Light on the cell cycle of the non-photosynthetic bacterium Ramlibacter tataouinensis

Ramlibacter tataouinensis TTB310, a non-photosynthetic betaproteobacterium isolated from a semi-arid region of southern Tunisia, forms both rods and cysts. Cysts are resistant to desiccation and divide when water and nutrients are available. Rods are motile and capable of dissemination. Due to the strong correlation between sunlight and desiccation, light is probably an important external signal for anticipating desiccating conditions. Six genes encoding potential light sensors were identified in strain TTB310. Two genes encode for bacteriophytochromes, while the four remaining genes encode for putative blue light receptors. We determined the spectral and photochemical properties of the two recombinant bacteriophytochromes RtBphP1 and RtBphP2. In both cases, they act as sensitive red light detectors. Cyst divisions and a complete cyst-rod-cyst cycle are the main processes in darkness, whereas rod divisions predominate in red or far-red light. Mutant phenotypes caused by the inactivation of genes encoding bacteriophytochromes or heme oxygenase clearly show that both bacteriophytochromes are involved in regulating the rod-rod division. This process could favor rapid rod divisions at sunrise, after dew formation but before the progressive onset of desiccation. Our study provides the first evidence of a light-based strategy evolved in a non-photosynthetic bacterium to exploit scarse water in a desert environment.

The betaproteobacterium Ramlibacter tataouinensis TTB310 was previously isolated from sand particles coating a meteorite fragment collected from a semi-arid region of southern Tunisia (Tataouine) and identified as a new genus and species 1,2 . One of the most unusual characteristics of R. tataouinensis is its occurrence as both rod-shaped and spherical cells. The spherical cells are cysts, and are embedded within thick extracellular polymeric substances (EPS) that provide long-term desiccation resistance [2][3][4] . The rod-shaped cells are highly motile, which allows the colonization of a new location where they revert into cysts ( Supplementary Fig. S1). In contrast to the classical cell cycle with a cyst stage that must differentiate into rods before division (e.g. Azotobacter and Azospirillum, among others), R. tataouinensis cysts are able to divide and separate when water and nutrients are available ( Supplementary Fig. S1). These spherical cells capable of dividing are considered cysts because of their tolerance to desiccation 5 . Their ability to divide is compatible with their cyst status as demonstrated for the "reproductive cysts" of some free-living ciliates 6,7 . Cyst division is also observed in Rhodospirillum centenum, an anoxygenic photosynthetic and thermotolerant bacterium, when depleted of nutrients 8 . The daughter cells no longer divide 36-48 h after cyst induction. However, "hypercyst" mutants, which can readily form cysts on nutrient-rich medium, still cluster after cyst division 9 .
The unique cell cycle of R. tataouinensis, which alternates between desiccation-tolerant cysts capable of division and desiccation-sensitive motile rods capable of dissemination and division, is well suited for life in the hot and dry conditions of a desert environment. Due to the strong correlation between light, heat and desiccation, light should be an important external cue allowing R. tataouinensis to anticipate desiccation events by induction of protective mechanisms. In this context, six genes encoding potential light sensors were previously identified in strain TTB310 5 , representing one of the highest proportions of light-sensing proteins exhibited by a chemotrophic non-phototrophic bacterium 10 . Two of these genes (Rta_25470 and Rta_28950) contain all of the hallmarks of a bacteriophytochrome, whereas the other four genes encode blue light receptors. Specifically, Rta_12790 encodes a blue light-sensing histidine kinase containing a LOV (Light Oxygen Voltage) domain (phototropin analogous), while Rta_31060, Rta_20590 and Rta_26080 encode putative "Blue Light Using FAD" (or BLUF) proteins, although it has been noted that a reactive tyrosine and many highly conserved BLUF residues are absent from Rta_26080 10 .
The N-terminal photosensory input of phototropins containing two flavin mononucleotides (FMN) is coupled to a C-terminal output region that comprises a classic serine/threonine kinase motif. The blue light photoactivation of one of the two FMN induces protein rearrangements leading to autophosphorylation of the C-terminal kinase domain. BLUF photoreceptors bind flavin adenine dinucleotide (FAD) as chromophore. The output domains are PAS domains, GGDEF or EAL domains. Bacteriophytochromes possess a N-terminal photosensory core domain (PCD) that autocatalytically attaches a linear tetrapyrrole, whose photo-isomerization links two photo-interconvertible states that absorb in the red (Pr form) or far-red (Pfr form). The C-terminal module often contains a histidine kinase motif, involved in dimerization and signal transduction via phosphotransfer to a response regulator 11,12 .
