Patterns in evolutionary origins of heme, chlorophyll a and isopentenyl diphosphate biosynthetic pathways suggest non-photosynthetic periods prior to plastid replacements in dinoflagellates

Proteins with evolutionarily diverse origins comprise the Karenia brevis and Karlodinium veneficum pathways

To convert glutamyl-tRNA to glutamate-1-semialdehyde, Karenia brevis and Karlodinium veneficum possesses two versions of glutamyl-tRNA reductase (GTR). We consider “Karenia-1” and “Karlodinium-1” sequences as VI-type, as they were nested within a clade with those of peridinin-containing dinoflagellates, and the clade as a whole received full statistical support in the GTR phylogeny (Fig. 2A). On the other hand, the second version of GTR in the two kareniacean species (Karenia-2 and Karlodinium-2) were most likely acquired from the haptophyte endosymbiont (i.e., EA-type). The two GTR sequences were nested within the haptophyte clade, and the haptophyte clade (including the Karenia-2 and Karlodinium-2 sequences) was supported by a MLBP of 66% and a BPP of 0.99 (Fig. 2A).

Aminolevulinic acid (ALA) is synthesized from glutamate-1-semialdehyde by glutamate-1-semialdehyde 2,1-aminomutase (GSAT). We identified a single version of GSAT in both Karenia brevis and Karlodinium veneficum. The possibility of the two GSAT sequences being acquired from the haptophyte endosymbiont can be excluded, as the haptophyte sequences (except that of Pavlova sp.) formed a robust clade and excluded the kareniacean sequences in the GSAT phylogeny (Fig. 2B). Nevertheless, the phylogenetic origin of either Karlodinium veneficum or Karenia brevis GSAT sequence could not be pinpointed any further. The GSAT sequences of K. brevis and peridinin-containing dinoflagellates grouped together in the both ML and Bayesian phylogenies, but their monophyly was poorly supported (MLBP of 8% and BPP < 0.50). The Karlodinium veneficum GSAT sequence showed no specific affinity to any sequence examined here. Thus, we withhold to conclude the precise origins of the GSAT sequences of Karenia brevis and Karlodinium veneficum in this study.

Synthesis of porphobilinogen from ALA is likely catalyzed by a single aminolevulinic acid dehydratase (ALAD) homologue in both Karenia brevis and Karlodinium veneficum. Both Karenia brevis and Karlodinium veneficum ALAD sequences formed a robustly supported clade with those of peridinin-containing dinoflagellates (MLBP of 99% and BPP of 1.0; Fig. 2C), suggesting that the two sequences are of VI-type.

We identified two versions of porphobilinogen deaminase (PBGD), which deaminates porphobilinogen to synthesize hydroxymethylbilane, in Karenia brevis (Karenia-1 and 2). The PBGD phylogeny (Fig. 2D) recovered (i) a clade of the sequences of peridinin-containing dinoflagellates and Karenia-1 sequence with a MLBP of 100% and a BPP of 0.99, and (ii) a clade of the haptophyte and Karenia-2 sequences with a MLBP of 73% and a BPP of 0.83. Thus, K. brevis seemingly uses both VI-type and EA-type enzymes for hydroxymethylbilane synthesis. We identified four PBGD sequences in K. veneficum (Karlodinium-1-4), which clearly share a single ancestral sequence. The Karlodinium veneficum clade appeared to be distant from the clade comprising the sequences of peridinin-containing dinoflagellates and the Karenia-1 sequence, suggesting that the Karlodinium-1-4 sequences are not of VI-type. The K. veneficum clade was connected to the haptophyte clade (including the Karenia-2 sequence) with a MLBP of 47% and a BPP of 0.67 (highlighted by an arrowhead in Fig. 2D). The statistical support for the particular node is insufficient to conclude or exclude the haptophyte origin of the Karlodinium sequences with confidence. Thus, we determined to leave the origin of the Karlodinium sequences uncertain.

