Functional remodeling in the zoospore to vegetative cell transition

To explore the changes in cellular structure and function as vegetative cells developed from zoospores, we tested which KOG classes (KOG--) or KEGG BRITE (ko---) groups were over- or underrepresented among the significantly differentially expressed proteins. S6 Fig and S1D Table display the full set of significantly enriched KOG/ko groups, including enrichment tests for each individual expression cluster (C1-C4). A curated subset of enrichment tests and KOG/ko groups of interest is shown in Fig 5. Gene Ontology (GO) enrichment was also performed for the cluster-based analysis (S1D Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Functional classification of differentially expressed proteins during the transition from zoospores to vegetative cells in A. limacinum.

Bars show the total number of significantly upregulated (red, right-facing) and downregulated (blue, left-facing) proteins (at any time point) relative to zoospores (time 0) for each selected KOG class or ko group. Unfilled triangles in the first column at right indicate functional groups significantly overrepresented (upward triangle) or underrepresented (downward triangle) among all differentially expressed proteins (Fisher’s Exact Test with 5% false discovery rate). Filled triangles in the last column indicate functional groups overrepresented among upregulated (C3 and C4, upward triangle) or downregulated (C1 and C2, downward triangle) genes in vegetative cells in comparison to zoospores. The comparisons shown here represent a curated subset of the full enrichment analysis shown in S6 Fig.

https://doi.org/10.1371/journal.pone.0326651.g005

Transition away from motility and environmental sensing

One of the most obvious morphological changes during the transition from free-swimming zoospore to non-motile vegetative cell is the loss of structures associated with flagellar motility. Consistent with the changes observed for alpha-tubulin (Fig 4), we observed a notable reduction in proteins linked with motility. Specifically, cilia and flagella-related proteins (such as ko03037 in Fig 5) and cytoskeleton proteins (such as ko04812 in Fig 5) were downregulated during the transition from zoospores to vegetative cells, likely associated with the loss of flagella during this phase.

KOG ‘signal transduction mechanisms’, which includes protein kinases (ko01001) and other pathways, was significantly overrepresented in early-downregulated cluster C1 and the combined set of downregulated proteins (C1 + C2) (Fig 5, S6 Fig). This is consistent with a decline in the environmental sensing systems needed by zoospores but not by vegetative cells. For example, the ‘RubyACR’ anion-conducting channelrhodopsin A35957 (called AlACR1 in [34]) was significantly downregulated in all vegetative cell time points compared to zoospores. Other channelrhodopsins, A7690 and A143491 (AlACR2 and AlACR3 in [34], respectively) were also downregulated, though not significantly (S2 Table). Notably, however, the signal transduction category was underrepresented among all differentially expressed proteins overall (Fig 5), suggesting that most signal transduction pathways remain relatively stable during the transition.

Several categories related to intracellular trafficking were also overrepresented among downregulated genes in the transition from zoospore to growing vegetative cell (KOG class ‘intracellular trafficking, secretion, and vesicular transport’; membrane trafficking ko04131; vesicle-mediated transport GO:0016192; intracellular protein protein transport GO:0006886; Fig 5 and S1D Table). Declines in specific proteins like the coatomer COPI subunits (A71350 and A142887 in cluster C1, A68400 and A139751 in cluster C2) are consistent with an enrichment of zoospores in vesicle-related activities involved in their initial attachment to surfaces [26]. These KOG/ko categories were also underrepresented among all differentially expressed proteins, again suggesting overall stability during this transition (Fig 5).

Remodeling of gene expression machinery

Changes in proteins associated with gene expression machinery suggest the zoospore could be poised for rapid proteome remodeling, both in the production of new proteins (relatively high ribosome content), and degradation of existing proteins (relatively high proteasome content). While we did not detect significant changes in ko groups, KOG classes, or GO terms associated with transcription, KOG class ‘RNA processing and modification’ and spliceosome proteins (ko03041) were overrepresented among upregulated proteins (Fig 5), particularly cluster C4 (S6 Fig). Proteins associated with translation, protein maturation, and protein trafficking showed strong responses. Ribosome (ko03011) proteins, along with KOG class ‘post-translational modification, protein turnover, chaperones’, proteasome (ko03051), and the GO terms ribosome (GO:0003735) and structural constituent of ribosome (GO:00037335), were downregulated in vegetative cells compared to zoospores, particularly in cluster C2 (Fig 5, S1D Table, S8A and S8B Figs). Chaperones and folding catalysts (ko03110) were also overrepresented in cluster C2, and several GO terms associated with protein folding and turnover (such as GO:0005839, proteasome complex, and GO:0006457, protein folding) were enriched among downregulated genes. Conversely, proteins involved in ribosomal biogenesis (ko03009), tRNA biogenesis (ko03016), and translation factors (ko03012), along with GO terms related to translation initiation and elongation, were enriched in genes upregulated in vegetative cells compared to zoospores, particularly in cluster C4 (Fig 5, S1D Table, S7 Fig).

