, serial dilutions of freshly grown cultures (OD 600 = 1.0) were spotted onto LB agar plates containing 100 µg ml −1 hygromycin B. Plates were exposed to a UV dose of 25 mJ/cm 2 using a Stratalinker UV Crosslinker 2400 (Stratagene). Control samples were not exposed to UV. Plates were incubated for 3 days at 37 °C. Survival was assessed by comparing colonies grown on the respective treated and untreated plates.
GFP-reporter assay to monitor recA P1 and P2 transcriptional activity. To monitor activities of the recA P1 and P2 promoter, a 419 bp DNA fragment, comprising 386 bp upstream sequence and the first 11 codons of recA was fused to the GFP gene and was inserted in a pMyNT-derived integrative plasmid. pMyNT was kindly provided by A. Geerlof (EMBL Hamburg). To assess the individual activity of each promoter, the P1 or P2 promoter motif was replaced with a stretch of the coding sequence of mouse histone H2Ab2 of identical size and similar GC content. The vectors were transformed into Msm SMR5 and Msm ΔpafBC. Three individual colonies of each reporter strain were grown in 7H9 supplemented with hygromycin B (50 µg ml −1 ) to an OD 600 of 0.7 at 37 °C. Cultures were exposed to DNA-damaging stress by the addition of mitomycin C (80 ng/ml) and were grown for an additional 4 h. Control cultures were grown in the absence of mitomycin C. Cells were harvested by centrifugation and cell-free lysates were prepared by resuspension of the pellet in 700 µl ice-cold lysis buffer (PBS supplemented with 1 mM PMSF, 1x c0mplete EDTA-free (Roche) protease inhibitor, 1 mM EDTA) and bead beating as described above for the RecA immunoblots. Protein content was determined by the Bradford assay. GFP fluorescence was measured in a black 96-well half-area plate (30 µl lysate/well) with a Synergy2 microplate reader (BioTek) instrument (excitation filter 485/20, emission filter 528/20).
RecA degradation assay. The gene encoding an N-terminally His 6 -tagged variant of Msm recA was cloned into the pMyC plasmid under the control of an acetamide-inducible promoter. Msm SMR5, Msm ΔpafBC and Msm ΔprcB carrying pMyC-His 6 -recA were grown in 7H9 medium to an OD 600 of 1.0. Acetamide (final concentration 0.02% w/v) was added to the cultures to induce recA expression for 1 h. Cells were washed three times with PBS supplemented with 0.05% TWEEN-80 (v/v). Then cells were resuspended in 7H9 medium without acetamide and incubation of cultures continued for 9 h. Samples were withdrawn at the given time points. Cell-free lysates were prepared as described above. Equal protein concentrations were loaded and separated by SDS-PAGE for subsequent immunoblot analysis with an anti-His 6 -tag antibody in order to monitor His 6 -RecA degradation.
RNA isolation and quantitative real-time PCR. For RNA isolation strains were grown to an OD 600 of 2.6 in 7H9 medium without OADC. Cells were pelleted by centrifugation and were resuspended in an equal volume of Trizol-reagent (Life Technologies). The samples were subsequently lysed by bead beating using Lysing MatrixB tubes and a Fast Prep-24 device (MP Biomedicals). Supernatants were used for isolation of RNA with the RNA mini kit (Bio & Sell) according to the manufacturer's instructions. DNA contaminations were removed by treating isolated RNA with DNase I (Thermo Fisher Scientifc). Total RNA concentration was determined