Keywords

1 Introduction

Oxidation of DNA has been known for over 40 years and considerable efforts have been made to describe the generation, metabolism, and repair of such lesions. The progress in this field is a recurrent topic in review articles (Cadet et al. 2010; Barciszewski et al. 1999). Unlike in DNA, RNA oxidative damage has not been in the focus of research until the past decade. This might be due to the assumed transient nature of RNA. Nonetheless it has been shown that most mRNAs have an average half-life of ~7 h in mouse cells and ~10 h in human cells, attributing the small percentages of mRNA with considerable lower half-lives to transcription factor mRNAs (Yang et al. 2003; Sharova et al. 2009). Moreover mRNA just accounts for a few percent of total RNA in cells, whereas the majority of RNA includes stable tRNA and rRNA which have considerably lower turnovers: the half-lives in vitro are estimated to be in the range of several hours to days for tRNA and several days for rRNA (Defoiche et al. 2009).

In the recent years the central dogma of RNA being “the messenger in the middle” between DNA and proteins had to be expanded by the disclosure of the activities of non-protein-coding RNAs (ncRNA): while over 85 % of our genome is transcribed to RNA, the known protein-coding gene exons only account for less than 3 % of the human genome (Hangauer et al. 2013). Such ncRNAs have been linked to diverse regulatory mechanisms (Nelson and Keller 2007).

Overall these data imply that oxidative damage to RNA could in principle have a strong and probably so far underestimated impact on different cell mechanisms. Recently, oxidatively damaged RNA has been closely linked to aging, a still growing number of neurodegenerative and other diseases (Poulsen et al. 2012). The data suggest that RNA oxidation is an early event in disease rather than the effect of cell decay (Shan et al. 2007; Chang et al. 2008). The progress in this field has been thoroughly reviewed elsewhere (Nunomura et al. 2009; Poulsen et al. 2012).

It is a long known fact that RNA is hydrolytically less stable than DNA due to the fact that the 2′-hydroxyl group of the ribose undergoes cyclophosphate formation with the 3′-O-phosphate under basic conditions which leads to strand cleavage. Compared to DNA, purified RNA is known to have a greater chemical oxidative stability (Thorp 2000). Nevertheless RNA is considered to be more prone to oxidation due to its higher abundance in cells, its extensive subcellular distribution (vicinity to mitochondria), and lesser association with protective proteins (Li et al. 2006). Indeed it has been found that in mitochondrial DNA there is a 16-fold higher abundance of 8-oxo-dG compared to nuclear DNA in rat liver. This has been explained by the spatial vicinity of the oxygen metabolism, the absence of histones in mitochondria, and the relatively inefficient DNA repair (Richter et al. 1988). Moreover it was shown recently that in H2O2-treated E. coli cells, levels of 8-oxo-rG in rRNA were the same or even higher than that for non-ribosomal RNA (Liu et al. 2012). According to the authors the results are consistent with the observation of rRNA oxidation in vivo and suggest that complex structures and association with proteins do not protect rRNA from oxidation.

It has been found that oxidative stress causes greater damage to RNA than DNA in different mammalian cells: the basic levels of 8-oxo-rG in RNA increase up to the fivefold after oxidative stress and are 10–25 times higher than for 8-oxo-dG in DNA (Fiala et al. 1989; Shen et al. 2000; Hofer et al. 2005, 2006).

2 Definition of Reactive Oxygen and Nitrogen Species and Their Properties

Reactive oxygen and nitrogen species are collective terms for either oxygen or nitrogen containing molecules, radical or nonradical, that act as oxidizing agents or that can be easily converted into radicals.

Reactive oxygen and nitrogen species (ROS, RNS) are either produced endogenously as natural by-products of the normal cellular metabolism of oxygen and nitrogen or from exogenous sources like tobacco, drugs, xenobiotics, and ionizing radiation. Under oxidative stress, which implies an imbalance between the generation of ROS and the system’s capability of neutralizing ROS or subsequent damage, these reactive intermediates may damage a wide spectrum of vital cell compounds like nucleic acids, amino acids, and unsaturated fatty acids.

