Origin of eukaryotic plasmalogen biosynthesis by horizontal gene transfer from myxobacteria

Significance

Plasmalogens are important membrane lipids in animals, yet their evolutionary origin has remained a puzzle due to their patchy distribution across the eukaryotic tree. We found that the core set of genes required for aerobic plasmalogen production was acquired by an early eukaryotic ancestor in a single event of horizontal gene transfer (HGT) (non-parent-progeny transfer of genetic material) from myxobacteria. Our results thus pinpoint a bacterial origin for a relevant eukaryotic membrane lipid biosynthetic pathway, explain its unusual evolutionary history, and highlight the importance of HGT in shaping early eukaryotic metabolic capabilities.

Abstract

Plasmalogens are a unique class of glycerophospholipids defined by a distinctive vinyl ether bond. While these lipids are abundant in animals and important for human health, their evolutionary history remains enigmatic, mostly due to their absence in some major eukaryotic lineages like plants. Here, we resolve the origin and evolution of the aerobic plasmalogen biosynthesis pathway in eukaryotes. Through comprehensive phylogenomic analysis and experimental validation of enzyme activity and plasmalogen presence, we demonstrate that the essential desaturase plasmanylethanolamine desaturase 1 (PEDS1)—and likely the fatty acyl-CoA reductase (FAR) and glycerone phosphate O-acyltransferase (GNPAT) enzymes also critical in the pathway—were acquired by an early eukaryotic ancestor through horizontal gene transfer (HGT) from myxobacteria. Our data show that this bacterial pathway was retained in the Amorphea and Discoba supergroups but lost or replaced in others. The findings yield insights into how HGT shaped metabolic pathways in early eukaryotes.

Lipids are key components of cell membranes that determine their rigidity, fluidity, and function (1). Among these, plasmalogens are a unique class of glycerophospholipids (GPs) defined by a vinyl ether bond at the sn-1 position, distinct from the ester bond typically found in phospholipids (Fig. 1A) (1–3). This chemical feature endows plasmalogens with unique physical-chemical features that affect cellular membrane properties and likely underlie their roles in diverse biological functions (4–7). Plasmalogens can represent up to a fifth of the human phospholipidome and their altered levels are linked to serious conditions ranging from neurodegenerative diseases to cancer (4–10). Discovery of the enzyme responsible for the vinyl ether bond formation (Fig. 1B) (11–13) has enabled the identification of specific plasmalogen-mediated functions, which include a bacterial response to light (11), myeloid cell biology (10), and neurodevelopment (14). However, and despite their importance in animal biology, the evolutionary history of plasmalogens remains enigmatic.

Fig. 1.

Plasmalogens and their aerobic biosynthesis pathway in mammals and in Myxococcus xanthus. (A) Chemical structures of plasmalogens (plasmenyl or alkenyl-acyl GPs), plasmanyl (or alkyl-acyl) GPs and diacyl GPs. The vinyl ether bond in plasmenyl GPs is shaded green, and the ether bond in plasmanyl GPs is shaded gray (the carbon–carbon bond that becomes desaturated by PEDS1 is marked in red). In diacyl GPs, both hydrocarbon chains are linked via ester bonds. The sn positions are indicated, with the hydrocarbon chain at sn-1 represented as R₁ and that at sn-2 as R₂. X indicates the head group. (B, Left) Biosynthesis of plasmalogens in mammals. Signature enzymes of the pathway (FAR, GNPAT, AGPS, and PEDS1) are indicated, with their Enzyme Commission numbers in brackets. For simplicity, steps in the biosynthesis pathway that are common for alkyl-acyl and diacyl GPs are omitted (refs. 4–6). G3P: glycerol 3-phosphate. (Right) The two M. xanthus operons, elbABCDE (elb for ether lipid biosynthesis; locus tags MXAN_1531 for elbA to MXAN_1527 for elbE) and MXAN_1676-MXAN_1675-MXAN_1674 (genes to be named), with roles in plasmalogen biosynthesis assigned based on sequence analysis; genes are depicted as thick arrows and ovals inside them indicate that they encode for putative AGPS, FAR, AGPAT/GNPAT (AT), acyl-CoA synthase (ACS), acyl-carrier protein (ACP), NAD-dependent epimerase/dehydratase family protein (NAD), haloacid dehalogenase-like phosphohydrolase (HAD) functions. Gene deletions of elbD or MXAN_1676 have been shown to block alkyl and alkenyl ether lipid production via the corresponding biosynthetic pathway in M. xanthus (6, 11, 15).

Plasmalogens show a patchy distribution across the tree of life. They have been shown to be absent in archaea, plants, and fungi, yet present in animals, some protists, and certain bacteria (largely restricted among obligately aerobic bacteria to the phylum Myxococcota, and to some anaerobic and facultatively anaerobic bacteria) (5, 6, 11, 16, 17). Most strikingly, the biosynthesis of these lipids appears to have evolved twice (5, 18, 19). Anaerobic bacteria produce plasmalogens via a reductive pathway suited for an anoxic world, which involves conversion of an ester to a vinyl ether bond (5, 17, 18). In contrast, animals rely on a completely different, oxygen-dependent pathway to generate an alkyl ether bond and convert it into a vinyl ether one, involving four key enzymes (Fig. 1B): fatty acyl-CoA reductase (FAR), glycerone phosphate O-acyltransferase (GNPAT), alkylglycerone phosphate synthase (AGPS), and plasmanylethanolamine desaturase (PEDS1), the enzyme mentioned above that creates the vinyl ether bond (4–6, 11–13, 16).

