Identification of a Novel Vinyl Reductase Gene Essential for the Biosynthesis of Monovinyl Chlorophyll in Synechocystis sp. PCC6803*

The vast majority of oxygenic photosynthetic organisms use monovinyl chlorophyll for their photosynthetic reactions. For the biosynthesis of this type of chlorophyll, the reduction of the 8-vinyl group that is located on the B-ring of the macrocycle is essential. Previously, we identified the gene encoding 8-vinyl reductase responsible for this reaction in higher plants and termed it DVR. Among the sequenced genomes of cyanobacteria, only several Synechococcus species contain DVR homologues. Therefore, it has been hypothesized that many other cyanobacteria producing monovinyl chlorophyll should contain a vinyl reductase that is unrelated to the higher plant DVR. To identify the cyanobacterial gene that is responsible for monovinyl chlorophyll synthesis, we developed a bioinformatics tool, correlation coefficient calculation tool, which calculates the correlation coefficient between the distributions of a certain phenotype and genes among a group of organisms. The program indicated that the distribution of a gene encoding a putative dehydrogenase protein is best correlated with the distribution of the DVR-less cyanobacteria. We subsequently knocked out the corresponding gene (Slr1923) in Synechocystis sp. PCC6803 and characterized the mutant. The knock-out mutant lost its ability to synthesize monovinyl chlorophyll and accumulated 3,8-divinyl chlorophyll instead. We concluded that Slr1923 encodes the vinyl reductase or a subunit essential for monovinyl chlorophyll synthesis. The function and evolution of 8-vinyl reductase genes are discussed.

chlorophyll a 3 and b (Fig. 1) (3,4). Among them, the biosynthesis of chlorophyll a, b, c, d and all bacteriochlorophyll species requires the reduction of the 8-vinyl group (2). All of these chlorophyll molecules have a common backbone structure. Modification of the side chains on the common backbone structure gives rise to the diversity of chlorophylls (5,6). Among the variety of chlorophyll species, bacteriochlorophyll a and monovinyl chlorophyll a ( Fig. 1) are most commonly used for the photochemistry in photosynthetic organisms. The exceptions to this are a group of marine cyanobacteria, Prochlorococcus species (7), and a cyanobacterial symbiont, Acaryochloris marina (8), which use 3,8-divinyl chlorophyll a ( Fig. 1) and chlorophyll d for their photochemistry, respectively. It is not clear why the majority of photosynthetic organisms prefer monovinyl chlorophylls instead of divinyl chlorophyll species. Akimoto et al. (9) recently suggested that the replacement of monovinyl chlorophylls by 3,8-divinyl chlorophylls in the dvr mutant of Arabidopsis thaliana significantly changed the antenna system and thus affected energy and electron transfer. In addition, our group also observed that the dvr mutant is highly susceptible to strong illumination (10). There may be some (yet unknown) advantage for electron and/or energy transfer in the reduction of the 8-vinyl to an 8-ethyl group.
For the synthesis of monovinyl chlorophyll, the reduction of the 8-vinyl group on the pyrrole ring B of 3,8-divinyl chlorophyllide (or possibly 3,8-divinyl protochlorophyllide) is a prerequisite. In our previous study (10), we identified a gene encoding 8-vinyl reductase that is responsible for the synthesis of monovinyl chlorophyll from Arabidopsis, and we termed the gene DVR. Homologues of the DVR gene were found in the genomes of higher plants, green algae, and Prasinophytes. In addition, most but not all green sulfur bacteria and some purple bacteria have DVR homologues that were termed bciA (19). Interestingly, there were no homologues identified in the genome of a red alga, Cyanidioschyzon merolae. Homologues were also found in five Synechococcus species but were not identified in any other cyanobacteria. Because C. merolae and the majority of cyanobacteria that lack DVR homologues synthe-* This work was supported by Grant-in-aid for Creative Scientific Research size monovinyl chlorophyll, it has been hypothesized that an unidentified 8-vinyl reductase is present in these organisms.
To identify the second DVR gene, we first searched for genes exclusive to C. merolae and cyanobacteria that do not possess DVR homologues but synthesize monovinyl chlorophyll a. For this purpose, we developed a bioinformatics tool, correlation coefficient calculation tool (CCCT), 4 to select the candidates of the cyanobacterial genes encoding the hypothetical 8-vinyl reductase. We subsequently carried out a molecular genetic analysis and finally identified the gene in cyanobacteria that is involved in the 8-vinyl reductase activity. The function of the gene product and its evolutionary history are discussed.