With regard to nucleic acid damage the main reactive oxygen species include hydrogen peroxide (H2O2), singlet oxygen (1O2), and the radical species superoxide (•O2 −) and hydroxyl radical (•OH). Reactive nitrogen species include peroxynitrite (ONOO−), nitric oxide (•NO), and nitrogen dioxide (•NO2). Endogenous sources of ROS mainly include the mitochondrial respiratory chain, NADPH oxidases (NOX), and 5-lipoxygenases (LOX) among others which mainly form the primary ROS •O2 − (Nathan and Cunningham-Bussel 2013). Mitochondria are among the most reactive organelles in the cell and consume nearly 90 % of the total oxygen content (Bolisetty and Jaimes 2013). Overall seven sites of superoxide production are known (Brand 2010). In the mitochondrial respiratory chain up to 5 % of the electrons flowing through the electron transport chain can be abstracted mainly at the levels of complex I (NADH/CoQ reductase) and complex III (ubiquinol/cytochrome c reductase) to form •O2 − (Novo and Parola 2008; Lenaz 2012). Superoxide, having a quite short half-life and low reactivity, is readily converted to hydrogen peroxide by superoxide dismutase (SOD), as depicted in Fig. 1. The further inactivation of H2O2 is catalyzed by a series of enzymes including glutathione peroxidase (GPX), peroxiredoxins (PRX), thioredoxin, or catalase in peroxisomes. Compared to •O2 −, hydrogen peroxide is electrophobic and can easily diffuse across membranes. It was found that the concentration of H2O2 in mitochondria is about 100 times higher compared to •O2 − (Cadenas and Davies 2000). In the presence of divalent metal ions H2O2 can further react in Fenton-type reactions to yield the highly reactive hydroxyl radical.

Fig. 1
figure 1

Main ROS (red) and RNS (blue) formed in cells. Superoxide is generated mainly in mitochondria or by NOX and LOX (a) and is further dismuted by superoxide dismutase (SOD) to form hydrogen peroxide (b). H2O2 is then further metabolized to water by glutathione peroxidases (GPX) or other enzymes (c). Reactive hydroxyl radicals are produced by Fenton-type reactions of H2O2 in the presence of Fe2+ (d). Nitric oxide is formed in the conversion of l-arginine to citrulline by nitric oxide synthases (NOS) (e) and readily forms peroxynitrite upon reaction with superoxide (f). Reaction of peroxynitrite with carbon dioxide yields nitrite and carbonate radicals (g), while peroxynitrous acid generates hydroxyl and nitrite radicals (h). Magnitudes of half-life times and the respective mobilities of the different species are given in parentheses

Full size image

Nitric oxide, which is formed through the conversion of l-arginine to citrulline by nitric oxide synthases (NOS), is a highly diffusible and relatively stable RNS. Upon reaction with •O2 − it forms the highly reactive peroxynitrite (ONOO−) at diffusion-controlled rates (Beckman et al. 1990; Radi et al. 2001). Peroxynitrite is a short-lived species and it can as well in its protonated form participate directly in oxidation reactions. Peroxynitrous acid itself immediately undergoes homolytic fission and forms a hydroxyl radical and nitrogen dioxide (Fig. 1, h) or a nitrate anion. Upon reaction with carbon dioxide, peroxynitrite forms [ONOOCO2 −] that decays to give either a carbonate radical and nitrogen dioxide (Fig. 1, g) or nitrate and carbon dioxide (Radi et al. 2001).

The magnitudes of half-life times and the respective mobilities of ROS (Pryor 1986; Karuppanapandian et al. 2011) and RNS (Pacher et al. 2007; Szabó et al. 2007) are depicted in Fig. 1.

3 Reactivity of ROS and RNS Versus Nucleic Acids

While H2O2 and •O2 − have been shown to be unreactive themselves with respect to nucleic acids (Lesko et al. 1980; Brawn and Fridovich 1981), •OH efficiently reacts at diffusion-controlled rates near the site of its formation with all four nucleobases and the (deoxy-)ribose sugar. Hydroxyl radical footprinting has been used to cleave nucleic acids to the nucleotide level and for revealing RNA folding pathways (Tullius and Dombroski 1985; Tullius and Greenbaum 2005). Mechanisms for the oxidation on the sugar have been widely explored for DNA (Pogozelski and Tullius 1998; Evans et al. 2004; Dedon 2007), whereas for RNA the existing data are limited to a few examples. Hydroxyl radicals can in principle react with sugar units by the abstraction of H-atoms of any of the five carbon atoms present. It was found that oxidation at the C1′-center of the ribose sugar is slower than for the 2′-deoxyribose sugar, presumably due to the polar effect of the 2′-hydroxyl group in RNA as the authors suggest (Neyhart et al. 1995). A similar behavior was found for the generation of C4′ radicals in model nucleotides: fragmentation of the C3′–O3′ bond by formation of a C3′ radical was shown to be far slower in the presence of a 2′-O substituent on the sugar moiety (Crich and Mo 1997). It has been further shown that iron-bleomycin (FeBLM), which efficiently cleaves DNA by the generation of C4′ radicals, is far less reactive but more sequence selective with RNA as the substrate (Hecht 1994). The data suggest, although not entirely conclusive, that the ribose unit seems to be more stable against hydroxyl radical attack compared to deoxyribose.

