Nitrogen fixation in a non-heterocystous cyanobacterial mat from a mountain river

In situ nitrogen fixation was investigated in a cyanobacterial mat growing on the bed of rocks of the Muga River, Spain. The filamentous non-heterocystous cyanobacterium Schizothrix dominated the mat, showing nitrogenase activity in the light at similar rates to those found in nearby heterocystous Rivularia colonies. N2 fixation in the light was significantly increased by an inhibitor of PSII and oxygen evolution, DCMU (3-[3,4-dichlorophenyl]-1,1-dimethylurea), and anaerobic conditions. However, no nitrogenase activity was found in the dark. Addition of fructose as a respiratory substrate induced nitrogenase activity in samples incubated under aerobic conditions in the dark but not in anaerobic conditions. Microelectrode oxygen profiles showed internal microaerobic microzones where nitrogen fixation might concentrate. Analyses of the 16S rRNA gene revealed only the presence of sequences belonging to filamentous non-heterocystous cyanobacteria. nifH gene diversity showed that the major phylotypes also belonged to this group. One of the three strains isolated from the Schizothrix mat was capable of fixing N2 and growing in the absence of combined N. This was consistent with the nifH gene analysis. These results suggest a relevant contribution of non-heterocystous cyanobacteria to nitrogen fixation in these mats.

Cyanobacteria are frequently key organisms in microbial mats in which the morphology, structure and colour are determined by the dominant species, the sediment features and other environmental factors. Several of the characteristic properties of this unique group of microorganisms, such as oxygenic photosynthesis and the ability of some cyanobacteria to fix atmospheric N 2 , have been proposed as important reasons why they are typically successful mat-building organisms 1 . N 2 fixation, an important process when mats grow in nutrient-depleted environments 2 , can be carried out by representatives of various bacterial taxa (called diazotrophs), and cyanobacteria are generally considered the principal diazotrophs in these ecosystems [3][4][5] . This process has often been studied in microbial mats from marine and hypersaline environments but has seldom been studied in rivers and streams 6,7 . Although substantial progress has been made towards understanding the ecological roles of cyanobacteria in freshwater benthic environments, little is known about their contribution to the N cycle via N 2 fixation 8 .
We have been investigating various aspects of cyanobacterial communities in Spanish rivers, such as species composition and distribution as indicators of nutrient conditions [9][10][11][12] , the genetic and morphological characterization of strains [13][14][15] , and P metabolism [16][17][18] . In the Muga River, a small calcareous river in northeast Spain, we found a pink-pigmented cyanobacterial mat that dominated the bed of rocks of the river and consisted overwhelmingly of the filamentous non-heterocystous cyanobacterium Schizothrix, as previously recorded in this river 17 . The widespread distribution of this mat and the low level of combined nitrogen in the river led us to hypothesize that the Schizothrix mat could fix N 2 . Although cyanobacteria are globally abundant in freshwater benthic environments, the magnitude and importance of N 2 fixation in lotic ecosystems is poorly understood 7,8 . Therefore, this study sought to investigate nitrogen fixation and potentially diazotrophic organisms in the Schizothrix mat as well as ecophysiological strategies related to the ability to perform nitrogen fixation in this habitat. In situ nitrogenase activity was measured, and sequences of 16S rRNA and nifH genes were examined to analyse diazotroph assemblages. Three cyanobacterial strains isolated from these mats were also characterized morphologically, genetically (16S rRNA and the nifH gene) and physiologically. Their ability to grow in a culture medium without combined N Morphological study of the cyanobacterial composition of the Schizothrix mat. Light microscopy studies indicated that the microbial mat was dominated by two distinct morphotypes of Schizothrix, although other filamentous non-heterocystous cyanobacteria of the genera Phormidium sp. and Leptolyngbya sp. were also present, although at much lower frequencies ( Fig. 1D-M). The most frequent morphotype was identified as S. coriacea, which showed long, densely intertwined filaments, typically with up to six trichomes, although it often only had one ( Fig. 1D-F). The cells were 1.7-2.3 μ m wide and the sheaths were firm, thin and non-lamellate, 4.4-9.1 μ m wide, and usually had a colourless appearance under light microscopy. The second morphotype, which was classified as S. vaginatus, had wider trichomes ( Fig. 1G-H) and was characterized by long, intertwined filaments that were enveloped by thick sheaths, usually colourless, 7-8 μ m wide, and formed of cells 2.4-2.8 μ m wide. The morphotype classified as Leptolyngbya sp. was characterized as having long trichomes, straight or flexuous and not attenuated at the ends, formed by cells up to 1 μ m wide, and with apical rounded cells, without calyptra ( Fig. 1I-J). Finally, two morphotypes of Phormidium sp. were observed. One exhibited straight trichomes with cells 3-3.5 μ m wide, cross walls not narrowed, and a rounded end cell, without calyptra (Fig. 1K). The other morphotype had wider, 4.6-5.0 μ m trichomes, which were slightly curved, gradually attenuated to varying extents towards the apex, with a rounded end cell and terminated by a calyptra (Fig. 1L-M). Heterocystous filaments comprised less than 0.1% of the biomass. Three cyanobacterial strains were isolated from these mats, which we designated strains MU50, MU51 and MU52. The morphology of these isolates was very similar, showing straight trichomes with unconstricted cross walls, cells 2.9-3.6 μ m wide, and pointed conical or rounded end cells ( Fig. 1N-P).
