**1. Introduction**

Microbes in aquatic habitats need to adapt to frequent changes in environmental parameters like temperature, O2-saturation, or light conditions. While phototrophic bacteria can take advantage of pigment-protein complexes to use light energy for ATP production, they face the special challenge of photooxidative stress: (bacterio-) chlorophylls can act as photosensitizers and transfer energy to the ground state triplet oxygen (3O2), causing a spin conversion in the π\*2p orbital that generates

highly reactive singlet oxygen (1O2). While other photosensitizers like humic acids also contribute to photooxidative stress, (bacterio-) chlorophyll *a* is regarded as the main cause of 1O2-generation in photosynthetic bacteria. Independently of light, processes like lipid peroxide decomposition or hypochloric acid reacting with hydrogen peroxide can generate 1O2 [1].

Facultative anoxygenic phototrophic bacteria of the genus *Rhodobacter* adjust their lifestyle to the light and oxygen conditions. Due to a high metabolic versatility, they do not rely on photosynthesis for ATP production, but can also perform aerobic or anaerobic respiration or fermentation. They do not form photosynthetic complexes at high oxygen concentrations, and at intermediate oxygen concentration, light inhibits the accumulation of pigment-protein complexes [2,3], which reduces the risk of photooxidative stress. Several protein regulators, including redox-responsive factors, photoreceptors, and even proteins with dual-sensing function like AppA [2], but also RNA regulators [4,5], contribute to the regulated formation of photosynthetic complexes.

Nevertheless, situations that cause photooxidative stress cannot be completely avoided, and consequently, mechanisms to defend this stress are important for survival. As seen across all kingdoms, 1O2 damages a wide variety of biomolecules, including nucleic acids, amino acids, fatty acid lipids or thiols, and glutathione [1,6,7]. Singlet oxygen can directly oxidize its targets or generate other reactive oxygen species (ROS) like endo- or hydroperoxides via (4 + 2) cycloaddition or the ene reaction [8]. Without a proper cellular response, 1O2-stress can be cytotoxic [9,10]. In the case of DNA, the mutagenic potential of 1O2 in *Escherichia coli* can be assigned to the oxidation of guanine sites to 8-oxo-7,8-dihydro-2--deoxyguanosine (8-OHdG), which is susceptible to single strand breaks [9,11]. Regardless of the occurrence of 8-OHdG, 1O2 can also a ffect RNA, specifically viral RNA, by mediating RNA-protein-crosslinking, as shown for photoinactivation of HIV-1 [12]. As 1O2 can form peroxides, it also targets unsaturated fatty acids, causing lipid peroxidation which impairs membranes in their potential, integrity, or transport activities [13,14]. However, due to their high abundance in the cell, proteins are the primary targets of 1O2 [15]. Protein damage by 1O2 may often be traced back to the oxidization of amino acids containing sulfur or aromatic compounds [7,16], but other ROS generated by 1O2 might target the proteome as well. Unfolded or aggregated proteins and the loss of enzyme activities are likely consequences of 1O2 [17].

For more than a decade, *R. sphaeroides* have served as the bacterial model organism to elucidate the photooxidative stress response. Increased expression of certain genes in response to 1O2 was demonstrated, and an important role of the alternative sigma factors RpoE, RpoHI, and RpoHII in this response was revealed [10,18–21]. An early step in 1O2-dependent gene activation is proteolytic degradation of the antisigma factor ChrR [22,23]. The released RpoE sigma factor directly activates a small number of genes like the DNA photolyase gene *phrA* or *cfaS* (cyclopropane fatty acyl-phospholipid synthase) [24,25], but also the *rpoHII* gene. RpoHII, together with RpoHI, activates a high number of genes upon photooxidative stress, but also in response to other stresses [20,21]. Genes that are activated upon photooxidative stress have functions, e.g., in the detoxification of toxic molecules like peroxides or methylglyoxal, in protein quality control and turnover, in 1O2 quenching, DNA repair, and transport [1,26]. Although carotenoids provide protection against 1O2 in *R. sphaeroides*, genes for carotenoid synthesis are not activated by 1O2 in this bacterium [27,28].

