Iron-mediated degradation of ribosomes under oxidative stress is attenuated by manganese

Protein biosynthesis is fundamental to cellular life and requires the efficient functioning of the translational machinery. At the center of this machinery is the ribosome, a ribonucleoprotein complex that depends heavily on Mg 2 1 for structure. Recent work has indicated that other metal cations can substitute for Mg 2 1 , raising questions about the role different metals may play in the maintenance of the ribosome under oxidative stress conditions. Here, we assess ribosomal integrity following oxidative stress both in vitro and in cells to elucidate details of the interactions between Fe 2 1 and the ribosome and identify Mn 2 1 as a factor capable of attenuating oxidant-induced Fe 2 1 -mediated degradation of rRNA. We report that Fe 2 1 promotes degradation of all rRNA species of the yeast ribosome and that it is bound directly to RNA molecules. Furthermore, we demonstrate that Mn 2 1 competes with Fe 2 1 for rRNA-binding sites and that protection of ribosomes from Fe 2 1 -mediated rRNA hydrolysis correlates with the restoration of cell viability. Our data, therefore, suggest a relationship between these two transition metals in con-trolling ribosome stability under oxidative stress.

Protein biosynthesis is fundamental to cellular life and requires the efficient functioning of the translational machinery. At the center of this machinery is the ribosome, a ribonucleoprotein complex that depends heavily on Mg 21 for structure. Recent work has indicated that other metal cations can substitute for Mg 21 , raising questions about the role different metals may play in the maintenance of the ribosome under oxidative stress conditions. Here, we assess ribosomal integrity following oxidative stress both in vitro and in cells to elucidate details of the interactions between Fe 21 and the ribosome and identify Mn 21 as a factor capable of attenuating oxidant-induced Fe 21 -mediated degradation of rRNA. We report that Fe 21 promotes degradation of all rRNA species of the yeast ribosome and that it is bound directly to RNA molecules. Furthermore, we demonstrate that Mn 21 competes with Fe 21 for rRNA-binding sites and that protection of ribosomes from Fe 21 -mediated rRNA hydrolysis correlates with the restoration of cell viability. Our data, therefore, suggest a relationship between these two transition metals in controlling ribosome stability under oxidative stress.
All organisms require a number of metal elements in trace amounts, with manganese, iron, copper, zinc, selenium, cobalt, and molybdenum all considered essential for plants and animals, whereas larger amounts of magnesium, calcium, potassium, and sodium are also required (1). Divalent metal cations have a long-established involvement in biomolecules including stabilizing structures and participating in the active sites of enzymes operating across a vast catalytic range (2)(3)(4).
Enzymes coordinating divalent metal ions act in important processes such as DNA replication (5)(6)(7), cellular metabolism and respiration (8,9), the phosphorylation underpinning much of cellular signaling (10), the oxygen-evolving function of photosystem II (11), and the oxygen-transporting function of hemoglobin (12). Additionally, metal cations have essential stabilizing roles in structures including phospholipid bilayers (13) and the ribosome (14). Although a given binding site may exhibit a preference for a given ion, competition between and substitution of ions occurs throughout metal-coordinating biomolecules with a variety of effects (15)(16)(17)(18). Many questions remain unresolved regarding how a preferred metal cofactor is selected and whether promiscuous coordination of ions is advantageous (19).
Among biological molecules that utilize divalent metal cations, magnesium is the most common (20)(21)(22), because of a combination of its abundance and amenable chemical properties such as its small radius and lack of redox activity (7,23). Although other divalent cations may be similar enough to be coordinated in place of Mg 21 , the differing properties can impact the biomolecule to which they are bound. For example, kinases that usually utilize Mg 21 can associate with other trace metal ions but suffer a loss in efficiency (10). DNA polymerases require divalent metal cations, and most often employ Mg 21 in this role. Coordinating other metal cofactors including Mn 21 and Co 21 can increase activity, but negatively affect fidelity, and can be carcinogenic as a result (5)(6)(7).
The ribosome is a massive ribonucleoprotein complex that is particularly dependent on Mg 21 (14) to fold and maintain stability because of the negative charge of the RNA backbone, and as many as 200 Mg 21 ions can be associated with just the large subunit, coordinated in six distinct geometries (23)(24)(25). How other ions with different properties are able to promote RNA folding is not well-understood (26). Ribosomal Mg 21 can be substituted with other divalent cations, including Fe 21 or Mn 21 (27)(28)(29)(30), and the ribosome can competently mediate translation in this state (30).
The intricately folded rRNA (rRNA) of the functional ribozyme core has remained largely the same since it evolved 3-4 billion years ago (31,32). Dozens of ribosomal proteins (r-proteins) interact with rRNA residues to assemble compact, flexible, and stable subunits (33), which are abundant and long-lived in a cytoplasmic environment populated with RNases. However, in physiologically challenging environments such as nutrient scarcity or oxidative stress, the ribosome can lose stability and undergo degradation (34,35), but the mechanisms initiating this process are still not well-understood.
Recent work has drawn on the interesting observation that although evolution has conserved the ribosomal core through billions of years, the earth environment has changed dramatically over this time and with it the bioavailability of the metals on which the ribosome structure depends. Metal levels in the aqueous environment prior to abiogenesis included high concentrations of Fe 21 (36), whereas Mg 21 was less abundant than today (30,37). With the evolution of photosynthesis and the increase in molecular oxygen, much of the iron was lost from the aqueous environment through precipitation (38). Similarly, abundant Mn 21 was oxidized and precipitated out, resulting in vast amounts of both transition metals in sedimentary rocks This article contains supporting information. ‡ These authors contributed equally to this work. * For correspondence: Natalia Shcherbik, shcherna@rowan.edu. (39). Therefore, the interaction of Fe 21 with rRNAs presents a potential stabilizing interaction used by ancient ribosomes in the differing conditions of the earth when the core structure evolved, a role that was then filled by Mg 21 as Fe 21 availability decreased. Although this establishes Fe 21 and Mn 21 interactions with the ribosome as evolutionarily relevant, little is known about these associations in the present-day oxidative environment and how ribosomal stability is impacted.