Cyanobacterial 16S rRNA gene analysis. The genetic cyanobacterial diversity of the Schizothrix mat was examined by analysing the sequences of amplified cyanobacterial 16S rRNA genes. A total of 29 16S rRNA sequences were recovered in addition to the three isolates, all of which belonged to the order Oscillatoriales (Fig. 2). In the phylogenetic analyses, based on near-complete sequences of 16S rRNA, the sequences formed several clusters (A-E). The majority of the retrieved sequences were clustered in group B, which also included the sequences of the isolated strains. Strain MU51 was grouped with nine other sequences from the mats in subcluster B1. Subcluster B2 contained eight sequences from the mats and those corresponding to the strains MU50 and MU52. All other phylotypes were grouped in cluster E, which was further resolved into subclusters E1-E4. None of our sequences were included in the heterocystous cluster (Fig. 2). These results confirmed the microscopic observations highlighting the dominance of oscillatorian cyanobacteria in the Schizothrix mat.
Physiological experiments with Schizothrix mats. Assays of nitrogenase activity (acetylene-reducing activity), photosynthesis ( 13 C incorporation) and uptake of combined nitrogen ( 15 NO 3 − and 15 NH 4 + incorporation) were performed in situ (Fig. 1B) at midday in ambient light (1400 μ mol photons m −2 s −1 ). To compare rates in this ecosystem, physiological activities were also measured in macroscopic colonies of heterocystous Rivularia, which were located close to the Schizothrix mats in the river 17 (Table 2). The Schizothrix mat showed nitrogenase  activity in the light at a similar rate to that of the Rivularia colonies, with no significant difference detected (t-test, P = 0.198) ( Table 2). No nitrogenase activity was observed in either the Schizothrix mat or the Rivularia colonies incubated in the dark after 30 min of preincubation (data not shown). Specific photosynthetic activity and the uptake of combined nitrogen ( 15 N-NO 3 − and 15 N-NH 4 + ) were lower in the Schizothrix mat than in the Rivularia colonies (t-test: photosynthesis, P = 0.009; nitrate uptake, P = 0.026; ammonium uptake, P = 0.007) ( Table 2). The Schizothrix mat had a low water content and a low ash-free dry weight/dry weight ratio, similar to those of the Rivularia colonies ( Table 2). The chlorophyll (chl) content was significantly lower in the Schizothrix mat, whether referenced to dry weight or to ash-free dry weight (t-test, P < 0.001). Similarly, C content and, especially, N content were lower in the Schizothrix mat (t-test, P < 0.001). As a consequence of having a low N content, the Schizothrix mat had a higher C/N ratio than the Rivularia colonies (Table 2). Nevertheless, converting ethylene reduced the amount of nitrogen fixed, assuming the theoretical value of 3 19 and comparing this value with those of the N-NO 3 − and N-NH 4 + incorporation in the Schizothrix mat and Rivularia colonies, we found that the contribution of N 2 fixation to the total N incorporation in the Schizothrix mat (8.2%) was in the range of that of Rivularia colonies (4.2%), highlighting the importance of N 2 fixation in the Schizothrix mat. The low physiological activity of the Schizothrix mat is in accordance with its low C and N content and suggests a slow growth rate for this mat.  Nitrogenase activity was at its maximum at midday, decreasing thereafter, with very low values at sunset and at night, but increasing again in the morning. These results suggest a major role of light-dependent nitrogenase activity in this mat and therefore the importance of phototrophic cyanobacteria.
