Horizontal transfer of bacterial polyphosphate kinases to eukaryotes: implications for the ice age and land colonisation

Recent reviews have proposed a widespread occurrence of horizontally transferred bacterial type polyphosphate kinase enzymes in eukaryotes [1]. Inorganic polyphosphate (poly P) has been present since pre-biotic times [2] and has been proposed as an energy distributor in a pre-ATP world [3]. Poly P is found in organisms that represent species from each domain in nature: Eukarya, Archaea and Bacteria [4–6]. Studied mainly but not exclusively in prokaryotes, poly P and its associated enzymes are vital in diverse basic metabolism, in at least some structural functions and, notably, in stress responses [7]. These numerous and unrelated roles for poly P are probably the consequence of its presence in life-forms from early in evolution [8]. The genomes of many bacterial species, including human pathogens, encode a homologue of a major poly P synthetic enzyme, polyphosphate kinase 1 (PPK1) based on a phospholipase D structure [9]. Loss of PPK1 produces reduced poly P levels, and deletion of the ppk1 gene in pathogens also results in a loss of virulence towards protozoa and animals [7]. A second PPK activity in bacteria, PPK2, is related to thymidylate kinase [10, 11]. PPK2 is distinguished from PPK1 by its preference for utilising poly P in the reversible generation of GTP. Polyphosphate-AMP-phosphotransferase (PAP) uses poly P to phosphorylate AMP to ADP [8]. Other enzymes that influence accumulation of poly P are the hydrolytic enzymes: exopolyphosphatase (PPX) that releases Pi from the ends of poly P [12] and endopolyPases (PPN) that cleave poly P to progressively shorter chains. These enzymes together maintain poly P metabolism and catabolism in bacteria. Poly P metabolism is less well characterised in eukaryotic systems. In Dictyostelium discoideum a PPK activity (DdPPK2) based on three actin like proteins has been documented [13]. Hothorn et al. 2009 [14] identified a PPK activity associated with a fourth class of enzyme, a vacuolar transport chaperone (VTC), which has a distribution largely limited to simple eukaryotes. Intriguingly, Reusch et al. (1997) [15] assigned a PPK activity to a Ca²⁺ ATPase activity in humans. Recently Lonetti et al. (2011) [16] demonstrated Saccharomyces cerevisiae to have undetectable levels of poly P in mutants unable to produce inositol pyrophosphates. Hence, a range of families of enzymes have been shown to have PPK activity. The enzymes responsible for poly P synthesis in most eukaryotes remain unidentified with the exceptions of an actin related protein in Dictyostelium discoideum[13], vacuolar transport chaperones in Saccharomyces cerevisiae[14] and a Ca²⁺ ATPase in Homo sapiens[15]. In this respect, the hypothesis that bacteria – like PPK enzymes based on phospholipase D [9] or thymidylate kinase [10, 11] exist in eukaryotes, requires investigation.

Horizontal gene transfer (HGT) is now recognised as an important force in eukaryote genome evolution [17]. Hooley et al. (2008) [18] summarised evidence that suggested a very limited distribution of bacterial type PPK1 and PPK2 enzymes in a small number of eukaryotes. Rao et al. (2009) [1] have claimed evidence for a surprisingly wide potential distribution of the bacteria type PPK1 and PPK2 in eukaryotes. Computer aided bioinformatics techniques can exploit genome project databases swiftly to summarise likely candidates for PPK activity. Similarity search tools such as BLAST [19] and multiple alignment programs like Clustal W/X, TCoffee and MUSCLE [20–22] allow rapid comparisons of sequence data. Phylogenetic analyses [23, 24] can infer evolutionary relationships between DNA or protein sequences. The current paper aims to examine the evidence for bacteria type polyphosphate kinases in eukaryotes and to consider their relationships to possible donor prokaryotes. The possible selective advantages in acquiring such prokaryote genes are discussed.

Nearly 200 eukaryote genomes have been examined in the present work for evidence of bacteria type PPK1 and PPK2. No single database contains a definitive list of eukaryote bacteria type PPKs but we can conclude that relatively few eukaryotes possess these enzymes (Table 1, Additional file 1). Hooley et al. (2008) [18] reported extensive conservation of structure between bacterial PPK1s and their eukaryote counterparts. Here we demonstrate a similar degree of conservation between the two eukaryote examples of PPK2 and bacterial counterparts (Figure 1). There is therefore a taxonomically discontinuous distribution of a limited number of bacterial type PPKs in diverse simple eukaryotes. The most parsimonious explanation for such eukaryote ppk genes, several of which contain no introns suggesting prokaryote origins, is a number of independent horizontal transfer events from bacteria.