As well as the drop in available iron, the oxygenation of the atmosphere produced a further reason for the translation machinery to select another cofactor. The proclivity for Fe 21 to participate in the Fenton reaction with H 2 O 2 (which produces hydroxyl radicals and Fe 31 ) is of consequence in cells under oxidative stress conditions because it can enhance oxidantinduced damage to biomolecules. In an increasingly oxygenrich environment, continuous close associations with Fe 21 would have become a dangerous proposition (30). Conversely, the higher reduction potential of manganese prevents its participation in Fenton chemistry (40). Aside from this difference, the biochemistry of these transition metal ions is linked by several similar parameters (including oxidation states, atomic radius, and preferred coordination geometries), as well as their abundant bioavailability in the environment (16,18,30).
In our previous work, we have demonstrated that Fe 21 is a factor that contributes to destabilization of ribosomes and provided evidence that it promotes chemical hydrolysis of rRNA. Based on the structural predictions of the ES7L region of the 60S subunit (41,42), we proposed that Fe 21 replaces Mg 21 at specific binding sites on a ribosome (41) and that these sites become primed to undergo rapid hydrolysis upon exposure to ROS. In agreement with this model, cellular systems that limit the amounts of either cellular oxidant species or ribosomeassociated Fe 21 act to protect ribosomal integrity (35,41,43).
Here we add detail to the understanding of divalent cation interactions with the ribosome by demonstrating that Fe 21 interacts directly with rRNA and in the presence of oxidants is sufficient to degrade all yeast rRNA molecules in reproducible patterns. Moreover, we identify Mn 21 as a factor that can protect the rRNA from these cleavage events by substituting Fe 21 at the ribosomal binding sites, which correlates with improved cell viability under oxidative stress.

ROS-induced degradation of rRNA is mediated by Fe 21 throughout the ribosome
We have demonstrated in a previous study (41) that Fe 21 has a direct role in cleaving the sugar-phosphate backbone of rRNA under oxidative stress conditions. Specifically, one prominent cleavage site was observed within the ES7L region of the 25S rRNA in the large (60S) ribosomal subunit. To further understand the roles of Fe 21 in rRNA degradation, we chose to investigate whether an iron-dependent oxidative mechanism is responsible for the fragmentation of the 5S and 5.8S rRNAs of the 60S subunit or the 18S rRNA of the 40S subunit. For these experiments, we employed a strain deleted for the gene encoding the mitochondrial glutaredoxin Grx5, loss of which results in increased endogenous levels of labile iron in the cytoplasm (44,45) and has proved to be a useful system for modeling the effects of increased Fe 21 availability in cells. In our previous work, we observed that all rRNAs are unstable in a grx5D mutant (41), but it remained unclear whether the detected cleavages resulted primarily from enzymatic digestion by ribonucleases or Fe 21 -dependent nonenzymatic reaction mechanisms.
To investigate the extent of fragmentation of different rRNA species by a Fe 21 -dependent mechanism, we utilized an in vitro Fe 21 /ascorbic acid assay that was developed as part of our previous study (41), further validation of which is shown in Fig. S1. In this assay, we use purified ribosomes as a substrate, ammonium iron(II) sulfate hexahydrate (Fe(NH 4 ) 2 (SO 4 ) 2 ) to supply extra Fe 21 ions, whereas ascorbic acid is used for its ability to generate H 2 O 2 by inducing redox cycling at metal-binding sites (46,47). Cellular lysates from exponentially growing WT cells were centrifuged through a sucrose cushion to obtain purified ribosomes, which were treated with ascorbic acid alone, ascorbic acid in combination with Fe(NH 4 ) 2 (SO 4 ) 2 , or ascorbic acid with both Fe(NH 4 ) 2 (SO 4 ) 2 and the iron chelator deferoxamine (DFO). In a complementary experiment, yeast grx5D cells were subjected to oxidative stress by treatment with menadione before RNA extraction. RNA samples from the in vitro Fe 21 /ascorbic acid assay and those from menadione-treated grx5D cells were resolved adjacent to one another on the same gel and analyzed by Northern hybridization with probes for each of the four rRNAs ( Fig. 1, A-D and probe hybridization schematic in E).
Consistent with our previous findings (41), purified ribosomes treated with ascorbic acid alone produced distinct degradation fragments detected using a probe complementary to the 59 end of the 25S rRNA (y540) (Fig. 1A). Addition of Fe 21 greatly increased the degradation, resulting in well-defined rRNA fragments and disappearance of full-length 25S rRNA. The inclusion of the iron chelator DFO in the reaction completely prevented degradation. Similarly, an increase in rRNA fragmentation was observed for the 18S rRNA ( Fig. 1B), 5.8S rRNA (Fig. 1C), and 5S rRNA (Fig. 1D) when treated with ascorbic acid and Fe 21 , and the degradation was again prevented by the inclusion of the chelator DFO. Interestingly, when rRNAs degradation products generated in vitro were compared side by side with those in menadione-treated grx5D cells, we noticed a clear resemblance between the rRNAs' fragmentation patterns, regardless of a probe used (Fig. 1, compare third and fifth lanes on each blot). The similarities were seen most clearly when using probes that detect well-resolved and defined rRNA fragments, such as probe y540 for 25S rRNA (Fig. 1A), probe y532 for 18S rRNA (Fig. 1B), probe y534 for 5.8S rRNA (Fig. 1C), and probe y506 for 5S rRNA (Fig. 1D). Further probes spanning the length of both the 25S and 18S rRNAs were also used to assess the fragmentation across the length of the molecules (Fig. S2C). In all cases, the addition of Fe 21 increased the degradation observed with ascorbic acid alone, and this degradation was prevented by inclusion of DFO. However, for both the 25S and 18S rRNAs, the degradation products appeared as a smear of low molecular weight fragments as probes moved away from the 59 ends (Fig. S2).