Vertical profiles of oxygen concentration and photosynthesis derived in the laboratory from mat samples under similar conditions to those in the field (Fig. 4) showed that oxygen concentration was more than 100% air saturation for the first 500 μ m, followed by a sharp decline in concentration to 30% air saturation at 1500 μ m. This decline was less pronounced thereafter, reaching 20% air saturation at maximum depth. Photosynthetic activity was maximal at approximately 500 μ m and declined progressively to negligible values below 1700 μ m (Fig. 4).
The addition of DCMU, an inhibitor of PSII electron transfer and oxygen evolution, to the Schizothrix mat significantly increased nitrogenase activity in the light (t-test, P = 0.041) (Fig. 5), but the addition of the respiratory substrate fructose to the DCMU-supplemented mat did not induce a further increase in nitrogenase activity in the light. Again, no nitrogenase activity was found in the assays carried out in the dark, but the addition of fructose to the samples incubated in the dark induced substantial nitrogenase activity. These experiments were carried out at midday under the climatic conditions of the river with saturating light (> 1000 μ mol photons m −2 s −1 ) and saturated oxygen concentration in the water (100.6%).
The Schizothrix mat samples collected in the field were used to assess nitrogenase activity under anaerobic conditions in the laboratory because nitrogenase is inhibited by oxygen. Anaerobic conditions were obtained by gassing the samples with argon for 2 h. DCMU was added to the anaerobic samples incubated in the light to prevent photosynthetic O 2 evolution. Incubation under anaerobic conditions did not induce nitrogenase activity in the dark even when fructose, as an exogenous carbon source, was added. However, the anaerobic conditions did induce a significant increase (t-test, P< 0.001) in nitrogenase activity in the light (data not shown).

Parameter
Schizothrix mat Rivularia colonies Ash-free dw/dw 0. 16   Scientific RepoRts | 6:30920 | DOI: 10.1038/srep30920 nifH gene analysis. We retrieved 28 nifH sequences from Schizothrix mats, all of which shared similarity with cyanobacterial sequences reported in the GenBank database. Most of the sequences were tentatively classified as non-heterocystous Oscillatoriales-like nifH based on homology searches performed by comparing the sequences with the public protein databases. A group of nine sequences had similarity to sequences reported in the GenBank database as belonging to the heterocystous order Nostocales. The divergence of the nifH clones examined by DNA sequence ranged from 8.0-29.4%. It would appear that these cloned sequences represent novel populations because no complete similarity was found between our clones and the sequences in the database.
Phylogenetic analysis enabled us to compare the nifH gene sequences obtained from the Schizothrix mat with those available in the public database representative of the orders Nostocales and Oscillatoriales. As most of the nitrogenase sequences published in the database belonged to uncultivated microorganisms that have not been identified, we also included sequences from the environmental samples and strains previously isolated from cyanobacterial communities in the Muga River (Fig. 6). This phylogenetic analysis highlights the clear separation of non-heterocystous and heterocystous cyanobacteria, in which the majority of nifH gene sequences recovered from the Schizothrix mat clustered with the sequences from the order Oscillatoriales. Most of these clustered with the sequence of the isolated strain MU51 (Fig. 6, asterisked). No PCR amplification of the nifH gene was noted for the two other isolates, MU50 and MU52. The Nostocales-like nifH sequences from the Schizothrix mat and the previously isolated heterocystous strains were distributed on different branches mixed with sequences of the database belonging to order Nostocales.
Physiological experiments with isolated strains. The three isolated strains had low growth rates when grown without gassing or stirring on a nitrate medium (Fig. 7A). There were no significant differences between the strains (ANOVA, P = 0.378). All three had a generation time of approximately 9 days. Only the MU51 strain was able to grow in a medium without combined N and maintained its characteristic green colour throughout the experiment. Growth rates were slightly lower than on a medium with nitrate (generation time, 13 days), although  Scientific RepoRts | 6:30920 | DOI: 10.1038/srep30920 the differences were not significant (t-test, P = 0.273). This result suggests that MU51 was able to fix atmospheric N 2 and grow with N 2 as the only source of nitrogen. The other two strains did not grow on the medium without combined nitrogen. They exhibited a progressive decline in chlorophyll concentration until they were pale yellow in colour at the end of the experiment.
