In contrast to the oxidized sediments in the vicinity of the basins, which had a beige-brown colour at the top, the anoxic sediments within the l'Atalante and Bannock basins were dark grey with a black surface layer approximately 1-cm thick, that was less compacted than the underlying layer and very easily disturbed. The top layer of the Discovery basin sediments was jet-black, and also not very compacted and very viscous owing to the high magnesium chloride content. There was no obvious layering of this sediment. The sediments of the Urania basin in contrast were of a light grey color. Gas bubbles developed in the Urania sediment shortly after retrieval of the multicorer, possibly due to out-gassing of methane or carbon dioxide under atmospheric pressure. The upper two centimeters of the anoxic sediments were nearly liquid, so they could only be sampled with a pipette and it was not necessary to mix the sediment with ambient water to form slurries before processing.

The highest colony-forming units (CFU) per ml of sediment were observed on artificial seawater, with estimated numbers of 1.1 × 10⁴ for l'Atalante, 4.5 × 10³ for Bannock, 1.3 × 10³ for Urania and 5 × 10² for Discovery basins. The number of CFU was generally one order of magnitude lower on medium with elevated salinity (12%, medium B/2) than on artificial seawater. No growth was observed on medium B with 24% salinity. Total cell counts were not available from this study, but Sass 24 obtained a total cell count from Urania basin sediment of 3.6 × 10⁷ cells ml^-1 which was similar to counts obtained by Danovaro et al. 14 for l'Atalante basin sediments (5.7 × 10⁷ cells ml^-1; converted using a density value of 1.23 g ml^-1 5), indicating that we cultivated about 0.004 to 0.02% of the microbes in the sediment.

The lowest counts were found in Discovery brine-lake sediment, and given that 5 M MgCl₂ inhibits growth but preserves cell structure 9, it can be assumed that this provides a baseline for inactive but culturable microorganisms in the basin. This is supported by the findings of Polymenakou et al. 15 that PLFA profiles from sediments in the Discovery basin were similar to those from 'normal' hemipelagic Mediterranean sediments, suggesting that the majority of the source organisms for the PLFA were introduced into the basin by sedimentation or lateral flows. Sediments of the other three basins (l'Atalante, Bannock and Urania) had PLFA patterns that clearly differed from those of sediments outside the brine lakes revealing distinct and probably active communities 15. Two questions emerge from these observations: were isolates similar between the four basins, and do they demonstrate the potential to be active in situ?

A total of 89 strains were isolated from the sediment samples (29 from l'Atalante, 30 from Bannock, 16 from Discovery, and 14 from Urania basins). Screening by Amplified Ribosomal DNA Restriction Analysis (ARDRA) and partial sequencing of the 16S rRNA genes revealed that 80 strains (90% of those isolated) were related to the Bacillales. The other nine strains were related to Halomonas aquamarina (4 strains), Pseudomonas sp. (4 strains) and Alteromonas sp. (1 strain). Since Alteromonas and Halomonas spp. were previously shown to be dominant organisms in the chemocline but less abundant in the brine 6 we assume that they were not indigenous sediment bacteria but introduced by sedimentation. Spore-forming bacteria however were more typically isolated from brine-lake sediments, and only rarely from the chemocline (data not shown). Therefore, in the present study we focused on isolates affiliating with the Bacillales. Out of the 80 strains from the Bacillales 25 representative strains with differing ARDRA profiles were selected for physiological characterization and detailed phylogenetic analysis. Of these 25 ARDRA groups, 17 were from one basin only, six were found in two basins, and two were found in three basins. There were no clear trends in the distribution of the isolates in the different basins; however, although numbers are relatively small, it is notable that one of the ARDRA groups represented by strain AS5, had 14 strains that were isolated only from l'Atalante and Bannock basins, the most geochemically similar, yet geographically distant basins.

Most of these Bacillus-like isolates are closely related to previously cultivated organisms (Fig. 2), many of which are moderately halophilic or alkaliphilic. Six strains (from l'Atalante, Urania and Bannock, but not from Discovery basin) belong to a cluster accommodating genera with many halotolerant representatives like Halobacillus, Virgibacillus and Pontibacillus (see Fig. 2) 2526. All of these strains were isolated on a medium with elevated salt concentration (12% NaCl).

