Redox signaling through zinc activates the radiation response in Deinococcus bacteria

Deinococcus bacteria are extremely resistant to radiation and other DNA damage- and oxidative stress-generating conditions. An efficient SOS-independent response mechanism inducing expression of several DNA repair genes is essential for this resistance, and is controlled by metalloprotease IrrE that cleaves and inactivates transcriptional repressor DdrO. Here, we identify the molecular signaling mechanism that triggers DdrO cleavage. We show that reactive oxygen species (ROS) stimulate the zinc-dependent metalloprotease activity of IrrE in Deinococcus. Sudden exposure of Deinococcus to zinc excess also rapidly induces DdrO cleavage, but is not accompanied by ROS production and DNA damage. Further, oxidative treatment leads to an increase of intracellular free zinc, indicating that IrrE activity is very likely stimulated directly by elevated levels of available zinc ions. We conclude that radiation and oxidative stress induce changes in redox homeostasis that result in IrrE activation by zinc in Deinococcus. We propose that a part of the zinc pool coordinated with cysteine thiolates is released due to their oxidation. Predicted regulation systems involving IrrE- and DdrO-like proteins are present in many bacteria, including pathogens, suggesting that such a redox signaling pathway including zinc as a second messenger is widespread and participates in various stress responses.

For all organisms it is important to respond efficiently to changes in environmental constraints leading to stress conditions within cells. In case of DNA damage-generating stress, many bacteria use the SOS response to induce expression of DNA repair genes. The molecular signal inducing the SOS response is the presence of singlestranded DNA 1 . RecA filaments formed on single-stranded DNA stimulate self-cleavage and inactivation of LexA, the repressor of the SOS response. Not all bacteria use the SOS response to induce DNA repair genes 2 . In Deinococcus bacteria, induction of recA and other DNA repair genes occurs in a RecA/LexA-independent manner and in the present work we aim to identify the intracellular molecular signal that induces this SOSindependent response.
Deinococcus species are famous for their extreme tolerance to DNA damage-and oxidative stress-generating conditions such as radiation and desiccation, and for their capacity to repair massive DNA damage [3][4][5][6] . This extreme tolerance results from a combination of multiple factors and mechanisms 7 , including limitation of oxidative protein damage 8-10 that has been correlated with a high intracellular Mn 2+ /Fe 2+ ratio (e.g. 0.24 in Deinococcus radiodurans versus < 0.01 in radiation-sensitive Escherichia coli) 11 and the accumulation of small antioxidant Mn 2+ -containing complexes (with intracellular Mn 2+ concentrations ranging from 0.2 to 2 mM, 70% of which not bound to proteins) 12 . An efficient radiation/desiccation response (RDR) mechanism for induced expression of a set of genes forming the RDR regulon is also essential for radiation tolerance [13][14][15][16] . The predicted RDR regulon in different Deinococcus species consists of about 20 genes, including classical DNA repair genes (e.g. recA, uvrB) and novel genes more specific to Deinococcus (e.g. ddrB, ddrO) 17 . We previously demonstrated that RDR regulon expression is controlled by two proteins highly conserved in Deinococcus species: IrrE and DdrO 14 .
