Ultrathin sections of KT71^T cells revealed a typical Gram-negative cell wall with an inner cytoplasmic membrane and an outer membrane. The nucleoide was obviously concentrated in a central region of the cell that is less electron dense compared to the surrounding cytoplasm. Small electron dense particles that were frequently observed in ultrathin sections could be identified as polyphosphate inclusions by electron energy-loss spectroscopy. Interestingly, these inclusions could be detected mainly in cells grown under chemoheterotrophic conditions (Figure 2A, B).

In cells that were cultured under conditions allowing optimal expression of photosynthetic pigments the development of an intracytoplasmic membrane system was detected by transmission electron microscopy of ultrathin sections (Figure 2D). The membrane invaginations were of the lamellar type and adhere to the inner surface of the cytoplasmic membrane, similar to those observed in Rhodopseudomonas palustris [21]. However, in contrast to Rhodopseudomonas palustris intracytoplasmic membranes were only poorly developed in KT71^T and no thylacoidal membrane stacks were formed. In chemoheterotrophically grown cells that did not express photosynthetic pigments a similar differentiation of the cytoplasmic membrane was hardly detectable (Figure 2B).

Whole cells absorption-spectra of photoheterotrophically grown cultures of KT71^T were determined and compared with spectra obtained with cultures of Rhodospirillum rubrum DSM 467^T incubated anaerobically in the light. The photosynthetic apparatus of R. rubrum comprises a reaction center (RC) enclosed by a circular light-harvesting antenna complex (LH1), but no secondary (i.e. peripheral) light-harvesting (LH2) antenna complex [22]. As shown in Figure 3A the in vivo spectra determined from both strains were nearly identical. A minor difference is that the two major absorption bands of strain KT71^T in the near-infrared region were located at 802 and 876 nm and thus are blue-shifted in respect to R. rubrum DSM 467^T (peaks located at 804 and 882 nm). This indicates that the photosynthetic apparatus of strain KT71^T comprises only one type of light-harvesting antenna complex (LH1, peak at 876 nm) and a photosynthetic reaction center (RC, peak at 802 nm) both forming a core complex. Even upon incubation under conditions that are unfavorable for photosynthesis (e.g. in the dark) no indication for a peripheral LH2 complex in KT71^T cells could be detected, which would give rise to a more intense peak around 800 nm and a shoulder or an additional peak at 820–850 nm in the respective spectra. Consequently, it can be assumed that no significant amount of the LH2 complex is expressed in KT71^T cells under laboratory growth conditions, although in the genome of this strain the genes pucA (annotated open reading frame (ORF): KT71_03077) and pucB (annotated ORF: KT71_03072) were found that encode structural polypeptides forming the LH2 complex [10].

Suboptimal conditions for photosynthesis and pigment production caused a significant decrease of the LH1 complex size in KT71^T cells (Figure 3B). This assumption is based on the observation of a smaller peak ratio between the long-wavelength (Q_y) absorption bands of the LH1 antenna bacteriochlorophyll a (BChl a) at 876 nm and the reaction center accessory BChl a at 802 nm compared to spectra of KT71^T cells grown under optimal conditions for photosynthesis or cells of R. rubrum DSM 467^T, which are known to express a LH1-RC core complex with a relatively constant stoichiometry of 32 BChl a molecules per reaction center [23]. Recent studies applying high resolution electron microscopy or atomic force microscopy on photosynthetic membranes support the view of a LH1 complex that is variable in size and shape, at least in some bacteria [24], [25]. Alternatively, the variable absorption of the LH1 antenna complex in KT71^T cells may be independent of the actual size but due to perturbations in the BChl a environment leading to a lower LH1-BChl extinction coefficient.

Identification of the major pigments in photoheterotrophically grown KT71^T cells was done by UV/visible spectroscopy of acetone/methanol extracts prepared from wet cell pellets. The results essentially confirm the finding of Fuchs et al. [10], who used HPLC in combination with a photodiode array detector to identify the photosynthetic pigments BChl a and spirilloxanthin in KT71^T cells. In this study BChl a was identified in crude extracts by characteristic peaks at 363, 600, and 770 nm and spirilloxanthin by peaks at 465, 495 and 529 nm. The values obtained coincide well with published data for BChl a [26] and spirilloxanthin [27]. For comparison aged cultures of R. rubrum DSM 467^T, which are known to contain BChl a and spirilloxanthin as major photosynthetic pigments, were extracted applying the same protocol as used for KT71^T. Under optimal conditions for photosynthesis the UV/visible spectra indicate an almost identical pigment composition for both strains. However, in contrast to R. rubrum the pigment stoichiometry in KT71^T was highly variable with BChl a to spirilloxanthin ratios ranging from about 2 to 10 (Figure 3C, D). Besides BChl a and spirilloxanthin no other photosynthetic pigments could be identified in KT71^T cells grown under various conditions. However, sometimes an additional pigment type could be detected under incubation conditions that lead to a repression of photosynthetic pigment synthesis. For instance, cell suspensions of cultures that were incubated in the dark showed frequently a bright yellow-orange pigmentation and an additional type of compound displaying peaks between 400 and 420 nm became apparent in acetone/methanol extracts (Figure 3D). It is unlikely that these compounds represent carotenoids due to the highly variable spectral fine structure. Extraction of freeze-dried KT71^T cells with chloroform/methanol and separation of extracts by thin layer chromatography (data not shown) resulted in the development of a distinct band comprised of yellow to orange-red compounds that could be easily distinguished from photosynthetic pigments, which had a much lower polarity. Following separation by thin layer chromatography the unknown pigments were again dissolved in acetone/methanol to determine UV/visible spectra. The spectra obtained displayed a broad absorption maximum in the visible range between 350 and 570 nm thus explaining the unusual high absorbance values in this region in the spectrum shown in Figure 3D. Therefore, under certain conditions production of such polar pigments may lead to an overestimation of the spirilloxanthin concentration by the spectral reconstruction method.

