Introduction

Surface oceanic waters around the tropical and subtropical latitudes are commonly oligotrophic. In such nutrient-poor environments, symbiotic relationships are commonly observed and enhance survival via mutual benefits for both partners. Nitrogen is the most abundant substance (∼78%) in the atmosphere and is essential for creating and sustaining life on earth. Yet, its inert form (N2) is metabolically unavailable to most organisms, making the role of nitrogen-fixing organisms (diazotrophs) crucial in N-limited environments.

Many symbiotic associations are known between diazotrophs and other organisms from terrestrial symbioses between Rhizobium and legumes to aquatic symbioses between cyanobacterial diazotrophs and phytoplankton such as Richelia intracellularis and diatoms. R. intracellularis is a small (∼10 μm in diameter) nitrogen-fixing cyanobacterium spatially segregating nitrogen fixation from photosynthesis and thus fixing atmospheric N during daytime. It consists of one polar heterocyst (site of nitrogen fixation) and three to 10 vegetative cells (Geitler, 1932). It is usually found as an endosymbiont within diatoms such as Rhizosolenia spp, Hemiaulus spp, or as an episymbiont attached to Chaetoceros spp. (Venrick, 1974; Taylor, 1982; Sundstrom, 1984; Gomez et al., 2005). Diatom–diazotroph associations containing the cyanobiont R. intracellularis and the filamentous non-heterocystous Trichodesmium spp., were traditionally considered the dominant nitrogen-fixing plankton in marine tropical oceans. The ‘New N’ supplied by these organisms, and a variety of small unicellular diazotrophs described recently (Zehr et al., 2001; Montoya et al., 2004; Zehr et al., 2007), contribute substantially to these systems (Venrick, 1974; Mague et al., 1977; Carpenter and Romans, 1991; Capone et al., 1997; Carpenter et al., 1999; Arrigo, 2005). In the tropical North Atlantic, extensive N2 fixation by blooms of diatoms and R. intracellularis produced nearly 70% of total N demand in surface waters (Carpenter et al., 1999; Janson et al., 1999).

The eastern Mediterranean Sea is extremely oligotrophic due to a general west to east anti-estuarine circulation, which exports nutrients from the Mediterranean into the North Atlantic (Bethoux et al., 1998). The oligotrophic nature results in a high water clarity, low values of chlorophyll a (Chl a), low rates of primary production (Berman et al., 1984; Azov, 1986; Li et al., 1993; Yacobi et al., 1995; Psarra et al., 2000; Tanaka et al., 2007), and dominance by small-sized phytoplankton (Li et al., 1993; Yacobi et al., 1995; Ignatiades et al., 2002). The Levantine basin, located in the easternmost section of the Mediterranean, is the most oligotrophic region within the Mediterranean Sea. Its characteristics are a prominent deep chlorophyll maximum (DCM) at 100–130 m depth (Li et al., 1993; Yacobi et al., 1995), low phytoplankton biomass (<50 ng Chl a l−1) in the surface layers (Li et al., 1993; Vidussi et al., 2001), an average salinity of ∼39 psu, and sea-surface temperatures between 16–28 °C.

Studies using geochemical and stable isotopic (δ15N) mass balances have suggested that nitrogen fixation supplies a significant source of new nitrogen to the Mediterranean (Bethoux, 1986; Bonin et al., 1989; Bethoux et al., 1992, 2002; Pantoja et al., 2002). Yet, only recently have actual N fixation rates been reported from the Mediterranean Sea. One study, from the Western Mediterranean, averaged ∼7 nmol N l−1 day−1 (Garcia et al., 2006) while measurements from a warm-core eddy near Cyprus yielded rates of ∼129 nmol N l−1day−1 (Rees et al., 2006). However, these studies lack any identification or characterization of the organisms responsible for this fixation.

We recently described a diverse diazotrophic community with differential patterns of nifH expression from an east–west transect off the Israeli coast (Man-Aharonovich et al., 2007). One of the organisms identified was 99% identical to R. intracellularis. Previous observations indicated the presence of the diatoms Rhizosolenia spp. and Hemiaulus hauckii (B Kimor, personal communication), which harbor these cyanobionts although, to date, no published studies exist on their dynamics anywhere within the Mediterranean.

