Environmental samples analyzed for Mesodinium and cryptophyte diversity were opportunistically gathered from various sources. In most cases (BR, NC, Bar-M4, GF-LL3a, GF-XVI, TV), samples were collected from surface water preserved using 5–10% Lugol’s fixative. However, samples were also collected onto a 0.2 μm Sterivex^TM filter (EMD Millipore, Billerica, MA USA) (CL) or centrifuged to form a cell pellet (LIS), and kept frozen until analysis. Previous research has shown that Lugol’s preserved material is sufficient for DNA extraction and quantitative (q) PCR assays, and while sensitivity of the assays decreased overtime, positive amplification was still possible for several months of sample storage at room temperature (Bowers et al., 2000). The sample from SGI, however, was unpreserved and shipping of the sample was slow due to the geographical isolation of the collection site. Upon the arrival of the SGI sample it was immediately centrifuged and the pellet was frozen until DNA was extracted.

Cryptophyte and Mesodinium spp. cultures were grown in the lab for DNA extraction and sequencing of rbcL or rRNA gene fragments for phylogenetic comparison with our environmental data. Cryptophyte cultures including Teleaulax acuta (SCCAP K-1486), T. amphioxeia (CCMP 1170, GCEPO1), Geminigera cryophila (CCMP 2564), Chroomonas sp. (CCMP 270), Hemiselmis pacific (CCMP 706), H. andersenii (CCMP 439), H. rufescens (CCMP 440), Hemiselmis sp. (SUR21-C3), Hemiselmis sp. (NR11), and Hanusia phi (CCMP 325) were grown in the lab in F/2-Si at 18°C on a 14:10 L:D cycle, as were M. rubrum cultures AND-A0711 and CBJR05. Additionally, the M. rubrum cultures were maintained with GCEPO1, by feeding the ciliate weekly with a 1–5% volume addition of log-phase cryptophyte culture. A Falcomonas sp. culture (CCMP 2293) was also grown on F/2 without added Si (F/2-Si), but at 4°C and constant light. All of the above cryptophyte and Mesodinium cultures were harvested for DNA extraction by centrifugation of 10 ml of dense culture within a 15 ml Falcon tube at 4000 RPM for 15 min, the supernatant decanted, and the cell pellet frozen until processed. Cultures of Teleaulax gracilis (Cr6EHU), T. minuta (Cr8EHU), T. cf. merimbula (Cr59EHU), aff. Teleaulax (Cr22EHU), Plagioselmis nannoplanctica (Cr50EHU), P. cf. prolonga (Cr10EHU, Cr127EHU, Cr143EHU, Cr194EHU) and Urgorri complanatus (Cr1EHU) were grown in the same light and temperature conditions described above, on a modified F/2-Si (enriched with soil extract and selenium at 0.006 mM Na₂SeO₃ final concentration) at a salinity of 30 (P. nannoplanctica strain grown in freshwater), and were preserved in ethanol prior to DNA extraction and analysis of the rbcL gene. A live sample of M. rubrum culture MR-INO200702 and a 5% Lugol’s-preserved sample of culture MR-MAL01 were also acquired for analysis. For all cultures, approximately 10–20 ml of dense stationary-phase culture was collected by centrifugation (4000 RPM, 10 min) and frozen until extraction.

