Localized control of oxidized RNA

The oxidation of biological molecules by reactive oxygen species (ROS) can render them inactive or toxic (Holmstrom and Finkel, 2014). Although DNA, lipids and proteins have long been considered as crucial targets of oxidation, recent evidence suggests that RNA oxidation has roles in oxidative stress, aging and certain neurodegenerative diseases (Wurtmann and Wolin, 2009). For example, the translation of oxidized mRNAs generates aberrant proteins (Nunomura et al., 2009; Tanaka et al., 2007) and elevated RNA oxidation in neurons is associated with Alzheimer disease, Parkinson disease and amyotrophic lateral sclerosis (Nunomura et al., 2009).

Molecular quality control systems specifically recognize oxidized molecules and subject them to repair, sequestration or degradation (Stoecklin and Bukau, 2013). Quality control systems have been characterized for oxidized DNA, proteins and lipids, but not for oxidized RNA (Wurtmann and Wolin, 2009). However, oxidized RNA quality control might involve YB-1 (also known as YBX1), Auf1 (also known as HNRNPD) and polynucleotide phosphorylase (PNPase) because these proteins bind oxidized RNA and control its accumulation (Li et al., 2014). Moreover, the intracellular location(s) of oxidized RNA quality control in eukaryotic cells have been hypothesized to include stress granules and processing bodies (Thomas et al., 2011; Walters and Parker, 2014; Wurtmann and Wolin, 2009). For stress granules, this is supported by their recruitment of YB-1 and Auf1, and their formation under oxidative stress conditions when excess ROS cause oxidative damage (Bravard et al., 2010; Onishi et al., 2008; Tanaka et al., 2014). Similarly, processing bodies increase in size and number during oxidative stress, and they contain mRNA degradation machinery, a component of RNA quality control (Thomas et al., 2011; Walters and Parker, 2014). Although our understanding of the quality control of oxidized RNA is advancing, the precise intracellular locations are still unknown.

In the chloroplasts of plants and green algae, antioxidant systems and molecular quality control are particularly important because photosynthesis produces ROS as hydrogen peroxide (H₂O₂), singlet oxygen, superoxide and the hydroxyl radical (Foyer and Shigeoka, 2011). Moreover, chloroplasts have a genome and a gene expression system that are targets of oxidation and mutagenesis by ROS (Wurtmann and Wolin, 2009; Zheng et al., 2014). Although chloroplasts have known quality control systems for oxidized DNA, proteins and lipids (Apel and Hirt, 2004), nothing is known about how they manage oxidized RNA.

An avenue to study RNA quality control and its localization in chloroplasts arose with our discovery of stress-granule-like bodies that form during oxidative stress in the chloroplast of the unicellular green alga Chlamydomonas reinhardtii (Uniacke and Zerges, 2008). These ‘chloroplast stress granules’ (cpSGs) were seen by confocal fluorescence microscopy as stress-induced foci containing mRNAs encoded by the chloroplast genome and stress granule marker proteins. cpSGs form at the inner perimeter of the pyrenoid, a micro-compartment in the chloroplasts of most algae where CO₂ fixation is catalysed by the Calvin cycle enzyme ribulose bisphosphate carboxylase/oxygenase (Rubisco) (McKay and Gibbs, 1991). cpSGs are enriched in the large subunit of Rubisco (RBCL), but not the Rubisco holoenzyme; they are not similarly enriched in the small subunit of this complex, RBCS (Uniacke and Zerges, 2008). The RBCL in cpSGs might function in RNA metabolism because, under oxidizing conditions, the protein has been shown to acquire an RNA-binding activity and form aggregates that might represent a biochemical form of cpSGs (Knopf and Shapira, 2005; Yosef et al., 2004). Taken together, these results suggest that cpSGs, RBCL and the pyrenoid have some undefined role(s) in chloroplast RNA metabolism during stress. This hypothesis and the previously reported role of RBCL in autoregulatory feedback translational repression are mutually compatible (Cohen et al., 2006).