The role of these light receptors has already been reported in several instances. In Pseudomonas syringae, a phototropin interacts with the bacteriophytochrome BphP1 to regulate swarming motility in response to light 13 . In Rhodobacter sphaeroides, photosystem synthesis is controlled by oxygen tension and light quantity. Both the BLUF protein AppA and the cryptochrome-like protein CryB are involved in this regulation 14,15 . The Bacillus subtilis phototropin YtvA is one of the stress input proteins that, when activated, leads to phosphorylation of parts of the macromolecular complex known as the stressosome 16,17 . It was also shown that infection by pathogenic bacteria is controlled by light via LOV domain histidine kinases in Brucella abortus 18,19 . More generally, it has been proposed that blue light receptors could regulate the transition between single and multicellular lifestyles 20 . Despite their large number in bacteria, the physiological role of most of these bacteriophytochromes remains unknown.
Far-red and red lights regulate the growth rate of cyanobacteria via the bacteriophytochromes Cph1 and Cph2, respectively 21 . Cph2 also inhibits phototaxis towards blue light 22 . In Rhodopseudomonas palustris and Bradyrhizobium, bacteriophytochromes regulate the synthesis of the photosynthetic machinery 23,24 . Carotenoid synthesis is regulated by a bacteriophytochrome in Deinococcus radiodurans 25 . A knockout mutant of the Pseudomonas aeruginosa bacteriophytochrome did not present a clear phenotype, although transcriptomic and proteomic analyses suggest the possible connection of bacteriophytochrome to quorum-sensing 26 . One of the two bacteriophytochromes present in Azospirillum brasilense Sp7 is required for the survival of on minimal medium under red light 27 . Its deletion increased sensitivity to the photooxidative stress 27 . In Agrobacterium fabrum, the conjugation is under the control of bacteriophytochromes 28 . More recently, it has been shown that bacteriophytochromes of Pseudomonas syringae pv. tomato were responsible for the repression of coronatine biosynthesis genes upon exposure to red light. This leads to the decrease of the stomatal reopening and bacteria entry 29 . Considering the fundamental signaling capability of light across all domains of life, and how extensively light-regulated pathways have been studied in animals, plants, and photosynthetic bacteria, we have only scratched the surface of light regulation in non-photosynthetic microbes. This and future studies are important for understanding a basic information processing problem.
In the present work, we studied the effect of light on R. tataouinensis TTB310, with a focus on the biochemical and biophysical properties of the two bacteriophytochromes and their roles in the cell cycle of this intriguing bacterium. Our results demonstrate that these two bacteriophytochromes initiate the rod division under red light, favoring a rapid division at sunrise before desiccation.

Results
Darkness, light and temperature effects on the growth of R. tataouinensis TTB310. Given the abundance of photoreceptors in strain TTB310, we sought to further characterize the role of light on growth in this bacterium. In a first set of experiments, we followed the development of strain TTB310 after 48-96 h of growth in darkness on agar plates at temperatures between 15 and 30 °C. Bacteria formed colonies composed essentially of cysts at all temperatures tested. However, at temperatures higher than 22 °C, the initial colony was surrounded by several equidistant colonies (Fig. 1A,B). This implies that a few mobile rod-shaped cells (emissaries) left the first colony to migrate and form new colonies, where a majority of cyst divisions occurs. Such rod-shaped cells have been observed at the periphery of colonies made of cysts 3 . Cell growth was monitored up to 7 days under continuous white light illumination provided by fluorescent tubes. Startlingly, we observed that relatively low-intensity (144 µmol m −2 s −1 ) white light illumination strongly inhibited the growth of strain TTB310 (Supplementary Table S1) in comparison to growth at the same temperature in the dark. This growth inhibition was partially reversible when cells were placed in the dark after exposure to the fluorescent light, as seen in previous experiments on day/night cycle exposure 5 . No effect was observed on the development of the E. coli control under the same illumination conditions. In contrast with our observation, we were expecting a high resistance to light in strain TTB310 that could be used to overcome the strong illumination prevailing in the desert.
