Three soils from the former Lake Texcoco (19˚30'N, 98˚59'W) with different electrolytic conductivity (EC) were sampled in triplicate. As such, nine soil samples were obtained. Soil was sampled by augering the 0 - 10 cm layer of a 1 m²-delimited area 20 times while the 0 - 2 cm layer was discarded. Soil samples were transported to the laboratory in black polyethylene bags on ice, 5 mm sieved separately and stored at −20˚C pending analysis.

Soil physicochemical characteristics were determined as described previously [25] . The EC was measured in a saturated solution extract and pH was measured in 1:2.5 soil-H₂O suspensions using a glass electrode (Table 1). The first soil had a low EC of 1.2 dS・m⁻¹ (considered the Tex-Low soil), the second soil had an EC ranging from 83.1 to 107.8 dS・m⁻¹ (considered the Tex-Med soil), while the third sample had EC ranging from 137.3 to 152.5 dS・m⁻¹ (considered the Tex-High soil). According to the FAO soil classification, a very strong saline soil has an electrolytic conductivity (EC) > 16 dS・m⁻¹ [26] .

Diazotrophic enrichment cultures were obtained by adding subsamples of 0.5 g soil to 125 ml serum-flasks containing 100 ml N-free semisolid medium [27] . The medium contained (g・l⁻¹): 0.1, MgSO₄・7H₂O; 0.5, sodium thioglycolate; 0.008, ferric citrate; 0.008, Na₂MoO₄・2H₂O, and agar 0.1% (w/v). The medium was autoclaved at 121˚C for 15 min. After sterilization, the medium was supplemented with 55 mM filter-sterilized glucose and phosphate buffer (36 mM K₂HPO₄ and 16 mM NaH₂PO₄). Serum flasks were stoppered with cotton plugs and incubated statically at 22˚C for several days. When turbidity appeared, the flasks were closed with sterile rubber stoppers, acetylene added and nitrogenase activity measured after one day. Positive cultures were stabilized by repeated transfers (1% inoculum) with an interval of two weeks.

^aEC: Electrolytic conductivity. ^bValues between parenthesis are standard deviations of the mean.

Nitrogen fixing bacteria were isolated in solid N-free culture medium. Serial decimal dilutions of enrichment cultures in sterile water were done and 10⁻⁵ to 10⁻⁹ dilutions were spread onto N-free Bridges agar and incubated in an atmosphere enriched with GasPack^® system at 28˚C for five days. Five representative colonies of each morphotype obtained were used to confirm nitrogenase activity by the acetylene reduction assay. The positives were conserved in 25% v/v glycerol at −70˚C.

Random amplification polymorphic DNA (RAPD) fingerprints were obtained to distinguish morphologically similar strains. Briefly, genomic DNA from the isolates was obtained from 48 h bacterial cultures with a QIAamp DNA minikit (Qiagen Inc., Valencia, CA) on the QIAcube® apparatus (Qiagen). Extracted bacterial DNA was eluted from the columns with 200 µl elution buffer and stored at −20˚C.

The RAPD fingerprints were generated using the primer OPB01 (5’-GTT TCG CTC C-3’) and the reaction conditions were used as previously reported [28] . The RAPD products were electrophoresed in 1.5% agarose gels in 1 ´ TAE (40 mM Tris, pH 8.3; 20 mM acetic acid; 1 mM EDTA) and stained with SYBR® Gold (Invitrogen Corporation, Carlsbad, CA).

At least two isolates of each group were selected for identification by a similitude and phylogenetic analysis of the 16S rRNA gene partial sequence. The amplification of 16S rRNA genes was done using the universal bacterial primer 8 forward and 1492 reverse [29] . The PCR products were purified using the QIAquick PCR purification kit according to manufacturer’s instructions (Qiagen Inc., Valencia, CA). Sequencing was done by Macrogen Inc. (DNA Sequencing Service, Seoul, Korea).

The nifH and nifD genes fragments were amplified from chromosomal DNA samples using primers designed for this work. These primers were chosen from conserved regions detected in multiple sequence (nucleotidic and aminoacidic) alignments of a broad collection of nifH and nifD genes and design degenerate primer pairs. The designed primers correspond to N19-N38 (5’-TAY GGI AAR GGI GGI ATH GG-3’) and N378-N398 (5’-GGI GAY GTI GTI TGY GGI GGI-3’) of nifH Methanocaldococcus jannaschii MJ-4000-136 (DQ516852), and N181-N200 (5’-CGC GGC TGC GCC TAY GCM GG-3’) and N1309-N1328 (5’-CCK TTC CGY CAG ATG CAY TC-3’) of nifD Klebsiella pneumoniae (X13303). The PCR reactions were done with an initial denaturation step at 94˚C for 10 min, 35 cycles at 94˚C for 60s, at 55˚C for 60 s, and at 72˚C for 60 s and a final extension at 72˚C for 10 min. Reactions contained 20 ng template DNA, 1´ reaction buffer, 50 mM MgCl₂, 0.25 mM of each dNTP, 10 pM of each primer and 1 U Taq polymerase, adjusted to 25 ml. Purified PCR products were sequenced by Macrogen Inc.

