Lake Superior Supports Novel Clusters of Cyanobacterial Picoplankton

Samples for phylogenetic analyses were collected during two research cruises in the western arm of Lake Superior in May and September 2004 (Fig. 1). Samples for flow cytometry and for photosynthetic rate measurements were obtained during cruises in March and August 2006. In May 2004, during isothermal mixing, water was collected from a depth of 5 m at stations Sterner B (SB; 46.81°N, 92.13°W), Castle Danger 1 (CD1; 47.30°N, 91.67°W), and Western Mid-lake (WM; 47.33°N, 89.90°W). In September 2004, when the water column was stratified, water was collected from the epilimnion and deep Chl maximum (DCM) from stations CD1 (5 m and 15 m respectively) and WM (5 m and 30 m, respectively) and from the surface mixed layer (5-m depth) from station SB and nearshore Keweenaw Waterway station Portage Deep (PD; 47.06°N, 88.50°W). Water from all stations except for SB and PD was collected using a trace metal clean pumping system (34). Water from stations SB and PD was sampled using an acid-clean Teflon-coated Go-Flo bottle (General Oceanics, Miami, FL). For phylogenetic analyses, water from all stations was passed through 0.45-μm polycarbonate filters (GE Osmonics, Trevose, PA), which were then placed into 1.6-ml microcentrifuge tubes containing TE buffer (10 mM Tris-HCl [pH.7.5], 1 mM EDTA-Na₂) and stored at −20°C for DNA extraction.

Whole water was collected as described above and drawn through a 0.2-μm polycarbonate filter, which was immediately placed onto a BG-11 agarose plate (1) containing 250 μg ml⁻¹ cycloheximide, and the plates were stored in the dark until the end of the cruise. All BG-11 media were modified with reduced nitrate as described previously (16). In the laboratory, plates were incubated at 18°C in 10-μmol quanta m⁻² s⁻¹. Strains were also obtained by amendment of surface water collected from station CD1 in May 2004 with 30 μM NaNO₃, 8 μM K₂HPO₃, 10 nM FeCl₃, and 250 μg liter⁻¹ cycloheximide_. After 4 weeks of room temperature incubation under continuous illumination at 5 to 10 μmol quanta m⁻² s⁻¹, the culture was centrifuged at 3,000 × g for 15 min to collect picocyanobacteria and the sediment was spread onto BG-11 plates containing cycloheximide as described above. When single colonies of picocyanobacteria formed on plates, they were transferred into liquid BG-11 medium and grown under the same light conditions. After several rounds of retransfer into fresh BG-11, all cultures became unialgal but still were contaminated with heterotrophic bacteria. The strains included in this study were designated LS0417, LS0425, LS0427, LS0503, LC0530, LS0588, and LS0590.

For the extraction of DNA from environmental samples, each filter was aseptically cut into slices and placed into 50-ml Falcon tubes. Three millilters of STE buffer (10 mM Tris-HCl [pH 8.0], 0.1 M NaCl, 1 mM EDTA-Na₂) was added to the tubes with lysozyme to a final concentration of 3 mg ml⁻¹. The tubes were incubated with shaking for 2 h at 37°C. Furthermore, 300 μl of sodium dodecyl sulfate (10% [pH 7.0]) was added to the samples and incubated at 37°C for 1 h, 95°C for 15 min, −70°C for 15 min, and 37°C for 15 min. The tubes were then subjected to two rounds of phenol-chloroform extraction, and the DNA was precipitated with 96% ethanol and resuspended in 50 μl of Tris-EDTA buffer. DNA from the isolated cyanobacterial strains was extracted essentially as described by Crosbie et al. (6).

Oxygenic phototroph-specific primers 106F (5′ CGGACGGGTGAGTAACGCGTGA 3′) and 789R (5′ GACTACAT/AGGGGTATCTAATCCCA/TTT 3′) (27) were used to amplify cyanobacterial 16S rRNA genes from DNA preparations. PCR amplifications were performed in a 50-μl volume containing 5 μl of 10× PCR buffer (500 mM KCl, 100 mM Tris-HCl [pH 9.0], 1% Triton X-100), 5 μl of the DNA preparation, 200 μM concentrations of deoxynucleoside triphosphates, 0.2 nM primers, and 0.25 μl of Taq DNA polymerase (Promega). The amplification steps included 94°C for 5 min, followed by 30 cycles of 94°C for 1 min, 58°C for 2 min, and 72°C for 3 min. In order to increase the efficiency of the following TOPO reaction, a final step of 30 min at 72°C was added to ensure the PCR products are 3′ polyadenylated. The same PCR and ligation conditions with the TOPO vector were used for the amplification of the 16S rRNA genes of isolated cyanobacterial strains. For amplification of cpcBA IGS sequences, the forward and reverse primers described previously were employed (6): forward, 5′ TAGTGTAAAACGACGGCCAGTTGYYTKCGCGACATGGA 3′; and reverse, 5′ TAGCAGGAAACAGCTATGACGTGGTGTARGGGAAYTT 3′. The PCR conditions were as described by Robertson et al. (31).

