Spatio-temporal insights into microbiology of the freshwater-to- hypersaline, oxic-hypoxic-euxinic waters of Ursu Lake

Ursu Lake is located in the Middle Miocene salt deposit of Central Romania. It is stratified, and the water column has three distinct water masses: an upper, freshwater-to-moderately saline stratum (0-3 m), an intermediate stratum exhibiting a steep halocline (3-3.5 m), and a lower, hypersaline stratum (4 m and below) that is euxinic (i.e. anoxic and sulphidic). Recent studies have characterised the lake’s microbial taxonomy, and given rise to intriguing ecological questions. Here, we explore whether the communities are dynamic or stable in relation to This article is protected by copyright. All rights reserved. taxonomic composition, geochemistry and biophysics, and ecophysiological functions during the annual cycle. We found: (i) seasonally fluctuating, light-dependent communities in the upper layer (≥0.987-0.990 water-activity), a stable but phylogenetically diverse population of heterotrophs in the hypersaline stratum (water activities down to 0.762), and a persistent plate of green sulphur bacteria that connects these two (0.958-0.956 water activity) at 3-3.5 m; (ii) communities which might be be involved in carbonand sulphur cycling between and within the lake’s three main water masses; (iii) uncultured lineages including Acetothermia (OP1), Candidate Phyla Radiation, Cloacimonetes (WWE1), Marinimicrobia (SAR406), Omnitrophicaeota (OP3), Parcubacteria (OD1), and SR1, in the hypersaline stratum (likely involved in the anaerobic steps of carbonand sulphur cycling); and (iv) that species richness and habitat stability are associated with high redox-potentials. Ursu Lake has a unique and complex ecology, exhibiting both dynamic fluctuations and stability, and can be used as a comparator system for other stratified hypersaline systems and a modern analogue for ancient euxinic water bodies.


Introduction
Meromictic lakes, permanently stratified inland water bodies, occur in a range of geological settings, and can host stratified microbial ecosystems within their chemically distinctive water layers (Boehrer and Schultze, 2008;Lee et al., 2018). Those developed on ancient salt deposits of Central Romania, such as Ursu Lake, have a unique hydrology (Alexe et al., 2018). Ursu Lake has a permanently stratified halocline, from low salt (near-fresh) water at the surface, to saturated bottom brine (Máthé et al., 2014;Andrei et al., 2015;Alexe et al., 2018). Whereas a number of deep-sea haloclines have been well-studied, their microbial ecosystem lacks primary production (e.g. Yakimov et al., 2019), and whereas hypersaline systems such as crystallizer ponds and deliquescing halite have been well-characterised (La Cono et al., 2019), these lack a freshwater upper stratum. Ursu Lake is considered the largest, deepest, geologically This article is protected by copyright. All rights reserved.
youngest, continental lake yet documented (Alexe et al., 2018) which is at the same time heliothermal (entrapping the sun's heat), meromictic, and euxinic (anaerobic and sulphidic). Furthermore, it concomitantly represents a low salt/ high water-activity habitat, intermediate salt-concentration habitat, and hypersaline habitat that are collectively able to host microbes that are primary producers, halophilic heterotrophs, and from other ecophysiological groups.
Ursu Lake has a surface area of 4.12 ha, which is slightly larger than that of the CaCl2saturated Don Juan Pond in Antarctica (approx. 3 ha). However, the considerable depth of Ursu Lake means that its volume is almost 500 000 m 3 (Alexe et al., 2018) compared with only 3 000 m 3 for the Don Juan Pond. The three primary water masses within Ursu Lake are the upper, freshwater-to-moderately saline layer known as the mixolimnion (from 0-3 m), the intermediate stratum exhibiting a steep halocline (from 3-3.5-4 m), and the lower, hypersaline stratum known as the monimolimnion (from about 4 m down to 11-18 m) that is euxinic. These water masses are maintained by a combination of freshwater inputs from rivulets, precipitation, and the underlying evaporite deposit of halite dating from Middle Miocene (Badenian) period, ca. 14 Ma ago (Peryt, 2008;Alexe et al., 2018). There is also evidence that physical chemistry of brines prevents mixing with overlying waters of lower salinity (Thorpe, 1969;Raup, 1970).
Among the 41 salt lakes of the Transylvanian Basin (Central Romania), Ursu Lake is the largest and deepest saline lake of natural origin, and is one of the largest hypersaline, meromictic heliothermal lakes in the world Alexe et al., 2018). Cultivationbased (Máthé et al., 2014;Baricz et al. 2015) and metabarcoding analyses Andrei et al., 2017) indicated higher prokaryotic diversity than in comparable lake systems, mainly owing to the high diversity of Bacteria. There is, however, a paucity of information about seasonal dynamics, biophysics, and ecophysiology of the various microbial communities present in the lake. We carried out the current study to elucidate the complexity of the Ursu Lake ecosystem. The specific aims were to: (i) assess the spatio-temporal dynamics of the prokaryotic communities along the salinity/ water activity and redox gradients, and during the annual cycle, and ascertain environmental variables that can act as determinants of microbial diversity; (ii) seek evidence of nutrient cycling with the lake system; (iii) draw inferences on the This article is protected by copyright. All rights reserved.
keystone species and functional roles of microbes in this unique freshwater-to-hypersaline lake; and (iv) assess stability versus fluctuation within the overall ecology of the system.