The origin of this aerobic pathway in eukaryotes has been obscure. Indeed, the identity of PEDS1 was only recently unmasked, when it was demonstrated that M. xanthus CarF and its animal homologs, as well as those from the bacterial phylum Spirochaetota, all correspond to PEDS1; whereas proposed candidates in plants or Alphaproteobacteria were not functionally equivalent to PEDS1 (11). Moreover, and in addition to PEDS1/CarF, M. xanthus has potential FAR, GNPAT, and AGPS activities (6, 11, 15), mirroring the mammalian pathway (Fig. 1B). This resemblance raises a fundamental question: did eukaryotes evolve this pathway independently, or was it acquired from bacteria?

Besides the report mentioned above (11), few other studies have tried to shed light into the origin and evolution of the aerobic plasmalogen biosynthesis pathway in eukaryotes. A recent study (20) identified PEDS1 homologs in some members of the Holozoa, a clade that includes animals and their closest unicellular relatives (21, 22), in some genera of Amoebozoa and Excavata, and even in a few fungi. However, taxon sampling in these studies was limited, largely constraining the potential interpretations. Here, we investigate the origin of eukaryotic plasmalogen biosynthesis by mapping the distribution of the four key enzymes (FAR, GNPAT, AGPS, and PEDS1) across the tree of life. We combined comprehensive bioinformatic and phylogenetic analyses, using the broadest possible taxon sampling, with experimental validation of PEDS1 activity in diverse taxa. We identified a stark divide in biosynthetic capacity: the complete aerobic pathway is conserved in the Amorphea and Discoba supergroups but absent in others. Our phylogenetic reconstruction indicates that the genes for this pathway were acquired by an early eukaryotic ancestor via horizontal gene transfer (HGT) from myxobacteria. These findings reveal a bacterial origin for a hallmark animal biosynthesis pathway and highlight how HGT has shaped the metabolic versatility of eukaryotes.

Results

Distribution of Plasmalogen Biosynthesis Pathway in Eukaryotes.

To unravel the evolution of the eukaryotic plasmalogen biosynthesis pathway, we first need to map its distribution across the tree of eukaryotes. For this, we performed an extensive search for the four key enzymes (FAR, GNPAT, AGPS, and PEDS1) in a dataset of 376 high-quality genomes encompassing the whole breadth of the known eukaryotic diversity [(23); SI Appendix]. Detection of FAR, GNPAT, and AGPS would suggest an ability to produce plasmalogen precursors, and the additional presence of PEDS1 would suggest plasmalogen synthesis. For improved detection of distant homologs, we followed an iterative search algorithm that accounts for genome variations over evolutionary time and assigns higher rates to conserved positions when evaluating the hits. This allowed identification of functionally conserved regions even in proteins of low overall sequence similarity.

Discoba.

The plasmalogen biosynthesis pathway exhibited significant variation among the five Discoba groups analyzed (Fig. 2A and Datasets S1–S5). We identified PEDS1 in Heterolobosea and Kinetoplastea, whereas Jakobida, Euglenida, and Diplonemea lacked this desaturase. GNPAT hits in Heterolobosea, Kinetoplastea, and Euglenida were about twice the size of mammalian homologs due to a large ~650-residue N-terminal region not assigned to any known domain. Such an extension had been previously reported in the Trypanosomatidae family but with no precise function determined (24, 25). Since M. xanthus has a predicted GNPAT fused to an N-terminal FAR module (MXAN_1675 in Fig. 1B), and Tetrahymena thermophila (Ciliophora) GNPAT was experimentally shown to occur as a C-terminal fusion to FAR in a single bifunctional 1,140-residue FARAT protein (26), we explored the possibility that the large GNPAT hits were really FARAT. Because sequence-based searches failed to identify FAR domains within these extensions, we used protein structure comparison, which has proven to be a sensitive alternative for protein domain identification and homology detection in Discoba (27, 28), effectively unmasking as FARAT all the GNPAT identified with the N-terminal extension (except for that in Angomonas deanei) (SI Appendix, Fig. S1 and Dataset S3). Another copy of FAR, but as an independent protein, was found in Euglena gracilis and in the Trypanosomatidae Strigomonas culicis and Vickermania ingenoplastis (Dataset S4). Finally, we observed AGPS in all groups except Jakobida, always co-occurring with the other enzymes except in Diplonema papillatum (the only Diplonemea analyzed), where we identified only AGPS (Dataset S5).

Fig. 2.

Distribution of plasmalogen pathway genes across major Eukaryotic clades. (A) A schematic phylogenetic tree illustrating the presence of key enzymes in the plasmalogen biosynthesis pathway (PEDS1, AGPS, FAR, and GNPAT) across diverse eukaryotic lineages. The tree is color-coded by supergroup, with numbers in each triangular branch indicating how many genomes were analyzed and triangles without numbers representing clades with a single genome. Open or filled circles indicate, respectively, that more than one-third or two-thirds of species in the clade have the indicated enzyme. FAR and GNPAT connected by a line indicates a bifunctional FAR-GNPAT fused FARAT protein, and the FAR protein/domain with an asterisk indicates FAR presence also as an independent protein. In the “Plasmalogen” column: filled hexagons or with a circle indicate, respectively, that plasmalogens were detected in this study or in other studies; outline-only hexagons indicate that plasmalogens were not detected in lipid analyses. In the “Validated PEDS1” column: hexagons mark cases where PEDS1 function was confirmed by heterologous complementation in M. xanthus (filled hexagon, validated in this study; with a circle, in other studies). The final column lists species whose PEDS1 was tested in this study: Trypanosoma cruzi (Tc), Leishmania major (Lm), Stygamoeba regulata (Sr), Dictyostelium discoideum (Dd), Chromosphaera perkinsii (Cp), Capsaspora owczarzaki (Co), and Pigoraptor chileana (Pc). The asterisk next to Sr indicates that PEDS1 was identified through transcriptomic data. The tree was obtained from EukProt (23). Phylogenetic positions inferred from (22, 29, 30). (B) The small subset of fungal species where PEDS1 was found in BLASTP analyses and the presence in these species of the other enzymes in the plasmalogen biosynthesis pathway. Symbols as in (A).

