The interplay between Mn and Fe in Deinococcus radiodurans triggers cellular protection during paraquat-induced oxidative stress

The bacterium Deinococcus radiodurans is highly resistant to several stress conditions, such as radiation. According to several reports, manganese plays a crucial role in stress protection, and a high Mn/Fe ratio is essential in this process. However, mobilization of manganese and iron, and the role of DNA-binding-proteins-under-starved-conditions during oxidative-stress remained open questions. We used synchrotron-based X-ray fluorescence imaging at nano-resolution to follow element-relocalization upon stress, and its dependency on the presence of Dps proteins, using dps knockout mutants. We show that manganese, calcium, and phosphorus are mobilized from rich-element regions that resemble electron-dense granules towards the cytosol and the cellular membrane, in a Dps-dependent way. Moreover, iron delocalizes from the septum region to the cytoplasm affecting cell division, specifically in the septum formation. These mechanisms are orchestrated by Dps1 and Dps2, which play a crucial role in metal homeostasis, and are associated with the D. radiodurans tolerance against reactive oxygen species.


Results and Discussion
Mn and Fe at the forefront of oxidative stress management. To evaluate the protective effect of manganese in D. radiodurans when subjected to oxidative stress, we used methyl viologen (MV, paraquat) at early-exponential growth to induce superoxide production, in the presence or absence of added manganese (Fig. 1A,  Supplemental Fig. 1).
After the oxidative stress stimulus, cells immediately stopped their normal growth but could recover after 30 hours (Fig. 1A). The addition of manganese before MV (Mn + MV) protected the cells against oxidative stress (Fig. 1A). Interestingly, Δdps1 strain does not feel the stress as wild-type (WT) cells do (Fig. 1B). This strain showed no growth arrest after the MV addition, and consequently, the pre-treatment with manganese had no visible effect. In fact, Δdps1 cells phenocopy the Mn + MV condition in WT cells (Fig. 1B). D. radiodurans mutant strains Δdps2 and Δdps1Δdps2 (Fig. 1C,D) show similar growth profiles under control and MV conditions, suggesting that the absence of Dps2 is sufficient to induce the observed phenotypes. Cellular growth profiles show that the cells are not able to recover to the same growth level as WT do after oxidative stress, even in the presence of extra manganese. Still, Δdps2 and Δdps1Δdps2 regain the ability to grow to higher optical densities in the presence of manganese alone, a trend observed in the WT strain but not in Δdps1 (Fig. 1B-D). The observed effect of higher optical densities in the presence of Mn has been previously reported for the D. radiodurans WT strain (e.g. 23,24 ).
To assess metal distribution during the cellular response to oxidative stress, we followed metal-intracellular localization using nano-XRF. Manganese, phosphorus, and calcium, in wild-type D. radiodurans, are co-localized in rich elemental cellular regions that resemble electron-dense granules, while iron is mostly localized in close proximity to or within the septum membrane ( Figs. 2A, 3). . Elemental areal density quantification bar (ng/mm 2 ) for each element is presented. In panel (C,D) in control conditions, the small regions where Mn is concentrated are marked with white arrows. Data presented are from three biological replicates (+/−sd). Each condition was measured by nano-XRF at least in 12 cell tetrads.
When an oxidative stress stimulus occurs, manganese is mobilized to the cytoplasm and is transported to the outer region of the membrane ( Fig. 2A, Supplemental Fig. 2), while the iron is mostly dispersed throughout the cell ( Fig. 2A). It is possible that the mobilization of manganese ultimately results in membrane protection, which would agree with the protective effect exerted by the external addition of manganese before MV addition, as shown above. This result corroborates previous studies, in which the addition of manganese before the exposure to radiation protects the cells 5 . Notably, phosphorous and calcium follow manganese redistribution ( Fig. 2A, Supplemental Fig. 2). It has been reported that manganese can form complexes with phosphate molecules (Pi) that can detoxify ROS, probably in a Mn 2+ -Pi form 10,25 . Analysis of D. radiodurans WT cells after being subjected to the addition of manganese alone showed that the regions rich is several elements, which we proposed to be electron-dense granules, tend to increase in size (Supplemental Fig. 3A), and exhibit higher amounts of Mn, P and Ca, when compared to control condition (Fig. 2, Supplemental Figs. 3, 4). So the formation of Mn 2+ -Pi complexes may not be the sole form of Mn-complexes in D. radiodurans.
