Light history influences the response of the marine cyanobacterium Synechococcus sp. WH7803 to oxidative stress.

Marine Synechococcus undergo a wide range of environmental stressors, especially high and variable irradiance, which may induce oxidative stress through the generation of reactive oxygen species (ROS). While light and ROS could act synergistically on the impairment of photosynthesis, inducing photodamage and inhibiting photosystem II repair, acclimation to high irradiance is also thought to confer resistance to other stressors. To identify the respective roles of light and ROS in the photoinhibition process and detect a possible light-driven tolerance to oxidative stress, we compared the photophysiological and transcriptomic responses of Synechococcus sp. WH7803 acclimated to low light (LL) or high light (HL) to oxidative stress, induced by hydrogen peroxide (H₂O₂) or methylviologen. While photosynthetic activity was much more affected in HL than in LL cells, only HL cells were able to recover growth and photosynthesis after the addition of 25 μM H₂O₂. Depending upon light conditions and H₂O₂ concentration, the latter oxidizing agent induced photosystem II inactivation through both direct damage to the reaction centers and inhibition of its repair cycle. Although the global transcriptome response appeared similar in LL and HL cells, some processes were specifically induced in HL cells that seemingly helped them withstand oxidative stress, including enhancement of photoprotection and ROS detoxification, repair of ROS-driven damage, and regulation of redox state. Detection of putative LexA binding sites allowed the identification of the putative LexA regulon, which was down-regulated in HL compared with LL cells but up-regulated by oxidative stress under both growth irradiances.

concomitant electron transport exceeds consumption of photochemically generated reductants. On the other hand, light and ROS seem to act synergistically in PSII photoinhibition, either through direct damage to PSII and/or by inhibiting repair. The influence of growth irradiance on the sensitivity to reactive oxygen stressors remains unclear. In particular, it is not well understood whether low light-(LL-) and high light-(HL-) acclimated cells, which show very different photophysiologies (Kana and Glibert, 1987a, b;Kana et al., 1988;Moore et al., 1995;Six et al., 2004, Six, et al.,2005 and different tolerance to stress factors (MacDonald et al., 2003;Bouchard et al., 2005;Garczarek et al., 2008), are equally able to withstand oxidative stress. Cells acclimated to LL have a slow electron transport rate, presumably low rates of primary ROS generation, but would not be expected to maintain full induction of ROS detoxification paths. Cells acclimated to high light have a faster electron transport rate, and presumably higher rates of primary ROS generation, but are in parallel expected to maintain stronger induction of ROS detoxification paths. The net cellular response to additional ROS therefore depends upon the relative magnitudes of the underlying effects of primary cellular ROS generation and the counteracting effects of ROS detoxification.
In order to address this question, we used Synechococcus sp. WH7803, a physiologically well characterized marine cyanobacterial isolate (see e.g. Kana and Glibert, 1987a, b;Toledo et al., 1999;Lindell and Post, 2001;Garczarek et al., 2008;Jia et al., 2010) for which the complete genome sequence (Dufresne et al., 2008) and a whole-genome microarray are available. We monitored PSII photophysiology and the global transcriptome response of LL-and HL-acclimated cells of this strain upon exposure to various concentrations of a direct reactive oxygen stressor, hydrogen peroxide (H 2 O 2 ), and a cellular-dependent O 2 •producing agent, methylviologen (MV).

Effect of H 2 O 2 and MV on PSII function of LL and HL cells
Synechococcus sp. WH7803 cells acclimated to low light (18 μ mol photons m -2 s -1 , hereafter LL cells) and high light (250 μ mol photons m -2 s -1 , hereafter HL cells) are clearly in different initial states. In particular, the initial PSII activity, as indicated by the F V /F M values before the application of any treatment, is lower in HL-acclimated cells than in LL-acclimated ones (Fig. 1). Incubation of LL-and HL-acclimated Synechococcus in the presence of different concentrations of H 2 O 2 and MV for 2 h under their respective growth light irradiances indicated that the higher the concentration of oxidizing agent, the stronger the decrease of PSII activity (F V /F M , Fig. 1). LL and HL cells however showed quite distinct photoinactivation patterns in response to oxidizing stress. For LL-acclimated detoxifying the cells from this oxidizing agent and restoring photosynthetic activity, and most probably other affected cellular processes.

Effect of light acclimation and electron transport rate on MV-mediated ROS production and PSII inactivation
If it is correlated with photosynthetic electron flow at PSI acceptor side, MV-mediated ROS production is expected to differ between LL-and HL-grown cells under their respective growth irradiance. To check this hypothesis, HL-acclimated cultures were either maintained under HL or shifted to LL conditions (inducing a sharp decrease in photosynthetic electron flow) 10 min before addition of either sub-saturating (1µM) or saturating concentrations (50 or 250 µM) of MV (Fig. 3).
For all MV concentrations tested, the shift to LL led to a much lower PSII inactivation after 2 h ( Fig. 3A) and an interruption of H 2 O 2 production, as assessed by the quasi constancy of the fluorescence of scopoletin, a compound sensitive to H 2 O 2 (Fig. 3B), compared to cells maintained under HL. This demonstrates that the MV-mediated PSII photoinactivation rate is highly dependent upon the photosynthetic electron transport rate. Surprisingly, the relative PSII photoinactivation in HL cells shifted to LL was only slightly lower than that of LL cells maintained in LL in the presence of MV (Fig. 3A), even though the H 2 O 2 production was significantly higher in the LL cultures ( Fig. 3B).

