Oxidative Stress in Bacteria and the Central Dogma of Molecular Biology

1The earliest single-celled life forms evolved on Earth in an anoxic environment around 4 billion years ago. It was only after the “great oxidation event” approximately 2.4 billion years ago that, due to the emergence of photosynthesis, the oxygen level of Earth’s atmosphere rose significantly. Consequently, cellular life forms had to adapt to the potential danger of harmful by-products of oxygen metabolism, better known as reactive oxygen species (ROS). ROS can damage all types of cellular components like nucleic acids, proteins and lipids. Two of the most widely studied forms of ROS, superoxide (O2 −) and hydrogen peroxide (H2O2), are constantly produced endogenously through the autoxidation of O2 on a range of both aerobic and non-aerobic respiratory flavoproteins, as well as on non-respiratory flavoproteins. To protect the cells against these harmful compounds, bacteria evolved enzymes termed superoxide dismutases (SODs) to convert O2 − into O2 and H2O2, as well as catalases and peroxidases to remove H2O2 to continuously neutralize the endogenously produced ROS. In fully aerated Escherichia coli, the interplay between endogenous production and scavenging enzymes results in a steady-state intracellular concentration of ∼0.2 nM O2 − and ∼50 nM H2O2. Internal concentrations can however be increased by exogenous sources. The membrane of bacteria is semipermeable for H2O2. Naturally produced H2O2 can therefore enter and potentially damage bacterial cells. Sources of H2O2 can be H2O2 generated through photochemistry in surface waters (including in common lab media), oxic-anoxic interfaces (e.g., near the intestinal epithelium), excretion from lactic acid bacteria, and potentially indirectly from phagocytes, which produce superoxide through an NADPH oxidase. O2 − itself is not permeable through cytosolic membranes at neutral pH and can therefore not penetrate the bacterial cell enclosed in the phagosomes. One hypothesis is that the O2 − dismutates spontaneously to H2O2 to exert the observed toxic effect on the captured bacteria in the phagocyte. Due to the membranes’ impermeability to superoxide, other methods than simple diffusion must be used both in the lab and in nature to increase intracellular O2 − concentration in bacteria. Redox-cycling antibiotics, such as the synthetic viologen paraquat or the naturally occurring phenazines and quinones, are able to penetrate bacterial cells. Inside the cell, they oxidize redox enzymes and produce O2 − by transferring the electrons to oxygen. A unique source of ROS is found in phototrophic organisms. During photosynthesis, both O2 − and H2O2 are produced at the photosystem I. Additionally, singlet oxygen (1O2) is generated at the photosystem II by a transfer of light energy to oxygen. Singlet oxygen, another relevant ROS species, has a short half-life and diffuses freely through cells. Measurements of singlet oxygen travel distances in its lifetime determined that in microorganisms, 1O2 can theoretically be distributed throughout the whole bacterial cell once it is produced in the photosystems of the cytosolic membrane. In cyanobacteria, photosystems I and II are assisted by light-harvesting, antenna-like protein complexes called phycobilisomes. Close investigation of the ROS generated at phycobilisomes isolated from Synechocystis sp. PCC 6803 revealed that both a Type 1 and Type 2 photochemistry reaction took place, resulting in O2 − and 1O2 respectively. While this list of ROS sources does not comprise the whole universe of potentially harmful ROS for bacteria (e.g., lipid peroxidation, ozone), this review focuses on the most widely studied forms of ROS to summarize our current understanding of oxidative stress damage and response in bacteria.