In all D. radiodurans dps mutant strains, the intracellular amount of Mn is similar to that of WT, while P amounts decrease (Supplemental Fig. 4). The ability that the mutant strains Δdps1, Δdps2, and Δdps1Δdps2, have to form putative electron-dense granules is compromised since manganese and phosphorus are homogeneously distributed in the cytoplasm. However, small regions where these metals are concentrated are still visible in the cells of Δdps2 and Δdps1Δdps2 (Fig. 2C, D-white arrows). At this point, it is not possible to assess if the nature of these regions is the same as that of the WT electron-dense granules or if they constitute another type of element-containing bodies. Furthermore, nano-XRF results show that iron is concentrated in the septum region in WT cells, while it looks more widespread throughout the cell in all knockouts mutants (Fig. 2 ,Supplemental Figs. 4,5). It is important to notice that the Mn/Fe is ca. 0.8 ± 0.15 for the conditions tested and in all D.radiodurans strains, except for when external manganese was added to the cell growth media (Supplemental Fig. 3).
Therefore, we can say that the resistance and recovery capabilities of D. radiodurans to oxidative stress and the intracellular distribution of elements are dependent on the presence of Dps1 and Dps2. The absence of Dps1 renders the cells insensitive to oxidative stress as they do not interrupt their normal growth when MV is added as WT cells do (Fig. 1A,B). On the contrary, the absence of Dps2 (Δdps2 and Δdps1Δdps2) is sufficient to compromise cell growth and the ability of the cells to recover from oxidative stress or to use Mn as a protecting agent (Fig. 1C,D). Thus we next asked what the role of both proteins in this process is. Different Dps protein forms determine their cellular localization. Although Dps are known as dodecameric proteins, we previously have shown that Dps1 can be dimeric, trimeric or dodecameric according to the condition that it was subjected 21 . Based on our previous data the dimeric Dps1 is associated with DNA 21 . Here, we show that under control conditions this association is observed during the exponential growth phase of the bacteria since GFP-Dps1 fluorescence signal superimposes with the DAPI staining of DNA (Fig. 4, Supplemental Fig. 6) and that, during this growth phase, Dps1 remains mostly dimeric (Fig. 5A) 21 . Immediately after Mn or MV addition, Dps1 changes its oligomeric form from a dimer to a trimer (Fig. 5A) as also shown previously 21 , and the protein appears dispersed throughout the cell with a stronger signal in specific regions that are consistent with the electron-dense granules (Fig. 4A). This change in oligomerization state is concomitant with the phosphorylation of the protein in, at least, five different residues (Fig. 5B), but does not depend on it 21 . The phosphorylation sites www.nature.com/scientificreports www.nature.com/scientificreports/ of Dps1 are not known, but interestingly, Dps1 has a putative phospho-motif (D 145 ART 148 QVADLV) that can be targeted by the radiation pyrroloquinoline quinone (PQQ) inducible protein kinase (RqkA) (DR2518) 26 .
