Cyanobacteria, one of the most diverse bacterial phyla found in a wide array of habitats, are adapted to a broad range of environmental conditions. The ability of some cyanobacteria to fix atmospheric nitrogen makes them unique because of their carbon and nitrogen autotrophy. Cyanobacteria play an important role in building soil fertility in rice fields . Several studies also underlined the role of cyanobacteria as useful biological agents in remediation and amelioration of soil environments . The practice of using cyanobacteria as inoculants, also known as cyanobacterisation, has been extensively studied [2,3]. Dash et al. , found that heterocystous cyanobacteria contributed to nitrogen economy better than unicellular ones. Also, indigenous cyanobacterial isolates may be better adapted to local conditions . Due to the importance that has been given to the cyanobacteria in the contribution of nitrogen in other rice fields around the world, it was important to have information about their function in the Uruguayan ricefields. These ricefields had a previous history of a three years pasture or fallow and are dry-seeded with flooding at tillering. The density of cyanobacteria in these rice fields was lower than those reported in other ricefields of the world , so the inoculation with isolates of native cyanobacteria appears as a possible nitrogen supplement in these ecosystems . Therefore, knowing what genera of cyanobacteria are present in Uruguayan ricefields is relevant. Traditionally, the identification of cyanobacteria was based on their morphological characteristics. Their morphological diversity has been used to divide the group into five subsections . The first three subsections comprises unicellular cyanobacteria and filamentous cyanobacteria that do not form heterocysts , the subsection IV (Nostocales) comprises filamentous heterocyst-forming cyanobacteria without true branching , and the subsection V (Stigonematales) comprises filamentous heterocyst-forming cyanobacteria with true branching . In the subsection IV and V, vegetative cells can differentiate into morphologically and ultrastructurally distinct cells, called heterocysts, that are specialized in nitrogen fixation under aerobic conditions, but also into akinetes, resting cells that survive environmental stresses, depending on growth conditions . The morphological classification does not include the genetic information, and the most of the genetic sequences present in the databases lacks a morphological description . In consequence, a review of the classification of these microorganisms was pointed as necessary . The nucleic acid-based analyses are the most accepted tool for assessing microbial diversity . The main gene used as molecular marker in bacteria is the 16S rRNA gene, because of its universality and conservation and the presence of hypervariables regions . The morphological classification of Cyanobacteria had not been supported by the analysis of the 16S rRNA gene . When this gene was analised, the cyanobacteria of subsections IV and V belong to the same monophyletic group . The resolving power of the 16S rRNA gene is at the species level or above it . For this reason, it is advisable to carry out a phylogenetic analysis with other molecular markers, such as the hetR gene, essential gene for the differentiation of the heterocyst , the nifH gene, that encodes for the Fe protein of the nitrogenase , the rbcL gene, that codifies the large subunit of the Ribulose 1, 5-bi- phosphate carboxylase / oxygenase , among others. However, the isolated analysis of each of the attributes is not enough to classify them and a polyphasic approach is more appropriate and reliable . This approach involves the analysis of morphological, molecular, biochemical and physiological characters. During heterocyst differentiation, there is an increase in the level of Glutamine Synthetase (GS) which is a key enzyme for the nitrogen assimilation . Other enzymes, as Nitrate Reductase (NR), work in the reduction of nitrate to ammonia . The analysis of N-assimilating enzymes in cyanobacteria may be besides a valuable tool for classifying these organisms, an important characteristic when selecting them to be used as biofertilizer. The objectives of this study were: • To characterize ten indigenous filamentous heterocystous nitrogen-fixing cyanobacteria isolated from an Uruguayan rice field, according to morphological features and phylogenetic analysis; • To evaluate the isolates in relation to growth rates, pigment composition, nitrogenase, nitrate reductase and glutamate synthase activities.
Materials and Methods
Cell growth and culture conditions
The ten heterocyst forming cyanobacteria, Su1, Su5, Su8, Su10, Su11, Su16, Su37, Su40, Su45, and Su54, have been isolated from Uruguayan rice paddy fields  and kept lyophilized. Cyanobacteria cultures were grown in Erlenmeyer flasks under constant agitation in BG11° medium  at pH 7.6 and buffered with 10 mM HEPES. Cultures were kept at 28°C ± 2°C under white light supplied by fluorescent lamps providing a photosynthetic photon flux density of 50 mE m-2 s-1 (16:8 light: Dark period). For estimating the growth rate, growth in BG11° liquid media was assessed via time series measurements each 2 days during 10 days using Optical Density (OD) at 750 nm as a pigment independent measurement with a Shimadzu UV-1201V spectrophotometer. All measurements for the rest of the determinations were done at the exponential phase of each culture, except when expressly indicated.
