Single-cell genomics unveiled a cryptic cyanobacterial lineage with a worldwide distribution hidden by a dinoflagellate host

Genome Sequence and Phylogenetic Analyses of Cyanobacteria Isolated from a Pelagic Dinoflagellate.

A single O. magnificus cell was obtained from seawater collected off Shimoda Bay, Japan, and cyanobacterial symbiont cells were isolated from a chamber of the host (SI Appendix, Fig. S2). The genomes of the isolated cyanobacterial cells were amplified and analyzed using two high-throughput sequencing platforms, Illumina MiSeq and Oxford Nanopore MinION. From scaffolds assembled using the short and long sequencing reads from the MiSeq and MinION systems, respectively, we successfully recovered ∼1.87 Mbp of sequence (SI Appendix, Table S1) representing the genome sequence of the cyanobacteria isolated from O. magnificus (hereafter referred to as OmCyn). The sequence showed particular similarity to genomes of the cyanobacterial lineage referred to as the marine picocyanobacterial clade, consisting of the genera Synechococcus and Prochlorococcus. This is consistent with the results of a previous study (24), which showed that the majority of prokaryotic 16S ribosomal RNA gene (16S rDNA) sequences amplified from marine dinoflagellates with cyanobacterial symbionts have high similarities to Synechococcus/Prochlorococcus species. Our phylogenetic analysis of 16S rDNA sequences confirmed the phylogenetic relationship between OmCyn and the marine picocyanobacteria (SI Appendix, Fig. S3). The OmCyn sequence made a monophyletic clade along with previously reported sequences. Most of those were a part of diverse 16S rDNA sequences directly amplified from dinoflagellate cells with cyanobacterial symbionts that are akin to O. magnificus (24). Biodiversity within the picocyanobacterial clade has been extensively studied, as these cyanobacterial species numerically dominate global oceans (7, 25). However, we detected no strong phylogenetic affinity between the clade of cyanobacterial sequences from dinoflagellates, including OmCyn, and previously explicitly described marine picocyanobacterial lineages (7) based on traditional phylogenetic markers for this group—namely, 16S rDNA (SI Appendix, Fig. S3) and the internal transcribed spacer (ITS; SI Appendix, Fig. S4). The two trees showed different phylogenetic positions of the clade comprising the cyanobacterial sequences from dinoflagellates: The ITS tree suggested the clade is basal to a well-defined Synechococcus clade, known as Synechococcus subcluster 5.1 (7), while the 16S rDNA tree placed the clade as sister to one of the clades within Synechococcus subcluster 5.1, named clade V. Nevertheless, since both backbones of the two trees did not receive high bootstrap support, we could not conclude on the phylogenetic position of OmCyn based on these single-gene phylogenetic analyses. To reveal the position of OmCyn within the phylogeny of marine picocyanobacteria, we carried out phylogenomic analyses using datasets consisting of 40 protein sequences (Fig. 2, Left, and SI Appendix, Fig. S5). The resultant trees clearly showed that OmCyn occupies a basal position to Synechococcus subcluster 5.1 in agreement with the ITS phylogeny (SI Appendix, Fig. S4). Synechococcus subcluster clade 5.1, together with OmCyn, was shown to be a sister to another monophyletic clade consisting of Prochlorococcus species; this topology was strongly supported with high bootstrap values (Fig. 2, Left, and SI Appendix, Fig. S5). Based on this phylogeny, the OmCyn lineage likely emerged soon after the split of the Prochlorococcus lineage and before the divergence of Synechococcus subcluster 5.1.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Maximum likelihood (ML) phylogenetic tree of a multiple-protein dataset and protein repertoire comparison between the marine picocyanobacteria and OmCyn. (Left) ML tree inferred from a concatenated dataset of 40 orthologous proteins. The dataset includes only taxa whose genomes have been completely sequenced. The ML tree for the full dataset comprising 50 taxa is shown in SI Appendix, Fig. S5. Thick branches indicate 100% bootstrap values based on ML nonparametric bootstrapping. (Right) Comparison of protein numbers in the nonredundant proteomes of marine picocyanobacteria and OmCyn. Each bar indicates a proteome size without functional redundancy (protein number in a species protein repertoire).