2Oxidative stress can result in damage of both the backbone and bases of nucleic acids, both free and incorporated oxidized amino acids, as well as cofactors of proteins. To mitigate the damage of oxidative stress on cell biology, different stress response regulons are activated in bacteria, depending on the type of stressor (reviewed in). OxyR, a transcription factor widely distributed among gram-negative bacteria, is directly induced by H2O2. In most cases, the activated form of OxyR recruits RNA polymerase to transcribe about thirty different stress response genes, but there are also cases found in some bacteria where OxyR acts as a repressor to prevent the transcription of these genes under unstressed conditions. PerR is an alternative transcription factor to OxyR, which is often found in gram-positive bacteria such as Bacillus subtilis, but reports on PerR homologues in gram-negative bacteria, such as Campylobacter jejuni and Synechocystis sp. strain PCC 6803, have been published as well. The regulon of PerR contains most of the same stress response genes as the OxyR regulon. Both induce enzymes to scavenge H2O2 and therefore mitigate further oxidative damage. However, only the OxyR regulon includes disulfide-reducing redoxins to repair already damaged proteins. Two successive transcription factors are employed by E. coli to defend cells against the threat of increased O2 − concentrations. SoxR is first activated by rising superoxide levels and induces the transcription of SoxS. SoxS in turn promotes the transcription of a defensive regulon encompassing 25 proteins to prevent the entry and accumulation of redox-active molecules in the cell. The SoxRS organization as found in E. coli is not a universal trait of all bacteria. In fact, performing a BLAST search for both SoxR and SoxS homologues in the bacterial domain revealed that SoxR is restricted to Proteobacteria and Actinobacteria, while SoxS was exclusively found in the family Enterobacteriaceae. The model bacterial system for the response to photooxidative stress through 1O2 is the anoxygenic anaerobic photosynthetic Rhodobacter sphaeroides. In R. sphaeroides, the 1O2 response regulon is activated through a cascade of transcription factors. First, the alternative sigma factor RpoE gets activated. While the RpoE regulon itself is rather small, it includes two additional sigma factors, RpoHI and RpoHII, which in turn activate the stress response regulon against singlet oxygen.

3These defensive systems can be highly effective, allowing an exponentially growing E. coli culture to survive and overcome an extracellular H2O2 concentration that is 106 times the normal intracellular H2O2 concentration produced by endogenous activity. Interestingly, cyanobacteria, which initially caused the great oxidation event, are much more susceptible to H2O2 as compared to eukaryotic aquatic microorganisms. This higher susceptibility to H2O2 might potentially be exploited to combat increased cyanobacteria blooms caused by global warming. Notwithstanding the effectiveness of the response system, damage through oxidative stress can be acquired extremely quickly and acts non-discriminatory on all cellular compounds. In the next chapters, we will review how exactly oxidative damage influences different classes of biomolecules. In particular, we will focus on the damage that oxidative stress inflicts on the different steps of the extended central dogma of molecular biology in bacteria, which describes the flow of genetic information in a cell (Figure 1).

4DNA cannot be damaged directly neither by H2O2 nor by O2 − alone. Nevertheless, mutants lacking SODs or catalases and peroxidases show increased mutagenesis rates. The mutagenic effect of H2O2 is easily explained by the Fenton reaction mediated by Fe2+-ions associated with the DNA:The resulting hydroxyl radical is short-lived due to its high reactivity. It can react with most biomolecules at almost diffusion-limited rates. When produced in close proximity to DNA, hydroxyl radicals can damage both the nucleobase or the deoxyribose moieties, leading to mutations and strand breaks. Fittingly, Rai et al. found Fe2+ preferably associated to specific short DNA sequences that had previously been found to be preferentially cleaved under oxidative stress conditions. A commonly used marker for oxidative stress damage is 8-oxo-7,8-dihydroguanosine (8-oxo-G), the most commonly observed oxidation product of a guanine base (Figure 1). Guanine has a lower reduction potential than the other DNA bases. Nearby oxidized bases can therefore readily be repaired by an electron transfer from the guanine base to the oxidized base, effectively transferring the oxidation site onto the guanine. 8-oxo-G can base-pair with both a cytosine or adenine base, thus affecting the coding potential of DNA. Crystal structures of a Bacillus stearothermophilus DNA polymerase with 8-oxo-G pairing to both the cognate and mismatched base showed that under these conditions the proofreading mechanisms of the DNA polymerase are no longer effective. In fact, Hsu et al. observed an inversion of the ordinary mismatch recognition. The in theory canonical 8-oxo-G:cytosine base-pair behaved as a mismatch, whereas the 8-oxo-G:adenine base-pair was recognized as cognate, which ultimately leads to a G to T transversion in the replicated DNA strand.

5The impact of O2 − on DNA is less obvious. In the end, the observed DNA damage is again due to the Fenton reaction described above. The involvement of superoxide is however only indirect. Keyer et al. showed that SOD deficient bacterial strains had increased free intracellular iron concentrations. The authors were also able to show that O2 − excised the iron from a family of dehydratases that contain [4Fe-4S] clusters. Additionally, O2 − was shown to reduce Fe3+ to Fe2+, regenerating ferrous iron for a continuous Fenton reaction. Finally, O2 − is converted to H2O2 through the actions of SODs. The increase in both Fenton reaction educts results in the observed higher mutagenesis rate. Taken together, these findings highlight the need of bacterial cells to properly regulate their intracellular pool of free iron as well as iron associated and bound to macromolecular complexes. Unsurprisingly, both the transcription factors OxyR and PerR induce the transcription of an iron import repressor (fur) and a DNA-binding mini-ferritin (dps, mrgA in B. subtilis) to minimize free iron under conditions of oxidative stress. The SoxRS system also activates Fur but has no effect on Dps. Instead it stimulates transcription of the endonuclease IV (nfo) gene for DNA repair. Similarly, iron-binding proteins are upregulated through the RpoE/RpoHI/II sigma factors cascade in R. sphaeroides, while uptake mechanisms for Fe2+ are minimized.