Here, we show that oxidized RNA localizes to the pyrenoid by analysing immunofluorescence microscopy experiments performed using an antibody against a major oxidized base in RNA and DNA, 8-hydryoxyguanine (8-oxoG) (Nunomura et al., 1999; Wurtmann and Wolin, 2009). Results of genetic analyses support a requirement for RBCL, but not Rubisco, in the management of oxidized RNA. An RBCL pool that could carry out such a ‘moonlighting’ function is revealed by results of biochemical fractionation experiments. We extend our discovery of the localization of oxidized RNA to human cells by showing that HeLa cells under arsenite-induced oxidative stress localize oxidized RNA to cytoplasmic foci that are neither stress granules nor processing bodies. Our results begin to shed light on how oxidized RNA is managed in an algal chloroplast and suggest that the compartmentalization of oxidized RNA quality control is a general phenomenon.

We characterized the distribution of oxidized RNA in Chlamydomonas cells by immunofluorescence microscopy using a commercial antibody against 8-oxoG. The immunofluorescence signal was specific to 8-oxoG because it was eliminated when the antibody was incubated with 8-oxoG prior to staining (Fig. 1A,B).

Inspection of cell images revealed that there was a higher 8-oxoG immunofluorescence signal in the chloroplast than in the central region with the nucleus and most cytosolic compartments (Fig. 1A). The 8-oxoG immunofluorescence signal was seen in two distinct patterns; throughout the pyrenoid and in punctate foci located in or near the chloroplast. In order to determine whether these patterns represent 8-oxoG in DNA or RNA, we used two approaches. First, we co-stained cells with DAPI, to visualize the nucleus and the chloroplast nucleoids (the latter contain the multicopy chloroplast genome). These DNA-containing structures only showed weak immunofluorescence staining for 8-oxoG (Fig. 1A) and were clearly distinct from the 8-oxoG immunofluorescence staining of the pyrenoid and foci. Therefore, most of the 8-oxoG detected with this method is not in the genomic DNA of either the nucleus or chloroplast. Second, we asked whether RNase or DNase treatment altered the 8-oxoG immunofluorescence staining of the pyrenoid, the foci or both. Cells (fixed and permeabilized) were exposed to either RNase or DNase prior to staining with the 8-oxoG antibody and DAPI. The 8-oxoG immunofluorescence signal in the pyrenoid was eliminated by treatment with RNase, but not DNase, revealing that it represents oxidized RNA (Fig. 1C). By contrast the 8-oxoG immunofluorescence signal in most punctate foci was eliminated by treatment with DNase, but not with RNase, revealing that it represents oxidized DNA (Fig. 1D). These foci could be the DNA of mitochondria, which are localized adjacent to the chloroplast (Rasala et al., 2014), and were not explored further. DNase treatment often generated a diffuse 8-oxoG immunofluorescence signal throughout the chloroplast for unknown reasons (Fig. 1D).

We focused on the 8-oxoG immunofluorescence staining of the pyrenoid because it represents oxidized RNA and was atypical; we did not see this pattern for any of the four chloroplast mRNAs or nine chloroplast proteins that we analysed previously (not including RBCL and RBCS, which were seen in the pyrenoid) (Bohne et al., 2013; Schottkowski et al., 2012; Uniacke and Zerges, 2007, 2008, 2009). Therefore, 8-oxoG immunofluorescence staining of the pyrenoid represents the specific localization of oxidized RNA and is not due, for example, to nonspecific entry into the pyrenoid of RNA from the surrounding stroma (the chloroplast compartment analogous to the cytoplasm).

When cells were exposed to H₂O₂, a ROS used to induce oxidative stress, the percentage of cells in which the pyrenoid was stained for 8-oxoG increased from 44% (n=215) to 64% (n=189). This result suggests that oxidized RNA localizes to the pyrenoid for quality control.

Oxidized RNA did not localize to cpSGs, which were stained for two marker proteins (RBCL or a protein of the 30S subunit of the chloroplast ribosome; Fig. 1E,F) (Uniacke and Zerges, 2008). cpSGs formed during exposure to H₂O₂, but the 8-oxoG immunofluorescence signal remained diffuse throughout the pyrenoid (Fig. 1E,F). This result suggests that cpSGs do not accumulate oxidized RNA.