IrrE, also called PprI, is a COG2856 domain-containing metalloprotease that induces expression of the RDR genes by cleaving and inactivating DdrO, the transcriptional repressor of the RDR regulon 14,[17][18][19][20] . DdrO has a helix-turn-helix (HTH) motif-containing DNA-binding domain of the XRE family 19 . The HTH_XRE-type domain is present in many transcriptional regulators and has been named after the PBSX repressor protein Xre from the defective Bacillus prophage PBSX 21 . DdrO cleavage has been demonstrated when IrrE and DdrO are co-expressed from plasmids in E. coli without exogenous stress, and in vitro with both proteins purified from E. coli. In Deinococcus itself, however, the cleavage is somehow induced when bacteria are exposed to radiation 14 . The N-terminal domain of IrrE is structurally similar to zinc metalloprotease thermolysin 22 , and includes the active site/metal ion binding motif HEXXH that is essential for protease activity 14 . No metal ion was observed in the crystal structure of IrrE, but briefly soaking apo-IrrE crystals in a Zn 2+ -containing solution resulted in binding of Zn 2+ in the expected site, accompanied by re-orientation of the side chains of two active site residues towards the Zn 2+ ion 22 . Addition of Zn 2+ to metal-free IrrE was sufficient to restore DdrO cleavage in vitro 17 . The in vitro protease activity of IrrE could also be restored after addition of Mn 2+ or Fe 2+ ions to metal-free enzyme 17 . Such in vitro re-activation not only by Zn 2+ but also by some other metals is commonly found for zinc metalloproteases, including thermolysin 23,24 . Besides its protease domain, the crystal structure of IrrE revealed an additional domain with structural similarity to GAF domains 22 , which may have a role in signaling or in protein-protein interactions 25,26 . Interestingly, predicted but largely uncharacterized COG2856/XRE protein pairs have been identified in many other bacterial genera, including pathogens and bacteria used in biotechnological industry 14,19,[27][28][29] , suggesting that stress response mechanisms involving such protein pairs are more widespread than currently recognized.
The molecular signaling mechanism by which radiation triggers cleavage of DdrO by the constitutively expressed IrrE in Deinococcus is unknown. Exposure of the cells to radiation or desiccation causes redox imbalance through formation of reactive oxygen species (ROS) 6 . ROS can react with and damage any molecule in the cell, including DNA. In turn, DNA damage can trigger ROS production 30 . Besides causing damage, ROS and other reactive species may also activate proteins through oxidative post-translational modification of cysteine residues, as exemplified by chaperone Hsp33 and transcription factor OxyR in E. coli and several other bacteria 31,32 . Here, we investigated if ROS are involved in IrrE activation, either by direct redox modification of IrrE or indirectly by generating another molecular signal acting downstream the stress-induced ROS and redox imbalance. We show that IrrE is a zinc-dependent enzyme in vivo in Deinococcus, that induction of IrrE protease activity still occurs when IrrE lacks its GAF-like domain, and that IrrE is activated directly by increased intracellular levels of free zinc ions following a zinc shock. Moreover, an increase in free zinc and IrrE protease activation is also observed after oxidative treatment of the cells. We propose that induction of the RDR regulon by radiation and oxidative stress involves a zinc signal generated by impaired redox homeostasis. Such role for zinc as a second messenger in regulation of cellular enzyme activities has been described previously for eukaryotes [33][34][35] , but to our knowledge this is the first time for prokaryotes.

Results
DdrO cleavage in Deinococcus is stimulated by oxidative treatments. Cleavage of RDR regulon repressor DdrO (129 aa) by metalloprotease IrrE (281 aa) in D. deserti has been observed after exposure of the cells to gamma or UV radiation 14 . The cleavage occurs between L106 and R107 of DdrO. Here we show that this cleavage is also induced after a short exposure to mitomycin C (MMC) ( Fig. 1; Supplementary Fig. S1). MMC is known in particular as an alkylating agent that generates interstrand DNA cross-links, but the metabolism of this quinone-containing antibiotic in the cell also results in oxidative damage to other macromolecules via the generation of various ROS 6,36-38 . The antioxidant compound thiourea scavenges hydroxyl and superoxide radicals and hydrogen peroxide (H 2 O 2 ) 39, 40 , and has been found to inhibit the effects of the MMC-induced ROS production in E. coli 36 . Here we observed that the MMC-induced DdrO cleavage in D. deserti was decreased in the presence of thiourea ( Fig. 1; Supplementary Fig. S1), suggesting the involvement of MMC-induced ROS in DdrO cleavage stimulation. Another antioxidant, TEMPOL, which catalyzes superoxide dismutation, facilitates catalase-like metabolism of H 2 O 2 and limits hydroxyl radical formation [41][42][43] , also inhibited DdrO cleavage in cells exposed to MMC (Fig. 1). When the cells were directly exposed to H 2 O 2 for 10 min, DdrO cleavage was also efficiently induced, which was inhibited by thiourea and TEMPOL ( Fig. 1), further confirming the participation of redox signaling in IrrE activation.
