A Cryptic Algal Group Unveiled: A Plastid Biosynthesis Pathway in the Oyster Parasite Perkinsus marinus

Abstract

Plastids are widespread in plant and algal lineages. They are also exploited by some nonphotosynthetic protists, including malarial parasites, to support their diverse modes of life. However, cryptic plastids may exist in other nonphotosynthetic protists, which could be important in studies on the diversity and evolution of plastids. The parasite Perkinsus marinus, which causes mass mortality in oyster farms, is a nonphotosynthetic protist that is phylogenetically related to plastid-bearing dinoflagellates and apicomplexans. In this study, we searched for P. marinus methylerythritol phosphate (MEP) pathway genes, responsible for de novo isoprenoid synthesis in plastids, and determined the full-length gene sequences for 6 of 7 of these genes. Phylogenetic analyses revealed that each P. marinus gene clusters with orthologs from plastid-bearing eukaryotes, which have MEP pathway genes with essentially the same mosaic pattern of evolutionary origin. A new analytical method called sliding-window iteration of TargetP was developed to examine the distribution of targeting preferences. This analysis revealed that the sequenced genes encode bipartite targeting peptides that are characteristic of proteins targeted to secondary plastids originating from endosymbiosis of eukaryotic algae. These results support our idea that Perkinsus is a cryptic algal group containing nonphotosynthetic secondary plastids. In fact, immunofluorescent microscopy indicated that 1 of the MEP pathway enzymes, 1-deoxy-D-xylulose 5-phosphate reductoisomerase, was localized to small compartments near mitochondrion, which are possibly plastids. This tiny organelle seems to contain very low quantities of DNA or may even lack DNA entirely. The MEP pathway genes are a useful tool for investigating plastid evolution in both of the photosynthetic and nonphotosynthetic eukaryotes and led us to propose the hypothesis that ancestral “chromalveolates” harbored plastids before a secondary endosymbiotic event.

Introduction

“Plastids” are intrinsic organelles in plants and algae, but gaps remain in our knowledge regarding their diversity and distribution. Plastids originally arose from endosymbiotic cyanobacteria and are now involved in processes including photosynthesis and other biochemical processes in plant cells. Among the protists, several lines of algae or plastid-bearing eukaryotes (PBEs), such as giant kelp and diverse bloom-forming algae are known. Furthermore, intracellular parasites of the phylum Apicomplexa, including the malarial parasite, have been recently shown to harbor nonphotosynthetic but essential plastids (Wilson 2005). In a very recent environmental sequencing study, a distinct group of PBEs, the picobiliphytes, were discovered (Not et al. 2007). Thus, it seems very likely that unknown PBEs still exist.

On the basis of their deduced evolutionary history, plastids can be divided into 2 classes: primary plastids with 2 bounding membranes, which are direct descendants of endosymbiotic cyanobacteria, and secondary plastids with more bounding membranes, which originate from past engulfed eukaryotic algae (Bhattacharya et al. 2004). Because secondary plastids remain “outside” with regard to membrane topology, proteins targeted to these compartments must be transported via a secretory pathway; they must contain an N-terminal bipartite targeting peptide, composed of a signal peptide (SP) to lead the polypeptide to the endoplasmic reticulum (ER), and a subsequent transit peptide (TP) to deliver the mature protein into the plastid lumen (van Dooren et al. 2001). Secondary PBEs are scattered over the protist phylogeny; however, it has been proposed that members of the so-called chromalveolate group, which consists of diatoms and other stramenopiles, haptophytes, cryptophytes, dinoflagellates, and apicomplexans, ancestrally contain secondary plastids of a single red algal origin (Cavalier-Smith 1999). Several lines of “evidence” support the “chromalveolate” hypothesis (Fast et al. 2001; Yoon et al. 2002), but critics note that the hypothesis assumes too many independent losses of plastids (Bodył 2005). Thus, a better understanding of plastid distribution in the basal chromalveolates is desirable to address these criticisms.

Perkinsus spp. are marine unicellular protists with a worldwide distribution that attack a wide range of mollusks, including clams and abalones, causing mass mortality (Villalba et al. 2004). Perkinsus marinus is the most notorious species of the genus because it parasitizes the eastern oyster Crassostrea virginica and has heavily impacted oyster fisheries and hence coastal water quality in the United States (Villalba et al. 2004). Molecular phylogenetic data have shown that this species is a basal chromalveolate derived from the ancestral dinoflagellates just after the split from apicomplexans (Cavalier-Smith and Chao 2004; Leander and Keeling 2004; Adl et al. 2005); thus, examining P. marinus for the existence of ancestral secondary plastids is of interest. Although electron microscopy (EM) observations have revealed no signs of plastids (Perkins and Menzel 1967; Perkins 1976, 1996; Sunila et al. 2001), 2 quite recent studies have suggested that Perkinsus spp. contain secondary plastids; P. marinus was shown to possess genes for a plant-type ferredoxin system that possibly encode plastid-targeting signals (Stelter et al. 2007) and an EM observation indicated a tiny organelle bounded by 4 membranes in Perkinsus olseni (=Perkinsus atlanticus) (Teles-Grilo et al. 2007). However, gene sequence or morphology in itself cannot prove the existence of vestigial plastids; at the least, localization data should be presented in order to confirm the presence of plastids.

