2.1 Congener collection and cultivation

Schmidingerella arcuata was collected from surface waters in Long Island Sound, USA (41.31°N, 72.06°W) and identified via comparisons to the rRNA gene sequences attributed to the species S. arcuata on GenBank. Schmidingerella meunieri was isolated from surface waters in Puget Sound, WA (cultures provided by Dr Suzanne L. Strom and Kelley Bright, Western Washington University), and is only operationally referred to as S. meunieri in this study; although no published rRNA gene sequences were available for S. meunieri for comparison, an isolate from the same culture was studied and identified as S. meunieri by Gruber et al. (2019). Individuals were isolated using a drawn glass pipette into six-well culture plates. Clonal cultures were grown in autoclaved seawater filtered through a 0.2-μm membrane and maintained at 18°C on a 12-h:12-h light:dark cycle at ~50 μmol photons m^–1 s^–1. All cultures were fed saturating concentrations (>3 × 10³ cells ml^–1) of the dinoflagellate Heterocapsa triquetra and the prymnesiophyte Isochrysis galbana (strain TISO), which were grown in F/2 medium (Guillard, 1975) under the same light and temperature conditions as the ciliates.

2.2 Lorica morphology

Specimens were observed with a compound microscope (Olympus BX50; 400–1000×) under bright-field and differential interference contrast illumination, either live or fixed with nonacid Lugol’s solution. Specimens were observed and measured before and after fixation, but no measurable effects of preservation on lorica dimensions were seen. Lorica measurements included total length, lorica aperture diameter, maximum diameter and posterior process (pedicel) length (Table 1). Measurements were obtained from individuals (S. arcuata, n = 163; S. meunieri, n = 83) that were in an exponential growth phase. Morphometric data and micrographs were collected using nis-elements advanced research imaging software version AR-3.00 (Nikon). Morphometric differences between S. arcuata and S. meunieri were evaluated via a t test (α = 0.05), box and whisker plots and principal component analyses using the R environment (R Core Team, 2013).

2.3 Growth and feeding experiments

All experiments were conducted in six-well plates, each well containing 10 ml of autoclaved, 0.2-μm-filtered seawater. Prior to the start of all experiments, 60 individuals were picked from culture, rinsed twice in 0.2-μm-filtered seawater and acclimated for 24 h under experimental conditions (containing the algal prey species intended for experiments) to avoid carryover from culture conditions. Following acclimation, 15 individuals were picked from the acclimation well, rinsed five time in 0.2-μm-filtered seawater, and added to each experimental well containing a known concentration of algal prey. Growth was evaluated with 11 different algal prey, including two potentially toxic dinoflagellates (Table 2). All experiments were done in triplicate. Experiments lasted 3 days and were completed simultaneously for S. arcuata and S. meunieri. At the end of the experiment, the ciliates were briefly observed live and then fixed with 2% nonacid Lugol’s iodine. Ciliates and prey were counted using an inverted microscope (Olympus IX70; 400×); treatments were scored for positive net growth, very low or no net growth, and complete mortality.

2.4 Numerical and functional responses

Numerical and functional response curves (growth or ingestion, respectively, vs. food concentration) were generated for two of the algal foods that supported net positive growth, the dinoflagellate Heterocapsa triquetra and the cryptophyte Rhodomonas salina. Experiments were done in six-well plates, as above, but with varying food concentrations, and no-ciliate controls to account for algal growth (Schoener & McManus, 2017). Ciliates were first acclimated for 24 h under experimental conditions, and then 15 ciliates were added to triplicate wells containing different food concentrations. Growth and ingestion were estimated from changes in ciliate and food abundance after 3 days. Immediately before and after the experimental incubation period, algal cell counts were completed via a haemocytometer using a compound microscope (Olympus BX50; 400–1000×) or a Coulter Particle Analyzer (Beckman Coulter), and concentrations ranged from 10² to 10⁵ cells ml⁻¹ (3.5 × 10³–5 × 10³ ng C ml⁻¹). Algal prey concentration was converted to carbon (C, in ng per cell) based on volume to carbon relationships. To accomplish this, we first estimated the biovolume of each algal species separately by measuring their size in reference to the taxon-specific geometric shapes and corresponding biovolume calculations provided in Hillebrand et al. (1999), and then we used the volume to carbon conversion (pg C per cell = 0.216 × biovolume^0.939) of Menden-Deuer and Lessard (2000).

