Introduction

Biological nitrogen (N2) fixation is a major source of nitrogen in the oligotrophic oceans (Capone, 2000; Carpenter, 1983; Karl et al., 2002). N2 fixation by unicellular cyanobacteria, including Crocosphaera watsonii, can account for a large fraction (>50%) of total N2 fixation in oligotrophic waters (Zehr et al., 2001; Montoya et al., 2004; Zehr et al., 2007b). These microorganisms are keystone taxa in oligotrophic surface waters, because they are typically outnumbered by co-occurring nondiazotrophic microorganisms by 103- to 105-fold, yet support the nitrogen demand of nitrogen-starved microbial food webs (Zehr et al., 2007b). C. watsonii is widely distributed in tropical surface waters of the northwestern Atlantic and Pacific oceans, where it is typically present at abundances <103 cells per ml (Church et al., 2005; Hewson et al., 2007; Zehr et al., 2007b). Although blooms of large colonial (for example, Trichodesmium) (Capone et al., 1997; Carpenter, 1983) and symbiotic (Carpenter et al., 1999) diazotrophs have been reported in tropical waters worldwide (Capone et al., 1997), unicellular cyanobacteria are less easily observed (Zehr et al., 2008), so little is known about the variability in abundance of these microorganisms.

Studies of C. watsonii gene expression in natural populations have been limited to a handful of proteins implicated in iron stress (iron-deficiency-induced protein A, idiA) (Webb et al., 2001) and inorganic nutrient acquisition (Dyhrman and Haley, 2006; Zehr and Turner, 2001). Whole genome transcript profiling by RNA oligonucleotide microarrays (de Saizieu et al., 1998) allows observation of genome-wide gene transcription patterns in cultivated microorganisms; however, it may be unsuitable for natural mixed-taxon assemblages.

Random community transcript (mRNA) sequencing (hereafter referred to as metatranscriptomics) offers promise to capture snapshots of active transcripts under in situ conditions. Recent applications of metatranscriptomics revealed differences in community gene expression in response to tidal and diel cycles (Frias-Lopez et al., 2008; Poretsky et al., 2005; Poretsky et al., in press) and in response to elevated CO2 concentrations (Gilbert et al., 2008). These studies revealed a large component of sequences that are of unknown origin (that is, no hits to gene databases), including possibly intergenic spacer-encoded riboswitches and termination factors (Frias-Lopez et al., 2008; Gilbert et al., 2008; Poretsky et al., 2005; Poretsky et al., in press). Because sequencing is random, most sequences obtained in metagenomic and metatranscriptomic studies are from abundant taxa, and rarer organisms, such as C. watsonii, may be absent altogether (Johnston et al., 2005). In most metagenomic sequencing efforts (Venter et al., 2004), microdiversity of dominant prokaryotic groups (Brown and Fuhrman, 2005; Garcia-Martinez and Rodriguez-Valera, 2000) complicates assembly and taxonomic identification of short genomic fragments. C. watsonii has an unusually high degree of genome conservation, with similar nucleotide composition between strains isolated two decades ago and bacterial artificial chromosome-cloned genomic fragments obtained from station ALOHA in 2005 (Zehr et al., 2007a). Because C. watsonii has a high degree of genome conservation, metagenomic and metatranscriptomic studies are relatively simple as sequence reads identical to the C. watsonii WH8501 genome can be isolated even among complex read libraries.

The objective of this study was to analyze the in situ gene expression to understand the ecophysiology of the N2-fixing unicellular cyanobacterium C. watsonii. We applied metatranscriptomics to capture snapshots of microbial community expressed gene transcripts under in situ conditions at a sampling site in the South Pacific that had high C. watsonii abundances. We identified several C. watsonii transcripts that were present in large numbers in the metatranscriptome, and developed quantitative reverse transcriptase polymerase chain reaction (qRT-PCR) assays to verify the in situ transcription pattern of these highly expressed genes.

Methods

Sampling location and station characteristics

Samples were collected on board the R/V Kilo Moana at station KM0703.025 (14°59.9′S, 175°0′E) on 13 April 2007. Additional samples were obtained at seven stations in the southwest Pacific and equatorial Atlantic oceans in June 2006 and April–May 2007 (Table 1). Station KM0703.025 seawater samples were collected from two CTD casts. The first cast, which sampled depths at 5, 30, 45, 75 and 106 m (deep chlorophyll maximum depth), was completed at 1246 hours (local time). The second cast to sample a subsurface bloom of C. watsonii with peak abundance at 37 m occurred at 1425 hours. Sunrise occurred at 0525 hours and sunset at 1838 hours. Sampling times for other stations are indicated in Table 1.

