Quality of Irrigation Water Affects Soil Functionality and Bacterial Community Stability in Response to Heat Disturbance

ABSTRACT

Anthropogenic activities alter the structure and function of a bacterial community. Furthermore, bacterial communities structured by the conditions the anthropogenic activities present may consequently reduce their stability in response to an unpredicted acute disturbance. The present mesocosm-scale study exposed soil bacterial communities to different irrigation water types, including freshwater, fertilized freshwater, treated wastewater, and artificial wastewater, and evaluated their response to a disturbance caused by heat. These effectors may be considered deterministic and stochastic forces common in agricultural operations of arid and semiarid regions. Bacterial communities under conditions of high mineral and organic carbon availability (artificial wastewater) differed from the native bacterial community and showed a proteobacterial dominance. These bacterial communities had a lower resistance to the heat treatment disturbance than soils under conditions of low resource availability (high-quality treated wastewater or freshwater). The latter soil bacterial communities showed a higher abundance of operational taxonomic units (OTUs) classified as Bacilli. These results were elucidated by soil under conditions of high resource availability, which lost higher degrees of functional potential and had a greater bacterial community composition change. However, the functional resilience, after the disturbance ended, was higher under a condition of high resource availability despite the bacterial community composition shift and the decrease in species richness. The functional resilience was directly connected to the high growth rates of certain Bacteroidetes and proteobacterial groups. A high stability was found in samples that supported the coexistence of both resistant OTUs and fast-growing OTUs.

IMPORTANCE This report presents the results of a study employing a hypothesis-based experimental approach to reveal the forces involved in determining the stability of a soil bacterial community to disturbance. The resultant postdisturbance bacterial community composition dynamics and functionality were analyzed. The paper demonstrates the relatedness of community structure and stability under cultivation conditions prevalent in an arid area under irrigation with water of different qualities. The use of common agricultural practices to demonstrate these features has not been described before. The combination of a fundamental theoretical issue in ecology with common and concerning disturbances caused by agricultural practice makes this study unique. Furthermore, the results of the present study have applicable importance regarding soil conservation, as it enables a better characterization and monitoring of stressed soil bacterial communities and possible intervention to reduce the stress. It will also be of valued interest in coming years, as fresh water scarcity and the use of alternative water sources are expected to rise globally.

INTRODUCTION

Bacterial communities in soil are affected by many anthropogenic inputs, including fertilization, organic carbon amendment, irrigation water quality, and the irrigation regime (1–10). The effects of these inputs on the bacterial community are considered deterministic forces (11), which are also referred to as “niche based” or “habitat filters” in the literature (12). Bacterial communities may change in their size, activity level, and composition depending on the given ecological selection impacting them. These types of changes were documented for treated wastewater-irrigated soil bacterial communities (2, 6, 7); thus, the incentive for this study was to measure the resulting stability of the bacterial community after irrigation with water of different qualities. Although changes in soil bacterial community composition due to anthropogenic inputs have been documented in many studies, it is difficult to determine whether these changes have a long-term deleterious effect on the ability of soil bacterial communities to provide ecological services. The addition of a substrate, such as in the case of treated wastewater irrigation, will encourage the proliferation or activity of organisms that utilize the given substrate, followed by a reduction in the availability of the resource. However, it is unclear whether the specific enrichment imposed by the substrates on the bacterial community may change a given bacterial community in a way that is not suitable for their habitat (on a temporal and physical large scale). These changes may further jeopardize the biological production and the ecological services provided by the soil environment.

In addition to anthropogenic inputs, soil bacterial communities may be interrupted by disturbances that are usually short-term, often intense, events (13). These random unpredicted changes to the soil environment, such as by certain types of pollution, fluctuating redox states, drying and rewetting, and freezing and heating, have disturbing effects on the microbiome and soil homeostasis and thus may be defined as stochastic (11, 14). These disturbances may reduce the soil microbiome function, size, and diversity or change the bacterial community composition (13, 15–17). Bacterial communities that better withstand a disturbance, by either retaining their characteristics in the face of the disturbance (higher resistance) or regaining their undisturbed form quickly after the disturbance ends (high resilience) (18), are more stable, contribute better to biological production (13, 15, 17), and are beneficial in agricultural operations. Therefore, understanding the effect of soil management as a force structuring a bacterial community's resistance and resilience to natural events or disturbances occurring as part of the management strategy is important for assessing the sustainability of a given management.