In this study, we focused on the diatom–diazotrophic associations in the eastern Mediterranean and have identified, characterized and followed the seasonal and spatial dynamics of the filamentous nitrogen-fixing cyanobacteria–R. intracellularis and its diatom hosts: Rhizosolenia spp. and Hemiaulus spp. at 2 monthly monitored stations off the Mediterranean coast of Israel.

Materials and methods

Sample collection and processing

Water samples were routinely collected during daytime in monthly cruises (June 2006–May 2007) on board the R/V Mediterranean Explorer. Two stations were sampled: one coastal station TB200 (32° 09.982′ N and 34° 34.509′ E) at ca. 200 m bottom depth and one pelagic station TB1000 (34°14.909′ E and 32°10.059′ N), at ca. 1000 m bottom depth. Water samples for molecular distribution, nitrogen fixation and nutrients concentration determination were collected using Niskin bottles (8 l) mounted on a Rosette equipped with a CTD (seabird 19 Plus), in-situ fluorometer (Turner designs, Cyclops7) that was calibrated against extracted Chl a in the laboratory.

Samples for cell counts were taken from the surface waters (∼5 m depth). A Nytex plankton-net (55–300 μm equipped with a digital flow meter (Hydrobios No 438110) was towed horizontally behind the research vessels and used to compute the total volume of water passing through the net.

Sample collection and processing for cell identification and enumeration

Samples were collected from the net and preserved in 50 ml tubes with 0.5 % gluteraldehyde and stored at 4 °C in the dark until further examination in the lab. Samples were identified and counted using an epifluorescence microscope (Nikon Eclipse 80i) equipped with three specific filters—chlorophyll fluorescence filter (Ex: 450 nm, Em: 680 nm), phycoerythrin fluorescence filter (Ex: 490 nm, Em: 580 nm) and phycocyanin fluorescence filter (Ex: 620, Em: 660 nm).

Five to 20 replicate samples (70 μl) per station/depth were counted microscopically (with a total of 200 to >3000 heterocysts counted depending on abundance). The chloroplasts of the host cell (diatom) were observed using the chlorophyll a filter only while the endosymbiotic cyanobacterium R. intracellularis was observed under both phycoerythrin and phycocyanin filter sets. Both the host and the endosymbiont R. intracellularis were identified and counted.

Nitrogen fixation rates

N fixation was measured on preconcentrated samples using the acetylene reduction technique which measures total nitrogenase activity (Capone, 1993; Montoya et al., 2004) and has also been applied in the field on concentrated samples (Falcon et al., 2002, 2004). Ten to 15 liters were pre-concentrated from surface waters (∼5 m) and size fractionated to 1–10 μm, 10–100 μm and >100 μm under low vacuum. Biomass was incubated in 125 ml serum bottles with 80 ml of 0.2 μm filtered seawater. Acetylene (9 ml) was added to the bottles and 2 ml of gas from the headspace were immediately removed to Vacuettes (Greiner Bio-One, Kremsmünster, Austria) for T0. Sample bottles (at neutral pressure), and appropriate blanks (0.2 μm filtered seawater and killed blanks) were incubated for 24 h in a flowing seawater bath at 25 °C and ∼100 μmol quanta m2 s−1. Ethylene production was measured with an FID-GC (SRI 310C) using known ethylene standards. Ethylene production was converted to N-fixation rates according to Capone and Montoya (Capone and Montoya, 2001).

Sample collection and processing for molecular analysis

Twenty-liter samples were collected with Niskin bottles from each depth in the euphotic zone (∼5, 60, deep chlorophyll maximum and 150 m). Each sample was placed into an acid-washed container and pumped using a peristaltic (Cole Parmer-Master Flex, Vernon Hills, IL, USA) pump onto 0.2 μm-pore-size Supor filters (Pall Gelman Inc., Ann Arbor, Michigan). The filter was then placed into a cryotube with 1 ml of lysis buffer (Massana et al., 1997) and stored at −80oC until further processing.

