Divergence of Biocrust Active Bacterial Communities in the Negev Desert During a Hydration-Desiccation Cycle

Rain events in arid environments are highly unpredictable and intersperse extended periods of drought. Therefore, tracking changes in desert soil bacterial communities during rain events, in the field, was seldom attempted. Here, we assessed rain-mediated dynamics of active bacterial communities in the Negev Desert biological soil crust (biocrust). Biocrust samples were collected during, and after a medium rainfall and dry soil was used as a control; we evaluated the changes in active bacterial composition, potential function, potential photosynthetic activity, and extracellular polysaccharide (EPS) production. We hypothesized that rain would activate the biocrust phototrophs (mainly Cyanobacteria), while desiccation would inhibit their activity. In contrast, the biocrust Actinobacteria would decline during rewetting and revive with desiccation. Our results showed that hydration increased chlorophyll content and EPS production. As expected, biocrust rewetting activated Cyanobacteria, which replaced the former dominant Actinobacteria, boosting potential autotrophic functions. However, desiccation of the biocrust did not immediately change the bacterial composition or potential function and was followed by a delayed decrease in chlorophyll and EPS levels. This dramatic shift in the community upon rewetting led to modifications in ecosystem services. We propose that following a rain event, the response of the active bacterial community lagged behind the biocrust water content due to the production of EPS which delayed desiccation and temporarily sustained the biocrust community activity.


Introduction
Drylands are the largest terrestrial biome on Earth and represents 35% of the world's landmass [70]. Rain in hot arid environments is rare and unpredictable, and the main source of water is dew [58], or fog [48]. This moisture is readily absorbed into the soil surface but quickly evaporates due to high temperatures and low humidity [20]. The long droughts in drylands limit plant growth, and in their stead, the soil is covered by microbial mats, named biological soil crust (biocrust). Biocrusts are soil surface matrices of phototrophic and heterotrophic microorganisms that bind soil particles, using extracellular polymeric substances [12,21,50], or fungal hyphae [11,52]. Biocrust phototrophs are the main primary producers in this habitat, and together with heterotrophs, they form a rigid and stable mat that can resist xerification and soil erosion [1,16].
Desert biocrusts are the main source of carbon and nitrogen for the soil microbiome [2] and are strong contributors to soil respiration [23]. It was recently shown that during long droughts, many biocrust microorganisms rely not only on photosynthesis but also on oxidation of atmospheric trace gases [55,61]. Once the biocrust is hydrated, phototrophs respond quickly by inducing their photosynthetic systems and related functions to take full advantage of the rare water abundance before the soil dehydrates [62]. In preparation to drought, photosynthetic members of the biocrust community either form a seed bank of species that can spring to life when the soil water content increases [45,53,62] or cryptogams that remain dormant in the biocrust until the next rain event [69]. However, the abrupt hydration may also cause osmotic shock that could result in massive cell lysis and the release of osmoregulatory solutes [37,39]. The period of water abundance is usually brief, and the soil quickly dehydrates, forcing microorganisms to cease their activity [62,68]. In summary, members of the biocrust community must respond quickly and efficiently not only to hydration but also to the subsequent desiccation.
Harsh desert conditions shift primary production from plants to oxygenic photosynthetic microorganisms, mostly Cyanobacteria [92]. However, biocrusts are dominated by heterotrophic microorganisms, mainly the phyla Actinobacteria and Proteobacteria [61,65]. Members of these phyla can meet their energy demands during prolonged droughts by harvesting sunlight or atmospheric trace gases [55]. Studies that focused on biocrust community shifts and cyanobacterial response to hydration-desiccation cycles were carried under controlled conditions [4,44,68]. To the best of our knowledge, these cycles were never monitored in the field to reflect the response to a rain event before, during, and after the rain. Under natural conditions, the biocrust community dynamics during the hydration-desiccation cycle may be affected by a plethora of variables, such as temperature, rain intensity, or soil structure, which could not be replicated in laboratory settings. Thus, it is imperative to elucidate the resuscitated community and its response to the gradual dehydration after a rain event in the field.
In this study, we followed the community structure and activity before, during, and after a rain event in the Negev Desert Highlands (Israel). We studied the active biocrust community by using SSU ribosomes as a proxy to active bacterial communities [83]. Although ribosomes do not quickly degrade in dormant or even dead cells [84,85], under field conditions, they present a reliable means to distinguish between active and inactive cells [5,9,83]. We hypothesized that the biocrust community would quickly respond to hydration and to desiccation. We predicted that high moisture content in the biocrust would trigger photosynthetic activity and carbohydrate production while a decrease in moisture content would lead to an inactivation of phototrophs. We further predicted that heterotroph's response to hydration-desiccation cycle would differ among phyla, as previously found for biocrust [4] and topsoil [83] collected at the same site. Specifically, we hypothesized a sharp decrease in relative abundance of the dominant Actinobacteria phylum during the rain event and subsequent increase in the Actinobacteria as the biocrust dries [5,13,83].