Irrigation with water of various qualities, such as freshwater, fertilized freshwater, treated wastewater, and wastewater, is a major soil management practice worldwide. This management practice was studied previously, and results show that both the application of mineral fertilizer (4, 5) and the use of wastewater of different treatment levels (6, 19–21) drive bacterial communities to an alternative stable state (according to the terminology used by Shade et al. [13]). This alternative state may consist of fewer operational taxonomic units (OTUs) of higher dominance and lower richness dominated by populations that are in an unusual (relative to natural conditions) but stable high abundance. The populations promoted by these practices are also possibly less adapted to cope with natural disturbance events such as intense heat and desiccation, which are common in arid and semiarid regions and expected to increase in frequency as a result of global climate changes (12, 22). Intense heat is also used as a means for pest control, as in the case of soil solarization (23), which is an important agricultural application with increasing popularity. Both irrigation as a deterministic force structuring the bacterial community and the disturbance (heat and desiccation) represent soil management methods in agriculture as well as common events in vast land areas.

The hypothesis, as suggested above, that bacterial communities structured by deterministic forces characterized by a high resource availability will be less stable in the face of a disturbance is based on the theory suggesting a trade-off between fast growth/resource utilization and survivorship in the face of stress (24). Several terms are used to classify bacterial life strategies (25), while the terms r- and K-selected organisms as used in previous studies (25–27) are most suited for this study. r-selected organisms are “adapted to maximize their intrinsic rate of growth when resources are abundant and K-selected organisms, on the other hand, are adapted to compete and survive when populations are near carrying capacity and resources are limited” (26, 27). These definitions are not directly related to how an organism will react to a disturbance but suggest the general behavior. For instance, an advantage might be given to a K-selected organism during a stress event on the basis of the definition. On the contrary, transient conditions of high resources will favor the r-selected organisms on the basis of their fast growth ability. In accordance with the hypothesis, the objective of this study is to understand whether a soil bacterial community structured by various amounts of organic and mineral resources will change its stability in response to a disturbance caused by heat stress.

RESULTS

Resistance.

The resistance of soil bacterial communities to different levels of heat treatment was evaluated by measuring the soil community fluorescein diacetate (FDA) hydrolysis rate and the bacterial community fingerprint. The potential hydrolytic activity of the soil increased proportionally with the amounts of total organic carbon (TOC) and total nitrogen (TN) introduced by irrigation water. Furthermore, the reduction in activity rates after heat stress of various intensities was also related to the irrigation water type (Fig. 1). While there was no significant reduction in the activity rate of soil irrigated with fertilized tap water (FW), activity rates in soils irrigated with fertilized treated wastewater (TW) and artificial wastewater (AW) were significantly reduced by the heat stress, validated by an analysis of variance (ANOVA) test (F_[4,15] = 17. 6; P < 0.05 and F_[4,16] = 7.16; P < 0.05, respectively). Samples irrigated with TW showed a significant reduction only when exposed to 65°C (1.18 ± 0.39 mg · kg⁻¹ · h⁻¹; P < 0.05) or 80°C (1.5 ± 0.33 mg · kg⁻¹ · h⁻¹; P < 0.05) compared to the activity rate measured in control soil (3.24 ± 0.56 mg · kg⁻¹ · h⁻¹). Samples irrigated with AW showed a marginally significant reduction already after exposure to 35°C (4.65 ± 1.55 mg · kg⁻¹ · h⁻¹; P < 0.08) and a significant reduction after exposure to 50°C (4.22 ± 1.43 mg · kg⁻¹ · h⁻¹; P < 0.05) compared to the control soil activity rate (8.27 ± 2.84 mg · kg⁻¹ · h⁻¹). As expected, these soils also showed significantly reduced activity rates after exposure to 65°C (2.21 ± 0.33 mg · kg⁻¹ · h⁻¹; P < 0.05) or 80°C (3.5 ± 1.01 mg · kg⁻¹ · h⁻¹; P < 0.05). A PCR denaturing gradient gel electrophoresis (DGGE) fingerprint analysis of soil bacterial communities showed that the active (based on 16S rRNA amplification) soil bacterial community was significantly affected by heat stress treatment (Fig. 2). All control samples clustered closely, with mild differences between the AW pattern and those of FW and TW, suggesting AW had a stronger selection influence on the bacterial community. After a mild heat stress of 35°C or 50°C, FW and TW irrigated samples clustered separately from the control samples, while samples exposed to either 65°C or 80°C clustered separately. On the other hand, AW samples exposed to the mild 35°C and 50°C stresses clustered closely to all other samples exposed to the high stress of 65°C. Samples exposed to 65°C clustered together, with the exception of a single TW sample. All samples exposed to the extreme stress of 80°C clustered separately from all other stressed samples, regardless of the irrigation water type.

FIG 1.

