Transcriptional analyses

Date palm (Phoenix dactylifera L.) seeds were collected from fruits of the Sukary cultivar, imported from Kingdom of Saudi Arabia. Seeds were isolated, washed, air-dried and stored in plastic bags at room temperature until required. Seeds were then germinated in vitro in a current of running water for two weeks before planting in culture boxes placed in an incubator at 30 ± 2°C and humidified daily and seedlings were obtained 15 days after sowing. Seedlings with a radicle length of 0.5 or 2 or 4.5 cm were referred as stage I, II and III, respectively. For stress experiments, 2,3,7,8-tetrachlorodibenzo-p-dioxin (2,3,7,8-TCDD dissolved in toluene at 10 μg mL^-1, purity 99%) was purchased from Supelco Inc., USA. TCDD was placed in a 10 mL capped glass tube and evaporated to dryness under nitrogen. For health and environmental safety reasons, residual TCDD was re-dissolved in a minimum volume (100 μL) of dimethyl sulfoxide (DMSO), and 5 mL of aqueous solutions of TCDD were prepared in deionized and distilled water to obtain initial concentrations of 0, 10, 50 and 100 ng L^-1 TCDD. For treatment of seeds with TCDD, seeds were germinated as described above and humidified daily with the prepared solutions of TCDD at various concentrations. Seedlings at stages 0, I, II and III were taken for further analysis. For drought stress, seedlings were pre-treated with water for 2, 4 and 6 days under the culture conditions described above. Osmotic stress was achieved by irrigating seedlings with water at concentrations of 150 and 300 mg L^-1 NaCl at each development stage.

Changes in relative transcriptional abundance of genes encoding LD-associated proteins in response to TCDD exposure were analyzed by reverse-transcription quantitative PCR (RT-qPCR). Briefly, frozen fine powder (1 g) samples from whole seedlings at stages 0, I, II and III were used to extract total RNA using an RNeasy kit (Qiagen) according to the manufacturer’s instructions. The quality of extracted RNAs was checked on agarose gels and concentrations measured by a Nanodrop device. Remaining traces of genomic DNA were digested by DNase I and the lack of trace genomic DNA in total RNA was confirmed by a control PCR using total RNA as the template. Aliquots of 1 μg total RNA were used for first-strand cDNA synthesis according to Hanano et al (2014). Real-time and qPCRs were performed as described by Hanano et al (2014) and primers are listed in S1 Table. The relative expression of target genes was normalized using two reference genes Act-1 and Tub-β [49]. Each measurement was performed in triplicate together with a dilution series of the reference gene. PCR efficiencies were between 95% and 105% (data not shown) and the average of C_T was taken. The relative quantification RQ of target genes was calculated directly using software from the qPCR system. Sequences of amplified regions were confirmed by sequencing on an ABI 310 Genetic Analyzer using a Big Dye Terminator kit (Applied Biosystems).

Oil palm transcriptome analysis was carried out according to [50]. The oil palm genome P5-build was used to read map the RNA seq data from Roche 454 reads (a full dataset is available from NCBI BioProject PRJNA201497). Briefly, reads from Roche/454-derived libraries were assembled into isotigs, which were blasted onto Arabidopsis thaliana gene models with a threshold of E-value < 10⁻⁵. The best-hit A. thaliana gene model was assigned to the homologue of the query isotig. To estimate expression levels of genes in mesocarp and kernel tissues, Illumina HiSeq 2000 reads from each library were mapped to assembled isotigs from all Elaeis guineensis reads by using the Burrows–Wheeler Aligner. Gene group expression levels were calculated as the number of mapped reads on each isotig divided by the total number of isotigs, multiplied by 100,000, and scaled by the number of genes in each gene group. Both copy number and read coverage were the mean of measures from two biological replicates. Data were analysed as described above for Roche/454 data, except that expression levels were calculated as transcripts per million tags. Identification of expression caleosin in oil palm was done using open source Tuxedo suite software [51].

Bioinformatics procedures

The bioinformatics procedures used in this study are summarised in the workflow depicted in S1 Fig. For data collection, a list of CLO/PXG–like genes was identified using one model species (A. thaliana) and three economic crops (P. dactylifera, E. guineensis, Musa acuminate), which were then used as the source of the CLO/PXG reference sequences in this study. Using reference genomes and proteomes from public databases (i.e. NCBI Entrez and Pubmed), we constructed a dataset of well-annotated full-length CLO/PXG genes. We then selected 34 species from the Viridiplantae based on the quality of their sequenced genomes, their economic and conservation importance, and their evolutionary significance. The selected species included representatives from all available major Viridiplantae groups, such as the algal Charophyte, Chlorophyceae and Trebouxiophyceae clades, plus the major Embryophyte (land plant) taxa including bryophytes, ferns, gymnosperms, basal angiosperms and the major extant monocot and dicot groups. Although the vast majority of currently available sequence data comes from the monocot and dicot groups, and particularly from crop species of economic importance, we are confident that there is sufficient genomic data from the sequenced non-angiosperm groups to enable robust conclusions to be drawn about the evolution of this gene family in the Viridiplantae.