Upon reaction with RNA, peroxynitrite was found to form 8-oxo-rG and 8-nitro-rG (Masuda et al. 2002). Peroxynitrite seems to react with RNA according to different mechanisms: it was found that bicarbonate caused a dose-dependent increase in the formation of 8-nitro-rG whereas it produced no apparent effect on 8-oxo-rG formation. While 8-nitro-dG in DNA was found to depurinate rapidly (Yermilov et al. 1995), in RNA it was shown that 8-nitro-rG was far more stable (Masuda et al. 2002). Data on reactions of peroxynitrite with DNA suggest two different mechanisms: an increasing bicarbonate concentration seems to favor the formation of •NO2 (Fig. 1, g) and a low pH seems to induce the formation of hydroxyl radicals which can lead to strand breaks (Fig. 1, h) (Yermilov et al. 1996). Moreover it has been shown that peroxynitrite reacts mainly with dG nucleosides to form 8-oxo-dG and 8-nitro-dG while dA, dC, and dT showed only minimal reactivity (Burney et al. 1999).

Peroxynitrite has also been shown to induce strand cleavage in DNA (Salgo et al. 1995; Kennedy et al. 1997; Szabó and Ohshima 1997; Niles et al. 2006). According data for RNA cleavage is still missing, but it can be suggested that RNA cleavage happens to a lower extent. Overall the data suggest that the main oxidative damage seems to be coming from H2O2-generated •OH and peroxynitrite itself, as well from ONOO−-generated •OH because of the high diffusibility of H2O2 and peroxynitrite. Other forms of ROS hardly get into the vicinity of nucleic acids because of their limited half-life and their lacking moving capability.

Singlet oxygen, another strong oxidant generated by light (photooxidative stress), has so far only been shown to be able to oxidize rG (Schneider et al. 1993). Overall the data show that the reactivity of RNA under the influence of ROS and RNS is comparable to DNA with respect to base modifications. On the other hand, RNA might be less vulnerable against sugar backbone oxidation and subsequent strand cleavage compared to DNA.

4 Oxidative Lesions in RNA

The four main RNA oxidative lesions (Fig. 2) have been isolated from Torula yeast RNA (Yanagawa et al. 1990, 1992). Two other base lesions that are generated indirectly by oxidative stress through lipid peroxidation products are 1,N 6-ethenoadenosine (ε-rA) and 3,N 4-ethenocytidine (ε-rC) (El Ghissassi et al. 1995). All of these lesions have subsequently been synthesized and chemically introduced into RNA: 8-oxo-7,8-dihydroguanosine (Kim et al. 1998; Koga et al. 2013), 8-oxo-7,8-dihydroadenosine (Kim et al. 2002), 5-hydroxyuridine (Cui et al. 2009), 5-hydroxycytidine (Küpfer and Leumann 2011), and ε-rA and ε-rC (Srivastava et al. 1994; Calabretta and Leumann 2013).

Fig. 2
figure 2

Main stable oxidative lesions found in RNA: 8-oxo-7,8-dihydroguanosine (8-oxo-rG), 8-oxo-7,8-dihydroadenosine (8-oxo-rA), 5-hydroxycytidine (5-HO-rC), 5-hydroxyuridine (5-HO-rU), 1,N 6-ethenoadenosine (ε-rA), and 3,N 4-ethenocytidine (ε-rC). Further oxidation products of 8-oxo-purines: 2,6-diamino-4-hydroxy-5-formamidopyrimidine (FapyG) and 4,6-diamino-5-formamidopyrimidine (FapyA)