Taxonomic assignations were done with the RDP classifier 2.2 at an 80% confidence threshold [30] and based on the Greengenes reference database (version 1210). A collection of taxonomically related sequences obtained from the national center for biotechnology information (NCBI) taxonomy homepage (http://www.ncbi.nlm.nih.gov/Taxonomy/taxonomyhome.html/) and Ribosomal Database Project-II Release 10 (http://rdp.cme.msu.edu) were used for a multiple alignment analyses with CLUSTAL X [31] . Only common 16S rRNA gene regions were included in the phylogenetic analyses. Maximum likelihood analyses were done using MEGA version 5 [32] . “Find best model” tool were used to evaluate the substitution models. For each model, AICc value (Akaike Information Criterion, corrected), Maximum Likelihood value (lnL), and the number of parameters (including branch lengths) were calculated [33] . The confidence at each node was assessed by 500 bootstrap replicates [34] . The similitude percentages were obtained by subtracting p-distance from one.

The nitrogen-fixation (NF) activity of the enrichment-cultures and pure cultures was examined by means of the acetylene reduction activity assay (ARA) as reported by Navarro-Noya et al. [28] . Briefly, strains were grown in 10 ml vials containing 5 ml N-free semisolid medium to determine their nitrogenase activity. A single colony of each bacterial strain was incubated at 30˚C for 72 h. The vials were sealed with rubber stoppers and acetylene was injected to a final concentration of 5% (v/v) by replacing an identical volume of air. The test tubes were incubated at 30˚C for 12 or 24 h. The produced ethylene was measured with an Agilent 4890-D GC-17 gas chromatograph (USA). The gas chromatograph was fitted with a 0.2 ml sample loop on the injector port, a 50/80 Porapak N column (182.88 cm by 3.175 mm) and a flame ionization detector. The temperature of the injector was 40˚C, the detector 250˚C and the oven 55˚C. Concentrations of ethylene were calculated every time samples were analysed by comparing peak areas against a standard curve prepared from a standard concentration of 250 ppm ethylene. Reference strain Klebsiella variicola ATCC BAA-830T was included as positive and Escherichia coli DH10b as negative control.

Total cell protein concentration in the semisolid medium used to determine ARA was measured by the coomassie brilliant blue assay [35] . Bovine serum albumin (sigma no. A5503, Grade V, 99% purity) was used as standard. Prior to the protein determination, the sample was incubated with 10% (w/v) trichloroacetic acid at 90˚C for 20 min, then cooled to room temperature and centrifuged at 8000´ g for 15 min [36] . The insoluble protein was disolved in 0.1 M NaOH at 55˚C for 1 h. Semisolid medium without inoculum were processed as control.

Tolerance to NaCl was determined by broth microdilution according to the CLSI document M27-A2 (2002). NaCl concentrations were varied from 0% to 25% with 1% increments. Briefly, an inoculum suspension was adjusted to a 0.5 McFarland standard and diluted first 1:50 and then 1:20 in R2A medium. Medium R2A contained (g・l⁻¹): 0.5, yeast extract; 0.5, proteose peptone (Difco no. 3); 0.5, casamino acids; 0.5 g, glucose; 0.5, soluble starch; 0.3, sodium pyruvate; 0.3, K₂HPO₄; 0.05, MgSO₄・7H₂O; pH 7.2 [37] . The minimum inhibitory concentration (MIC) microplates were incubated at 28˚C until growth in the medium without NaCl.

The sequence data reported in this paper have been deposited in the GenBank database, under accession numbers from JQ436863-JQ436919.

Nitrogen fixation rate in the Tex-Low soil varied from 172 to 296 nmol C₂H₄ h⁻¹・g⁻¹ soil, while in the Tex-High soil from 7.4 to 30.4 nmol C₂H₄ h⁻¹・g⁻¹ soil (Table 1). Stabilized cultures were obtained after six transfers in N-free medium. Nitrogenase activity in the stabilized cultures of the Tex-Low was 14.2 C₂H₄ h⁻¹・culture⁻¹, in Tex-Med 4.3 C₂H₄ h⁻¹・culture⁻¹ and in Tex-High 4.8 nmol C₂H₄ h⁻¹・culture⁻¹ (data not shown). Gram-positive spore-forming bacilli dominated in all the enrichment cultures.