It is known that the very early (first ~30 minutes) steps of zoospore germination in the chytrid Blastocladiella emersonii do not require protein or RNA synthesis; chytrid zoospores have stocked mRNA and proteins in the nuclear cap that support a bout of protein synthesis essential for completion of germination [35]. A structure like the nuclear cap has not been reported in labyrinthulomycete zoospores, but our data suggests they may be similarly pre-programmed for early steps of settlement and transformation, as has also been observed in some oomycetes [7]. Comparison with published RNA-seq data from A. limacinum zoospores is consistent with protein stockpiling (see Conclusions, below).

Metabolic reprogramming supports cell growth

Oxidation of endogenous fatty acids is commonly reported to be an important energy source for zoospores [36–38]. Many organisms, including many stramenopiles, store fatty acids in the form of triacylglycerols and sterol esters, sequestered in specialized organelles, called lipid droplets, which can serve as an energy reserve [39–41]. Thraustochytrid zoospores are commonly, though not universally, reported to contain lipid droplets [42]. Morita et al. [20] observed a decline in the number of lipid droplets in the first 4 hours after zoospore settlement and reported that the zoospores had much greater content of sterol esters (as % total fatty acid) than vegetative cells.

Consistent with this shift away from lipid-based energy reserves, proteins involved in lipid transport, storage, and metabolism were significantly downregulated during the transition to vegetative cells. This includes KOG class ‘lipid transport and metabolism’ (LTM) (Fig 5) and several GO terms associated with fatty acid metabolism (S1D Table). Three proteins associated with the lipid droplet [43] were significantly downregulated (acyl-CoA synthetase, A142380; a Rab1/YPT1 GTPase, A77077; short-chain acyl-CoA dehydrogenase, A115689; all in cluster C2). The downregulation of peroxisomal beta-oxidation enzymes such as enoyl-CoA hydratase (HSD17B4, A11390, C2) and 3-hydroxyacyl-CoA dehydrogenase (fadN, A41946, C2), along with mitochondrial lipid metabolism proteins such as carnitine/acylcarnitine translocase (CACT, A46635, C2), short-chain 2-methylacyl-CoA dehydrogenase (ACADSB, A116973, C2), trifunctional enzyme subunit alpha (HADHA, A136855, C1), carnitine O-palmitoyltransferase 2 (CPT2, A138285, C2), and acyl-CoA dehydrogenase (ACADM, A142917, C2), suggest a shift away from the breakdown of stored triglycerides and oxidation of long-chain fatty acids in the peroxisome and mitochondria during the transition to a vegetative life-stage ([44]; S8C Fig). The first step of peroxisomal fatty acid beta-oxidation produces H₂O₂, and the decline of a peroxisomal catalase (A138555, C1) would be consistent with a decrease in H₂O₂ production by that pathway. Also downregulated was the mitochondrial electron-transfer flavoprotein ubiquinone oxidoreductase (ETFQO, A46910, C2), identified as essential for mitochondrial fatty acid beta-oxidation and branched-chain amino acid degradation [45].

At the same time, components of the lipid transport and metabolism class linked to anabolic biosynthesis, such as isoprenoid biosynthesis, were upregulated during the transition. These include the mevalonate isoprenoid biosynthetic pathway (GO:0008299; S8D Fig, S1D Table) and related reactions, including two putative sterol 24-C-methyltransferases (A40066, C4 and A136674, C3; both GO:0003838), and another methyltransferase (A43854, C4) previously reported to be associated with the lipid droplet [43].