Nucleic-acid extraction and reverse transcription

Nucleic acids were extracted from the samples using a phenol–chloroform extraction according to Massana et al. (1997). Nucleic acids were treated with RNase-free DNase I (Ambion, Cambridgeshire, UK) for 30 min at 37 °C to remove DNA residue. DNase was inactivated by heat denaturation at 75 °C for 10 min and samples were stored at −80 °C until further use. Total RNA (100–300 ng) was then reverse transcribed using nifH degenerate reverse primer nifH3 (5′-ATRTTRTTNGCNGCRTA-3′) and nifH723R (5′-GATGTTCGCGCGGCACGAADTRNATSA-3′) using Bio-RT (Bio-Lab Ltd, Jerusalem, Israel) according to manufacturer's instructions.

PCR amplification of nifH transcripts

Nitrogenase Fe protein transcripts (nifH) were amplified from cDNA using the PCR primers of Zehr and McReynolds, (1989). A fragment of the nifH gene (approximately 360 bp) was amplified with a nested PCR strategy. The PCR mixtures contained 2.5 μl of 10 × ExTaq buffer (Takara, Madison, WI, USA), 2 μl of a mix of deoxynucleoside triphosphates (2.5 mM each), 18.75 μl of DEPC water, 0.25 μl of 100 μM nifH3 (5′-ATRTTRTTNGCNGCRTA-3′), 0.25 μl of 100 μM nifH4 (5′-TTYTAYGGNAARGGNGG-3′) and 0.25 μl of ExTaq DNA polymerase (5 U μl−1; Takara). The reaction mixtures were amplified for one denaturation step (5 min at 94 °C), followed by 30 cycles of 94 °C for 1 min, 50 °C for 1 min and 72 °C for 1 min, and one final 7-min extension cycle at 72 °C. To the nested PCR master mix, 0.5 μl of the first PCR product and 0.5 μl of DEPC water were then added. nifH primers previously designed (Zehr and McReynolds, 1989): 0.25 μl of 100 μM nifH1 (TGYGAYCCNAARGCNGA) and 0.25 μl of 100 μM nifH2 (ADNGCCATCATYTCNCC). Second-round reactions were amplified with one denaturation step of 5 min at 94 °C, 30 cycles of 94 °C for 1 min, 57 °C for 1 min and 72 °C for 1 min, and a final 7-min extension cycle at 72 °C. All of the tests for the presence of contaminating DNA in the RNA samples or in the reagent were carried out according to (Zani et al., 2000). PCR products were cloned using the QIAGEN_PCR cloning kit and selected clones were sequenced. The nifH sequences reported in this study are deposited with GenBank under accession numbers EU807710–EU807732.

Phylogenetic analyses of nifH sequences

nifH sequences were assembled and translated into amino-acid sequences using the SEQUENCHER program (version 4.5). All diversity and phylogenetic analyses were performed using deduced amino-acid sequences. NifH protein sequences were aligned using CLUSTALX version 1.81 (Thompson et al., 1997). Neighbor-joining and maximum-parsimony analyses were conducted on amino-acid data sets using Paup 4.0b10 (Swofford, 2002). Default parameters were used in all analyses. Bootstrap re-sampling of neighbor-joining (1000 replicates) and maximum-parsimony (1000 replicates) were performed.

Results and discussion

Study area

To examine spatial and temporal changes in R. intracellularis, one coastal (TB200) and one pelagic (TB1000) station, off the continental shelf, were sampled monthly during 2006–2007 (Figure 1). There were significant differences in the physical and chemical properties of the coastal and pelagic stations (Figure 2). The most apparent difference was the depth of the upper mixed layer during winter (January–March). At the coastal station, the winter deep mixing, which almost reached the bottom, could provide nutrients from the sediments while at the pelagic station, the deep mixing could not supply similar amount of nutrients from deeper layers.

Figure 1
figure 1

Map of the station locations: Station TB200, coastal station (on the continental shelf), bottom depth of 200 m (32° 09.982′ N and 34° 34.509′ E) and a pelagic station, TB1000 with a bottom depth of 900 m, off the continental shelf (34°14.909′ E and 32° 10.059′ N).