2.5 DNA barcode sequencing and distance analysis

Single-cell sequencing (n = 4 and 6 for S. arcuata and S. meunieri, respectively) of PCR products for SSU, large subunit (LSU) and ITS rRNA genes was completed as detailed in Santoferrara et al. (2013, 2015). In short, single individuals were sequentially transferred into autoclaved seawater and then into 0.2-ml tubes with 20 μl of DNA buffer (1% SDS, 0.1 m EDTA at pH 8), incubated with 1 μl of Proteinase K (20 mg ml^–1) for 12 h at 55°C, and extracted using the DNA Clean & Concentrator-5 kit (ZymoResearch). The SSU rRNA gene was amplified with universal eukaryotic primers (Medlin et al., 1988) and the D1–D2 region of the LSU rRNA gene and 5.8S-ITS were amplified with primers reported by Ortman (2008) and Wang et al. (2014), respectively. PCR products were purified using ExoSAP-IT PCR Product Cleanup Reagent with incubation at 37°C for 15 min and 80°C for 15 min. Sanger sequencing of products was completed at the Keck DNA Analysis Facility, Yale University. Sequences were quality-checked and assembled in mega version 6 (Tamura et al., 2013), which was also used to align and estimate genetic distances within and between the two isolates for each marker. Sequences were added to curated alignments of tintinnid SSU, ITS and LSU gene sequences (Santoferrara et al., 2017), and the LSU sequences were aligned with mafft version 7 using default parameters (Katoh & Standley, 2013) for neighbour-joining inference as implemented in mega version 6 (1000 bootstraps).

2.6 Single-cell ‘omics

Whole-transcriptome amplification (WTA) was achieved using the SMART-Seq2 v4 Ultra Low input RNA kit (Cat: 634889; Takara) following the manufacturer’s protocols, with the exception that one-quarter of the reaction volume was used. The Repli-g single-cell kit (Cat: 150343; Qiagen) was used for whole-genome amplification (WGA) of both isolates following the manufacturer’s protocols. The Qubit dsDNA assay (Invitrogen) was used to quantify the concentrations of cDNA and genomic DNA (gDNA) from WTA and WGA, respectively. PCR was used to test cDNA for eukaryotic SSU rRNA (Medlin et al., 1988) and genus-specific ITS (Costas et al., 2007); universal 16S rRNA primers (Lane, 1991) were used to test for prokaryotic contamination. Libraries were prepared with the Illumina Nextera XT kit (Cat: FC1311096; Illumina) then sequenced with an Illumina HiSeq 2500 at Macrogen Sequencing.

2.7 Transcriptome and genome assembly

Raw WTA and WGA Illumina sequencing data were trimmed for quality and size (Q28 and minimum length of 200 and 1500 bp, respectively) using bbmap (version 38.39; Bushnell, 2014). WTAs were assembled using rnaspades (version 3.13.1; Bankevich et al., 2012), and WGAs were assembled using megahit (version 1.2.9; Li et al., 2015). Assemblies were processed through custom python scripts (https://github.com/maurerax/HTS-Processing-PhyloGenPipeline.git) for the removal of rRNA gene sequences and prokaryotic transcripts (Cerón-Romero et al., 2019; Maurer-Alcalá et al., 2018).

2.8 Genome architecture analysis

As the Repli-g single-cell kit (for WGA) preferentially amplifies long sequences, those data are expected to represent the large scaffolds (>1500 bp) of the MIC genome, while WTA transcripts (>200 bp) serve as a proxy for the gene-sized MAC chromosomes (Maurer-Alcalá et al., 2018). WTA and WGA assemblies were analysed to reveal patterns of genome architecture through custom python scripts, as detailed previously in Smith et al. (2020). In brief, WGA sequences were identified as putative micronuclear loci for macronuclear regions by mapping the cDNA from WTA transcripts to WGA scaffolds, using thresholds requiring 60% of the transcript length and a 97% identity. Micronuclear loci were sorted into architectural categories of “scrambled” and “unscrambled,” based on the arrangement and order between the micronuclear loci and macronuclear regions. MIC regions where MDS are separated by IES, called MDS–IES boundaries, were identified through the analysis of MAC–MIC alignments and the existence of tandem sequence repeats (pointer regions) (Figure 1b).