Full size image

These oxidative modifications described for RNA only represent a small fraction of the lesions found in DNA (Cadet et al. 2010). For pyrimidines potentially interesting oxidative lesions include 5,6-dihydroxy-5,6-dihydropyrimidines (pyrimidine glycols) and 5-hydroxymethyl- and 5-formylpyrimidines (hm5C, hm5U, f5C and f5U). While pyrimidine glycols are known precursors for 5-HO-rC and 5-HO-rU, 5-hydroxymethyl- and 5-formylpyrimidines could be potential oxidation products of 5-methylcytidine and 5-methyluridine, which are found in tRNA and mRNA (for m5C) or rRNA and tmRNA (for m5U) (Cantara et al. 2011). For 8-oxo-purines, the low redox potentials make them especially vulnerable to undergo further oxidation which leads to 4,6-diamino-5-formamidopyrimidine (FapyA) and 2,6-diamino-4-hydroxy-5-formamidopyrimidine nucleosides (FapyG) (Fig. 2) among a number of other derivatives for G including imidazolone, oxazolone, and spiroiminodihydantoin derivatives (Cadet et al. 2010). For FapyG it could be shown that the N-glycosidic bond of the isolated ribonucleoside is remarkably stable against anomerization and cleavage (Burgdorf and Carell 2002). Furthermore FapyG and FapyA were just recently quantified among other oxidative lesions in the analysis of RNA damage in Alzheimer’s disease patients and it can therefore be assumed that they are also present in the pool of oxidatively damaged RNA lesions (Bradley-Whitman and Lovell 2013). Also it is known from DNA that lipid peroxidation products lead to a wide variety of substituted etheno-, propano-, as well as malondialdehyde adducts. So far there is no data available on the hybridization properties or possible mutagenicity of these lesions in RNA.

4.1 Base-Pairing Properties of Lesion Containing Duplexes

The influence of the oxidized bases on duplex stability was elucidated by UV melting experiments with RNA duplexes and RNA/DNA hybrids. The results of these T m measurements in the same sequence context are summarized in Table 1.

Table 1 T m data (°C, 260 nm) of duplexes of oligoribonucleotide 5′-AUGCUXAGUCGA-3′ with 5′-UCGACUYAGCAU-3′ (RNA) or 5′-TCGACTYAGCAT-3′ (DNA) as complement; values in parentheses indicate the difference in T m compared to the corresponding natural duplex; duplex concentration: 2 μM in 10 mM NaH2PO4, 150 mM NaCl, pH 7.0
Full size table

5-Hydroxypyrimidines lead to no notable destabilization in neither of the duplexes compared to the natural cases and all formed base pairs are in accordance with Watson and Crick. An important fact to note is that in the case of 5-hydroxycytidine a biphasic melting behavior was observed (Küpfer and Leumann 2011).Footnote 1

8-Substituted purine nucleosides have long been known to adopt the syn conformation (Uesugi and Ikehara 1977). For 8-oxo-rG it has been shown that it exists mainly in the neutral 6,8-diketo form with a pK a of 8.5 for the 6-enolate-8-keto form and a pK a of 11.2 for the 6,8-dienolate form (Cho et al. 1990). In contrast to 8-oxo-rG, the pK a of 8.7 for the 8-enolate form of 8-oxo-rA is quite low (Cho and Evans 1991). The different tautomeric forms of the oxidized nucleosides with their respective pK a values are depicted in Fig. 3.

Fig. 3
figure 3

pK a values for the different tautomeric forms of 8-oxo-rG, 8-oxo-rA, 5-HO-rU, and 5-HO-rC. Values for the 8-oxo-purines are given for the ribonucleosides and values for the 5-hydroxypyrimidines are given for the deoxynucleosides

Full size image

For 8-oxoguanosine the duplexes are destabilized in the case where 8-oxo-rG pairs to pyrimidine bases. When 8-oxo-rG is paired against purine bases the duplexes gain in stability compared to the natural mismatch duplexes: for 8-oxo-rG/dG the duplex has the same stability as the mismatch duplex with rG/dG; for 8-oxo-rG pairing against rG, rA, and dA the duplexes gain considerably in stability. Especially in the case of 8-oxo-rG/rA the duplex stability equals almost the stability of the 8-oxo-rG/rC duplex. These results have just recently been confirmed showing that the T m of 8-oxorG/rA is even slightly higher than for the natural case (Koga et al. 2013).Footnote 2 This remarkable stabilization for 8-oxo-rG pairing rG and rA could be explained by data measured previously in DNA: it has been found that the Hoogsteen face of 8-oxo-dG pairs with dA forming a syn-8-oxo-dG/anti-dA base pair (Kouchakdjian et al. 1991; Gannett and Sura 1993). Such a syn-8-oxo-rG/anti-rA base pair would be expected to add to the overall duplex stability compared to an anti-rG/anti-rA mismatch. Furthermore it has been shown that rG can pair with the Hoogsteen face of rA forming an anti-rG/syn-rA base pair (Pan et al. 1999). It has been suggested that in a similar manner this could also give rise to an anti-8-oxo-rG/syn-rG base pair (Koga et al. 2013). The data suggest that 8-oxo-rG easily switches between anti and syn conformations to comfort the opposing base (Fig. 4).

