Background

Cyanobacterial ancestors gave rise to plastids (chloroplasts) in the ancestor of a eukaryotic lineage. The birth of the plastid had an impact on eukaryotic genome evolution, by way of endosymbiotic gene transfer (EGT), a particular form of lateral gene transfer (LGT) from endosymbionts into the phylogenetically discontiguous host genome [1]. Subsequently, an algal ancestor gave rise to secondary plastids in several punctate lineages of eukaryotes. A number of these secondarily phototrophic lineages lost their photosynthetic ability and further diverged into secondarily heterotrophic, plastid-lacking protists [2, 3].

Although the position of the root of eukaryotes is still uncertain, the presence of gene fusions and insertion/deletion sequences in the marker genes have allowed us to sort eukaryotes into at least three large groups; Opisthokonta, Amoebozoa and bikonts (Archaeplastida, Chromalveolata, Rhizaria and Excavata) [4–10] (Figure 1). Most phototrophic eukaryotes harboring plastids derived from primary endosymbiosis (primary plastids) are classified into the super-group Archaeplastida (i.e. glaucophytes, green plants and red algae) [10]. Although it is widely accepted that primary plastids share a single origin [[11–13], but see [14, 15]] and the Archaeplastida are monophyletic [[3, 16], but see [17, 18]], the evolutionary history of the primary plastids is still debatable [19–21]. In plastid-lacking protists, 'plastid imprints' can be exemplified by genomic information, i.e. genes with affinity to extant cyanobacterial or algal genes. These genes were supposed to have originated from EGT events, and this assumption should be affirmed by the resulting phylogenetic relationship between 'imprint' genes and the extant relatives of the putative endosymbionts. The biggest challenge and the limitation of this 'imprint' searching process is that the inevitable incompleteness of genome information on lineages of interest and the ever-developing phylogenetic methodologies make it difficult to distinguish EGT and ancient LGT [22]. Thus, although available eukaryotic genome data are increasingly accumulating, gene and genome phylogenies should be carefully interpreted to infer evolutionary scenarios.

Figure 1
figure 1

A schematic representation of eukaryotic phylogeny. The current consensus phylogeny and rooting of eukaryotes based on previous studies [4, 23, 40]. Arrows and stars indicate plastid acquisition via endosymbiosis and alternative hypothetical time points which primary endosymbiosis occurred, respectively. The root of the eukaryotic tree on the unikonts/bikonts boundary is hypothesized, but still controversial [5, 7–9, 19]. Archaeplastida are represented as a monophyletic group, but see also [19].

Full size image

Chromalveolata is a large taxonomic group of eukaryotes, encompassing secondary phototrophs and secondarily heterotrophic protists [10], and the 'chromalveolate hypothesis' argues that this group originated from a common ancestor harboring the chlorophyll c-containing secondary plastid derived from a red alga (Figure 1) [23]. Among the secondarily heterotrophic chromalveolates, several lineages have retained remnant chloroplasts for non-photosynthetic metabolic pathways, e.g. apicoplasts in apicomplexan parasites [24]. Recent genomic surveys revealed the presence of plastid-derived genes, and further suggested the presence of cryptic secondary plastids in non-photosynthetic alveolate protists [25, 26]. Furthermore, re-examination of the whole genome sequences suggested the existence of algal genes in ciliates, another plastid-lacking alveolate lineage, which could support the photosynthetic ancestry of ciliates [27]. Oomycetes are plastid-lacking stramenopiles, or chromists, classified into Chromalveolata [10]. Although whole genome sequence analysis showed that a number of genes with affinity to photosynthetic organisms (cyanobacteria and algae) are encoded in the nuclear genome, most of these 'plastid imprints' candidates were only suggested by similarity search and phylogenetic analyses have not yet led to fully recovering the expected tree topology [28]. Considering the uncertain phylogenetic affinity of the 'best hit' in similarity search [29], reassessment of the genome information is important to determine whether the evolutionary history of oomycetes is comparable to ciliates [27].