To verify that differential patterns of genome architecture (i.e., MIC–MAC alignments) were not artificially produced by variations in the MIC and MAC assemblies or due to sequencing errors, the full assembled MAC region of each MAC–MIC alignment was mapped to the raw reads of both the WGA (MIC) and WTA (MAC) of both ciliates. Reads were mapped using bbmap (version 38.39; Bushnell, 2014) and visualized using geneious prime (version 2021.1.1). Only those patterns with high read support (10,000–600,000 raw reads mapped) were retained for further analysis.

2.9 Architectural barcodes

Isolate-specific primers were designed using MIC sequences that surround MDS–IES boundaries in differential MAC–MIC alignments. Primers were tested for specificity using single-cell PCR of individuals from both isolates to verify that cross amplification of the other isolate did not occur. The sensitivity of each primer set was evaluated by adding one cell from culture to a freshly collected plankton sample concentrated on a filter. Further, a primer that targets the LSU rRNA gene of both isolates but avoids the amplification of other ciliates was designed to verify absence when isolate-specific primers showed negative PCR results. Primer details are given in Table S2. The primers were used to test environmental samples to investigate co-occurrence of the isolates. To test for presence/absence of the isolates over time, biweekly water samples were collected for 18 months (January 2016 to June 2017) from surface water in the Long Island Sound estuary (41.31°N, 72.06°W). In brief, plankton from 20 L of seawater were collected on 5-μm pore size glass fibre filters, then stored in lysis buffer at −20°C. DNA was extracted from filters using The Quick-DNA Fecal/Soil Microbe Kit (Zymo Research), and amplified through PCR using isolate-specific primers (Table S2), and products were visualized via gel electrophoresis. To ensure that only the intended DNA target was amplified during troubleshooting, positive PCR products were purified using exosap-it (Thermo Scientific), Sanger sequenced and aligned.

3.1 Morphology and rRNA gene barcodes

Schmidingerella arcuata and Schmidingerella meunieri have similar lorica morphologies and dimensions (Figures 2 and 3). The loricae of both ciliates are hyaline and present a conspicuous suboral bulge, indicating that they are congeners under the genus Schmidingerella (Agatha & Strüder-Kypke, 2012). In culture, both exhibit protoloricae and spiralled (coxlielliform) paraloricae, and form resting cysts with an emergence pore and plug, features previously observed in this genus (Meunier, 1919) and which are included in its diagnosis (Agatha & Strüder-Kypke, 2012). Despite their overall similar size, S. arcuata and S. meunieri differ significantly in oral diameter (S. arcuata: 63.71 μm; S. meunieri: 78.34 μm; t test, p < .0001) and total length (S. arcuata: 137.81 μm; S. meunieri: 188.73 μm; p < .0001; Table 1, Figures 3 and 4), but are not significantly different in maximum diameter or posterior process length. In a principal component analysis (PCA), the first factor was robust, with an eigenvalue of 3.41, and it accounted for 85.31% of the variance in the data (Figure 5). Factor 2 had an eigenvalue of 0.33 and it accounted for 8.3% of the variance. The eigenvalues for factors 3 and 4 were 0.20 and 0.06 respectively, together accounting for a further 6.42% of the variance. The SSU rRNA gene and 5.8S-ITS sequences are identical between S. arcuata and S. meunieri and the LSU rRNA gene differs by 0.5% (three substitutions out of 657 bp characterized; Figure 6).