Results and discussion
Ursu Lake is a unique freshwater-to-hypersaline ecosystem The density-separated, stratified water column of Ursu Lake was maintained throughout the seasons of the annual cycle in relation to the physical, chemical and biological parameters ( Fig.   1, Tables S1-S6). Overall, the ionic composition of Ursu Lake resembles that of sea water, with NaCl as the main salt component . Determinations of water activity down through the water column according to samples taken in summer 2015 and spring 2016 validated the salinity gradient from surface (0.987-0.990 water activity, relatively close to freshwater) down to 11 m depth (0.763-0.764 water activity, equivalent to saturated brine) (Table S1). By its continuous density-stratification, Ursu Lake can be classified as meromictic (Hammer, 1986;Boehrer and Schultze, 2008).
The upper, slightly saline layer of the mixolimnion (down to 2.5 m and roughly 4.6-7.7% w/v total salinity; Fig 1, Table S1) was characterised by water-activity values ( 0.975-0.990) that are optimal for growth of many microbes. The most-severe halocline ranged between 3 and 4 m in depth (i.e., the oxic/anoxic transition zone, 0.786-0.958 water activity); and a hypersaline monimolimnion extended from 4 m downwards (0.762-0.776 water activity; 35-37% w/v total salinity) Table S1). The low-to-moderate saline mixolimnion of Ursu Lake is oxygenated and euphotic, and is prone to seasonal fluctuations in chemistry, water activity, and temperature (Table S1). Ursu Lake is heliothermal as the temperature of the underlying water mass (2.5-4 m) reaches~40 o C during summer (Table S1). The upper, slightly saline layer of the mixolimnion (down to 2.5 m and roughly 4.6-7.7% w/v total salinity; Fig 1, Table S1) was characterised by water-activity values (0.975-0.990) that are optimal for growth of many microbes. The mostsevere halocline ranged between 3-m depths (i.e., the oxic/anoxic transition zone, 0.786-0.958water activity); from 4 m downwards, a hypersaline monimolimnion (0.762-0.776 water activity, 35-37% w/v total salinity) was evidenced (Table S1).
This article is protected by copyright. All rights reserved.
The low-to-moderate saline mixolimnion of Ursu Lake is oxygenated and euphotic, and is prone to seasonal fluctuations in chemistry, water activity, and temperature (Table S1). Ursu Lake is heliothermal as the temperature of the underlying water mass (2.5-4 m) reaches ~40 o C during summer (Table S1). Plots of salt concentration, dissolved oxygen and redox potential versus depth of the water column indicate that the halocline is more gradual at depths of 3-4 m in the summer season whereas an oxycline and redoxcline occur between 2 and 3 m but are both closer to surface than during other seasons (TableS1). The intermediate stratum (3-4 m deep) is characterised by steep gradients of salinity, pH, light, dissolved oxygen, and redox potential. Beneath this stratum, the lake water becomes more temperature-stable (~20 o C), pHstable (pH~6.0), hypersaline (> 30% w/v salt), and oxygen-depleted (< 0.2 mg O2 L -1 ), and is aphotic, highly reduced (> -300 mV), and sulphide-rich (~ 5 mM H2S) (Fig 1, TableS1).
The stratification of salinity, water activity, and concentrations of total carbon, total nitrogen, sulphate, and ammonium are temporally stable (Fig 1, Tables S2-S5). Total carbon, total inorganic carbon, dissolved carbon and dissolved inorganic carbon concentrations were highly correlated with each other (Pearson's r > 0.9, p = 0.05) and with total nitrogen and total dissolved nitrogen (Fig. S1). Total organic carbon in the mixolimnion was relatively constant during autumn, winter and spring. During the summer, however, total organic carbon increased, probably due to inputs of allochthonous carbon that originate from littoral macrophytes and decayed biomass of invertebrates (Alexe et al., 2018). Throughout the study period, particulate carbon and dissolved organic carbon (also, therefore, total carbon) appeared to accumulate within the 3-5 m stratum (Fig. 1, Table S2). Below 5 m, the concentration of total organic carbon also remained more or less constant. The phosphorus concentration (made up of dissolved and particulate fractions) appeared uniform down to a depth of 2 m, increased below 2 m and was highest between 3 and 5 m and then decreased slightly at lower depths ( Fig. 1, Table S4).
