Genome-wide analysis on Chlamydomonas reinhardtii reveals impact of hydrogen peroxide on protein stress responses and overlap with other stress transcriptomes

SUMMARY

Reactive oxygen species (ROS) are produced by and have the potential to be damaging to all aerobic organisms. In photosynthetic organisms, they are an unavoidable byproduct of electron transfer in both the chloroplast and mitochondrion. We employ the reference unicellular green alga, Chlamydomonas reinhardtii, to identify the effect of H₂O₂ on gene expression by monitoring the transcriptome changes in a timecourse experiment. Comparison of transcriptomes from cells sampled immediately prior to addition of H₂O₂, and 0.5 and 1 h subsequently revealed 1278 differentially abundant transcripts. Of those transcripts that increase in abundance, many encode proteins involved in ROS detoxification, protein degradation and stress-responses, whereas among those that decrease are transcripts encoding proteins involved in photosynthesis and central carbon metabolism. In addition to these transcriptomic adjustments, we observe that H₂O₂ addition is followed by an accumulation and oxidation of the total intracellular glutathione pool, and a decrease in photosynthetic O₂ output. Additionally, we analyze our transcriptomes in the context of transcript abundance changes in response to singlet O₂ (O₂*), and relate our H₂O₂-induced transcripts to a diurnal transcriptome, where we demonstrate enrichments of H₂O₂-induced transcripts early in the light phase, late in the light phase and 2 h prior to light. On this basis several genes that are highlighted in this work may be involved in previously undiscovered stress remediation pathways or acclimation responses.

INTRODUCTION

Reactive oxygen species (ROS) are potentially harmful but unavoidable byproducts of aerobic respiration and oxygenic photosynthesis. Among the dominant sources of ROS in the cell is the reduction of univalent O₂ generating superoxide (O₂⁻) in complexes I and III of the mitochondria. In plants, the plastid is an additional source of ROS, where energy transfer within the reaction center of PSII generates singlet oxygen (O₂*), and photoreduction of O₂ to O₂⁻ commonly occurs at PSI (reviewed in (Apel and Hirt 2004)). Although several independent detoxification systems exist, the transient accumulation of ROS is a vital signal that allows the cell to regulate the metabolic processes that generate these toxins and ensure that appropriate protective measures are taken to avoid irreversible damage to proteins, lipids and nucleic acids. The cell can orchestrate the production of enzymes to reduce ROS concentrations and repair damage; otherwise, if the damage is irrevocable, autophagy or apoptosis may ensue.

The ability of H₂O₂ to elicit specific induction or repression of gene expression is well documented (Foyer and Noctor 2009, Gough and Cotter 2011, Stone and Yang 2006). Many of these changes serve to increase the cell's capacity to detoxify H₂O₂, usually through the activity of catalases (to produce H₂O and O₂) or peroxidases (to produce H₂O). Unlike catalase, peroxidase must be reduced by an external electron donor (such as glutathione for glutathione peroxidases) to reduce H₂O₂. Therefore, a common response to H₂O₂ stress is the biosynthesis of antioxidants such as glutathione and ascorbate. Other responses are not necessarily directed at managing H₂O₂ levels but at mitigating the damage produced by H₂O₂. Conserved transcriptional responses include expression of molecular chaperones and increased capacity for protein degradation (Vandenbroucke et al. 2008). Poisoning appears to originate with the Fenton reaction (Walling 1975), whereby H₂O₂ oxidizes solvent-exposed Fe²⁺, damaging Fe-S clusters (Jang and Imlay 2010) and inactivating mononuclear Fe proteins (Anjem and Imlay 2012, Sobota and Imlay 2011). In addition to the injury imposed on these susceptible Fe-dependent proteins, the Fenton reaction produces the hydroxyl radical (^•OH), which cannot be enzymatically detoxified and reacts at virtually diffusion-limited rates with most biomolecules. In many ways, H₂O₂ stress can resemble an unfolded protein response, likely due to the misfolding or inactivation of oxidized proteins.

In this study, we employed the single-celled green alga Chlamydomonas reinhardtii as a robust system for understanding the global impact of H₂O₂ stress on the transcriptomic landscape in a photosynthesizing cell. In recent years, Chlamydomonas has become a premiere organism for the interrogation of transcriptional responses to stress through the use of RNA-sequencing (Blaby et al. 2013, González-Ballester et al. 2010, Schmollinger et al. 2014, Wakao et al. 2014). In addition to surveying the global gene expression response to H₂O₂, we validated several of our observations by physiological and biochemical approaches. The wealth of data from previously published transcriptomes in this organism enabled us to contextualize our results with other stresses, allowing us to identify both overlapping and unique responses.

MATERIALS AND METHODS

Strains and culture conditions

Strain CC-4532 (Mets strain 2137 mt⁻) was used throughout this study. Routine growth was performed in Tris-acetate-phosphate (TAP) using Hutner's trace element mix (Hutner et al. 1950). Cultures were grown in Innova incubators (New Brunswick Scientific, Edison, NJ) at 24°C, agitated at 180 rpm with continuous light (95 μmol m⁻² s⁻¹, 6 cool white fluorescent bulbs at 4100K and 3 warm white fluorescent bulbs at 3000K per incubator). Unless indicated otherwise, freshly aliquoted H₂O₂ (Fisher) was added to cultures to a final concentration of 1 mM at ~2 × 10⁶ cells ml⁻¹. Culture densities (in cells per ml) were determined with a hemocytometer.

H₂O₂ RNA-Seq read realignment and analysis

Transcriptomic data analyzed in this study was initially acquired as described in (Urzica et al. 2012a). The sequence reads resulting from this previous study were aligned to v5.5 of the reference genome (Merchant et al. 2007) (available at ftp://ftp.jgipsf.org/pub/compgen/phytozome/v9.0/Creinhardtii/) using STAR (Dobin et al. 2013), default parameters plus –alignIntronMax 10000, and expression estimates were determined and normalized in terms of fragments per kb of exon per million fragments (FPKM) using cuffdiff (Trapnell et al. 2010) default parameters.

Fv/Fm measurements

Fv/Fm measurements were determined using a FluorCam 800MF (Photon Systems Instruments, Czech Republic). 3 × 10⁷ cells were sampled from cultures and concentrated onto a Whatman 25 mm circular filter paper disc using a vacuum. Cell-containing filter discs were dark-adapted for 15 min before performing Fv/Fm measurements using FluorCam6.0 software (Photon Systems Instruments, Czech Republic).