Restoration of integrity of rRNAs by an iron chelator in the presence of Fe 21 and ascorbic acid (Fig. 1, A-D) supports the hypothesis that Fe 21 is responsible for inducing rRNA cleavage in vitro. Additionally, probes targeting regions throughout the 18S and 25S rRNAs were able to hybridize with degradation products, suggesting that Fe 21 -mediated degradation occurs at multiple sites in these molecules (Fig. S2). The similarity of rRNA degradation patterns observed between menadionetreated grx5D cells and those from the in vitro Fe 21 /ascorbic acid assay indicates that cellular RNases do not substantially contribute to these cleavage events (Fig. 1, A-D) and that a Fe 21 -mediated mechanism is primarily responsible for cleavages initiated in the 5.8S, 5S, and at the 59 end of the 25S and 18S rRNAs when cellular redox state is perturbed. Thus, iron plays a pivotal role in the degradation of ribosomes occurring during oxidative stress. Having established the broad ability of Fe 21 to promote degradation of rRNAs, we next examined whether Fe 21 is coordi-nated by the rRNA itself or by a protein component of the ribosome. For instance, Fe 21 ions could be brought into proximity with rRNAs' sugar-phosphate backbone by one or more ribosome-associated proteins, which coordinate Fe 21 without inhibiting its redox activity. Alternatively, Fe 21 might bind rRNA directly, likely by substituting magnesium ions in rRNA structures, as was proposed previously (41,48,49). To distinguish between these two possibilities, we purified ribosomes from WT yeast cells and treated them with proteinase K under nondenaturing conditions to remove all accessible ribosomal and ribosome-associated proteins prior to assaying rRNA degradation in the presence of ascorbic acid and Fe 21 ions (scheme in Fig. 2A). To assess degradation of proteins within purified ribosomes following proteinase K treatment, the samples were analyzed by SDS-PAGE and stained with Coomassie Blue (Fig. 2B), demonstrating complete degradation of protein content at both enzyme concentrations. We detected no difference in the appearance of 25S rRNA degradation products between ribosomes treated with proteinase K and those that remained untreated (Fig. 2C). Consistent with previous data (Fig. 1), degradation was observed with ascorbic acid treatment only and . Adjacent to these is total RNA from grx5D cells that were treated with 50 mM menadione for 2 h at 30°C. The sequences of all probes used in this study are listed in Table S1. Representative hybridizations are shown. E, schematic representation of the annealing sites of the probes used. See Table S1 for probe sequences.
was increased with the addition of Fe 21 (Fig. 2C). Finding that ascorbic acid alone was able to induce hydrolysis indicates that Fe 21 remained associated with the purified ribosomes after the degradation of proteins and therefore must associate directly with the rRNA. Furthermore, the observation that additional Fe 21 could enhance ribosomal degradation implies that protein-free rRNAs retain the ability to coordinate Fe 21 ions independently of any protein component.

Ribosome-bound iron primes rRNA for cleavage
The generation of the similar patterns of rRNAs' degradation products both in vitro and in cells ( Fig. 1) suggests that Fe 21mediated rRNA cleavages are not random but site-specific, implying the existence of ribosomal iron-binding loci. We aimed to further establish that Fe 21 binds to ribosomes and that these bound ions contribute to the observed cleavage events.
Having observed that ascorbic acid alone was able to induce degradation of rRNAs from purified ribosomes ( Fig. 1) (41), we reasoned that Fe 21 ions present in the ribosome remain associated with it through the purification process and are active in the ascorbic acid-induced cleavage reaction. To test this hypothesis, purified ribosome pellets were either treated with DFO to chelate any bound iron or were left untreated (scheme in Fig. 3A). Subsequently, the pellets were purified through a second sucrose cushion and washed to remove excess reagent, before being subjected to the Fe 21 /ascorbic acid assay to analyze the effect of iron chelation of purified ribosomes on the ascorbic acid-induced degradation of their rRNA (scheme in Fig. 3A). The y540 probe for the 59 end of the 25S rRNA was used because of the distinct degradation fragment pattern detected by this probe, which was also sensitive enough to allow identification of rRNA fragments when treated with ascorbic acid only ( Fig. 1) (41). Northern hybridization analysis indicated that chelation of ribosomes with DFO completely inhibited 25S rRNA hydrolysis with ascorbic acid (Fig. 3B, compare lanes 2 and 5), whereas addition of Fe 21 along with ascorbic acid restored 25S rRNA cleavages in both DFO-treated and untreated ribosomes (Fig. 3B, lanes 3 and 6).
To better understand the extent of the observed rRNA degradation, we quantified the amounts of full-length 25S rRNA in the various reaction conditions (Fig. 3C). To do this, we P P Figure 2. rRNA hydrolysis in vitro occurs independently of ribosome-associated proteins. A, schematic representation of the workflow to test the effect of removing the protein from ribosomes on rRNA cleavage in the presence of Fe 21 and ascorbic acid. Ribosomes were purified from WT cells by centrifugation through a 50% sucrose cushion, and ribosomal pellets containing 10 mg of RNA were treated with the indicated amounts of proteinase K for 5 min on ice or remained untreated. B, samples of purified ribosomes treated with proteinase K at the indicated concentrations were analyzed by SDS-PAGE and stained for total protein content with Coomassie Blue. C, samples of purified ribosomes treated with proteinase K at the indicated concentrations were treated with ascorbic acid and Fe(NH 4 ) 2 (SO 4 ) 2 , as in Fig. 1, and analyzed by Northern hybridization using the 25S rRNA probe y540. A representative hybridization is shown.
converted probe hybridization signals for the 25S rRNA and its degradation products into phosphorimager units and calculated the ratio of full-length rRNA to the total rRNA present (Fig. 3D). The resulting number multiplied by 100 represents a percentage of undegraded full-length rRNA in a sample, and we termed this the "rRNA stability factor," K RS . Using this approach, we determined that in ribosomes untreated with the chelator, approximately half of the 25S rRNA is degraded by addition of ascorbic acid, whereas DFO-chelated ribosomes are protected from hydrolysis because there is no significant loss of the full-length rRNA compared with the untreated sample. In both untreated and DFO-chelated ribosomes, the addition of Fe 21 (along with ascorbic acid) resulted in degradation of over 75% of the 25S rRNA (Fig. 3, B and C).