Finding and assessing candidate CLO/PXG sequences from representative genomes

Using sequences from the 34 reference species shown in S2 Table, we performed local alignment searches within each species using blastp from the BLAST+ toolset [52]. At this stage the total number of candidate sequences was >500. These sequences were analysed using InterProScan (http://www.ebi.ac.uk/interpro/) to confirm the presence of the calcium binding and EF hand domains. CLO/PXG sequences were also visually inspected using Geneious version 10.0.9 (http://www.geneious.com, and CLC Genomics Workbench 10.0.3 (https://www.qiagenbioinformatics.com/) to confirm the presence of the full range of canonical CLO/PXG domains [53, 54]. After confirming the presence of major caleosin domains, we derived a list of candidate caleosin genes and assigned names to these sequences, which are shown in S2 Table with corresponding species names and clade/groups. We then performed a set of experiments using published and peer-reviewed toolsets on these candidate sequences. The experiments included motif discovery, physical and chemical property analysis, domain sequence conservation analysis and consensus study and evolutionary pattern studies. This enabled us to narrow down the dataset to 131 CLO/PXG-like sequences. A list of physical and chemical properties of CLO/PXG proteins (molecular weight (MW), isoelectric point (pI), amino acid length) is shown in S2 Table. The physical and chemical properties of CLO/PXG proteins were computed using ExPASY (http://www.expasy.org/protscale/) [55, 56]. For further detailed analysis we selected 67 of the 131 sequences that presented themselves as strong candidates to belong to the CLO/PXG family (S2 Table where sequences marked in green are the 67 selected CLO/PXGs). A list of the 34 species with their corresponding CLO name, Taxon ID and number of CLO/PXG isoforms per species is given in S3 Table.

Motif, transmembrane (TM) domain, secondary structure prediction and intron-exon analysis

Motif analysis was performed on the 67 sequences from 34 species to identify conserved domains that might elucidate the biological activities of these multifunctional proteins. The Multiple En for Motif Elicitation (MEME) software package was utilised to discover and analyse motifs across sequences [57, 58]. Protein datasets were used to analyse motifs for each sequence. Sequences were analysed using the discriminative mode and a window size of 15–50, which enabled identification of seven distinctive motifs. The list of motif logos and the motif distribution patterns across species sequences are shown in Fig 1. Transmembrane (TM) domain and secondary structure predictions were performed using the Geneious package and both Geneious and ClustalW aligners were used to align and prepare the sequences for display. The following parameters were used to align protein sequences: a) Cost matrix BLOSUM 45, b) Gap open penalty 12, c) Gap extension penalty 3, and d) refinement iteration of 2 and the resulting data are shown in S2 Fig. The location (start and end point) of each transmembrane domain is shown in S4 Table and S2 Fig. The lengths of the TMs are consistent at 21 residues. The secondary structure predictions are shown in S3 Fig. The Scipio program version 1.4 was used to identify intron-exon [59]. A summary graph of overall intron/exon numbers identified using the Scipio program in each sequence is shown in S4 Fig while S5 Table shows the locations of intron/exons in the full list of the 67 analysed genes.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Motif analysis.

(A) Consensus distribution of motifs in all Viridiplantae sequences. (B) Distribution of motifs across 67 representative Viridiplantae CLO/PXG proteins grouped in eight taxonomic clades as follows. (i) Trebouxiophyceae, (ii) Chlorophyceae, (iii) Charophyta, (iv) Non-seed plants, (v) Gymnosperms, (vi) basal Angiosperms, (vii) Monocots and (viii) Dicots.(C) Sequences of the 7 major motifs found in Viridiplantae CLO/PXG proteins.

https://doi.org/10.1371/journal.pone.0196669.g001

MSA (multiple sequence alignment) and phylogenetic analyses

Multiple sequence alignment and domain analyses were performed using ClustalOmega software, version 1.2.2, using the default parameters [60, 61]. The alignments were inspected using the CLC Genomics Workbench 10.0.3 (https://www.qiagenbioinformatics.com/). Complete alignments with RasMol colour codes [62] are shown in Fig 2. The amino acid sequence alignments were used to construct a phylogenetic tree using the ClustalW2 program version 2.1. The tree was generated following Bayesian Inference (BI), Neighbour-joining (NJ), and Unweighted Pair Group Method with Arithmetic Mean (UPGMA) methods. The tree topologies constructed using the three different methods showed complete consistency. The NJ tree was constructed using ClustalW2 program version 2.1 [60]. The constructed tree was inspected using FigTree and is shown in Fig 3 [63]. Phylogenetic analysis shown in Fig 3A was built from CLO/PXG sequences across Viridiplantae while the tree shown in Fig 3B was built with CLO/PXG sequences across Viridiplantae plus a range of representative basal and more advanced Fungal species, namely Aspergillus flavus, Ustilago maydis, Rozella allomycis, Coprinus cinereus, Rhizophagus irregularis, and Rhizopus delemar. Note that the addition of these six fungal sequences did not affect the phylogeny of the plant sequences.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Sequence alignments of CLO/PXG protein families from three representative Charophyte species plus 67 sequences from 34 species across the Viridiplantae.