The amount of BChl a and spirilloxanthin produced per cellular dry weight depended largely on the incubation conditions. Under optimal conditions for photoheterotrophic growth A_{880 nm}/A_{660 nm} values of 0.94–1.05 were obtained in cultures of KT71^T. In acetone/methanol extracts of the harvested cells the following pigment concentrations were determined (15 samples, standard deviation in parentheses): 3.02 (±0.42) nmol BChl a/mg (dry weight, dw) and 0.81 (±0.19) nmol spirilloxanthin/mg (dw). The cellular dry weight was calculated from the A_{660 nm} value of the harvested culture. Accordingly, the BChl a to spirilloxanthin ratio averages 3.7 under optimal conditions for pigment expression. A quantification of photosynthetic pigments in extracts of photoheterotrophically grown R. rubrum cultures (A_{880 nm}/A_{660 nm} values of 1.22–1.44) prepared by the same method resulted in the following values (5 samples): 4.48 (±0.46) nmol BChl a/mg (dw) and 1.84 (±0.21) nmol spirilloxanthin/mg (dw). The dry weight of the material extracted was calculated from the A_{680 nm} value of the harvested culture using a multiplication factor of 0.87 mg [28]. As aged cultures of R. rubrum were used for pigment extraction spirilloxanthin represented the prevailing carotenoid in the respective spectra (Figure 3C, D), so that the molar extinction coefficient of spirilloxanthin could be used for the estimation of the carotenoid abundance. The calculated average BChl a/spirilloxanthin ratio in whole cells of R. rubrum was 2.4 and hence considerably lower than in KT71^T cells. Within the boundaries of experimental error the values obtained for the type strain of R. rubrum are in good agreement with published data about this strain [e.g. 29], thereby indicating that the values determined for KT71^T should be also valid. Consequently, KT71^T represents an exception to the common assumption that aerobic anoxygenic photosynthetic bacteria are in general characterized by a low bacteriochlorophyll to carotenoid ratio [7].

Redox carrier proteins and quinones represent an essential part of the photosynthetic apparatus in all anoxygenic photosynthetic proteobacteria. Open reading frames proposed to represent genes encoding proteins that could play a role in photosynthetic and respiratory electron transport are listed in Table S1. As periplasmic cytochrome types, which are known to participate in the electron transport chain, only the types c ₄ and c ₅ were found, whereas no cytochromes c ₂ or c ₈ were annotated. A gene coding for a reaction center cytochrome c subunit is located within the photosynthesis genes superoperon (PGS). In addition, two genes were identified encoding different high-potential iron-sulfur proteins (HiPIPs), which are known to play a role in respiration and photosynthesis [30]. Genes for an ubiquinol-cytochrome c-oxidoreductase (cytochrome bc₁ complex) and three putative terminal oxidases were also found. Hence, all elements necessary for the operation of a Q-cycle mode of energy conservation could be localized in the KT71^T draft genome. In order to identify compounds of the electron transport chain that are essential for photosynthesis the cytochrome (cyt) composition in KT71^T was analyzed in photoheterotrophically grown cells (5 mM DL-malate as substrate) and compared with the composition in unpigmented cells grown chemoheterotrophically in various media. We used N,N-dimethyldodecylamine-N-oxide (LDAO) for the extraction and solubilization of cytochromes from whole cells, because this requires only a small amount of biomass. It should be noted that the solubilization of certain membrane-bound cytochromes requires rather harsh methods like the extraction of membranes with butanol [31]. Hence, it is possible that not all cytochrome types were solubilized completely by using the mild detergent LDAO, but instead remained partly in the insoluble material that was separated from the supernatant. Redox difference spectroscopy was used to analyze the expression of cytochromes in batch cultures that were incubated under various initial oxygen concentrations ranging from 3 to 21 vol%. Common characteristics of the obtained redox difference spectra were a Soret or gamma band at 422–426 nm, a beta band located at 521–525 nm and an alpha peak at 552–553 nm (Figure S1A). These spectral features are characteristic for c-type cytochromes [32], which therefore represent the dominating fraction of cytochromes in KT71^T cells. By far the highest expression levels of cyt c were found in cells expressing a photosynthetic apparatus. The appearance of redox difference spectra derived from solubilized extracts of these cells was quite similar to those from chromatophore membranes of phototrophic bacteria expressing a reaction center tetraheme cytochrome, like for instance Allochromatium vinosum [33] or Rhodoferax fermentans [34]. The intensities of the cyt c peak (Δε_552-538) in ascorbate-reduced minus ferricyanide-oxidized difference spectra of KT71^T cells correlated well with the expression of photosynthetic pigments. This effect was independent of the substrate utilized and confirms the presence of a reaction center cytochrome c subunit. In Figure 4A the correlation of the expression level of the photosynthetic apparatus with the abundance of high-potential cyt c and BChl a is shown. The data obtained could be fitted to single potential curves with coefficient of determination (R²) values of 0.95 for cyt c (12 samples) and 0.88 for BChl a (83 samples). At maximal expression of the photosynthetic apparatus the ratio between BChl a and high-potential cyt c approximates 10∶1. In fully pigmented KT71^T cells only about 75% of solubilized cytochromes c were reduced by ascorbate compared to a complete reduction with dithionite, so that for total cyt c the ratio would be 10∶1.25. Assuming that the solubilized heme c in photoheterotrophically grown cells is predominantly bound to the tetraheme cytochrome the ratio would be 10 BChl a to 0.3125 reaction center cytochrome c or 32∶1, which is the theoretically expected value for a photosynthetic reaction center embedded in a LH1 complex containing 28 BChl a molecules.

