Introduction

Cyanobacteria comprise an ancient group of oxygenic photosynthetic organisms. Their photosynthetic apparatus is associated with thylakoid membranes (Gantt 1994). Their photosynthetic pigments include chlorophyll a (and b in the case of prochlorophytes), carotenoids and the phycobiliproteids phycocyanin and phycoerythrin. In order to protect the chlorophyll molecules from excess light, the presence of carotenoids is essential (Frank and Cogdell 1996). With respect to the carotenoid composition, two groups can be distinguished. They differ by the occurrence of zeaxanthin (ß,ß-carotene-3,3′-diol) or canthaxanthin (ß,ß-carotene-4,4′-dione) as the major carotenoid. Other carotenoids that may be present in substantial amounts include echinenone (ß,ß-carotene-4-one), myxoxanthophyll (a 3′,4′-didehydro-1′,2′-dihydro-ß,ψ-carotene 3,1′-diol with a sugar moiety attached to C-2′) and ß-carotene (ß,ß-carotene) (Stransky and Hager 1970; Hertzberg et al. 1971; Goodwin 1980). The latter is a precursor for all cyanobacterial carotenoids except myxoxanthophyll. This glycoside is derived from monocyclic γ-carotene. For several cyanobacteria like unicellular Synechocystis (Fernández-Gonzalez et al. 1998; Steiger et al. 1999), Synechococcus (Masamoto and Furukawa 1997) and other filamentous species, a regulatory effect of light on the overall synthesis of carotenoids has been demonstrated.

The pathway to ß-carotene involves three types of reactions. They include condensation of two molecules of geranylgeranyl pyrophosphate to phytoene catalyzed by phytoene synthase, four desaturation steps and cyclisation of both ends of the molecule to ß-ionone rings (Sandmann 2001). In contrast to other bacteria, cyanobacteria have acquired two homologous desaturase genes crtP and crtQb that mediate the formation of ζ-carotene and lycopene, respectively (Sandmann 2002). However, the desaturation intermediates and products possess cis configurations. Therefore, the resulting 7,9,7′,9′-tetracis lycopene must be isomerized to the all-trans configuration by the isomerase CrtH before cyclisation is possible (Breitenbach et al. 2001a; Masamoto et al. 2001). Bacteria other than cyanobacteria (including fungi) catalyze the formation of all-trans lycopene from phytoene by a single enzyme, the crtI gene product (Sandmann 2001). However, the distribution of the CrtI versus the CrtP/CrtQb/CrtH desaturation pathway is not clear cut, since it was recently shown that the green sulfur bacterium Chlorobium tepidum possesses the same genes found for the cyanobacterial desaturation pathway (Frigaard et al. 2004). In cyanobacteria, ß-carotene is typically modified to oxo derivatives involving hydroxylation at positions 3 and 3′ of the ionone rings by CrtR (Masamoto et al. 1998), monoketolation at position 4 to echinenone by CrtO (Fernández-Gonzalez et al. 1997) or diketolation at positions 4 and 4′ to canthaxanthin (Steiger and Sandmann 2004).

According to its morphology, the rod-shaped unicellular cyanobacterium Gloeobacter violaceus has been assigned to the order Chroococcales. Nevertheless, molecular phylogenetic analysis based on 16S rRNA demonstrated an isolated position away from other groups of cyanobacteria for G. violaceus (Honda et al. 1999). Special structural features separate G. violaceus from all other cyanobacteria. G. violaceus is the only oxygenic organism that lacks thylakoids as the site for photosynthesis. Its photosynthetic apparatus is associated with the cytoplasmic membrane (Rippka et al. 1974). The absence of thylakoids may be the reason why sulfoquinovosyl diglyceride is missing (Selstam and Campbell, 1996). Photosystem I of G. violaceus contains a reduced number of subunits compared to other cyanobacteria (Inoue et al. 2004). Furthermore, the composition of light-harvesting phycobilisomes is untypical. They possess a bundle structure with six parallel rod-shaped elements attached to the inner surface of the cytoplasmic membrane (Guglielmi et al. 1981) and contain phycourobilin in contrast to most other cyanobacteria. The carotenoid composition of G. violaceus is still obscure. The presence of oscillaxanthin, echinenone and ß-carotene has been proposed (Rippka et al. 1974; Jöstingmeyer and Koenig 1998). However, detailed analytical data are missing to date, and no information is available on the reactions of the biosynthetic pathway and the genes and enzymes involved.

