We synthesized poly(A) primed cDNA using RNA extracted from cultures of D. acuminata taking advantage of the trans-spliced leader sequence present on mature dinoflagellate transcripts. Dinoflagellate transcripts are modified in vivo by the addition of the identical 22 bp trans-spliced leader sequences to the 5' end of all mRNAs [27,28]. Use of this dinoflagellate-specific leader sequence as a 5' primer site during the cDNA amplification step significantly biased the D. acuminata cDNA pool toward full-length, nuclear-encoded, dinoflagellate transcripts. The cDNA was randomly sheared and sequenced using 454 FLX Titanium pyrosequencing. Sequencing produced 10.8 megabases of data that assembled into 5,991 unique contigs. There are no sequenced dinoflagellate genomes available to aid in estimating the number of genes in D. acuminata, however, transcriptional profiling of the dinoflagellate Alexandrium tamarense identified 30,917 unique gene signatures, suggesting that only a fraction of the total transcriptome of D. acuminata was sequenced [29]. From the finished assembly, 816 contigs were fully annotated by Blast2GO, and 16 contigs were identified as potentially plastid-related (i.e., cellular compartment GOslim term of plastid or thylakoid, see additional File 1). Of the 16 candidate contigs, three were determined to be nuclear-encoded, plastid-targeted genes: a photosystem II subunit (psbU), plastid ferredoxin (petF), and a gene encoding an auxiliary light-harvesting protein (LI818). The remaining 12 contigs were either plastid-encoded and introduced into the cDNA pool through mispriming of the oligo dT primer (photosystem I subunit E), or only peripherally related to plastid function (e.g., sec61 protein translocator). Two additional contigs, a second photosytem II gene (psbM) and a plastid phosphate transporter (TPT), were identified as plastid-targeted through sequence similarity searches but were not annotated by Blast2GO because of their high e-value scores. Full-length cDNA sequences complete with the dinoflagellate-specific, trans-spliced leader motif, 5' untranslated region (UTR), and 3' UTR were obtained by PCR from D. acuminata for all five genes and used for subsequent phylogenetic analyses and targeting peptide predictions (Table 1).

In addition, we synthesized G. cryophila and M. rubra poly(A) primed cDNA that was sequenced by the same method, and the data assembled into 17,997 and 27,723 contigs, respectively. These contig numbers are likely overestimates of the transcriptome sizes of these organisms because multiple contigs can represent a single transcript due to gaps in the assembly. However, the 17,997 contigs are consistent with G. cryophila having a gene number similar to other sequenced unicellular algae, which have 5,000-15,000 genes [30]. Likewise, the M. rubra contig number is also in agreement with gene numbers from the sequenced ciliate genomes [31,32]. BLASTN comparisons of the three assemblies showed that none of the nuclear-encoded genes in G. cryophila or M. rubra matched those in D. acuminata at the nucleotide level, indicating that the D. acuminata dataset is not contaminated with ciliate or cryptophyte nuclear transcripts. Cryptophyte homologs of all the nuclear-encoded, plastid-related genes of D. acuminata were identified using BLASTX, with the exception of ferredoxin because it is plastid encoded in cryptophytes and therefore not amplified in the oligo-dT primed cDNA synthesis.

Of the five nuclear-encoded, plastid proteins identified in D. acuminata, only photosystem II subunit M appears to be of cryptophyte origin. The psbM protein is a low molecular mass subunit (33-38 amino acids on average) thought to be involved in photosystem dimer formation [33]. Due to the short length of the alignment, the maximum likelihood phylogenetic analysis of this protein was inconclusive (Figure 2a). However, the Bayesian analysis supports the grouping of D. acuminata and G. cryophila, and the neighbour-joining analysis weakly supports grouping D. acuminata with crytophytes (Figure 2b). The C-terminal end of psbM was also highly similar to the cryptophyte homologs. This region was not included in phylogenetic analyses because of poor conservation among algal groups (for alignment see additional File 2).

