, Volume 20, Issue 5, pp 747–757

Analysis of the bacteriorhodopsin-producing haloarchaea reveals a core community that is stable over time in the salt crystallizers of Eilat, Israel

Original Paper

DOI: 10.1007/s00792-016-0864-4

Cite this article as:
Ram-Mohan, N., Oren, A. & Papke, R.T. Extremophiles (2016) 20: 747. doi:10.1007/s00792-016-0864-4


Stability of microbial communities can impact the ability of dispersed cells to colonize a new habitat. Saturated brines and their halophile communities are presumed to be steady state systems due to limited environmental perturbations. In this study, the bacteriorhodopsin-containing fraction of the haloarchaeal community from Eilat salt crystallizer ponds was sampled five times over 3 years. Analyses revealed the existence of a constant core as several OTUs were found repeatedly over the length of the study: OTUs comprising 52 % of the total cloned and sequenced PCR amplicons were found in every sample, and OTUs comprising 89 % of the total sequences were found in more than one, and often more than two samples. LIBSHUFF and UNIFRAC analyses showed statistical similarity between samples and Spearman’s coefficient denoted significant correlations between OTU pairs, indicating non-random patterns in abundance and co-occurrence of detected OTUs. Further, changes in the detected OTUs were statistically linked to deviations in salinity. We interpret these results as indicating the existence of an ever-present core bacteriorhodopsin-containing Eilat crystallizer community that fluctuates in population densities, which are controlled by salinity rather than the extinction of some OTUs and their replacement through immigration and colonization.


Haloarchaea Temporal analysis Seasonal Community stability Thalassohaline Eilat 


Taxonomic stability in microbial communities is impacted by environmental influences and various patterns are seen in different niches. For instance, microbial communities inhabiting pineland soils exhibited changes in diversity on a seasonal basis, but exhibit no differences in the microbial biomass (Rogers and Tate 2001). Environmental factors have been shown to shape the microbial communities inhabiting different river ecosystems (Liu et al. 2011, 2013; Winter et al. 2007; Wu et al. 2011). Changes in water temperature and conductivity, for example, were identified to influence temporal partitioning in bacterial communities of a subtropical river. Similar studies on prokaryotic communities from different lakes revealed seasonal variation in species abundances and functional capacity of the communities (Dickerson and Williams 2014; Zaccone et al. 2014). The bacterioplankton community within the Salton Sea was shown to undergo seasonal variations with minimal overlap in the detected community composition between the sampled seasons presumably in response to environmental instability (Dillon et al. 2009). In contrast to these studies, community stability was observed throughout the seasons for cyanobacterial populations defined by temperature that inhabit a microbial mat from Octopus Spring in Yellowstone National Park, most likely due to a constant abiotic environment not prone to disturbances (Ferris and Ward 1997). While it is clear that the environment has a huge impact on communities, changes in structure due to perturbations, however, may result only in abundance fluctuations of indigenous populations, rather than opening new niches for the invasion of non-native species, as seen in Florida beach sands before, during and after oil contamination due to the Deep Water Horizon spill (Rodriguez et al. 2015).

One niche, broadly described as hypersaline environments is characterized by salt concentrations higher than seawater. These are divided into two major categories based on the ionic composition: thalassohaline and athalassohaline environments (Oren 2006). Thalassohaline environments typically result from the evaporation of seawater, which concentrates salts and ion ratios similar to that of its origins, until specific salts (e.g., calcium sulfate) reach saturation and precipitate (Oren 2006). Sabkhas, salt marshes, and sea salt production facilities are examples of thalassohaline environments. Many hypersaline lakes are examples of athalassohaline environments, and can be formed when a water body is landlocked and terminal (e.g., Dead Sea). Salinity increases as minerals are transported into the lake and evaporation occurs. This process dictates that the ionic composition is unique in each athalassohaline environment. However, some inland lakes are thalassohaline, e.g., Tuz Lake, Turkey (Mutlu et al. 2008) and Lake Tyrrell, Australia (Podell et al. 2014). Physicochemical studies on hypersaline environments, both thalassohaline and athalassohaline, have shown that owing to the high salinity, these environments are subject to low solubility of gases, diffusion rates and very low water activity (Litchfield 1998). These hypersaline environments also vary in pH and temperature (Oren 2002b) making them too extreme for most organisms.

