Natural communities are open systems that are influenced by local processes (e.g., competition, niche differentiation, or predation), by regional processes (e.g., dispersal from/to a locality), as well as by historical factors (e.g., speciation or colonization events) (Cornell, 1999; Hillebrand and Blenckner, 2002). These processes determine the diversity and composition of communities and potentially their response to environmental change and disturbances. Communities can respond differently to disturbances depending on the type, intensity and frequency of the disturbance as well as on the capacity of the different species to tolerate them (Sousa, 1984). This relates to central concepts of stability including (i) resistance, which is the ability of a community to withstand a disturbance, (ii) recovery, which is the ability of a response variable to return to the initial state, and (iii) resilience (more specifically engineering resilience), which is the rate of recovery of a community to pre-disturbance conditions (Pimm, 1984). During the last decade, an increasing number of studies have also looked at effects of disturbances on bacterial community composition (BCC) and functioning in diverse set-ups, ranging from experiments at the lab-scale to field studies, investigating changes along environmental gradients. The consensus of this work that has been summarized by Allison and Martiny (2008) and Griffiths and Philippot (2013) for soils and more generally by Shade et al. (2012) is that bacterial communities are in most cases not resistant to disturbances. Moreover, compositional changes concur with functional changes in some cases, but not in others, indicating that compositional and functional resistance and resilience are not necessarily linked. For example, functional redundancy among bacterial taxa can result in a situation where composition changes in response to the disturbance whereas functioning remains stable. Whether or not functional redundancy is found will also be influenced by the type of function that is measured (Langenheder et al., 2006; Peter et al., 2011) and the strength and type of the disturbance that is applied (Langenheder et al., 2005; Berga et al., 2012; Comte et al., 2013). Finally, disturbances may also reduce species richness (alpha-diversity) of bacterial communities, which might have negative consequences for the magnitude and stability of ecosystem processes. There is, however, contrasting evidence if this is actually the case (Peter et al., 2011; Delgado-Baquerizo et al., 2016; Roger et al., 2016).

Generally spoken, disturbance strength (intensity) is a key feature that will determine how communities of macro- as well as microorganisms respond to the disturbance (Sousa, 1984; McCabe and Gotelli, 2000; Berga et al., 2012; Shade et al., 2012; Gibbons et al., 2016). Disturbance strength can influence the community response by altering mechanisms such as physiological plasticity, adaptation, species extinctions, or changing community composition (Sousa, 1984; Lake, 2003; Griffiths and Philippot, 2013). To date, there are only few empirical studies that have investigated how disturbance strength influences compositional turnover or alpha-diversity (species richness) in bacterial communities. Berga et al. (2012) found gradual changes in bacterioplankton composition in response to a salinity disturbance of increasing intensity. Gradual reductions in richness in response to increasing disturbance frequencies in soil bacterial communities have also been observed (Kim et al., 2013). In contrast, Gibbons et al. (2016) found evidence for unimodal disturbance–diversity relationships in response to disturbance intensity and frequency of two disturbance types (biomass removal, UV exposure) in a complex culture system. Their results supported the Intermediate Disturbance Hypothesis, which predicts that diversity in communities is highest at intermediate rates of disturbance frequency and intensity (Connell, 1978).