One candidate of 'plastid imprints' in oomycetes has been confirmed by studies reporting the phylogeny of gnd genes, which encode 6-phosphogluconate dehydrogenase, showing that some plastid-lacking protists have plant-like, cyanobacterium-derived gnd genes [20, 21, 30]. These analyses suggested that the gnd genes with cyanobacterial ancestry were acquired early in eukaryotic evolution, either via ancient eukaryote-to-eukaryote LGT, or primary EGT that occurred earlier than had ever been thought [21]. Additionally, the phylogeny of gnd genes demonstrated that cyanobacterial genes are also present in several Excavata protists, e.g. the heterolobosean amoebo-flagellate Naegleria gruberi. Naegleria gruberi is a non-parasitic heterotrophic species related to N. fowleri, which is the causative agent of primary amoebic meningoencephalitis in mammals [31]. Although the phylogenetic relationship within Excavata is still unclear, Heterolobosea, together with Jakobida, is likely to be a sister group of Euglenozoa [18, 32].

To address how many genes have cyanobacterial ancestry in plastid-lacking protists, and whether cyanobacterial ancestry is limited to this gnd gene or also found in other genes, we conducted a phylogenomic analysis using genome sequence data of a taxonomically wide range of plastid-lacking eukaryotes. Here we present a gene mining study with a novel pipeline automatically producing and summarizing one-by-one phylogenetic trees, and show phylogenetic analyses of resultant candidate genes with cyanobacterial ancestry, using the whole genome sequence data from a wide range of eukaryotic lineages.

Results

To address how many genes are derived from cyanobacteria in non-photosynthetic protists, we conducted cyanobacterial gene mining using the genome sequence data of a wide range of the plastid-lacking eukaryotes (Additional file 1). Using the whole genome data, we conducted BLAST searches against all 'Bacteria' and selected queries showing the highest similarity to genes in the available cyanobacteria genome sequences. We then drew the neighbor-joining (NJ) trees for genes showing homology to cyanobacterial counterparts. After the first tree construction step, we selected the gene trees where cyanobacteria and eukaryotes formed a monophyletic group excluding other prokaryotes. As a result, we obtained a shorter list of candidates, which we termed 'genes with cyanobacterial affinity'. Subsequently we re-analyzed the eukaryotic genes with cyanobacterial affinity by visually checking and re-drawing the Bayesian and maximum likelihood (ML) trees after manually trimming operational taxonomic units (OTUs). In total, we identified 12 plastid-lacking protist genes 'with cyanobacterial ancestry' in the genomes of the wide range of eukaryotes: two plastid-lacking bikonts (the oomycete P. ramorum and the heterolobosean N. gruberi) and three unikonts (the slime mold D. discoideum, the budding yeast S. cerevisiae and the choanoflagellate M. brevicollis) (Table 1). These were the eukaryotic genes with cyanobacterial ancestry that shared the same origin with Archaeplastida and other eukaryotes. They were placed within a monophyletic subclade mostly composed of photosynthetic organisms (cyanobacteria and plants/algae) and showed an apparent cyanobacterial ancestry as far as was determined by tree topology (Table 1; Figures 2, 3, 4 and 5; and Additional files 2, 3, 4, 5, 6, 7, 8 and 9). We found another type of gene with cyanobacterial ancestry, which were the protist genes forming monophyletic groups mostly with genes from extant cyanobacteria (prokaryote-type genes with cyanobacterial ancestry). Among a number of candidate genes found through the first screening, we have presented three typical trees that were resolved with significant support values (Additional files 10, 11 and 12). We postulate that these prokaryote-type genes are remnants of the bacterium-to-eukaryote LGT, which occurred 'recently' in evolution. Interestingly, while Phytophthora ycf21 homologs, probably transferred from a relative of the extant cyanobacterial species via LGT, were placed within the cyanobacterial gene clade, the ciliate Tetrahymena ycf21 homolog showed affinity to Archaeplastida (Additional file 11). This gene was neither found in the Paramecium genome nor in the list of the recently identified algal genes in ciliates [27].