Our previous work shows that Dps2 is a stable dodecamer and it can store iron or manganese, but when is in the presence of equal amounts of iron and manganese, the protein is selective towards iron, incorporating approximately double amounts of iron than of manganese 21 . It is interesting to note that in WT cells grown in control conditions, the localization of Dps2 coincides with the iron distribution, mostly at the septum region (Figs. 2A, 4B). Consistently, Dps2 was primarily detected in the membrane fraction of these cells (Fig. 5C, Supplemental Fig.  8A). Upon MV addition, Dps2 and iron became homogeneously distributed throughout the cells; Mn appeared to leave the electron-dense granules and also became widespread all over the cell, and appeared in the outer part of the plasma membrane ( Fig. 2A, Supplemental Fig. 2). Upon addition of external Mn, Dps2 was no longer localized in the setptum membrane, instead was localized in the cytosol (Figs. 4B, 5C, Supplemental Fig. 8A). Thus, we propose that in vivo conditions in the presence of an excess of manganese, Dps2 can store manganese metal as well as iron. Also, it is interesting to note that the Dps2 crystal structure previously determined, showed a metal iron site coordinated by D 132 (X)D 133 (X 60 )D 193 (X)N 195 (X 5 )I 200 27 . This metal site is close to the putative Dps2 Mn-binding motif, N(X 3 )D 28 . However, from the crystal structure, the binding motif includes N(X 5 )I instead of N(X 3 )D, in which the coordination of I 200 is through the oxygen from the main chain. Thus it is plausible that Dps2 will manage an increase in cytoplasmic manganese occurring with the external addition of this metal or as a result of oxidative stress, acting then as a manganese transporter in the cell. In this sense, both iron and manganese mobilization would be mediated by Dps2. The absence of Dps2 results in the disruption of the WT patterns of distribution of both metals (Figs. 2, 3). www.nature.com/scientificreports www.nature.com/scientificreports/ Phosphorous and calcium distributions are also affected in Δdps2 and follow the manganese distribution pattern. While the formation of Mn 2+ -Pi complexes 7,8 can explain phosphorous co-distribution with manganese, the calcium profile was unexpected. If calcium distribution follows the manganese one, which depends on Dps2, then Dps2 should also influence calcium localization. Indeed, Dps2 can bind Ca 2+ 29 , which further supports this hypothesis.
A direct functional link between Dps1 and Dps2. Remarkably, the detachment of Dps2 from the membrane is dependent on the presence of Dps1, since in Δdps1 cellular extracts all Dps2 is membrane-associated, and this can be reversed by the addition of recombinant Dps1 (Fig. 5C-lanes 6-9, 11-14 Supplemental Fig. 8A). However, Dps1 is not directly responsible for Dps2 N-terminal cleavage, since adding recombinant Dps1, in the trimeric form (Dps1 T ), to the full-length dodecameric form of recombinant Dps2 (Dps2 M ) does not lead to a change in Dps2 molecular mass (Supplemental Fig. 7). Also, Dps1 phosphorylation may have a role in this process since the trimeric form of Dps1 detected in D. radiodurans cellular extracts is always phosphorylated.
The direct consequence of the above mechanism is that, when Dps1 is not present, Dps2 cannot leave the membrane, and if Dps2 is shuttling metals, such as iron and manganese, their homeostasis will be affected. Accordingly, the absence of Dps1 is sufficient to destabilize the formation or growth of the electron-dense granules, where manganese, phosphorus, and calcium are concentrated, as our data indicates (Fig. 2B). In the initial phase of WT growth (time 0), a small proportion of the Dps1 trimeric form is detected (Fig. 5A), which may be www.nature.com/scientificreports www.nature.com/scientificreports/ important to initiate the formation of the granules, through the action of cytoplasmic Dps2 (Dps2 C ). Thus, in Δdps1, these regions are much smaller than in the WT cells ( Fig. 2A,B).
While Dps1 affects Dps2 localization, the absence of Dps2 does not affect the oligomerization states of Dps1 (Supplemental Fig. 8B). However, in all the three dps knockout mutants, the correct distribution of iron and manganese to the septum and granules, respectively, is abolished. According to our model, the absence of MV-sensitivity by Δdps1 cells may be due to the exclusive presence of Dps2 M that results in widespread distribution of manganese, which would protect the cells against oxidative stress, as the addition of external manganese does in WT cells.