We assessed the evolutionary origins of two distinct versions of UROS in Karlodinium veneficum (Karlodinium-1 and 2), and a single UROS sequence detected in an expressed-sequence tag data from Karenia brevis (Fig. 3A). The Karenia sequence, Karlodinium-1 and peridinin-containing dinoflagellate sequences grouped together with a MLBP of 76% and a BPP of 0.95. Thus, the Karenia and Karlodinium-1 sequences are most likely of VI-type. The Karlodinium-2 appeared to be remote from the dinoflagellate or haptophyte clade, suggesting this sequence was of LA-type (Fig. 3A). The UROS phylogeny connected the Karlodinium-2 sequence and that of a red alga Rhodosorus marinus with a MLBP of 95% and a BPP of 0.93, and the “Karlodinium + Rhodosorus” clade was nested within the bacterial sequences. The particular clade could have been generated by two separate gene transfers from a single bacterium (or two closely related bacteria) to K. veneficum and Rhodosorus. Alternatively, the combination of the first gene transfer from a bacterium to either K. veneficum or Rhodosorus, and the second one between the two eukaryotes may have produced the “Karlodinium + Rhodosorus” clade in the UROS phylogeny. In either of the two scenarios, we can regard the Karlodinium-2 sequence as LA-type. We need to increase the number of red algal UROS sequences in the future studies to retrace the precise origins of the Karlodinium and Rhodosorus sequences.

Four acetyl side chains in uroporphyrinogen III are removed by uroporphyrinogen decarboxylase (UROD) to generate coproporphyrinogen III. Pioneering studies revealed that photosynthetic eukaryotes with complex plastids possess evolutionarily distinct, multiple versions of UROD (Kořený et al., 2011; Cihlář et al., 2016). The UROD sequences of peridinin-containing dinoflagellates were split into three clades in the UROD phylogeny (designated as D1, D2 and D3 clades in Fig. 3B), suggesting that the three distinct versions have already been established in the ancestral dinoflagellate. Likewise, haptophytes were found to possess three distinct versions of UROD (designated as H1, H2 and H3 clades in Fig. 3B). We here identified five and four UROD sequences in Karenia brevis and Karlodinium veneficum, respectively. Among the five sequences identified in K. brevis, the “Karenia-1, 2 and 4” sequences were considered as EA-type, as they were placed within the haptophyte sequences in the UROD phylogeny (Fig. 3B). H1 clade including the Karenia-1 and 2 sequences received a MLBP of 100% and a BPP 1.0. The Karenia-4 sequence and the haptophyte sequences (except that of Pavlova) formed H3 clade, of which monophyly was supported by a MLBP of 79% and a BPP of 0.99. The “Karenia-3” sequence grouped with the sequence of a euglenid Eutreptiella gymnastica with a MLBP of 100% and a BPP of 0.99. As the Karenia-3 sequence showed no affinity to the dinoflagellate or haptophyte clade in the UROD phylogeny, we categorize it in LA-type. The intimate phylogenetic affinity between the Karenia-3 and E. gymnastica sequences hints either (i) gene transfer between the two organisms or (ii) two separate gene transfers from an as-yet-unknown organism to K. brevis and E. gymnastica. It is also important to improve the sequence sampling from euglenids in future phylogenetic studies to understand precisely the biological event generated the union of the Karenia-3 and E. gymnastica sequences. The “Karenia-5” sequence is most likely descended from one of the UROD versions established in the ancestral dinoflagellate (i.e., VI-type), as this sequence participated in D3 clade, of which monophyly received a MLBP of 100% and a BPP of 1.0. We also assessed the origins of four K. veneficum sequences (Karlodinium-1-4) based on the UROD phylogeny (Fig. 3B). The Karlodinium-1 and 2 sequences were nested within H1 clade, and the Karlodinium-4 sequence were placed within H3 clade. Thus, we conclude that the three UROD sequences are of EA-type. The Karlodinium-3, one of the L. chlorophorum sequences (see below), and diatom sequences formed a robust clade (MLBP of 100% and BPP of 1.0; highlighted by an arrowhead in Fig. 3B), suggesting that the Karlodinium-3 sequence were laterally acquired from a diatom (i.e., LA-type).

Coproporphyrinogen III is oxidized by CPOX to yield protoporphyrinogen IX. We identified a single version of CPOX in Karenia brevis, but none in Karlodinium veneficum. The CPOX phylogeny (Fig. 3C) recovered a clade comprising the sequences of peridinin-containing dinoflagellates and K. brevis with a MLBP of 98% and a BPP of 0.99. Thus, K. brevis uses a VI-type CPOX.