The isolates were viewed under a Nikon Labophot-2 microscope and the nature of filaments and the shape and size of vegetative cells, heterocysts and akinetes, were analysed and assigned to different genera, using the keys of the Bergey’s Manual of Determinative Bacteriology . For the observation of akinetes and hormogonia, the cultures were left in the same medium during a month without agitation.
Molecular and phylogenetic analyses
Genomic DNA was isolated according to Zhou et al. . The purity of DNA was checked by measuring A260/ A280 ratio in a NanoDrop 2000 (Thermo Scientific) spectrophotometer. Genomic DNA quality was checked in a 0, 8% agarose gel and photographed in a gel documentation system (Bio-Rad, USA). Partial 16S rRNA gene sequence was amplified employing a universal primer for Eubacteria 27F1 (5’-AGAGTTTGATCCTGGCTCAG-3’)  and a specific primer for Cyanobacteria 809R (5’-GCTTCGGCACGGCTCGGGTCGATA-3’) . PCR reaction mixture was prepared with 2.5 μl of 10× Taq buffer, 1 μl of 10 mM of each deoxynucleotide, 0.5 μl of 10 μM of each primer, 0.20 μl of 5U Taq DNA polymerase and 2 μl of DNA extract. The PCR program was: Initial denaturation at 94°C for 5 min, 30 cycles of denaturation at 94°C for 1 min, annealing at 55°C for all the isolates, except for Su5 (temperature of annealing: 62°C) for 1 min and extension at 72°C for 1 min, and finally, a final extension at 72°C for 10 min. The PCR was done in a Thermal cycler Px2 (Thermo Electron Corporation). The hetR gene region was amplified using specific primers for cyanobacteria HETR-F1 (5’-TATCTGGCTTTTAGYGCCATG-3’) and HETR-R1 (5’-CTTGGTGATATTTATCWGCCC-3’) . PCR was performed in a total reaction volume of 25 μl according to the following program: Initial denaturation at 95°C for 5 min, 35 cycles of denaturation at 95°C for 30 s, annealing at 58°C for 30 s and extension at 72°C for 1 min. A final extension at 72°C for 10 min was done. Sequencing of the PCR products was done at Macrogen Inc. (Korea). The sequences were edited manually with the software Chromas Lite versión 2.1. . The partial 16S rRNA gene sequences have been deposited in the NCBI GenBank database under the accession numbers: MH271072, MH271073, MH271074, MH271075, MH271076, MH271077, MH271078, MH271079, MH271080, MH271081. The partial hetR gene sequences were deposited in the NCBI GenBank database under the accession numbers: MH424314, MH424315, MH424316, MH424317, MH424318, MH424319, MH424320, MH424321, MH424322, MH424323.The nucleotide sequences were aligned against all the sequences presents in the NCBI GenBank database by BLASTN (http://www.ncbi.nlm.nih.gov/BLAST). For the phylogenetic analysis, trees were built with the Neighbor-joining method and the Maximum Parsimony (MP) algorithm in the MEGA program 4 . Chlorobium limicola was used as an out group taxa for the phylogenetic tree of the 16S rRNA gene, and Trichodesmium sp. for the phylogenetic tree of the hetR gene.
Chlorophyll a (Chl a) was quantified according to Wellburn . 1 mL of culture was centrifugated at 12,000 rpm during 5 min, previous homogenization with a Potter pestle. The pellet was resuspended in absolute ethanol and incubated overnight at 4°C. After centrifugation, the supernatant absorbance was read at 665 nm. The phycobiliprotein content was determined according to Bennet & Bogorad  by the method of enzymatic digestion with lisozyme and freezing-thawing cycles. The amount of carotenoids was estimated by the procedure of Liaaen-Jensen . 2 mL of culture was centrifugated at 12,000 rpm during 5 min, previous homogenization with a Potter pestle. The pellet was resuspended in acetone 85% (v/v) and incubated 48 h at 4°C. After centrifugation, the supernatant absorbance was read at 450 nm.
Photosynthetic oxygen evolution
Photosynthesis was measured as O2 evolution (nmol min-1 μg-1 Chl a) for 6 min with a Clark-type O2 electrode (Hansatech Instruments Ltd.) as described in Irisarri et al. . 2 mL aliquots of cell suspensions were placed in a controlled cuvette at 27°C and illuminated with a quantum flux density of 400 mE m-2 s-1.
