For the assay development, we used cultures of A. malmogiense (four strains), B. baltica (four strains), G. corollarium (three strains), and other dinoflagellates that had been isolated from samples collected at the SW coast of Finland and have been maintained at the Tvärminne Algal Culture collection, now included in the Algal Culture Collection of SYKE Marine Research Centre (Table 1 and Figure 1). Except for Protodinium simplex, which was ordered from SCCAP (Denmark) and preserved with Lugol iodine solution upon arrival, all cultures used in this study (Table 1) were grown in temperature-regulated incubators at 50 μmol photons m⁻² s⁻¹ and a 14:10 h light:dark cycle. As a growth medium, F/2-Si enrichment (Guillard and Ryther, 1962) at 6.5 salinity was used; throughout this paper, the salinity values are given in practical salinity units (PSU). All cultures were grown at 4°C, except for Heterocapsa triquetra, which was maintained at 17°C. Subsamples of the cultures in the exponential growth phase were taken, preserved with Lugol iodine solution and stored chilled in the dark until they were used for microscopy analysis and DNA extraction.

In addition to the monocultures, field samples were collected in a coastal area (Himmerfjärden Bay, 58°.59′N, 17°.43′E; bottom depth 32 m) and used for the assay validation (Figure 1). Two time periods were selected for the sample collection to ensure (1) the presence of abundant and diverse dinoflagellate communities (March 15–June 8, 2010), when all target species are present ubiquitously in the water column; and (2) the absence of any of the target species in the plankton (September 28, 2010); by this time all of them should have undergone complete encystment and disappear from the water column (Kremp et al., 2009). All samples were collected in the photic zone (0–14 m) during the daytime using a plastic hose (inner diameter of 25 mm), and the entire sample volumes were transferred to a bucket. After thorough mixing, two subsamples of the collected plankton were drawn from each sample and designated for (1) microscopy analysis of the Apocalathium/Biecheleria/Gymnodinium complex; these samples (200 mL) were immediately preserved with acetic Lugol iodine solution (0.8 mL), and (2) qPCR assay; these samples (1 L) were also preserved with Lugol iodine (4 mL). All samples were stored chilled in darkness until analyses.

For each species, we designed primers and a hydrolysis probe (minor groove binder, MGB), which carries a fluorophore at the 5′-end and a quencher at the 3′-end; see Supplementary Text S1. The hybridized probe is cleaved by the 5′ exonuclease activity of the Taq polymerase resulting in fluorescence increase. The species-specific primers and those probe sequences characterized by the highest number of mismatches to non-target sequences were tested in silico for their specificity by a BLAST sequence similarity search¹ against the GenBank nucleotide collection. Due to the additional specificity provided by the presence of the TaqMan probe, this approach was selected because it is less subject to false positives than the intercalating dye method (Kutyavin et al., 2000), which is crucial for mixed plankton samples.

The target for designing primers and probes was the rDNA ITS, regions ITS1 and ITS2. Sequencing of B. baltica (strain WHTV-C1) and G. corollarium (strain GCTV-B4) were performed in collaboration with the Department of Biomolecular Sciences, University of Urbino, Italy. The 5.8S rRNA gene and flanking ITS regions were amplified and sequenced using the BigDye Terminator Cycle Sequencing Kit (Applied Biosystems, Cheshire, United Kingdom) and an ABI PRISM 310 Genetic Analyzer instrument. The sequences have been deposited into the GenBank database (accession numbers MN525428 and MN525429 for G. corollarium GCTV-B4, and B. baltica WHTV-C1, respectively). For B. baltica, WHTV-C1 sequence and GenBank sequence DQ167868 were used to generate a consensus sequence. For A. malmogiense, a consensus sequence was created based on the GenBank sequences for the following strains: SHTV-2 (AY970655), SHTV-5 (AY970656), SHTV-6 (AY970657), SHTV-1 (AY499515), and 702 (EF205037).

The primer-probe systems (Supplementary Table S1) were first tested in silico for specificity. For each target species, alignments were made using the software BioEdit Sequence Alignment Editor version 7.0.9.0 (Hall, 1999). Non-target species were included based on taxonomic relatedness to the target species, strong sequence match, and occurrence of the species in the Baltic Sea spring phytoplankton community. The alignments were searched manually to determine unique sequences within the ITS regions relevant for primer and probe design, which was then performed using Primer3 (Whitehead Institute and Howard Hughes Medical Institute, Maryland). A database similarity search was made using BLAST to ensure in silico specificity of primers and probes. Primers were synthesized by Sigma-Aldrich Sweden AB (Stockholm, Sweden) and probes by Applied Biosystems (Cheshire, United Kingdom). The probes were dual-labeled with the fluorophore 6FAM and the quencher MGBNFQ at the 5′ and 3′ ends, respectively.

