Abundance and Biogeography of Picoprasinophyte Ecotypes and Other Phytoplankton in the Eastern North Pacific Ocean

Abstract

Eukaryotic algae within the picoplankton size class (≤2 μm in diameter) are important marine primary producers, but their spatial and ecological distributions are not well characterized. Here, we studied three picoeukaryotic prasinophyte genera and their cyanobacterial counterparts, Prochlorococcus and Synechococcus, during two cruises along a North Pacific transect characterized by different ecological regimes. Picoeukaryotes and Synechococcus reached maximum abundances of 1.44 × 10⁵ and 3.37 × 10⁵ cells · ml⁻¹, respectively, in mesotrophic waters, while Prochlorococcus reached 1.95 × 10⁵ cells · ml⁻¹ in the oligotrophic ocean. Of the picoeukaryotes, Bathycoccus was present at all stations in both cruises, reaching 21,368 ± 327 18S rRNA gene copies · ml⁻¹. Micromonas and Ostreococcus clade OI were detected only in mesotrophic and coastal waters and Ostreococcus clade OII only in the oligotrophic ocean. To resolve proposed Bathycoccus ecotypes, we established genetic distances for 1,104 marker genes using targeted metagenomes and the Bathycoccus prasinos genome. The analysis was anchored in comparative genome analysis of three Ostreococcus species for which physiological and environmental data are available to facilitate data interpretation. We established that two Bathycoccus ecotypes exist, named here BI (represented by coastal isolate Bathycoccus prasinos) and BII. These share 82% ± 6% nucleotide identity across homologs, while the Ostreococcus spp. share 75% ± 8%. We developed and applied an analysis of ecomarkers to metatranscriptomes sequenced here and published -omics data from the same region. The results indicated that the Bathycoccus ecotypes cooccur more often than Ostreococcus clades OI and OII do. Exploratory analyses of relative transcript abundances suggest that Bathycoccus NRT2.1 and AMT2.2 are high-affinity NO₃⁻ and low-affinity NH₄⁺ transporters, respectively, with close homologs in multiple picoprasinophytes. Additionally, in the open ocean, where dissolved iron concentrations were low (0.08 nM), there appeared to be a shift to the use of nickel superoxide dismutases (SODs) from Mn/Fe/Cu SODs closer inshore. Our study documents the distribution of picophytoplankton along a North Pacific ecological gradient and offers new concepts and techniques for investigating their biogeography.

INTRODUCTION

Marine phytoplankton are responsible for roughly half of global net primary production (1), and photosynthetic picoeukaryotes (cells ≤2 μm in diameter) have been shown to play important roles in primary production (2–6). Although these small cells come from diverse lineages, their distinguishing morphological features are limited and difficult to visualize (7, 8). Therefore, documentation of basic ecological parameters, such as abundance and distribution of different taxa, is still limited.

Green algae that belong to the prasinophytes include many marine picoeukaryotes. Class II prasinophytes (i.e., Mamiellophyceae) are the most widespread (7, 8). For example, a circumglobal study on diversity of marine eukaryotes found class II prasinophytes in all 45 samples (9). Molecular phylogenies and field studies have revealed multiple clades or ecotypes within three major class II genera, Bathycoccus, Micromonas, and Ostreococcus (10–12). Genome analyses demonstrate extensive diversity among Micromonas clades that reflect species level divisions (12, 13) and lesser but still notable divergence between Ostreococcus species (14). The Bathycoccus genus is thought to consist of a single species (Bathycoccus prasinos), and only one cultured strain (Bban7, also known as RCC1105) has a sequenced genome (15). Cultured Bathycoccus strains and environmental clones have 100% 18S rRNA gene identity, unlike the Micromonas and Ostreococcus clades (16–18). However, targeted metagenomic data from the tropical Atlantic suggest that different Bathycoccus ecotypes may exist, based largely on an intron/intein presence/absence polymorphism in the gene PRP8 (18, 19). Notably, the internal transcribed spacer (ITS) from the tropical Atlantic wild Bathycoccus metagenome also branches separately from cultured strains (18), while the ITS from cultured strains and two coastal Chilean Bathycoccus metagenomes are similar (20). Sequence variations have led to the proposal that two (18) or more (20) ecotypes exist and that the ecotypes may partition between mesotrophic and oligotrophic environments (18). However, little is known about overall genetic distances and whether the observed sequence variants do indeed correspond to ecotype-level differences.

Although much is known about other picophytoplankton, especially the cyanobacteria Prochlorococcus and Synechococcus, less data are available on the abundance of different class II prasinophytes in nature (10, 11, 21–23). Here, we enumerate picoeukaryotes as well as their cyanobacterial counterparts along an ecological gradient in the eastern North Pacific. Using quantitative PCR (qPCR), we enumerated two Ostreococcus clades (OI and OII), Micromonas, and Bathycoccus at coastal, mesotrophic, and oligotrophic ocean sites. The ecotypes were further resolved by establishing genetic distances among homologs identified in three genome-sequenced Ostreococcus species, the B. prasinos genome (15), and three Bathycoccus targeted metagenomes (18, 20). This information was used to investigate the oceanographic distributions of Ostreococcus and newly defined Bathycoccus ecotypes in metatranscriptomes generated here and in prior metatranscriptomic and metagenomic data from the same region. The metatranscriptomes were also used to explore gene expression in relation to the environmental gradient under study. Collectively, the results document the distributions and genetic divergence of widespread picoeukaryotic phytoplankton.

MATERIALS AND METHODS

Field sampling.

The Transect survey was carried out in 2009 (cruise WFAD09), and the Drift study was performed in 2010 (cruise CANON10, also referred to as S410). Conductivity-temperature-depth (CTD) casts sampled 16 and 13 stations during the Transect and Drift surveys, respectively. During the Transect survey, the same stations were sampled on both the outward-bound (westward direction) and inbound legs of the cruise; these were separated by 8 days, during which other cruise activities and sampling occurred. Water for nutrients, flow cytometry, and DNA were collected using Niskin bottles mounted on a rosette with a CTD instrument and fluorometer. Five-hundred or 1,000 ml of seawater was filtered onto a 0.2-μm-pore-sized Supor filter for DNA (Pall Scientific, Port Washington, NY, USA) and frozen at −80°C. RNA samples from WFAD09 (50 liters) were collected using Niskin bottles, gravity filtered through 20-μm Nitex mesh, and subsequently gravity filtered onto an 0.8-μm-pore-size 142-mm-diameter Supor filter using a peristaltic pump. Bisected filters were transferred to 50-ml tubes and frozen at −80°C approximately 1 h after CTD instrument retrieval. Flow cytometry samples were preserved at a final concentration of 0.25% using an EM-grade glutaraldehyde (Electron Microscopy Sciences, Hatfield, PA, USA) (24). After 20 min of incubation at room temperature in the dark, the samples were frozen in liquid N₂. Dissolved iron profiles were generated from 3- to 10-ml samples collected using a trace metal clean rosette during CN207, a cruise 2 years prior to WFAD09 for which metagenomics data were available (analyzed here) and that had water characteristics similar to those of WFAD09 (19, 25, 26). Samples for iron were gently N₂ pressure filtered directly from the Teflon-lined Niskin bottles through a 0.2-μm acid-clean AcroPak Supor filter in a class 100 clean hood.

Nutrients, iron, chlorophyll, and flow cytometry.