Regarding the photoprotective function of carotenoids, both *R. capsulatus* and *R. sphaeroides* accumulate mainly spheroidene (SE) under anaerobic conditions and spheroidenone (SO) under (semi-) aerobic conditions [27,29–33]. An oxygen-activated spheroidene monooxygenase (CrtA) causes this shift and incorporates a keto-group into SE to form SO [31,32]. The shift from SE to SO helps *Rhodobacter* to counteract 1O2 [34]. The introduced keto-group stabilizes the intramolecular charge transfer state of excited carotenoids by binding to the reaction center of the light-harvesting complex I. Due to its low energy, the intramolecular charge transfer state of carotenoids enables the quenching of 1O2 without sacrificing light-harvest. This could explain why di fferent carotenoid-deficient mutants of *R. sphaeroides* showed decreased survival rates under photooxidative stress when (hydroxy-) SO amounts were very low [27].

Many small RNAs (sRNAs) are induced by 1O2 [35], and for some, the regulatory function could be elucidated. The sRNA-mRNA interactions are often stabilized by the RNA chaperone Hfq, which is another crucial element of the photooxidative stress response in *R. sphaeroides* [36]. Some of these sRNAs (CcsR1-4, Pos19) are involved in balancing the glutathione pool and in the downregulation of the pyruvate dehydrogenase complex and aerobic electron transport, a primary source of ROS [37,38]. Other sRNAs (SorY, SorX) reduce the metabolic flux into the tricarboxylic acid (TCA) cycle or affect polyamine transport [39,40]. a switch from glycolysis to the pentose phosphate cycle and reduced activity of the TCA cycle upon oxidative stress reduce the production of the pro-oxidant NADH and increase production of the protective NADPH. An integrative "omics" approach supports the importance of posttranscriptional regulation in the 1O2 response of *R. sphaeroides* [28].

*R. capsulatus*, another member of the *Rhodobacteraceae*, shares a very similar life style with *R. sphaeroides*, and was also intensely studied with regard to its adaptation to different oxygen- and light conditions [41,42]. Under high oxygen tension, *R. capsulatus* cultures show more pigmentation than *R. sphaeroides*, implying a faster adaptation to phototrophic conditions but a higher risk of 1O2 production. However, the response to 1O2 has not been elucidated in *R. capsulatus*. In this study, we applied omics approaches to analyze and compare the response of the two *Rhodobacter* species to photooxidative stress. Although both species share the same habitats, our findings sugges<sup>t</sup> individual strategies to defend against photooxidative stress in addition to a common core response.

#### **2. Materials and Methods**

#### *2.1. Bacterial Strains and Growth Conditions*

*Rhodobacter* strains (Table S1) were cultivated at 32 ◦C in minimal medium containing malate as a carbon source [43]. For microaerobic conditions (~25 μM O2), cultures were incubated in Erlenmeyer flasks with a culture volume of 80% and shaking at 140 rpm. To cultivate *Rhodobacter* under aerobic conditions (160–180 μM O2), cultures were grown either in baffled flasks with shaking at 140 rpm and a culture volume of 20%, or in flat glass bottles gassed with air. To establish phototrophic growth, airtight flat glass bottles were completely filled with medium and cultures were illuminated continuously with white light (60 <sup>W</sup>·m<sup>−</sup>2; fluorescent tube: Omnilux 18W). In order to shift *Rhodobacter* between two different growth conditions, exponentially growing cultures (OD660 of ~0.4) were diluted to an OD660 of 0.2.

#### *2.2. Photooxidative Stress Experiments*

Photooxidative stress experiments were carried out as previously described in Glaeser and Klug, 2005 [27]. In short, pigmented cultures from microaerobic cultivation were shifted to aerobic conditions in air-gassed flat glass bottles in the dark. Methylene blue was added at a final concentration of 0.2 μM. After an OD660 of ~0.4 was reached, cultures were exposed to 800 W·m<sup>−</sup><sup>2</sup> white light to generate 1O2 (photooxidative stress).

