In 2011, water samples from the sub-surface (1 m depth) were collected at Arcachon Bay in France (Figure 2) between July and October for microarray analysis (Table 1). The sampling site termed Tès (1°10'00 W, 44°40'00 N), is directly located in front of the town of Arcachon inside the bay. Data of toxic, harmful, and other phytoplankton abundances is provided by IFREMER (Ifremer/Quadrige²/Rephy DATA) from the paired station named Teychan (1.5 km from Tès). Cell counts were done as previously described by Medlin and Schmidt [15] and Kegel et al. [16].

For the microarray analysis, a minimum of three liters (Table 1) were filtered onto 3 µm nitrocellulose filters (47 mm) in triplicate. For each sampling date, the first and second replicated filter was transferred into cryogenic vials containing 1 mL of TRI Reagent (Sigma-Aldrich). Those samples were snap frozen and stored at –80 °C until further process for RNA extraction. Toxicity was measured by one of the Partners (Queens University Belfast, UK) with a newly developed multiplex optical Surface Plasmon Resonance biosensor (Multi SPR) in parallel with the enzyme-linked immunosorbent assay (ELISA) [17]. The target toxins are domoic acid (DA) for amnesic shellfish poisoning (ASP), okadaic acid (OA) and dinophysistoxins (DTXs) for diarrhetic shellfish poisoning (DSP) and saxitoxin (STX) for paralytic shellfish poisoning (PSP) toxin analogs. Therefore, the third replicated filter was transferred into cryogenic vials without TRI Reagent and sent frozen to Queens University Belfast who was responsible for the toxin measurements.

RNA extraction was done with minor changes to that presented in Kegel et al. [16]. Briefly, acid-washed glass beads (300 µm) and 500,000 cells of Dunaliella tertiolecta-strain UIO226 (stored in TRI Reagent) as a control were added to the samples and the samples were bead-beaten twice for 1 min at 4,800 oscillations/min (BioSpec Mini Bead Beater). Cell-TRI Reagent mixture was transferred into a new microcentrifuge tube, vortexed for 15 s and left to stand at room temperature (RT) for 10 min. After another 15 s of vortexing, samples were incubated at 60 °C for 10 min in a Thermoshaker at maximum speed. Samples were vortexed again for 15 s and then transferred into pre-spun Phase Lock Gel Heavy 2 mL tubes (5 Prime; 12,000 g for 30 s). After the addition of 100 µL of 1-bromo-3-chloropropane (BCP) to the samples, the tubes were shaken thoroughly for 15 s. Samples were incubated at RT for 5 min and centrifuged (12,000 ×g) for 15 min at 4 °C. The upper phase was mixed gently with 200 µL chloroform and centrifuged (12,000 ×g) for 2 min at 4 °C. The aqueous phase was then transferred to a fresh 2 mL RNase-free tube. Equal volumes of isopropanol were added, vortexed for 15 s and incubated for one hour at −20 °C. After incubation, samples were centrifuged (12,000 ×g) for 15 min at 4 °C. Supernatant was quickly removed and pellets were washed three times with 1 mL ethanol (75%): ethanol was added, vortexed for 5 s, centrifuged (12,000 ×g) for 10 min at 4 °C, and supernatant carefully removed. Following the third wash, the supernatant was completely removed and the pellet was air-dried for 5 min. The pellet was dissolved in 100 µL of RNase-free water. To get rid of TRI Reagent residuals, samples were precipitated with 0.5 volume of 7.5 M NH₄Ac and 2 volumes of ice-cold ethanol (absolute, stored at −20 °C). The mixture was vortexed and incubated at −80 °C for 1.5 h. Immediately after incubation, samples were centrifuged at 4 °C and max. speed for 20 min. The supernatant was removed; the pellet was washed in 500 µL of 70% ice-cold ethanol (stored at −20 °C) and centrifuged for 5 min at max. speed. The washing step was repeated and the pellet was air-dried for 30–60 min. The RNA was re-suspended in 50 µL nuclease-free water and its concentration and integrity was measured by NanoVue spectrophotometer (GE Healthcare) and Agilent Bioanalyzer 2100 (Agilent Biotechnologies). Samples were snap-frozen in liquid nitrogen and stored at −80 °C until further use.

