Introduction

Photosynthetic microbial mats are ecosystems with high densities of functional and phylogenetic diversity, resulting in a strongly coupled cycling of elements (Canfield and Des Marais, 1993; Bebout et al., 1994; Ley et al., 2006). Therefore, these ecosystems have been studied extensively to gain fundamental insights into global processes with implications ranging from the early evolution of microorganisms and metabolic diversity to microbial interactions and nutrient cycling (Des Marais, 1990; Paerl et al., 2000; Des Marais, 2003). The cycling and flux of carbon, sulfur and nitrogen are particularly intertwined in these mats. The availability of nitrogen determines overall mat productivity, because photosynthetic primary production creates a high demand for fixed nitrogen, which requires high rates of N2 fixation (Herbert, 1999). However, the distribution of N2 fixation activity among phylogenetically and functionally diverse microorganisms in the mats remains largely unclear.

In most microbial mats, the uppermost layer is visually dominated by unicellular and nonheterocystous filamentous cyanobacteria, such as Microcoleus or Lyngbya spp. (D’Amelio et al., 1989). N2 fixation in these mats has a distinct diel pattern: N2 fixation is largely absent during the day, increases after sunset and is maximal at night or in the early morning prior to sunrise (Stal et al., 1984; Bebout et al., 1987, 1994; Omoregie et al., 2004a, 2004b). Daytime suppression of N2 fixation has been attributed to high levels of O2 from oxygenic photosynthesis, which leads to inhibition of the O2-sensitive nitrogenase (Bothe, 1982). The mats in the Elkhorn Slough estuary (Moss Landing, CA, USA), that were the focus of this study are dominated by filamentous cyanobacteria, in particular by Microcoleus spp. (Burow et al., 2011). However, the N2 fixation activity pattern and the diazotrophic community in Elkhorn Slough mats were so far unknown.

Traditionally, cyanobacteria were thought to have a major role in the total N2 fixation in microbial mats (Paerl et al., 1991; Bebout et al., 1993). A few cyanobacteria that were isolated from microbial mats fixed N2 in cultivation-based studies and corroborated this hypothesis (Stal and Krumbein, 1981; Paerl et al., 1991). However, concern of general cultivation bias has motivated researchers to pursue cultivation-independent methods to gain more comprehensive insight of the diazotrophic community in complex ecosystems (Zehr et al., 1995).

N2 fixation is mediated by the nitrogenase enzyme complex, consisting of nifH-encoded dinitrogenase reductase, which transfers electrons to dinitrogenase, encoded by nifD and nifK, ultimately catalyzing the reduction of N2 to NH3. This unique process has been studied at different levels using various methods such as acetylene reduction assay (ARA) and nifH surveys. The fortuitous transformation of acetylene to ethylene by nitrogenase makes the ARA a useful, indirect measure for nitrogenase activity in cultures as well as in complex communities (Stewart et al., 1967). The high sensitivity of the assay has enabled measurements of rapid changes in N2 fixation activity in response to rapidly changing environmental conditions. Employment of this technique in diel cycle studies of microbial mats has provided invaluable insights into the patterns of N2 fixation depending on irradiance and thus O2 concentrations, such as the above-mentioned patterns for Microcoleus- and Lyngbya spp.-dominated mats (Stal et al., 1984; Bebout et al., 1987; Bebout et al., 1994; Paerl et al., 1996). More recently, the nifH gene has been used as a phylogenetic and functional marker for N2 fixation and allows investigating the phylogenetic distribution of the genetic potential for N2 fixation in complex microbial communities. Surveys of nifH in microbial mats suggested that heterotrophic bacteria might also have an important role in microbial mat N2 fixation in addition to cyanobacteria (Zehr et al., 1995; Steppe et al., 1996; Zehr et al., 2003; Omoregie et al., 2004a; Severin et al., 2010; Severin and Stal, 2010b). Analysis and quantification of nifH transcripts have helped to identify the fraction of diazotrophs actively expressing this essential gene for N2 fixation and has given insights into gene-expression dynamics in the environment (Omoregie et al., 2004b; Moisander et al., 2006; Severin and Stal, 2010a). Previous studies (Steunou et al., 2008; Severin and Stal, 2010a) have revealed discrepancies between the expression of nifH by diazotrophic groups and nitrogenase activity patterns measured by acetylene reduction, illustrating that gene expression does not necessarily correspond to activity.

In contrast to the above-mentioned methods, stable isotope probing with 15N2 provides a direct and unambigious measure of N2 incorporation activity (Montoya et al., 1996). While 15N incorporation is measured in bulk by isotope ratio mass spectrometry (IRMS), secondary ion mass spectrometry (SIMS) and the recently developed CAMECA NanoSIMS for high-resolution SIMS have enabled the connection of ecosystem level processes to activities at the level of single cells. These technologies have been used for the stable isotope probing of the metabolic activities of cell aggregates (Orphan et al., 2001) or single cells (Lechene et al., 2006; Popa et al., 2007; Finzi-Hart et al., 2009; Ploug et al., 2010; Foster et al., 2011), respectively. In combination with fluorescence in situ hybridization (FISH) targeting 16S rRNA, SIMS studies enable direct linkages of phylogeny to function in natural communities (Orphan et al., 2001; Behrens et al., 2008; Li et al., 2008; Musat et al., 2008; Halm et al., 2009).