Nucleic acids of environmental samples were extracted from frozen cells collected by centrifugation (4000 RPM, 10 min) or on Sterivex^TM filters (EMD Millipore) using either the PowerWater Sterivex DNA Isolation Kit (MO BIO Laboratories, Inc.) or a hot detergent lysis method as described by Gast et al. (2004), modified to exclude zirconia-silica bead disruption. All cultures as well as the North Carolina Bloom sample were extracted using the Qiagen DNeasy Plant Mini Kit. PCR was conducted using GoTaq (Promega) or GoTaq G2 Hot Start mix in 50 mL reactions, with a final concentration of 2.5 mM MgCl₂, 200 μM dNTPs, 2.5 U GoTaq Flexi polymerase, and 0.1 μM primers for normal and 0.2 μM for hot start. Primers for Mesodinium spp. (Table 1) were designed to amplify the majority of the SSU and LSU rRNA genes, and the entire ITS region, resulting in a ∼1880 bp amplicon. Gene fragments of the large subunit of the ribulose-1,5-bisphosphate carboxylase oxygenase gene (rbcL) were PCR amplified from DNA extracts using primers targeting cryptophyte plastids. A new primer, crypt_rbcLR2 (5′-CAGTGAATACCACCTGAAGCTA-3′), designed using an alignment of cryptophyte rbcL sequences (See Supplementary Data Sheet 1 for accession numbers) was used in combination with L2F (Hoef-Emden et al., 2005). PCR conditions were: 95°C for 5 min followed by 40 cycles of 95°C for 60 s, 55°C for 60 s, and 72°C for 90 s followed by 72°C for 7 min. The genus-specific primers MESO_245F and MESO_28S_R were used to amplify a combined fragment of the Mesodinium spp. 18S-ITS-28S genes. PCR conditions were: 95°C for 5 min followed by 40 cycles of 95°C for 60 s, 57°C for 60 s, and 72°C for 90 s followed by 72°C for 7 min. PCR products were visualized by agarose gel electrophoresis and later excised and purified from the gels using the Zymoclean Gel DNA Recovery Kit (Zymo Research). Clone libraries were constructed from gel purified fragments using the pGEM-T Easy Vector in the pGEM-T Easy Vector System II cloning kit (Promega Corporation) according to the manufacturer’s protocol. Selected clones were submitted for Sanger sequencing with a single primer to either Beckman Coulter Genomics (Single Pass Sequencing) or the W. M. Keck Ecological and Evolutionary Genetics Facility at the Marine Biological Laboratory (Woods Hole). Full-length Sanger sequencing of select Mesodinium clones were run at Genewiz (Boston) (see Supplementary Data Sheet 1 for accession numbers). Sequences were edited and assembled into contigs using Sequencher (Gene Codes Corporation). We sequenced and analyzed 687 cryptophyte rbcL clones (accession numbers in progress) from environmental samples. For Mesodinium spp., a total of 903 clones were sequenced (accession numbers in progress) from all stations. Using a sequence similarity criterion of 98 and 99% for cryptophytes and Mesodinium, respectively, we constructed independent contigs and generated consensus sequences for use in our global alignment for each sample. This process helped to reduce the number of sequences used in our phylogenetic analyses while maintaining meaningful diversity data.

Consensus sequences from assembled contigs were used for separate alignments of Mesodinium spp. and cryptophytes, in combination with available sequences in Genbank and sequences generated from cultures in our laboratory (see above). Alignments were constructed using the Clustal X algorithm (Larkin et al., 2007), and refined by eye using MacClade 4.08a (Maddison and Maddison, 2000). All maximum likelihood (ML) phylogenetic trees were executed with PhyML 3.0 (Guindon et al., 2010) using 100 bootstrap replicates and the general time reversible (GTR) substitution model, while estimating the gamma distribution parameter, proportion of invariable sites, and the transition/transversion ratio. Phylogenetic trees were constructed using TreeDyn 198.3 (Chevenet et al., 2006). Assignment of Mesodinium and cryptophyte species was determined by their nearest neighbor match of a known species within ML phylogenetic trees.

The most widely encountered variant was D (M. major), found at 6 of the 10 sampling sites, while variants A and F were found at 5 (Figure 2). Sequences from variants C and H, as well as the heterotrophic M. pulex were each only recovered from one station, while no clade E sequences were found in this study. Analysis of samples from red-water events of Mesodinium revealed that the bloom phenomenon within this species complex is not associated with one particular variant. Sequences from subclades B and D dominated (>85%) clones recovered from the four blooms analyzed in this study, and in two cases sequences consisted only of M. major (Figure 2). Variant B dominated two blooms from the eastern United States, and was not found at non-bloom sites (Figure 2). Clones from variants F and H dominated the bloom sample from SGI, however, the sample was not preserved when shipped and thus the relative proportion of M. major/rubrum variants is uncertain. Samples from the Baltic Sea were from coastal and open water stations, and were not associated with bloom events. Collectively, the Baltic Sea had the greatest diversity of Mesodiniidae sequences, with clones recovered from heterotrophic M. pupula, the mixotrophic M. chameleon, as well as from multiple M. rubrum/major clades. The CC sample was also diverse, with three variants of M. major/rubrum as well as M. chamaeleon (Figure 2). In contrast, a non-bloom sample from an off shore station in the Barents Sea was composed entirely of M. major.