Site Description
The study was conducted in the long-term ecological research station in the Negev Desert Highlands site of Avdat (30°86′N, 34°46′E, Israel; Fig. 1). In this site, annual rainfall extends from October to April (https:// deims. org/ fcc28 bb3-551a-4396-819c-0589a bc6be 6f) and ranges from 20 to 180 mm (average of 90 mm). The minimum annual temperature is 2.2 °C, the maximum is 41.2 °C, and the average annual temperature is 20 °C. The soil in Avdat is mostly loess sediments in levelled Byzantian agricultural terraces cleared of rocks [17]. Vegetation cover is dominated by dwarf shrubs, mostly Haloxylon scoparium [79]. Dew and fog in the area were estimated at 0.1-0.2 mm per day [41,46] and occur year-round for approximately 200-250 days [94]. In this site, the biocrust was identified as type 2 or 3 in some places, as described by Kidron et al. [49], which means that it is a mature cyanobacterial biocrust [18,49,88].  The biocrust becomes greener after hydration and dries out quickly 1 3 of 2.07 mm. The lightest rain was 0.1 mm, and the heaviest rain was 12 mm. The rain event (5.6 mm, maximum average temperature 14.6 °C) occurred 29/01/18 (T[R]), and samples were collected until the biocrust dried (T [1], T [2], T [3]; Fig. 1). We chose this rain event to have a medium rain event -not light, nor heavy rain as both carry a challenge that would have impaired our results. As control (T[0]), biocrust samples were collected in May 2018, at the end of the wet season after a month of drought (https:// ims. data. gov. il/ ims/1). The control samples were collected after the rain season to ensure that they would be completely dry. It had rained 7.1 mm during the week prior to our chosen rain event, and therefore, the soil would be too wet to obtain an accurate representation of dry soil profile. The biocrust was sampled using a trowel, and each sample was ~ 5 mm thick. For each time point, five samples (each ~ 200 g) were collected, each at least 10 m apart (N = 25 samples). Sampling in multiple sites is not really feasible as rain in the desert is local. Each biocrust sample was homogenised through a 2-mm sieve, and then, four subsamples were stored: (1) in − 80 °C for molecular analysis, (2) in − 20 °C for chlorophyll extraction, (3) in 60 °C for 3 days and then kept at room temperature for chemical analysis, and (4) were used immediately to evaluate the water content.

Physico-chemical Analyses
Water content, total organic carbon, and total nitrogen were measured in the biocrust samples. Biocrust water content was determined by the gravimetric method. The soil was weighed before and after drying at 105 °C; then, the percentage of moisture in the soil was determined [22]. Organic carbon content was determined using the loss-on-ignition method. A portion of 30 g of dry biocrust sample was burnt at 380 °C for 6 h, and the fraction of organic carbon content was calculated as previously described [22,42]. Total nitrogen was measured in 50 mg of soil using the FlashSmart CHNS/O elemental analyser (ThermoFischer, Waltham, MA, USA). The standards used to calibrate the instrument were BBOT (2,5-Bis (5-tert-butyl-benzoxazol-2-yl) thiophene), Tocopherol Nicotinate, and a soil reference material.