Table 1 Locations and sampling dates and times for samples collected at stations away from the Crocosphaera watsonii bloom
Full size table

Sample collection and nucleic acid recovery

Station KM0703.025 seawater samples were collected using 8 × 10 l Niskin bottles that were triggered at the sampling depth and returned to the research vessel. For the first cast (deepest sample collected at 1226 hours and CTD retrieved at 1246 hours), small samples (3–4 l) for analysis of gene and transcript abundance over a vertical profile were taken. Four samples were collected for daytime expression from each depth (5, 30, 45, 75 and 106 m) into acid-washed and seawater-rinsed 4.5 l polycarbonate bottles and processed by 1320 hours. Four additional 4.5 l bottles were filled with seawater from each depth for nighttime expression. The night expression bottles were placed immediately into an on-deck flow-through incubator where they remained at ambient temperature before sample processing at 0025 hours on 14 April. Water for DNA and RNA samples was filtered using peristaltic-pump-driven positive-pressure filtration serially through 10 μm polycarbonate (GE Osmonics, Trevose, PA, USA) and 0.2 μm Supor (Pall Gelman, Ann Arbor, MI, USA) filters. For RNA, 3 l were filtered, and the filters were placed immediately into microcentrifuge tubes containing 100 μl glass beads and 350 μl RLT buffer (RNEasy kit, Qiagen, Valencia, CA, USA). Samples for DNA (4 l) were collected from all depths, filtered per the RNA samples and placed into microcentrifuge tubes containing 200 μl TE buffer. The filters were then frozen in liquid nitrogen for transport to the laboratory at the University of California Santa Cruz.

For the second cast (samples collected at 1420 hours and CTD retrieved at 1425 hours; 37 m), water samples for metatranscriptome sampling were transferred from the Niskin bottles into acid-washed and seawater-rinsed 60 l polypropylene and 40 l polycarbonate carboys. The 40 l carboy was immediately transferred to an on-deck flow-through incubator for nighttime expression analysis. Water in the 60 l carboy was serially filtered immediately by air-driven positive-pressure filtration (<2 kPa) through a 142-mm-diameter 5 μm pore-size filter (Isopore; Millipore, Billerica, MA, USA) then through a 0.2 μm pore-size Durapore (Millipore) filter. The filters were housed in 142-mm-diameter polycarbonate filter holders (Geotech Environmental, Denver, CO, USA) and kept cool during the filtration, which took approximately 30 min. The incubated 40 l carboy sample was processed starting at 2145 hours in the same manner as the day sample. During filtration, the bottle and filter holders were kept in darkness. Filtration ended at 2205 hours. After filtration the filters were placed into 50 ml conical tubes (BD Falcon, Franklin Lakes, NJ, USA) and frozen immediately in liquid nitrogen for transport to the laboratory. Samples from the southwest Pacific and equatorial Atlantic oceans were processed similarly. Smaller samples (1.8 l for RNA and 4 l for DNA) were also collected during the 1403-hour cast and filtered for DNA and RNA as described above for the 1215-hour cast.

Particle concentrations and dissolved organic carbon and nitrogen measurements

Particle concentrations were inferred from particulate backscatter measurements at 650 nm (collected with a WET Labs Inc. (Philomath, OR, USA) hyperspectral absorbance and attenuation meter (ac-s)). All profiles were corrected for absorbance and attenuation of 0.2 μm filtered seawater profiles and bbp650 was calculated as the difference of cp650 and ap650 as per Jonasz and Fournier (2007). Dissolved organic carbon and nitrogen samples collected from standard depths at each station were measured by high-temperature combustion methods described by Carlson et al. (1998).

Microscopy of Crocosphaera-like cells

Samples for microscopy of Crocosphaera-like cells were taken on both casts at the same depths as small-volume RNA samples. Two seawater-rinsed 50 ml conical tubes were filled directly from Niskin bottles, and preserved immediately with 2% (final concentration) 0.02-μm-filtered formaldehyde. The samples were kept at 4 °C before processing. Subsamples (40 ml each depth) were filtered through black 1.0 μm pore-size polycarbonate filters (Whatman) and mounted sample-side up on glass slides. The filters were then mounted with a drop of immersion oil and covered with a coverslip. Duplicate filters were prepared from independent samples at each depth. The slides were kept at −20 °C before microscopic examination.

Slides were examined under × 40 magnification using green light excitation with a Zeiss Axioplan epifluorescence microscope. Crocosphaera-like cells were identified as unicellular cyanobacteria between 2 and 5 μm in diameter and by phycoerythrin fluorescence. Two hundred cells were counted in 20–40 fields (depending on cell density) on each replicate slide.

Microscopy of virus-like particles and bacteria

Samples (50 ml) of seawater from 0, 37, 75 and 106 m were collected in sterile centrifuge tubes and immediately amended with 2%, 0.02-μm-filtered formalin. The samples were processed immediately for SYBR Green I microscopy following established protocols (Noble and Fuhrman, 1998; Patel et al., 2007), using 1–3 ml of preserved sample in the slide preparation. Slides were observed using epifluorescence microscopy under blue excitation at × 100 magnification. More than 200 total virus-like particles and bacteria were counted in 20 microscopic fields for each slide.