In the present investigation, the carotenoid composition of G. violaceus was determined and its up-regulation by light established. Since the genome of G. violaceus has been sequenced (Nakamura et al. 2003), genome analysis in combination with functional expression studies using individual candidates homologous to carotenogenic genes from other organisms led to the identification of all genes involved in carotenoids biosynthesis of G. violaceus with the exception of the lycopene cyclase gene.

Material and methods

Organisms and cultivation

The cyanobacteria G. violaceus and Synechocystis PCC6803 were cultivated in BG11 medium (Rippka et al. 1979) at 26°C in white light of 40 μmol m−2 s−1. Under our conditions, the doubling time was about 50–60 h. Light intensity was varied in the experiments as indicated and concentrations of inhibitors applied were 100 μM diphenylamine and 10 μM norflurazon. In the light-response experiments, a G. violaceus starting culture was divided and diluted to a cell density of 0.3 mg dry weight per ml prior to diphenylamine addition and growth continued for 3 days. Escherichia coli XL1-Blue MRF’ (Stratagene) was cultivated overnight at 37°C in LB medium (Sambrook et al. 1989). E. coli JM101 carrying different plasmids (see Table 1) in complementation experiments was grown for 2 days at 28°C and was induced with 0.5 mM IPTG. The medium was supplemented according to the plasmids with ampicillin (25 mg/ml) and chloramphenicol (34 mg/ml).

Table 1 Expression constructs with genes from Gloeobacter violaceus and complementation plasmids used for functional characterization
Full size table

Plasmid construction and DNA techniques

Unless indicated otherwise, molecular standard methods were used according to Sambrook et al. (1989). Restriction enzymes and other nucleic acid modifying enzymes were used according to the manufacturers’ instructions. DNA fragments were analyzed on agarose gels and purified using the DNA Isolation Kit Elu-Quick from Applichem (Darmstadt, Germany). DNA was isolated from G. violaceus cells after lysozyme treatment (2 mg/ml) for 2 h at 37°C using the Wizzard Genomic DNA Purification kit from Promega (Mannheim, Germany). All open reading frames indicated in Table 1 were cloned using the oligonucleotides listed therein to amplify the genes by PCR with genomic DNA from G. violaceus as template. All resulting DNA fragments with T-overhanging ends were ligated into the XcmI-digested vector pMON3820 (Borokov and Rivkin 1997) as in-frame fusions and selection of the correct orientation.

Complementations and carotenoid analysis

Carotenoids from G. violaceus and Synechocystis were analyzed by HPLC. They were extracted from freeze-dried cells with methanol for 15 min at 60°C and partitioned against ether/petroleum ether (1:9 by volume). Separation of carotenoids was carried out by HPLC on a Nucleosil C18, 3 μm column with acetonitrile/methanol/2-propanol (85:10:5 by volume) and a flow rate of 1 ml/min. Absorbance spectra were recorded on-line with a Kontron photodiode array detector 440. Carotenoids were identified by their retention times and their spectra. They were compared to the carotenoids from Synechocystis that have been identified earlier (Bramley and Sandmann 1985; Takaichi et al. 2001) and to authentic standards.