In addition to cyt c, cytochromes containing heme a or b as prosthetic groups were identified. Low-potential cytochromes b were of low abundance and probably overshadowed by the more abundant c-type cytochromes in dithionite-reduced minus ferricyanide-oxidized difference spectra. Cytochromes containing heme a became apparent in redox difference spectra by a shoulder of the Soret band at 445 nm along with a broad alpha peak at 601–604 nm. This assumption was confirmed by a distinct band at 588 nm in pyridine hemochrome difference spectra (Figure 4B). It can be deduced from the annotated genome sequence of KT71^T that heme a is most likely bound to a cytochrome caa₃ oxidase (Table S1). Surprisingly, only 50% of the total heme a was reduced by ascorbate, which could be explained by a resting form of the oxidase that is transferred into the fully active (or pulsed) form by reduction with dithionite. A similar effect was previously described for a novel aa₃-type oxidase from Pseudomonas aeruginosa [35]. Interestingly, the lowest expression level of cytochromes containing heme a were repeatedly found in photoheterotrophically growing cells, whereas in cells grown in SYPG medium under microaerobic conditions a maximal expression level of cytochrome aa₃ was observed and estimated to be 0.03 nmol/mg (dw).

Lipid soluble isoprenoid quinones are involved in electron transfer and transport of protons across the cytoplasmic membrane during photophosphorylation. Chemical analysis of the quinone composition in KT71^T cells grown chemoheterotrophically in SYPG medium as well as photoheterotrophically with 5 mM DL-malate as substrate revealed ubiquinone 8 as the sole respiratory lipoquinone, so that it can be assumed that this quinone type is also present in the photosynthetic reaction center. Likewise, the composition of polar lipids, which were mainly represented by phosphatidylethanolamine and phosphatidylglycerol along with an unidentified phospholipid, remained unchanged under chemoheterotrophic and photoheterotrophic growth conditions.

To demonstrate the generation of metabolically useful energy from light, growth curves of KT71^T cells expressing various amounts of photosynthetic pigments were determined in light and darkness. A 40 W tungsten filament incandescent lamp was used for illumination resulting in a light intensity of about 1400 lux (equivalent to 28 µE m⁻² s⁻¹). The effect of light on growth was quantified by comparing dry weights from total biomass produced under light and dark incubation. In addition, the production of photosynthetic pigments in growing cells was followed by recording A_{880 nm}/A_{660 nm} values. In Figure 5 representative growth curves of KT71^T cultures incubated in two different media are shown. KT71^T cells growing in defined medium with L-malate synthesized photosynthetic pigments continuously in the light and darkness, but reached a much higher cell density in the light (Figure 5A). This finding is in clear contrast to results obtained with laboratory cultures of several representative aerobic photosynthetic alphaproteobacteria, which accumulated BChl a exclusively in the dark [7]. Consequently, generalizing statements postulating a stringent inhibition of BChl synthesis in marine aerobic anoxygenic photosynthetic bacteria even by low light intensities [e.g. 36], [37] have to be reconsidered. Interestingly, our data are in line with early results of Kolber et al. [38] who found only a minor effect of light on pigment synthesis in aerobic bacteriochlorophyll-containing bacteria in marine ecosystems, but a diel growth cycle in oxygenic phototrophs with chlorophyll synthesis restricted to periods of darkness. In contrast to growth on L-malate pigment synthesis of KT71^T cells was hardly detectable during growth in the complex medium SYPG. It appears that trace amounts of pigments were only synthesized during the stationary phase thereby inducing a minimal increase of cell density (Figure 5B). Thus, the data presented indicate a light-dependent stimulation of growth due to photosynthesis. Further evidence for this assumption was obtained by growing pigmented and unpigmented KT71^T cells under identical conditions in defined semiaerobic medium containing 6 mM DL-malate as substrate (Table 1).

In Table 2 the percentage of substrate assimilation upon growth with various single carbon sources is compared for different strains. The percentage of carbon assimilation obtained by KT71^T in the dark with acetate or L-malate as substrate was unusually low, at least compared to Silicibacter pomeroyi and R. rubrum. The low growth yield could either indicate an incomplete utilization of these substrates or an uncoupling of substrate oxidation and energy generation. This observation could provide a first clue on the function of photosynthesis in this strain, if we assume that additional energy generated from light is required for the efficient utilization of certain substrates with a low energy yield. A light-dependent increase in growth yield in cultures of KT71^T was only observed if a photosynthetic apparatus was expressed, but there was no arithmetic correlation of the growth yield increase with the abundance of the photosynthetic apparatus, so that some additional factors may have influenced the efficiency of photophosphorylation. Based on the dry weight values obtained growth yield estimates could be made for KT71^T and compared with data obtained with Silicibacter pomeroyi or R. rubrum. It was found that in KT71^T the molar growth yield on malate (Y_malate) was 18.6 g (dw) in the dark and 44.2 g (dw) in the light. Assuming that the amount of biomass formed per mol ATP is constant for a certain strain cultivated in the same medium [39] it can be deduced that under optimal conditions KT71^T cells incubated in the light produce around 2.4 times more ATP than cultures incubated in the dark. With malate as substrate the growth yield of KT71^T cultures in the light was considerably higher than in the non-photosynthetic species Silicibacter pomeroyi, but smaller than in R. rubrum cultures grown anaerobically in the light. Nevertheless, aerobic photosynthesis still increased the growth yield of KT71^T cultures growing on malate by a factor of approximately 1.4 compared to Silicibacter pomeroyi, which is close to the estimate of Kolber et al. [1] for the light-induced growth yield increase of aerobic anoxygenic photosynthetic bacteria in the upper open ocean.