Mesodinium spp. have an unusual 18S rRNA gene with numerous deletions and substitutions within conserved regions compared to other ciliates (Johnson et al., 2004), and they appear to lack a separate 5.8S rDNA region, suggesting that it may be fused to the 28S rDNA gene (Herfort et al., 2011b). This characteristic is also found in the microsporidia and bacteria (Hoef-Emden, 2005). While Mesodinium show typical litostome secondary structure in their V4 region, including reduction of helices 23_1, 23_8, and 23_9 and the absence of 23_5 (Strüder-Kypke et al., 2006), phylogenies based on the SSU rRNA gene consistently result in a novel, deep basal branch within the ciliates (Johnson et al., 2004; Chen et al., 2015a). From ultrastructural observations, only the ciliary transition region supports their placement within the litostomes (Garcia-Cuetos et al., 2012). A recent multigene phylogeny of the Mesodiniidae also suggests a deep basal ciliate branch and strongly implies that they may represent a distinct ciliate class, the Mesodiniea (Gao et al., 2016). However, a phylogenomic analysis of non-model ciliates using transcriptomic data failed to resolve the monophyly of the Mesodiniea (Chen et al., 2015b).

Our phylogenetic analysis of taxa within the Mesodiniidae agree with Garcia-Cuetos et al. (2012), with strong support for four major clades distinguishing the two heterotrophic lineages of M. pulex and M. pupula, the mixotrophic M. chamaeleon, and the predominantly phototrophic M. major/rubrum complex. At least one additional major clade of Mesodinium-like ciliates are found in freshwater habitats, and phylogenies based on the SSU rRNA gene group them as a sister group to M. pulex (Bass et al., 2009). Since we did not sequence samples from freshwater habitats and no cultures of these species were available, comparisons of their full rRNA gene cassette to marine Mesodiniidae were not made here. Our analysis resolved two novel subclades of M. rubrum: (1) variant G was identified from sequence data of a culture from Chesapeake Bay, USA, along with environmental sequences from the North Carolina (NC) and St. George Island (SGI) bloom events, and (2) variant H, which is only represented by environmental sequences from the SGI sample. One sequence from a culture isolated in Spain (Riobó et al., 2013) grouped with a Danish culture from the F subclade (Garcia-Cuetos et al., 2012) as well as clones from the Baltic Sea and SGI. Together, variants F and G grouped strongly with variant A to form one major branch of the M. rubrum/major complex. We found M. rubrum sequences from the Baltic Sea, CC, NC, and SGI that grouped with variant A, which includes a well-studied cultured strain from Antarctica (Gustafson et al., 2000; Johnson et al., 2007) and sequences from coastal Oregon (Herfort et al., 2011b). Cultures isolated from Korea (Yih et al., 2004) and Japan (Nishitani et al., 2008) were found to belong to subclade B, which included sequences from blooms in LIS and NC, as well as the recurrent blooms of the Columbia River Estuary (Herfort et al., 2012). M. major (subclade D) was found in the Baltic Sea, Bar-M4, and in blooms from coastal BR and CL, and has also been found in the Columbia River Estuary (Herfort et al., 2011b) as well as coastal Denmark (Garcia-Cuetos et al., 2012). Two other M. rubrum variants, C and E, remain uncultured and have only been found in the Northeastern Pacific (here and Herfort et al., 2011b). Subclades B, C, D, E, and H formed the second major branch of the M. rubrum/major complex. While C, E, and H have not yet been cultured, subclade B appears to be highly variable in size, both within and between strains (Table 2). A bloom of M. rubrum variant B in LIS was found to contain cells of typical size for M. rubrum cultures (Tables 2 and 3).