6In contrast to O2 −, singlet oxygen can directly react with DNA. Unlike HO∙, which targets all nucleobases, 1O2 exclusively interacts with guanine, which contains conjugated, electron-rich double bonds. The resulting oxidation product of the base is however again 8-oxo-G.

7In bacteria, the genetic information stored in DNA is transcribed to RNA by a single RNA polymerase (RNAP) structurally similar to eukaryotic RNA polymerase II. Specificity of gene expression is modulated by the use of different sigma (σ) factors and an array of transcription factors such as the above mentioned OxyR, PerR, and SoxR/S (Figure 1). The fact that the stress response genes in their regulons are actively transcribed and translated as a response to oxidative stress is evidence that the RNAP is not incapacitated under these conditions. The aforementioned transcription factors themselves are activated by oxidation. For example, H2O2 oxidation of OxyR induces the formation of an intramolecular disulfide bond, which results in a significant conformational change of OxyR into the DNA binding form. PerR, which is often found in gram-positive bacteria, generally works as a repressor. Upon oxidation, it loses the ability to bind DNA and thus enables the stress regulons to be transcribed. PerR subunits have two metal ion binding sites; the structural site which irreversibly binds Zn2+ and the regulatory site which reversibly binds Fe2+. PerR is deactivated by H2O2 through metal-catalyzed oxidation of histidine residues that coordinate the Fe2+ ion in the regulatory site, resulting in the dissociation of PerR from DNA. The homodimer SoxR, which is first triggered in the SoxRS cascade, is activated through the oxidation of [2Fe-2S]+ clusters when cells are exposed to redox-cycling compounds. Whether it is O2 − that directly induces SoxR oxidation is however disputed. Activation of the alternative transcription factor RpoE in R. sphaeroides by singlet oxygen is sensed by ChR, an anti-transcription factor in complex with RpoE. ChR is targeted by 1O2, which results in the disassembly of the complex, releasing active RpoE.

8Additional evidence for the functionality of RNAP under oxidative stress is the transcription of small regulatory noncoding RNAs (sRNAs). One example found in E. coli is OxyS, which is induced by OxyR. This 109 nucleotides (nt) long RNA has been implicated in the regulation of as many as 40 genes and deletion of the oxyS locus results in higher mutagenesis rates compared to wildtype cells when treated with H2O2. Barshishat et al. showed that OxyS induces a temporary cell cycle arrest, therefore allowing for proper DNA damage repair. The cell cycle arrest is prompted through the repression of the nusG gene upon binding of up to two OxyS molecules around the ribosome binding site on the nusG mRNA. NusG is a cofactor of Rho-dependent transcription and has been implicated in the suppression of prophages and other horizontally acquired portions of the genome. The downregulation of NusG results in elevated expression of the rac prophage which encodes for the kilR gene. KilR prevents proper assembly of FtsZ, a ring-like structured protein essential for cell division, contributing to the observed cell cycle arrest upon oxidative stress. Several additional sRNAs have been found to be induced by both 1O2 and O2 − in R. sphaeroides. A subset of these sRNAs were shown to be directly regulated by the RpoE/RpoHI/II sigma factors cascade as a response to photooxidative stress. Later reports confirmed the roles of these sRNAs in oxidative stress resistance.

9An inverted role was described for the sRNA IsrR in the cyanobacterium Synechocystis sp. PCC 6803. IsrR was found to regulate the expression of the iron deficiency-stress induced protein A (IsiA), which was postulated to shade the photosystem II for oxidative stress prevention. However, in this case, IsrR is not induced by oxidative stress but is constitutively transcribed, resulting in diminished isiA mRNA levels. Upon oxidative stress, IsrR levels are reduced, allowing for isiA mRNA accumulation and translation.