The low growth rates of these strains made it difficult to obtain sufficient biomass to assess their ability to fix N 2 with the acetylene reduction assay. Therefore, to avoid a long incubation with acetylene, which can be toxic for the cells 20 , we used the most direct and accurate method of incorporation of 15 Consistent with the findings presented above and confirming its ability to fix N 2 , the MU51 strain showed a small but significant increase in δ 15 N with respect to natural abundance (t-test, P = 0.038) when the cultures were . Neighbor-joining (NJ) phylogenetic tree inferred from nitrogenase amino acid sequences (sequences translated from a 374-bp internal fragment of nifH). GenBank accession numbers for database sequences are in parentheses following the name. Sequences obtained in this study are indicated in bold. The sequence from the isolated strain MU51 is indicated by a dot. Nucleotide sequences obtained from this study were compared with information available from the National Center for Biotechnology. Information database using the Basic Local Alignment Tool (BLAST) (http://www.ncbi.nlm.nih.gov/BLAST), and the most closely sequences related to the sequences obtained were downloaded where available. These sequences, our own ones, and that of an outgroup were used for the phylogenetic analysis. Bootstraps values > 50% are indicated at the nodes. The scale bar represents 0.02 mutations per nucleotide positions.
Scientific RepoRts | 6:30920 | DOI: 10.1038/srep30920 maintained under aerobic conditions for 24 h with 10% 15 N 2 (99% enrichment) (Fig. 7B). The other two strains did not show any increase in δ 15 N, indicating that they were not able to incorporate 15 N 2 . This was confirmed by their lack of growth in the absence of combined nitrogen.

Discussion
Non-heterocystous cyanobacteria as dominant organisms in the microbial mats. Nutrient loading, concentration and ratios are strong selective forces shaping cyanobacterial communities in running waters. Large increases in the proportion of N 2 -fixing cyanobacteria in periphyton communities have been found in response to low concentrations of dissolved inorganic nitrogen (DIN) in river water, especially N-NO 3 − and N-NH 4 +9,12,21-23 . Similar to other mountain rivers, the Muga River showed a low level of combined N. However, the dominant community is a pink-pigmented cyanobacterial mat (see Fig. 1A,B) for which microscopic examination showed it to be overwhelmingly composed of non-heterocystous cyanobacteria identified morphologically as belonging to the genus Schizothrix, as previously recorded in this river 17 . This finding is consistent with that of Stal 1 , who reported that the vast majority of microbial mats are composed of non-heterocystous cyanobacteria. However, the Schizothrix mat in this river has a simple structure because it develops directly on the bed of rocks and lacks the deep layers of sediment characteristic of most microbial mats. This simplicity is reflected in the 16S rRNA gene diversity, the analysis of which revealed only the presence of sequences belonging to filamentous non-heterocystous cyanobacteria (Oscillatoriales). The sequences of our Schizothrix cultures as well as the majority of the other cloned sequences from this mat were included in the same cluster. Another Schizothrix sequence from the database, which was also included in the tree, is not closely related to the sequences obtained from the mat, most likely related to its occurrence in a different (marine) habitat 24 . Similar results have been found in freshwater benthic mat-forming cyanobacteria isolated from New Zealand 25 and other environments, including extreme polar environments and thermal springs 26,27 , whereby filamentous types of cyanobacteria from the order Oscillatoriales with thin trichomes were found to be the most abundant group. It has been argued that the absence of heterocystous cyanobacteria from the microbial mats is due to their lack of gliding motility, which is mainly a characteristic of non-heterocystous cyanobacteria 1 . This property is important for optimal vertical positioning and allows for the compensation of the rapidly shifting physicochemical gradients in microbial mats 1 .
The fact that non-heterocystous cyanobacteria were dominant in a river with low levels of combined nitrogen is striking because to the best of our knowledge, no diazotrophic abilities have been described for non-heterocystous cyanobacteria from running waters. Given its potential for nitrogen fixation and similar values in nitrogenase activity to those found in nearby heterocystous Rivularia colonies, other specific traits may Figure 7. Growth rate (A) and increase in δ 15 N with respect to natural abundance (B) in three strains isolated from the Schizothrix mat. Cells were grown for two weeks in a medium with nitrate (green boxes) or in a medium without combined nitrogen (brown boxes). The incorporation of 15 N 2 was determined in cells cultured for 24 h in a medium without combined nitrogen under aerobic conditions in air supplemented with 10% 15 N 2 (99% enrichment). The raw data are shown as scatter plots.