Metagenomic Analyses Reveal an Intimate Association of OmCyn with the Host Dinoflagellate and Its Worldwide Distribution.

Marine picocyanobacteria occupy a wide variety of ecological niches (7, 26). We attempted to determine the OmCyn distribution pattern using publicly available marine metagenomic data to reveal the habitat and behavior of the newly identified marine picocyanobacterial lineage in the natural environment. A read-mapping analysis using Tara Oceans short-read data (27⇓–29) against OmCyn and 19 picocyanobacteiral genome sequences revealed a marked difference in the occurrence pattern among sample size fractions between OmCyn and the remaining species examined. Reflecting their small cell size (<2 μm), the genome sequences of marine picocyanobacteria were most frequently (>70%; Fig. 3A) identified in samples from the 0.8–5-μm fraction relative to other fractions (5–20, 20–180, and 180–2,000 μm). In contrast, only 7% of OmCyn genome sequences were detected in samples from the 0.8–5-μm fraction, with 83% detected in the 20–180-μm fraction (Fig. 3A), inconsistent with the size of the symbiont cells (<5 μm) (30). This distribution pattern is clearly equivalent to that of the 18S rDNA of O. magnificus, which hosts the OmCyn lineage, based on Tara Oceans metabarcoding analysis data (Fig. 3A) (31). The majority of the 18S rDNA V9 amplicons corresponding to the O. magnificus 18S rDNA sequence were from the 20–180-μm fraction (Fig. 3A), with only 3% from the 0.8–5 μm fraction. This distribution pattern for the host is reasonable, considering that O. magnificus cells are 75–115 μm in size (32) and the low proportion of reads detected in the smaller fraction (0.8–5 μm) may have originated from free DNA in the water sample (e.g., DNA released from disrupted cells). OmCyn sequences were detected from the 20–180-μm samples from 51 out of 57 Tara Oceans geographical stations, suggesting its wide distribution across the world's oceans (Fig. 3B). Consistent with the comparison of sample size fractions, global distribution patterns of the OmCyn genome sequence and the O. magnificus 18S rDNA V9 amplicon were shown to be the most correlated with each other compared with free-living marine picocyanobacterial genomes (Fig. 3C). The result of our principal coordinate analysis (PCoA) on the distribution patterns also showed clear isolation of the OmCyn genome and the O. magnificus 18S rDNA V9 amplicon from the marine picocyanobacterial genomes (SI Appendix, Fig. S6A). Moreover, the sequence occurrence patterns for both the OmCyn genome and the O. magnificus rDNA amplicon among Tara Oceans metagenomic samples were predicted to be similarly associated with environmental physicochemical parameters (SI Appendix, Fig. S6B), supporting further the co-occurrence of OmCyn and O. magnificus.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Distribution patterns in natural environments estimated on the basis of metagenomic and metabarcoding data. (A) Relative abundances among four size fractions. While the marine picocyanobacterial lineages were abundant in the 0.8–5-μm fraction, OmCyn was estimated to be most abundant in the 20–180-μm fraction, matching the abundance pattern of O. magnificus, the host dinoflagellate. The relative abundances of cyanobacterial lineages, including OmCyn, and O. magnificus were calculated based on Tara Oceans metagenomic data and 18S rRNA V9 metabarcoding data, respectively. (B) OmCyn genome-sequence read abundance in the 20–180-μm fraction from 57 Tara Oceans stations. Circle sizes indicate abundance of detected reads per 5 million metagenomic reads. (C) Correlation matrix of distribution patterns among marine picocyanobacteria, OmCyn, and O. magnificus.