NH₄⁺ was analyzed as in reference 27, and NO₃⁻, NO₂⁻, PO₄³⁻, Si(OH)₄, and chlorophyll a (Chl a) were analyzed as in reference 28. Dissolved Fe samples were preconcentrated in line and analyzed using modifications of the sequential injection analysis (SIA) method described in reference 29. The following parameters have previously been reported from these cruises: temperature, salinity, NO₃⁻, NH₄⁺, PO₄³⁻, and Chl a at stations H3, 67-70, and 67-155 on cruise CN207 (19); temperature, salinity, and Chl a for the Drift study (30); temperature, salinity, NO₃⁻, NO₂⁻, NH₄⁺, and Chl a for 15 stations on cruise WFAD09 (31); and PO₄³⁻ for stations H3, 67-70, and 67-155 on cruise WFAD09 (32). Flow cytometry samples were run on an Influx (Becton Dickinson, San Jose, CA, USA) at approximately 25 μl · min⁻¹, as described in reference 5. Winlist 6.0 and 7.1 (Verity Software House) were used to analyze listmodes and characterize Prochlorococcus, Synechococcus, and eukaryote populations as in references 33 and 34.

Quantitative PCR.

DNA was extracted using a modification of the DNeasy plant kit (Qiagen, Valencia CA, USA), including the addition of a bead-beating step (10). Bathycoccus, Micromonas, and Ostreococcus clades OI and OII were then enumerated in triplicate reactions, along with inhibition tests and no-template control reactions. These assays were performed in 25-μl volumes using a TaqMan master mix, as described in reference 10. Inhibition tests were performed using 2 μl of DNA template and an additional 2 μl of 18S rRNA gene plasmid (0.5 × 10⁴ or 0.5 × 10⁵ copies · μl⁻¹) per reaction. Based on these tests, environmental template solutions were diluted between 1:4 and 1:40 to prevent inhibition. Thermal cycling consisted of 10 min at 95°C (initial denaturation), followed by 45 cycles of 15 s at 95°C and 1 min at 60°C, using an AB7500 (Applied Biosystems, Foster City, CA, USA). Data were collected during the annealing phase. Ten-fold plasmid serial dilutions were used to generate standard curves. Threshold and baseline values were calculated using AB7500 software. Copy numbers per reaction were obtained by regression of the cycle threshold (C_T) against log scale copy numbers of standards and converted to 18S rRNA copies · ml⁻¹ of seawater based on the volume of seawater filtered (and extracted) and the amount of template used. Bathycoccus prasinos RCC1105, Ostreococcus lucimarinus, Ostreococcus sp. strain RCC809, Micromonas pusilla CCMP1545, and Micromonas sp. strain RCC299 each have two complete rRNA operons in current genome assemblies (13–15), while Ostreococcus tauri appears to have one (35). Therefore, the 18S rRNA copies · ml⁻¹ herein can be considered equivalent to twice the number of cells · ml⁻¹ (depending on the cell cycle stage).

Statistical analysis.

After testing for normality (using the Kolmogorov-Smirnov test), t tests or Mann-Whitney rank sum tests implemented in SigmaPlot (version 13.5) were used to compare water mass characteristics where, e.g., each prasinophyte taxon was detected versus not detected. Only photic zone samples (defined using a Secchi disk) were included. For the principal-component analysi (PCA), a correlation matrix was used to standardize photic zone data; observations with missing data were omitted. For the analysis using NH₄⁺ data, the PCA contained 16 variables and 27 observations. Omitting NH₄⁺ allowed for 39 observations with 15 variables. Analysis was performed using Matlab (OriginLab, Northampton, MA, USA).

Metatranscriptome library construction and sequencing.

cDNA libraries were constructed from samples collected during the Transect survey (Fig. 1) from the transition zone (TZ; C45) and oligotrophic ocean deep chlorophyll maximum (OO_DCM; C28) at 13:02 and 14:08, respectively, at 19:37 for the coastal ocean (CO; C5), and at 09:11 for the oligotrophic ocean surface (OO_SURF; C26). Extraction supplies were obtained from Life Technologies (Grand Island, NY) unless otherwise noted, and the procedure allowed DNA (not used) and RNA to be extracted from the same material. Buffers were prepared with nuclease-free H₂O and filter sterilized; plasticware was also nuclease free/sterile. Filters were placed in petri dishes, covered with ∼2 ml of lysis buffer (5 ml of RNAlater, 25% sucrose, 2.5 mg · ml⁻¹ lysozyme, 5 mM Tris [pH 8], and 27.5 mM each EDTA and ethylene glycol tetraacetic acid; Sigma, St. Louis, MO, USA), sliced into ∼1-cm² pieces, and transferred with the buffer into polypropylene tubes. The lysis buffer was brought to 20 ml, and samples were incubated for 1 h at 37°C. Four milligrams of proteinase K (Qiagen, Valencia, CA, USA) was added, and samples were frozen (liquid N₂) and thawed (55°C) three times. Samples were incubated at 55°C for 2 h under agitation with 4 mg of proteinase K and sodium dodecyl sulfate (Sigma, St. Louis, MO, USA) to 1% (vol/vol) and centrifuged 2 min at 4,500 × g. Supernatants were transferred to fresh tubes, and filter pieces were stored at −80°C for later reextraction. Extractions used pH 8 phenol, and nucleic acids were recovered by isopropanol precipitation, as in reference 36. Pellets were resuspended in 1 ml of RNA extraction buffer (4 M guanidine thiocyanate, 25 mM sodium citrate, 0.5% Sarkosyl; Sigma) and acidified with 50 μl of 2 M sodium acetate, pH 4 (Sigma). The resulting solution was extracted with pH 4 phenol, chloroform, and isoamyl alcohol (ratios of 125:24:1, respectively; Sigma), and RNA was precipitated with isopropanol (36). Pellets were washed twice with 75% ethanol and resuspended in water. RNA was purified using the RNeasy kit and the RNase-free DNase kit (Qiagen). To maximize recovery, the procedure was repeated on the refrozen filter pieces. The resulting two RNA aliquots per sample were combined, yielding 1.8 to 12.3 μg of total RNA from ∼25 liters of seawater. Then, 500-ng aliquots of RNA were amplified in two rounds of in vitro transcription using the Message AMP II antisense RNA (aRNA) amplification kit and manufacturer's protocols, except for the use of primer T7-BpmI ₁₆VN (5′-GCCAGTGAATTGTAATACGACTCACTATAGGGGCGACTGGAGTTTTTTTTTTTTTTTTVN-3′) instead of primer T7 Oligo (37), yielding between 112 and 237 μg of aRNA per sample. A total of 60 μg of aRNA per sample was converted to single-stranded cDNA using the Superscript III first-strand synthesis system (Invitrogen, Carlsbad, CA, USA) with random hexamer primers. Double-stranded cDNA was produced by incubating DNA for 2 h at 16°C with 40 U of DNA polymerase I, 10 U of DNA ligase, and 200 μM deoxynucleoside triphosphates (dNTPs) in 5× second-strand buffer. RNA was digested with 2 U of RNase H. cDNAs were blunt ended by adding 5 U of T4 DNA polymerase and incubating them at 16°C for 5 min. Purification used Agencourt AMPure XP magnetic beads (Beckman Coulter, Indianapolis, IN, USA). The resulting material was size selected (300 to 3,000 bp) on a low-melting-point 1% agarose gel and extracted using the QIAquick gel extraction kit (Qiagen). Poly(A) tails were digested using BpmI (NEB, Ipswich, MA, USA) according to the manufacturer's protocol, and cDNA was purified with AMPure XP beads. cDNA quantity (1.2 to 1.5 μg) and quality were assessed with Qubit and Bioanalyzer (Agilent, Santa Clara, CA, USA), respectively. Sequencing was performed on a 454 FLX+ instrument using titanium chemistry (Roche/454, Branford, CT, USA).

FIG 1.

Locations sampled during the North Pacific Transect and Drift cruises. (A) Remotely sensed high-resolution Chl a concentrations during the month of the Drift study, October 2010 (MODIS/Aqua). Locations sampled for metatranscriptomes during the line 67 Transect survey (black crosses) and the overall trajectory of the drifter (pink) during the Drift study are depicted. Drift profiles for which qPCR was performed are also indicated (⊙; C44 and C51). (B) In vivo fluorescence (Chl a derived) along the Transect survey (2009). Horizontal lines above the plot show locations sampled in the three regime types (oligotrophic ocean, OO; transition zone, TZ; and coastal ocean, CO). Metatranscriptome sample sites (black crosses) were 25, 172, and 785 km from shore, which corresponded to casts C2, C42, and C20, respectively. r.f.u., relative fluorescence units.