DNA extraction

DNA was extracted from filters using a protocol modified from the DNEasy plant kit (Qiagen). DNA filters in 200 μl TE buffer were first submerged in a dry ice–ethanol bath to freeze. The tubes were then amended with buffer AP1 (Qiagen) and replaced in the dry ice–ethanol bath for 1 min before transferring to a heat block at 70 °C for 2 min. This freeze-thaw was repeated three times. After freeze-thaw, the tubes were bead-beaten in the BioSpec Products Mini Beadbeater for 2 min, followed by brief (pulse) centrifugation in a benchtop microcentrifuge. Afterward, 45 μl Proteinase K (supplied in the DNEasy plant kit; Qiagen) was added to each reaction and tubes were incubated at 55 °C for 1 h in a hybridization oven. The tubes were then amended with 4 μl RNase A (supplied in the DNEasy plant kit; Qiagen) and incubated at 65 °C for 10 min. After incubation, the filters were removed from the tubes using sterile steel needles, 130 μl of buffer AP2 (Qiagen) added and the samples vortexed and placed on ice. The tubes were then centrifuged at 14 000 g for 2 min, and the supernatant was processed per the manufacturer's protocol. The final extraction volume was 50 μl.

RNA extraction

Small-volume samples for quantitative PCR (qPCR) were extracted using RNEasy kits (Qiagen) according to manufacturer's protocols. Genomic DNA (gDNA) contamination in the small-volume extracts was eliminated by treatment with the Turbo DNA-free kit (Applied Biosystems, Austin, TX, USA) according to manufacturer's recommendations.

RNA for metatranscriptomic analysis was extracted from filters using the RNA Mini Isolation Kit II (Zymo Research, Orange, CA, USA), with the following modifications: 6 ml ZR buffer (Zymo Research) was added to the filters in centrifuge tubes while the filters remained frozen. Glass Beads (0.1 mm diameter, 100 μl total) were also added to the centrifuge tubes, and tubes were vortexed to ensure coverage of the filter by buffer. The tubes were then bead-beaten for 2 min, after which the filters were mechanically homogenized using sterile disposable serological pipettes for a further 2 min. After votexing the samples again, the tubes were centrifuged to remove filter debris from the buffer. The ZR buffer supernatant was then passed through the spin columns provided in the kit, and then the extraction followed manufacturer's recommendations. Contaminating gDNA was removed from extracts using the Zymo DNA-free RNA kit.

Metatranscriptomic sample processing

DNA-free total RNA for metatranscriptomic analysis was first subjected to terminator exonuclease using the mRNA-Only Kit (Epicentre, Madison, WI, USA), which was stopped by the addition of 1 μl kit-supplied stop solution. The resulting RNA mixture was subjected to a second mRNA purification protocol, using the MicrobExpress kit (Applied Biosystems), which employs capture hybridization by magnetic separation. At the conclusion of the mRNA-Only and MicrobExpress protocols, mRNA was depleted by 40–60% from total RNA as measured by spectrophotometry. The enriched mRNA sample was amplified by in vitro transcription using the MessageAmp II-Bacteria kit (Applied Biosystems) following the manufacturer's recommendation without amendment. The resulting amplified messenger RNA (aRNA) was 100–200 times the concentration of the mRNA after removing rRNA. The aRNA was converted to single-stranded (ss) cDNA using SuperScript III reverse transcriptase (Invitrogen, Carlsbad, CA, USA). The ss cDNA was then converted to double-stranded (ds) cDNA by treatment with Escherichia coli DNA polymerase I and RNase H. A total of 12 and 8 μg of ds cDNA was produced from the day and night samples, respectively, at the conclusion of processing.

Pyrosequencing

ds cDNA (5 μg) were subjected to picoliter reactor pyrosequencing at 454 Life Sciences, resulting in 123 686 individual reads. The average read length was 174 bp for day metatranscripts and 155 bp for night metatranscripts, representing 11.9 Mbp of day and 8.6 Mbp of night nucleotide sequence. Sequences from this study are deposited as libraries TA_35115 (day) and TA_35117 (night) under accession no CAM_P0000051 to the Community Cyberinfrastructure for Advanced Marine Microbial Ecology Research and Analysis (CAMERA).

Sequence analysis

Pyrosequences were compared against the draft C. watsonii WH8501 genome by BLASTn analysis (Altschul et al., 1997). Sequences with <99% similarity and E-values >10−20 to the genome contigs were discarded. This resulted in an effective sequence length threshold of 105 bp. The sequences with ⩾99% similarity to the C. watsonii genome were then compared by BLASTx analysis against the complete nonredundant (nr) protein database. Sequences with matches <95% on the amino-acid level to gene products were discarded from further analyses. Similarly, rRNA sequences were removed. The remaining non-rRNA sequences matching ⩾95% to proteins were assigned to metabolic pathways by BLASTx analysis against KEGG (Kanehisa et al., 2004).