Gas chromatographic quantification of ethylene formed (acetylene reduction activity, ARA) was utilized as an index of nitrogen fixation. Commercially available standard ethylene was used for quantification. ARA was performed in 25 mL aliquots of cell suspensions placed in 60 mL vials incubated with acetylene for 4 hours. Nitrogenase activity was expressed as U mg-1 total proteins, where U was μmol ethylene min-1. For total soluble protein quantification, cultures were centrifuged (6000 g for 10 min) and pellets were resuspended in an extraction buffer (100 mM Tris; 50 mM EDTA; NaCl 100 mM; PMSF (phenylmethylsulfonyl fluoride) 1%, pH 8.0). The protein extract was quantified according to  and bovine serum albumin was used as standard. To perform ARA assay, we selected four isolates, Su5, Su8, Su10 and Su16, due to their morphological differences.
Nitrate reductase activity
NR activity was determined as described by Herrero et al. , which is based on the ability of this enzyme to reduce NO3- to NO2-. Cyanobacteria were previously cultured for 5 days in a BG11 medium with KNO3 20 mM. NR activity was expressed as U mg-1 total proteins, where U was μmol NO2- min-1.
Glutamine synthetase activity
GS activity was performed as described by Dobrogosz . The GS transferase activity was determined measuring γ-Glutamil Hidroxamate (γ-GH) with glutamine as substrate in the presence of ADP and arsenate. GS activity was expressed as U mg-1 total proteins, where U was μmol γ-GH min-1.
All experiments described were conducted independently at least twice to confirm the reproducibility of the results. Analysis of variance and t-test were used to establish differences among the different isolates parameters determined. A confidence level of 95% was used in all analysis. The software used was InfoStat/E .
Results and Discussion
The cyanobacterial isolates used in this study, were selected based on their differential colony growth on solid medium. The morphological characters of the cyanobacterial isolates were observed microscopically. Ten isolates could be identified microscopically according to their morphological features. All of them belong to the order Nostocales (unbranched filaments exhibiting cell differentiation into vegetative cells, heterocysts and akinetes). Within the order, the isolates were divided into two different morphotypes (Figure 1). Most of them had simple isopolar filaments without branching, belonging to the Nostocaceae, Su1, Su5, Su8, Su11, Su16, Su40 and Su54 , and their genus can be ascribed to Nostoc. The isolates Su16, Su37 and Su45 belong to the family Rivulariaceae, as they had whip-like tapering filaments and to the genus Calothrix. Most of the isolates formed straight filaments, except Su40 and Su54 that presented slightly coiled filaments. Heterocysts and akinetes were always found solitary but not in a particular position within the filament. The information provided by all these features, supported the heterogeniety of the isolates. Morphological diversity may be explained by cyanobacteria great acclimatazing ability in response to growth conditions and can lead to miss identification . Zapomĕlová et al. , reported high variability of cyanobacterial morphological traits but an effect of growth conditions was shown only in some of these traits as the occurrence of heterocysts. Traditionally, the classification of cyanobacteria has been based on morphological characters such as trichome width, cell size, division planes, shape and arrangement, pigmentation and the presence of characters such as gas vacuoles and a sheath . These authors estimated that more than 50% of the strains in culture collections are misidentified. Some of these diagnostic features can be lost during cultivation . Such limitations of phenotypic characters have highlighted the requirement for more reliable methods and promoted molecular approaches in cyanobacterial taxonomy .
