From Green to Red: Horizontal Gene Transfer of the Phycoerythrin Gene Cluster between Planktothrix Strains

Eight Planktothrix strains isolated from lakes Steinsfjorden and Kolbotnvatnet (Norway) have been investigated. All strains have been kept in continuous, nonaxenic cultures at the Algal Culture Collection of the Norwegian Institute for Water Research (NIVA), in Z8 medium and light at a photon flux density of 10 μmol m⁻² s⁻¹, and a light-dark cycle of 12:12 h. Prior to genomic DNA isolation, cells were centrifuged and resuspended in Tris-EDTA (TE) buffer. DNA was extracted by the following procedure: cells were treated with lysozyme (final concentration, 15 mg/ml), followed by RNase A and proteinase K (5 mg/ml, 1% lithium dodecyl sulfate [LDS]) treatment. Samples were then incubated at 60°C (shaking at 300 rpm) for 60 min. In cases where the solution did not clear after 60 min, the incubation time was prolonged with additional proteinase K. Subsequently, 1 volume phenol-chloroform-isoamyl alcohol (25:24:1, vol/vol) was added. The solution was mixed by inversion at 37°C for 30 min to remove pigments and proteins. After centrifugation, the upper layer was treated twice with 1 volume chloroform-isoamyl alcohol (24:1, vol/vol). DNA was precipitated using 0.1 volume 3 M sodium acetate and 2.5 volumes ice-cold 96% ethanol on ice for 1 h. The DNA pellet was washed twice with ice-cold 70% ethanol, dried at room temperature, and dissolved in Tris-HCl buffer (pH 8.0). All DNA samples were purified using Amicon Ultra-0.5 ml 50K centrifugal columns to ensure high-quality DNA for paired-end library preparation using the 454 Life Sciences protocols. DNA was concentrated according to the manufacturer's instructions and washed twice using Tris-HCl buffer, pH 8.0.

Seven out of eight Planktothrix genomes were sequenced using 454 pyrosequencing at the Norwegian Sequencing Centre (http://www.sequencing.uio.no/). Strain NIVA-CYA 34 was initially sequenced using Sanger sequencing and 454 pyrosequencing at the Max Planck Institute for Chemical Ecology. For Sanger sequencing, the DNA sample of NIVA-CYA 34 was amplified using the REPLI-g kit (Qiagen). The resulting DNA was randomly sheared, and the fragment size range from 2,500 to 3,000 kb was selected for cloning into pUC18 standard vectors. The resulting clones were sequenced from both ends with the standard sequencing primers on an ABI 3700 machine generating 20,066 sequencing reads comprising 16 Mb. An additional 454 shotgun library was prepared and sequenced at the Norwegian Sequencing Centre to ensure satisfactory quality of the NIVA-CYA 34 genome assembly. For the remaining seven strains, both shotgun and paired-end libraries were prepared and sequenced to 23- to 41-fold coverage (see Table S1 in the supplemental material for information about libraries and number of reads/bases sequenced for each strain).

The newbler program (gsAssembler; Roche-454, Branford, CT, USA) was used to assemble the 454 reads into scaffolds and contigs (the newbler version used for assembly of each strain together with assembly statistics is shown in Table S1 in the supplemental material). Since all strains had been cultivated nonaxenically, reads from cocultured bacteria were present in the data set. To decrease the chance of coassembling these contaminating reads with the genome of the strain of interest, stringent overlap settings with a minimum of 98% overlap identity and a minimum of 60-bp overlap length were used. However, the assemblies of most strains still contained some short scaffolds with low average read depth (below 5- to 10-fold) and/or GC percentages that diverged from high-read-depth scaffolds. These scaffolds were considered to be derived from cocultured bacteria present in the sample used for DNA extraction (42). The low-read-depth scaffolds were compared to the nonredundant NCBI protein database using BLASTX, and all noncyanobacterial matching scaffolds were removed before annotation.