At the outset, it was important to eliminate possible artefacts from the analysis. There are several clear examples of likely incorrect identifications of PPK encoding genes. For example the 100% DNA sequence identity of the proposed Populus ppk1 and ppk2 with Ralstonia/Delftia seems an obvious case of bacterial contamination. Several hits to bacteria – type PPK1s have been claimed for insects [1, 26]. These exciting suggestions must be examined in the light of the possible contamination of gene banks with bacterial sequences [27]. Conversely it is possible that vector search programs may erroneously eliminate real examples of horizontal transfers. It is important to consider that DNA sequencing and annotation errors may give misleading gene descriptions [28]. Table 4 summarises a relatively small number of arthropod matches that identify potential PPK1 and PPK2 encoding sequences. Of these, four have clearly been identified as likely bacterial contaminants by the original workers. Just two of the remaining four have some EST support for their presence as genuine eukaryote genes. Their lack of similarity to any other arthropod DNA sequences and high DNA identity to bacteria still makes their annotation questionable.

Host/parasite relationships provide obvious opportunities for gene exchange [17]. However, in a genome project there is the potential for mistaking a horizontally transferred gene for a bacterial DNA contaminant acquired during gene bank construction [29]. The Wolbachia insect symbiont has integrated 30% of its genome into the Callosobruchus beetle genome; most of these genes are disrupted and transcriptionally inactive [30]. Klasson et al. 2009 [31] demonstrated the expression of Wolbachia genes in Aedes mosquito. These observations may be consistent with early Blastn reports of ppk1 matches at the DNA level to some invertebrates [1, 26]. However, bacterial DNA may be horizontally transferred but not active as a PPK1 product. The twenty Wolbachia genomes available on NCBI were searched via BLASTp using the E. coli PPK1 (NP_416996 ) and P. aeruginosa PPK2 (NP_248831) protein sequences. No significant PPK1 (lowest e – value 2.3, just 29% identity over 34 amino acids) or PPK2 (two hits at e – value < 0.05, best being e = 0.042 with just 27% identity over 82 amino acids) matches were found. The source of potential PPK encoding sequences, whether active or not, in invertebrates remains a puzzle. Claims for the widespread occurrence of bacteria type PPKs in eukaryotes [1] are overstated and these enzymes show a much more restricted distribution.

Figure 2 demonstrates quite distinct clusters of PPK1 enzymes in eukaryotes. These clusters are consistent with the taxonomic groupings of these eukaryotes. P. patens, O. tauri and O. lucimarinus form one group (green plants and green algae), the three Dictyostelium species, P. pallidum and T. trahens (non-photosynthetic eukaryotes) a second group, with P. yezoensis and C. merolae (red algae) forming a third group. All of these groups are well supported by bootstrapping values. The most obvious donor for horizontally transferred PPK1s in eukaryotes is an ancestor in common with the cyanobacteria as shown by the association of the Cyanothece sp. PPK1 with the eukaryotic grouping. The cyanobacteria formed the original endosymbionts generating chloroplasts. The two Ostreococcus species are generally considered to be very divergent, with an average of only 70% amino acid identity between orthologous proteins, making O. tauri and O. lucimarinus amongst the most dissimilar members of the same genus in any eukaryote [32]. The most parsimonious explanation may be the acquisition of PPK1 when Ostreococcus and Physcomitrella last shared a common ancestor, with subsequent losses in other lines. Attempting to trace a common or single origin of a specific ppk gene may be unrealistic, particularly in light of the complex evolution of algae with the potential for secondary horizontal gene transfer events [33].

There are only two convincing eukaryote PPK2s found on the Interpro database (Figure 1). Phylogenetic analysis suggests that these two have quite different origins (Figure 3). Interestingly, O. lucimarinus appears not to have a ppk2 – presumably the O. tauri example was gained after the species separated or a ppk2 sequence inherited from a common ancestor was subsequently lost by one species. Derelle et al. 2008 [34] describe one possible candidate virus, OtV5, as an agent of horizontal transfer in this genus. Raymond and Blankenship (2003) [35] emphasise the importance of HGT in evolution of eukaryotic algae with endosymbiosis extending beyond the original event of engulfment of cyanobacteria to create plastids to include acquisition of genes from other algae at other times. Rohwer and Thurber (2009) [36] give further examples of HGT into metazoans within the marine environment including viral vectors moving genes between animals and plants.