Genetic distances among putative Bathycoccus and Ostreococcus ecotypes.

To identify genes for characterizing Bathycoccus and Ostreococcus genetic diversity and for use as ecomarkers in environmental samples, groups of homologous sequences were first identified in the predicted proteomes from B. prasinos (strain Bban7) and three Ostreococcus genomes (taxa listed above). Bathycoccus sequences from three targeted metagenomes (19, 20) were also used after identifying all possible open reading frames (ORFs; i.e., sequence between stop codons with a minimal length of 60 amino acid residues). The analysis was not performed for Micromonas because, of the seven established clades (12), isolates from only two have complete genome sequences. Clusters of homologs were computed using OrthoMCL (38) based on all versus all BLASTP (39) searches of Bathycoccus and Ostreococcus proteomes, using a cutoff E value of 10⁻¹⁰⁰ E. Clusters containing one sequence from each input source were kept to ensure analysis of exclusively single-copy orthologs (i.e., clusters composed of sequences for the four and three Bathycoccus and Ostreococcus predicted proteomes, respectively). Clusters producing alignments shorter than 100 amino acid residues were discarded. Final data sets were composed of 1,104 Bathycoccus clusters and 3,212 Ostreococcus clusters. Protein sequences from each cluster were aligned with T-Coffee v9 (40) and back translated to codon alignments to improve gene nucleotide alignments. Columns with gaps were removed from the nucleotide alignments, and identities of each codon alignment were used to estimate percent identity at the genus level. To determine if the genetic distances were significant among Bathycoccus and Ostreococcus ecotypes, t tests were conducted on nucleotide identity differences between pairs of ecotypes in R (41).

To identify ecotype/species ecomarkers, the prior all versus all BLASTP results were run again in OrthoMCL with the Bathycoccus and Ostreococcus data together. Only groups composed of single-copy genes and populated by all seven sequence sources (i.e., homolog groups containing seven ORFs, one from each distinct Bathycoccus and Ostreococcus sequence source) were retained. To avoid potential ORF length biases in sequence recruitment (mainly due to large indels or to Bban7 having longer ORFs likely related to gene model differences), the sequences within a cluster were corrected by discarding indels. Sequences from the same cluster were aligned using T-Coffee, and positions with more than three gaps were removed. Only groups of corrected sequences for which the longest sequence was ≤30% longer than the smallest were retained. Sequences with mean identities <50% were discarded, because their alignments were error prone, inflating the distance among taxa. Sequences with >95% identity were also discarded, because they were too similar to be accurately classified for the purposes of this study. A total of 112 homolog groups were identified that could be used as ecomarkers to classify meta-omic reads into different Bathycoccus and Ostreococcus types. To test if the Bathycoccus and Ostreococcus ecotype distributions differed significantly from each other across the sampling locations, pairwise t tests were conducted on their ecomarker relative abundances, and P values were adjusted for multiple comparisons using Bonferroni correction.

Metatranscriptome analyses.

Twelve published metatranscriptomes (30) from samples collected during the Drift study were reanalyzed; specifically, samples were obtained at 4-h intervals starting at 14:00 on 16 September 2010 (36°2.712, −123°1.302) and ending 44 h later (35°47.412, −122°42.54). A 13th metatranscriptome from the prior publication on the Drift study (30) was excluded, because it led to overrepresentation of one time point. Additionally, unlike the Transect survey metatranscriptomes, the Drift study sequences were generated using methods tuned for prokaryotic transcript recovery, and fewer reads were generated per sample. Redundant sequences were removed from Transect and Drift metatranscriptomes using CD-Hit v4.6 (42), because they were considered to be 454 artifacts (43), and rRNAs were removed using riboPicker (44) (see Table S1A in the supplemental material). In both cases, default settings were used. The 454 reads were assigned to taxonomic groups using BLASTX (39) against predicted proteomes from sequenced genomes and transcriptomes from cultures of the following: Aureococcus anophagefferens, B. prasinos, Bigelowiella natans, Chlamydomonas reinhardtii, Ectocarpus siliculosus, Emiliania huxleyi, Guillardia theta, Micromonas sp. RCC299, Micromonas pusilla CCMP1545, Micromonas sp. strain CCMP2099, O. tauri, O. lucimarinus, Ostreococcus sp. RCC809, Pyramimonas parkeae, Phaeodactylum tricornutum, Thalassiosira pseudonana, algal viruses, and vascular plants. Predicted proteins from genome projects were retrieved from GenBank or the Joint Genome Institute genome portal (45). The results were filtered with cutoffs of >60% identity, a bit score >50, and an E value of ≤10⁻²⁰ with the reference protein sequence. Sequences that passed this filter were used as BLASTX queries against the GenBank nonredundant (nr) database (46). This filtering step was performed only for the subset of algae with the most reads, i.e., all those analyzed further, because of its computational intensity (Table S1B). Those that matched a different taxon ID in nr with a better overall score were removed from the set (4.9% ± 3.6% of those initially assigned to genome sequenced picoprasinophytes) (Table S1B). To count reads per reference protein, assigned reads were divided into species-specific sets, and the proteins of the corresponding species were used as TBLASTN queries against each of the parsed read sets. Hits were only counted if they matched the protein uniquely or if the protein was present in duplicate. Reads mapped to a taxon were summed, and gene counts in that sample were expressed as a percentage of the taxon total sum (in that sample). For Drift study metatranscriptomes, this percentage was computed from the two 14:00 (combined) and two 22:00 (combined) data sets to compare to the Transect survey samples collected at the same time of day. The 12 Drift study metatranscriptomes were also summed to maximize B. prasinos transcript counts, and the summed values are used for comparisons here.

The minimum cutoff for comparative analyses was ≥2 reads per predicted protein (on average for a taxon) in two or more samples. B. prasinos and O. lucimarinus met these criteria, but the 67-155 15-m sample did not for the eukaryotic phytoplankton taxon analyzed and was therefore excluded from expression comparisons. Final comparisons were analyzed between the genes in the sample with the fewest total reads to a taxon (but ≥2 per gene on average, i.e., B. prasinos in the 105-m OO_DCM sample), with a focus on those with higher relative percent expression than in samples that had more overall reads assigned to that taxon. Because a greater number of overall Bathycoccus reads were available from the CO and the Drift study, gene expression detection sensitivity was expected to be superior at those sites compared with reads from the OO_DCM. Thus, higher relative percentage of reads from a particular gene in the OO_DCM (compared to the CO and the Drift study) should be reflective of OO_DCM responses without additional rarefraction steps.

Super oxide dismutases (SODs) and nitrogen transporters were identified in metatranscriptomic data after first performing reciprocal BLASTP (39) searches using previously annotated proteins (47) against nr and the Joint Genome Institute (JGI) prasinophyte browsers to recruit previously unreported orthologs. For the ammonium transporter (AMT) phylogeny, Bathycoccus and Ostreococcus RCC809 sequences were added to alignments from reference 47, the alignment was manually adjusted, and ambiguously aligned or nonhomologous positions were masked. Phylip was used to construct neighbor joining (NJ) distance trees (48), and maximum likelihood (ML) methods were performed in PhyML (49) with 100 bootstrap replicates.