Taken together, these data provide evidence that Fe 21 is bound directly to yeast ribosomes and is retained through the sucrose cushion purification procedure. These ribosomebound Fe 21 ions appear to be active participants in rRNA hydrolysis, because treatment with a chelator abolished the RNA cleavages, and are likely responsible for the degradation observed following treatment with ascorbic acid only. This agrees with the reported interaction of Fe 21 with ribosomes (48) and indicates that these ions in the cleavage events demonstrated in vitro and in cells in Fig. 1.
Genetic mutations that decrease cellular Fe 21 levels protect rRNA from degradation in yeast cells subjected to low-level oxidative stress Having observed that chelating Fe 21 bound to purified ribosomes protects rRNA from degradation (Fig. 3), we next investigated whether a decrease in Fe 21 availability in cells would have the inverse effect on rRNA stability to that observed with the increased Fe 21 availability in grx5D cultures. To achieve this, we selected two genes responsible for the maintenance of cellular Fe 21 levels: AFT1, which encodes a transcription factor central to iron uptake when intracellular iron is depleted (50,51), and FET3, an Aft1-regulated gene that encodes an accessory protein to the cell-surface Fe 21 transporter Ftr1 (51)(52)(53).
To confirm the decreased intracellular iron levels in null mutants for each of these genes, we measured iron in wholecell extracts from equal cell numbers of WT and mutant cells using inductively coupled plasma (ICP)-MS (Fig. 4A). Cellular iron levels in the grx5D mutant were more than 8-fold greater than in the WT, whereas both aft1D and fet3D mutations resulted in significantly lower cellular iron levels, supporting the reported iron dysregulation in each of these mutant strains.
The null strains were next assayed by Northern blotting for degradation of the 25S and 18S rRNAs under oxidative stress, alongside cultures of the increased intracellular iron mutant . . Figure 3. Treatment of purified ribosomes with the iron chelator DFO abolishes rRNA fragmentation. A, schematic representation of the double-cushion workflow for the chelation of ribosome-bound iron, followed by the assessment of this metal's presence within ribosomal structures via an in vitro ascorbate reaction. B, ribosomes purified from WT yeast cells were treated on ice with 0.5 mM DFO for 20 min or remained untreated and pelleted through a second 50% sucrose cushion to remove DFO excess. Equal ribosome amounts, equivalent to 3.5 mg of RNA, were incubated with 0.5 mM ascorbic acid and 1 mM Fe (NH 4 ) 2 (SO 4 ) 2 . The reaction products were analyzed by Northern hybridization with probe y540; a representative hybridization is shown. C, assessment of rRNA integrity by calculating the coefficient of rRNA stability (K RS ) values. Three replicates of the reactions shown in B were used for the calculation. The error bars represent S.D. ***, P , 0.001 (two-tailed two-sample unequal variance t test). D, calculation of the coefficient of rRNA stability (K RS ) exemplified here for 25S rRNA. Radioactive signal of the probe hybridized to undegraded, full-length 25S rRNA (FL-RNA, area shown in red) is divided by the total hybridization signal in the same lane of the blot (TOTAL RNA, blue) and multiplied by 100.
grx5D and the WT strain (Fig. 4B). Each lane of the Northern blotting was quantified and converted to K RS , and the values for the menadione-treated samples were normalized to the untreated sample from the same strain (Fig. 4C).
As previously observed, extensive degradation of the 25S and 18S rRNAs was detected in the grx5D mutant (Fig. 4, B, lanes 7 and 8, and C, yellow columns). In contrast, low levels of iron in aft1D and fet3D strains were accompanied by decreased menadione-induced fragmentation of both rRNAs tested (Fig. 4B,  lanes 3-6). This manifested as an increase in K RS , indicating that a greater proportion of the total RNA remained full length following treatment (Fig. 4C, pink and red columns) when compared with WT cells (Fig. 4C, blue columns). The proportion of total 25S rRNA that remained full length was decreased by ;35% by menadione treatment in WT cells, whereas in grx5D cells it was decreased by ;75%. In aft1Dand fet3D-derived samples, this decrease was only 20-25% (Fig. 4C). The extent of degradation of 18S rRNA was less dramatic but showed the same pattern as 25S rRNA, and differences were statistically significant using probes against both, 18S and 25S rRNA (Fig. 4C). The levels of cellular iron in these strains (Fig. 4A) inversely correlates with ribosomal stability (Fig. 4C).
The data in Figs. 3 and 4 together demonstrate that iron associates with the ribosome and that there is a correlation between the level of available intracellular iron and rRNA stability. This suggests that available iron in the cell associates with binding sites on the ribosome and influences cleavage events in a manner that depends on redox activity.

Mn 21 protects rRNAs from Fe 21 -mediated ROS-induced cleavage in vitro
Our data indicate that Fe 21 is bound directly to rRNA and in the presence of ROS induces hydrolysis at adjacent sites on the sugar-phosphate backbone. We have previously postulated (41) that the observed Fe 21 -dependent rRNA cleavage events may be a result of the generation of reactive hydroxyl (OH) radicals from H 2 O 2 via the localized Fenton reaction (54).
Given the examples of Mn 21 and Fe 21 binding at the same sites, but with the crucial difference that Fe 21    radicals via the Fenton reaction, then rRNA in which the Fe 21 is substituted with Mn 21 would not generate radicals, and as such, the rRNA stability in the presence of ascorbic acid (as a source of H 2 O 2 ) would be increased.
To address this hypothesis, ribosomes purified from WT cells were simultaneously treated with ascorbic acid and different concentrations of MnCl 2 in vitro and analyzed for 25S rRNA degradation by Northern blotting using the y540 probe. As previously, the results of Northern blots were quantified and expressed as K RS (Fig. 3D) as a measure of rRNA stability. First, we tested the effect of treating purified ribosomes with Mn 21 instead of Fe 21 (Fig. 5A; quantification in Fig. 5B). In contrast to the strong rRNA-hydrolyzing activity of Fe 21 , 25S rRNA from ribosomes were treated instead with Mn 21 (at 1 or 400 mM concentrations), and ascorbic acid largely remained intact and full-length (Fig. 5A). A small amount of 25S rRNA degradation was detected from the addition of ascorbic acid alone (Fig.  5A, compare lanes 3 and 4). It was also observed that in the absence of Fe 21 , 1 mM Mn 21 improved stability in the presence of ascorbic acid, whereas at 400 mM Mn 21 there was a small increase in degradation compared with the 1 mM treatment.