Scientific RepoRts | 6:30920 | DOI: 10.1038/srep30920 explain niche differentiation. Many mat-forming Schizothrix populations grow in aquatic environments that are subject to alternating periods of great moisture or even being submerged with periods when they are almost or completely dry 28 . Schizothrix mats from the Muga River typically grow in shallow areas with only a thin film of water under low flow and are conspicuous on the dry riverbank. The brick colour typical of the outer part of the mats exposed to full daylight 28 was observed in these parts of the river (see Fig. 1A), most likely as a result of the UV-protection of the extracellular red pigment gloeocapsin in the sheath of trichomes 29 . Therefore, in addition to potential nitrogen fixation, specific ecological niche occupancy in a system with changing flows, such as the Muga River at this site 17 , may explain the dominance of Schizothrix mats in these running waters. Cantonati et al. 30 found clear patterns in the depth distribution of benthic cyanobacteria and algal pigments that indicated adaptation to environmental pressures such as water level fluctuation and light attenuation.

Nitrogen fixation and diazotrophic organisms in the Schizothrix mat. The results show that the
Schizothrix mat fixes N 2 in the light at similar rates to those of a heterocystous cyanobacterium located nearby in the river. It should be noted that these rates are lower than those reported for other cyanobacterial mats, such as those from marine habitats 31 . However, rates of nitrogenase activity in running waters from mountain ecosystems or similar systems with low water temperatures are typically low 32 . The results of our N 2 fixation survey were similar to those obtained from subalpine oligotrophic streams in the Rocky Mountains of North America 33,34 and from Antarctic streams 35,36 . A few exceptions have shown that N 2 fixation rates in streams can be higher. Grimm and Petrone 31 reported high rates in desert streams during summer and autumn. Kunza and Hall 37 found that average N 2 fixation rates were low in nine out of twelve streams in Grand Teton National Park, Wyoming. Variations in N 2 fixation rates have been explained in terms of the availability of inorganic N but also with respect to temperature and light 31,33,34,37 . Nitrogenase activity is greater at lower NO 3 − concentrations in the stream waters, although this relationship was shown to be nonlinear 34,37 . Nonetheless, it should also be noted that N 2 fixation rarely exceeds nitrate or ammonium uptake 37 . Marcarelli and coworkers 7 compared N 2 fixation rates from a wide variety of streams, reporting that the median N fixation was 80-130 times lower than median DIN uptake fluxes. They highlighted the fact that in-stream N 2 fixation rarely contributes 5% of the annual N input to N budgets. This is a similar value to those that we have found. On the other hand, ambient stream water temperature directly affects N 2 fixation. Increased temperatures stimulate N 2 fixation in all streams analysed [31][32][33]37 . However, it is difficult to discern which factor exerts the strongest influence because these parameters tend to covary 31 . Further research is needed to quantify the incidence of such characteristics, which may help explain the spatial and temporal variation in N 2 fixation rates.
The finding that nitrogenase activity occurs in the light was unexpected because it has been widely reported that cyanobacterial mats that are dominated by non-heterocystous cyanobacteria usually only show substantial rates of nitrogenase activity in the dark 38,39 . These results raised the question about what kind of diazotrophs are responsible for the N 2 fixation in this mat. The addition of DCMU to the Schizothrix mat produced a significant increase in nitrogenase activity, as usually occurs in non-heterocystous cyanobacteria 1,3,39 , whereas the addition of DCMU to heterocystous cyanobacteria caused a significant decrease in nitrogenase activity in the light 1,4,40 . The results of the laboratory experiments also indicated that at least one of the main strains of non-heterocystous cyanobacteria from the Schizothrix mat can fix N 2 . This is also consistent with the results of the analysis of the nifH gene. The majority of nifH sequences retrieved from the samples were related to Oscillatoriales-type nifH. Some of these sequences belonged to one of the dominant Schizothrix morphotypes isolated, which in turn was able to grow in the absence of combined nitrogen and to exhibit nitrogenase activity. These results suggested that nitrogen fixation in these mats may be attributed to non-heterocystous cyanobacteria. However, it has been argued that different mechanisms such as nifH-degenerated primers, polyploidy in cyanobacteria and differential DNA extraction could bias nifH clone libraries against heterotrophic diazotrophs 41 . Therefore, the importance of these nitrogen-fixing organisms may have been underestimated in analyses of early successional biological soil crust communities 41 . Therefore, although our results indicate that the primary diazotrophs in the mats are cyanobacteria because all the recovered sequences belonged to this group, in spite of the use of general primers for aerobic and microaerophilic diazotrophs, we cannot discount the potential for fixation by non-phototrophic bacteria stimulated by cyanobacteria-released metabolites in the light 42 .