In summary, Dps1 and Dps2 appear to act in a concerted way in D. radiodurans to regulate manganese and iron homeostasis and act in the formation of the regions rich in elements such as manganese and phosphorus that we propose to correspond to electron-dense granules.
Dps-dependent iron localization is essential for septation. Under control conditions, cell division occurs as an orderly process, where tetrads are generated by the formation of two septa in opposite sides perpendicularly to the cell wall, forming two cells 1,30 . Moreover, manganese is concentrated in the rich elemental regions, whereas iron is localized in the membrane septum ( Fig. 2A). Interestingly, all the dps knockout mutants show a strong phenotype in cell division, as cells seem to have difficulty in performing septation to form the new tetrads (Supplemental Fig. 9). We hypothesize that the presence of Dps2 in the membranes of D. radiodurans may play a role in cell septation since its tendency to distribute to the cytoplasm correlates with the severity of the phenotype. Both Δdps2 and Δdps1Δdps2 mutants have the most pronounced phenotype. Their cells have irregular forms, and the septum formation seems to be affected. This is visible even under control conditions (Supplemental Fig. 9). Accordingly, both WT and Δdps1 strains have similar doubling times of ca. 2 hours, while Δdps2 and Δdps1Δdps2 grow slower with a doubling time of ca. twice that of WT and Δdps1 (Supplemental Fig. 10). Even though the most striking difference amongst WT and the dps knockout mutants strains is that the mutants cells tend to aggregate (Supplemental Fig. 9), they still show a tetrad-like organization in the Δdps1 mutant cells, which is gradually lost in Δdps2 and more evidently in Δdps1Δdps2. These phenotypes can merely reflect the cells' inability to accurately form the septum when Dps2 is not present or fully functional (in Δdps1, Dps2 cannot exit the membrane). Thus, in both Δdps2 and Δdps1Δdps2, cellular division occurs in a disordered manner even under control conditions. This effect is much more pronounced when cells are subjected to a stress condition, where the septum region is profoundly damaged (Fig. 2C,D, Supplemental Fig. 9). This can explain why in both Δdps2 mutants the cells never recover after being subjected to MV promoted stress condition, even when pre-treated with manganese (Fig. 1C,D). This effect is reflected in the lag-phase of the different strains (Supplemental Fig. 10). Also, both Δdps2 mutants have longer lag-phases when compared with the WT cells, while the Δdps1 mutant has a slightly shorter lag-phase. It is interesting to observe that the cell septation phenotype observed in the Δdps2 mutants is similar to the one seen in the autolysin knockout mutant from murein hydrolase of Staphylococcus aureus, (AtlA) 31 . Although autolysins from D. radiodurans have not been studied, there are at least four genes predicted for N-acetyl muramoyl-L-alanine amidases, namely DR2394, DR1387, DR2567, and DR1632. The role of iron in the cell separation process remains unclear. Nevertheless, we cannot rule out that the effect observed in the cell division is related to the damage of DNA since Dps2 protects DNA under in vitro conditions 21 . If that is the case, the inhibition of the septum formation due to DNA damage could resemble the SOS system from Escherichia coli (e.g. 32 ).
In summary, these data indicate that there is a link between Dps function, the presence of iron in the septum region, and cell septation. The association of Dps with the cell division process is only very poorly studied, although DpsA from Streptomyces coelicolor was proposed to be involved in the condensation of DNA to ensure an appropriate degree of nucleoid compaction during cell division 33 . Nevertheless, the question remains of whether Dps in D. radiodurans are also involved in the DNA compaction before cell division, DNA protection in addition to controlling Fe localization in the septum region.

electron-dense granules are multi-ion stores involved in stress response.