To further investigate the interplay between the two transition metal ions, Mn 21 and Fe 21 , we examined whether Mn 21 can compete with Fe 21 for 25S rRNA fragmentation in the presence of ascorbic acid. We carried out the in vitro reaction in the presence of 1 mM Fe 21 and ascorbic acid alone or with a range of Mn 21 concentrations up to 400 mM ( Fig. 5C; quantification in Fig. 5D). Consistent with our initial observation (Fig.  5, A and B), Mn 21 suppressed the Fe 21 -mediated hydrolysis of 25S rRNA in a dose-dependent manner (Fig. 5, C and D). At the highest treatment level used, Mn 21 concentration was in excess of Fe 21 by 400-fold, resulting in the most prominent inhibition of Fe 21 -mediated rRNA cleavage (Fig. 5C, compare lanes 2 and  8). A 20-fold excess of Mn 21 over Fe 21 had a moderate protective effect (Fig. 5C, compare lanes 2 and 6), whereas equimolar amounts of metals had only minor effects (Fig. 5C, compare  lanes 2 and 4). Quantifications illustrated that Mn 21 is capable of restoring 25S rRNA in a dose-dependent manner (Fig. 5D), whereas the highest protection rate, at ;30%, was detected with 400 mM MnCl 2 .
To corroborate our biochemical data, we analyzed the purified ribosomal fraction prepared from WT and grx5D cells by ICP-MS to assess the presence and amounts of iron and manganese on ribosomes. As a positive control, we also measured magnesium, because it binds and coordinates ribosomal structures promoting rRNA folding and ribosome assembly and therefore is present in large quantities (23). As expected, ICP-MS detected high concentrations of magnesium in purified yeast ribosomes from both strains (Fig. S3). We were also able to detect both iron and manganese in the WT and grx5Dderived ribosomal fractions, with ribosomes purified from grx5D cells containing higher Fe 21 levels than those purified from the WT cells. Interestingly, the manganese level was lower in grx5D cells, which may reflect the increased Fe 21 in this strain out-competing Mn 21 ions for ribosomal binding sites. These data support the possibility that the observed effect of Mn 21 on rRNA in vitro may have a physiological effect.
Collectively, these results show that Mn 21 does not induce 25S rRNA fragmentation in the presence of ascorbic acid in the manner that Fe 21 does and in fact inhibits Fe 21 -mediated rRNAs hydrolysis when both metal ions are present. This would suggest that Mn 21 occupies and competes for the same sites as Fe 21 on the ribosome and prevents the Fe 21 -dependent Fenton reaction mechanism for the hydrolysis of the sugarphosphate backbone of rRNAs. We also uncovered a necessity for Mn 21 to be greatly in excess of Fe 21 to confer a significant protective effect on rRNA integrity.

Mn 21 protects ribosomal RNAs from the effects of dysregulated Fe 21 in cells
To examine whether the protection conferred to 25S rRNA by Mn 21 in vitro (Fig. 5) can occur in cells, we used the yeast grx5D strain (because it contains high levels of Fe 21 ) and supplied Mn 21 exogenously by supplementing the medium with MnCl 2 . As described previously, menadione was used to induce oxidative stress ( Fig. 1) (41). Mid-log grx5D cultures grown in rich medium supplemented with a range of concentrations of Mn 21 (from 10 mM to 2 mM) were treated with 50 mM menadione or remained untreated, and total RNA was analyzed by Northern hybridization using probes against the 25S rRNA ( Fig. 6A; quantification in Fig. 6B). None of the Mn 21 concentrations tested had an effect on the integrity of 25S rRNA in cells without menadione treatment (Fig. 6, A, odd lanes, and B,  left columns). However, upon addition of menadione, the Fe 21dependent 25S rRNA degradation was counteracted by Mn 21 in a dose-dependent manner (Fig. 6, A, even lanes, and B, right  columns). This effect was most apparent at higher concentrations of Mn 21 (1 and 2 mM), in which K RS is as much as double that in those without Mn 21 (Fig. 6B). This stark difference is consistent with the in vitro data (Fig. 5D).
We next asked whether the ribosome-protective effect of manganese is a general effect or a peculiarity of excess-iron grx5D cells. To address this question, we examined 25S and 18S rRNAs stability and cell viability in WT cells subjected to oxidative stress. Unlike for oxidant-sensitive grx5D cells, a higher concentration of menadione was required for WT cells to induce oxidative stress. Therefore, a menadione concentration range of 100-500 mM was used to treat WT cultures for growth with or without MnCl 2 in the medium. As in grx5D cells, Mn 21 did not affect the integrity of 25S rRNA in cells without menadione treatment (Fig. 6, C, lanes 1, 5, and 9, and D, columns 1-3). Again, upon addition of menadione, rRNA degradation was decreased by Mn 21 in a dose-dependent manner at all menadione concentrations tested (Fig. 6, C and D). At the highest menadione concentration (500 mM), at which the degradation of full-length rRNA is most extensive (Fig. 6C, lane 4), the presence of 2 mM Mn 21 was able to almost completely restore the integrity of rRNA, because we detected full-length 25S rRNAs in these samples (Fig. 6C, lane 10-12), which was consistent with increase of K RS by three times (Fig. 6D, compare columns  10 and 12).
Additionally, we carried out the Northern blotting analyses from Fig. 6 using probes for the 18S rRNA (Fig. S4). Although degradation of 18S rRNAs induced by menadione treatment occurs to a lesser extent compared with 25S rRNAs in both WT and grx5D cells (Fig. 6 and Fig. S4), the addition of Mn 21 was again able to rescue 18S rRNA from degradation (Fig. S4, A-C; quantifications are in Fig. S4, B-D). To test whether the observed RNA degradation in response to Fe 21 and ascorbic acid treatments was specific to ribosomal RNAs, we also analyzed degradation of the noncoding snoRNA U3 and found that stability was not affected (Fig. S5).