It is also important to highlight those groups of organisms which are notable by their absence from the short list of eukaryote PPKs. Horizontally transferred genes have been shown to affect the metabolism of numerous fungal species [37]. Yet no examples of PPK1 or PPK2 were observed in the annotated or partially completed 116 genomes of fungi investigated here. However, since fungi do not engulf organisms via phagotrophy, it may suggest an additional clue to the origin of the horizontally transferred PPKs in other eukaryotes. Similarly, most simple eukaryotes and almost all photosynthetic organisms examined did not contain these bacteria type enzymes.

Horizontally transferred genes from eubacteria would be expected at least initially to contain no introns. Assuming that the bacteria type PPK1 and PPK2 enzymes have been acquired from bacteria, how and why do horizontally transferred genes acquire introns? Table 2 reveals that some eukaryote ppk1 and ppk2 examples are indeed intron free. Spliceosomal introns are typically found in nuclear genomes, and their presence indicates a major role in evolution, although no overall general function is known. It is known that organisms with shorter life cycles tend to have less intronization perhaps reflecting selection for reduced processing time for mRNAs. Intron length has been positively correlated to gene expression in unicellular eukaryotes but negatively correlated in multicellular eukaryotes whilst mildly deleterious elements may accumulate in more complex organisms that have small populations [38, 39]. In the examples shown in the present work there is a range from ppk1 genes with no introns (such as in Ostreococcus and D. discoideum) to the 21 and 22 introns found in P. patens (Table 2). Neither example of eukaryote ppk2 has introns (O. tauri, N. vectensis). Rotifers, which also predate bacteria, have acquired many bacterial genes and whilst some are defective, others are expressed and these may include genes with introns [40]. This compares with relatively little evidence of intron gain in Entamoeba horizontally transferred genes [41] or in the recent evolutionary past of higher plants [42]. There is no single source of data expressing intron density for each of these eukaryote species in a common format, with some authors and databases quoting introns per gene, introns per transcript, introns per kb of coding sequence or introns per spliced gene. The C. merolae and D. discoideum genomes have means of just 0.005 and 1.31 introns per gene respectively [39] so absence or a low number of introns in their ppk genes is not remarkable. A possible mechanism of intron gain could be increased transposon activity that activated double stranded DNA repair [43].

In P. patens an accepted horizontally transferred glycerol/water channel gene has 5 introns [44]. There are four recognised domains for PPK1 [9] so multiple introns are unlikely to have evolved as a means to promote domain shuffling. Stenoien (2007) [45] suggests that highly expressed genes in P. patens have shorter introns than genes with low expression levels and so acquisition of small introns may reflect mechanisms to regulate gene expression (Table 3). It has been suggested that there has been an ancestral duplication of the P. patens genome [27, 46] with essential genes such as rad51 (recombination repair) being present in two copies, thus allowing pseudoallelism and the protection of an essential function in the gametophyte (haploid) plant. Hence, the presence of two copies of a ppk1 gene in this species suggests an essential function. Table 3 implies that the two P. patens ppk1 genes have an unusually high number of introns. However, Csuros et al. 2011 [43] suggest that this species has a mean of 5.5 introns per kb of coding sequence. On this basis, the apparently large number of introns may, at least partially, reflect the unusually large size of these two genes for the species. Sucgang et al. 2011 [47] describe mean intron values of 1.9 and 1.5 per spliced gene for D. discoideum and D. purpureum respectively. D. discoideum and D. purpureum diverged approximately 400 million years ago [47] so the acquisition of PPK1 in these and the other slime moulds presumably predates this speciation event. The colonisation of the land by plants around 470 million years ago was followed by the divergence of the line leading to Physcomitrella from that leading to Selaginella and higher plants about 430 million years ago [48]. Hence, it is reasonable to suggest that one horizontal transfer event may have occurred in a common ancestor of these species and the genes have been maintained by common selective pressures. In individual species such as Selaginella the ppk genes have subsequently been lost. Although this is the most parsimonious explanation, the occurrence of multiple acquisitions and losses should not be ruled out.