To better understand the respective roles of the red/far-red and blue light receptors in this unexpected light inhibition, growth of strain TTB310 was monitored under red-or blue-filtered illumination. The light intensities were respectively 110 and 18 µmol m −2 s −1 for the resulting blue and red lights. Only the blue light induced light inhibition of growth (Supplementary Table S1), whereas the red light did not impair growth but did drastically increase rod division frequency (Fig. 1C). Similar results were obtained for illumination by red (660 nm) or far-red (780 nm) LEDs. In both cases, light induced the formation of a large number of motile rod-shaped cells that divided and disseminated on the agar surface, and only a few cysts were present after one week of observation. Varying the light intensity of the illumination at either 660 or 780 nm demonstrates that the induction of rod divisions occurs in both cases even at a very low intensity (0.5 µmole m −2 s −1 ).
These results suggest that blue light is responsible for growth inhibition of strain TTB310, whereas the induction of rod division could be mediated via light specifically absorbed by the bacteriophytochromes.
Due to the extreme growing conditions in the desert, the TTB310 bacteriophytochromes may present unusual photochemical properties, as already observed for several other bacteriophytochromes 24 . We therefore expressed and purified recombinant holo bacteriophytochromes of strain TTB310 to determine their spectral characteristics.

Expression and purification of recombinant holobacteriophytochromes. The genes encoding
RtBphP1 and RtBphP2 were co-expressed with a previously characterized heme oxygenase gene required for chromophore synthesis 30 in Escherichia coli and subsequently purified by affinity chromatography. Gel electrophoresis and zinc fluorescence demonstrated that the biliverdin (BV) is covalently bound in both proteins,  www.nature.com/scientificreports www.nature.com/scientificreports/ as previously reported for bacteriophytochromes of various bacterial species. Both bacteriophytochromes are dimeric, as indicated by cross-linking experiments (data not shown).

Spectral properties.
In agreement with the results of Baker et al. 31 we observed, in a dark-adapted sample of purified recombinant RtBphP1, an absorption spectrum characteristic of the Pr form of a bona fide BphP, with absorption maximum centered around 708 nm ( Fig. 3, top, black line). Upon illumination with 660-nm light, a typical Pr/Pfr transition was observed with the large bleaching of the main absorption band and the appearance of a broad band centered at 740 nm ( Fig. 3, top, red line). The light-minus-dark difference spectrum is typical for bacteriophytochromes (Fig. 3 bottom). Unexpectedly, illumination with a 780-nm light, a wavelength only very weakly absorbed by the Pr state of a dark-adapted sample of RtBphP1 (Fig. 3, top, black line), induced the formation of approximately one-third of the full Pfr state. This is clearly shown in Fig. 4 (left) where we measured the light-induced absorbance changes at 705 nm following a 780-nm illumination of a dark-adapted sample of RtBphP1. Illumination by both 780-nm and 660-nm lights induced approximately 90% of Pfr state of RtBphP1. Complete transformation to the Pfr state was obtained by a 660-nm illumination (Fig. 4, left). Starting from this state, a 780-nm illumination rapidly restored the Pr state to level previously obtained after the first 780-nm illumination (Fig. 4, left). The dark reversion from the Pfr to the Pr state is a multiphasic process (Fig. 4, left), as previously reported by Baker et al. 31 . A complete reversion required more than 90 min. Since the extinction coefficient of the Pfr form at 780 nm is much larger than the Pr form, we concluded that the induction of the Pfr state by this illumination could only occur if the quantum yield for the Pr state is much higher than the Pfr state. After analyzing the effects of illumination by 640 or 780-nm light on the Pr state of RtBphP1 (Fig. 4, left, Supplementary  Fig. S3), we deduced, using the equations derived from Giraud et al. 32 , that the quantum yield of the transition from the Pr state to the Pfr state is 6-7 times greater than the light-induced reverse transition. Such a large difference in quantum efficiency between Pr and Pfr forms has already been observed for example for AtBphP1 (or Agp1), a bacteriophytochrome present in Agrobacterium tumefaciens 33 . One consequence of this large difference in quantum yield between the two RtBphP1 forms is that excitation by any wavelength between 550 and 800 nm induces a partial or complete formation of the Pfr state. Therefore, RtBphP1 acts as a simple detector of a broad spectrum of red light.  