The isolates were Gram-positive and, with the exception of strain BS3, formed endospores. All but four isolates were found to be motile by peritrichous flagella (Table 1). The pairs of strains that were phylogenetically most similar, such as AS2 and AS3, US2 and US4, US7 and BS29, and AS6 and BS6 (Fig. 2) were also phenotypically most similar (Fig. 3). The isolates were physiologically diverse (Table 1) particularly when considering their enzyme activities and substrate utilization capacities (Table 1). Generally, the isolates grew well on yeast extract, peptone, casamino acids and a range of carbohydrates and amino acids, while fatty acids or alcohols were less commonly used. A few strains grew also on aromatic compounds or n-alkanes. Compared with the other isolates, strains DS1 and DS5 had limited catabolic capabilities, growing only on five and three substrates, respectively. Interestingly, one of them (strain DS1) grew well on n-alkanes.

All strains were mesophilic, growing between 12 and 37°C, indicating that they could grow within the temperature range experienced at the surface of the Mediterranean and the deep-sea, including the brine lakes which have a temperature of 13.9 to 16.7°C 5. Some isolates also grew at 4°C, whereas others grew at 45 or even 53°C (Table 1). With the exception of strain US13 (minimum salinity for growth of 0.5% NaCl), all strains grew in media containing only 0.1% NaCl, and most had a growth optimum between 1 and 5% NaCl, suggesting that they were marine and incapable of growth in the brine-lake sediments. None of the five isolates from Discovery brine-lake sediment grew in media with an MgCl₂ concentration exceeding 7.5%, and therefore were not adapted to in-situ conditions. However, the eight strains isolated on medium B/2 with 12% NaCl were generally the most halotolerant (Table 1). Three of these strains (AS2, AS3 and US13) were capable of growing at extremely high salt concentrations, similar to those in the basins. They grew up to 30% NaCl with an optimum of 10–15% NaCl. Strains AS2 and AS3 also grew by fermentation in anoxic artificial seawater as well as in medium with 20% NaCl (Table 1), and all three strains (AS2, AS3 and US13) could use dimethylsulfoxide and trimethylamine N-oxide as terminal electron acceptors (data not shown).

The high proportion of Bacillus-related strains among the isolates in this study, although no pasteurisation was performed to select for spore formers, suggests that many the Bacillus-like isolates might originate from dormant endospores and that most of the non-spore-forming marine microorganisms that might drift into the brine do not survive. This is supported by the general lack of correlation between the affiliation of the isolates and the basin of origin, and especially by failure of many strains to grow at in-situ salinities. Endospores can be stained by the commonly used dyes like acridine orange and DAPI 27, and so can be expected to contribute to the total cell counts. Endospore numbers were estimated from dipicolinic acid (DPA) contents 28 in sediment samples from the Urania (3.8 × 10⁵ g^-1 dw sediment), Bannock (9.0 × 10⁵ g^-1 dw sediment) and Discovery (1.2 × 10⁶ g^-1 dw sediment) basins. These numbers are lower than those obtained from highly active coastal sediments where endospores were estimated to represent up to 3% of the total cell counts 28. However, considering the low total cell counts in the range of 2.6 and 5.7 × 10⁷ cells ml^-1 (note the different unit) in the brine-lake sediments it is apparent that the contribution of endospores to the brine-lake sediment microbial communities is in the same range (up to 5% of the total cell counts).

The capacity to form endospores may explain why Bacillus-like organisms can be isolated from almost all types of environment, e.g. soils, sediments, foods and water 29, including extreme environments like solar salterns 3031, deep-sea hydrothermal vents 32 or brine-lake sediments (this study). Whether they play an important ecological role in most of these environments remains a matter of debate. In studies using a combination of culture-dependent and -independent methods members of the Bacillales are often found among the isolates, but rarely in DNA-based detection methods 3334, suggesting that spore-forming bacteria are easily cultivated with standard methods although they represent only a small proportion of the microbial community. This might be explained by the fact that bacteria in most environments are strongly energy-limited, and it was shown that starved cells can be severely harmed by a substrate shock after exposure to high substrate concentrations as found in many standard media 35. In contrast, endospores do contain enough energy for germination and are specifically adapted to quickly respond to substrate availability and by forming a vegetative cell they are able to proliferate. Therefore, it can be expected that they can easily outgrow other organisms after transfer onto microbiological media. On the other hand, endospores possess a very rigid and resistant cell wall and it is unclear to what extent endospores are successfully lysed and extracted using standard nucleic-acid extraction procedures [e.g. 36]. The spores themselves may have an important ecological role, especially as they contribute to such a high proportion of total counts in sediments. For example, spores have been shown to catalyse the oxidation of Mn²⁺ 37.