Quantitative PCR and RT-PCR

Gene and transcript abundances of three previously unexamined genes (isiA, Transposase IS4 and Cwat_DRAFT_0061) and one previously examined gene (nifH) (Church et al., 2005) were estimated by qPCR and qRT-PCR. Primers and TaqMan hybridization probes were designed for the three unexamined genes using Primer3 software (Rozen and Skaletsky, 2000), and checked against the nr and env_nr database by BLASTn to identify possible cross-amplification with environmental genes. The primer and probe sequences used are given in Table 2. qPCR and qRT-PCR were conducted as described elsewhere (Church et al., 2005) for the four genes, using oligonucleotide standards for IS4 and Cwat_DRAFT_0061 (Table 2) and plasmid standards for isiA and nifH over eight orders of magnitude. The copy numbers of genes and transcripts were calculated as described elsewhere (Church et al., 2005). Extracted small-volume RNA was converted to ss cDNA using SuperScript III reverse transcriptase (Invitrogen) before use as template in qPCR (qRT-PCR), as described by Hewson et al. (2007). The PCR conditions for all four genes were identical to those used in previous studies of nifH (Church et al., 2005).

Table 2 Primer and TaqMan probe sequences used in qRT-PCR analysis of highly abundant metatranscripts of Crocosphaera watsonii WH8501
Full size table

Results and discussion

Characteristics of C. watsonii population

The mixed layer depth, as calculated per Lorbacher et al. (2006) was 25.9 m, where surface salinity was depressed by ∼0.75 relative to waters immediately below. Crocosphaera-like cell abundances at station KM0703.025 reached 3.84–4.70 × 103 cells per ml between 37 and 45 m depth (the C. watsonii-enriched horizon), but were much less abundant at the surface, at 75 m, and were not detected by microscopy at the depth of the deep chlorophyll maximum (106 m, see Figure 1). This was the highest abundance of C. watsonii observed anywhere to date (Campbell et al., 1997, 2005; Charpy, 2005; Church et al., 2005; Foster et al., 2007; Hewson et al., 2007; Zehr et al., 2007b). The C. watsonii-enriched horizon corresponded to distinct peaks of chlorophyll a fluorescence and total particle concentration. Thus, the observed bloom was not simply an increase in chlorophyll per cell, but was a localized enhancement of photosynthetic pigment and particles. This C. watsonii bloom feature was at the 27% light level just below the base of the mixed layer and was concomitant with the maximum percent oxygen saturation observed at this station. This indicates significant oxygen evolution rates coincident with the C. watsonii maxima, because photosynthetic picoeukaryotes and picocyanobacteria were at least an order of magnitude less abundant than C. watsonii (Figure 1).

Figure 1
figure 1

(a) Salinity and temperature profiles, (b) particle concentrations as estimated by particulate backscatter at 650 nm relative to fluorometric chlorophyll a measured by a CTD mounted Seapoint fluorometer, (c) dissolved organic carbon and dissolved organic nitrogen profiles and (d) the microscopically determined abundance of Crocosphaera watsonii-like cells relative to percent O2 saturation at station KM0703.025 (14°59.9′S, 175°0′E) on 13 April 2008. (d) The bloom depth horizon fluctuated between 37 and 45 m over the course of station occupation. Note that microscopy counts for 37 m were taken approximately 2 h after the other depths. The depth horizon of the Crocosphaera maximum coincides with dissolved organic carbon and nitrogen enrichments. Error bars on abundance=s.e.

Full size image

Total prokaryote and viral abundances were much higher than the abundance of C. watsonii and typical of oligotrophic open ocean conditions (Hewson and Fuhrman, 2006, 2007). Bacterial abundance was highest at the enriched depth, but the range of abundance at the depths sampled was relatively narrow (2.63–5.88 × 105 cells per ml). Viral abundance was highest at the C. watsonii-enriched horizon (1.29 × 107 viruses per ml), but was about an order of magnitude less abundant at the surface and deep chlorophyll maximum. The ratio of viruses to bacteria was higher in the enriched horizon (ratio 22) compared to surface waters (ratio 12) and the deep chlorophyll maximum (ratio 14).

Total prokaryote and viral abundance is typically greater in more productive conditions (Hewson et al., 2001, 2006). The large abundance of both prokaryotes and viruses in the C. watsonii-enriched horizon is consistent with an environment in which a greater amount of dissolved organic matter is present due to photosynthetic activity. In fact, measured dissolved organic carbon and nitrogen were enhanced at the C. watsonii maximum (Figure 1). The higher virus:prokaryote ratio within the bloom may be a consequence of either higher viral production by prokaryotes under a productive regime (Hewson et al., 2001) or production of viruses by the abundant C. watsonii. To date there have been no reports of viruses specific to C. watsonii; however, viral lysis has been described in other pelagic nondiazotrophic and diazotrophic cyanobacteria including Trichodesmium (Hewson et al., 2004; Ohki, 1999).