Ecophysiological strategies. The ability to fix nitrogen in the light requires the development of mechanisms for protecting the O 2 -labile nitrogenase 38 . Paerl and coworkers 43,44 emphasized the importance of the formation of anoxic microzones in microbial mats and other systems in which N 2 fixation could proceed in an uninhibited manner. In our study, the results from oxygen microelectrodes indicate the existence of internal O 2 -reduced microzones where nitrogen fixation may be concentrated. The saturated levels of oxygen found in the upper layers of the photosynthetically active mat would most likely restrict nitrogenase activity to the deeper layers (1300-1600 μ m in depth) where the O 2 concentration is well below saturation. Stal et al. 45 found the highest nitrogenase activity in the lowest layer of a cyanobacterial mat (2-3 mm), indicating a spatial separation of N 2 fixation (lower layers) and oxygenic photosynthesis (top layer).
In addition, we propose that enhancement of oxygen-scavenging mechanisms, such as high respiration rates, may be supplementary ways of protecting nitrogenase activity, as has also been proposed for Trichodesmium 43 . The addition of DCMU had an alleviating effect due to the decreased O 2 production at the water-splitting site of PSII, thereby allowing nitrogenase activity to attain its maximal rate. Under these conditions, ATP had to be supplied mainly by cyclic phosphorylation and the reductant supplied by aerobic respiration, which in turn decreased O 2 concentration. Under strictly anaerobic conditions in the light, nitrogenase also showed maximal rates of activity. Here, again, ATP had to be supplied by cyclic phosphorylation, whereas the reducing power could be derived from fermentative metabolism. Fermentative NADH most likely feeds electrons to the nitrogenase enzyme through PSI, as was reported in isolated heterocysts by Schrautemeier and Böhme 46 . Dark aerobic conditions did not facilitate significant levels of nitrogenase activity in the Schizothrix mat unless exogenous fructose was supplied. The addition of fructose increased respiratory metabolism, thereby decreasing oxygen tension and increasing the supply of ATP and reductant to a level compatible with maximal rates of N 2 fixation. However, under anaerobic conditions, exogenous fructose failed to support nitrogenase activity, indicating that fermentative metabolism by itself cannot provide enough ATP and reductant for nitrogen fixation.
The Mehler reaction, which occurs in cyanobacteria via direct photoreduction of O 2 to water (without the production of significant amounts of reactive oxygen species, as occurs in eukaryotes 47 ), may also alleviate the negative effect of increasing O 2 production at PSII. In addition, increased Mehler activity could generate excess ATP through cyclic electron transport, which is needed to support nitrogen fixation [48][49][50] .
In conclusion, although we cannot discount the possibility that non-phototrophs contribute to nitrogen fixation in the Schizothrix mat, all of our results (the microscopic observations, the 16S rRNA and nifH gene analyses, the effect of DCMU on nitrogenase activity, and the physiological experiments on isolated strains) are consistent with our hypothesis that non-heterocystous cyanobacteria are driving nitrogen fixation in the mats. These diazotrophs may be important contributors to the nitrogen fixed in this river and merit further scrutiny.

Methods
Study site and sampling. Field studies were carried out in the Muga River, Girona, northeast Spain, at a site 10 km downstream from its source in the Pyrenees (42°22′ N, 2°42′ W; UTM ED50 47 48 24 4684320; 310 m.a.s.l.). The field site and water sampling and analysis procedures have previously been described in detail 17 . Samples of the cyanobacterial communities were collected with the help of a knife blade because their structure does not allow the use of a metal core. The fresh weight of all the samples was determined with a field balance after gently placing the samples on filter paper to eliminate water droplets. Samples for determinations and genetic analysis were frozen in liquid nitrogen before transportation. The samples used to measure nitrogenase activity in the laboratory were kept in darkness at 4 °C until further use. Air irradiance during the experiments was measured every 10 min using a Li-Cor LI-1000 data logger equipped with a 2π quantum sensor. Fieldwork was carried out from 9 to 11 May 2009. The river flow during this time was moderate (average of 0.16 m s −1 ).

Community and biomass analyses.