H₂O₂ degradation (FOX) assays

The concentration of H₂O₂ in culture flasks was determined by FOX assays, as described by Nourooz-Zadeh et al (Nourooz-Zadeh 1999). Briefly 0.1 mL culture samples were taken from cultures at a cell density of 2 × 10⁶ cells mL⁻¹, or non-inoculated flasks for cell-free control samples. Samples containing cells were centrifuged (16,100 ×g, 3 min) and the supernatant transferred to fresh tubes. 0.9 mL FOX assay reagent A (25 mM 93% sulfuric acid, 2.5 mM ammonium iron sulfate, 1 mM xylenol orange) was added, the mixture incubated at room temperature for 10 min, and the absorption determined at 560 nm. H₂O₂ concentrations were determined by reading absorption measurements off a standard curve prepared with known concentrations. Measurements were performed on three separate cultures (biological replicates). Measurements were taken at 0 h (immediately after addition of 1 mM H₂O₂), and at 0.5, 1, 2, 3, and 4 h thereafter.

O₂ consumption and evolution

Oxygen evolution and consumption rates were measured on a standard Clark-type electrode (Hansatech Oxygraph with a DW-1 chamber) and the data analyzed with the Hansatech OxyLab software v1.15. Experiments were performed using 2 mL of culture at 2 × 10⁶ cells mL⁻¹ in the presence of 20 mM acetate and 10 mM KHCO₃. Samples were taken 0.5 h prior to addition of H₂O₂ (added to a final concentration of 1 mM), immediately after addition (t0), and at 0.5, 1, 2 and 4 h subsequent to addition of H₂O₂. Measurements were performed on three separate cultures (biological repeats). The rate of respiration rate was measured as oxygen consumption over a period of 4 min in the dark. The rate of photosynthetic O₂ evolution was measured for 4 min in the light (75 μmol m⁻² s⁻¹) after a 4 min acclimation period and was calculated as the difference between oxygen evolution in the light and oxygen consumption in the dark.

Glutathione pool measurements

Chlamydomonas cells treated with 1 mM H₂O₂ were collected at 0, 0.5, 1, 3 and 4 h by centrifugation (5000 ×g, 5 min), washed once in 50 mM sodium phosphate (pH 7.5) solution, resuspended in 0.2 N HCl and lysed by two cycles of freeze/thaw at −80°C. Crude extracts were cleared by centrifugation at 15000 ×g for 20 min at 4°C. 500 μl of sample was neutralized by adding 50 μl of 50 mM NaH₂PO₄ (pH 7.5) and 0.2 N NaOH to a final pH of between 5 and 6. The neutralized sample was directly used for measuring total glutathione (reduced (GSH) plus oxidized (GSSG) glutathione) by the recycling assay initially described by Tietze (Tietze 1969) and adapted from Queval and Noctor (Queval and Noctor 2007). The method relies on the GR (Glutathione Reductase)-dependent reduction of 5,5’-dithiobis(2-nitro-benzoic acid) (DTNB; Sigma, D8130). Oxidized glutathione was measured after treatment of neutralized sample with 10 mM 4-vinylpyridine (VPD, Sigma V320-4) for 30 min at 25°C. To remove excess VPD, the derivatized sample was centrifuged twice at 15000 ×g for 20 min at 4°C. To measure total glutathione or GSSG, sample was added to a mix containing 120 mM NaH₂PO₄ (pH 7.5), 300 μM DTNB, 500 μM NADPH, 1 mM EDTA (pH 8), 1 U ml⁻¹ GR (Sigma, G3664), and DTNB reduction was measured at 412 nm. Different GSH (Sigma, G4251) concentrations ranging from 0 to 5 μM were used as standards. For each time point, three independent biological replicates were measured with at least 3 technical replicates for each sample. Data are represented as mean ± SD (n≥9).

Methods for comparative transcriptome analysis

The same cutoffs as those used for the H₂O₂ analysis (i.e. ≥10 FPKM/RPKM, ≥2.0-fold change) were applied to the expression estimates of previously published datasets (Castruita et al. 2011, Hemschemeier et al. 2013, Malasarn et al. 2013, Ramundo et al. 2014, Urzica et al. 2012b, Wakao et al, 2014). The list of up-regulated transcripts from each dataset was then compared to the list of transcripts that increased in abundance before and 1 h after H₂O₂ addition (referred to as H₂O₂-responsive). For the ClpP-depletion-H₂O₂ overlap, our dataset was composed of transcripts that met out cutoffs and increased in strain DCH16 after 43 h of vitamin addition. For the rapamycin analysis, our dataset was composed of transcripts that met our cutoffs and increased in A31 (wild-type strain in that work) after 8h of rapamycin addition. The p-value for each overlap was calculated using R, with the command sum(dhyper((q:m, k, 17301-k, m))), where q = number of transcripts in the overlap, m = number of transcripts that increased in abundance following H₂O₂ addition, and k = the number of transcripts that increased in abundance in the dataset being compared. Because the datasets we performed these comparisons with were all aligned to the v4 assembly of the Chlamydomonas genome, we converted loci IDs in the H₂O₂ dataset from v5 to v4 (which contains 17301 loci, which is the total population in the equation) using the correspondence table available at http://genome.jgi.doe.gov/pages/dynamicOrganismDownload.jsf?organism=PhytozomeV10. For each transcript in the diurnal cycle, we determined at which time point the max FPKM value is reached. Transcripts that did not reach 10 FPKM at any time point were excluded. We then determined the number of H₂O₂-responsive transcripts in each list of max FPKMs and calculated the p-value associated with the overlap as above.

Accession numbers

Sequenced reads are available on the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) database under accession number GSE34826.

RESULTS

GENOME-WIDE RESPONSES TO H₂O₂

To capture both intermediate and persistent transcriptional responses, cDNA-sequencing (RNA-Seq) was performed on RNA isolated immediately before (0′), 0.5 h and 1 h after the addition of 1 mM H₂O₂ (Figure 1A). Based on the rapid degradation of H₂O₂ from the culture (Figure 1B), we reasoned that primary transcriptional responses would occur within the first 1 h. Indeed, within 0.5 h, the supernatant concentration of H₂O₂ from the cultures was reduced to ~60% and reduced to background levels after 4 h (Figure 1B). The addition of 1 mM H₂O₂ had no noticeable effect on cell number or growth rate (Figure 1C).