Relation between D1 protein turnover and ROS-dependent photoinactivation of PSII
In order to compare the PSII repair capacities of LL and HL cultures exposed to 25 µM H 2 O 2 , we plotted the time course variations of the relative D1 protein cell content (Fig. 4). While both LL and HL cells were able to increase their D1 protein content in response to oxidative stress during the first 8 h, only the HL cells sustained a high level of D1 protein over two days, thus enabling sufficient D1 protein turnover until complete recovery of PSII function. In contrast, PSII activity of LL cells dropped sharply and ceased after 30 hours, suggesting that even a low dose of H 2 O 2 provoked an irreversible impairment of PSII repair cycle within about one day, and likely subsequent cell death.
To assess the initial D1 repair capacities of HL and LL cells and the effect of H 2 O 2 on the balance between PSII inactivation and repair, HL and LL Synechococcus cells were either maintained under their initial growth irradiance or transferred to dark, and exposed to H 2 O 2 and/or lincomycin, an inhibitor of protein synthesis. The results showed that no PSII damage was detectable after 2 hours in LL-acclimated cells without addition of any oxidant, neither in the dark nor under light, independently of the presence of lincomycin (Fig. 5A). In contrast, HL cells suffered a strong progressive light-dependent PSII inactivation in the presence of lincomycin (Fig.  5B), reaching 70 % inactivation after two hours compared to control cultures without lincomycin.
This demonstrates that, in contrast to LL cells, PSII complexes from HL Synechococcus undergo considerable photoinactivation under their growth irradiance, which must be counteracted by the induction of a fast PSII repair cycle at a much higher rate than in LL-acclimated cells.
Secondly, the stronger effect on LL cells of 750 µM H 2 O 2 plus lincomycin compared to lincomycin alone (Fig. 5C) Table   S1).
Principal Component Analysis (PCA, Fig. 6A) as well as hierarchical clustering dendrograms of the set of genes significantly differentially expressed in at least one of the pairwise comparisons ( Fig. 6B) indicated good reproducibility among biological replicates used for microarrays analyses since all replicates grouped together but for the HL stressed ones. This suggests a similar effect of H 2 O 2 and MV on HL cells (Fig. 6AB). The most highly differentiated datasets were the oxidative stressed cells vs. control ones, independently of the reactive oxidizing agent used. The contribution of the stress factor seems to be dominant in the PCA component 1, responsible for 35.4 % of the variability of the data (Fig. 6A). The second major factor appears to be light acclimation, which segregated LL and HL datasets within both control and stressed conditions. This effect would be mainly represented by the PCA component 2, responsible for 18.08 % of the data variability.
Furthermore, the higher distance between HL and LL datasets in stressed conditions compared to control ones suggests that this component 2 takes into account the synergistic effect of oxidative stress and HL acclimation.
In order to identify the proportion of genes responding to the different treatments, Venn diagrams were performed using the same datasets (Fig. 7). These analyses showed that MV and H 2 O 2 induced a similar effect on both LL-and HL cells (Fig. 7A) Table 1. Among discordant genes in HL cells, the few ones which are also discordant in LL cells are retrieved. Noteworthy, these common genes include lexA and recA that play a key role in the control of DNA repair (Kuzminov, 1999;Butala et al., 2009;Kolowrat et al., 2010). The expression of these two genes was activated by oxidative stress but repressed by HL acclimation.
Most of the other discordant genes encode ribosomal subunits or proteins without attributed function.

Identification of functional groups responding to oxidative stress and light acclimation
Two-dimension hierarchical clustering analysis allowed the identification of 5 gene clusters behaving similarly in the different treatments ( most obvious ones included ribosomal and photosynthesis genes that were strongly down-regulated by oxidants in both light conditions (group E), suggesting a decrease of photosynthetic and protein translation activities. Indeed, most of the genes encoding the subunits of the complexes driving these processes (ATP synthase, photosystems I and II, phycobilisomes, Calvin cycle, ribosomes) exhibited a similar response. This group E also includes numerous genes involved in phosphate uptake (pstB, pstSII, phnDE, etc), storage (ppx, ppa) and regulation (phoB) as well as DNA repair genes such as mutM and most rec genes (except recA, located in group B). Another functional group with a characteristic behavior consists in genes involved in photoprotection, which are mainly induced by oxidative stress in LL and/or HL conditions (groups A to C). Among those, genes encoding HL inducible proteins (HLIP) are particularly well represented in group B corresponding to transcripts induced in both light conditions, which also include the ocp gene coding for the so-called orange caroteno-protein (Kerfeld et al., 2005;Wilson et al., 2006;Kirilovsky, 2007). In contrast, genes involved in the synthesis of other carotenoids (crtDEHQR genes), including zeaxanthin, are located in the group C of genes induced by ROS in HL cells only.
As expected, chaperones and proteases encoding genes (clpBCPX, dnaJK, groLS, etc.) were induced by ROS and were also mainly represented in groups A to C.