Members of the class Halobacteria (Domain: Archaea; Phylum: Euryarchaeota), usually called haloarchaea to distinguish them from halophilic bacteria, are typically thought of as the dominant inhabitants of the hypersaline crystallizer ponds, where NaCl precipitates at ~32 to 37 % and is then harvested for commercial purposes (Antón et al. 1999; Benlloch et al. 2001, 2002; Ghai et al. 2011; Grant et al. 1999; Litchfield and Gillevet 2002; Martínez-Murcia et al. 1995; Maturrano et al. 2006; Ochsenreiter et al. 2002; Øvreås et al. 2003; Sabet et al. 2009; Walsh et al. 2005). Several cultivation and molecular based studies on crystallizer ponds have led to some general conclusions about haloarchaeal communities. Individual communities tend to be comprised by a small number of dominating genera, with the square archaeon, Haloquadratum walsbyi, often reported as having the largest population sizes (Antón et al. 1999; Benlloch et al. 1995, 2001, 2002; Legault et al. 2006; Martínez-Murcia et al. 1995; Oh et al. 2010) sometimes comprising >60 % of all the archaea (Antón et al. 1999). Dominance by the genus Haloquadratum is observed in some hypersaline lakes also (Ghai et al. 2011; Mutlu et al. 2008). This, however, does not appear to be the case in every hypersaline environment. Snapshot analyses identified Halorubrum- related phylotypes as the most frequently retrieved genus from a saltern in Slovenia (Pašic et al. 2005) and Haloquadratum 16S rRNA gene sequences were not recovered from a saltern in San Diego (Bidle et al. 2005). Additionally, in some of these studies, frequently observed halobacterial clones retrieved did not cluster with any cultivated and described halobacterial species.

Given that saturated brines require of microorganisms specific adaptive characteristics to survive in this unique habitat, and that these habitats typically exist in hot, dry climates where environmental conditions are generally stable, it is predicted that many of them will maintain a steady taxon community structure like that seen in some hot springs (Ferris and Ward 1997). Studying such communities through time will provide insight into their community stability. This study focuses on the structure of the bacteriorhodopsin-containing halobacterial community inhabiting the sea salt crystallizer ponds in Eilat, Israel through time. The Israel Salt Industries Ltd. in Eilat, operational since 1977, produces about 170,000 tons of salt a year and, harvesting is year round. Ponds 301–304 (Litchfield et al. 2009) and, more recently, ponds 305–307 are the salt crystallizing ponds with salt concentrations of 300 gL−1 and above. In order to study the Eilat salt crystallizer pond, the haloarchaeal specific gene bacteriorhodopsin (bop) was used as the molecular marker. Bacteriorhodopsin, a member of the haloarchaeal rhodopsin family, is a light-driven proton-pump that generates an electrochemical gradient for the production of ATP (Oesterhelt and Stoeckenius 1973). Our analyses indicate the existence of a non-random, bacteriorhodopsin-producing core haloarchaeal community that fluctuates in taxon abundance in correlation with salinity.

Materials and methods

DNA isolation and PCR amplification

Brine samples were collected from the reddest pond among the salt crystallizers (301–304) at the salt works in Eilat over a period of 3 years at five time points (see Table 1). Four liters of the water collected from the salt crystallizer pond was centrifuged at 6500 rpm for 30 min in a large Sorvall rotor and the cell pellet was collected. DNA from these pellets was isolated using the protocol published in the Halohandbook (Dyall-Smith 2008). Briefly, 400 μl of distilled and deionized water was added to the pellet to lyse cells by osmotic shock. Lysates were then placed in a heating block maintained at 70 °C for 10 min to inactivate proteins. A working solution of the template DNA for PCR amplification was prepared by making a 1:200 dilution of the crude environmental DNA stock.
Table 1

Sample information



Sample name

% Salinity

Atmospheric temperature (°C)