Ecomarker analysis was also used to identify Bathycoccus and Ostreococcus reads in the WFAD09 and CANON2010 metatranscriptomes and in line 67 metagenomes available from 2007 (19). The 112 Bathycoccus/Ostreococcus ecomarkers were used as TBLASTN (39) queries against environmental reads. TBLASTN hits (bit-score, >50; E value, <10⁻⁵) were then used as BLASTX queries against nr. Only reads having their best BLASTX hit to one of the Bathycoccus/Ostreococcus ecomarkers (and bit-score, >50; E value, <10⁻⁵) were retained. These candidate reads were then used as BLASTN queries against the four Bathycoccus and three Ostreococcus ORF data sets. ORF data sets from Micromonas CCMP1545 and RCC299 (13) were used to remove false positives (i.e., reads with best BLASTN hits to Micromonas were removed). Reads were classified as being Bathycoccus or Ostreococcus if they produced a BLASTN hit with a bit-score of ≥250 and nucleotide identity ≥95% and binned to one of the four Bathycoccus or three Ostreococcus types based on their best BLASTN hit. Relative contributions were then computed for each sample. Those with the same statistics for two or more distinct Bathycoccus or Ostreococcus sequences were discarded (7.6% of WFAD09 candidate reads, 4.7% of CANON2010, and 4.9% of CN207).

Nucleotide sequence accession number.

Metatranscriptomes generated in this study are available in the NCBI Sequence Read Archive under project number PRJNA300413.

RESULTS AND DISCUSSION

Environmental conditions.

Two cruises sampled picophytoplankton in the northeast Pacific Ocean off central California during boreal autumn (Fig. 1A). The Transect study sampled three ecological regimes along California Cooperative Oceanic Fisheries Investigations (CalCOFI) line 67: the coastal ocean (CO), a mesotrophic/transition zone (TZ), and the oligotrophic ocean (OO) (Fig. 1B) (cruise WFAD09) (32). Along line 67, surface waters at coastal stations are typically colder and more nutrient rich, reflecting coastal upwelling (19, 50), while sea surface temperatures increase and nutrient concentrations decrease offshore (see Data Set S1 in the supplemental material). The Drift study followed a semi-Lagrangian drifting environmental sample processor (ESP) to resample the same mesotrophic water mass near line 67 (Fig. 1A) (cruise CANON10).

Nutrient and chlorophyll concentrations as well as other physicochemical parameters illuminated broad differences across the Transect survey (Fig. 1 and 2; see also Fig. S1 in the supplemental material). NO₃⁻ and NH₄⁺ declined from a euphotic zone maximum of 20 μM and 0.51 μM (31), respectively, to below detection limits (0.020 μM and ∼0.010 μM, respectively; lower NH₄⁺ values were detected but may not have been quantitative), while PO₄³⁻ was always detectable (Table 1; see also Data Set S1 in the supplemental material). Chl a decreased and the chlorophyll maximum deepened with increasing distance from shore (Fig. 1B). At the most offshore stations, the deep chlorophyll maximum (DCM) ranged from 85 to 105 m, and DCM NO₃⁻ and NH₄⁺ ranged from 0.001 to 0.048 μM and 0.003 to 0.009 μM, respectively (Fig. 2A; see also Data Set S1). During the 8 days between sampling the TZ on the westward leg of the Transect survey cruise (casts C8 and C10) and on the inbound leg (C42), the water column structure became less stratified and Chl a became more homogeneous from 0 to 35 m (see Fig. S1 and Data Set S1). Drift survey NO₃⁻ concentrations were between those of the Transect survey CO and TZ. Chl a extended deeper in the water column during the Drift study than in the CO, but it did not exhibit the more pronounced subsurface maxima observed at TZ sites (Fig. 2A and Table 1; see also Data Set S1).

FIG 2.

Profiles from the OO, TZ, and CO locations sampled during the Transect survey and Drift study. Note different y-axis scales for OO casts (C20 and C19) from other plots and different x-axis scales for various organismal abundance plots. (A) NO₃⁻, NH₄⁺, and in vivo fluorescence (Chl a derived). (B) Prochlorococcus, Synechococcus, and photosynthetic picoeukaryote abundance by flow cytometry. (C) Ostreococcus clade OI, clade OII, Micromonas, and Bathycoccus 18S rRNA gene copies · ml⁻¹ by qPCR with error bars representing the standard deviation of technical triplicates. Ostreococcus clade OII was assayed but rarely detected. Note that Transect survey profiles for metatranscriptome sample sites are represented by C2, C42, and C20; outward-bound or westward-leg TZ casts are C10 and C8, while inward-bound sampling was C42, 8 days later.

TABLE 1.

Environmental parameters in representative Transect survey and Drift study samples^a

Parameter	Transect survey (2009)b				Drift study (2010) (start to end of drift)
	OODCM	OOSURF	TZ	CO
Latitude	33.286	33.286	36.126	36.740	36.045 to 35.790
Longitude	−129.429	−129.429	−123.490	−121.02	−123.022 to −122.709
Depth (m)	105	15	10	10	25
Time of day	14:08	09:11	13:02	19:37	Diel
Date (mo/day)	10/7	10/6	10/11	10/2	9/16 to 9/18
NO3− (μM)	0.001	0.000	0.257	12.909	8.51 ± 1.87
NH4+ (μM)	0.003	0.000	0.047	0.263	Not measured
Si(OH)4 (μM)	2.32	1.35	2.92	9.912	7.76 ± 2.00
PO43− (μM)	0.34	0.47	0.49	1.153	1.49 ± 0.19
Total Chl a (μg · liter−1)	0.75	0.10	1.79	2.12	0.85 ± 0.08
Synechococcus (cells · ml−1)c	159	2,797	126,858	21,512	111,300 ± 18,357
Prochlorococcus (cells · ml−1)c	82,502	156,236	135,649	1,938	35,969 ± 1,272
Total eukaryotes (cells · ml−1)c	3,674	1,086	28,137	10,574	64,824 ± 13,216

^aTransect data come from the depth/locations used for metatranscriptome sequencing. For the Drift study, ranges or averages are provided for samples corresponding to the 12 metatranscriptomes (30) reanalyzed here. See Data Set S1 in the supplemental material for all data.

^bTransect nutrients and Chl a were measured at approximately the same depth on a cast adjacent to that of metatranscriptome water collection.

^cFlow cytometry counts for the Drift study are the averages and standard deviations for casts C42 and C48, the 06:00 cast for which qPCR was performed.

Phytoplankton abundance.

Phytoplankton groups varied dramatically across the regimes investigated during the Transect survey. Small eukaryote abundance increased 2- to 3-fold between the OO surface and DCM, while the abundance of Synechococcus declined beneath 60 m and Prochlorococcus dominated numerically throughout the water columns (maximum of 195,118 cells · ml⁻¹) (Fig. 2B; see also Data Set S1 in the supplemental material). During TZ sampling on the westward leg of the cruise, abundance of small eukaryotes in the upper 40 m (e.g., C10, ranging from 15,336 to 19,646 cells · ml⁻¹) was an order of magnitude higher than in the OO (Table 1; Fig. 2B). Prochlorococcus, Synechococcus, and picoeukaryotes were more abundant in the TZ during the inward-bound leg, when metatranscriptome sampling was performed, than during the westward-bound leg. Small eukaryotes had a maximum of 28,137 cells · ml⁻¹ at 0 m. Prochlorococcus increased to 139,435 cells · ml⁻¹ (0 m) and was roughly equal to Synechococcus (117,037 cells · ml⁻¹, 0 m). In the CO, small eukaryotes and Synechococcus were abundant only in the upper 10 m, while few Prochlorococcus cells were present (Fig. 2B; see also Data Set S1). The Prochlorococcus cells that could be distinguish using cytometry amounted to <2,000 cells · ml⁻¹. Eukaryotes and Synechococcus were more abundant in the Drift study than in the Transect survey. For example, at the surface of Drift study profile C44, Synechococcus and eukaryotes were 241,697 cells · ml⁻¹ and 117,062 cells · ml⁻¹, respectively. All three phytoplankton groups extended deeper in the water column in the Drift study than in the CO (Fig. 2B), including Prochlorococcus populations which were easily definable in the former but not in the latter.