Accumulation of total carbon, nitrogen, and phosphorous at 3 to 5 m depths is likely to be a consequence of slower settling velocity of organic matter at density interface (Prairie et al., 2015), and low rates of carbon mineralization under the anaerobic, hypersaline conditions (Boehrer et al., 2017). As the highest total cell counts were also recorded in the intermediate stratum (at 3 m depth; these were ~2 x 10 8 cells mL -1 in summer, autumn, and winter, and ~1.3 This article is protected by copyright. All rights reserved. x 10 7 cells mL -1 in spring), we assumed that the organic matter accumulation at this depth is, in part at least, due to active populations of microbes (Table S6). Incident light, measured as photosynthetically active radiation (PAR), penetrates Ursu Lake to a depth of 3-3.5 m, where measurements of total photosynthetic pigments indicated the formation of deep-chlorophyll layer (Table S6). Counts of picocyanobacteria and picoeukaryotes revealed maxima at 1 m, and 3-4 m depth (40 -47 x 10 6 cells mL -1 during spring), respectively (Table S6). Relative abundances of picoeukaryotes correlated positively with total organic carbon, total dissolved nitrogen and phosphorous while total cell counts correlated with total organic carbon and photosynthetic pigments (Table S7). The scarcity of oxygen below the oxic/anoxic interface is consistent with the elevated ammonium, methane, sulphide, and sulphate concentrations ( Fig.   1, Tables S1, S2, S5). The persistence of simultaneous oxygen-depleted and high-sulphide conditions below a depth of 5 m indicates that the habitat is euxinic (Meyer and Kump, 2008).
As sulphides and methane, originating in the monimolimnion, where not detected above 3 m depth, it is probable that their oxidation occurs within the 3-3.5 m layer. In combination, these characteristics of Ursu Lake make it distinct from other meromictic ecosystems worldwide, as this lake i) is permanently heliothermal and density-stratified; ii) with a water column spanning from near-freshwater-to-hypersaline brine, with euxinic deeper water, and iii) has substantial depth with a large surface area (Alexe et al., 2018).
Contrasting microbiology is found within the three-strata water column of Ursu Lake To assess stability and fluctuation within the microbial communities of the mixolimnion, intermediate stratum and monimolimnion, we performed metabarcoding of samples taken at eight discrete depths along the water column during four successive seasons. The sampling points were chosen in correspondence with the stratification of physicochemical parameters, based on in situ measurements, to cover all three main water masses of the meromictic Ursu Lake: mixolimnion (0.5-2 m), intermediary stratum (3 m), hypersaline stratum (3.5-9 m), a sampling strategy in line with those used for other stratified hypersaline systems (Klepac-Ceraj et al., 2012;Yau et al., 2013;Baatar et al., 2016). Our previous analysis  revealed that that the permanently stratified Ursu Lake appeared as having a physicochemically This article is protected by copyright. All rights reserved. and biodiversity-stable bottom water, while the upper part of water body appeared more dynamic; this is why sampling was carried out at the two monimolimnetic depth sample points, and more-frequent sample points within the mixolimnetic and transition water strata. By including a second dimension (time), the current study, supported by data from the 32 samplings, provides relevant spatio-temporal insights into the stratified ecosystem of the Ursu Lake.
After quality-filtering and operational taxonomic unit (OTU)-clustering at the 97% sequence identity, we obtained 1,244,800 good-quality reads grouped into 1,649 OTUs.
Amongst these, 46 OTUs were affiliated to the Archaea and 1,168 OTUs to the Bacteria. To discriminate between individual OTUs, we grouped the retrieved phylotypes in 'Abundant OTUs' (i.e. ≥ 1%, equivalent to ≥ 389 reads abundance in at least one sample), and the relatively 'Rare OTUs' (i.e. <1% or below 389 reads abundance in all samples). Although not reaching a clear asymptote ( Fig. 2A), rarefaction curves indicate that the sequencing effort detected the majority of species possibly present in our samples and increasing sequencing depth uncovers only low abundance taxa. We found 143 Abundant OTUs and 1,506 Rare OTUs accounting for 86.3% and 13.7% of the total number of sequences, respectively. A significant proportion of reads (21.6% of the total, hereinafter referred to as 'unassigned') presented low similarities to characterized prokaryotic lineages. There was considerable variability in the composition of microbial communities down through the water column; 62 Abundant OTUs were unique to the mixolimnion, 21 were unique to the intermediate stratum, and 40 were found only in the monimolimnion. Few Abundant OTUs were common to all three water masses (Fig.   2B), a finding that is consistent with microbial tolerance to different biophysical and nutritional conditions (Yau et al., 2013;Stevenson et al., 2015;Hamilton et al., 2015;Meyerhof et al., 2016).
This article is protected by copyright. All rights reserved.
Taxa in the Planctomycetes (Planctomyces sp.) and Verrucomicrobia (Verrucomicrobiaceae, Coraliomargarita sp.) are generally less abundant but nevertheless increased in abundance following the summer peak in phytoplankton. Genomes of Planctomyces have been found to be enriched in carbohydrate-active enzymes (i.e. sulfatases) and are presumed to be involved in the initial breakdown of the sulphated heteropolysaccharides produced by various algae (Woebken et al., 2007), subsequently using their carbon skeletons as an energy source (Glöckner et al., 2003). Similarly, Puniceicoccaceae representatives have been shown to This article is protected by copyright. All rights reserved. assimilate exudates from Synechococcus sp. in euphotic waters (Nelson and Carlson, 2012).