No. of clones sequenced

3rd January






20th June






25th August






15th May






22nd of April






The bop gene was amplified using primers bop401F and bop795R adapted from (Papke et al. 2003) and modified to carry the M13 forward and reverse sequences and renamed, respectively, to bopF_poly (5′-GTA AAA CGA CGG CCA GTG ACT GGT TGT TYA CVA CGC C-3′) and bopR_poly (5′-AAC AGC TAT GAC CAT GAA GCC GAA GCC GAY CTT BGC-3′). Using bop as the molecular marker circumvents issues of low taxonomic resolution and multiple divergent copies of the 16S rRNA gene observed in many genera of halobacteria (Boucher et al. 2004; Mylvaganam and Dennis 1992). Bacteriorhodopsins are present in significant quantities in solar salterns (Litchfield et al. 2000; Oren and Shilo 1981) and are widely expressed among halobacteria living in light-filled environments. The advantages of the bacteriorhodopsin gene as a molecular marker for halobacterial communities has led to several publications (Dillon et al. 2013; Gomariz et al. 2015a; Papke et al. 2003; Pašic et al. 2007) and was applied here to each of our sample sites using the primers to amplify bacteriorhodopsin directly from the community DNA.

The DNA polymerase Phusion (New England BioLabs) was used to produce high fidelity copies of the template. The following PCR cycle protocol was used: One minute initial denaturation at 94 °C, followed by 30 cycles of 30 s at 94 °C, 30 s at 53 °C and 45 s at 68 °C. Final elongation occurred at 68 °C for 5 min. PCR reactions contained Phusion polymerase [2 units/µl], 0.25 µl; 5 × GC buffer, 5 µl; DMSO [100 %], 2.5 µl; dNTPs [100 mM], 1 µl; forward and reverse primers [10 µM], 0.5 µl each; genomic DNA [20 ng/µl], 1.0 µl; and 1.0 µl of dH2O. Acetamide [25 % w/v], 13.25 µl was added to the reaction to improve specificity of primer binding and DMSO was used to facilitate denaturation of the high G + C template.

Cloning, plasmid isolation and sequence acquisition

PCR products of the expected length were isolated from a 1 % (w/v) agarose gel using the SV Gel & PCR Clean-Up System (Promega) and then cloned using the Zero Blunt TOPO Cloning Kit (Invitrogen) according to the manufacturer’s directions. The recombinant plasmid was isolated from the clones for each sample using the Wizard Plus SV Minipreps DNA Purification System (Promega) according to manufacturer’s instructions. Plasmids were tested for presence of an insert by restriction digestion with EcoRI. The purified plasmids with cloned inserts were sent to GENEWIZ Inc. for sequencing. The sequences obtained in this study were submitted on GenBank under the following accession numbers: KT028773–KT029121.

Sequence analysis

The obtained raw sequences were manually curated using the commercial software Geneious 4.8.3. Relevant phylogenetic context to be included in community composition detection was extracted from GenBank using TBLASTX (Altschul et al. 1990) (http://blast.ncbi.nlm.nih.gov/Blast.cgi) against all of the currently available, 109 draft and closed halobacterial genomes and the non-redundant nucleotide database. BLAST hits with the lowest e value and the highest query coverage were added to our analyses. All bacteriorhodopsin sequences were aligned using MUSCLE (Edgar 2004). Multitaxa alignments were edited using MacClade 4.08 (Maddison and Maddison 2003).

Bacteriorhodopsin-producing halobacterial community analysis

The aligned bacteriorhodopsin sequences were analyzed using Mothur v.1.20.0 (Schloss et al. 2009) to estimate the species richness, community diversity indices and the similarity between the bacteriorhodopsin-producing halobacterial communities, referred to simply as haloarchaeal communities, recovered from each sampling time point. Operational Taxonomic Units (OTUs) were determined at 100, 99, 97 and 95 % sequence similarities using the average neighbor clustering, and the different OTUs were used for further analyses. 99 % sequence similarity was designated as a stringent estimate for OTU clustering. A previous study showed that bop gene sequences with less than 1 % variation formed species-like clusters (Papke et al. 2007). 95 % was chosen as a liberal estimate for the OTU definition. The presence of outliers in the observed number of OTUs over time was tested three ways—the extreme Studentized deviation (ESD identifier) (Rosner 1983); Hampel identifier (Hampel 1971); and the standard boxplot rule (Tukey 1977). Species richness estimators Chao (1984), which is nonparametric and bases values on the number of singletons and doubletons, and Ace (Chao and Lee 1992), which is based on the number of rare groups of observed OTUs (OTUobs), were calculated for each time point. Shannon index (H) and Simpson index of diversity (1 − D) were also estimated as a measure of the species diversity and evenness at each time point. Rarefaction curves were generated within Mothur to estimate the sampling completeness and efficiency for each time point.