Our data show the presence of rich elemental regions with high concentrations of manganese, phosphorous, and calcium that we propose to correspond to the electron-dense granules in D. radiodurans. Other elements, like potassium or copper, can also be found in these regions, but to a lesser extent (Supplemental Fig. 3). This is in agreement with previous work, in which under control conditions manganese is mostly not associated with proteins 8,10 . After a stress stimulus, manganese, phosphorus, and calcium are dispersed throughout the cell and into the membrane structures ( Fig. 2A, Supplemental Fig. 2). As manganese and phosphorus migrate together, it is tempting to speculate that they probably exist in the form of Mn 2+ -Pi complexes and that these are directed to the membrane to play a role in the protection of macromolecules against oxidation, as previously proposed 9,10 . Noteworthy, no data is supporting the in vivo formation of Mn 2+ -Pi complexes in D. radiodurans, but manganese can form complexes with pyrophosphate or orthophosphate, in other organisms 9,34,35 . Also, D. radiodurans has enzymes, e.g., polyphosphate kinases (DR1939, DR0132) and exopolyphosphatase (DRA0185), capable of hydrolyzing polyphosphate 2,36 . As shown above, calcium follows the same redistribution pattern as phosphorus and manganese. As in other organisms, calcium mobilization to the membranes may serve to initiate stress-responsive signaling pathways involving kinases and should be investigated in the future. Our data do not allow to discriminate between free Ca 2+ and Ca 2+ -complexes, and thus, we cannot rule out that calcium phosphate complexes can also exist intracellularly as previously observed in mitochondria 37 .
The rich elemental regions are devoid of sulfur, while calcium, manganese, and phosphorus, are compartmentalized in specific regions within the cell ( Fig. 2A, Supplemental Fig. 3). The fact that in the dps knockout mutants sulfur does not appear as excluded from these regions suggests that the observed structures in these mutants are different from those observed in the WT cells. www.nature.com/scientificreports www.nature.com/scientificreports/ In sum, the data presented here show the presence of rich elemental regions in D. radiodurans that act as reservoirs of calcium, manganese, and phosphorus and can be mobilized upon stress.

DrDps as central players in oxidative stress response.
To summarize our findings, we propose a model (Fig. 6) showing the role of DrDps in manganese and iron homeostasis that contributes to the intracellular protection against oxidative stress of the radiation-resistant bacterium Deinococcus radiodurans. The model proposes that under control conditions this organism stores manganese in the rich elemental regions that we propose to be electron-dense granules together with other elements, namely phosphorus, and calcium, whereas iron is mostly localized in the septum region. Under these conditions, Dps1 associates with the DNA ( 21 and the data here presented), whereas Dps2 preferentially store/transport iron 21 , and localizes mostly at the septum region.
After being subjected to a stress stimulus, in our case oxidative stress promoted by MV addition, iron becomes dispersed throughout the cell, probably halting cell division. Manganese is also distributed throughout the cell, probably in the form of Mn 2+ complexes with small molecules such as phosphate, by Dps2. Moreover, manganese appears in the outer part of the membrane, where it can protect membranes from further damage. Indeed, Mn-complexes have been proposed to play a crucial role in the scavenging of ROS upon stress conditions 9,10 . Calcium follows a similar distribution and may be responsible for the activation of stress-responsive signaling pathways involving kinase activation that may affect the phosphorylation state of Dps1.