Are HL acclimation and oxidative stress cumulative?
The physiological experiments performed in this study revealed that under their growth irradiance, HL-acclimated cells were more affected than LL ones by H 2 O 2 stress with regard to PSII photoinactivation, and this was verified across the whole range of H 2 O 2 concentrations tested ( Fig.   1 and 2). A possible hypothesis explaining this interesting observation could be that cells grown at 250 µmol photons m -2 s -1 were already facing a stress situation, so that they could not handle an additional stress. However, marine Synechococcus exhibit a large physiological flexibility that allows them to grow under a large range of irradiance in laboratory conditions (Kana and Glibert, 1987a, b;Six et al., 2004;Six et al., 2005) and with maximum growth rates usually observed between 200-300 µmol photons m -2 s -1 (Kana and Glibert, 1987a;Moore et al., 1995;Six et al., 2004). Thus, the HL irradiance used in this study should not represent highly stressful growth conditions. Furthermore, microarray analyses revealed that only 121 genes were differentially expressed in the HL-Ct vs. LL-Ct pairwise comparison, while 4 to 6 times more genes responded to addition of oxidizing agents. None of the 121 genes seemed to be involved in ROS defense or detoxification mechanisms, except for a ferredoxin (petF_1580) and a peroxiredoxin (prxQ_1236) encoding genes that were both significantly upregulated in HL compared to LL cells. Interestingly, lexA and recA genes, which are involved in SOS response (Butala et al., 2009;Kolowrat et al., 2010), were downregulated in HL cells, supporting the interpretation that those cells were not under high light stress that would require activation of this DNA repair system.

Why are the HL cells more sensitive than LL cells to MV-mediated oxidative stress?
MV is a very low potential redox compound that primarily accepts electrons from the F A and F B iron-sulfur clusters of PSI centres at the expense of ferredoxin reduction (Supplemental Fig. S1; Fujii et al., 1990). By reacting very quickly with O 2 , the reduced MV •+ radical is spontaneously reoxidized back to MV with formation of the membrane impermeable superoxide anion. The latter compound undergoes further dismutation to form H 2 O 2 that can readily pass through biological membranes (Bus and Gibson, 1984;Goldstein et al., 2002). PSI electron transport includes both linear electron flow (LEF) from PSII through PSI to NADP + , and cyclic electron flow (CEF) from the PSI acceptor side back to cytochrome b 6 f complex (Supplemental Fig. S1; Bendall and Manasse, 1995;Joliot and Joliot, 2002). By reacting with any electrons coming from PSI, MV inhibits CEF and diverts LEF (Yu et al., 1993;Herbert et al., 1995) towards ROS production (Jia et al., 2008;Fan et al., 2009). In the presence of MV, the ROS production in a cell is thus mostly dependent upon the absolute number of electrons passing through PSI complexes per time unit, which results from two parameters that are both strongly influenced by growth irradiance: the electron transport rate at PSI complexes, and the number of PSI per cell. LEF is thus a critical parameter when considering the differential response of LL and HL cells to MV-mediated oxidative stress.
At their growth irradiance, HL cells mediate much faster LEF than do LL cells, thereby producing more ROS per PSI. For PSI content, previous studies have shown that in cyanobacteria, acclimation to high irradiance is accompanied by a strong decrease of the thylakoid surface and chlorophyll a content per cell (Kana and Glibert, 1987a;Moore et al., 1995;Six et al., 2004). In cyanobacteria chlorophyll a is almost exclusively bound to reaction centers (especially to PSI complexes), and HL cells must therefore contain fewer photosystems than LL cells. Thus, HL cells exhibit a high PSI electron transport rate favoring a large MV-mediated ROS production rate, but a lower photosystem cell content, and therefore a lower number of potential ROS formation sites than LL cells.
Addition of MV concentrations higher than 25 µM induced a saturation of PSII photoinactivation in both LL and HL cells (Fig. 1B concentration was in contrast almost 10-fold lower in HL cells than in LL cells (Fig. 1B). If the number of PSI per cell was a major factor influencing ROS production, a severe PSII photoinactivation would have been expected in LL cells, given their high PSI cell content. Since this is not the case, this rather suggests that the LEF rate is the major factor controlling the amplitude of the MV-induced oxidative stress. Furthermore, when HL cells were suddenly shifted to LL conditions, lowering LEF without affecting the PSI cell content, the production of H 2 O 2 almost stopped (Fig. 3B), with a consequent slowing down of PSII photoinactivation compared to the cells maintained in HL conditions (Fig. 3A). This clearly highlights the importance of LEF in the amplitude of MV-induced oxidative stress in HL cells. Still, it is worth noting that this only partially explains the higher sensitivity of HL cells to oxidative stress. Firstly, despite the sharp reduction in MV-dependent ROS production, the HL cells shifted to LL still exhibited a 25% photoinhibition (Fig. 3AB). Moreover, even though this stronger PSII inactivation could, at least in part, be due to a higher rate of endogenous ROS formation in the presence of MV, this justification does not hold true for H 2 O 2 , since the same oxidant concentration, added exogenously, also produces a much stronger photoinhibition in HL cells than in LL ones (Fig. 1A). Thus, besides inducing a higher MV-dependent ROS production, HL acclimation also seems to result in a higher sensitivity to ROS-driven damages.