EXPERIMENTAL PROCEDURES
Materials and Growth Conditions-Synechocystis sp. PCC6803 was grown at 30°C in BG-11 medium (11) in liquid and solid (0.5% agar) media under continuous illumination with normal light (50 mol photons/m 2 s, white fluorescence lamp) or high light (1000 mol photons/m 2 s).
Spectrometric Measurements-Absorption spectra of cells were measured by using an opal diffuser at room temperatures with a Hitachi U-3310 spectrophotometer (Hitachi, Tokyo, Japan).
Electron Microscopy-Wild type and mutant Synechocystis sp. PCC6803 cells were grown under normal light conditions. The cells were harvested by centrifugation and embedded in 3% low melting agar. The agar-containing cells were initially fixed in 2.5% glutaraldehyde in 0.1 M cacodylate buffer (pH 7.4). Agar-embedded specimens were postfixed in 1% OsO 4 in 0.1 M cacodylate buffer and were dehydrated in a graded ethanol series. Dehydrated specimens were subsequently embedded in an Epon resin mixture (TAAB Epon 812, TAAB Laboratories Equipment Ltd., Berkshire, UK) (13). Ultra-thin sections were mounted on 200-mesh size grids. Specimens were stained for 20 min with 2% (w/v) aqueous uranyl acetate and briefly with a 0.2 g/liter solution of lead citrate. Specimens were viewed with a JEOL 1200EX transmission electron microscope (JEOL Ltd., Tokyo, Japan) at an accelerating voltage of 60 kV.
Gene Cloning from Synechocystis sp. PCC 6803-To inactivate the putative 8-vinyl reductase gene of Synechocystis sp. PCC6803, the Slr1923 gene was amplified by PCR using primers 5Ј-GCCCAGTGGACAGGCATTGT-3Ј and 5Ј-ATGGG-TAACGTTACCGCCTT-3Ј. The PCR product was cloned into the pGEM T-easy vector (Promega). The kanamycin resistance gene from Escherichia coli was inserted into the SmaI site that is present in the inserted DNA fragment.
Transformation and Segregation Analysis of Synechocystis sp. PCC 6803-Transformation of Synechocystis sp. PCC6803 was carried out according to the methods described previously (14). Transformants were propagated on BG-11 agar plates containing 20 g/ml kanamycin. For the promotion of segregation, transformants were incubated under dim light (1 mol photons/m 2 s) with 5 mM glucose supplementation in agar medium. The segregation state of the transformants was analyzed by PCR.
Sequence Retrieval and Blast Analysis-Sequences were obtained either from the GenBank TM data base or from Joint Genome Institute. The stand-alone BLAST program (version 2.2.10) (15) was obtained from NCBI, and default parameters were used. Mathematica for Mac OS X, version 6.0.1 (Wolfram Research, Inc.) was used to control the BLAST program and to calculate the correlation coefficient as described under "Results." We will freely distribute the program to calculate the correlation coefficient upon request.
Phylogenetic Analysis-The deduced amino acid sequences of DVR and Slr1923 homologues were aligned using the Clust-alW software (16) with Gonnet residue weights. A neighbor joining tree (17) was constructed on the basis of the sequence alignment with the ClustalW software.

RESULTS
Basic Characterization of 8-Vinyl Reductase in Synechocystis sp. PCC 6803-Prior to the start of our bioinformatics approach to identify 8-vinyl reductase in Synechocystis sp. PCC 6803, we examined whether this organism possesses specific activities for the reduction of the 8-vinyl group of the tetrapyrrole. Thylakoid membranes of wild type cells were purified and solubilized by Triton X-100. 3,8-Divinyl chlorophyllide a was incubated with the solubilized membranes, and the formation of monovinyl chlorophyllide a was monitored by HPLC after a 15-min incubation. When NADPH was added to the reaction mixture, the formation of monovinyl chlorophyllide a was observed. Without the addition of NADPH, only 3,8-divinyl chlorophyllide a was detected. 5 This experiment provided evidence which confirmed that 8-vinyl reductase is present in this organism, and its activity is NADPH-dependent. Thus, we initiated the following bioinformatics approach to identify the gene encoding the second 8-vinyl reductase. Selection of Possible 8-Vinyl Reductase Genes from the Cyanobacterial Genome-For the following bioinformatics approach, a few assumptions were made. 1) All cyanobacteria producing monovinyl chlorophyll should have a specific enzyme(s) that reduces the 8-vinyl group of the tetrapyrrole. 2) Prochlorococcus species lack 8-vinyl reductase, because they do not produce monovinyl chlorophyll. 3) There are only two types of 8-vinyl reductase in cyanobacteria. The last assumption was made in order to make our bioinformatics approach possible. Without this assumption, the list of the candidate genes for those encoding 8-vinyl reductase would become too long for the subsequent reverse-genetics approach. We should note that 5 of 16 Synechococcus species whose genomes have been sequenced contain a DVR homologue. Therefore, it was not possible for us to accurately predict whether Synechococcus species have the second 8-vinyl reductase or not. Thus, we excluded those five species from the bioinformatics analysis. Taken together, if our assumption is correct, the second 8-vinyl reductase should exist in all cyanobacteria that produce monovinyl chlorophyll a except for the five Synechococcus species, and it should be absent from Prochlorococcus species. For the bioinformatics analysis, the cyanobacteria that were expected to have the second 8-vinyl reductase were placed into the "plus" group for bioinformatics analysis and the Prochlorococcus species were placed into the "minus" group. The second 8-vinyl reductase should be absent from this minus group. A red alga, C. merolae, was also included in the plus group, because its sequenced genome does not contain a DVR homologue. The addition of a eukaryotic organism to the analysis considerably increases the amount of information on the distribution of each gene.