The function of the individual genes was assessed by co-transformation in E. coli with a geranylgeranyl pyrophosphate background mediated by pACCRT-E, phytoene background by pACCRT-EB, lycopene background by pACCRT-EBI or ß-carotene background by pACCAR16ΔcrtX (see Table 1 for details). The newly formed carotenoids from the complemented E. coli were extracted and analyzed in the same way as described above.

Sequence analysis/software

Sequence data for G. violaceus were obtained from the Kazusa database (http://www.kazusa.or.jp/cyano/gloeobacter.html). Database searches were carried out with the similarity search tool BLAST P 2.2.10 (Altschul et al. 1997). Phylogenetic analysis of amino acid sequences were performed with the program Clustal X (Thomson et al. 1997) and the alignments visualized with TreeView.

Results and discussion

Carotenoids in G. violaceus

Carotenoids synthesized by G. violaceus were analyzed by HPLC (Fig. 1). Two peaks with carotenoid spectra were detected in lipophilic extracts (trace A). The minor peak 1 at 7 min gave a bell-shaped absorbance spectrum at 460 nm, the major peak 2 with a retention time of 29 min exhibited three maxima at 425 (shoulder), 450 and 478 nm. The peaks 1′ and 2′ were attributed to the corresponding cis isomers by similar spectra and a cis peak at 338 nm for ß-carotene. In trace B, a carotenoid extract of Synechocystis of known composition (Bramley and Sandmann 1985; Takaichi et al. 2001) was chromatographed as a reference to identify echinenone and ß-carotene. The other Synechocystis carotenoids, myxoxanthophyll (peak 3) and zeaxanthin (peak 4), were absent in G. violaceus. Reports on the occurrence of the acyclic glycoside oscillaxanthin (Jöstingmeyer and Koenig 1998) or any other carotenoid glycoside (Rippka et al. 1974) could not be confirmed. Instead, we extracted a reddish pigment with spectroscopic properties of a phycobilin. Based only on its color and its low Rf value in thinlayer chromatography, the nature of this pigment may have been misinterpreted as oscillaxanthin.

Fig. 1
figure 1

HPLC analysis of carotenoids extracted from Gloeobacter violaceus untreated (a), from Synechocystis PCC6803 treated with 10 μM norflurazon (b), G. violaceus treated with 10 μM norflurazon (c) and G. violaceus treated with 100 μM diphenylamine (d). Peak 1 echinenone, peak 2 ß-carotene, peak 3 myxoxanthophyll, peak 4 zeaxanthin and peak 5 phytoene

Full size image

The Synechocystis culture was treated with norflurazon, which is an inhibitor of the cyanobacterial type phytoene desaturase (Linden et al. 1990). As a consequence of this partial block of the pathway, phytoene (peak 5) was accumulated. In contrast to Synechocystis, application of norflurazon to G. violaceus even at a concentration resembling 100 times the I50 value for phytoene desaturase inhibition in other cyanobacteria (Linden et al. 1990) had no effect on the carotenoid composition including the accumulation of phytoene (trace C). Instead, diphenylamine an inhibitor of the CrtI-type phytoene desaturase (Sandmann and Fraser 1993) of bacteria other than cyanobacteria and of fungi was inhibitory to the conversion of phytoene leading to the accumulation of this carotene (trace D, peak 5). This result is the first indication for the existence of a phytoene desaturase in G. violaceus that is totally different to the desaturases from other cyanobacteria (Sandmann 2002).