Spores of Bacillus-like organisms can rest dormant for long time periods and they are resistant to damage through desiccation and radiation for example 23. Bacillus-like organisms have been isolated from deep subsurface sediments 2122 and from inclusions inside materials like amber 38, salt crystals 39 or glacial ice 4041 where the spores must have been included since the time of deposition. The age of the inclusions in the salt crystals and amber was estimated in the range of several million years 3839, the age of those in glacial ice in the range of 5 to 750,000 years 4041.

Therefore, the spores in the basin sediments may have drifted into the brines from environments far from the deep-sea hypersaline anoxic basins, accumulated in the brine-lake sediments and remained viable. The high salinity within the brine lakes could even have enhanced the preservation of the spores 9144243. Alternatively it can be envisaged that they originate from the evaporites, as viable spores were reported from salt deposits tens of millions of years old 39, far older than the Messinian evaporates beneath the Mediterranean Sea.

About two thirds of the strains tested did not grow under in-situ salinities and probably derived from dormant spores, whereas seven strains were very halotolerant and grew at salinities close to in-situ values. However, in the fermentation assay only two of these (strains AS2 and AS3) formed cultures turbid enough for photometric analysis. This is not unusual; many anaerobic bacteria form only low biomass. This can be explained by the low energy yield available from many fermentations. Otherwise many bacteria may rely on the activity of a syntrophic partner which is lacking when grown in pure cultures. An example of such bacteria are strains related to the genus Rhizobium isolated under anoxic conditions from Mediterranean sediments 44. None of these formed turbid cultures under anoxic conditions. Rhizobium-related organisms were unexpected for deep-sea sediments, but were proven to be abundant members of the sediment microbial communities by quantitative PCR 45. Whether the same is true for the Bacillus-like organisms in the brine-lake sediments needs to be investigated further. In any case, our data confirm that it is possible to isolate Bacillus-like strains that are physiologically capable of growing under the conditions in the brine, i.e. extremely high salinity combined with anoxia and a sulfide concentration of several millimol per liter.

Four different brine lakes in the Eastern Mediterranean were sampled between 20^th and 31^st August 2001 on a cruise with the R/V Urania. Three of them, l'Atalante, Discovery and Urania basin, are situated very close to each other on the north-eastern edge of the Mediterranean Ridge 47, only a few kilometers apart, whereas the Bannock basin is located on the southern edge of the ridge 48 (Fig. 1). Samples from the upper two centimeters of the sediments were taken with a box corer (Discovery basin) or a multicorer (all other basins). Immediately after retrieval, the sediment was aseptically transferred into serum bottles. The headspace was flushed with oxygen-free nitrogen and sealed by means of butyl rubber stoppers. Until processing, the samples were stored in the dark at room temperature. Dilution series from the sediment samples were made with a sterile solution of 25% NaCl.

For enrichment and isolation of bacteria six different oxic media were used, representing a combination of three different salt concentrations and two mixtures of organic substrates. Medium B reflected the salt composition of the Bannock basin 48; the salinity was approximately 24%. It contained as major salts (in g l^-1): NaCl 181.2, MgCl·6H₂O 61, KCl 7.5, CaCl₂·2H₂O 2.2, Na₂SO₄ 14.2, MgBr₂·6H₂O 1.2, H₃BO₃ 0.3, TRIS 2.4. Minor constituents were added from stock solutions (in mM final concentration): SrCl₂ 0.05, LiCl 0.075, NaF 0.07, NH₄Cl 0.5, KH₂PO₄ 0.1. One ml of a trace element solution (SL10, 49) was added and the pH adjusted to 7.6 with HCl. A solution of 10 vitamins 50 was added after autoclaving. Medium B/2 resembled medium B, but contained only half of the amount of the major salts. Medium SW consisted of artificial seawater which contained (in g l^-1): NaCl 24.3, MgCl₂·6H₂O 10, KCl 0.75, CaCl₂·2H₂O 1.5, Na₂SO₄ 4. The minor constituents, buffer and pH of the artificial seawater were as described above, additionally MgBr₂ and H₃BO₃ were added from stock solutions to a final concentration of 0.4 mmol l^-1 each.

