An Unusual Inverted Saline Microbial Mat Community in an Interdune Sabkha in the Rub' al Khali (the Empty Quarter), United Arab Emirates

Salt flats (sabkha) are a recognized habitat for microbial life in desert environments and as analogs of habitats for possible life on Mars. Here we report on the physical setting and microbiology of interdune sabkhas among the large dunes in the Rub' al Khali (the Empty Quarter) in Liwa Oasis, United Arab Emirates. The salt flats, composed of gypsum and halite, are moistened by relatively fresh ground water. The result is a salinity gradient that is inverted compared to most salt flat communities with the hypersaline layer at the top and freshwater layers below. We describe and characterize a rich photosynthetically-based microbial ecosystem that is protected from the arid outside environment by a translucent salt crust. Gases collected from sediments under shallow ponds in the sabkha contain methane in concentrations as high as 3400 ppm. The salt crust could preserve biomarkers and other evidence for life in the salt after it dries out. Chloride-filled depressions have been identified on Mars and although surface flow of water is unlikely on Mars today, ground water is possible. Such a near surface system with modern groundwater flowing under ancient salt deposits could be present on Mars and could be accessed by surface rovers.


Introduction
Hypersaline environments are often found in deserts where intense evaporation and low levels of water input create concentrations of salt. Such environments are of interest as examples of life in extremes and are also relevant to the question of life on Mars-a cold desert world. Salt flats are relevant analogs for habitats for life on Mars both because salt can stabilize water as liquid at low temperature and pressure and because a salt crust can preserve evidence of past life. Accordingly, there has been considerable discussion of the relevance of coastal salt flats of marine origin as possible environments for life on Mars and as sites of preservation of biomarkers [1][2][3][4]. Desert salt flats have been studied as well. Davila et al. [5] reported on hydroscopic salts trapping atmospheric water to supply photosynthetic microbes in salt domes in the driest regions of the Atacama Desert [6]. Douglas and coworkers [7][8][9] conducted a series of studies in the mineralogical and microbiological properties of the Badwater salt flat in Death Valley, California. Barbieri et al. [10] considered organic preservation and biosignatures in dry salt flats of gypsum from Upper Pleistocene evaporite deposits of the Chott el Gharsa, a wide continental "sabkha" (a transliteration of the Arabic word for a salt flat) in southern Tunisia. Coastal sabkha are present along the Arabian Gulf (a.k.a. the Persian Gulf) along the northern coastline of the United Arab Emirates (UAE) [11,12] as subtidal flats, intertidal flats, and in particular as extensive supratidal flats [13]. There are also sabkhas deep inland in the UAE in the Liwa Oasis within the dunes of the Rub' al Khali (the Empty Quarter) which is an extensive area of aeolian dunes covering much of the Arabian Peninsula, including part of the UAE and eastern Saudi Arabia [14]. Small basins between the dunes connect to the water table, creating a flat level below which aeolian deflation cannot readily occur [14]. The interdune flats may be sites of erosion or deposition, and may be classified according to moisture content as dry, damp, wet, or evaporitic [14] and can contain salt crusts [14][15][16]. These inland sabkha are an interesting analog for salt deposits on Mars and are the focus of this study.
Salt deposits are important targets in the search for life on Mars. Recent orbital data have indicated brine deposits associated with the seasonal dark streaks on Mars known as recurring slope lineae [17]. The indicated salts are calcium perchlorate and magnesium perchlorate. However chloride salts are expected on Mars as well. Orbital spectral data indicates that there are chloride-bearing deposits, likely formed in an evaporitic environment in the ancient geologic regions of Mars [18][19][20]. The surface conditions at these locations are currently too dry to support life and the pressure is too low to allow liquid water. The deposits on Mars are at the low levels in the basin but the elevation of the deposits is high, 1260 m above the average Martian surface [18]. The inland sabkhas of the Liwa Oasis may be an analog for how these Martian salt flats could support life just below the surface in the present climate of Mars.