Genomes of both Heterolobosea species analyzed, Naegleria gruberi and Neovahlkampfia damariscottae, encode AGPS and FARAT, but only N. gruberi additionally harbors PEDS1. Among the Kinetoplastea, the pathway distribution was a mosaic, with ten out of fifteen taxa containing PEDS1, seven having all four enzymes and two (Trypanosoma vivax and Perkinsela spp.) lacking them entirely. That homologs to the four enzymes occur in several species is consistent with previous experimental reports detecting plasmalogens in various kinetoplastid taxa, including Trypanosoma (31–33), Strigomonas (34), Crithidia (33), and Leishmania (33, 35). Notably, species such as Bodo saltans and Phytomonas sp. Hart1 lack PEDS1 but have the other three enzymes, suggesting a secondary loss of PEDS1 and ability to synthesize only plasmanyl GPs.

The Euglenida class, represented by E. gracilis, is unique among Discoba in having FAD4 (Fatty Acid Desaturase 4) instead of PEDS1. The functionally divergent FAD4 emerges in PEDS1 queries due to sequence similarity, and was first described in Arabidopsis thaliana as a chloroplast membrane‐associated enzyme (36, 37), in contrast to PEDS1 localization to the endoplasmic reticulum (11). Unlike PEDS1, which generates a vinyl double bond at sn-1 (Fig. 1A), FAD4 produces an unusual trans double bond between carbon 3 and 4 of an sn-2 palmitoyl group in phosphatidylglycerol to generate a GP typically found in the thylakoid membranes throughout the plant kingdom, from green algae to land plants (37). Lack of PEDS1 in E. gracilis is consistent with its reported lipid analyses detecting plasmanyl GPs but no plasmalogens (33, 34, 38).

Amorphea.

The Amorphea supergroup also exhibited a heterogeneous distribution of the plasmalogen biosynthesis pathway (Fig. 2A and Datasets S1–S5), indicative of a complex evolutionary history shaped by differential gene retention. Within its Amoebozoa clade, the sole representative analyzed in phylum Discosea, Acanthamoeba castellanii, revealed not only AGPS and FARAT (as well as FAR) but also PEDS1, consistent with plasmalogens reported in this species (33). In the phylum Evosea, the complete pathway was ubiquitously detected in the genomes of the Eumycetozoa (slime molds), except for Synstelium polycarpum, which lacks PEDS1. By contrast, no evidence of the plasmalogen pathway was found in the sister lineage to Eumycetozoa, Archamoebae, suggesting that the entire pathway was secondarily lost in this group. As in Discoba, FAR and GNPAT occur recurrently in Amoebozoa as FARAT, whose detection was straightforward since PSI-BLAST and InterProScan reliably identified the FAR domain. FARAT among Amoebozoa was previously described only in Dictyostelium discoideum that, consistent with having a complete functional pathway, produces plasmalogens (39). Only AGPS was detected in the Apusomonad Thecamonas trahens.

Among Holomycota, all four enzymes were absent in Fungi, but we found FAR in their closest relative Fonticula alba (Rotosphaerida). Fungi are of particular interest, since our analysis of 53 representative fungal genomes indicated complete lack of the pathway (with two exceptions having AGPS; Dataset S5), consistent with the established notion that Fungi do not produce plasmalogens (5, 19). However, a recent report (20) suggested the presence of PEDS1 in three fungal species: Thozetella sp., Trichoderma semiorbis, and Mollisia scopiformis, based on BLASTP analyses against the NCBI database. We therefore further investigated PEDS1 occurrence in fungal taxa and analyzed the gene structure and synteny of fungal PEDS1 homologs that we retrieved with BLASTP against the NCBI nr database (Fig. 2B, SI Appendix, Fig. S2A, and Dataset S6). Apart from those mentioned, we found PEDS1 homologs in the genomes of fungi Oculimacula yallundae and eight species of genus Trichoderma, all listed as monoisolates in NCBI; and detected AGPS, FAR, or GNPAT generally encoded by neighboring genes in PEDS1-containing Thozetella sp. and Trichoderma species, but none of the three in O. yallundae and M. scopiformis. Interestingly, FAR and GNPAT were found fused as FARAT in these fungal species, with two FARAT copies in Thozetella sp. Thus, the plasmalogen biosynthesis enzymes occur in a restricted number of Fungi.