The pCR 2.1-TOPO vector (Invitrogen) was used to clone the PCR amplicons as described in the user's manual. TOP10F′ chemically competent Escherichia coli cells (Invitrogen) were transformed with the TOPO cloning reaction with Luria-Bertani plates containing 100 μg liter⁻¹ ampicillin used to select for transformants. The colonies were screened for positive clones by colony PCR with the primers used for the initial amplification. Positive colonies were stored in 25% glycerol at −70°C for subsequent sequencing.

High-throughput sequencing of Escherichia coli cultures transformed with TOPO vector bearing inserts of environmental 16S rRNA gene and cpcBA IGS PCR amplicons was performed at the sequencing center of the Clemson University Genomic Institute. Four 96-well plates containing glycerol stocks were frozen at −70°C prior to shipping to Clemson. Sequencing was performed in an ABI 3730XL 96-capillary sequencer using 2 pM per reaction of M13 reverse sequencing primer (Invitrogen) for the pCR 2.1-TOPO vector. Sequences were aligned using the alignment program CLUSTALX 1.83 (14). The phylogenetic trees for each sampling site were constructed from 615 unambiguously aligned bases by the neighbor-joining method using MEGA 3.1 software (20).

To measure photosynthetic rates, NaH¹⁴CO₃ (60 μCi; specific activity, 58 mCi mmol⁻¹ [MP Biomedicals, Solon, OH]) was added to size-fractionated samples of 30 ml following a 30-min dark adaptation. The cell suspension was distributed as 1-ml aliquots into 7-ml glass scintillation vials that were incubated simultaneously under 24 different light intensities for 2 to 3 h using a temperature-controlled photosynthetron (CHPT Mfg. Inc., Georgetown, DE) as described by McKay et al. (23). The reaction was terminated by the addition of 50 μl of formaldehyde to each sample. Acid-stable ¹⁴C assimilation was measured by liquid scintillation counting following the addition of 4.5 ml of Ecolite (+) cocktail (MP Biomedicals) to each vial. Total activity of the added ¹⁴C was determined by adding 20 μl of the sample at time (t) = 0 to scintillation cocktail containing 200 μl of β-phenylethylamine (Sigma). Background activity was determined at t = 0 by dispensing a sample aliquot directly into formaldehyde prior to adding scintillation cocktail. Experiments were performed on unfiltered water and on a fraction that passed through a 5-μm (March) or 2-μm (August) polycarbonate membrane under gentle vacuum filtration (100 mm Hg). Rates of photosynthesis were determined by using a nonlinear regression curve-fitting function (SigmaPlot 9.0; Systat Software, Inc., San Jose, CA) based on the equation of Platt et al. (28). Chl a used to normalize rates of photosynthesis was extracted overnight at −20°C in 90% acetone and measured by fluorometry (37).

Water samples were either analyzed by flow cytometry immediately following collection or preserved in 2% (wt/vol) paraformaldehyde. APP abundance was determined by flow cytometry using a FACSCalibur instrument (15-mW air-cooled argon-ion laser fixed at 488 nm [Becton Dickinson, Mountain View, CA]) with CellQuest software v5.1.1 (BD Biosciences). Double-distilled water was used as a sheath fluid, and the analyzed volume was calculated by measuring the sample flow rate as recommended in the Becton Dickinson manual and recording the time each sample was counted. Samples were run at the “Hi flow” setting (ca. 50 μl min⁻¹) following methods based on Corzo et al. (4). Both eukaryotic APP and Synechococcus sp. cells were enumerated using FL2-versus-FL3 plots. To remove background electronic noise, an FL3 threshold of 52 was set.

The 16S rRNA gene and cpcBA IGS sequences obtained during this study were deposited in GenBank under accession no. DQ519592 to DQ520013, DQ525026 to DQ525059, DQ526396 to DQ526425, and DQ533710.

Water samples were collected from a depth of 5 m at stations CD1, WM, and SB during spring isothermal mixing in May 2004 and from the epilimnion and DCM at stations CD1 and WM and mixed surface waters at stations SB and PD during September 2004 (Fig. 1). High-throughput sequencing of 480 plasmid clones resulted in 368 successful reactions. Out of the entire pool of sequences, 41 did not cluster within the oxygenic phototroph lineage in a neighbor-joining tree, and 29 clustered together with the plastids of the green alga Chlorella mirabilis (Chlorophyta) and diatoms (data not shown). All sequences that clustered within the plastid lineage and the majority (39 sequences) of those that did not cluster within the oxygenic phototroph lineage were from the May 2004 libraries.