The hydrology of the Liwa Oasis has been well-studied because the shallow aquifer is an important source of fresh water for human use. Thus, the age and structure of the aquifer, the flow and composition of the groundwater, and the nature of the salt flats, are well-described [16,[21][22][23][24][25][26] and summarized briefly here. The location of the Liwa Oasis is shown in Fig 1 [27]. The landscape is dominated by large sand dunes with interdune spaces [14,15,28]. The sand is deposited on a flat layer of middle Miocene-age carbonates, which is essentially at sea level [16]. The aquifer contains water that is held in the sand dunes bounded below by the impermeable carbonate layer [16]. This is shown schematically in Fig 2. The water table crests in the interior of the dune field and slopes downward following the topography of the dunes. In the interdune areas, particularly near the edge of the dune field, the level of the ground water can reach the surface and wet surfaces are produced. Over time these wet areas accumulate solutes from evaporation and a salt flat (sabkha) is formed.
Radiocarbon dating and models of the groundwater formation for the Liwa Oasis [16] imply that the groundwater formed when the rainfall was approximately 20 ± 5 cm yr -1 between 6,200 and 9,000 yr before present and that the shift from wet to dry took place relatively rapidly between 5,500 and 6,200 yr ago. The 1990 predevelopment elevations of this water table are reported at 105 m for the crest and 85 m for the locations of discharge near the edges of the sand dunes [16]. The composition of the salts in the ground water is dominated by CaSO 4 and NaCl [16]. Measurements of the Cl content of the groundwater near the crest of the aquifer [16] indicated levels~1-2% of seawater. Human consumption of this fresh water source is exceeding recharge and lowering the groundwater table [16].
The structure of the Liwa Oasis interdune salt flats are inverted compared to most salt flats. In most locations with layers of saturated salt a continuing inflow from rain or tides brings a surface flow of relatively low salinity water, typically fresh water or seawater, onto the top of the salt-saturated layers. Thus, the salinity generally increases downward from either fresh or seawater at the surface to saturated salt at the bottom. The Liwa sabkha are unusual in that the surface is a thin layer of saturated salt and the relatively fresh water is flowing in from below  [27] showing the approximate field location (red dot) of the Liwa Oasis Sabkhas. White areas are hyper-arid, stippled grey areas are arid, dark areas are semi-arid and grey areas are sub-humid. Very light grey is the Arabian Gulf. The UAE boader is shown as a dotted line.  the surface. The ground water level throughout the area of the sabkha is at, or just below, the surface. The future lowering of the level of the fresh ground water will alter the nature of the interdune sabkha communities. At the present time, the ground water reaches the surface of the sabhka site we studied and thus, just a few tens of cm below the salt flat surface, is relatively fresh water. This pattern is due to the thin translucent salt crust surface, the upward flow of fresh water at the site, and the lack of surface water flow in the desert environment.
In this paper we report on studies of the physical and microbiological composition of the inverted salt flat community in a Liwa Oasis sabkha. The interaction between the salt flat and the ground water provides a sheltered habitat for microbial life in the desert. We document the spatial distribution and identity of the organisms that live in this salinity and light gradient based on 16S ribosomal RNA (rRNA) gene sequences as well as the sequences of the mcrA (methyl coenzyme-M reductase) gene, which is present only in methane-producing microorganisms. Microbial communities of other arid environments, predominantly hypersaline microbial mats, are known to contain methanogens, and to produce methane [4,[29][30][31]. This is of interest both because methane is an important greenhouse gas in Earth's atmosphere and because of the possible in situ detection of Martian methane by the Curiosity rover on Mars [32].

Study site
The field studies did not involve endangered or protected species. A permit to conduct research in the United Arab Emirates was obtained through the University of UAE via our co-author A. A. Fig 1 shows the location of the Liwa Oasis and the extent of the hyper-arid zone in southwestern UAE. The study site is located at N23.05088°E053.77005°at an elevation of 86 m.  At the study site there is a salt crust on the surface. This is seen in the foreground image in

Biological samples
Endoevaporitic mats were gathered by breaking away pieces from the overlying crust (as seen in the insert in Fig 4). The mat shown in Fig 5 was brought back moist to NASA Ames Research Center and sectioned at different depths based on the colored layers and then frozen at -80°C for later extraction and sequencing to characterize the bacterial community in each layer.