Phylogenetic reconstruction

All 349 sequences were used to construct a maximum likelihood tree from distances calculated under the Generalized Time-Reversible model (Tavaré 1986) within the PHYML 3.0 (Guindon et al. 2010) phylogenetic program. Any OTU at 99 % that contained sequences from at least two time points was termed a ‘shared OTU’. An OTU was defined as a ‘cumulative shared OTU’ when it contained two or more 99 % shared OTUs that combined into a single OTU at 95 %. The shared OTUs at 99 % and the cumulative shared OTUs at 95 % were labeled ‘sOTU number’ and ‘csOTU number’, respectively. The tree was viewed, edited and midpoint rooted using FigTree (http://tree.bio.ed.ac.uk/software/figtree/), a tree editing software.

Community comparisons

The halobacterial communities recovered from the sampling time points were compared three ways. First, an online tool—UNIFRAC (Lozupone et al. 2006) was employed. The mid-point rooted maximum likelihood tree (not shown) of all the sequences and an environmental file listing the sequences from each sampling time point were uploaded onto the UNIFRAC tool to perform the JackKnife Environmental Clustering Analysis. Based on the clustering of the sequences observed on the maximum likelihood tree, UNIFRAC estimated a community level pairwise distance matrix. This was used to develop a UPGMA dendrogram of the communities and the JackKnifing provided support for the clustering of the sample sites. Second, the LIBSHUFF command within Mothur was employed. The command within Mothur implemented the original LIBSHUFF program (Singleton et al. 2001). It tested for similarity in structure between two or more communities by incorporating the Cramer-von Mises test statistic (Anderson 1962) and returning a significance value for the difference between each pair in consideration. Finally, the OTUs defined at 99, 97, and 95 % sequence similarity were manually curated to determine the OTUs with sequences from different time points clustering together. Pairwise OTU correlations were estimated by determining the Spearman’s rank correlation coefficient (Spearman 1904) within Mothur. The correlation coefficient estimated the relatedness between the relative abundances of each pair of OTUs.


Sample compositions and abundance of genera through time

The top BLAST hit for the representative sequence of each OTU at 95 % was used to determine the taxonomic affiliations of the community members. Normalized bar plots (Fig. 1) were constructed with these top BLAST hits from the non-redundant database (Fig. 1a) and then only from available genomes (Fig. 1b) with >50 % identity to pictorially represent the observed community structure through time. While each time point retrieves top BLAST hits from the genus Halorubrum, the non-redundant database has many sequences from environmental studies and therefore the top BLAST hit for over 60 % of each community is of unknown taxonomic affiliation (Fig. 1a). Other recovered top BLAST hits from this database of known affiliation were Haloarcula, Halomicrobium and Halosimplex and each of these was recovered only once. To more accurately assign sequences to taxa, we retrieved the top BLAST hits querying only the sequenced haloarchaeal genomes, which revealed additional details regarding taxonomic affiliations. Figure 1b shows that 11 groups were identified with two genera (Halomicrobium and Halosimplex) recovered only once, four genera were recovered twice (Haloquadratum, Natronorubrum, Haloplanus and Halobiforma), one genus was recovered thrice (Haloferax), one group was recovered four times (Haloterrigena) and three genera were recovered in every sample (Halorubrum, Haloarcula, and a taxonomically unclassified group with sequence identity <50 % of the bacteriorhodopsin genes in the genomes). With the exception of Halomicrobium and Halosimplex, each genus was detected from more than one sample period, varying in their relative sequence abundances by our methods, which suggests that every genus detected is a long-term member of the existing community, but below our ability to detect them in some samples.
Fig. 1

Community composition over time. Normalized bar plots were constructed at each time point based on the top BLAST hits for representative sequences of each OTU at 95 %. a Top BLAST hits from tBLASTx against entire non-redundant nucleotide database. b Top BLAST hits from the 109 available genomes by performing a stand alone tBLASTx