During the Transect survey, Prochlorococcus HLI and LLI ecotypes dominated the OO and TZ samples based on a previous report (32). Clade IV dominated Synechococcus in the OO, and clades I, V/VI/VII, EPC1, and EPC2 were present at 3% to 10% of TZ cyanobacterial 16S amplicons (32). However, knowledge on the abundance of small eukaryotic taxa is limited to Ostreococcus (10) in this region. We quantified the picoeukaryotic prasinophytes Bathycoccus, Micromonas, and Ostreococcus in the profiles using qPCR (Fig. 2C). Ostreococcus clade OII was detected only in the OO, with a maximum of 72 18S rRNA copies · ml⁻¹ at the DCM. Ostreococcus clade OI was not detected in the OO but was present in CO and TZ waters. Laboratory studies suggest that isolates from clades OI and OII correspond to differently light-adapted ecotypes (51–53). However, our data support prior field-based results which indicate that spatial factors may govern Ostreococcus ecotype distributions, with clade OI present in mesotrophic waters and clade OII in oligotrophic waters (10). Here, Ostreococcus clade OI abundances also varied between the westward leg (up to 4,375 ± 511 18S rRNA copies · ml⁻¹) and inward-bound (<1,000 18S rRNA copies · ml⁻¹) TZ samplings. Moreover, OI abundance increased with depth in C8 and C10 (westward leg; Fig. 2C). Maximum Ostreococcus clade OI abundance was observed at the surface during the Drift study (36,713 ± 1,485 18S rRNA copies · ml⁻¹).

The maximum Micromonas 18S rRNA copies · ml⁻¹ was observed at the CO (14,613 ± 196), with lower abundances in the Transect survey TZ (949 ± 170) and Drift study (8,080 ± 641). These values likely represent contributions from multiple Micromonas clades, since seven have been characterized to date (12) and several of these cooccur in the southern California bight (17). Micromonas and Ostreococcus clade OI were detected in the same 29 qPCR samples (41 were analyzed), and both were undetectable in the OO. The average water temperature for these 29 samples (which include both Transect survey and Drift study) was 14°C ± 2°C.

Unlike Micromonas or Ostreococcus clade OI and clade OII, Bathycoccus was detected in all 41 qPCR samples. Bathycoccus dominated the prasinophyte groups in the majority of Transect survey profiles and had the highest number of 18S rRNA copies · ml⁻¹ observed (21,368 ± 327, surface CO) (Fig. 2C). In the OO and TZ, maximum Bathycoccus abundance was near the DCM (e.g., 7,251 ± 289 18S rRNA copies · ml⁻¹, OO_DCM; and 6,722 ± 328 18S rRNA copies · ml⁻¹, westward-leg TZ, 40 m), whereas, in the CO and inward-bound TZ, as well as the Drift study, the maximum occurred at the surface (Fig. 2C). These distribution patterns are generally in accord with the deepening of the DCM offshore (Fig. 1B). During the Drift study, Bathycoccus reached 10,759 ± 436 18S rRNA copies · ml⁻¹ (0 m, C44), but Ostreococcus clade OI was the dominant prasinophyte (Fig. 2C).

Nine other quantitative studies on picoprasinophytes have enumerated more than one genus. In the coastal Mediterranean Sea, Bathycoccus was undetectable in autumn but contributed 12% to 20% of the total picoeukaryotic population in winter (absolute copies · ml⁻¹ were not provided) (54). Few (≤5%) Ostreococcus 18S rRNA copies · ml⁻¹ were recovered in either season in this study, which used different qPCR primers (54) than ours (10). Using tyramide signal amplification-fluorescent in situ hybridization (TSA-FISH) probes, Micromonas and Bathycoccus together accounted for 60% of the green algal cells in the Norwegian and Barents Seas in late summer, while Ostreococcus was not detected (21). Similarly, a TSA-FISH study demonstrated that prasinophytes dominate the autotrophic picoeukaryotic community in the English Channel, specifically Micromonas, Bathycoccus, and, to a lesser extent, Ostreococcus (averages of 45%, 8%, and 1% of picoeukaryotes, respectively) (55). At the Bermuda biological time series station (BATS), Bathycoccus, Ostreococcus clade OII, and Micromonas were present at 96, 2,273, and 1,344 18S rRNA copies · ml⁻¹, respectively, at 40 m in a profile collected during spring mixing (56). However, during summertime, only Bathycoccus remained detectable (∼100 18S rRNA copies · ml⁻¹ in the DCM). One difference between the Pacific OO region studied here and BATS is the PO₄³⁻ availability, which is below detection (<0.010 μM) at BATS in summer but is ∼0.36 μM in the Pacific OO surface waters (see Data Set S1 in the supplemental material). Perhaps the most similar study in terms of general environs is one at an upwelling influenced coastal site off Chile, which used TSA-FISH to quantify picoprasinophytes (23). This study found that Ostreococcus was more abundant than Micromonas and Bathycoccus in 17 of 23 samples analyzed over the year and, overall, was the most abundant member of the eukaryotic picophytoplankton.

Phytoplankton abundance in relation to environmental data.

The clearest patterns in our data were the lack of Ostreococcus clade OI and Micromonas in the OO and the absence of Ostreococcus clade OII at study sites inshore of the OO. In the CO, which had the shallowest photic zone of all the stations, maximum 18S rRNA copies · ml⁻¹ occurred at the surface for Bathycoccus, Ostreococcus clade OI, and Micromonas (Fig. 2A and C; see also Data Set S1 in the supplemental material). During westward-leg sampling of the TZ (C8, C10), the Bathycoccus, Ostreococcus clade OI, and Micromonas maxima were associated with increased NO₃⁻ and NH₄⁺ deeper in the water column (Fig. 2; see also Data Set S1). The inward-bound TZ (C42) had more homogenous Chl a between the surface and 40 m than the outward-bound samples and exhibited an increase in Bathycoccus and Ostreococcus clade OI 18S rRNA copies · ml⁻¹ at the surface compared to other depths, as seen in the CO and the Drift study (C44). Notably, NH₄⁺ was also higher (0.04 μM) at the surface in this inward-bound TZ cast than for the westward-leg TZ casts (e.g., 0.01 μM).

Patterns were investigated further by partitioning photic zone samples according to whether a taxon was detected by qPCR or not. It should be noted that our sample set was biased for mesotrophic and coastal waters more than for oligotrophic waters. We found that PO₄³⁻, NO₃⁻, and NH₄⁺ were significantly higher (P < 0.003) in samples containing Micromonas and Ostreococcus clade OI (which always cooccurred) than in samples in which these taxa were not detected. Median nutrient values for photic zone samples with Micromonas and Ostreococcus clade OI were 0.802 μM PO₄³⁻ and 4.903 μM NO₃⁻ (n = 27). For the photic zone samples without these two picoprasinophytes (n = 11), the median PO₄³⁻ was 0.433 μM and NO₃⁻ was below detection. Likewise, NH₄⁺ was higher for samples with these taxa (0.080 μM, n = 14) than for those without (0.002 μM, n = 12). Samples that contained Ostreococcus clade OII had significantly (P < 0.01) lower NH₄⁺ (median, 0.005 μM) than those with Micromonas and Ostreococcus clade OI (median, 0.080 μM), but other measured parameters (PO₄³⁻, NO₃⁻, temperature, and salinity) were not significantly different. Significant differences were not observed for any of these parameters for the samples containing Bathycoccus (all samples) and the subset that also contained Micromonas and Ostreococcus clade OI.