Figure 2
figure 2

Uroporphyrin III methyltransferase gene phylogeny showing the presence of genes with cyanobacterial ancestry in oomycetes. The MrBayes consensus tree with Bayesian posterior probabilities (BI) (70% or more) and maximum likelihood (ML) bootstrap support values (50% or more) is shown. Thick branches represent BI and ML values not lower than 100 and 95, respectively. Different phylogenetic affiliations are represented as follows: green, green plants; magenta, red algae; blue-green, glaucophytes; orange, Chromalveolata; dark blue, Excavata; yellow, Rhizaria; gray, unikonts; sky blue, cyanobacteria. Stars indicate plastid-lacking eukaryotes.

Full size image
Figure 3
figure 3

Cobalamin-independent methionine synthase genes in oomycetes are monophyletic with algal and cyanobacterial homologs. See legend for figure 2.

Full size image
Figure 4
figure 4

Amino acid decarboxylase genes in Heterolobosea, within a subfamily with cyanobacterial ancestry. See legend for figure 2.

Full size image
Figure 5
figure 5

Naegleria genes are a member of the multiple gene families of TIC55-like oxidoreductase genes. See legend for figure 2.

Full size image
Table 1 Summary of eukaryote-type genes with cyanobacterial ancestry identified in this study
Full size table

We found that Uroporphyrin III methyltransferase gene homologs (Figure 2) consisted of two large subfamilies of genes with cyanobacterial ancestry, and that oomycete genes were included only in one of them. Given that both subfamilies include green plants, red algae, chromalveolates and cyanobacteria, it is likely that they diverged within the ancestral cyanobacteria and transferred into eukaryotic hosts via primary and secondary endosymbioses. Both of the subfamilies were concurrently present in the cyanobacterial and green algal genomes. In land plants, red algae, diatoms, haptophytes and the plastid-lacking oomycetes, one of the subfamilies might be lost along with the loss of the plastid. The Thalassiosira homolog formed a monophyletic group with green plants, rather than red algae, suggesting that it was acquired independently of the secondary plastid of the red lineage. In this study, the bacteriovorous choanoflagellate Monosiga brevicollis gene and the proteobacterial genes (Gluconobacter, Alteromonas and Nitrosomonas) were treated as 'apparently LGT-derived genes', incongruously showing affinities to photosynthetic bikonts [33].

Genes encoding cobalamin-independent methionine synthase in green and red algae, diatoms, and oomycetes formed a monophyletic group with cyanobacterial homologs, while the land plants and the red alga Cyanidioschyzon homologs were placed in different clades unrelated to cyanobacteria (Figure 3). Close association between diatom and oomycete genes suggested the deep ancestry of the genes in the chromalveolate lineage. We failed to find the homologs in the prasinophytes Ostreococcus and Micromonas, suggesting that this gene family was dispensable in some plant lineages.

One of the genes with cyanobacterial ancestry found in N. gruberi is pyridoxal-dependent amino acid decarboxylase gene (Figure 4). The tree indicated that green plants were split into different eukaryotic clades. Naegleria and chromalveolate genes showed robust monophyly with green plants, included in a cyanobacterial gene clade. The tree showed that land plants possessed another subfamily, associated with red algal and fungal genes, apparently of non-cyanobacterial origin. We also identified genes with cyanobacterial ancestry from Naegleria in an oxidoreductase gene family that included genes encoding Rieske iron-sulfur cluster 55 kDa protein of chloroplast inner membrane translocon (TIC55), chlorophyll a oxidase (CAO), Lethal-leaf spot 1 (LLS1, which is synonymous with pheophorbide a oxygenase (PAO)) and accelerated cell death 1 (ACD1) (Figure 5) [34, 35]. All the members of this family in land plants were hypothesized to be located at the inner membrane of the chloroplast, and to be involved in chlorophyll metabolism [34]. The phylogenetic tree of the TIC55-like gene family showed intricate distribution of cyanobacterial, green plant and chromalveolate genes.

In other trees of the genes identified in this study (Additional files 2, 3, 4, 5, 6, 7, 8 and 9), gene clades with cyanobacterial ancestry were mostly composed of bikonts genes, besides the choanoflagellate M. brevicollis genes (see Discussion).