Protoporphyrinogen IX is further oxidized by PPOX to obtain protoporphyrin IX. The PPOX sequences of peridinin-containing dinoflagellates were separated into two distinct clades labeled as “D1” and “D2” (Fig. 4A), and both received MLBPs of 100% and BPPs ≥ 0.97. The PPOX sequences in D2 clade are likely cytosolic version, as these dinoflagellate sequences, as well as those of other photosynthetic eukaryotes bearing complex plastids (chlorarachniophytes, Euglena gracilis and V. brassicaformis), formed a robust clade with the PPOX sequences of heterotrophic eukaryotes (MLBP of 93% and BPP of 0.99; highlighted by an arrowhead in Fig. 4A). We identified two versions of PPOX in Karenia brevis (Karenia-1 and 2), while a single version was identified in Karlodinium veneficum. The Karenia-1 sequence fell into D1 clade (Fig. 4A), suggesting that this PPOX sequence is of VI-type. On the other hand, the PPOX phylogeny united the Karenia-2 and a single PPOX of Karlodinium veneficum with the sequences of stramenopiles, haptophytes and L. chlorophorum with a MLBP of 100% and a BPP of 0.99 (highlighted by a double-arrowhead in Fig. 4A). As the bipartitions within the clade were principally unresolved (Fig. 4A), we cannot exclude the possibility of the Karenia-2 and Karlodinium-1 sequences group with the haptophyte sequences in this clade. The Karenia-2 and Karlodinium-1 sequences are definitely not of VI-type, but the phylogenetic resolution was not sufficient to classify the two sequences into of EA-type or LA-type. Thus, we decide to leave the origins of the Karenia-2 and Karlodinium-1 sequences uncertain.

In the last step in the heme biosynthesis, ferrochelatase (FeCH) converts protoporphyrin IX to protoheme. None of the FeCH sequences identified in Karenia brevis and Karlodinium veneficum appeared to be of VI-type. K. brevis possesses three distinct versions of FeCH (Karenia-1-3). The FeCH phylogeny (Fig. 4B) united the Karenia-1 sequence with one of four versions of FeCH in L. chlorophorum (Lepidodinium-3) with a MLBP of 100% and a BPP of 1.0, and this union was then connected specifically to γ-proteobacterial sequences with a MLBP of 75% and a BPP of 0.99 (highlighted by an arrowhead in Fig. 4B). This subtree can be explained by two sequential gene transfer events, namely the first gene transfer from a γ-proteobacterium to either L. chlorophorum or K. brevis, and the second one between the two dinoflagellates. Consequently, the Karenia-1 and Lepidodinium-3 sequences can be of LA-type, and traced back to the bacterial sequence. The Karenia-2 sequence is unlikely to be of VI-type, as the sequences of peridinin-containing dinoflagellates (and one of the sequences of L. chlorophorum) were united with a MLBP of 100% and a BPP of 1.0 (Fig. 4B). The Karenia-2 sequence, which showed no clear affinity to the haptophyte sequences, is unlikely to be of EA-type (Fig. 4B). Thus, we conclude that the Karenia-2 sequence is of LA-type, although the precise donor remains uncertain. Finally, the FeCH phylogeny grouped the Karenia-3 sequence and a single FeCH sequence identified in Karlodinium veneficum with the cyanobacterial sequences, and their monophyly was supported by a MLBP of 76% and a BPP of 0.94 (highlighted by a double-arrowhead in Fig. 4B). We conclude that the two dinoflagellate sequences are of cyanobacterial origin (i.e., LA-type).

Pioneering studies on kareniacean dinoflagellates revealed the evolutionary chimeric natures of the plastid proteomes of Karenia brevis and Karlodinium veneficum, which comprises VI-, EA- and LA-type proteins (Ishida & Green, 2002; Nosenko et al., 2006; Patron, Waller & Keeling, 2006; Burki et al., 2014). This study further enables us to evaluate the precise contribution of EA-type proteins on the evolution of the heme biosynthetic pathway in kareniacean dinoflagellates—EA-type proteins found only in three out of the nine steps, namely: (i) GTR in Karenia brevis and Karlodinium veneficum, (ii) PBGD in Karlodinium veneficum, and (iii) UROD in Karenia brevis and Karlodinium veneficum.