Most species of aerobic anoxygenic photosynthetic bacteria are routinely cultured in only one type of medium, so that the effect of carbon source utilization on the expression of the photosynthetic apparatus is largely unknown. In addition, often media were used that contain complex nutrients like yeast extract or peptones, so that the substrate utilization pattern cannot be exactly determined. Strain KT71^T did not express significant amounts of photosynthetic pigments during subcultivation in complex SYPG medium, but expressed a photosynthetic apparatus if malate was used as sole carbon source. In some experiments an induction of pigment synthesis was also observed after transfer from SYPG medium to defined media containing other carbon sources like succinate or oxoglutarate, but not with oxaloacetate or pyruvate. A detailed analysis of the induction of pigment synthesis in KT71^T was however hampered by several adversities. For instance, the transfer of cultures to defined media containing single carbon sources did not always lead to the induction of pigment synthesis and caused normally a prolonged lag phase, so that results could not be easily reproduced. Hence, we decided to study the effect of carbon source utilization with pigmented cells that were pre-grown under photoheterotrophic conditions. Using this approach a change of pigment expression occurred promptly and could be reliably determined. It was found that the level of pigment expression in cultures grown under dim light in SMFC medium at an initial oxygen tension of 6 vol% was quite stable and ranged between 0.85 and 0.88 (A_{880 nm}/A_{660 nm} values in stationary phase), so that preparatory cultures were routinely obtained by subcultivation in this medium. In a series of experiments more than twenty different substrates were identified that can be utilized by KT71^T as a single carbon source and resulted in different levels of pigment expression. In Figure 6 the cell density and pigmentation of stationary phase cultures is shown following transfer from SMFC medium in defined medium containing 2 mM of carbon source. KT71^T is able to utilize a wide range of substrates including citric acid cycle intermediates and sugars (Figure 6A), fatty acids (Figure 6B) and amino acids (Figure 6C). Overall, the obtained data indicate that the expression of photosynthetic pigments in defined medium depends mainly on the position at which substrates are funneled into the citric acid cycle. Most substrates that enter the central metabolism at the starting point of the citric acid cycle, i.e. as acetate or oxaloacetate, caused a severe repression of pigment production, which resulted in unpigmented cells after several transfers in the same medium. In contrast, fumarate or malate that both are formed at the end of this cycle stimulated the expression of a photosynthetic apparatus. The amino acids L-serine and L-alanine caused no significant repression of pigment synthesis (Figure 6C), although they were likely degraded to pyruvate, which acted as a strong inhibitor of photosynthesis. The reason could be an unexpected pathway for the assimilation of these amino acids or a more complex interaction of pigment expression and carbon metabolism in KT71^T.

It was originally thought that members of the marine OM60/NOR5 clade represent obligately oligotrophic bacteria unable to grow in media with high nutrient concentrations [17], [40]. However, we have found that strain KT71^T can be gradually adapted to growth in liquid or solid media containing more than 5 g l⁻¹ of complex substrates. Although, an increase in the length of the lag phase was observed upon transfer to media with very high nutrient concentrations this effect was probably not due to oligotrophic metabolism, but instead may indicate a slow adaptation of the gene expression pattern to changing growth conditions. An extended lag phase was also observed for example, if cultures were transferred to media containing different carbon sources in low concentrations. In addition, more than twenty phylogenetically diverse isolates affiliated to this group were isolated recently from Wadden Sea sediment off Sylt, Germany (unpublished results), which grew on isolation very well in SYPG medium and to a higher density as in the oligotrophic medium described by Eilers et al. [12]. Hence, the results obtained with strain KT71^T are presumably not based on a genetic drift caused by repeated subcultivation in the laboratory. However, independent of the general growth response there could be still a possible effect of substrate concentration on the expression level of photosynthesis genes as assumed by Kolber et al. [1] and Cho et al. [17]. To test this hypothesis we used batch cultures of KT71^T that were incubated under fully aerobic, semiaerobic and microaerobic conditions with various amounts of malate as sole carbon source. Photosynthetic pigments were extracted from cultures grown to stationary phase and quantified by spectrophotometry. First we tested, if an increase of the carbon source concentration has a negative effect on the growth yield. The results are shown in Figure 7 and indicate no negative effect on the growth yield by an increase of the malate concentration provided the oxygen concentration is not a growth limiting factor. At an initial oxygen concentration of 6 vol% the growth yield with all tested concentrations of malate was similar, but merely because the availability of oxygen limits growth. This becomes clear from the observation that the growth yield increased proportional with the malate concentration under an air atmosphere. In Figure 7 it is shown that the amount of BChl a produced increased with the substrate concentration under semiaerobic growth conditions and reached maximal levels with 6 or 8 mM malate. Thus, in the presence of an oxygen surplus the expression level of photosynthetic pigments increased with the amount of substrate utilized, which is in clear contradiction to the proposed stimulation of aerobic anoxygenic photosynthesis by nutrient limitation. In general, it appears that the growth yield and pigment production depended on the initial malate/oxygen ratio and were optimal under semiaerobic conditions, provided that neither the carbon nor oxygen concentration is a growth limiting factor. In contrast to our finding Cho et al. [17] reported in their study of strain HTCC2080, which is a member of the OM60/NOR5 clade and closely related to KT71^T (Figure 1), that light-dependent growth stimulation occurred only in medium with very low concentrations of organic substrates. However, it has to be noted that the substrate composition in their oligotrophic medium (based on autoclaved sea water containing 80 µM dissolved organic carbon) was most likely not identical to the artificial mixture of substrates that was used for the preparation of carbon-rich media containing 3.0 or 14.7 mM dissolved organic carbon, so that the effect on photosynthesis may well have been caused by a different substrate utilization pattern.