Carbohydrate metabolism also underwent substantial reprogramming, shifting from glycolysis and TCA cycle activity in zoospores toward gluconeogenesis in vegetative cells. Mitochondria-associated phosphoenolpyruvate carboxylase (A46884, C2), and GO terms associated with glycolysis (such as GO:0006096, including phosphofructokinase A139427, C1; S1D Table, S8E Fig), were more abundant in zoospores. In contrast, the cytoplasmic first committed step of gluconeogenesis, phosphoenolpyruvate carboxykinase (A84665, C3), along with fructose bisphosphatase (A70359, C4), was more abundant in vegetative cells. Also more abundant in the zoospores were several proteins associated with the TCA cycle, including pyruvate dehydrogenase (GO:0004739, A46525 and A117291 in C2), acetyl-CoA synthase (A83581, C1), citrate synthase (A36111, C2), and 2-oxoglutarate dehydrogenase (A142771, C2). On the other hand, malate dehydrogenase (A79000, C3) was more abundant in vegetative cells (S8F Fig). This pattern may reflect both a developmental transition and adaptation to the growth medium, which contains glycerol rather than glucose, along with the glucogenic amino acid glutamate.

The metabolic reprogramming extended to pathways supporting active cell growth and reproduction, marked by upregulation of amino acid and nucleotide metabolism. This included upregulation of the KOG classes ‘amino acid transport and metabolism’ and ‘nucleotide transport and metabolism’, the BRITE group amino acid related enzymes (ko01007), and several related GO terms (S8G and S8H Figs, S1D Table). Several enzymes in one-carbon metabolism were also upregulated (S8I Fig). The upregulation of proliferating cell nuclear antigen (PCNA, A115200, C3) by T2 could indicate an early onset of DNA replication. Yet many basic components of cell structure and function, such as ubiquitin system (ko4121), DNA repair and recombination proteins (ko03400), and KOG class ‘nuclear structure’, were underrepresented among significant proteins, suggesting that the relative protein investment in these systems is not much affected by the transition from zoospore to vegetative cell.

Finally, changes in energy production systems indicated a shift in the way ATP synthesis and regeneration are balanced during vegetative cell growth. KOG class ‘Energy production and conversion’ is overrepresented among differentially expressed proteins, with a similar number upregulated and downregulated (Fig 5; S1C Table). Several components of F-type ATPase (GO:0046933) were upregulated, while components of V-type ATPase were downregulated (S8J Fig). Also downregulated in vegetative cells were all three predicted ATP:guanido phosphotransferases (A42973, C2; A45810, C1; A140607, C2; GO:0004054), which are components of the phosphotaurocyamine pathway that may play a role in providing ATP to the flagella [46]. Adenylate kinases may also help replenish ATP [46]. Among eight putative adenylate kinases predicted in the A. limacinum genome, seven were detected in the proteome, and only one (A65144, C2) was differentially expressed, being more abundant in the zoospore stage (T0).

Taken together, these results indicate that the transition from zoospore to vegetative cell is characterized by coordinated metabolic reprogramming–from lipid catabolism to anabolic processes that support sustained cell growth.

Candidate proteins for bothrosome and ectoplasmic network

The bothrosome is a membranous organelle ~200 nm in diameter, located at a cup-like plasma membrane invagination. The basal layer of electron-dense material of the bothrosome is connected on one side with the cytoplasm and endoplasmic reticulum, and on the other with the interior of the EN [47,48]. The EN is a wall-less, membrane-bound compartment that may contain ‘net elements’ described as vesicles or internal membrane cisternae, but no cytoplasmic organelles like ribosomes or mitochondria [49]. New EN is thin, but thickens and develops more complex internal net elements after contacting a food particle and is actively involved in the hydrolysis of substrates and uptake of nutrients [24,25].