Effect of ROS on PSII damage and repair
The PSII repair cycle is a crucial process in the response of photosynthetic organisms to various stressful conditions (Park et al., 1995;Nishiyama et al., 2006;Six et al., 2007a). Several hypotheses have been proposed to explain the detrimental effect of ROS on PSII function, either through direct damage to PSII reaction centers with consequent degradation of the D1 protein (Hideg et al., 2007;Vass and Cser, 2009) occurred right after stress initiation (Fig. 4). This de novo synthesis of D1 was not sufficient to maintain PSII quantum yield indicating that most of these D1 proteins were not included in functional PSII centers (Garczarek et al., 2008).
While PSII repair cycle inhibition was partial at 25 µM, it seemed to be complete in presence of 750 µM H 2 O 2 , since this oxidant had the same effect on HL cells with and without lincomycin ( Fig. 5D). Furthermore, this H 2 O 2 concentration also induced strong direct PSII damage since the level of photoinactivation was about twice higher than when the repair cycle was blocked by lincomycin alone. These ROS-driven damages seemed mostly light-independent, since the photoinactivation in the presence of 750 µM H 2 O 2 alone was similar in the dark. Such damages could not be restricted to PSII only but could be attributed to a more global degradation of cellular components by ROS (Imlay, 2003), including translation machinery (Kojima et al., 2007). At high H 2 O 2 concentrations, the decrease of PSII quantum yield in HL cells reflected both an impairment of PSII repair cycle and direct damages to PSII complexes.
In contrast to HL cells, LL cells initially showed a nearly constant D1 pool in the presence of 25 µM of H 2 O 2 , which appeared sufficient to maintain PSII quantum yield at its initial level during 6 h (Figs. 2 and 4). This suggests that 25 µM H 2 O 2 did not provoke additional PSII damage beyond the very low photoinactivation rate at this light level, which was below detection. However, an effect leading to PSII damage was clearly noticed at 750 µM of H 2 O 2 (Fig. 5C). In contrast to HL cells, these damages were light-dependent, as almost no damage occurred when H 2 O 2 was added in the dark (Fig. 5C-D). A possible explanation for the light-dependency of these damages relies on the formation of reactive hydroxyl radicals OH • through oxidation by H 2 O 2 of reduced metals, such as the Fe or Mn atoms located in reaction centres II (Imlay, 2003). Indeed, PSII non-heme iron, one of the prosthetic group of the D1 protein, has been reported to be oxidized by H 2 O 2 and is thought to be involved in a specific OH • -induced cleavage of this protein (Diner and Petrouleas, 1987;Lupinkova and Komenda, 2004). Such reactions are expected to be light-dependent since the redox state of the non-heme iron depends on photosynthetic electron transfer (Ishikita and Knapp, 2005).
Similarly, at PSII donor side, the redox status of the Mn cluster, that can also be oxidized by H 2 O 2 (Lupinkova and Komenda, 2004), is also dependent on PSII electron transport. In darkness, the oxidised state of such PSII-bound metals might prevent a reaction with H 2 O 2 and therefore the resulting formation of OH • leading to D1 protein cleavage.
Interestingly, a cumulative effect of light and oxidative stress on PSII function does not agree with the results obtained on Synechocystis sp. PCC 6803 for which oxidative stress induced by H 2 O 2 did not accentuate light-induced damages but only affected the rate of PSII repair (Allakhverdiev and Murata, 2004). It should however be noted that there are major experimental differences between those results and our study. Indeed, the Synechocystis experiments were mainly performed by monitoring the recovery of PSII activity after cultures were briefly exposed to very high light conditions (from 250 to 2,000 µmol photons m -2 s -1 ), inducing strong PSII photoinactivation that the cells were only able to withstand for a short time. In contrast, in our study, Synechococcus sp. WH7803 cells were fully acclimated to the different growth irradiances.
Furthermore, a large range of H 2 O 2 concentrations was tested, as the dose proved to be important to differentiate between ROS-induced direct damage and inhibition of repair. For instance, the effect of H 2 O 2 on PSII repair activity could only be detected at 25 µM and on HL cells, i.e. when some light-but no ROS-mediated damages were generated. Although our approach did not allow the assessment of the PSII repair cycle when no detectable photodamage had occurred, it had the advantage of taking into account the differential damage resistance/repair capacities of these cells that cannot be appreciated through a light shift. For instance, one could expect that HL cells, which possess a higher ratio of D1:2 to D1:1 protein isoforms than LL cells (Garczarek et al., 2008), also possess a higher PSII resistance to photoinactivation (Campbell et al., 1995;Campbell et al., 1998;Tichy et al., 2003). The fact that HL cells were in contrast even more sensitive to photoinhibition indicates that oxidative stress has a strong effect on direct PSII damages independently of the dominant D1 isoform present in the PSII. Altogether, these data suggest that the cumulative effect of light-and ROS-driven damages in HL cells and the resulting stronger dependence on D1 repair, are probably responsible for the higher PSII sensitivity of those cells to oxidative stress. For cells shifted from HL to darkness we observed a moderate, progressive decline in F V /F M (Fig. 5B).
Addition of lincomycin prevented this decline, which therefore depends upon protein synthesis in the dark. This decline in F V /F M in the dark resulted mostly from an increase in the F 0 level of fluorescence in cells shifted from HL to darkness (Supplemental Figure S2). Similar increases in F 0 occurred in the absence or presence of H 2 O 2 .