Disturbances can also affect communities at the regional scale and change beta- and gamma-diversity as well as community assembly mechanisms, in particular whether communities are stochastically or deterministically assembled. If communities are stochastically assembled their composition is determined by random birth, death as well as immigration and emigration events, whereas species sorting by local environmental conditions leads to deterministically assembled communities (Nemergut et al., 2013). In general, effects of disturbances at the regional scale are less studied and focused mainly on laboratory experiments with simple microbial model communities (e.g., Cadotte, 2007) or field studies/experiment with larger organisms (Chase, 2007; Lepori and Malmqvist, 2009; Vanschoenwinkel et al., 2010). The consensus of these studies is that even at this scale disturbance strength appears to be important and it has been shown that beta- and gamma-diversity peak at intermediate disturbance strength (Cadotte, 2007), where co-existence of species was highest. Experiments performed under more realistic conditions in the field showed that strong disturbances affecting all local communities in a region or metacommunity decrease beta-diversity (Chase, 2007; Lepori and Malmqvist, 2009; Vanschoenwinkel et al., 2010) and this effect has also been found for bacterial communities in soils as a response to a disruption related to land conversion (Rodrigues et al., 2013). Such decreases in beta-diversity can occur as a result of strong deterministic effects of the disturbance, because it filters out species that are unable to resist harsh environmental conditions (Chase, 2007; Jiang and Patel, 2008). However, several studies with microbial communities have also shown that the timing of the sampling in relation to the disturbance is important and that the importance of stochastic assembly processes peaks at early stages after the disturbances, whereas deterministic processes became more important at later stages of the succession processes (Ferrenberg et al., 2013; Zhou et al., 2014; Zhang and Zhang, 2015).

The response of bacterial communities to disturbances may also be influenced by dispersal (e.g., Shade et al., 2012; Vuono et al., 2016). For example, dispersal can introduce new species to local communities that are specialized to the disturbance conditions or enable the re-colonization by taxa that became locally extinct in response to a disturbance, thereby enhancing community resilience. Most studies that have tested the resistance and resilience of complex bacterial communities have been done at the scale of laboratory experiments (Shade et al., 2012). They do often have the disadvantage that they preclude natural dispersal sources as well as recruitment processes from internal seed-banks, both of them being important features of natural communities (Lennon and Jones, 2011). In general, field experiments addressing how bacterial communities respond to disturbances are limited and have mostly focused on local communities (see Jones et al., 2008; Shade et al., 2012). Shade et al. (2011) performed an experiment with planktonic bacterial communities in lakes where they studied the effect of lake mixing and found that bacterial communities were resilient to nutrient and oxygen changes caused by the mixing. Jones et al. (2008) studied the response of freshwater bacterial communities to typhoon-induced water mixing and concluded that bacterial communities were able to recover fast. However, these studies were not able to address the effects of disturbance strength on bacterial communities in multiple ecosystems, i.e., across space, at the same time. We also lack understanding about how both compositional and functional stability are affected by differences in disturbance intensity at the field scale. For this purpose, rock pools are ideal natural model systems since they are highly variable systems where disturbances occur naturally and frequently (Srivastava et al., 2004; Jocqué et al., 2007; Langenheder et al., 2012). Evaporation and reduced precipitation cause drying events as well as increases of salinity by concentrating the salt that pools receive by sea-spray. Salinity causes differences in BCC among rock pools (Langenheder and Ragnarsson, 2007; Langenheder et al., 2012) and is generally a very strong environmental (Lozupone and Knight, 2008) and evolutionary filtering force (Logares et al., 2009). Moreover, the fact that disturbances occur frequently and at different intensities offers the opportunity to study whether adaptation to these conditions occurs. Finally, rock pools are small, which makes it easy to manipulate them, and they are located close to each other so that they are all influenced by the same climate regime. They are filled during rainfall and thus influenced by bacteria dispersed with rain.

Here we performed a whole ecosystem manipulation study where the salinity of rock pools was increased at three different levels and decreased back to their original salinity levels after rainfall. The main aim was to determine how disturbance intensity, i.e., salinity of different intensities, affects bacterial communities in terms of composition, diversity and functioning, both at the local and regional scale. We specifically hypothesize (1) that compositional changes along the disturbance gradient are stronger than functional changes due to functional redundancy; and (2) that increasing salinity acts as a strong environmental filter and thereby reduces (a) alpha-diversity and (b) beta-diversity because deterministic community assembly processes are enforced.