Little impact of endosymbiotic gene transfer on the L. chlorophorum pathway

We identified three versions of GTR in L. chlorophorum (Lepidodinium-1-3). The Lepidodinium-1 sequence grouped with the sequence of peridinin-containing dinoflagellates (and the Karenia-1 and Karlodinium-1 sequences), and this “dinoflagellate” clade received a MLBP of 100% and a BPP of 1.0 (Fig. 2A). Thus, we conclude that the Lepidodinium-1 sequence is of VI-type. The Lepidodinium-2 and 3 sequences were tied to each other with a MLBP of 100% and a BPP of 1.0, and this clade was connected with the “dinoflagellate” clade described above (Fig. 2A). However, the phylogenetic affinity between the two clades received little statistical support (MLBP of 38% and BPP < 0.50; highlighted by an arrowhead in Fig. 2A), suggesting that the Lepidodinium-2 and 3 sequences were unlikely of VI-type. The GTR sequences of land plants and green algae (plus a euglenid E. gymnastica) formed a clade supported by a MLBP of 100% and a BPP of 0.99, and appeared to be distant from the two sequences of L. chlorophorum (Fig. 2A). Thus, we can exclude the green algal (endosymbiont) origin of the Lepidodinium-2 and 3 sequences. Combined, the two sequences are regarded as LA-type, although the donor of the GTR gene to L. chlorophorum remains unclear.

We failed to classify two out of the three versions of GSAT identified in L. chlorophorum. In the GSAT phylogeny (Fig. 2B), the “Lepidodinium-1” and “Lepidodinium-2” sequences fell separately into the cluster of the sequences of peridinin-containing dinoflagellates and K. brevis, but this clade as a whole received no significant statistical support. Thus, we left the origins of the Lepidodinium-1 and 2 sequences uncertain in this study. On the other hand, the “Lepidodinium-3” sequence was excluded from the clade comprising the sequences of diverse photosynthetic eukaryotes and cyanobacteria with a MLBP of 100% and a BPP of 1.0, and connected to the sequences of Streptomyces coelicolor, Mycobacterium tuberculosis and Corynebacterium diphtheriae with a MLBP of 71% and a BPP of 0.91 (highlighted by an arrowhead in Fig. 2B). This tree topology suggests that the Lepidodinium-3 sequence was acquired from a bacterium (i.e., LA-type), albeit we cannot pinpoint the bacterium donated a GSAT gene to L. chlorophorum.

Two versions of ALAD were identified in L. chlorophorum (Lepidodinium-1 and 2). The sequences of peridinin-containing dinoflagellates, the Lepidodinium-1 sequence and the sequences of Karenia brevis and Karlodinium veneficum clustered with a MLBP of 99% and a BPP of 1.0 in the ALAD phylogeny (Fig. 2C). Thus, we conclude the Lepidodinium-1 sequence as VI-type. The Lepidodinium-2 sequence was distantly related to the sequences of peridinin-containing dinoflagellate or green algae/land plants, but showed no strong affinity to any clade/sequence in the ALAD phylogeny (Fig. 2C). Thus, we propose the Lepidodinium-2 sequence as LA-type, albeit its precise origin was unresolved in the ALAD phylogeny.

Two versions of PBGD and a single version of UROS identified from L. chlorophorum. The PBGD phylogeny (Fig. 2D) united the two sequences of L. chlorophorum together with a MLBP of 100% and a BPP of 0.99, but this clade showed little phylogenetic affinity to the sequences of peridinin-containing dinoflagellates, green algae/land plants or any sequences considered in the phylogenetic analysis. Likewise, the UROS phylogeny (Fig. 3A) recovered no clear affinity of the sequence of L. chlorophorum to other sequences including those of peridinin-containing dinoflagellates or green algae/land plants. Thus, L. chlorophorum likely uses LA-type PBGD and UROS, but their precise origins remain uncertain.