Water samples
Water collected from the two sampling pools (Fig 4 insert) was measured for salinity and pH. Gases collected as bubbles were analyzed for O 2 , N 2 , CO 2 and CH 4 content and 13 C/ 12 C isotope values. Water samples were collected from underneath the crust into sterile 50 mL conical tubes. pH was measured on site using a photometer (Lovibond, CheckDirect, UK) and phenol red tablets. Salinity was measured using a refractometer.

Gas samples
Gas samples were collected from the same volume of material as the water samples, down to 20 cm, below the mats in the shallow sabkha water pools by prodding with a metal rod. Gas bubbles produced by this disturbance were captured by water displacement in an inverted funnel fitted with a syringe seal held over the area of sediments being disturbed. A syringe needle was inserted through the seal, and the captured gas in the funnel was removed and re-injected into evacuated serum bottles. Three gas samples were collected at each of the two shallow pools.  Methane (CH 4 ), carbon dioxide (CO 2 ), oxygen (O 2 ) and nitrogen (N 2 ) concentrations were determined at NASA Ames Research Center using a Shimadzu 14A gas chromatograph (GC). Methane concentrations were measured with a flame ionization detector fitted with a Porapak N column, with N 2 as the carrier gas. CO 2 was measured using a thermal conductivity detector (TCD) fitted with a Porapak N column with helium as the carrier gas. N 2 and O 2 were analyzed using a TCD fitted with a molecular sieve packed column, with helium as the carrier gas. Matheson Tri-gas 100 ppm CH 4 , 20% O 2 /79% N 2 , and 1% CO 2 standards were run along with samples.
Gas samples for stable carbon isotopic analysis of methane were processed at Florida State University. Methane concentrations were determined using a Shimadzu Mini-2 gas chromatograph (GC) with a flame-ionization detector, fitted with a 1.83-m, 1/8in. (3.1 mm) stainless steel tubing packed with HayeSep Q 80/100 mesh. All samples were handled by directly injecting the gas (~250 μL) into the GC, where the sample was carried by UHP helium for analysis. Multiple methane standards (Scott Gas, PA, USA) were run along with samples. The analytical errors for methane concentration analyses are ± 0.5 ppm. Stable carbon isotope values were obtained using a modified cryofocusing method to amplify the methane peak [4,33]. Isotope analyses were performed using the methods described in detail in Tazaz et al. [4] and are only briefly summarized here. Duplicate analyses were performed on all gas samples. Carbon isotope data are reported in the ''del" notation (e.g., δ 13 C) [4]: where sam refers to the sample and std to the standard, Pee Dee belemnite (PDB). The units of δ are permil (‰). The analytical errors for stable isotopic analyses are ± 0.4‰ for δ 13 C in methane.

Nucleic acid extractions
Genomic DNA was extracted from each layer of the mat, hereafter referred as top, middle, bottom and sediment layers (Fig 5), in triplicate using the MO BIO Power Biofilm DNA isolation kit (#24000-50; MO BIO Laboratories, Inc., Carlsbad, CA, USA) and a bead beater to lyse the cells according to the product manual. The extracted DNA was analyzed on an agarose gel to assess the quality of the extractions, and quantified on a spectrophotometer (P330 nanophotometer; Implen Inc., Westlake Village, CA, USA).