Fig. 4
figure 4

Base-pairing modes for 8-oxo-rG: canonical 8-oxo-rG/rC base pair according to Watson and Crick (left), 8-oxo-rG/A base pair where 8-oxo-rG pairs with its Hoogsteen side (middle) and the 8-oxo-rG/G base pair where the Watson–Crick side of 8-oxo-rG pairs to the Hoogsteen side of rG

Full size image

For 8-oxoadenosine moderate to strong destabilization of the duplexes was found with the strongest decrease for the pairing to pyrimidine bases in both RNA/RNA and RNA/DNA duplexes. In DNA 8-oxo-dA was shown to pair dT according to Watson and Crick (Guschlbauer et al. 1991). In analogy to 8-oxo-dG it has also been found that 8-oxo-dA can pair to dG via its Hoogsteen face forming a syn-8-oxo-dA/anti-dG base pair (Leonard et al. 1992). Later on base pairs of syn-8-oxo-dA with dC and dA have been postulated (Kamiya et al. 1995). Nevertheless 8-oxoadenosine still discriminates the binding partners in favor of the matched base.

These findings suggest that all oxidized bases are capable of forming stable base pairs but with a lower discrimination of mismatches. This might subsequently lead to the synthesis of modified proteins via the interaction of oxidatively damaged mRNA with tRNA as was suggested earlier (Kong and Lin 2010).

Etheno-adenosine and -cytidine lesions show major destabilization in RNA duplexes and RNA/DNA heteroduplexes because of the loss of their ability to form standard Watson–Crick hydrogen bonds. Duplexes formed with a G opposite the lesion are slightly more stable than the others, most likely because of stronger base stacking interactions of the etheno bases (Calabretta and Leumann 2013).

4.2 Mutagenicity of Oxidatively Damaged Nucleobases

The potential mutagenicity of oxidized RNA bases was elucidated by reverse transcription assays. In these assays a DNA primer is elongated on a modified RNA template using reverse transcriptases and the efficiency of the incorporation of natural 2′-O-deoxynucleoside monophosphates (dNMP) is determined.

4.2.1 5-Hydroxypyrimidines

For a 5-HO-rU containing RNA template it was found that reverse transcriptasesFootnote 3 incorporated preferentially dAMP opposite the lesion. Notable misincorporation of dGMP was dependent on the reverse transcriptase used: MMLV-RT incorporated up to one-third, Superscript™II-RT up to one-fourth, and AMV only trace amounts of dGMP compared to incorporated dAMP (Cui et al. 2009).

In the case of a 5-HO-rC containing template there was found substantial misincorporation of dAMP besides the expected incorporation of dGMP: HIV1-RT incorporated dAMP almost as efficiently as dGMP while for the AMV-RT and MMLV-RT this process is two- to threefold less efficient. Interestingly, HIV1-RT also incorporated dTMP with about half the efficiency of dGMP (Küpfer and Leumann 2011). In DNA 5-HO-dU and 5-HO-dC have also been tested for potential mutagenicity during DNA replication and it was found that 5-HO-dU and 5-HO-dC cause C/T transition mutations (Kreutzer and Essigmann 1998; Suen et al. 1999). The mutagenicity has been associated with tautomeric variability of the oxidized base: while the amino form of 5-hydroxycytidine pairs to guanosine, the imino form would be able to pair to adenosine (Suen et al. 1999; La Francois et al. 2000).

4.2.2 8-Oxo-7,8-Dihydropurines

Reverse transcription assays with 8-oxo-rG containing templates were performed using MMLV-RT and AMV-RT (Kim et al. 1999) and HIV-RT and RAV2-RT (Kim et al. 2004). While MMLV-RT seemed to incorporate preferentially dTMP opposite 8-oxo-rG, AMV inserted the correct dCMP residue. Using HIV-1 RT and RAV2-RT, almost no dAMP and dGMP were incorporated opposite 8-oxo-rG. HIV-1 RT incorporated almost exclusively dCMP opposite 8-oxo-rG. Using RAV2-RT, dCMP was only incorporated 1.5-fold of dTMP opposite 8-oxo-rG. These results have to be looked at carefully since there is evidence that the authors might have used the not fully deprotected 8-oxo-rG-DMC derivative (see Footnote 2). It was recently shown that duplexes of 8-oxo-rG-DMC containing RNA with DNA have considerably lower T m values compared to 8-oxo-rG containing heteroduplexes and clearly discriminate in favor of dC opposite 8-oxo-rG-DMC (Koga et al. 2013).

