General comparison of the carotenoids biosynthetic genes from cyanobacteria
Similarity search between query sequences and cyanobacterial genomes were performed by BLASTP program. The distribution of genes involved in carotenoid biosynthesis across 18 cyanobacterial genomes is summarized in Figure 2. We can see geranylgeranyl pyrophosphate synthase (crtE/GGPS) and phytoene synthase genes (crtB/pys) are widely distributed among all the species. The cyanobacteria share the same carotenoid biosynthetic pathway to lycopene except for G.violaceus 7421.
Multiple alignments of the predicted amino acid sequences from the homologous carotenoid biosynthetic genes from cyanobacteria were constructed. A similar degree of difference in these proteins among cyanobacteria is noted here. Consistent with the hypothesis that the early reactions of carotenoid biosynthesis are conserved , the present study also reveals the enzymes are more conserved in the upstream pathway. In spite of the difference in the lycopene biosynthetic pathway between G. violaceus PCC 7421 and other species, the enzymes in the formation of phytoene have the close phylogenetic relationship supported by more conserved domain. Aside from G. violaceus PCC 7421, the crtP shares more than 60% amino acid identity across different species. With exception of the crtQa from Anabaena sp. PCC 7120, ζ-carotene desaturase also have highly similarity with the amino acid identity from 55% to 99.3% among various cyanobacteria. While the carotene ketolase and carotene hydroxylase in the late steps are significantly less conserved than other enzymes in the pathway.
The diversity of enzyme involved in the desaturation step
Phytoene is converted to lycopene by four-step desaturation and use two related enzymes phytoene desaturase (CrtP/Pds) and ζ-carotene desaturase (CrtQ/Zds) in the most of cyanobacteria; However, G.violaceus PCC7421, like most bacteria and fungi, uses only one enzyme, phytoene desaturase (CrtI), catalyzing four-step in this pathway. Surprisingly, homologues of CrtI from G. violaceus PCC7421 are also found in Anabaena variabilis ATCC 29413 (e=0/Identities=57%), Anabaena sp. PCC 7120 (e=0/Identities57%), Trichodesmium erythraeum IMS101 (e=0/Identities=57%), Crocosphaera wastsonii WH8501 (e=0/Identities=55%), Synechocystis sp.PCC6803 (e=0/Identities=56%), which are not involved in the lycopene biosynthetic pathway. Thus, although the crtI homologs in these cyanobacteria appear to be involved in carotenoid biosynthesis, their functions are different from that of crtI in G. violaceus PCC 7421 and bacteria. We therefore propose that these enzymes originated in a same ancestor and then evolved into a different enzyme in different cyanobacteria that produces novel carotenoids that acquire new physiological function. The carotenoid biosynthetic pathway in G. violaceus PCC 7421 is unique contrast to other cyanobacteria. The molecular phylogenetic analysis based on 16S rRNA also demonstrated an isolated position away from other groups of cyanobacteria for G. violaceus . This organism is thought to retain traces of the ancestral properties of cyanobacteria.
crtQa from Anabaena sp. PCC 7120 which had been functionally identified to convert ζ –carotene to lycopene  while crtQb is involve in this desaturation step in other species . By BLASTP program, we also found homologue of crtQb from Anabaena sp. PCC7120, but the information on its function is not available yet. crtQa was found no homologues in other species. Nevertheless both crtQa and crtQb convert ζ –carotene to lycopene, they have no similarity in sequence, and only crtQb displays high conservation with the plant counterparts. The crtQb and crtP from cyanobacteria show high similarities in their amino acid sequence and both contained partial amine oxidoreductase domain. It is very likely that they evolved from the same ancestor. Surprisingly, crtQa is share little sequence similarity to the 'plant-type' phytoene desaturase (crtP gene product), but it has considerable conserved with the bacterial-type enzyme (crtI gene product). It is possible that the cyanobacterial crtQa gene and crtI gene of other microorganisms originated in evolution from a common ancestor.
The evolutionary analysis of crtL- type cyclase and its absence in some species
The cyclization reaction of lycopene to β-carotene is also related to different enzymes. The ends of the resulting acyclic lycopene may be cyclized to β-ionone, or ε-ionone rings. The formation of β-ionone rings and of ε-ionone rings in plants is catalyzed by two different enzymes, the β-cyclase and the ε-cyclase. The same case is in some cyanobacteria. Both enzymes show high similarities in their amino acid sequence and it is very likely that they evolve from the same ancestor . The phylogenetic relationship among the crtL from cyanobacteria, green algae and higher plants is depicted in Fig.3. From the phylogenetic tree, we can see the enzymes fall into two groups. The cyclase from cyanobacteria separating from other cyclase formed monophyletic group divided into two subclusters containing the β-cyclase and the ε-cyclase. So we suppose that gene may be duplicated after the speciation of the cyanobacteria, chlorophytes and plants. In order to well understand the evolution of the β-cyclase and the ε-cyclase, we examined the tertiary structure using the lycopene cyclases from Arabidopsis thaliana and Prochlorococcus marinus str. MIT 9312 as an example (Fig. 4). A comparable analysis for the tertiary structure of cyclase from cyanobacteria and plants reveal β-cyclase and ε-cyclase have similar structure folds from the same organisms. A single loop formed with five β-strands and one α-helix has conserved in four models, which may be related to binding domain. Several antiparallel β-strands both contained in the tertiary structure of plant-type β-cyclase and ε-cyclase are lacking in that of the cyanobacteria. We supposed lycopene cyclase in a given lineage may evolve through gene duplication that happened after cyanobacteria and chlorophytes/plants speciation event.