Transcriptome response to oxidative stress
Effects of oxidative stress on PSII do not alone determine the survival capacities of the cells.
Indeed, while HL cells were able to recover from a 25 µM H 2 O 2 stress, LL ones eventually died even though their photosynthetic activity seemed initially less affected (Fig. 1). We performed transcriptome analyses to observe the global response of the cells and to understand why HL cells were capable of better resistance and recovery than LL cells when facing oxidative stress. concordance between LL and HL cells, meaning that most genes differentially expressed in both light conditions responded similarly in HL and LL cells (Fig.7B). Hierarchical clustering analyses ( Fig. 8) however, allowed highlighting some differences associated with one or the other light condition.
The photosynthetic genes are among the most affected by oxidative stress in both LL and HL cells ( Fig. 8 and Supplemental Table S1), including the four psbA gene copies encoding the D1 protein of the PSII core in Synechococcus sp. WH7803. At the transcriptomic level, microarray data showed that the two transcripts making up most of the psbA mRNA pool in this strain (Garczarek et al., 2008), are inversely regulated in response to oxidative stress, indicating an exchange between D1:1 (encoded by psbA_0784) and D1:2 (encoded by psbA_0790) isoforms in both HL and LL cells. The resulting D1 pool would then be dominated by the D1:2 isoform, which is thought to provide resistance to photoinhibition (Krupa et al., 1991;Clarke et al., 1993a;Clarke et al., 1993b;Campbell et al., 1995;Campbell et al., 1998;Tichy et al., 2003). Accordingly, ftsH_1216, the ROSinduced direct ortholog of Synechocystis sp. PCC 6803 slr0228 gene (Li et al. 2004;Houot al., 2007), encoding the FtsH2 protease involved in the clearance of damaged D1 proteins from inactivated PSII (Komenda et al., 2006;Komenda et al., 2010), belongs to the cluster of genes mainly activated in response to ROS in Synechococcus sp. WH7803 (cluster B; Fig. 8). It is worth noting that, as previously described, the D1 pool in the PSII of HL cells is dominated by the D1:2 isoform (Garczarek et al., 2008). HL cells would then be better prepared to sustain ROS-driven photoinhibition not only thanks to the initial dominance of the transcript coding for D1:2 isoform but also by completing the isoform switch in response to additional stress. Most other photosynthetic genes belong to some of the main gene clusters inhibited by oxidative stress (clusters D and E; Fig. 8). This includes most of the ATP synthase, photosystems I and II subunits as well as the genes involved in phycobilisome synthesis and Calvin cycle, suggesting a decrease in energy supply in response to stress.
Although most of the photosynthetic genes were not significantly differentially affected by oxidative stress in LL and HL cells, some differences in gene induction however gave us hints to explain the survival of HL cells in oxidative stress conditions that were lethal for LL cells. This includes a number of genes involved in photoprotection mechanisms such as genes encoding the orange carotenoid protein (ocp, Kerfeld et al., 2005;Wilson et al., 2006;Kirilovsky, 2007), whose induction was significantly higher in HL than in LL cells (Supplemental Table S1). In addition, the induction of the crtDEHQR genes involved in carotenoid synthesis (Masamoto et al., 1998;Klassen, 2010) mainly belong to the cluster C of genes more activated in HL cells than in LL ones www.plantphysiol.org on August 26, 2017 -Published by Downloaded from Copyright © 2011 American Society of Plant Biologists. All rights reserved. (Fig. 8), with crtR, a gene involved in the conversion of from β -carotene into zeaxanthin, being clearly the most up-regulated. Carotenoids are thought to dissipate energy from photosensitized chlorophyll or from singlet oxygen and may have intrinsic antioxidant properties (Edge et al., 1997). In particular, zeaxanthin-deficient mutants of Synechocystis sp. PCC 6803 and Synechococcus sp. PCC 7002 are more sensitive than their respective wild types to HL and oxidative stress, and it has been shown that such mutants indeed accumulate ROS (Schafer et al., 2005;Zhu et al., 2010). The genes differentially regulated by oxidative stress also include the highlight-inducible proteins (HLIPs), three of which being induced at higher levels in HL cells than in LL ones (hli_0937, hli_0797, hli_1441). HLIPs have been shown to be involved in protecting cells against photodamages either by direct or indirect dissipation of excess absorbed light energy (Montané and Kloppstech, 2000;He et al., 2001;Havaux et al., 2003), or by binding and storage of free chlorophylls (Funk and Vermaas, 1999;Vavilin et al., 2007;Kufryk et al., 2008).
Among the genes clearly induced by oxidants in HL cells (cluster C, Fig. 8), those encoding the NADPH dehydrogenase subunits are highly represented. This complex is involved in a variety of cellular processes including respiratory electron flow, CEF around PSI, indirect control of redox state of the plastoquinone pool, as well as in CO 2 uptake (Cooley and Vermaas, 2001;Ogawa and Mi, 2007;Battchikova et al., 2010). The active induction of ndh genes would allow maintaining energy production of HL cells under oxidative stress. Notably, a direct effect of NDH-dependent CEF in alleviating photo-oxidative stress has also been suggested in tobacco chloroplast by compensating the stromal over-reduction that induces formation of ROS (Wang et al., 2006).
Tolerance to the deleterious effects of ROS may result from the activity of a number of mechanisms that maintain ROS at a harmless level, or that actively repair ROS-induced damages.
ROS scavenging enzymes such as catalases, peroxidases and superoxide dismutases, as well as antioxidants such as glutathione, tocopherol and ascorbic acid, are involved in the maintenance of low levels of ROS (Noctor and Foyer, 1998;Storz and Imlay, 1999;Mittler, 2002;Masip et al., 2006). Such antioxidant activities involved in ROS scavenging and redox homeostasis have been reported in freshwater cyanobacteria such as Synechocystis sp. PCC 6803 and Synechococcus sp.  Table S1; Langford et al., 2002;Culotta et al., 2006). In contrast, in Synechocystis sp. PCC 6803, that only carries the sodB gene, the latter seemed to be induced in response to H 2 O 2 (Li et al. 2004;Houot al., 2007). In Synechococcus sp. PCC 7942, the cytosolic Fe-SOD does not protect the cells from ROS generated within the thylakoid membrane (Thomas et al., 1998), suggesting that other SOD might confer such a protection. Thus, the increase in periplasmic Cu/Zn-SOD activity would confer a resistance to the light-driven ROS production in HL cells. Another gene involved in ROS detoxification, dpsA, encoding a DNA binding protein, was also upregulated in response to oxidative stress in HL cells only. DpsA has been shown to be involved in the detoxification of Fe 2+ ions and H 2 O 2 and to more specifically protect DNA from oxidative damages (Chiancone and Ceci, A number of chaperone and protease encoding genes were also affected by ROS in only one of the light acclimation condition. Among those, the 3 genes encoding the GroEL/GroES system (groS, groL1, groL2) and htpG were particularly activated in LL cells. These two major chaperone systems in bacteria are known to be activated in response to cold stress, high irradiance and MV in possible exchange of Clp proteolytic core complexes in response to stress (Stanne et al., 2007); Adrian Clarke, pers. com.). Finally, degQ, which encodes a periplasmic protease known to confer some stress resistance (Barker et al., 2006), was also somewhat more induced in HL cells than in LL ones.
An element which could contribute to the better ROS resistance of HL cells is the capacity to induce an SOS response, which is involved in protecting the cells from damaging effect of increased mutation rates. Although some authors have suggested that cyanobacteria could lack an SOS response similar to that of Escherichia coli (Domain et al., 2004;Michel, 2005 WH7803. According to our analysis, all genes from the LexA regulon, i.e. that carry a putative LexA box in their regulatory region, were very similarly regulated in response to HL acclimation and to oxidative stress, strengthening the hypothesis that LexA may indeed regulate these genes and that the gene products with unknown function might be involved in the SOS response. In contrast, genes that are thought to be part of the LexA regulon in E. coli, but that do not have a LexA binding motif in their promoter region in Synechococcus sp. WH7803, are not similarly regulated and are most probably not under the control of LexA. Using the WebLogo tool (Crooks et al., 2004), the nine putative binding sites found in Synechococcus sp. WH7803 allowed to define the consensus of the LexA-Box in this strain, which seems to be degenerated from the palindromic TACAN 2 TGTA consensus (Supplemental Fig. S3). Interestingly, the expression of lexA was more induced in HL than in LL cells, suggesting that the SOS response may indeed play a role in the better resistance of HL cells to oxidative stress. Although it might appear surprising that both repressor-(lexA) and derepressor-(recA) encoding genes are simultaneously induced, the constant production of LexA would allow its re-accumulation to repress SOS genes once DNA damages have been repaired (Michel, 2005). At last, the expression of uvrA which was clearly differently regulated compared to all other SOS genes, was strongly repressed in response to oxidative stress in HL cells, and that may be due to the fact that uvrA is only expressed in very early stages of the SOS response (Kuzminov, 1999;Michel, 2005).