These data indicate that Mn 21 does not affect rRNA stability during normal growth without stress but has a protective role during oxidative stress in cells, similar to the observations from our in vitro experiments (Fig. 5). Additionally, the protective effect of Mn 21 in cells is shown here for both the 25S rRNA and 18S rRNA. Considering that 18S rRNA was found to be a target for Fe 21 -dependent oxidant-induced degradation ( Fig.  1B and Fig. S2B) (41), it is logical that Mn 21 also competes with Fe 21 in the small ribosomal subunit.

Mn 21 restores viability in yeast cells subjected to oxidative stress
To address the physiological relevance of the Mn 21 -protective effect on yeast ribosomes, we tested whether manganese would rescue the menadione-induced decrease in viability of grx5D cells (Fig. 7A). In fact, previous studies have shown that treatment of grx5D cells with low-dose oxidants results in inhibition of cell growth (41,55). We incubated grx5D cultures in medium supplemented with either 1 mM (Fig. 7A, left panel) or 2 mM (Fig. 7A, right panel) MnCl 2 for 4 h prior to adding 50 or 100 mM menadione. As a control, we used grx5D cultures that were treated with menadione in the absence of Mn 21 (Fig. 7A,  upper panels). Serial dilutions of the tested cultures were plated on the rich-medium agar plates, and cell growth was monitored over time. In the absence of Mn 21 , grx5D cell viability was almost entirely lost with 50 mM menadione treatment, and no growth was observed with 100 mM menadione treatment. However, either concentration of Mn 21 tested allowed increased cell viability following menadione treatment. The rescue effect of manganese was comparable with that of the iron chelator 1,2-phenanthroline (PHL; Fig. S6), suggesting that redox activity of Fe 21 contributes to the decreased viability and that preventing this activity via chelation has the same outcome as substituting Fe 21 with Mn 21 at the ribosome. The effect of Mn 21 on WT cell viability was also tested using the higher menadione concentrations (Fig. 7B), as in the rRNA degradation assay (Fig. 6, C and D). Growth of WT cells was completely abolished at 200 mM menadione treatment in the absence of Mn 21 in the growing culture, whereas 1 or 2 mM Mn 21 treatments resulted in partial restoration of the cell viability (Fig. 7B).
Although these data suggest that a decrease in redox-active Fe 21 at rRNAs is responsible for the protective effects of Mn 21 , we cannot rule out other antioxidant functions of the additional P . ' P P P P P ' P P P P . P Figure 6. Exogenously supplied manganese protects ribosomes from menadione-induced 25S rRNA degradation in cells. A, mid-log grx5D cultures grown in YPDA were shifted to medium supplemented with the indicated concentrations of MnCl 2 and grown for 4 h at 30°C. The cultures were adjusted to the same A 600 of ;0.6 and grown for an additional 2 h in the presence (1) or absence (2) of 50 mM menadione. RNA was isolated and analyzed by Northern hybridization with the 25S rRNA probe y540. A representative blot from three biological replicates is shown. B, quantification of the Northern hybridization data from A. The K RS values for 25S rRNA were determined for every culture treated or untreated with menadione and/or MnCl 2 . The data show mean values of three independent experiments; error bars represent S.D. The differences between K RS values determined for samples treated with menadione (right columns) grown in MnCl 2 -free medium and those grown in 1-10 mM, 0.5 mM, 1 mM, and 2 mM MnCl 2 -containing medium were significant. ***, P , 0.001; **, P ,0.01; *, P , 0.05; NS, not significant. Two-tailed two-sample unequal variance t test was used for statistical analysis. C and D, the experiment was performed and the data were quantified as in A and B, except that WT cells were used, MnCl 2 ranged from 0 to 2 mM, and menadione concentrations were 0, 100, 200, and 500 mM as indicated. Representative blots from three biological replicates are shown for all experiments. ****, P , 0.0001; ***, P , 0.001; **, P , 0.01; *, P , 0.05; NS, not significant (two-tailed two-sample unequal variance t test).
Mn 21 in this assay. Nevertheless, our biochemical characterization of Fe 21 /Mn 21 competition in the rRNA degradation process (Fig. 5, C and D) provides strong evidence that Mn 21 protects the entire ribosome from Fe 21 -dependent H 2 O 2 -induced degradation and that this contributes to the restoration of cell growth in oxidative stress.

Discussion
In this study, we have demonstrated that Fe 21 is coordinated directly by rRNA and from this position promotes fragmentation of the rRNA under oxidative conditions. These data support a model whereby Fe 21 -dependent, Fenton reaction-mediated hydrolysis is attenuated by Mn 21 , which competes with and excludes Fe 21 from binding sites on the rRNA. The working model for how the competition between these cations affects the integrity of ribosomes under oxidative stress is outlined in Fig. 8.
The polypeptide synthesis activity of ribosomes with Fe 21 and Mn 21 binding has been previously addressed in conditions that simulate the primordial earth environment (30), raising the question about the interplay between these two transition metal ions and the preferred divalent cation of the rRNA, Mg 21 , in present-day ribosomes. In our previous study (41), which focused on a specific cleavage event mediated by bound Fe 21 in the ES7 region of the 25S rRNA, it was noted that three Mg 21 -binding sites exist close to the cleavage site and that Fe 21 is likely to bind in place of Mg 21 in this context. Other work has shown that sites of RNA cleavage that occurs with Fe 21 treatment correlate with Mg 21 sites (56). We therefore consider it likely that the Fe 21 -dependent cleavages we observed throughout multiple rRNAs correspond to what are normally Mg 21 -binding sites. Given the large number of Mg 21 ions bound to rRNA, it is then perhaps surprising that any large fragments remain following Fe 21 /oxidant damage. These fragments may simply represent larger sections of the rRNA that do not readily coordinate any metal ions. Alternatively, Fe 21 cannot readily replace Mg 21 at certain binding sites. We also observed a lower efficiency of Mn 21 in protecting small ribosomal subunits. This may reflect fewer metal-coordinating elements present in the smaller 18S rRNA molecule (42).