Genome sequencing has revealed that biosynthesis of isoprenoid precursors is a key metabolic role of both photosynthetic (Matsuzaki et al. 2004; Derelle et al. 2006) and nonphotosynthetic (Gardner et al. 2002, 2005) plastids. Isoprenoids are a diverse and versatile group of compounds including sterols, carotenoids and other terpenes, and the side chains of quinones and chlorophylls. Isoprenoids are all derived from isopentenyl diphosphate and its isomer, both of which are synthesized by the methylerythritol phosphate (MEP) pathway in plastids but the mevalonate (MVA) pathway in the cytosol of many eukaryotes, including higher plants and animals (Rodríguez-Concepción 2004). It has been suggested that in eukaryotes the MEP pathway only exists in PBEs, with a discussion of the evolutionary origins of the genes based on the orthologs of higher plants and several bacteria, although only 5 out of 7 MEP genes were known at the time (Lange et al. 2000). Many eukaryotic genomes have been sequenced since then, and to date the MEP pathway genes have been found to be specific to PBEs and would seem to be necessary for most given that they lack the MVA pathway genes; thus, it seems that the MEP pathway is a specific feature of plastids, photosynthetic, or otherwise. Therefore, the 7 MEP pathway genes (shown in fig. 1) would be good molecular markers for the study of plastid distribution and evolutionary history; however, to the best of our knowledge, no prior studies have involved the widespread sampling of PBEs.

A preliminary database of the P. marinus genome at The Institute for Genomic Research (TIGR) contains partial sequences of the MEP pathway genes (the existence of ispC, ispG, and ispH has already been discussed by Stelter et al. [2007]). In this study, we attempted to clone full-length MEP pathway genes of P. marinus to elucidate the evolution of secondary plastids by phylogenetic analyses. A new analytical method for investigating protein sorting signals was also developed to compare signals between genes. Finally, immunofluorescent microscopy was performed to demonstrate the existence of plastids in P. marinus.

Materials and Methods

Culture Conditions

Perkinsus marinus strain CRTW-3HE was obtained from the American Type Culture Collection (number 50439; ATCC, Manassas, VA) and maintained at 26 °C in ATCC Medium 1886. Discontinued products for the medium components were substituted as follows during the course of the study: Instant Ocean Sea Salt (Aquarium Systems, Mentor, OH) for artificial seawater (S1649; Sigma, St Louis, MO) and Lipid Mixture (1000×; L5146; Sigma) for Lipid Concentrate (100×; 21900-014; Gibco, Grand Island, NY).

Sequencing MEP Pathway Genes

Total RNA was extracted from cell pellets using TRIzol Reagent (Invitrogen, Carlsbad, CA), and the mRNA was enriched using the PolyATtract mRNA Isolation System III (Promega, Madison, WI). Complementary DNA was synthesized using the CapFishing Full-length cDNA Premix Kit (Seegene, Seoul, Korea) with random hexamers or the oligo dT adapter as a primer. Reverse transcriptase–polymerase chain reaction and rapid amplification of cDNA ends were performed using PrimeSTAR HS DNA polymerase (Takara Bio, Tokyo, Japan) and the primers listed in supplementary table S1 (Supplementary Material online). Polymerase chain reaction products were purified using Wizard SV Gel and PCR Clean-Up System (Promega) and then cloned with the ZeroBlunt TOPO PCR Cloning Kit for Sequencing (Invitrogen). Inserts were sequenced using vector-specific or gene-specific (shown in supplementary table S1, Supplementary Material online) primers. Full-length sequences have been deposited in DDBJ/EMBL/GenBank under the accession numbers AB284361–AB284366 (shown in supplementary table S3, Supplementary Material online). All experiments described here were performed according to the manufacturers' instructions. The amino acid sequences were inferred from the most upstream ATG with the universal codon table.

Phylogenetic Analysis

All possible MEP pathway orthologs for bacteria for which complete genome sequences were available as of 1 November 2006 were retrieved from the National Center for Biotechnology Information Refseq database. Very large multiple sequence alignments were constructed using MUSCLE 3.6 (Edgar 2004) and then 50 bacteria were chosen (supplementary table S2, Supplementary Material online) based on the alignments to retain at least a single species for each bacterial phylum and to eliminate species with organism-specific indels when possible. Amino acid sequences for 8 eukaryotes were deduced based on the sequenced genome databases (supplementary table S3, Supplementary Material online). Data matrices were constructed as follows: structure-based alignments were created for bacterial sequences using the Expresso (Armougom et al. 2006) Web service (http://www.tcoffee.org/), then eukaryotic sequences were added using ClustalX 1.83 (Thompson et al. 1997), and the results were manually refined and sites chosen for analysis. Although the Expresso Web service was used, ispG and ispH alignments were constructed by conventional methods because no suitable structural data were available for these genes. First, maximum likelihood (ML) trees were inferred with a WAG substitution matrix (Whelan and Goldman 2001) using Phyml 2.4.4 (Guindon and Gascuel 2003) with bootstrap values (100 replications), and distant problematic bacterial sequences (underlined in supplementary table S2, Supplementary Material online) were eliminated. Final ML trees were then inferred within additional operational taxonomic units (OTUs) for dinoflagellates and haptophytes, which have been obtained and synthesized from expressed sequence tags (EST) (source organisms and the original accession numbers are shown in the supplementary table S3, Supplementary Material online). Apicomplexan orthologs were excluded because they contain a vast number of changed or inserted residues, making it difficult to robustly infer phylogeny. In fact, trees drawn to include them resulted in placing them at deeper branches, disturbing tree topology, or reducing statistical support to greater or lesser degrees (data not shown).