For zone of inhibition assays, exponentially growing cultures were diluted into soft agar (0.8%, *w*/*v*) and poured onto malate minimal salt medium agar (1.6%, *w*/*v*). Five microliters of methylene blue (10 μM) were spotted onto a filter paper disk, which was placed in the center of the agar plate. Cultures were incubated for 48 h at 32 ◦C under illumination with 20 W·m<sup>−</sup><sup>2</sup> white light.

#### *2.3. Analysis of Pigmentation*

BChl *a* and carotenoids were extracted and measured as described in Glaeser and Klug, 2005 [27]. Briefly, 1 mL samples of *Rhodobacter* cultures were harvested at 17,000× *g* for 5 min. Pellets were resuspended in 50 μL ddH2O and mixed with 500 μL of acetone/methanol (7/2, *v*/*v*) by vortexing for 30 s. Samples were centrifuged at 17,000× *g* for 5 min, and the absorption of the supernatant was measured in a Specord 50 Plus spectrometer (Analytik Jena, Jena, Germany), using acetone/methanol (7/2, *v*/*v*) as a reference. The carotenoid and BChl *a* concentrations were calculated from the absorptions

at 484 and 770 nm, respectively, with extinction coe fficients of 128 mM−1·cm<sup>−</sup><sup>1</sup> for carotenoids [44] and 76 mM−1·cm<sup>−</sup><sup>1</sup> for BChl *a* [45]. Concentrations were normalized to the OD660.

#### *2.4. Measurement of Reactive Oxygen Species*

Singlet oxygen levels were measured using the fluorescent probe Singlet Oxygen Sensor Green (SOSG, Molecular Probes, Eugene, OR, USA). a SOSG stock solution of 100 μM was prepared in HEPES bu ffer (40 mM, pH 7, 1% methanol). Six microliters of the SOSG stock solution were added to 114 μL culture samples (final SOSG concentration of 5 μM). Technical duplicates were incubated for 30 min at 32 ◦C and 450 rpm in a Vibramax 100 shaker (Heidolph Instruments, Schwabach, Germany). Samples were either kept in the dark or illuminated with 800 W·m<sup>−</sup><sup>2</sup> red light. Samples without SOSG served as background controls. Cells were centrifuged at 8,000 rpm for 5 min and resuspended in 100 μL HEPES bu ffer (40 mM, pH 7, 1% methanol). Fluorescence intensities (excitation 500 nm, emission 532 nm) were measured in an Infinite M200 microplate reader (Tecan, Crailsheim, Germany). After subtraction of the background control, fluorescence intensities were normalized to BChl *a* levels. Ratios between illuminated samples and dark controls were subsequently calculated.

General ROS levels were measured as previously described [43], using the oxidation-sensitive fluorescent probe 2,7-dihydrodichlorofluorescein diacetate (H2DCFDA, Molecular Probes, Eugene, OR, USA). Culture samples of 100 μL were incubated at 32 ◦C with H2DCFDA (final concentration of 10 μM) for 30 min and shaking at 140 rpm in technical triplicates. a culture sample without H2DCFDA served as background control. Fluorescence intensities (excitation 492 nm, emission 525 nm) were measured in an Infinite M200 microplate reader (Tecan, Crailsheim, Germany). After subtraction of the background control, fluorescence intensities were normalized to the OD660.