OTUs are shared between samples

A five-way Venn diagram was constructed to demonstrate community overlap using different OTU definitions at 99, 97 and 95 % (Fig. 2). Identified in this analysis is the existence of a set of OTUs at the 95 % definition that are found in all five samples through time. These four OTUs called the ‘core’ encompass 179 of the 349 overall sequences (~52 %) obtained in this study and based on top BLAST hits belong to Halorubrum (cs03 and s01), Haloarcula (cs01), and one unknown (cs06). Many of the other 95 % OTUs were detected at more than one time point. These mercurial OTUs represent ~37 % of the overall sequence data. Within this 37 %, ~3 % cluster into 1 OTU (s08), belonging to Haloterrigena, that occurs in four out of five time points. Approximately 19 % is shared between three time points and form five OTUs (s21, cs02, s30, s04, and cs07). ~15 % cluster into 13 OTUs (s06, s07, s09, s10, s03, s31, s27, s22, s23, cs05, cs04, s19, and s20) and are found at two time points. A fraction (~11 %) of the total sequences formed 15 95 % OTUs that were only retrieved at one time point in our study (Aug’11: 9; May’12: 1; Apr’13: 4; Jan’10: 1). All sampling time points had unique OTUs detected, except in the June’10 sample. With one exception, all OTUs unique to a sample are represented by a single sequence (singleton) (12), two sequences (doubleton) (1) or three sequences (tripleton) (1): the exception being from the August’11 sample which had a unique 95 % OTU (Aug’11-02) that was comprised of 22 sequences. The rarely retrieved OTUs, including Halomicrobium, Halosimplex, Haloquadratum, Natronorubrum, Haloplanus, Halobiforma, and Haloferax and unknown genera, are likely due to sampling limitations.
Fig. 2

Sharing of OTUs between time points. Number of shared OTUs represented vertically at 99, 97, 95 % sequence similarities respectively. (x)a—total number of sequences obtained at that time point: (x, y, z)b—number of OTUs unique to the sampling time point represented for 99, 97, 95 % sequence similarities, respectively

To determine if there was a dependent, non-random abundance relationship between members of core OTUs the Spearman’s rank correlation coefficient (ρ) was estimated for each pair compared. ρ was determined for 100 % identical sequences (100 % OTUs) found within the core 95 % OTUs and only p values <0.05 were considered significant. The values range from −1 to 1, indicating a negative and positive correlation, respectively, between the relative abundances of the two 100 % OTUs compared. Supplemental figure S1 is a plot of the statistically significant correlation coefficients, which shows most 100 % OTU pairs have a strong positive correlation (ρ = 0.88–1.00). Three 100 % OTU pair comparisons, which belong to the core 95 % OTUs cs01 and cs03, returned a strong negative correlation (ranging from ρ = −0.89 to −0.91), while the remaining significant interactions of those two core OTUs showed a positive correlation. Comparison between other core OTUs resulted in negative correlation coefficients but were not statistically significant. The high numbers of shared OTUs and the strong correlation between them indicate that though PCR and cloning is not a quantitative technique, the sequence data retrieved were not random with respect to the abundances detected, and therefore dynamics seen in sequence abundance probably reflect real fluctuations in natural population sizes and not an indication of newly colonized cells invading the habitat.

Sampling efficiency and richness estimations

To estimate sampling efficiency, rarefaction curves were generated for each sampling time point for OTU definitions of 99 and 95 % sequence similarity (Supplemental figure S2). Though the curves never flatten for the August’11, May’12, and the April’13 samples at 99 % suggesting further sampling is required, they do begin to level, and at 95 % OTU definition, the plots plateau suggesting reasonable sampling at 95 % sequence similarity. The January’10 and June’10 curves at both 99 and 95 % sequence similarity indicates abundant sampling from these time points, suggesting that all time points are well sampled, but not completely, at 95 % OTU definitions.