In general, attempts to identify orthologues by similarity searches using tools such as BLAST (15) face a common problem in determining the threshold of similarity with which orthologues are selected. If only genes showing very high similarity to the query sequence are selected as orthologues, there is a high probability for true orthologues to go undetected within the search. On the contrary, if genes showing lower similarity to the query sequence are selected as orthologues, false genes might be identified as orthologues. To avoid this problem, CCCT ranked possible orthologues by calculating the correlation coefficient between the ability of monovinyl chlorophyll synthesis and the E value of each gene to the query sequence. Fig. 2 shows an example of CCCT analysis. If the query sequence shows high similarity (low E value) to one of the genes that is found only in the genomes of the plus group, CCCT gives higher values (correlation coefficient) closer to 1. In contrast, when another sequence shows high similarity to a gene found within the genomes of the minus group, the program generates a lower value that is closer to Ϫ1. With this strategy, both the pattern of gene distribution and the phenotype can be compared and scored.
In this study, all ORFs from the Synechocystis sp. PCC 6803 genome were used as a query sequence, and BLAST searches were performed against each genome from 32 organisms that were categorized either in the plus or minus group (see Fig. 2 for the list of organisms). For each genome, an ORF exhibiting the lowest E value to each query sequence was selected, and its value was used for the calculation of the correlation coefficient. Finally, all ORFs were ranked according to the order of higher correlation coefficient. Fig. 2, B-D, shows the E value profiles of the candidate proteins that were identified through our CCCT analysis. The highest correlation coefficient was found in Slr0335 (r ϭ 0.998), which encodes a phycobilisome LCM core membrane linker polypeptide. The second and third highest were Slr1923 (hypothetical protein, r ϭ 0.993) and Sll0757 (amidophosphoribosyltransferase, r ϭ 0.957), respectively. As shown in Fig. 2, B and C, the distribution pattern of two genes, Slr1923 and Slr0335, matched our categorization pattern of the plus and minus groups. The distribution of the third gene, SLL0757, also matched our categorization pattern except for red algae "C. merolae" (Fig. 2D). Slr0335 was excluded from further analysis because the gene encodes a known component of phycobilisome. The Slr1923 gene has sequence similarity to the F420 reducing hydrogenase (FRH) ␤-subunit of Methanobacterium thermoautotrophicum (18). Although the reaction catalyzed by FRH is not very similar to that presumed for the reaction by 8-vinyl reductase, we thought it was possible that Slr1923 was involved in the reduction of the 8-vinyl group, because an oxidoreductase family can potentially catalyze a wide range of reactions. Therefore, we decided to pursue this promising gene for further functional analysis.
Disruption of the Slr1923 Gene-To examine whether the Slr1923 gene product participates in the reduction of the 8-vinyl group on pyrrole ring B, we disrupted the Slr1923 gene by homologous recombination in Synechocystis sp. PCC 6803 (supplemental Fig. 1). Using an HPLC system incorporating a photo-diode array detector, we analyzed the pigment compositions of the slr1923 mutant and wild type cells (Fig. 3). In the mutant cells, a chromatographic peak corresponding to the monovinyl chlorophyll (8.56 min, peak 3-2) disappeared. Instead, another peak appeared at an earlier elution time (8.44 min, peak 3-1) in the mutant cells. The early peak in the mutant has the same retention time and absorption spectrum to those of 3,8-divinyl chlorophyll, which accumulate in the dvr mutant of A. thaliana (10). Furthermore, this peak provided a [M ϩ H] ϩ m/z value (870.5), which corresponds to the Mg 2ϩ -dechelated 3,8-divinyl chlorophyll a (3,8-divinyl pheophytin a). In contrast, the HPLC peak corresponding to monovinyl chlorophyll of wild type gave a [M ϩ H] ϩ m/z value of 872.5, which represents Mg 2ϩ -dechelated monovinyl chlorophyll a (pheophytin a). It is important to note that the Mg 2ϩ ion was removed from the chlorophyll molecules during the mass spectrometry analysis. Taken together, these results clearly demonstrate that Slr1923 encodes an essential component of 8-vinyl reductase, which is required for the monovinyl chlorophyll synthesis in this organism.