Assembly of the NIVA-CYA 34 genome was done using both the Sanger and 454 reads. Newbler was given a trimming file (-vt option) to remove pUC18 plasmid vector sequences from the Sanger reads. The “-stopjoin” and “-repfill” options were used for assembly. One low-coverage noncyanobacterial scaffold was removed from the newbler assembly. Four of the unscaffolded contigs were identified as containing cyanobacterial sequences; these were added to the scaffolds before annotation.

According to classification of Genome Sequence Standards (43), we consider the assemblies to be improved high-quality drafts. Annotation of all genome sequences was performed using the U.S. Department of Energy (DOE) Joint Genome Institute (JGI) integrated microbial genomes database and comparative analysis system (IMG) (44).

The IMG system was used for pairwise comparison of genomes and calculation of Pearson correlation coefficients of COG (Cluster of Orthologous Groups) profiles. Homologous genes were defined as genes having a minimum of 80% sequence identity and identified using BLASTP.

Hierarchical clustering by COG profiles was performed by the IMG system using the tool Cluster (http://www.falw.vu/∼huik/cluster.htm). For construction of the pairwise hierarchical tree, a function profile vector (gene counts per COG) was generated for each genome. The distance metrics between these profile vectors were calculated by means of uncentered correlation. Pairwise single-linkage clustering was performed, first grouping the two closest profile vectors to form a group and then grouping the next pair of closest groups or vectors until all genomes were incorporated into the hierarchical tree.

Phylogenetic analyses of genomes were performed using a set of core genes. Core genes were defined as genes present in all genomes and having a minimum of 90% sequence identity, resulting in a data set of 3,690 genes. The final data set (after removing 100% identical genes in all genomes together with transposases, oligopeptide synthetases, retron-type reverse transcriptases, phage-associated proteins, and proteins shorter than 50 amino acids [aa]) contained 2,914 genes. Ten subsets of genes for generating phylogenies were created by random sampling of 20 genes from the core gene set (2,914 genes) repeated 10 times using R (R Development Core Team [http://www.R-project.org]). Sequences were aligned using MEGA 5 software (45) and concatenated for each of the 10 data sets. Lists of genes for each data set are shown in Table S2 in the supplemental material. ML analyses were carried out using RAxML (46) with 1,000 bootstrapped resamplings with the GTRCAT model. The resulting individual phylogenetic trees are shown in Fig. S1 in the supplemental material. Finally, the aligned sequences of all 200 genes were concatenated, resulting in a 217,546-bp-long alignment (available through https://doi.org/10.6084/m9.figshare.719100), and ML analyses were carried out using RAxML (46) with 1,000 bootstrapped resamplings with the GTRCAT model.

All gene sequences analyzed (phycoerythrin, phycocyanin, and flanking genes) were downloaded from IMG. Sequences were aligned using MEGA 5 software (45). Nucleotide diversity was analyzed using the computer program DNA Sequence Polymorphism (DnaSP) (47). Maximum likelihood analyses of cpc genes were carried out using RAxML (46) with 1,000 bootstrapped resamplings. GTRCAT was determined to be the best evolutionary model using ModelTest (48).

Recombination events were detected by visual analyses of informative sites (variable sites where each variant occurs in at least two sequences) as described by Rudi and coworkers (7). In order to detect the recombination breakpoints, concatenated sequences of cpcBA (reverse complement) and CHAP domain genes were analyzed using the RDP 4 software (49). Sequences from strains NIVA-CYA 406 and NIVA-CYA 15 were discarded as these are identical with sequences from strains NIVA-CYA 98 and NIVA-CYA 34, respectively. Recombination signals were accepted if at least three different methods detected statistically significant (P < 0.05) evidence of recombination.