Given their widespread distribution and heterotrophic lifestyle, members of the Verrucomicrobiaceae might be involved in the utilization of phytoplankton-derived organic matter, either by protein assimilation or by polysaccharide hydrolysis (Orsi et al., 2016).
Down through the intermediate stratum, there is not only a transition from low salinity to hypersaline conditions, but also from euphotic to aphotic, and oxic to anoxic conditions. This stratum is, therefore, highly heterogeneous and accommodates ecophysiologically diverse  Table S9). This finding is consistent with patterns observed in other meromictic lakes where anoxygenic phototrophs dominate the intermediate stratum (Lehours et al. 2005;Degermendzhy et al., 2010;Lauro et al., 2011). The predominance of green sulphur bacteria reads at 3-3.5 m was corroborated by the high chlorophyll concentrations, high total cell counts, and high concentration of bacteriocholorophyll pigments, according to thin-layer chromatography (data not shown). These photoautotrophs contribute considerably to oxidative sulphur cycling and carbon photo-assimilation in stratified lakes (Guyouneaud et al., 2001).
Furthermore, the diagenetic products of bacteriochlorophylls and other pigments (e.g. isorenieratene) derived from green sulphur bacteria and found in sediments, have been used as biomarkers for ancient (e.g. Quaternary) photic-zone euxinia events (Damsté et al., 2001;Meyer and Kump, 2008). Micro-oxic conditions were recorded in the intermediate layer (3-3.5 m); nevertheless, we believe that the aerobic phylotypes identified exist within the necromass that sinks downwards from the mixolimnion. Similarly, the aerobic taxa detected in the euxinic This article is protected by copyright. All rights reserved.
(lower part) of the monimolimnion (i.e. Cyanobacteria, represented by Synechococcus sp.related reads, were abundant in each of the three distinct water masses) may represent necromass and/or inactive biomass that has drifted down from the mixolimnion. Within the intermediate stratum (3-3.5 m), we detected taxa involved in methanotrophy and Fe/Mn metabolism (Tables S8 and S10). Methylococcales (represented by OTU244) oxidise and assimilate C1 compounds (Paul et al., 2016) while the rarely microaerophilic Defferibacterales (represented by OTU50) can use various terminal electron acceptors including Fe(III), Mn(IV), S(0), Co(III), and nitrate, and some could use fermentative metabolism. Although there have been earlier reports of these taxa in saline habitats (Schneider et al., 2013;Paul et al., 2016), whether they are metabolically active within the hypersaline stratum of Ursu Lake remains unclear. The presence of Abundant OTUs assigned to sulphate-reducing bacteria (Desulfovibrionales and Desulfobacterales related OTUs) in the samples from the intermediate stratum (3 m depth) is consistent with studies on differential oxygen tolerance of sulphatereducing bacteria, some taxa being able to survive in the oxic zones of microbial mats (Jonkers et al., 2003). The high-salt (>30% w/v), oxygen-depleted stratum (3.5-9 m) favours abundant phylotypes related to Methanobacteria, Bacilli (Lactobacillales order), Lentisphaerae, and candidate phylum Acetothermia (OP1) that may be involved in anaerobic carbon cycling.
Alphadiversity patterns revealed by our analyses indicate that prokaryotic richness is enhanced in the stable conditions of euxinic monimolimnion (Fig. S4). Community stability over time is supported by analysis of similarity (ANOSIM, R=-0.007, p=0.445). The four-way set difference analysis (Fig. 2C) indicated that 38 Abundant OTUs (accounting for 63% of reads associated to all Abundant OTUs) were detected all year-round (Table S10). However, temporal variation was apparent from the distribution of OTUs unique to each season (11 in July and 13 in November 2015; 19 in February and 18 in April 2016). Albeit that these OTUs were phylogenetically related to phylotypes encountered throughout the year (Fig 2C, Table S10), the mixolimnion appears as a more seasonally fluctuating environment compared to the deeper strata, as evidenced by the succession of aerobic heterotrophic Bacteroidetes, Proteobacteria and Verrucomicrobia bacterial clades (Fig. S3, Table S10). Aeromonadales affiliated OTUs were abundant in spring, Chromatiales in spring and autumn, while Gammaproteobacteria This article is protected by copyright. All rights reserved.
