In the original Phytophthora genome analyses, 855 genes were considered to be of algal or cyanobacterial origin, and 30 most likely candidates were subject to further detailed analyses (Tyler et al., 2006). We downloaded the protein sequences of these 30 genes of Phytophthora ramorum from http://www.jgi.doe.gov/Pramorum, and used them as queries to search the National Center for Biotechnology Information (NCBI) non-redundant (nr) protein sequences database (E-value cutoff 1e-7). Additional searches were also performed against over 650 transcriptomes in the Marine Microbial Eukaryote Transcriptome Sequencing Project (MMETSP) (Keeling et al., 2014), the fungal genome database at the Joint Genome Institute¹, and our internal customized database (Supplementary Table 1). Particularly, a total of six red algal genomes of five genera were used to search for P. ramorum gene homologs, including Cyanidioschyzon merolae, Porphyridium purpureum, Chondrus crispus, Galdieria sulphuraria, G. phlegrea, and Pyropia yezoensis. Complete genome sequence data of multiple photosynthetic stramenopiles (including Aureococcus anophagefferens, Ectocarpus siliculosus, Fragilariopsis cylindrus, Nannochloropsis gaditana, Phaeodactylum tricornutum, Saccharina japonica, and Thalassiosira pseudonana) were also searched in our analyses.

In the original Phytophthora genome paper (Tyler et al., 2006), the search of “algal” genes was based on significant matches to sequences from Plantae and cyanobacteria (that is, these sequence matches had the highest bit scores and the lowest E-values outside the stramenopiles). The Phytophthora “algal” genes identified from the BLAST search were shared by other stramenopiles (or chromalveolates), and they had stronger BLAST matches to homologous genes of red algae and/or cyanobacteria than to sequences from archaea, opisthokonts or non-cyanobacterial bacteria. Additionally, a complementary approach based on Smith–Waterman alignment was also used to identify candidates with significantly higher similarities to red algal or green plant homologs than to those from opisthokonts or amoebozoans. Because Cyanidioschyzon merolae, which also happens to have a streamlined genome, was the only red alga whose complete nuclear genome sequence was then available, the matches to green plant homologs were included and interpreted as resulting from the lack of sufficient red algal genome data (Tyler et al., 2006).

In the current study, BLAST search was performed against NCBI nr, MMETSP and our internal customized databases for each of the 30 most likely “algal” genes, followed by re-analyses of its phylogenetic distribution and gene structure. Following the criteria used by the Phytophthora genome paper (Tyler et al., 2006), we compared the best BLAST matches (represented by the highest bit scores) between homologs from red algae, cyanobacteria, photosynthetic stramenopiles and those from archaea, opisthokonts, amoebozoans, and non-cyanobacterial bacteria. Because more red algal genomes and transcriptomic data were included in our analyses, the matches to green plant homologs were no longer included and used as proxy for red algal homologs.

For each of the 30 “algal” genes identified in Phytophthora genome sequencing project, we performed further phylogenetic analyses. In order to attain a broad and balanced sampling, we selected protein sequences from representative groups of three domains of life (eukaryotes, bacteria, and archaea). The same sampling strategy was also used to ensure sufficient coverage of representative taxa within each major eukaryotic group. This was done using a Perl script followed by manual inspection and additional sequence sampling if needed. Particular attention was paid to groups under-sampled in the previous analyses, such as chromalveolates and other protists. Multiple alignments of sampled sequences were performed using MUSCLE (Edgar, 2004), followed by careful manual inspection of alignment quality, gene structure, shared insertions/deletions (indels), and conserved amino acid residues. Gaps, ambiguously aligned sites, and sequences whose real identity could not be confirmed were removed from alignments. Phylogenetic analyses were performed with maximum likelihood method using PhyML 3.1 (Guindon et al., 2010) and distance method using neighbor of PHYLIP-3.695 (Felsenstein, 2013). The optimal model of protein substitution and rate heterogeneity were chosen based on the result of ModelGenerator (Keane et al., 2006). Bootstrap analyses were performed using 1,000 replicates.

If the previously identified “algal” genes in Phytophthora are indeed derived from a red algal endosymbiont acquired by the ancestor of stramenopiles, their homologs might also be found in photosynthetic stramenopiles. Given their presumably red algal nature, these stramenopile sequences theoretically should have a closer relationship to homologs from red algae (or red algae and other photosynthetic eukaryotes plus cyanobacteria) than to those from other organisms (e.g., opisthokonts, amoebozoans, non-cyanobacterial bacteria, and archaea). Our BLAST results with a larger taxonomic sampling showed that, for all of the 30 most likely “algal” genes previously identified in Phytophthora, only 10 of them (about 33%) were more similar to sequences of red algae, photosynthetic eukaryotes and/or cyanobacteria (Table 1); these 10 genes were also the viable candidate genes of red algal origin.