In this study, we used (to the best of our knowledge) an unprecedented breadth of methods—nitrogenase activity measurements, analysis of nifH gene diversity and expression, 15N2 tracer experiments, NanoSIMS, catalyzed reporter deposition (CARD)-FISH and cultivation experiments—to identify active N2-fixing microorganisms in a complex microbial mat ecosystem. By this combined approach, we were able to characterize a novel group of diazotrophic cyanobacteria in Elkhorn Slough microbial mats, and demonstrated their ecophysiological importance in N2 fixation.

Materials and methods

Study site

The sampling site is located in the Elkhorn Slough estuary at 36°48′46.61″N and 121°47′4.89″W. The Elkhorn Slough is a shallow seasonal estuary that extends inland 11 km from Monterey Bay with mixed semidiurnal tides; tidal exchange and sporadic surface water input during winter rainy seasons are the main water transport mechanisms (Chapin and Johnsin, 2004).

Mat sampling and diel cycle studies setup

Microbial mats collected at Elkhorn Slough (10 pieces of ca. 144 cm2 of 2 cm thickness including a 1 cm sediment layer) were sampled on 20 October 2009 and transported to a greenhouse facility transparent to ultraviolet radiation at NASA Ames Research Center within 1–2 h. In the greenhouse, mat pieces were placed in acrylic aquaria transparent to ultraviolet radiation and covered with in situ water (circulated and aerated) for ca. 20 h before the beginning of a diel cycle study (starting at 1200 hours and ending at 1500 hours the following day). Two successive diel cycle studies with the same mats were carried out (21/22 and 23/24 October 2009) under natural solar irradiance, and the water temperature was kept constant at ca. 18 °C (in situ average).

Biogeochemical analysis (ARAs and 15N2 incubations)

Nitrogenase activity was measured with the ARA as previously described (Bebout et al., 1993). Mat cores (10 mm diameter, 10 mm thick) were sampled in triplicate every 3 h, and subsequently incubated with acetylene for 3 h. Triplicate water samples without mat served as negative controls. Ethylene was quantified in a Shimadzu GC-14A gas chromatograph (Shimadzu, Kyoto, Japan). For measuring the depth distribution of nitrogenase activity in the mats, triplicate mat cores of 10 mm diameter and 10 mm thickness were horizontally sectioned into three layers (uppermost layer 0–2 mm depth, second layer 3–6 mm and third layer of 7–10 mm depth) and the layers were separately incubated as mentioned above.

To measure 15N2 incorporation, mat cores of 10 mm diameter and 10 mm thickness were transferred to a 14 ml serum vial, covered with 1 ml of in situ water, capped with gas-tight rubber stoppers and 8 ml of the headspace was exchanged with 15N2 gas (98+ atom% 15N2; Cambridge Isotope Laboratories, Andover, MA, USA). Mats were incubated for 10 h in the dark (2030 hours until 0630 hours the next day), and subsequently, half of the mat cores were sectioned for bulk isotope analysis in the same depth intervals as mentioned above. The other portions of the sectioned cores were preserved for NanoSIMS analysis by fixation in 4% paraformaldehyde for ca. 16 h at 4 °C. Fixed cores were washed twice in 1 × phosphate-buffered saline (PBS) (pH 7.6) and stored in PBS/ethanol (40/60, vol/vol) at −20 °C for further analysis. Unlabeled mat sections served as controls. Isotope ratios for 15N/14N were determined by IRMS (ANCA-IRMS; PDZ Europa Limited, Crewe, England) at the University of California, Berkeley. Additionally, mat cores were first sectioned and then incubated with 15N2 to verify the depth distribution of N2 fixation by IRMS.

IRMS and ARA data of vertical sections were analyzed for statistical differences using an analysis of variance with a Tukey's HSD mean separation at P<0.05 using the R program version 2.13.1 (http://www.r-project.org/index.html).

Companion samples for nucleic acid extraction and subsequent molecular analysis (six independent cores of 10 mm diamater and 10 mm thickness per time point) were flash frozen in liquid nitrogen and stored at −80 °C.

Cultivation and 15N2 incubation experiments of Elkhorn Slough Filamentous Cyanobacterium-1 (ESFC-1)

Cyanobacteria were enriched from microbial mat samples as described previously (Prufert-Bebout and Garcia-Pichel, 1994). Media formulations and enrichment conditions are described in the Supplementary Information. After enrichment, the cyanobacterial cultures were assessed for N2 fixation activity after three washes in N-free media (ASN-) and pre-incubated in ASN- for 4 days. Cultures were then transferred into 14 ml serum vials, which were filled with fresh ASN- media to eliminate gas headspace. The vials were capped with gas-tight rubber stoppers and 35 μl of 15N2 was added, whereas an equal volume of medium was vented with a needle. Vials were incubated at 22 °C for 24 h (dark/light cycle, 8 h/16 h, light intensity ca. 40 μmol photons m−2 s−1). At the end of the incubation, the cyanobacterial biomass was rinsed in 1 × PBS and frozen at −80 °C for IRMS analysis.