PCR amplification
Bacterial 16S rRNA genes for all DNA extractions were amplified using the universal primers 8F (5' -AGAGTTTGATCCTGGCTCAG -3') and 1492R (5' -GGTTACCTTGTTACGACTT -3') [34,35]. Triplicate PCR reactions were run for each sample, with each PCR mixture containing 0.6 μL of each primer (50 μM), 12.5 μL of GoTaq Green master mix (M7122; Promega, Madison, WI, USA), 1 μL of DNA template (~20 ng μL -1 ) and 10.3 μL of nuclease-free water for a maximum reaction volume of 25 μL. The 16S rRNA PCR program was set to: 95°C for 10 min, followed by 28 cycles each at 95°C for 0.5 min, 55°C for 1 min, 72°C for 1.5 min, ending with one final extension cycle at 72°C for 12 min in a Peltier thermocycler PTC-200 (MJ Research Inc., Waltham, MA, USA). The mcrA gene was amplified using the mcrA primers described in Luton et al. [36]. Triplicate PCR reactions were run for each sample, with each PCR mixture containing 1 μL of each primer (10 μM), 12.5 μL of GoTaq Green master mix (M7122; Promega, Madison, WI, USA), 1 μL of DNA template (~20 ng μL -1 ) and 9.5 μL of nuclease-free water for a maximum reaction volume of 25 μL, using the following PCR program: 95°C for 1 min, 35 cycles at 94°C for 30 s, 55°C for 30 s, 72°C for 1 min with a final extension cycle at 72°C for 5 min according to Garcia-Maldonado et al. 2014 [37]. A second amplification was performed for 25 cycles using 2 μL of mcrA PCR product from the first amplification to get more distinct bands. All amplifications were analyzed along with positive and negative controls for quality control. Replicate PCR products were pooled and run on an agarose gel for each gene. The 16S rRNA and mcrA fragments were excised and purified using the Omega E.Z.N.A. gel extraction kit (D2500-01; Omega Bio-Tek Inc., Norcross, GA, USA) according to manufacturer instructions. Prior to cloning, a poly(A) tail reaction was performed on all 16S rRNA samples by adding 6 μL of PCR product, 1 μL 10X PCR buffer, 1 μL MgCl 2 , 1 μL dNTP, and 1 μL Taq polymerase to a total volume of 10 μL per sample and heating the samples to 72°C for 20 min in a thermo cycler.

Cloning
The Life Technologies TOPO TA Cloning Kit for Sequencing (#K457501) (Life Technologies, Carlsbad, CA, USA) was used to ligate 4 μL of the purified PCR product to 1 μL of cloning vector (pCR4-TOPO) in addition to 1 μL of salt solution, followed by the vector-PCR insertion into One Shot Top10 Chemically Competent E. coli cells according to the product manual. The cells were plated at different densities on Luria Broth (LB) agar plates containing 50 μg μL -1 of ampicillin, 0.2 mM IPTG and 40 μg mL -1 X-gal, and were incubated at 37°C overnight. For each layer, white colonies were picked and transferred to individual wells in a deep 96-well plate containing 500 μL of LB medium, 10% glycerol and 100 μg mL -1 ampicillin, and were grown overnight at 37°C. Ninety-six clones per layer were shipped in dry ice to Beckman Coulter Genomics (Danvers, MA, USA) for Sanger sequencing of the 16S rRNA gene on an ABI PRISM 3730xl DNA analyzer (Applied Biosystems, Foster City, CA, USA). Clones with the mcrA gene insert were verified by PCR using plasmid specific M13 primers, purified using the PureLink Quick Plasmid Miniprep kit (#K210010) (Life Technologies, Carlsbad, CA) and submitted to ELIM Biopharmaceuticals, Inc. (Hayward, CA, USA) for Sanger sequencing on an ABI PRISM 3720xl DNA analyzer.

Nucleotide sequence data and classification
Full length Sanger reads were obtained through Beckman Coulter Genomics single pass standard sequencing read service using the primers T3 (5' -ATTAACCCTCACTAAAGGGA-3'), T7 (5' -TAATACGACTCACTATAGGG-3') and 515 (5' -GTGCCAGCAGCCGCGGTA A-3') for the 16S sequences. The three reads per sample were quality-filtered, trimmed and assembled to full 16S sequences (~1.5 kb) using the software Geneious (V. 7.1.6; Biomatters Ltd., Auckland, NZ). Assemblies containing less than 3 reads were discarded. Full 16S sequences were aligned with the E. coli strain K-12 (U00096) as a reference using the SILVA Incremental Aligner (SINA) (v1.2.11) program provided by the SILVA ribosomal RNA project; http://www.arb-silva.de/aligner/) as described in Pruesse et al. [38]. Columns in the alignments with common gaps were removed using the filter.seqs command in the open-source software MOTHUR (http://www.mothur.org/) [39]. Chimeras were identified using the software Mallard (v. 1.02) with a cut-off line set to 99.9%, as described in Ashelford et al. [40]. The nucleotide sequence of outliers (samples with deviation from expectation (DE) values above the set cut-off line in Mallard) were queried manually against the NCBI non-redundant nucleotide sequence database using the BLASTN program (http://blast.ncbi.nlm.nih.gov/Blast.cgi) [41] to further assess if these were real chimeras, and were discarded if chimeric breakpoints in sequences were evident in BLAST results. After quality control, we ended with a total of 317 16S sequences, with 71, 76, 85, 85 for the top, middle, bottom and sediment layer, respectively.