Figure 1. H₂O₂ is rapidly decomposed by Chlamydomonas.

(A) For RNA-Seq analysis, strain CC-4532 was grown to a density of 2 × 10⁶ cells mL⁻¹ before collecting an initial sample for RNA preparation (0′). Immediately after this collection, H₂O₂ was added to a final concentration of 1 mM. Subsequent samples were taken at 0.5 and 1 hour after addition. (B) H₂O₂ concentrations were determined by FOX assays. For flasks containing cells, cultures were grown to a density of 2 × 10⁶ cells mL⁻¹. Cell culture supernatant was assayed for H₂O₂ concentration prior to (0′), immediately following (0 h) and at timepoints subsequent to H₂O₂ addition. Cell containing cultures are shown in black, cell-free cultures are shown in white. (C) Cell density measured following H₂O₂ addition. 0′ is defined as the time point immediately prior to H₂O₂ addition, which was added at 0 h. Error bars represent one standard deviation of three measurements.

The sequenced reads were aligned to the v5 assembly of the Chlamydomonas genome (strain CC-503 cw92 mt⁺), and expression estimates were determined for 17,741 (Table S1) loci using the JGI v5.5 gene models (Blaby et al. 2014, Merchant et al, 2007). Roughly 7% of predicted transcripts (1278) were differentially abundant (defined as ≥2.0–fold change and ≥10 FPKM) between any two of the three timepoints (Table S2). The abundance of 291 transcripts changed between 0′ and 0.5 h (68 down and 223 up), 1251 transcripts between 0′ and 1 h (511 down and 740 up) and 201 transcripts between 0.5 h and 1 h (68 down and 133 up).

Based on manual annotation of these transcripts using Pfam domains (Finn et al. 2014), orthology to Arabidopsis thaliana and/or Saccharomyces cerevisiae (based on reciprocal best blastp hits), or using literature-based functional curation, the highest enriched categories include protein metabolism, nucleotide/nucleic acid metabolism and transcripts that encode proteins related to organelles of the endomembrane system (Figure 2; Table S2). Stress and signaling also factored highly among up-regulated functional groups, whereas the tetrapyrrole synthesis and photosynthesis groups decreased in mRNA abundance in all three permutations of comparisons (i.e. 0′–0.5, 0.5–1 and 0′–1). Roughly 13% (73 genes) of the Greencut, a phylogenetic inventory of chloroplast-related genes (Heinnickel and Grossman 2013, Karpowiczet al. 2011), was differentially regulated across the timecourse (between 0′ and 1 h), with 28 decreasing and 45 increasing in transcript abundance respectively (Table S3). Several of those transcripts increasing in abundance encode heat-shock proteins and other stress-related functions, and it is possible that others of presently unknown function have a role in stress remediation (see below and Table S3). Six genes belonging to the CiliaCut (conserved genes amongst ciliated organisms (Merchant et al. 2007)) were differentially expressed, all of which decreased in transcript abundance between 0′ and 1 h (although only 165 of the 191 genes in the published CiliaCut, originally defined using the v3 genome assembly, could be reliably mapped to v5.5 gene models).

Figure 2. Functional classification of transcripts involved in the H₂O₂ response.

(A) Functional characterization of genes differentially expressed between any two timepoints. The percentage of total is shown in parentheses (B) number of genes in each subcategory within the protein metabolism class of panel (A). Increased and decreased mRNA abundances are shown in blue and red respectively. (C) relative transcript abundance of genes encoding components of the proteasome (normalized to the max abundance). (D) relative transcript abundance of genes encoding ubiquitin and proteins that mediate the ligation (or hydrolysis) of ubiquitin to target proteins. mRNA abundances are presented normalized to maximum. (E) relative transcript abundance of genes encoding putative and known proteases and peptidases (normalized to peak abundance). FPKM values can be found in Tables S1 and S2.

Protein stress

Using both manual annotation and automated MapMan ontologies, the largest proportion of transcripts fell into protein metabolism (Figure 2A–B; Figure S1; Table S2), suggesting a reprioritization of cellular machinery. We found that the transcripts encoding all putative subunits of the 26S proteasome increased in abundance following H₂O₂ addition (Figure 2C; 1.8 to 3.3 fold between 0′ and 1 h). This coincided with an increase in 31 transcripts encoding proteins putatively involved in ubiquitination, and a reduction in transcript abundance for two putative ubiquitin-specific proteases (Figure 2D). An impact on protein degradation following H₂O₂ addition was not limited to the proteasome-ubiquitin pathway. We also found that the transcript abundance for 22 putative proteases and peptidases increased (while 7 decreased) between 0′ and 1 h (Figure 2E). Other indications of protein stress include increased abundance of 38 transcripts encoding either known or putative molecular chaperones and chaperonins (such as HSP20, HSP90, HSP70, DnaJ proteins, and FKBP-type peptidyl-prolyl cis-trans isomerases) across the timecourse (Table S2). While pathways were induced, presumably to assist in protein folding and initiate protein turnover, 48 transcripts encoding proteins associated with the ribosome, enzymes involved in rRNA processing and ribosome modification/assembly were down-regulated, as were transcripts encoding proteins responsible for transcription and several tRNA modifications (Table S2).

ROS detoxification

The dual role of H₂O₂ as a harmful by-product of essential oxygenic metabolic reactions and as a signaling molecule necessitates high-fidelity modulation of its presence and concentration within each cellular compartment. Consequently, all aerobic and photosynthetic organisms encode an array of systems charged with ensuring intracellular H₂O₂ concentrations do not become toxic. Upon exposure to H₂O₂, we observed that several ROS detoxification systems were induced. Although transcripts for some ROS remediation proteins increase in this experiment, they were already relatively highly abundant at 0′, including transcripts encoding catalases and peroxiredoxins (Figure 3 and Table S4). Of the six putative superoxide dismutases (FSD1 and MSD1–MSD5), FSD1, MSD1 and MSD2 moderately increased in transcript abundance ~1.5-fold with maximal expression estimates at 839, 382 and 116 FPKM respectively. Transcript level of MSD3 increased 8-fold between 0′ and 1 h, although transcript abundance remained relatively low (peak 24 FPKM). Neither MSD4 nor MSD5, both predicted to localize to the secretory pathway, were expressed (Table S4). The abundance of transcripts encoding three of five glutathione peroxidases did not change. However, GPX4 and GPX5 transcripts (encoding non-selenocysteine GPX, and considered to use thioredoxin rather than glutathione as electron donor (Dayer et al. 2008)) increased in abundance (peak FPKM values of 20 and 193, respectively). GPX5 expression was previously shown to be induced by multiple ROS, including H₂O₂, and the corresponding protein was found to reduce H₂O₂ using cytosolic TRXh1, but not GSH, as an electron donor (Fischer et al. 2009, Leisinger et al. 2001). Transcripts encoding proteins putatively contributing to the remediation of ROS damage that were also increased included those encoding nucleoredoxin (NRX2 and NRX3), thioredoxin (TRXh1), glutaredoxin (GRX4), methionine sulfoxide reductase (Cre10.g464850), and, as noted previously (Urzica et al. 2012a), four steps in the Smirnoff-Wheeler pathway for ascorbate biosynthesis (VTC2, MDAR1, PNO1 and DHAR1).

