The genusPerkinsus occupies a precarious phylogenetic position. To gain a better understanding of the relationship between perkinsids, dinoflagellates and other alveolates, we analyzed the nuclear-encoded spliced-leader (SL) RNA and mitochondrial genes, intron prevalence, and multi-protein phylogenies. In contrast to the canonical 22-nt SL found in dinoflagellates (DinoSL),P. marinus has a shorter (21-nt) and a longer (22-nt) SL with slightly different sequences than DinoSL. The major SL RNA transcripts range in size between 80–83 nt inP. marinus, and ∼83 nt inP. chesapeaki, significantly larger than the typical ≤56-nt dinoflagellate SL RNA. In most of the phylogenetic trees based on 41 predicted protein sequences,P. marinus branched at the base of the dinoflagellate clade that included the ancient taxaOxyrrhis andAmoebophrya, sister to the clade of apicomplexans, and in some cases clustered with apicomplexans as a sister to the dinoflagellate clade. Of 104Perkinsus spp. genes examined 69.2% had introns, a higher intron prevalence than in dinoflagellates. Examination ofPerkinsus spp. mitochondrial cytochrome B and cytochrome C oxidase subunit I genes and their cDNAs revealed no mRNA editing, but these transcripts can only be translated when frameshifts are introduced at every AGG and CCC codon as if AGGY codes for glycine and CCCCU for proline. These results, along with the presence of the numerous uncharacterized ‘marine alveolate group I' andPerkinsus-like lineages separating perkinsids from core dinoflagellates, expand support for the affiliation of the genusPerkinsus with an independent lineage (Perkinsozoa) positioned between the phyla of Apicomplexa and Dinoflagellata.

Perkinsus marinus and dinoflagellate cultures, RNA isolation and cDNA construction

Perkinsus marinus isolate ATCC 50439 andP. chesapeaki ATCC PRA-65 were grown in tissue culture flasks with liquid media, samples (3–4×10⁶ cells) were collected by centrifugation and total RNAs were isolated as reported previously[14]. DinoflagellatesAmphidinium carterae (CCMP1314) andKarlodinium veneficum (CCMP2778) were grown in f/2 seawater medium at 20°C at a 12 h∶12 h light∶dark photocycle with a photon flux of approximately 50 µE·m⁻²s⁻¹. When the cultures were in the exponential growth phase, ∼1×10⁶ cells were harvested and total RNAs isolated according to Zhang et al.[8]. These RNAs were used for cDNA synthesis as described previously[8].

Identification of the SL RNA genes from theP. marinus genome project

Perkinsus spp. were suspected to possess a SL sequence similar to that of dinoflagellates (DinoSL;[8]). Two types of SL sequences were detected at the 5′ end ofP. marinus full-length cDNAs ofpcna andcyclins[14], PmaSL1,5′-ACCGTAGCCATCTTGGCTCAAG-3′ (22 nt) and PmaSL2,5′-ACCGTAGCCATCTGGCTCAAG-3′ (21 nt). These twoPerkinsus SL sequences were used to queryP. marinus whole-genome shotgun reads [http://www.ncbi.nlm.nih.gov/genomeprj/46451] to identify SL RNA genes. For hits with 85–100% identity to the queries, the genome sequences were collected for alignment with one another and with SL RNAs from dinoflagellates. Type-specific primers were designed for amplifying the putative SL RNAs (Table 1).

Download:

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

Table 1. SL RNA-related oligonucleotides used in this study.

https://doi.org/10.1371/journal.pone.0019933.t001

RNA blot analyses of SL RNA

Total RNA from ∼10⁶ cells of bothPerkinsus species and four strains of dinoflagellates in our previous studies[8],[9], includingProrocentrum minimum (CCMP696),Polarella glacialis (CCMP2088),Karenia brevis (CCMP2228) andKarlodinium veneficum (CCMP1975) were used for RNA blots. RNA samples were loaded onto an 8% acrylamide/8 M urea gel, a medium resolution gel optimal for RNAs below 350 nt, electrophoresed, and transferred to nylon membranes[22]. Oligonucleotide probes used for hybridization included dinoSLa/s for detection of the general dinoflagellate SL RNAs and the two types ofPerkinsus SL RNAs (PmaSL-La/s and PmaSL-Sa/s hybridizing to exons and PmaSL-Li and PmaSL-Si to introns) (Table 1). The cDNA clones containing the twoP. marinus SL RNAs were dot blotted to serve as positive controls for detection of the specific substrate SLs on RNA blots. Total RNA fromLeishmania tarentolae cells was included to provide size markers. Oligonucleotide probes were labeled with γP³²-ATP for hybridization[22].

Rapid amplification of cDNA 3′ end (3′ RACE) and folding analysis

Poly (A) mRNA was depleted fromP. marinus total RNA and a poly (A) tail was added to the remaining population usingEscherichia coli Poly (A) Polymerase (Takara Mirus Bio) as reported[8]. First-strand cDNA synthesized using GeneRacer Oligo dT primer (Invitrogen) was used as PCR template. Two rounds of touch-down PCR were carried using the same conditions as above, with the extension time of 5 sec at 72°C. The first round of PCR was performed using PmaSL-LSF1 and GeneRacer3 primers. The PCR products were diluted 100-fold and used in the second round PCR with PmaSL-LSF2, PmaSL-LNF2, PmaSL-LNF3, PmaSL-S2F2, or PmaSL-S2F3, each paired with GeneRacer3, as the nested primers (Table 1).