Targeting Presequence Analysis

The N-terminal extensions of the translated amino acid sequences for each MEP pathway gene were examined using SignalP-HMM (Nielsen and Krogh 1998; Bendtsen et al. 2004) for SP and SOSUI (Hirokawa et al. 1998) for the transmembrane region. The distribution of protein sorting signals was examined using newly developed sliding-window iteration of TargetP (SWIT) analysis. Amino acid sequences were inferred from the first in-frame start codon, and subcellular localization was predicted by submitting the first 130 residues to the TargetP server (http://www.cbs.dtu.dk/services/TargetP/) (Nielsen et al. 1997; Emanuelsson et al. 2000); this was then repeated successively after eliminating the first residue from the N-terminus of the previously analyzed amino acid sequence (i.e., sliding a 130-residue window with a 1-residue step). Iterations were performed using the newly developed Ruby script, facilitated by the BioRuby library (http://www.bioruby.org/) (Goto et al. 2003). TargetP yields scores for SP, mitochondrial TP (mTP), chroloplast TP (cTP), and an other category for each iteration. The 4 scores were plotted with window positions on the y and x axes. Two superoxide dismutase genes, PmSOD1 and PmSOD2 (AY095212 and AY095213, respectively), were used for comparison (Wright et al. 2002).

Antibody

Rabbits were immunized with 2 synthesized peptides, TATVEDALKHPNWS and YTLAYPQRLHHDGS, which were partial fragments of the deduced ispC peptide sequence. A fraction of IgG (270 μg/ml) from the sera was obtained by affinity purification using the peptide cocktail to capture specific antibody. Recombinant ispC protein (with an N-terminal 6×His tag) lacking the predicted bipartite targeting peptide was expressed in Escherichia coli BL21-AI (Invitrogen) and was purified using HisTrap FF crude (GE Healthcare, Little Chalfont, Buckinghamshire, UK). We discovered that the cytosolic fraction of P. marinus contained a protein that nonspecifically reacted with normal rabbit IgG (Santa Cruz Biotechnology, Santa Cruz, CA); therefore, all antibodies were first adsorbed with proteins from the cytosolic fraction of P. marinus that were obtained by precipitation with 50–70% saturated ammonium sulfate.

Centrifugal Fractionation

Cells were collected by centrifugation (200 × g for 5 min at ambient temperature), washed with phosphate-buffered saline (PBS), resuspended in 50 mM N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid buffer (pH 7.4) supplemented with protease inhibitor cocktail (Complete; F. Hofmann-La Roche, Basel, Switzerland), and disrupted by sonication on ice. The homogenate was subjected to differential centrifugal fractionation at 4 °C and divided into 4 fractions: 200 × g sediment (5 min), 2,000 × g sediment (10 min), 20,000 × g sediment (15 min), and the supernatant. The 20,000 × g sediment was incubated for 30 min at 4 °C in buffer containing 0.5% Triton X-100 and was then centrifuged at 20,000 × g for 15 min. All fractions were subjected to western blotting performed according to the following protocol: proteins separated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (10% gel) were transferred to PVDF membrane (Immobilon-P; Millipore, Billerica, MA), membranes were blocked for 1 h in blocking buffer (3% bovine serum albumin and 0.05% Tween 20 in PBS), and primary antibody (1:1000 dilution of the affinity-purified antibody) was added for 1 h. Next, alkaline phosphatase-conjugated Goat Anti-Rabbit IgG (Fc) (Promega S3731) in blocking buffer (1:7500) was added for 1 h and then the blots were developed by addition of BCIP/NBT Color Development Substrate (Promega S3771).