We first presumed that stress-generated ROS directly oxidize IrrE. The sulfur-containing amino acids methionine (Met) and cysteine (Cys) are particularly sensitive to oxidation, and fulfill a key role in redox signaling. Thus for some proteins oxidation of these residues leads to their activation (e.g. HypT, OxyR, Hsp33) 44 . Met and Cys residues highly conserved in deinococcal IrrE proteins were mutated in D. deserti IrrE. None of the mutations affected radiation resistance and DdrO cleavage in D. deserti ( Supplementary Fig. S2), excluding a role of these residues in IrrE activation upon stress exposure. This is in line with the previously reported in vitro cleavage of DdrO by IrrE, using purified proteins for which any modification (e.g. oxidized residues) was not detected by mass spectrometry 14,17 . Together, these data suggest that ROS stimulate IrrE activity in an indirect manner.

IrrE is a zinc-dependent metalloprotease in Deinococcus.
DdrO cleavage by IrrE is not only stimulated by radiation and other stresses that induce ROS production, but also rapidly (within 5-10 min) and efficiently after exposure of Deinococcus cultures to an excess of zinc ions, but not of other metal ions including Mn 2+ and Fe 2+17 , as confirmed here ( Supplementary Fig. S1). The stimulation of IrrE protease activity by the zinc shock might be direct, if IrrE activity indeed requires Zn 2+ as co-factor in vivo, or indirect, if the zinc shock, in common with radiation, causes ROS formation and DNA damage. To better understand the zinc shock-induced activation mechanism, we first investigated the metal dependency of the IrrE activity in vivo by applying different treatments in the presence of TPEN or dipyridyl, which are membrane-permeable chelators with high affinity for zinc (Zn 2+ ) and iron (Fe 2+ ) ions, respectively. Dipyridyl (DIP) was included in these tests because of oxidative stress through ROS formation often involves Fe 2+ . DIP (500 μM) inhibited DdrO cleavage only in H 2 O 2 -exposed cells ( Fig. 2; Supplementary Fig. S3), showing that the observed cleavage stimulated by H 2 O 2 , but not by UV or MMC, requires Fe 2+ , most likely via the Fenton reaction where H 2 O 2 reacts with Fe 2+ generating oxygen-radical species. TPEN (25 to 50 μM) totally blocked DdrO cleavage in D. deserti and D. radiodurans exposed to each of the applied treatments ( Fig. 2; Supplementary Fig. S3). Additional experiments showed that adding 10 μM TPEN to the culture was sufficient to inhibit stress-induced DdrO cleavage ( Supplementary Fig. S3 45 . Together, the data strongly support that the in vivo protease activity of IrrE is dependent on zinc ions. The zinc shock-induced cleavage is not accompanied by ROS production and DNA damage. Next, we analyzed whether the zinc shock-induced cleavage occurs indirectly via generation of ROS and/or DNA damage. We have previously shown that exposure of D. deserti to radiation results in an increase of point mutations as detected by an increased number of rifampicin resistant mutants 46 . Such mutations can arise through direct DNA damage or after incorporation of oxidized nucleotides. Efficient cleavage of DdrO was observed when D. deserti was exposed to either zinc shock or H 2 O 2 treatment. However, in contrast to H 2 O 2 exposure, zinc shock was not accompanied by induced mutagenesis (Fig. 3a). In addition, ROS were not detected following the zinc shock when a fluorescent ROS detector was used ( Supplementary Fig. S4). These results indicate that DdrO cleavage by IrrE after the zinc shock is induced directly by the sudden increase of the Zn 2+ concentration, and not indirectly via the generation of ROS and DNA damage. An increase in intracellular free Zn 2+ following zinc shock was indeed observed using Fluozin-3 AM, a membrane-permeable molecule that   Fig. S5). The direct activation by Zn 2+ is further supported by the observation that the antioxidants thiourea and TEMPOL did not inhibit the zinc-induced DdrO cleavage (Fig. 3b). Moreover, also dipyridyl did not inhibit the zinc-induced cleavage, showing that this induction did not occur indirectly via a step that involves Fe 2+ (Fig. 3b).