Structures were modeled for the two dominant types of SL RNA transcripts using the MFOLD online program [http://mobyle.pasteur.fr/cgi-bin/MobylePortal/portal.py?form=mfold]. Folding was performed using the default setting except that the temperature was set at 27°C to match theP. marinus culture conditions.

Mitochondrial gene analyses

The mtcox1 andcob sequences were PCR-amplified from both genomic and cDNA templates using universal andPerkinsus-specific primers designed in this (Table 2) and previous studies[12],[23],[24]. PCR was performed with 30 cycles of 95°C for 15 sec, 50–58°C for 30 sec, and 72°C for 40 sec. PCR products were sequenced either directly or after cloning into a T-vector, with 5–10 clones randomly chosen for sequencing. To obtain the ends of the mt genes, we designedPerkinsus-specific primers for bothP. marinus andP. chesapeaki (Table 2) based on the mtcox1 and the partialcob sequences obtained from the newly releasedP. marinus genome shotgun sequence.

Download:

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

Table 2. Mitochondrial gene primers designed in this study.

https://doi.org/10.1371/journal.pone.0019933.t002

Generation of full-length gene sequences

Ribosomal proteins (RPs) from dinoflagellates[21] were used to query theP. marinus genome and GenBank databases to retrieve RPs fromP. marinus, apicomplexans, ciliates, diatoms and other eukaryotic representatives. Since many of the dinoflagellate RP cDNAs available were not full-length, to maximize phylogenetic information from these genes, 22 full-length cDNAs of RPs from dinoflagellatesAmphidinium carterae CCMP1314 andK. veneficum CCMP2778 were cloned using dinoflagellate-specific SL coupled with 3′ RACE as described previously ([8]; GenBank accession # GU372975-GU373034). To diversify the gene markers for phylogenetic analyses, another 12 conserved gene sequences were collected from our ongoing cDNA sequencing project for these two species, and their 5′ and 3′ ends achieved using RACE as necessary. Using these as queries, homologs were collected from GenBank forP. marinus and other species mentioned above. The absence of histones, long considered a benchmark of typical dinoflagellates, is erroneous (see[21] for review); thus, histone genes were retrieved from thePerkinsus genome project database. Full-length or nearly full-length mtcox1 andcob sequences were also obtained fromP. marinus andP. chesapeaki. The 3′ end ofcob for bothPerkinsus spp. was obtained using the 3′ RACE technique withPerkinsus cob primers paired with GeneRacer3 primer (Invitrogen). All of these genes were used in phylogenetic analyses.

Multi-protein phylogenies

Predicted aa sequences of each gene were aligned with homologs from related organisms using CLUSTAL W (1.8) and inspected manually. Phylogenetic relationships ofP. marinus with alveolate relatives and other eukaryotes were inferred using Neighbor Joining (NJ), Maximum Likelihood (ML), and MrBayes (MB) analyses. NJ analysis was performed online [http://clustalw.ddbj.nig.ac.jp/top-e.html] with the default setting. For ML tree reconstruction, the datasets were run through ProtTest[25] to identify the best-fitting aa substitution models (Table 3), which were then employed in the phylogenetic analysis using PhymLv3.0[26]. MB analysis was carried out with 20,000–1,000,000 MCMC generations depending on when the average standard deviation of split frequencies reached below 0.01, a tree sampling frequency of 10–100, and 25% of generations discarded as burn-in[27]. To verify the reliability of the tree topologies, branch support was estimated based on bootstrap (1,000 resamplings) in NJ, approximate Likelihood Ratio Test (aLRT) in ML, and posterior probability in MB.

Download:

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

Table 3. Best substitution models selected by ProtTest for phylogenetic analyses.

https://doi.org/10.1371/journal.pone.0019933.t003

Analysis of Intron Frequency

Thirty-six and 37 unique full-length cDNAs fromP. marinus andP. chesapeaki, respectively[8], were used as the queries to nBLAST-search againstP. marinus genomic sequences to obtain the corresponding genomic DNA. The recently published full-length cDNAs and genomic DNAs for proliferating cell nuclear antigen (pcna) and two types of cyclins fromP. marinus[14], as well as 36 other common protein-coding genes ofP. marinus such as tubulins,gapdh, centrin,hsp90 and ribosomal proteins reported in GenBank were compared (Table S1). Canonical GT/AG intron/exon boundaries validated the deduced intron start and end positions. The percentage of genes within this cohort that contained introns was determined.