Discussion

We identified eight and seven genes with cyanobacterial ancestry in the genome sequences of the oomycete P. ramorum and the heterolobosean N. gruberi, respectively (Table 1). It was reported that the apicomplexan Cryptosporidium 'recently' lost their secondary plastid, and retained two to seven putative plastid-derived genes in the genome [36]. This number is comparable to our result of the gene mining study using oomycete and heterolobosean genomes. In addition, our system resolved the hidden diversity of the gene family repertoire in eukaryotic genomes by one-by-one gene phylogenies.

Secondary EGT scenario

Although the phylogenetic positions of Cryptophyceae and Haptophyta are still debatable [e.g. [17, 37–41]], the chromalveolate hypothesis has been reinstated to support the evolutionary scenario that the plastid-lacking protists oomycetes and ciliates once might have had a plastid [27]. According to this hypothesis, the genes with cyanobacterial ancestry found in the oomycete genomes were acquired via secondary EGT in the common ancestor of Chromalveolata, from the red algal ancestor of secondary plastids. This explanation is also applicable under the alternative hypothesis for chromalveolate plastids, which proposes that a tertiary endosymbiont of the haptophyte/cryptophyceae lineage is the origin of the stramenopile/alveolate plastids [22]. The phylogenetic tree of the photorespiratory glycerate kinase genes, suggesting the red algal origin of the Phytophthora genes (Additional file 7), is consistent with the chromalveolate hypothesis. However, several other gene trees in this study showed oomycete genes with green lineage affinity, not red algae (e.g. Additional files 2, 3 &4). Recently, Frommolt et al. [42] demonstrated that, out of 16 genes involved in carotenoid biosynthesis from chromalveolate algae, one third (5/16) of plastid-targeted, nuclear-encoded genes are most closely related to green algal homologs. Reyes-Prieto, Moustafa and Bhattacharya [27] identified 16 genes of possible algal origin in the ciliates Tetrahymena thermophila and Paramecium tetraurelia, and 7/16 of their trees show a close relationship between green plants and Chromalveolata. Frommolt et al. [42] attributed the close relationships between green plants and chromalveolate genes to the secondary endosymbiosis of an ancient green plant (e.g. prasinophyte), based on the hypothesis on the monophyly of the Archaeplastida [16, 40]. This explanation might be also applicable to the plant-like genes in ciliates [27].

While Heterolobosea and Euglenozoa are often united as the morphologically defined taxon, Discicristata, within Excavata [10], recent morphological and molecular phylogenetic analyses suggest that the heteroloboseans (e.g. Naegleria) never possessed the secondary plastid of green lineage and share the same origin with Euglenida [43]. Molecular phylogenetic analyses showed that Excavata is separated from other secondary plastid-containing eukaryotes (Chromalveolata and Rhizaria) [18, 40]. Therefore, it is unlikely that the genes with cyanobacterial ancestry found in the heterolobosean nuclear genomes originated from the plastid cognate with any known secondary plastids in extant photosynthetic eukaryotes. The amino acid decarboxylase gene (Figure 4) and the gnd gene (Additional file 3) [21] trees demonstrated the presence of genes with cyanobacterial ancestry in other heterolobosean species than N. gruberi, suggesting that the ancestor of the genus Naegleria possessed this gene family. Furthermore, although ML bootstrap support or Bayesian posterior probability (BI) values were not always sufficient, the Naegleria genes occupy relatively basal phylogenetic positions within the bikonts clade in all seven trees (Figures 4 and 5; Additional files 3, 4, 6, 8, and 9). Thus it is possible that the genes with cyanobacterial ancestry were introduced en bloc in the ancestor of Heterolobosea, via a batch gene transfer, in a concerted manner. One possible origin of such a concerted gene transfer is secondary EGT from a photosynthetic eukaryote with a basal phylogenetic position within bikonts. However, as discussed above, it is unlikely that Heterolobosea experienced secondary endosymbiosis and acquired genes common to the extant secondary plastid-containing eukaryotes via secondary EGT.