Full size image
Figure 2
figure 2

Representative depth profiles of the two sampling stations (TB200 and TB1000) during a stratified month (upper panel) and a mixed month (lower panel). In both panels parameter shown are: Temperature (blue and light blue), salinity, PSU (pink and red); Chl a, μg l−1 (green and light green); and % light at depth (purple and orange).

Full size image

The pelagic upper layer (100 m) had NO3+NO2, PO4 and Si(OH)4 concentrations of <0.08−0.85 μM, <0.008–0.11 μM and 0.51–1.41 μM, respectively. Slightly higher concentrations of PO4 (up to 0.13 μM) and Si(OH)4 (up to 2.14 μM) were measured at the coastal station. These ranges of concentrations are consistent with previous studies from the Levantine basin (Berman et al., 1984; Azov, 1986; Bonin et al., 1989; Herut et al., 1999; Kress et al., 2005).

During stratified months (June–September) Chl a concentrations peaked to 0.45 μg l−1 at the DCM (∼110–150 m) while during winter mixing, the DCM (∼0.4 μg Chl a l−1) was widely spread between 50–130 m at the pelagic station and was not developed at the coastal station showing uniform vertical Chl a distribution (∼0.2–0.3 μg l−1). The highest photosynthetic biomass at both stations, estimated by Chl a concentrations and by variable fluorescence, occurred during the winter mixing.

Identification and characterization of the diatom–diazotroph associations

Using epifluorescent microscopy we identified the presence of the two main known diatom hosts of R. intracellularis from our two sampling stations: Rhizosolenia clevei var. communis and Hemiaulus hauckii. The cyanobiont R. intracellularis was present in all samples containing H. hauckii whereas it was only sporadically found in Rhizosolenia.

H. hauckii found in our study period was characterized by chains composed of 3–15 cells. Each cell contained one or two R. intracellularis trichomes though occasionally H. hauckii cells contained up to four R. intracellularis trichomes (see Figures 3a–d). The R. intracellularis trichome consisted of 3–5 vegetative cells and one or two heterocysts at the end of each trichome depending on the size and stage of the H. hauckii–R. intracellularis life cycle. The diameter of the H. hauckii varied and averaged ∼30 μm whereas the diatom chain could be as long as 200 μm. The average diameter of R. intracellularis inside H. hauckii was ∼10 μm.

Figure 3
figure 3

Diatom- Richelia symbioses. The diatom Hemiaulus hauckii and its endosymbiotic cyanobacterium Richelia intracellularis (a–d). The diatom Rhizosolenia spp and its endosymbiotic cyanobacterium Richelia intracellularis (e–i). (b, g) bright-field microscopy; (a, f) chlorophyll a fluorescence (ex: 450 nm, em: 680 nm ); (d, i) phycoerythrin fluorescence (ex: 490 nm, em: 580 nm); (c.h) phycocyanin fluorescence (ex: 620, em: 660 nm).

Full size image

Under bright-field light microscopy only the H. hauckii cells were apparent (Figure 3b). The distinguishing spindles connecting the H. hauckii cells and their length (40 μm), enabled identification of the Hemiaulus as H. hauckii (Heinbokel, 1986; Kimor et al., 1987, 1992; Villareal, 1994). Under chlorophyll excitation only H. hauckii chloroplasts were visible (Figure 3a). The high chlorophyll autofluorescence of the diatom masked the chlorophyll contained in the cyanobiont vegetative cells. R. intracellularis was located in the diatom intercellular space (Villareal, 1991; Rai et al., 2000). The physical proximity between the cyanobiont and the diatom's chloroplasts may be due to the cyanobionts' high demand for carbohydrates (Wolk et al., 1994) required for the energetic costs of nitrogen fixation (Villareal, 1992; Janson et al., 1995a; Rai et al., 2000). Yet, we are not aware of any documented evidence for this carbon exchange.