The direct stimulation of IrrE-mediated DdrO cleavage by zinc ions suggests that radiation/redox-induced cleavage may also involve an increase in free zinc ions. A ROS-induced increase in intracellular free zinc ions is strongly supported by several previous studies. Indeed, many oxidants and other reactive species can cause rapid release of zinc ions from redox-sensitive cysteine-containing zinc sites such as zinc fingers and related structures in proteins, which allows novel interactions between zinc ions and other proteins 31,33,34,[47][48][49] . Therefore, we hypothesized that oxidative treatments leading to DdrO cleavage, such as exposure to H 2 O 2 , could modify the free zinc ion content in Deinococcus. Using Fluozin-3 AM we showed that H 2 O 2 concentrations in the range of 10 to 30 mM indeed lead to substantial increases in the level of free zinc ions (Fig. 4). These data strongly support a role for zinc in oxidative-stress-induced activation of IrrE.
Identification of Deinococcus deserti proteins with zinc/cysteine sites. This section describes possible sources of ROS-induced zinc mobilization in D. deserti. In addition to zinc/cysteine (Zn/Cys) sites in proteins, zinc ions can be bound by bacillithiol (BSH), a low-molecular-weight thiol produced by Deinococcus and several other bacteria, with oxidation of BSH (to BSSB) causing release of the zinc 50 . The zinc bound to BSH could thus be an important source for IrrE in oxidative conditions. However, with respect to DdrO cleavage upon stress and DdrO re-accumulation during recovery, no difference was observed between wild-type and BSH biosynthesis deficient mutant strains of D. deserti ( Supplementary Fig. S6), showing that BSH is not crucial for regulating IrrE activity.
Release of zinc ions from Zn/Cys sites after oxidation or other modifications of the Zn 2+ -coordinating thiolates has been reported for several bacterial proteins 49,51 . Numerous proteins that bind zinc through Cys residues, or through a combination of Cys and histidine (His) residues, have been described or predicted. The zinc-binding residues are generally present as two pairs of closely spaced Cys/His (e.g. the most frequently found CXXC motif, but also others such as CXC, CXXXC and CXXH) separated by a longer spacer within the protein sequence 35,52 . To find D. deserti proteins that could be potential sources of oxidative stress-induced increase of intracellular free zinc, we searched its genome for proteins containing Zn/Cys sites. We found that it encodes homologs of at least 62 proteins harboring Zn/Cys sites (Supplementary Table S1), with a large fraction (27%) consisting of proteins involved in DNA replication and repair (Table 1). Similar Cys motifs and thus potential Zn/Cys sites are present in more than 30 additional D. deserti proteins, including uncharacterized small proteins containing four or more cysteine residues (e.g. the radiation-induced 74-residue-long DdrS with two CXXC motifs) (Supplementary Table S1). In a ddrS mutant, no difference was observed for DdrO cleavage induction upon stress when compared to the wild-type strain (Supplementary Fig. S6). Taking into consideration all these data, we presume that zinc ions are released from various different proteins following disturbed redox homeostasis. Indeed, the list of identified Zn/Cys site-containing D. deserti proteins (Supplementary Table S1) contains homologs of at least The GAF-like domain of IrrE is not essential to induce DdrO cleavage. The results described above indicate that IrrE protease activity is stimulated by increased availability of zinc ions following exposure to radiation and other agents that cause ROS production and redox imbalance. Then what could be the role of the C-terminal GAF-like domain of IrrE? GAF domains may bind small signaling molecules or may have a role in protein-protein interactions 25,26 . IrrE/DdrO-related systems are found in other bacteria, but a GAF-like domain is not present in all IrrE-like proteins 19 . Therefore, it is interesting to investigate whether or not this domain plays a role in relation with zinc signaling. When a truncated IrrE lacking this domain was co-expressed with DdrO in E. coli, DdrO cleavage was observed (Fig. 5a), showing that the GAF-like domain is not strictly necessary for the cleavage by the N-terminal zinc peptidase domain. Nevertheless, the cleavage was clearly less efficient than with entire IrrE. The truncated IrrE could not be purified from E. coli, indicating that the C-terminal domain has at least a structural role important for protein stability, and possibly contributes in the interaction with DdrO, as suggested by modeling of the IrrE-DdrO complex ( Supplementary Fig. S7) 19 . IrrE lacking its C-terminal domain was also expressed, under control of the irrE promoter, in a D. deserti irrE deletion mutant. Strikingly, some DdrO cleavage was observed after exposure of the cells to zinc shock, showing that the GAF-like domain is not absolutely required for the induction of IrrE protease activity in D. deserti ( Fig. 5b; Supplementary Fig. S8).