All Planktothrix genomes have been deposited as Whole Genome Shotgun projects at DDBJ/EMBL/GenBank under the accession numbers AVFS00000000 to AVFZ00000000 (see Table 1 for details). The versions described in this paper are versions AVFS01000000 to AVFZ01000000. The 454 sequencing data are available from the NCBI Sequence Read Archive (NIVA-CYA 98, SRP028838; NIVA-CYA 406, SRP028560; NIVA-CYA 407, SRP028561; NIVA-CYA 56/3, SRP028564; NIVA-CYA 34, SRP028840; NIVA-CYA 405, SRP028540; NIVA-CYA 540, SRP028566; NIVA-CYA 15, SRP028535). Annotated genomes are publicly available at IMG (https://img.jgi.doe.gov/cgi-bin/er/main.cgi).

Genomes of eight Planktothrix strains were sequenced: four red and four green strains, including a red (NIVA-CYA 540) and two green (NIVA-CYA 34 and NIVA-CYA 405) strains associated with Cht 9/10 (Table 1).

All eight genomes were of similar size, ranging from 5.49 Mb to 5.71 Mb, with a GC content of 39.5% (Table 1). The IMG annotation system predicted between 4,790 and 5,088 protein-coding genes in each genome (Table 1). Of the predicted protein-coding genes, 65 to 67% could be annotated using the COG functional annotation database, and 19 to 20% were annotated to a KEGG pathway (Table 1).

Analyses of gene content (both COG-annotated and all protein-encoding genes) and phylogenetic analyses were performed for comparison of eight sequenced genomes. Nucleotide diversity analysis of 200 randomly selected core genes revealed high sequence similarity between eight compared Planktothrix genomes, with only 6,243 polymorphic sites detected within the 217,546-bp-long alignment and nucleotide diversity (Pi) of 0.01222. Pairwise comparisons of COG profiles suggest that the 8 Planktothrix genomes are also functionally highly similar (Pearson correlation coefficients range from 0.97 to 1.00 [Fig. 1A]). However, the hierarchical clustering of COG-annotated genes (Fig. 1B) reveals that some genomes are more closely related than others. Proximity of groupings indicates the relative degree of similarity between the COG-annotated gene complements and shows that strains within a chemotype are more closely related, as seen in, for instance, Cht 1 (NIVA-CYA 98 and NIVA-CYA 406) and Cht 9/10 (NIVA-CYA 34, NIVA-CYA 405, and NIVA-CYA 540), than are strains of different chemotypes. The amount of homologous genes in 8 genomes was detected by examining all protein-coding genes, including the 33 to 35% not assigned to COGs. Pairwise gene complement comparisons revealed that at least 82% of protein-encoding genes have a homolog in all eight genomes (Fig. 1C). Furthermore, 98 to 99% of the protein-encoding genes in the red strain NIVA-CYA 540 have a homolog in the green strains NIVA-CYA 34 and NIVA-CYA 405. Phylogenetic analyses of 200 core genes support the finding that strains within a chemotype (Cht 1 and Cht 9/10) are more closely related (Fig. 1B; see also Fig. S1 in the supplemental material), essentially showing the same topology as the COG clustering. All together, these results suggest that the occurrence of red strains within Cht 9/10, which mostly consists of green strains, is probably caused by acquisition of phycoerythrin-encoding genes by green strains.

Eight different phycocyanin (PC)-encoding genes were identified in all Planktothrix genomes studied. Similar to other cyanobacteria, genes coding for alpha PC (cpcA) and beta PC (cpcB) subunits are clustered with three rod linker proteins (cpcC₁, cpcC₂, and cpcD) in a 3.2-kb cpcBAC₁C₂D operon (Fig. 2A). Additionally, the genes cpcE and cpcF (encoding two subunits of phycobilin lyase) form another small gene cluster. Neither the cpcEF cluster nor the cpcG gene (coding for phycobilisome rod-core linker) is located in the vicinity of other photosynthesis genes. The cpc genes are highly conserved, with between 0.00085 and 0.03754 substitutions per base (Table 2).