Morphological observations of microbial mats were made under a dissecting microscope (Leica, Leica Microsystems) and an Olympus BH2-RFCA photomicroscope equipped with phase-contrast and epifluorescence. Morphometric measurements were made from digital images captured with an Olympus DP 20 digital camera (high resolution; 1600 × 1200 pixels) interfaced with an Olympus CX41 microscope. Morphological identification followed the taxonomic keys of Komárek and Anagnostidis 51 and Whitton 28 . All of the important key features that were taken into consideration to identify natural samples, such as cell width, cell length, attenuation of apical cells, and sheath morphology, were studied in samples of 25-50 cells. Dry weight was measured by weighing the samples after drying at 60 °C for 24 h. The ash-free dry weight was measured by weighing the dried samples after mineralization at 450 °C for 2 h and calculating the difference between the dry weight and the weight after mineralization. Chlorophyll a (Chl a) content was determined by overnight extraction with 90% acetone using samples that were previously sonicated for 10 min in the dark. After centrifugation at 9,600 g, absorbance was measured at 665 nm using a Hitachi U2000 spectrophotometer. This procedure was repeated with the same sample until no more pigment could be extracted. To determine the C and N content, the samples were dried at 60 °C for 24 h, ground by mortar and pestle and analysed using an elemental analyser (LECO CHNS-932) with a thermal conductivity detector. The percentage abundance in natural samples was evaluated by counting the presence of each morphotype (as cells in a filament or as equal numbers of individual cells).

Cyanobacterial isolation.
To simulate the chemical conditions of the river water, cyanobacteria were isolated on Petri dishes with a 1.5%-modified CHU No 10 52 agar medium in which P was reduced to 1.5 mg l −1 and calcium was increased to a final concentration of 0.15 g l −1 . Three strains were successfully isolated from the Schizothrix mats from the Muga River, each from a single trichome. Photosynthetic activity. Photosynthetic activity was measured using the stable isotope technique ( 13 C uptake) as previously described 53 . To prepare the assay, 10% of the natural dissolved inorganic carbon (DIC) concentration of the water was added as H 13 CO 3 (99% 13 C atoms) (Cambridge Isotope Laboratories, MA, USA). 13 C isotopic enrichment was measured with an IRMS Micromass-Isochrom mass spectrometer. The DIC content of the water was calculated from its alkalinity, determined by acid titration, taking the pH and temperature into consideration.
Microelectrode studies of light-saturated photosynthesis. Gross oxygenic photosynthesis was estimated at depth intervals of 200 μ m through the vertical profile of the microbial mat samples with a polarographic Clark-type oxygen microelectrode (diameter 50 μ m) (Diamond General, Ann Arbor, MI, USA) as previously described 36 . Two profiles of photosynthetic activity were derived at different positions on each of the three replicates of each mat.
Uptake of ammonium and nitrate. The uptake of N from ammonium and nitrate was measured using the stable isotope technique as previously described 53  Nitrogenase activity. Nitrogenase activity of the Schizothrix mats was measured using the acetylene reduction technique as previously described 4 . In the field experiments, samples of the mats were incubated in situ in triplicate in 75 ml flat plastic bottles (1 g fresh weight per bottle) containing 15 ml filtered river water. Control bottles containing river water without mat samples were also incubated to check for other possible sources of ethylene aside from the cyanobacterial communities. The plastic bottles were capped with reversible rubber stoppers and covered with parafilm to ensure gas-tightness. A total of 10% of the air inside the bottles was replaced by the same volume of chemically pure acetylene (Carburos Metalicos, Madrid, Spain), which contained approximately 0.05-0.075 nmol/ml of contaminant ethylene. The samples were incubated for 4 h on the streambed to maintain natural conditions of temperature and illumination (see Fig. 1B). The samples were secured to the streambed as described above. Special care was taken to ensure that the samples were positioned with the surface layer facing upwards and to avoid overlapping. At the start and at the end of incubation, duplicate gas samples were collected with double needles in 10 ml pre-evacuated BD Vacutainer tubes (Belliver Industrial Estate, Plymouth, UK). These tubes give a peak at the same retention time as a concentration of ethylene equivalent to 0.05 nmol/ml. The ethylene concentration was determined in 1 ml gas samples using a gas chromatograph (Shimadzu model GC-8A) equipped with a flame ionization detector and Porapak N 80/100 column. The detection limit of the chromatographic method was 0.01 nmol ethylene/ml. To improve the reliability of the data, any values below 0.05 nmol ethylene/ml were not taken into consideration. To eliminate any errors arising from contaminants, ethylene production was quantified as the difference between the final and initial samples. A mat sample was considered to exhibit nitrogenase activity if differences in ethylene concentration between the initial and final gas samples were significant (paired t-test). Samples incubated in the dark were covered with aluminium foil and kept in darkness at ambient water temperature for 30 min before adding the acetylene. Fructose (1 mM) and/or DCMU (3-[3,4-dichlorophenyl] -1,1 dimethylurea) (10 −5 M) were added 15 min before adding the acetylene.