Ancient eukaryote-to-eukaryote LGT or primary EGT scenarios

Alternatively, we can argue for two other explanations: a concerted eukaryote-to-eukaryote LGT scenario or a more ancient primary EGT scenario. The Naegleria genes with cyanobacterial ancestry shown in Table 1 are basally positioned within bikonts, but not intruding into any of gene clades from extant photosynthetic eukaryotes (Figures 4 and 5; Additional files 3, 4, 6, 8, and 9). Thus, if we assume that these genes were acquired via non-endosymbiotic LGT, they may originate from unknown ancient photosynthetic lineages basally positioned within bikonts. Meanwhile, under the primary EGT scenario, in which the primary endosymbiosis occurred in the common ancestor of bikonts (Figure 1) [[19–21], but see Ref. [9] for further discussion on the root of eukaryotic tree of life], ancient primary EGT occurred much earlier than the conventional hypothesis, from the cyanobacterium-like prokaryote to the common ancestor of bikonts. Primary plastids were subsequently lost in many lineages of bikonts, except for the Archaeplastida lineages, but some genes originating from the cyanobacterial ancestor of the primary plastids have been retained in the nuclear genomes of the plastid-lacking lineages of bikonts (Figures 1 and 6). The loss of the plastid might have triggered the loss of genes that specifically functioned within the plastid. Only a portion of the plastid-derived genes, which we can find now in the plastid-lacking protist genomes, might have escaped from or survived through eliminative pressure in a lineage-specific manner, by acquiring additional functions with other components and/or in other cellular compartments. This might account for the observed punctate distribution of gene families among the eukaryotes [44, 45].

Figure 6
figure 6

An evolutionary history of the genes with cyanobacterial ancestry. Thick continuous arrows represent gene flow via EGT. Thin broken arrows indicate gene expression or intracellular transport into organelles. Dashed line circles and boxes indicate that they have been lost in the evolutionary history. Note that the genes with cyanobacterial ancestry (white), which had been derived from the plastid genome via EGT, were retargeted into the plastid. After rounds of gene family duplication, some genes (magenta) gained additional functions in other cellular compartments (cytosol, mitochondrion, etc.). In some plastid-lacking protists, a number of genes were retained in the nuclear genomes after the plastid loss events. Mt, mitochondrion.

Full size image

Recently, a hypothesis for the non-monophyly of Archaeplastida was proposed based on the phylogenetic analyses of slowly evolving nuclear-encoded genes [17, 19]. This non-monophyly hypothesis could be also considered within the scope of the primary EGT scenario. It is notable that a number of the trees in this study (Figure 2; Additional files 2, 3, 4, 6, and 8) showed intriguing topologies, depicting the split of Archaeplastida and inclusion of Chromalveolata and Excavata genes within it, as shown in the previously reported multiple slowly-evolving gene phylogeny [19] and gnd gene phylogeny [20, 21]. These results are consistent with the hypothesis for the non-monophyly of Archaeplastida, and suggest that the oomycete and heterolobosean genes with cyanobacterial ancestry might reflect the host nuclear genome phylogeny. On the other hand, the genes found in the marine choanoflagellate M. brevicollis were positioned within the bikonts clade, but not associated with the genes from other Opisthokonta relatives (Metazoa and fungi), suggesting that the tree topologies were probably not reflective of the host phylogeny [46] but eukaryote-to-eukaryote LGT (Figure 2; Additional files 4, 8, and 9). No gene with cyanobacterial ancestry was found in D. discoideum (Amoebozoa), and only one gene in S. cerevisiae (Opisthokonta). These results are also consistent with the ancient primary EGT scenario.