Among Holozoa, which includes Metazoa and their closest unicellular relatives, we report the presence of the full plasmalogen pathway in the filasterean Pigoraptor chileana and confirm it in Capsaspora owczarzaki (20). Both species have FARAT and an independent FAR copy. In six Ichthyosporean species analyzed, the four enzymes were detected only in Chromosphaera perkinsii, with separate GNPAT and FAR proteins. No evidence of the pathway was found in the remaining Ichthyosporea (except for AGPS in Creolimax fragrantissima), in Pluriformea (or Corallochytrea), or in Choanoflagellatea, the closest unicellular relatives of animals. In Metazoa, where plasmalogens have been widely reported (10–12, 33, 40–42), the complete pathway was found in ~70% of the 58 genomes analyzed, including all early-branching metazoan lineages. Interestingly, PEDS1 was absent in Metazoa with a parasitic lifestyle (~10% of analyzed metazoan genomes). Notably, among Metazoa, GNPAT hits are typically 600 to 750 residues long (Dataset S2), and we did not identify any FARAT. The 1,248-residue GNPAT in the jellyfish Sanderia malayensis had an N-terminal extension but with no detectable FAR domain based on BLASTP and InterProScan analyses, or structural predictions. Independent GNPAT and FAR genes are thus the norm in Metazoa, contrasting with the two occurring predominantly as FARAT in other Amorphea groups.

Diaphoretickes.

PEDS1 was undetected among Diaphoretickes and the other enzymes rarely coexisted in a species or group (Fig. 2A and Datasets S1–S5). In one notable exception, phylum Ciliophora within the SAR group, we found AGPS co-occurring with FAR and GNPAT mostly fused as FARAT, in six of the seven species analyzed (SI Appendix, Fig. S3 and Datasets S2 and S3). All proteins identified initially as GNPAT were really FARAT (except one in Paramecium tetraurelia and another in Pseudocohnilembus persalinus). PSI-BLAST identified at least one FARAT as a GNPAT or FAR hit in P. tetraurelia, Stylonychia lemnae, and T. thermophila (with two, three, and one FARAT, respectively). Overall, these results imply that Ciliophora can produce plasmanyl GPs but not plasmalogens, in accord with lipid data reported for T. thermophila (26).

In several Diaphoretickes groups, absence of PEDS1 contrasts with presence of FAD4, the desaturase unrelated to plasmalogen biosynthesis retrieved in the search for PEDS1 homologs, as noted earlier (11, 37). FAD4 was detected in a) three (Diacronema lutheri, Emiliania huxleyi, and Phaeocystis antarctica) of the nine Haptophyta species analyzed, all of which lacked PEDS1, AGPS, GNPAT, and FAR; b) Cryptomonadales, including Guillardia theta and Hanusia phi, with a sporadic AGPS, GNPAT, and FAR distribution (Fig. 2A and Datasets S1, S2, S4, and S5). By comparison, FAD4 was prevalent throughout Archaeplastida, encompassing red and green algae together with plants (Fig. 2A and Dataset S1). In Rhodophyta, eight of the ten species analyzed contained FAD4, three also with AGPS; and both enzymes co-occurred in Glaucophyta, represented by Cyanophora paradoxa. Within Chloroplastida, we detected FAD4 in ~80% of the 67 species from Chlorophyta, Streptophyta, and Palmophyllophyceae. In most of these, AGPS and/or FAR were absent, except for six species with both AGPS and FAR co-occurring with FAD4 (Fig. 2A and Datasets S1, S2, S4, and S5). Finally, within the SAR clade, FAD4 was restricted to Ochrophyta (~64% of 45 species analyzed) and Colpodellida (both representative species had FAD4) (Fig. 2A and Dataset S1). By contrast, AGPS, GNPAT, and FAR displayed a patchy distribution across SAR except in Ciliophora, where neither FAD4 nor PEDS1 (see above) were present (Fig. 2A and Datasets S1, S2, S4, and S5).

Experimental Analysis of Selected PEDS1 Homologs.

Since PEDS1 shows a patchy distribution in the tree of eukaryotes (Fig. 2), we wondered whether the bioinformatically retrieved hits function as genuine PEDS1. Therefore, we selected representative taxa to verify enzymatic activity and capacity for plasmalogen production. Because PEDS1 homologs from various Metazoa have been experimentally tested in previous studies (10–12, 33, 40–42), we focused on other taxonomic groups for a broader perspective of PEDS1 functionality across key eukaryotic lineages. These included the following (SI Appendix, Fig. S4A and Dataset S7): three close unicellular relatives of animals [the filastereans C. owczarzaki and P. chileana, as well as the ichthyosporean C. perkinsii (43)]; two amoebozoans (the Eumycetozoa D. discoideum and the Discosea Stygamoeba regulata); and two representatives from Kinetoplastea (T. cruzi and L. major).

To test for PEDS1 function, we performed heterologous complementation analysis in M. xanthus, where plasmalogens are specifically required to induce a photoprotective carotenogenic response that causes cells to turn from yellow in the dark to orange in the light (6, 11, 44). The M. xanthus plasmalogen, MxVEPE, bears branched hydrocarbon chains (iso15:0 or i15:0) at sn-1 and sn-2 (11). Nonetheless, functional eukaryotic PEDS1 homologs generate MxVEPE when expressed in the carF-deleted (ΔcarF) M. xanthus strain, even though they produce different plasmalogens, usually with straight chains, in their natural organism (11). Complementation can thus be monitored by the light-induced colony color change, and by lipid analysis to detect MxVEPE as the iso15:0 dimethylacetal (i15:0-DMA) derivative via fatty acid methyl ester gas chromatography-mass spectrometry (FAME GC-MS) (11). M. xanthus strains expressing each selected PEDS1 homolog with an N-terminal FLAG-tag (for immunoblot analysis) in the ΔcarF genetic background produced MxVEPE, albeit in different amounts, likely due to the observed differences in protein levels (Fig. 3A) and to possible intrinsic variations in activity and substrate preferences. Light-induced orange color intensity correlated with MxVEPE levels produced by each homolog, the color in strains expressing C. perkinsii, P. chileana, or T. cruzi homologs closely matching that for M. xanthus CarF (Fig. 3A).