Species richness estimators for each sampling time point were calculated and plotted (Supplemental figure S3). At the 95 % sequence similarity definition, the observed number of OTUs is similar to the estimations of species richness for both Chao and Ace, whereas at the 99 and 97 %, the observed number of OTUs was lower than the richness estimates. Though variation in richness estimations exists, the observed OTUs through time do not drastically change. There are no outliers in the data range as measured by the ESD identifier, Hampel identifier, and the standard boxplot rule, which together indicates that community diversity is statistically equivalent through time. The communities at each time point appear rich and diverse in OTUs (Table 2). The Shannon index (H) is a commonly used diversity index that ranges between 1.5 and 3.5 and takes into account both abundance and evenness of species observed in the community. H estimations suggest a diverse community at each time point, similar to the findings from the rarefaction curves. Simpson’s index of diversity (1 − D) ranges between 0 and 1, and describes the sample diversity. The estimated 1 − D values corroborate the findings from the rarefaction curves. Since sampling efficiency is good but not complete, the absence of an OTU in some samples while present in others is not evidence for the absence of the sequence from the community. To test this, we performed a correlation test between the species richness estimations and salinity and showed that the Chao (correlation coefficient = 0.72) and Ace (correlation coefficient = 0.85) estimations of the low abundance members changed with respect to the salinity: therefore, increases in salinity possibly due to higher temperatures likely facilitated changes in population sizes as seen in other studies (Boujelben et al. 2012; Gomariz et al. 2015b; Podell et al. 2014; Rodriguez-Brito et al. 2010) and had an effect on our ability to detect members of the community.
Table 2

Species diversity and evenness estimations at the different sampling time points


99 %

97 %

95 %






























August ’11




















April ’13










Mean ± SD

23.8 ± 8.32


17.4 ± 6.81


16 ± 5.83


UNIFRAC and LIBSHUFF analyses show statistical similarity in OTUs through time

We statistically evaluated the observed between sample similarities at different times using UNIFRAC, which analyzed a maximum likelihood tree containing all 349 bop sequences and an environmental file listing the sampling time point each sequence. The pairwise UNIFRAC distances calculated are listed in supplementary table S1. The UNIFRAC distances between each sampled bacteriorhodopsin-containing community pair at different times ranged from 0.36 to 0.62, which statistically confirms the qualitative observation for the existence of core OTUs, and that a large number of non-core OTUs are shared between samples. Jackknifing analysis was performed and the UPGMA dendrogram that was derived from the UNIFRAC output data is shown in Fig. 3. The small UNIFRAC distances, and the Jackknifing statistic for samples indicates that the January’10 and May’12 samples, as well as the April‘13 and June‘10 samples are robustly and statistically similar. These analyses indicate the presence of similar bacteriorhodopsin-containing communities through the years, since the highly supported clustering seen between the January’10 and May’12 samples, and the June‘10 and April‘13 samples is a function of detecting many of the same OTUs between those time points.
Fig. 3

Community dynamics. UPGMA dendrogram of the environments based on the clustering of the sequences from each time point on the maximum likelihood tree (all 349 sequences). Jackknifing provided the branch support. Statistically insignificant values (>0.0025) from LIBSHUFF pairwise analysis are plotted. Unidirectional dashed arrows indicate that one community is a subset of the other (brown May’12; blue January’10; red June’10) and bidirectional solid arrows indicate two communities are subsets of each other

LIBSHUFF carries out pairwise comparisons to determine if one data set is a subset of the other. Allowing for a 5 % false detection rate and applying the Bonferroni correction for multiple library comparisons, only p values less than 0.0025 are considered statistically significant for inferring that two samples are different. Results from LIBSHUFF analyses are presented in supplementary table S2 and are mapped onto the UNIFRAC derived similarity data in Fig. 3. The January’10 and May’12, and August’11 and May’12 data sets are subsets of each other. Each sample except for August’11 is a statistically robust subset of April’13. These results indicate that the bacteriorhodopsin-containing haloarchaeal community composition is statistically similar through time and that differences seen in OTU absence/presence likely reflect the natural rise and fall in taxon abundances above and below our detection limits, rather than the colonization of new taxa.