Similar assays with 8-oxo-rA containing RNA were performed using AMV-RT, MMLV-RT, and RAV2-RT (Kim et al. 2002). All reverse transcriptases incorporated preferentially dTMP opposite the lesion. AMV-RT incorporated also dGMP and dAMP to a three- and fourfold lesser extent and MMLV- and RAV2-RT incorporated dGMP to a fivefold lesser extent compared to dTMP.

In DNA 8-oxo-dG is known to lead to G/T transversions in vitro (Wood et al. 1990) and in vivo (Moriya et al. 1991; Cheng et al. 1992). DNA polymerases incorporate dAMP and dCMP differently opposite 8-oxo-dG: DNA pol α, pol δ, and pol III preferentially incorporate dAMP, whereas DNA pol β and pol I incorporate dCMP (Wang et al. 1998). Replication of 8-oxo-dA containing DNA templates revealed that exclusively dTMP is inserted opposite the lesion by bacterial DNA polymerases pol I (Klenow fragment) and Taq DNA pol (Guschlbauer et al. 1991). Nonetheless, 8-oxo-dA is found to be mutagenic in mammalian cells where DNA pol α is found to misinsert dGMP and recombinant rat DNA pol β can misinsert dAMP and dGMP (Shibutani et al. 1993; Kamiya et al. 1995).

4.2.3 ε-Adenosine and ε-Cytidine

For reverse transcription assays it was found that only the error-prone HIV-1 RT was able to bypass the lesions. In case of ε-rA containing RNA templates HIV-1 RT incorporated deoxynucleoside triphosphates in the order dAMP > dGMP ≫ dCMP and dTMP. For ε-rC containing templates the incorporation order observed was dAMP > dTMP ≫ dCMP > dGMP. In both cases full-length cDNA was obtained in the presence of all deoxynucleoside triphosphates (Calabretta and Leumann 2013).

The mutagenicity of ε-dA and ε-dC has been shown to be strongly mutagenic when tested in mammalian cell lines, whereas in E. coli minor mutagenic properties were observed. In mammalian cells predominantly A/G transitions and A/T transversions were found for ε-dA and C/T transitions for ε-dC (Pandya and Moriya 1996; Levine et al. 2000). In E. coli only minor mutagenic properties were found for ε-dA (Basu et al. 1993; Pandya and Moriya 1996) and only low frequent C/A transversions for ε-dC (Moriya et al. 1994).

4.2.4 Oxidative Damage to the Nucleotide Pool

Oxidative damage to the nucleotide pool can be induced by direct oxidation of ribonucleotides or by degradation of oxidatively damaged RNA. Leukocytes that were put under oxidative stress revealed increased levels of 8-oxo-rG in the nucleotide pool (Shen et al. 2000). Quality control of RNA synthesis is thus crucial and mechanisms have evolved to reduce incorporation of oxidatively damaged ribonucleotides. It was shown that 8-oxo-rGTP was incorporated into RNA by E. coli polymerase at a rate of ~10 % of that of rGTP (Taddei et al. 1997). To prevent the incorporation of 8-oxo-rGTP, MutT protein in E. coli degrades the di- and triphosphates of 8-oxo-rG to the monophosphate 8-oxo-rGMP (Ito et al. 2005). The reutilization of 8-oxo-rGMP is prevented since it was found that guanylate kinase does not phosphorylate 8-oxo-rGMP to 8-oxo-rGDP and therefore no 8-oxo-rGTP can be synthesized (Sekiguchi et al. 2013). A similar pathway has been disclosed for human cells. Here 8-oxo-rGTP is incorporated by human RNA polymerase II at a rate of 2 % of that of rGTP (Hayakawa et al. 1999). MTH1 protein has been shown to hydrolyze 8-oxo-rGTP to 8-oxo-rGMP. Additionally it could be shown that MTH5 and NUDT5 hydrolyze 8-oxo-rGDP to 8-oxo-rGMP (Ishibashi et al. 2005). Normally GTP is synthesized from GDP through phosphorylation by nucleotide diphosphate kinase (ND kinase) and GDP is itself produced through phosphorylation of GMP by guanylate kinase (GK). While ND kinase has been found to phosphorylate 8-oxo-rGDP as well, GK is not able to phosphorylate 8-oxo-rGMP, thus preventing the synthesis and incorporation of 8-oxo-rGTP into RNA (Hayakawa et al. 1999).