Because the EN and bothrosome are morphologically unique to Labyrinthulomycota, the core genes involved in orchestrating the EN/bothrosome likely include both widely conserved proteins (e.g., actin) and potentially novel factors specific to Labyrinthulomycota. We hypothesized that proteins uniquely conserved in Labyrinthulomycota and upregulated during EN formation are promising candidates for bothrosome and EN functions. Specifically, we looked for candidate proteins which are upregulated during the transition from zoospores to the EN-producing vegetative cells, and which have identifiable homologs in other Labyrinthulomycota, but no high-scoring homologues in other Eukaryota. In total, 18 such genes were identified, listed in Table 1 and S1E Table. None have been experimentally tested, and most lack functional annotations by standard sequence similarity methods. To glean insights into their potential roles, we analyzed these proteins using “deep homology” approaches (Phyre2 and HHPred; see methods; note that candidates do not have recognizable homology outside labyrinthulomycetes using “standard” homology searches such as BLAST). The 18 proteins could be divided into three categories: 1) proteins likely performing general cellular functions (e.g., transcription, translation, metabolism) that are upregulated during vegetative growth, and may be unusually divergent in Labyrinthulomycota, but are less likely to be involved in the EN/bothrosome; 2) proteins with features suggestive of direct EN/bothrosome involvement; and 3) proteins too divergent to assign confident structure or function. In particular, we are interested in candidates that fall into the latter two categories.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Potential EN/bothrosome candidate proteins.

https://doi.org/10.1371/journal.pone.0326651.t001

Three candidate proteins (A0A6S8AHQ2, A142921, and A42866) have bioinformatic predictions suggestive of direct involvement in EN formation or function. A0A6S8AHQ2 exhibits some similarity to micronemal proteins from parasitic protists. Phyre2 suggested A0A6S8AHQ2 matches micronemal protein 4 from Toxoplasma gondii (94.5% confidence) and SML-2 from Sarcocystis muris (93.8% confidence) (Table 1 and S1E Table). HHPred similarly identified micronemal protein 4 as its highest-confidence match (probability 95.5%). Microneme proteins are adhesins, which function in host cell recognition, attachment, and penetration [50,51]. Adhesins can assemble into hair-like appendages, such as pili and fimbriae, which help parasites bind to receptors on their hosts [52,53]. InterProScan and MEMSAT both predicted an N-terminal signal peptide, and a non-cytoplasmic/extracellular domain for A0A6S8AHQ2. Such features suggest this protein may be extracellular or secreted, raising the intriguing possibility of its involvement in the EN, perhaps performing a role related to EN adhesion or substrate interaction.

The second candidate protein with features suggesting a potential role in the EN was A142921, which Phyre2 suggested bore similarity to type VI secretion system protein (T6SS) (65.2% confidence). T6SS are contractile proteins used by predatory bacteria to inject cytotoxic effector proteins into prey cells [54,55]. MEMSAT predicted that A142921 possesses an N-terminal signal peptide, an extracellular domain, and a transmembrane domain. Thus, A142921 may have a role associated with the EN, perhaps as a secreted or contractile protein.

The third EN candidate protein potentially involved in the EN was A42866. MEMSAT predicted a transmembrane domain and a short extracellular C-terminus. Moreover, GPS-Lipid predicted a palmitoylation site, suggesting potential association between A42866 and a membrane. The highest confidence match suggested by Phyre2 was trafficking protein particle complex subunit 11 (TRAPPC11) (69.1% confidence), a subunit of the TRAPPIII complex involved in vesicular trafficking and autophagy [56–58]. HHPred indicated similarity to BAR protein (64.82% probability), as well as the BAR-domain containing protein endophilin (55.3% probability), which can be involved in modulating/sensing membrane curvature, vesicle membrane fission, and vesicle uncoating [59–62]. These predictions suggest that A42866 may have a similar function related to membrane remodeling or vesicular trafficking.

Two additional candidate proteins (A47522 and A141237) had clearly recognizable structural or functional domains, but ambiguous predicted functions; we wonder whether these ‘missing’ functions may be bothrosome or EN related. The highest confidence match to A47522 from both Phyre2 and HHPred was “SXP/RAL-2 family protein” (74.9% confidence and 48.81% probability, respectively), which are proteins of unknown function secreted by parasitic nematodes [63]. Supporting the possibility that A47522 might be extracellular, MEMSAT predicted an extracellular N-terminus, followed by a transmembrane domain.

A141237 was associated with the Pfam/InterPro/PANTHER domain “late embryogenesis abundant protein” (LEA, PF02987, IPR004238, PTHR47372:SF49). LEA proteins confer desiccation tolerance in plants and animals [64]. HHPred suggested high confidence similarity to various apolipoproteins, with the highest probability match of apolipoprotein E (98.53% probability). Apolipoprotein E is involved in the binding of cell-surface receptors with lipoproteins [65]. Thus, although an exact function of A141237 will ultimately need to be determined by experiment, these findings hint of a possible extracellular or EN-associated role.