In several cases where we detected large distinct fragments following Fe 21 -dependent rRNA cleavage, the rRNA degradation occurring inside cells was compared with that which occurred in the in vitro experiments (Fig. 1). That these same distinct fragment bands appear in both assays indicates the close similarity in cleavage mechanisms in the cell and those in the limited chemical environment of the in vitro assay. The results are presently less explicit, with several of the probes that did not allow resolution of individual fragments and produced a low-molecular-weight smear on the Northern blots. Overall, however, the degradation profile of yeast rRNAs produced with Fe 21 and ascorbic acid ( Fig. 1 and Fig. S2) argues that not only is iron necessary for initiating ROS-induced rRNA degradation, but it is also sufficient for the multiple instances of fragmentation to happen, with little to no enzymatic hydrolysis involved.
The addition of Mn 21 clearly decreases Fe 21 -mediated rRNA degradation (Fig. 5, C and D), but Mn 21 concentrations that are orders of magnitude higher than that of Fe 21 are required to have a major protective effect. An excess of 400 times results in an ;30% increase in full-length rRNA, suggesting that Mn 21 has a lower affinity for these ribosomal binding sites than Fe 21 . The stability of complexes with divalent transition metal cations tends to increase with atomic number (57), supporting Mn 21 having a weaker affinity than Fe 21 . In fact, both ions are reported to have relatively low dissociation constants when bound to proteins compared with other divalent cations (19) and can be readily replaced by Mg 21 (58). The selection between Mn 21 and Fe 21 in other biomolecules that can comfortably coordinate either ion, including SOD enzymes (17,59), appears to depend primarily on available ion concentration and may reflect shifts in metal homeostasis under stress conditions (19,40). Magnesium-binding sites in the ribosome have been classified by distinct geometries (23), and it is possible that other cations have different affinities depending on the arrangement of side-chain ligands. Investigating the bindingsite architecture and conformations of coordinated Fe 21 and Mn 21 in rRNA structures may clarify these differences.
A protective effect of Mn 21 being substituted for Fe 21 has been observed previously in nonredox enzymes in bacteria (60,61). This phenomenon occurs following oxidation of the Fe 21 in a Fenton reaction, causing it to dissociate from the active site and reaction of the hydroxyl product with the adjacent residues. In this Escherichia coli system under oxidative stress conditions, the cells enact transcriptional adjustments that increase both uptake of manganese and the sequestration of iron, contributing to the substitution of Fe 21 with Mn 21 at active sites to protect against iron-dependent oxidative damage (15). Furthermore, the protective effect was also demonstrated with cobalt, which unlike manganese does not have any inherent antioxidant activity (60). Cobalt is not imported by E. coli but can enter nonspecifically when included in the medium at high concentrations. This supports the model that the protective effect of manganese is due to substitution of iron rather than other oxidant defensive activities. It may be interesting to investigate whether cobalt is able to replicate the manganese protection of ribosomes, because this would confirm that the effect is due to iron displacement rather than general antioxidant functions associated with manganese.
Our data from live-cell experiments ( Fig. 7 and Fig. S6) do not exclude the possibility of Fe 21 or Mn 21 interactions also occurring away from the ribosome, which could influence the outcome of the oxidative stress experienced by the cell. The grx5D mutant exhibits a 6-fold increase in cellular iron (45), and although the oxidant sensitivity of this strain may in part be attributed to the increase in rRNA hydrolysis, the effects of loss of GRX5 also include impaired synthesis of the iron-sulfur clusters and lack of mitochondrial respiration (62). Also of note is the manganese-dependent Sod2 enzyme, which can incorporate Fe 21 instead of Mn 21 (63), resulting in a decreased oxidant defense. Conversely, Sod2 activity may increase with Mn 21 availability, thus enhancing the cell oxidant defense capability. However, SOD enzymes protect the cell from damaging super-oxide anion but in doing so produce H 2 O 2 , the substrate of the Fenton reaction (64). Nonproteinaceous manganese antioxidant species are also up-regulated under oxidative stress (40). The in vitro data presented show that the RNA stability effects are so closely recreated in purified ribosomes as to suggest that additional antioxidants are not required for the protective capacity of manganese on ribosomes, but we cannot exclude that there are further protective influences in the complex cellular environment.
Although we were not able to demonstrate that Mn 21 could rescue the ability of oxidatively stressed ribosomes' to synthesize polypeptides immediately after the stress was removed, we suspect that this is due to the broad effects of oxidant insult throughout cells. Logically, protecting rRNA stability has only a limited capacity to maintain translational activity after such pervasive stress. In fact, the damaging effects of oxidative species extend beyond nucleic acids and affect lipids and proteins throughout the cell, thus damaging other components of the translational machinery (65)(66)(67).
Both iron and manganese are essential elements in humans while also being toxic at high doses, and thus their import and metabolism is of medical relevance. Metal ion imbalances in cells can occur as a result of either environmental exposure or genetic mutations influencing metabolism and cellular availability (68)(69)(70). A guidance value for toxic dose of manganese is only approximately five times the recommended nutritional intake for humans (8), which leaves a relatively small range of acceptable levels. The pathologies associated with dysregulation of manganese are interlinked with those of iron because of their overlapping interactions in biological systems (71). Chronic overexposure to manganese is associated with broad neurological effects including impaired motor control and cognition (72). Iron accumulation in the brain is associated with Alzheimer's and Parkinson's diseases (73), and neurodegenerative effects are also linked to manganese accumulation, although direct links to Alzheimer's and Parkinson's diseases are less clear (72,74). Therefore interdependence between iron and manganese (as well as other metals (74)) creates a delicate situation to navigate in the maintenance of homeostasis.
In conclusion, it has been well-established that manganese can protect cells from oxidative stress, whereas iron can enhance oxidant-induced damage. That these two transition metal neighbors can substitute each other at the same sites in biomolecules presents a potent target for modulation of cellular machinery under oxidative stress. Here we report a novel observation of the protective role for Mn 21 at the ribosome, wherein it can defend against oxidative damage by displacing Fe 21 and preventing hydrolytic cleavages throughout the ribosomal RNAs. We propose that an ancient rivalry between divalent metal cations continues to influence the stability of the translation machinery. These observations highlight the multilayered competition that defines the relationship between iron and manganese, from the regulation of importers, competition for uptake, to further competition at binding targets inside the cell including the ribosome.