Discussion
The deinococcal irrE and ddrO genes have been described for the first time more than 15 years ago, with irrE being required for radiation resistance and radiation-induced expression of recA 53 , and ddrO as one of the genes highly induced after radiation or desiccation 16 . The precise function of IrrE and DdrO remained unknown for many years until we showed that IrrE and DdrO function together, which led to the description of a novel stress response mechanism where metalloprotease IrrE cleaves and inactivates transcriptional repressor DdrO when the cells are exposed to radiation 14 , thereby inducing expression of RDR regulon genes (e.g. recA and other DNA repair genes) 17 . Very recently we showed that DdrO consists of two domains: the predicted N-terminal DNA-binding domain, and a C-terminal dimerization domain showing a new fold 19 . Cleavage of this C-terminal domain by IrrE abolishes dimerization and DNA binding of DdrO. Insight in the protease activity of IrrE in Deinococcus, and its stimulation, was obtained in the present study, offering a more complete view of the regulatory mechanism (Fig. 6). We showed that DdrO cleavage by IrrE in Deinococcus cells is dependent on zinc ions, and, after having investigated several possible activation mechanisms, that this in vivo activity is stimulated directly by an increase in available zinc ions very likely resulting from modification in the cell redox status. IrrE-mediated DdrO cleavage in Deinococcus is induced after exposure to radiation, desiccation, mitomycin C, H 2 O 2 and zinc shock, indicating that these conditions share a common intracellular signal stimulating the cleavage. Radiation, desiccation, mitomycin C and H 2 O 2 have in common that they induce DNA damage and oxidative stress 6 . The latter may involve formation of oxygen-radical species through the Fe 2+ -catalyzed Fenton reaction. However, the zinc shock-induced DdrO cleavage was not accompanied by DNA damage or ROS production and was not inhibited by an Fe 2+ chelator.
Many proteins and enzymes bind Zn 2+ , this ion usually having a catalytic or structural role. At catalytic sites, such as in IrrE, the zinc ion is often bound to histidine, glutamic acid and aspartic acid residues. At structural sites, the Zn 2+ is often bound to the thiolates of four cysteines (e.g. two pairs of the CXXC motif) or to a  49,56 . These proteins have been studied because of their role in sensing and responding to oxidative stress. Their Zn/Cys centers function as redox switches, and oxidation/reduction regulates the activity of these proteins. For example, under non-stress condition the four conserved Cys residues of Hsp33 bind a zinc ion and Hsp33 is inactive.  Supplementary Fig. S9). A third independent experiment is shown in Supplementary Fig. S8. P, 20 ng purified DdrO.  31,59 .
Besides the redox-induced modification of the activity of these proteins, the released zinc may participate in other pathways. Of note, Hsp33 proteins are highly conserved in Deinococcus 7 . Although there are lower levels of Fe 2+ -catalyzed oxidative protein damage in Deinococcus compared to radiation-sensitive species 9 , it can be assumed that proteins like Hsp33 function in the same way in Deinococcus as in other bacteria, that is, involving oxidation of Cys residues (with concomitant release of zinc ions). Many other Deinococcus proteins, such as FUR family members and Trx2 (see Results), may release zinc upon oxidation of their Zn/Cys sites. Zinc shock and oxidative stress have thus in common that they transiently augment the intracellular concentration of free zinc ions (i.e. not bound to proteins). We propose that released zinc acts as a second messenger that signals redox imbalance and, by becoming more available to IrrE, increases the amount of active zincbound IrrE to induce the radiation response (Fig. 6). Under standard conditions, bacterial cells maintain an extremely low concentration of free zinc (less than one atom per cell), and most zinc ions are bound to proteins or other molecules 60 . Even a small increase in free zinc can thus be a very potent signal, which can be sensed by zinc-responsive proteins (e.g. IrrE in Deinococcus). A role of released zinc as second messenger in regulation of enzyme activities has been described for eukaryotic cells [33][34][35] , but such a role in induction of an oxidative stress/ DNA damage response in bacteria has, to our knowledge, not been described previously. Remarkably, many DNA replication and repair proteins have Zn/Cys sites (Table 1). It is tempting to speculate that destruction of their zinc-binding sites inhibits their function, while the mobilized zinc stimulates IrrE activity and expression of DNA repair genes.