FDA activity in soils after exposure to various heat stress intensities. Treatments marked by the lowercase letter a are significantly different (P < 0.05) from unmarked groups within the same irrigation water treatment, whereas samples marked with an asterisk (*) showed marginal significance (P < 0.08) from unmarked groups on the basis of an ANOVA test. Bars represent the standard deviations. FW, TW, and AW represent freshwater with additional minerals (N, P, and K), treated wastewater with additional minerals, and artificial wastewater, respectively.

FIG 2.

Denaturizing gradient gel electrophoresis (DGGE) fingerprint of communities irrigated with the three water types without stress (indicated as control) and after exposure to various heat stress intensities (35, 50, 65, and 80°C). Control treatment samples were analyzed in replicates of four, 80°C treatment was analyzed in three replicates, and 35, 50, and 65°C treatments were analyzed in duplicates; replicate numbers are indicated in the figure. FW, TW, and AW represent freshwater with additional minerals (N, P, and K), treated wastewater with additional minerals, and artificial wastewater, respectively.

Resilience.

The irrigation type, representing different intensities of selection in the soil environment, caused changes in the total (16S rRNA genes) and active (16S rRNA) bacterial community compositions at day 0 relative to the native soil composition (T₋₄₂, unaffected by the irrigation treatment) (Fig. 3). These comparisons showed that all water treatments modified the soil native bacterial community, both total (R² = 0.31, P = 0.028) and active (R² = 0.44, P = 0.001) compositions, while AW irrigation had a stronger effect. As the nonmetric multidimensional scaling (NMDS) suggested (Fig. 3), irrigation water influenced the active bacterial community composition of soils (see Fig. S3 in the supplemental material) by increasing and reducing the populations' abundances. Samples irrigated with tap water (W) were characterized by very dominant Bacilli groups, while AW samples were highly abundant with proteobacterial groups. FW- and TW-irrigated soils showed intermediate abundance values (relative to W and AW extremes) of Bacilli and Proteobacteria and relatively high abundances of Flavobacteriia. Samples at day 0 irrigated with AW had a lower OTU richness in the total bacterial community than samples irrigated with W, FW, and TW (F_[3,11] = 3.9, P = 0.05) (see Fig. S4). The active bacterial communities in all samples showed lower OTU richness than that found in the total bacterial community, while soil irrigated with W and AW had the lowest measured OTU richness (F_[3,11] = 8.3, P < 0.05). The lower richness in AW-irrigated soils was coupled with a significantly higher Bray-Curtis similarity between the total bacterial community and the active bacterial community than that of samples irrigated with W and TW (F_[3,11] = 5.05, P < 0.05) (Fig. S4).

FIG 3.

Total (16S rRNA gene) and active (16S rRNA) community bacterial compositions at day 0 (before applying heat stress) and at T₋₄₂ (before irrigation started). W, freshwater; FW, freshwater with additional minerals (N, P, and K); TW, treated wastewater with additional minerals (N, P, and K); AW, artificial wastewater. The red crosses represent the different OTUs. NMDS stress = 0.065.

With a reliance on the results of the resistance experiment, we chose to expose soil samples to 65°C heat disturbance in the resilience experiment. Despite the fact that this temperature exceeds the normal exposure of most soil habitats, all soil bacterial communities were affected similarly at this temperature (Fig. 1). After the exposure, all soil samples showed decreased activities relative to their undisturbed controls. Furthermore, the potential activity measured by FDA hydrolysis and dehydrogenase (DEH) and nitrification potentials indicated that the samples did not regain the baseline activity potential by the end of the experiment (day 21). The recovery of soil functionality was assessed using a resilience (RL) index which is bounded by +1 and −1. The value 1 at the time of measurement indicates full recovery, while lower values indicate a lower rate of recovery (16). The results of the DEH RL index (Fig. 4A) analyzed using the Kruskal-Wallis test showed that AW-irrigated soils recuperated to a higher extent than W-, FW-, and TW-irrigated soils at day 4 (χ² = 20.5, P < 0.01), day 9 (χ² = 40.2, P < 0.01), and day 21 (χ² = 19.3, P < 0.01). On the other hand, the FDA RL index (Fig. 4B) showed the same trend only at day 4 (χ² = 30.9, P < 0.01), while at day 9 and 21 there were no significant differences between the water treatments. The lack of change in the FDA RL index was due to higher activity levels of AW and TW stressed samples than the undisturbed control, whereas W and FW showed lower activity levels than their undisturbed controls (data not shown). The RL index does not discriminate between these two cases, as both indicate that the disturbance is still ongoing. The nitrification potential RL index was low in all treatments due to activity rates that were higher than the baseline measurement (Fig. 4C). Furthermore, due to the high variance between replicates, no significant differences were detected using this method. Although significant differences were not achieved, the mean values from both TW and AW treatments were higher than those from W- and FW-irrigated soils, in agreement with results from the DEH assays and to some extent those of the FDA assays.