Two principal-component analyses (PCA) were performed on photic zone data. One was limited to samples that had NH₄⁺ data (n = 27; Fig. 3A), which was not measured during the Drift study. The other included all photic zone samples with nutrient measurements (n = 39) (Fig. 3B) apart from NH₄⁺. The first, second, and third components accounted for 34.1%, 29.2%, and 11.3% of the variance, respectively (74.6% cumulative variance), for samples with NH₄⁺ measurements and 32.3%, 26.6%, and 13.2% of the variance, respectively (72.1% cumulative variance), for the broader set. PCA scores for each station clustered loosely according to geographic location, with TZ samples falling along a gradient between OO and CO stations (Fig. 3A). Notably, Drift study samples showed the greatest variability but, overall, were closest to Transect survey CO samples. Ostreococcus clade OII and Prochlorococcus had distinct patterns from the other groups studied (i.e., Synechococcus, Ostreococcus OI, Bathycoccus, Micromonas, and small eukaryotes). The abundance of Ostreococcus OII showed a positive relationship with OO depth, while Prochlorococcus abundance was positively associated with higher temperatures in surface waters. The other phytoplankton groups were positively related to each other (Fig. 3B). NH₄⁺ and NO₂⁻ concentrations were also positively associated, likely reflecting ammonia oxidation by bacteria and/or archaea (31). Higher NH₄⁺ and NO₂⁻ standing stocks were associated with increased abundance of Synechococcus, Ostreococcus OI, Bathycoccus, Micromonas, and picoeukaryotes, while other inorganic nutrients (e.g., NO₃⁻) were less so. Overall, the PCA results highlight the association of cold, nutrient-rich, upwelling-influenced surface water near the coast with high abundances of Ostreococcus clade OI, Bathycoccus, Micromonas, and Synechococcus and highlight the transition to warmer, nutrient-poor California current waters containing Prochlorococcus, Bathycoccus, and Ostreococcus clade OII.

FIG 3.

Principal-component analysis (PCA) using available data from all cruises and stations. The points represent scaled PCA scores for observations, and the vectors indicate loadings. PCA, including NH₄⁺ (n = 27, 16 dimensions) (A) and PCA excluding NH₄⁺ (n = 39, 15 dimensions) (B). Samples are in red for the Transect survey OO (+), TZ (●), and CO (□) and black for the Drift study (★). In both analyses, the first component is driven primarily by the concentrations of macronutrients and their inverse relationships to Prochlorococcus and temperature, and the second component reflects differences in biology and chemistry linked to depth (or covarious characteristics) in the water column.

Among the picoprasinophytes we quantified, only Bathycoccus was present at all stations. This lack of biogeographical pattern might reflect the presence of multiple Bathycoccus ecotypes (which would both be amplified by our qPCR primer-probe set). Moreover, at the inward-bound TZ station, a Bathycoccus abundance peak was observed at the surface and a second slight peak was present below 40 m (Fig. 2C), as might be expected if the ecotypes were adapted for different depth-associated parameters, such as light or nutrients.

Bathycoccus ecotype genetic distances.

We sought to test whether the omnipresence of Bathycoccus in the euphotic zone reflected success of a single cosmopolitan species, B. prasinos, or if it reflected differential contributions from more than one ecotype. First, we needed to establish genetic distances between putative Bathycoccus ecotypes (18, 20) to enable their identification in -omics field data. Because a definition of picoeukaryote ecotype-level divergence does not exist, we first analyzed three Ostreococcus genome sequences to establish a benchmark for interpreting Bathycoccus genetic distances. The available Ostreococcus genomes come from species that belong to Ostreococcus clades for which distributions have been characterized in parts of the Pacific and Atlantic Oceans (8, 10). The Ostreococcus clade OI qPCR primer-probe set used here was designed to amplify both O. lucimarinus and O. tauri (10) to distinguish them from Ostreococcus clade OII. However, O. lucimarinus and O. tauri share only ∼90% of their protein-encoding genes (14) and form separate clades using 18S rRNA gene sequences or other markers (12); hence, we have termed O. tauri clade OI/C according to reference 16. A comparative genome analysis has not been published that includes Ostreococcus sp. RCC809 (clade OII), but, as shown here (Fig. 2) and previously (10), OI and OII have distinct biogeographies that support clear niche differentiation. Thus far, cooccurrence of clades OI and OII has only been observed where Atlantic continental shelf waters intersect with the Gulf Stream current (10). We identified 3,212 single-copy homologs present in O. lucimarinus, O. tauri, and Ostreococcus sp. RCC809 that could be used to define nucleotide distances between these three species. Our analyses showed that these homologs had, on average, 72% to 75% nucleotide identity (Fig. 4A and C; see also Data Set S2 in the supplemental material) and were significantly different (P < 0.001).

FIG 4.

Bathycoccus and Ostreococcus ecotypes defined by homolog cluster analysis and natural distributions. (A and B) Percent nucleotide identity distributions for 3,212 single copy homologs in the three genome-sequenced Ostreococcus species (A) and 1,104 single copy homologs in the B. prasinos genome and Bathycoccus targeted metagenomes (B) (15, 18, 20). The black dots represent distribution outliers. Note the difference in y-axis scales. The divergence across homologous genes observed here allows differentiation of the Ostreococcus and the Bathycoccus ecotypes. (C and D) Ostreococcus (C) and Bathycoccus (D) average nucleotide identities based on the homolog groups identified in genomes and prior Bathycoccus targeted metagenomes (15, 19, 20). (E and F) Ecomarker analysis for Ostreococcus (E) and Bathycoccus (F) ecotypes in the Transect survey and Drift study metatranscriptomes.

The same type of analysis was performed for available Bathycoccus genome-level data. In this case, 1,104 homologous proteins were identified in the B. prasinos Bban7 genome (15) and the Bathycoccus targeted metagenomes from the tropical Atlantic (19) and coastal Chile (20) (see Data Set S2 in the supplemental material). The genes encoding these proteins in the coastal Chilean metagenomes (T142 and T149; sorted from a well-mixed coastal water column, 2 days and 15 km apart, at 5 and 30 m, respectively) have on average 97% ± 4% nucleotide identity, and their identities are not statistically different (P > 0.05) from B. prasinos, which was isolated in the coastal Mediterranean (Fig. 4B and D). Thus, we concluded that the coastal Chilean metagenomes and B. prasinos were reflective of a single ecotype, rather than multiple species or ecotypes; here, these three are considered representatives of the clade BI ecotype. The tropical Atlantic Bathycoccus metagenome came from warm oligotrophic waters and had lower identity to B. prasinos (82% ± 6%; P < 0.001) and to the coastal Chilean metagenomes (Fig. 4B and D). Thus, while the tropical Atlantic Bathycoccus metagenome and B. prasinos shared higher identity than the three Ostreococcus species, their divergence was significant and notable (Fig. 4); hence, we termed it the clade BII ecotype. The conclusion that two Bathycoccus ecotypes exist is supported by phylogenetic analysis of their internal transcribed spacer (18), a marker for which divergence has been related to speciation (57).

Taxon distributions in -omics data.

We analyzed the relative contributions of clades BI and BII (which could not be delineated by our qPCR primer-probe set) in samples from the Transect survey and the Drift study, as well as clades OI and OII (which were delineated using qPCR), to gain insight into their distributions. Metatranscriptomes were generated from surface samples for the CO and TZ as well as the OO surface and DCM (OO_SURF and OO_DCM) samples, with 1.4 ×10⁶ reads on average per sample (Table 2; see also Table S1 in the supplemental material). For the Drift study, 12 published metatranscriptomes were reanalyzed, because the initial publication (30) did not include reference genomes from Bathycoccus and several other phytoplankton. The samples were collected between 23 m to 25 m, which was either in the mixed layer or below (Fig. S1). Metatranscriptome reads from the Transect survey and Drift study were assigned to predicted proteomes available for the nuclear genomes of eukaryotic phytoplankton, including two nonmarine species which were used to recruit reads from green algae that are not represented in current genome databases. Bathycoccus recruited the most reads in the Transect survey (Table 2). In the Drift study, Ostreococcus clade OI recruited the most reads. Multiple Micromonas clades were present, and, when their reads were summed, they became the second and third most-represented genera in the CO and Drift metatranscriptome data, respectively (Table 2). Thus, the higher relative abundance of Ostreococcus clade OI in the Drift study and of clade OII in the OO, and the generally high abundance of Bathycoccus in all samples but the OO_SURF (Table 2), matched the patterns observed by qPCR at the corresponding depths (Fig. 2C).