Figure 3. Relative abundance of transcripts putatively involved in ROS detoxification subsequent to H₂O₂ addition.

The size of each pie chart is proportional to the total FPKM. Genes and corresponding FPKM values used in this analysis can be found in Table S4.

By contrast, mRNA abundances for 2-cys peroxiredoxin, ascorbate peroxidase or catalase encoding genes did not increase following H₂O₂ addition. Indeed, both CAT1 and CAT2 transcripts decreased in transcript abundance over the timecourse, with CAT1 transcript being highly abundant at 0′ (197 FPKM) (Table S4). In the case of 2-cys peroxiredoxins (abundance of PRX1 and PRX2 transcripts was already high at 0′ (FPKM > 1300)), the capacity for H₂O₂ decomposition may be increased by the induction of two putative sulfiredoxins (encoded by Cre05.g232800 and Cre17.g729950), both of whose transcripts increase in abundance. These proteins can reactivate over-oxidized 2-cys peroxiredoxin, a mechanism considered to play an important role in H₂O₂ signal transduction in mammals (reviewed in (Jeong et al. 2012)) PRX1 and the sulfiredoxin Cre17.g729950 are predicted to localize to the plastid, while PRX2 and the second putative sulfiredoxin (Cre05.g232800) have no predicted signal peptide (Table S4).

We observed little impact on mRNA abundances of tocopherol (vitamin E), polyamine or carotenoid biosynthesis genes. Each of these pathways has been demonstrated to function in ROS detoxification in other organisms (Havaux et al. 2005, Schneider 2005, Zhu et al. 2010), although the tocopherol acts primarily as a scavenger of O₂*. Our data suggest that in Chlamydomonas these metabolites may not be important for acclimating to short term H₂O₂ exposure.

Effect of H₂O₂ on glutathione metabolism

As in other organisms, the glutathione redox cycle is expected to play a central role in maintenance of the cellular redox balance. We observed that abundance of GSH1 mRNA, encoding γ-glutamylcysteine synthetase, the first enzyme in glutathione biosynthesis, increased more than 2-fold between 0′ and 1 h (FPKM 122). In contrast, the abundance of the mRNA encoding the subsequent enzymatic step, glutathione synthetase (GSH2), remained low, and decreased slightly across the timecourse (Figure 4A; Table S4). To investigate the impact that these transcriptional responses had on glutathione levels, we determined the glutathione pool size in response to H₂O₂ and the proportion present in the oxidized state. We observed both oxidation and accumulation of glutathione after H₂O₂ treatment. The glutathione pool size increased linearly and had more than doubled within 4 h of H₂O₂ addition (Figure 4B). The redox state of glutathione was also altered concomitantly due to GSSG accumulation that resulted in a 50% decrease of the GSH/GSSG ratio within 0.5 h. Within 4 h, this oxidation is partially reversed as the relative proportion of the reduced glutathione (GSH) fraction begins to increase (Figure 4C).

Figure 4. Effect of H₂O₂ on transcripts involved in glutathione metabolism and the glutathione pool.

(A) transcript abundance of glutathione synthetase 1(GSH1), glutathione synthetase 2 (GSH2) and four paralogs of glutathione-S-transferase (GST1-3 and FAP179) at 0′ (i.e. immediately prior to 1 mM H₂O₂ addition), 0.5 and 1 h is shown as a heatmap. The scale bar is in log(FPKM). (B) total glutathione, normalized to cellular content at 0′. (C) relative proportion of reduced and oxidized glutathione. Error bars represent the standard deviation of at least three independent cultures with 3 technical replicates for each (n≥9).

We also observed that the abundance of GST1 and GST2 transcripts, encoding two putative glutathione S-transferases, was significantly elevated (~100-fold and ~10-fold respectively between 0′ and 1 h, each peaking at >700 FPKM) (Figure 4A; Table S4). The expression of both genes was previously found to increase in response to ROS (Fischer et al. 2005). Moreover, consistent with increased sulfur requirements for cysteine production as a precursor for glutathione biosynthesis, we observed slight increases (2–fold at 1 h vs. 0′) in expression of a putative cysteine transporter (AOT4) and generally for genes encoding enzymes involved in sulfur metabolism.

Photosynthesis

Previous transcriptomic studies on H₂O₂ exposure in land plants and photosynthetic microbes have noted downregulation of genes encoding proteins with roles in photosynthesis (Desikanet al. 2001, Vandenabeele et al. 2003, Zeller et al. 2005). In our dataset, although the change in abundance for most transcripts did not meet our cut-offs, we observed that in general mRNA abundance corresponding to nuclear genome-encoded subunits of photosynthetic complexes was reduced by 30–50% in response to H₂O₂ (Table S5). The abundance of transcripts encoding enzymes of chlorophyll biosynthesis are also generally reduced to ~50% at 1 h vs. their abundances at 0′. In contrast, CHL1, which encodes the ortholog of chlorophyllase I in Arabidopsis, increased in transcript level 3-fold over the same time period. The impact of H₂O₂ on these transcripts correlated with a reduction in photosynthetic efficiency evaluated by measuring PSII maximum efficiency and O₂ evolution following H₂O₂ addition (Figure 5A and B). Both of these parameters return to pre-H₂O₂ levels after 4 h, at which time the H₂O₂ content of the cell-free supernatant has returned to background levels (Figure 1B), consistent with observations with tobacco (Vandenabeele et al. 2003).

Figure 5. H₂O₂ negatively impacts photosynthetic and respiratory rates, which recover by 4 h.