Two major types ofPerkinsus SL RNA

From the reportedP. marinus genome database we identified two major types of SL RNA genes: PmaSLRNA-L or L-type, and PmaSLRNA-S or S-type (Figure 1A), with the SL exons corresponding to the two SL sequences found previously inpcna andcyclins[14]. These sequences were similar to DinoSL (Figure 1B). For the L-type, we identified seven sequences (Table 3), and all but one (AAXJ01000089, containing two units of SL RNA) are 1–1.8 kb in length containing a single SL RNA gene. For the S-type, 42 sequences were identified with lengths ranging 1 to >14 kb (Table 4); of these, some were arrayed as tandem repeats or as a single unit clustered with both or either of the U2 and U4 snRNA genes downstream of the SL RNA gene; others were single or 2-unit tandem-repeat sequences not associated with U2 or U4 snRNA genes (Table 4).

Download:

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

Figure 1. Alignments of spliced leader (SL) RNAs ofPerkinsus marinus and dinoflagellates.

A) Representative genomic sequences of two types ofP. marinus SL RNA. B)P. marinus SL RNAs with the reported representatives of dinoflagellate SL RNA genomic sequences (modified according to[9]; the number of identical clones retrieved for each type is indicated by “@number” following the species abbreviation and type number). The SL region (boxed) is shown in uppercase letter, intron and the flanking regions are shown in lowercase letters, * indicates the conserved nucleotide (nt). The first ‘A’ of SL is numbered as nt 1. SL RNA transcripts mapped by 3′ RACE analyses are denoted by arrows to indicate the terminal positions, thickness with darkness of the arrows denote relative frequency of clones that ends where it is indicated. Note that the PCR-amplifiedAmoebophrya sp. genomic sequences contain only one unit of SL RNA gene, the partial SL sequence is of the primer used. Per,P. marinus, Amo,Amoebophrya sp.; Har,Heterocapsa arctica; Kbr,Karenia brevis; Kve,Karlodinium veneficum; Ppi,Pfiesteria piscicida; Pgl,Polarella glacialis; Pmi,Prorocentrum minimum. SL refers to SL RNA sequences obtained from SL-only repeats; SL-5S indicates SL RNA sequences from genes associated with 5S rRNA genes. *: sequences from[8]; **: sequence from[46]; #: sequences from[9], $1-4: GQ178071-GQ178074; •: sequences missing in the original reports. Shaded are conserved positions defined as identical in over six sequences in at least three species. A non-canonical C in the splice donor site of KbrSL-3 is boxed. Gaps introduced in the sequence alignment are shown as ‘–’.

https://doi.org/10.1371/journal.pone.0019933.g001

Download:

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

Table 4. Genomic sequences containing SL RNA genes identified fromP. marinus genome data.

https://doi.org/10.1371/journal.pone.0019933.t004

The majorP. marinus SL RNA transcripts are 80–83 nt

The sequences containing the two types ofP. marinus SL RNA genes (PmaSLRNAs) were conserved in the first 82–83 nt, with the SL exon of the L type 1-nt longer than that of the S type. Sequence similarity diminished in the downstream intron region. The sequence upstream of SL was more complex: for the L-type PmaSLRNAs, upstream sequences were uniform, whereas those of the S type were diverse, with some resembling the L type (Figure 1A). When PmaSLRNAs were aligned with the representatives of known dinoflagellate SL RNAs, PmaSLRNAs showed similarity in the exon (i.e. the 21/22-nt SL region) and moderate similarity in the beginning of the intron region (i.e., immediately downstream;Figure 1B). As in dinoflagellates, the predicted Sm-binding sequence was located in the SL exon of PmaSLRNAs, and the 3′ termini of the majority of substrate transcripts mapped within poly-T tracts, reminiscent of the termination element in SL RNAs of kinetoplastid[22], some dinoflagellates[9], and of other small RNA genes.

The SL RNAs of twoPerkinsus spp. and four dinoflagellates were analyzed by gel electrophoresis and hybridization. Ethidium bromide staining revealed that the twoPerkinsus species have similar small RNA molecule profiles with commonalities to the dinoflagellateP. minimum (Figure 2A). Hybridization of an RNA blot of this gel with the 19-nt dinoflagellate SL probe DinoSLa/s (including 14 nt of SL and 5 nt of intron;Table 1) showed the dinoflagellate SL RNA pattern with major transcripts of <56 nt for the four dinoflagellates as reported previously[8],[9]; no hybridization was detected for the twoPerkinsus species (Figure 2B). Probing the blot withP. marinus L-type or S-type SL probes (PmaSL-La/s and PmaSL-Sa/s respectively;Table 1), strong bands of >72 nt appeared in bothPerkinsus species for both probes, with a minor band of slightly shorter length in theP. marinus sample for probe PmaSL-Sa/s; neither probe hybridized to dinoflagellate SL RNA (Figure 2C, 2D), indicating that the >72-nt bands are specific to the genusPerkinsus, and thatPerkinsus SL RNAs are longer than those of typical dinoflagellates. Consistent with the similar RNA levels seen on the gel for the twoPerkinsus species, probe PmaSL-La/s detected equivalent levels of this SL RNA variant (Figure 2C) in the two species. However, the band ofP. chesapeaki was weaker than that ofP. marinus with probe PmaSL-Sa/s (Figure 2D), possibly reflecting reduced expression or impaired hybridization due to a nucleotide alteration(s) in the exon region inP. chesapeaki. The minor band in theP. marinus lane may represent degraded SL RNA products. To further distinguish the two types of PmaSL RNA transcripts and to explore whetherP. chesapeaki SL RNAs have similar introns to those ofP. marinus, additional probes were designed for the PmaSLRNA L-type and S-type intron sequences (PmaSL-Li and PmaSL-Si;Table 1). Both intron probes revealed bands at >72 nt and some minor bands of <72 nt inP. marinus (Figure 2E, 2F), but no bands inP. chesapeaki, suggesting thatP. chesapeaki SL RNAs have different intron sequences thanP. marinus. An additional band appeared at ∼150 nt with PmaSL-Si for bothPerkinsus spp. (Figure 2F), a likely result of non-specific hybridization to the abundant 5.8S ribosomal RNA (Figure 2A). To validate the specificity of the probes, 3′ RACE cDNA clones of the L- and S-type SL RNA were used to create dot blots that were hybridized separately with each probe. Each yielded a positive signal only when the corresponding probe was used (Figure 2G, 2H).