Finally, three candidate proteins (A46070, A442, and A43122) were too divergent for clear bioinformatic inference (other than MEMSAT predicted that all three possess transmembrane and extracellular domains). These enigmatic proteins are particularly interesting candidates for EN-related functions, since the EN itself is unique to labyrinthulomycetes, potentially explaining why deep homology tools were ineffective. Another reason might be related to the fact that Phyre2 and InterProScan indicated they were likely highly disordered proteins. Future experiments are required to resolve their functions.

The candidate proteins identified here provide a starting point for unraveling the molecular basis of EN and bothrosome biology in thraustochytrids. Their predicted features, such as membrane association and extracellular or vesicular trafficking-related roles, align with what we know about the unique structural and functional characteristics of the EN and bothrosome. While their precise roles remain speculative, these proteins represent prime targets for uncovering evolutionary innovations tied to the emergence of these specialized organelles, and await future experimental validation.

Zoospore isolation and timing of vegetative cell development

Aurantiochytrium limacinum ATCC MYA-1381 [28,67] was grown at 25 °C in a 50 mL liquid culture of GPY media (3% D-glucose, 1.5% peptone, 0.5% yeast extract, 1.8% Instant Ocean sea salt) on an orbital shaker at 100 rpm. After 24 hours, the culture was spread on d-GPY agar plates (2% D-glucose, 1% peptone, 0.5% yeast extract, 1.75% Instant Ocean, 1% agar) and grown at 25 °C for 24 hours. Zoospore release was achieved by flooding the plate with 15 mL of artificial seawater (ASW: 1.8% Instant Ocean in dH₂O, filter sterilized), and waiting 2 hours when zoospore concentration in ASW pipetted off the flooded plate was estimated to be 2.75 million cells/mL using a Neubauer chamber. For microscopy, 500 µL of freshly isolated Aurantiochytrium limacinum zoospores were inoculated into 35 mm Nunc™ Glass Bottom Dishes (ThermoFisher Scientific, USA) containing 5 mL sterile A1 media (containing glycerol, maleic acid, and sodium glutamate in artificial seawater; [68]). The petri dish was left undisturbed for 15 minutes to allow the zoospores to settle down and attach to the glass surface. Then the petri dish was placed on the stage of a custom-assembled spinning disk confocal microscope consisting of an automated Zeiss frame, a Yokogawa CSU10 spinning disc, a Ludl stage controlled by a Ludl MAC6000, and an ASI filter turret attached to a Photometrics Prime 95B camera controlled with Metamorph software (version: 7.10.2.240). Settled A. limacinum cells were visualized using an oil-immersion Plan Apochromat 63x/1.4 NA DIC objective. Time-lapse imaging, including z-stack, was performed on 5–7 different fields for 12 hours with a 5 minute interval. Zoospore settling and vegetative cell growth were assessed for motility, presence of flagella, cell size, and EN development.

Proteomics

Zoospores were collected from a 1-day-old petri dish culture on d-GPY agar medium in 15 mL ASW as described above. 1 mL of zoospore solution was inoculated into each of twelve 150 mm diameter Petri dishes containing 35 mL liquid A1 media. Zoospores from the remaining 3 mL were collected by centrifugation (10 minutes, 2,400 x g) for time 0 (T0). After 2, 4, 6, and 8 hours (T2, T4, T6, and T8, respectively), cells were released from three replicate petri dishes using a cell scraper and collected by centrifugation. The biomass of each pellet was measured using a precision balance (Mettler Toledo AL54) and pellets were stored at −80 °C for protein extraction the following day. Each of five time points was represented by 3 biological replicates. Samples at Time 0 (T0) served as a reference point against which changes at subsequent time points (T2, T4, T6, T8) were compared.

The frozen cell pellets were resuspended in 1 mL of lysis buffer (500 µL 1M KCl, 1000 µL 25 mM MgCl₂, 500 µL 1M Tris pH 8.3, 0.45% NP40, 0.45% Tween, 100 µL 0.5% SDS, 8210 µL dH₂O), and vortexed for 20 minutes at high speed. Then, the mixture was centrifuged at 15,000 xg, 4 °C for 15 minutes. The supernatant was collected, placed on dry ice, and sent for protein analysis by Creative Proteomics (Shirley, NY, USA). Protein concentration was quantified using a bicinchoninic acid (BCA) assay (Thermo Fisher Scientific, USA). 60 µg of protein from each sample were prepared using dissolution buffer, denaturant, and reducing reagent from the BCA kit. Cysteine residues were blocked using cysteine blocking reagent, and the proteins were digested using trypsin solution from bovine pancreas (Promega). Peptides were labeled per replicate using iTRAQ-8plex (Sigma) tags 113, 114, 115, 116, and 117 for time points 0, 2, 4, 6, and 8 respectively. The labeled peptides were combined into one tube per replicate.