Chlorophyll a Concentration
Chlorophyll a from each sample was extracted using a protocol based on Ritchie [76] and Castle et al. [24] adjusted to arid soil. The extraction was done by using methanol at a ratio of 3:9, followed by a 15-min incubation at 65 °C, and a 2-h incubation at 4 °C. The samples were then centrifuged, and the absorbance was measured in the supernatant by spectrophotometry (Infinite 200 Pro, Tecan, Männedorf, Switzerland) at 665 nm. The concentration of chlorophyll a was calculated as previously described [76]. Dried Spirulina samples were used as a positive control at a concentration of 0.003 g per g of soil, while distilled water (DW) was used as a negative control. In this study, the chlorophyll a concentrations are presented in mg chlorophyll a per g of soil (mg chla /g soil).

Carbohydrate Extraction and Measure of Polysaccharide Content
Extracellular polysaccharides (EPS), specifically the tightly bound carbohydrates that are attached to the soil particles, were extracted using a 100-mM EDTA solution and incubated at room temperature for 16 h. 20 mL of tightly bound carbohydrates were extracted from 2.5 g of soil and stored at − 20 °C until further processing. The polysaccharide content was measured using a phenol-sulfuric acid assay and estimated according to a glucose standard curve, as previously described [32]. Briefly, each EPS fraction was mixed with equal volume of 5% w/v phenol and 2.5 folds sulfuric acid. The mixture was vortexed, incubated (45 min at room temperature), and absorbance was measured at 490 nm (Tecan). The polysaccharide concentrations are presented in μg C (glucose equivalent) per g of soil (µg C/g soil).

RNA Extraction and Preparation for Sequencing
RNA was extracted from 0.5 to 1 g of soil as previously described [3]. Briefly, phenol-extracted total nucleic acids were treated with DNase (Takara, Shiga, Japan) to remove the DNA. The remaining RNA was cleaned using the MagListo RNA Extraction kit (Bioneer, Daejeon, S. Korea). To assure that the RNA is free of DNA, the ribosomes were amplified using primers of 16S rRNA encoding gene (see reaction protocol below). Ribosome samples that served as templates for amplification were cleaned again with the MagListo kit (Bioneer) and re-tested by amplification. This step was repeated till no amplicon was detected. The DNAfree RNA was reverse transcribed to cDNA using Superscript IV (ThermoFischer) at 50 °C and purified using the PCR purification kit (Bioneer) in accordance with the manufacturer's instructions. The cDNA was used as template to amplify the V3-V4 regions of the 16S rRNA using 341F and 806R primers (Table S1), in triplicates. Library preparations and sequencing were performed at the Research Resource Centre at the University of Illinois with pair end (2 × 300 bp) MiSeq platform (Illumina, San Diego, CA, USA).

Community Analysis
Reads were merged, quality checked, and trimmed following the NeatSeq-Flow pipeline [80]. The sequences were analysed using the QIIME2 platform [14] and de-noised into Amplicon Sequence Variants (ASVs) using Dada2 [19]. The denoising statistics from Dada2 can be found in Table S1. Taxonomy was assigned to the reads using Silva v138 [73]. At input, the average number of reads is 95,317 with a minimum number of 73,141 reads and a maximum number of 124,251 reads. The denoising statistics can be found in Table S2. All raw sequences used in this study can be found in BioProject (https:// www. ncbi. nlm. nih. gov/ bioproject) under the submission number PRJNA718159.