Download:

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

Figure 2.Perkinsus spp. substrate SL RNA larger than typical of dinoflagellates.

A) Denaturing 8% polyacrylamide/8 M urea gel of total cell RNA fromPerkinsus spp. and other organisms. Lane 1,Leishmania tarentolae; 2,P. chesapeaki; 3,P. marinus; 4,Prorocentrum minimum; 5,Polarella glacialis; 6,Karenia brevis and 7,Karlodinium veneficum. B–F) Probing of the blot shown in A) using oligonucleotides DinoSLa/s, PmaSL-La/s, PmaSL-Sa/s, and designed from intron regions in the PmaSL-L and PmaSL-S genotypes, respectively. Arrows highlight the SL RNA transcripts. G) and H) Dot blots of the PmaSLRNA-L (G) and PmaSLRNA-S (H) cDNA clones using the same probes as in (E) and (F), respectively.

https://doi.org/10.1371/journal.pone.0019933.g002

A 3′ RACE analysis gave an assortment of 3′ ends for both PmaSLRNAs. Of the 48 PmaSLRNA cDNA clones mapped, 25 ended at the 2^nd T, 11 clones at the 1^st T, and 4 clones ended at the 3^rd T of the poly-T tracts present in both SL genes, representing 83% of the ends obtained. Thus, most PmaSLRNA transcripts were 80–83 nt in length, corresponding to the major band observed in the RNA blots. The minor end classes of <72 nt may have contributed to the minor products seen by RNA blotting, possibly representing degraded or misprocessed SL RNA products.

PmaSL present in protein coding genes and other genomic locations

BLAST analysis using PmaSL1 and PmaSL2 hit some cDNA or genomic DNA sequences apparently coding for proteins (e.g. EH076923, EH059894, EH059894, EH059894). In addition, over 100 genomic sequences were retrieved from the genome data that contained recognizable PmaSL1 (>60, e.g. AAXJ01000048, AAXJ01000335, AAXJ01000111, AAXJ0100359, AAXJ01004662, AAXJ01000077) and PmaSL2 (>40, e.g. AAXJ01000111, AAXJ01000162, AAXJ01000192, AAXJ01000237, AAXJ01000370) but no recognizable intron downstream. Most of these SL sequences started with T, and were arrayed in tandem repeats, and their downstream regions were variable. To investigate whether those SL RNA-like genomic sequences were also expressed, we designed primers (Table 1) containing a partial SL and downstream nucleotides or the downstream sequences alone and applied them to 3′ RACE and RNA blotting analyses. Neither of the approaches yielded clear products, indicating that these SL-like sequences are not functional SL RNA genes.

Predicted PmaSLRNA structures and Sm-binding site locale

Similar to the situation in dinoflagellates, no apparent Sm-binding site sequence was found in the predicted intron regions of either of the PmaSLRNAs. Instead, AUUCUGG (L-type) or AUCUGG (S-type) found within the SL was the only recognizable candidate Sm-binding site, as in the DinoSL (AUUUUGG). The predicted intron region was similar between the two PmaSLRNAs, in contrast to the conserved intron in DinoSL RNAs, with the exception of the ancient parasitic genus of dinoflagellatesAmoebophrya that showed considerable variation (Figure 1B). In the structural simulation using the default conditions for all but temperature, which was adjusted to the culture temperature of 27°C, the splice-donor dinucleotide (‘gu’ in ‘Gguag’) was double-stranded and the putative Sm-binding site (AUUCUGG/AUCUGG) single-stranded, forming a small terminal loop. The simulation yielded one comparable structure for both types of PmaSLRNAs (Figure 3). The predicted structures were similar to typical dinoflagellate SL RNA structures, having two stem-loops[8],[9], with the ‘extra’ intron region situated in a bulge of unpaired sequence connecting the two stem loops.

Download:

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

Figure 3.Perkinsus marinus SL RNA secondary structure similar to DinoSL RNA.