Heterocytous cyanobacteria form a monophyletic cluster among cyanobacteria on the basis of their 16S rRNA gene sequences . In concert to this, a partial amplification of the 16S rRNA gene was carried out of the 10 isolates. Two clusters were detected with the analysis of an 800 bp fragment of the 16S rRNA (Figure 2). Cluster A held most of the isolates, except Su37 and Su45. Both Su37 and Su45 (cluster B) had been ascribed to Calothrix genus according to their morphological characteristics. Only Su45 gathered with a Calothrix species. The 16S rRNA gene sequence of isolate Su37 did not group with any of the sequences used to make the phylogenetic three (Figure 2). But with the BLASTn against all the sequences in the NCBI GenBank database, this isolate shared 96% of identity with the 16S rRNA gene of an uncultured bacterium (AB659975.1), retrieved from soil of a rice paddy field in Japan . The isolate Su16 (Calothrix-morphotype) presented a 99% identity to Tolypothrix sp. (AM230706.1) and was intermixed with Nostoc species in cluster A (Figure 2).All Nostoc-morphotype isolates were grouped in cluster A (Figure 2). The BLAST output showed that Su10 shared 100% 16S rRNA gene sequence similarity with the strain Nostoc piscinale CENA21 (AY218832.2), coming from the Brazilian Amazon floodplain . Su40 shared an identical 16S rRNA gene sequence with the previously described Nostoc calcicola TH2S22 (AM711529.1), isolated from a Thai not-flooded ricefield . The other isolates were grouped with Nostoc sp. (Figure 2), except for Su5 that grouped with Nostoc piscinale and Su1 that gathered with Nostoc elgonense. Su1, showed a 98% of sequence similarity with a 16S rRNA gene of a Nostoc elgonense (AM711548.1) present in the NCBI GenBank database . It has been proposed that two isolates are members of the same species, if they share at least 97% of similarity of the 16S rRNA gene . On the other hand, it is also suggested that the percentage of similarity should be, at least, 98.7% . So in this case, we cannot say that the isolate Su1 belongs to the specie N. elgonense. A good correlation between morphological identification and genomic clustering was found that enabled the identification of two isolates, Su10 and Su40, at the species level.For most isolates, a good correlation was found between morphological identification and genomic clustering and it was possible to identify two isolates at the species level (Su10 and Su40). The other gene analysed, hetR, is a transcriptional regulator that directs heterocyst differentiation . The amplification product of hetR gene had about 740 bp. The comparison of the edited sequences of the partial hetR gene of the cyanobacteria studied with the NCBI GenBank database did not show any homology equal or greater than 95%, except for Su37 and Su45 that presented a percentage of similarity with the hetR gene of Calothrix sp. (95% and 96%, respectively). According to this gene phylogenetic analysis, the ten isolates may be divided into two major clades, containing the majority of the isolates in one clade and only Su54 in the other (Figure 3). The major clade included isolates of different morphotypes, the Nostoc morphotypes and the Calothrix morphotypes are mixed. This mixture could be explained because the database for hetR gene sequences is smaller compared to the 16S rRNA database . The 16S rRNA gene database is large and so provides a more robust phylogenetic reconstruction which results in a better taxonomic identification. Apart from this, the 16S rRNA gene does not suffer from horizontal gene transfer and is more conserved in function and in structure than the genes that code for proteins (reviewed by Barker et al. ), like the hetR gene.The classification of cyanobacteria is troublesome due to the presence of species based solely on morphology without any genetic information and by the sequences of cyanobacterial species in databases without morphological description . Considering the two genes analysed, the 16S rRNA provided more information for the identification of these cyanobacteria isolates. In spite of this, the isolate Su16 did not posess false-branch filaments, a characteristic attributed to Tolypothrix sp. In the hetR phylogenetic tree this isolate was closer to Calothrix sp. The isolates Su5 and Su10 presented morphological differences as different heterocyst location in the filament, reinforcing that taxonomic identification of cyanobacterial species could be improved through a combination of morphological analysis with the use of molecular techniques . Castenholz & Norris  suggested that more stress must be given to morphological features in the species definition in the case of cyanobacteria. In order to go deeper in the phylogenetical characterization of cyanobacterial isolates, other gene marker such as Internally Transcribed Spacer between the 16S and 23S rRNA sequences (ITS) could be used. This marker has been recently used for the study of intra-species Mycrocistis diversity by Next-Generation Sequencing . ITS regions carry several operons so conventional PCR can randomly amplified any of these operons, implying that some caution should be taken when using them .
Generation time and specific growth rate
The generation time and specific growth rate was compared among the 10 cyanobacterial isolates. All the isolates had a generation time higher than 35 h with the growth conditions used in this work. The growth rate presented a high standard deviation and only Su5 and Su16 exhibited higher growth rate (0.55 ± 0.22 d-1 and 0.67 ± 0.42 d-1, respectively) than Su40 (0.13 ± 0.07 d-1). The growth rate of the rest of the isolates could not be differentiated (data not shown). The growth rate was not related to the genus of cyanobacteria, as one Nostoc sp. and one Calothrix sp. isolates presented the higher values. Isolates within the same genus presented the highest and lowest growth rates. The generation time is a factor to consider when thinking in growing these cyanobacteria to use them as inoculant for rice cultivation.