Molecular analysis

Detailed information about DNA and RNA extractions can be found in Supplementary Information. Briefly, RNA and DNA of microbial mats were co-extracted from the uppermost 2 mm of three pooled mat cores by combining phenol–chloroform extraction with parts of the RNeasyMini and QIAamp DNA Mini Kit (Qiagen, Valencia, CA, USA), respectively. RNA was reverse transcribed into single-stranded cDNA using the SuperScript III First-Strand Synthesis System (Invitrogen, Carlsbad, CA, USA).

Detailed information about the construction of 16S rRNA gene/transcripts and nifH gene/transcript clone libraries can be found in the Supplementary Information. In summary, (1) a 16S rRNA gene clone library was constructed from the cyanobacterial enrichment culture ESFC-1 (total of 36 sequences). (2) general 16S rRNA clone libraries of the upper 2 mm of Elkhorn Slough mats were generated from mat samples collected on 12/13 January 2009. The following clone libraries were constructed: 12 January, 2100 (total of 329 sequences) and 13 January, 0700 hours (total of 243 sequences). (3) a nifH gene clone library was generated from the cyanobacterial enrichment culture ESFC-1 (total of 39 sequences) (4) clone libraries of the nifH genes (DNA) and transcripts (cDNA) were constructed from the uppermost 2 mm of mat cores sampled during two consecutive diels: 21 October 2009, at 2250 hours (DNA: 93 sequences and cDNA: 92 sequences) and 24 October 2009, at 0310 hours (DNA: 75 sequences and cDNA: 88 sequences). Sequence analyses of the clone libraries listed above are described in Supplementary Information.

Amplicons of the 16S rRNA V8 hypervariable region were constructed from seven time points in the year 2009 (13 January, 30 April, 1 July, 19 August, 16 September, 21 October and 13 November 2009) (see Supplementary Information for more detailed description). Analysis of the V8 amplicon sequences is described in Supplementary Information.

Sample preparation for NanoSIMS analysis

Material from the uppermost 2 mm of fixed microbial mats from the October 2009 samples (15N2-incubated and sampled for natural abundances) were transferred with tweezers onto silicon wafers (Ted Pella, Redding, CA, USA), teased apart and attached by drying. In experiments where CARD-FISH was combined with NanoSIMS analysis, wafers were coated with VectaBond (Vector Laboratories, Burlingame, CA, USA). Wafers were mapped with reflected light and scanning electron microscopy (SEM) for orientation in the NanoSIMS. SEM images of higher magnification were also collected to match increased 15N/14N ratios subsequently measured by NanoSIMS with microbial cells. Higher magnification images were also taken from filamentous cyanobacteria to ensure that the investigated regions of cyanobacterial filaments were free of attached microorganisms and that increased 15N/14N ratios could be attributed to the cyanobacterial bacterial filament and not to associated epibionts.

CARD-FISH for NanoSIMS analysis of ESFC-1 filaments in mat samples

Design and optimization of ESFC-1-specific oligonucleotide probes are described in Supplementary Information. CARD-FISH was conducted as decribed previously (Pernthaler et al., 2002) with hybridizations conducted at 46 °C and washing at 48 °C. Hybridization was conducted on silicon wafers coated with VectaBond; embedding in agarose was omitted. Hybridizations were performed with the following probes specific for the ESFC-1 cluster: ESFC1_172 and ESFC1_177 (Supplementary Table 1), and with NON338 (Wallner et al., 1993) as a negative control and EUBI-III as positive control probe (Amann et al., 1990; Daims et al., 1999). Cells were counterstained with 4′,6-diamidino-2-phenylindole (DAPI). Wafers were mapped with epifluorescence microscopy, ESFC-1 filaments were identified and their location on the wafer imaged. Wafers were mapped for NanoSIMS analysis as above-mentioned.

NanoSIMS

SIMS was performed at Lawrence Livermore National Laboratory using a Cameca NanoSIMS 50 (Gennevilliers Cedex, France) as previously described (Popa et al., 2007). Secondary ions 12C14N and 12C15N were detected by pulse counting to generate 10–20 serial quantitative secondary ion images (that is, layers). Samples were also imaged simultaneously by secondary electrons. Samples were sputtered to a depth of ∼100 nm to achieve sputtering equilibrium before collecting data (Ghosal et al., 2008). The depth of analysis during a measurement was between 50 and 200 nm. Measurements were repeated on selected cells to ensure measurement accuracy. Selected samples were also sputtered at high-beam currents (∼1 nA) between repeat measurements to determine if isotopic composition changed; no significant changes were found with cell depth. Natural abundance samples of the filamentous cyanobacteria were run and analyzed, and those values were used to correct the values of 15N2-incubated samples.