Metatranscriptome characteristics

At the station with a high abundance of C. watsonii (the ‘bloom’ station), 68 405 day and 55 281 night total RNA sequences were obtained, including 18 217 day and 12 105 night putative mRNAs that did not match rRNA libraries. The mean day mRNA read length was 173 bp and night 153 bp, with mean G+C contents of 46%. Of putative mRNAs, 672 day and 327 night sequences were >105 bp and were ⩾99% identical at the nucleotide level to the draft C. watsonii WH8501 genome (AADV02000000). An additional 53 day and 53 night sequences >105 bp were between 97% and 99% identical to the C. watsonii WH8501 genome sequence and were more similar to C. watsonii WH8501 than to other microorganisms; however, we did not include these in further analyses. Of ⩾99% C. watsonii WH8501 nucleotide sequence matches, 450 day and 148 night sequences (67% and 45%, respectively) shared homology with annotated nonribosomal genes of C. watsonii WH8501. The remainder of transcripts with C. watsonii WH8501 nucleotide matches did not share homology with protein-encoding regions. The largest number of nonprotein coding mRNAs (47 day and 29 night metatranscripts) matched C. watsonii contig 288 (of the September 2006 draft annotation) in a region between conserved hypothetical genes. The function of these intergenic region transcripts is unknown, however it is possible they represent rho-independent termination factors or riboswitches that are transcribed from intergenic regions beyond the protein-encoding region. These results are in line with previous studies showing large numbers of unidentified transcripts in surface seawater (Frias-Lopez et al., 2008; Gilbert et al., 2008).

Of the 5698 total protein-encoding genes in the C. watsonii WH8501 genome, 226 (4%) were detected in the combined day and night metatranscriptomes. The ratio of C. watsonii WH8501 nucleotide matches to total sequences (1.17% and 0.69% in day and night, respectively) is similar to the ratio of C. watsonii cell abundance to total prokaryote cell abundance. Gene length normalized frequencies (Frias-Lopez et al., 2008) demonstrated genome-wide diel expression of C. watsonii WH8501 genes, indicating that transcription of genes was not restricted to one part of the genome, or our sampling biases recovery of particular genome regions (Figure 2).

Figure 2
figure 2

Length normalized frequency of detection of Crocosphaera watsonii WH8501 genes. Each point represents the number of transcripts detected for each of the 5958 genes in the draft genome assembly. Solid symbols indicate transposases. Sequence reads that share >95% homology to C. watsonii proteins by BLASTx analysis are indicated.

Full size image

The low proportion of recovered C. watsonii WH8501 genes in the metatranscriptome indicates that sampling was undersaturated, which was a consequence of their low abundance relative to other microorganisms or due to differential expression of genes. A previous study of metatranscriptomes at station ALOHA indicated differential expression of Prochlorococcus genes (Frias-Lopez et al., 2008). Because Prochlorococcus is a dominant component of microbial assemblages in the open ocean, comparisons with a complementary metagenome were possible. However, for less-abundant components like C. watsonii WH8501, which comprise a small fraction of cell abundance and thus would have poor metagenomic coverage, such comparisons would require deep sequencing to obtain complimentary RNA and DNA frequency. Because we used a random survey approach, genes have equal probability of recovery. Hence, unless the genes are present in multiple copies in the C. watsonii WH8501 genome, comparison of transcript frequency in the random survey, in the absence of complementary gene frequency, allows identification of the most actively expressed genes.

The largest fraction of C. watsonii WH8501 transcript sequences obtained from the bloom station in the day (30%) and night (18%) was conserved hypothetical proteins or proteins of unknown function, which is in line with C. watsonii WH8501 transcripts recovered from other stations in the southwest Pacific and equatorial Atlantic oceans (30.7%), and previous metagenomic studies of microorganisms (Tyson et al., 2004; Venter et al., 2004; Rusch et al., 2007) and viruses (Angly et al., 2006; Breitbart et al., 2002, 2003). The most highly expressed C. watsonii hypothetical protein (Cwat_DRAFT_0061; ZP_00519306.1) was present in the bloom day (15% of transcripts) and night (5% of transcripts) metatranscriptomes. Cwat_DRAFT_0061 was also recovered at station ALOHA (Poretsky et al., in press) and at a station in the southwest Pacific (KM070304) (Supplementary Table 1). These highly expressed genes of unknown function may be attractive targets for expression studies because they appear to be important in the ecophysiology of C. watsonii.

The transcripts from the bloom with assigned KEGG pathways were predominately associated with genetic information processing and energy metabolism, with fewer sequences involved in carbohydrate, amino acid, lipid and cofactors and vitamins metabolism, and cell information processing and signaling (Figure 3). The C. watsonii WH8501 transcripts from energy metabolic genes included those associated with photosynthesis and N2 fixation.

Figure 3
figure 3

Metabolic KEGG pathway categories of Crocosphaera watsonii WH8501 genes recovered at station KM0703.025 at 37 m on 13 April 2008. The pie charts represent 442 day and 144 night sequences. The lower pie chart fragments show different energy metabolic pathways.