www.nature.com/scientificreports www.nature.com/scientificreports/ The dark-adapted state of RtBphP2 absorbs maximally at 755 nm, which is typical for a Pfr form (Fig. 5, top). Illumination of RtBphP2 by infra-red light (780 nm) resulted in the rapid and complete formation of a Pr form with an absorption maximum centered at 690 nm (with an unusually low amplitude), whereas the amplitude of the Soret band was not affected (Fig. 5). Among the numerous phytochromes with a dark Pfr ground state, this peculiar property of the Pr form of RtBphP2 was previously observed for the A. tumefaciens bacteriophytochrome AtBphP2 (or Agp2) 34,35 and for a truncated form of the R. palustris bacteriophytochrome RpBphP1 36 . A 780-nm illumination completely bleached the 755-nm band, whereas a 660-nm illumination induced approximately two-thirds of the bleaching (Fig. 4, right). Since 660 nm is close to the isosbestic point of the light-dark difference spectrum of RtBphP2 (Fig. 5, bottom), the formation of two-thirds of the Pr state by a 660-nm illumination implies that the quantum yield for the transition from Pfr to Pr for RtBphP2 is about two times greater than the quantum yield of the reverse reaction. The reversion to the Pfr state occurs in the dark with a half-time of 3 min, and a 660-nm illumination slightly accelerates this reversion (Fig. 4, right). Due to the 2.5-fold larger extinction coefficient of the Pfr state as compared to the Pr state and the higher quantum yield of the Pfr state, any wavelength in the red and near infra-red induces a significant transition from the Pfr state to the Pr state. Like RtBphP1, RtBphP2 acts as a simple detector of a broad spectrum of red and far-red light.
phenotypes of mutants inactivated in heme oxygenase, bacteriophytochromes and cognate regulators. The unusual photochemical properties of RtBphP1 and RtBphP2 described above suggest that any wavelength of excitation in the red region, between 600 and 800 nm, will convert both RtBphP1 and RtBphP2 into a significant portion of their light-induced forms, respectively their Pfr and Pr forms. Therefore, we cannot predict whether only one bacteriophytochrome or both bacteriophytochromes are involved in the induction of rod divisions, based on the effect of the 660-nm or 780-nm illumination. To clarify their role in this differentiation process, we used plasmid insertion, the unique genetic tool that we have succeeded in developing in R. tataouinensis, without being able to complement the mutants without additional selective pressure (see Materials and Methods). Mutants inactivated for each of the bacteriophytochrome genes were constructed (∆RtBphP1 :  Rta_25470 and ∆RtBphP2: Rta_28950). Inactivation of the heme oxygenase gene (∆RtHmuO: Rta_25460) and the atypical hybrid histidine kinase gene (∆RtBHK: Rta_25490) was also achieved (Fig. 2B), whereas creation of a response regulator mutant (∆RtBRR: Rta_25480) has failed despite multiple attempts.
The heme oxygenase gene mutant (∆RtHmuO) cannot synthesize the BV and therefore the two bacteriophytochromes cannot respond to light excitation. As expected, illumination of this mutant ∆RtHmuO with red light (660 or 780 nm) did not induce rod divisions (Fig. 6H), contrary to what was observed in the WT (Fig. 6B), and only cyst divisions were observed in both the dark and the light during the first 48 h of growth (Fig. 6G,H).
Inactivating the gene encoding RtBphP1 or RtBphP2 resulted in a surprising phenotype. Specifically, both mutants (∆RtBphP1 and ∆RtBphP2) displayed no cyst division in the dark as observed for WT but only motile rod divisions (Fig. 6C,E vs. A). Rod divisions were also observed under 660 or 780-nm illumination, as in the case of the WT (Fig. 6D,F vs. B). This phenotype could not result to plasmid excision due to the presence of kanamycin to maintain the selection pressure. We also exclude a potential polar effect on downstream gene expression since the genes of the heme oxygenase and of RtBphP1 belong to the same operon, and nevertheless their mutant presented different phenotypes. Moreover, the phenotype linked to the inactivation of RtBphP1 or RtBphP2, which www.nature.com/scientificreports www.nature.com/scientificreports/ are genetically isolated, is similar. We are therefore confident in the fact that the phenotype observed for the mutants RtBphP1 and RtBphP2 are due to their inactivation.