We identified seven versions of UROD in L. chlorophorum, and six of them showed clear affinities to the sequences of peridinin-containing dinoflagellates. In the UROD phylogeny (Fig. 3B), the sequences of peridinin-containing dinoflagellates formed three distinct clades (D1-3 clades), and each of these clades enclosed at least one sequence of L. chlorophorum, namely: (i) the “Lepidodinium-2” sequence in D1 clade, (ii) “Lepidodinium-3 and 4” sequences in D2 clade, and (iii) “Lepidodinium-5, 6 and 7” sequences in D3 clade. Thus, the six sequences described above are concluded as VI-type. The “Lepidodinium-1” sequence and sequences of diatoms (and K. veneficum) formed a clade with a MLBP of 100% and a BPP of 1.0 (highlighted by an arrowhead in Fig. 3B), suggesting that this version was acquired from a diatom (i.e., LA-type).

Three versions of CPOX were identified in L. chlorophorum, but none of them was of VI-type or EA-type. The CPOX phylogeny (Fig. 3B) placed the “Lepidodinium-1” sequence in a remote position from the sequences of peridinin-containing dinoflagellates or green algae/land plants. Instead, the Lepidodinium-1 sequence grouped with the bacterial sequences, as well as the eukaryotic sequences for the cytosolic pathway, with a MLBP of 77% and a BPP of 0.87 (highlighted by an arrowhead in Fig. 3C). This protein most likely bears no N-terminal extension (Table S5). Altogether, we propose that the Lepidodinium-1 sequence encodes a cytosolic CPOX enzyme involved in C4 pathway, and omitted from the discussion below. The “Lepidodinium-2” and “Lepidodinium-3” sequences share high sequence similarity in the mature protein region, while their N-terminal regions are distinct from each other (Table S5). In the CPOX phylogeny, the two Lepidodinium sequences showed a specific affinity to the diatom sequences (MLBP of 86% and BPP of 0.98; highlighted by a double-arrowhead in Fig. 3C), instead of the dinoflagellate or green plant sequences. Thus, the ancestral CPOX of the two L. chlorophorum sequences was most likely of LA-type.

We conclude that L. chlorophorum possesses two distinct VI-type (Lepidodinium-1 and 3) and a single LA-type PPOX (Lepidodinium-2). The PPOX sequences of peridinin-containing dinoflagellates were split into two distinct clades (D1 and D2), and the two clades received strong statistical support from both ML bootstrap and Bayesian analyses (Fig. 4A). The Lepidodinium-1 and 3 sequences were included in D1 and D2 clades, respectively. As the PPOX sequences (including the Lepidodinium-3 sequence) in D2 clade can be considered as the cytosolic version, we did not discuss the Lepidodinium-3 sequence further. The PPOX phylogeny recovered a robust clade comprising the Lepidodinium-2 sequence and, the sequences of haptophytes, stramenopiles, Karlodinium veneficum and Karenia brevis (MLBP of 100% and BPP of 0.99; highlighted by a double-arrowhead in Fig. 4A). Thus, the Lepidodinium-2 sequence was acquired from an organism distantly related to dinoflagellates or green algae/land plants (i.e., LA-type).

In L. chlorophorum, we identified four versions of FeCH. The “Lepidodinium-1” sequence is of VI-type, as this sequence apparently shared the origin with the sequences of peridinin-containing dinoflagellates, and their monophyly was supported with a MLBP of 100% and a BPP of 1.0 (Fig. 4B). On the other hand, we consider the rest of the sequences of L. chlorophorum as LA-type, as the “Lepidodinium-2,” “Lepidodinium-3” and “Lepidodinium-4” sequences appeared to be distantly related to the dinoflagellate clade described above or the green algae/land plant sequences in the FeCH phylogeny (Fig. 4B). The precise positions of the Lepidodinium-2 and Lepidodinium-4 sequences were unresolved, and it remains unclear how L. chlorophorum acquired the two versions of FeCH. On the other hand, the FeCH phylogeny united the “Lepidodinium-3” and Karenia-1 sequences together (MLBP of 100% and BPP of 1.0), and this dinoflagellate clade was then connected to two γ-proteobacterial sequences with a MLBP of 75% and a BPP of 0.99 (highlighted by an arrowhead in Fig. 4B). We have already proposed the two scenarios for the origin of the Lepidodinium-3 and Karenia-1 sequences, in which two lateral gene transfers were invoked (see the previous section for the details).