Data were processed as quantitative isotopic ratio images using custom software (LIMAGE, L Nittler, Carnegie Institution of Washington), and were corrected for detector dead-time and image shift from layer to layer. Regions of interest (ROIs) were defined in each image, and the isotopic ratio for each ROI were calculated by averaging over all of the replicate layers. For filamentous cyanobacteria, typically 3–10 connected cells within a filament were analyzed, and the isotopic ratio displayed in the Results section for the filamentous cyanobacteria is the average ratio of all filaments with standard error. Cells were analyzed for significant enrichment using a one-tailed Student t-test with equal variance.

Data are presented as 15N/14N ratio values and δ15N values (calculated after equation (1)) and are presented as means±standard error (SE). All reported enrichment values are corrected with natural abundance values.

For a preliminary estimation of the contribution of small filamentous cyanobacteria and single cells to the observed 15N incorporation, we analyzed 15 small filamentous cyanobacteria and 42 single cells in a total of 30 images of each 900–2500 μm2 size. The 15N/14N ratio of enriched cells and filaments was measured and the atom percent enrichment (APE) of the ROI was calculated according to equation (2).

The cell area of enriched single cells and filaments were measured based on the secondary electron image, and the 15N enrichment was calculated as APE per cell area or filament area. The origin of enriched signals was confirmed with SEM images that were collected prior to NanoSIMS analysis.

Nucleotide accession numbers

16S rRNA gene and transcript sequences obtained in this study are deposited under GenBank accession numbers JQ013010–JQ013029. Sequences of nifH genes and transcripts are deposited under GenBank accession numbers JQ013030–JQ013038.

Results

N2 fixation patterns in Elkhorn Slough cyanobacterial mats

Cyanobacterial mat samples were collected near the mouth of the Elkhorn Slough on 20 October 2009. Microscopic examination of the mats revealed that the upper green layer was dominated by filamentous, non-heterocystous cyanobacteria, primarily by Microcoleus spp.; Oscillatoria spp., smaller filamentous cyanobacteria and Lyngbya spp. were also observed (Supplementary Figure 1).

Nitrogen fixation has not been studied previously in Elkhorn Slough mats. To elucidate the N2 fixation patterns in these mats, mat samples were assayed for nitrogenase activity and incorporation of fixed N2 using acetylene and 15N2, respectively, as substrates (Bebout et al., 1993; Montoya et al., 1996). ARAs with complete mat cores in two successive diel cycles revealed that nitrogenase activity was more than 8-fold higher at night relative to the daytime (Figure 1). The maximum daytime nitrogenase activity was 3.7 μmol C2H4 m−2 h−1 on 21/22 October and 4.2 μmol C2H4 m−2 h−1 on 23/24 October. The nighttime activity reached 30.2 μmol C2H4 m−2 h−1 and 62.4 μmol C2H4 m−2 h−1 on 21/22 and 23/24 October, respectively. These values are within the activity range of previously studied mats from Bird Shoal, North Carolina, and Guerrero Negro, Mexico (9 μmol C2H4 m−2 h−1 to >600 μmol C2H4 m−2 h−1), which are dominated by similar cyanobacteria (Bebout et al., 1993, 1994; Omoregie et al., 2004a).

Figure 1
figure 1

Nitrogenase activity measured in a diel cycle on 21/22 October 2009 of Elkhorn Slough microbial mats. ARAs were conducted with entire mat cores of 10 mm diameter and 10 mm thickness.

Further ARAs on layer-separated mat samples obtained in 2009 showed that >95% of the potential nitrogenase activity measured in the whole mat was recovered in the upper 2 mm of the ca. 1-cm-thick mats (Supplementary Figure 2). The uppermost layer (0–2 mm) had significantly higher activities than the other two deeper layers (P<0.013, 0–2 mm versus 3–6 mm and P<0.013, 0–2 mm versus 7–10 mm) (Supplementary Figure 3). In contrast, the comparison of 3–6 mm versus 7–10 mm did not show significant differences in activity (P<0.999). When incubated with 15N2 in the dark, IRMS measurements performed on these layer-seperated mat samples confirmed that the uppermost layer was significantly enriched in 15N relative to the 3–6 mm (P<0.001) and 7–10 mm (P<0.001) sections (Supplementary Figure 4), whereas no difference was observed comparing the 3–6 mm and 7–10 mm layers (P<0.998). After a 10–11 h incubation, the 15N/14N ratio in the upper layer (0–2 mm) reached 0.00424±9.2 × 10−5 (156.2±25.2‰), compared with 0.00368±2.5 × 10−6 (3.0±0.7‰) in the 3–6 mm sections and 0.00368±3.5 × 10−6 (4.1±1.0‰) in the 7–10 mm sections (Supplementary Figure 2). Therefore, the investigation of diazotrophs in Elkhorn Slough mats focused on these uppermost 2 mm.