To determine whether the compartmentalization of oxidized RNA could be a general phenomenon, we asked whether it occurs in human cells. HeLa cells were treated with the oxidative stressor arsenite and then stained for 8-oxoG. The 8-oxoG immunofluorescence signal was detected in multiple cytoplasmic foci in more than 40% of these cells, as compared to 5% of untreated cells (Fig. 2A). These foci were distinct from stress granules and processing bodies, which were stained for marker proteins specific to each of these RNA granule types (Fig. 2A,B). They were also not within mitochondria and therefore could not be oxidized RNA or DNA of the mitochondrial genetic system (Fig. 2C). Nuclei also had 8-oxoG-containing bodies, which appear to be nucleoli (Lee et al., 2014). RNA with 8-oxoG, and not DNA, was detected in these bodies, both cytoplasmic and nuclear, because they did not stain for DNA with DAPI and they were not detected in RNase-treated cells (Fig. 2A). These results reveal that mammalian cells compartmentalize oxidized RNA to cytoplasmic structures, which we name here ‘oxidized RNA bodies’ (ORBs). These results also substantiate the presence of oxidized RNA in nucleoli, as reported previously (Nunomura et al., 1999).

Returning to the chloroplast, we hypothesized that RBCL functions in the quality control of oxidized RNA based on the localization of 8-oxoG RNA to the pyrenoid, the fact that RBCL is a major protein of the pyrenoid and the evidence that RBCL has a dual function involving RNA during stress (Yosef et al., 2004). To test this hypothesis, we asked whether an RBCL-knockout mutant, MX3312 (hereafter ΔrbcL), had an elevated level of 8-oxoG in RNA. The results of immuno-slot-blot analyses revealed a significantly higher mean level of 8-oxoG in RNA from ΔrbcL than in RNA from the wild-type strain (Fig. 3A). The elevated level of oxidized RNA in ΔrbcL is not due to Rubisco deficiency because it was not detected in another Rubisco-deficient mutant, which lacks RBCS, but has RBCL (ΔRBCS-CAL005.01.13, hereafter ΔRBCS) (Fig. 3A) (Dent et al., 2005). Therefore, our results support a role for RBCL in the management of oxidized RNA that either involves its known RNA-binding activity or is indirect, occurring through other factors that interact with RNA (Yosef et al., 2004). This oxidized RNA is probably in the chloroplast because RBCL is a chloroplast protein and most 8-oxoG RNA was detected in this organelle (Fig. 1).

We were surprised to find that RNA from ΔRBCS cells showed an even lower mean level of 8-oxoG than did RNA of the wild-type strain (Fig. 3A). This phenotype is not due to the genetic background because transformation of ΔRBCS with a wild-type copy of RBCS2 restored the mean level of 8-oxoG RNA to nearly that of the wild-type strain (Fig. 3A). The low level of 8-oxoG in RNA from ΔRBCS is also not due to a deficiency for some unknown RBCS function because a double mutant lacking both RBCL and RBCS was found to have a high mean level of 8-oxoG in RNA, similar to that of ΔrbcL (Fig. 3A). Therefore, the low 8-oxoG RNA level in ΔRBCS is RBCL-dependent and does not reflect some unknown function of RBCS.

Exposure of the cells to H₂O₂ did not significantly change the mean level of 8-oxoG RNA from wild-type, ΔrbcL or ΔRBCS strains (Fig. 3A). That these RBCL-dependent effects were detected in cells under non-stress or stress conditions supports a constitutive nature of the proposed moonlighting function of RBCL. The moonlighting function of RBCL probably does not involve oxidized DNA because similar mean levels of 8-oxoG were detected in total DNA from ΔrbcL and the wild-type strain (Fig. 3B). Although the level of 8-oxoG was higher in DNA from ΔrbcL than in DNA from ΔRBCS, this difference was only 1.4-fold versus 10-fold for the same comparison of 8-oxoG in RNA (described above) (Fig. 3A). Moreover, as stated above, chloroplast nucleoids did not stain for 8-oxoG by immunofluorescence microscopy (Fig. 1A,C). Treatment of cultures with H₂O₂ increased the variability in the level of 8-oxoG in DNA between biological replicate experiments but did not increase the mean levels relative to DNA from the non-treated cultures (Fig. 3B). No differences in mean levels of oxidized protein were found between the wild-type strain, ΔrbcL or ΔRBCS mutant strains, when measured as carbonylated amino acid residues in total protein (Fig. 3C). Therefore, our results do not support a function of RBCL related to oxidized DNA or oxidized protein.