FIG 4.

Recovery of community activity after 65°C stress. (A) RL indices from a dehydrogenase (DEH) assay. DEH indicates oxidative potential in soil irrigated with different water types (W, FW, TW, and AW). (B) RL indices from a fluorescein diacetate (FDA) hydrolysis assay. FDA hydrolysis indicates hydrolytic enzyme activity in soil irrigated with different water types. (C) RL indices of nitrification potential at day 21. W, FW, TW, and AW represent freshwater, freshwater with additional minerals (N, P, and K), treated wastewater with additional minerals (N, P, and K), and artificial wastewater, respectively.

Throughout the experiment, both the total (based on DNA sequencing of rRNA genes) and the active (based on cDNA sequencing of rRNA) bacterial communities were analyzed. However, the dynamics (change of a population's relative abundance) were better represented by the active bacterial communities and will therefore be the focus here. Stress (R² = 0.23, P = 0.01), time from stress (R² = 0.1, P = 0.01), and irrigation (R² = 0.09, P = 0.01) were all factors that showed significant influences on the active bacterial community, validated by a permutational multivariate analysis of variance (PERMANOVA) test and presented by the NMDS ordination (Fig. 5). The irrigation water type determined the bacterial community composition in the control samples (significant at days 4, 9, and 21) (see Table S1). The samples subjected to heat stress were more dynamic (change in composition over time), and their compositions were determined by both the time (days) after the stress and the irrigation type. Immediately after the heat stress and at day 4 there was no influence of the irrigation water, while at days 9 and 21 of the stress, irrigation played a significant role in determining the bacterial community composition (Table S1). Heat stress played a significant role in determining bacterial community composition, and W-, FW-, and AW-stressed samples were significantly different from their controls at both early (days 0 and 4) and late (days 9 and 21) periods. On the other hand, TW-irrigated soils did not show significant differences between stressed and control samples (see Table S2).

FIG 5.

NMDS of active bacterial community compositions throughout the experiment based on Bray-Curtis similarities of samples calculated by Hellinger-transformed 16S rRNA gene distributions. The plot illustrates the distribution of samples on axis 1 (horizontal) and axis 2 (vertical). The symbol shapes indicate the four irrigation types, and the colors indicate the stress and time. Class relative abundance data and environmental factors that significantly correlated with axes are plotted as vectors that indicate the direction and strength (R²) of the correlations. NMDS stress = 0.078.

The correlations of environmental variables and class-relative abundances to the axes of the NMDS, viewed by the ordination, may indicate which forces drove the bacterial community shift. DEH activity correlated with control samples, as the heat stress significantly reduced the soil community's functional capabilities. TN correlated with all stressed samples, and TOC highly correlated with samples at day 0 after stress, indicating the exposure to heat stress resulted in the release of soluble organic materials and minerals to the environment. Furthermore, these results show that the released TOC was rapidly consumed, while the nitrogen released was either less available or consumed more slowly. To illustrate the populations' dynamics, we used both the correlations between the classes' relative abundance and the NMDS axes and the abundances of these classes relative to that in the unstressed control soil (see Fig. S5 to S8). These two descriptive strategies were used to reveal specific population dynamics following the stress and to characterize their behavior. Bacilli and Clostridia are two classes that correlated with the early (days 0 and 4) stressed samples (positive on axis 1 and negative on axis 2) (Fig. 5). These two groups were termed resistant (see Fig. S5), as their relative abundances increased immediately after the stress relative to those of the unstressed controls. As the bacterial community developed, different bacterial classes were shown to be recuperating. The abundances of classes Gammaproteobacteria, Cytophagia, Saprospiria, and Sphingobacteriia were all correlated with the late stressed samples (positive on axes 1 and 2). These classes were termed resilient and opportunistic, as all of them were negatively influenced by the stress at day 0 or 4 but exceeded the abundances relative to those in the controls by day 21 (see Fig. S6). Alphaproteobacteria, Betaproteobacteria, and Deltaproteobacteria, as well as Acidimicrobiia, were termed resilient classes, as they were negatively affected by the stress at day 0 but returned to their undisturbed relative abundance (in at least half the treatments) by day 21 (see Fig. S7). These classes were correlated negatively on axis 1 and positively on axis 2. The bacterial classes that were sensitive to the heat stress also correlated negatively on axis 1 and positively on axis 2 (Fig. 5). These classes included the Actinobacteria, Thermoleophilia, Planctomycetia, and Anaerolineae and did not return to their original abundances by the end of the experiment (see Fig. S8). Two classes showed diverse responses to the stress, namely, Pedosphaerae and Solibacteres (see Fig. S9). Although the various groups showed high recuperation, it is evident that the major growth/activity of soil bacteria in FW-, TW-, and AW-irrigated soils after the heat stress is related to the proteobacterial members and in W, to Bacilli members, as these were the dominant classes in these samples (Fig. 6).

