Polycystine radiolarians associate with diverse phytoplankton

Abstract

Radiolarians are marine protists with a global distribution. Epipelagic radiolarians host photosynthetic endosymbionts, but the identity and specificity of this relationship appears to vary between radiolarian subgroups. While the class Acantharea and the order Collodaria both possess stable and relatively specific relationships with the haptophyte Phaeocystis and the dinoflagellate Brandtodinium nutricula, respectively, the orders Nassellaria and Spumellaria (which comprise the solitary Polycystinea) might have greater flexibility in terms of the identity of their photosymbionts. However, little molecular data has been generated to identify the phytoplankton with which polycystines can associate. Here, we performed short-read 16S and 18S rRNA gene sequencing with universal primers on single polycystine cells collected from the Sargasso Sea to determine common members of the polycystine holobiont. While previous work on polycystine photosymbioses suggested that they almost always exclusively associate with B. nutricula, we determined that polycystines instead associated with a wide diversity of phytoplankton, and the diversity of the polycystine holobiont is distinct from the diversity of environmental samples. Finally, we found that a substantial proportion of the reads associated with cell samples were of opisthokont origin (mostly copepods), revealing other possible interactions between an uncultivable and difficult-to-study protist with its environment.

INTRODUCTION

Oligotrophic regions of the ocean surface are hotbeds for mixotrophy (Ward et al., 2011; Hartmann et al., 2012; Leles et al., 2017). Mixotrophy is a trophic strategy which allows an organism to acquire nutrients via both phagotrophy and photoautotrophy (Mitra et al., 2016). This metabolic plasticity is likely advantageous in highly variable or seasonal environments (Lichtman, 2007), but also in nutrient-deplete environmental conditions such as oligotrophic gyres. Here, mixotrophs can outperform photoautotrophs by supplementing the nutrient demands of photosynthesis with heterotrophy, allowing the mixotroph to take advantage of both organic and inorganic nutrients (Zubkov and Tarran, 2008; Ward et al., 2011). Consequently, much of the primary productivity in oligotrophic gyres comes from mixotrophs (Mitra et al., 2016; Faure et al., 2019).

Radiolarians are marine planktonic protists that are especially abundant in oligotrophic gyres. Epipelagic radiolarians are mixotrophs in that they possess microalgal endosymbionts. Our knowledge about the biology and ecology of this group is limited due to the present inability to cultivate them, and while some of them are large enough to identify and count with an Underwater Vision Profiler in situ (Biard et al., 2016), most radiolarians are ~ 200 μm or smaller (Biard, 2022), making them inaccessible to study except by single-cell methods. Nonetheless, we know that Radiolaria are heterotrophs that feed on microzooplankton, sinking particles, other protists and copepods, likely among other prey lineages (Anderson et al., 1984; Swanberg and Caron, 1991; Suzuki and Not, 2015).

Radiolarians possess intricate tests, or skeletons, and the group can be subdivided based on the composition of their skeleton. Acantharea are a radiolarian class whose members possess a skeleton made of strontium sulfate, while the class Polycystinea consists of radiolarians with a silica skeleton. Polycystinea are further subdivided into three orders: Nassellaria, Spumellaria and the colonial Collodaria, all of which can be found throughout the water column.

The photosymbiotic Radiolaria are estimated to fix significantly more carbon than the plankton community around them (Caron et al., 1995), and these symbioses have been studied and observed for centuries. For example, in the 1800’s, Thomas Huxley noted the presence of “yellow cells” within the collodarian Thalassicolla (Huxley, 1851). The nutritive value of the yellow cells for their host was later confirmed by Ernst Haeckel (1871) with the discovery of significant amounts of intracellular starch in each symbiont cell. They have since been observed and studied extensively by morphology, and while radiolarian photosymbionts are often yellow, they can also be green or red and vary in size, usually between 5 and 20 μm across. Furthermore, some researchers have micromanipulated radiolarian cells to remove and cultivate the symbionts, subsequently using light and electron microscopy to study them (Probert et al., 2014; Yuasa et al., 2016), and others have studied the transfer of photosynthates or other nutrients from symbiont to host with ¹⁴C isotope evidence and later, nanoSIMS (Anderson et al., 1983; Decelle et al. 2021).

Despite the long history of study into radiolarian photosymbioses, most of the information we know about these relationships is limited to Acantharea and Collodaria. All photosymbiotic acanthareans, i.e. those belonging to Acantharea clades E & F, host the bloom-forming haptophyte genus Phaeocystis (Decelle et al. 2012a; Mars Brisbin et al., 2018), however, there is no signal for co-diversification between the two groups (Decelle et al., 2012a). Acanthareans can secondarily host the haptophyte Chrysochromulina, either instead of or in addition to Phaeocystis (Mars Brisbin et al., 2018), and one acantharean genus, Acanthochiasma, is additionally capable of hosting a wide array of dinoflagellate lineages, often with multiple photosymbiont lineages per cell (Decelle et al., 2012b).