Functional Predictions
Functional predictions of the 16S amplicons were done using PICRUSt2 [7,28,31,56,93] and the KEGG database (May 2020) [43]. We selected genes in accordance with previous reports [10,61] and associated them with each step in the KEGG database to build our own database (Table S3). Metabolic pathways for different functions were selected, including energy harvesting (organotrophy, lithotrophy, and phototrophy) [26,36,54,86], nitrogen metabolism [34], and survival during stress, like DNA conservation and repair, or sporulation and reactive oxygen species (ROS-damage prevention) [15,38,40,71,72,74,75,81]. The assignment of function from our own database to the KEGG numbers of the abundance table from PICRUSt2 was done in R using phyloseq [60]. The significance of temporal differences in predicted functionalities was evaluated using a non-parametric test (Kruskal-Wallis test and a post-hoc Dunn test [30,33,51]).

Statistical Analysis
All statistical analyses were done with R (version 4.0.0, The R Core) using the phyloseq [60] along with ggplot2 [89], vegan [66], magritt [6], dplyr [91], scales [90], and grid [63] packages. The alpha diversity was measured using Shannon's index, as calculated by the "estimate-richness" function of the phyloseq package. The beta diversity was represented using a PCoA (principal coordinate analysis), based on a Bray-Curtis distance matrix. The significance of differences between time points for data that were not normally distributed were determined using a non-parametric test like Kruskal-Wallis tests and Dunn tests [30,33,51]. Normally distributed data was analysed with pairwise t-test between the timepoints.

Temporal Changes in the Biocrust Chlorophyll a, Carbohydrates, and Soil Parameters
We followed changes in the biocrust before, during, and after a rain event. We noted that a day after the rain (T[1]), the biocrust was visibly greener than at any other time point (Fig. 1). The average chlorophyll a concentrations along with the soil water content in the biocrust at each sampling point were monitored ( Fig. 2A; Table S4a and S4b). The biocrust water content was lowest after a month of drought (at T[0]) and significantly higher during the rain event T[R] (2.26% and 16.2%, respectively, p = 0.05, χ 2 = 27.6; Table S5). After the rain event, soil moisture significantly decreased to 6.22% at T[1] (p < 0.05, χ 2 = 27.6). Chlorophyll a concentrations significantly increased right after the rain event (from 8.45 to 14.57 mg chla/g soil, during the rain event, p = 0.0002, χ 2 = 8.9; Table S4a, S4b, and S5), and decreased significantly in later days (from 14.57 to 11.17 mg chla/g soil, three days after the rain, p > 0.02, χ 2 = 8.9; Table S4a, S4b, and S5). The carbohydrate concentration significantly increased after the rain event (from 83 to 143 µg/g soil, p < 0.05, F-value = 2.1; Table S4a, S4b, and S5; Fig. 2B).

Fig. 2 A Chlorophyll a content (boxplot) and water content (in %)
(dashed line and points), as well as B carbohydrate concentration (boxplot) in Negev Desert biocrusts before, during, and after a rain event. Boxes are drawn from the first to the third quartile of the data, solid lines across the box represent the median, and the tips show the minimum and maximum values excluding the outliers (1.5 times less or more than the lower or upper quantiles) represented by dots outside of the boxes. The different letters denote the differences between the time points (a-c for the chlorophyll a content, and d-e for water content) After the first day, the concentration decreased slowly until day 3, where it was significantly lower (from 143 to 72 µg/g soil, p < 0.05, F-value = 2.1; Table S4a, S4b, and S5; Fig. 2B). The total organic carbon (Fig. S1) and total nitrogen (Fig. S2) showed slight temporal changes (Table S4a and S4b) that were not significant (Table S5). Figure 3 shows the active bacterial community composition at the order level for each sampling point. The community is mostly composed of the phyla Cyanobacteria, Actinobacteriota, Bacteroidetes, and Proteobacteria in addition to six other phyla ( Fig. 3; Table S6). During the dry season (T[0]), the biocrust community composition differed significantly from the community depicted during the rain event (T[R]) ( Fig. 4; Table S7). The differences were shown mostly in orders belonging to the Actinobacteriota and Cyanobacteria phyla ( Fig. 3; p < 0.05, χ 2 = 8.9 and χ 2 = 14.3, respectively Table S7). The relative abundance of Cyanobacteria, dominated by the Cyanobacteriales, increased during the rain event (T[R]) (from 3 to 39%, Table S6; p < 0.05, χ 2 = 14.3, Table S7). While the relative abundance of the Actinobacteriota, dominated by the species IMCC26256, decreased during the rain event (T[R]) (from 37 to 0.8%; Table S6; p < 0.05, χ 2 = 8.9; Table S7).