FIG 6.

Active community compositions of soil samples after heat stress, described by class level relative abundance. Samples at day 0 (12 h after the stress), day 4, day 9, and day 21 under different irrigation water qualities. The data are based on sequences classified using the Greengenes database. W, FW, TW, and AW represent freshwater, freshwater with additional minerals (N, P, and K), treated wastewater with additional minerals (N, P, and K), and artificial wastewater, respectively. Groups with average relative abundances lower than 0.001 are not shown in the analysis.

DISCUSSION

It is hypothesized that bacterial communities under increasing mineral and organic matter input will have a lower stability in response to a disturbance, both in the resistance to the stress as well as in their resilience after the stress is over. The hypothesis is based on two secondary hypotheses: the first is that bacterial communities with the least intervention will have higher species richness, and the second, that bacterial communities receiving less input will have a large fraction of the community in an inactive or dormant state. Previous findings show that bacterial communities with a high diversity and richness utilize available resources more efficiently (28) and cope better in fluctuating environments (28–30). Furthermore, the dormant community is described as the bacterial “seed bank” that increases the stability of the community (31). A larger inactive fraction of the bacterial community results in a larger fraction surviving in the case of a disaster, as inactive bacteria are relatively unsusceptible to stress.

In this study, the soil bacterial community resistance was indeed higher following low inputs of minerals and organic matter. However, the communities under a high input level (AW, according to the difference from time zero as shown in Fig. 3) had a higher functional resilience after the disturbance ended, which contradicts the second initial hypothesis. The high organic and mineral load introduced to the AW-irrigated soil enriched it with a high proportion of active populations. This presumably r selection lifestyle-dominated community, which consists of bacterial, archaeal, and eukaryotic organisms, had a significantly higher activity rate but a lower resistance, evident by the fingerprinting results of only the bacterial fraction. The soil r-selected populations outnumbered the soil K selection lifestyle bacteria that have lower growth rates but a possibly higher stress resistance. The dominance of bacteria with a low stress resistance resulted in a strong shift of the community composition immediately after the heat stress ended. However, 4 days after the stress ended, fast-growing groups, which have low competitiveness in an oligotrophic environment, started to dominate the bacterial community and restore the soil functional capabilities. These fast-growing groups used the same substrate source that initially reduced the community resistance. The external substrate was also supplemented with the dead microbial biomass evident by the TOC values (presented as correlations in Fig. 5). On the basis of the sequencing results (Fig. 5; see also Fig. S5 to S9 in the supplemental material), these r-selected populations include some Bacteroidetes and mostly Proteobacteria OTUs. The occurrence of various proteobacterial classes in high mineral and organic matter soils was described in previous studies (3, 6, 32, 33).

Soil communities maintain functional stability even with decreased bacterial species richness, due to the high functional redundancy this environment holds (34) but possibly to the presence of archaeal and eukaryotic-resistant organisms. In this study, we used FDA hydrolysis and DEH assays that determine the activity of highly redundant functions (35, 36), as well as potential nitrification, which is a phylogenetically more narrowly distributed trait of soil organisms (36). FDA hydrolysis indicates the potential activity of intracellular and extracellular enzymes. The DEH assay measures the intracellular reduction of TTC under anaerobic conditions as an indicator of the microbiological redox system (37) and has a good correlation with the soil community size (35). On the other hand, the potential nitrification rate is an indicator of the sizes of the bacterial and archaeal populations that readily oxidize ammonia. The results of FDA hydrolysis and nitrification potential assays showed (using an RL index) that one of the disturbance effects was the overactivity (relative to unstressed controls) of the TW- and AW-irrigated soils. The results from the DEH assay further show that not only the extracellular activity but also the intracellular activity was affected, which might indicate a higher growth rate in AW-irrigated samples. Most of the 16S rRNA found in AW-irrigated soils was annotated as Proteobacteria (Fig. 6), indicating a high activity but possibly also a high growth rate. It is possible that the activity of organisms that were not detected in the analysis was involved as well; however, the focus here is on the bacterial fraction.