Maximum likelihood phylogenetic analyses of concatenated cpcA and cpcB sequences revealed that NIVA-CYA 540 clusters with red strains NIVA-CYA 98, NIVA-CYA 406, and NIVA-CYA 407. However, NIVA-CYA 540 clusters together with green NIVA-CYA 34 and NIVA-CYA 405 strains in the cpcC₁C₂ tree (Fig. 3). Comparison of cpcB sequences showed that the NIVA-CYA 540 gene is identical to that of all red strains (see Fig. S2 in the supplemental material). Interestingly, examination of the cpcA gene located downstream of cpcB revealed that in strain NIVA-CYA 540, the 5′ end of the gene is identical with red strains while a small part of the 3′ end of the gene resembles the green strains (Fig. 2B). This finding indicates that in strain NIVA-CYA 540, the 5′ (upstream) part of the cpcBAC₁C₂D operon was replaced during insertion of the phycoerythrin gene cluster using genic homology as the recombination locus.

The genes involved in synthesis of phycoerythrin (PE) are organized into one 19.7-kb bidirectional gene cluster located upstream of the phycocyanin gene cluster cpcBAC₁C₂D (Fig. 2A). The phycoerythrin gene cluster comprises 20 genes in Planktothrix (Table 2; Fig. 2A), and comparisons with other cyanobacteria indicate that 18 of the genes have homologs in previously characterized phycoerythrin operons. The genes pebAB and cpeAB code for phycoerythrobilin (PEB) and PE alpha and beta subunits, respectively. Six genes code for the PE-specific phycobilisome rod linkers, and four genes encode lyases (Table 2). Two of the five genes encoding hypothetical proteins, ORF3 and ORF4 (Fig. 2A; Table 2), have no homologs in PE gene clusters in genomes available through the IMG system.

The phycoerythrin gene cluster is otherwise highly conserved at the sequence level. Only three genes (cpeA, cpeB, and cpeU) show any nucleotide diversity at all; the remainder of the genes are identical in all four red strains (Table 2). In these three genes, the strain NIVA-CYA 540 has a sequence identical to that of strain NIVA-CYA 407, which might indicate that the phycoerythrin gene cluster in NIVA-CYA 540 originates from this strain or a strain associated with the same chemotype.

Visual inspection of the phycocyanin gene cluster cpcBAC₁C₂D revealed that one of the breakpoints of the homologous recombination event leading to insertion of the phycoerythrin operon in strain NIVA-CYA 540 is most likely located within the cpcA gene. In order to identify the second recombination breakpoint, an approximately 9-kb-long region downstream of cpeZ in the genomes of the red strains (corresponding to a 9-kb-long region upstream of cpcB in the genomes of the green strains) was examined. The genes encoding the following proteins flank cpeZ/cpcB in all eight genomes: SMC chromosome segregation protein, protein of unknown function DUF88, short-chain alcohol dehydrogenase, 4-hydroxyphenylpyruvate dioxygenase, and CHAP domain (Fig. 2A). Examination of informative sites (see Fig. S3 in the supplemental material) revealed several recombination events in all genes except for the gene coding for protein of unknown function DUF88, which is highly conserved. Interestingly, the 8.5-kb sequence comprising all genes except for the gene encoding the CHAP domain is almost identical in strains NIVA-CYA 540 and NIVA-CYA 407, differing in only one position. Visual examination of informative sites of the gene coding for the CHAP domain revealed that the strains NIVA-CYA 540 and NIVA-CYA 407 have different sequences at the 5′ end of the CHAP domain gene (Fig. 2C). We therefore suggest that putative recombination breakpoints are located in cpcA and the gene coding for the CHAP domain, indicating that the DNA fragment inserted through homologous recombination is approximately 30 kb long. This entire 30-kb fragment differs at only a single nucleotide between strains NIVA-CYA 540 and NIVA-CYA 407, which again supports the notion that the DNA fragment inserted into an originally green strain originates from strain NIVA-CYA 407 or a closely related strain associated with the same chemotype as NIVA-CYA 407.