Bacteriorhodopsin, a member of the halobacterial rhodopsin protein family, is a light-driven proton-pump that generates an electrochemical gradient for the production of ATP (Oesterhelt and Stoeckenius 1973), and it is present in significant quantities in hypersaline environments (Javor 1983; Oren and Shilo 1981; Stoeckenius et al. 1985). In this study, and others, it is shown to recover the familiar genera and diversity in halophilic environments when compared to the 16S rRNA gene (Dillon et al. 2013; Pašic et al. 2005) and provides excellent support for binning haplotypes (Dillon et al. 2013). Similar to the studies using 16S rRNA genes to survey Haloarchaea in hypersaline environments (e.g., (Benlloch et al. 1995; DeMaere et al. 2013; Dillon et al. 2013; Fernández et al. 2014; Ghai et al. 2011; Mutlu et al. 2008; Ochsenreiter et al. 2002; Oh et al. 2010; Zhaxybayeva et al. 2013), we did find unknown diversity: ~85 % of the sequences returned an uncultured haloarchaeon from the non-redundant database and ~24 % were considered unknown from comparing against the 109 sequenced genomes. Haloarchaea have multiple, often highly divergent 16S rRNA gene copies, that greatly biases the interpretation of data in environmental analyses (Boucher et al. 2004; Wright 2006). Further, because the 16S rRNA gene is highly conserved, and recombines easily between haloarchaeal species (Boucher et al. 2004; Papke et al. 2007) diversity is hidden, as two species could easily share the same sequence. By using the bacteriorhodopsin gene, we circumvented many of those complications.

Haloquadratum was previously suggested to be a dominant organism in this salt crystallizer pond using morphological criteria (Oren 2002a) and polar lipid composition profiling (Oren et al. 1996). However, in this study, it was recovered from only two samples: August’11 and May’12. There seems to have been a possible bloom of Haloquadratum in the August’11, correlating positively to the increase in salinity (Table 1). These data are in agreement with other studies showing that ion concentrations are correlated with Haloquadratum abundance (Podell et al. 2014). Other ecological conditions like, rain, wind, and temperature remained stable for weeks prior to each sampling time point. One month prior to each sampling, there is no record of rain, wind speeds fluctuated from 6.4 to 19 km h−1 and the difference in temperature between the coolest and warmest day was 19 °C. (see http://www.weatherundergound.com). PCR biases that cause differences in the ability to amplify DNA are known to exist (Suzuki and Giovannoni 1996). However, the primers used have recovered Haloquadratum sequences in previous studies (Dillon et al. 2013; Papke et al. 2003; Pašic et al. 2005) with Haloquadratum-related sequences being the most abundant in (Dillon et al. 2013). In total, these results suggest that Haloquadratum is a member of this community, often below our detection limits, but that sometimes it dominates when the salinity conditions are favorable.

Many studies on soils, rivers, lakes or other aquatic microbial communities indicate a lack of temporal stability (Bardgett et al. 1999; Dickerson and Williams 2014; Dillon et al. 2009; Liu et al. 2011, 2013; Rogers and Tate 2001; Väätänen 1980; Wu et al. 2011; Zaccone et al. 2014). However, those communities also experienced large fluctuations in environmental conditions. Further, dramatic community shifts in response to perturbations may only cause the natural abundances of native taxa to rise and fall without the gain or loss of taxa (e.g., Rodriguez et al. 2015). The seasonal environmental changes in southern Israel are markedly less fluctuating, and our results indicate the existence of a core set of bacteriorhodopsin OTUs in the Eilat salt crystallizer pond that experiences some abundance fluctuations in reaction to salinity changes. In a previous study surveying the impact of salinity and seasonal changes on microbial diversity in the Israel Salt Industries Ltd. in Eilat, similar findings were reported: two 16S rRNA gene OTUs were identified as present in every sample regardless of the salinity or season with varying relative ratios between the two (Litchfield et al. 2009). In this study, four OTUs were retrieved from all five sampling time points over 3 years and comprised ~52 % of the overall sequences obtained. Further, OTUs covering 89 % of the total sequences were found in more than one time point and, the June’10 sample shared every one of its OTUs with another sample suggesting the OTUs were present throughout, sometimes escaping detection. Statistical analyses support this interpretation. The Spearman’s correlation shows that the individuals detected are not present by chance; the detected number of OTUs at each sampling time is statistically neither over nor under abundant regardless of OTU definition; OTU abundances are significantly negatively and positively correlated indicating detected sequences likely capture natural fluctuations of the bacteriorhodopsin-containing populations; both LIBSHUFF and UNIFRAC analyses determine that the datasets are highly similar and each is typically a subcomponent of the other. An alternative explanation for the data is that the differences seen in composition between sample times are due to the colonization of dispersed cells from other environments that then outcompete the established Eilat Saltern cells for niche space and rise in population frequency to above our detection limits. This alternative explanation with frequent colonization events is non-parsimonious, and is not supported by the statistical analyses.