The studies of Herfort et al. (2011b) and Garcia-Cuetos et al. (2012) revealed that “M. rubrum” is a species complex of closely related genetic variants, most of which are morphologically indistinguishable. While cell size clearly distinguishes M. major from M. rubrum variants, cell size within M. rubrum strains appears to be highly variable considerably (Table 2). Furthermore, sequence variation of rRNA genes and the ITS region in M. major relative to M. rubrum, is equal to or lower than variation among M. rubrum variants (Table 4), making the distinction of species vs. variant somewhat ambiguous based on these sequences alone. While we are beginning to identify trends in genotype distribution of M. major/rubrum ciliates and their association with bloom events, we know little about distinctions in the physiological or behavioral diversity among these variants and if they may possess distinct ecological niches.

While our results strongly suggest that heterotrophic species have less intracladal diversity in marine ecosystems, our samples focused on bloom events and are therefore biased toward the mixotrophic species. Stations where heterotrophic species were abundant, such as two Gulf of Finland (Baltic Sea) samples rich in M. pupula (>40% of Mesodinium clones), revealed essentially identical sequences within and between these populations. Furthermore, M. pupula sequences from the Baltic Sea were nearly identical to those from SGI. In addition, M. pulex sequences from SGI and M. chamaeleon sequence data from SGI and the Baltic Sea were remarkably similar to strains isolated from Danish waters.

Taxonomic relationships within the cryptophytes, based on traditional morphological and structural traits of their pigments and cell surface, have recently been tested using gene phylogenies (Fitt et al., 2001; Hoef-Emden et al., 2002). These studies have found that while biliprotein type is congruent with molecular phylogenies, the type of inner periplast is not (Hoef-Emden et al., 2002). These phylogenies revealed seven separate lineages within plastid-bearing cryptophytes, all except Cryptomonas, containing representatives from marine environments. An additional lineage, represented by the brackish-water species U. complanatus, was later discovered (Laza-Martínez, 2012). Previously Hoef-Emden et al. (2005) commented on the potential for accelerated evolutionary rates when constructing rbcL phylogenies of cryptophytes, due to shifts in the number of amino acids showing changes in codon usage among early and late diverging taxa. This study was based only on genus Cryptomonas, an exclusively freshwater group containing some highly divergent taxa. Yet, despite finding high evolutionary rates for rbcL, resulting cryptophyte phylogenies were largely congruent with those of 18S rDNA (Hoef-Emden et al., 2005). Our results, which used the most diverse rbcL dataset to date, are also congruent with previously published phylogenies of nuclear 18S rRNA genes (Hoef-Emden et al., 2002). The eight lineages of plastid-bearing cryptophytes were recovered in the rbcL phylogeny. Moreover, the Proteomonas lineage appeared as the sister group of TPG with high support. While major clades are well-supported in previous phylogenies, the relationships among lineages have not been well-resolved. The clustering of Proteomonas and TPG was only partially supported in nucleomorph, but not nuclear, 18S rRNA genes phylogenies in a clade that also included Guillardia/Hanusia (Hoef-Emden et al., 2002).

The traditionally used morphological trait of periplast structure, as a continuous sheet or polygonal, has been shown to vary with life stage in the genera Cryptomonas and Proteomonas and is thus phylogenetically uninformative (Hill and Wetherbee, 1986; Hoef-Emden and Melkonian, 2003). Similar observations of periplast dimorphism have also been made within the Teleaulax group (Garcia-Cuetos, 2011), and this trait has traditionally been used to distinguish this genus from Plagioselmis. Furthermore, the midventral band, a morphological trait used to distinguish Teleaulax (thought to be absent) and Plagioselmis, was recently shown to be polyphyletic within this group (Laza-Martínez et al., 2012). The finding of strains with the T. amphioxeia and P. prolonga morphology sharing nearly identical rbcL sequences can be interpreted from the perspective of a dimorphic species, making both names synonym. The Teleaulax/Plagioselmis dimorphism can also account for incongruences between microscope-based morphological identifications and environmental DNA sequence identities, as in Bazin et al. (2014), where only T. acuta sequences were retrieved from a P. prolonga red tide sample. These observations, combined with the polyphyletic phylogeny of Teleaulax shown here and previously using 18S rRNA (Deane et al., 2002; Rial et al., 2015), suggest that these genera are in need of taxonomic revision.