Given the strong correlation of sequence occurrence patterns for OmCyn and O. magnificus, along with reports that the symbiotic cyanobacteria of Ornithocercus species can be passed on to the symbiotic chambers of daughter cells through host cell division (12, 33), OmCyn cells are likely to exclusively reside within the symbiotic chambers of these dinoflagellates in the natural environment. This assumption could explain why the lineage including OmCyn was never explicitly described in previous surveys that extensively explored the genetic diversity of the marine picocyanobacterial clade. This may be attributable to the fact that studies focused on marine picocyanobacteria often use samples of small fraction size (i.e., filter size <20 μm) (34⇓⇓⇓–38) to increase the sensitivity of analyses by filtering out nontarget organisms with larger cell size; such procedures may lead to oversight of species attached to larger organisms. This implies that we may still be missing symbiotic cyanobacterial lineages other than OmCyn. Interestingly, our metagenomic read-mapping analyses detected a large number of metagenomic reads, which did not exactly match but showed high similarity to the OmCyn genome sequence from the eukaryote size fraction (20–180 μm; Fig. 4). Considering that Ornithocercus species other than O. magnificus and their relatives also have cyanobacterial symbionts (39, 40), and that some 16S rDNA sequences from dinoflagellate cells made a monophyletic clade with the OmCyn sequence (SI Appendix, Fig. S3), these metagenomic reads may partially represent the genomic diversity of cyanobacteria in symbiotic relationships with dinoflagellates.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Metagenomic-read recruitment plot of the OmCyn genome; 250 million Tara Oceans metagenomic reads from each of four filter size fractions (0.8–5, 5–20, 20–180, and 180–2,000 μm), totaling 1 billion metagenomic reads, were used for the mapping. Each marker in the left plot indicates reads with ≥100-bp alignment lengths. The histogram on the right shows the number of mapped reads. The asterisk indicates the group of reads that did not exactly match the OmCyn genome sequence but had high similarity (>90% identity). The color code represents filter size fractions from which each metagenomic read was obtained.

Genome Streamlining.

Another remarkable feature of the OmCyn genome is its streamlined protein-coding gene content. Given the high genome-sequencing coverage (mean, 436-fold), and the fact that the sequence contains tRNA genes for all 20 amino acids, the scaffolds likely represent the complete, or nearly complete, OmCyn genome. The genome sequence of OmCyn was predicted to contain 1,846 protein-coding genes, which could be clustered into 1,726 protein types by orthologous protein clustering (41). This protein repertoire of OmCyn (1,726 protein types; Fig. 2, Right, and SI Appendix, Table. S2) was significantly smaller (detected as an outlier by Grubbs’ test; α = 0.05) relative to the repertoires of other species within Synechococcus subcluster 5.1 (2,061–2,498 protein types; Fig. 2, Right). The size of the protein repertoire of OmCyn is comparable to those of Prochlorococcus species, which have the most reduced genomes among free-living cyanobacteria (Fig. 2, Right). The limited protein repertoire suggests that many protein-coding genes were lost during the evolution of the OmCyn lineage. To estimate the extent of OmCyn protein repertoire reduction, we assessed how the protein repertoires in each lineage have been altered from the ancestral state by comparing them with the putative protein repertoire of a common ancestor of the marine picocyanobacterial clade. The ancestral protein repertoire was defined as the protein set in common between total protein repertoires for each group of genomes included in Synechococcus subcluster 5.1 and the Prochlorococcus clade. This putative ancestral protein repertoire of marine picocyanobacteria consists of 2,114 proteins, and comparisons showed that, while the Synechococcus genomes have retained an average of 86% of ancestral proteins, the OmCyn genome possesses only 65% (Fig. 2, Right), supporting that a genome reduction event occurred in this lineage (see Dataset S1 for a list of proteins missing from OmCyn). The same assessment revealed that Prochlorococcus species with small genomes (Prochlorococcus other than P. marinus MIT 9313) have retained only 69% of ancestral proteins. While genome size and retention of putative ancestral proteins are apparently equivalent in OmCyn and Prochlorococcus species with reduced genomes, genome reduction, including gene losses, appears to have occurred independently in the two lineages based on their phylogenetic relationship (Fig. 2, Left, and SI Appendix, Fig. S5). In support of this notion, the OmCyn genome retains most genes encoding the components of phycobilisome, a large light-harvesting antennae complex, that were characteristically lost during Prochlorococcus evolution (26). To evaluate the impact of gene loss on metabolic function in OmCyn, we further analyzed the proteins that were predicted to be discarded from the OmCyn genome using the Kyoto Encyclopedia of Genes and Genomes (KEGG) orthology system (42). The analysis showed that the metabolic functional category that experienced the most severe reduction was membrane transport (SI Appendix, Fig. S7), which is responsible for interactions with extracellular environments. Similar trends in gene loss have been documented in the genomes of other bacterial symbionts (22, 43⇓–45), implying that the symbionts have been adapted to relatively stable environments (e.g., the symbiotic chamber of dinoflagellates).