The most prominent feature in the L. chlorophorum pathway is that no gene from the green algal endosymbiont was detected in the heme biosynthesis. Instead, genes transferred from organisms related to neither host (dinoflagellate) nor endosymbiont (green alga) largely contributed to the L. chlorophorum pathway. The impact of lateral gene transfer on the heme biosynthesis is most prominent in the steps catalyzed by PBGD, UROS and CPOX, in which only LA-type proteins were identified.

Large impact of EGT on the Karenia brevis and Karlodinium veneficum pathways

Mg chelatase (MgCH), which comprises three subunits ChlD, ChlH and ChlI, inserts Mg²⁺ to protoporphyrin IX. In Karlodinium veneficum, ChlI is plastid-encoded (Gabrielsen et al., 2011) and the rest of the subunits were nucleus-encoded (see below). Although no plastid genome data is available for Karenia brevis, chlI is predicted to reside in the plastid genome of Karenia mikimotoi based on a pioneering study on the plastid transcriptome of this species (Dorrell, Hinksman & Howe, 2016). In this study, we identified the partial chlI contig in the transcriptomic data of K. brevis (contig No. 0173787962). The 3′ terminus of this contig was not completed, we could not conclude whether the chlI transcript received the poly(U) tail, which is the unique RNA modification occurred in peridinin-containing plastids as well as the non-canonical plastids of Karenia mikimoti and Karlodinium veneficum (Dorrell & Howe, 2012; Richardson, Dorrell & Howe, 2014). Although the precise genome harboring chlI in K. brevis remains uncertain, we assume that the Karenia brevis chlI sequence was transcribed from the plastid genome as demonstrated in Karenia mikimotoi by Richardson, Dorrell & Howe (2014). As the current study focuses on nucleus-encoded proteins involved in plastid metabolisms, we stopped examining the origin and evolution of ChlI in the two kareniacean species (and L. chlorophorum; see below) any further.

We here examine the evolutionary origins of two nucleus-encoded subunits of MgCH, ChlH and ChlD, in Karenia brevis and Karlodinium veneficum. A single version of ChlD was identified in Karenia brevis and Karlodinium veneficum. The ChlD phylogeny (Fig. 5A) recovered a clade of the sequences of haptophytes, Karenia brevis and Karlodinium veneficum with full statistical support, suggesting that the kareniacean ChlD sequences are of EA-type. Both Karenia brevis and Karlodinium veneficum possess two distinct versions of ChlH (Karenia-1 and 2, and Karlodinium-1 and 2). ChlH sequences can be split into two distinct clades, “ChlH-1” and “ChlH-2” (Fig. 5B) as a previous study reported (Lohr, Im & Grossman, 2005). ChlH-1 sequences are ubiquitously distributed in photosynthetic organisms, while ChlH-2 sequences have been found in restricted lineages. The sequences of green algae/land plants formed two distinct clades (Gp1 and 2 clades). In the ChlH phylogeny (Fig. 5B), the haptophyte ChlH-1 sequences, the Karenia-1 and Karlodinium-1 sequences formed a clade supported with a MLBP of 98% and a BPP of 1.0. In contrast, the Karenia-2 and Karlodinium-2 sequences were nested within the Gp2 clade containing ChlH-2 sequences of green algae and a euglenid, and their monophyly received a MLBP of 81% and a BPP of 0.98 (highlighted by an arrowhead in Fig. 5B). Thus, we conclude that Karenia brevis and Karlodinium veneficum possess ChlH-1 sequences acquired from the haptophyte endosymbiont (i.e., EA-type), while their ChlH-2 sequences are of green algal origin (i.e., LA-type).

S-adenosylmethionine: Mg-protoporphyrin O-methyltransferase (MgPMT) converts Mg-protoporphyrin IX to Mg protoporphyrin IX monomethyl ester. The MgPMT sequences of Karenia brevis and Karlodinium veneficum grouped with the haptophyte sequences (except the one of Pavlova), and their monophyly was supported by a MLBP of 67% and a BPP of 0.99 (highlighted by an arrowhead in Fig. 5C). The two kareniacean species most likely use the MgPMT acquired from the haptophyte endosymbiont (i.e., EA-type).