The Slr1923 gene is located in a gene cluster consisting of Slr1923, Slr1924, and Slr1925 in this order. Slr1924 and Slr1925 encode carboxypeptidase and a protein involved in cobalamin biosynthesis, respectively. It was possible that disruption of Slr1923 affected the expression of Slr1924 and Slr1925. To examine whether possible disturbance in the expression of Slr1924 and Slr1925 had affected the synthesis of chlorophyll, we inserted the kanamycin resistance gene into Slr1924. We

8-Vinyl Reductase for Chlorophyll Biosynthesis
found that the Slr1924 disruptant did not show any observable phenotype, including accumulation of 3,8-divinyl chlorophyll (supplemental Fig. 2). These results indicate that the Slr1924 protein is not involved in the reduction of the 8-vinyl group. Furthermore, the same experiment indicates that Slr1925 is also not involved in this reaction. If the accumulation of 3,8-divinyl chlorophyll was caused by the interference of the expression of Slr1925 in the slr1923 mutant, then the disruption of the Slr1924 gene should have resulted in accumulation of 3,8-divinyl chlorophyll, but this was not the case (see supplemental Fig. 2). Collectively, we concluded that the accumulation of 3,8-divinyl chlorophyll in the slr1923 mutant was not the result of possible disturbance in Slr1924 and Slr1925 gene expression.
Phenotype of the Slr1923 Knockout Mutant-Disruption of Slr1923 not only affected chlorophyll synthesis but also carotenoid synthesis. With respect to the wild type, the ratio of zeaxanthin to chlorophyll increased 2-fold; however, the ␤-carotene-to-chlorophyll ratio remained unchanged in the mutant cells. As mentioned under "Discussion," it is possible that the increase in the zeaxanthin contents is induced by the light stresses. Electron microscopy showed that the average size of wild type cells was about 1.33 Ϯ 0.04 m (n ϭ 4), whereas the slr1923 mutant cells was about 1.15 Ϯ 0.07 m (n ϭ 4) (Fig. 4). With respect to wild type, the number of thylakoid membranes per cell was greatly reduced in the mutant cells, and they also exhibited a wider space between thylakoid membranes. These morphological changes in the mutant cells may also indicate that they are experiencing stressful conditions. Part of the stress could be that the 3,8-divinyl chlorophyll does not bind well to the proteins and is partly present in free form, which would be highly phototoxic.
In our previous paper, we reported that the dvr mutant of A. thaliana, whose monovinyl chlorophyll was replaced by 3,8-divinyl chlorophyll, was completely bleached within a day under exposure to high light conditions (10). Therefore, we intended to determine whether cyanobacteria containing 3,8-divinyl chlo- CCCT calculated the correlation coefficient between the two variables, that is the presence of homologues to the query sequences (Synechocystis ORFs) and that of the hypothetical 8-vinyl reductase. The organisms whose sequences were used for CCCT analysis are indicated on the x axis. ϩ or Ϫ after the label for each organism indicates the hypothetical presence or absence of the second (non-plant type) 8-vinyl reductase. A, theoretical model shows the ideal pattern in which the hypothetical presence of the second 8-vinyl reductase (absence of plant-type DVR homologues) and the presence of homologues (which is expressed by the E value of the BLAST search) to the query sequence perfectly match with each other. B-D, the graphical representation of the correlation patterns for three loci that are listed at the top by CCCT analysis. APRIL 4, 2008 • VOLUME 283 • NUMBER 14 rophyll also exhibit a similar response to strong illumination. Wild type and mutant cells were initially cultured in liquid medium under normal light conditions (50 mol photons/m 2 s) and were subsequently transferred to high light conditions (1000 mol photons/m 2 s). Wild type cyanobacteria turned slightly yellow but were able to survive exposure to high light illumination (Fig. 5). In contrast, slr1923 mutant cells were completely bleached after 1 day of high light treatment (Fig. 5). These convincing results may indicate that the replacement of monovinyl chlorophyll with 3,8-divinyl chlorophyll induces a similar response through an analogous mechanism between higher plants and cyanobacteria.