5 Consequences of Oxidatively Damaged RNA

During the last few years increasing evidence emerged that oxidatively damaged RNA causes severe effects on cellular function. Messenger RNA oxidation in vitro was shown to cause reduction in protein expression and ribosome stalling on the transcripts (Shan et al. 2007). In another study the authors found that oxidized mRNA induces translation errors (Tanaka et al. 2007). It was also shown that nitric oxide and peroxynitrite accelerate mutation of RNA viruses in vitro and in vivo, probably via the formation of 8-oxo-rG and 8-nitro-rG (Akaike et al. 2000).

Preliminary studies on oxidatively damaged ribosomes have shown that if the catalytic adenine, A2451 in the peptidyl transferase center (PTC) of the 23S ribosome subunit of E. coli, is replaced by an 8-oxo-adenine unit, peptide synthesis is abolished (Polacek, unpublished results). This is quite remarkable since it has been shown earlier that several base modifications on A2451 only have a minor influence on the activity of the PTC as long as the 2′-hydroxyl group of the ribose is present (Erlacher and Polacek 2008).

Ongoing studies on the efficiency of translation on site specifically damaged mRNA show that in rabbit reticulocyte lysate ribosome stalling as well as read-through can be observed. Whether oxidized mRNA leads to mutations in the synthesized proteins is yet unclear (Leumann, unpublished results).

6 Fate of Oxidized RNA

High levels of oxidized RNA are not tolerated by cells. Several studies have shown that levels of oxidized RNA dropped after removal of oxidative stress inducing conditions (Shen et al. 2000; Kajitani et al. 2006). Recently it could be shown that H2O2-induced 8-oxo-rG levels in E. coli cells dropped rapidly after switching to a H2O2-free medium (Liu et al. 2012). These results suggest that oxidatively damaged RNA is actively removed from total RNA. In theory this might happen either by repair of the oxidized lesions or by degradation. So far there is no strong evidence supporting the existence of RNA repair mechanisms. The only repair mechanism reported for RNA is the dealkylation of alkylated RNA by DNA demethylases (Aas et al. 2003). Therefore it seems more likely that oxidatively damaged RNA is removed by degradation.

Removal of damaged RNA can be achieved by ribonucleases. Polynucleotide phosphorylase (PNPase) is a 3′–5′ exoribonuclease that is important for the degradation of RNA. In E. coli it has been shown that PNPase specifically binds 8-oxo-rG containing RNA with a higher affinity than undamaged RNA (Hayakawa et al. 2001). Undamaged RNA has been shown to be degraded efficiently while 8-oxo-rG containing RNA is tightly bound to PNPase and thus protected from nuclease attack. Human PNPase is located mainly in mitochondria and also specifically binds 8-oxo-rG containing RNA. Under induced oxidative stress, levels of hPNPase decreased rapidly while the levels of other proteins were left unchanged (Hayakawa and Sekiguchi 2006). The authors speculate that the PNP might thus play an important role in maintaining the high fidelity of translation by sequestering the oxidatively damaged RNA molecules and possibly direct them to degradation. In a further study it has been shown that PNPase-deficient cells are hypersensitive to H2O2 and less viable than wild-type cells (Wu et al. 2009). Introduction of a plasmid-borne PNP gene restores the viability against oxidative stress. Furthermore overexpression of hPNPase in HeLa cells reduces 8-oxo-rG in RNA and improves cell viability against H2O2 treatment (Wu and Li 2008).

Another protein that shows a similar binding capacity for 8-oxo-rG containing RNA is the mammalian Y box-binding protein 1 (YB-1 protein). YB-1 has multiple regulatory activities and was originally identified as a transcription factor which binds to the Y box. YB-1 is a major core protein of messenger ribonucleoprotein (mRNP) and has a weak RNA binding capacity. It was found that YB-1 selectively binds 8-oxo-rG containing RNA to form stable complexes. Cells of E. coli expressing YB-1 were found to acquire resistance against oxidative stress. These data imply that YB-1 protein may be involved in sequestering oxidatively damaged RNA from normal cellular processes (Hayakawa et al. 2002).

Oxidized nucleobases are known to have low one-electron redox potentials and are therefore prone to undergo further oxidation processes (Yanagawa et al. 1992). In DNA 8-oxo-dG and 8-nitro-dG showed a higher reactivity with ONOO− than dG itself (Burney et al. 1999).