#### *2.5. Transcriptome Analysis by RNA-Sequencing*

#### 2.5.1. Sample Preparation for RNA-seq

Cultures of *R. capsulatus* were shifted from microaerobic to aerobic growth in the dark followed by photooxidative stress as described by Bergho ff and colleagues [28]. Samples of 20 mL before (0 min) and after stress (10 min) were collected, cooled on ice, and centrifuged at 10,000× *g* for 10 min at 4 ◦C. Cell pellets were resuspended in 1 mL minimal medium and centrifuged at 10,000× *g* for 10 min at 4 ◦C. RNA was extracted via the hot phenol protocol [46]. The RNA was resolved in RNase-free water (Roth) and treated with DNaseI (Invitrogen, Carlsbad, CA, USA) to remove traces of DNA. a test PCR (rpoZ-for: 5--GAT GAT CTG CGC GAG CGT CT-3-; rpoZ-rev: 5--CCT TGC GCG TCC ATC AAT GC-3- ) was performed to ensure that the RNA was free of DNA. RNA integrity was assessed using the Agilent RNA 6000 Nano Kit on the Agilent 2100 Bioanalyzer (Agilent Technologies, Santa clara, CA, USA) to ensure high quality RNA (RIN ≥ 9) for downstream processing. rRNA was depleted from 5 μg of total RNA using the Ribo-Zero rRNA Removal Kit (Gram-Negative Bacteria, Epicentre Biotechnologies, Madison, WI, USA) as recommended by the manufacturer. One microliter of either 1:10 diluted ERCC ExFold RNA Spike-in Mix 1 or Mix 2 (Ambion, Austin, TX, USA) was added to 1 μg of rRNA-depleted RNA. To create 5--monophosphorylated RNA, rRNA-depleted RNA (including ERCC Spike-in Mixes) was treated with RNA 5- polyphosphatase as recommended by the manufacturer.

#### 2.5.2. Strand-Specific Library Preparation and Illumina Sequencing

Strand-specific RNA-seq cDNA library preparation and barcode introduction was based on RNA adapter ligation as described earlier [47]. The quality of the libraries was validated using an Agilent 2100 Bioanalyzer (Agilent Technologies) following the manufacturer's instruction. Cluster generation was performed using the Illumina cluster station. Single-end sequencing on the HiSeq2500 followed a standard protocol. The fluorescent images were processed to sequences and transformed to FastQ format using the Genome Analyzer Pipeline Analysis software 1.8.2 (Illumina, San Diego, CA, USA). The sequence output was controlled for general quality features, sequencing adapter clipping, and demultiplexing using the fastq-mcf and fastq-multx tool of ea-utils [48].

#### 2.5.3. Read Mapping, Bioinformatics and Statistics

The quality of the sequencing output and potential contamination was analyzed using FastQC (Babraham Bioinformatics, http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Identified adapter contamination and remaining artificial sequence (barcode) were removed using program fastx\_trimmer from the FASTX-210 toolkit version 0.0.13 (http://hannonlab.cshl.edu/fastx\_ toolkit/). On the 3--end, reads were trimmed if the per base Phred score fell short of 20. Trimmed reads with a remaining length < 20 nucleotides were discarded. All sequenced libraries were mapped to the *R. capsulatus* genome (accession no. NC\_014034) and the pRCB133 plasmid (accession no. NC\_014035.1) using Bowtie2 (version 2.1.0) in end-to-end alignment mode [49]. After read mapping, the resulting bam files were filtered for uniquely mapped reads using SAMtools (both strands) [50]. The determined uniquely mapped read counts served as inputs to DESeq2 [51] for the pairwise detection and quantification of di fferential gene expression. For DESeq2 parametrization, we used a beta prior and disabled Cook distance cut o ff filtering. All other parameters remained unchanged. In addition, RPKM (reads per kilobase max. transcript length per million mapped reads) values were computed for each library from the raw gene counts. The list of DESeq2 determined di fferentially expressed genes (DEGs) was filtered with a conservative absolute log2 fold change cuto ff of at least 1 and a cuto ff for a multiple testing corrected *p*-value of at most 0.05.

#### 2.5.4. ERCC Spike-in Control Analysis

To assess the platform dynamic range and the accuracy of fold-change responses, ERCC RNA Spike-in controls were used. Spike-in control sequences were added to the *R. capsulatus* reference genome/annotation prior to read alignment and read counts for Spike-in controls were determined, along with normal gene counts with program htseq-count. Further data analyses and the generation of dose- and fold-change response plots were performed as described by the manufacturer (Ambion, Carlsbad, CA, USA).

#### 2.5.5. RNA-seq Data Accessibility

The RNA-seq analysis can be found in Table S2. Raw RNA-seq data have been deposited in NCBI's Gene Expression Omnibus, and are accessible through GEO Series accession number GSE134200.