TABLE 2.

Numbers of metatranscriptomic reads assigned to genome-sequenced eukaryotes by sample location and depth based on a BLASTX search against a database of genome-level information from eukaryotic algae and viruses^a

Eukaryotic species	Taxon (genome sequenced)	Predicted proteomeb	Transect survey finding				Drift
			OODCM	OOSURF	TZ	CO
Class II prasinophytes (Mamiellophyceae)	Bathycoccus prasinos Bban7	7,921	16,958	602	12,079	30,613	75,887
	Ostreococcus lucimarinus	7,605	451	473	1,604	12,437	180,370
	Ostreococcus tauri	7,987	403	534	528	968	15,392
	Ostreococcus RCC809	7,492	1,095	802	935	1,634	32,956
	Micromonas RCC299	10,109	982	1,223	1,413	11,388	30,984
	Micromonas CCMP2099c	19,316	2,648	2,955	1,831	4,495	12,923
	Micromonas CCMP1545	9,702	591	735	702	3,365	7,059
Class I prasinophytes	Pyramimonas parkeaec	20,299	3,493	3,871	3,391	2,666	15,939
Chlorophytes	Chlamydomonas reinhardtii	17,113	3402	3504	3908	3367	16,392
Unassigned green	Arabidopsis thaliana	35,386	3,503	4,862	4,126	3,951	25,117
Stramenopiles	Aureococcus anophagefferens	11,501	10,283	5,751	12,684	5,913	48,783
	Ectocarpus siliculosus	16,269	3,181	4,110	3,758	3,138	14,123
	Phaeodactylum tricornutum	10,417	1,271	1,851	1,521	6,417	8,220
	Thalassiosira pseudonana	11,673	1,431	2,819	2,984	9,392	16,886
Cryptophytes	Guillardia theta	24,840	2,433	2,935	2,260	4,711	11,185
Rhizarians	Bigelowiella natans	21,708	2,059	2,661	2,116	2,408	10,961
Haptophytes	Emiliania huxleyi	33,340	6,681	6,778	6,425	6,893	22,894

^aSee Materials and Methods; Drift study data from 12 samples in reference 30 were summed and are reported here as a single sample.

^bThe number of proteins predicted in genome projects for each species is shown as the predicted proteome.

^cThe predicted proteome for this taxon is based on assembled transcriptomes available at iMicrobe under project number CAM_P_0001089. Predicted protein counts are likely inflated due to assembly methods.

Other taxa were detected in the metatranscriptomic data as well (Table 2; see also Table S1B in the supplemental material). In the OO_SURF, haptophyte and pelagophyte sequences dominated (represented by Emiliania huxleyi and Aureococcus anophagefferens, respectively) (Table 2). Reads assigned to A. anophagefferens were present at several sites and likely came from the pelagophyte Pelagomonas calceolata, which can be abundant along line 67 (58). Finally, reads assigned to the diatom Thalassiosira pseudonana were notable in the CO (Table 2).

In a separate analysis of the metatranscriptomes, distributions of the Ostreococcus and Bathycoccus ecotypes were analyzed using information from the homolog analysis (Fig. 4A to D). First, we identified 112 genes in the homolog groups that were present in all the picoprasinophyte genomes and targeted metagenomes analyzed (Table S2; see also Data Set S2 in the supplemental material). These 112 genes are here termed ecomarkers. As seen in qPCR data, reads from O. lucimarinus (clade OI) ecomarkers dominated the Drift study and in the CO and TZ samples (Fig. 4E). Likewise, Ostreococcus clade OII was observed exclusively in the OO. Although some reads were assigned to O. tauri clade OI/C in the BLASTX analysis (Table 2; see also Table S1 in the supplemental material), none were assigned to O. tauri ecomarkers, reflecting the higher stringency of the nucleotide-based approach. The results provide deep-sequencing support for prior inferences from clone libraries that O. tauri is likely restricted to a few bays and lagoons, while O. lucimarinus is widespread in mesotrophic/coastal environments (8, 10).

Unlike results for Ostreococcus, ecomarkers from Bathycoccus clade BI (B. prasinos) and clade BII cooccurred in the CO, TZ, and OO_DCM samples (Fig. 4F). BII dominated (73%) the OO_DCM Bathycoccus signal, and BI reads were 27% of total. In contrast, only BI ecomarkers were detected in the Drift study. It should be noted that sample rarefication was not performed (59), and ecotype distributions across samples should be compared with that in mind, especially for samples in which few ecomarkers were recovered (see Table S2 in the supplemental material). We examined the observed patterns further by applying the ecomarkers to 2007 line 67 (cruise CN207) metagenomic data generated at the same time of year (19). In 2007, Ostreococcus ecotypes followed the same regional distributions observed in the ecomarker analysis of the metatranscriptomes (and qPCR), where clades OI and OII did not overlap and O. tauri was undetected. For Bathycoccus, the ecomarker analyses of metagenomic and metatranscriptomic results also agreed, although Bathycoccus was not detected in the 2007 OO_SURF. Remarkably, proportions of Bathycoccus ecotypes were stable between the two line 67 cruises (and the two data types), resulting in 76.2% ± 4.6% BII reads and 23.8% ± 4.5% in the OO_DCM, versus 99.5% ± 0.7% and 99.7% ± 0.5% BI reads in the TZ and CO, respectively. Thus, the high Bathycoccus abundances observed in our study do not appear to be from a single cosmopolitan type. Rather, the two Bathycoccus ecotypes appear to have differential distributions but with greater niche overlap than for the Ostreococcus ecotypes.

Exploratory gene expression analyses.

Exploratory gene expression analyses were performed for Bathycoccus and Ostreococcus in our unreplicated metatranscriptomes. These taxa were selected because ≥2 reads (on average) were mapped per predicted protein in the genome, at two or more sites (Table 2; see also Table S1 in supplemental material). In fitting with the cell cycle of phytoplankton and prior metatranscriptome results (60), 12 and 10 photosynthesis-related genes were in the top 100 Bathycoccus genes expressed in the TZ and OO_DCM (midday) samples, respectively. In the CO and Drift 22:00 (evening) samples, ribosomal proteins were prominent (see also Data Set S3 in the supplemental material). In the OO_DCM, 26 of the 100 genes recruiting the most Bathycoccus reads were of unknown function (Data Set S3 and Table S4). In the TZ and CO, 19 of the top 100 expressed were of unknown function; for the summed Drift metatranscriptomes, this number was 11. Ostreococcus genes recruiting the highest percentage of reads in a sample had functions similar to those from Bathycoccus in CO and Drift data (Fig. 5; see also Data Set S4), but Ostreococcus reads were too few to be analyzed in the TZ and OO_DCM.

FIG 5.

Highly expressed Bathycoccus proteins of known function that recruited ≥0.1% of OO_DCM Bathycoccus reads and comparison to the percentage of reads recruited at other sites. Protein identification numbers (ORCAE database) (ID) and InterPro Scan assigned functions (analyzed here) are provided. Only reads recruited to the Bathycoccus prasinos protein set (from the sequenced genome) were analyzed. The five shades of blue represent read recruitment percentages in categories (dark to light) of ≥1.00%, <1.00% to ≥0.100%, <0.100% to ≥0.010%, <0.010% to ≥0.001%, and <0.001% to 0.000%.