3.2 Ecophysiology experiments

Of 11 algae tested as single prey, five supported growth in S. arcuata and four supported growth in S. meunieri (Table 2). Only the cryptophyte Rhodomonas salina and the dinoflagellates Heterocapsa triquetra and Scrippsiella sp. supported growth in both ciliates (Table 2). In contrast, both ciliates experienced zero or net negative growth under the dinoflagellates Prorocentrum micans, P. minimum, Amphidinium carterae, Alexandrium minutum and the potentially toxic dinoflagellate Alexandrium pacificum (Table 2). Treatments that yielded contrasting growth results between ciliate species (including net positive vs. negative/zero) include Tetraselmis sp., Isochrysis galbana and Alexandrium fundyense. The prey Tetraselmis sp. resulted in net positive growth in S. meunieri and total mortality in S. arcuata. Conversely, I. galbana enabled net positive growth in S. arcuata at medium to high concentrations (300–3000 ng C ml⁻¹), while S. meunieri was only capable of positive growth at very high prey concentrations (3000 ng C ml⁻¹), suggesting that I. galbana is of low nutritional value to S. meunieri. In addition, the potentially toxic dinoflagellate A. fundyense enabled positive growth in S. arcuata, while S. meunieri experienced 100% mortality within the 24-h acclimation period, suggesting that mortality was a result of toxin content vs. starvation or nutritional deficiency. Additionally, backwards swimming was observed in S. meunieri during the acclimation period with A. fundyense.

Numerical and functional response data revealed both significant (H. triquetra treatment) and nonsignificant (R. salina treatment) differences between the ciliates (Figure 7, Tables 3 and 4). When H. triquetra was used as prey, the maximum growth and the half-saturation level were significantly higher (μ_max: z = 3.57, p < .0001; k_m: z = 1.97, p = .0250, 95% confidence interval [CI]) in S. meunieri than in S. arcuata. Similarly, the maximum ingestion rate (ng C day⁻¹) was significantly higher in S. meunieri (I_max: z = 7.88; p < .0001; 95% CI), although the half-saturation parameter k_m was only significantly different from S. arcuata at a α = 0.10 (k_m: z = 1.80; p = .0670; 90% CI) (Table 5). Conversely, the R. salina feeding experiment yielded no significant differences in the growth response (μ_max: z = 0.75, p = .4530; k_m: z = 1.52, p = .8790) or ingestion response (I_max: z = 1.258, p < .2080; k_m: z = 0.959, p = .3380) between ciliates (Table 4).

3.3 Genome architecture

The assemblies of both the micronuclear genome (MIC) and the macronuclear transcriptome (a proxy for the macronuclear genome [MAC]) were much larger for S. arcuata (MIC: 49 Mb; MAC: 6.3 Mb) than for S. meunieri (MIC: 19 Mb; MAC: 1.8 Mb), probably a reflection of differential sequencing depth (Table 5). Of the 11,673 transcripts for S. arcuata, about 15% could be mapped to the MIC assembly, while in S. meunieri about 8% of the 7476 transcripts were mapped to the MIC. Comparing the micronuclear-mapped transcripts from S. arcuata (1718) and S. meunieri (598), we found some differences in MDS and IES length, as well as pointer region %GC content, although these differences may partially reflect shallower sequencing depth in S. meunieri (Table 5). Of 350 micronuclear-mapped transcripts found in both ciliates, we present five examples that match at 98.1%–99.2% sequence identity in the transcriptome (MAC proxy) of both ciliates (Figure 8; Table S1) but exhibit substantially different architecture in the MIC of each species (Figure 8). We found no evidence for alternative processing (more than one MAC sequence resulting from a single MIC region; Katz and Kovner, 2010).

3.4 Architectural barcodes

To evaluate the presence of one or both species in surface waters of the Long Island Sound estuary (41.31°N, 72.06°W), we used five species-specific primer sets to analyse DNA extracted from filtered plankton samples; this includes two sets specific to the amplification of each ciliate, plus a single “Genus” primer that amplifies both ciliates (Table S2). Positive PCR gel bands (confirmed with Sanger sequencing) indicated the presence of S. arcuata on five of the 23 sampling dates, and the presence of S. meunieri on three (Table 6). On only one sampling date (June 6, 2016; 18.4°C) were both ciliates present. S. arcuata was identified in samples that ranged in temperatures from 14.1 to 18.4°C, while S. meunieri was identified in samples from 8.2 to 18.4°C.