The organic substrates are denoted as CPS (casamino acids 0.5 g l^-1, bactopeptone 0.5 g l^-1, soluble starch 0.5 g l^-1, glycerol 13 mmol l^-1) or as M, a mixture of the following substrates (in 1 mM final concentrations, pH 7): glucose, glycerol, sodium acetate, sodium lactate, sodium fumarate, sodium succinate, alanine, glycine, sodium glutamate and sodium thiosulfate.

After six weeks of incubation at 20°C colonies from each positive plate were selected after their appearance, and isolated by streaking out repeatedly on the respective medium. Isolated strains were maintained on CPS-SW at 20°C. The number of CFU ml^-1 sediment was estimated from the mean number of colonies on the seawater-based medium with organic components CPS and M, using the calculation of Cavalli-Sforza 51.

Tests for the presence of oxidase and catalase, Gram-staining and staining of flagella and spores were performed according to standard procedures 52. For further enzyme tests the API 20 E-system (bioMérieux, Marcy L'Étoile, France; 53) was used following the manufacturer's instructions. The degradation of Tween 80 and the following polymers was determined on CPS-SW agar (without starch) at 30°C, unless indicated otherwise. All results were confirmed with two independent replicates of each test. Amylase activity was detected by zones of clearing around colonies on medium with 2 g l^-1 of soluble starch (BDH Laboratories) after flooding plates with Lugol's iodine solution. TWEEN 80 esterase activity was detected by formation of precipitate around colonies on medium containing 1% w/v TWEEN 80 (BDH Laboratories). Gelatin liquefaction was tested by adding 12% w/v gelatin (Sigma^®, Type A; 300 Bloom) instead of agar to CPS-T-SW (excluding starch); the solidified medium was inoculated with a needle and incubated at room temperature. Casein hydrolysis was demonstrated by zones of clearing around colonies on media to which 3 g l^-1 of sterile skimmed milk (autoclaved separately at 110°C for 40 min) had been added. Cellulase activity was detected by a zone of clearing around colonies after flooding plates containing 1% w/v low-viscosity carboxymethyl cellulose (Sigma) with 1 g l^-1 of Congo Red (Sigma). Ground chitin (50 g) from crab carapaces was dissolved by stirring for one hour in 2.5 l of 32% HCl. The solution was filtered through a cloth before adding sufficient deionised water (approximately 20 l) to precipitate the chitin. The precipitate was washed by centrifugation, until a neutral pH was obtained. The prepared chitin (0.5 g l^-1) was added to CPS-T-SW (excluding starch) agar, and added as an overlay medium on top of standard CPS-T-SW, which was inoculated and incubated at 30°C. Chitinase activity was detected by a hydrolytic zone appearing as clear halos around colonies. DNAse activity was detected by zones of clearing around colonies growing on DNase Test Agar (Merck) with added NaCl (19.4 g l^-1) after flooding plates with 1 M HCl.