Full size image

Surprisingly, transposase transcripts (predominately IS605 and IS4) were among the most highly represented C. watsonii WH8501 transcripts of known function, representing 10.5% and 15.9% of C. watsonii WH8501 sequences in the day and night, respectively. This corresponded with a greater number of C. watsonii WH8501 recombination and repair orthologs (7.1–10.7% of all transcripts) than in the dominant sympatric cyanobacterium, Prochlorococcus marinus MIT9301 (2.0–2.7% of all transcripts), which lacks transposase genes altogether. Our results are consistent with a metatranscriptome sequencing effort from station ALOHA in the North Pacific Ocean, where 3 transposase sequences were recovered from 50 total C. watsonii WH8501 transcripts (Zehr et al., 2007a). Transposases were also recovered at other locations in the southwest Pacific Ocean and equatorial North Atlantic Ocean, but represented a smaller percentage of total C. watsonii WH8501 transcripts (Supplementary Table 1). At station KM0703.04 in the southwest Pacific Ocean, transposases represented 5.5% and 2.2% of C. watsonii WH8501 metatranscripts, respectively, whereas at station SJ0603.09 in the eastern portion of the equatorial North Atlantic Ocean, one transposase sequence was recovered among four total C. watsonii WH8501 sequences.

The abundance of C. watsonii transposase transcripts was surprising, as their function in C. watsonii ecology is unclear. It has been hypothesized that genetic diversity in C. watsonii may be maintained by genome rearrangement, rather than by gene sequence divergence because the diazotroph has low levels of genome DNA sequence variability between habitats and over time (Zehr et al., 2007a). High transposase activity in C. watsonii is similar to that found in populations of the archaeon Ferroplasma acidarmanus (Allen et al., 2007), and positive selection of transposases has been linked to increased fitness or invasion of new environments (Mes and Doeleman, 2006; Allen et al., 2007; Franguel et al., 2008). The C. watsonii WH8501 genome has a large number of transposases (419 transposases in the ∼6.3 Mb genome), some of which interrupt reading frames of genes (Zehr et al., 2007a) and may result in phenotypic diversity. The activity of transposases may be linked to gene modification or silencing. In other cyanobacteria, the activity of transposases is linked to gene modification or perhaps even gene silencing through insertions within protein-encoding regions (Mlouka et al., 2004). The biosynthetic investment in transposase expression suggests a vital function in the cyanobacterium for population fitness in oligotrophic environments.

Another possible function of transposase activity in C. watsonii is defense against lysogenic conversion, because enhanced genomic diversity results in lower frequency of insertion sites for phage integration. The effect of transposases may be similar to clustered regularly interspaced short palindromic repeats (CRISPRs) (Jansen et al., 2002), the frequency of which corresponds to phage attack frequency (Barrangou et al., 2007). CRISPRs have been implicated both as gene silencers in prokaryotes (similar to siRNAs in eukaryotes) and as a mechanism of viral resistance (Barrangou et al., 2007; Tyson and Banfield, 2008). The C. watsonii WH8501 genome contigs contain 1 confirmed (lying within a gene encoding for peptide synthetase; NP_486684.1) and 12 questionable CRISPR-like structures, more than found in common cultivated marine microorganisms (Grissa et al., 2007). One of the questionable structures shares sequence homology with heat-shock protein DnaJ, three sequences of which were retrieved as metatranscripts in the day sample. C. watsonii are typically rare and therefore are unlikely to have a multiplicity of infection congruent with lytic phage attack under non-bloom conditions (Wilcox and Fuhrman, 1994). However, Synechococcus in Tampa Bay demonstrate greater frequency of lysogenic conversion in summer months when abundances are greater (McDaniel et al., 2002). Thus, when C. watsonii abundance is elevated, transposase activity may be enhanced to defend against phage attack.

Another dominant transcript recovered was the photosystem I-associated iron-stress-induced protein A (isiA; annotated in genome as the photosystem II psbC-encoding COP43 homolog, but with highest amino-acid sequence similarity to isiA in Cyanothece and Trichodesmium). isiA is formed under prolonged iron stress in cultures of cyanobacteria (Laudenbach et al., 1988; Laudenbach and Straus, 1988) as a mechanism to prevent high-light-induced oxidative damage by increasing cyclical electron flow around the photosynthetic reaction center photosystem I (Michel and Pistorius, 2004; Ohkawa et al., 2000). isiA sequences were recovered from transcriptomes at other stations in the southwest Pacific Ocean and the eastern portion of the equatorial North Atlantic Ocean (Supplementary Table 1), suggesting that this transcript may be a useful marker for the physiological status of natural populations of C. watsonii and may be more sensitive than the idiA (Webb et al., 2001), transcripts of which were not detected in our libraries. The energy metabolic gene expression pattern of C. watsonii was different from the non-N2-fixing cyanobacterium P. marinus WH9301, which had a lower percentage of transcripts for nitrogen metabolism (16.7- to 27.6-fold higher in C. watsonii), but greater number of transcripts of genes in the glycolysis and gluconeogenesis pathway (1.3- to 2.3-fold lower in C. watsonii).