OTUs affiliated to the Legionellales, Methylococcales, and Thiohalorhabdales were abundant in autumn. Most of Bacteroidetes -affiliated OTUs exhibited high abundances in all seasons, except Sphingobacteria which was abundant in spring, and Cytophagia-Roseivirga which was abundant in spring and summer. Defferibacteres and Planctomycetia OTUs were detected at higher numbers in spring, SR1 was abundant in autumn and winter, while Cloacimonetes (WWE1) was abundant in winter (Table S10) We have detected increases in abundance for OTUs affiliated to Flavobacteriaceae (Psychroflexus sp. and Polaribacter sp.) spring and winter, and a significant proportion of reads pertaining to Cryomorphaceae throughout the year (Tables S8). Members of Flavobacteria, such as Polaribacter sp. can degrade complex algal polysaccharides via the production of diverse glycoside-hydrolases (Teeling et al., 2016). On the other hand, Delmont et al. (2015) suggest involvement of Cryomorphaceae in algal decomposition, as a particular functional gene pool related to virulence and host invasion was found in Cryomorphaceae bacteria associated to a Phaeocystis bloom. We observed seasonal fluctuations of Abundant OTUs affiliated to and Vibrio sp.
(Gammaproteobacteria) ( Table S8). It is possible that the succession of distinct bacterial population to be related to the diversity of algal (phytoplankton) exudates, as no individual bacterial species is known to possess all of the genes required for complete degradation of all naturally occurring polysaccharides (Teeling et al., 2016).
Salt concentration and water activity, pH, and redox determine the fine-scale ecology of Ursu Lake For samples taken down through the 1-37% (w/v) salt water column, downstream alpha-and beta-diversity analyses were performed. All 32 samples (at eight discrete depths and during four seasons) for which a total of 2,469,892 reads (with a mean length of 441 bp per read) were recovered. Due to the variations in the number of reads per sample, the rarefaction and diversity analyses were performed at the smallest sequencing depth (i.e. 38,900, representing a total of 1,244,800 high-quality reads). In hypersaline systems, biodiversity is determined, and can be constrained, by salt-induced cellular stress which, in turn, relates to adaptation to ionic This article is protected by copyright. All rights reserved.
conditions, osmotic stress and water activity (Laron and Belovsky, 2013;Alves et al., 2015;Stevenson et al., 2015;Lee et al., 2018, Merlino et al., 2018. Prokaryotic richness was highest in the extreme conditions, including low water activity values of the hypersaline stratum (0.763-0.764 at 11 m; Table S1). This finding is consistent with recent evidence showing that the wateractivity of NaCl-saturated brines is thermodynamically mid-range rather than extreme Lee et al., 2018). In Ursu Lake, the impact of environmental parameters on prokaryotic diversity is illustrated by high spatial variations with increases of up to two-fold in calculated richness as depth decreases, e.g. the chao1 indices ranged between 350-587 for the mixolimnion (0.5-2 m) and 785-1055 for the hypersaline stratum (4 -9 m), and the observed species (OS) indices increased from ~300 in the mixolimnion to >700 in the hypersaline stratum (Fig. S4, Table S11). Increasing richness with depth (higher salinity, lower water activity) was reported in other meromictic saline lakes, including high-latitude coastal lakes (Lauro et al., 2011) and inland alkaline saline (Dimitriu et al., 2008), pH-neutral saline (Baatar et al., 2016) and hypersaline lakes (Baricz et al., 2014;Andrei et al., 2015). The richness indices correlated strongly with salinity, total carbon, total nitrogen, sulphate, sulphides, ammonium, and phosphate concentrations (Fig. 5A, 5B, Table S12). Limited temporal variations in richness and evenness raise the prospect that no significant changes occurred during the sampling period, the seasonal homogeneity endorsing the independence of microbial community from seasonal blooms of phytoplankton.
Evenness indices (Shannon-H and Simpson-S) showed similar values down through the water column and there were no apparent correlations with any of the measured physicochemical parameters (Fig. 5A) However, the reduced evenness (Table S11)  the anoxygenic phototrophs affiliated to green (Lauro et al., 2011;Laybourn-Parry et al., 2014) or purple sulphur bacteria Klepac-Ceraj et al., 2012) become dominant.
This article is protected by copyright. All rights reserved.
The Mantel test performed for the physicochemical data and beta-diversity (based on Bray-Curtis distance matrix) indicated the significant impact (Mantel r = 0.8219, p = 0.001) of selected environmental factors on the prokaryotic community composition. PCoA ordination plot, based on Bray-Curtis dissimilarity matrices, was used to test for patterns and to further identify the contribution of environmental factors in shaping the spatial and temporal variability of the microbial community. PCoA ordination (carried out to order/arrange samples characterized by values on multiple variables) separated samples into three distinctive depthdependent zones (Fig. 5B). The mixolimnetic microbial community is impacted by the effects of pH (that can be used as a proxy for photosynthetic activity of phytoplankton) and redox potential (a proxy for dissolved oxygen), whereas the monimolimnetic microbiome is affected by salinity and availability of reduced nutrients. The depth-related decrease of pH and dissolved oxygen are presumably driven by mineralization of organic matter. Accordingly, depth-related decreases in pH, oxygen and nutrients, and increases in salinity apparently select for anaerobic halophilic community (Meyerhof et al., 2016).