In order to contribute to assess the diversity of the different heterocystous cyanobacteria isolates we measured their different pigments content (Table 1). The maximum Chl a content was recorded in Su8, but its Chl a content was not different to the content of Su5 (two of the Nostoc-morphotype isolates). Su5, Su8 and Su54 exhibited the highest carotenoid content while Su40 the lowest (Table 1). The content of phycobiliproteins was most variable among isolates. These chromoproteins can represent up to 50% of the protein mass of cyanobacteria cells . Phycoerythrin is not always present in cyanobacteria , although this phycobiliprotein appeared in all the 10 isolates studied (Table 1). Allophycocyanin could not be detected in Su40, but this pigment is reported to be present in all cyanobacteria in nature . The highest phycocyanin/phycoerytrin ratio corresponded to Su37, Su40 and Su45. The colour of each strain would be determined by the combination of the different phycobiliproteins . Cyanobacteria pigmentation changes with environmental light quality and this is known as complementary chromatic adaptation. Bennett and Bogorad  showed that this adaptation was the result of altered phycobilisome pigment-protein composition. For this reason the pigment content has not taxonomic relevance although it can be used to compare different cyanobacteria strains grown in the same conditions .
The photosynthetic activity of the isolates was estimated by their oxygen production. The rates of oxygen photoevolution were similar for all the isolates, although Su45 that exhibited the maximum rate was different, only, to Su11 and Su40 (Figure 4). These two latest cited isolates may not be considered for use as potential biofertilizers. Accordingly, Su40 showed one of the lowest generation times and the lowest carotenoid content which could explain its low rate of photosynthesis. This feature suggests Su40 may not be a good choice to select as inoculant.
Nitrogenase activity was measured with the acetylene reduction assay for the isolates Su5, Su8, Su10 (Nostoc- morphotype) and Su16 (Calothrix-morphotype). Su10 recorded the highest average nitrogenase activity, followed by Su5 and Su16, although the activity of these 3 isolates was not significantly different (p>0.05) (Figure 5). Only Su8 exhibited lower nitrogenase activity than Su10 (Figure 5). A higher heterocyst frequency could not be detected in the isolates with the highest nitrogenase activity (data not shown) as was reported by Kumar et al. . Hrčková et al. , reported higher nitrogenase activity for strains of Tolypothrix than for strains of Nostoc. These authors explained that Nostoc strains produce large amounts of mucus and create thick envelopes for cell protection , reducing diffusion and light penetration and limiting nitrogenase activity. Our results did not show that the nitrogenase activity was related to cyanobacteria morphotype. We found that one isolate belonging to the Nostoc-morphotype presented the highest nitrogenase activity (Su10) and other isolate that belongs to the same morphotype presented the lowest nitrogenase activity (Su8). In spite of the drawbacks of the acetylene reduction assay to estimate nitrogenase activity , the isolates with higher activity may be considered to be used as biofertilizers.
Nitrate Reductase (NR) activity
In spite of the low nitrate reductase activity registered (Figure 6), Su40 showed the highest NR activity, while Su37 the lowest. Most of the isolates did not show significant differences among them. The incubation time with nitrate, the inducer of the activity, may not have been enough .
Glutamine Synthetase (GS) activity
Although some cyanobacterial strains from ricefields were shown to release small quantities of ammonia during their active growth phase , most of the fixed products are made available mainly through microbial mineralization . Strains of cyanobacteria that release ammonium continuously were suggested as a solution to control the N availability to rice plants . Some cyanobacterial mutants were obtained with a totally deficient GS and requering exogenous glutamine , but a number of mutants of nitrogen-fixing strains with low GS activity that continuously release ammonia have been isolated, too . Despite heterocystous-cyanobacteria features are good N2-fixers, their strategies to thrive in the rice fields have to be studied before considering them as potential candidates as biofertilizers. For instance, their strategies to cope with UV-B and osmotic stress were studied by our group [65,66]. It is also important to identify cyanobacterial strains before recommending them to be used as biofertilizers. The poliphasic approach of this work was recommended to identify cyanobacteria to the level of genera . Komárek  also stated that the registration of different stable eco-species is necessary and highly desirable but there are not yet prescriptions for such a practice. In conclusion, our results point to Su5 and Su16 as the best candidates to be used as biofertilizers. These cyanobacteria belong to different morphotypes but they registered the highest growth rate and nitrogenase activity. However, their GS activity was high indicating they may not release ammonium. According to our results and although only two of the isolates were identified at species level, the morphological identification agreed with the 16S rRNA gene phylogenetic analysis. The ten isolates could be ascribed at the genus level to Nostoc or Calothrix and the 16S rRNA provided more information than the hetR gene for the identification of the isolates.
We are thankful to CSIC, UdelaR and ANII (VC grant) for their financial support.