About half of our samples were from Mesodinium blooms, and these samples (excluding SGI) were dominated (>95%) by T. amphioxeia, as expected from previous studies (Nishitani et al., 2010, #874; Herfort et al., 2011b, #893). However, clone libraries from non-bloom samples were also largely comprised of TPG sequences, particularly T. amphioxeia (Figure 4). Exceptions were the high latitude BS and SGI samples, which were largely composed of T. acuta and G. cryophila-like sequences, respectively. G. cryophila appears to primarily be a cold-water species, dominating cryptophyte assemblages in Antarctic Dry Valley Lakes (Bielewicz et al., 2011), and was co-isolated with an Antarctic M. rubrum strain (Gustafson et al., 2000). T. gracilis and T. minuta have also been shown to sustain in vitro M. rubrum (clade F) growth (Rial et al., 2015). However, only few sequences of the former and none of the latter were recovered. These results strongly suggest that TPG species may frequently dominate cryptophyte communities in coastal ecosystems, and play critical roles in supporting the productivity of M. major/rubrum.

Baltic Sea planktonic communities have been shown to be rich in TPG cryptophytes, and are also known to have Hemiselmis and Rhodomonas spp. (Hill et al., 1992; Carstensen and Heiskanen, 2007; Hajdu et al., 2007). While none of our Baltic Sea samples were associated with a Mesodinium bloom, Mesodinium cell counts were not available and thus it is unclear to what degree their plastids may be contaminating the cryptophyte rbcL community signal. In addition to TPG species, we also found Falcomonas, Hemiselmis, and U. complanatus sequences in the Gulf of Finland. U. companatus was originally described from French, Spanish, and Portuguese estuaries and has a maximum growth rate at a salinity of 10 psu (Laza-Martínez, 2012); thus its presence within these brackish waters is not surprising. A single clone of U. complanatus was also recovered from our BS sample. While the BS was dominated by T. amphioxeia, M. major was found in this sample at ∼10 cells/ml, which would equate to about 400–600 T. amphioxeia plastids ml^-1 based on plastid numbers reported by Garcia-Cuetos et al. (2012). Since our community samples were not size fractioned, in this case our clone libraries may have been highly enriched with plastids from this species since it is the preferred prey of the M. major/rubrum complex (Yih et al., 2004; Park et al., 2007; Myung et al., 2011; Hansen et al., 2012).

Peaks in cryptophyte abundance have been found to both proceed (Johnson et al., 2013) and co-occur (Kim et al., 2007; Weber et al., 2014) with high levels of M. rubrum-like ciliates in coastal ecosystems. However, most historical accounts of Mesodinium blooms did not note the abundance or composition of co-occurring cryptophyte communities. While certain aspects of M. major/rubrum blooms have been well-studied within a variety of ecosystems, such as the Southampton estuary (Crawford et al., 1997), upwelling zones from Ecuador to Chile (Ryther, 1967; Avaria, 1976; Jimenez and Intriago, 1987), Baltic Sea Fjords (Lindholm, 1978; Lindholm and Mörk, 1990; Sjöqvist and Lindholm, 2011), and portions of the Gulf of California (Gárate-Lizárraga et al., 2002; Bulit et al., 2004), only within the Columbia River estuary (Herfort et al., 2011a) have studies focused on cryptophyte-Mesodinium dynamics. Within this system, cryptophytes are high in abundance both prior to and during early M. rubrum blooms, but decline as ciliate concentrations increased (Peterson et al., 2013). Observations of M. rubrum cells from these communities revealed what appears to be a novel feeding strategy on cryptophyte algae, with numerous prey cells attached to either the cirri and/or feeding tentacles of the ciliates (Peterson et al., 2013). These observations suggest that some M. rubrum populations may be able to act opportunistically and quickly consume a surplus of cryptophytes in order to sustain their productivity. In the Columbia River, M. rubrum cells from blooms were analyzed for 16S rRNA gene diversity during two consecutive years and found to exclusively possess T. amphioxeia-like plastids (Herfort et al., 2011b). In contrast, surface clone libraries from the estuary were nearly devoid of cryptophyte sequences, suggesting they primarily reside at depth (see Peterson et al., 2013).