The large number of C. watsonii N2 fixation gene transcripts at the bloom station is consistent with the environmental conditions at the sampling location, where N2 fixation rates at the bloom depth horizon as measured by average 15N2 uptake over 24 h were 0.5–1.5 nmol l−1 h−1, accounting for 60% of the total areal rate at the station (Montoya JP, unpublished data). We recovered transcripts for most components of the C. watsonii nitrogenase gene cluster in the day and some at night in the bloom sequence libraries (Figure 4). Only a single nitrogenase transcript was found at another southwest Pacific station (KM0703.04) and none were found elsewhere. Nitrogenase activity has been measured predominately at night in cultures of C. watsonii WH8501, and C. watsonii WH8501-related cyanobacterial diazotroph nifH transcripts are in low abundance during the light phase (Church et al., 2005; Hewson et al., 2007). The disproportionate number of sequences observed between day and night metatranscriptomes prevents direct comparison of day and night expression levels. The abundance of nitrogenase gene cluster transcripts in the day may reflect constitutive expression of some parts of the cluster that do not directly lead to N2 fixation activity. In Cyanothece sp. ATCC 51142, differential expression of nitrogenase cluster genes occurs over diel cycles, with expression of some genes during the light, even though most genes are more highly expressed in the dark (Colon-Lopez and Sherman, 1998; Stoeckel et al., 2008; Toepel et al., 2008). The Fe–Mo structural protein nifK, which is encoded for by the nifHDK polycistronic mRNA (Chen et al., 1998), had almost as many transcripts represented as nifH, which is the gene commonly used to examine diazotroph abundance and nitrogenase expression in environmental samples (Church et al., 2005; Short and Zehr, 2007; Zehr et al., 2007b). The nifB gene, which forms part of the nifBQ cluster and is necessary for the processing of the Fe–Mo cofactor encoded by nifHDK (Roberts et al., 1978), was also highly represented among the daytime transcripts, in line with expression observed in Cyanothece sp. ATCC 51142 (Toepel et al., 2008). Finally, a hypothetical protein flanked by nifZ and nifP (Cwat_DRAFT_3843) had the greatest number of transcripts; however, its function in nitrogenase is unknown and it may be an attractive target for expression studies. These nitrogenase genes may provide more sensitive targets for study of environmental nitrogenase transcription in future studies. Because C. watsonii WH8501 also has the genomic potential to assimilate NH4+, and reduce NO3− and NO2−, the biosynthetic cost to C. watsonii of gene transcription for N2 acquisition at this station appears necessary to maintain high abundances under otherwise favorable but oligotrophic environmental conditions.

Figure 4
figure 4

Graphical interpretation of Crocosphaera watsonii WH8501 nitrogenase cluster gene length normalized metatranscript frequency in day (light bars) and night (dark bars). The nitrogenase cluster is on a single contig of the unfinished genome; the direction of arrows indicates the orientation of the gene in the cluster. The frequency of metatranscripts is indicated by the height of the bars above the cluster representation.

Full size image

Quantitative PCR and RT-PCR

Based on metatranscriptomic sequence frequency within libraries, we developed qPCR and qRT-PCR assays for Transposase IS4, iron-inducible stress protein A (isiA), and the conserved hypothetical protein similar to COP23 in Synechococcus (Cwat_DRAFT_0061). We applied these and the previously developed group B cyanobacterial nifH assay (of which C. watsonii WH8501 is a representative) (Church et al., 2005) to vertical profiles of microbial DNA and RNA collected during day and night phases at the bloom location. These genes were targeted because their transcript abundance suggested a key function in the physiology of C. watsonii in natural populations; however, there was no information on their expression levels over diel cycles.

The abundance of Crocosphaera-like cells at the bloom depth determined microscopically (Figure 1) and confirmed by flow-cytometric analysis (data not shown) was several orders of magnitude higher than gene abundance (Table 3) at each depth, which may be a result of microbial DNA extraction efficiency or the identification of cells as C. watsonii that were actually morphologically similar cyanobacteria. The qRT-PCR results of transcript abundance were consistent with observations of metatranscriptomic sequence frequency (Figure 5). Transcript abundance data normalized to cell abundance (determined microscopically) indicated that most transcripts were present as multiple copies per cell at the depth sampled for metatranscriptomics; however, at shallower and deeper depths the number of transcripts was less, often <1 transcript per cell (Figure 5). This is likely due to low RNA extraction efficiency, but may also indicate the presence of inactive cells within the assemblage, because inactive and active cells are not distinguished by microscopy. As expected, but consistent with the observed metatranscript frequency, the ratio of expression between day and night transcripts (Figure 5) was the least for nifH. This gene was previously shown to have strong circadian rhythm, with night expression typically 2–3 orders of magnitude greater than day expression (Church et al., 2005; Hewson et al., 2007; Zehr et al., 2007b).