Predicted structures of SL RNA forP. marinus L-type (A) and S-type (B) based on the most abundant cDNAs obtained. Model simulation was run using MFOLD: Prediction of RNA secondary structure modeling program (http://bioweb.pasteur.fr/seqanal/interfaces/mfold-simple.html) under default settings except that the folding temperature was set at 27°C, the culture temperature.

https://doi.org/10.1371/journal.pone.0019933.g003

Unique sequences and anomalous frameshifts inPerkinsus mt genes

All the possible combinations forcob primers designed based on dinoflagellatecob (Table 2;[12],[23],[24]) were tested but failed to PCR amplify any products. BLAST searching usingcob aa sequences from apicomplexans and dinoflagellates against theP. marinus whole genome shotgun sequencing database (tblastx) hit one contig (860 bp, AAXJ01022806) containing the 5′ end of acob-like sequence. The corresponding mRNA of this sequence and its 3′ end were obtained for both species ofPerkinsus by PCR and 3′ RACE usingPerkinsus-specific primers paired with the GeneRacer3 primer (GenBank accession numbers HQ670239, HQ670241;Figure 4,Table 2).

Download:

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

Figure 4.Perkinsus spp.cob and the predicted aa sequences.

Sequences ofP. marinus, and forP. chesapeaki only the sites with different nt/aa sequences, are shown. Four invariant His residues that are ligands for heme β are highlighted in blue. ‘-’ indicates missing sequence; the potential quadruplet codons ‘aggy’ for glycine are marked in grey, and quintuplet codons ‘uaggc’ and ‘ucggu’ for glycine are boxed.

https://doi.org/10.1371/journal.pone.0019933.g004

Using dinoflagellatecox1 primer sets dinocox1F5-R3[24] and universalcox1 primer set cox1_5b-3a (Table 2), DNA fragments were amplified from genomic and cDNA templates ofP. marinus (0.96 kb) and cDNA ofP. chesapeaki (0.33 kb), respectively. Direct sequencing of these fragments proved that they werecox1 sequences with 50–60% identity to that of dinoflagellates and apicomplexans. When the 0.96-kbP. marinus cox1 sequence was used to BLAST against theP. marinus genome database, one 3147-bp sequence (AAXJ01004741) was obtained with 100% identity to theP. marinus DNA fragment we found. Nearly full-length cDNAs ofcox1 were generated by PCR amplification usingPerkinsus-specificcox1 primers for bothPerkinsus species (GenBank accession numbers HQ670238, HQ670240;Figure 5,Table 2). Both thecob andcox1 cDNA sequences matched the corresponding genomic DNAs, indicating that no mRNA editing events occurred in either transcript.

Download:

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

Figure 5.Perkinsus spp.cox1 and the predicted aa sequences.

Sequences ofP. marinus, and forP. chesapeaki only the sites with different nt/aa sequences, are shown. Six invariant His residues that are ligands for heme α, Cu_B and heme α3 are highlighted in blue. ‘-’ indicates missing sequence; the potential quadruplet codons ‘aggy’ for glycine and ‘ccccu’ quintuplet codons for proline are marked in grey, and a standard ‘ggu’ codon for glycine inP. chesapeaki cox1 is marked in red.

https://doi.org/10.1371/journal.pone.0019933.g005

Comparison of nt and deduced aa sequences ofPerkinsus cob andcox1 with counterparts in other alveolates revealed that correct translation ofPerkinsus mt genes required the Mold/Protozoan/Coelenterate mt codon table (TGA codes for tryptophan) in general. To be fully translatable without internal stop codons, however, frameshifts had to be introduced at every AGG and CCC codon, the equivalent of using AGGY to code for glycine (six sites incob and 7–8 sites incox1) and CCCCU for proline (twice incox1) (Figures 4,5). An analogous result was reported by Masuda et al.[20] for theP. marinus cox1. Multiple cDNAs and genomic sequences substantiated these unusual reading frames, as well as the direct sequencing of PCR products. An interesting difference was found between the twoPerkinsus species: at one site incox1, glycine was encoded by an AGGU codon inP. marinus, but by a standard GGU codon inP. chesapeaki (Figure 5). With the introduction of these invoked anomalous quadruplet and quintuplet codons, the deduced aa sequences of the twoPerkinsus COX1 were 98% identical to each other, 46–50% similar to the homologs in apicomplexans, 42–49% to dinoflagellates, 29–31% to ciliates, and 38–42% to other organisms (Figure 6). Forcob (Figure 4), besides the quadruplet codon AGGY, glycine was also encoded by the quintuplet codons UAGGC (forP. marinus) and UCGGU (forP. chesapeaki). After these adjustments, the deduced COB aa sequences of the twoPerkinsus spp. shared 97% similarity to each other, 34–36% to apicomplexans, 22–44% to dinoflagellates, 15–17% to ciliates, 27–33% to other organisms (Figure 6).

Download:

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

Figure 6. Phylogenetic affiliation ofP. marinus with apicomplexans based on mitochondrial COB and COX1.