Immunofluorescent Microscopy

Cells were incubated for 30 min in culture medium containing 200 nM CMXRos (MitoTracker Red; Invitrogen M7512), washed for 15 min, and then fixed for 30 min at −20 °C with 4% (w/v) paraformaldehyde in 85% (v/v) methanol. Fixed cells were washed with PBS and then mounted on Matsunami adhesive silane–coated slides (Matsunami Glass Industries, Kishiwada, Japan) and air dried. The slides were blocked for 1 h with BlockAce (Dainippon Sumitomo Pharma, Osaka, Japan) containing 4 μg/ml normal goat IgG (Santa Cruz Biotechnology), labeled for 1 hr with a 1:1000 dilution of affinity-purified antibody, and then incubated for 1 h with 1:400 dilution of Alexa Fluor 488 Goat Anti-Rabbit IgG (H + L; Invitrogen A11008) in Can Get Signal immunostain Solution A (Toyobo, Osaka, Japan). After counterstaining with 1 μg/ml 4′,6-diamidino-2-phenylindole (DAPI), the slides were observed under an epifluorescence microscope (Olympus BX60) equipped with cooled digital color camera (Olympus DP70). For negative control experiments, the affinity-purified antibody was omitted or substituted with the same concentration of normal rabbit IgG.

Results

MEP Pathway Genes

We first searched for MEP pathway genes using a similarity search service at the P. marinus genome database of TIGR (http://www.tigr.org/tdb/e2k1/pmg/) using red alga, green alga, and apicomplexan homologs as queries for each gene (identifiers are shown in supplementary table S3, Supplementary Material online). Irrespective of species used as queries, all homologs resulted in the same significant hits: single contigs for dxs, ispC, ispE, and ispF and 2 contigs for ispG and ispH. For the latter genes, the hits were located at the termini of contigs and were assumed to be single genes astride contig gaps. Thus, all these genes were thought to exist only in single copy. In contrast, MVA pathway genes were not found when we searched for them in a similar fashion.

Six out of 7 MEP pathway genes (fig. 1) were then sequenced in their entirety using gene-specific primers designed based on the contig sequences. No homolog of the gene responsible for the third step, ispD, was found in the database, and an attempt to amplify the gene using degenerate primers failed. Untranslated regions were short, 19–45 and 15–97 nt for the 5′ and 3′ ends, respectively.

Evolutionary Origin and Relationships

To investigate the evolutionary origin of the sequenced MEP pathway genes, we performed phylogenetic analyses using PBEs and a wide range of bacteria. The dxs ortholog of P. marinus clusters with those of PBEs, and they form a clade with the alpha-proteobacterial counterparts with 100% bootstrap support, which is independent of cyanobacterial orthologs (fig. 2). The ispC orthologs of P. marinus and PBEs form a clade with the cyanobacterial orthologs with 100% bootstrap support (fig. 3). For the ispE tree, eukaryotic orthologs, including that of P. marinus and chlamydial counterparts, form a weakly supported (76%) clade (fig. 4). The ispF ortholog of P. marinus clusters with those of PBE with weak bootstrap support (58%), but the clade has failed to show a relationship with specific bacteria (fig. 5). This is probably because of the shorter data matrix (104 sites), corroborated by poor phylogenetic resolution of the overall tree. For the ispG tree, the P. marinus ortholog and eukaryotic orthologs other than those from red algae form a clade with the chlamydial orthologs with 100% bootstrap support, whereas red algal orthologs form another clade with the cyanobacterial orthologs (fig. 6). The ispH orthologs of P. marinus and PBEs form a clade with the cyanobacterial orthologs with 100% bootstrap support (fig. 7). All these results indicate that each of P. marinus orthologs clusters with the corresponding PBE orthologs; thus, the P. marinus orthologs are unlikely to have been independently transferred from bacteria.

Some of the P. marinus orthologs showed weak phylogenetic affinity to the dinoflagellate orthologs. The dxs ortholog formed a highly supported clade with the dinoflagellate Heterocapsa (fig. 2). The ispC ortholog formed a clade with a synthetic dinoflagellate OTU with <50% bootstrap support (fig. 3). Additionally, when a short sequence (105 sites) from the dinoflagellate Amphidinium ortholog was included in the ispG analysis, it was sister to the Perkinsus ortholog but with weak bootstrap support (54%). However, bootstrap supports for the adjacent nodes became weak, possibly due to the short dinoflagellate sequence (data not shown). The statistical weakness seems attributed to missing data derived from using partial ESTs of the dinoflagellate orthologs. In fact, the dxs tree, which has the strongest support, has the lowest amount of missing data (6% of sites) in the dinoflagellate ortholog. Therefore, the strength of this phylogenetic model may improve if full-length sequences become available. Reliable relationships for ispE and ispF orthologs of P. marinus were not resolved, and this was probably due to long-branch attraction or the shorter data matrix (figs. 4 and 5). The ispE tree (fig. 4) shows a highly supported close relationship between orthologs of the haptophyte Emiliania and the dinoflagellate Karlodinium. Although Karlodinium has a tertiary plastid derived from haptophytes, the close relationship should not be interpreted as a result of endosymbiotic gene transfer, but rather as an artifactual grouping of stray OTUs, because these 2 OTUs are derived from partial ESTs and have no overwrap on the sequence alignment. To summarize, in total, the MEP pathway genes of P. marinus have weak phylogenetic affinity to those of dinoflagellates, and this likely shows that the MEP pathway is descended from the common ancestor of P. marinus and dinoflagellates.