We assessed Bathycoccus genes that might help to understand acclimation to habitat-specific factors, especially in the OO_DCM (see Materials and Methods). Fourteen known-function and 14 unknown-function Bathycoccus genes with ≥0.01% read levels in the OO_DCM were not detected elsewhere (see Table S3 in the supplemental material). Further, 147 Bathycoccus genes each received ≥0.1% OO_DCM reads, of which 112 had InterPro-assigned (61) functions (Fig. 5; see also Data Set S3 in the supplemental material). Proteins recruiting ≥0.1% reads in the OO_DCM appeared to often have higher relative transcript abundance than in other samples (Fig. 5), including proteins involved in folding and translation (e.g., EF1B, HSP70, HSP20, and a family 10 glycosyl transferase) as well as a nickel superoxide dismutase (NiSOD). Four SOD isozymes are known, each binding different metals (Mn, Fe, Ni, and Cu/Zn) but catalyzing the same dismutation reaction. SODs provide protection from reactive oxygen species (ROS) that are formed during photosynthesis, under stress and during other cellular processes (62). To interpret this finding, we first annotated SODs in the B. prasinos and Ostreococcus sp. RCC809 genomes and improved previous O. lucimarinus, O. tauri, M. pusilla CCMP1545, and Micromonas sp. RCC299 annotations (13, 14). We identified all four SOD metalloforms in the class II prasinophyte genomes except O. tauri. This includes previously overlooked NiSODs and different Cu/ZnSODs which formed three homolog groups Cu/ZnSOD1, Cu/ZnSOD2, and Cu/ZnSOD3. In the metatranscriptome data, the Bathycoccus Mn/FeSOD was represented at equivalent levels (≥0.01% category) (Fig. S2) at all Transect sites but appeared to be less abundant in the Drift study. Results in the CO and the Drift study for Ostreococcus mirrored those of Bathycoccus (Fig. S2). In contrast, Bathycoccus Cu/ZnSOD2 and Cu/ZnSOD3 were found in the CO and the Drift study samples but were not detected in the OO_DCM and the TZ (Fig. S2). Although these results must be interpreted with caution, given that the OO_DCM and TZ samples had lower Bathycoccus read counts than the CO and summed Drift samples, similar field results have been reported for diatom Cu/ZnSODs (63). Moreover, the Bathycoccus NiSOD was close to 10-fold higher at the OO_DCM than at other sites and had higher relative read counts (again ∼10-fold) than the other SOD metalloforms. We quantified Fe in the 2007 transect cruise (CN207) and found that, above 100 m, it declined from ∼200 pM in TZ stations to between 50 and 90 pM in the OO (Data Set S1 and Fig. S3). These results point to Ni acquisition and NiSOD usage as a strategy for avoiding Fe dependence (and potentially Mn and Cu dependence) under limiting conditions.

The study region encompassed a strong gradient in photic zone NO₃⁻ and NH₄⁺ concentrations (Table 1; Fig. 2A). We annotated ammonium and nitrate transporters in B. prasinos and Ostreococcus sp. RCC809 (clade OII), because they had not been characterized previously (see Table S5 in the supplemental material). The six and four NH₄⁺ transporters (AMTs) in B. prasinos and O. lucimarinus, respectively, had read percentages >0.01% in most samples, and two showed expression in the ≥0.1% category (see Fig. S4 in the supplemental material). We then performed a phylogenetic analysis of the AMTs. This showed that four Bathycoccus and three Ostreococcus sp. RCC809 AMT genes are present in single copies and branch in clades containing known AMTs from M. pusilla CCMP1545, Micromonas sp. RCC299, O. lucimarinus, and O. tauri (47) (Fig. S5). The two other Bathycoccus AMT1 proteins (AMT1.3a, GenBank accession number CCO19599; AMT1.3b, GenBank accession number CCO17111) form a sister group to Micromonas sp. RCC299 AMT1.3, for which close homologs were unknown (47). The overall AMT count in the B. prasinos genome is higher than in other genome-sequenced class II prasinophytes (Fig. S5). Expression of AMT1.2, AMT2.1 (plastid targeted), and AMT2.2 was low or undetectable in the OO_DCM, while AMT1.1, AMT1.3a, and AMT1.3b generally contributed >0.02% of total Bathycoccus reads across the regimes (Fig. S4) (note, again, there were lower overall Bathycoccus reads in the OO_DCM and TZ than in the CO and summed Drift, which could influence reliability of results from genes with low read counts in the OO_DCM and TZ). The latter two transporters showed read percentages of 0.026 ± 0.015 across the sites, indicating use under a variety of nitrogen concentrations (Table 1; see also Data Set S1 in the supplemental material). AMT2.2 relative read percentages were higher in the Drift study (<0.100% to ≥0.010% category) and the CO (≥0.100% category), whereas NH₄⁺ concentrations were higher than the OO_DCM and TZ, in which Bathycoccus AMT2.2 percentages were lower (0.006% and <0.001%, respectively) (Fig. S4; see also Data Set S1). Thus, AMT2.2 is likely a low-affinity transporter that could serve as an indicator of organismal thresholds with respect to high NH₄⁺ availability.

We identified a putative low-affinity NO₃⁻ permease (NRT1) and a high-affinity NO₃⁻ transporter (NRT2) in the B. prasinos and Ostreococcus sp. RCC809 genomes by homology (see Table S5 in the supplemental material). Affinity assignments for class II prasinophyte transporters have thus far been based only on homology to verified transporters in, e.g., Chlamydomonas reinhardtii and land plants. Here, Bathycoccus NRT1 and NRT2.1 expression was observed in all samples except the CO (see Fig. S4 in the supplemental material). However, relative read percentages for these genes were low (i.e., in the <0.010% to ≥0.001% category) (Fig. S4), except at the OO_DCM, where the NRT2.1 read percentage was 0.029% and the concentration of NO₃⁻ was two orders of magnitude lower (<0.010 μM) than at other stations (Table 1). Ostreococcus NRT2.1 showed no expression in the CO and the Drift study. Thus, we surmise that NRT2.1 is a high-affinity NO₃⁻ transporter and provides a useful indicator of organism thresholds for low nitrate availability.

Conclusions.

Picoprasinophytes are often abundant in coastal zones, and we show that Bathycoccus, Micromonas, and Ostreococcus clade OI can be abundant in the eastern North Pacific mesotrophic transition zone. Overall, Ostreococcus clade OI showed high 18S gene counts in the upwelling-influenced and coastal waters, but not elsewhere. Micromonas and Ostreococcus clade OI cooccurred. Bathycoccus was present at all sites and abundant even at the DCM in the oligotrophic ocean. However, the Bathycoccus cells enumerated by qPCR do not come from B. prasinos alone but, rather, are from two ecotypes, named here BI and BII, based on the new ecomarker methodology we established for analyzing these groups. Prior reported sequence variations in Chilean Bathycoccus targeted metagenomes were initially thought to represent other ecotypes but likely represent single nucleotide polymorphisms (frequency of <3%) present in the BI ecotype population. Furthermore, ecomarker analysis demonstrated that O. lucimarinus-like strains dominate the clade OI signal and that clade OI/C (O. tauri-like) is not present. Similar to previous studies, the Ostreococcus clade OI and clade OII ecotypes showed distinct spatial distributions related to system nutrient levels (10). The Bathycoccus ecotypes overlap more frequently and are present over larger gradients in NH₄⁺ and NO₃⁻ concentrations, extending to the oligotrophic ocean. Our studies indicate that NRT2.1 is a high-affinity nitrate transporter and that AMT2.2 is a low-affinity ammonium transporter in the class II prasinophytes. Two additional AMT transporters in Bathycoccus that are absent from its closest relative (Ostreococcus) appear to be persistently expressed in the samples we studied that extended to the open ocean. Collectively, these results provide targets for future field analyses and testable hypotheses for the taxa that are in culture.