(A) Klebsormidium nitens alignment. (B) Oryza sativa alignment. (C) Arabidopsis thaliana alignment. (D) 67 protein sequences alignment from 34 Viridiplantae species. The five major structural domains are shown respectively as the N-terminal H-caleosin, Ca 2⁺ binding EF Hand, Lipid-binding, Heme binding and kinase phosphorylation and C-terminal kinase phosphorylation domains. The proline knot region and two conserved Histidines are also shown.

https://doi.org/10.1371/journal.pone.0196669.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Phylogenetic analysis of 67 CLO/PXG sequences.

(A) 67 CLO/PXG sequences from 34 species across the Viridiplantae. (B) 67 CLO/PXG sequences from 34 species across the Viridiplantae plus six selected fungal species.

https://doi.org/10.1371/journal.pone.0196669.g003

Bioinformatics analyses

As of November 2017 we found >1310 sequence hits from the Viridiplantae that were classified in public databases, such as NCBI, as being members of the ‘caleosin superfamily’. These sequences were mostly annotated in databases as either ‘caleosin’ or ‘peroxygenase’ although some were labelled as ‘hypothetical protein’, and a small number were labelled as ‘ABA-induced protein’ or ‘EF hand protein’. An example of the annotation list for the CLO/PXG enzymes (EC:1.11.2.3) in A. thaliana can be found at http://www.brenda-enzymes.org/all_enzymes.php?ecno=1.11.2.3&table=Source_Tissue#TAB. It should be noted that although this list is annotated as ‘plant seed peroxygenase’, the source tissue list shows that the relevant CLO/PXG genes are expressed throughout the plant and not just in seeds. We found that some of the putative CLO/PXG genes in the various public genome databases were present as incomplete or partial sequences with missing and/or corrupted versions of the major canonical CLO/PXG domains known to be essential for protein function. Such sequences, which may be from non-functional pseudogenes, were normally discarded in our analysis although several examples were included to illustrate points of gene expansion as discussed below. A list of the chosen representative 34 species across the Viridiplantae and the CLO/PXG sequences used in subsequent analyses in shown in S2 Table.

As shown in Fig 1C, motif analysis using MEME enabled the identification of seven highly conserved CLO/PXG protein regions, which varied in length from 15 to 50 residues, and are found throughout the Viridiplantae. The motifs are numbered 1 to 7 in order of their motif score, which reflects their length and their extent of conservation across all the species analysed. The distribution of these motifs beginning at the N terminus of CLO/PXG proteins was 4-1-7-6-2-5-3 as shown in Fig 1A. In Fig 1B, the selected CLO/PXGs are grouped in eight distinctive taxonomic clades that represent the major Viridiplantae groups. In all cases these highly divergent plant groups show highly conserved organisation of the seven Motifs listed above. However, it should be noted that Motif 4, which is present in H-domain variants, is absent from at least one CLO/PXG sequence from all angiosperm species, i.e. Groups (vi), (vii) and (viii), and it is these Motif 4-lacking proteins that make up the L-isoforms of CLO/PXG [1, 2, 64, 65]. Motif 4 is a 30–50 residue domain present close to the N-terminus and is found in many CLO/PXG sequences throughout the Viridiplantae (and Fungi). This Motif is characteristic of the so-called H-caleosins (where H = high molecular weight), as previously reported from several labs [1, 2, 27, 28, 64–66]. We found that all Angiosperm genomes sequenced to date contain at least one copy each of the L- and H-caleosin sequences and although the evidence is less clear for the other Viridiplantae, it seems that most species also contain both isoforms. One reason for the existence of these two isoforms that are differentiated only by the 30–50 residue N-terminal insertion may be found in their respective pI values. The H-isoforms have low pI values typically below 6 while the L-isoforms have values typically above 8, which indicates that they may function optimally in different subcellular compartments [1].

Motif 1 includes the canonical calcium binding EF hand domain as found in all CLO/PXG sequences in all species. Motifs 7 and 6 contain the proline-rich and lipid binding domains respectively, which are well conserved in the Viridiplantae. Although the MEME software identified Motifs 7 and 6 as separate features, they can probably be regarded functionally as a single well conserved lipid-binding motif due to the presence of the group of about 20 non-polar residues that make up the putative TM or LD-binding domain. Motif 2 includes heme-binding and kinase phosphorylation domains and includes a relatively lengthy 32-residue consensus sequence, NIHKCKHGSDSGVYDTEGRFVPEKFEEIFSKY, which we also found as a very highly conserved domain in most plant and fungal CLO/PXGs. Motif 5 is a shorter and weaker feature that is of unknown function. Motif 3 is a well-conserved C-terminal domain with a characteristic casein kinase phosphorylation box, DGSLFE, as reported elsewhere [12, 40]. However, the full version of Motif 3 includes a larger 27-residue consensus sequence, LYWLCKDKDGFLHKETVRRCYDGSLFE, which is relatively highly conserved across the Viridiplantae.