The present phylogenetic trees indicate that the MEP pathway genes of PBEs and P. marinus have essentially the same pattern of mosaic origins in Cyanobacteria, Proteobacteria, and Chlamydia (table 1). In PBEs, dxs (fig. 2) and ispE (fig. 4) had a sister relationship with the proteobacterial and chlamydial orthologs, respectively, whereas ispC and ispH seemed to be derived from Cyanobacteria (figs. 3 and 7). A chlamydial origin was also observed for ispD orthologs (supplementary fig. S1, Supplementary Material online). The evolutionary origin of ispF remains unclear, probably owing to insufficient informational sites being used (fig. 5). However, only red algae had cyanobacteria-like ispG orthologs, and P. marinus and other PBEs harbored genes that were closely related to their chlamydial counterparts (fig. 6). This difference is not a phylogenetic artifact because there were long and well-aligned insertion sequences that were shared with PBEs other than red algae and Chlamydia (supplementary fig. S2, Supplementary Material online). The present phylogenetic analyses demonstrated that P. marinus and all PBEs analyzed here, with a curious exception of red algae, obtained their MEP pathway genes from an identical source.

Targeting Presequence and Subcellular Localization

The inferred amino acid sequences of the 6 MEP pathway genes in P. marinus have obvious N-terminal extensions relative to their bacterial homologs (fig. 1). N-terminal extensions often function as targeting presequences to deliver the peptide into certain subcellular compartments (e.g., ER, mitochondria, and plastids). First, SignalP predicted that all genes would have SPs (fig. 1; supplementary fig. S3, Supplementary Material online), although the probability of cleavage was not very high. Each predicted SP indicated an abnormally long (10–20 residues) hydrophilic n-region, but h- and c-regions were normal (supplementary fig. S3, Supplementary Material online). Because elimination of a few residues from the N-termini of the n-regions greatly improved the cleavage probability (data not shown), we consider that SPs with minor modification exist and they are likely cleaved off. The cleavable SP indicates that the MEP pathway enzymes are trafficked via the secretory pathway. However, TargetP predicted that all MEP pathway orthologs (except dxs) would be targeted to mitochondria. A lesson learned from the SignalP result is that simple predictions using full-length sequences may yield dubious results if the organism under question is distantly related to organisms after which the predictor has been modeled. Therefore, the need exists to monitor trends and the robustness of results obtained by TargetP analysis to better understand whether the mitochondrial targeting was significant and to further investigate the characteristics of the N-terminal extension.

We developed a new method of exploiting TargetP in order to robustly examine the distribution of targeting preferences. The method iteratively invoked TargetP with a sliding 130-residue window from the N-terminus; thus, we named the method SWIT. Figure 8 shows the SWIT results for the MEP pathway genes and 2 P. marinus superoxide dismutases, PmSOD1 and 2 (Wright et al. 2002). For example, for dxs, a prediction with the first 130 residues yielded scores of 0.002, 0.062, 0.165, and 0.064 for cTP, mTP, SP, and others (y intercepts for green, blue, yellow, and gray lines), respectively; this is equivalent to the simple TargetP prediction using the full-length sequence. As the window slides downstream, the SP score increases up to 0.393 and then decreases to near zero. Although the first result (y intercept) alone, or the simple TargetP prediction, seems rather unreliable, we recognize that the N-terminal region likely has SP preference (yellow arrow). As the window slides along, a significant prominence for cTP appears (green arrow) in the vicinity of the predicted SP cleavage site (yellow triangle). The MEP pathway genes shared this characteristic SWIT trend, that is, a peak in the SP score at the N-terminal region and a subsequent increase in the cTP score. The trends seem to suggest that these enzymes are localized to secondary plastids because an N-terminal bipartite (SP + cTP) presequence is a characteristic feature of proteins of secondary plastids (van Dooren et al. 2001). Thus, we hypothetically considered residues between the N-terminus and the predicted SP cleavage site and between the SP cleavage site and the most downstream predicted site for cTP cleavage (green triangles) as putative SP and cTP, respectively.

One issue is that all aforementioned genes, except dxs, have considerable mTP scores at the y intercepts (fig. 8), which correspond to the simple TargetP results (see above). Both PmSOD1 and 2 also have high mTP scores at their N-termini, similar to the 5 MEP pathway genes (fig. 8); however, PmSOD2 localizes to an unknown, nonmitochondrial compartment, whereas PmSOD1 localizes to the mitochondria (Schott and Vasta 2003). According to the SWIT result, the mTP score for PmSOD1 is maintained along the N-terminal extension, and this corresponds to its mitochondrial localization; however, the score for PmSOD2 rapidly declines and SP and subsequent cTP preferences appear alternatively. Given that SP and cTP preferences do not appear for PmSOD1, they can be used to discriminate localization to a compartment other than the mitochondria; additionally, a high mTP score at the N-terminus has no biological significance when there are significant SP and cTP preferences. The MEP pathway genes share the SP + cTP feature with PmSOD2; thus, the MEP pathway genes appear to have protein sorting signals that are characteristic of proteins targeted to secondary plastids and to be localized to a compartment other than the mitochondria.

