Resistance to compositional change in the face of disturbances is enhanced if a microbial community contains many individuals that have versatile physiologies, or physiological plasticity (Evans and Hofmann, 2012). Bacteria often navigate environmental change by expressing a range of metabolic capabilities (e.g., Meyer et al., 2004; Swingley et al., 2007), and therefore the existing community can confront new conditions through gene expression by individual cells. From an evolutionary standpoint, adaptive gene expression refers to natural selection acting on gene expression (Whitehead and Crawford, 2006), and this phenomenon has been observed, for example, in a yeast model of experimental evolution (Ferea et al., 1999). Furthermore, mixotrophy, or the ability to use many different carbon and energy sources, may be a common phenomenon in natural microbial communities (Eiler, 2006), and provides further support for the notion of individual flexibility in fluctuating environments. Cellular stress responses also provide protection for individual cells from damaging physical factors such as reactive oxygen species, temperature, and ultraviolet light (e.g., Craig, 1985; Ziegelhoffer and Donohue, 2009; Kolowrat et al., 2010). Stress protection in E. coli is associated with a progressive decrease in nutritional competence, or the breadth and range of carbon and nutrient resources that a cell can use (Ferenci, 2005; Ferenci and Spira, 2007), which ultimately may reduce the population’s ability to confront other environmental changes. Therefore, there may be fitness costs to microbial stress responses that are manifested in seemingly unrelated metabolic pathways, or costs of maximizing protection against one stress over another. Stress tolerance can contribute to microbial community resistance to pulse or press disturbances, but will depend on the intensity and duration of the disturbance relative to an individual’s levels of stress tolerance (Figure 4).

Compositional stability is mediated, in part, by the ability of individual microorganisms to respond to, accommodate, and exploit environmental change. This highlights a difference between microbial communities and communities of larger organisms: prokaryotes have a degree of physiological plasticity that is unparalleled in the eukaryotic world. The unique ability to shift to an entirely different lifestyle within a short time (as in the classic example of Rhodobacter sphaeroides, which can grow anaerobically as a phototroph but also grow aerobically as a chemoheterotroph) is likely to increase the compositional stability while simultaneously reducing the functional stability. Current ecological theory based on plants and animals may not accommodate the enormous physiological plasticity of prokaryotes, which may necessitate development of new principles applicable to microbial communities. We hypothesize that physiological plasticity can contribute to a microbial community’s resistance and resilience to either pulse or press disturbances. But, similar to stress tolerance, the contribution of plasticity to community stability will depend on the intensity and duration of the disturbance relative to an individual’s physiological response.

Dormancy is a bet-hedging strategy that allows organisms to enter a reduced state of metabolic activity (Jones and Lennon, 2010); see Lennon and Jones, 2011 for a recent review). A substantial fraction of community members may be dormant or inactive at any given moment (e.g., Pedrós-Alío, 2006; Jones and Lennon, 2010). Dormancy strategies may be common among communities living in temporally dynamic environments, promoting overall compositional stability in fluctuating conditions. Furthermore, the proportion of inactive taxa (and, inactive individuals within a taxon) may signify important environmental differences among communities from similar habitats, as shown, for example, in gut microbial communities from humans with and without irritable bowel syndrome (Rehman et al., 2010).

Dormancy has most likely evolved across the tree of life as a means for contending with temporarily variable environments. It is an advantageous strategy under unpredictable conditions because it allows individuals to maximize their long-term, geometric fitness (de Jong et al., 2011). Recent studies suggest that, in a wide range of ecosystems, a substantial fraction of microbial communities may be metabolically inactive (Lennon and Jones, 2011). This observation has important implications for the resistance and resilience of microbial communities. First, on a population-level, dormancy may prevent the extinction of certain taxa from a system. For example, active individuals of E. coli succumb when exposed to certain antibiotics. However, the population can ultimately be rescued, not necessarily by the survival of mutants, but rather by subpopulations of dormant persister cells that become reactivated when antibiotic effects are attenuated (Lewis, 2006). Evidence suggests that microorganisms may use dormancy in a variety of other situations as well, for example when challenged by unfavorable temperatures (Whitesides and Oliver, 1997), starvation (del Giorgio and Gasol, 2008), or predator-induced mortality (Pearl et al., 2008). Second, dormancy has the potential to affect the stability of communities and ecosystem processes. Dormant individuals can be long-lived and contribute to seed banks. It is well documented that seed banks can maintain species diversity (Chesson, 2000), and this diversity may directly contribute to stability of microbial communities via niche complementation and/or functional redundancy (Loreau and Hector, 2001; Petchey and Gaston, 2002). Seed banks may also aid in the recovery of microbial communities from severe disturbance events. For example, dormant seeds often contribute to the reestablishment of plant communities following fire, flooding, or wind storms (Bond and Midgley, 2001). In many cases, microbial communities show similar signs of rapid recovery following catastrophic disturbances (Jones et al., 2008; Shade et al., 2012). Although dormancy may contribute to the stability of these communities, this remains to be tested.

We propose that dormancy is more important for maintaining community stability under pulse disturbance scenarios (Figure 4), as dormancy would be advantageous in a temporarily disturbed ecosystem, but likely less so in a continuously disturbed ecosystem. The exception to this would be a long-lived seed bank of dormant cells that is sustained beyond the effects of a press disturbance. Also, depending on the specific ecosystem changes caused by a press disturbance, environmental cues for “waking” from dormancy (Epstein, 2009) may be altered or absent if a disturbance is continuous.