(A) Fv/Fm measurements were taken in triplicate (independent cultures) immediately prior to addition of 1 mM H₂O₂ (0′), immediately after addition (0 h) and at 1, 4 and 8 h subsequently. Error bars indicate ± standard deviation of the three measurements. (B) O₂ evolution rates in standard growth conditions were determined 0.5 h prior to addition of 1 mM H₂O₂ (0’), immediately after addition (0 h) and at 1 and 4 hours subsequently. Error bars indicate standard deviation between triplicates (independent cultures), at a starting cell density of 2 × 10⁶ cells mL⁻¹. During measurement, cells were exposed to 75 μmol m⁻² s⁻¹ light. (C) O₂ consumption was determined 0.5 h prior to addition of 1 mM H₂O₂ (0’), immediately after addition (0 h) and at 1 and 4 hours subsequently. Error bars indicate standard deviation between triplicates (independent cultures), at a starting cell density of 2 × 10⁶ cells mL⁻¹. Significance, indicated by *, was determined by t-test (n = 3, p < 0.05)

Exceptions to the downward trend include LHCB7, a minor antenna protein associated with PSII (Peers and Niyogi 2008), which increased in abundance by 2.5-fold, peaking at 27 FPKM at 1 h, and TBA1, a putative oxidoreductase proposed to regulate psbA translation in response to redox status (Somanchi et al. 2005), which increased in abundance by 3.3-fold, peaking at 61 FPKM. Consistent with their known/predicted role mitigating photo-oxidative stress, several LHC-like genes rose in transcript abundance during the timecourse, including ELIP2 (2-fold induction, peak 13 FPKM) and ELIP8 (~3-fold induction, peak 162 FPKM) (Table S5).

Respiration

Having observed a reduction in photosynthetic O₂ evolution and in transcript abundance of many genes encoding photosynthetic subunits, we were interested to determine how the mitochondrial electron transport chain (ETC) was affected. Accordingly, we observed a decline in O₂ consumption, with a trough at 0.5 h before recovering to pre-H₂O₂ rates by 2–4 h. However, despite this measurable reduction, the abundance of most transcripts encoding proteins of the mitochondrial ETC does not change (although transcript level need not necessarily correlate with changes in protein level) (Table S6). Deviating from this trend are transcripts encoding assembly factors for complex III (BCS1) and complex IV (SCO1, PET191, and CMC1), whose abundances increase between 0′ and 1 h. These increases coincide with the increased abundance of transcripts encoding the alternative enzymes NDA1 (a type II NAD(P)H dehydrogenase) and AOX1 (an alternative oxidase), which increase 2.2-fold (peaking at 31 FPKM) and 4.3-fold (peaking at 96 FPKM), respectively.

Carbon metabolism

The abundances of many transcripts encoding enzymes of central carbon metabolism, including the glyoxylate cycle, glycolysis/gluconeogenesis and the Calvin Benson (CB) cycle, generally decrease moderately over the timecourse (Table S7). Coincident with a reduction in O₂ consumption, we observed major reductions in mRNA abundance of a few specific metabolism-related genes, specifically those involved in acetate assimilation, including a putative acetate transporter, Cre17.g702900 (Goodenough et al. 2014), ACS3 and MAS1 encoding acetyl-coA synthetase and malate synthase respectively. Transcripts from these three genes exhibit some of the largest reductions in abundance in the dataset (Table S2) and may reflect reduced flux of metabolites towards acetyl-CoA, and respiration. In contrast, we observed a slight increase in mRNA abundance of several lipid and starch synthesis genes, likely induced as part of a general stress response (Table S7). Transcript levels of GPD2, encoding glycerol-3-phosphate dehydrogenase, and linking carbon assimilation via the CB cycle and gluconeogenesis to lipid synthesis, increased 5-fold between 0′ and 0.5 h (peak FPKM 9 and 66 respectively). Previously, it has been speculated that there may be a link between oxidative stress and the carbon concentrating mechanism (CCM), based on observations that CAH1 and CCP1 were repressed in response to a number of reactive oxygen species (Ledford et al. 2007). Our data recapitulate these observations, and we observed reduced abundance of CAH1 transcript (reduced from >500 FPKM at 0′ to 138 at 1h), and also of CAH3 transcript (127 and 87 FPKM at 0′ and 1h respectively), although this appears not to be a general phenomenon of the CCM as the abundances of other known transcript encoding components were either stable or increased slightly.

COMPARATIVE TRANSCRIPTOME ANALYSES

Overlap between H₂O₂ and singlet oxygen

We observed a significant overlap in the number of transcripts that increased in abundance following H₂O₂ and rose bengal (RB) addition, which generates O₂* (Figure 6; (Wakao et al. 2014)). Roughly 15% of the induced H₂O₂ transcripts between 0′ and 1 h were induced by rose bengal (27% of the RB transcripts in H₂O₂; Table S2). The three most highly represented categories common to both datasets were protein metabolism, proteins related to the ER/Golgi/secretory pathway and lipid metabolism (Figure 6A). However, only the ER/Golgi/secretory and lipid metabolism groups were significant (p-value < 0.001). We noted that the effect on proteasome-subunit transcript abundance and transcripts encoding molecular chaperones (such as heat shock proteins) was specific to H₂O₂; the only transcripts encoding HSPs affected by both stresses were HSP70E, HSP70A and CPN60C. The analysis also found that PTOX1 was up in both datasets, as was GPX5 and two nucleoredoxins. Of the top 10 fold-changing transcripts in each pairwise comparison in our H₂O₂ dataset (i.e. 0′ vs. 0.5 h, 0′ vs. 1 h and 1 h vs. 0.5 h), only two transcripts were also in the RB dataset: a glutathione S-transferase and a protein bearing similarity to the ThiJ family (encoded by GST1 and Cre01.g004900 respectively). H₂O_2, but not RB, affected transcripts involved in tetrapyrrole metabolism (Table S2).

Figure 6. Overlap between H₂O₂ dataset and published stress transcriptomes.