NanoSIMS analysis of Elkhorn Slough mat microbial community

The upper layers of the 15N-labeled October 2009 nighttime samples were also analyzed by NanoSIMS in order to assess 15N enrichment by different cell morphotypes. Significant and high 15N enrichments were observed in small cyanobacterial filaments (P<0.001), which were ⩽150 μm in length and composed of individual cells that were ca. 3 μm long and ca. 2 μm wide. Enrichment was also observed in multiple single cells (P<0.001). The enriched small filaments had a 15N/14N ratio of 0.01436±0.00176 (2912.9±480.8‰, n=15), and enriched single cells had a 15N/14N ratio of 0.03519±0.00807 (8588.6±2200.1‰, n=42) (Figure 2). In contrast, filaments of the abundant cyanobacterial group Microcoleus spp. and of Oscillatoria spp. showed very low levels of enrichment (15N/14N=0.00382±2.4 × 10−5 or 40.9±6.5‰, n=26 (P<0.001) and 15N/14N 0.00373±1.9 × 10−5 or 17.1±5.1‰, n=10 (P<0.005), respectively). For a preliminary estimation of the contribution of the small filamentous cyanobacteria and single cells to the observed 15N incorporation in the upper mat layer, cellular 15N enrichments were analyzed in over 30 images of 900–2500 μm2 size each. Based on this screen, >80% of the total incorporated 15N was observed in smaller cyanobacterial filaments and ca. 20% in the single cells.

Figure 2
figure 2

(a, b) Representative NanoSIMS secondary electron (e−) and nitrogen isotope ratio (15N/14N) images of Elkhorn Slough microbial mats from October 2009 after incubation with 15N2 in the dark. Small filamentous cyanobacteria that fixed nitrogen incorporated the 15N tracer (clarified by the arrows in b). The color scale bar represents 15N/14N ratios (0.00367 corresponds to natural abundance). Scale bars represent 5 μm.

Identification of an uncharacterized nifH phylotype in Elkhorn Slough mats

To determine the diazotroph community in the mats and determine which community members were actively expressing the functional gene for N2 fixation, nifH clone libraries were constructed from DNA and cDNA recovered from mat samples in October 2009 (21 October 2009, 2250 hours and 24 October 2009, 0310 hours). DNA-derived nifH sequences recovered from both time points clustered with nifH cluster I and cluster III of the nifH tree as defined previously (Chien and Zinder, 1996; Zehr et al., 2003) (Table 1). Sequences of nifH cluster III were found to be numerically dominant in both samples: 21 October (79.6%) and 24 October (70.7%). In contrast, the majority of the expressed nifH sequences recovered from cDNA of these samples were found to group with cluster I (75% and 85.2%, respectively) and a minor fraction with cluster III (25% and 14.8%, respectively). Diversity indices indicated a reduced diversity of expressed nifH sequences (Simpson index OTU95, 1.32 and 2.01, 21 and 24 October, respectively) versus sequences from nifH genes (Simpson index, OTU95 3.17 and 2.50).

Table 1 Taxonomic affiliation of nifH sequences from DNA and cDNA in October 2009 Elkhorn Slough microbial mat samples

Of the expressed nifH sequences in samples from the 21 and 24 October, 52% and 23.9%, respectively, grouped with cyanobacterial sequences in cluster I. Almost all (97%) of these cluster I cyanobacterial sequences formed a monophyletic lineage distinct from other cyanobacterial nifH sequences in the database (Figure 3a). The sequence identity within this monophyletic lineage was >94.7% on the deduced amino-acid level. This cluster does not include a cultured cyanobacterium. The only closely related sequence in publicly available databases that clustered with this lineage was a translated nifH sequence (DQ821979) recovered from a mixed Lyngbya culture enriched from mats collected in Guerrero Negro, Mexico (Moisander et al., 2007) (Figure 3b). The amino-acid sequences of the monophyletic lineage from Elkhorn Slough were 96.2–100% identical to this sequence.

Figure 3
figure 3

(a) Maximum likelihood tree of cyanobacterial deduced dinitrogenase reductase (nifH) sequences from Elkhorn Slough microbial mats collected in October 2009 (bolded) and their closest relatives. The novel lineage represents 37% of the sequences recovered from cDNA clone libraries; DQ821979 is the only closely related sequence. Bootstrap values calculated with the PhyML algorithm that were ⩾50% are displayed in the tree. Scale bar represents 10% estimated sequence divergence. (b) Maximum likelihood tree of the ESFC-1 nifH phylotypes (bolded) grouping with the cyanobacterial nifH sequences from Elkhorn Slough microbial mats and DQ821979. Representatives of OTU97 are depicted along with the number of sequences within each OTU in parentheses. Bootstrap values calculated with the PhyML algorithm that were ⩾50% are displayed in the tree. Scale bar represents 10% estimated sequence divergence.