The aligned and concatenated sequences of cpcBA (reverse complement) and the CHAP domain genes were examined for recombination signals with the RDP4 software. Four methods implemented in RDP (Geneconv, MaxChi, Chimaera, and 3 Seq) detected NIVA-CYA 540 as recombinant, with the green strain NIVA-CYA 34 as major parent and the red strain NIVA-CYA 98 as minor parent (Fig. 2D). Recombination breakpoints were detected in cpcA (position 303) and the CHAP gene (position 13) and correspond well with positions of the most 5′ red-like informative site in cpcA (position 356) and the most 5′ green-like informative site in the gene encoding the CHAP domain (position 141).

Since classification of Planktothrix strains into chemotypes is based on the profile of oligopeptide isoforms produced, all genomes were examined for the presence of the genes needed for oligopeptide production. Oligopeptide isoforms produced by each strain (according to the work of Rohrlack and coworkers [35]) and the presence of genes encoding biosynthetic pathways are shown in Table 3. In general, genes encoding synthesis of all oligopeptides but microginin are present in every genome examined, even though several oligopeptides are not produced by all strains. The microginin gene cluster is present only in the genome of microginin-producing strain NIVA-CYA 98.

Based on oligopeptide profiles, Rohrlack and coworkers (35) have classified 87 Planktothrix strains from Northern Europe into 17 distinct chemotypes. Here, we compare genomes of eight Planktothrix strains from six of these 17 chemotypes. However, Cht 10 is, according to Rohrlack and coworkers (35), closely related to Cht 9 and was considered a subgroup of the latter in the present study. Both comparison of protein-coding genes and phylogenetic analyses of 200 core genes revealed that the genome of NIVA-CYA 540, the only Cht 10 strain described by Rohrlack and coworkers (35), is highly similar to the genomes of the Cht 9 strains NIVA-CYA 34 and NIVA-CYA 405, supporting the hypothesis that Cht 10 is a subgroup of Cht 9. In the light of genome similarity and the fact that cyanopeptolin isoforms are the only oligopeptides which are different in Cht 9 and Cht 10, we expect that a mutation within the cyanopeptolin synthetase gene cluster leads to the formation of Cht 10. Additional analyses are also required to elucidate why several oligopeptide gene clusters are present in genomes of strains that do not produce (detectable amounts of) these oligopeptides; it may be that the operons have been inactivated by mutations, as shown for microcystin synthetase genes (9, 50).

The most significant finding of this study is the transfer of the entire 19.7-kb phycoerythrin gene cluster between Planktothrix strains isolated from Lake Steinsfjorden. We propose that the DNA fragment containing the phycoerythrin gene cluster was transferred by transformation or transduction and integrated into the chromosome through homologous recombination.

Transformation and transduction are considered the most likely mechanisms for DNA uptake, since no gene transfer agent systems have been documented in cyanobacteria (18) and no genes coding for cell-to-cell junctions were found in any of the Planktothrix genomes. Cyanobacteria have been shown to be naturally competent (e.g., references 51 and 52); therefore, it is conceivable that Planktothrix strains have the ability to take up naked DNA from the environment. Free DNA is released into the aquatic environment from decomposing cells or viral particles or through excretion from living cells and can persist for days to years (53) in fragments exceeding 10 kb in size (54). As several chemotypes of Planktothrix have been shown to coexist in Lake Steinsfjorden (34, 35), DNA from different chemotypes is most likely continuously available for natural transformation. Cyanophage-mediated transduction has also been shown to be important (e.g., references 55 and 56), and the majority of marine cyanophages have been reported to carry one or more photosynthetic genes (57, 58). Notably, no cyanophages containing the entire phycoerythrin gene cluster have been reported so far, and no obvious traces of cyanophages were found in the vicinity of phycocyanin and phycoerythrin gene clusters.