CONCLUSION
Our data showed that the susceptibility of Synechococcus cells to oxidative stress varies depending on culture light history. HL cells are more immediately affected by oxidant exposure than are LL cells, but HL cells then show better capacity to ultimately acclimate and recover from low concentration of hydrogen peroxide. By testing a range of oxidant concentrations and treatment durations on LL and HL cultures, we show that, depending upon the oxidant concentration, PSII damage observed in oxidant-stressed HL cells results both from cumulative effects of direct ROS damage to PSII, and ROS inhibition of the rapid PSII repair needed to counter the high photoinactivation rate. These findings largely reconcile earlier findings that ROS directly inactivates PSII with other studies showing ROS inhibition of PSII repair. Even though the active PSII repair cycle of HL cells is initially inhibited by oxidative stress, it nonetheless allowed them to fully recover when H 2 O 2 exposure did not exceed 25 µM. Although the oxidative stress induced a similar global transcriptomic response in HL and LL cells, in HL cells we detected the induction of other cellular processes including more efficient photoprotection mechanisms, some ROS detoxification and a better redox/energy homeostasis, that most likely helped prevent the death of HL cells. Thus, in HL cells, physiological changes induced by ROS were reversible when the intracellular level of ROS decreased, while in LL ones, similar ROS treatment led to cell death.
More generally, identification of the factors that set the threshold at which a cell makes the transition from successful acclimation/resistance to oxidative stress-induced cell death will be of major interest to understand the cumulative effect of environmental stress conditions.  Synechococcus sp. WH7803 growth was monitored by flow cytometry using a BD FACS Canto flow cytometer (Becton Dickinson Biosciences, San Jose, CA, USA) as previously described (Marie et al., 1997;Marie et al., 1999) and by phycoerythrin fluorescence emission maximum (excitation at 530 nm) using a spectrofluorometer (LS-50B Perkin Elmer, Waltham, MA).
In order to measure PSII damage and repair, 500 µg mL -1 lincomycin (Sigma Aldrich), an inhibitor of protein synthesis, was added or not to the culture immediately before submitting the cells to oxidative stress, following a procedure similar to the one described by Six et al. (2007a).
When cultures were shifted from one light condition to another, cells were shifted to the new light condition 10 min prior to addition of oxidising agent, immediately after the lincomycin.