The apparent contradiction between the evidence that light specifically absorbed by bacteriophytochromes induced rod divisions and the observation that their corresponding inactive mutant induced the same type of  www.nature.com/scientificreports www.nature.com/scientificreports/ division, even in the dark, is considered in the Discussion section. We will also discuss the puzzling observation that mutation of one or the other of the two bacteriophytochromes induced the rod-rod division phenotype.
The phenotype of the atypical hybrid histidine kinase gene mutant (∆RtBHK; Rta_25490) is similar to that of the two mutants ∆RtBphP1 and ∆RtBphP2, in that rod divisions occurred in the dark (Fig. 6I) and the light (Fig. 6J).

Discussion
R. tataouinensis TTB310 presents a distinctive cell cycle and a complex network development on dilute nutrient agar. In the dark, the main process observed is a cyst-cyst division ( Supplementary Fig. S1 "cyst-cyst division", Fig. 1A,B). Few motile rods, which derive from peripheral cysts of the original colony, form distantly new colonies of cysts ( Supplementary Fig. S1, "cyst-rod-cyst differentiation") 5 . Illumination with white light of low intensity strongly inhibits the growth of R. tataouinensis TTB310 cells. This growth inhibition is specifically induced by blue light whereas red light leads to rod divisions ( Supplementary Fig. S1 "rod-rod" division, Fig. 1C).
Although the growth conditions we used in the present work did not replicate the desert environment, these particular effects of blue and red lights could provide some clues into the lifestyle of R. tataouinensis TTB310. In the desert, water is available essentially in the form of dew. This occurs when the temperature in the desert drops to the dew point of the ambient air, i.e. several hours before the end of the night. These conditions are met in the desert only a few days a year. In the dark and in the presence of water, cyst divisions and (less frequently) a cyst-rod-cyst cycle occurred (Figs 1 and 7). At sunrise (i.e. red light conditions), about one hour after dew formation, rod divisions are induced (Figs 1 and 7), allowing rapid cell division before the progressive increase of sunlight coupled to the onset of desiccation. Under these conditions, blue light from the sun should inhibit cell division and allow desiccation tolerance for those cells localized in the first millimeters below the sand surface (Fig. 7). This blue light inhibition probably involves one or several of the blue light receptors detected in the strain TTB310 genome. The precise mechanism of this regulation will be investigated in future studies. At sunset, the temperature of the sand is higher than the dew point of the ambient air. Therefore, despite the presence of red light, rod divisions should not occur due to the absence of water (Fig. 7).
The observation that rod division is stimulated by red light specifically absorbed by bacteriophytochromes demonstrates that these chromoproteins are involved in this regulation. The phenotype of the mutant inactivated for heme oxygenase, which does not exhibit the light-induced rod division, is in agreement with this conclusion.
The regulation of cyst formation by light is not restricted to R. tataouinensis TTB310. Deletion of Ppr, an unusual bacteriophytochrome present in Rhodospirillum centenum, inhibits the light regulation of a polyketide synthase gene (PKS) 37-39 whose regulation is strongly correlated with stress-induced cyst formation 8 . Ppr, which acts as a simple detector of light 38,39 as we describe here for RtBPhP1 and RtBPhP2, is therefore probably involved in the light regulation of cyst formation in R. centenum. Regulation of bacterial cyst formation by light, via bacteriophytochromes, could therefore be a general mechanism present in bacteria.