The PlatinumBright Infrared Labeling Kit from KREATECH (Amsterdam, Netherlands) was used to label 1.5 µg RNA of field sample using 2 µL ULS dye and 2 µL 10× labeling solution in a total volume of 20 µL. Samples were labeled by incubation for 30 min at 85 °C. After incubation, samples were placed on ice and spun down and then purified with KREApure columns (KREATECH) according to the manufacturer’s instructions. Concentration and incorporation of the dye was measured by a NanoVue (GE Healthcare). The DoL (degree of labeling) was calculated and was between 1.9 and 2.2% (Table 1). RNA was fragmented by adding 1/10 volume of RNA fragmentation buffer (100 mM ZnCl₂ in 100 mM Tris-HCL, pH 7.0) and an incubation of 15 min at 70 °C. The reaction was stopped with the addition of 1/10 volume of 0.5 M EDTA (pH 8.0) and the samples were placed on ice. The RNA was fragmented to reduce the effect of the secondary structure on the accessibility of the probe. Despite this fragmentation, we still have heterogeneous probe sensitivity, which reflects the influence of the secondary structure and we can only partially overcome this by fragmenting the RNA to remove the strongest secondary structure formations.

Probe design was done with the open software package ARB [18]. All oligonucleotides including the positive and negative controls were synthesized by Thermo Fisher Scientific (Ulm, Germany) with a C6 aminolink at the 5' end of the molecule. The probes had a length between 18 and 25 nt and a 15 nT-long poly (dT) tail following the NH₂ link at the 5' end. Table 2 shows a list of the probes and their targets. The complete hierarchy for each probe can be found in the GPR-Analyzer which is available online at http://folk.uio.no/edvardse/gpranalyzer. The probe sequences are patent pending and a commercial kit will soon be available from Kreatech containing the array and all reagents for hybridization. Epoxy-coated slides (Genetix or Schott) of MIDTAL version 3.2 were printed using a pin printer VersArray ChipWriter Pro (Bio-Rad Laboratories GmbH, Munich, Germany) and split pins (Point Technologies, Inc., CO) as described by Kegel et al. [16]. One array contained 136 different probes and 4–8 replicates, as well as three negative (NEGATIVE1_dT, NEGATIVE2_dT, NEGATIVE3_dT), one positive control (TBP = TATA-box binding protein), Poly-T-Cy5 (spotting control), and two internal controls (DunGS02_25_dT and DunGS05_25_dT for Dunaliella tertiolecta) (MIDTAL ver3.2). After spotting, slides were incubated for 30 min at 37 °C and then stored at −20 °C.

Printing of MIDTAL slides version 3.3 was done by Scienion AG using a sciFlexarrayer S11 and epoxy-coated slides from Genetix. One array contained eight replicates of 140 different probes including the seven controls stated above. After printing, the slides were transferred to a 75% humidity chamber, kept there overnight at RT, and stored afterwards in a sealed aluminum bag refilled with argon at 4 °C.

Before use, slides were blocked by incubating the DNA chips in a blocking solution (0.02% SDS, 2× SSC) for 20 min at 50 °C and ~70 rpm in the dark. The slides were washed once in ddH₂O for 10 min at 50 °C and twice always in fresh ddH₂O for 15 min at RT and ~70 rpm in the dark. The slides were dried by centrifugation in a glass dish for 3 min at 900 rpm and stored in the fridge (possible for up to two month).

Labeled field samples (1 µg RNA) were mixed with 30 µL of 2× hybridization buffer, 3 µL Poly-dA (1 µM), 10 ng TBP-control and adjusted with nuclease-free water to 45 µL. Poly-dA is added to block the poly-T spacer on the probe and TBP is the TATA box gene fragment added as the positive hybridization control. The labeled RNA was then denatured for 5 min at 94 °C. After denaturation, the samples were shortly placed on ice and 15 µL of KREAblock (background blocker from KREATECH) was added. Slides were placed into an array holder; coverslips (LifterSlips, Erie Scientific, USA) were cleaned and placed onto the microarrays. Half of the hybridization mixture (30 µL) was added to one microarray. Hybridization was carried out for 1 h at 65 °C in a 50 mL Falcon tube containing a wet Whatman paper. The DNA chips were washed three times and shaken (~70 rpm) in the dark under stringent conditions. The washings were always undertaken for 10 min. The incubation in the first washing buffer (2× SSC/10 mM EDTA/0.05% SDS) and the second washing buffer (0.5× SSC/10 mM EDTA) was done at room temperature. The incubation in the third washing buffer (0.2× SSC/10 mM EDTA) was done at 50 °C.