Fig. 3.

Functional validation of selected PEDS1 homologs. (A) Complementation analyses using the M. xanthus ΔcarF strain expressing M. xanthus CarF (Mx) or homologs from C. owczarzaki (Co), C. perkinsii (Cp), P. chileana (Pc), D. discoideum (Dd), S. regulata (Sr), T. cruzi (Tc), and L. major (Lm) (all FLAG-tagged for immunoblot analysis). (Top) Abundance of i15:0-DMA relative to 100% for the strain expressing M. xanthus CarF. Values are mean ± SD (n = 3 biological replicates). Middle: Light-induced colony-color assay for the corresponding strains above. (Bottom) Immunoblot of the corresponding cell extracts probed with monoclonal anti-FLAG antibodies and, as loading control, monoclonal antibodies against the RNA polymerase (RNAP) β subunit. (B) Representative FAME GC-MS chromatogram (relevant section shown) of total lipid extracts (Top) from C. perkinsii and C. owczarzaki. Electron impact mass spectrum and diagnostic m/z peaks (below) in FAME GC-MS analysis for the characteristic 16:0-DMA and 17:0-DMA FAME products (peaks in red) are also shown. Under the growth conditions used, minor quantities of 18:0-DMA (0.1%) were also detected in C. perkinsii. Plasmalogen abundance (as DMAs) is calculated as a percentage of the total lipid content in the FAME GC-MS analysis. (C) Complementation analyses as in (A), with the PEDS1 homologs from T. harzianum (Th), T. aggressivum (Tag), Thozetella sp. (Thz), and M. scopiformis (Ms).

Consistent with our validation of PEDS1 activity for the homologs tested, plasmalogens have been reported in T. cruzi, L. major, and D. discoideum, with C16 or C18 sn-1 vinyl ether-linked chains (31–33, 35, 39, 45, 46), as is commonly observed in Metazoa (10, 11, 33, 40–42). Our FAME GC-MS analyses of C. perkinsii, the only Ichthyosporea with PEDS1 identified to date, detected plasmalogens primarily with 16:0 at sn-1 (as 16:0-DMA; Fig. 3B), a chain length common among eukaryotes. By contrast, plasmalogens with 17:0 at sn-1 were observed in C. owczarzaki (Fig. 3B), as also reported in a recent study (20). Differences in chain lengths among species are likely related to substrate availability and species-specific requirements of FAR, GNPAT, and/or AGPS (26, 47, 48). Although complete lipidomics data are not available for P. chileana or S. regulata, the presence of a functional PEDS1 together with AGPS and FARAT (or separate GNPAT and FAR proteins) (Fig. 2A) suggests a capacity for plasmalogen biosynthesis in the two species.

In view of the unexpected identification of PEDS1 homologs in some Fungi (Fig. 2B, SI Appendix, Fig. S2A, and Dataset S6), a group traditionally considered to lack plasmalogens, we also performed complementation analyses of the M. xanthus ΔcarF strain with N-terminal FLAG-tagged fungal PEDS1 hits (SI Appendix, Fig. S4B and Dataset S7) from: Trichoderma harzianum (strain CBS 354.33), Trichoderma aggressivum (CBS 100526), and Thozetella sp. (PMI_491), where we identified PEDS1, AGPS, and FARAT; and M. scopiformis (CBS 120377), where we identified only PEDS1. MxVEPE production was observed with the Thozetella and M. scopiformis homologs, the only ones detected in immunoblots, though less than with M. xanthus CarF (Fig. 3C). The other two fungal homologs tested were expressed poorly in M. xanthus (despite codon usage optimized for M. xanthus), likely explaining the negligible MxVEPE produced with T. harzianum and T. aggressivum homologs (Fig. 3C). Nonetheless, these results confirm PEDS1 activity for at least two fungal homologs, when expressed in M. xanthus. For further experimental evidence, we examined the presence of plasmalogens by FAME GC-MS analysis in two of the above fungal species with PEDS1, AGPS, and FARAT homologs whose specified strains were available from culture collections, T. harzianum (CBS 354.33) and T. aggressivum (CBS 100526), and whose PEDS1 were analyzed in M. xanthus (Fig. 3C). Thozetella sp.PMI_491, whose PEDS1 functioned best in M. xanthus, was unavailable in culture collections. Both species examined had several of the characteristic FAMEs described in various Trichoderma and other fungi (49), such as those derived from the more abundant C16 and C18 fatty acids (SI Appendix, Fig. S2B). However, neither plasmalogens nor their precursor ether lipids were detected. The presence of key biosynthetic enzymes, yet no plasmenyl or plasmanyl GPs, may stem from poor enzyme expression and/or activities, which can be addressed in future studies.

Evolutionary Origin of PEDS1.

Given the evident similarities between plasmalogen biosynthesis in eukaryotes and in bacteria from phylum Myxococcota, we studied the origin of this metabolic pathway in eukaryotes. The Genome Taxonomy Database (GTDB) (50), which provides a phylogenetically consistent and rank-normalized genome-based taxonomy for prokaryotes, was searched for PEDS1, AGPS, GNPAT, and FAR (or FARAT) using as queries the complete set of eukaryotic sequences identified in the previous section. Although individual genes were detected across several bacterial phyla, only six phyla contained at least one genome with the four enzymes. Overall, we identified 73 bacterial genomes encoding the full pathway: 61 in phylum Myxococcota, and the remaining 12 distributed across five additional phyla (SI Appendix, Fig. S5 and Datasets S1, S2, S4, and S5). The capability to synthesize plasmalogens and their precursors aerobically in bacteria is thus currently concentrated mostly within Myxococcota.