Genes encoding IrrE/DdrO-like protein pairs consisting of a putative metalloprotease (COG2856; ImmA/IrrE family) and a transcriptional regulator (XRE family) have been identified in many bacteria and bacteriophages 14,27,29 . Only a few have been studied experimentally, which showed a stress-induced inactivation of the XRE family repressor by the COG2856 domain-containing protein, resulting in induction of prophage or transposon excision 27,28,61,62 . In each case, however, it is unclear how stress activates the COG2856 protein.
Of note, these other studied COG2856 domain proteins are smaller than IrrE and, unlike IrrE, do not contain www.nature.com/scientificreports/ an additional GAF-like domain 19 . As we observed protease activation of truncated IrrE lacking the GAF-like domain, this supports the proposed mechanism of IrrE activation and suggests that oxidative stress-induced release of zinc may also be involved in activation of GAF-lacking IrrE-like proteins. One of these metalloproteases, MpaR in Listeria monocytogenes, was found to be activated during intracellular growth of this pathogen in macrophages 61 , during which bacteria experience oxidative and nitrosative stress. Activated MpaR triggers excision of a prophage, thereby promoting virulence of L. monocytogenes 61 . The potential role of zinc signaling in MpaR activation is supported by another recent study that showed that nitric oxide produced by macrophages causes release of zinc from Zn/Cys sites of at least 15 proteins in the pathogen Salmonella enteritica 51 , including homologs of eight Zn/Cys site-containing proteins conserved in Deinococcus. The COG2856 protein Rir in Streptococcus thermophilus, a species used dairy industry, is involved in prophage induction causing cell lysis by interacting with and preventing oligomerization of repressor Crh without proteolytic cleavage of the latter, despite the presence of a predicted HEXXH zinc-binding active site in Rir 28 . In this case, it can be hypothesized that zinc binding stabilizes Rir or the Rir/Crh interaction. Taken together, a role for zinc as second messenger following stress exposure of bacteria may be more universal than currently recognized.
After the incubations, cells were collected by centrifugation for 1 min at 10,000g and directly lysed by heating for 10 min at 95 °C in Novex NuPAGE LDS Sample Buffer supplemented with Novex NuPAGE Sample Reducing Agent (Invitrogen) (volume corresponding to 100 µl sample buffer for 1 ml of culture with an OD 600 of 1) and by passing through a syringe needle, and then frozen at − 20 °C. Protein separation by SDS-PAGE (20 μl sample per lane) and immunoblotting to analyze DdrO cleavage were performed as described previously 14 . A lane with 20 ng DdrO, purified as described 17 , was included on the Western blots. The experiments were performed in triplicate (biological replicates), and representative results are shown.