The transmembrane domain (TM) predictions (S4 Table) and (S2 Fig) show the distribution of TM domains in each CLO/PXG sequence. Note that, since many CLO/PXG isoforms bind to LDs either instead of or in addition to bilayer membranes, these TM domains can be regarded as generalised lipid binding domains and not necessarily only involved in transmembrane functions. A single TM domain was present in each case except for five sequences where there were two TM domains. These five CLO/PXG sequences were ApCLO1 from Trebouxiophyceae, VcCLO1 and GpCLO1 from Chlorophyceae, SmCLO2 from lower plants and EgCLO3 from the monocot group. There were also six CLO/PXG sequences where no predicted TM domain was present, namely CrCLO2 from Chlorophyceae, MpCLO1 from lower plants, PdCLO4 from monocots, StCLO5, NtCLO5 and GmCLO5 from dicots. In all cases these were members of larger CLO/PXG families in each species and, since the predicted proteins had all the other features of CLO/PXGs, these may be non-lipid-binding, soluble isoforms. Gene structures were predicted using the Scipio program version 1.4. Identified intron-exons were inspected using Webscipio program and are presented in S5 Fig [67, 68]. A summary graph of overall intron/exon numbers identified using webscipio program in each sequence is shown in S4 Fig while S5 Table shows the detailed locations of intron/exons in the full list of the 67 analysed genes.

This shows that their gene organisation is relatively divergent in the Viridiplantae as a whole, although the intron/exon structures are relatively conserved within the more recently diverged dicot group of species.

In Fig 2, the protein sequence alignments are shown for CLO/PXG protein families from three representative Charophyte species, namely the alga Klebsormidium nitens (Fig 2A) which contains 5 isoforms, the monocot Oryza sativa (Fig 2B) with 10 isoforms, and the dicot A. thaliana (Fig 2C) with 7 isoforms. For a full sequence alignment of all 67 representative CLO/PXG proteins from 34 Viridiplantae species see Fig 2D. All of these alignments show high levels of sequence conservation, especially within the seven key motifs depicted in Fig 1. The three alignments in Fig 2 demonstrate how CLO/PXG gene families can be relatively large and the encoded proteins rather diverse, even within a single species. This is consistent with transcriptome and functional evidence that different CLO/PXG proteins are found in different subcellular and tissue locations, are subject to different forms of regulation and also may have different types of enzymatic or structural activities in plants [8, 27, 28, 69, 70]. In the case of the alga K. nitens (Fig 2A), the major structural domains are conserved in all five isoforms and while there are large insertions in three of the isoforms, these do not affect the integrity of the key canonical domains. For example KNCLO1 has a 100+ residue extension at the N terminal, which has no similarity with any other published sequence, but the remainder of the protein is obviously a member of the CLO/PXG family.

The monocot O. sativa (japonica rice) genome contains 10 annotated CLO/PXG-like sequences (Fig 2B), not all of which are likely to be functional proteins. In particular, OsjCLO10 lacks the Ca⁺-, heme- and lipid- binding domains and contains a 150-residue insert at the N-terminus. This insert has some similarity with the missing CLO/PXG-like domains and may therefore be due to a transposition of part of the genomic sequence, most likely due to exon shuffling. In contrast, OsjCLO2 contains a 68-residue insert at the N-terminus with no homology to other database sequences but the remainder of the protein includes all the normal CLO/PXG motifs. Apart from these unusual sequences, the remaining 8 rice isoforms show the normal protein architecture although their H-domains are much less conserved compared to those of A. thaliana as discussed below. Transcriptome data show that 8 of the rice CLO/PXG-like genes are expressed in a wide variety of vegetative and reproductive tissues and are modulated by drought, salt and cold stresses [71].

The model dicot species A. thaliana contains 7 CLO/PXG isoforms (Fig 2C). According to transcriptome data, all of the genes encoding these isoforms are expressed, although AtCLO7 is only present at very low levels relative to the other 6 genes [1, 65]. As in other Brassicaceae species, it is clear that segmental duplication is the main driver of the accumulation of the large gene family in A. thaliana [1, 65]. However, a more detailed inspection of the AtCLO7 sequence shows that the protein is not a true caleosin because it lacks critical motifs such as the Ca⁺-binding EF hand and the two heme-binding histidine residues. In comparison with the other six genes, AtCLO7 is the least expressed and shows little or no response to environmental stimuli [1], which means that it is probably a pseudogene. The other six sequences fall into two well-defined classes, namely the H-isoforms, AtCLO1, 2 & 3 plus AtCLO8, and the L-isoforms, AtCLO4 & 6.

Phylogenetic analyses of a representative 67 CLO/PXG sequences from 34 species across the Viridiplantae are shown in Fig 3A. The sequences separate into four clearly distinct clusters that are labelled as follows: A) Chlorophyte and Streptophyte green algae, B) non-angiosperm Embryophytes; C) H-isoform angiosperms; and D) L-isoform angiosperms. This phylogeny supports the conclusion from the Motif analysis (see above) that H-caleosin isoforms are ancestral to the L-caleosin isoforms. In Fig 3B, the phylogenetic analysis has been extended to include sequences from six selected basal and advanced species that represent all of the major Fungal taxa as well as the 67 Viridiplantae sequences. The fungal sequences used were: A. flavus (AflCLO1); U. maydis (UmCLO1); R. allomycis (RaCLO1); C. cinereus (CcCLO1); R. irregularis (RiCLO1); and R. delemar (RdCLO1). Note that the CLO/PXG sequences for each of these very diverse fungal species that probably diverged from one another >1 billion years ago (Bya) [72, 73] are located within a single branch of the green algal cluster and fall within the same group as the Trebouxiophyte, (shown as the yellow group A). This indicates that the fungal CLO/PXG sequences are more closely related to those of algae than to the land plants, which did not appear until after 500 million years ago (Mya), and the fungal CLO/PXGs may therefore have been derived from Trebouxiophyte algae. The origin, distribution and biological functions of fungal CLO/PXGs will be the subject of a subsequent paper.