8-Vinyl Reductase for Chlorophyll Biosynthesis
The growth rate of the slr1923 mutant was indistinguishable from that of the wild type cells under low light conditions (10 mol photons/m 2 s), and it was slightly lower than the wild type cells under standard light conditions (40 mol photons/m 2 s) (Fig. 6). Under high light conditions (150 mol photons/m 2 s), the growth of slr1923 mutant was significantly retarded. These results also showed that the mutant accumulating 3,8-divinyl chlorophyll was sensitive to high light irradiation.
Spectral Analysis of the Mutant-To further identify the effects of the accumulation of 3,8-divinyl chlorophyll, absorb-ance spectra of the mutant cells were measured (Fig. 7). The absorbance peak in the soret region was red-shifted by 6 nm in the mutant because of the accumulation of 3,8-divinyl chlorophyll. Absorption corresponding to chlorophyll (around 440 and 660 nm), carotenoids (around 490 nm), and phycobiliproteins (around 630 nm) was observed in both wild type and mutant cells. With respect to wild type, absorption corresponding to chlorophyll decreased in the mutant. However, absorption corresponding to phycobilisomes increased in the mutant. These data indicate that the ratio of phycobiliproteins to chlorophyll increased in the slr1923 mutant. Because a decrease in chlorophyll content was observed with in the dvr mutant of Arabidopsis (10), it is possible that the same phenomenon may have occurred in the slr1923 mutant. It should be noted that phycobilin synthesis was not inhibited in the slr1923 knock-out

8-Vinyl Reductase for Chlorophyll Biosynthesis
mutant as judged by its absorption spectrum (Fig. 5). Considering that phycobilins are synthesized by oxidative ring opening of heme, these data indicate that the common pathway of the heme and chlorophyll biosynthesis was not significantly affected in the mutant. Instead, it is likely that the chlorophyll branch of tetrapyrrole synthesis is predominantly inhibited compared with the heme branch. A possible explanation is that chlorophyll proteins are destabilized, causing the amount of free chlorophyll or chlorophyllide to rise, which in turn feed back to reduce the flux of intermediates through the pathway. Phylogenetic Analysis of 8-Vinyl Reductase-We analyzed the distribution of Slr1923 homologues in photosynthetic organisms (Fig. 8A). We could not find any Slr1923 homologues in the Synechococcus species, in which we found DVR homologues, and in Prochlorococcus species. Instead, we found Slr1923 homologues in other cyanobacteria that have been deposited in the GenBank TM data base. In addition, Slr1923 homologues were found in higher plants, green sulfur bacteria, green filamentous bacteria, and purple bacteria, and their predicted amino acid sequences are highly conserved (supplemental Fig. 3A for the alignment of the sequences). Despite this high conservation, there is no evidence showing that these Slr1923 homologues encode functional 8-vinyl reductase genes, with the exception for those belonging to the cyanobacterial clade. According to the public microarray data base in NASC, the Arabidopsis homologue of Slr1923 was predominantly expressed in the photosynthetic organs, and its expression was increased by the induction of leaf senescence in Arabidopsis. Interestingly, our phylogenetic analysis suggested that a red algal sequence belongs to the cyanobacterial clade.
For comparison with the phylogenetic tree of Slr1923 homologues, we constructed a phylogenetic tree for DVR homologues. We found DVR homologues in higher plants, green algae, prasinophytes, diatoms, green sulfur bacteria, cyanobacteria (Synechococcus species), and purple bacteria (Fig. 8B). We expressed recombinant DVR proteins from Arabidopsis (10) and Synechococcus WH8102 5 in E. coli and confirmed that these protein have 8-vinyl reductase activity for 3,8-divinyl chlorophyllide a. Thus, it would be reasonable to assume that all DVR homologues in the green lineage clade and the Synechococcus clade encode functional 8-vinyl reductase. In addition, Chew and Bryant (19) also detected the 8-vinyl reductase activity with a recombinant BciA protein from Chlorobium tepidum. Therefore, it is likely that all the proteins belonging to the green sulfur bacterial clade are genuine 8-vinyl reductase enzymes. To the best of our knowledge, there is no current evidence which shows that proteins belonging to other clades are also functional 8-vinyl reductase enzymes, even though their predicted amino acid sequences are well conserved (supplemental Fig. 3B for the alignment of the sequences).