What are the advantages to a eukaryote in maintaining a bacteria-type PPK? As poly P is found in all cells there must be alternative mechanisms for manufacture in eukaryotes, perhaps based on the actin related PPK3 system [18]. Horizontally acquired PPK1 or PPK2 must replace or supplement such native enzymes. Indeed, the primitive red alga C. merolae has a VTC1p homologue in a poly P containing vacuole. Poly P is a key store of phosphate in this acid and heat tolerant species rather than phytic acid commonly found in higher plants [49]. Similarly poly P is the stored form of phosphate in green algae [50], hence acquiring PPK1 or PPK2 may provide a greater flexibility in nutrient stress responses in algal cells. O. tauri is a unicellular green alga that is an important member of the global phytoplankton and has cell dimensions of around 1 μm diameter, equivalent to a prokaryotic cell. In O. tauri nitrogen starvation results in an increase in PPK activity [51]. So poly P accumulation via enhanced PPK activity is clearly a valuable response to nutrient depletion stress, possibly to maintain phosphate reserves. Such activity may withdraw more reactive and soluble phosphate molecules from metabolism or potential efflux from a tiny cell, like O. tauri or C. merolae, with a high ratio of surface area to volume.

Simple multicellular eukaryotes may face periods of water immersion followed by desiccation. For example, mosses such as P. patens and slime moulds both have a need to escape aqueous environments to sporulate [52]. Desiccation tolerance in individual cells and tissues is required in plants such as P. yezoensis and P. patens, which lack the efficient vascular systems of higher plants [27]. Under such circumstances there may be selection pressure to use horizontally transferred genes available in the surrounding bacterial population. Higher plants have more sophisticated water transport, compatible solute manufacture and waxy cuticles so that individual cells are no longer desiccated. Similarly, filamentous fungi have thick chitinous walls and the VTC PPK system [14], so may be more desiccation tolerant anyway. Hence slime moulds and mosses may represent special cases of incomplete adaptations to dry land colonisation.

Eukaryote PPK1s are characterised by being of higher mass than prokaryote homologs [18] – N terminal extensions perhaps reflect differences in targeting the enzyme, e.g. to a membrane or subcellular compartment, or in assuming quaternary structures. However, Target P and Signal P analysis [53] failed to show any signal peptides or common chloroplast or mitochondrial targeting sequences amongst the eukaryote PPK1s or PPK2s. Zhao et al. (2008) [54] demonstrated that poly P influences intron splicing protein localisation and concentration of poly P subapically in E. coli and plays a crucial role in establishing cell polarity in cytokinesis. Subcellular localisation of PPK activity in eukaryotes is then also presumably important. Inorganic phosphate and smaller molecules such as ATP and GTP are highly reactive and PPK provides a mechanism for storage of phosphate in the less reactive high molecular weight poly P. For some eukaryotes, such as those adapting to novel stressful environments without the benefit of a developed vascular system, the acquisition of a bacteria type PPK and the evolution of additional ppk genes under the control of new promoters may provide additional opportunities for the control of poly P synthesis within the cell. A ppk1 deletion mutant of the social slime mould D. discoideum had reduced levels of poly P and was deficient in development, sporulation and predation [55].

Lenton et al. (2012) [56], using P. patens as an experimental organism, have recently described the exciting hypothesis that non-vascular plants colonising rock surfaces accelerated chemical weathering, releasing phosphate for enhanced growth of oceanic phytoplankton, to the extent that falls in atmospheric carbon dioxide precipitated the global growth of ice sheets. Central to this concept is the release of phosphate to the ocean to fuel oceanic carbon fixation. Interestingly, two other eukaryotes that have bacteria type PPK1 and PPK2 are the abundant picophytoplankton Ostreococcus tauri and O. lucimarinus (Table 1[18]). In this respect, horizontal transfer of bacterial ppk genes to early plants such as P. patens colonising the land and to marine phytoplankton exploiting the consequent increase in oceanic phosphate, would have been a key factor in the slow decline in atmospheric carbon dioxide in the Ordovician.

Convincing database examples of eukaryote PPKs derived from bacteria type PPK1 and PPK2 enzymes are rare and currently confined to a few simple eukaryotes. These enzymes likely represent horizontal transfer events occurring before the time of the colonisation of land by plants, with the possibility of subsequent multiple losses and further gains in different lineages. It is proposed that the retention of such horizontally transferred sequences is an advantage for stress tolerance in eukaryotes without sophisticated multicellular adaptations to stresses such as desiccation or nutrient depletion. The enhanced acquisition, release and storage of phosphates facilitated by bacterial PPKs may have promoted the colonisation of land by early plants and fuelled the growth of oceanic phytoplankton. There is very limited evidence for DNA sequences encoding PPKs in a wider range of eukaryotes, notably some invertebrates, though it is less clear that these represent functional genes.