Our evidence of a Rubisco-independent moonlighting function of RBCL suggests the existence of an RBCL pool that is distinct from the RBCL in the Rubisco holoenzyme. Therefore, we carried out biochemical fractionation experiments to identify such a pool. In addition to the major soluble RBCL form of the Rubisco holoenzyme, three forms of RBCL are known. Two insoluble forms of RBCL have been detected in cells undergoing oxidative stress or senescence: one was in insoluble aggregates (Knopf and Shapira, 2005) and the other was associated with membranes (Knopf and Shapira, 2005; Marin-Navarro and Moreno, 2006). In addition, RBCL that is newly synthesized and unassembled has been proposed to autoregulate rbcL translation during oxidative stress through its RNA-binding activity (Cohen et al., 2005).

We developed a differential centrifugation scheme to separate soluble and insoluble proteins into three fractions (Fig. 4B): fraction S16 has soluble proteins; fraction P16-TS has insoluble proteins that can be solubilized by Triton X-100 (e.g. membrane proteins), and fraction P16-TI has insoluble proteins that cannot be solubilized by Triton X-100. Analyses of the fractions from the wild-type strain revealed that S16 had both RBCL and RBCS, representing the Rubisco holoenzyme (Fig. 4C). Detection of other RBCL forms in fractions of wild-type strains was hampered by contamination of most subcellular fractions by the Rubisco holoenzyme due to its extreme high abundance (Spreitzer, 2003). For example, fraction P16-TI from the wild-type strain contained both RBCL and RBCS, presumably in the Rubisco holoenzyme (Fig. 4C). This contamination was not a problem with ΔRBCS because it lacks the Rubisco holoenzyme, and it accumulates RBCL to only 1–10% of the wild-type level (Fig. 4A) and has an enhanced level of the RBCL function relating to oxidized RNA (Fig. 3A). Therefore, we presumed that most or all RBCL in the ΔRBCS stain represents a pool dedicated to this function. Analysis of the fractions obtained from ΔRBCS cells revealed RBCL primarily in P16-TI and, as expected (because this mutant lacks soluble Rubisco holoenzyme), only in trace amounts in S16 (Fig. 4D). The RBCL in P16-TI from ΔRBCS is not newly synthesized because this form is soluble and fractionated to S16, as revealed by ³⁵S-pulse-labelling (Fig. 4E). Furthermore, this RBCL is probably not insoluble due to membrane association, because it was not present together with detergent-solubilized membrane proteins in P16-TS (Fig. 4C,D). These results suggest that the major form of RBCL in ΔRBCS differs in physicochemical properties from the soluble RBCL pools of the Rubisco holoenzyme and the newly synthesized protein. Therefore, this RBCL could represent a pool of the protein that is dedicated to its function relating to oxidized RNA in the chloroplast.

To address the possibility that the RBCL in the P16-TI fraction from ΔRBCS represents the form in cpSGs, we asked whether P16-TI has another feature of cpSGs; enrichment in the 30S subunit of the chloroplast ribosome over the 50S subunit (Uniacke and Zerges, 2008). Indeed, results of immunoblot analyses revealed a greater proportion of the 30S subunit pool in P16-TI fractions, whereas the 50S subunit pool was not similarly enriched. This difference was observed in analyses of both ΔRBCS and the wild-type strain (Fig. 4C,D). However, although stress induces the recruitment of RBCL and 30S subunits to cpSGs, neither shifted from the soluble pool (S16) to P16-TI when cells were exposed to H₂O₂ (Fig. 4C,D). This result provides further support of a constitutive nature of the proposed moonlighting function of RBCL related to oxidized RNA, and raises the possibility that cpSGs are a manifestation of an RBCL-containing ribonucleoprotein (RNP) particle that exists at the submicroscopic level under non-stress conditions (see Discussion).