The remaining 298 sequences that did cluster within the cyanobacterial radiation were divided into nine libraries according to the sampling sites and dates. The phylogenetic analysis of the epilimnion (5-m depth) library from CD1 is presented in Fig. 2. Included in the analysis are strains LS0417, LS0425, LS0427, LS0503, LS0530, LS0588, and LS0590, which were isolated from pelagic (LS0417, LS0425, LS0427, and LS0588) and nearshore (LS0503, LS0530, and LS0590) stations in Lake Superior during the years 2004 and 2005. Other strains are Cyanobium gracile and Synechococcus freshwater isolates, as well as marine Synechococcus and Prochlorococcus, which together constitute the major clusters of the picocyanobacterial clade. The cluster designations are in accordance with previous studies on freshwater Synechococcus spp. (6, 7, 31). All major picocyanobacterial clusters described in previous studies (6, 7) were recovered by our analysis and retained high bootstrap values (Fig. 2). Thirty-one out of 32 unique sequences clustered within the picocyanobacterial clade sensu Urbach et al. (36) with 99% bootstrap support. The majority of the sequences clustered within two new picocyanobacterial groups, Lake Superior clusters I and II (LSI and LSII, respectively), with average pairwise sequence similarities of 99.4% and 99.7%, respectively. The two groups are closely related to each other and Crosbie group H (6), which consists of a number of PE-rich isolates from Lake Mondsee, Austria, and a PE-rich strain, LPB1, isolated from Lake Biwa, Japan. However, LSI and LSII sequences formed two independent clusters that did not include any of the strains isolated from any other lake previously sampled. Only three of the picocyanobacterial sequences did not cluster within groups LSI and LSII. LS373 was in group B (6), which consists of PE-rich isolates from several oligotrophic lakes in Europe; LS381 grouped with a PE-rich isolate, MH301, from Lake Mondsee, Austria, which remained unclustered in the 16S rRNA gene tree by Crosbie et al. (6). Another picocyanobacterial sequence (LS353) did not cluster with any of the known picocyanobacteria. One sequence in the library, LS365, was outside the picoplankton lineage and was most similar to the 16S rRNA gene of Synechocystis sp. strain PCC 6803 (data not shown). Similarly, the 16S rRNA sequences from the DCM at CD1 formed two new clusters, LSI and LSII (average pairwise sequence similarities of 99.5% and 99.7%, respectively), within the picoplankton clade that were closely related to but well separated from group H (data not shown).

LSI and LSII clusters were also present on the trees obtained for the September epilimnion libraries from WM and SB and the DCM library from WM (data summarized in Fig. 4; see Fig. S1 to S3 in the supplemental material). The same trend was observed in the WM DCM library, although there were fewer LSI sequences compared to the sequences obtained from the epilimnion at that station.

Libraries constructed from the spring (May 2004) samples contained fewer cyanobacterial sequences than did the September libraries. The distribution of sequences among picocyanobacterial clusters from station CD1 resembles that of the September libraries but with fewer sequences in LSI and more group B sequences compared to the epilimnion libraries (summarized in Fig. 4). Unlike the LSII cluster from the September trees, LSII had moderate bootstrap support. However, in a neighbor-joining tree of LSI and LSII sequences from libraries prepared from stations CD1 and PD, all LSII sequences clustered together with a 70% bootstrap (data not shown). The SB and WM libraries from May contained sequences from LSI and II, subalpine cluster I, the MH301 cluster (summarized in Fig. 4), as well as two more sequences clustering with Oscillatoria limnetica (19).

Overall, clusters LSI and LSII contain the most Lake Superior cyanobacterial 16S rRNA gene sequences in our data set from open lake stations. Both groups are closely related to Crosbie group H (6) but form independent clusters supported by high or moderate bootstrap values.

Unlike the cyanobacterial 16S rRNA gene sequences from SB and pelagic stations CD1 and WM, only 2 out of 30 unique sequences from surface waters at station PD clustered with the Lake Superior cluster LSI and cluster LSII was not represented in the library (Fig. 3). In contrast, members of each of the major freshwater picocyanobacterial clusters described in previous studies, except for the C. gracile cluster and group I sensu Crosbie (6), were represented in the PD library (Fig. 3). In addition, a new cluster, MH305, which previously consisted of a single PE-rich strain, MH305 (5, 6), contained two PD sequences (LS11 and LS45) and a PC-rich Lake Erie isolate, ARC11 (GenBank accession no. DQ026832 ; A. R. Cupp and G. S. Bullerjahn, unpublished data). Lastly, two minor PD-specific clusters named PDI and PDII, with average sequence similarities of 99.8 and 99.7%, respectively, were present and did not include any previously characterized cyanobacterial strains.