Conditions of severe carbon or oxygen limitation led to a reduction in growth yield and inhibition of pigment production in KT71^T. This effect may be explained by a correlation of the membrane redox state with the efficiency of energy generation. At an optimal carbon to oxygen ratio the membrane redox carriers are likely to be in a balanced state, which could promote the establishment of an active Q-cycle. In contrast, under conditions of severe oxygen limitation respiration is inhibited and redox carriers will be predominantly in a reduced state, whereas the combination of a high oxygen concentration with a very low carbon concentration will give rise to a potential overoxidation of the electron transport chain. Thus, the observed growth yield optimum under semiaerobic conditions is in agreement with the assumed involvement of a Q-cycle in photophosphorylation, which requires a balanced membrane redox state for optimal efficiency [41]. On the other hand, it can be deduced that KT71^T has only a limited ability to balance the cellular redox state actively in order to optimize the membrane redox state for the generation of energy.

A major environmental parameter that controls the expression of photosynthesis genes in anoxygenic photosynthetic bacteria is light. In most studies it was found that illumination with dim light (below 1000 lux) stimulates expression of a light-harvesting apparatus, whereas high light-intensities (above 10,000 lux) cause a repression of pigment synthesis [42], [43]. The reason behind this type of regulation is probably to increase the amount of absorbed photons under non saturating conditions for photosynthesis and on the other hand to avoid photooxidative damage by strong light in the presence of oxygen. In some species of aerobic anoxygenic photosynthetic bacteria affiliated to the Alphaproteobacteria an almost complete repression of photosynthetic pigments even at intermediate light intensities of 4000 lux [44] or 20 µE m⁻² s⁻¹ [7] was observed. In contrast, some preliminary results obtained with members of the OM60/NOR5 clade indicated that expression of pigments is repressed in darkness and stimulated by light [10], [17]. To find out if light is indeed a requirement for the expression of pigments in KT71^T we incubated various defined media in complete darkness after inoculation with cells grown photoheterotrophically in SMFC medium. To determine the stability of pigment expression cultures were transferred to fresh medium for at least five times in the dark. The results obtained are illustrated in Figure S2 and indicate that after several transfers in darkness the level of pigment expression decreased in all tested media. While the expression of pigments was stable in cultures of KT71^T growing with malate in the light, continuous incubation in the dark resulted in the almost complete loss of photosynthetic pigmentation (A_{880 nm}/A_{660 nm} value of 0.57) after six successive transfers. The unpigmented phenotype was stable even during prolonged incubation in the light and hence could be used for a comparison with photoheterotrophically growing cells in a series of experiments described below. In Figure S2 it is also shown that the significance of this effect depended largely on the carbon source utilized. In cultures growing on succinate or fumarate a high amount of photosynthetic pigments was still produced after several transfers in darkness, whereas with propionate the synthesis of pigments was completely repressed after three successive transfers, with oxaloacetate or pyruvate as carbon source already upon the first transfer. Interestingly, cultures growing with 2-oxoglutarate showed a variable repression of pigments during incubation in darkness, which indicates that the regulation of pigment synthesis is unstable under these conditions and depends not only on illumination. Hence, it can be concluded that light is not a prerequisite for pigment expression, but stimulates it indirectly, perhaps by interaction with the photosynthetic electron transport system as suggested by Happ et al. [45].

On the other hand, it could be shown in several preliminary experiments that light has a negative effect on pigment expression, provided the intensity is above 10,000 lux. Illumination of cultures growing in SMFC medium under semiaerobic conditions with high-intensity white light (11,000 lux, equivalent to 154 µE m⁻² s⁻¹) emitted from fluorescent lamps (Osram T8 L 58W/865 Lumilux Daylight) led to strong inhibition of pigmentation or even prevented growth (data not shown). Consequently, the effect of light on the level of photosynthetic pigment expression in KT71^T is not fundamentally different from the situation in most facultative anaerobic anoxygenic phototrophs. To determine the impact of light on the synthesis of pigments in more detail, cultures growing in SMFC medium were incubated in darkness or under illumination with various LED spotlights emitting either red, green or blue light with peak wavelengths of 627, 518 and 466 nm, respectively. An effective illuminance of 5500 lux was measured, if the green LED lamp was used. Based on the results summarized in Figure 8 it can be deduced that only blue light had a significant effect on pigment expression, which was independent of the initial oxygen concentration. Consequently, it is likely that any possible sensors that are involved in the light-dependent regulation of pigment synthesis in KT71^T sense blue light.