However, it is interesting that only in genus Prochlorococcus, both of lycopene β-(crtL-b) and ε-cyclase (crtL-e) enzymes were found, while, in Synechococcus only one enzyme has good hit with the query sequence. Although there is not only no detectable crtL-e- but also no crtL-b-like lycopene cyclase gene in the genomes of Synechocystis sp. 6803, Thermosynechococcus elongatus, Trichodesmium erythraeum, Gloeobacter, Crocosphaera wastsonii WH8501, Nostoc punctiforme and Anabaena, the related carotenoids had been detected in some species[5-7]. It would be of interest to know which enzymes converting lycopene to β-carotene in these cyanobacteria. Recently, Takaichi et al (2005)  found Anabaena sp. PCC 7120 alr3524 has sequence homology to a new type lycopene cyclase CruA from Chlorobium tepidum . Then we used alr3524 from Anabaena sp.7120 and CruA from C. tepidum as query sequence, it is interesting to found homologous enzymes were identified in Thermosynechococcus elongates, Anabaena, Nostoc, Synechocystis sp. PCC 6803, Trichodesmium, C.watsonii WH 8501, G.violaceus PCC7421 other than in Prochlorococcus and Synechococcus (Table1), but their functions have yet to be investigated.
Conserved domain between crtW-type ketolase and crtR-type hydroxylase
Two distinct β-carotene ketolase genes, crtW and crtO, were found in the genome sequences of cyanobacteria. Anabaena sp. PCC 7120, N.punctiforme PCC 73102, Anabaena ATCC 29413 and G.violaceus PCC 7421 and Synechocystis sp. PCC 6803 were found contain crtO homologous gene, Synechococcus WH8102 and Synechococcus sp. CC9902 were found contain crtW homologous genes. Although these two enzymes involve the same β-carotene ketolation, the characteristics of enzymes are different. CrtO and crtW do not share significantly amino acid sequence homology. CrtOs have six conserved regions including the FAD binding motif  and show partial amino oxidase domain, while crtWs sharing three typical histidine rich motifs (Table2) show some characters of fatty acid desaturase. Carotenoid hydroxylases (crtR) in cyanobacteria bears little or no relationship to the carotennoid hydroxydrases from plants and bacteria. It shows some similarity to crtW-type ketolase, especially conserved in the three H-Boxes (Fig.5), which reveal the crtR and crtW might have a common ancestor and acquire the different function during the evolution. The origin and phylogenetic position of crtR and crtW relative to other members of the three H–boxes FA protein family is of considerable interest.
Structure of crt gene cluster in the cyanobacterial chromosomes
To elucidate the complete genomic structure of the crt genes, we mapped them onto cyanobacterial genomes (Fig. 6). The structure of the crt gene clusters varies greatly among species. In Prochlorococcus, crtB, crtP and crtQb often clustered and transcribed in the same direction. Actually, in many cases, crtB and crtP are directly adjacent to each other on the chromosome and may form an operon. crtE and crtL-b formed an operon in Prochlorococcus NATL, Prochlorococcus MED4, Synechococcus WH8102 and Synechococcus sp. CC9902. While other crt genes are arranged in random in the genome and they are not always transcribed in the same direction. It is interesting cyanobacteria is distinct from other eubacteria in the organization of crt clusters although forming a coherent systematic group, genes for carotenoid biosynthetic enzymes are frequently clustered into large operons [28-30] in typical bacteria, but this does not appear to be the case in cyanobacteria. Although the cartoenoid biosynthetic pathway in G. violaceus PCC 7421 is similar with other eubacteria, the genomic structure of crt genes is not distinct from other cyanobacteria.
The crt genes are arranged in random in the cyanobacteria chromosomes. These loosely organized operon structures are sometimes considered ''destructed'' due to genome rearrangement, and secondary in origin . While genome rearrangement and even gene displacement can be common during operon evolution , fragmentation of a well adapted operon will at least require the evolution of regulatory elements for newly generated gene clusters. The crt genes will acquire the new regulatory elements respectively to adapt for new environments.
Each of these enzymes is a single-gene produce in most cases. Multiple copies of ketolases were only identified in the filamentous species. Actually, two carotenoid ketolase genes crtW38 and crtW148 were cloned from the cyanobacterium, Nostoc punctiforme PCC 73102 and functionally characterized . Scanning the genomics of all species for crt genes by the similarity search we also found two crtO ketolases and two crtW existed in Nostoc punctiforme PCC 73102 and Anabaena ATCC 29413 respectively. There are no paralogous copies of crt genes other than in filamentous cyanobacteria. Most of filamentous cyanobacteria exhibit a wide range of ecological tolerance and are found in freshwater, marine and terrestrial habitats. The increased number of isozymes associated with pigment biosynthesis in filamentous cyanobacteria relative to unicellular species may be related to increased regulatory demands and perhaps also to different local environments.