Since the carotenoid composition of cyanobacteria may be light-regulated (Fernández-Gonzalez et al. 1998; Steiger et al. 1999), G. violaceus was cultivated under very low, medium or high intensity light. No new carotenoids were detected under high light conditions but higher intensities resulted in lower levels of colored carotenoid (Table 2). The highest amounts of ß-carotene and echinenone were found in the culture grown at very low light. The determined carotenoid levels represent a steady-state value resulting from biosynthesis and simultaneous degradation. In parallel experiments, the carotenoid biosynthesis capacity was evaluated in the presence of the phytoene desaturase inhibitor diphenylamine (Table 2). As shown by similar experiments with Synechcocystis, the level of colorless phytoene, which exhibits a better light stability than colored carotenoids, is a useful indicator to determine de-novo synthesis independent of photo-degradation (Steiger et al. 1999). Under high light, the carotenoid biosynthesis activity leading to phytoene was about 1.6-fold higher than under medium light intensity, or 2.4-fold higher than under very low light. These results obtained here for G. violaceus point at a light-dependent regulation of carotenogenesis at the level of phytoene synthesis. An up-regulation of phytoene synthase under high light was also found in Synechocystis (Fernández-Gonzalez et al. 1998) and in leaves of higher plants (Simkin et al. 2003). Surprisingly, degradation of retained colored carotenoids was lowest in the 5 μmol m−2 s−1 culture of Gloeobacter after inhibition (Table 2). One explanation may be a lack of substantial amounts of antioxidants. Their synthesis may be turned on under higher light intensities when photosynthesis is optimal.

Table 2 Carotenoid content of Gloeobacter violaceus treated with and without diphenylamine under different light intensities
Full size table

Functional identification of carotenogenic genes

Homology search in the G. violaceus genome database at Kasuza DNA Research Institute for genes involved in biosynthetic steps leading to keto carotenoids revealed several candidates (Table 1). Open reading frames of a single putative phytoene synthase gene glr1744 was detected together with three different putative crtI-type phytoene desaturase genes, a crtO-related mono ketolase gene gll0394 and a crtW-related diketolase gene gll1728. The presence of the latter was quite unexpected since a diketo carotenoid like canthaxanthin was absent in G. violaceus cells (Fig. 1). No cyanobacterial type desaturase gene crtP or crtQb are present in the genome, and no crtL gene encoding a cyanobacterial lycopene cyclase was found.

All genes from Table 1 (top) were cloned in an E. coli expression vector to express an in-frame fusion to about 35 amino acids encoded by the lacZ gene. The function of each gene was analyzed by conversion of the putative substrate in E. coli and HPLC determination of the reaction product. The peaks of the residual substrates that were not fully converted and that were present as only carotenoids in control experiments are marked in Fig. 2 as dotted lines. Plasmid pMONTglr1744 was co-transformed with a geranylgeranyl pyrophosphate expressing plasmid (trace A). The analysis revealed the formation of phytoene (peak 1) with its typical spectral absorbance. A control transformant with an empty pMON3820 vector was devoid of carotenoids. Plasmids pMONTglr0867, pMONTgll2133 and pMONTgll2874 were all expressed in E. coli with an established phytoene background. Only plasmid pMONTglr0867 mediated the desaturation of phytoene in a 4-step process yielding lycopene (trace B, peak 2) with an absorbance at 446, 472, and 504 nm. The products of orfs gll2133 and gll2878 lack any desaturation activities in E. coli: they were unable to introduce double bonds into phytoene, neurosporene, 1-HO-neurosporene or ß-carotene (data not shown), indicating that they do not function as an acyclic carotene desaturase like CrtI, a 1-hydroxycarotene 3,4-desaturase like CrtD or 1,2,3,4-desaturase of a ß-ionone end group like CrtU. In addition, they were tested as prolycopene isomerase. None of the encoded proteins catalyzed the isomerization of prolycopene to all-trans lycopene, a reaction step typical for cyanobacterial carotenogenesis. Therefore, the existence of a prolycopene isomerase gene crtH (Breitenbach et al. 2001a; Masamoto et al. 2001) is unlikely.