Low-abundance microbes flourish in Ursu Lake
The species distribution pattern indicates the presence of a considerable number of lowabundance phylotypes in the water column of Ursu Lake (Fig. 2D). Rarefaction analysis ( Fig.   2A), as well as richness indicators (Fig S4, Table S11), suggest an increase in the prokaryotic diversity down through the water column, that may be seen as the outcome of increasing proportion of Rare phylotypes. Out of the 1649 OTUs, 1506 (91%) are Rare (i.e. present in This article is protected by copyright. All rights reserved. abundances below 1%). We found that 401 Rare OTUs (24.3% of total number of OTUs) are shared between all three water strata (Fig. 2E). Interestingly, the proportion of Rare  (Fig. 2E).
Within the detection limits of the analysis, approximately 48% (790 OTUs) of all Rare OTUs identified in Ursu Lake were confirmed as present during each season (Fig. 2F) S5). It is nevertheless likely that some community changes may occur below the threshold for detection. Any populations that diminish to levels that are undetectable (≤1-10 reads) may be dormant or quiescent, and / or stressed by biotic or abiotic factors, and either in decline or lag phase until they recover during subsequent seasons. It is also possible that there are occasional extinctions (and, indeed, occasional introductions or evolution of new phylotypes) for some of the Rare taxa within the microbiome. (Fig. S5). One intriguing finding was that several lineages that are thought uncultivable (i.e. Parvarchaeota, OP3, OP8, OP9, OP11, TM6, TM7) were found only as Rare microbes. This is corroborated by the phylogenetic diversity of uncultivated microbes found in the bottom organic-rich (sapropelic) sediment of Ursu Lake (Andrei et al., 2017).
This article is protected by copyright. All rights reserved.
Our analyses indicated a 1:10 ratio between Abundant-and Rare taxa within the microbial communities of Ursu Lake; a ratio that is not surprising given that low abundance phylotypes, by definition, represent the majority of the microbial diversity usually present in individual ecosystems (Pedrós-Alió, 2012; Lynch and Neufeld, 2015). Down through the three strata of the Ursu Lake water column, the majority of Rare phylotypes were affiliated to Actinobacteria, Bacteroidetes, Firmicutes, Proteobacteria, and Verrucomicrobia (Table S13).
The phylogenetic similarities at high-taxonomic level (i.e. class) of Abundant-and Rare microbes in Ursu Lake samples (Table S8, Table S13) provides evidence that rare phylotypes are active in situ, even when the methodology does not distinguish directly between cells that are metabolically inactive and active (Galand et al., 2009). The increased diversity of the Rare phylotypes in the euxinic monimolimnion (4-9 m), compared to the upper strata, could be related to higher metabolic diversity and various substrates (i.e. highly diverse carbon sources, electron donors and terminal electron acceptors) availability in the anoxic strata of stratified freshwater (Lehours et al., 2005;Gies et al. 2014), alkaline (Dimitriu et al., 2008) or hypersaline lakes (Klepac-Ceraj et al., 2012;Andrei et al. 2015).
Whereas low water activity and hypersalinity select for halophilic taxa (Lee at al., 2018), the salt-saturated waters of Ursu Lake have a remarkably high bacterial diversity (Fig. 3-4).
Among the 1649 OTUs, only two archaeal OTUs were detected as Abundant, one affiliated to  . The rest of the Ursu Lake archaeal phylotypes, affiliated to Bathyarchaeota (Crenarchaeota of TACK superphylum), Aenigmarchaeota and Parvarchaeota (DPANN superphylum) were identified solely among the Rare OTUs (Table S13). It has been suggested that the Rare taxa may serve as 'seed bank' that harbours and provides ecophysiologically diverse microbes that maintain or enhance the metabolic diversity and functional roles of the in situ communities (Lynch and Neufeld, 2015). Yet, compared to the surface waters, the increased number of Rare OTUs in the stable conditions of Ursu Lake's monimolimnion suggests that they are less efficient, yet active as resource competitors (Hugoni This article is protected by copyright. All rights reserved. phylotypes (Elusimicrobia, Fibrobacteres, Lentisphaeria) and candidate phyla (BRC1, Cloacimonetes, Dependentiae, Hydrogenedentes, Omnitrophicaeota, Marinimicrobia, WPS2, (Table S13). It is clear that these microbes are ubiquitous within anoxic, sulphur and organic carbon-rich environments, but their ecophysiological roles have yet to be determined (Hamilton et al., 2016;Llorens-Marès et al., 2016).
Putative spatio-temporal function of the Ursu Lake ecosystem The geochemically stratified waters of meromictic lakes host disparate microbial communities, and microbe-driven processes, so can perform the complete cycling of nutrients (Lauro et al., 2011;Barberán and Casamayor, 2011;Meyerhof et al., 2016). The high taxonomic diversity of Ursu Lake's prokaryotic community argues for complex interactions between chemistry, biophysics and ecophysiology of the lake's microbiome (Hallsworth, 2018), some of which can be elucidated using SSU rRNA gene-based diversity (Aßhauer et al., 2015;Knight et al., 2018).