The DNA was obtained by Chelex extraction (Giraffa et al., 2000) using Lugol-preserved samples of dinoflagellates and wild plankton; this method was selected based on pilot tests evaluating several extraction protocols (Supplementary Text S2). For all extractions, the appropriate culture/sample volume was filtered onto a 5.0 μm Millipore TMTP Isopore polycarbonate membrane filter (Ø 25 mm) (Millipore, MA, United States) using a 15 mL Millipore filtration tower (Ø 16 mm) fitted to a vacuum filtration manifold. The filter was placed in a petri dish and cut diagonally with a scalpel into eight pieces which were all placed in a 2 mL tube. Nuclease-free water (200 μL; Qiagen, Hilden, Germany) was added, and samples were centrifuged at 10,000 rpm for 5 min, and then frozen at −80°C; see Supplementary Text S2 for the details and justification of this procedure. After thawing, each sample received approximately 100 mg of acid-washed glass beads (≤106 μm; Sigma-Aldrich, MO, United States) and was homogenized for 2 × 20 s in a FastPrep^®-24 Instrument (MP Biomedicals, CA, United States). Then, 200 μL Chelex^® 100, 10% w/v (Bio-Rad, CA, United States) were added followed by heating at 50°C for 30 min and then at 105°C for 8 min. The sample was centrifuged at 10,000 rpm for 5 min, and the supernatant (100 μL) containing the DNA was transferred into a 1.5 mL tube and stored at 8°C overnight before processing.

The quantity of the extracted DNA was evaluated by reading the whole absorption spectrum (220–750 nm) with a Nanophotometer^™ (Implen), and the DNA quality was assessed using absorbance ratio at both 260/280 and 230/260 nm. Both ratios were acceptable in most of the samples (1.8–2.0 and 2.0–2.2, respectively). In addition to the concentrations measured spectrophotometrically, the DNA concentration in each sample was also quantified fluorometrically by staining with Hoechst dye 33258 (Sigma-Aldrich) to avoid an overestimation of the DNA concentration measured by Nanophotometer (Bhat et al., 2010). The fluorometry-based DNA concentrations were applied to estimate template concentration in the TaqMan real-time PCR reaction.

Gene-based standard curves were constructed for all three species to determine the assay’s efficiency, using a synthetic gene fragment approach (Vermeulen et al., 2009). A synthetic DNA oligonucleotide (Invitrogen Ltd.) comprising the target sequences (Table 2) and cloned into a plasmid was used as a standard (Figure 2). For each target species, the synthetic gene was assembled and cloned into pMA-T plasmid using Sfil and Sfil cloning sites. The plasmid DNA was purified from transformed bacteria (E. coli, K-12), and concentration was determined by UV spectroscopy. The final construct was verified by sequencing, and the congruence within the restriction site was 100%. When preparing the DNA for the standard dilutions, 5 μg of the lyophilized plasmid DNA were dissolved in 50 μL distilled water. For the standard curves based on the synthetic gene fragments, a 10-fold serial dilution ranging 75 to 75 × 10⁵ copies per reaction for all species of the plasmid construct containing the target sequence, were generated and used as a template. Each reaction was performed in duplicate.

One strain of each species, A. malmogiense (SHTV-5), B. baltica (WHTV-S1), and G. corollarium (GCTV-B4), were used to extract genomic DNA and construct cell-based standard curves. These extractions were serially diluted and used to generate target DNA representing a known cell number per reaction versus Ct data. The cell-based standard curves were constructed using 10-fold dilutions of the DNA extract using known cell concentrations ranging 1 to 8 × 10³ cells mL⁻¹ for A. malmogiense, 1 to 7 × 10³ cells mL⁻¹ for B. baltica, and 1 to 9 × 10³ cells mL⁻¹ for G. corollarium. Each reaction was performed in triplicate.

To determine the mean number of the rRNA gene copies per cell, the dilution series with a known cell count and gene-based Ct values were used as input for calculation. The slope of the linear regression between the rRNA gene copies and cell numbers in the reaction were used to determine the copy number per cell. This analysis was conducted for three strains of each target species (Table 1).