Meanwhile, collodarians have an obligate and specific photosymbiosis with the dinoflagellate Brandtodinium nutricula (Peridiniales, Dinophyceae) (Probert et al., 2014). Symbiotic dinoflagellates are housed within the shared ectoplasm of the collodarian colony, where they divide (Anderson, 1976). The dinoflagellate can nevertheless revert from the athecate, symbiotic morphology to its thecate, free-living form when removed from the collodarian, indicating that the relationship is not obligate for the dinoflagellate (Probert et al., 2014). Like the Acantharea-Phaeocystis photosymbiosis, this relationship is widespread across collodarian lineages.

While collodarian and acantharean photosymbioses have been well-studied in the last decade, those belonging to the other two polycystine lineages (Nassellaria and Spumellaria) are less documented. Brandtodinium nutricula was determined to be a prevalent photosymbiont to all of Polycystinea and not just Collodaria (Probert et al., 2014). However, the methods of this study preferentially selected for dinoflagellate symbionts, which precluded the possibility of detecting symbionts from other phytoplankton groups. Others have used similar methods to determine that, in addition to B. nutricula, the dinoflagellate Gymnoxanthella radiolariae (Gymnodiniales, Dinophyceae) is also frequently in association with Nassellaria and Spumellaria (Yuasa et al., 2016). A handful of studies have additionally reported the presence of non-dinoflagellate photosymbionts in polycystines, including the haptophyte Chrysochromulina, the chlorophyte Tetraselmis and the prominent and widespread cyanobacterium Synechococcus (Gast and Caron, 2001; Foster et al., 2006a, 2006b; Yuasa et al., 2012, 2019). In fact, one spumellarian genus (Dictyocoryne) has been identified to have a broad range of symbionts (Foster et al., 2006a; Yuasa et al., 2012, 2019), and even multiple lineages co-existing within a single cell (similar to the acantharean Acanthochiasma), further highlighting the diverse symbiotic partners possible among polycystine radiolarians (Anderson and Matsuoka, 1992).

Polycystinea has the potential to become an excellent candidate group for studying a broad range of concepts relating to photosymbiosis due to the seemingly varied photosymbiotic capabilities present across lineages, where Collodaria forms exclusive associations with one alga, while the rest seem more flexible. However, most of the above findings were determined for just one or a couple of polycystine species and warrant further investigation. Foundational information relating to the diversity, specificity and frequency of polycystine photosymbioses stands in the way of asking conceptually-based questions about their photosymbioses. Here, we address this knowledge gap by describing the holobiont community of 54 radiolarian cells that span the diversity of Polycystinea, collected from the oligotrophic Sargasso Sea. Using single-cell metabarcoding of 16S and 18S rRNA gene markers, we found that a wide diversity of microalgae, which we hypothesize are polycystine photosymbionts, are consistently present in association with solitary polycystines (Nassellaria and Spumellaria).

MATERIALS AND METHODS

Sample collection

Environmental plankton samples were collected with vertical net tows from 100 m (the approximate depth of the deep-chlorophyll maximum) to the surface with a 50 cm diameter, 50 μm mesh net. Samples were collected at four locations in the vicinity of Bermuda (Supplementary Table 1) over the span of eight days from May to June of 2022. Samples were immediately processed upon returning to land, where they were concentrated by filtering through a 0.5 μm mesh. Polycystine cells were isolated using a drawn-glass micropipette and imaged with DIC optics on an Olympus IX83 inverted microscope. Each cell was washed thrice in sterile filtered seawater to remove contaminants before being stored in 10 μL of RNAlater. Concentrated environmental samples were also stored in RNAlater at a ratio of 1:5. Isolated cells and environmental samples were stored at 4°C overnight to allow penetration of the RNAlater, and then transferred to −20°C until further processing.