This study employed whole ecosystem manipulations to investigate the mechanisms by which freshwater bacterial communities responded to environmental change, specifically increasing salinities, at local as well as regional scales, and aimed to identify whether functional and compositional recovery occurs. The major finding was that community composition is the most sensitive parameter to salinity changes, whereas community functions (respiration and CSUR) did not change. These results support our first hypothesis because compositional resistance decreased with increasing salinity levels, whereas functioning was stable due to functional redundancy. However, community composition showed a high level of recovery, i.e., became similar to pre-disturbance conditions once salinity changes were reversed. On the contrary, the second hypothesis had to be rejected because alpha-diversity (richness and evenness) and beta-diversity were resistant to any of the applied levels of salinity change.

Our study is in line with results of previous lab-based studies that have shown that the intensity of a disturbance or environmental change determines its effect on BCC (Ager et al., 2010; Berga et al., 2012). Hence, BCC changed at an increasing magnitude with increasing salinities. Moreover, we found changes in the relative abundance of major bacterial classes reflecting a shift from those with preferences for freshwater vs. marine conditions with increasing salinities in the pools (Zwart and Crump, 2002; Newton et al., 2011). For example, Betaproteobacteria that are typically more abundant in freshwater compared to marine environments decreased after salt additions (Nold and Zwart, 1998; Newton et al., 2011), whereas Gammaproteobacteria that are often found in lower abundances in freshwater compared to marine environments (Barberán and Casamayor, 2010; Herlemann et al., 2011), increased in response to salt additions. On the contrary, increasing salinities did not change broad bacterial community functions such as respiration rates or the rate of carbon utilization (CSURs), which were also found to be resistant or resilient to salinity or nutrient disturbances in previous studies (e.g., Bowen et al., 2011; Baho et al., 2012; Berga et al., 2012). Functional redundancy may, however, depend on the type and specificity of a function (Langenheder et al., 2006; Peter et al., 2011), so that we cannot preclude that different results, including concomitant changes in community composition and functioning along disturbance gradients, are found if different functions are studied.

An interesting result of this study is the observed resistance of bacterial communities to small increases of salinity such as 3 psu. Resistance of BCC to disturbances has been observed in other studies (e.g., Bowen et al., 2011), however, here it was surprising since previous results from laboratory experiments have shown that salinity increases of a similar magnitude can induce changes in BCC (Langenheder et al., 2003; Berga et al., 2012; Székely et al., 2013). Our findings here are, however, in congruence with a study that has been implemented along the salinity gradient in the Baltic Sea, where communities that were found in localities of salinities up to 3.2 psu were compositionally very similar to those found at 0 psu (Herlemann et al., 2011). In general, the resistance to small increases in salinity could be related to adaptation to fluctuating environmental conditions that include recurrent changes in salinity in rock pools. Support for this idea comes also from the observed high recovery capacity of the communities that encountered a salinity change of 6 and 12 psu. Shade et al. (2011) also reported high recovery rates after a disturbance, more specifically to water mixing, in aquatic bacterial communities and Girvan and Campbell (2005) observed recovery of soil bacteria after copper additions. It has been suggested that the disturbance history of a community can also influence its sensitivity and resilience to additional disturbances (Hawkes and Keitt, 2015). Additionally, it has also been shown that bacterial communities are less affected by strong disturbances when they have experienced smaller disturbances before (Bressan et al., 2008) or even that some bacterial strains can anticipate environmental changes because of pre-exposure to the disturbance (Mitchell et al., 2009). Hence, “stress priming” may be commonly found in bacterial communities (Rillig et al., 2015; Andrade-Linares et al., 2016), in particular also in response to osmotic stress (Andrade-Linares et al., 2016). Rock pools are small and often temporary ecosystems that show strong fluctuations in environmental conditions over time. Thus, it seems possible that their disturbance history and historical environmental filtering could explain why bacterial communities were relatively little affected by salinity. In our studies, the communities did not change unless they experienced strong increases in salinity (6 and 12 psu) which they have probably rarely experienced before.