A photorespiratory gene with cyanobacterial ancestry in fungi

Our analysis using the genome data of the budding yeast S. cerevisiae identified one gene with cyanobacterial ancestry, encoding the glycerate kinase for photorespiration (Additional file 7). Given that photorespiration is essential for cyanobacteria and plants, it is likely that the glycerate kinases in plants and cyanobacteria are phylogenetically and physiologically related to photorespiration [47, 48]. A previous study on glycerate kinases showed that, regardless of the complete absence of photorespiratory metabolism in fungi, the gene product from the budding yeast Saccharomyces showed similar enzymatic activity and substrate specificity compared with the Arabidopsis gene, suggesting that the plant and fungal genes catalyze the same reaction in different contexts of the metabolic pathway [47]. Another example of plant-type genes in fungi was reported in a phylogenetic study of the genes encoding high-affinity nitrate transporter NRT2, which suggested that fungi probably acquired the NRT2 genes via LGT from one of the chromalveolate lineages [49]. Meanwhile, our data showed that the fungal clade was located outside the clade of plants plus oomycetes (Additional file 7), suggesting that fungal glycerate kinase genes with cyanobacterial ancestry likely originated from an LGT event from an ancestor of cyanobacteria, or eukaryote-to-eukaryote LGT from an ancestor of Archaeplastida (or bikonts). One likely explanation for the presence of photorespiratory genes in oomycetes is that the ancestor of Chromalveolata possessed this gene family, but some photosynthetic descendants lost this gene family or replaced it with other genes during the course of lineage-dependent customization of photorespiratory pathways [[50, 51]; for discussion on carbon assimilation in diatoms], while oomycetes retained the genes without any replacement.

Gene family expansion and differential reduction

Another conclusion of this analysis is that rounds of gene family expansion and selective reduction are important factors in making eukaryotic genome phylogeny look like a complicated mosaic (Figure 6). It is likely that the alteration of gene family repertoire contributed to the restructuring of the intracellular metabolome and a reduction of the dispensable gene families. Our data showed that all the genes identified in this study were members of multiple gene families. Algae and plastid-lacking protists retained only members of subfamilies (e.g. Figure 2 and Additional file 8), suggesting that the punctate distribution might be a corollary of the common mechanism by which genes with cyanobacterial ancestry were retained in their genomes. The presence of genes from multiple subfamilies in one organism supports this idea (e.g. two Uroporphyrin III methyltransferase subfamilies in prasinophytes and Volvox in Figure 2). Discontinuous loss or gain of a metabolic pathway in a lineage might be another factor in punctate distribution; e.g. the oxidative pentose pathway, and the cyanobacterial gnd genes functioning therein, were present in most bikonts but lost in the ciliate Tetrahymena [21, 52]. A recent study on pyridoxal-dependent amino acid aminotransferase reported that, besides the ancestrally eukaryotic enzymes, land plants possess a distinct subfamily of prokaryote-type chloroplast-targeted enzymes [53]. Our data with richer taxon sampling identified another prokaryote-type subfamily with cyanobacterial ancestry (Additional file 8), illustrating the hidden evolutionary diversity of protist and algal metabolomes.

Future prospects

Our results showed that many genes with cyanobacterial ancestry identified in this study were found only in complete genome sequences, suggesting that these genes might be difficult to discover by expressed sequence tag (EST) library sequencing, probably due to the low-level expression of these genes. Although the whole genome data from excavate parasites (e.g. Trypanosoma, Giardia and Trichomonas) are available, they seem to be unsuited for the gene mining study because of the unusual nucleotide substitutions (see Methods). At the stage of starting the present gene mining study, N. gruberi was the only species with whole genome data released within the non-parasitic excavates, and thereby the excavate genes with cyanobacterial ancestry were mostly from N. gruberi. More genome data from plastid-lacking protists from Excavata and Rhizaria as well as Archaeplastida, especially red algae and glaucophytes, are needed to unravel the evolutionary history of plastids, and plastid-lacking protists.

Conclusion

The comparative analyses of the genome sequence data of the plastid-lacking eukaryotes demonstrated the potentially significant contributions of ancestral or extant cyanobacteria to the eukaryotic genomes, which probably occurred via LGT or ancient primary EGT events. Furthermore, the automated phylogenetic analyses revealed the diversity and punctate distribution of gene families within the genomes in the unicellular microbes. More genome data of the plastid-lacking Excavata and Rhizaria will make the evolutionary history clear and support our hypotheses.