4.1 Genome architecture inferences

Previous studies suggest that genomic organizational patterns may serve as mechanisms for rapid reproductive incompatibility between ciliates (Collens & Katz, 2021; Gao et al., 2014, 2015; Nowacki et al., 2011; Weiner & Katz, 2021). We consider this concept in our analyses of the micronuclear genome and macronuclear transcriptome of S. arcuata and S. meunieri, which reveal different patterns of micronuclear organization, including the gene regions presented in Figure 8, which are >98% identical in the MAC of both ciliates (Table S1). We suggest that the differences in micronuclear architecture found between S. arcuata and S. meunieri, namely the differential arrangement of protein-coding segments within the germline nucleus of these ciliates, indicate that they may be incapable of producing viable post-conjugation macronuclei. This implies that S. arcuata and S. meunieri may be reproductively isolated, thus representing separate biological species under established species concepts (Mayr, 1957, 1970). The ecophysiological differences in growth and food preferences somewhat support this from the functional niche separation point of view (Weisse, 2017).

A possible mechanism whereby minor variations in genome rearrangement processes can provide opportunities for rapid diversification and a path to speciation is discussed by Gao et al. (2015). In the ciliate Chilodonella uncinata, gene duplication coupled with loss of macronuclear-destined segments has led to germline loci that require precise processing to generate functional genes (Maurer-Alcalá et al., 2018). As duplicated genes in the MIC diverge over time, duplicates in different populations may diverge enough to result in incompatible genotypes (Gao et al., 2015). If the divergence arises in a subdivided population, crosses between members would be incapable of generating the same functional gene family members, resulting in a reproductive barrier that, if sustained, could cause rapid but restricted genetic differentiation via reproductive isolation, leading to the speciation of lineages (Gao et al., 2015). Differentiation via these small changes in genome architecture could be expected to occur more rapidly than that caused by stochastic mutations (Cabej, 2013; Chalker et al., 2013).

Further, as morphological differentiation in incipient species is considered an evolutionary character that already lags in comparison to mutational molecular divergence (Derycke et al., 2016; Struck et al., 2018), it is possible that genome architectural incompatibility may be a mechanism that allows for rapid genetic incompatibility between populations without detectable morphological differences (i.e., cryptic species). This may help explain the apparent disconnect between degrees of morphological and molecular diversity in ciliates, particularly in cryptic taxa such as S. arcuata and S. meunieri, and might also explain why evolutionary divergence between ciliate individuals may not scale according to observable morphological differences (Santoferrara, Grattepanche, et al., 2016). Additionally, if barriers to gene flow are impermanent and hybridization opportunities fluctuate, this process may represent a mechanism for speciation in sympatry, when gene flow is present (Caron & Hutchins, 2013; Struck et al., 2018).

4.2 Ecophysiological differences

Morphological differences between S. arcuata and S. meunieri (Figure 3, Table 1) may have ecological relevance. In tintinnids, the oral diameter of the lorica aperture is considered to be a significant character controlling ingestion capability, including the long-standing observation that tintinnids select prey that are no larger than ~30% of their oral diameter (Dolan, 2010; Montagnes et al., 2008). This appears to be at least somewhat supported in this study as the major differences between (nontoxic) prey responses of S. arcuata and S. meunieri were in algal prey of substantially different size (the 4-μm-diameter I. galbana supported growth in the smaller S. arcuata; the 10-μm-diameter Tetraselmis sp. supported growth in the larger S. meunieri).

Numerical and functional response curves (i.e., growth and ingestion rates) represent traits that determine the effect of a species on ecosystem function; they may indicate a functional distance or differential niche occupancy between similar taxa (Lavorel et al., 2013; Weisse, 2017). For example, the growth rates of some ciliates can equal or exceed those of their algal prey, suggesting that these ciliates can exploit or control sudden prey blooms, influencing phytoplankton dynamics and trophic transfer (Banse, 1982; Montagnes, 2013; Montagnes & Lessard, 1999). In the present study, the significantly higher growth and ingestion rates of S. meunieri vs. S. arcuata on identical prey (H. triquetra) suggests a selective advantage for the former, which may allow for the proliferation of separate niches based on the availability of specific prey. Similarly, the ability for S. arcuata to maintain positive growth in lower prey concentrations than S. meunieri for some prey (I. galbana) may provide an advantage during times of reduced prey availability. Further, (nontoxic) prey treatments where positive growth was observed in only one ciliate (e.g., Tetraselmis sp.) suggest a discrepancy in nutritional requirements that may translate to a difference in ecological trait space (Weisse, 2017). In the A. fundyense treatment, S. arcuata was able to maintain positive growth throughout the entire experiment incubation period, while S. meunieri exhibited mortality within the 24-hr prey acclimation stage (Table 2). This may indicate a differential susceptibility to the potentially toxic A. fundyense, rather than a difference in nutritional requirements, as S. meunieri cultures can regularly survive 24 h without food. Ciliate susceptibility to toxic algae has been reported in the literature, particularly between dinoflagellates such as A. fundyense and tintinnid congeners under the genus Favella (Fulco, 2007; Hansen, 1989; Kamiyama et al., 2006; Rosetta & McManus, 2003) some of which are now have been redesignatead under the genus Schmidingerella (Agatha & Strüder-Kypke, 2012). Studies on Favella ehrenbergii and F. taraikaensis also reflect our observation of reverse-swimming direction for S. meunieri prior to mortality, further indicating a toxic effect (Hansen, 1989; Kamiyama & Arima, 1997). Overall, these ecophysiological data suggest that S. arcuata and S. meunieri vary in some ecological traits that define their effect on ecosystem functioning (Weisse, 2017), an indication that they may occupy separate ecological niches.