Substrate utilization was tested in microtiter plates using liquid SW medium and the following compounds as sole carbon and energy source (final concentrations in brackets): (1) The poly- and disaccharides dextran, alginic acid, laminarin (each at 2 g l^-1), cellobiose, sucrose, maltose, lactose, and trehalose (each at 5 mM), (2) the monosaccharides and their derivates glucose, fructose, galactose, mannose, arabinose, ribose, xylose, rhamnose, N-acetyl-glucosamine, glucosamine, gluconic acid, sorbitol, mannitol (each at 10 mM) and meso-erythritol (20 mM), (3) the alcohols glycerol (20 mM), methanol (20 mM), ethanol (20 mM), n-propanol (10 mM) and n-butanol (10 mM), (4) the fatty acids formate (20 mM), acetate (20 mM), propionate (10 mM) and butyrate (10 mM), (5) the carboxylic acids lactate, pyruvate, malate, succinate, fumarate, citrate (each at 20 mM) and α-ketoglutarate (10 mM), (6) the 20 common amino acids glycine, alanine, cysteine, proline, serine, threonine, valine (at 20 mM each), arginine, asparagine, aspartic acid, glutamine, glutamic acid, histidine, isoleucine, leucine, lysine, methionine, phenylalanine, tryptophan and tyrosine (at 10 mM each), (7) miscellaneous compounds: betaine (20 mM), benzoate (10 mM) and salicylate (10 mM), (8) the complex substrates yeast extract, casamino acids and bactopeptone (each at 1 g l^-1), and (9) the n-alkanes, n-dodecane and n-hexadecane (1% v/v) as sole carbon and energy source were tested in oxic SW medium in sealed serum bottles. Growth was assessed visually after two weeks incubation at 30°C.

The range for growth at different temperatures, salt concentrations and pH values was tested in liquid CPS-SW medium in glass test tubes. The medium was amended with yeast extract (1 g l^-1), and, below pH 7.0, PIPES (10 mM) was used as a buffer instead of TRIS. The salt concentrations tested ranged from 0.1 to 30% w/v NaCl, with the other major salts at a concentration of 0.5 mM. Growth was measured by turbidity at 430 nm.

Anaerobic growth was tested in screw cap bottles using liquid anoxic CPS-SW medium buffered with NaHCO₃ (30 mM, autoclaved separately) replacing TRIS, Na₂S (2 mM, autoclaved separately at 109°C) as reducing agent, resazurin (0.25 mg l^-1) as redox indicator, and filter-sterilized glucose (4.5 g l^-1) replacing starch.

The results of the 106 phenotypic tests were coded for numerical analysis in binary format and a similarity dendrogram was drawn using the XLSTAT™ software package (version 2008.1.02; Addinsoft, Paris, France) using the Agglomerative Hierarchical Clustering (AHC) method, calculated with the Jaccard coefficient and UPGA (unweighted pair-group average) linkage.

Freshly grown colonies were suspended in 10 mM TRIS, pH 8, centrifuged and washed. The cells were then lysed by five freeze-thaw cycles (1 min in liquid nitrogen, followed by 5 min at 60°C). The lysate (0.5 – 1 μl) was used directly in the PCR. The 16S rRNA genes were amplified in a Gene Amp PCR System (Applied Biosystems, Foster City, CA, USA) using 0.05 units·μl^-1 final concentration of TaqPol DNA polymerase (Qiagen, Crawley, UK). The primers used were 27F (5'-AGA GTT TGA TCM TGG CTC AG-3') and 1492R (5'-TAC GGY TAC CTT GTT ACG ACT T-3') with an annealing temperature of 55°C. The PCR product was purified with the QIAquick system (Qiagen, Crawley, UK) and approximately 100 ng of the cleaned product was digested for 2.5 h at 37°C with the restriction enzymes Alu I (1 unit·μl^-1 final concentration, Roche Diagnostics, Mannheim, Germany). The resulting DNA fingerprints were analyzed after electrophoresis in a 3% w/v agarose gel (1 h, 100 V, 100 base pair ladder as standard) using a digital image analyzer and the Quantity One software package. One strain of each fingerprint-type was selected for partial sequencing of the 16S rRNA genes, and 25 strains with different ARDRA profiles and related to Bacillaceae were selected for phenotypic characterization.

Purified PCR products were subjected to sequence analysis using the ABI PRISM Big Dye Terminator Cycle Sequencing kit version 2.0 and an ABI 310 automatic sequencer (Applied Biosystems, Foster City, CA, USA). The primers used to obtain sequence fragments were 518R (5'-CGT ATT ACC GCG GCT GCT GG-3'), 338R (5'-CTG CTG CCT CCC GTA GGA GT-3'), 357F (5'-ACT CCT ACG GGA GGC AGC AG-3'), 907R (5'-CCG TCA ATT CMT TTR AGT TT-3'), 1389R (5'-ACG GGC GGT GTG TAC AAG-3'). Sequence fragments were checked using the Chromas 1.45 software package 54 and contigs assembled using BioEdit 5.0.9 55. Assem