The search and classify tool in SINA was used for taxonomic assignment of the 16S sequences based on the last common ancestor method. Representative sequences for each phylum were further queried against the NCBI nucleotide database using the BLASTN program. Sequences identified as belonging to the genus Salinibacter were further analyzed in MOTHUR [39]. The cluster command was used to identify operational taxonomic units (OTUs) with a 0.01 cutoff using the average neighbor algorithm.
All sequences determined in this study were deposited in the NCBI database under Gen-Bank accession numbers KU308828-KU309144 for 16S rRNA sequences, and KU51439 5-KU514406 for mcrA sequences.

Results
Water in the shallow pools was uniformly 30°C at the time of measurement. The results are shown in Table 1 as the average and standard deviation of three replicate samples at two nearby sites. Although the two samples only differ only slightly in salinity, the oxygen levels differ by a factor of~2, but with large variation between replicates. Methane concentration is high in both samples indicating a source of methane in the salt flat. Carbon dioxide and methane are distinct between the two samples (no overlap of the error bars) and they co-vary.
The presence of methanogenic microorganisms at this site was confirmed through the amplification of the mcrA gene (~470 bp) from the bottom and sediment layers; the gene was not amplified from the top and middle layers. Twelve sequences obtained from these samples yielded close matches (>91% nucleotide identity scores) to methanogenic archaea: the extremely halophilic Methanohalobium evestigatum (CP002069) [44], the moderately halophilic Methanolobus vulcani (U22245) and Methanohalophilus portucalensis (AB908273) [45,46] based on BLAST analysis.
Further analysis of the salt crust revealed the presence of a diverse microbial community with unusual layering with primary productivity based on photosynthesis. Based on the distribution of identified environmental sequences, the top and middle layers of the endoevaporitic mat were dominated by the phylum Bacteroidetes (~63.4%), with a gradual decrease in abundance with increasing depth (Fig 7). Sequences most similar to those from the genus Salinibacter, an extreme halophile that thrives at salinity levels of 20-30% [47], was the dominant representative in the top three layers.
Sequences for mild to moderate halophiles of the genera Gracilimonas, Marivirga (0.5-10%) and Owenweeksia (1-7.5%) were only identified in the sediment layer (Table 2) [48][49][50]. In contrast, cyanobacterial sequences were only found in low abundances in the top and middle layers (2.8 and 5.3%, respectively), and higher abundance in the bottom and sediment layers at 42.4 and 33%, respectively (Fig 7). Representative cyanobacterial sequences identified were most similar to the unicellular, halotolerant, coccoid Cyanothece sp., Euhalothece sp. and Dactylococcopsis salina. The taxonomic distribution of members of the Bacteroidetes and Cyanobacteria phyla along the vertical profile are consistent with the presence of a salinity gradient in this sabkha, where the top layer of the endoevaporitic mat has a high salinity, and the lower layer is influenced by a lower salinity due to the presence of fresh groundwater.
Sequences from the phylum Proteobacteria were present in all four layers, but were almost twice as abundant in the sediment layer, which was mostly composed of fine sand. Based on genus-level identities provided by the SINA analysis, this phylum was the most diverse in our clone libraries in terms of richness. Within these sequences, representatives of the Alphaproteobacteria, Gammaproteobacteria, and Deltaproteobacteria classes were identified ( Table 2).  Table 2. Select identities of closest cultured matches based on longest sequences for the phyla identified in each layer, using the NCBI BLASTN program. Proteobacteria representatives are additionally identified as belonging to the class Alphaproteobacteria (A), Deltaproteobacteria (D), or Gammaproteobacteria (G). The sequence similarity is indicated by ID% with similarities of 99% marked with bold type. Sequences with low-identity to the Gemmatimonadetes and sequences most similar to Verrucomicrobia were less common and only present in the lower two layers. Sequences with unknown identity were categorized as "other" and varied from 2.8 to 8.2% in abundance of total sequences. Plastid sequences belonging to phototrophic eukaryotes were also identified and placed in a "plastids" category. These were present in the top (14.1%), middle (5.3%) and bottom (4.7%) layers (Fig 7). The closest identities to these plastids were for the unicellular, flagellated marine Chlorophytes Dunaliella tertiolecta and Crustomastix stigmatica ( Table 2). Dunaliella was also observed through light microscopy in fresh mat samples. Due to their motility via flagella, these phototrophic eukaryotes may not only be able to move between layers, but also populate the water below the endoevaporitic crusts.