Extraction of RNA from cells and preparation of ribosomeenriched fractions
To purify total RNA from cells for Northern blotting analysis, we used the published formamide extraction method (75). To purify ribosomes from cells for the in vitro ascorbic acid assay, the cells were grown to mid-log phase (A 600 = ;0.8), collected, and washed in buffer A (30 mM HEPES-KOH, pH 7.4, 100 mM KOAc, 3 mM MgOAc) that was pretreated with Chelex 100 for 15 min at room temperature with shaking to remove trace iron. The cells were then lysed in buffer A by glass bead shearing. Lysates were clarified by centrifugation at 21,000 3 g for 10 min and layered on 0.5 ml of 50% (w/v) of sucrose cush-ion prepared in Chelex-treated buffer A. Ribosomes were precipitated by centrifugation for 90 min at 150,000 3 g (55,000 rpm, Beckman tabletop Optima ultracentrifuge, rotor TLA-55). Ribosomal pellets were washed twice with buffer A and kept frozen at 280°C. For an ascorbic acid assay, pellets were defrosted on ice, resuspended in buffer A supplemented with 0.15 unit/ml of RiboLock (Thermo Fisher), and the total RNA concentration was measured spectrophotometrically.
In vitro ascorbic acid assay and Fe 21 /Mn 21 competition assay The in vitro iron/ascorbic acid reactions were described in Ref. 41. In brief, we used a total of 3.5 mg RNA per in vitro reaction. Solutions of ascorbic acid, Fe(NH 4 ) 2 (SO 4 ) 2 , and/or MnCl 2 were added to RNA as indicated in the figure legends. The reaction volume was adjusted to 100 ml with buffer A; reactions were incubated for 10 min on ice, followed by precipitation of RNA with 50% isopropanol. RNA was pelleted by centrifugation at 21,000 3 g for 1 h at 4°C, washed with 80% ethanol, and dissolved in 5-10 ml of FAE (98% formamide, 10 mM EDTA).

Iron chelation and proteinase K treatment of ribosomes
For iron chelation, ribosomal pellets prepared by ultracentrifugation through the sucrose cushion (as described above) were dissolved in buffer A, DFO was added to a final concentration of 0.5 mM, and the solution was incubated on ice for 20 min. DFO-treated ribosomes were collected by another round of ultracentrifugation (90 min, 150,000 3 g) to remove the excess of DFO. Ribosomal pellets were washed and stored at 280°C or analyzed further.
For treatments with proteinase K, ribosomal pellets prepared by ultracentrifugation through the sucrose cushion were resuspended in buffer A, and 3 mg of total RNA was treated with 0.7 or 2 units of proteinase K (Roche) for 5 min on ice. The reactions were divided into two aliquots each: one aliquot of the proteinase K-treated ribosomes was resolved by SDS-PAGE and analyzed by Coomassie staining, whereas the second aliquot was used for an ascorbic acid assay.

RNA analysis
We used 1.2% agarose gels containing 1.3% formaldehyde (76) to analyze large rRNA species (25S and 18S rRNAs), whereas for small rRNAs, we used 8% polyacrylamide gels containing 8 M urea as described previously (41). RNA was transferred to nylon membranes (Hybond N, GE Biosciences). Individual rRNA species and their degradation derivatives were hybridized with 32 P-labeled oligonucleotide probes (Table S1) as described in Ref. 77. Hybridizations were visualized with the GE Amersham Biosciences Typhoon 5 imager and analyzed with ImageQuant software (GE Healthcare). For quantification, the volume of the hybridization signal corresponding to the area of interest was converted to phosphorimaging units. For quantification of rRNA stability, we calculated the ribosome stability factor (K RS ) by dividing the number of phosphorimager units corresponding to nondegraded full-length rRNA by the number of phosphorimager units derived from the entire sample lane. Where indicated, the K RS values determined in cells treated with oxidants were normalized to K RS values determined in untreated cells. The P values were calculated using two-tailed, two-sample unequal variance t tests.

Cell viability assays
Overnight cultures were diluted with YPDA to A 600 of ;0.2, grown for 4 h at 30°C in the presence or absence of 1 or 2 mM MnCl 2 , or in medium supplemented with 80 mM PHL. For oxidant treatment, the cells were incubated with various concentrations of menadione for 2 h at 30°C with shaking. For treatment of WT cells, we used 100, 200, and 500 mM menadione; for grx5D cells, we used 50 and 100 mM menadione. Following the treatments, 5-fold serial dilutions of each culture were plated on YPD agar plates and incubated at 30°C for 3-5 days.

ICP-MS analysis
The cell samples were extracted by addition of 200 ml of subboiled nitric acid (HNO 3 ), vortexed for 30 s, and let sit for 30 min at room temperature. Subsequently, 10 ml of sub-boiled water was added, and the sample was centrifuged for 8 min at 1,000 3 g before the supernatant was removed and amended with 50 ml of sub-boiled HCl. Purified ribosomes were resuspended in 120 ml of distilled H 2 O, extracted with concentrated nitric acid (32% HNO 3 v/v in extraction) at room temperature for 3 h, and then centrifuged at 17,000 3 g for 10 min; the supernatant was then diluted and amended with HCl to achieve final acid concentrations of 2% HNO 3 and 0.5% HCl (v/v). Acidified aqueous samples were analyzed via ICP-MS (Agilent 7900) using a reaction-collision cell to minimize polyatomic interferences. A calibration curve was created using standards from Inorganic Ventures and was analyzed at the beginning of the batch and for determination of the limit of quantification. Batch quality assurance measured also included assessment of seven blanks (to determine the detection limit) and on-line internal standards for assessment of signal stability. A sample was considered quantitative only if all quality assurance/quality control standards were achieved, including: above the limit of quantification, reproducible element and internal standard values from triplicate measurements, and appropriate internal standard recovery. Sample dilutions were employed if the concentration was above the calibration curve or if the sample matrix interfered with analysis (i.e. altered internal standard recovery).

Data availability
All data are presented in the article.