DISCUSSION
Bioinformatics Analysis-Three bioinformatics strategies may be potentially utilized to aid the identification of genes that are responsible for certain cellular functions. The first strategy is based upon a motif search. Proteins that participate in specific cellular processes often have distinctive motifs, such as the Rieske center binding motif. This strategy has been proven successful when searching within an organism containing only a limited number of proteins possessing the motif of interest (20,21). This strategy is only applicable when the protein (or gene) of interest is predicted to have a distinctive motif. In the case of 8-vinyl reductase, it was difficult to predict whether or not this protein possesses a distinctive motif that could be identifiable through a motif search. For a second strategy, which aids in the identification of target genes, one can employ a co-expression analysis (22). This type of gene expression analysis compares the expression patterns of a selected gene with the other genes from a model organism such as A. thaliana. This method is powered by comparative analysis of data that has been deposited on public DNA microarray data bases. This type of gene expression analysis is only applicable when a large microarray data base is available for the organism of interest. In addition, it is limited to the study of genes exhibiting a predictable expression pattern. Therefore, this method was not applicable to the search for the novel 8-vinyl reductase gene. The third method is based on the comparison of the gene sets between organisms. Generally speaking, a particular function of a certain organism should be provided by some specific gene products. If the correlation between the occurrence of a particular function and a specific gene is estimated among various organisms, it facilitates the prediction of the gene that is responsible for the function of interest. Previously, we successfully identified the gene encoding chlorophyllide a oxygenase (an enzyme responsible for chlorophyll b biosynthesis) in Prochlorococcus species by comparing the gene sets between a chlorophyll b-containing organism, Prochlorococcus marinus and chlorophyll b-less Synechococcus species (23). However, the bioinformatics tool that was used in the aforementioned study was only applicable to the closely related organisms (23). This tool also had limitations because it cannot analyze the genomes of more than four organisms. In this study, we developed a new tool, CCCT, which can compare the gene sets from an unlimited number of organisms. With this method, we successfully identified the gene involved in the reduction of the 8-vinyl group of the tetrapyrrole. Because the number of available genome sequences is rapidly

8-Vinyl Reductase for Chlorophyll Biosynthesis
increasing, our new method may be applied for a wider range of genes of interest.
Function of Slr1923-Disruption of Slr1923 resulted in the replacement of monovinyl chlorophyll by 3,8-divinyl chlorophyll. These data indicated that Slr1923 is involved in the reduction of the 8-vinyl group. However, when recombinant Slr1923 protein was expressed in E. coli, we were not able to detect any enzymatic activity. 5 It is possible that the Slr1923 gene product did not form a proper conformation because of the heterologous gene expression system of E. coli and was therefore rendered unable to exhibit enzymatic activity. It is also possible that additional proteins are required for its catalytic activity. Slr1923 has sequence similarity to the gene encoding the ␤ subunit of FRH in M. thermoautotrophicum (24). Three heteromeric FRH subunits (␣, ␤, and ␥) form an (␣␤␥) 8 complex in this organism. The ␤ subunit was expected to bind the substrate molecule and has a catalytic function. The ␣ and ␥ subunits participate in the electron transfer from a hydrogen molecule to the ␤ subunit (18). The genome of Synechocystis sp. PCC6803 contains homologous genes to ␣ (Sll1226) and ␥ (Sll1224) subunits, both of which form a gene cluster. We disrupted Sll1224 and Sll1226 by sitedirected mutagenesis; however, 3,8divinyl chlorophyll did not accumulate in both mutants (supplemental Fig. 4). These data indicate that Sll1224 and sll1226 are not involved in the reduction of the 8-vinyl group. Therefore, we concluded that the Slr1923 gene product does not form a complex similar to FRH. These results are plausible if we consider the predicted biochemical properties of the Slr1923 protein with that of FRH. In the reaction of FRH, the ␣ and ␥ subunits of FRH receive two electrons from an unknown electron donor one by one, and transfer them to the FAD cofactor in the ␤ subunit one by one  (25). Then the ␤ subunit reduces the substrate by transferring two electrons at a time from FADH 2 . In contrast, we found that a two-electron donor, NADPH, is involved in the reaction of 8-vinyl reductase. Therefore, it is likely that two electrons are transferred to the FAD molecule that possibly resides in the Slr1923 protein, and then the electrons are transferred to the 8-vinyl group of the tetrapyrrole simultaneously. If this assumption is correct, the accessory proteins that play the same roles as the ␣ and ␥ subunits of FRH would not be necessary for the 8-vinyl reduction. Further studies are necessary to clarify the function of the Slr1923 gene product.