To obtain evidence that the differential levels of oxidized RNA in the wild-type and mutant strains (Fig. 3A) have relevance in vivo, we tested ΔrbcL for impaired tolerance to stress induced by exogenous H₂O₂. ΔRBCS was again used to control for the effects of Rubisco deficiency. Cultures of the wild-type strain, ΔrbcL and ΔRBCS were exposed to a toxic concentration of H₂O₂ (4.0 mM) and the percentage of live cells was monitored over 8 h (Fig. 5A). The results revealed that ΔrbcL cells died significantly faster than did the wild-type cells. Therefore, the elevated level of oxidized RNA in ΔrbcL correlates with impaired H₂O₂ tolerance. ΔRBCS exhibited H₂O₂ hypertolerance, measured both as cell survival (Fig. 5A) and in a more stringent assay of viability, the percentage of colony forming units compared with the initial values prior to H₂O₂ exposure (Fig. 5B). Therefore, in ΔRBCS, the low mean level of oxidized RNA and H₂O₂ hypertolerance are both opposite to the loss-of-function phenotypes for these traits in ΔrbcL. Similarly, the rescued ΔRBCS mutant (by transformation with RBCS2) showed wild-type H₂O₂ tolerance (Fig. 5C) and a wild-type mean level of oxidized RNA (Fig. 3A). Finally, like ΔrbcL, the double mutant for both RBCL and RBCS showed impaired H₂O₂ tolerance and a high oxidized RNA level (Fig. 5D). These differences in H₂O₂ tolerance did not reflect inherent differences in growth rate, transcript levels of oxidative stress marker genes or the rate of H₂O₂ degradation in the medium (Fig. S1). Thus, the biological relevance of the different levels of oxidized RNA detected in vitro is supported by the role of RBCL in H₂O₂ tolerance in vivo.

Our results reveal that oxidized RNA is compartmentalized in the pyrenoid of an algal chloroplast and cytoplasmic ORBs in human tissue culture (HeLa) cells (Figs 1 and 2). These findings in such phylogenetically distant genetic systems suggest that the compartmentalization of oxidized RNA has fundamental significance. Compartmentalization of DNA and protein quality control is well documented and believed to have several functions: the sequestration of damaged molecules prevents them from interfering with the processes in which they normally function; it prevents the degradation or attempted repair of undamaged substrates; and, finally, compartmentalization could enhance local concentrations of substrate molecules and quality control factors to establish thermodynamic parameters that favour forward reactions, for example, in repair or degradation (Adjibade and Mazroui, 2014; Stoecklin and Bukau, 2013; Walters and Parker, 2014). Our results open avenues to study the role of compartmentalization in RNA quality control.

Our results show that RBCL has a moonlighting function related to oxidized RNA in the chloroplast (Fig. 3) and H₂O₂ tolerance (Fig. 5), and that this function is independent of its role as a subunit of the Rubisco holoenzyme. In ΔRBCS, the low level of oxidized RNA and high H₂O₂ tolerance could reflect an enhanced level of this moonlighting function because these phenotypes are dependent on RBCL. For example, more RBCL might be available for the management of oxidized RNA when it cannot be assembled into the Rubisco holoenzyme. The Rubisco-independency of the proposed oxidized-RNA-related function of RBCL could explain the evolutionary retention of RBCL in plant and algal lineages that have lost photosynthesis (Krause, 2008). We also identify an insoluble RBCL pool that is not part of the Rubisco holoenzyme and could carry out this dual moonlighting function (Fig. 4). Other evidence of a Rubisco-independent RBCL pool has been reported in Chlamydomonas recently; RBCL has been shown to accumulate to a level that is several fold above the equal stoichiometric amounts with RBCS that are required in the Rubisco holoenzyme complex (Recuenco-Muñoz et al., 2015). A Rubisco-independent pool of RBCL with a moonlighting function related to oxidized RNA could explain the non-colocalization of the immunofluorescence signal from 8-oxoG RNA and the strongest patches of RBCL immunofluorescence signal, which are probably the Rubisco holoenzyme (Fig. 1E).

Our results and results reported previously support the inclusion of RBCL in a class of metabolic proteins with dual functions as RNA-binding proteins (Yosef et al., 2004). These proteins have been proposed to coordinate metabolism and gene expression and to enhance the functional diversity of proteomes (Hentze and Preiss, 2010). For only a few such proteins has evidence of the dual function been demonstrated in vivo, as we have done here for RBCL (Figs 3 and 5).