Measurement of H 2 O 2 production during MV stress
Hydrogen peroxide production by cells upon MV-induced oxidative stress was monitored by the decrease of scopoletin fluorescence (Patterson and Myers, 1973;Tichy and Vermaas, 1999), as this molecule is degraded by H 2 O 2 in the presence of horseradish peroxidase. 0.5 to 1 x 10 8 cells from exponentially growing cultures were pelleted (6,000 x g for 10 min, Eppendorf 5424R), resuspended in 3 ml of 25 mM HEPES (pH 7.0) and incubated in a cuvette for 30 min under their initial HL-or LL-acclimation conditions. After addition of 1.33 µM scopoletin (Sigma Aldrich) and 25 U mL -1 horseradish peroxidase (Sigma Aldrich), the cells where incubated for 10 min under their initial irradiance or shifted to LL irradiance and 0, 50 or 250 µM of MV were then added. After 30 min the scopoletin fluorescence emission was monitored at 460 nm (excitation at 350 nm) using a spectrofluorometer (LS-50B Perkin Elmer, Waltham, MA). After subtracting the dye-free background signal, fluorescence was expressed as percentage of initial value.

D1 protein quantification by immunoblotting
A culture volume of 50 mL was centrifuged in presence of 10 mg L -1 of the non-ionic surfactant Pluronic F-68 (Sigma Aldrich) to avoid cell adhesion at the surface of the tubes and the pellet was flash frozen in liquid nitrogen and stored at -80°C until analysis. Total protein assays, polyacrylamide gel electrophoresis, protein transfer onto a polyvinylidene fluoride membrane, immunoreactions and chemoluminescence detection were performed as described by Garczarek et al. (2008). D1 protein amounts were expressed as a percentage of initial conditions.

RNA sampling and extraction
For the transcriptomic analyses, Synechococcus sp. WH7803 cultures were submitted to oxidative stress by addition of H 2 O 2 or MV and harvested when PSII quantum yield fell to half of the initial value. For H 2 O 2 experiments, this level of PSII photoinactivation was reached 2 h after submitting LL and HL cultures to 750 µM and 25 µM respectively (Fig. 2). Because of the large divergence in dose and kinetics responses to MV between LL-and HL cells, it was not possible to find MV concentrations leading to 50 % decrease of quantum yield at the same time for both light acclimations. Thus, array analyses for MV were performed on HL and LL cultures incubated at the same MV concentration (50 µM) but harvested once PSII quantum yield was halved, i.e. after 1 and 3.5 h of stress respectively (Fig. 2).
For every experiment, a 300 mL volume of stressed culture was harvested by centrifugation (10,000 x g for 7 min, Eppendorf 5417R) in presence of 10 mg L -1 of the nonionic surfactant Pluronic F-68 (Sigma Aldrich). Pellets were then flash-frozen in liquid nitrogen and stored at -80°C until analysis. Four independent biological replicates of the experiments were carried out.
Total RNA was extracted with acid pH-guanidinium thiocyanate/phenol/chloroform (TRIzol reagent, Invitrogen) following a modified procedure from Millican and Bird (Millican and Bird, 1998) and purified on column (QIAgen RNeasy Mini Kit). Briefly, frozen cells were resuspended in at 65°C and flash-freezing in liquid nitrogen. Following 10 min incubation at 65°C with vortexing, the clear lysate was vigorously mixed with 100 µL of chloroform and incubated at room temperature for 5 min in microtubes containing an interphase gel barrier (Phase Lock Gel Heavy, 5 PRIME). After centrifugation at 9,000 x g for 15 min, upper aqueous fraction was collected and the total RNA was purified following the QIAgen RNeasy Mini Kit procedure. Two consecutive DNAse treatments (QIAgen RNase-Free DNase Set) were performed on the RNeasy column during the purification step, as recommended by the manufacturer. RNA was retrieved by two consecutive elutions with 30 μ l of RNA-free water (Ambion) and its quality checked by capillary electrophoresis on a Bioanalyzer 2100 using the Procayote total RNA nano Chips (Agilent, Santa Clara, CA, USA).

Microarray analysis -Array design, cDNA labelling and hybridization
Microarray experiments were performed using a homemade array targeting 2,497 protein coding genes out of the 2,586 genes identified so far in the Synechococcus sp. WH7803 genome (Dufresne et al., 2008), as well as 126 potential small RNA (WR Hess, personal communication).
The 60-mer oligonucleotides were designed and synthesized by Eurogentec (Liege, Belgium), resuspended at a final concentration of 20 µM into a 1X spotting buffer (Schott Nexterion, Jena, Germany) and spotted in duplicates on Schott Nexterion slides using facilities of the Rennes transcriptomic platform (France).
cDNA synthesis from 5 µg of total RNA and indirect CyDye cDNA labelling were performed using the ChipShot Indirect Labeling and Clean-Up System (Promega, Madison, USA) following the manufacturer procedure. After resuspension, labeled cDNA was quantified using a NanoDrop 1000 spectrophotometer (Thermo Scientific, Wilmington, DE, USA) and vacuum concentrated to use equal amounts of Cy3-and Cy5-labelled probes (20 to 100 pmol) for hybridization. All hybridizations were performed on 4 independent biological replicates. A pool of RNA from all samples investigated in this study was used as reference sample. This allowed to maximize the number of spots with a significant signal. Furthermore, in order to minimize bias due to differential dye bleaching or unequal incorporation of the Cy3 and Cy5 dyes during RT reactions, 2 out of the 4 replicate samples for each condition were hybridized in dye swap experiments. hybridization chamber (Telechem, Atlanta, USA) and hybridized for 17 h in a water bath at 55°C.
The washing consisted in steps of 5 min at 55°C in 4 X SSC, 5 min in the same buffer at room temperature, 1 min in 0.2 X SSC at room temperature and finally 1 min in 0.1 X SSC at room temperature. Immediately after washing, slides were briefly rinsed with distilled water before being dried by centrifugation at 300 × g for 90 s. Scanning was performed immediately after this step.