Our interpretations above are further corroborated by other studies on hypersaline environments and indicate haloarchaeal communities are likely to be globally stable with minor oscillations in population abundances that are correlated with environmental factors. Temporal studies on the thalassohaline Lake Tyrrell in Australia (Podell et al. 2014); the saturated brines from Sfax solar saltern in Tunisia (Boujelben et al. 2012); the Bras del Port solar saltern in Santa Pola, Spain (Gasol et al. 2004; Gomariz et al. 2015a, b); and the South Bay Salt Works in Chula Vista, USA (Rodriguez-Brito et al. 2010), all showed that the detected haloarchaeal community membership at the sequence, OTU, and/or genus level was largely constant across sampled time points. Fluctuation in taxon relative abundances was observed in Lake Tyrrell, and there was an ion concentration-dependent negative correlation in the abundances of some genera (e.g., Halorubrum and Haloquadratum), while other community members like the Nanohaloarchaea (Ghai et al. 2011; Narasingarao et al. 2012; Zhaxybayeva et al. 2013) and Halorhabdus showed no correlation (Podell et al. 2014). Similar results were observed in Sfax solar saltern where 95 % of the OTU and sequence fluctuations correlated with ecological parameters (Boujelben et al. 2012). This is consistent with what is observed in our study where abundances within core OTUs varied with salinity, and most non-core OTUs were found in multiple samples. Bacterial components of hypersaline communities also appear to have similar outcomes (Gasol et al. 2004; Gomariz et al. 2015a), indicating a general phenomenon of extreme hypersaline environments. By and large, hypersaline environments appear to provide constant ecological conditions, probably because they are typically found in hot dry climates, and appear to promote locally stable microbial communities across the globe.


This study recovered the presence of a stable OTU core representing the bacteriorhodopsin-producing haloarchaeal community inhabiting the salt crystallizing ponds in Eilat, Israel. These four core OTUs comprised ~52 % of overall sequences with only ~11 % of the sequencings being unique to any one time point, and contained the following genera—Halorubrum, Haloarcula, Haloterrigena, Halomicrobium, Halosimplex, Haloquadratum, Natronorubrum, Haloplannus, Halobiforma, and Haloferax. Fluctuations in relative OTU abundances corresponded to natural variations of salinity. Because evidence from many temporal studies on hypersaline habitats, especially saturated brines, also demonstrate the existence of stable haloarchaeal and bacterial communities worldwide, we hypothesize that communities are resistant to dispersed invading species, at least temporarily, and might lead to the formation of endemic hypersaline adapted communities and populations.


The authors thank Salt of the Earth Eilat Ltd. for allowing access to the Eilat salterns; the Interuniversity Institute for Marine Sciences of Eilat for logistic support; the UConn Bioinformatics Facility for providing computing resources; and the reviewers for their invaluable comments. This research was supported by the National Science Foundation (award numbers, DEB0919290 and DEB0830024), the US-Israel Binational Science Foundation (Grant Number 2013061) and NASA Astrobiology: Exobiology and Evolutionary Biology Program Element (Grant Numbers NNX12AD70G and NNX15AM09G).

Supplementary material

792_2016_864_MOESM1_ESM.pdf (555 kb)
Supplementary material 1 (PDF 555 kb)

Copyright information

© Springer Japan 2016

Authors and Affiliations

  • Nikhil Ram-Mohan
    • 1
  • Aharon Oren
    • 2
  • R. Thane Papke
    • 1
  1. 1.Department of Molecular and Cell BiologyUniversity of ConnecticutStorrsUSA
  2. 2.Department of Plant and Environmental SciencesThe Alexander Silberman Institute of Life Sciences, The Hebrew University of JerusalemJerusalemIsrael

Personalised recommendations