The molecular mechanisms involved in the dual function of RBCL related to oxidized RNA remain to be determined. They could involve the protection of undamaged RNA from ROS or the selection of oxidized RNA for repair, degradation or sequestration from the translated pool. That the immunofluorescence signal from 8-oxoG RNA was not enriched in cpSGs (Fig. 1E,F) is inconsistent with sequestration of oxidized RNA, but compatible with the other mechanisms. For example, cpSGs might not accumulate oxidized RNA if they protect non-oxidized RNA from ROS or if they rapidly degrade or repair oxidized RNA. Nevertheless, it remains to be determined whether or not cpSGs function in the quality control of oxidized RNA. It also remains to be determined whether RBCL controls the level of oxidized RNA directly or indirectly, that is, through other factors and pathways.

Our findings have potential relevance to stress granules and processing bodies. These RNA granules have been implicated in fundamental cell biological processes, but their functions and physicochemical properties are only partially understood. Our results reveal a Rubisco-independent pool of RBCL which is insoluble, a feature that might be expected of cpSG proteins because cpSGs are aggregates of RNA and protein, and insolubility in Triton X-100 is a property of processing body proteins (Teixeira et al., 2005). Furthermore, the P16-TI fraction with this RBCL pool, like cpSGs, has an excess of the 30S subunit of the chloroplast ribosome over the 50S subunit (Fig. 4). Even under non-stress conditions, when most cells do not have cpSGs, RBCL was detected in the insoluble form and shown to affect the level of oxidized RNA in vivo (Fig. 3). Taken together, these results suggest that RBCL carries out its moonlighting function constitutively, and at the submicroscopic level, that is, independently of cpSGs. Whether or not stress granules and processing bodies also function as submicroscopic RNP assemblies under non-stress conditions remains to be determined.

Oxidized RNA arises under optimal growth conditions and the mean level did not increase during stress (Fig. 3). RNA oxidation might result from ROS produced as a byproduct of photosynthesis, which occurs under both non-stress and stress conditions (Foyer and Noctor, 2009). Moreover, a presumed detrimental nature of oxidized RNA in the chloroplast seems at odds with the wild-type growth rate of ΔrbcL despite its having an elevated level of oxidized RNA (relative to wild type) (Fig. 3A; Fig. S1A). Our results are consistent with a certain level of oxidized RNA being benign under optimal conditions (Fig. S1A) and becoming detrimental in the presence of H₂O₂ (Fig. 5). For example, oxidized bases in chloroplast mRNAs could result in the synthesis of aberrant iron-binding proteins with Fenton activity (conversion of H₂O₂ into the highly toxic hydroxyl radical) and thereby potentiate the toxicity of H₂O₂ (Foyer and Noctor, 2009).

The wild-type strain was 4A+ (CC-4051); ΔrbcL was CC-4696 (MX3312) (Dr Genhai Zhu, Pioneer Hi-Brid, Redwood City, CA) (Satagopan and Spreitzer, 2004); ΔRBCS was CAL005.01.13 (dim1) (Dent et al., 2005). ΔrbcL and ΔRBCS are non-photosynthetic, but fully viable under heterotrophic conditions (Satagopan and Spreitzer, 2004). Because Rubisco mutants are highly light-sensitive (Johnson, 2011), all cultures were grown and tested under heterotrophic conditions [in the dark on Tris-acetate phosphate (TAP) medium (Gorman and Levine, 1965), at 24°C, with orbital shaking]. To generate the complemented ΔRBCS strain, the wild-type RBCS2 gene (on plasmid pSS2) (Khrebtukova and Spreitzer, 1996) was introduced into ΔRBCS by glass-bead-mediated transformation as described previously (Purton, 2007). The double mutants for rbcL and the RBCS locus (RBCS1 and RBCS2) were obtained from a cross between ΔrbcL (mt+) and ΔRBCS (mt−). It was crucial to use cultures in the exponential growth phase and in the density range of 2×10⁶–4×10⁶ cells/ml.