18S and 16S short-read amplicon sequencing

DNA was purified from single isolated cells and environmental samples using the MasterPure DNA & RNA Purification Kit (Epicenter, Madison, Wisconsin, USA). Amplicon libraries were generated using a two-step PCR protocol with dual index sequencing (Kozich et al., 2013). The V4-V5 region of the plastid, bacterial or cyanobacterial 16S rRNA gene was PCR amplified using nex-515Y/926R primers (Parada et al., 2016). The V4 region of the 18S rRNA gene was also PCR amplified using nex-TAREuk primers (Stoeck et al., 2010) or UNonMet primers (Bass and del Campo, 2020), where all samples were first subjected to PCR amplification with nex-TAREuk primers, and if this was unsuccessful, PCR amplification with UNonMet primers was attempted instead (Supplementary Table 2). Samples amplified with UNonMet primers subsequently underwent a nested PCR with nex-TAREuk primers and limited cycles to generate PCR products of the correct length for Illumina sequencing (Supplementary Table 2). The UNonMet primers were used for 5/7 of the environmental samples and 22/54 of the single cell samples. PCR products were then purified with AMPure beads (Beckman Coulter, Brea, CA, USA). The final PCR attached indexing barcodes to both ends of each amplicon (Hamady et al., 2008). PCR products were again purified with AMPure beads, quantified with Qubit broad range dsDNA fluorescent dye (Invitrogen, Carlsbad, CA, USA) and pooled at approximately equimolar concentrations before sequencing on the Illumina MiSeq using v3 2 × 300 paired-end chemistry.

Amplicon data processing

Sequence quality was assessed with FastQC v.0.11.7 using default settings. Data processing was implemented with Qiime2 version 2022.11 (Bolyen et al., 2019) and quality-filtered with DADA2 using default settings (Callahan et al., 2016), where the first 20 and the last 10–60 reads were trimmed to remove primers and improve sequence quality. Taxonomy was assigned to 18S and 16S ASVs using the naïve Bayes classifier plugin within Qiime2 (default settings), trained with the PR2 reference database (version 4.14.0) and the PhytoRef reference database (Decelle et al., 2015), respectively. Non-target sequences, defined here as ASVs which did not receive a taxonomic assignment or received a taxonomic assignment within Proteobacteria, Streptophyta (land plants) or termite-symbiotic Metamonada (lab-specific contamination), were filtered out of the dataset before downstream analyses.

The taxonomic assignments of ASVs were assessed genus by genus and designated as potential photosymbiont lineages if they were assigned to a known phytoplankton group that is small enough to be plausibly maintained within radiolarian cells (i.e. excluding those > 200 μm). The definition of “potential photosymbiont” was intentionally left vague so that the full spectrum of possible associations were taken into account, without leaving any algal lineages out. For quality control, lineages with a read depth below 20 per cell were filtered from the cell samples before analyses. Non-metric multidimensional scaling (NMDS) was used to visualize the organization of sample beta diversity in 2D space using the “metaMDS” function of the R package vegan (Oksanen, 2022). For this, unweighted UniFrac distance matrices were first calculated in Qiime2 with default settings. PERMANOVA tests, calculated within Qiime2 with 999 permutations, were used to assess whether the difference of between-group and within-group means (where groups included aposymbiotic cells vs. symbiotic cells, polycystine subgroups and environmental vs. cell samples) from the aforementioned distance matrix were statistically significant.

The proportion of cells containing each symbiont genus was converted to z-scores (the number of standard deviations from the mean) and then plotted as a heatmap, both of which were performed using the R package gplots. Differential abundance of taxa between all cell samples and all environmental samples was assessed with DESeq2, implemented within R, using Wald significance tests and parametric fitting of dispersions to the mean intensity. For this, only taxa with a read depth of 100 or greater were analyzed because comparing the abundance of a taxon with low read depth across two treatments could result in a biased interpretation of underrepresentation in one treatment, when the taxon is nearly absent in the dataset of both treatments. Results were considered significant with an alpha threshold of 0.05.

Sample categorization

Isolated cells were categorized as either photosymbiotic or aposymbiotic. This distinction was made based on a combination of molecular and morphological evidence, where the presence of potential photosymbiont lineages (as defined above) in the metabarcode dataset of a cell sample counted as molecular evidence. The photosymbiotic nature of some cells was visually evident, in that they possessed countless algal cells in the polycystine cytoplasm. Similarly, the aposymbiotic nature of some cells was visually evident, in that the polycystine cytoplasm possessed no algal cells, nor any pigmentation. In contrast, the symbiotic nature of some cells was visually ambiguous, in that pigmented cells in the cytoplasm of the radiolarian were limited but present, or it was difficult to make sense of cell contents within a complex skeleton with light microscopy alone. For the purposes of this study, cells were deemed photosymbiotic only if they had both the molecular evidence of photosymbionts, and they were either visually photosymbiotic or were visually ambiguously photosymbiotic.