Obtained fluorescent signals and the surrounding background intensity were calculated by superimposing a grid of circles (midtal_ver32_20110429.gal or MIDTAL_V3.3.gal) onto the scanned image using the GenePix 6.0 software. First results were processed through the phylochip analyzer program to generate a hierarchy file to establish the hierarchical levels of the probes on the chip [19]. The hierarchy file and hybridization results were then progressed with the GPR-Analyzer version 1.27 and the hierarchy file version 1.06 [20]. A signal-to-noise ratio (S/N ratio) above two was taken as a cutoff for a positive signal. To compare values from different hybridizations, signals were normalized using the internal control DunGS02_25_dT (corresponds to Dunaliella tertiolecta), and replicates averaged. The mean of the total signal intensity and its standard deviation (SD) for the replicates of each probe, which are depicted in the graphs below, can be found in supplementary S2. All microarray results were uploaded to the MIDTAL database at http://www.mba.ac.uk/midtal. Specific instructions can be found in the MIDTAL manual [14] to open a new account from this site.

The samples were characterized by a mixed assembly of species (Table S1) and dominated mainly by diatoms and cryptomonads (Table 3). Dominant taxa in the five samples were Chaetoceros spp., Cryptomonadales, Asterionellopsis glacialis and Cylindrotheca closterium. The last sample (6A) by the end of October showed also a bloom of Nitzschia spp. With respect to potentially toxic algae (Table 4, Table S1), it was possible to observe several developments of Pseudo-nitzschia species. Based on morphological characteristics of the valves and on previous distinctions made by other authors [21,22], species of Pseudo-nitzschia were grouped and counted using four identification groups: the “slender” (seriata complex, i.e., P. multiseries + pungens), the “thin” (valve < 3 µm, delicatissima complex, i.e., P. calliantha + delicatissima + pseudodelicatissima), the “wide” (valve > 3 µm, seriata complex, i.e., P. australis + fraudulenta + seriata + subpacifica), and the “sigmoid” (P. multistriata). The sigmoid group was observed with low abundances in August (sample 3A, 400 cells·L⁻¹) and higher abundances in October (samples 4A and 6A) with a maximum of 40,600 cells·L⁻¹ in the last sample. In July (sample 1A), it was possible to observe a high concentration of the “wide” Pseudo-nitzschia (P. australis, fraudulenta, seriata, and subpacifica) with 30,200 cells·L⁻¹, as well as 2 cells of Alexandrium spp.. At the beginning of August (sample 2A) and the beginning of October (sample 4A), species of Prorocentrum (P. cf. minimum, balticum, and cordatum) were detected with 400 cells·L⁻¹. Furthermore, 7,000 cells·L⁻¹ of Pseudo-nitzschia (mainly from the “thin” group) were observed in sample 2A. Except for the aforementioned Pseudo-nitzschia multistriata (400 cells·L⁻¹), no potentially toxic species were observed at the end of August (sample 3A). In both October samples (4A and 6A), it was possible to identify Heterosigma akashiwo with 600 and 400 cells·L⁻¹, respectively. In the late October sample (6A), 30 cells·L⁻¹ of Dinophysis caudata were also counted.

Despite few events of potentially toxic algae blooms during our study period, we can point out the presence of five genera in our samples (Pseudo-nitzschia, Heterosigma, Prorocentrum, Alexandrium, and Dinophysis) that are all represented by probes on the MIDTAL microarray.

The insertion of a taxonomic hierarchy file in the GPR-Analyzer [20] gave us the advantage to distinguish false positives among the species-specific probes in the microarray analysis and exclude them prior to data interpretation. Briefly, for a species to be present, the entire taxonomic hierarchy leading to that species must also be present. The slopes of culture calibration curves of each species incorporated into the GPR-Analyzer allow for the transformation of microarray signals into cell abundances.