10Upon a short H2O2 burst, Liu et al. found that in E. coli 8-oxo-G levels were higher in RNA than in DNA, meaning that oxidative stress has a stronger effect on RNA. Additionally, they showed that highly structured RNA like tRNAs (transfer RNA) and rRNAs (ribosomal RNA) are not protected from oxidative damage. This is in line with findings of other groups that oxidative stress generally inhibits the translation process. This chapter will now discuss in more detail how the different cogwheels of the translation machinery are affected by oxidative stress (Figure 1).

11mRNAs carry the information for the synthesis of a protein. Equal to its mutagenic effect in DNA replication, 8-oxo-G can affect the decoding process during translation. Using an E. coli in vitro translation system, Simms et al. showed that an oxidized base in the A-site codon can lead to the incorporation of a non-cognate amino acid. 8-oxo-G preferably adopts the syn conformation of the glycosidic bond in the ribosomal A-site, resulting in an 8-oxo-G:A base pair during the decoding step. However, regardless of whether the incoming tRNA had a C or an A base pairing with the 8-oxo-G, translation rates at these codons in the in vitro system were decreased to an extent that suggests strong translational stalling in vivo instead. Incorporation of a non-cognate amino acid could lead to misfolding and protein aggregation. In agreement with the conclusion of Simms et al., a later study found no increase in protein aggregation under oxidative stress conditions. Simms et al. additionally found that oxidation at the wobble position, where usually non-cognate base pairing is tolerated more freely, equally affected peptide bond formation rates as oxidation at the other two positions of the codon did. Summarizing their findings, Simms et al. concluded that oxidations of mRNA bases inhibit the decoding step by preventing the small subunit from adopting the active conformation required for proper tRNA selection regardless of the position of the oxidative lesion in the codon. This observation, however, seems to be exclusive to mRNA-tRNA interactions in the decoding center, since peptide release mediated by bacterial release factors (RF1 and RF2) was not affected by oxidized stop codons. The hypothesis that mostly the elongation step of translation seems to be affected by oxidative stress is supported by findings of other groups. Zhong et al. found an increase of polysomes under oxidative stress conditions, an observation that can be explained by uninhibited translation initiation and stalling ribosomes on the mRNA. Similar results were found in a cyanobacterium. In a more recent publication, Zhu et al. quantified the elongation rate in E. coli upon addition of H2O2. They found an immediate and dose-dependent decline of the elongation rate upon stress induction (down to one amino acid per second at 5 mM H2O2), followed by a gradual recovery to a normal elongation rate of approximately 16 amino acids per second. Furthermore, the immediate drop in elongation rate and the H2O2 dose-dependent gradual incline of the rate to normal levels correlated well with an observed dose-dependent growth arrest phase.

12Generally, stress response mechanisms and changes in growth rate can have diverse effects on mRNA stability in bacteria in both global and gene-specific ways. To our knowledge, there is no numerical data available on mRNA degradation rates in bacteria under oxidative stress conditions. Using a DNA microarray analysis of H2O2 stressed E. coli cells, Zheng et al. reported an upregulation of the OxyR regulon and other stress response genes, as well as a downregulation of many ribosomal protein genes, cold shock genes, ATP synthase genes, and transporter genes. There are however some reports on sRNAs regulating mRNA stability upon oxidative stress conditions. As discussed above, sRNAs can positively affect oxidative stress resistance. While in the case of IsrR the degradation of the sRNA was shown to stabilize isiA mRNA, a different mechanism was described for the sRNA OsiR found in Deinococcus radiodurans. OsiR also positively regulates oxidative stress tolerance, but its transcription is increased upon oxidative stress. OsiR targets the mRNA of KatE2, one of two KatE-type catalases found in D. radiodurans. Under oxidative stress conditions, OsiR base-pairs with the katE mRNA, which results in increased mRNA transcript levels and increased translation of KatE.

13RNA modifications have the potential to regulate mRNA stability. For example, N6-methyladenosine (m6A) can mark mRNAs for degradation in humans. Whereas m6A has been found to be an abundant mRNA modification in gram-negative bacteria, no change in the m6A/A ratio was observed under oxidative stress conditions in both E. coli and P. aeruginosa. Adenosine-to-inosine (A-to-I) RNA editing is another form of post-transcriptional RNA modification found in all domains of life. A recent report linked a specific A-to-I editing event on the mRNA of the flagellar filament protein fliC to an increased virulence in the bacterial plant pathogen Xanthomonas oryzae pv. oryzicola (Xoc). Plants often generate ROS as a defensive reaction to either directly kill invading pathogens or, more importantly, work as signaling molecules for gene expression activation in the context of pathogen defense. The authors identified a serine to proline mutation (S128P) in FliC after exposing the bacteria to H2O2. The S128P editing event can potentially change filament structure and enhance biofilm formation, leading to an increased oxidative stress tolerance.