Following the physical transfer of exogenous DNA into a new cell, foreign DNA is usually integrated into the recipient genome through recombination. Functionally related genes (islands) that are frequently transferred between strains are often located in proximity to tRNAs or transposons (e.g., references 59 and 60). In this case, however, we did not find such hallmarks of island shuttles. For homologous recombination, the incoming sequences must contain a region of sufficient length that is similar to sequences in the recipient genome (61). The Planktothrix genomes investigated in this study show a high degree of sequence identity, and therefore, homologous recombination events between these eight strains could be expected to occur frequently. Moreover, homologous recombination can facilitate the transfer of heterologous genes, particularly when these are located between regions of sequence similarity shared by the donor and the recipient (62). In strain NIVA-CYA 540, the DNA fragment containing phycoerythrin genes is inserted at the same location as in other red genomes, suggesting homology-based incorporation of DNA. In the genomes of red strains, the phycoerythrin genes are located between the phycocyanin operon and genes upstream of the operon (putative gene encoding a CHAP domain). Transfer and incorporation of such a DNA fragment from a red genome into a green genome will lead to transfer of the phycoerythrin gene cluster and production of the pigment.

The DNA fragment inserted to NIVA-CYA 540 is nearly identical to the equivalent genome region in strain NIVA-CYA 407 (Cht 5), suggesting that the DNA fragment containing the phycoerythrin operon originates from a strain associated with Cht 5 and that the transfer event is recent. The latter is also supported by long-term field study of Lake Steinsfjorden since the red type of Cht 9 was first reported in 2004 (34).

Bacteria tend to lose nonfunctional or otherwise unneeded DNA from their genomes (63, 64); therefore, fixation of the acquired DNA within the genome is highly dependent on its functionality and/or advantage to the recipient under selectable environmental conditions (65–67). An operon, like the phycoerythrin operon transferred to NIVA-CYA 540 from a closely related donor, contains all genes and regulatory regions required for a function to be conferred and is therefore more likely to be fixed than a single gene or genes from a distant relative after HGT (20). The red color of NIVA-CYA 540 filaments supports the functionality of the horizontally transferred phycoerythrin gene cluster. Consequently, the fixation of the operon within the chemotype is likely to depend on the competitive gain associated with expressing the phycoerythrin gene cluster and the contribution to the ecological fitness of the given Planktothrix chemotype. Since phycocyanin and phycoerythrin absorb different colors of light, red and green cyanobacteria can coexist under white light as shown by competition studies (68). Previous studies have also shown that growth of Planktothrix in Lake Steinsfjorden is probably first of all controlled by irradiance (69). Therefore, coexistence of strains associated with the same chemotype (Cht 9/10) but producing different light-harvesting pigments might increase the range of niches in which this chemotype is able to thrive. At the same time, green- and red-colored Planktothrix strains have a similar depth distribution in Lake Steinsfjorden, indicating that growth under low-irradiance conditions and formation of blooms in the metalimnion do not necessarily require phycoerythrin (69). Moreover, production of phycoerythrin involves a metabolic cost and does not necessarily led to a higher growth rate (70). Further sampling of Planktothrix strains from Lake Steinsfjorden or from other lakes considered to form the same metapopulation (29) and competition experiments under laboratory conditions are necessary to investigate whether the red type of Cht 9 has an advantage or disadvantage compared to the green type of Cht 9. Another important question is whether transfer of other genes or gene clusters is necessary to become a successful red Planktothrix strain or if transfer of a phycoerythrin gene cluster is all that is needed.

Based on the higher degree of sequence similarity detected in phycoerythrin genes than in phycocyanin genes, it is appealing to suggest that the ancestral state of Planktothrix was green and that, during evolution, some strains have acquired genes coding for phycoerythrin as previously suggested for other cyanobacteria (40, 41). However, analysis of both phycoerythrin and phycocyanin sequences from a larger number of red Planktothrix strains is necessary before conclusions here can be drawn.

In summary, we show that the filamentous cyanobacterium Planktothrix is able to take up and integrate large (up to 30-kb-long) DNA fragments. Our results demonstrate that through a single HGT event, highly similar Planktothrix strains can acquire a complex gene cluster coding for synthesis of an additional photosynthetic pigment and thus an opportunity to exploit a new ecological niche(s).