To explore the evolutionary origins of the oxygen-dependent desaturase PEDS1, we performed phylogenetic analysis including all sequences retrieved from both prokaryotic and eukaryotic datasets. For eukaryotes, we incorporated all PEDS1 hits (Fig. 2). Phylogenetic reconstruction recovered the topology of the desaturase gene distribution, revealing two distant monophyletic groups, one comprising the PEDS1-like homologs in Amorphea and Discoba and the other FAD4-like homologs in Diaphoretickes (Fig. 4).

Fig. 4.

Phylogenetic tree for PEDS1 homologs in Bacteria and Eukaryota. Maximum likelihood phylogenetic tree inferred using PEDS1 homologous protein sequences (137 amino acid positions) obtained from the GTDB and EukProt databases plus Fungi sequences retrieved by BLASTP NCBI. Colors correspond to relevant taxonomic classifications. For prokaryotes, the colors represent phylum Myxococcota. For eukaryotes, the represented clades are Amorphea (Holozoa; and Amoebozoa, as those sequences within Amorphea that do not correspond to Holozoa), Diaphoretickes (Chloroplastida, SAR) and Discoba. Holozoa within Amorphea and Chloroplastida or SAR within Diaphoretickes are highlighted. Black dots indicate ultrafast bootstrap (UFB) values greater than 95%. For visualization purposes, a fully bacterial clade (pink) comprising 229 sequences was collapsed.

Within the portion of the tree encompassing Amorphea and Discoba, the topology is consistent with known species relationships. First, all eukaryotes from these two clades form a monophyletic group (Fig. 4), which aligns with the presence/absence patterns described earlier (Fig. 2A). These are, in fact, the only major eukaryotic clades having PEDS1. Within this monophyletic group, Holozoa and Amoebozoa (Fig. 4 and SI Appendix, Fig. S6 for details) each form distinct, closely related clusters, while a more basal lineage comprises Discoba. Two exceptions are N. gruberi, which instead of clustering with Discoba appears on an independent branch between Amoebozoa and Holozoa, and E. gracilis, placed with the FAD4-like Diaphoretickes group (Dataset S8, for details). Notably, the entire Amorphea and Discoba monophyletic assemblage of eukaryotes is closely related to PEDS1/CarF from the Myxococcota phylum. Taken together, the observations suggest that an early eukaryote may have acquired PEDS1 through an HGT event from Myxococcota. Fungal PEDS1 sequences retrieved from the NCBI formed a monophyletic group separate from the other eukaryotes (Fig. 4) and more closely related to bacterial PEDS1 (SI Appendix, Fig. S6), suggesting an independent HGT event.

The only monophyletic group with FAD4-like proteins is Chloroplastida, with two notable outliers: C. paradoxa (Glaucophyta) and, most strikingly, E. gracilis, which belongs to supergroup Discoba, lacks PEDS1 and is photosynthetic, a trait acquired through a secondary endosymbiosis event involving a green alga (51). This well-documented endosymbiosis is consistent with E. gracilis FAD4 clustering alongside Ostreococcus and Micromonas sequences in the phylogenetic tree (SI Appendix, Fig. S7). The remaining organisms harboring FAD4-like enzymes (Haptophyta, Cryptomonadales, Rhodophyta, Ochrophyta, and Colpodellida) do not form clear monophyletic clusters, suggesting a more complex evolutionary history. Whether these phylogenetically related proteins are truly FAD4 remains to be examined.

Evolutionary Origin of the Other Enzymes of the Plasmalogen Biosynthesis Pathway.

We next examined the origin of the remaining enzymes in the pathway. Since the AGPS sequences did not provide sufficient phylogenetic signal, we focused on FAR and GNPAT, which exist as a fused FARAT protein in most eukaryotic groups, as shown above. A comprehensive phylogenetic analysis using all retrieved GNPAT sequences was first performed, with redundancy in the extensive dataset reduced by sequence clustering based on identity and coverage (Dataset S9). Since including the few fungal FARAT sequences diminished ultrafast bootstrap (UFB) values, the analysis was performed with and without them. With fungal sequences excluded, we observed a single monophyletic group encompassing all eukaryotes, including those in Discoba, Amorphea, and Diaphoretickes (represented solely by Ciliophora; Fig. 5 and SI Appendix, Fig. S8) implying, in principle, a single evolutionary origin of FARAT in eukaryotes. Eukaryotic FARAT proteins appeared closely related to sequences from Myxococcia, suggesting FARAT acquisition by HGT from phylum Myxococcota, as proposed for PEDS1. Moreover, the phylogenetic tree supports a single HGT event also for FARAT, likely before the Discoba and Amorphea lineages diverged, given the monophyletic relationship and clade topology. However, the relationships among subgroups deviate from those expected for strict vertical inheritance, suggesting a more intricate evolutionary history. The most basal eukaryotic clade is composed by Discoba, specifically by Kinetoplastea and Euglenida, which display a topology consistent with known species relationships (52, 53). By contrast, the Heterolobosea N. gruberi and N. damariscottae clustered closer to Ciliophora than to Discoba, and both appeared more related to Amoebozoa (SI Appendix, Fig. S8 and Dataset S10). A similar pattern observed for N. gruberi PEDS1 suggests its FARAT and PEDS1 were replaced through a secondary HGT event, possibly from an Amoebozoa. For Ciliophora, lack of PEDS1 (for direct comparison with FARAT) precluded a more definitive assessment of this scenario. Nonetheless, since Ciliophora is the only Diaphoretickes with FARAT (Fig. 2A), high UFB in the tree would support an HGT event among eukaryotes (SI Appendix, Fig. S8), possibly from an amoeba to either Ciliophora or Heterolobosea, which could have then acted as an intermediary for subsequent HGT to its sister group. As noted, while Amoebozoa, Discoba, and SAR possess FARAT, having separate GNPAT and FAR appears to be exclusive to Holozoa (with some exceptions). Additionally, FAR appears to have undergone expansion in Holozoa, forming a gene family, an aspect not explored further in this study.