(A) functional classification of transcripts whose abundance increased after H₂O₂ and rose bengal (RB) additions. Numbers in parentheses correspond to the number of transcripts in each category. (B) p-values (P) calculated (using the hypergeometric distribution) for enrichment of H₂O₂ transcripts (induced after 1 h) in several published datasets: wild type (WT) and the sak1 mutant after RB addition (Wakao, Chin, Ledford, Dent, Casero, Pellegrini, Merchant and Niyogi 2014), ClpP depletion (Clp), rapamycin addition (RAP; (Ramundo et al. 2014)), growth in dark anoxia (anoxia; (Hemschemeier et al. 2013)), iron limitation (Fe lim.; (Urzica et al. 2012b)), iron deficiency (Fe def.; (Urzica et al. 2012b)), zinc limitation (Zn; (Malasarn et al. 2013)) and Cu deficiency (Cu; (Castruita et al. 2011)).

We found a larger (and more significant) overlap between the H₂O₂ response and the RB response in the sak1 mutant than in the parent (4A⁺). SAK1 is a putative transcription factor, and disruption of the corresponding gene leads to defective transcriptional responses when the mutant is exposed to O₂* stress (Wakao et al. 2014). Roughly 25% of the induced H₂O₂ transcripts between 0 and 1 h were also induced by RB in sak1, vs. 15% in the parent strain (Figure 6B). To qualify the overlap between these datasets, we also determined the overlap between the H₂O₂ response and several other transcriptomes that were noted for their enrichment of transcripts related to oxidative stress (ClpP-depletion and poor Fe nutrition (Ramundo et al. 2014, Urzica et al. 2012b)) or not (rapamycin-treatment, anoxia, poor Cu nutrition and poor Zn nutrition (Castruita et al. 2011, Hemschemeier et al. 2013, Malasarn, et al. 2013)) (Figure 6B). We observed both ClpP-depletion and rapamycin treatment (Ramundo, et al. 2014) had a more significant enrichment of H₂O₂-induced transcripts than did RB addition (based on the p-value calculated using the hypergeometric distribution). Conversely, both anoxia and copper depletion had relatively little overlap (Figure 6B). Among these comparisons, the Fe dataset had a moderate overlap, while this analysis revealed a surprisingly large enrichment of H₂O₂-induced transcripts in the Zn deficient dataset (Figure 6B).

Oxidative stress during the Chlamydomonas cell cycle

A recent diurnal transcriptome experiment identified a “light-stress cluster” of 280 genes that was transiently expressed at the onset of the light phase, but remained near-undetectable across remaining timepoints (Zones et al. In press). Of these 280 transcripts, 99 were significantly differentially abundant in response to H₂O₂ in our experiment (56 increased mRNA abundance, 43 decreased between 0′ and 1 h). Next, we determined the time point representing the peak transcript level for each gene in the Chlamydomonas genome (using a cutoff of at least 10 FPKM at that point) and analyzed the overlap between the transcripts in each time point and those induced by H₂O₂ addition. Again, we identified an enrichment of H₂O₂-induced transcripts at the first time point in the light. We also observed an enrichment of H₂O₂-induced transcripts with peak values towards the end of the day. We did not observe a significant enrichment during the night except, interestingly, at the 22 h time point (2 hours prior to the lights coming on; Figure 7A). For transcripts that we could assign a general function, the 1 h overlap contained transcripts involved in protein folding and nucleic acid metabolism and binding, whereas towards the end of the day there was an increasing enrichment of transcripts related to protein metabolism in particular the proteasome. At the 22 h time point there was an enrichment of transcripts related to ROS detoxification/regulation (Figure 7B; Table S8).

Figure 7. H₂O₂ responsive transcripts during the diurnal cycle.

(A) analysis of the peak transcript abundance of H₂O₂ –induced transcripts during a 12 hour diurnal cycle (Zones et al. In press). p-values (P) calculated (using the hypergeometric distribution) for enrichment of H₂O₂ – induced transcripts at each time point is presented as a bar graph while the number of H₂O₂ – induced transcripts as a percentage of the total number of H₂O₂ –induced transcripts is shown as a circle. The 12 hours in light is shaded with yellow and highlighted with a sun, while the 12 hours of dark is shaded grey and highlighted with a moon. (B) the top 5 functional categories containing H₂O₂ –induced transcripts at the most significant time points in the cycle. (C) same as for panel A except that the percentage and associated probability of rose bengal-induced transcripts from WT and the sak1 mutant is shown. (D) same as panel A except that percentage and associated probability of anoxia-induced transcripts is shown.

For comparative purposes, we repeated this analysis using the RB treatment (parent and sak1) and anoxic transcripts in place of our H₂O₂ dataset. As with the comparisons described above, we saw some enrichment of RB-induced transcripts at time point 1 h and toward the end of the day in both strains, but, in contrast to the H₂O₂ dataset, the most significant overlap was at the 22 h time point. As expected from the redox status of these cells, we saw no enrichment for anoxia transcripts until the last hour of night.

DISCUSSION

In this study, we have leveraged our transcriptomic analysis to provide a comprehensive overview of the impact H₂O₂ addition has on transcript abundance in Chlamdyomonas. A specific focus on photosynthetic metabolism and ROS detoxification mechanisms provides robust predictions of stress-responsive and remediating genes. In addition, by placing these data in the context of related transcriptomes (Castruita et al. 2011, Hemschemeier et al. 2013, Malasarn et al. 2013, Ramundo et al. 2014, Urzica et al. 2012b, Wakao et al. 2014, Zones et al. In press) we have been able to identify overlapping responses, potentially highlighting the contribution of ROS stress to independent stimuli. The ability to survey the level of mRNA abundance of all genes provides a powerful tool to assess the involvement of genes responding to a condition, and thus serves as a means towards hypothesizing roles for genes of previously unknown function.