Analysis of the phenotype of the mutants ∆RtBphP1 and ∆RtBphP2 provides some insights to the mechanism of the regulation of cell cycle regulation by red light in R. tataouinensis TTB310. One intriguing result is the similar phenotype of these two mutants, for which rod division occurs in both the light and the dark. Since both RtBphP1 and RtBphP2 are both autophosphorylated in the dark 31 (and data not shown), and this activity is partially suppressed by 700-nm light in the case of RtBphP1 31 , we propose that the two autophosphorylated www.nature.com/scientificreports www.nature.com/scientificreports/ bacteriophytochromes repress rod division in the dark by transferring their phosphate to their response regulators (Fig. 7). This repression of rod divisions results in cyst division and the cyst-rod-cyst cycle (including the presence of rods generated from cyst differentiation) (Fig. 6A). In the light, the bacteriophytochromes are dephosphorylated. The response regulator cannot be phosphorylated and the repression of rod division can no longer occur (Figs 6B and 7), so rod divisions become possible. In the absence of either bacteriophytochrome (∆RtBphP1 or ∆RtBphP2) in the dark as well as in the light, phosphorylation of the response regulator cannot occur, resulting in rod divisions (Figs 6C-F and 7). In the absence of the BV (by inactivation of the heme oxygenase gene, ∆RtHmuO mutant), the two bacteriophytochromes are phosphorylated both in the dark and in the light, and rod division is repressed under both conditions (Figs 6G,H and 7).
Although the mechanism described above explains the phenotype of ∆RtBphP1 or ∆RtBphP2 mutants, explaining the rod-rod division phenotype when only one of the two bacteriophytochromes is inactivated requires their direct or indirect cooperation. This raises two possibilities. In the first situation (direct cooperation), the active in vivo bacteriophytochrome is a heterodimer of RtBphP1 and RtBphP2, phosphorylated in the dark. In the absence of RtBphP1 or RtBphP2 (i.e. in ∆RtBphP1 or ∆RtBphP2 mutants), the active heterodimer cannot be formed, RtBRR cannot be phosphorylated in the dark, and rods are the dominant phenotype of these mutants (Figs 6I and 7). Interestingly, the presence of phytochrome heterodimers has been clearly demonstrated in higher plants 40,41 . For example, phytochromes PhyC and PhyB form heterodimers in rice 41 and Arabidopsis thaliana 40 . In rice, when PhyB is absent, PhyC exists as a monomer that is spectrophotometrically active but biologically inactive 41 . To test the possibility of heterodimerization between RtBphP1 and RtBphP2, we coexpressed their two genes in tandem in E. coli together with the heme oxygenase gene, but detected no bacteriophytochrome heterodimers formed in these conditions when our standard metal affinity chromatography protocol was carried out (data not shown). Although this experiment does not completely exclude the possibility of heterodimer formation in vivo, we do not favor this first hypothesis.
The second hypothesis is to consider an indirect cooperation between the two bacteriophytochromes via the cognate response regulator RtBRR (Supplementary Fig. S4). In a first step, RtBphP1 phosphorylates RtBRR in the dark 31 . Due to the "arm-in-arm" arrangement of the RtBRR dimer 31 , this phosphorylation could induce a conformational change in the second monomer of the dimer. After this conformational change, the second monomer can only be phosphorylated by RtBphP2. This hypothesis takes into account the unusual "arm-in-arm" ultrastructure of RtBBR and the absence of RR gene in the vicinity of the Rta_28950 gene (RtBphP2) (Fig. 2B). In a more complex situation, each bacteriophytochrome reacts with its own response regulator. In this case, the gene regulation requires the cooperation of these two regulators possibly via the atypical hybrid histidine kinase RtBHK, whose inactivation leads to the same phenotype as bacteriophytochrome inactivation (Fig. 6I,J).
In conclusion, our results establish that the two bacteriophytochromes of R. tataouinensis TTB310 are involved in the rapid rod division phase of the cell cycle. The cooperation of their signaling pathway and their particular optical properties constitutes a strategy that can be exploited to significantly broaden their efficiency from 550 to 800 nm. This mechanism, which occurs at sunrise while dew remains on sand particules, may favor a rapid dissemination of rods in order to avoid the noxious effects of strong light and dehydration. This is the first demonstration that (red) light controls the cell cycle of a non-photosynthetic bacterium, a phenotype likely coupled to this microbe's tolerance to water scarcity. This is likely a complementary system to the internal circadian rhythm (the kaiC gene was detected in the genome of strain TTB310 5 ) using the light as external cue.
Recently, a report has highlighted the roles of photoreceptors of Pseudomonas syringae pv. tomato in optimizing plant leaf invasion as a function of lighting conditions 29 . Involvement of photoreceptors in optimizing the development, dissemination and growth of bacteria as a function of the presence and quality of light may be a more general trait than previously anticipated.