It can be assumed that the correlation of pigment expression and carbon metabolism in KT71^T is a complex interaction, because two different stable phenotypes growing with malate as sole carbon source could be obtained in the course of this study. The unpigmented phenotype growing on malate was obtained from photoheterotrophically growing cultures by repeated subcultivation in the dark (Figure S2). There is a possibility that the unpigmented phenotype is based on a loss of photosynthesis genes, but this is not probable, because an expression of photosynthesis genes in unpigmented cells could be demonstrated by reverse transcription-PCR of mRNA, albeit at a lower level (unpublished results). To gain more information about the possible role of the intermediary carbon metabolism for the expression of photosynthesis genes in KT71^T, we tried to analyze the citric acid cycle activity in cells growing on various substrates by using the specific inhibitor fluoroacetate. After entering the cell fluoroacetate is converted to fluoroacetyl-CoA by an acetyl-CoA synthetase (annotated ORF: KT71_08034). This compound is used by citrate synthase instead of acetyl-CoA to form fluorocitrate, which then inhibits specifically the enzyme aconitase thereby leading to an interruption of the citric acid cycle [46]–[48]. Fluoroacetate was used previously in several studies to eliminate citric acid cycle activity in resting cells and to determine the fraction of reduction equivalents that can be produced independently of this pathway [e.g. 49]. In our experiments fluoroacetate was added to cultures that already entered the exponential growth phase, so that a substrate specific intracellular carbon balance was established before fluoroacetate entered the cells. The response of KT71^T to the addition of fluoroacetate was tested using eight different medium compositions. The data obtained are summarized in Table 1. In Figure S3 representative growth curves are shown, which were obtained by incubation of KT71^T in four different media supplemented with varying amounts of fluoroacetate. Growth inhibition caused by fluoroacetate did not correlate directly with the level of pigment expression observed in parallel cultures incubated without fluoroacetate. Interestingly, the effect of fluoroacetate on cells of the unpigmented phenotype growing with malate as carbon source was markedly different to the effect caused on pigmented cells growing in the same medium, which probably indicates a different pattern of substrate utilization. Supposing that the amount of intracellulary synthesized fluorocitrate is independent of the incubation conditions, the data obtained would imply that generation of NAD(P)H by the citric acid cycle depends mainly on the utilized substrate and is almost absent in cells growing chemoheterotrophically with malate, pentanoate or in SYPG medium. On the other hand, this pathway would be indispensable for cells growing chemoheterotrophically in SMP medium or in defined media containing sucrose, pyruvate or oxaloacetate. Furthermore, growth in the absence of significant citric acid cycle activity would indicate that these cells do not have a respiratory metabolism, which depends on the production of NADH by the oxidative citric acid cycle. However, this appears unlikely, because it was shown that KT71^T is an obligately aerobic microorganism and expresses a respiratory electron transport chain with terminal oxidases under all tested incubation conditions. In addition, growth behavior in various deep agar cultures indicated an aerobic metabolism with all tested substrates (Figure S4). Based on these results we conclude that the metabolism is always respiratory and thus requires an active citric acid cycle for the production of the required amount of reduction equivalents. Hence, there must be another reason for the dependence of fluoroacetate inhibition on substrate utilization. We propose the following explanation: Kelly [48] has shown that fluoroacetate-resistant cells of Halothiobacillus neapolitanus (formerly Thiobacillus neapolitanus) are characterized by a lack of acetyl-CoA synthetase activity. Furthermore, he demonstrated that an excess of extracellular acetate prevents inhibition by fluoroacetate, but only if it is present in the medium prior to addition of fluoroacetate. This would mean that acetate has to be metabolized before fluoroacetate enters the cell in order to prevent growth inhibition. We conclude from these observations and our data, that substrates that lead to a depletion of the intracellular coenzyme A pool and an increase of acetyl-CoA or other coenzyme A thioesters induce a resistance against fluoroacetate by repression of the acetyl-CoA synthetase activity, which is necessary for the formation of fluoroacetyl-CoA. On the other hand, substrates that lead to an increase of the intracellular coenzyme A to acetyl-CoA ratio would stimulate acetyl-CoA synthetase activity and fluoroacetate toxicity. This theory can explain why KT71^T cells growing with pentanoate are rather resistant to fluoroacetate, whereas cells growing with oxaloacetate are highly sensitive. Pentanoate is metabolized to acetyl-CoA and propionyl-CoA thereby leading to a depletion of the coenzyme A pool, whereas an excess of oxaloacetate would lead to a decrease of the acetyl-CoA concentration by the citrate synthase reaction. Based on this assumption, we can conclude that the high fluoroacetate sensitivity of cells growing on pyruvate or sucrose is probably due to an intracellular accumulation of oxaloacetate. Oxaloacetate can be synthesized from pyruvate or phosphoenolpyruvate, which are intermediates of glycolysis, by carboxylation reactions. The key enzyme for this pathway in KT71^T is phosphoenolpyruvate (PEP) carboxylase (annotated ORF: KT71_12695), which forms oxaloacetate and phosphate by the carboxylation of PEP. It is known that this anaplerotic reaction represents a major route for CO₂ assimilation in heterotrophic and mixotrophic bacteria [6], [9]. In addition, it can be deduced that the repression of pigment synthesis in cultures growing with malate in the dark correlates with an increase of the acetyl-CoA concentration, so that the intracellular activation of fluoroacetate is prevented. A rise of the acetyl-CoA concentration in KT71^T cells growing with malate could be caused for example by an induction of malic enzyme (annotated ORF: KT71_12670), which decarboxylates malate to pyruvate that can be subsequently converted to acetyl-CoA by the pyruvate dehydrogenase complex. On the other hand, it is likely that the presence of pyruvate in SMP medium leads to a feedback inhibition of malic enzyme and stimulation of pyruvate orthophosphate dikinase (annotated ORF: KT71_12340) and PEP carboxylase, thereby leading to the accumulation of oxaloacetate and depletion of acetyl-CoA.

Based on these considerations, we propose that the level of pigment expression in KT71^T depends partly on the intracellular carbon metabolism and is maximal under conditions of a balanced carbon flux, that is when neither acetyl-CoA nor oxaloacetate are accumulating.

Previously, it was found that the oxygen concentration within a layer of cells that formed in deep agar cultures of KT71^T is around 30 µmol O₂ l⁻¹ and quite stable, although its position relative to the air/agar interface varied and correlated with the amount of substrate utilized [10]. Hence, it is likely that the position of the visible band of bacteria in the agar medium reflected the oxygen demand of growing cells. Cells which require a high amount of oxygen to balance their redox state will be situated closer to the agar surface than cells requiring only a small amount of oxygen. Based on these considerations, we expected that inoculation of semisolid agar media containing various substrates will result in different patterns of band formation depending on the potential of substrates to stimulate oxygen consumption.