Candidate primers for A. malmogiense, B. baltica, and G. corollarium were tested using conventional PCR and DNA extracts from the monocultures of SHTV-5, WHTV-S1, and GCTV-B4, respectively. Each PCR reaction contained 2 μL of extracted DNA, 5 μL of Standard Taq Reaction Buffer (Final concentration 2.5×, 3.75 mM MgCl2, New England BioLabs, MA, United States), 1 μL of dNTPs (Final concentration 0.5 mM, New England BioLabs), 0.5 μL of Taq polymerase (corresponding to 2.5 units, New England BioLabs), 2 μL of each primer (final concentrations 1 μM), and 7.5 μL of nuclease-free water (Qiagen, Hilden, Germany) resulting in a total reaction volume of 20 μL. Nuclease-free water was used instead of DNA for the non-template controls (NTC). Samples were amplified using MJ Research MiniCycler for 2 min at 94°C followed by 30 cycles of 15 s at 94°C, 15 s at 60°C and 45 s at 72°C with a final extension step for 7 min at 72°C. PCR products were mixed with 6× loading dye solution and ran on a 1.5% agarose gel. GeneRuler 100 bp DNA Ladder (Thermo Scientific, MA, United States) was used as a marker. The gel was stained with ethidium bromide (0.2 μg/L) for 20 min, rinsed in distilled water for 7 min and visualized under UV light. All gels were examined to ensure that the PCR products had the expected size (bp) and no amplification occurred in NTC. Two main criteria for primer selection for qPCR were used: (1) efficient amplification as indicated by clear and sharp bands in the gels, and (2) the product size, with shorter amplicon size being favored as more suitable for qPCR (Penna and Galluzzi, 2013).

To ensure proper binding to the target species, functional primers with the corresponding probes were tested by qPCR using a StepOne Real-Time PCR instrument (Applied Biosystems). The following reagents were added for a 20 μL reaction mixture: 10 μL of TaqMan Gene Expression Master Mix (Applied Biosystems), 1.5 μL of each primer (0.75 μM), 0.5 μL of the probe (0.25 μM), 2 μL of the DNA extracted from a target species, and 4.5 μL of nuclease-free water. For the NTC, nuclease-free water was used instead of DNA. The thermal cycling conditions consisted of 2 min at 50°C and 10 min at 95°C followed by 40 cycles of 15 s at 95°C and 1 min at 60°C. Fluorescence data were collected at the end of each cycle, and the cycle threshold was determined automatically by the instrument software. Extracted DNA from the cultures SHVT-5, WHTV-S1 and GCTV-B4 was used in six-step tenfold serial dilutions (1×–10⁵×) and primer-probe sets were evaluated based on the qPCR efficiency and correlation coefficients (R²). All samples, standards, and NTC were analyzed in triplicates.

The method reproducibility (inter-assay variation) was assayed by calculating the CV_Ct (coefficient of variation for cycle threshold) of gene-and cell-based standards in independent experiments run on different days using different sets of dilutions; five assays were used for gene-based standard curves, and four assays were used for cell-based standard curves. The replicate standard curves were produced with the final set of primers and probes using a serial dilution series over the six-step tenfold serial dilutions (1×–10⁵×). The inter-run variation was calculated for Ct values as described by Rutledge and Côté (2003). For each primer-probe set the pooled slope and intercept were calculated and used for downstream applications of the gene-based standard curve.

Each assay was tested with several strains of the respective target species (Table 1). The specificity of each assay was confirmed by using DNA from the non-target species as a template. More specifically, cross-reactivity between primers and probe for A. malmogiense were tested with DNA extracted from the other two species, and the same was done for the assays developed for B. baltica and G. corollarium. The specificity of each primer-probe set was further tested by screening against non-target species Heterocapsa triquetra, Peridiniella catenata and Protodinium simplex (Table 1) that are phylogenetically close to the target species (Saldarriaga et al., 2001; Zhang et al., 2007). All primer-probe sets were also tested with DNA from a field sample collected in spring 2010—a time when all target species were expected to be present in the water column. The qPCR products were sequenced (KIGene, Karolinska University Hospital, Stockholm, Sweden) to confirm the specificity of the assays. Furthermore, a field sample collected in the fall, not containing the target species, was also used in the qPCR assay. As positive controls, the target species samples were included in all assays.

Out of seven primer pairs (Supplementary Table S1), one pair (S416F/S494R) produced an unspecific product in the NTC. All other pairs amplified the desired DNA region with no amplification in the NTC. However, the pair S366F/S542R was eliminated from further testing because of the relatively large amplicon, which was considered suboptimal for qPCR. The remaining five pairs were combined with probes and tested in the qPCR assays. Among those, the best reaction efficiency (94.8) and R² (~1.00) were observed in the qPCR assay with primers S459F/S542R in combination with probe S484 (Table 2).

Out of six primer pairs (Supplementary Table S1), two pairs (B86F/B211R and B504F/B601R) produced unspecific products in the NTC and were eliminated from further testing. The remaining four pairs amplified the target region with no product in the NTC. Two of these, B51F/B211R and B134F/B211R, were chosen for downstream qPCR testing in combination with compatible probes (B169 in both cases). In the qPCR tests, the superior efficiency and R² values, 93.3 and ~1.00, respectively, were obtained for the primers B51F/B211R (Table 2).