14Aminoacylated tRNAs read the different codons on mRNAs with their anticodon to deliver the correct amino acid to the ribosome. Interestingly, Liu et al. found that tRNAs folded into their native structure displayed even slightly higher 8-oxo-G levels than denatured tRNAs when treated with H2O2 in vitro. tRNAs are next to rRNA the second most abundant class of RNA and generally considered stable transcripts. However, recent reports indicate that tRNA stability can be swiftly altered as a response to certain stress conditions and general demand In accordance, both Zhong et al. and Zhu et al. found a global reduction in full-length tRNA availability after the induction of oxidative stress by H2O2 in E. coli. The H2O2 dose-dependent reduction in full-length tRNAs could at least partially explain the strongly reduced elongation rate during oxidative stress conditions. Further evidence for this hypothesis is given by the observation that an overexpression of RNase D, resulting in an already initially reduced tRNA pool, lead to a hypersensitivity to H2O2. The opposite was true, when the tRNA pool was enriched by addition of a plasmid carrying extra copies of rare tRNA genes, cells became more resistant to oxidative stress. Surprisingly, Leiva et al. published somewhat contradictory results to the two previous reports. Instead of a global reduction of the tRNA pool, the authors found that from ten tested tRNA species only one specific tRNA (tRNAGly) was downregulated under both H2O2 and paraquat stress. One explanation for this discrepancy might be the fact that all three publications used different E. coli strains in their studies, hinting to the possibility that the response to oxidative stress could even be strain specific. Leiva et al. also investigated whether the tRNA modification pattern was altered in response to the oxidative stress. They found no change at all known modification sites for the tRNAGly isoacceptors before and after addition of the stress. However, the authors also pointed out that abasic sites and unknown oxidation products of modified tRNAGly nucleotides might have escaped their analysis. Another study, using UV-induced oxidative stress, did describe degradation of many post-transcriptional modifications under oxidative stress conditions.

15tRNAs are acylated with the correct amino acids by aaRSs. To ensure the cognate amino acid is loaded onto the 3′-CCA tail of the tRNA according to the genetic code, aaRS have evolved pretransfer and posttransfer proofreading functions (reviewed in). While the aminoacylation active site of aaRSs exclude most non-cognate amino acids (pretransfer), especially structurally similar amino acids can escape this first proofreading sieve and be mistakenly added to a non-cognate tRNA. The posttransfer editing sites can correct these erroneous loadings. Oxidative stress has been shown to affect both the free amino acids (reviewed in) and the editing sites of aaRSs. For example, the aromatic amino acid phenylalanine (Phe) is a prime target for oxidation through ROS. One of the potential oxidation products of Phe, meta-tyrosine (m-Tyr), has been shown to be incorporated into E. coli proteins at Phe codon positions if the editing site of PheRS was mutated. Additionally, the editing deficient PheRS E. coli strain was shown to be more sensitive to oxidative stress. Contradicting these findings, in Salmonella enterica serovar Typhimurium (S.Typhimurium) the PheRS displays increased proofreading activity of m-Tyr-tRNAPhe under oxidative stress conditions. Salmonella have been shown to be able to survive inside of ROS producing macrophages. A restructuring of the S. Typhimurium PheRS under oxidative stress conditions to a hyperaccurate editing version can at least partially explain the observed resilience of Salmonella against human ROS-mediated immune defense mechanisms.

16In the case of threonyl-tRNA synthetase (ThrRS), which is prone to incorporate the near-cognate serine (Ser), the editing site contains an essential cysteine residue. Surface cysteine can readily be oxidized by ROS. Consistently, oxidative stress has been found to deactivate the editing site of ThrRS and lead to the misincorporation of Ser at Thr codons. If protein mistranslation due to loading of non-cognate amino acids on tRNAs under oxidative stress happens in vivo, increased protein aggregates due to misfolded proteins might be expected. As mentioned above, no such aggregates were detected in a later study. One explanation could be that the misfolded proteins are readily captured and degraded by proteases before they aggregate. Indeed, a protease-deficient E. coli strain was shown to be more susceptible to H2O2. An effect that was exaggerated when Ser was added in excess. These mistranslation events and the resulting need for chaperones show why genes encoding the HSP20 chaperones have been found to be induced by H2O2 in an OxyR independent manner.