After additional curation to remove shorter sequences, 110 of the clones identified by SINA as most similar to Salinibacter were further examined using an OTU-based approach. For the top, middle, bottom, and sediment samples, there were 40, 45, 22 and 3 Salinibacter sequences, respectively. At 0.01 cutoff (99% similarity), 11 distinct OTUs were identified ( Table 3). The largest OTU contained 62 sequences (56.4%), and four OTUs contained only a single sequence. Although the present survey only allows for qualitative comparisons of the depth distribution of these OTUs, there appeared to be a by-sampling partition for OTUs 1, 2 and 5 (mainly in the top and middle samples), and OTUs 3 and 4 (mainly found in the bottom sample). OTU 6 appeared to be cosmopolitan with representation in all samples, but with only 5 sequences total, this is not definitive (Table 3).
A phylogenetic analysis of representative sequences from the OTUs with multiple sequences supported the use of a 0.01 cutoff for the clustering, but did not reveal any divergence patterns consistent with the location of the OTU within the sabkha (not shown).

Discussion
From the physical setting of the salt flat we expect that light levels and salinity should systematically decrease with depth. The case of oxygen is more complex, the deep ground water contains oxygen, and oxygen is present in the atmosphere at the surface. However, microbial processes in the mat consume oxygen and can produce anoxic layers. Hence, while oxygen levels may decrease with depth from the surface, layers deep enough to be in the unaltered groundwater will have high oxygen content. The two salty samples (320 and 340 ppt) in Table 1 represent shallow subsurface solutions that have interacted with the salt crust and have been exposed to some extent to the ambient atmosphere. The high methane and carbon dioxide levels argue against the oxygen being preserved from the ground water; there is no evidence of vigorous mixing of ground water with material in the biological layers. The sampling method was inadequate to directly characterize the salinity profile below the salt crust with depth in a systematic way and we make no attempt to assign a depth difference to these samples. Nor can we exclude the possibility that our samples represent material from a range of depths mixed together by the creation of the sampling pond, which would imply that the differences in salinity and oxygen are not significant. The salinity in both sampling pools indicated nearly saturated conditions (solubility of NaCl in water at 30°C is 360 ppt). In contrast, groundwater that had been pumped into a nearby holding tank for irrigation use had a salinity of 17 ppt in agreement with published data [16] indicating, as mentioned above, the groundwater has low salinity.
The methane content differs between the two samples, increasing to 3442 ppm for the lower salinity, lower oxygen, higher carbon dioxide sample. It seems plausible that methane content would be higher in the lower salinity layers with lower oxygen levels. However, we note that we cannot rule out migration of methane gas through the system against, or with, the flow of water. Clearly further measurements are required to understand the depth profile.
The presence of methanogens at this site was strongly suggested by the amplification of the mcrA gene, unique to methanogens, from the bottom two layers of the mat and high sequence identities to three phylogenetic groups of methanogenic archaea. The presence of methanogens was also confirmed by the detection of methane (0.3% v/v) in bubbles collected from this site, with δ 13 C values comparable to those reported in endoevaporitic mats [4,30,51]. The unusually enriched δ 13 C values (-41 and -38‰) are higher than the traditional δ 13 C values expected for biogenic methane (~<-50‰) [4,[51][52][53], but may be explained by lower isotopic fractionation in systems where substrate is limited, such as in high-salinity endoevaporitic mats [4,30,51]. Methane oxidation may also result in more enriched δ 13 C values, but methanotrophs are thought to be mostly absent in hypersaline environments [54]. However, our 16S sequences did include one representative with similarity to the methanotrophic Methylohalomonas lacus in the sediment layer ( Table 2).