The consensus trees with support from NJ (bootstrap, only >84% are shown), ML (aLRT), and MB (posterior probability). Brackets indicate clades of apicomplexans (AP), dinoflagellates (DI) and ciliates (CI).

https://doi.org/10.1371/journal.pone.0019933.g006

High density ofcis-introns relative to dinoflagellates

The correspondingP. marinus genomic sequences of 39 and 29 full-length cDNAs fromP. marinus andP. chesapeaki, respectively[8],[14], were obtained. Comparison of these 68 cDNAs with the genomic DNA sequences revealed the presence of introns in 42 genes, yielding a 61.8% intron rate. Through GenBank database searches, we obtained an additional 36 common genes with known genomic structures, 30 of which have intron(s) (Table S1). Overall, the intron rate forP. marinus genes was 69.2% (72 out of 104). The intron-containing genes harbored between one and ten introns with the lengths ranging from 39 to 1622 bp, the majority of which were <100 bp.

Multi-protein phylogeny ofPerkinsus and other lineages

Twenty-two ribosomal proteins were obtained forPerkinsus and various organisms; Maximum Likelihood (ML) trees inferred from the individual sequences gave varied tree topologies (Figures S1,S2,S3,S4). In general,P. marinus, dinoflagellates, apicomplexans, and ciliates formed a monophyletic group, while in several cases the heterokont diatoms, the closest relative of the alveolates, branched with some of the alveolate lineages, but without bootfostrap support.Perkinsus spp. allied with dinoflagellates in some cases (e.g.Figures S1C, 1F,S2E), and with apicomplexans (e.g.Figure S1D) or the diatoms (e.g.Figure S2D) in others, often with weak or no bootstrap support in these cases, indicating an unstable phylogenetic affinity. In contrast, NJ trees based on the 12 conserved protein sequences (actin, b-tubulin, GAPDH, α-tubulin, centrin, HSP90, EF1-α, ADP ribosylation factor, TIF5A, SmD1, cytochrome C and 14-3-3) produced similar tree topology, withP. marinus clustering with dinoflagellates in most of the cases (Figures S5,S6). For mt genes,Perkinsus spp. clustered with ciliates in COB tree, while allied with dinoflagellate/apicomplexan cluster in COX1 tree (Figure 6). When the concatenated RP sequence data (3,142 aa) was used, analyses using NJ, ML, and MB produced trees of similar topologies in whichP. marinus branched at the base of the dinoflagellate clade (Figure 7). This was true for the analyses both without (Figure 7A) and with (Figure 7B) the ancient dinoflagellate lineageOxyrrhis marina. The only exception was the MB tree in whichP. marinus was allied with the clade of apicomplexans (Figure 7A). Similarly, when the other 12 protein sequences were concatenated (3,879 aa) the consensus tree inferred from the three algorithms showed the close relationship betweenP. marinus and dinoflagellates (Figure 7C). In most of these concatenated trees, the alliance ofP. marinus and dinoflagellates was supported.

Download:

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

Figure 7. Phylogenetic affiliation ofP. marinus with dinoflagellates and apicomplexans based on 30 conserved protein sequences.

The consensus trees of concatenated genes of 19 ribosomal proteins (RPs) for 17 taxa (A), 8 RPs for 18 taxa includingOxyrrhis (B) and 11 non-RP proteins (C). Supports of nodes are from NJ (bootstrap), ML (aLRT), and MB (posterior probability). Brackets indicate clades of apicomplexans (AP), dinoflagellates (DI) and ciliates (CI).

https://doi.org/10.1371/journal.pone.0019933.g007

Multiple sequences were obtained for each of theP. marinus histones; in most of the phylogenetic trees, these sequences clustered together and allied with apicomplexans except for the H3 tree, in which oneP. marinus H3 grouped with the apicomplexanToxoplasma gondii, the other with dinoflagellate/ciliate clade (Figures 8,9). Histone 2A in many organisms has acquired an isoform referred to as H2A.X. In both dinoflagellates andP. marinus, H2A.X seems to be the dominant, if not the only, form. The homolog retrieved from theP. marinus genome was clustered with H2A.X in the clade of apicomplexans (Figure 8).

Download:

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

Figure 8. Phylogenetic affiliation ofP. marinus with apicomplexans based on histone H2A.

The canonical H2A and the isoform H2A.X consensus tree with support from NJ (bootstrap), ML (aLRT), and MB (posterior probability). Brackets indicate clades of apicomplexans (AP), dinoflagellates (DI) and ciliates (CI).

https://doi.org/10.1371/journal.pone.0019933.g008

Download:

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

Figure 9. Phylogenetic affiliation ofPerkinsus marinus with apicomplexan based on three histone proteins.

Histone H2B(A), H3 (B), and H4 (C) consensus trees with support from NJ (bootstrap, only >70% are shown), ML (aLRT), and MB (posterior probability). Brackets indicate clades of apicomplexans (AP), dinoflagellates (DI) and ciliates (CI).

https://doi.org/10.1371/journal.pone.0019933.g009

Perkinsus SL RNAs mark earlier emergence and more complex evolution oftrans-splicing in alveolates

PmaSLRNA sequences are similar to those of dinoflagellate SL RNAs, including the location of an apparent Sm-binding domain in the exon rather than in the intron, as is the case typically in other SLtrans-splicing eukaryotes (see[8],[9] for review). The SL has left its footprints in other parts of the dinoflagellate genome in the form of single and tandem exon repeats located adjacent to the 5′ UTRs of protein coding genes[28]. This apparently unproductive phenomenon is postulated to occur when SL-containing mRNA are reverse-transcribed and integrated to the genome[28] but could also be a result of chromosome cross-over recombination[16]. Likewise, SL exons in single or multiple units were found in someP. marinus genes. The S-type SL with L-type intron was also suggested to exist based on PCR-amplified cDNA sequences ofP. marinus SL RNA[29], although it requires verification by further genomic analysis.