In Figure S4 the positions of formed bands in deep agar cultures containing various substrates is shown. In cultures growing in SYPG or SMP medium or with sucrose or pyruvate as substrates the position of the cell layer was close to the agar surface, whereas with acetate, oxaloacetate or malate (unpigmented phenotype) it was located several millimeters below the surface, thereby indicating a lower oxygen demand. The position of the band that formed in cultures with cells growing photoheterotrophically with malate as substrate was located at an intermediate position. Hence, we conclude that unpigmented cells growing on acetate, malate or oxaloacetate have a lower respiratory activity than cells growing with sucrose or pyruvate as substrate.

The cytochrome c oxidase activity in bacteria can be roughly quantified by the amount of N,N,N′,N′-tetramethyl-p-phenylenediamine (TMPD) that becomes oxidized to a blue colored derivative immediately upon contact with whole cells. It is assumed that this reaction is specific for cells expressing an active cytochrome c oxidase, because strains lacking such an enzyme usually show only a weak reaction upon exposure to TMPD, which can be used as electron donor instead of cytochrome c by terminal oxidases [50]. We have found that the oxidase activity in strain KT71^T was highly variable and changed with the initial oxygen tension as well as with the substrate utilized (Figure 9), which indicates a correlation with the cellular redox state. An identification as redox monitor was however complicated by the fact that the amount of oxidized TMPD correlated not proportionally with the oxygen concentration, but in most cases displayed a minimum at semiaerobic growth conditions. In Figure 9 it is shown that the oxygen concentration at which a minimal cytochrome c oxidase activity was found depended on the substrate utilized and was relatively high (<21 vol% O₂), if cells were grown in the complex media MB or SYPG and rather low (<6 vol% O₂), if they were grown chemoheterotrophically in defined medium containing malate as sole carbon source. We deduced from this observation that the oxidase activity reaches minimal values under conditions of a balanced redox state and increases as the redox state becomes more reduced or oxidized. Thus, at oxygen-limiting conditions (6 vol% initial oxygen concentration) high oxidase activities should indicate either an overreduced or overoxidized cellular redox state. It can be deduced from Figure 9 that cells growing with acetate are probably in an overreduced state, which is unexpected because the amount of reduction equivalents that can be derived from this carbon source should be lower than from most other substrates. It is also noteworthy, that the oxidase activity in photoheterotrophically growing cells remained almost completely repressed under various initial oxygen concentrations ranging from 6 to 21 vol%, which could indicate that the redox state is actively kept in a balanced state.

The observed variation of the cytochrome c oxidase activity in KT71^T cells should be reflected in a variable biochemical composition of the electron transport chain. Hence, we analyzed the diversity of electron transport chains in this strain by redox difference spectroscopy of whole cells or extracts treated with various inhibitors that interrupt the electron transfer at different sites. In Figure S1 results are shown that were obtained either with cells grown photoheterotrophically on malate or with unpigmented cells that were grown chemoheterotrophically in SMP or SYPG medium or with malate as substrate. Potassium cyanide was used to inhibit terminal oxidases (Figure S1B). Addition of cyanide to solubilized cell extracts should result in the reduction of the inhibited oxidase and its primary electron donor [51]. It could be shown that in solubilized extracts of cells grown in SYPG medium about 90% of the total high-potential cyt c could be reduced by the addition of cyanide. On the other hand, in photoheterotrophically grown cells only a small fraction of solubilized cyt c was reducible by the addition of cyanide and a characteristic featureless bleaching of the spectrum in the region between 450 and 550 nm was observed. This feature is typical for reduced iron-sulfur proteins [52] and could indicate that under conditions of photoheterotrophic growth cyt c is replaced by an alternative periplasmic redox carrier, likely a high-potential iron-sulfur protein (HiPIP). In a complementary experiment stigmatellin was used to inhibit reduction of cyt c by the ubiquinol-cytochrome c-oxidoreductase (Figure S1C). A significant oxidation of cyt c was only found in difference spectra (dithionite and stigmatellin minus dithionite) of chemoheterotrophically growing unpigmented cells. In cells expressing a photosynthetic apparatus the spectrum was dominated by redox active photosynthetic pigments, whereas signals indicating presence of c-type cytochromes were hardly detectable. To analyze the expression of terminal oxidases carbon monoxide (CO) was added to whole cell suspensions (Figure S1D). CO can be used for the detection of oxidases because it binds specifically to heme moieties that have oxygen binding capacity. However, the identification of distinct oxidases in whole cells is difficult and usually restricted to cells expressing only one type of oxidase. Nevertheless, a pronounced variability in difference spectra (dithionite and CO minus dithionite) obtained with pigmented and unpigmented cells was evident. Apparently, signals derived from CO-binding c-type cytochromes (trough at 550–553 nm) were most intense in photosynthetic cells and less pronounced in unpigmented cells grown in SMP and SYPG medium or with malate as substrate. In chemoheterotrophically grown cells a visible trough at 443–447 nm in the Soret region indicated expression of an oxidase containing heme a. However, an additional characteristic trough at around 610 nm caused by the ferrocytochrome a₃-CO complex was not detected. The presence of terminal oxidases containing heme b was deduced from a negative shoulder around 564 nm and a trough at 428 nm in the Soret region of the difference spectrum, which could be most clearly detected in photoheterotrophically grown cells. Based on the obtained results, we conclude that the intensity of the oxidase reaction correlates with signals indicating presence of an active aa₃-type oxidase in redox difference spectra. Hence, it is possible that during photoheterotrophic growth or under conditions of a balanced redox state expression of the aa₃-type cytochrome c oxidase is inhibited, whereas a cb-type oxidase characterized by a low TMPD turnover rate is stimulated. In Figure 10 the estimated amounts of cytochromes c and high-potential iron-sulfur proteins participating in the respiratory electron flow are shown and can be correlated with the results of the TMPD assay. It appears that growth conditions that are characterized by a participation of HiPIPs in the respiratory chain correlate with a low TMPD oxidase activity. These data would indicate that HiPIPs in KT71^T are not only likely candidates for the reduction of the reaction center cytochrome c (especially due to the lack of alternative donors like cyt c ₂ or c ₈), but may also play a role as electron donors for a cb-type terminal oxidase under conditions of a balanced redox state or during photoheterotrophic growth.