The immunofluorescence protocol for Chlamydomonas was as reported previously (Uniacke et al., 2011). To induce cpSG formation, live cells were treated with 2.0 mM H₂O₂ for 15 min prior to fixation. Where indicated, fixed and permeabilized cells were treated for 1 h at 37°C with 10 µg/ml RNase A (Fermentas) or 50 µg/ml DNase I (Invitrogen). 8-oxoG was immunodetected with a monoclonal mouse antibody (dilution 1:500, QED Bioscience Inc., clone 15A3). The specificity of this antibody in situ was confirmed by incubating with its antigen 8-hydroxy-2′-deoxyguanosine (1.0 mg/ml) for 2 h before immunofluorescence-staining. Rabbit antibodies were used to immunodetect RBCL (Agrisera AS03037, dilution 1:2000) and the 30S subunit chloroplast ribosomal-protein (S-20, dilution 1:2000) (Randolph-Anderson et al., 1989). Secondary antibodies were Alexa-Fluor-488-conjugated goat anti-mouse-IgG and Alexa-Fluor-568-conjugated goat anti-rabbit-IgG antibodies (Invitrogen Inc.). Images were acquired by epifluorescence microscopy using a Leica DMI 6000 microscope (Leica Microsystems) with a 63×/1.4 objective, a Hamamatsu OrcaR2 camera, and Volocity acquisition software (Perkin-Elmer). z-stacks were taken by series capture at a thickness of 0.2 µm per section. Stacks were deconvoluted with AutoQuant X3 (Media Cybernetics Inc.) (Abramoff et al., 2004).

HeLa cervical cancer cells were obtained from the American Type Culture Collection (Manassas, VA; ATCC). Cells were cultured at 37°C in Dulbecco's modified Eagle's medium (DMEM; Sigma-Aldrich, St Louis, MO) supplemented with 10% fetal bovine serum, penicillin and streptomycin (all from Sigma-Aldrich). Fixed and permeabilized cells were treated with 10 µg/ml RNase for 1 h, where indicated. 1.0 mM sodium arsenite treatments of live cells were for 1.0 h under the conditions described above.

The protocol was as described previously (Fournier et al., 2013). Rabbit antibodies were used to immunodetect G3BP1 (Dr Imed Gallouzi, McGill University, Montreal, Canada; dilution 1:2000), RCK (Santa Cruz Biotechnology, dilution 1:150), TOM20 (FL-145) (Dr Josée Lavoie, Laval University, Quebec City, Canada, dilution 1:100), and 8-oxoG (clone 15A3, from QED Bioscience Inc., dilution 1:500). The secondary antibodies were Alexa-Fluor-488-conjugated goat anti-mouse-IgG, Alexa-Fluor-598-conjugated goat anti-mouse-IgG and Alexa-Fluor-568-conjugated goat anti-rabbit IgG antibodies (Invitrogen Inc.). Immunofluorescence signals were visualized by using an LSM 700 confocal laser scanning microscope (Zeiss), controlled with 2009 ZEN software for image acquisition and analysis. Images were acquired using the following settings: 63× oil objective (zoom 1.0), 0.06 µm for pixel size and 1.00 airy units as a pinhole.

In each biological replicate experiment, RNA, DNA and protein were isolated from the same culture. Where indicated, live cells were exposed to 2.0 mM H₂O₂ for 15 min. Total RNA was extracted using TRI-reagent (Sigma-Aldrich) according to the manufacturer's protocol. RNA was shown to be free of DNA and quantified by analysis with a 2100 Bioanalyzer (Agilent). Total DNA was extracted using hexadecyltrimethylammonium bromide, as described previously (Murray and Thompson, 1980). RNA was removed from DNA preparations by treatment with DNase-free RNase (10 µg/ml, Fermentas) at 37°C for 1 h, followed by precipitation of the DNA from RNA fragments with polyethylene glycol (Sambrook and Russell, 2001). DNA concentrations were quantified by UV spectrophotometry. Total RNA (5 µg) and DNA (1 µg) samples were transferred to a nitrocellulose membrane with a Minifold-II slot blot system (Schleicher & Schuell). Membrane filters were reacted with the commercial antibody against 8-oxoG (1:500, QED Bioscience Inc., clone 15A3) overnight, at 4°C (Sambrook and Russell, 2001). A goat anti-mouse-IgG secondary antibody (KPL) was used and ECL detection was performed with a commercial kit (Millipore). To isolate total protein, cells were pelleted (3000 g, 3 min) and broken in 50 mM Tris-HCl pH 8.0, 1 mM EDTA, 50 mM NaCl, 1 mM PMSF and 100 mM DTT by bead beating (Hopkins, 1991). Protein samples (20 μg) were analysed with the OxyBlot Kit according to the manufacturer's protocol (Millipore). Total protein concentration was determined as described previously (Smith et al., 1985). Quantification was carried out as described previously (Bohne et al., 2013). For statistical analyses, each biological replicate was compared to a corresponding ΔrbcL strain prior to the addition of H₂O₂, which was designated as 100% oxidation. Statistical differences were determined in each case using a one-sample Student's t-test.