Transcriptional analyses of CLO/PXG in date and oil palm

The oil palm genome contains six CLO/PXG-like sequences and a heatmap analysis of their expression patterns in 22 different tissue libraries is shown in Fig 4A. Transcripts of EgCLO2 were particularly highly expressed in developing fruit mesocarp and kernel (seed) tissues where there is considerable accumulation of storage lipids. In contrast EgCLO1 was more highly expressed in shoot and floral tissues, while EgCLO3 showed high but relatively even levels of expression in all tissues in a manner similar to that of a constitutive gene profile. The remaining three genes, EgCLO4, 5, 6, all responded to the same probe so these data represent the sum of their expression patterns. In comparison with the other genes, these three had much lower expression in most conditions although they showed significant upregulation in later stages of kernel, mesocarp and pollen development. These three tissues are all actively accumulating storage lipids at the stages where EgCLO4, 5, 6 are upregulated which is consistent with a role in LD formation as found in other CLO/PXG proteins in other plant and algal species. The overall differentially regulated patterns of CLO/PXG-like gene expression in oil palm supports the conclusion that the encoded proteins carry out a range of roles in various tissues throughout the plant and that these roles occur at different developmental stages.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Transcriptome analysis of CLO/PXG gene expression in date palm and oil palm tissues.

(A) The oil palm genome P5-build was used to read map the RNA seq data from Roche 454 reads (a full dataset is available from NCBI BioProject PRJNA201497). Reads from Roche/454-derived libraries were assembled into isotigs, which were blasted onto Arabidopsis thaliana gene models with a threshold of E-value < 10−5. The best-hit A. thaliana gene model was assigned to the homologue of the query isotig. To estimate expression levels of genes in mesocarp and kernel tissues, Illumina HiSeq 2000 reads from each library were mapped to assembled isotigs from all Elaeis guineensis reads by using the Burrows–Wheeler Aligner. Gene group expression levels were calculated as the number of mapped reads on each isotig divided by the total number of isotigs, multiplied by 100,000, and scaled by the number of genes in each gene group. Both copy number and read coverage were the mean of measures from two biological replicates. Data were analysed as described above for Roche/454 data, except that expression levels were calculated as transcripts per million tags. Identification of expression caleosin in oil palm was done using open source Tuxedo suite software[50]. (B) Transcriptional analysis of CLO/PXG gene expression in date palm tissues and treatments as follows: a) exposure to 0, 10 and 100 ng.L^-1 of the dioxin, TCDD; b) drought for 2, 4 and 6 days; c) exposure to 0, 150 and 300 ng.L^-1 NaCl. Seedlings with a radicle length of 0.5 or 2 or 4.5 cm were referred as stage I, II and III respectively.

https://doi.org/10.1371/journal.pone.0196669.g004

The date palm genome contains five CLO/PXG-like sequences but only three of these genes, PdCLO2, 3 and 4, were expressed (Fig 4B) at moderate to high levels following the various treatments while PdCLO1 and 5 were expressed at much lower levels in all cases. Exposure to the hydrophobic organic toxin, dioxin, led to a 20-fold upregulation of PdCLO2 and 3, while PdCLO3 and 4 were transiently upregulated by >30-fold following drought stress and were also >20-fold upregulated following increasing exposure to NaCl. Expression of PdCLO4 gene was higher in the young emerging shoot (plumule) than in radicle while PdCLO2 was more expressed in radicle tissues than in plumule and very small levels of transcripts for both genes were found in petioles [74]. This expression pattern, with a strong upregulation in response to various abiotic stresses, is similar to that on other plants. It is possible that the less expressed genes PdCLO1 and 5 have different roles, such as in abiotic stress or seed development, that were not tested in this transcriptome panel.

Evolution of the CLO/PXG gene family in the Viridiplantae

The occurrence of CLO/PXG-like genes is depicted in the evolutionary tree shown in Fig 5. In all currently sequences Archaeplastida, CLO/PXG-like genes are only found in the Viridiplantae, which is one of four taxa that make up this group. The Archaeplastida (Plantae) are widely, albeit not universally, recognised as a monophyletic group of photosynthetic organisms descended from an endosymbiotic association between a heterotrophic eukaryote and a cyanobacterium that took place >2 Bya [75]. This supergroup includes all extant red and green algae plus the land plants [76]. It is now recognised that several other groups of non-Archaeplastidal photosynthetic organisms, most notably the Stramenopiles, which include diatoms and brown algae, originated separately from non-plant/algal ancestors that secondarily acquired red or green algal endosymbionts [77]. These organisms include diatoms, the genomes of which definitely lack CLO/PXG orthologs, although some species have recently been shown to contain a very different class of LD-binding proteins that appear to play analogous roles to CLO/PXG in LD accumulation in response to stress [78–80].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Presence of CLO/PXG sequences across the Viridiplantae and their estimated evolutionary divergence times.