Full-length 18S sequencing and processing

To improve phylogenetic placement, the full-length radiolarian 18S rRNA gene was amplified with radiolarian-specific primers or with custom Spumellaria primers designed for this study (Supplementary Table 2), although the full-length 18S of the solitary Polycystinea is notoriously difficult to amplify. PCR products were either purified and sequenced directly or ligated into the pCR4-TOPO vector using the TOPO TA Cloning Kit and cloned with One Shot TOP10 chemically competent Escherichia coli (Invitrogen, Carlsbad, California, USA) following the manufacturer’s protocols. Inserts from positive transformant colonies were amplified using the sequencing primers M13F and M13R and then purified. All PCR products were sequenced at the Arizona State University Genomics Core Facility on both strands using M13F and M13R, plus TAREukF/R as internal sequencing primers (Stoeck et al., 2010) (Supplementary Table 2).

Phylogenetic analyses

Radiolarian 18S clone sequences were trimmed of vectors and assembled using Geneious (version 2023.2, www.geneious.com). Directly sequenced 18S sequences were also assembled using Geneious. The identity of new sequences was initially checked with a BLASTn search against the NCBI NR database, and then aligned using MAFFT default settings (Katoh et al. 2002) with previously published sequences from their respective lineages: Nassellaria including Collodaria (Biard et al., 2015; Sandin et al., 2019) and Spumellaria (Sandin et al., 2021). Radiolarian 18S V4 amplicons were aligned secondarily to the full-length 18S alignments using the hmmalign feature of HMMer v.3.3.1 with default settings (Eddy, 2008). Ambiguously aligned sites were removed using TrimAl v.1.4.1 default settings (Capella-Gutiérrez et al., 2009). Maximum likelihood phylogenies were estimated using IQ-Tree v.1.6.12 (Nguyen et al., 2015). Support for nodes was assessed from 1000 ultrafast bootstrap replicates (Hoang et al., 2018).

RESULTS

Sampled radiolarians span the diversity of Polycystinea

In order to identify polycystines and their photosymbionts in the Sargasso Sea, 109 radiolarian cells were initially imaged, isolated, and subjected to molecular analyses. The 18S V4 region was successfully amplified and sequenced from 54 of those cells, of which 24 were also successful for 16S V3-V4 sequencing and seven were successful for full-length radiolarian 18S Sanger sequencing (Supplementary Table 1, Supplementary File 1).

The taxonomic assignation of sequenced cells was made based on a combination of morphological features and the phylogenetic position of the associated molecular data, if available. Using the taxonomic frameworks provided by Sandin et al. (2019, 2021), 37 of the isolated cells could be assigned to genera. The remaining 17 could only be assigned to the superfamily level due to the limitations of discerning internal skeletal structures with light microscopy alone, and/or lack of radiolarian molecular data. The phylogenetic position of the radiolarian V4 ASVs and full-length 18S sequences (when available) was consistent with morphology, though 20 cells did not yield a radiolarian ASV at all and were thus assigned taxonomy solely based on morphology.

Of the 54 sequenced polycystines, 31 were visually photosymbiotic, and the algal cells present varied in size, shape, quantity and color within the polycystine cytoplasm (Fig. 1A–H). Eight cells were visually aposymbiotic (Fig. 1L–M). The rest (15 cells) were visually ambiguous in terms of their symbiont status (Fig. 1I–K). Of the visually ambiguous cells, 12 possessed molecular evidence for the presence of potential photosymbionts while the remaining three did not. One cell (bios087) appeared visually photosymbiotic; however, we did not recover algal ASVs for this cell, so it was considered aposymbiotic. In sum, 43 cells were categorized as photosymbiotic and 11 were categorized as aposymbiotic.

Sequenced polycystines fell within seven and six known nassellarian and spumellarian superfamilies, respectively (Fig. 2, Supplementary File 2). The 18S rRNA gene phylogeny of Polycystinea produced for this study agrees with previously reported 18S rRNA phylogenetic trees and places Collodaria as a subgroup to Nassellaria, as has been previously inferred (Ishitani et al., 2012; Yuasa and Takahashi, 2016). Two aposymbiotic cells (one of which is represented by a full-length 18S sequence) additionally formed a clade that branched sister to the rest of Spumellaria. The sampled epipelagic polycystines from the Sargasso Sea thus span a substantial portion of the phylogenetic diversity of Polycystinea. Of these, the unknown, deep-branching spumellarian lineage and the superfamily Stylodictyoidea were the only two lineages to be exclusively aposymbiotic in our sequenced cells. The rest of the aposymbiotic radiolarians fell within the spumellarian superfamilies Spongopyloidea, Stylodictyoidea and Pylonioidea, and the nassellarian superfamilies Eucyrtidioidea and Plagiacanthoidea (Fig. 2). Among the superfamilies that contained a combination of symbiotic and aposymbiotic cells, aposymbiotic cells comprised between 20% and 67% of each individual superfamily (Fig. 2).