17At the ribosome, proteins are formed by linking together the amino acids brought to the ribosome by tRNAs according to the codons of the mRNA present in the ribosomal decoding center. Bacterial ribosomes are enormous heterologous complexes consisting of a large (50S) and a small (30S) subunit. Both subunits contain RNA and protein components. Even though the ribosomal proteins outnumber the rRNA molecules (54 proteins to 3 rRNA molecules in E. coli), the rRNA content prevails and establishes the structural and functional core of the ribosome. As mentioned before, the high structural complexity of the ribosomes does not protect rRNA from oxidation. In a more recent study, we showed that the oxidation of the 50S subunit had more of a negative impact on in vitro translation than the oxidation of the 30S subunit (; Figure 1). We observed the same trend with in vitro reconstituted 50S and 30S subunits using H2O2 treated 23S rRNA and 16S rRNA, respectively, in combination with ribosomal proteins isolated from unstressed Thermus aquaticus. In both cases, in vitro translation of a genuine mRNA was reduced. However, the reduction of translational activity was more severe using oxidized 23S rRNA as compared to 16S rRNA. The 23S rRNA of the 50S subunit harbors the peptidyl transferase center (PTC) of the ribosome. Interestingly, the investigation of the influence of oxidized PTC nucleotides using an atomic mutagenesis approach showed varied results. Whereas the oxidation of 23S rRNA residues A2451 and U2585 resulted in a significant loss of translational activity in vitro, the oxidation of other active site nucleotides had no effect at all (U2506, A2602, G2447) or resulted in even slightly stimulated protein synthesis as in the case of 5-OH-C2063. While these varied results highlight the complexity of oxidative stress damage, the question remains whether these particular nucleotides are actually prone to oxidation in vivo. Therefore, Willi et al. generated a crude oxidation map of the 23S, 16S, and 5S rRNAs using an 8-oxo-G antibody for immunoprecipitation of oxidized RNA followed by sequencing and alignment to rRNA genes. Fitting to the previous findings, we reported non-random occurrences of oxidation hotspots which were mostly found on the 23S rRNA, including in domain V which harbors the PTC nucleotides. Remarkably, 16S rRNA regions forming the decoding site and the anti-Shine-Dalgarno sequence of the small subunit appear to remain largely oxidation-damage free in vivo, which fits with previous reports that translation initiation was not affected by oxidative stress. Efforts for a higher resolution oxidation map of the rRNAs comprising also other rRNA oxidative lesions are currently underway. As mentioned above, oxidation of rRNA inside the ribosome does not necessarily interfere with ribosome functions and oxidation of some residues even resulted in accelerated rates of protein synthesis in vitro. This illustrates that rRNA nucleobase oxidations can potentially be utilized by bacteria to fine-tune ribosome functions during different growth or stress conditions. Indeed, the natural occurrence of the bacterial rRNA modification 5-OH-C2501 gives further credence to this hypothesis. While in the polyextremophile Deinococcus radiodurans the modification is found ubiquitously on the bacterium’s ribosomes, it is completely absent in the rather closely related thermophile Thermus thermophilus. In E. coli, the extent of the modification is growth phase dependent. While in early exponential phase only about 30% of ribosomes are modified, the frequency increase in stationary phase to about 70%. Recent research on the 5-OH-C2501 modification identified RlhA as the modifying enzyme, thereby completing the set of rRNA modifying enzymes in E. coli. The exact biological role of this cytosine hydroxylation inside the ribosome is however so far unknown.

18RNA molecules have a negatively charged phosphate backbone. Magnesium ions (Mg2+) are well known for their ability to counteract this negative charge and therefore stabilize the folding of large RNA molecules. Unsurprisingly, there are between 100–1,000 Mg2+ ions found on bacterial ribosomes that are essential for the structural integrity and functional activity of the ribosome. However, it has been shown that these Mg2+ ions can be replaced by Fe2+ and Mn2+ without significant loss of activity, mimicking the conditions of the early anoxic Earth. While it has been shown before that exchange of Mg2+ ions in the ribosome can have a detrimental effect on rRNA stability, research on the specific role of ribosome bound Fe2+ for oxidative lesions generated through localized Fenton reactions is scarce in bacteria. Most of the work has been focused on eukaryotes and human neurodegenerative diseases specifically. However, since the core structures and functions of the ribosome are conserved, it is not far-fetched to assume that similar findings can be expected for bacteria. Indeed, our own unpublished data shows a significant increase of 8-oxo-G occurrences at nucleotides that are <2.4 Å from Mg2+ binding sites when ribosomes were treated with Fe2+ and H2O2.