In the presented work, we identified significant changes in transcript abundance of 1278 genes, of which 175 have a primary gene symbol and 932 have no known biological role (Table S2). Given this, many of these genes may be of particular interest. They could encode previously undiscovered stress response/ROS detoxification systems, especially considering the majority increased in transcript abundance; between 0′ and 0.5 h, 223 transcripts increased, as were 740 between 0′ and 1 h, whereas 68 and 511 transcripts decreased between these same timepoints, respectively (Table S1). One gene that exhibits one of the greatest increases in transcript abundance (at both 0′ vs. 0.5 h, and at 0′ vs. 1 h) is Cre03.g152750, induced >80-fold. The predicted protein sequence encodes a BAG domain, associated with apoptosis and programmed cell death (Kabbage and Dickman 2008, Kang et al. 2006), and a calmodulin-binding domain. That this gene is very lowly expressed (R/FPKM ≤1) in all but one other published Chlamydomonas transcriptome, including RB exposure, suggests it may be responding specifically to H₂O₂ exposure. This notion is supported by its expression profile in a recent diurnal experiment, in which mRNA is detectable at the first light time point, but at no other, and was a member of a light-stress cluster in that work (Zones et al. In press). Cre12.g542050 and Cre04.g226138 are increase in abundance, but do not contain predicted conserved domains and bear no significant similarity to proteins encoded on genomes besides Chlamydomonas and the related green alga Volvox carteri. The absence of predicted protein domains or clear orthologs in other characterized organisms makes these genes recalcitrant to functional predictions, and future investigation employing approaches such as classical/reverse genetics will be required to confirm their involvement in ROS detoxification. This analysis also uncovered a previously unrecognized putative ferritin encoding gene in the Chlamydomonas genome. Cre01.g033300 encodes a protein with high similarity to bacterial DPS (DNA-binding protein from starved cells), which are ferritin-like proteins responsible for sequestering Fe ions as a means to protect DNA from oxidative damage (Almirón et al. 1992, Pulliainen et al. 2005). Interestingly, in response to H₂O₂ Cre01.g033300 transcript was increased roughly 10 fold, but the expression estimates remained low (peak 10 FPKM; Table S2), whereas after RB treatment, this transcript increased roughly 80 fold (peak 223 FPKM; (Wakao et al. 2014)). Also, Cre01.g033300 is clearly regulated by SAK1 (Wakao et al. 2014), suggesting that Fe-catalyzed DNA damage may be a larger threat from singlet oxygen stress compared to H₂O₂. Seventy-five genes (45 of unknown function) belonging to the GreenCut inventory, whose phylogenetic profile suggests they are of importance to green photosynthetic organisms, increased in mRNA level in this study. This implicates these genes in having some role in ROS-stress management; indeed, included within this group are several genes previously identified in a light-stress cluster (Zones et al. In press), as well as numerous characterized stress response enzymes (the largest increase in transcript level is HSP22A, whose mRNA abundance increases 176-fold to 58 FPKM at 1 h from 0.3 at 0′). Furthermore, the Arabidopsis ortholog of one GreenCut member, LHCB7, a PSII antenna protein, has been shown to be expressed under conditions of oxidative stress (Alboresi et al. 2011) (mRNA level increase of 2.5–fold in our data across the timecourse).

In terms of redox detoxification, the transcripts for many quintessential H₂O₂-reducing enzymes did not respond to H₂O₂ treatment, such as ascorbate peroxidase and catalase. Both CAT1 and CAT2 transcript abundance were actually reduced over our timecourse. This is in contrast with other unicellular organisms, such as S. cerevisiae or Synechocystis PCC 6803, but similar to land plants and mammals (Vandenbroucke et al. 2008). The accumulation and oxidation of glutathione in response to H₂O₂ is commonly observed in most organisms (Noctor et al. 2012, Vandenbroucke et al. 2008). The first step of glutathione synthesis, catalyzed by γ-glutamylcysteine synthetase, is the rate-limiting step (Lu 2013, Masip et al. 2006, Noctor et al. 2012). The increased expression of GSH1 in response to H₂O₂ in Chlamydomonas may therefore account for the observed accumulation of glutathione. A similar increased expression of this gene was reported in yeast and mammals under diverse stress conditions (Lu 2013, Wu and Moye-Rowley 1994). By contrast, neither GSH1 nor GSH2 expression is induced by H₂O₂ in Arabidopsis but glutathione accumulation appears to involve induction of genes encoding enzymes involved in cysteine synthesis in the chloroplast (Noctor et al. 2009). Interestingly, several of these genes were also induced in Chlamydomonas after H₂O₂ addition. All of these results suggest that the mechanisms of redox signaling and detoxification of Chlamydomonas are unique since some regulations are shared with higher eukaryotes while others are related to other unicellular organisms. The specificities of Chlamydomonas redox metabolism and its responses to hydrogen peroxide may also be linked to the presence, as in mammals, of selenoproteins including glutathione peroxidases, thioredoxin reductases and methionine sulfoxide reductases. Moreover, the site of H₂O₂ accumulation also effects the transcriptional response, as recently shown in Arabidopsis (Sewelam et al. 2014), and exogenous H₂O₂ may not influence the ROS levels within some compartments significantly. However, the addition of H₂O₂ was clearly able to elicit a generalized reduction in transcripts corresponding to nucleus-encoded photosynthetic complexes and significant remodeling of transcripts related to the endomembrane system. Alternatively, the abundance of catalase transcripts may not be directly related to catalase activity considering the hypothesized role of these proteins in gating H₂O₂ signaling, and the decrease in catalase activity upon H₂O₂ addition, most probably linked to ROS-dependent post-translational regulation (Michelet et al. 2013, Shao et al. 2008).

We noted a large proportion of transcripts encoding proteins involved in protein folding and degradation increased coordinately with a decrease of transcripts encoding proteins involved in protein synthesis. These trends suggest that the stress caused by H₂O₂ addition signals a general reduction in cellular activity during the period of exposure, and suggests the potential for global protein remodeling and/or clearance of oxidized proteins following H₂O₂ stress. One possible conclusion is that these systems are induced to clear oxidized proteins with the simultaneous induction of protein degradation and protein folding while protein synthesis is downregulated. Another possibility is that the induction of proteasome transcripts is a result of feedback, since the proteasome and the ubiquitin activating/conjugating system have been shown to be inhibited by H₂O₂ (Davies 2001) and by glutathionylation (Demasi et al. 2014). The observed induction of protease transcripts may be responsible for the degradation of oxidatively damaged proteins in compartments and/or a means to compensate for reduced activity of the proteasome.

Moderate reductions in transcript abundance of CB cycle and carbon metabolism genes are consistent with previous reports that algae experience reductions in CO₂ fixation in response to H₂O₂ exposure (Takeda et al. 1995). Reduced abundances of RNAs encoding subunits of the photochemistry machinery were supported physiologically by demonstrating reduced O₂ output subsequent to H₂O₂ exposure. The general reduction observed in transcript level of the photosynthetic machinery is correlated with reduced photosynthetic efficiency at 1 h. Reduction of transcript abundance of photosynthesis related genes may serve to preserve redox balance and prevent production of additional ROS from photosynthetic electron transport.