Cells from 75-ml cultures were pelleted by centrifugation (5000 g, 5 min) at room temperature and resuspended in 5.0 ml ice-cold MKT buffer [25 mM MgCl₂, 20 mM KCl, 10 mM Tricine pH 7.5, 1.0% (v/v) protease inhibitor cocktail (Sigma-Aldrich)]. Cells were broken by three passages through a chilled French pressure cell at 1000 psi. The cell lysate was centrifuged at 3200 g for 1 min to pellet unbroken cells. The supernatant was collected and centrifuged in a benchtop microcentrifuge at 16,000 g for 20 min at 4°C. The resulting supernatant was designated as S16. The pellet (P16) was resuspended in MKT buffer with 2% (v/v) Triton X-100 and incubated for 15 min at room temperature with gentle agitation to solubilize membranes. These samples were then centrifuged at 16,000 g for 20 min at 4°C. The supernatant was designated as P16-TS. The pellet (P16-TI) was resuspended in MKT buffer. Protein samples from each fraction were subjected to SDS-PAGE and immunoblot analysis. Protein loading was based on equal proportions of each fraction. Rabbit antibodies were used to immunodetect RBCL (Agrisera, AS03037; dilution 1:30,000), RBCS (Dr Robert Spreitzer, University of Nebraska, Lincoln, NE; dilution 1:5,000), CP43 (Agrisera, 111787; dilution 1:2,000), S-20 (30S r-protein; dilution 1:30,000) (Randolph-Anderson et al., 1989), L-30 (50S r-protein; dilution 1:20,000 (Randolph-Anderson et al., 1989) and HSP70B; dilution 1:30,000 (Schroda et al., 1999). Secondary staining used goat anti-rabbit-IgG antibody (KPL) incubated for 1 h at room temperature.

³⁵S-pulse-labelling reactions were performed with ³⁵SO₄ as described previously (Uniacke and Zerges, 2007). Cells (∼1.2×10⁷ per sample) were then pelleted by centrifugation (5000 g, 5 min), washed once with 500 μl 50 mM Tris-HCl pH 7.4, resuspended in 80 μl MKT buffer and broken by bead beating (Hopkins, 1991). The cell lysates were centrifuged (16,000 g, 5 min), and the pellet and supernatant fractions were subjected to SDS-PAGE (7.5–15% acrylamide, 6.0 M urea). ³⁵S-pulse-labelled proteins in dried gels were visualized with a phosphoimager (Typhoon, GE Healthcare).

Cell survival and viability were assayed following addition of H₂O₂ to 4.0 mM. Cell survival was determined by counting the proportion of live cells with the Trypan Blue exclusion assay (Sigma-Aldrich). Viability was determined by assessing the number of colony forming units on agar-solidified TAP medium. For statistical analyses of survival and viability, a mixed analysis of variance (ANOVA) was conducted with strain as a between factor and time as a within factor. When appropriate, post hoc analyses were conducted using Tukey's honest significant difference test.

We thank A. Piekny and C. van Oostende for assistance with microscopy, W. Brake for assistance with statistical analyses, M. Schroda for the anti-HSP70B antiserum; I. Gallouzi for the anti-G3BP1 antiserum; G. Zhu, and Pioneer Hi-Bred (Redwood City, CA) for ΔrbcL-MX3312, and R. Spreitzer for the pSS2 plasmid and the anti-RBCS antiserum. This work was carried out, in part, in the Centre for Microscopy and in the Centre for Structural & Functional Genomics (Concordia University, Montreal, Canada).