Fig. 5.

Phylogenetic tree illustrating FARAT homologs in Bacteria and Eukaryota. Maximum likelihood phylogenetic tree generated using GNPAT homologous protein sequences (470 amino acids positions) obtained from the GTDB and EukProt databases. Colors correspond to relevant taxonomic classifications. For prokaryotes, the colors represent phylum Myxococcota. For eukaryotes, the represented clades are Amorphea, Discoba, and Diaphoretickes. Black dots indicate ultrafast bootstrap (UFB) values greater than 95%. For visualization purposes, a fully bacterial clade (pink) comprising 361 sequences was collapsed.

Regarding the fungal sequences, their inclusion in the initial phylogenetic reconstructions significantly lowered UFB values across several key nodes, particularly those supporting the primary HGT event (SI Appendix, Fig. S9). These sequences introduced phylogenetic uncertainty and obscured the evolutionary signal within the broader eukaryotic tree. When included, fungal FARAT sequences form a monophyletic group that branches at an incongruent basal position relative to the rest of the eukaryotic sequences. The phylogenetic incongruence, coupled with the statistical instability they introduce, together with the fact that FARAT was identified only in a small closely related group of fungal species, suggests that its presence in Fungi could be the result of a secondary, independent HGT event, a scenario that mirrors the evolutionary pattern observed for PEDS1.

Discussion

This study represents the most comprehensive taxonomic investigation to-date of genes essential for plasmalogen production across the tree of life. By integrating phylogenomics with structural and functional validation, we resolve the long-standing puzzle of why this lipid class is distributed so unevenly across the tree of life. Our results partition the eukaryotic diversity into two groups: those employing the myxobacteria-derived PEDS1 system (Amorphea and Discoba), which potentially synthesize plasmalogens, and those relying on the unrelated FAD4 desaturase (Diaphoretickes), which potentially cannot synthesize them.

A striking feature of our data is the mutual exclusivity of PEDS1 and FAD4: they never co-occur in the same organism, at least those sampled here. This is best exemplified by E. gracilis (Discoba). While most Discoba clades possess all the defining enzymes of the aerobic pathway, E. gracilis retains AGPS and FARAT, but lacks PEDS1 and instead has a plant-like FAD4. Studies in M. xanthus and various Metazoa firmly established that plasmalogen synthesis absolutely requires PEDS1 in the final indispensable step of vinyl ether bond formation (10–13, 42), clearly explaining why E. gracilis lacks plasmalogens yet produces the plasmanyl GP precursors (33, 34, 38). FAD4 in E. gracilis likely originated from the green algal endosymbiont (51), consistent with our finding that FAD4 is prevalent among the Chloroplastida and with green algae producing the unique thylakoid membrane lipid whose biosynthesis requires FAD4, like land plants (37). That FAD4, but not PEDS1, occurs in E. gracilis supports the hypothesis that having both enzymes is mutually exclusive, perhaps reflecting an incompatibility between their respective lipid products in cellular membranes. If this were the case, a similar replacement process as in E. gracilis may have occurred in Diaphoretickes. The striking absence of PEDS1 in photosynthetic organisms like E. gracilis could be related to the susceptibility of the plasmalogen vinyl ether bond to cleavage by reactive oxygen species like singlet oxygen (¹O₂), which in turn produces other reactive toxic species (11, 54, 55). Photosynthesis could thus exert selective pressure against plasmalogens and PEDS1 because it generates ¹O₂ as a byproduct of chlorophyll photoexcitation and energy transfer to oxygen at much higher levels than do other cellular processes (56). Consistent with PEDS1 absence in photosynthetic organisms, a recent study reported the unexpected discovery of aerobic anoxygenic photosynthesis in some members of phylum Myxococcota (57), and our search of their genomes (BioProject ID PRJNA943119) failed to reveal any PEDS1/CarF homologs.

Beyond PEDS1, our analysis of the associated enzymes (FAR, GNPAT, AGPS) reveals a dynamic evolutionary history. In Metazoa, FAR and GNPAT function as separate enzymes (Fig. 1A). However, we demonstrate that this is the exception, not the rule. In most PEDS1-containing eukaryotes—including Amoebozoa, Discoba, Filasterea, and Ciliophora—these activities are fused into a single bifunctional protein, FARAT (26, 39). Whereas the FAR and GNPAT activities of FARAT localize to the peroxisome (26, 39), metazoan GNPAT occurs and acts in the peroxisome, but the peroxisomal membrane-located FAR acts on the cytosolic side (58). Physical separation of the FAR and GNPAT activities in Metazoa might ensure tighter spatial control, and the dual activity of FARAT in the peroxisome might enable more efficient coordinated regulation. Our analysis suggests that FARAT represents the ancestral eukaryotic state, inherited directly from a bacterial progenitor. The split system observed in animals likely results from a secondary gene fission event that occurred within the Holozoa. Thus, the mammalian pathway organization is a derived trait, underscoring the importance of studying diverse protists to reconstruct metabolic evolution.