The cyanobiont cells were visible only under phycocyanin or phycoerythrin (PE) excitation filters (Figures 3c,d,e,h, and i). These pigments are unique to the cyanobacterial endosymbionts (Villareal, 1992; Janson et al., 1995a; Rai et al., 2000) and are not found in the diatoms. Heterocysts appeared to contain high concentrations of PE (discernable by very high fluorescence under PE filters). This may be due to the pigment role in energy transfer to PSI or to the lack of protein degradation during heterocyst differentiation (Peterson, 1981; Fay, 1992; Janson et al., 1995a). Phycobiliproteins are also known to have a role in cyanobacterial nitrogen storage (Allen et al., 1984; Kromkamp, 1987). This increase in fluorescence might indicate that the heterocysts are using PE for an intracellular nitrogen storage facility (Figures 3d and i) (Wyman et al., 1985; Janson et al., 1995b; Wingard et al., 2002; Zhang et al., 2006). Nitrogen storage as PE, combined with the close physical proximity between cyanobiont and host, may imply that nitrogen is rapidly assimilated once fixed by the R. intracellularis.

In Rhizosolenia spp. the trichomes of R. intracellularis lie within the diatom's periplasmatic space (Villareal, 1992; Janson et al., 1995b) and not between the host chloroplasts. Thus, under bright-field light microscopy, the vegetative cells of R. intracellularis were visible inside the Rhizosolenia and contained one to two trichomes with the heterocyst located toward the end of the Rhizosolenia valve (Figure 3e). The diameter of the Rhizosolenia cells varied between 20–30 μm whereas the R. intracellularis diameter was ∼10 μm. R. intracellularis contained ∼10 vegetative cells with one enlarged heterocyst (Figures 3h and i). The heterocysts were visible only when the organisms were excited with the PE or PC filter sets. This might be due to low concentrations of Chl a in these specialized nitrogen-fixing cells where no PSII activity occurs (Zhang et al., 2006). The diatom host can acquire their cyanobiont via two major pathways: either directly from the aquatic environment, or, through the parental cells following cell division (Villareal, 1989; Rai et al., 2000). The initiation mechanism is not known at present and the published literature on transmission and division is scarce (Villareal, 1989). Moreover, during our study period we did not observe any free living R. intracellularis as have been observed in other studies (Gomez et al., 2005; Foster and O'Mullen, in press). We observed the propagation of the symbiosis only via cellular division (Figure 4) and not by direct acquisition of symbionts from the seawater. Bright-field and epifluorescent microscopy revealed the division of the parent H. hauckii into two daughter cells (Figures 4a and b). Intertwined and visible under PE fluorescence, we also observed concurrent division of the symbiont R. intracellularis (Figure 4c). These results contrast with the asynchronous division reported from the only known culture of Rhizosolenia-Richelia that had been isolated, maintained, and was documented up to seven generations (Villareal, 1989). What controls the timing of those processes is still unknown. We hypothesize that nutrient cycling may play a significant role in cellular division of the DDA's, yet further work is required to validate this.

Figure 4
figure 4

Cellular division of the diatom host and cyanobiont. Division of the diatom Hemiaulus hauckii under bright-field microscopy (a) with the diatom's chloroplast under Chlorophyll a fluorescence (b), and the cyanobiont Richelia intracellularis under phycoerythrin fluorescence (c).

Full size image

Molecular identification and nifH expression of R. intracellularis

We confirmed the microscopical identification of the diatom–diazotroph associations using molecular tools. In both stations (TB200 and TB1000) water samples were collected during 2006 throughout the euphotic zone for molecular characterization of the symbiotic associations. Our recent study at an east–west transect in the Levantine basin identified expressed nifH with high identity to R. intracellularis from size-fractionated samples between 11 to 60 μm (Man-Aharonovich et al., 2007). Here, we expanded and sequenced a total of 52 randomly picked clones containing nifH inserts from samples >0.2 μm collected during daylight hours (as heterocystous cyanobacteria including Richelia are known to actively fix during the day). From the total clones that were sequenced, 54.5% had between 98–99% identities at the DNA level to R. intracellularis, while only a small number of clones from each of the other diazotrophs were found representing any of the other diazotroph groups. This suggests that R. intracellularis is the main daytime nifH-expressing organism throughout the water column during our study period. We could not decipher from our phylogenetic analyses which organism (Rhizosolenia spp or H. hauckii) was the host of R. intracellularis (Figures 3a–d). However, as >95% of the microscopical counts of the diatom containing the endosymbiont were of H. hauckii, we assume that R. intracellularis found in this study was predominantly from H. hauckii.