Of the 30 “algal” genes in Phytophthora, nine (30%) showed stronger BLAST matches (represented by higher bit scores) to homologs from opisthokonts, amoebozoans, non-cyanobacterial bacteria or archaea than to those from photosynthetic stramenopiles (Table 1). Another gene encoding methylthioadenosine phosphorylase (P. ramorum Gene ID 86425) had no detectable homologs in sequenced photosynthetic stramenopiles. Although the possibility of differential gene losses cannot be ruled out, genes with such a distribution pattern may also suggest an independent origin in oomycetes, such as horizontal gene transfer (HGT) from prokaryotes or other eukaryotes to oomycetes. Moreover, 16 of these 30 genes (about 53%) had more significant matches to homologs from opisthokonts, amoebozoans, non-cyanobacterial bacteria or archaea than to those from red algae and cyanobacteria (Table 1). Particularly, four genes had the strongest BLAST matches in non-cyanobacterial bacteria, and two in opisthokonts or amoebozoans. This observation based on simple pairwise comparisons suggests that many of these “algal” genes in oomycetes have no stronger similarity to photosynthetic stramenopile, red algal or cyanobacterial sequences. If sequence similarity is largely correlated with sequence relatedness, as commonly believed, the nature of these “algal” genes might be seriously questioned.

We further performed phylogenetic analyses on each of these 30 “algal” genes to evaluate its origin. If an oomycete gene is of red algal origin, the gene and its stramenopile (or chromalveolate) homologs are expected to form a clade sister to red algal and/or cyanobacterial sequences. This, however, was not the pattern uncovered in our study. Tree topologies for 21 (70%) genes were poorly supported overall (or the position of oomycete sequences couldn’t be confidently determined), thus providing no sufficient evidence for any evolutionary scenarios (Supplementary Materials). These poorly supported tree topologies might be caused by multiple issues, for example, insufficient phylogenetic signal or heterogeneity in evolutionary rates. Nevertheless, such topologies, combined with the information of phylogenetic distribution from BLAST search, should not be interpreted as evidence for a red algal origin of involved Phytophthora genes. The remaining genes had relatively well-resolved phylogenies and will be detailed in the following sections.

In our analyses, several of these “algal” genes indeed showed a close affinity with algal or cyanobacterial sequences. In addition, for 12 “algal” genes previously identified in Phytophthora, the protein products of their plant and/or algal homologs are localized in plastids (Tyler et al., 2006), as predicted by TargetP (Emanuelsson et al., 2000) (Table 1). It is well known that proteins of organelles-derived genes are often re-imported into the original organelles (mitochondria or plastids) to participate in related biochemical activities (Bogorad, 1975; Ellis, 1981; Weeden, 1981; Timmis et al., 2004). This information has been frequently used as supplemental evidence for genes of organellar origin. However, such affinity with algal/cyanobacterial sequences or functionality in other plastids might not necessarily support the suggestion of a historical red algal endosymbiont in the ancestral stramenopile.

The most likely “algal” gene uncovered in our current study encodes cobalamin-independent methionine synthase (MetE). Our phylogenetic analyses indicated that MetE sequences from oomycetes, photosynthetic stramenopiles, chlorarachniophytes, chromerids and cryptophytes formed a large group with homologs of red algae, green algae, and cyanobacteria (Figure 1). Within this group, oomycete MetE sequences formed a strongly supported clade with red algal instead of other stramenopile homologs. Although the overall molecular phylogeny of MetE is consistent with an algal origin of oomycetes and, to a certain extent, an algal/cyanobacterial origin of all stramenopiles, the strength of this evidence is somewhat compromised by the fact that oomycete and other stramenopile sequences didn’t form a monophyletic group (see Discussion). Two other similar cases are related to the genes encoding prolyl oligopeptidase II (Supplementary Figure 1) and cAMP-binding mitochondrial solute carrier (Figure 2). For both genes, their molecular phylogenies showed that oomycete and red algal sequences were closely related. Particularly in the latter case, a NLPC_P60 and two CAP_ED domains are uniquely shared by oomycetes and red algae, but are absent from other stramenopiles (Figure 2). A parsimonious explanation for these findings would be that oomycetes obtained this gene from red algae directly or vice versa.