A recent study proposed that PEDS1 was acquired by Filozoa, Excavata, Amoebozoa, and some small number of Fungi through four independent HGT events, each from an unknown donor (20). In contrast, our expanded taxon sampling and refined search strategies supports a more parsimonious scenario: a single HGT event from the phylum Myxococcota into an early eukaryotic ancestor. The topology of our trees (Figs. 4 and 5) shows the eukaryotic PEDS1 and FARAT sequences forming a monophyletic clade nested within, or sister to, the myxobacteria. This implies that the entire functional module was likely transferred en bloc. Indeed, HGT from Myxococcota to eukaryotes has also been proposed for key genes in the biosynthesis pathway of steroids, a class of fundamental eukaryotic membrane lipids (59–61). This pathway is also oxygen-dependent, as that for plasmalogens, and regulation of the two pathways is linked in mammals (62), with neither pathway reported in Archaea. Overall, these findings reinforce an essential relationship of early eukaryotes with aerobic myxobacteria in the evolution of eukaryotic membrane lipid complexity. Acquisition of the plasmalogen biosynthesis pathway, with consequences in membrane composition and function, may have conferred adaptive advantages.

The presence of PEDS1 in a handful of fungal species (e.g., Thozetella) appears to be the only exception to this single-origin model. If derived from the same ancestral HGT event, fungal PEDS1 would be expected to cluster near a PEDS1-like group, ideally closer to Holozoan sequences, but instead they lie apart and are more closely related to bacterial PEDS1. This suggests an independent more recent HGT event, also from Myxococcota, into this small group of Fungi, several of which are soil-dwelling, a trait shared with myxobacteria and known to promote HGT (63, 64). The apparent lack of plasmalogens in these handful of Fungi with PEDS1, AGPS, and FARAT may be because these recently acquired genes via HGT are not yet fully assimilated and optimized for function, consistent with the low levels of intron gain in the corresponding genes (Dataset S6), which reportedly correlates with low gene expression (63–65).

In our analysis, the patchy distribution of the pathway in eukaryotes is mainly explained by several independent events of secondary loss of the plasmalogen synthesis genes, an evolutionary process relevant and frequent in eukaryotic evolution (66, 67). A complete loss of the pathway could therefore account for its absence in anaerobic organisms like Metamonada, or Archamoebae, the sole anaerobic group within Amoebozoa. Secondary loss of the pathway is the most parsimonious evolutionary scenario in Archamoebae, whose aerobic sister groups do possess the pathway and is supported by our analysis reconstructing the evolutionary history of PEDS1 and FARAT genes. The alternative explanation that Metamonada never acquired the aerobic pathway is plausible if this group diverged prior to the common ancestor of Discoba and Amorphea. Future studies may build upon our findings on gene loss and gain to further elucidate the functional role of plasmalogens in eukaryotes.

It is important to acknowledge that, like previous studies, our work still grapples with the unresolved root of the eukaryotic tree (29). The relationships among the supergroups Metamonada, Discoba, Amorphea, and Diaphoretickes remain partially unresolved (29), which makes it challenging to draw further inferences. Several possibilities arise from the observation that the genes we studied are almost entirely restricted to Discoba and Amorphea. The simplest scenarios include either an HGT event to a common ancestor of both Amorphea and Discoba (if they are sister groups) or one involving the LECA followed by secondary loss in Diaphoretickes and Metamonada. We favor the latter scenario, drawing on recent phylogenomic analyses suggesting that excavate features represent the ancestral state of eukaryotes (30) (therefore Amorphea and Discoba would not be sister groups), and the accumulating evidence of HGT events from Myxococcota to LECA or pre-LECA (59, 68–70). While HGT events that introduce new functions are likely to be fixed by selection, these functions are often reversible and transient. Nonessential traits tend to be lost over evolutionary time when selective pressures shift (71). HGTs in eukaryotes are generally rare and short-lived, with many acquired genes providing only temporary benefits before being lost. However, the genes that are retained over long evolutionary periods, as observed in our study, are likely to be functionally significant or relevant (72). In sum, our work on the origin and evolution of aerobic plasmalogen biosynthesis points to a myxobacterial origin for this important eukaryotic membrane lipid biosynthetic pathway, explains its unusual evolutionary history, and underscores how HGT has shaped early eukaryotic metabolic capabilities.

Materials and Methods

A detailed description of methods, protocols, and experimental analysis can be found in SI Appendix. These include details on homology searches, phylogenetic analyses, protein structure prediction and alignment; growth of M. xanthus, selected protists, and fungi, complementation analysis in M. xanthus and immunoblot analysis of its cell extracts for expressed PEDS1 homologs, total cellular lipids isolation from M. xanthus, C. owczarzaki, C. perkinsii, T. harzianum (CBS 354.33), and T. aggressivum (CBS 100526), and FAME GC-MS fatty acid analysis. PEDS1 (and FAD4), GNPAT, FAR, and AGPS sequences and structures used for analyses in this study are detailed in Datasets S1–S6 provided as separate files, while plasmids and M. xanthus strains generated in the study for complementation analysis of PEDS1 homologs are listed in Dataset S7.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Dataset S04 (XLSX)

Dataset S05 (XLSX)

Dataset S06 (XLSX)

Dataset S07 (XLSX)

Dataset S08 (TXT)

Dataset S09 (XLSX)

Dataset S10 (TXT)

Dataset S11 (TXT)

Dataset S12 (TXT)

Dataset S13 (TXT)

Dataset S14 (TXT)

Dataset S15 (TXT)

Data, Materials, and Software Availability

All study data are included in the article and/or supporting information.