In the laboratory experiments, samples of the Schizothrix mat were incubated in triplicate in 40-ml Erlenmeyer flasks (approximately 0.5 g fresh weight per flask) containing 10 ml of filtered river water. Samples were incubated at room temperature (22 °C) with a constant saturating illumination of 450 μ mol photons m −2 s −1 . Anaerobic conditions were obtained by gassing the samples with argon for 30 min. DCMU (10 −5 M) was added to those anaerobic samples incubated in the light to prevent photosynthetic oxygen evolution. Before starting the experiment, samples were kept in filtered river water at room temperature for 20 h to allow full rehydration and acclimation to room temperature. DCMU or fructose (1 mM) was added at the same time as the filtered water. Gas samples of 1 ml were taken at the start and finish of the incubation period, and ethylene production was measured as described above.
Nitrogen fixation of the isolated strains was measured using the stable isotope 15 N 2 . The samples were incubated in 10 ml Vacutainer tubes, which were filled to overflowing with culture medium without nitrogen before being carefully sealed with a septum cap. A gas-tight syringe was used to inject 0.5 ml of 15 N 2 (99% of 15 N 2 atoms) (Cambridge Isotope Laboratories, MA, USA) into each tube. Finally, a 0.5-ml volume of the solution was withdrawn to equalize the pressure across the septum. Two replicates plus one control (without 15 N 2 gas) were incubated with light:dark periods of 16:8 h at a temperature of 18 °C and a light intensity of 20 μ mol photons m −2 s −1 during the light period. After incubation, samples were rinsed three times with culture medium without nitrogen to eliminate any unincorporated 15 N. The samples were dried at 60 °C and ground with a mortar and pestle. Additional samples of each culture were taken before the start of the experiment to determine the natural abundance of 15 N. The natural abundance of 15 N and isotopic enrichment were measured using an IRMS Micromass-Isochrom Mass Spectrometer. Genetic analyses. DNA was extracted as previously described 11,21 . The nifH segment (374 bp) was amplified using two primer sets designed by Olson et al. 54 . One of the primer sets was designed to amplify aerobic and microaerophilic diazotrophs (forward primer: 5′ -ATXGTCGGXTGXGAXCCYAARGC-3′ ; reverse primer: 5′ -ATGGTGTTGGCGGCRTAZAKYGCCATCAT-3′ , where X = C or T, Y = G or C, R = G or A, Z = C, G, or A, and K = G or T), while the other preferentially amplified cyanobacterial nifH fragments (forward primer: 5′ -CGTAGGTTGCGACCCTAAGGCTGA-3′ ; reverse primer: 5′ -GCATACATCGCCATCATTTCACC-3′ ). The PCR conditions are described in Díez et al. 55 and Zehr et al. 56 . Almost the entire 16S rRNA gene, the intergenic transcribed spacers, and part of the 23S rRNA gene (approximately 1800 bp) were also amplified from the DNA using the universal 16S ribosomal DNA primer pA (forward primer: 5′ -AGAGTTTGATCCTGGCTCAG-3′ ) 57 and cyanobacteria-specific B23S (reverse primer: 5′ -CTTCGCCTCTGTGTGCCTAGGT-3′ ) 58 under previously described conditions 59 . The amplified products were separated by cloning using a Qiagen PCR Cloningplus kit following the manufacturer's instructions. Recombinant clones carrying the correct size insert were purified and sequenced as previously described 11 . The MALLARD program 60 was used to identify anomalous 16S rRNA gene sequences within multiple sequence alignments. The nifH sequences were translated and the amino acid sequences were aligned. Multiple sequence alignments were performed using CLUSTAL W 61 in the current version of the BIOEDIT program 62 . The phylogeny of the deduced nitrogenase amino acid sequences or 16S rRNA gene sequences was analysed with MEGA software v 6.06 63 . Trees were constructed using neighbour-joining (NJ) algorithms 64 . Bootstrap analysis of 1000 replicates was performed for each consensus tree 65 . All clone sequences were submitted to GenBank under accession numbers JQ514111-JQ514149 and KF544966 for nifH sequences and JQ612137-JQ612154 and KF544967-KF544980 for 16S rRNA sequences.
Statistical analysis. ANOVA with Tukey post hoc testing, independent sample t-tests, and paired sample t-tests were used to compare group differences. The SigmaStat program V2.03 was used for all statistical procedures.