Peptide fractionation was performed per replicate with 15 components using HPLC. We used an Ultimate 3000 nano UHPLC system coupled with a Q Exactive mass spectrometer (Thermo Fisher Scientific, USA) with an ESI nanospray source. For iTRAQ-labeled samples, the full scan was performed between 350−1,650 m/z at the resolution 120,000 at 200 Th, and the automatic gain control target for the full scan was set to 3e6. The MS/MS scan was operated in Top 15 mode using the following settings: resolution 30,000 at 200 Th; automatic gain control target 1e5; normalized collision energy at 32%; isolation window of 1.2 Th; charge state exclusion: unassigned, 1, > 6; dynamic exclusion 40 s.

Raw MS files were analyzed separately against the two available predicted proteomes for Aurantiochytrium limacinum, from JGI (https://mycocosm.jgi.doe.gov/Aurli1/Aurli1.home.html) [30] and from MMETSP (from the 2020_06 uniprot_trembl release) [69], using Maxquant (1.6.2.14). The parameters were set as follows: protein modifications were carbamidomethylation (C) (fixed), oxidation (M) (variable), itraq-8plex; enzyme specificity was set to trypsin; maximum missed cleavages were set to 2; precursor ion mass tolerance was set to 10 ppm and MS/MS tolerance was 0.6 Da. 3006 of the 14859 (20.2%) JGI-predicted proteins, and 3298 of the 14622 (21.9%) MMETSP-predicted proteins were detected (Maxquant output is available at Dryad DOI https://doi.org/10.5061/dryad.2z34tmpxj). To combine the results into a single dataset containing all detected and significant proteins, we bidirectionally compared all the predicted MMETSP and JGI proteins using Diamond v2.0.10.148 [70]. 9,673 proteins formed pairs of reciprocal best hits. In addition, 3,356 JGI-predicted proteins with no MMETSP hit,1,886 MMETSP-predicted proteins with no JGI hit, and 1,866 MMETSP plus 3,099 JGI proteins that did not form reciprocal best hit pairs were initially combined (totally 19,880 proteins). After applying additional filtering to remove ambiguous or redundant assignments, the final non-redundant predicted proteome comprised 19,844 proteins. In total, 3,783 unique proteins were detected (S2 Table).

Throughout the text and tables, protein models from the JGI Aurli1 genome annotation are referenced using their original protein ID numbers with a prepended “A” (e.g., A43122). These identifiers can be searched at the JGI Aurli1 gene model portal by removing the “A” prefix and entering the numeric portion (e.g., “43122”) in the “Protein ID” field (https://phycocosm.jgi.doe.gov/Aurli1/). In contrast, the MMETSP-derived proteins use UniProt accession numbers (e.g., A0A6S8DQS1). These accessions follow the UniProt standard format, beginning with “A0A” followed by seven alphanumeric characters. These entries are available from the UniParc archive (https://www.uniprot.org/uniparc).

Differential abundance analysis

Differential abundance analysis was performed using the Bioconductor package DEP v1.20.0 and following their best practices [71] in parallel for the MMETSP and JGI datasets. For each of the two datasets, we created a summarized experiment object with the time points and replicates using the SummarizedExperiment Bioconductor package. A batch correction was performed using Combat from the package SVA, intensity values were normalized using variance stabilizing transformation (VSN), and limma was used to calculate empirical Bayes moderated t-statistics for time points 2, 4, 6, and 8 with respect to time point 0. The significance thresholds were adjusted to p-value <0.1 and fold-change greater than 1.0001 for upregulation or the inverse 1/1.0001 = 0.9999 for downregulation. Of the 3006 detected JGI proteins, 394 were significantly differentially expressed for at least one of the later time points when compared to time point 0, and 530 of the 3298 MMETSP proteins were significant. In total, 623 unique proteins were significantly differentially expressed (S2 Table).