In contrast to these gene losses, the cyanobacterial “core” protein set that is conserved among all complete genomes of free-living cyanobacteria analyzed in this study is mostly retained (96.9%; Fig. 2, Right) in the OmCyn genome. The high degree of universality of the core protein set appears to indicate the importance of these proteins for growth and function of cyanobacteria. The genome of OmCyn has likely been streamlined while retaining minimum metabolic integrity, including photosynthetic ability. A total of 20 cyanobacterial core proteins are missing in OmCyn (SI Appendix, Table S3); however, to our knowledge, none of these proteins are critical for growth or photosynthetic ability. It is noteworthy that the list of the missing cyanobacterial core proteins includes five genes for DNA recombination and repair: recF, recG, recN, recO, and recR. Since such loss of genes for genetic recombination has been observed in various bacterial symbionts with reduced genomes (44, 46⇓–48), it is likely to be a common phenomenon associated with reductive genome evolution. In contrast to OmCyn, which retains most of the cyanobacterial core protein set, intracellular cyanobacterial symbionts, which metabolically depend on their host cell for growth (16, 19, 21, 44), retain only 74–88% (SI Appendix, Table S4) of core proteins. Retention of the conserved protein set implies that OmCyn retains metabolic independence from its host cell, contrasting sharply with obligate intracellular cyanobacterial symbionts.

The genome streamlining of OmCyn cyanobacteria can be explained by the intimate association between the symbiont and its host; if the habitat of the cyanobacteria is limited to inside the O. magnificus symbiotic chamber, the effect of genetic drift should increase due to shrinkage of the effective population size (49). An increased effect of genetic drift would accelerate the loss of genes dispensable in the relatively mild and stable environment inside the symbiotic chamber.

The characteristics of OmCyn also shed light on another important question regarding the function of the symbiont within dinoflagellate cells. Some symbiotic cyanobacteria provide fixed nitrogen to their hosts (17, 21, 50); however, this cannot be the case for the relationship between OmCyn and O. magnificus since all species of the marine picocyanobacterial clade, including OmCyn, lack genes encoding proteins required for nitrogen fixation. As the OmCyn genome retains genes required for photosynthetic activity, and since Ornithocercus is a nonphotosynthetic dinoflagellate, a reasonable explanation for this symbiotic relationship would be that the symbiont provides a source of organic carbon for its host in open oceans, an oligotrophic habitat. It remains ambiguous how the host cell takes up nutrients from the symbionts, which reside in an extracellular space (i.e., the symbiotic chamber). An intriguing possibility that Ornithocercus species and their relatives ingest and digest their cyanobacterial symbionts has been raised, based on transmission electron microscopy observations (33, 51). In symbiont-bearing dinoflagellate cells, food vacuoles with inclusions resembling degraded cyanobacterial symbionts were observed (33). If true, given the intimate association between OmCyn and its host, cyanobacterial symbionts could be considered a “domesticated breed” farmed within the specialized chambers of the dinoflagellates.