We hypothesize that this cb-type oxidase could represent a novel HiPIP∶oxygen oxidoreductase, which cannot utilize TMPD as electron donor. The existence of a HiPIP∶oxygen oxidoreductase in betaproteobacteria was also previously proposed by Hochkoeppler et al. [53], who found that respiration in the facultative phototroph Rhodoferax fermentans depends on the oxidation of HiPIP by a cb-type terminal oxidase. Alternatively, a low TMPD oxidase activity of whole cells could be caused by the expression of ubiquinol oxidases. However, no genes encoding potential ubiquinol oxidases could be identified in the KT71^T draft genome sequence.

A detailed analysis of the cellular fatty acid composition under various incubation conditions revealed a significant dependence on substrate utilization and the availability of oxygen. This indicates that the production of distinct cellular fatty acids could reflect the cellular redox state. In Table S2 it is shown that the position of the double bond in the prevalent unsaturated fatty acids changed from ω6 (counted from the methyl end) at microaerobic conditions to ω7 under conditions close to oxygen saturation, which could indicate two different pathways for the synthesis of unsaturated fatty acids that are controlled by the cellular redox state. Accordingly, under the same oxygen tension the amount of synthesized ω6 fatty acids depended on the medium used and reached higher levels upon incubation with the nutrient-rich MB medium than with malate as carbon source (Table S2). In Figure 11 the ratios of ω6c to ω7c unsaturated fatty acids in cells that were grown under various incubation conditions are shown. The data presented could indicate that cells growing with sucrose as carbon source have a lower redox state than cells growing in SMP medium or with acetate and that photoheterotrophically growing cells have a more reduced redox state than unpigmented cells growing chemoheterotrophically in the same medium. In addition, one may assume from the data obtained that the overall redox state in unpigmented cells growing with acetate or in SMP medium under semiaerobic conditions is comparable to that in photoheterotrophically growing cells. This would indicate that the expression of photosynthesis genes in strain KT71^T is not only controlled by the overall redox state, but depends also on the composition or activity of the electron transport chain, which has been shown to be different in photosynthetically active and unpigmented cells.

It is known that the redox state of the electron transport chain is reflected by the reduction of the quinone pool, which in turn is influenced by the type of substrate, availability of oxygen and light intensity in photoheterotrophic bacteria [54]. It has been shown, for instance, that oxygen depletion or an increase of the substrate concentration led to an increased reduction of the quinone pool in Escherichia coli [55] and R. rubrum [56]. In Figure 12 the pigment composition of photoheterotrophically growing KT71^T cells is correlated with a variation of growth conditions suggested to alter the reduction of the cellular ubiquinone pool. The data presented can be interpreted in such a way that conditions, which lead to an increased reduction of the quinone pool lead to a decrease of the BChl a to spirilloxanthin ratio. Consequently, transfer of photoheterotrophically grown cells to medium containing acetate as substrate resulted in a more oxidized quinone pool than transfer to a medium with sucrose. As both substrates induced a drastic reduction of the pigment production, it was not possible to correlate the repression of photosynthesis genes with a distinct change of the BChl a to spirilloxanthin ratio. Thus, one might assume that the reduction of the quinone pool represents a signal that controls mainly the pigment stoichiometry, but has only a minor influence on the level of pigmentation.

Interestingly, we have also found that illumination has a strong influence on the pigment stoichiometry. An analysis of the pigment composition in cells grown on various carbon sources revealed that upon growth in dim light the molar ratio of BChl a to spirilloxanthin was on average 62% higher than upon incubation in complete darkness, while the expression level of the photosynthetic apparatus (deduced from A_{880 nm}/A_{660 nm} values) remained almost constant. Although, the values determined varied considerably the tendency of this effect was quite stable and could be reproduced in numerous independent experiments. As shown in Figure 12 this effect was not restricted to an illumination with blue light, but increased with decreasing wavelength of the light source. Hence, it appears that the pigment stoichiometry in KT71^T is influenced by signals that depend on the reduction of the quinone pool instead of being controlled by a light-sensing protein. Klamt et al. [54] have proposed that the quinone pool in photosynthetically active proteobacteria is more reduced in darkness than under illumination, which would be in line with our findings.

A possible explanation for the redox-sensitive regulation of the pigment stoichiometry in strain KT71^T could be as follows: As shown above this strain does not express a peripheral light-harvesting (LH2) complex, which could be used to adapt the absorption of photons to light intensity as observed for example in Rhodobacter species that produce variable amounts of the LH2 complex in response to changing light intensities [57], [58]. Instead, strain KT71^T may control the amount of photons absorbed not only by the abundance of the LH1 complex, but also by the spirilloxanthin content of the light-harvesting apparatus. Some meters below the water surface in coastal areas the majority of the penetrating photons is in the range between 400 and 550 nm [9] and thus absorbed by carotenoids and not bacteriochlorophyll a, so that a reduction of the carotenoid content could help to avoid photooxidative damage by illumination under conditions of a high redox state. This finding is unexpected, because it has been a frequently repeated assumption that aerobic anoxygenic photosynthetic bacteria accumulate large amounts of carotenoids in order to prevent photooxidative damage caused by reactive oxygen species. However, a large part of the carotenoids produced in these species is not involved in light harvesting and only loosely associated with the photosynthetic apparatus [59]. Hence, the strategy of KT71^T may be more economical under conditions of energy deprivation and nutrient limitation.