Co-expression of IrrE or IrrEΔGAF with DdrO in E. coli. Two plasmids were used for co-expression in E. coli: a pET-TEV derivative encoding IrrE or IrrEΔGAF (both with N-terminal His-tag) and a pET22b derivative encoding DdrO (with C-terminal His-tag). E. coli BL21 (AI) cells freshly transformed with the two plasmids were grown at 37 °C overnight to saturation in 10 ml of LB medium containing kanamycin and ampicillin. This pre-culture was used to inoculate (start OD 600 0.05) 100 ml of LB medium with the antibiotics and grown at 37 °C with aeration. At OD 600 of 0.6-0.7, IPTG (0.1 mM final concentration) and L-arabinose (0.2% final concentration) were added and the cells were further incubated at 37 °C for 3 h. Five hundred µl of induced cells were taken, centrifuged ( www.nature.com/scientificreports/ (100 µl of sample buffer for 1 ml of culture at OD 600 1), and heated for 10 min at 95 °C. Samples (20 μl per lane) were loaded on SDS-PAGE gels (Novex NuPAGE 10% Bis-Tris Gels; Invitrogen), and migrated in 1X NuPAGE MES SDS Running Buffer (Invitrogen) for 1 h at 130 V. In vitro cleavage control was performed as described previously 14 . The proteins were visualized by staining with Imperial protein stain (Pierce).
Induced mutagenesis. D. deserti strains were grown to OD 600 0.4 and then 250 µM ZnCl 2 or 10 mM H 2 O 2 or nothing (control) was added, and cultures were incubated further for 10 min at 30 °C. One part (1 ml) of the cultures was taken to prepare samples for immunoblotting to analyze DdrO cleavage as described above. The other part of the cultures was centrifuged for 5 min at 5000 g, and the cell pellets were re-suspended in fresh growth medium and incubated for 20 h at 30 °C with shaking to permit fixation of mutations. Then 1 ml of each culture were spread in duplicate on plates with 10 µg ml −1 rifampicin. In parallel, 0.1 ml of serial dilutions 10 −6 and 10 −7 were spread in duplicate on plates without rifampicin. The entire experiment was performed in triplicate (biological replicates). Colony-forming units (CFU) were determined after incubation for 3-4 days at 30 °C. The mutation frequency was calculated by the ratio "number of Rif R CFU/ml" / "total number of CFU/ml".

Zinc measurements with fluorescent probe FluoZin-3. Experiments to detect free intracellular Zn 2+
were based on a previously described protocol 65 . D. deserti was grown to OD 600 0.4. Bacteria were resuspended in fresh TSB/10 and then incubated with 5 µM FluoZin-3, AM, cell permeant (Thermo Fisher Scientific, F24195) for 60 min at room temperature in the dark to allow complete de-esterification of intracellular acetoxymethyl (AM) esters. The bacteria were then washed three times in TSB/10. Fluorescence measurements were performed in a microplate at 28 °C and using a spectrofluorimeter (Infinite 200 Pro; Tecan). After excitation of the sample at 485 nm, fluorescence emission was recorded at 519 nm before and after addition of H 2 O 2 or ZnCl 2 .
ROS measurements with fluorescent probe H 2 DCFDA. ROS detection using H 2 DCFDA (2′,7′-Dichlorodihydrofluorescein diacetate; Sigma-Aldrich, D6883) was performed after adaptation of a previously described method 66 . D. deserti was grown to OD 600 0.4. Bacteria were washed once in 0.85% KCl and then incubated with 25 µM H 2 DCFDA for 60 min at room temperature in the dark, and then washed again in 0.85% KCl. Fluorescence measurements were performed in a microplate at 28 °C and using a spectrofluorimeter (Infinite 200 Pro; Tecan). After excitation of the sample at 480 nm, fluorescence emission was recorded at 526 nm before and after addition of H 2 O 2 or ZnCl 2 .
Construction of D. deserti gene deletion mutants. D. deserti strains in which ddrS (Deide_04721), bshA (Deide_11680) or bshC (Deide_13730) are deleted and replaced by a kanamycin resistance gene, were constructed as described previously for other deletion mutants 46 . Briefly, DNA fragments corresponding to upstream and downstream regions of the gene to be deleted, were cloned upstream and downstream of a kanamycin resistance cassette in a pUC19 derivative. Resulting plasmids, which do not replicate in D. deserti, were introduced in D. deserti, and double homologous recombination events and complete deletion of the gene of interest were verified by diagnostic PCR.

Data availability
The IrrE and DdrO genes and proteins in this study correspond to SwissProt entries C1CZ84 and C1CYP4, respectively. All data that support the findings of this study are available from the corresponding author on request.