Surprisingly, all of the sequences that had high identity to R. intracellularis (∼99%) fell into two nifH clusters, referred as Clusters A, B (Figure 5). Both subclades were statistically supported (bootstrap values >80) in the phylogenetic analysis based on amino-acid sequences. To examine the reasons for variation in the different clades from the two stations, we explored changes in the physical/chemical characteristics of the water column (see area of study and Figures 2 and 6). Cluster A was comprised of R. intracellularis nifH sequences recovered from the pelagic station (TB1000) during the mixed period (February, March 2006) and throughout the euphotic zone from surface samples to 120 m depth. Cluster B was comprised of R. intracellularis nifH sequences recovered from the coastal station (TB200) during the mixed-period (February, March 2006) and at the onset of stratification (May 2006). The only sequences resulting from these months were from surface-layer samples.

Figure 5
figure 5

A Phylogenetic tree reconstructed from nifH-deduced amino-acid sequences. The sequences were obtained from the two sampling stations TB200 and TB1000 during 2006. The two groups obtained from the clone libraries sequences are marked in boldface letters. Bootstrap values (NJ/MP) exceeding 80% are indicated above the branches. The clusters (a) (open Ocean station—TB1000) and (b) (coastal station—TB200) represent the two groups.

Full size image
Figure 6
figure 6

Spatial and temporal R.intracellularis heterocyst counts from surface waters (∼5 m) at our two representative stations TB1000-open Ocean (triangles) and TB200—coastal (circles) from May 2006 through August 2007. Insert shows correlation between above heterocyst counts and N-fixation rates during stratified months only (May–September).

Full size image

We hypothesize that the regional circulation pattern (Rosentraub and Brenner, 2007) may play a significant role in the creation of the two different clades. The sharp continental slope located between the continental and pelagic station could generate a hydrological barrier (Rosentraub personal communication) that separates the phytoplankton community and thus also the R. intracellularis endosymbionts. Thus, while heterocyst abundance was similar at the two stations, the phylogenetic analysis revealed two distinct clades of R. intracellularis (Figure 5). Bacterial diversity may not always exhibit a specific taxa-area relationship due to small size, horizontal gene transfer, physiological flexibility and resilience to the environments (Horner-Devine et al., 2004a, 2004b). Yet, taxa-area relationships have been found for bacterial populations driven by environmental heterogeneity (Horner-Devine et al., 2004b). The two distinct clades we found for Richelia from the two stations concurs with the above findings by (Horner-Devine et al., 2004b) or may present a case of allopatric speciation similar to isolated island populations that have emerged from one source population (Turelli et al., 2001). The presence of Richelia within the diatom-hosts in a complex association must also be accounted when examining distribution and diversity. Further investigation into the genetic profiles of the host diatoms would be necessary.

Abundance and Nitrogen fixation

H. hauckii and R. intracellularis were present throughout the year at both stations (TB200 and TB1000) (Figure 6) and were also counted during several night-time samplings. The rare sightings of Rhizosolenia–R. intracellularis associations in our stations correspond to results from a study in the Caribbean and North Atlantic Ocean where 91–100% of Hemiaulus hosts contained the R. intracellularis symbionts while the Rhizosolenia–Richelia associations were rarer (5–254 times less abundant than Hemiaulus–Richelia associations) (Villareal, 1994). Recently, Foster et al. (2007) reported that H. hauckii and R. intracellularis phylotypes were more abundant compared with the other diatom–diazotrophic symbioses in most stations and depths sampled during a cruise to the Western tropical North Atlantic Ocean. Accordingly, they suggested that under similar conditions, the smaller and more rapidly growing H. hauckii and R. intracellularis have a competitive advantage over the other diatom–diazotroph assemblages (Foster et al., 2007).