The significance of the methane is twofold. First, it represents a source of methane in a desert environment on Earth contributing to the global inventory of methane sources. Secondly, methane production in this sabkha provides one possible model for methane production on Mars, as discussed below.
Studies of orbital images over the past 40 years show no detectable change in the position of the dunes or the location of the sabkhas. This indicates that the timescale for covering and uncovering new sabhka flats is in excess of hundreds of years. Based on satellite images and GPS ground observations over several years, Lorenz and Radebaugh [55] computed a net motion of~0.1 m yr -1 advance for the large dunes of the Liwa Oasis. The sabhka studied here was about 500 m in diameter (Fig 2) suggesting a lifetime against burial by the large dunes of 5000 years. It may be possible to discover previously buried sabkha in this area exposed by recent movement of the dunes and thus investigate the preserved remains of past mat systems.
An important contrast between this environment and previous studies of salt ponds and coastal sabkhas is the salinity gradient. In most hypersaline environments in deserts or coastal sites the salt arrives with the water supply from above the surface and accumulates as the water evaporates. In the Liwa Oasis sabkha studied here, the water flow is in the subsurface; there is very little surface flow. This creates an unusual layered structure with saturated salt layers on top and relatively freshwater layers below. The structure of the microbial mat (Fig 5) is the clearest indication we have of this vertical structuring due to light, salinity, and oxygen gradients. Clone libraries constructed from four depths within the sabkha crust showed changes in taxonomic composition consistent with decreasing salinity with depth (e.g., reduced total and proportional abundances of Salinibacter sequences), but further research is required to understand the role this gradient has on structuring the microbial community. This gradient is shown schematically in Fig 6. The interdune sabkhas in Liwa Oasis are of particular interest as a Mars analog. The salt crust is a relic deposited by previous evaporation. Although widespread, the salt crust is not easily visible by remote sensing because it is mixed with sand. Similar relic salt flats may exist on Mars today, remnants from a wetter period in the past buried by the present shifting sands. The sabkha studied here suggests an alternative way these salt deposits could support life. If there was a source of deep ground water flowing upward from below, a near-surface habitable layer might be formed protected from the low pressure, arid surface conditions. The results from the Liwa Oasis suggest a microbial model for such a habitat and for methane production at such a site. Light can penetrate the salt crust providing energy for photosynthesis but the crust provides a barrier to loss of water into the dry atmosphere due both to physically blocking exchange between the water and the atmosphere and by lowering the vapor pressure of water due to solution. This reduction of water loss by the salt layer operates in the Liwa Oasis as it would on Mars.
In the Liwa Oasis, methane is being released from the salt community continuously as the salt pan is fractured in many places. This would not be the case on Mars as the pan would have to be tightly sealed to preserve the liquid water habitat below. On Mars methane would build up below the salt pan until the gas pressure exceeded the pressure holding the salt pan down, the vapor pressure of water, assuming it is cold, would barely exceed the ambient pressure and hence alone would not suffice to break the surface. The pan would then crack with a sudden release of gas. Rapid evaporation and flash freezing of the outflowing water on the surface would quickly reseal the crack with ice, which would slowly sublime leaving behind sediment and salt. The cycle might repeat on timescale of years or even decades. Such a system could conceivably explain large outbursts of methane on Mars as has been reported [56]. However it does not provide an explanation for the subsequent rapid loss of the methane plumes also reported [56]. This rapid loss is inconsistent with known photochemical processes on Mars [57]. A salt pan system such as studied here can prolong habitability for photosynthetic surface biology when only ground water is available and surface conditions are unsuitable for stable liquid water. Another important aspect of this system is the preservation of biomarkers, including methane and other evidence for life in the salt after it dries out. Such a near surface system could be accessed by surface rovers. If an inverted salt mat system fed by groundwater existed on Mars in the past but is no longer operative it may still be of considerable astrobiology interest. It has been suggested [2] that salt deposits can preserve evidence of past life. As the flow of water to the inverted mat ceased, the active microbial system would be encased in salt and preserved over geological time, relatively protected from radiation and oxidation damage.