Fig. 2
figure 2

Genetic complementation in Escherichia coli of carotenogenic genes from Gloeobacter violaceus and HPLC analysis of the carotenoid products. a JM101/pACCRT-E/pMONTglr1744, b JM101/pACCRT-EB/pMONTglr0867, c JM101/pACCAR16ΔcrtX/pMONTgll0394, d JM101/pACCAR16ΔcrtX/pMONTgll1728. Peak 1 phytoene, peak 2 lycopene, peak 3 ß-carotene, peak 4 echinenone, peak 5 canthaxanthin. Dotted peaks indicate the substrate provided for conversion by the individual gene products

Full size image

ß-Carotene conversion in E. coli was tested for gll0394 and gll1728. The first transformant synthesized echinenone (trace C, peak 4). The other carrying pMONTgll1728 accumulated the diketo carotenoid canthaxanthin (trace D, peak 5, bell shaped spectrum at 472 nm) in addition to the monoketo derivative echinenone. Thus, gll0394 encodes a ß-carotene ketolase crtO and gll1728 a ß-carotene diketolase crtW.

Recently, a novel lycopene cyclase gene CT0456 was identified in C. tepidum by inactivation (Frigaard and Bryant 2004). The resulting mutant was devoid of cyclic carotenes. The homolog gll3598 from G. violaceus looks like a promising candidate for the missing lycopene cyclase gene. It was cloned in an expression vector and co-transformed in E. coli with a lycopene background. However, the transformant did not form cyclic carotenoids like γ-carotene or ß-carotene (data not shown).

Evolution of carotenoid biosynthesis

Phytoene synthase is a rare example of a carotenogenic enzyme with a continuous line of evolution from archea and bacteria via cyanobacteria to plants (Sandmann 2002). In the phylogenetic tree of Fig. 3a with the enzymes from cyanobacteria, photosynthetic purple bacteria and from Pantoea ananatis (formerly Erwinia uredovora) as a representative non-photosynthetic gram negative bacterium, phytoene synthase from G. violaceus (G.v.glr1744) groups with the cyanobacteria. Protein identity among this cluster is around 50%. Phytoene synthase from G. violaceus is more distant to the enzyme from P. ananatis and from photosynthetic bacteria. The phytoene desaturase G.v.slr0867 is a bacterial CrtI-type. It clusters in the lower part of Fig. 2b with bacterial phytoene desaturases. The closest relationship is to phytoene desaturase from species of the gram positive Deinococcus/Thermus group with a 60% and higher amino acid identity. The closest phytoene desaturase from a photosynthetic bacterium, Rubrivivax gelatinosous shares only 47% identity. The second cluster (top) represents cyanobacterial prolycopene isomerases that evolved from the bacterial phytoene desaturases (Sandmann 2002). The related orfs gll2133 and gll2874 of unknown function are separated from both clusters. Since all carotenogenic enzymes in carotenogenesis of G. violaceus are defined except for a lycopene cyclase for which both orfs do not encode and other functions could not be assigned (data for functional examination not shown), both DNA sequences may resemble non-coding evolutionary relicts.

Fig. 3
figure 3

Phylogenetic tree of phytoene synthases CrtB (a), enzymes related to phytoene desaturase CrtI or prolycopene isomerase CrtH (b) and ß-carotene diketolase CrtW (c). Genes are indicated by species followed by the gene accession number. Abbreviation of species: Bra. Bradyrhizobium ORS278, D.r. Deinococcus radiodurans, E.u. Pantoea ananatis (formerly Erwinia uredovora), G.v. Gloeobacter violaceus, Nos. Nostoc PCC7120, N.p. Nostoc punctiforme PCC73102, Par. Paracoccus PC1 (formerly Alcaligenes), P.mc. Paracoccus marcusii, P.m. Prochlorococcus marinus MIT9313, R.c. Rhodobacter capsulatus, R.g. Rubrivivax gelatinosus, R.p. Rhodopseudomonas palustris, R.r. Rhodospirillum rubrum, R.s. Rhodobacter spheroides, Sco Synechococcus WH8102, Scy Synechocystis PCC6803, T.t. Thermus thermophilus HB27, Xan. Xanthobacter Py2

Full size image

The diketolase CrtW from G. violaceus shows the closest relationship to one of the ketolases from Nostoc punctiforme with 41% protein identity and to ketolases from other cyanobacteria (Fig. 3c). Although the gene product is active in E. coli (Fig. 2), it has no obvious diketolase function in G. violaceus—it may either work in this organism as a monoketolase or may not be expressed due to the lack of a functional promoter. The possibilty of a cryptic diketolase gene would resemble the situation in non-carotenogenic Streptomyces griseus with a silent carotenogenic gene cluster. In this bacterium, carotenoid biosynthesis could be established by the integration of a suitable promoter (Schumann et al. 1996).