The predicted TP regions suggested by SWIT analysis were further examined in comparison to the TPs of PBEs. The predicted regions were overall hydrophilic and with a net positive charge, as in PBEs, resulting from depletion of acidic residues (5.3% compared with 12.9% in the homologous region) and accumulation of hydroxylated residues (especially Ser; 16.2% compared with 6.4%) and positive-charged residues (especially Arg; 7.1% compared with 5.2%). All but ispF were shown to possess Phe residues in a hydrophobic context in proximity to the SP cleavage site, and a single transmembrane helix followed by an Arg-rich region within the TP region (fig. 1; supplementary fig. S3, Supplementary Material online). This feature resembles a similar feature in the class I TP of peridinine dinoflagellates, and no apicomplexans reported so far harbor the transmembrane helix (Patron et al. 2005). No hydrophobic regions were predicted for ispF or PmSOD2, which makes them similar to the class II TP on the other hand (Patron et al. 2005). Collectively, the predicted TP regions resemble those of the peridinine dinoflagellates with respect to physicochemical features and class duality.

To examine whether the predicted bipartite targeting sequence is functional, a polyclonal antibody recognizing the second step enzyme coded by ispC was prepared. The affinity-purified antipeptide antibody successfully detected recombinant ispC (fig. 9A, lane 1), and it detected a protein of approximately 43 kDa in P. marinus cell lysates (lane 2). Although smaller in size than predicted molecular weight of ispC (51 kDa), the size corresponds to an estimated molecular weight of the mature protein after cleavage of the predicted bipartite targeting peptides (asterisk in fig. 8). Detection at 43 kDa suggests that ispC is transported using targeting peptides that are subsequently cleaved. SOSUI predicted that the mature ispC protein was soluble. However, centrifugal fractionation showed the ispC protein specifically in the 20,000 × g sediment (lane 5), indicating that the protein was not cytosolic. The protein was easily solubilized by a mild detergent (lane 8), suggesting that it may be associated with membranes or membrane-bounded organelles. Finally, immunofluorescent microscopy revealed dotted signals for ispC (fig. 9B), and these signals were not associated with mitochondria (fig. 9C) or DNA (fig. 9D). Cells typically had multiple fluorescent spots that were frequently located near mitochondria (fig. 9E). Staining for extranuclear DNA overlapped with mitochondrial fluorescence (fig. 9F) but not with ispC fluorescence (fig. 9G).

Discussion

Plastids in P. marinus

The presence of 6 MEP pathway genes (fig. 1), together with the lack of any MVA pathway genes, suggests that the MEP pathway is responsible for de novo isoprenoid synthesis in P. marinus. A little possibility cannot be ruled out that the MEP pathway is not functional because we have been unable to find an ispD ortholog, but it does not impact on our logic for proving the existence of plastids in P. marinus (see below). Of course, activity of the MEP pathway still requires characterization in order to further discuss plastid function; however, this is beyond the scope of the present study.

The P. marinus genes reported here are very likely relevant to plastids because the MEP pathway is specific to PBE, all PBEs have obtained the genes from essentially the same source (table 1), and all MEP pathway genes of P. marinus consistently group with them (figs. 2–7³⁴⁵⁶⁷). Furthermore, the fact that all these genes have the bipartite targeting sequence (figs. 1 and 8), plus the fact that the predicted targeting sequences are cleaved (fig. 9A, lane 2), supports our idea that these enzymes traffic using machinery homologous to those of secondary PBEs. Finally, a subcellular fractionation experiment indicated that ispC is not cytosolic and is associated with membranes or organelles (fig. 9A, lanes 5 and 8), and immunofluorescent microscopy showed punctate localization of ispC near mitochondria (fig. 9E). Generally said, if a group of proteins accumulates at a subcellular compartment by means of machinery reasonably homologous to that of plastids, that compartment should be identified as a plastid. Although recent studies have shown the evidence for P. marinus plastid (Stelter et al. 2007; Teles-Grilo et al. 2007), no direct evidence that completely fulfills this criterion has been presented. To our knowledge, we are the first to use proper evidence in order to demonstrate that P. marinus harbors secondary plastids.

We revealed the existence of P. marinus plastids by detecting ispC using immunofluorescent microscopy. However, our SWIT analysis showed that ispF and PmSOD2 had similar but slightly different bipartite presequences (figs. 1 and 8). Although PmSOD2 has also shown punctate localization (Schott and Vasta 2003), it remains to be elucidated whether ispC and PmSOD2 are colocalized. Another point to be examined is that the ispC spots do not overwrap with DNA signals (fig. 9G). Given that staining of mitochondrial DNA is definitive (fig. 9F), the putative plastids seem to contain very low quantities of DNA or may even lack DNA. Our preliminary search of the TIGR genome database suggests that P. marinus has genes for mitochondrial DNA and RNA polymerases, elongation factors, and ribosomal proteins but seemingly lacks those for plastid counterparts (data not shown). There are no examples in the literature of DNA-lacking plastids; however, we know of many independent reports that show modified or degenerated mitochondria that lack DNA (Hackstein et al. 2006). It is thus worth considering the possibility of DNA-lacking plastids in P. marinus, although this remains to be confirmed experimentally.