Two of the plastid-related proteins, ferredoxin and the triose-phosphate transporter (TPT), group with peridinin dinoflagellates (i.e., containing the ancestral dinoflagellate plastid characterized by the photopigment peridinin) in phylogenetic analyses. These genes have either been retained from a peridinin-containing ancestor of Dinophysis or have been acquired from these dinoflagellates through gene transfer (Figure 3). The plastid TPT is involved in transport of fixed carbon out of the plastid [34]. This protein may provide the mechanism by which D. acuminata benefits from the photosynthesis occurring within the plastid by exporting the products of the Calvin cycle (e.g., glyceraldehyde-3-phosphate) to the cytoplasm. Plastid ferredoxin (petF) is the second dinoflagellate-derived plastid protein in D. acuminata and is responsible for distributing the electrons generated by photosystem I to various reactions in the plastid stroma [35]. The petF gene is encoded on the plastid genome in cryptophytes, and a copy is presumably present in the cryptophyte plastid genome of Dinophysis [36], however plastid gene transcripts are not polyadenylated and therefore are not amplified in oligo dT cDNA synthesis. Although the G. cryophila plastid petF sequence is unknown, the nuclear-encoded copy in D. acuminata is clearly distinct from petF in the cryptophytes G. theta and Rhodomonas salina in our tree (Figure 3b). In addition, the D. acuminata petF cDNA sequence contains a 5' spliced leader, a 3' UTR, and poly (A) tail strongly suggesting it is expressed from the nucleus and not the organelle.

The remaining two proteins, a light harvesting protein (LHP) and psbU, appear to be derived from either haptophytes or fucoxanthin dinoflagellates (i.e., dinoflagellates that have replaced the peridinin plastid with one derived from haptophytes and containing the photopigment fucoxanthin). LHPs shuttle the light energy captured by chlorophyll and accessory pigments to the photosystems, and algal groups have different LHPs depending on their combination of chlorophyll and accessory pigments [37]. We identified only one LHP in D. acuminata, a member of the distinct LI818 LHP family involved in stabilizing the photosystem in response to heat or photodamage [38-40] and may, in some situations, act as a substitute for other LHPs [39]. D. acuminata LI818 weakly groups with homologues from fucoxanthin dinoflagellates, Karlodinium micrum and Karenia brevis (Figure 4a). However, despite large EST datasets for cryptophytes and peridinin dinoflagellates, an LI818 family member has yet to be found in either of these groups of organisms, excluding them as a source of the LI818 gene in D. acuminata. Transcriptome sequencing in G. cryophila produced nine different LHPs, all of which grouped with other cryptophyte or red algal homologs within the Lhcz and Lhcc protein families (for the phylogenetic tree see additional File 3).

The last nuclear-encoded plastid protein identified in D. acuminata is the photosystem II protein psbU. Phylogenetic analyses moderately support the grouping of this protein with haptophytes and the dinoflagellates K. micrum and K. brevis (Figure 4b). PsbU, along with psbO and psbV, is extrinsically associated with photosystem II on the luminal side of the complex and enhances the oxygen evolution activity and structural stability of the complex [41,42]. PsbU is specifically involved in protecting the photosystem from heat and photodamage and may have an increased functional interaction with photosystem II when PsbO is absent [43]. PsbO is nuclear-encoded in eukaryotes and thus far missing from the D. acuminata transcriptome dataset.

Dinophysis and Karenia/Karlodinium are not considered close relatives; therefore, it is likely that the genes encoding LI818 and psbU were acquired through horizontal gene transfer (HGT) [44]. Interestingly, another member of this genus, Dinophysis mitra, is reported to have haptophyte-like plastids, suggesting that these genes could have been acquired during an earlier association with a haptophyte in the ancestor of D. acuminata [45]. HGT of plastid-related genes has been shown to be widespread in chromalveolates, particularly in heterotrophic taxa [46,47].

If D. acuminata nuclear-encoded proteins function in the cryptophyte plastid, they should contain targeting peptides that facilitate their import into the organelle. Primary plastids, such as those in red algae and land plants, are surrounded by two membranes, and these proteins require an N-terminal transit peptide for plastid import. Proteins targeted to secondary plastids with four (e.g., cryptophytes, haptophytes) or three membranes (e.g., peridinin dinoflagellates) possess a bipartite leader sequence composed of a signal peptide, to target the protein to the endoplasmic reticulum, followed by a plastid transit peptide [48-50]. Phylogenetic analyses show that the plastid proteins of D. acuminata are derived from algae with three- or four-membrane bound plastids, suggesting they ancestrally contained both signal peptide and transit peptide. However, only two membranes surround the plastid in D. acuminata. Therefore, we expect the targeting peptides of plastid genes in D. acuminata to resemble the transit peptides found in organisms with primary plastids.

All of the plastid genes in D. acuminata possess putative transit peptides (for sequences see additional File 4). Only ferredoxin is predicted to contain a bipartite leader composed of both a signal peptide and transit peptide. The other four genes have simple transit peptides as predicted by plastid ultrastructure. PsbU, a protein that functions within the plastid lumen, also contains a twin-arginine signal peptide that directs it through the twin-arginine translocase into the plastid lumen [51]. ChloroP predicted transit peptide cleavage sites for all five proteins, but only scored those of psbM and psbU as statistically significant (score > 0.5). WoLF PSORT, a second tool for predicting protein subcellular localization, classified all five D. acuminata proteins as plastid-targeted. The putative transit peptide of ferredoxin contains a phenylalanine motif that is found in red algae and chromalveolates [49,52]. Plastid membrane proteins like TPT possess transit peptides structurally different from those for proteins targeted to the stroma or thylakoid membrane and therefore are not identified by programs like ChloroP [53].