Materials and Methods
Bacterial strains and growth conditions. Ramlibacter tataouinensis TTB310 T2 was cultured at 30 °C in the dark in tenfold diluted tryptic soy broth (TSB/10, Difco Laboratories) or tryptic soy agar (TSA/10).  Table S1), bacterial growth was estimated five times from serial dilution of a preculture grown in the dark with shaking (TSB/10, 30 °C) and from the number of colonies developing from each 5 µL drop (4 replicate drops for each dilution) deposited and cultivated at 22 °C on TSA/10 under different fluorescent lighting conditions (No Filter, Blue Filter, or Red filter) compared to the 100% control in dark conditions (using two layers of aluminum foil): the 100% control is equal to 2.10 7 bacteria.mL −1 .
Morphological characterization of bacteria. Wild type (WT) and various mutants were observed with an inverted microscope (Willowert Wetzlar 21-D6330, X4 objective), and photographed (X10 objective) with a BX50 Olympus microscope equipped with a differential interference contrast (DIC) device and a digital camera (Olympus F-view II). When necessary, a small piece of agar supporting several colonies was covered with a glass cover slip, cut and placed on a microscope slide to observe the presence of cysts or rods with the 100 X oil immersion objective (UPlanApo, Olympus) 4  www.nature.com/scientificreports www.nature.com/scientificreports/ expression and protein procedures. The Rta_25470 and Rta_28950 genes, designated respectively as RtBphP1 and RtBphP2, were PCR-amplified from genomic DNA with the Herculase II Fusion DNA polymerase (Agilent Technologies) using primers designed to add appropriate restriction sites (Supplementary Table S2) for expression as His 6 -tagged versions in the pBAD/HisB expression vector (Invitrogen). An upstream alternative start codon that elongates the protein N-terminal sequence containing the presumed cysteine ligand of the biliverdin chromophore, which was absent from the previous annotation, was used to begin amplification of the Rta_25470 gene. These fragments were cloned into the pTOPO-XL cloning vector and sequence-confirmed, before digestion with the restriction endonucleases BglII/EcoRI for RtBphP1 or BamHI/EcoRI for RtBphP2 (which already contains a BglII restriction site). Fragments were subsequently cloned into the pBAD/ HisB-RpBphP2-HmuO plasmid [containing the hmuO gene that encodes the Bradyrhizobium sp. ORS278 heme oxygenase 30 ], predigested with the BglII/EcoRI endonucleases, and inserted in place of the RpBphP2 gene. Overnight cultures of E. coli TOP10 containing pBAD/HisB-RtBphP1-HmuO or pBAD/HisB-RtBphP2-HmuO were grown with shaking (200 rpm) at 37 °C in LB medium supplemented with ampicillin (50 µg mL −1 ). The culture was diluted 1:100 into fresh LB medium with ampicillin 50 µg mL −1 (1 L in a 2-L flask) and grown at 37 °C with shaking (200 rpm) until an OD 600 of 0.6 was obtained, at which point L-arabinose was added to a final concentration of 0.66 mM (0.01%). Protein expression was allowed to proceed overnight (16-18 h), with shaking at 30 °C before harvesting the cells by centrifugation (6,000 g for 5 min at 4 °C). The cell pellet, generally green in color, was used immediately or stored at −20 °C until initiation of protein purification. Cells were resuspended in buffer A (Hepes 10 mM pH 7.0 or Tris-HCl 50 mM pH 8.0, 0.5 M NaCl, 20 mM imidazole) with the addition of DNase I (10 mg mL −1 ), and broken by French Press lysis (2 kbar or 200 MPa) or by using a One Shot cell disruptor (CellD). The cell lysate containing the soluble holo-BphP was then clarified by centrifugation (6,600 g for 30 min at 4 °C) to remove insoluble material and cell debris, followed by a 0.22-µm filtration. The soluble fraction was applied onto a 1-mL HisTrap HP column (GE Healthcare) pre-equilibrated against buffer A and washed with 10 mL buffer A. Green-colored proteins were eluted with 2.5 mL buffer B (Hepes 10 mM pH 7.0 or Tris-HCl 50 mM pH 8.0, 0.5 M NaCl, 0.5 M imidazole) and immediately eluted with Tris-HCl 50 mM pH 8.0, 0.25 M NaCl using a PD-10 Desalting Column (GE Healthcare).