To date, we did not observe any substantial differences in heterocyst counts between the coastal station and the pelagic station except during October 2006 (11 heterocysts l−1 at the pelagic station and 52 heterocysts l−1 at the coastal station). In both stations heterocyst counts varied between 1–55 heterocysts l−1 and no significant bloom of H. hauckii–R. intracellularis was observed throughout the study period (Figure 6). Large blooms of the diatom–cyanobacterial association have been described in the tropical Atlantic, where heterocyst counts inside the bloom area ranged from hundreds to >1000 heterocysts l−1 compared with 39 heterocysts l−1 outside the blooms (Villareal, 1994; Carpenter et al., 1999; Jahnke et al., 1999). Recent determinations of the H. hauckii–R. intracellularis associations by quantitative (Q)PCR of nifH gene copies showed similar abundances (Foster et al., 2007). The highest numbers of heterocysts found at our coastal station (TB200) were counted in June and October 2006 (∼50 heterocysts l−1) and at the pelagic station (TB1000) in June and November 2006 (∼50 heterocysts l−1). Figure 6 clearly shows a seasonal peak of heterocyst concentrations during autumn (October–November), coinciding with the deepening of the mixed-layer depth at both stations. Dust events, delivering nutrients, in particular P, are also characteristic during these months (autumn and spring) (Herut et al., 2002, 2005) and may enhance diatom–diazotroph occurrence.

Although total Richelia heterocyst counts were low, the actual presence of heterocysts and the expression of nifH detected by reverse transcribed-PCR (Figure 5), indicates active nitrogen fixation. According to our microscopical and molecular identification, R. intracellularis was the dominant diazotroph in the 10–300 μm fractions. For this size fraction N-fixation rates averaged ∼1 nmol N l−1 day−1 at both stations, and throughout the photic zone during the sampling period, and comprised 50%–70% of the total N-fixation rates measured (0.4–4.5 nmol N l−1 day−1 for >1 μm size-fraction). This is consistent with our findings that 54.5% of all sequenced clones were identified as R. intracellularis (see previous section Molecular identification and nifH expression).

A positive correlation was found between heterocyst numbers and the generally low nitrogen fixation rates only during stratified months (Figure 6, insert). We note that, at all times during the study period, to obtain any measurable rates of N fixation or statistically significant numbers of diazotrophs at our stations, we had to pre-concentrate large volumes of water. Species-specific rates of N-fixation by R. intracellularis as well as the total contribution of new N by DDA's and other diazotrophs to the N-budget of the Levantine basin remain uncharacterized and should be carefully examined at a wider scale. Moreover, our sampling and analyses for nifH expression performed predominantly during daytime does not take into account possible contributions of unknown active diazotrophs during night time. Yet, the low rates of N-fixation measured during this study, combined with the low Richelia abundances and lack of DDA blooms (Figure 6), may reflect a scenario controlled by limited availability of essential nutrients as has been recorded for bacterial and primary productivity in the eastern Mediterranean (Herut et al., 2002; Flaten et al., 2005; Zohary et al., 2005; Tanaka et al., 2007).

In present day oceans, diatoms are almost the sole photoautotrophs with an absolute requirement for silicon in the form of silicic acid that is synthesized to form their frustules. Previous studies show that there is no significant difference in the seasonal silicic acid depth integrated values between the 120 (shelf) and 400 m (continental slope) stations (Herut et al., 2000). Moreover, the surface silicic acid concentrations in a westward section off the Mediterranean Israeli coast show no consistent gradient of lower concentrations towards the open sea, though seasonal changes are depicted in the range of 0.5–1.7 μmol l−1 (Kress and Herut, unpublished data).

We examined whether the seasonal and depth-related changes in Si(OH)4 and NO3+NO2 influenced diatom–diazotroph abundance (Figure 7). The ratio between Si(OH)4/NO3+NO2 was used to indicate the relative availability of these nutrients. For non-symbiotic diatoms, blooms can occur when Si, P and bioavailable nitrogen sources (other than atmospheric nitrogen) are high (leading to a low Si(OH)4/NO3+NO2 ratio). A high ratio of Si to bioavailable N could provide a competitive advantage to the diatom–diazotroph associations where N is supplied via biological N-fixation.