No MgPME cyclase has been identified in any dinoflagellates regardless of plastid-type, and we could not examine the origin and evolution of this enzyme (see above). However, DVR, which recognizes the product of MgPME cyclase (divinyl protochlorophyllide) as the substrate and generate protochlorophyllide, were identified in both dinoflagellates bearing peridinin and those bearing non-canonical plastids. We identified both N-DVR and F-DVR sequences in Karenia brevis, while only N-DVR sequence was found in Karlodinium veneficum. The F-DVR phylogeny (Fig. 5D) recovered the clade of the sequences of K. brevis and haptophytes with robust statistical support, suggesting that K. brevis was acquired from the haptophyte endosymbiont (i.e., EA-type).

The N-DVR phylogeny (Fig. 5E) united the sequences of Karenia brevis and Karlodinium veneficum together with a MLBP of 98% and a BPP of 0.99, and the kareniacean clade showed no clear phylogenetic affinity to either haptophyte sequences or other dinoflagellate sequences. Instead, the kareniacean clade grouped with the sequences of stramenopiles, haptophytes, and two chromerids (Chromera and Vitrella) supported with a MLBP of 100% and a BPP of 1.0 (highlighted by an arrowhead in Fig. 5E). In this large clade, the affinity between the kareniacean clade and haptophyte sequences remained uncertain, and we are not sure of the precise origin of the two kareniacean sequences.

In land plants, conversion of protochlorophyllide to chlorophyllide a is catalyzed by the light-dependent and/or light-independent forms of POR. The light-dependent POR is nucleus-encoded, while the light-independent form comprises three plastid-encoded subunits (ChlB, ChlL and ChlN). No gene for light-independent POR was found in the sequenced region of the Karlodinium veneficum plastid genome (Gabrielsen et al., 2011) or plastid transcriptomic data of Karenia mikimotoi (Dorrell, Hinksman & Howe, 2016), implying that kareniacean species lack the light-independent version. We identified two and three distinct versions of the light-dependent POR in Karenia brevis and Karlodinium veneficum, respectively (Karenia-1 and 2, and Karlodinium-1-3), as demonstrated in Hunsperger, Randhawa & Cattolico (2015). In the POR phylogeny (Fig. 6A), the Karenia-1 and Karlodinium-1 sequences grouped with the sequences of stramenopiles, cryptophytes and haptophytes, and their monophyly was supported by a MLBP of 95% and a BPP of 0.98 (highlighted by an arrowhead in Fig. 6A). The Karenia-1 and Karlodinium-1 sequences cannot be of VI-type, as the two sequences are distantly related to other dinoflagellate sequences included in the alignment. However, it is difficult to classify the Karenia-1 and Karlodinium-1 sequences into EA-type or LA-type, as the relationship between the two dinoflagellate sequences and the haptophyte sequences was unresolved in the particular clade. Thus, we leave the origins of the two sequences uncertain in this study. The Karenia-2, Karlodinium-2 and Karlodinium-3 sequences robustly grouped together within the haptophyte sequences, and this “haptophyte” clade received a MLBP of 98% and a BPP of 0.99 (Fig. 6A). Thus, we conclude that these sequences were acquired from the haptophyte endosymbiont (i.e., EA-type).

The final step of the Chl a biosynthesis is catalyzed by Chl synthase (CS). A single version of CS was identified in Karenia brevis and Karlodinium veneficum. The two kareniacean sequences were placed within the haptophyte clade in the CS phylogeny, and the “haptophyte” clade as a whole was supported by a MLBP of 98% and a BPP of 1.0 (Fig. 6B). Thus, the sequences of Karenia brevis and Karlodinium veneficum are considered as EA-type.

The phylogenetic analyses described above revealed that EA-type proteins operate in all the steps converting protoporphyrin IX to Chl a in Karlodinium veneficum and/or Karenia brevis (except the step catalyzed by MgPME cyclase; see above). In addition, the common ancestor of Karenia brevis and Karlodinium veneficum should have possessed LA-type ChlH-2, POR and N-DVR, which were acquired from phylogenetically diverse eukaryotes distantly related to dinoflagellates or haptophytes.