All four primer pairs (Supplementary Table S1) amplified the desired target region with no amplification in the NTC. In the qPCR assay, the primer set G458F/G602R with G535 probe produced standard curves with the highest efficiency (96.9) and R² = 0.99 (mean values for both parameters, Table 2). In no case was a positive amplification observed in the NTC of qPCR.

The gene-based standard curves covered linear detection over six orders of magnitude, with a mean qPCR efficiency of 96, 94 and 91% for G. corollarium, A. malmogiense, and B. baltica, respectively (Table 3 and Figure 2A). The detection limit tested was 75 gene copy numbers in all assays. The Ct for the lowest gene copy number tested varied from 33.6 in GCTV to 34.7 in WHTV, so it is not likely that the sensitivity lower than 75 gene copy numbers can be achieved. The inter-assay variation was low, with CV_Ct mean values of the standard curves being 1.3, 0.8 and 0.7% for B. baltica, A. malmogiense, and G. corollarium, respectively (Table 5 and Figure 3A). The CV_Ct was not related to the copy number in the reaction in all assays. The pooled slope and intercept, calculated for the respective target species standard curve, are provided in Table 3.

The cell-based standard curves covered linear detection over six orders of magnitude, with a mean qPCR efficiency of 94% for A. malmogiense and B. baltica, and 92% for G. corollarium (Table 4 and Figure 2B). For all three species, the detection limit tested was less than one cell in the reaction. The CV_Ct mean values of the standard curves for the target species were 1.7, 1.1 and 1.0% for B. baltica, A. malmogiense and G. corollarium, respectively, and remained at ≤1.0% for all species in the low cell abundance range (1 to 10² cells mL⁻¹). In B. baltica, the CV_Ct was significantly positively related to the cell abundance, whereas the relationship for A. malmogiense was nearly significant (Figure 3). In G. corollarium, the CV_Ct was the lowest among the three species (Table 6) and unrelated to the cell abundance (Figure 3).

No cross-reactivity for any of the laboratory strains of A. malmogiense, B. baltica and G. corollarium was observed (Table 7). Similarly, no amplification was observed for DNA extracted from the cultures of Heterocapsa triquetra, Peridiniella catenata, and Protodinium simplex, except for a few cases where a non-specific product was observed after more than 35 cycles (i.e., beyond the dynamic range of the assays; Figure 2A). Sequencing of the qPCR-products from the spring bloom field sample DNA confirmed the presence of all three target species and the specificity of the assays. Moreover, when the environmental DNA extracted from the field sample collected in September 2010 was used as a template, no specific amplification was observed with any of the primer-probe sets selected for the species-specific qPCR assays for A. malmogiense, B. baltica and G. corollarium, which was consistent with the absence of these species in the phytoplankton community at the time. However, for the B. baltica and G. corollarium primer-probe sets, some unspecific amplification occurred at Ct >34, corresponding to the end of the standard curve and outside of the standard curve, respectively.

There was a very strong (R² > 0.98 in all cases) relationship between the cell abundance estimates based on qPCR and the Utermöhl counts (Table 8). However, the qPCR assays consistently overestimated the number of cells in the culture samples used in the laboratory testing (Figure 4). Moreover, neither the slopes nor the intercepts of the regression lines for the test strains were significantly different (slopes: F_2,12 = 2.597, p > 0.12; the pooled slope equals 5.12; intercepts: F_2,14 = 0.059, p > 0.94; the pooled intercept equals −217).

To evaluate the adequacy of the developed qPCR assays for environmental screening, they were applied to environmental community DNA extracts from plankton collected in a coastal area of the Northern Baltic Proper during spring bloom 2010 (Figure 1). For consistency, the pooled slope and intercept for the plasmid standard curve for each species (Table 3) were used when estimating cell abundances in the field samples. To convert the Ct values to the cell abundance estimates, we applied the average rRNA gene copy number observed for each species (Figure 5). Variable cell abundance was observed for the dinoflagellates comprising the Apocalathium/Biecheleria/Gymnodinium species complex, with clear peaks for G. corollarium and B. baltica (Figure 6). The most common species was G. corollarum detected at high abundance on all sampling occasions, whereas A. malmogiense was detected at low abundances and only twice, and moderate quantities of B. baltica were detected on all sampling occasions. The regression for the qPCR vs. cell counts was highly significant, although the qPCR-based estimates in most cases exceeded those based on the light microscopy (2.4-fold on average; Figure 7). Within the observed range of the dinoflagellate abundance (4 to 4 × 10⁵ cells L⁻¹), there was a significant positive relationship between the fold difference (between the qPCR and cell counts) and the total dinoflagellate abundance when the extreme value of 5.4-fold difference observed on the 15th of March was excluded from the regression.