Temporal Changes in the Active Microbial Community Composition
The biocrust water content decreased after the rain, but no major changes were detected in the biocrust community composition (Figs. 3 and 4; Table S6 and S7). This pattern was supported by the alpha and beta diversity analyses (Fig. 4). The community diversity significantly differed before the rain (T[0]), during (T[R]), and after the rain (T[1-3]) ( Fig. 4a; p < 0.05, T-value = 51.83; Table S10 and S11), but among the later timepoints (T[R, 1-3]), the diversity did not differ (p > 0.04; Table S11). Similarly, the PCoA showed that T[0] was separately clustered from the other time points (Fig. 4b; F- Table S9) and the phototrophy-related pathways (χ 2 = 209.59, p-value < 0.05; Fig. 5; Table S8 and S9). The other pathways remained unaffected by the change in hydration (p-value > 0.05; Table S8 and S9). The results suggest that the rain event enhanced the autotrophic activity of the biocrust communities but did not affect other persistent functions, such as DNA conservation and repair ( Fig. 5; Table S3).

Discussion
Biocrust bacterial communities change during hydration [4,44,87]. Most apparent was the change in relative abundance of Cyanobacteria, which increased while Actinobacteria decreased (Fig. 3). In arid soils, rain events entail a decrease in the relative abundance of Actinobacteria both in the biocrust [4] and topsoil [8,83]. Members of the Actinobacteria were shown to be well adapted to harsh environments [35,95], and abundant in the Negev Desert biocrust [61]. Similar results were obtained under controlled conditions where the biocrust was hydrated to saturation [4,82]. However, we did not detect an increase in Firmicutes (Fig. 3 after the rain event (T [1][2][3] as reported in controlled experiments of hydrated biocrusts collected from the semi-arid Moab Desert [44], even with similar dry communities as ours. Here, in addition to Cyanobacteria, we have detected an increase in the Alphaproteobacteria and Cytophagales following the rain event (Fig. 3), emphasizing the disparity between filed survey and artificial (laboratory-based) hydration.
The filamentous cyanobacterium Leptolyngbya ohadii isolated from Negev Desert biocrust was shown to respond  [67,68]. Slight increases in biocrust moisture, triggered by dew simulation, induced DNA repair and associated regulatory genes, activating the photosynthetic system of L. ohadii [62,74]. In the field, the rain event significantly increased soil moisture ( Fig. 2A). Increase in soil water content activated the Cyanobacteria (Fig. 3) and triggered functions required for resuscitation of their photosynthetic system (Fig. 5). This resulted in a sharp rise in chlorophyll a ( Fig. 2A) and carbohydrates (Fig. 2B) concentrations. The concentration of chlorophyll a was suggested to be linked to soil water content [69] that could activate the biocrust primary producers, i.e., Cyanobacteria and/or green algae [78].
Though chlorophyll a content (a proxy of potential photosynthesis activity) increased with soil moisture ( Fig. 2A), no significant changes were detected in the total organic C or N content ( Fig. S1 and S2; Table S4a, S4b, and S5). This observation could either suggest that the immediate change in these parameters is negligible compared to existing soil reservoirs, or that they were immediately consumed by heterotrophs in the biocrust [61]. Alternatively, it was recently proposed that in arid biocrusts, the members of the dominant Cyanobacteria phylum exchange C for N with copiotrophic diazotrophs, thus rapidly utilizing available nutrients to enable their colonisation in the oligotrophic dryland soils [27]. Regardless of the explanation, the lack of change in organic C and N excludes them from serving as indicators for resuscitation of arid biocrust microbial communities during rain events.