Table 3 Cell abundance—normalized gene abundance of Fe–Mo protein of nitrogenase (nifH), conserved hypothetical protein Cwat_DRAFT_0061 (most similar to COP23 circadian-related protein in Synechococcus sp. PCC 8801), iron-stress-induced protein A (isiA), and Transposase IS4 at bloom station KM0703.25
Full size table
Figure 5
figure 5

Gene and transcript abundance of Fe–Mo protein of nitrogenase (nifH) (Church et al., 2005), iron-stress-induced protein A (isiA), conserved hypothetical protein Cwat_DRAFT_0061 and Transposase IS4, as measured by qPCR and qRT-PCR; Church et al., 2005). Each data point represents the mean of three analytical replicates; s.e., standard error. Cwat_DRAFT_0061 transcripts were not detected in day and night at 15 m, transcripts of Transposase IS4 were not detected at night at 15 m and nifH genes were not detected at 106 m. No night transcript samples were obtained from 30 m. Note that samples for 37 m were taken approximately 2 h after the other sample depths.

Full size image

The C. watsonii photosystem I-binding isiA iron-stress protein had the greatest variation from day to night at the bloom depth (Figure 5), with most transcripts found during the day. The opposite pattern was observed in shallower and deeper depths, suggesting a key function for isiA at the C. watsonii-enriched horizon. isiA protects the photosynthetic complex from superoxide damage during photosynthesis, hence the strong diel change is consistent with photosynthetic patterns. However, the observation of isiA genes at high copy number at the bloom depth was somewhat unexpected because the photosynthetically active radiation at 37 m was only 27% of surface irradiance and isiA transcripts were also observed in the night metatranscriptome. Furthermore, this expression of isiA was surprising because Fe concentrations in the region are typically higher than in oligotrophic waters elsewhere due to undercurrent flux originating near Papua New Guinea (Wells et al., 1999). This suggests that although isiA has strong diel cycle, it may have a function other than photosynthetic activity or Fe starvation in C. watsonii cells at the bloom depth.

Transposase IS4 demonstrated an ∼3-fold increase in transcript abundance determined by qRT-PCR at night at the bloom depth (Figure 5). In contrast, transposase transcript abundance at shallower and deeper depths was higher during day than night. While consistent with metatranscript frequency, the diel cycling of the gene is unclear. Most phage infection occurs at night in surface waters of the ocean (Winter et al., 2004), hence it is possible that some transposase activity may be related to phage defense. Because most DNA damage in prokaryotic cells occurs as a consequence of UV irradiation (Jeffrey and Mitchell, 1997), it is also possible that transposase activity may be related to genome maintenance during darkness. No information is available on the growth characteristics of C. watsonii during day–night cycles; however, it is possible that enhanced cell division at night may cause higher transposase activity. Regardless of the mechanism by which it occurs, the increased transcript number per cell ratio at depths with higher abundance indicates that transposase activity is linked to enhanced fitness and enables C. watsonii cells to form dense concentrations.

The strong diel changes in transcripts of Cwat_DRAFT_0061 at all depths (Figure 5) suggest that this protein is linked to diel cycles. Cwat_DRAFT_0061 is similar (39% identical on the amino-acid level) to a circadian oscillating polypeptide COP23 precursor of Synechococcus sp. PCC 8801/RF-1, a cell membrane protein that is stable in arrhythmic cells but degrades rapidly under light–dark cycles (Chen et al., 1996). Our observation of a higher number of day transcripts than night transcripts of Cwat_DRAFT_0061 in the bloom suggests that it may have similar function to Synechococcus sp. PCC 8801 COP23.

The time between sample collection and processing (∼45 min) may have resulted in degradation of some mRNAs (Andersson et al., 2006). Energy metabolic transcripts may be overrepresented, and putative enzymes and translation transcripts underrepresented in degraded metatranscriptomes (Selinger et al., 2003). Hence, we cannot discount the possibility that the highly expressed isiA and nifH genes are a consequence of mRNA degradation during sample processing.

Conclusion

The different metabolic pathway gene transcription pattern of C. watsonii compared to co-occurring cyanobacteria may indicate a unique strategy for maintaining its niche within the open ocean ecosystem. C. watsonii has genome expression patterns that reflect the nutrient-limited environment of the oligotrophic ocean that provides insight into highly expressed genes important for the microorganism under favorable (that is, bloom) conditions. Whereas the function of highly expressed transposase and conserved hypothetical proteins observed in the bloom and elsewhere in the open ocean is unknown, our results demonstrate that the transcription (that is, transcripts per gene copy) of these genes were enhanced in the bloom of C. watsonii (Figure 5). The function of transposases in the population biology and physiology of C. watsonii needs further study, and these genes may provide markers for cell growth or physiological stress in the environment. The heavy investment of C. watsonii cells in genes that respond directly to ambient nutrient availability suggests that nutrient acquisition is a dominant activity of the diazotroph. Finally, the genes observed under ambient conditions may only be observed using our in situ gene expression approach, because cultivation studies may demonstrate different gene expression in confinement and highlight the need for more targeted transcriptomic studies to understand the autecology of individual key microorganisms in nature.