The model now presented shows how Dps, including their different proteoforms, Dps1 DNA /Dps1 P , Dps2 M / Dps2 C , control the localization and stress-dependent relocalization of manganese together with other elements (phosphorous, calcium, and iron) (Fig. 6). Dps1 regulates processes such as the formation of the rich elemental regions and the detachment of Dps2 from the membrane, while Dps2 C plays a role as a metal transporter (Fe vs. Mn). The DrDps crosstalk and the interplay between iron and manganese are critical protection players in alleviating oxidative stress damage, in D. radiodurans. As the formation of superoxide anion and hydrogen peroxide

GFP-Dps1 and GFP-Dps2 bacterial strains and transformation. D. radiodurans strains expressing
Dps proteins fused to GFP-tag were constructed by the tripartite ligation method 39 . Plasmid pFAP246 was the source for the cassette containing the GFP-tag and the resistance gene to chloramphenicol 40 . The gene sequence and the purity of the resulting tagged genes were verified by PCR and sequence analysis. Strains expressing tagged proteins were grown with aeration in TGY2x at 30 °C. Media were supplemented with chloramphenicol (3.5 µg/ ml final concentration). D. radiodurans Δdps1, Δdps2 and Δdps1Δdps2 strains were previously described 41 . Deinococcus radiodurans growths. D. radiodurans cells wild-type, Δdps1, Δdps2, Δdps1Δdps2, GFP-Dps1 and GFP-Dps2 strains were grown in M53 medium (1.0% (w/v) casein peptone, 0.5% (w/v) yeast extract, 0.5% (w/v) glucose and 0.5% (w/v) NaCl) at 30 °C. Four independent growths were performed, and cells were collected at different time points. Methyl viologen (paraquat) is a compound that increases the intracellular production of superoxide anion and thus induces an oxidative stress condition 42 . Different concentrations of methyl viologen (MV) were initially tested from 0.5 mM to 5 mM, without and in the presence of 0.5 mM manganese, which was added 2 hours before or after MV addition (Supplemental Fig. S1). Based on these initial tests, the effect of different compounds was tested, namely: 0.5 mM manganese (II) (Mn) chloride, 1 mM MV and 0.5 mM manganese (II) chloride followed after 2 hours by the addition of 1 mM methyl viologen (Mn + MV). These compounds were added to the media at an optical density (OD 600nm ) of 0.3. A control (water addition) growth was performed and followed simultaneously. The different growths were obtained from bacterial inocula in later exponential phase and reliable growing conditions, important to achieve a reproducible lag time. Cells were collected at different time points: 0 (before adding any compound -OD = 0.3), 15′ (early exponential phase), 2 hours (mid-exponential phase), and 20 hours (stationary phase) after adding the different compounds. The viability of the cells was assessed using 10 μg/ml propidium-iodide dye. The cells were visualized on a Leica DM RA2 microscope.

DrDps detection from soluble and membrane fractions. Cellular extracts from D. radiodurans
wild-type in control conditions, and after manganese, methyl viologen and Mn + MV addition at time points 0, 15' , 2 h and 20 hours were ultracentrifuged at 200,000 × g, at 4 °C. Soluble and membrane fractions were quantified using Bradford (Bio-Rad) 43 and modified Biuret methods 44 , respectively. A total of 30 µg of protein from each condition was loaded on a 12% PAGE, and the Dps1 protein bands were detected by Western-Blotting, as previously described 21 . Four biological replicates (+/−sd) were used.
Detection of Dps2 in soluble and membrane fractions in D. radiodurans wild-type, Δdps1 strains and after addition of Dps1 recombinant pure protein to the cellular extract of Δdps1 incubated during 1 h was also performed. The soluble and membrane fractions were isolated from control and manganese conditions at time 2 hours. Two biological replicates (+/−sd) were used.