We have shown the presence of secondary plastids in P. marinus by immunofluorescent microscopy; however, no corresponding structure has been found in EM specimens sampled from the same culture (Kuroiwa H, unpublished data). Upon conducting a literature review, we found that 3 candidate ultrastructures have been observed in allied organisms. One is a “coiled membrane system” near the mitochondria that was found in the oldest observations of P. marinus zoosporulation (Perkins and Menzel 1967). Although this structure has been considered a morphological variant of the mitochondria, it could also be interpreted as a multimembrane-bounded plastid resembling the apicoplast of the malarial parasite Plasmodium falciparum (Hopkins et al. 1999). Another candidate structure is the “enigmatic body” found in a related organism, Rastrimonas subtilis (Brugerolle 2002, 2003), which is limited by 2 or more membranes and is akin to the apicoplast of Toxoplasma gondii in spherical appearance and juxtanuclear localization (Matsuzaki et al. 2001; Köhler 2005); however, the assumed relationship between R. subtilis and Perkinsus spp. is based only on morphology and is currently unsupported by molecular information. Finally, a putative plastid bounded by 4 membranes was recently reported in P. olseni (=P. atlanticus) (Teles-Grilo et al. 2007). If this is the plastid to which the MEP pathway enzymes target, connection between the TP with a transmembrane helix and a plastid bounded by 3 membranes should be reconsidered. It is interesting that all 3 candidate structures have been observed in the respective organisms' flagellated stage and that no relevant structures have been reported at the EM level from nonflagellated cultures. Our results indicate that the nonflagellated cultures contain plastids (fig. 9B). It is possible that the plastid is in a degenerated form and is therefore overlooked in the nonflagellated stages at the EM level and develops during the course of differentiation into the flagellated stages. Note that ultrastructure alone (e.g., number of membranes) is inadequate to identify plastids (Köhler 2006); immunological labeling of the ultrastructure is required for confirmation, for which our affinity-purified antibody to ispC would be useful.

That the putative P. marinus plastid and the peridinine plastids share a common origin was suggested by the phylogenetic affinity between some MEP pathway genes and those of the peridinine dinoflagellates and the predicted features of the P. marinus TPs. The P. marinus TPs predict that the putative plastid is bounded by 3 membranes because TP with transmembrane helix seems to correspond to plastids with 3 bounding membranes, such as those found in dinoflagellates and euglenoids (Patron et al. 2005). Given that Perkinsus spp. are located basally to dinoflagellates (Cavalier-Smith and Chao 2004; Leander and Keeling 2004), it is reasonable to suppose that a common ancestor harbored a plastid that diverged to form the P. marinus and peridinine plastids. Perkinsus marinus has TPs that are reminiscent of those found in peridinine dinoflagellates (fig. 1), although unusual TPs have been found in dinoflagellates containing haptophyte-derived plastids (Patron et al. 2006); thus, the result supports the hypothesis that haptophyte-derived plastids are a derived feature (Yoon et al. 2005). Study of the MEP pathway genes in other nonphotosynthetic taxa of Dinozoa, such as Oxyrrhis, Syndinium, and Noctiluca, may reveal the presence of cryptic plastids. On the other hand, the Perkinsus MEP pathway genes failed to show a relationship with apicomplexan orthologs (data not shown) possibly because the apicomplexan OTUs were a strong source of a long-branch attraction artifact. The relationship between the putative plastid of P. marinus and the apicoplast is thus still open question. Because the nonphotosynthetic free-living flagellates, Colpodella spp., are phylogenetically basal to the apicomplexans (Kuvardina et al. 2002; Cavalier-Smith and Chao 2004; Skovgaard et al. 2005), investigating whether Colpodella contains plastids would be interesting. The putative plastid in P. marinus will be a key to verifying the chromalveolate hypothesis and will partially rebut its critics (Bodył 2005).

Implications for Plastid Evolution

Our phylogenetic analyses showed that the MEP pathway of PBEs has an apparently mosaic origin in Cyanobacteria, Proteobacteria, and Chlamydia (table 1). Although an earlier study with limited sampling gave similar results for 5 genes (dxs to ispF) (Lange et al. 2000), the present results reinforce the evolutionary origins by using a wide range of bacteria. We also showed that the mosaic pattern was essentially the same among PBEs, including P. marinus, as is often the case with other plastid-related mosaic pathways, such as the Calvin cycle, heme biosynthesis, and the shikimate pathway (Matsuzaki et al. 2004; Oborník and Green 2005; Richards et al. 2006). Thus, we assume that the mosaic pattern was established before radiation of a wide variety of plastids, again supporting a single origin for plastids (Bhattacharya et al. 2004; Matsuzaki et al. 2004). It is natural that the cyanobacterial genes would be introduced during the primary endosymbiotic event, but the contribution of chlamydial genes seems rather peculiar. The apparent affinity could be interpreted as an artifact of interbacterial gene replacement (Lange et al. 2000; Rujan and Martin 2001); however, we consider that the observed mosaic pattern reflects an actual contribution by Chlamydia, because Chlamydia-related higher plant genes are biased toward plastid functions (Brinkman et al. 2002), and the bias is difficult to explain without assuming a connection between Chlamydia and plastids.