8-Vinyl Reductase for Chlorophyll Biosynthesis
Homologues of Slr1923-We found the homologues of Slr1923 in various organisms, including eukaryotes, cyanobacteria, and photosynthetic bacteria. Interestingly, a highly homologous gene exists in the Arabidopsis genome (At1G04620). The predicted amino acid sequence of the At1G04620 gene product shows 53% amino acid identity and 68% similarity with that of the Slr1923 gene product. Besides, At1G04620 is predicted to encode a transit peptide for chloroplast targeting. Based on these collective observations, it is tempting to speculate that At1G04620 is involved in the reduction of the 8-vinyl group in Arabidopsis. However, in the null mutant of the Arabidopsis dvr gene, all detectable monovinyl chlorophyll was replaced by 3,8-divinyl chlorophyll (26). This observation suggests that the DVR protein is the sole 8-vinyl reductase in Arabidopsis. There are two possible explanations for this discrepancy. One explanation is that the At1G04620 product could function as 8-vinyl reductase, but the contribution of this enzyme to the synthesis of monovinyl chlorophyll may be very limited. As a result, the dvr mutant predominantly accumulated 3,8-divinyl chlorophyll. Another explanation is that the At1G04620 locus does not encode 8-vinyl reductase in Arabidopsis despite the high level of sequence similarity to Slr1923 (see supplemental Fig. 3). At present, we cannot conclude whether the At1G04620 locus encodes a functional 8-vinyl reductase or not in Arabidopsis. Similarly, it is not clear whether the genes (or proteins) belonging to the higher plant clade encode 8-vinyl reductase (Fig. 8).
It is noteworthy that the Slr1923 homologue in C. merolae was categorized into the cyanobacterial clade (Fig. 8A). Because this organism does not contain a DVR homologue, and because the Slr1923 homologue of this organism is very similar to those of cyanobacteria (see supplemental Fig. 3), it would be reasonable to assume that the Slr1923 of C. merolae encodes a functional 8-vinyl reductase. The classification of the Slr1923 of C. merolae in the cyanobacterial clade is in contrast to that of the other eukaryotic homologues in a distinct higher plant clade (Fig. 8A). This result indicates that the Slr1923 homologue of C. merolae has a different evolutionary history from those of the other eukaryotic organisms. It is possible that the cyanobacterial genes for Slr1923-related 8-vinyl reductase are horizontally transferred from the genome of C. merolae or vice versa.
Because of a lack of functional experimental evidence, it is also unclear whether the Slr1923 homologues in photosynthetic bacteria encode functional 8-vinyl reductase. The function of DVR homologues is also not known for purple bacteria and diatoms. To discuss the evolutionary history of 8-vinyl reductase genes, we summarized the distribution of DVR and Slr1923 homologues in representative photosynthetic bacteria and cyanobacteria in Table 1. In this table, a homologue belonging to a phylogenetic clade in which at least one gene was shown to encode the functional enzyme is indicted with a ϩ. Homologues belonging to a clade in which functionality has not been experimentally confirmed is indicated as Ϯ. Finally, organisms which lack a homologue are denoted by Ϫ.
Two extreme hypotheses were developed to discuss the evolutionary history of 8-vinyl reductase genes. In the first hypothesis, we assumed that all DVR (BciA) and Slr1923 homologues indicated in Fig. 8 encode functional 8-vinyl reductase. In other words, homologues of all ϩ and Ϯ organisms encode the functional enzyme. This assumption led us to conclude that only Roseiflexus species lack both BciA and Slr1923 homologues among the organisms in which the 8-vinyl group is reduced. Thus, upon this assumption, we predict that there is an unknown type of 8-vinyl reductase in the group of Roseiflexus. Upon the same assumption, we can also conclude that in certain green sulfur bacteria including Chlorobium phaeobacteroides DSM266, a BciA homologue is replaced by an Slr1923 homologue. C. phaeobacteroides DSM266 may have acquired the Slr1923 homologue by a horizontal gene transfer from purple bacteria or cyanobacteria. Alternatively, we hypothesized that only DVR in the cyanobacterial and eukaryotic clades and Slr1923 in the cyanobacterial clade encode the functional 8-vinyl reductase. In this second hypothesis, we need additional 8-vinyl reductase genes to explain the distribution of the ability to reduce the 8-vinyl group. One possible candidate for the third 8-vinyl reductase is the BchJ-encoded protein, which was suggested to encode 8-vinyl reductase in Rhodobacter sphaeroides (27). However, it is probably not the case, because the bchJ disruptant of R. sphaeroides still produced certain amounts of 8-vinyl BChl a (27). In addition, Chew and Bryant (19) did not detect 8-vinyl reductase activity with the recombinant BchJ protein of Chlorobium tepidum. Therefore, we should assume a yet unknown protein for the second hypothesis. Similar to the first hypothesis, horizontal gene transfer could explain the absence of Slr1923 or BciA homologues in Roseiflexus species and in C. phaeobaceroides. Taken together, both hypotheses with opposite assumptions indicate the presence of additional 8-vinyl reductase genes and horizontal gene transfer.