The discovery of plastid-targeted TPT and ferredoxin reveals a link between the metabolism of Dinophysis and its cryptophyte plastid. Presumably, endogenous metabolite transporters would be present in the plastid membrane of newly-acquired plastids, but having a nuclear-encoded transporter protein may allow Dinophysis to extend the useful life of the plastid by replacing damaged proteins. Likewise, proteins involved in stabilization of the photosystem (LI818, psbU, psbM) also may explain why the plastid can be maintained for a long period of time by protecting components of the photosystem.

Although this study has found that Dinophysis does have nuclear-encoded genes that presumably function in the plastid, the results are not consistent with this organism having the ability to permanently maintain the plastid with its native gene complement. We identified far fewer nuclear-encoded plastid genes (only 16 by GO annotation and only 5 when highly curated) than are typically found when sequencing the transcriptome of a truly autotrophic alga (Figure 5). In addition, unlike the transcriptomes of dinoflagellates with permanent plastid replacements, only one gene of the five identified potentially originated from the source of the plastid. In K. brevis and K. micrum, dinoflagellates that have plastids derived from haptophytes, a large proportion of the nuclear-encoded plastid genes were derived from the plastid donor [54,55]. Although our unigene dataset for Dinophysis is not comprehensive, the results of cDNA sequencing from dinoflagellates and other algae indicate that plastid related genes are highly expressed in algae, and we would expect to have discovered many more plastid genes if Dinophysis possessed the full complement necessary for plastid maintenance.

Early steps for establishing a permanent organelle may have occurred in D. acuminata, including the development of a mechanism for metabolite exchange under host control, however, it may be difficult for another critical step in plastid acquisition, massive transfer of genes from the endosymbiont to host nucleus, to happen in Dinophysis. Not only is the plastid not acquired directly from the cryptophyte donor, but also the cryptophyte nucleus and nucleomorph are not retained. Genes encoded on these genomes would be needed to establish a permanent organelle, but because of the indirect method of plastid capture, Dinophysis may not experience the frequent exposure to the cryptophyte genomes that would be necessary for large-scale gene transfer. It is possible that Dinophysis could acquire the necessary genes from other sources, as they have done with four of the genes that we identified. However, this would appear to be more difficult than obtaining a co-evolved set of genes from a single source. The indirect mode of plastid acquisition in D. acuminata may ultimately be a barrier to establishment of a permanent plastid.

D. acuminata strain DAEP01 was established from Eel Pond, Woods Hole, MA in September of 2006. Cells were cultured using the two step feeding system described by Park et al. [56] where D. acuminata (DAEP01) is fed the ciliate, M. rubra (CCMP2563), which is fed the cryptophyte, G. cryophila (CCMP2564). The three algae are cultured in modified f/2-Si medium [57] at 4°C on a 14:10 h L:D cycle. G. cryophila is fed to M. rubra upon reaching a cell density of 500,000 cells mL^-1 (2 mL of G. cryophila is added to 3 mL M. rubra in 20 mL f/2 medium). Once M. rubra cultures are completely clear of G. cryophila, the ciliate is fed to D. acuminata (3 mL M. rubra at ~14,000 cells mL^-1 is added to 2 mL D. acuminata at ~1800 cells mL^-1 in 20 mL f/2 medium). Weekly cell counts of D. acuminata fixed in Utermöhls solution [58] were performed at 100× magnification in a Sedgewick rafter counting chamber. Additional M. rubra cells were added to D. acuminata cultures when the Myrionecta/Dinophysis cell ratio fell below 1:1. Cultures of D. acuminata were inspected for the presence of M. rubra prey by light microscopy and harvested for RNA extraction at least one week after M. rubra was no longer observed in the cultures.