Radiolarians associate with diverse phytoplankton lineages

Taxa identified in our single cell 18S amplicon data were considered possible symbiont lineages by the broad criteria of being photoautotrophic and small enough to be a symbiont. These included the eukaryotic lineages Dinophyceae, Chlorophyta, Haptophyta and the cyanobacterium Synechococcus (Fig. 3). Dinoflagellates were the most common photosymbionts in radiolarian cells, both in terms of species richness and the proportion of isolated radiolarian cells in which they were detected. A total of six and nine dinoflagellate genera were identified in the 16S and 18S datasets, respectively; however, only Azadinium and Prorocentrum were identified in symbiotic cells from both datasets. 25 of the 43 symbiotic cells exclusively contained dinoflagellate symbionts (Fig. 4). Dinoflagellates were also the most abundant symbionts in terms of read counts; Brandtodinium nutricula comprised 100% of the reads in 10 cells, while Gymnoxanthella radiolariae and Lepidodinium comprised 100% of the reads present in two cells each. There were only two symbiotic radiolarian cells that did not contain dinoflagellate ASVs (Fig. 4). Of all the symbiont lineages, Lepidodinium and Brandtodinium were the most common by host cell count and read depth, where they were present in a total of 25 and 20 cells, respectively (Fig. 3). Interestingly, two dinoflagellate genera (Lepidodinium and Amphidinium) were only detected in the 16S datasets and the environmental 18S datasets, but not the single cell 18S datasets. The presence of plastid genes and the absence of nuclear genes in single-cell datasets, in combination with the unequivocal presence of the two dinoflagellate genera in the surrounding water column, might indicate the potential of polycystines to perform kleptoplasty, although other reasons (e.g. sequencing artifacts) could also result in the loss of the 18S signal in the cell.

Chlorophytes were the second most common eukaryotic photosymbiont, found in a total of 11 radiolarian cells across the 16S and 18S datasets. The chlorophyte genera considered to be possible photosymbionts to polycystines included Tetraselmis, Micromonas, Chlorella, Myrmecia and Trebouxia, although Myrmecia and Trebouxia only occurred in one cell each (Fig. 3). Chlorophyte genera were generally present in low read abundance; however, two nassellarian cells contained solely chlorophytes as photosymbionts (in one cell, Chlorella, and in the other cell, Trebouxia and Tetraselmis). Finally, haptophytes were only found in three radiolarian cells across the 16S and 18S datasets, and never exclusively (Fig. 4). The haptophyte genera detected were Phaeocystis, Chrysochromulina and Prymnesium, but all three were only present in one cell each (Fig. 3).

The cyanobacterium Synechococcus was found in 11 of the 24 cells that underwent 16S amplicon sequencing (Fig. 3). Like the haptophyte symbionts, Synechococcus was never the exclusive photosymbiont to a polycystine cell (Fig. 4). Synechococcus was the fourth-most prevalent photosymbiont (behind the dinoflagellates Brandtodinium, Lepidodinium and Amphidinum) (Fig. 3). It was also the fourth-most abundant in terms of total read depth, behind the dinoflagellates Lepidodinium, Brandtodinum and Gymnoxanthella, in that order.

Polycystine photosymbionts are variable across radiolarian superfamilies

The total read depth of B. nutricula, the known and widely accepted photosymbiont to Polycystinea, was 3-fold higher within Nassellaria compared to Spumellaria. 10 of the 33 nassellarians belonging to the superfamilies Archipilioidea, Carpocaniidae, Collodaria, Plagiacanthoidea and Pterocorythoidea exclusively hosted B. nutricula. Meanwhile, the total read depth of G. radiolariae, another widely accepted photosymbiont to Polycystinea, was 4-fold greater within Spumellaria compared to Nassellaria. Only three of the 15 spumellarians possessed just one microalgal lineage: two of which belonged to the superfamily Lithocyclioidea and exclusively hosted G. radiolariae, and one of which belonged to the superfamily Hexastyloidea and exclusively hosted the dinoflagellate Azadinium.

The nassellarian superfamilies Acanthodesmoidea, Archipilioidea and Eucyrtidioidea, in addition to the spumellarian superfamilies Pylonioidea, and Hexastyloidea had particularly diverse photosymbiont communities, hosting photosymbionts from each major lineage (Dinophyceae, Chlorophyta, Haptophyta and Cyanobacteria) (Fig. 3). While 17 polycystine cells had just one photosymbiont lineage, the average number of symbiont lineages per cell was 3.1, indicating that polycystines are likely capable of hosting diverse symbiont communities. Furthermore, there does not appear to be a relationship between the relatedness of radiolarian superfamilies and the number or identity of their symbionts (Fig. 2); however, future sampling efforts might eventually reveal one. Similarly, while the beta diversity of algal communities in cell samples was significantly different from environmental samples (PERMANOVA; P = 0.001), the beta diversity of algal communities present in polycystine cells is not significantly different between Nassellaria and Spumellaria (PERMANOVA; P = 0.067) (Fig. 5, Supplementary File 3).