As with most plant genomes, Chlamydomonas encodes many gene duplications with paralogs encoding multiple proteins of the same family (Merchant et al. 2007, Wu et al. 2015). As with many other biological processes, the genes encoding enzymes with roles in ROS detoxification are present in up to 5 or 6 copies. Unsurprisingly, known ROS-detoxification systems were generally increased in transcript level, although this pattern of expression was not uniform across all associated paralogs, with transcript abundance corresponding to one gene copy far exceeding that of other copies. For example, expression estimates for GST1 and GST2 surmount 700 FPKM, compared to GST3 peak of 70. Similar patterns were observed for many paralogous genes. While mRNA abundance does not necessarily translate to protein abundance, and therefore metabolite flux, it does suggest the possibility that some copies have more prominent roles than others.

The importance of understanding ROS–induced damage responses and protective mechanisms has resulted in H₂O₂-exposure gene expression studies in numerous organisms (for example (Chen et al. 2003, Chuang et al. 2002, Girardot et al. 2004, Mostertz et al. 2004), including several photosynthetic organisms (Desikan et al. 2000, Kobayashi et al. 2004, Li et al. 2004, Vandenabeele et al. 2003). In a recent meta-analysis, a number of conserved gene families was found to be induced in response to H₂O₂ across lineages and kingdoms (Vandenbroucke et al. 2008). Our data are largely in agreement with this study: the most highly conserved induced genes were DNAJ-like heat shock proteins, which also increased in mRNA level in our dataset. Three other protein families were found to be induced in all eukaryotes investigated (A. thaliana, S. cerevisiae, Schizosaccharomyces pombe and Homo sapiens), including GTP-binding proteins, protein kinases and ubiquitin-conjugating enzymes. In our dataset, genes encoding each of these functions were differentially expressed. However, in our transcriptomes, mRNA abundances for genes encoding both GTP-binding proteins and several protein kinases decreased over the timecourse, suggesting the direction of altered expression is not universal. Vandenbroucke et al. also observed a conserved induction of short-chain dehydrogenases/reductases and peroxiredoxins in all unicellular organisms in their study (Synechocystis, S. cerevisiae and S. pombe). Our data in Chlamydomonas, also unicellular, partially support this observation, where transcripts encoding proteins broadly annotated as dehydrogenases/reductases were induced several fold between 0′ and 1 h.

We also compared our H₂O₂ transcriptomes to other relevant Chlamydomonas RNA-Seq experiments. A recently published O₂*-response transcriptome allowed us to make a global comparison of these two oxidative stress conditions. Clearly H₂O₂ and RB addition elicit independent responses, but we did find a significant overlap highlighting a conserved cluster of stress-related genes. In particular, both ER stress and lipid stress ranked highly among the overlapping transcripts. Indeed, Cre01.g004900 transcript level rose in response to both ROS (>50-fold between 0′ and 1 h, peaking at 380 FPKM and 3 fold in response to RB) and bears sequence similarity to yeast HSP31, encoding a methylglyoxylase, which is also induced in response to oxidative stress (Dubacq et al. 2006). This induction may serve to reduce methylglyoxal build-up resulting from lipid peroxidation during both stresses.

Interestingly, the overlap between H₂O₂ and RB treatment of the SAK1 mutant was greater than RB treatment of the wild type. SAK1 encodes a key regulator of the RB response, and O₂* exposure of the SAK1 mutant appears to generate additional oxidative stress that was not observed in the wild type. One possibility is that mis-regulation of SAK1-target transcripts in the presence of RB either exacerbates or leads to a type of oxidative cellular damage that is in common with H₂O₂. In particular we noticed a substantial increase in the number of transcripts encoding proteins putatively involved in protein turnover in the set of transcripts shared between sak1 and H₂O₂ (18 transcripts) versus those shared among sak1, wild type and H₂O₂ (7 transcripts), suggesting that protein damage could be the point at which the H₂O₂ and sak1 signalling intersect.

We also observed a large overlap between H₂O₂-responsive transcripts and both ClpP-depletion and rapamycin addition. ClpP is a stromal protease that is thought to play an essential function in the chloroplast (Nishimura and van Wijk 2015), while rapamycin inhibits TOR signaling, a pathway that promotes cell growth and inhibits autophagy (Dennis et al. 1999). These two transcriptomes were originally performed in conjunction to identify transcriptional changes in the ClpP-depleted strain that are independent of its autophagy-like phenotype (as rapamycin treatment induces autophagy). In this study, we leveraged these datasets to reveal a potential contribution of oxidative stress to the observed transcitptional changes. Indeed, we identified roughly 197 transcripts that increase in abundance following H₂O₂ exposure and ClpP-depletion, while nearly 144 transcripts increase in abundance following H₂O₂ exposure and rapamycin treatment. The increase of 94 transcripts was shared between all three datasets. At this point we can only speculate as to whether the significant overlap we observed reflects the ability of ClpP-depletion and rapamycin treatment to elicit an oxidative stress response. For instance an obvious physiological intersect is induction of autophagy by the three conditions.

In the diurnal cycle, we found a significant number of H₂O₂-induced transcripts with peak transcript abundance after the transition from dark to light and as cells age during the day. Enrichment within the light stress cluster at the beginning of the day and towards the end of the day was shared with singlet O₂-induced transcripts, but the presence of H₂O₂-induced transcripts was associated with a lower p-value. Chlamydomonas cells experience an extended G1 phase during the day during which biomass is accumulated. Cell division is arrested until the night. Based on functional classification of the corresponding proteins, the H₂O₂-induced transcripts with maximal abundance at the end of day largely encode proteins involved in protein turnover. Unexpectedly, we also observed a spike in ROS-induced transcripts at 2 hours prior to the transition to day, possibly highlighting clock-regulated transcripts, which are maximally abundant in anticipation of the transition to the light period. Although there was a significant enrichment of H₂O₂-induced transcripts with peak abundance at this time point, we found a higher significance associated with RB-induced transcripts, possibly highlighting anticipation of photooxidative damage associated with the shock of switching from dark to light. The abundance of the Chlamydomonas ortholog of Arabidopsis CCA1 (Cre06.g275350; circadian clock-associated 1), demonstrated to regulate the induction of oxidative stress responses in that organism (Lai et al. 2012), is slightly reduced throughout our timecourse. In the Chlamydomonas diurnal experiment peak abundance occurs immediately prior to the light period, suggesting Cre06.g275350 is regulated by circadian rhythms here also, and is not influenced by ROS presence.