Only one monoketolase CrtO from Synechocystis has been functionally characterized (Fernández-Gonzalez et al. 1997). Similar genes from other cyanobacteria can be found in the database and are likely to exist. But due to the close relationship to CrtI genes, gene comparison is not sufficient and may be misleading. Therefore, gll0394 was compared only to BA10561 from Synechocystis PCC6803. Both monoketolases share an amino acid identity of 62%.

The carotenoid pathway of G. violaceus established in this investigation is outlined and compared to the typical cyanobactetrial pathway in Fig. 4. All genes necessary for the synthesis of ß-carotene and echinenone except for lycopene cyclase have been identified. With gll3598 an orf homologous to CT0456 has been found (Table 1). Although the inactivation of this gene in C. tepidum resulted in a phenotype devoid of cyclic carotenoids (Frigaard and Bryant 2004), the enzyme expressed from gll3598 was not functional in lycopene cyclization. One can speculate that this may be due to a heterodimer nature of this very special lycopene cyclase as it is the case for another but absolutely unrelated lycopene cyclase from Brevibacterium linens (Krubasik and Sandmann 2000).

Fig. 4
figure 4

General carotenoid biosynthetic pathway to ß-carotene and echinenone in cyanobacteria and specifically in Gloeobacter violaceus. Gene products are indicated at the individual reactions. The gene product in parentheses is the most like candidate for lycopene cyclase by comparison to CT0456 although complementation gave a negative result

Full size image

Phytoene synthase and both ketolases from G. violaceus are closely related to the corresponding gene products from other cyanobacteria. However, a major discrepancy with carotenogenesis of other cyanobacteria is the involvement of a bacterial phytoene desaturase CrtI and the absence of the cyanobacterial CrtP and CrtQb desaturases and of a prolycopene isomerase CrtH. 15-cis Phytoene is converted in four steps to all-trans lycopene by a single enzyme. Unlike in other cyanobacteria (Breitenbach et al. 2001b), dicis and tetracis carotenes are no intermediates of the pathway in G. violaceus. In a recent review on the phylogeny of carotenogenic genes (Sandmann 2002), it was concluded that desaturation in cyanobacterial carotenogenesis occurs by an exclusive utilization of newly acquired desaturases CrtP and CrtQb. This view has now to be modified. Not only was a green photosynthetic bacterium found since then, possessing the cyanobacterial type of carotene desaturase and isomerase genes (Frigaard et al. 2004), but also G. violaceus a cyanobacterium with the original crtI-type desaturase found in archea and other bacteria exists. In addition to the unique reactions of the biosynthetic pathway, the low amount of carotenoids in G. violaceus that resembles only one tenth of the total carotenoids of Synechocystis (Steiger et al. 1999), and the carotenoid pattern with ß-carotene as the major carotenoid and echinenone as the only xanthophyll is unique to cyanobacteria and other organisms with oxygenic photosynthesis (Goodwin 1980). This carotenoid inventory offers only limited protection for the photosynthetic apparatus against high light intensities and may be one of the reasons why G. violaceus grows poorly under strong light (Rippka et al. 1974). Both features, the carotenoid composition and the bacterial type carotenogenic pathway, are additional metabolic peculiarities of G. violaceus and support its isolated position among the other cyanobacteria by an early divergence from this group (Nelissen et al. 1995).