The community of organisms associated with Polycystinea is distinct from the environment

A plurality of reads sequenced in both the cell samples (27.9%) and the environmental samples (37.8%) were of copepod origin, likely due to the copepod bloom present in the Sargasso Sea during the month of sampling (Fig. 5). Radiolarian ASVs had the second-highest read depth across the single-cell samples (26.1%). Possible photosymbiont ASVs comprised 21.8% of reads present in cell samples, while this group only comprised 1.6% of environmental sample reads. Also frequently present in cell samples were ASVs belonging to a variety of other Metazoa (14.4% of all cell sample reads); however, the proportion of reads belonging to this group was substantially higher in environmental samples (30.0%) (Fig. 5).

Although metazoans comprised a large proportion of the ASVs present in cell samples, the composition of non-target ASVs in cell samples compared to environmental samples is significantly different between the two sample types (PERMANOVA; P = 0.001) (Fig. 5). Meanwhile, the composition of non-target ASVs between isolated aposymbiotic and symbiotic cells is not significantly different (PERMANOVA, P = 0.148) (Fig. 5), indicating that the non-target ASVs are more likely to be present in or on radiolarian cells selectively, i.e. as prey items, than as contamination.

According to the results of our DESeq2 differential abundance analysis, most of the taxa responsible for the significant difference between environmental samples and cell samples were potential photosymbiont lineages (Fig. 6). The following non-algal lineages were also significantly more abundant in cell samples compared to environmental samples: the copepod genus Undinula, marine fungi, stramenopiles from the little-studied group of heterotrophic flagellates Bicosoecida, and marine arachnids. Among the taxa which were significantly more abundant in the environmental samples include diatoms, stramenopiles from the order Sarcinochrysidales, the copepod genera Acartia and Candacia, along with other opisthokonts like mollusks and annelids, in addition to the ubiquitous parasite lineage Syndiniales (specifically, MALV Group IV) (Fig. 6).

DISCUSSION

What does the identity of the photosymbionts say about the polycystine photosymbiosis?

Here, we provide confirmation that B. nutricula is the most common photosymbiont to Polycystinea as had been concluded in previous literature (Probert et al., 2014; Liu et al., 2019; Yuasa et al., 2022); however, B. nutricula was only present in 58% of the total symbiotic cells isolated. Previous work reported the possibility that many radiolarian photosymbionts identified as B. nutricula or Gymnodinium could be misidentified and are instead Gymnoxanthella radiolariae (Yuasa et al., 2016); however, we found that just 14% of our samples contain this dinoflagellate lineage. These two results might indicate that solitary polycystines are less selective about the identity of their photosymbiont than has otherwise been reported (Probert et al., 2014). This is evident in that the average number of photosymbiont lineages present per cell was three, and in that the most common symbiont lineage was only present in approximately half of all symbiotic cells.

Our results further indicate that while solitary polycystines have photosymbiotic preference for dinoflagellates, especially B. nutricula, they are additionally capable of associating with Chlorophyta, Haptophyta and Cyanobacteria. This is not entirely new information; Chrysochromulina (Haptophyta) and Tetraselmis (Chlorophyta) have both been identified in radiolarians before (Gast et al., 2000; Yuasa et al., 2019). Similarly, trends in chlorophyll autofluorescence of polycystine symbionts support the notion that Synechococcus is an endosymbiont to the spumellarian genera Spongaster and Dictyocoryne (Zhang et al., 2018). The above studies, however, only report these symbiont lineages in one or two radiolarian genera. Our findings support the presence of these symbionts across diverse polycystines.

Several possible photosymbiont lineages detected here, to our knowledge, have not been identified previously as photosymbionts: Prorocentrum, Margalefidinium and Lepidodinium. While it is certainly possible that the association between these lineages and radiolarians is more in line with predation or even simply contamination during the single-cell isolation procedure, obscure dinoflagellate lineages, including Azadinium and Heterocapsa, that we found in our solitary polycystine samples have also been found as photosymbionts to the acantharean genus Acanthochiasma (Decelle et al., 2012b). These lineages were previously never reported as symbionts either. Similar to Decelle et al. (2012b) we found that these lineages occur at significantly higher relative read depths in hospite than in the environment. Maybe the tendency of radiolarians to host obscure and uncommon photosymbiont lineages is reflective of their “professional host” qualities, in that the benefit of the metabolic plasticity derived from transient photosymbioses with whichever available phytoplankton is greater than the benefit of an evolved, established photosymbiosis with just one partner. Such professional host qualities would not be unique to Polycystinea; many marine invertebrate lineages are known to host diverse, seemingly functionally redundant symbionts, including marine sponges, ascidians and coral (Baker, 2003; Evans et al., 2017; O’Brien et al., 2019; Oliveira et al., 2020). Similar to Polycystinea, coral also tend to preferentially host one dinoflagellate genus as their main symbiont, and can additionally host many fringe species at lower abundance (LaJeunesse, 2002). However, at present, our results can only act as hypotheses for potential symbioses present in Polycystinea.