The ispG alignment (supplementary fig. S2, Supplementary Material online) and phylogenetic tree (fig. 6) show that the origin of the red algal orthologs clearly differed from that of orthologs from other PBEs. This seems unusual, but other genes also show a similar difference between red algae and other PBEs, for example, enoyl acyl carrier protein reductase (Matsuzaki M, unpublished data) and plastid ferrochelatase (Oborník and Green 2005). This discrepancy in the phylogenetic positions of genes in red algae and the chromalveolates can be explained by the evolutionary histories of the symbiotic hosts (fig. 10). Two competing inferences exist for PBE phylogeny: the chromalveolates are independent of primary PBEs and are inexperienced with primary plastids (Rodríguez-Ezpeleta et al. 2005) (fig. 10A) and the chromalveolates originate from an in-group of primary PBEs and are experienced with primary plastids (Nozaki et al. 2003, 2007) (fig. 10B). Because it is widely accepted that the plastid progenitor in the chromalveolates is a kind of, or belongs to a sister group of, ancestral red algae (Fast et al. 2001; Yoon et al. 2002, 2005), the former topology predicts that red algae and the chromalveolates contain genes of the same cyanobacterial origin and requires extra loss, gain, or transfer events to explain the observed differences (fig. 10A shows a transfer event from the green plant lineage, which is the simplest). For the latter topology, the chromalveolates have 2 possible sources for plastid genes, the host ancestor and the symbiont. In this scenario, genes related to those of green plants can be inherited as constituents of the host genome and can be retained and used after the secondary endosymbiotic event. We prefer the latter model (fig. 10B) because it does not require additional transfer events other than the 2 widely accepted endosymbiotic events. In addition, this scenario can explain the frequently observed incoherence of the phylogenetic affinity of diatom genes, for example, the MEP pathway (figs. 2–7³⁴⁵⁶⁷), the shikimate pathway (Richards et al. 2006), and the heme biosynthesis pathway (Oborník and Green 2005), in which the diatom ortholog for each gene has phylogenetic affinities to either green plants or red algae. Genes with an affinity to green plants and red algae could be derived from the host eukaryote and the plastid progenitor at the secondary symbiotic event, respectively. Furthermore, this scenario implies that the host eukaryote of the secondary endosymbiotic event that forms the chromalveolate ancestor must have retained the primary plastid originating from the ancient primary endosymbiosis because the MEP pathway seems to be specific to PBEs; then, it was replaced with the secondary plastid obtained from the engulfed red alga.

Conclusion

The existence of the plastid and its affinity to those of dinoflagellates, as suggested in the present study, greatly changes our view of P. marinus. It was first described as a pathogenic protist, then subsequently reclassified as a fungal or sporozoan species, and has recently been treated as an alveolate flagellate (Villalba et al. 2004; Adl et al. 2005); however, P. marinus is a cryptic alga in terms of plastid existence. Our identification of a Perkinsus spp. plastid will make this algal organism key in discussions of plastid evolution. Furthermore, just as discovery of the apicoplast has revolutionized malaria chemotherapy, the algal nature of Perkinsus spp. may permit novel approaches for controlling perkinsosis in fisheries.

Preliminary sequence data for P. marinus were obtained from The Institute for Genomic Research via their Web site at http://www.tigr.org/. Sequencing of the P. marinus genome was accomplished with support from the US National Science Foundation. The genome sequence data for Galdieria sulphuraria were obtained from the Michigan State University Galdieria Database (http://genomics.msu.edu/galdieria/). The genome sequence data for Chlamydomonas reinhardtii, Ostreococcus lucimarinus, Phaeodactylum tricornutum, and Thalassiosira pseudonana were obtained from the US Department of Energy Joint Genome Institute (http://www.jgi.doe.gov/) and are provided for use in this publication only. The authors thank Dr Shin-ichiro Maruyama and Dr Takashi Nakada (the University of Tokyo) for helpful discussions and comments on figure presentation and Dr Shigeharu Sato (National Institute for Medical Research, UK) for providing valuable information on related topics. This study was supported by Grants-in-Aid for Creative Scientific Research (16GS0304 to H.N. and 18GS0314 to K.K.), for Scientific Research (17370087 to H.N. and 18073004 for K.K.), and for the Japan Society for the Promotion of Science (JSPS) Fellows (17-10695 to M.M.) from the JSPS and the Ministry of Education, Culture, Sports, Science, and Technology (MEXT); by a matching fund subsidy for “Academic Frontier” Projects (to T.K.) from MEXT; and by the Program for the Promotion of Basic Research Activities for Innovative Biosciences (to T.K.). M.M. is a Research Fellow of the JSPS.

References

Author notes