The increase of water content led to a "metabolic window" [55] for quick energy reservation following the "pulsereserve" paradigm proposed for the adaptation of plants to desert ecosystems [64]. In this framework, hydration of the biocrust increased potential activity of gene groups linked to assimilation of C and N ( Fig. 5; Table S9). We also detected an increase in potential motility of the community after hydration (Fig. 5) that could facilitate cell interactions in the soil aqueous phase [29]. While the community exploited the brief water abundance, it also prepared for the unavoidable desiccation and its associated stresses by increasing potential stress mitigation mechanisms like ROS-damage prevention and DNA repair ( Fig. 5; Table S9). In addition, the community prepared to persist during the long drought through potential activation of sporulation and DNA conservation mechanisms ( Fig. 5; Table S9). These strategies correspond to reports of sizeable fractions of spore-forming bacteria in desert biocrusts [61,65].
In the Negev Desert highlands, soil microbial communities have to quickly responded to hydration [4], but the response to desiccation may be slower despite the rapid drying of the soil surface ( Fig. 2A). Immediately after the rain, the biocrust dried up due to strong radiation,  Table S3). Time points are colour coded. The y-axis represents the abundance in copy number (CN) normalized to the 16S rRNA copy number. The different letters denote significant differences between the time points Boxes are drawn from the first to the third quartile of the data, solid lines across the box represent the median, and the tips show the minimum and maximum values excluding the outliers (1.5 times less or more than the lower or upper quantiles) represented by dots outside of the boxes. ROS reactive oxygen species evaporation, high winds, and low air humidity [47]. Unlike the rapid response to dew hydration and subsequent desiccation [67,68], the community did not immediately react to the decrease in the biocrust water content (Figs. 2 and  4). In a previous study [83], it was shown that the topsoil community bounces back to its original structure as the soil dries. In the biocrust, while desiccation was associated with slow decrease in chlorophyll a and carbohydrate concentrations (Fig. 2), it was not reflected in the composition of the active community (Figs. 3 and 4). It was reported that biocrust Cyanobacteria secrete copious amounts of EPS that bind soil particles [47,50] and retain water, slowing down the drying process [77]. Biocrust EPS may create microhabitats that retain humidity [25], thus protecting residing microorganisms from desiccation [57,59]. We hypothesize that the Negev biocrust was similarly impacted by EPS production (Fig. 2B), which may have created microhabitats that retain moisture, enabling an extended active phase of the biocrust communities following a rain event. This "bonus" active period allowed longer biosynthesis of energy resources providing access to organic molecules that may justify the Cyanobacteria's massive investment in EPS production [57] over proliferation [4]. EPS is known as a key component in the Negev Desert biocrust, maintaining its' structural integrity [50] but may also help in sustaining the activity of the inhabiting communities. However, we should mention that the carbohydrates and chlorophyll a measurements could reflect the presence of leftover compounds rather than potential photosynthetic activity or EPS production. Higher resolution measurements of active chlorophyll a and EPS production would be attempted in future studies.

Conclusions
In desert biocrusts, bacterial communities must respond quickly and efficiently to hydration and take advantage of the short window of opportunity to sequester nutrients. This fleeting water abundance requires the bacterial community to be equally adapted to the onset of desiccation to prevent cell damage. Our findings reinforce controlled studies that showed that biocrust hydration changed the active bacterial community by increasing Cyanobacteria relative abundance over Actinobacteria. Here, we have shown that the response of bacterial communities to biocrust desiccation following a rain event is slower than expected, allowing primary producers to be active even after the soil moisture decreased. This lag in response to dehydration could be associated with water retention by the newly secreted EPS, mediated by the Cyanobacteria activity surge. This grace period may justify the exhaustive metabolic cost of carbohydrates production that quickly follows rain events in the desert.