A total of 39 strains with positive AR activity were isolated from the enrichment cultures. However, after RAPD screening 21 different RAPD-profiles were found (Data not shown). Seven different profiles were found in the Tex-Low soil, 10 in the Tex-Med soil and four in the Tex-High soil.

Strains Tex01-S1, Tex01-S2, Tex01-S3, Tex01-S5, Tex22-S1, Tex02-S4, Tex03-P3, Med01-S2 and Hgh02- P1 were isolated in consortium with negative NF isolates. Consortiums were more frequent in soil with low EC.

Approximately 1430 nucleotides of 16S rRNA gene sequence of each isolate were used for identification based on pair alignments with sequences from databases and relationships in a phylogenetic tree. A majority of the isolates belonged to the genus Paenibacillus (P. durus, P. graminis, P. wynnii, P. massiliensis, P. borealis and P. fujiensis), with similarity percentages of 99.1 - 100 (Table 2). Only three of the isolates were identified as Azospirillum brasilense (Tex03-P3, Tex02-S3 and Tex02-S3c) and were detected in the Tex-Low sample soil. Paenibacillus durus and P. fujiensis were isolated from the Tex-Low soil, P. durus and P. graminis from Tex- Med and P. borealis, P. massiliensis and P. wynii from Tex-High soil.

Consortium members with negative NF activity were also analyzed for identification. Thus, Agrobacterium tumefaciens, Celullomonas hominis, Bacillus circulans, Achromobacter xylosoxidans and Pantoea agglomerans were found (Table 2).

Nitrogen fixation capacity of the isolates was investigated by PCR-amplification of nifH and nifD genes and quantified by ARA (Table 3). Except for the Hgh03-P4 isolate, nif genes were detected in all strains. In general, higher NF activities were found in isolates from the Tex-Low soil. Strains with a high activity (2811 - 4358 nmol C₂H₄ h⁻¹・mg・protein⁻¹) were Tex01-S5 and Tex02-S5 identified as P. fujiensis (first two) and Md01-S1 as P. durus. Strains Tex01-S4, Tex02-P4 and Tex02-S4 showed an NF activity of approximately 1500 nmol C₂H₄ h⁻¹・mg・protein⁻¹.

The isolates grew in medium containing 2% - 10% NaCl. Isolates with a tolerance to 10% NaCl were Hgh02- S1, Hgh02-S1 and Tex01-S5. There was no relationship between source of isolation (soil with low, medium or

^aThe best match was selected using the closest sequence from the phylogenetic tree. ^bSimilarity percentage was estimated by considering the number of nucleotide-substitutions between a pair of sequences divided by the total number of compared bases ×100.

^aNaCl tolerance was determined in R2A medium amended with NaCl.

high EC) and NaCl tolerance.

Maximum-likelihood phylogenetic trees of 16S rRNA, nifH, and nifD genes were constructed to determine relationships between the sequences of the isolates and related organisms from the GenBank database. Phylogenetic trees obtained with 16S rRNA sequences of the diazotrophic isolates and negative NF co-isolates showed that nitrogen-fixing species of the genus Paenibacillus were grouped in three different clusters (Figure 1). Cluster A grouped Tex-Low and Tex-Med isolates, cluster B grouped Tex-Med and Tex-High isolates and cluster C grouped Tex-High isolates. The 16S rRNA phylogram also included sequences belonging to the Proteobacteria and the Actinobacteria phyla, which corresponded to ribosomal sequences of the negative NF co-isolates.

Topology of the Maximum-Likelihood nifH tree displays a similar tendency with isolate sequences grouped in three phylogenetic clusters (Figure 2). Cluster A was formed by 7 nifH gene sequences of isolates from the Tex-Med soil and was related to P. graminis nifH gene (84% of identity). Cluster B grouped 3 nifH gene sequences of isolates from the Tex-High soil and were related to P. wynnii nifH (Hgh02-S4 and Hgh02-P4) and P. massiliensis nifH (Hgh02-S2). Cluster C comprised five isolate sequences from the Tex-Low soil and two from the Tex-Med soil. These nifH sequences were related to a compressed group formed by P. forsythia, P. sabinae, P. durus, P. zanthoxyli and P. fujiensis.