The fact that we did not observe any negative effect of increasing salinities on species richness could be related to the magnitude of the change, which might not have been strong enough to suppress or filter out many bacterial species. Herlemann et al. (2011) showed that, opposite to what is expected for macroorganisms, where richness is lowest at intermediate salinities, i.e., brackish conditions (Telesh and Khlebovich, 2010), increases in salinities did not affect bacterial species richness. This could be a result of a fast adaptation of bacteria to brackish conditions, for example, by generalist taxa (such as OTUs 1659 and 10659 which were unaffected by salinity increases and abundant in all treatments; Supplementary Table 4). Alternatively, fast replacement of sensitive bacteria with opportunists could also explain why we did not observe decreases in species richness with increasing salinity in our study. Support for this comes from the observation that a considerable number of OTUs switched from being rare to abundant after the disturbance, and, vice versa, there was a similar number that “switched” in the other direction. Likewise, beta-diversity did not decrease along the disturbance intensity gradient, which suggests that the applied salinity changes were not strong enough to decrease variability among local communities due to the filtering imposed by the disturbance as has been observed in animal metacommunities (Chase, 2007; Lepori and Malmqvist, 2009). This seems likely because the importance of deterministic vs. stochastic assembly processes did not change along the disturbance gradient. The overall dominance of deterministically assembled communities that we observed suggests that environmental selection is the major structuring factor in the pools as also seen in Langenheder and Ragnarsson (2007), Langenheder et al. (2012), Székely and Langenheder (2014), whereas stochastic (neutral) assembly processes are relatively less important than in other types of aquatic ecosystems (Östman, 2011).

The experimental manipulation included the coverage of the pools during the disturbance period and an initial mixing of water that led to changes in other environmental parameters such as DOC and TP concentrations. This occurred equally in all treatments, including the controls, and may explain some of the temporal patterns that we observed in all pools irrespective of the salinity level. Firstly, the pattern observed for species richness, with higher values at T65 and close to initial values at T160 could be explained by the manipulation of the pools. For example, the mixing of the pool water, which resulted in the re-suspension of precipitated material and bacteria found in the sediments, could have “seeded” the water column and thereby induced an increase of richness in all the pools. On the other hand, another possible explanation for this pattern could be that light exclusion, due to the covering of the pools, affected primary producers in the pools. This seems unlikely because Chl a concentration did not decrease during the period when the pools were covered, however, stress responses of phototrophic organisms, such as changes in their heterotrophic activity (mixotroph organisms) or formation of resting stages might have directly affected the bacterial communities, and in particular evenness, by alterations in intra-specific competition with autotrophs (McKnight et al., 2000; Eiler, 2006; Hornák et al., 2006). The increase in CSURs after the manipulation and subsequent return to initial levels by the end of the experiment, could also have been caused by increases in nutrient concentrations and substrate diversity due to the water mixing that may have interfered with or facilitated the use and/or rates of use of the carbon substrates in the Biolog Ecoplates. An alternative explanation is that the increase in CSUR is related to functional complementarity due to increases in species richness (Venail et al., 2008), which was also found in response to the manipulation, supporting previous findings of positive relationships between bacterial richness and ecosystem functions (e.g., Bell et al., 2005; Langenheder et al., 2010; Peter et al., 2011).

To summarize, our results show that freshwater bacterial communities were (a) functionally resistant to salinity changes and (b) compositionally resistant to salinity changes up to 3 psu and resilient to salinities ≥6 psu. OTU richness, evenness, as well as beta-diversity or the underlying community assembly mechanisms seemed to be unaffected by increases of salinity. More studies are needed to evaluate whether our findings generally apply to different types of natural systems or are specific for environments with strongly fluctuating environmental conditions and strong legacy effects, such as different types of small and temporary aquatic habitats. Future studies should therefore aim to understand how the perturbation history of communities influences their functional and compositional resilience at local and regional scales. To conclude, the important finding from our whole ecosystem manipulation experiment is that the stability of bacterial communities was generally high compared to many previous studies carried at the laboratory scale (e.g., Shade et al., 2012).

MB designed and performed the experiment, samples processing, data processing, and writing. YZ and AS performed the experiment, sample processing, and writing. SL designed and performed the experiment, samples processing, writing, and financial support.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.