Mobility shift detection of phosphorylated proteins -Mn 2+ -phos-tag Western blotting. The analyzed samples correspond to wild-type strain protein extracts, in manganese and control conditions, from time point 2 hours. Cellular extract containing 60 µg of total protein was centrifuged at 18,000 × g, 4 °C and the soluble fraction was injected on a 12% PAGE. Two gels were simultaneously run, without (control) and containing a 2 mM MnCl 2 and 100 µM Phos-tag AAL-107 solution (Wako Chemicals USA, Inc.) in the resolving gel. These two solutions lead to the formation of a di-nuclear metal complex (1,3-bis[bis(pyridin-2-ylmethyl)amino] propan-2-olato di-manganese (II) complex -Mn 2+ -Phos-tag) acting as a selective phosphate-binding tag molecule 45 . Afterwards, the Dps1 protein bands were detected by Western-Blot. Two biological replicates (+/−sd) were used. Dps1 and Dps2 cellular localization. D. radiodurans cells GFP-Dps1 and GFP-Dps2 strains in control, manganese and methyl viologen conditions were collected by centrifugation (11,000 × g, 1 min) at time points 0 (OD = 0.3) and 2 hours. Cells were resuspended in phosphate buffer saline (PBS), fixed with 3.7% paraformaldehyde for 15 min at room temperature followed by 30 min on ice, and stained with 2 µg/ml of 4,6-diamidino-2-phenylindole dihydrochloride (DAPI -Invitrogen) and 10 μg/ml of N-(3-Triethylammoniumpropyl)-4-(6-(4-(Diethylamino) Phenyl) Hexatrienyl) Pyridinium Dibromide (FM4-64 -Invitrogen). DAPI is a dye that stains nucleoid DNA with blue color (wavelengths of excitation/ emission -350/470 nm), and FM 4-64 is a dye that stains membranes with red color (excitation/emission − 515/640 nm). FITC filter (fluorescein-isothiocyanate-excitation/emission 495/517 nm) was used to visualize GFP-tag constructs. Cells were visualized on a Leica DM RA2 microscope, and the images were captured with a charge-coupled device (CCD) camera. The images were further processed using ImageJ software. Two biological replicates (+/−sd) were used. Synchrotron X-ray phase contrast and fluorescence nano-imaging. An aliquot of 1 ml from a growth of D. radiodurans cells wild type, Δdps1, Δdps2 and Δdps1Δdps2 strains in control and methyl viologen conditions were collected at time point 2 hours. Cells were fixed using 3.7% paraformaldehyde, then washed 2 times in PBS and stored frozen at −20 °C. Cells were afterwards resuspended in 25 µl of water, and then 1 µl was added onto the surface of a Si 3 N 4 membrane (Silson Ltd) with the size of 1.5 mm × 1.5 mm and the thickness of 500 nm. Cells were air dried and then mounted on the cell support. Correlative imaging of both X-ray phase contrast and X-ray fluorescence (nano-XRF) was performed on the same sample to investigate its morphology and elemental content. Experiments were performed under vacuum (~1e −7 mbar) at room temperature on the Nano-Imaging beamline ID16A-NI of the European Synchrotron Radiation Facility (ESRF, Grenoble). The X-ray excitation energy was 17 keV, and all relevant elements were detected using their K-level emission lines. A multilayer coated fixed curvature Kirkpatrick-Baez (KB) focusing mirror system 46 provides the nanofocus (~30 nm) and a very high flux of 4.1 × 10 11 ph/s from the broad bandpass (1%) at 17 keV. X-ray phase contrast imaging was firstly performed by recording magnified Fresnel projection images with an equivalent pixel size of 15 nm. Quantitative phase maps were retrieved 47 and converted to mass density (µg/mm 2 ) in all figures. Nano-XRF measurements were performed subsequently with a step size of 40 nm or 50 nm and a dwell time of 50 ms 48 . The summed spectrum recorded with a 6-element silicon drift detector (Sensortech, UK) was fitted with the open source software PyMCA 49 . The absolute calibration to the elemental areal density (ng/mm 2 ) was determined by a thin film standard (AXO Dresden GmbH). The average areal density' corresponds to the mean inside the cells of the amount per unit surface of a given element. The total intracellular amount of an element corresponds to the integral of the areal density over the cell area (surface). The total amount is equal to the average areal density multiplied by the area covered by the cell. As the size of the cells does not vary significantly, both quantities are more or less equivalent. Three biological replicates (+/−sd) were used. Each condition was measured by nano-XRF at least in 12 cell tetrads.

Interaction between
Significance statement. Re-distribution of intracellular Mn and Fe by Dps as key players for oxidative protection in Deinococcus radiodurans.