Figure 4 shows the overall distribution of 16S rRNA gene clones obtained in this study among different lineages within the cyanobacterial radiation. One striking feature of this data set is the apparent difference in the genetic diversities of cyanobacteria between the nearshore station PD and other sampling sites, including pelagic stations CD1 and WM and the western arm station SB near Duluth, MN. LSI is ubiquitous at CD1, WM, and SB (minimum of 50% of the clones), whereas at PD the LSI group accounted for only 4% of the sequences. Furthermore, cluster LSII is absent from the PD library but has representatives at all pelagic stations and SB. Conversely, picocyanobacterial clusters MH305, Crosbie group H, the Lake Biwa cluster, and PD clusters I and II are only present at PD. The major Lake Superior cluster, LSI, is more abundant in the mixed layer during the fall than in the deeper DCM layer or in the surface waters during the spring. The O. limnetica group (19) was present only in the spring libraries, as was the C. gracile cluster. Lake Superior strains LS0504 and LS0503, which cluster within these two groups, respectively, were also isolated in culture from the spring water samples.

DNA sequencing and neighbor-joining analysis of 63 cpcBA IGS clones from nearshore station PD and 105 from pelagic station CD1 revealed the same overall patterns seen with the 16S rRNA gene sequences. Indeed, phylogenetic analysis revealed the presence of novel, low-diversity samples at station CD1 (Fig. 5), whereas the cpcBA IGS library from station PD both revealed high diversity and included the major picocyanobacterial clusters described previously (Fig. 6). The cpcBA analysis at station CD1 revealed that close to 98% of the sequences fell within a unique cluster, compared to 90% of 16S rRNA gene sequences comprising the novel LSI and LSII clusters. In contrast, none of the cpcBA sequences represented in the major CD1 cluster was represented in the PD cpcBA clone library. This disparity between the 16S and cpcBA data is likely due to biases introduced by the PCR. Nonetheless, these data support the conclusion from the 16S rRNA sequences indicating that the APP of pelagic Lake Superior are phylogenetically distinct from other Synechococcus sp. strains identified and characterized around the globe.

Photosynthesis versus irradiance (PE) curves were determined for the total phytoplankton community and for the APP assemblage at station CD1 during March and August 2006. At both times, the APP were an abundant component of the total plankton fraction at this site, comprising >50% of Chl biomass. For the total community, biomass-normalized rates of light-saturated photosynthesis (P_m^Chl) were nearly 2× higher in August than in March (Fig. 7). Measurements of photosynthesis were conducted at in situ temperatures; thus, higher rates in August were consistent with the elevated temperature of midsummer surface waters (20°C) compared to March (4°C). Whereas P_m^Chl did not vary between total phytoplankton and a <5-μm fraction in March, the picoplankton in August exhibited ca. 33% of the P_m^Chl of the total community (Fig. 7). An assessment of photosynthetic efficiency (α) calculated from the initial slope of the PE curves showed no difference between total plankton and the <5-μm fraction during March but modest variation in August, with a picoplankton photosynthetic efficiency ca. 25% lower than the total. A temporal comparison clearly showed higher efficiency for those samples collected during March (α = 5.9 × 10⁻²) compared to August (α = 2.4 × 10⁻² to 3.2 × 10⁻²), consistent with deep mixing and the resultant light-limiting conditions experienced by phytoplankton prior to the onset of thermal stratification at this site in late June to early July (17). Also consistent with the light-limiting conditions synonymous with vernal holomixis were depressed values of E_k, a parameter derived from P_m^Chl/α_Chl to estimate the irradiance at which photosynthesis becomes light saturated. During March, E_k ranged between 17 and 18 μmol quanta m⁻² s⁻¹, whereas in August, E_k increased to 26 and 60 μmol quanta m⁻² s⁻¹ for the picoplankton and total phytoplankton, respectively.

The phylogenetic studies in this paper, along with detailed characterization of cultured freshwater Synechococcus spp., suggest that PE-rich picocyanobacteria comprise the major fraction of Lake Superior APP. However, flow cytometry of water samples obtained during March and August 2006 indicates that the composition of the APP undergoes a large shift from late winter to summer. Indeed, winter to early spring APP included abundant picoeukaryotes and PC-rich picocyanobacteria, whereas summer APP were exclusively PE rich (Fig. 8). Typical abundance of the PE-rich population exceeded 10⁴ cells ml⁻¹ at the surface in summer, with 10⁵ cells ml⁻¹ in the DCM (Table 1). We propose that the PE-rich APP is dominated by picocyanobacteria of the LSI and LSII groups described herein.