Figure 7
figure 7

Correlations between Si(OH)4/(NO3+NO2) and counts of Richelia heterocysts (het l−1) from the coastal station TB200 (a) and the pelagic station TB1000 (b). Heterocysts were always counted within the diatom hosts.

Full size image

We observed a positive correlation between the heterocyst counts (all heterocysts counted were within diatom hosts) and the Si(OH)4/NO3+NO2 (Figure 7). Although during the observations of the H. hauckii–R. intracellularis association the diatom/heterocyst ratio was not necessarily 1:1, we found a positive relationship between heterocyst and diatom counts. As Si(OH)4 concentrations did not increase or decline significantly at the two sampling stations to concentrations that would limit diatom growth, we deduce that the reduction in bioavailable N sources, such as nitrite and nitrate, stimulated increased N-fixation and enhanced the growth of symbiotic diatom–diazotroph associations.

For large-scale blooms of diatom–diazotroph symbioses such as H. hauckii, Rhizosolenia and R. intracellularis to form and persist requires two other essential nutrients: Fe and P. In the Eastern Mediterranean Fe availability may actually be quite high (Guerzoni et al., 1999). In contrast, P concentrations are at the nanomolar scale in the upper layers (Krom et al., 2005) causing intense competition between autotrophic and heterotrophic plankton for this limited resource (Psarra et al., 2005; Thingstad et al., 2005; Zohary et al., 2005).

Diatoms, with relatively small surface area/volume (SA/V), are competitively disadvantaged compared with the picoplankton's large SA/V ratio and high nutrient uptake rates that are suited for this ultraoligotrophic system. Thus, in the Eastern Mediterranean open Ocean where P is extremely limited (Flaten et al., 2005), we would not expect to find large-scale blooms of diatoms (Psarra et al., 2005) or the associations of H. hauckii, Rhizosolenia and R. intracellularis. This is consistent with our findings of limited abundances of both the diatom hosts and the diazotrophic cyanobacterial symbiont requiring high P.

Data from the late Pleistocene Mediterranean sapropels, including samples from the Levantine basin, show an annually-repeating seasonal flux dominated by Rhizosolenid diatoms including H. hauckii (Kemp et al., 1999) Richelia endosymbionts were not actually preserved in the sapropels. Yet the isotopic signatures of 15N/14N from these sapropels, suspended particles and extracted chlorophyll, have prompted claims that large diazotrophic blooms occurred and contributed 70–90% of new N to the Eastern Mediterranean (Sachs and Repeta, 1999; Pantoja et al., 2002). Those blooms could have formed in the past due to a shallower pycnocline (resulting from larger freshwater fluxes into the basin with high Si and P) and thus nutrient-enrichment of the euphotic layer (Sachs and Repeta, 1999; Pantoja et al., 2002). Moreover, such enhancement of carbon export may have triggered or amplified denitrification rates generating nitrogen deficient-water that would fuel subsequent nitrogen fixation (Deutsch et al., 2007).

Will large blooms of DDA's occur again in the Mediterranean? Future scenarios predict the enhancement of the scale and frequency of anthropogenic and/or atmospheric (for example, via Saharan dust) input of nutrients having combined N:P ≫16 and high Fe. These conditions could result in one of two scenarios. If P is still limiting, and limited denitrification continues, nitrogen-fixation in this area would remain low. Alternatively, the geophysical characteristics of the semi-enclosed basin of the eastern Mediterranean combined with higher inputs of nutrients could increase the inventories of limiting resources such as P as well as those of dissolved Fe and shift the system towards mesotrophy. Moreover, the global shift in temperatures may cause reduced circulation and the formation of anoxic conditions as well as an increase in shelf areas due to enhanced coastal flooding. These conditions could promote increased denitrification rates and subsequent loss of dissolved nitrate source, thereby enhancing nitrogen-fixation and primary productivity (Deutsch et al., 2007) and enabling diatom–diazotroph blooms.