19Ribosomal proteins have no apparent shielding effect on proximate rRNA against oxidative stress lesions (, unpublished data). Nevertheless, they are also not spared from oxidative damage. In a global approach to detect redox sensitive cysteines before and after the addition of oxidative stress, Leichert et al. identified both small and large ribosomal subunit proteins that had at least one cysteine oxidized. Another study found increased carbonylation of ribosomal proteins in viable but non-culturable E. coli cells found in a stationary phase culture. A specific example of oxidative damage actually inhibiting the biological function of a ribosomal protein was described in two publications in 1978. In the stalk region of the ribosome, protein uL10 (nomenclature according to) and multiple copies of the protein bL12 form one of the main interaction sites of the ribosome with translation factors, also known as the GTPase-associated center. The multiple copies of the bL12 protein have been shown to occur in two dimers on the uL10, whereas mutation of the uL10 to a single dimer-binding form showed strongly reduced translational initiation and elongation. Oxidation of the three methionine residues found in bL12 to methionine sulfoxide resulted in a loss of dimerization. This leads to inhibited binding of bL12 to uL10 on the ribosome, which in turn leads to a decrease of EF-G dependent translation (see also below). Another publication shows an interesting connection between the ribosomal protein uS4, mistranslation, and the resistance to oxidative stress. The authors found that E. coli cells with error-prone ribosomes due to the I199N mutation of uS4 show a significantly lower H2O2 sensitivity than E. coli cells with wildtype ribosomes. Mistranslation of proteins leads to the induction of the general stress response transcription factor RpoS. The regulon of RpoS includes the catalase KatE and the peroxidase OsmC, both of which were found to be upregulated in the error-prone E. coli strain, explaining the higher tolerance to H2O2. The exact link between mistranslation and induction of RpoS is however not completely clear. The authors speculate that mistranslation of proteins buffers away the protease ClpP, which usually degrades RpoS. This would therefore provide a new pathway to activate the general stress response as a defense mechanism against H2O2 that is independent of the OxyR regulon, relying on protein mistranslation only. Additionally, the authors find that the deletion of the sRNA DsrA, which is known to positively regulate RpoS translation, results in the loss of the protective effect observed against H2O2. Mechanistic explanations of this observation remain to be elucidated.

20Ribosome biogenesis is a particularly energy intensive process for the bacterial cell. Recycling the whole ribosome after oxidative damage would therefore appear very wasteful. To prevent the incorporation of oxidized free nucleotides in the first place, E. coli cells use the protein MutT to hydrolyze both 8-oxo-dGTP and 8-oxo-rGTP to the monophosphate forms, reducing mutagenic errors in DNA and RNA respectively. While it is yet unclear whether the damaged rRNA can actually be repaired in cells, there is at least evidence that the polynucleotide phosphorylase (PNPase), which specifically binds oxidized RNA transcripts, also plays a role in proper ribosome assembly. In the case of damaged ribosomal proteins, there is evidence that a subset of ribosomal proteins can be exchanged in the fully assembled ribosome, including the above discussed uL10 and bL12.

21Elongation factors (EF) are crucial for an efficient translation process. As was discussed above, mostly the elongation step of translation is inhibited by oxidative stress (Figure 1). It would therefore not be surprising, if also the essential elongation factors would be affected by ROS. Indeed, when Tamarit and co-workers analyzed the carbonylation pattern of proteins in E. coli after the treatment with H2O2, EF-G was found to be one of the primary targets. The same was true in Bacillus subtilis for both EF-Tu and EF-G. EF-Tu assists the decoding step of elongation. It delivers aminoacylated tRNAs to the free A-site of the ribosome in a ternary complex with GTP. If correct base-pairing of the codon and anti-codon in the decoding center of the 16S rRNA is established, the GTP is hydrolyzed in the GTPase-associated center on the large subunit and EF-Tu is released. In the cyanobacterium Synechocystis sp. PCC 6803 (Synechocystis), H2O2 oxidizes a conserved cysteine of EF-Tu. The authors further showed that in an E. coli in vitro translation system the use of the oxidized EF-Tu resulted in strongly inhibited translation. Similar findings were described for EF-G in both Synechocystis and E. coli. EF-G catalyzes the tRNA/mRNA translocation step of elongation, where after a new peptide bond is formed between two amino acids, the ribosome moves along the mRNA to the next codon. Like EF-Tu, EF-G is a so-called G-protein. Its overall structure resembles the structure of the aa-tRNA/EF-Tu complex and it utilizes the hydrolysis of GTP triggered at the ribosomal GTPase-associated center as energy source. Upon oxidation with H2O2, two conserved cysteine residues formed a disulfide bridge which resulted in reduced GTPase activity and suppressed dissociation of EF-G from the ribosome. Interestingly, mutations of the two target cysteines to serines rendered the EF-G mutant resistant to oxidative stress and did not show significant loss of activity. Therefore, the oxidation of these two conserved cysteines is likely a desired and expected mechanism for regulation of translation under adverse condition rather than random oxidative damage.