r-selected bacteria are adapted to maximize their intrinsic growth rates when resources are abundant (27). Many members, though not all, of the large bacterial Proteobacteria phylum are considered as such. Proteobacteria were abundant in soils with high levels of organic carbon of various sources, including those irrigated with treated wastewater (1, 6, 33). In the present study, we found indications that members of this group react quickly to the mineral and organic inputs by growing and as a result, lower the soil's resistance to heat stress. On the other hand, we found that members of this group were positively correlated with samples with a high functional resilience, possibly also due to their fast growth. In contrast, organisms from the phylum Firmicutes, specifically from the classes Bacilli and Clostridia, were found in high relative abundances in FW and W samples and showed very high survivorship after the heat stress. Firmicutes showed a high abundance and potential activity in the early stress samples, indicating their importance to the functional resistance of the soil environment. Other soil groups, such as members of the Actinobacteria phylum that were abundant in W, FW, and TW controls or Flavobacteriia that were abundant in FW and TW controls, were sensitive to the heat stress. These results suggest that Actinobacteria spores have lower survivorship than the Firmicutes, as previously described for Streptomyces (38). This also shows that they do not have the same growth rate capacities as Proteobacteria as previously proposed by Van der Voort et al. (39). These groups may have importance in the stability of soil communities in response to modest stresses; however, they did not withstand the 65°C heat stress applied in the resilience experiment.

Spore formation and dormancy (in its general term) are key characteristics contributing to the soil resistance (31), as demonstrated by Firmicutes in this study. Also, fast growth and high functional capabilities are essential for maintaining resilience, as in the case of Proteobacteria. This demonstrates the importance for the coexistence of different bacteria with either trait (dormancy and resistance or fast growth) in order to maintain stability. This concurs with the intermediate disturbance hypothesis (IDH), assuming that there is a trade-off between the ability to compete and the ability to withstand a disturbance. The IDH states that intermediate levels of disturbance result in a higher biodiversity level with the coexistence of organisms having different life strategies (i.e., r and K selection), which ensure ecosystem stability (24). Here, we demonstrate that AW has a negative effect on the soil resistance but increases its functional resilience. TW, containing less degradable and more diverse organic matter but large amounts of nitrogen, did not result in the same trend. Furthermore, the active bacterial community of heat-stressed TW-irrigated soil did not have a significant difference from that of undisturbed control bacterial communities. These results indicate that small amounts of diverse organic material may contribute to the soil bacterial community stability. In this study, it appears that TW may be considered an intermediate selection force compared to the other treatments and contributes to the community stability.

MATERIALS AND METHODS

Experimental design.

Two separate experiments were conducted to estimate the resistance and resilience of soil bacterial communities to heat disturbance under different irrigation water types. Both experiments were conducted at a mesocosm scale in a greenhouse semicontrolled environment. Plastic containers (100 ml) with water drainage were filled with 80 g sieved (2-mm pore size) sandy loam soil (81% sand, 6% silt, and 13% clay), originating from the West Negev region in Israel (31°21′N, 34°27′E). Mesocosms were manually irrigated, once a day, with 25 ml water of the four types; tap water ([W] used only in the second experiment), fertilized tap water (FW), fertilized treated wastewater (TW), and artificial wastewater (AW). W, FW, and AW were supplemented with 200 mg/liter NaCl to complement for the salinity found in the TW, obtained from the Shafdan wastewater treatment plant (http://igudan.org.il/english_site/wastewater_treatment.html, accessed 26 May 2016) after secondary treatment. N, P, and K were added as NH₄Cl (15.6 mg/liter), H₂PO₄ (1.4 mg/liter), and KCl (45 mg/liter) to FW, TW, and AW. AW was created using tap water and standard chemicals as described previously (40). The respective TOC and TN concentrations in irrigation water were 0.6 and 0.4 mg/liter in W, 0.8 and 4 mg/liter in FW, 9 and 5 mg/liter in TW, and 26 mg/liter and 8 mg/liter in AW.

To determine soil resistance, 3 weeks of irrigation was applied using three water types (FW, TW, and AW) until an alternative stable state bacterial community developed (verified using cultivation-independent molecular analyses of bacterial community structure [i.e., PCR DGGE fingerprinting]). After the irrigation period, the soil samples were subjected to 12 h of various heat stress intensities: control (no heat; 21 to 28°C), 35, 50, 65, and 80°C (see Fig. S1 in the supplemental material). The samples were then weighed, and deionized water was added to samples according to their estimated weight loss compared to that of the control samples. A 2-h period was given after the wetting of all samples to regain potential activity before their sampling (41). The soil samples were then analyzed for hydrolysis activity by fluorescein diacetate (FDA) breakdown (described below), and aliquots for nucleic acid purification (RNA and DNA) were flash frozen in liquid nitrogen and stored at −80°C. For the DGGE analysis, all four replicates of the control samples were used; 3 replicates of the 80°C stressed samples and duplicates from the remaining stress treatments were analyzed, due to gel size considerations.