Methods

Data preparation

The genome sequence data of P. ramorum, N. gruberi and M. brevicollis was produced by the US Department of Energy Joint Genome Institute (JGI) [54]. D. discoideum genome data (9 Nov 2007) at dictyBase [55] and S. cerevisiae genome data [56] were used for phylogenetic analysis. Red algal data were retrieved from the Cyanidioschyzon merolae [57], Galdieria sulphuraria [58] genome databases, and other algal data were from Aureococcus anophagefferens, Emiliania huxleyi, Micromonas pusilla, Micromonas sp. RCC299, Ostreococcus tauri, Ostreococcus sp. RCC809, Phaeodactylum tricornutum, Phytophthora sojae, Thalassiosira pseudonana and Volvox carteri genome databases on JGI. EST sequences of several protists were obtained from TBestDB [59] and all other sequences were from the NCBI GenBank refseq database [60]. We excluded amitochondrial and/or parasitic eukaryotes, which might cause long branch attraction due to unusual nucleotide substitutions [61, 62]. Fragments of N. fowleri amino acid decarboxylase gene [DDBJ: AB491948] were amplified from genomic DNA using degenerated primers based on the conserved amino acid motif YHHFGYP for the forward primer (TAYCAYCAYTTIGGITAYCC) and WQLACEG for the reverse primer (CCYTCRCAIGCIARYTGCCA). PCR products were directly sequenced using an ABI PRISM 3100 Genetic Analyzer (Applied Biosystems, Foster City, CA, USA) with a BigDye Terminator Cycle Sequencing Ready Reaction kit v. 3.1 (Applied Biosystems).

Phylogenomic analysis

A genome-wide phylogenetic program was made with several bench-made BioRuby scripts (Additional file 1), referring to the previously reported phylogenomic pipeline used in the macronuclear genome analysis of Tetrahymena thermophila [63]. For the first screening, query amino acid sequences were automatically subjected to BLAST searching using NCBI netblast [64] and EFetch utilities [65], extracting the genes showing the highest E-value to a cyanobacterial counterpart among 'Bacteria' by BLASTP. For the second step, these genes were subjected to BLASTP analysis against 'refseq-protein' to fetch homologous sequences with E-values less than 0.001, up to 500 hits at a maximum. Multiple alignments were then performed using MUSCLE [66], which automatically removed ambiguously aligned sites or sequences with too many gaps. Bootstrapped neighbor-joining trees were produced using QuickTree [67]. Trees were output in the PostScript format using the newicktops program in the NJplot package [68] with sizes and colors of OTU names modified according to the NCBI taxonomy database [69] to simplify the subsequent visual checking process. Genomes of several bacterial genera were intensively sequenced and many homologous sequences from closely related species and strains (e.g. Escherichia, Bacillus) appeared on the trees. To diminish the sampling bias, the output files of QuickTree were also used to parse tree topology and detect a monophyletic clade exclusively composed of OTUs from a single genus using Bio::Tree class methods in BioRuby scripts. One representative OTU was automatically selected in such single-genus clades, the other OTUs were removed, and the trees were re-constructed for visual checking. In addition to the automatic process, trees for genes listed in the putative photosynthetic endosymbiont-derived genes [28], but not detected in our analysis, were manually re-constructed. Non-cyanobacterial prokaryotic genes taxonomically unrelated to, but placed within, the cyanobacterial clade were interpreted as 'apparently LGT-derived genes' with cyanobacterial ancestry.

Candidate cyanobacteria-related genes were manually selected, their homologs were collected from major groups of the three domains of life, and then subjected to multiple protein sequence alignments using MUSCLE. Phylogenetic analyses were performed with a maximum likelihood (ML) method using RAxML [70] and with a Bayesian interference (BI) method using MrBayes [71]. ML and BI were based on the WAG substitution matrix with options of four gamma-distributed rate categories and estimate of invariable sites (plus empirical base frequencies in ML). ML branch support was evaluated with 1000 bootstrap replicates, and BI posterior probability values were calculated from the MCMC run data, which summarized when the average standard deviation of split frequencies reached less than 0.01. Except for cyanobacterial genes of which no homologs were found in other prokaryotes (e.g. Additional file 2), or of which monophyly was confirmed by previous studies (e.g. Additional file 3), threshold values to assess the monophyly of cyanobacterial gene clades were 50% on ML bootstrap or 0.9 on BI posterior probability values.