The major taxa that contain CLO/PXG sequences are shown as blue-shaded boxes. Individual species with one or more CLO/PXG sequences are shown in blue while other species where CLO/PXG sequences are definitely absent from their genomes are shown in brown. The estimated evolutionary divergence times of selected key taxa are shown as the number of million years ago (My). Starred major taxa are those with good evidence of monophyletic status while non-starred taxa are probably polyphyletic or paraphyletic.

https://doi.org/10.1371/journal.pone.0196669.g005

The two major groups that make up the Archaeplastida are the Viridiplantae (from Latin for ‘green plants’), and the Rhodophyta (red algae) plus a third much smaller group, the Glaucophyta [81]. These groups of photosynthetic organisms are estimated to have diverged from each other between 1–1.6 Bya [6, 82]. The Rhodophyta crown group diverged relatively early, at 1.0–1.6 Bya [75] but the earliest divergence times of the major two clades of the Viridiplantae, the Chlorophyta and the Streptophyta, both of which are generally but not universally recognised as monophyletic taxa, is less certain [6, 82]. While most estimates of divergence times of these two groups are in the range of 0.85–1.2 Bya [6, 77, 83], a more recent study suggests a later divergence at about 0.5–1.0 Bya [75]. Although molecular clocks place the origins of the Archaeplastida at between 0.9 and 1.9 Bya [75], evidence from recent molecular fossil data suggest that Cyanobacteria remained the dominant group of photosynthetic organisms until as recently as about 650 Mya, after which the eukaryotic Archaeplastida rapidly emerged as major components of aquatic, and later terrestrial, ecosystems [84].

The CLO/PXG gene family in the Chlorophyte algae

Since CLO/PXG-like genes are widely distributed in both Chlorophyta and Streptophyta, the gene family probably dates from >1 Bya in a common ancestor of the entire Viridiplantae taxon. Interestingly, it is now emerging that at this stage of algal evolution many of the Viridiplantae species, including all of those that express CLO/PXG-like genes, began to occupy low-salt, i.e. freshwater, habitats [75]. Given the virtually universal upregulation of CLO/PXG-like genes in response to salt stress [14, 17, 85], it is possible that one of the primary reasons for the evolution of CLO/PXG-like genes is related to a move towards non-saline environments. One of the characteristic features of higher plant CLO/PXGs is the occurrence of both H- and L- isoforms [12, 40, 64, 86] and this was found in at least one of the Chlorophyte species, Auxenochlorella protothecoides (Fig 1B), which indicates that CLO/PXG-like genes had already diverged into the two isoforms at a very early stage of their evolution. Although genome sequence data are still much less complete in green algae compared with land plants, the available information enables us to trace the presence, and sometimes the absence of CLO/PXG-like genes in a way that allows some evolutionary inferences to be made.

The Chlorophyta are conventionally divided into four groups, namely Trebouxophyceae, Chlorophyceae, Ulvaphyceae and Prasinophyceae. There is good evidence for the Chlorophyceae being a monophyletic clade while the other three groups are highly paraphyletic[5, 83]. CLO/PXG-like genes are present in at least four species of Chlorophyceae (Chlamydomonas reinhardtii, Gonium pectorale, Monoraphidium neglectum and Volvox carteri) and four species of Trebouxophyceae (Auxenochlorella protothecoides, Chlorella variabilis, C. vulgaris and Coccomyxa subellipsoidea). Furthermore, in at least five of these algal species the biological activity of these genes is supported by transcriptome and/or biochemical studies [14, 20, 44, 87]. However, CLO/PXG-like genes appear to be absent from several other sequenced genomes in both the Chlorophyceae (e.g. Dunaliella and Haematococcus spp.) and Trebouxophyceae (e.g. Lobosphaera spp.) [88, 89]. To date no genomes of the Ulvaphyceae have been sequenced and at present only a few complete genome sequences are available from the Prasinophyceae (e.g. Ostreococcus tauri, Micromonas spp. and Bathycoccus prasinos), none of which contain CLO/PXG-like genes. However, this does not necessarily mean that the CLO/PXG gene family is absent from other unsequenced members of these taxa. Overall, the most parsimonious interpretation of the currently available data is that CLO/PXG-like genes were originally present in the ancestor of the putatively monophyletic [83] Chlorophyta clade. Similar genes are still present in several of the present day descendants of these ancestral Chlorophyta, most notably in some, but not all, Chlorophyceae and Trebouxiophyceae. However the CLO/PXG genes appear to have been lost in other Chlorophyta species during the >1 billion years of their subsequent evolution.

The CLO/PXG gene family in the Streptophyte algae

The Streptophytes, including the Charophytes and Embryophytes are a monophyletic clade that diverged from the Chlorophytes >1.0 Bya [90, 91]. To date, full genomic sequence data are only available for one species of the Charophyta, namely the terrestrial and freshwater filamentous alga, Klebsormidium nitens, where as many as five CLO/PXG-like genes are also present. In addition, transcriptome data from the related species, K. crenulatum show the desiccation-induced expression of no fewer than six CLO/PXG-like genes [92]. Klebsormidium algae have primitive body plans and are made up of multicellular non-branching filaments that can survive on the land with substantial tolerance to novel stresses, such as drought and freezing, which do not normally occur in aquatic environments [93]. Therefore it is possible that as some Streptophyte algae became increasingly adapted to terrestrial conditions, new functions emerged for the CLO/PXG proteins inherited from their marine ancestors. Given the ubiquity of the occurrence of CLO/PXG-like genes in the Embryophytes (see below), it seems likely that similar genes will be found in other Charophyte species once more sequence data become available in the future.