Cultivation of a cyanobacterium belonging to the novel lineage from Elkhorn Slough

The above described NanoSIMS analysis identified small filamentous cyanobacteria to be highly active in incorporating 15N2 into biomass. Sequencing of nifH transcripts revealed that a substantial proportion of the expressed nifH sequences clustered in a novel cyanobacterial lineage. Therefore, we initiated cultivation experiments targeting small, non-heterocystous filamentous cyanobacteria in an attempt to identify the cyanobacteria associated with these expressed nifH sequences. One of these cultivations yielded a unicyanobacterial enrichment dominated by a filamentous cyanobacterium, named ESFC-1, which showed a similar morphology to the highly 15N-enriched filaments visualized by NanoSIMS (Figures 2 and 4b). The maximal filament length observed in culture was 600 μm, and average cell sizes were 2.75±0.07 μm length and 1.78 ±0.02 μm width. When tested for 15N2 incorporation by IRMS, we found that the ESFC-1 culture was highly enriched in 15N after 24-h incubation (15N/14N=0.00522±4.9 × 10−06 or 421.2±1.4‰).

Figure 4
figure 4

(a) Maximum likelihood tree of the cyanobacterial 16S rRNA from the ESFC-1 enrichment culture (bolded) with its closest relatives, linking the new nifH sequence cluster from Elkhorn Slough mats to a nearly full-length 16S rRNA gene sequence. Bootstrap values calculated with the PhyML algorithm that were ⩾50% are displayed in the tree. Scale bar represents 10% estimated sequence divergence. (b) Inset micrograph depicts ESFC-1 filaments from the unicyanobacterial culture. Scale bar represents 10 μm. (c) Maximum likelihood tree of the nearly full-length 16S rRNA gene sequence from ESFC-1 (bolded) together with the related sequences from Elkhorn Slough mats collected in January 2009, illustrating the diversity of this novel group. In all, 46% of the sequences from 12 January 2009 (‘SIUS’, cDNA sample, 2100 hours, 151 out of 329 sequences) and 14.8% of the sequences from 13 January 2009 (‘SIUT’, cDNA sample, 07:00 hours, 36 out of 243 sequences) grouped monophyletically with the 16S rRNA gene sequence type from ESFC-1. OTU98 representatives of Elkhorn Slough sequences along with the number of sequences within an OTU in parentheses are depicted in the tree. Bootstrap values calculated with the PhyML algorithm that were ⩾50% are displayed in the tree. Scale bar represents 10% estimated sequence divergence.

The nifH sequences of the ESFC-1 enrichment culture were recovered by PCR amplification and two highly similar nifH phylotypes were identified (OTU cut-off of 97% at the amino-acid level, 39 sequences analyzed). These sequences differed in only three amino acids and both are possibly different genomic copies of ESFC-1. Both phylotypes clustered with the novel cyanobacterial lineage from Elkhorn Slough microbial mats (Figure 3b) with up to 100% sequence identity at the deduced amino-acid level and up to 99.4% identity at the DNA level. Cyanobacterial 16S rRNA gene sequences recovered from the ESFC-1 culture yielded one phylotype (36 sequences analyzed, ⩾99.6% sequence identity) (Figure 4a). The ESFC-1 16S rRNA sequence was only distantly related to sequences in publicly available databases with the closest sequence (DQ289927, 92.5% sequence identity) recovered from South Atlantic Bight sediments off the coast of Savannah, GA (Hunter et al., 2006). The sequence of the closest cultured representative to ESFC-1 was Spirulina strain CCC Snake P. Y-85 (Y18793) (92.3% sequence identity), isolated from Yellowstone hot springs.

ESFC-1 is a representative of a prevalent and diverse cyanobacterial group

To investigate the prevalence and diversity of the novel cyanobacterial group in Elkhorn Slough mats, pyrotag sequences of V8 amplicons of the 16S rRNA gene from multiple time points in 2009 and nearly full-length sequences of the 16S rRNA gene of one particular time point were obtained. ESFC-1 affiliated sequences were identified in six out of the seven time points from samples collected throughout 2009. The sequences were proportionally higher in amplicon libraries generated from cDNA (0.28%–36%) than from DNA samples (0–5%) (Figure 5) and substantially higher in cDNA libraries recovered from mat samples collected in January 2009. Therefore, cDNA from these samples were selected to generate nearly full-length 16S rRNA clone libraries to investigate the diversity of environmental 16S rRNA sequences that clustered with the ESFC-1 cyanobacterial group. Of the 572 near full-length 16S rRNA sequences recovered from the January 2009 samples, 187 (32.7%) grouped monophyletically with the 16S rRNA gene sequence type from ESFC-1. This monophyletic clustering was observed with multiple treeing algorithms (maximum likelihood, maximum parsimony and neighbor joining) and was supported with a bootstrap value of 79% (Figure 4b). The sequence diversity within this cluster of ESFC-1 related sequences (92.5–100%) revealed a great diversity of ESFC-1-related cyanobacteria.