Microarray analysis -Image Acquisition and statistical analysis
Scanning of the arrays was performed using a Genepix 4000A scanner (Molecular Devices, Sunnyvale, USA). Photomultiplier gain at 570 nm (Cy3) and 660 nm (Cy5) was adjusted automatically using a threshold saturation of 0.001%, as implemented in Genepix 6.0 (Molecular Devices, Sunnyvale, USA). Addressing and segmentation of spots were automatically detected, manually corrected and intensities were quantified using the Genepix 6.0 software. All microarray experiments were MIAME compliant and raw data were deposited under accession number E- We particularly focused our attention on genes which were statistically significant using the Student t-test and/or LIMMA with discovery rate (FDR) lower or equal to 0.05, also taking into account a fold change (FC) cut-off between two conditions (log 2 (FC) < -1 or > 1).

MTAB
Finally, hierarchical clustering analyses (HCA, (Bolstad et al., 2004) and Principal Component Analyses (PCA, (Raychaudhuri et al., 2000) were performed to investigate the technical and biological reproducibility of our results and highlight the main groups of genes, excluding control oligonucleotides, that shared similar expression patterns over multiple conditions. The clustering was performed using the hclust function from the stats R package (R_Development_Core_Team, 2009), using the clustering method "ward" and a Pearson correlation, and PCA analysis was carried out using the FactoMineR package (Husson et al., 2008).
Analyses were performed on a subset of genes corresponding to the genes significantly differentially expressed (FDR ≤ 0.05 using t-test and/or LIMMA) in at least one of the pairwise comparisons (1,202 genes left), except for two dimension (genes vs. arrays) HCA analysis where no FDR restriction had been applied but where oligonucleotides targeting unknown proteins, hypothetical proteins and small RNA were removed (1548 genes left).

Real time quantitative PCR
To validate microarray data, real time quantitative PCR (hereafter qPCR) was performed on ten Synechococcus sp. WH7803 genes. This set includes genes that were either differentially expressed in microarray analyses or representative of key processes, including ROS detoxification processes, and photosystem II turnover. Gene specific primers for both reverse transcription and qPCR were designed using PrimerExpress™ software V2.0 (Applied Biosystems ; Supplemental Table S2) and the linearity of the cDNA content to C T (cycle at threshold) ratio was checked for every set of primers within the dilution range used. Reverse transcription was carried out on 100 ng RNA using SuperScriptII reverse transcriptase (Gibco-BRL, Gaithersburg, MD) as previously described (Six et al., 2007b). qPCR was performed on the 1:100 diluted cDNA obtained, using the DNA Engine/Chromo4 Real Time PCR-Detector (BioRad, Hercules, CA, USA) and using absolute SYBR Green ROX Mix (Abgene, Epsom, UK), as previously described (Garczarek et al., 2008), in the presence of 200 nM primers, except for the sodB primers which were used at 500 nM. The relative expression of each gene between two conditions was calculated using the 2 -ΔΔCT method (Livak and Schmittgen, 2001), using the rnpB gene as an internal standard (Mary and Vaulot, 2003). All genes and pairwise comparisons tested showed a similar response (up-or downregulation) in qPCR and microarray experiments (Supplemental Table S2; Pearson's correlation coefficient of 0.85).       genes responding to H 2 O 2 and/or MV) between LL-and HL-acclimated cells. C. Comparison of the pooled responses to oxidative stress (+/-reactive oxygen stressors) in LL (upper signs/values) and HL (lower signs/values) cells and of the genes differentially expressed in HL-compared to LLacclimated conditions. Concordant and discordant genes (i.e. common genes regulated in opposite way between two conditions) of this diagram are listed in Table 1.

Figure 8
Two-way cluster analysis diagram of gene expression data. This analysis is based on 1,548 genes, corresponding to all genes present on the array but microarray controls, small RNA oligonucleotides as well as unkown and hypothetical protein encoding genes. Each row in the diagram represents a gene and each column a light acclimation or a stress condition. The dendrogram at the top is similar to Fig. 6 except that each column was obtained from the mean of the 4 biological replicates performed in each condition. The color saturation represents differences in gene expression across the samples; red indicates higher than the median expression (black), and green indicates lower than median expression. The color intensity indicates degree of gene regulation. The boxes to the right of the diagram represent gene clusters A−E and show some of the genes present in these clusters and belonging to specific functional categories. Genes showing a statistically significant differential expression in at least one condition (Student's t-test and/or LIMMA with FDR≤0.05) are indicated in blue.

Supplemental Figure S1
Effect of methylviologen on ROS production through photosynthetic electron transport.
In the absence of methylviologen (MV), linear electron flow (LEF) from water to NADP throught Photosystems (PS) II and I, and cyclic electron flow (CEF) around PSI, merge at the Cytochrome  Supplemental Figure S3 Graphical representation of the predicted LexA motif identified in Synechococcus sp. WH7803.
The overall height of stacked symbols indicates the sequence conservation at the corresponding position, while the height of each symbol indicates the relative frequency of nucleotide. The logo has been obtained from the WebLogo application (http://weblogo.berkeley.edu).

Supplemental Table S1
Complete set of gene expression data as measured by microarray analyses. This