It is also worth noting that, while we captured a wide diversity of possible symbionts in this study, the diversity of photosymbiont lineages associated with Polycystinea might be higher still. We sampled radiolarians from just one region of the ocean during just one season; thus, we simply captured a snapshot of the symbionts possible to Polycystinea during these specific environmental conditions. Previous work that included sampling locations throughout the North Pacific revealed a similarly diverse repertoire of dinoflagellate and haptophyte genera present in the radiolarian holobiont (Nakamura et al., 2023). Interestingly, the phytoplankton lineages detected in Nakamura et al. (2023) only partially overlap with the ones detected here, where overlapping lineages include the haptophytes Prymnesium and Chrysochromulina, and the dinoflagellates Gymnoxanthella and Brandtodinium. The fact that most lineages detected here and in Nakamura et al. (2023) are unique to our two separate bodies of work further supports that polycystines are capable of associating with an even greater diversity of phytoplankton than we report here.

We additionally determined that the dinoflagellate Lepidodinium is present in the environment, but among our cell samples, it is exclusively present in the plastid 16S rRNA dataset. The absence of Lepidodinium in the 18S dataset of our cell samples could indicate that polycystines only retain the Lepidodinium plastid and digest the rest of the cell (also known as kleptoplasty). This would explain why the nuclear 18S of Lepidodinium is absent but the plastid 16S is present. The possibility that solitary polycystines steal the plastids from the dinoflagellate Lepidodinium, while preliminary and limited, is enticing. Lepidodinium is a dinoflagellate genus that possesses plastids of chlorophyte origin, also due to kleptoplasty (Matsumoto et al., 2011; Kamikawa et al., 2015). Solitary polycystines could therefore be grand thieves, stealing plastids which were already stolen from the original owner. This relationship would, thus, mirror the dynamics of the stolen plastids present in the dinoflagellate genus Dinophysis, which performs kleptoplasty on the ciliate Myrionecta rubra, which receives its plastids from the prey cryptophyte, Teleaulax acuta (Johnson and Stoecker, 2005; Nagai et al., 2008; Minnhagen et al., 2011); however, our data merely demonstrate this relationship as a possibility in Polycystinea. Whether our findings indicate a system of kleptoplasty or instead represent methodological biases, like differential amplification of a lineage based on primer selection, is yet to be determined.

There are many similarities between polycystine symbioses and the relationships present in symbiotic Foraminifera, which are either sister to or nested within Radiolaria (Sierra et al., 2013, 2022; Krabberød et al., 2017). Benthic Foraminifera from the order Rotallida are known to steal plastids from diatoms (Pillet et al., 2011; Jauffrais et al., 2018; Pinko et al., 2023). Benthic Foraminifera (and some planktonic Foraminifera) are also host to diverse photosymbiont lineages, some of which are shared between Foraminifera and Radiolaria, such as Tetraselmis (Gast et al., 2000) and Pelagodinium (Decelle et al., 2012b). Like polycystines, Foraminifera host photosymbionts from Chlorophyta, Haptophyta, Cyanobacteria and Dinophyceae, but unlike polycystines, they additionally host rhodophytes and diatoms (Lee, 2006). While Foraminifera as a whole are host to a diverse suite of phytoplankton lineages, individual lineages within Foraminifera have a general pattern of high fidelity to a single algal lineage, which is more similar to the photosymbioses present in Collodaria. Considering the close relationship between Foraminifera and Radiolaria, in addition to the shared traits between symbioses of the two groups, we therefore hypothesize that the phytoplankton present in our single-cell metabarcode datasets represent true photosymbiont diversity to Polycystinea; however, future work might illuminate these relationships more clearly.