To determine soil bacterial resilience, a second experiment was conducted. Soil samples were irrigated for 1 week with W to achieve a homogenous starting point; this starting point is termed T₋₄₂ (i.e., 42 days before the stress) (see Fig. S2). Samples were then irrigated with the four water types (W, FW, TW, and AW) for 6 weeks before a 12-h heat stress of 65°C was applied; this time point is termed day 0. At day 0, the soil was sampled 1 h before the heat stress and 2 h after it ended (as described for the resistance experiment). The mesocosms were further irrigated after the heat stress for 3 weeks, and sampling was performed at days 4, 9, and 21 after the stress. For all irrigation treatments, at each time point, samples that went through the stress procedure as well as no-stress controls were analyzed. In this experiment, biochemical assays (dehydrogenase oxidative potential and FDA hydrolysis), soil chemical analysis of TOC, TN electric conductivity (EC), and pH were analyzed using four replicates, while for the sequencing, only three were used.

Soil biochemical assays and chemical analysis.

Two soil biochemical assays were selected, as they correlate well with soil microbial community size (35). The methods were described previously by Elifantz et al. (2). Briefly, the soil oxidative potential was estimated by measuring the DEH activity (42) using 2,3,5-triphenyl tetrazolium chloride (TTC) as the substrate. Hydrolytic activity was measured by the fluorescein diacetate (FDA) hydrolysis assay as described by Schnürer and Rosswall (43). The soil nitrification potential was also measured using methods previously described at Elifantz et al. (2). The soil dissolved organic carbon and total nitrogen contents as well as salinity, electric conductivity (EC), and pH were measured as previously described by Frenk et al. (7). Biochemical activities were normalized to the rate (mg · hour⁻¹ · kg⁻¹ dry soil) and the soil chemical properties were normalized to dry soil weight, based on 24-h drying at 105°C.

Nucleic acid extraction.

RNA and DNA extractions were performed using a method previously described by Angel et al. (44). Total nucleic acids were extracted by disrupting 0.5 g of soil by bead beating in a Fastprep unit. The extraction solution included phosphate buffer, 10% hexadecyl-trimethyl-ammonium bromide (CTAB; Sigma-Aldrich, MO, USA), and water-saturated phenol (pH 7.8). The extract was then purified using a standard phenol-chloroform-isoamyl alcohol solution (25:24:1) followed by chloroform-isoamyl alcohol (24:1) purification. The nucleic acids were precipitated using 30% polyethylene glycol 8000 (Fluka [analytical grade]; Sigma-Aldrich, MO, USA) and 0.02 mg of glycogen (Fermentas, Vilnius, Lithuania), washed once with ice-cold 85% ethanol (diluted with diethyl pyrocarbonate [DEPC]-treated water), and resuspended in Tris-EDTA (TE) buffer. The extracted nucleic acids were then purified using a OneStep PCR inhibitor removal kit (Zymo Research, Irvine, CA, USA) and treated (for RNA only) with a DNase 1 DNA degradation kit (Sigma-Aldrich, St. Louis, MO, USA). The RNA was reverse transcribed using an ImProm-II kit (Promega, Madison, WI, USA) with random hexamer primers (Promega).

DGGE.

PCRs and denaturing gradient gel electrophoresis (DGGE) analysis were conducted as described previously (45). PCR was performed using a primer set for a general bacterial community, namely, 341F-GC containing a 40-bp GC clamp to enhance separation in DGGE, CGCCCGCCGCGCCCCGCGCCCGTCCCGCCGCCCCCGCCCGCCTACGGGAGGCAGCAG (46), and 907R, CCGTCAATTCMTTTGAGTTT (46). Gel gradients ranged from 20% to 70% urea and formamide (a 100% denaturant corresponds to 7 M urea and 40% [vol/vol] formamide) in TAE running buffer (2 M Tris base, 1 M glacial acetic acid, 50 mM EDTA). Gels were stained with 0.1 μl/ml GelStar (Cambrex Bio Science, Rockland, ME, USA) and photographed using a UV transillumination table (302 nm) with a MiniBIS-Pro digital camera (DNR Bio-Imaging Systems Ltd., Israel). The resulting gel images were subjected to computational analysis using Fingerprinting II software (Bio-Rad Laboratories, Hercules, CA, USA). The normalized banding patterns, based on a densitometric curve for each lane, were used to generate a dendrogram by calculating the Pearson product moment correlation coefficient (47) and by unweighted pair group method with arithmetic averages (UPGMA) clustering (48). The outcome of the analysis is composed of two different gel runs, which we combined together by using a control DNA band pattern every four lanes.