Figure 5
figure 5

Relative abundance of ESFC-1 affiliated 16S rRNA V8 pyrotag amplicons in 2009. Sequences affiliated with ESFC-1 were found in samples from 13 January (0700 hours), 1st July (0400 hours), 19 August (0115 hours), 16 September (1200 hours), 21 October (2250 hours) and 13 November (0000 hours) with a higher proportion in cDNA samples compared with DNA samples.

Investigation of in situ N2 fixation by ESFC-1 and related cyanobacteria

To confirm that cyanobacteria of the ESFC-1 cluster were actively incorporating 15N2 as suggested by nifH sequencing and our initial NanoSIMS survey of the October 2009 samples, further NanoSIMS investigations were combined with CARD-FISH. Oligonucleotide probes specific for the 16S rRNA of the ESFC-1 cluster identified in the January 2009 samples were designed and stringent hybridization conditions for the newly designed probes were determined as described in Supplementary Information (Supplementary Table 1 and Supplementary Figures 5 and 6). Samples from October 2009 incubated with 15N2 were hybridized with the ESFC-1-specific probes and imaged by epifluorescence microscopy. CARD-FISH combined with NanoSIMS analysis demonstrated that filaments of the ESFC-1 cluster were enriched in 15N (Figure 6). In all, 63% of 76 analyzed ESFC-1 ROIs were significantly enriched in 15N based on a 95% confidence interval, with levels ranging from low enrichments (15N/14N=0.00374 or 19.3‰) to a 15N/14N ratio of 0.025 (5802.3‰).

Figure 6
figure 6

NanoSIMS and epifluorescence images of filaments stained with the ESFC-1-specific CARD-FISH probe ESFC-1_177 after incubation of mat cores with 15N2. (a) Epifluorescence image of DAPI-stained cells and filaments. (b) Epifluorescence image of filaments stained with ESFC-1-specific probe ESFC1_177. (c) Secondary electron (e−) image of the same region. (d) Nitrogen isotope ratio (15N/14N) image depicting an ESFC1_177-stained filament highly enriched in 15N (15N/14N 0.025). Scale bar represents 5 μm.

Discussion

In this study, we used a combination of biogeochemical, molecular and NanoSIMS analysis in conjunction with targeted cultivation experiments to investigate the active diazotrophic community in microbial mats from the Elkhorn Slough estuary. With this approach, we discovered a novel filamentous cyanobacterial group, ESFC-1, that represented almost the entire fraction of the cyanobacterial expressed nifH genes. This expression data along with NanoSIMS analysis of ESFC-1 and other mat cyanobacteria led us to conclude that members of this novel group were the most active N2-fixing cyanobacteria in the investigated mat.

Biogeochemical analysis of the mats collected in October 2009 revealed patterns of N2 fixation activity similar to those observed previously in mats from Guerrero Negro in Mexico, Bird Shoal in North Carolina, and the island Mellum in the North Sea, which were dominated by filamentous non-heterocystous cyanobacteria; nitrogenase activity was restricted to the upper 2 mm of the mat (Stal et al., 1984) and the highest at night under anoxic conditions (Stal et al., 1984; Bebout et al., 1987, 1993; Paerl et al., 1996; Omoregie et al., 2004a, 2004b). Despite these biogeochemical parallels, molecular data indicated that a previously unrecognized group of diazotrophs was present and highly active in Elkhorn Slough mats.

Overall, the nifH gene sequences that we obtained from the Elkhorn Slough mats grouped with the nifH clusters I and III. This pattern was previously observed in mats from Guerrero Negro and the island Schiermonnikoog in the Wadden Sea (Omoregie et al., 2004a, 2004b; Moisander et al., 2006; Severin et al., 2010). A reduction in nifH transcript diversity compared with nifH gene diversity was observed in clone libraries recovered from Elkhorn Slough. The same pattern was previously reported in mats from Guerrero Negro (Moisander et al., 2006). These observations support the hypothesis that only a subset of the microbial community with the genetic potential for N2 fixation actively expresses nifH in these cyanobacterial mats. Interestingly, a high number of the transcripts from Elkhorn Slough in October 2009 formed a novel monophyletic sequence cluster, suggesting that a new cyanobacterial group might be an important diazotroph in these mats. However, N2 fixation is regulated on multiple levels ranging from transcription (Chen et al., 1998) to posttranslational protein modification (Kim et al., 1999), so the detection of nifH gene transcripts does not necessarily imply active N2 fixation in situ.