construction of mutants. R. tataouinensis is a new bacterial model system for which elaborate genetic tools
were not yet available. The ability to construct the single gene knockout mutants presented in this work is a major step forward and these are the first mutants to be reported for R. tataouinensis. As part of developing genetic tools specific to R. tataouinensis TTB310, resistance to several antibiotics was examined. Only the broad host range pBBR1MCS-2-Km r plasmid 42 was successfully introduced in electrocompetent TTB310 cells and detected by the acquisition of kanamycin resistance. None of the other tested pBBR1MCS derivatives (Cm r , Sm r /Sp r , Amp r , Tet r , Gm r ) conferred any antibiotic resistance (data not shown). Furthermore, preliminary attempts to inactivate a targeted gene in strain TTB310 by marker exchange mutagenesis failed, but did result in a plasmid integration that could be detected by the acquisition of kanamycin resistance encoded on the plasmid (data not shown). Therefore, we constructed a plasmid for all mutants that was unable to replicate in strain TTB310. This plasmid, pSW23T-Km r (derived from pSW23T 43 , modified in this study with the insertion of a PCR-amplified kanamycin resistance encoding gene from pBBR1MCS-2, cloned at the BamHI site), contains a fragment internal to the target gene whose integration into the chromosomal copy should disrupt the gene, resulting in a non-functional protein. Fragments of the genes RtHmuO (Rta_25460; 0.5 kb), RtBphP2 (Rta_28950; 1.1 kb), RtBphP1 (Rta_25470; 0.3 kb), RtBRR (Rta_25480 RR; 0.26 kb), and RtBHK (Rta_25490 HK; 0.32 kb) were PCR-amplified from genomic DNA using the Herculase II Fusion DNA polymerase (Agilent Technologies) and primers listed in Supplementary  Table S2. These fragments were then cloned into the pTOPO-XL cloning vector (Thermo Fisher Scientific). After sequence confirmation, each fragment was digested and cloned at the XbaI site of pSW23TKm r , except for RtBHK fragment, which was cut from the two EcoRI sites flanking the pTOPO-XL multiple cloning site, and cloned at the EcoRI site of pSW23TKm r . The resulting ligated plasmids were transferred to the E. coli strain S17-1 λpir (Sm r ). They were then purified and transferred by transformation to E. coli strain β3914, which is auxotrophic for DAP (Km r , Em r , Tc r ) 44 , and transferred to strain TTB310 by overnight conjugation at 30 °C on TSA/10 complemented with DAP. The conjugation drop was made by resuspending a 10-mL pellet from a 24-72 h TTB310 culture in 200 µL TSB/10, and then using 75 µL of this solution to resuspend the pellet of 1-mL E. coli overnight culture. Transconjugants were subcultured on TSA/10 without DAP and selected with Km (200 µg mL −1 ). Km-resistant R. tataouinensis exconjugants were confirmed by PCR-amplification between plasmidic Km r and the surrounding chromosomal region neighboring the PCR-amplified fragment, and the insertion site was sequence-confirmed for each mutant.
Absorbance measurements. Purified bacteriophytochrome spectra were recorded with a Cary 50 spectrophotometer either in the dark or under continuous illumination (at 640, 660, 700 or 780 nm) provided by LEDs with irradiance between 60-70 µmol m −2 s −1 . Light-induced absorbance changes were performed with a laboratory-built spectrophotometer 45 similar to the one developed by Joliot et al. 46 . The absorption level was sampled using monochromatic flashes emitted 1 ms to several seconds after actinic excitation, provided by a 2-µs Xenon flash. construction of the phylogenetic tree. As the two RtBphPs and their 100-most homologous sequences are histidine kinases (sensor part of two-component system), phylogenetic analysis was based on the alignment of the Photosensory Core Domain (PCD) extracted from the P2CS database 47 . When homologous intra-species sequences were detected, only one was kept to make a simplified tree. This tree was generated by using the "one click mode" on phylogeny.fr website 48 .