4.3 Environmental sampling and cryptic co-occurrence

Isolate-specific primers showed that S. arcuata and S. meunieri were found in the same sampling area throughout the year but on only one day were both ciliates present (June 7, 2016; 18.4°C). The presence of both species within the same water column at mostly different sampling dates may suggest that the populations are constrained by seasonal or environmental variables that restrict either species’ range. While S. arcuata was identified over a broader range of the year, in both summer (June–July) and autumn (October–November) sampling dates, S. meunieri was found in only three samples from late spring to early summer (April–July), but the dates in which S. meunieri was found encompass a broader range of water temperatures (S. arcuata: 1.69 SD; S. meunieri: 4.28 SD). In samples where the species do not co-occur but are still present within the same sampling month or close temperature range, the presence of either ciliate may reflect differences in predator–prey interactions and prey bloom dynamics, which can fluctuate rapidly in coastal environments (Rekik et al., 2021).

The co-occurrence of both ciliates in the same community raises questions about niche separation and small-scale isolation events. Cyclical shifts in water temperature, light, salinity and nutrient availability create dynamic barriers to gene flow that may allow for isolation and periodic or episodic co-occurrence on varied timescales (Hohenlohe, 2004). For example, members of the tintinnid genus Helicostomella were found to have cryptic forms that only co-occurred during post-bloom periods in spring and early summer (Santoferrara et al., 2015). In other studies, the divergence of ciliate sister species has been directly attributed to changing salinity (Stock et al., 2013) or pH regimes (Weisse et al., 2013), both of which are fluctuating ocean variables that control species ranges and restrict population distributions. When seasonal or other fluctuations in these variables cause species ranges to either overlap or isolate, they can act as impermanent barriers of reproductive isolation, as populations are separated and rejoined episodically, without time to hybridize, differentiate or outcompete (Weiner et al., 2012; Godhe et al., 2016). In combination with rapid genetic incompatibility, possibly driven by changes in genome architecture (explored above), these lines of isolation impermanence may act as loose barriers to gene flow that allow the rapid molecular speciation of closely related ciliates to coexist in reproductive isolation with little morphological differentiation. This may help to explain how so many morphologically and ecologically indistinguishable ciliate species are able to co-occur despite the competitive exclusion principle (Hardin, 1960).

4.4 Tintinnid diversity

Due to their centuries-long history of classification, tintinnid ciliates have been useful models to develop concepts regarding patterns of plankton biogeography. Prior to sequence repositories, these patterns were based solely on morphology-based identification and helped lead to assumptions regarding the cosmopolitan nature of microplankton (Dolan & Pierce, 2013; Fenchel et al., 1997). However, as DNA sequences are now available for most tintinnid families and common genera with existing morphological descriptions, patterns have emerged that instead suggest some level of endemism and dispersal limitation (Ganser et al., 2021). Today, global tintinnid SSU rRNA data reveal distribution patterns related to salinity, climate, bathymetry and latitude (Santoferrara et al., 2018; Santoferrara & Mcmanus, 2020). As these shifting perceptions appear to be largely driven by the increased taxonomic resolution in sampling data (Ganser et al., 2021), the ability to discriminate cryptic species such as S. arcuata and S. meunieri is critical for revealing these biogeographical patterns.