NanoSIMS investigations provided additional data to support the assignment of a highly active diazotrophic cyanobacterial group in Elkhorn Slough mats predicted by marker gene analysis. NanoSIMS of 15N2-incubated Elkhorn Slough mat samples from October 2009 showed that among the filamentous cyanobacteria, a particular morphotype that had cells of ca. 2 μm width and ca. 3 μm length was highly enriched in 15N. We hypothesized that these cyanobacterial filaments might harbor the novel and highly expressed nifH sequences. To identify these cyanobacteria, we conducted targeted cultivation experiments for diazotrophic cyanobacteria searching for the suspected morphotype, and indeed, we obtained a similar morphotype (ESFC-1) in one enrichment that fixed 15N2 in culture. Cyanobacteria in this culture harbored the novel nifH sequence type and enabled us to link the nifH sequence to the corresponding 16S rRNA gene sequence. These data also enabled us to design specific oligonucleotide probes for CARD-FISH/NanoSIMS experiments to examine 15N incorporation into the ESFC-1 populations in 15N2-incubated Elkhorn Slough mats. These studies revealed that ca. 60% of the ESFC-1-related filaments actively incorporated 15N. However, 15N enrichment in these populations was highly variable, with a small fraction of the ESFC-1 community dominating the total 15N incorporation. Large variations in 15N enrichments have previously been noted in NanoSIMS investigation of Chlorobium phaeobacteroides-related cells in Lake Cadagno (Switzerland) (Halm et al., 2009), Aphanizomenon sp. in the Baltic Sea (Ploug et al., 2010) and shipworm symbionts (Lechene et al., 2007). This variability may be due to spatial heterogeneity in the local environment, such as small-scale variations in nutrient concentrations, diffusion rates or redox potentials. The microbial mat environment is highly heterogeneous (Jørgensen et al., 1983; Des Marais, 2003), and as filaments experience different environmental conditions, they may differ in their physiology and N2 fixation activity, resulting in different 15N enrichment levels. We noticed that even adjacent filaments (Figure 6) differed in the degree of 15N-enrichment, suggesting that the local environment may not be solely responsible for the variable activity of the filaments, but also could be due to strain or species-level variation in ESFC-1 populations. The high level of diversity in the ESFC-1 clade at the 16S rRNA level supports this hypothesis. The functional significance of such fine scale variation has been extensively studied in low-diversity acid mine drainage biofilms and wastewater treatment reactors (Denef et al., 2010). Furthermore, the filaments could be in different physiological states with only a part of the population actively growing. These data also illustrate the power of single-cell techniques to reveal functional heterogeneity at the single-cell level within closely related populations.

Amplicon pyrosequencing of the V8 region across multiple samples throughout the year 2009 revealed that the ESFC-1 cluster is prevalent in Elkhorn Slough mats. ESFC-1-related sequences were much less abundant in the total microbial community (DNA) compared with the active community (cDNA), representing an example of a low abundance member of a microbial community that may perform an important ecosystem function, as was observed in NanoSIMS investigations of phototrophic bacteria in a meromictic lake (Musat et al., 2008; Halm et al., 2009). An ESFC-1-related nifH gene sequence has also been detected in a mixed Lyngbya culture enriched from mats collected in Guerrero Negro, Mexico (Moisander et al., 2007) suggesting that ESFC-1-related cyanobacteria may be present in other microbial mats, but may not have been identified due to low abundance at the time of sampling. The pyrotag sequence data were supported by near full-length 16S rRNA sequences from the January 2009 cDNA samples, in which ca. 33% of all cDNA sequences (187 total sequences) formed a broad monophyletic clade with a degree of diversity that may represent multiple species. Deep sequencing of mat samples from other environments may identify ESFC-1-related populations and help broaden our understanding of the ecophysiology of this novel cyanobacterial group.

Interestingly, we found no evidence that Microcoleus spp., which are ubiquitous and abundant members of marine microbial mats globally, had an important role in N2 fixation in the Elkhorn Slough mats. Microcoleus chthonoplastes, the type strain for mat-dwelling Microcoleus spp., was for many years not thought to be a diazotrophic cyanobacterium. However, the recently completed genome of M. chthonoplastes PCC 7420 revealed a nif-gene cluster (Bolhuis et al., 2010), which grouped the nifH gene with δ-proteobacterial genes in cluster III. In the nifH clone library of DNA samples from one of the two October time points (24 October 2009), we identified three sequences that grouped closely with these M. chthonoplastes sequences in cluster III. However, no nifH transcripts related to Microcoleus spp. were recovered from the Elkhorn Slough mat samples described in this study. Additionally, very low levels of 15N enrichment (average 15N/14N ratio of 0.00382±2.4 × 10−5 or 40.9±6.6‰) in Microcoleus spp. filaments were observed in NanoSIMS analysis of 15N2-incubated mat samples from October 2009 with the highest 15N/14N ratio being 0.00401. This trend was also noted in other samples with nighttime N2 fixation (December 2007) in which Microcoleus spp. filaments had enrichment values very close to natural abundance (15N/14N 0.00368±2.4 × 10−5, 3.4±6.6‰). These low levels of 15N enrichment observed may be due to cross-feeding from active diazotrophs. In contrast, ESFC-1 filaments were observed with 15N/14N ratios up to 0.025 (5802‰). Although Microcoleus spp. may fix N2 under specialized conditions, we could not find evidence that Microcoleus spp. were actively fixing 15N2 in the investigated Elkhorn Slough mats. This finding lends further support to our hypothesis that ESFC-1-related cyanobacteria were the dominant active cyanobacterial diazotrophs in the investigated Elkhorn Slough microbial mats.