PmaSLRNAs are distinct from dinoflagellate SL RNAs. In the apparent Sm-binding site, instead of a “TTTT” motif conserved in dinoflagellates, PmaSL has “TCTT” or “TCT”. The intron region of the SL RNA in dinoflagellates is conserved, but the similarities diminish inAmoebophrya, a parasitic lineage currently considered to represent the most ancient dinoflagellate[30]. SL RNAs inP. marinus display similar divergence from dinoflagellates, with a substantially longer intron relative to the core dinoflagellates andAmoebophrya, suggestive of an earlier divergence forP. marinus. The SL RNAs in other SLtrans-splicing eukaryotes range from 46 nt in the urochordateCiona intestinalis to 142 nt inTrypanosoma brucei. The SL RNA transcripts inP. marinus range from 80–83 nt, and are ∼83 nt inP. chesapeaki. Thus,Perkinsus SL RNAs have unique features in comparison to dinoflagellates, andPerkinsus spp. may represent the earliesttrans-splicing lineage within Alveolata, separated from the non-trans-splicing Ciliophora and Apicomplexa[8], yet distinct from the Dinoflagellata. Given the high diversity of the parasitic Syndiniales class of dinoflagellates[31], the uncharacterized marine alveolate group I that lies betweenPerkinsus and the core dinoflagellates ([7] and references therein) should be examined for the presence of additional types of SL RNA.

Perkinsus is a distinct pre-dinoflagellate taxon

As SLtrans-splicing occurs in both basal (e.g.Amoebophrya andOxyrrhis) and advanced (e.g.Alexandrium) lineages of dinoflagellates but not in apicomplexans and ciliates[8], the two closest relatives of dinoflagellates, the occurrence of this distinct mRNA processing mechanism is considered a defining indicator for dinoflagellates[12]. The presence of SL RNAtrans-splicing inPerkinsus spp. indicates its inclusion in or alliance with the phylum of dinoflagellates, in accord with previous molecular phylogenetic studies (e.g.[2]–[4],[30]). Likewise, our multi-protein phylogenies consistently show thatP. marinus is related to dinoflagellates among other representative eukaryotes with moderate-to-strong bootstrap support. Among the many single-gene phylogenetic trees, the majority is in agreement with the concatenated protein trees. In all trees,P. marinus was positioned as the earliest divergent even whenOxyrrhis, a genus hypothesized to be a pre-dinoflagellate[32] or an ancient lineage[12], was included. In addition,P. marinus was always placed basal toAmoebophrya, another ancient lineage of dinoflagellates.

Yet some degree of uncertainty exists in the phylogenetic position ofPerkinsus. Contrary to the long-held notion that dinoflagellates did not possess nucleosomes and canonical histones, genes of all four major histones have recently been found in dinoflagellates (for review see[21]); however, dinoflagellate histones usually have unique sequences with insertions/deletions in several places, resulting long branches in the phylogenetic trees (Figures 8,9). Comparing to dinoflagellates,P. marinus histones have typical eukaryotic sequences and group with apicomplexans in the phylogenetic trees. Besides histone trees, some other individual protein trees (Figures S1D,S2A, 2B, 2C,S3B,S5C) also show an alliance ofPerkinsus spp. with apicomplexans, in agreement with earlier morphological and cytological studies[2]. In rare cases,P. marinus is clustered with diatoms, apparently because the protein sequence was too short to provide strong support of any topology.

The current analysis is limited in that only the sequences from one or two species ofPerkinsus were available.Perkinsus appears more distant from apicomplexans than from dinoflagellates; however its generally close relationship with the clade of dinoflagellates could be due to the absence of taxa from intermediate lineages such as marine alveolate group I, additional taxa from the Perkinsozoa (e.g.Parvilucifera spp.), and dinoflagellates of the class Syndiniales.

Cis-splicing is thought to be uncommon in dinoflagellates[2]; however, only a few dinoflagellates have been examined for the presence of introns (e.g. form II Rubisco inSymbiodinium[33], luciferase C inPyrocystis lunula[34]). We have examined more than 30 genes such aspcna, form II Rubisco, 14-3-3, and centrin for several dinoflagellates ([35],[36] and our unpubl. results), and did not find introns. A relatively high intron density for a dinoflagellate is found inAmphidinium carterae, in which a survey of 31 genes yields a 48%cis-splicing rate[37]. Our analysis of 104Perkinsus genes yielded a 69.2%cis-splicing rate, a level contrasting those found in most dinoflagellates, and closer to the >50% level found in apicomplexans[38],[39].