The CLO/PXG gene family in the Embryophytes

Embryophytes are a monophyletic group probably diverged from Charophytes about 500 Mya and there is robust fossil evidence of their presence in a variety of terrestrial environments by 430–490 Mya [4, 94–98]. The most basal groups of Embryophytes are the Liverworts and Bryophytes (mosses) each of which is represented by just one sequenced genome, namely Marchantia polymorpha and Physcomitrella patens. In both cases their genomes contain only L isoform CLO/PXG-like sequences, suggesting that the H isoform has been lost in these relatively primitive non-vascular, multicellular plants. However, both L and H CLO/PXG isoforms are present in lycophyte, Selaginella moellendorfii, which is a fern-like seedless vascular plant that can form true roots and has the kind of ABA signalling pathways for which the A. thaliana CLO/PXG isoform, ATCLO4, acts as a negative regulator [98, 99]. Although no true fern genomes have been sequenced, such data are available for four species of Gymnosperm, which are seed-bearing vascular, but non-flowering, higher plants. In all cases, only H isoform CLO/PXG are present in these genomes (Fig 1B). Therefore, in the case of this admittedly small sample of non-flowering land plants, most species contain only one of these two CLO/PXG isoforms.

To date CLO/PXG-like genes have been found in all of the >300 angiosperm species for which sequence data are available. The split between Angiosperms and other vascular plants occurred >200 Mya [100] and the former now make up the majority of the terrestrial flora in terms of biomass and species diversity[101, 102]. In all of these species our analysis suggests that there is at least one copy each of the L and H isoforms of CLO/PXG and that in many cases these genes have become highly duplicated to form relatively large families of CLO/PXG-like genes, many of which are differentially expressed in various developmental and/or environmental conditions. We and others have previously shown that in a range of higher plants including Arabidopsis, maize, and rice, some CLO genes are induced by a variety of biotic and abiotic stresses (e.g. RD20, AtPXG1, AtPXG2, EFA27, OsCLO-2, OsCLO-6, ZmCLO7) while other CLO genes are unresponsive to such stresses but may be highly expressed in lipid-storing tissues such as seeds [8, 28, 69, 70]. It is also possible to divide higher plant CLO/PXGs into two functional groups, one of which has very low epoxygenase activity when expressed under standard conditions and tends to be seed specific, while the other has high epoxygenase activity and is expressed throughout the plant[28]. In many well-characterised higher plant genomes there are large numbers of CLO/PXG-like genes, such as six in Arabidopsis, 11 in the diploid Brassicas, nine in rice and 12 in maize. These sequences include highly conserved regions also found in algae, plus more variable regions some of which may have arisen after the monocot/dicot divergence of about 150–160 Mya [13]. In some cases the annotated CLO/PXG sequences appear to be pseudogenes as they are not expressed and in some cases lack crucial CLO domains and may also contain additional non-CLO domains. An example of the latter is the presence of bZIP and PKinase domains in the non-expressed putative maize sequences, ZmCLO1 and ZmCLO2a [13]and in an Arabidopsis CLO sequence [103].

The expansion and increasing complexity of CLO/PXG-like gene families in land plants is consistent with the evolution of a wider range of functions by the various CLO/PXG protein isoforms. Some of the functions relating to nutrient stress responses and LD packaging are also seen even in unicellular green algae and are probably universal for CLO/PXGs [43, 87, 104]. However, other functions such as oxylipin-based signalling [105, 106] are unique to multicellular organisms and probably evolved later, while other peroxygenase activities are related to the production of extracellular waxes such as cutin [107–109] and desiccation tolerance [35, 92, 110] so these functions would not have been required until plants became terrestrial. It is also becoming evident that CLO/PXGs have central roles in plant-pathogen responses, especially with fungi where the proteins may be involved in both host and pathogen crosstalk via oxylipin pathways [8, 9, 15, 98, 111]. In summary, the evolution of the CLO/PXG gene family in the Viridiplantae involves over one billion years of gradually expanding functions as these organisms increased in size and complexity and had to adapt to new forms of biotic and abiotic stress as they colonised new habitats and faced new challenges both in the sea and on land.