The radiolarian holobiont is distinct from the surrounding plankton community

While it is difficult to study the biological interactions among uncultivable organisms, metabarcode sequencing of single cells provides some insight into these interactions (Bird et al., 2017; Greco et al., 2021; Boscaro et al., 2023). As such, the substantial presence of copepod and cnidarian ASVs could be indicative of a predator–prey relationship, in which radiolarians phagocytose the metazoan at some point in its life stage. This possibility is supported by the fact that both Radiolaria and Foraminifera are known to be heterotrophs with a diverse diet (Culver and Lipps, 2003; Suzuki and Not, 2015), and that small invertebrates including copepods are a known prey to Foraminifera (Culver and Lipps, 2003). Still, the high relative read abundance of metazoans in cell samples could be the result of several factors, including (i) preferential binding of universal eukaryote primers to the metazoan 18S, (ii) higher cell count of metazoans in the Radiolaria holobiont relative to the single-cell constituents, like photosymbionts or parasites or (iii) the copepod bloom that was present in the Sargasso Sea at the time of sampling, which could cause a greater composition of the detritus consumed by some radiolarians to be copepod fecal pellets, or a higher likelihood of copepod eDNA to be present on the sticky exterior of the radiolarian.

Considering that the non-algal beta diversity of cell samples was significantly different from that of the environmental samples, there appears to still be some selection in terms of which metazoans were consumed or present in association with the isolated radiolarians. Our DESeq2 analyses indicated that just one copepod genus (Undinula) was significantly more abundant in cell samples compared to environmental samples, while other copepod genera (Acartia and Candacia) are more abundant in the environment. Similarly, marine fungi were significantly more abundant in cell samples compared to environmental samples, likely because they were present in the detritus consumed by radiolarians. These results indicate that the large proportion of opisthokont ASVs in our cell samples are more likely prey than eDNA because there would not likely be lineage selectivity in the opisthokonts associated with polycystines compared to environmental opisthokonts otherwise. However, further evidence is needed to confirm the prey preferences of Polycystinea.

Diatoms and MALV Group IV were among the most significantly abundant lineages in environmental samples compared to cell samples (Fig. 6). The former is interesting because one could argue that diatoms should be just as capable of becoming photosymbionts as any other lineage we identified; they are, after all, symbionts to Foraminifera (Lee, 2006). We, however, did not detect them in our cell samples. While polycystines appear to associate with a significant diversity of phytoplankton, this finding might support that there nevertheless exists some symbiotic preference for dinoflagellates, chlorophytes, haptophytes and cyanobacteria compared to other algal lineages. The latter (MALV Group IV) is interesting because others have identified radiolarians as a major reservoir to MALVs (Guillou et al., 2008; Bråte et al., 2012); however, our results do not support this finding. We speculate that this discrepancy might just represent the dynamic nature of the polycystine holobiont, in that holobiont constituents can vary significantly, as evidenced by the differences in the phytoplankton lineages present in our single-cell datasets compared to those present in the radiolarian, single-cell datasets of Nakamura et al. (2023) mentioned above.

The findings that the environmental phytoplankton community is distinct from the photoautotrophic holobiont communities of sampled polycystines, and that the relative abundance of possible photosymbiont sequences is significantly greater in cell samples compared to environmental samples, similarly indicates selectivity in terms of the phytoplankton lineages that associate with polycystines. While it is possible that these lineages are prey to polycystines like the metazoans likely are, many of these phytoplankton lineages are known photosymbionts to at least one radiolarian species, including Brandtodinium, Gymnodinium, Tetraselmis, Gymnoxanthella, Synechococcus and Phaeocystis or to other organisms, e.g. Chlorella, Symbiodinium and Tetraselmis (Baker, 2003; Kodama and Fujishima, 2010; Ishitani et al., 2014; Thomas et al., 2024). Still, metabarcode data is only one line of evidence indicating associations between the lineages listed here; therefore, fluorescence microscopy and cell staining to determine the intracellular presence and nature of the relationships between presented lineages is necessary to confirm the identities of the photosymbionts that we hypothesize in this study.

ACKNOWLEDGEMENTS

The authors would like to acknowledge Hannah Gossner for help during sample collection, in addition to Serena Aguilar and Daniel Jasso-Selles for help with benchwork.

FUNDING

This work was supported by the Samuel Riker Foundation through BIOS Grant-in-Aid program (to NLC), the Simons Foundation International’s BIOS-SCOPE program (NLC and LBB), the NSF (OCE-2227766 to LBB) and the School of Life Sciences at Arizona State University.

AUTHOR CONTRIBUTIONS

NLC and GHG conceived of the project. NLC and LBB performed sample collection. NLC performed benchwork and analyses. GHG and LBB advised and contributed to analyses. NLC wrote the manuscript. GHG and LBB contributed significantly to manuscript drafting. NLC, LBB, and GHG all approved of the final, submitted manuscript.

DATA AVAILABILITY

The data generated in this article are available in the GenBank nucleotide database under accession number ranges PQ394142-PQ394148 and in the GenBank short read archive database under BioProject PRJNA1165325.

References