That such ongoing oxidation induces abasic sites could be shown by the reaction of RNA treated under Fenton conditions with the aldehyde reactive probe (ARP) (Tanaka et al. 2011a, b). The ARP has been shown earlier to react specifically with the aldehyde functions of abasic sites and formylcytidine but no other oxidative lesion, e.g., 8-oxo-rA, 8-oxo-rG, or pyrimidine glycols (Ide et al. 1993). RNA abasic sites have been shown to be remarkably more stable than DNA abasic sites (Küpfer and Leumann 2007). That such abasic sites can be substrates for enzymatic degradation has been shown by the disclosure of a new class of enzymes: the ribosomal RNA apurinic site-specific lyase (RALyase) (Ogasawara et al. 1999; Ito et al. 2002). These enzymes exhibited a very high specificity for the cleavage of the abasic site in the sarcin–ricin loop of intact ribosomes. On the contrary, RALyase failed to cleave DNA abasic sites as well as protein-free abasic rRNA. Thus, rigorous proof for broader RNA-specific lyase activity is still missing, as no other members of this class of enzymes could be identified so far.

7 Redox Chemistry of Oxidized Nucleosides

The one-electron redox potentials of the four main oxidative lesions have been reported to be far lower than for the standard nucleosides: 8-oxo-rG: 0.58 V (1.29 V for rG), 8-oxo-rA: 0.92 V (1.42 V for rA), 5-HO-rC: 0.62 V (1.6 V for dC), and 5-HO-rU: 0.64 V (1.7 V for dT) versus NHE (Yanagawa et al. 1992; Steenken and Jovanovic 1997); for 8-oxo-dG: 0.74 (Steenken et al. 2000). So far two examples of exploiting the low redox potentials of hydroxynucleosides have been published.

For oxidized RNA nucleosides it could be shown that they catalyze the oxidoreduction of NADH and K3Fe(CN)6 (Yanagawa et al. 1992). In this system 5-HO-rC showed by far the highest activity followed by 5-HO-rU and 8-oxo-rG which showed roughly half and a quarter of the reactivity compared to 5-HO-rC. The authors conclude that these hydroxyribonucleosides might also serve as redox cofactors at active sites of ribozymes with redox activity. So far this was not experimentally proven.

8-Oxo-dG, incorporated in proximity to a cyclobutane pyrimidine dimer (CPD), can mimic the function of a flavin in photorepair. The study supports a photolyase-type mechanism in which in a first step the excited state of 8-oxo-dG transfers an electron to the CPD effecting cleavage of the cyclobutane and in a second step the back electron transfer regenerates 8-oxo-dG and the repaired pyrimidines (Nguyen and Burrows 2011, 2012b). The free tri-acetylated 8-oxo-rG and ribofuranosyluric acid nucleosides also showed enhanced repair of T=T and U=U CPDs (Nguyen and Burrows 2012a).

The possible role of oxidized RNA nucleosides as redox catalysts in a “RNA world” is not known. Their use in ribozymes might be a future prospect.

8 Conclusions and Perspective

For DNA a large number of oxidative lesions, their metabolism, and distinct repair mechanisms have been reported in the past. As reviewed here, only half a dozen lesions that arise under oxidative stress have been described, along with their biophysical properties for RNA. Nevertheless, in accordance with DNA it can be assumed that more lesions exist which are still lacking a detailed description.

While 85 % of the genomic DNA is transcribed into RNA, the function of 97 % of this RNA is still largely unknown. Thus, oxidatively damaged RNA cannot only interfere with the translation machinery but possibly also with a large number of mostly unknown regulation mechanisms of ncRNA. The number of such long intergenic noncoding RNAs (lincRNA) is still growing and their functions are yet elusive (Hangauer et al. 2013).

The data show that RNA is at least as prone to oxidation as DNA due to its spatial vicinity to the sources of ROS and RNS and the (so far) absence of a distinct repair machinery. Moreover the amount of RNA in a cell is considerably higher than that of DNA. Within this context, RNA cannot be viewed only as a transient messenger anymore: average lifetimes of RNA are probably long enough to have an impact on different cell mechanisms.

Oxidatively damaged RNA has been linked over the past decade to a still increasing number of neurodegenerative and other diseases (Poulsen et al. 2012). The data show that the oxidative damage is rather an early event in the development of such diseases than a consequence of cell decay. A deeper understanding of the biophysical properties of oxidative lesions in RNA would undoubtedly help to understand the more complex biological impact and mechanisms of such diseases.