Putative CLO/PXG genes/proteins in a bacterium and a basal opisthokont

We found that public databases contain several DNA sequences with weak to moderate similarity to CLO/PXG in other major taxa apart from Viridiplantae or Fungi. In the majority of cases these can be dismissed as being potential functional members of the CLO/PXG family because the derived proteins lack canonical domains required for biological functions, such as calcium-, haem- or lipid-binding motifs. However, there are several database entries that do contain the major CLO/PXG motifs and these may be non-orthologous versions, i.e. xenologs, of the well-established Viridiplantae and Fungal sequences [112]. In particular, there are several annotated putative CLO/PXG-like sequences in the genomes of the myxobacterium, Sorangium cellulosum, and the basal opisthokont, Capsaspora owczarzaki. As shown in Fig 6A, 6B and 6C, the derived protein sequences from these two organisms contain a calcium-binding EF hand motif plus the two invariant histidine residues involved in heme binding and a lipid-binding domain, all of which are in the same locations in the protein as in the well established plant and fungal sequences. There are additional smaller regions of sequence similarity, including in putative kinase domains, which may indicate that these proteins may be regulated by phosphorylation and the predicted proteins are also of similar length to most verified CLO/PXGs. One notable absence in the S. cellulosum and C. owczarzaki sequences is the H-domain, a motif that is found close to the N terminus of many, but not all, algal and land plant CLO/PXGs (see Fig 6B). This may indicate that if these sequences were acquired via horizontal gene transfer (HGT) then they originated as plant-derived L-caleosins.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Alignments of anomalous CLO/PXG sequences from two non-Viridiplantae species.

(A) Alignments of seven CLO/PXG sequences from S. cellulosum. (B) Alignments of two CLO/PXG sequences from C. owczarzaki with a range of green algal sequences. (C) Alignments of five putative CLO/PXG sequences from the Metazoan nematode genus, Panagrolaimus spp, with sequences from A. thaliana.

https://doi.org/10.1371/journal.pone.0196669.g006

Interestingly, no CLO/PXG-like sequences were found in the genomes of any of the close relatives of either S. cellulosum, C. owczarzaki or Panagrolaimus spp for which data are available. This may indicate that the genes were originally acquired as isolated instances of HGT and then became duplicated and were retained as multiple-copy gene families due to their usefulness for the particular lifestyles of these organisms. For example, S. cellulosum

Soce56 is a highly unusual aerobic myxobacterium that has the largest bacterial genome sequenced to date at 14.8 Mb [113]. The genome of this bacterium has undergone massive expansion due to gene endo-duplication and HGT, and its lifestyle involves high levels of social behaviour and production of a wide range of secondary metabolites with complex regulatory networks including many kinases that facilitate responses to rapidly fluctuating environments [113]. With their known roles in stress responses, signal transduction, and oxylipin metabolism it can be seen why the presence of as many as seven CLO/PXG genes might be adaptive for S. cellulosum.

In the case of the nematode genus, Panagrolaimus spp, a BLAST search with the five putative CLO/PXG sequences gave both higher plants and fungal CLO/PXGs in the top few hits. Therefore, if the original gene was acquired via HGT, as seems likely it is not clear whether the donor was a plant or a fungus. One interesting feature of the putative CLO/PXG genes from the nematodes is that they were only found in parthenogenetic species in the genus and are functionally linked to cryptobiosis and especially to desiccation tolerance [114]. As discussed above, the algal caleosins are highly upregulated following salt stress and may also have played a role in the transition of more complex plants from aquatic to terrestrial environments with the concomitant requirement for improved tolerance to desiccation.

A third well-supported putative CLO/PXG gene family is present in another unusual organism, namely the eukaryote, C. owczarzaki, which is a filose amoeboid symbiont of the pulmonate snail, Biomphalaria glabrata. Genomic sequence data suggest that C. owczarzaki is a member of an opisthokont lineage, the Holozoa, which is more closely related to the Metazoa than to the other major multicellular opisthokont group, the Fungi [115]. Interestingly, during specific phases of its life cycle, C. owczarzaki cells accumulate and then extrude LDs, although it is not known whether these extracellular LDs contain CLO/PXGs as is found in several algae (see above) [116, 117].

A less likely candidate CLO/PXG protein has been reported in the dinoflagellate, Symbiodinium [43] In this case an LD-associated 20kDa protein in Symbiodinium cells cross-reacted with anti-CLO/PXG antibodies generated against purified sesame and cycad proteins, which could possibly imply the presence of a CLO/PXG protein in this species. However, Symbiodinium, which is a photosynthetic endosymbiont within the reef building coral, Euphyllia glabrescens is a dinoflagellate that is unrelated to either the Viridiplantae or Fungi, having acquired its plastids from red algae, which do not have CLO/PXG genes (see above). Therefore, in the absence of definitive sequence data for the 20kDa LD protein or its gene, it cannot be assumed to be a bona fide CLO/PXG.

Finally, we found a few isolated database entries of CLO/PXG-like sequences in genomes of taxa from which these genes are thought to be absent, most notably the Metazoa. For example, we found a CLO/PXG-like sequence with very strong identity to higher plants in the genome entry for the Gulf Coast tick, Amblyomma maculatum (GenBank: AEO32378.1). These ticks are parasites of small mammals and birds. The entry was from a shotgun transcriptome assembly that had involved the collection of several hundred million sequences from ticks kept in culture. Correspondence with the lab concerned established that the ticks used in the study also harboured some algal growths on their external surfaces and that this may have been the source of the CLO/PXG-like sequence. It can be concluded that this particular CLO/PXG-like database entry is highly likely to be erroneous and due to contamination from plant or algal sequences rather than to HGT. The take-home messages from the two examples of Symbiodinium and A. maculatum, are that a) considerable caution should be exercised in the use of antibody cross reactivity in the absence of sequence data and b) anomalous sequence entries in databases should be carefully scrutinised before definite conclusions are drawn.