Genetic influx from phylogenetically diverse organisms shaped the L. chlorophorum pathway

As discussed in the previous section, the Chl a biosynthetic pathway in Karenia brevis and Karlodinium veneficum appeared to be shaped by the genes transferred from the endosymbiont (i.e., a haptophyte in the above systems). Curiously, this is not the case for the same pathway in L. chlorophorum, of which plastid was derived from a pedinophyte green alga. Note that we present no result from the L. chlorophorum ChlI, which turned out to be plastid-encoded (Kamikawa et al., 2015a). We identified eight proteins involved in the Chl a biosynthetic pathway in L. chlorophorum, and assess their phylogenetic origins individually (Figs. 5A–5E and 6A–6B). Among the eight proteins examined here, we conclude MgPMT as a sole EA-type protein among those involved in the L. chlorophorum pathway. The MgPMT phylogeny (Fig. 5C) placed the L. chlorophorum sequence within a radiation of the sequences of green algae, land plants and chlorarachniophytes, and their monophyly was supported by a MLBP of 88% and a BPP of 0.98.

Our surveys and phylogenetic analyses revealed that VI-type proteins were almost entirely eliminated from the L. chlorophorum pathway. We identified only one of the two versions of POR (Lepidodinium-1) as VI-type. The POR phylogeny (Fig. 6A) recovered two distinct clades of the sequences of peridinin-containing dinoflagellates (D1 and D2 clades), and placed the Lepidodinium-1 sequence within D1 clade. D1 clade containing the Lepidodinium-1 sequence as a whole received a MLBP of 97% and a BPP of 0.99.

We could not clarify the origin of the N-DVR sequence of L. chlorophorum, of which position was resolved in neither ML nor Bayesian phylogenetic analyses (Fig. 5E). The Lepidodinium sequence was excluded from the robust clade of the sequences of peridinin-containing dinoflagellates, suggesting that this sequence cannot be of VI-type. However, it is difficult to pursue the origin of the Lepidodinium sequence any further, as the N-DVR phylogeny failed to exclude a potential affinity between the sequence of interest and the green algal/land plant sequences (Fig. 5E).

We classified ChlD, ChlH, one of the two versions of POR (Lepidodinium-2), F-DVR and CS into LA-type. In the ChlD phylogeny, the sequence of L. chlorophorum appeared to be excluded from the sequences of peridinin-containing dinoflagellates and those of the green algal/land plant sequences (Fig. 5A), suggesting that this sequence cannot be of VI-type or EA-type. Instead, the sequence of L. chlorophorum, as well as those of chromerids, cryptophytes and Eutreptiella, were placed within the radiation of the stramenopile sequences, and their monophyly was supported by a MLBP of 96% and a BPP of 0.99. This tree topology prompts us to propose that L. chlorophorum ChlD was acquired from a stramenopile (i.e., LA-type).

To our surprise, the ChlH, POR and CS phylogenies placed the sequences of L. chlorophorum within the radiation of the chlorarachniophyte sequences, and their monophylies received MLBPs > 96% and by BPPs > 0.62 (Figs. 5B, 6A and 6B). We here propose multiple gene transfers from chlorarachniophytes to the ancestral Lepidodinium cell to interpret the aforementioned tree topologies. The putative chlorarachniophyte origins of the three genes are not contradict to a recent molecular clock analysis, in which the green algal endosymbiosis leading to the Lepidodinium plastids was predicted to occur more recently than that leading to the chlorarachniophyte plastids (Jackson et al., 2018).

The F-DVR phylogeny reconstructed a robust affinity between the sequence of L. chlorophorum and the chlorarachniophyte clade within the land plant/green algal sequences (Fig. 5D). This tree topology requires a combination of an EGT and LGT, as Lepidodinium species and chlorarachniophytes acquired their plastids commonly from green algae. Either ancestral Lepidodinium cell or the ancestral chlorarachniophyte acquired an F-DVR gene from the green algal endosymbiont (i.e., EGT), followed by the second gene transfer between the two organisms (i.e., LGT). Thus, we have to leave the evolutionary origin of the L. chlorophorum F-DVR uncertain.

The phylogenetic analyses described above revealed that EGT was much less significant in the L. chlorophorum pathway than the kareniacean pathway (see above). Instead, chlorarachniophytes seemingly donated the genes encoding the proteins involved in three out of the five steps in the L. chlorophorum pathway.