The unique reading frame shifting and the lack of mRNA editing for mt genes again markP. marinus as distinct from typical dinoflagellates. BothP. marinus cob andcox1 mRNAs are identical to their genomic DNAs, indicating that no mRNA editing occurs to correct the frameshifts in these mt genes. Masuda et al.[20] reported the full-length mtcox1 mRNA fromP. marinus, showing that this mRNA was not translatable with standard codon usage, due to a reading frame that had to be shifted a total of 10 times at every AGG and CCC codon to yield a consensus COX1 protein. One or two sites of +1 frameshifting have been documented in animal mt genes (for review, see[40]), but such extensive +1 and +2 frameshifts are unique. In retroviruses, a –1 frameshift is corrected by tRNA back-slippage over homopolymeric codons adjacent to a local secondary structure that may include a pseudoknot (for review, see[41]). Masuda et al.[20] suggest two feasible mechanisms for the translational frameshifts inPerkinsus: a ribosomal frameshift in which stalled ribosomes skip the first bases of these codons (similar to the model hypothesized by Beckenbach et al.[42]), or specialized tRNAs recognizing non-triplet codons AGGY and CCCCU to code for glycine and proline, respectively. In this study, we addcox1 forP. chesapeaki andcob sequences forP. marinus andP. chesapeaki, which share the unusual AGGY codon withcox1 and use other unusual codons (UAGGC and UCGGU) to encode glycine as well. Specialized tRNAs in thePerkinsus mitochondrial system recognizing non-triplet AGGY and CCCCU codons, and likely UMGGY as well, may be more likely than the ribosomal frameshifting scenario, as naturally occurring tRNA mutants suppress +1 frameshiftsvia an extended anticodon loop inEscherichia coli (e.g.[43]), and quadruplet codons are used in protein mutagenesis[44].

ThePerkinsus lineage is remarkably distinct from, while close to, dinoflagellates, and is most likely an independent lineage, supporting the postulate thatPerkinsus spp., along withParvilucifera spp., constitutes an independent phylum dubbed Perkinsozoa, the fourth phylum in Alveolata[6]. Although not addressed directly, a number of recent phylogenetic trees containing taxa from marine alveolate group I andPerkinsus-related parasitic alveolates such asParvilucifera spp. reinforce grouping ofPerkinsus spp. as an independent phylum[7],[45],[46]. Future phylogenies with broader taxon sampling that include species fromParvilucifera spp., Syndiniales in addition toAmoebophrya, and marine alveolate group I representatives will refine the phylogenetic relationships amongPerkinsus, dinoflagellates, and other alveolates.

Figure S1.

ML phylogenetic trees of six of the 22 ribosomal proteins. A, RPL11; B, RPL17; C, RPL18A; D, RPL18; E, RPL21; F, RPL22. Groupings of major clades are labeled on the right. DI, dinoflagellates; AP, apicomplexans; CI, ciliates; PE,Perkinsus.

https://doi.org/10.1371/journal.pone.0019933.s001

(TIF)

Figure S2.

ML phylogenetic trees of six of the 22 ribosomal proteins. A, RPL26; B, RPL32; C, RPL34; D, RPL35A; E, RPL44; F, RP_P1. Groupings of major clades are labeled on the right. DI, dinoflagellates; AP, apicomplexans; CI, ciliates; PE,Perkinsus.

https://doi.org/10.1371/journal.pone.0019933.s002

(TIF)

Figure S3.

ML phylogenetic trees of six of the 22 ribosomal proteins. A, RPS3a; B, RPS5; C, RPS7; D, RPS10; E, RPS11; F, RPS13. Groupings of major clades are labeled on the right. DI, dinoflagellates; AP, apicomplexans; CI, ciliates; PE,Perkinsus.

https://doi.org/10.1371/journal.pone.0019933.s003

(TIF)

Figure S4.

ML phylogenetic trees of four of the 22 ribosomal proteins. A, RPS17; B, RPS25; C, RPS26; D, RPS27a. Groupings of major clades are labeled on the right. DI, dinoflagellates; AP, apicomplexans; CI, ciliates; PE,Perkinsus.

https://doi.org/10.1371/journal.pone.0019933.s004

(TIF)

Figure S5.

NJ phylogenetic trees of six of the 12 non-RP proteins. A, actin; B, β-tubulin; C, GAPDH; D, α-tubulin; E, centrin; F, HSP90. Groupings of major clades are labeled on the right. DI, dinoflagellates; AP, apicomplexans; CI, ciliates; PE,Perkinsus.

https://doi.org/10.1371/journal.pone.0019933.s005

(TIF)

Figure S6.

NJ phylogenetic trees of six of the 12 non-RP proteins. A, EF1-α; B, ADP ribosylation factor; C, TIF5A; D, SmD1; E, cytochrome C; F, 14-3-3. Groupings of major clades are labeled on the right. DI, dinoflagellates; AP, apicomplexans; CI, ciliates; PE,Perkinsus.

https://doi.org/10.1371/journal.pone.0019933.s006

(TIF)

Table S1.

cis-introns inPerkinsus marinus.

https://doi.org/10.1371/journal.pone.0019933.s007

(XLS)