To support interpretation, several annotations were added to the dataset. First, depending on the expression level at time point 0, all proteins with above median expression were classified as a set of highly expressed proteins. Second, the datasets were joined to Labyrinthulomycete orthogroup cluster tables collected from JGI (https://phycocosm.jgi.doe.gov/clm/run/Honfer1-comparative-qc.4746;bx5D-n?organismsGroup=labyrinthulomycota) to determine if a protein had homologs in other Labyrinthulomycete species; a protein was considered to be conserved if its orthogroup was present in 3 or more of the 5 genomes. Pfam annotations were added from the JGI proteome (https://phycocosm.jgi.doe.gov/mycocosm/annotations/browser/pfam/summary;I9DIJz?p=Aurli1). Predicted subcellular localizations were obtained for the detected proteins using MultiLoc2 [72]. KEGG assignments were obtained from BlastKOALA [73], GO terms were obtained using OmicsBox/Blast2GO [74], and other domain predictions from Panther [75], ProSite [76], Gene3D [77], and MobiDBLite [78] were acquired using InterProScan v5.55.88 [79].

Potential bothrosome/EN-related candidate proteins were identified as proteins which were upregulated in vegetative cells and did not have close homologs in organisms lacking a bothrosome/EN – that is, outside of Labyrinthulomycota. We used both a BLAST-based method and an orthogroup-based method to identify homologs of detected proteins. The first identified proteins which had BLASTp hits in labyrinthulomycetes, but not other stramenopiles or other organisms. We compared all A. limacinum proteins with a stramenopile protein database (all proteins in the Genbank nr database with taxID 33634; downloaded Sept. 27 2022) using BLASTp. We selected the top five hits for each protein based on the lowest e-value and highest percent identity. We grouped the query proteins into four groups: (1) no hits in any stramenopiles (aurli_only), (2) hits only in Hondaea fermentalgiana (laby_only), (3) at least 5 better hits in stramenopiles other than H. fermentalgiana (stramenopile_not_hondaea), or (4) hits in both H. fermentalgiana and other stramenopiles (both). For the last group, we further analyzed the differences between H. fermentalgiana and other stramenopiles by subtracting their respective scores (bitscore, e-value and percent identity). We categorized the bitscore difference distribution between hits to H. fermentalgiana vs other stramenopiles into different quantiles (0–85%, 85%−90%, 90%−95%, 95%−97.5% and 97.5%−100%), where proteins with a larger difference were more similar to H. fermentalgiana proteins than other stramenopile proteins. The second method identified A. limacinum protein homologs by using OrthoFinder v.2.5.5 [80] to cluster sequences from a protist-enriched database [12]. We filtered detected proteins with the following four conditions: (1) the protein was present in at least two Labyrinthulomycota genomes, (2) the protein was not present in more than one non-Labyrinthulomycota genome, and (3) the protein was upregulated in timepoints T4, T6, or T8, and (4) the protein had at least one copy in Hondaea fermentalgiana. Using the BLAST or OrthoFinder methods, 18 EN/bothrosome candidate proteins were identified, listed in Table 1. Further bioinformatic analyses were performed in order to hypothesize potential protein functions of the candidates. Two remote protein homology detection tools were used, Phyre2 and HHPred [81,82], using default parameters. Further protein features were predicted using GPS-Lipid, HAMAP, and PsiPred with MEMSAT [83–88], using default parameters, except GPS-Lipid which used the high stringency threshold setting. The output of these remote homology detection searches are summarized in S1E Table.

Functional enrichment statistics and visualization

Enrichment analyses for KOG classes and ko groups were performed using Fisher’s Exact Test per cluster, as well as combining down- and up-regulated clusters. These tests were performed using the hypergeometric function dhyper from the R package stats, and corrected for multiple testing using the Benjamini-Hochberg procedure. Differential enrichment among clusters for gene ontology (GO) terms was conducted with R (v4.4.0) package topGO (v2.56.0) using the weight01 algorithm with Fisher’s Exact Test, with a Benjamini–Hochberg correction (5% false discovery rate).

PCA plots, heatmaps, volcano plots, and bar plots for proteins of interest were created with DEP v1.20.0. Venn diagrams were created with ggVennDiagram v1.5.2 [89].

All code is available at https://github.com/phylogenomic/laby_proteomics/tree/main