Two genes in our analyses were found to be specifically related to green plant sequences, which is in disagreement with the suggestion of a red algal plastid in the ancestral stramenopile. The gene encoding NCAIR mutase does not have detectable homologs in red algae. Phylogenetic analyses of NCAIR mutase supported a monophyletic group including sequences from oomycetes, photosynthetic stramenopiles, dinoflagellates, green algae and cyanobacteria (Figure 3). Because of the lack of detectable NCAIR mutase homologs in red algae, a red algal origin of this gene in all stramenopiles cannot be concluded. On the other hand, an independent green algal endosymbiont in stramenopiles might potentially explain such a distribution pattern (Moustafa et al., 2009; Dorrell and Smith, 2011). The other green plants-related gene in oomycetes encodes a probable folate-biopterin transporter (Supplementary Figure 22). Our analyses showed that sequences from oomycetes, diatom Thalassionema frauenfeldii and land plants formed a strongly supported clade, whereas other photosynthetic sequences, including red algae and cyanobacteria, formed another large group with only modest support.

In addition to primary algae and cyanobacteria, several groups of eukaryotes that have secondary plastids through higher-order endosymbioses might also be potential donors for genes in oomycetes. For instance, phylogenetic analyses of glucokinase indicated that sequences from oomycetes, haptophytes and ciliates formed a well-supported clade, which in turn grouped with homologs from photosynthetic stramenopiles, red algae, green algae, choanoflagellates and dinoflagellates (Figure 4). A similar case was also observed for the gene encoding ketol-acid reductoisomerase (Figure 5). As it is known that ciliates contain sequences of algal origin (Reyes-Prieto et al., 2008), this topology might suggest HGT from haptophytes-related groups to oomycetes, and again provides no support for a common photosynthetic origin between oomycetes and other stramenopiles.

As indicated above, a large fraction of “algal” genes in oomycetes/stramenopiles showed stronger matches in our BLAST search to homologs from opisthokonts, amoebozoans or non-cyanobacterial bacteria rather than those from red algae and cyanobacteria. For several of these genes, this relationship was also confirmed by subsequent phylogenetic analyses.

One of these genes encodes 2-isopropylmalate synthase in leucine biosynthesis and was previously detailed in the Phytophthora genome paper (Tyler et al., 2006). According to the authors, this gene was subject to at least two transfer events in eukaryotes: sequences of primary photosynthetic eukaryotes and stramenopiles (including oomycetes) were derived from cyanobacteria, whereas sequences of fungi were from α-proteobacteria. Specifically, diatom sequences were found to group with green plant rather than red algal homologs, which was interpreted as a separate ancestry or artifacts due to incomplete sampling (Tyler et al., 2006). Our current analyses support the previous conclusion that this gene in stramenopiles might have different origins, but also suggest a potentially more complicated evolutionary scenario. While sequences from brown algae and cryptophytes indeed grouped with red algal homologs, those from diatoms and Aureococcus with green plant sequences instead (Figure 6). The relationships between brown algae, cryptophytes and red algae uncovered here is in line with the suggestion of serial endosymbioses by Stiller et al. (2014), where a red alga was first adopted by a cryptophyte that was in turn engulfed by ochrophytes. The sequence affiliation between diatoms, Aureococcus and green algae might point to separate origins of this gene in other photosynthetic stramenopiles [e.g., from a potential green algal endosymbiont (Moustafa et al., 2009; Dorrell and Smith, 2011) or an independent HGT event]. Nevertheless, unlike previously reported in the Phytophthora genome paper (Tyler et al., 2006), oomycete sequences grouped with labyrinthulomycetes, another group of heterotrophic stramenopiles, and other eukaryotes, rather than being affiliated with diatoms, primary photosynthetic eukaryotes and cyanobacteria (Figure 6).

The gene encoding 3′-phosphoadenosine 5′-phosphosulfate reductase (PAPR), an enzyme in the sulfate assimilation pathway, is another example highlighting the potential pitfalls of insufficient sampling. PAPR and adenosine 5′-phosphosulfate reductase (APR) are homologous proteins and have a complex evolutionary history in eukaryotes (Kopriva et al., 2002; Kopriva and Koprivova, 2004; Patron et al., 2008). The APR gene was previously thought to exist in land plants, algae, and phototrophic bacteria. PAPR, on the other hand, was initially identified mainly in fungi and bacteria (Kopriva et al., 2002). Several more recent studies reported PAPR sparely in phototrophic eukaryotes, suggesting potential HGT events (Kopriva and Koprivova, 2004; Kopriva et al., 2007; Patron et al., 2008). Particularly, the study of Patron et al. (2008) indicated a potential bacterial origin of PAPR in P. sojae. With a much larger taxonomic sampling, our analyses showed that sequences from some stramenopiles (including oomycetes), bacteria (both cyanobacteria and non-cyanobacteria) and Paulinella chromatophora formed a major PAPR clade (Figure 7). The cyanobacterial origin of PAPR in P. chromatophora is somewhat expected, as this species contains an independently evolved plastid organelle (cyanobacterial endosymbiont) (Marin et al., 2005). As the overall topology of this clade is poorly supported, whether PAPR in stramenopiles was derived from a red algal endosymbiont or a separate HGT event could not be answered by our study.