22Once translated and properly folded, proteins fulfill a wide array of cellular functions, ranging from structure to catalysis to metabolism and to reproduction. As discussed above in piecemeal, there are several ways how oxidative stress can damage proteins. Carbonylation represents an irreversible and unrepairable damage to proteins. It mostly affects the amino acids proline, arginine, lysine, and threonine and has been used in organisms of all domains of life as an indicator of oxidative damage. Histidine can be oxidized to 2-oxo-histidine as has been shown in the deactivation of PerR through oxidative stress. The amino acids cysteine and methionine both contain an electron-rich sulfur atom in their side chains. Unsurprisingly, they are prone to oxidation through ROS as was seen in the case of OxyR activation, the loss of editing activity in the ThrRS, the loss of dimerization of bL12, and the loss of both EF-Tu and EF-G activity. Cysteine and methionine oxidation is further reviewed in. In the case of singlet oxygen, the impact on free amino acids has been studied extensively in vitro. 1O2 reacts selectively with only a subset of amino acids, namely cysteine, methionine, histidine, tyrosine, and tryptophan, and the interaction of 1O2 with these residues was found to be strongly solvent dependent. Studies under more biological relevant conditions, however, confirmed these findings. Importantly, the accessibility of the five vulnerable amino acids in folded proteins (i.e., is the amino acid solvent exposed or buried inside of the protein) can alter the reactivity of a specific residue toward singlet oxygen. A final mechanism of ROS impact on proteins discussed above was the oxidation of iron-sulfur (Fe-S) clusters, as was seen in the case of SoxRS activation. Fe-S clusters are ancient prosthetic groups that could assemble spontaneously on polypeptides of the primordial, anoxic environment found during the first billion years of life on Earth. Today, they are still one of the most ubiquitous cofactors found in nature. Their functions are extremely versatile, ranging from simple iron storage to the regulation of gene expression. Both O2 − and H2O2 can damage Fe-S clusters by either directly subtracting an electron or by generating reactive HO∙ through a Fenton reaction. Additionally, most often cysteine is the coordinating amino acid of the polypeptide backbone forming the Fe-S cluster. Oxidation of the coordinating cysteine by any generated ROS could lead to the inactivation of the Fe-S protein. Finally, H2O2 poisons the Isc system, which is responsible for Fe-S cluster synthesis. Therefore, in E. coli the OxyR regulon includes the activation of the Suf system, an alternative Fe-S cluster synthesis pathway.

23Once oxygen levels rose on Earth, bacterial cells had to develop systems to continuously remove endogenously generated ROS to keep oxidative damage to a minimum. Nonetheless, ROS levels inside cells can rise due to a variety of external reasons (Figure 1). Even though bacteria have efficient scavenging and repair systems, ever-emerging research in the field of oxidative stress has shown that all the steps of the central dogma of molecular biology are nonetheless susceptible to oxidative damage. In this review, we focused exclusively on bacterial systems. However, several parts of the central dogma of molecular biology are well conserved among all domains of life thus allowing extrapolation of insights gained in bacteria to other kingdoms. Especially the ribosome, the central hub of translation, has a highly conserved core found everywhere from bacteria to humans and findings from bacteria can, to an extent and with reasonable caution, be extrapolated to more complex organisms. Nevertheless, there is a plethora of research on oxidative stress directly performed in mammalian systems to be found. Of the outmost interest is the apparent link between oxidative stress and neurodegenerative diseases such as Alzheimer’s and Parkinson’s disease. Furthermore, there is an emerging view that these diseases and the associated oxidative stress are influenced by the interplay of the human host and the gut microbiome, highlighting once more the important role of bacteria in human health.