Within the euphotic zone of Ursu Lake, we identified a wide range of heterotrophic microorganisms (Actinobacteria, Alphaproteobacteria, Flavobacteria, and Gammaproteobacteria; Fig. 3), capable of working as consortia to degrade complex organic substrates derived from halotolerant phytoplankton (Teeling et al., 2016). Sulphur oxidisers are known to be active under moderately high salinities (up to 4 M NaCl) and microaerophilic conditions (Oren, 2011), so we believe that the sulphur oxidisers identified in Ursu Lake in the 0.5-3 m (Fig. 6) are active in situ. The large population of Prosthecochloris t a depth of 3 m that was observed regardless of season suggests the oxic/anoxic interface might be the main site of sulphur compounds turnover in Ursu Lake.
The Halanaerobiaceae is known to contain some of the most extreme halophiles, such as This article is protected by copyright. All rights reserved.
Halanaerobium lacusrosei that is capable of metabolism and growth at water activity values below that of saturated NaCl . Sulphate-reducing bacteria, mainly Deltaproteobacteria, and methanogenic functional types affiliated to Methanobacteriales, occur together in the hypersaline stratum although are known competitors for acetate; the available substrate.
Compared to the upper 3 m of the water column, in the 3.5-9 m portion, uncultivated microbes such as Parcubacteria (OD1) and Acetothermia (OP1) are abundant (Fig. 3, 6, Table   S8). Moreover, 26 out of the 143 Abundant OTUs (21% of total reads) exhibited low similarities to characterized prokaryotic lineages (referred to as 'unassigned'). The phylogenetic tree constructed by selecting the closest 20 sequence neighbours from the SSU/LSU 132 SILVA database (Fig. 6) indicated that a fraction of unassigned OTUs (Fig. 6, Table S14), Abundant in the monimolimnion, may be related to the Candidate Phyla Radiation (CPR). CPR bacteria have small genome size (often < 1 Mb) and exhibit multiple metabolic limitations, such as partial tricarboxylic acid cycle, lack of electron transport chain complexes, incomplete biosynthetic pathways for nucleotides and amino acids (Brown et al., 2015). All these suggest a codependent lifestyle in which these putative obligate fermenters obtain the much needed metabolites from the symbiotic bacterial or archaeal partner and reciprocate by providing fermentation end products (Wrighton et al., 2014;Nelson et al., 2015;Brown et al., 2015;Castelle et al., 2018). However, CPR representatives with respiratory and fermentative capacities as well as nitrogen and fatty acid metabolism have been detected by genome analysis (Castelle et al., 2017). The Acetothermia (OP1) affiliated OTUs in the monimolimnion of Ursu Lake are related to the candidate KB1 bacterial group, that has been consistently found in deep-sea hypersaline anoxic lakes (Nigro et al., 2016). Brine enrichments and radiotracer experiments that used 14 C-labeled glycine betaine (Yakimov et al., 2013), support the idea that these uncultured microorganisms assimilate glycine betaine as carbon and energy source.
According to Yakimov et al. (2013), in anoxic environments, the KB1 bacteria perform the reductive cleavage of glycine betaine, simultaneously producing acetate and trimethylamine, the latter supporting extremely halophilic methylotrophic methanogens. In addition to fermentative metabolism and co-dependent lifestyle (Hu et al., 2016;Hao et al., 2018), This article is protected by copyright. All rights reserved.
In the phylogenetic tree, the 15 of the unassigned OTUs cluster together (Fig. 6, Table   S14). The closest 20 sequence neighbours from the SSU/LSU 132 SILVA database for these 15 OTUs are affiliated to the Patescibacteria superphylum, which is known to contain Parcubacteria (OD1), Saccharibacteria (TM7), Gracilibacteria (GN02), and Microgenomates (OP11) (Peura et al., 2012). These fermentative microbes are known to thrive in anoxic, organic matter-rich environments, and are likely involved in cycling of C, H, and S (Wrighton et al., 2014). This is consistent with studies of nutrient cycling by Parcubacteria and Microgenomates that are associated to sulphur-rich ecosystems (Harris et al., 2004;Peura et al., 2012).
Throughout the year (Table S8, Table S14), Abundant Marinimicrobia (SAR406)-and Cloacimonetes (WWE1)-related OTUs were found in the 3-9 m portion of Ursu Lake's water column. The Marinimicrobia are thought to anaerobically digest recalcitrant polysaccharides such as cellulose (Limam et al., 2014) and Cloacimonetes proposed to perform S cycling within marine systems (Wrighton et al., 2014). Both candidate phyla were predicted to take part in syntrophic interactions in methanogenic environments; i.e. fermentative degradation of amino acids (Marinimicrobia) or propionate (Cloacimonetes) (Nobu et al., 2016). Genome-level studies indicated an eukaryote (amoeba)-dependent (parasitic) lifestyle of Dependentiae (TM6) (Yeoh et al., 2016), which could explain their maximum abundance in the intermediate stratum during the phytoplankton bloom in November 2015, while being part of the monimolimnetic's Rare taxa. Omnitrophicaeota (OP3) apparently thrive in anoxic environments, so share the habitat of methanogens. Metagenomic analyses indicated that OP3 is a diverse group comprising syntrophic bacteria with an anaerobic-respiring metabolism fuelled by formate or H2 (Glöckner et al., 2010).