Amplicon sequencing.

Genomic DNA and cDNA were amplified and sequenced using the primers 515F and 806R, recommended by the Earth Microbiome Project (49, 50). The V4 region of the 16S rRNA gene was amplified with domain-level primers that included the Fluidigm linker sequences, CS1 and CS2 (51). The primer set CS1_515F (ACACTGACGACATGGTTCTACAGTGCCAGCMGCCGCGGTAA) and CS2_806R (TACGGTAGCAGAGACTTGGTCTGGACTACHVGGGTWTCTAAT) was used. Subsequently, the amplicons were quantified, normalized, and then sequenced in a MiSeq (Illumina, San Diego, CA, USA) run according to procedures described by Caporaso et al. (49).

Sequence processing and analysis.

The ∼15 million sequences were prepared using the MOTHUR package version 1.33.3 according to the standard operating procedure (SOP) described on the software package website (52). Briefly, the sequences were first matched for contigs and aligned to the Silva-based reference alignment database provided on the software website. Unique sequences were clustered, and the chimeric sequences were removed using the Uchime algorithm (53). The sequences were classified using the greengenes database reference. Archaeal, eukaryotic, mitochondrial, chloroplast, and unknown sequences were removed from the analysis. The sequences were randomly subsampled to 10,000 sequences per sample and assigned to an OTU based on 97% similarity, after which rare OTUs (<10 appearances) were removed from further analyses.

Statistical analysis.

All data frames were subjected to a Shapiro-Wilk test for normal distribution (54) using R 3.1.3 (55). To measure the communities' resistances, a one-way analysis of variance (ANOVA) with Tukey's post hoc test for multiple comparisons was performed on FDA activity using SPSS statistical software for Windows (SPSS Inc., Chicago, IL, USA). The null hypothesis of this test was that the soil community is resistant. A significant difference between the control mean FDA rates and the rate following the disturbance results in a rejection of the null hypothesis. On the other hand, functional resilience was calculated using a resilience (RL) index, as the comparison needed to account for both predisturbed differences within irrigation types and differences to the control at a measurement time after the disturbance. RL index was calculated as RL(t_x) = 2∣C₀ − D₀∣/(∣C₀ − D₀∣ + ∣C_x − D_x∣) − 1 (16, 56), where C₀ is the control at time zero, D₀ is the disturbed sample at time zero, and C_x and D_x are the control and disturbed soils at time x, respectively. The RL indices for FDA and DEH were normalized to the activity rate in the W control at each time point. The average RL value was calculated from the comparison of all control samples to all stressed samples for each time point. The difference between irrigation treatments at each time point was validated using a Kruskal-Wallis one-way analysis of variance (57) along with Dunn's test for multiple comparisons (58) and Benjamini-Hochberg correction for false discovery rate (59) using R with the “dunn.test” package.

The similarities between active and dormant bacterial communities to quantify dormancy were determined by an inverted Bray-Curtis dissimilarity measure for community distances (1 − distance). Differences were verified with an ANOVA with a Tukey's post hoc test using SPSS.

Multivariate analyses were carried out using the Vegan package (60) in R. The ordination plots were calculated for the OTU abundance data. The data were transformed using a Hellinger transformation, and a Bray-Curtis distance matrix was generated from which NMDS plots were calculated. Environmental features, including data from TOC, TN, EC, pH, and enzymatic activity assays, as well as the classes' relative abundances, were correlated to the ordination axes using the vector fitting regression Vegan envfit function. This function plots arrows on the ordinates, where the length represents the strength of the correlation (R²) and the direction of the maximal correlation between the NMDS configurations. The environmental features that yielded significant correlations (Benjamini-Hochberg corrected) and which bacterial classes were significantly correlated with the axes and were highly ranked (at least 1% in at least 4 samples) were plotted. These classes were also analyzed separately for their post-stress succession and were tested for differences using the Kruskal-Wallis test followed by a Dunn's multiple comparison and a Benjamini-Hochberg correction.

A statistical comparison of the bacterial community compositions was done using a nonparametric analysis of variance (ANOVA) (61), also known as a PERMANOVA, based on the Bray-Curtis distance matrix (999 permutations). The test was performed using time, irrigation type, and stress as grouping variables. Furthermore, the samples were grouped by day and irrigation and the data set was separated by stress to determine if the irrigation type was a differentiating factor. Finally, a test was performed to determine if the controls for a water treatment in the early (day 0 united with day 4) and late (day 9 united with day 21) periods were different than the stressed samples.

Accession number(s).

Raw sequence data were submitted to the sequence read archive (SRA) under accession number SRP058583.