The mosaic distribution pattern of 8-vinyl reductase genes among photosynthetic organisms is not similar to most of the currently identified genes that are involved in the chlorophyll biosynthesis (2). Among the genes for chlorophyll biosynthesis, those encoding glutamyl-tRNA reductase, glutamate-1-semialdehyde aminotransferase, 5-aminolevulinate synthase, coproporphyrinogen III oxidase, Mg-protoporphyrin IX monomethyl ester cyclase, and protochlorophyllide oxidoreductase are not ubiquitously found in photosynthetic organisms. The genes for 5-aminolevulinate synthase are found in some purple bacteria, whereas in other photosynthetic organisms, the genes for glutamyl-tRNA reductase and glutamate-1-semialdehyde aminotransferase are present. A rare exception is Euglena gracilis. This organism seems to contain all three enzymes (28), and this unique distribution may reflect the fact that its genome is a hybrid of photoautotrophic and heterotrophic genomes (29). In the case of coproporphyrinogen III oxidase, Mg-protoporphyrin IX monomethyl ester cyclase, and protochlorophyllide oxidoreductase, both aerobic and anaerobic types of enzymes exist in photosynthetic organisms. One or both types of enzymes are used in an organism to cope with either the aerobic or anaerobic environments that are necessary for survival. The distribution of these enzymes among photosynthetic organisms does not seem to contradict the evolutionary tree of organisms (30,31). Thus, it would be interesting to learn why the distribution of 8-vinyl reductase genes is so unique among the genes that encode the enzymes in chlorophyll biosynthesis. One possible explanation is that the DVR-type 8-vinyl reductase has just evolved recently probably in eukaryotic organisms, and then the gene is currently spreading among photosynthetic orga-nisms by dominating the pre-existing Slr1923 genes. This process might be traceable by evaluating the distribution of 8-vinyl reductase genes in cyanobacteria. The phylogenetic tree of Slr1923 homologues strongly suggests that an ancestor of cyanobacteria had the Slr1923-type 8-vinyl reductase (Fig. 8A). In lineages of Synechococcus species, the Slr1923 homologues might have been driven away by DVR homologues, if a DVRtype enzyme was superior to the Slr1923 type for unknown reasons. Collectively, the genes for 8-vinyl reductase may provide an interesting model to examine the process of horizontal transfer of genes among photosynthetic organisms.
Phenotype of the Cyanobacterial 8-Vinyl Reductase Mutant-The knock-out mutant for the Slr1923 locus accumulated a higher level of zeaxanthin. This carotenoid species protects cells from photoinhibition (32,33). The level of zeaxanthin increases by strong illumination (32). Therefore, it is possible that oxidative stress was generated in the slr1923 mutant cells even under normal light conditions.
We also observed a reduction in cell size and the number of thylakoid membranes. These phenomena may be partly due to the stress induced by the accumulation of 3,8-divinyl chlorophyll a. It is consistent with the observation that heat shock stress enlarged intra-thylakoid membrane spaces in Synechocystis sp. PCC 6803 cells (34). Stress conditions would affect the organization of thylakoid membrane.
Akimoto et al. (9) reported that the time-resolved fluorescence spectra were significantly changed in the dvr mutant of Arabidopsis. Moreover, the delayed fluorescence from the photosystem II was not observed in the dvr mutant (9). These data suggest that both the antenna system and the electron transfer system were affected in the mutant. Further analysis of mutants producing 3,8-divinyl chlorophyll may reveal why most photosynthetic organisms utilize monovinyl-type chlorophylls for their photosynthesis.
In summary, we identified the gene encoding the 8-vinyl reductase or its essential subunit in most cyanobacteria in this study. A phylogenetic analysis of this gene led us to hypothesize that a third enzyme involved in the 8-vinyl reductase exists in photosynthetic bacteria. It also illustrates the unique distribution patterns of the genes involved in the reduction of the 8-vinyl group. Further phylogenetic analyses of the DVR (BciA) and Slr1923 genes and functional analyses of the 8-vinyl reductase mutants may increase our understanding to the evolution of pigment biosynthesis in photosynthetic organisms.