The dinoflagellate specific cDNA synthesis and amplification was performed using the Clontech Super SMART PCR cDNA Synthesis Kit. The first-strand synthesis reaction included 1 μg of total RNA and 84 pmoles of modified 3' SMART CDS Primer IIA (5' AAG CAG TGG TAT CAA CGC AGA GTT TGT TTT TTT TTC TTT TTT TTT TVN 3'). The reaction was incubated at 42°C for 90 min. The first-strand cDNA was column purified using the Clontech NucleoSpin Extract II Kit according to Super SMART cDNA synthesis protocol. The cDNA amplification was performed using the Clontech Advantage 2 PCR kit. The amplification reaction included 40 μl of purified first-strand cDNA, 20 pmoles 5' trans-spliced leader primer (5' TCC GTA GCC ATT TTG GCT CAA G 3'), and 20 pmoles PIIA PCR primer (5' AAG CAG TGG TAT CAA CGC AGA GT 3'). Cycling parameters included an initial denaturation step at 95°C for 1 min followed by 26-29 cycles of 95°C for 15 sec, 65°C for 30 sec, 68°C for 6 min. PCR products were visualized on an agarose gel to confirm expected cDNA size range and cleaned using the Clontech CHROMA SPIN -1000 size selection columns.

First strand synthesis of M. rubra and G. cryophila cDNA was performed using the Invitrogen Superscript III First-Strand Synthesis System. The first-strand reaction included 1 μg of total RNA and 50 pmoles modified oligo dT primer with PIIA tag (see above). The reaction was incubated at 50°C for 90 min. PCR amplification of the first-strand cDNA was performed using the Clontech Advantage 2 PCR kit. The reaction included 10 μl of first-strand reaction and 10 pmoles PIIA PCR primer (5' AAG CAG TGG TAT CAA CGC AGA GT 3'). Cycling parameters included an initial denaturation step at 95°C for 1 min followed by 18 cycles of 95°C for 30 sec, 58°C for 30 sec, 68°C for 6 min. PCR products were visualized on an agarose gel to confirm expected cDNA size range and cleaned using the Clontech CHROMA SPIN -400 size selection columns.

The cDNA was sequenced with a 454 FLX pyrosequencing machine at the Arizona Genomics Institute (Tucson, AZ, USA) and data was assembled with SeqMan (DNASTAR, Madison, WI, USA) or gsAssembler (Roche NimbleGen, Inc., Madison, WI, USA). D. acuminata contigs were queried against the M. rubra and G. cryophila assemblies using BLASTN to determine if sequences from these species were present in the D. acuminata dataset. Contigs were annotated using Blast2GO [59], and D. acuminata plastid gene fragments were identified using those annotations as well as BLASTX searches against the non-redundant protein database at NCBI. Full-length transcripts were independently verified by PCR from new and unamplified D. acuminata cDNA generated using the SuperScript III First-Strand Synthesis System (Invitrogen Co., Carlsbad, CA, USA). Gene specific internal primers were designed from 454 sequence fragments and paired with either the 5' trans-spliced leader primer or an anchored oligo dT primer for PCR amplification of the 5' and 3' ends, respectively. PCR products were sequenced directly or cloned into pGEM-T Easy vectors (Promega, Madison, WI, USA) and sequenced using BigDye dye terminator sequencing (Applied Biosystems, FosterCity, CA USA) on an automated DNA sequencer (ABI 3730 × l, Applied Biosystems). D. acuminata plastid protein sequences were searched against the G. cryophila transcriptome assembly using TBLASTN to identify G. cryophila homologs. The plastid cDNA sequences have been deposited in Genbank (Accession numbers HM125141-HM125145).

Amino acid sequences of D. acuminata and G. cryophila were aligned with algal sequences from Genbank using ProbCons with default parameter settings [60]. Gblocks was used to remove poorly aligned regions of the alignments [61]. Distance analyses were performed in PAUP* v4.0b10 [62] with 100 bootstrap replicates using a neighbor-joining search with minimum evolution as the objective function and uncorrected distances. The best-fit model for each alignment was identified by ProtTest v1.4 [63] using the AIC model selection criterion and a BIONJ tree. The ProtTest best-fit evolutionary model for each data set was applied to the maximum likelihood (ML) and Bayesian analyses. ML trees were inferred using the Cipres web portal RAxML rapid bootstrapping and ML search [64,65]. Bayesian analyses were performed in BEAST v1.4.7 [66] assuming an uncorrelated lognormal relaxed molecular clock model using the substitution and site heterogeneity models determined by ProtTest and with a Yule process speciation tree prior. Two independent runs (5 million - 10 million steps) were performed for each analysis and terminated once examination of the Markov chain Monte Carlo (MCMC) samples in Tracer v1.4.1 http://beast.bio.ed.ac.uk/Tracer indicated convergence of the chains with estimated sample sizes greater than 200. The maximum clade credibility tree topology was determined from resulting MCMC tree samples using TreeAnnotator v1.4.7 [66]. Amino acid sequences were screened for targeting peptides and associated cleavage sites using SignalP v3.0, ChloroP v1.1, TatP v1.0, and Wolf PSORT [67].