Overall, our analyses revealed that the three water masses (characterised by differences in the thermodynamic parameter water activity) influence community stability and shape of Ursu Lake's microbiome, which is dominated by a taxonomically diverse microbial community involved in carbon and sulphur cycling.
This article is protected by copyright. All rights reserved.

Conclusions
Ursu Lake has revealed the complexity of its microbial assemblages and community structure, and that salinity, pH, redox, and reduced nutrients are correlated with spatio-temporal variability of microbial populations in its unique ecosystem. In the low-to moderate salinity (1 to 10% w/v salt), euphotic upper two meters we found seasonal fluctuations in dissolved chemical species, water activity as well as temperature, which favoured a diverse community of mesophilic and physicochemical parameters, based on in situ measurements, to cover all three water strata of the meromictic Ursu Lake (mixolimnion, intermediate stratum, hypersaline stratum). For environmental monitoring, the physical and chemical parameters were measured from 0.1 to 11 m of depth. Water samples were collected using a submersible 12-V electric pump, with a flow rate of 9.9 L·min -1 , and stored in sterile 2 L polypropylene bottles. To avoid crosscontamination, prior to collecting the samples, the inside of the pump-tubing system was purged with water from each depth for at least 15 minutes (∼150 L total purged volume) until physicochemical parameters stabilized following a procedure adapted from Harter et al. (2014).
Constant flow rate (9.9 L min -1 ) was applied with no change between purging and water sampling. The samples were kept cool and light-protected and transported to laboratory in less than 6 hours.

In situ measurements and chemical profiling
In situ measurements (e.g. temperature, pH, dissolved oxygen, and oxidation/reduction potential) were performed using a portable water multi-parameter system HI 9828 (Hanna Instruments, USA). Salinity (g L -1 ) was estimated based on electrical conductivity values measured with a Multi 340i multi-parameter (WTW, Germany), with a built-in temperature correction. Vertical profile of photosynthetic active radiation (PAR, 400-700 nm) was recorded This article is protected by copyright. All rights reserved.
with a spherical irradiance probe connected to a calibrated quantum irradiance meter (ULM-500, Walz GmbH, Germany). Main ionic content of water column in Ursu Lake was detailed elsewhere . More than 90% of total ions were represented by Na + and Clfollowed by K + , Mg 2+ , Ca 2+ , and carbonates and sulphate as minor anions and cations, respectively . Total alkalinity was measured by titration with HCl to the methyl orange indicator endpoint. Total nitrogen (TN) and total dissolved nitrogen (TDN) including free ammonia, ammonium, nitrite, nitrate, and organic nitrogen were analysed by catalytic combustion followed by oxidation of nitrogen monoxide to nitrogen dioxide with ozone and subsequent chemiluminescence detection. This article is protected by copyright. All rights reserved.
samples (10 discrete depths) were measured as described in Stevenson et al. (2015) (Caporaso et al., 2010;Edgar, 2013). The pair-end reads were joined in QIIME, followed by quality filtration, dereplication and singleton removal in Usearch v8. Both de novo and reference chimera checking were performed in Usearch v8, using the last version of the Greengenes database ('13_8') as a reference (DeSantis et al., 2006). The taxonomy was assigned for the representative OTUs (441 bp) in QIIME using the default classifier (Greengenes) against the last version of the Greengenes database ('13_8'), using 'specieslevel' OTU cut-off (97% sequence identity). The taxonomy was added to the OTU-table with the biom-format package, and the mitochondrial and plastidial sequences were filtered out of the final OTU table. To minimise the uncertainties in OTU affiliation, sequences from each OTU were also queried against SILVA-LTP (Yilmaz et al., 2014), RDP (Cole et al., 2014) and EMBL-This article is protected by copyright. All rights reserved.
ENA databases (Leinonen et al., 2011). For 16S rRNA genes classified as chloroplasts taxonomy was assigned using Greengenes database. The relative phylogenetic placement of the Abundant unassigned OTUs was determined using the online tool SINA 1.2.11 (Pruesse et al., 2012) and the SSU/LSU 132 SILVA database (Quast et al., 2013). The closest 20 neighbours for each query sequence were retrieved from the database, and were further used for a phylogenetic tree construction with the Neighbor-Joining method (Saitou and Nei, 1987) in Mega7 (Kumar et al., 2016). The Kimura 2-parameter method (Kimura, 1981)   This article is protected by copyright. All rights reserved.    This article is protected by copyright. All rights reserved. Accepted Article This article is protected by copyright. All rights reserved.     Accepted Article