Microorganisms generally feature rapid growth, high population densities, and high mutation rates and are capable of recombination via lateral gene transfer, which facilitates response to disturbance events (e.g., Lenski and Bennett, 1993). As such, disturbance often can provide selection pressure that drives diversification (Travisano and Rainey, 1998; Cohan, 2002). A number of studies have shown that rapid evolution of population traits can influence the temporal dynamics of microbial communities. For example, microbial predator-prey dynamics are strongly affected by selection (Yoshida et al., 2003), especially for traits that provide defense against predators (Little and Currie, 2008). Also, rapid evolution can counterbalance the top-down effects of a novel predator on nutrient cycling. For example, viruses introduced into continuous cultures of the picocyanobacterium Synechococcus dramatically reduced population densities, thereby increasing availability of the growth-limiting resource, phosphorus. However, Synechococcus population densities rebounded with the growth of a virus-resistant Synechococcus mutant. Host resistance coincided with reduced nutrient availability (Lennon and Martiny, 2008). Together these studies suggest that rapid evolution is an important mechanism of compositional and functional resistance and resilience to certain disturbances, and is particularly important for response to press disturbances (Cohan, 2002). By contrast, there may be advantages of bet-hedging strategies, such as phenotypic plasticity or dormancy, to contend with disturbance in systems that experience transient pulse disturbances. Adaptability is likely important for microbial community resistance and resilience given repeated pulse disturbances (such as in a disturbance regime, e.g., fire or flooding), as well as press disturbances (Figure 4).

A tradeoff between growth rate and resource use efficiency (and hence competitive ability) may underlie the capacity of microbial populations to respond to disturbance (Stevenson and Schmidt, 2004). Enhanced growth rate is accompanied by a higher rate of synthesis of ribosomal components and is faster in microorganisms with more copies of rRNA-encoding genes, reducing the response time to favorable growth conditions for such organisms. When pulse disturbances are followed by favorable growth conditions, the fastest responders will multiply and alter community composition (Klappenbach et al., 2000), resulting in low resistance. By contrast, microorganisms with fewer copies of rRNA-encoding genes maximize efficiency of resource use (progeny per mole substrate; Lee et al., 2009), and may increase community resistance to press disturbances that result in long-term resource limitation. Genes other than rRNA genes distinguish bacteria optimized to grow at high nutrient (copiotrophic) and low nutrient (oligotrophic) conditions (Lauro et al., 2009). While the identity of rapid responders may be idiosyncratic across environments, quantifying the capacity of a microbial community for rapid growth or efficient resource utilization could inform hypotheses regarding each community’s compositional responses to pulse and press disturbances. Growth rate likely is important for microbial community resilience from pulse disturbances, as a few surviving individuals could grow quickly to recover to pre-disturbance population sizes after a sudden pulse, mortality-inducing disturbance, especially if the disturbance makes new resources available.

The relative growth rates of interacting microbial communities also can have implications for microbial resilience and resistance to press disturbances. A recent study compared the drought responses of a fungal-based food web in grassland soil with that of a bacterial-based food web in agricultural soil (de Vries et al., 2012). The results of this study suggested that relatively slower-growing fungi were more resistant to drought, but less resilient, while relatively faster-growing bacteria were less resistant but more resilient. The authors built structural models to assess the impact of fungal and bacterial drought responses on microarthropod (grazers of bacteria and fungi) richness, soil respiration, nitrogen dioxide production, and nitrogen leaching. This study demonstrated that press disturbances that alter microbial food webs may also influence soil resource availability. It also showed that the stability of microbial food webs was contingent, at least in part, on the ratio of slower-growing fungi to faster-growing bacteria.

Another means by which a microbial population can respond quickly to environmental change is through stochastic gene expression (Avery, 2006; Raj and van Oudenaarden, 2008), a process that samples multiple phenotypes and hence, like dormancy, is also considered a bet-hedging strategy. The presence of persister cells – dormant variants within a bacterial population that are tolerant to antibiotics (e.g., Lewis, 2010) – is one example of an alternative phenotype that increases fitness in an environment experiencing a transient selective pressure. Stochastic gene expression is among the multiple pathways that can lead to the formation of persisters (Lewis, 2010), a phenomenon that may have many parallels in non-pathogenic microorganisms. Stochastic gene expression may be key for microbial community resistance to pulse disturbances, as it offers a short-term strategy for survival of individuals that can re-populate a community after disturbance.

The large population sizes and rapid dispersal abilities (e.g., Finlay, 2002) of microorganisms can play an important role in community recovery after disturbance. Dispersal is a key feature of metacommunity theory, which recognizes communities as collections of interacting local communities linked by the movement of individuals in heterogeneous landscapes (e.g., Leibold et al., 2004; Logue et al., 2011). Disturbances initiate iterations of community re-assembly by killing or inactivating local resident taxa, releasing resources, and creating empty niches. These empty niches can be filled by local taxa (resistant taxa or taxa retrieved from seed banks) or by immigrants that arrive from other localities within a metacommunity.

Colonists dispersed from nearby localities (patches) can “re-seed” microbial populations that have become locally extinct after disturbance, thereby facilitating community resilience (Figure 4). For example, freshwater bacterial communities disturbed by a pulse salinity increase were both compositionally and functionally resilient because of continuous dispersal of microorganisms from an undisturbed source community (Baho et al., 2012). As another example, protozoan and rotifer population densities were more resilient after a recurring pulse disturbance (replacing 99% of a mesocosm’s contents with sterile media) when the disturbed mesocosm was connected to an undisturbed one (Altermatt et al., 2011a) than when the disturbed mesocosm had no refuge.

On the other hand, if disturbances are wide-spread such that they affect entire regions (e.g., climate effects, such as heat waves), dispersal could promote the dissemination of disturbance-tolerant taxa among localities (Eggers et al., 2012), thereby changing the dominant membership of a community and decreasing overall community resilience at both local and regional scales. For example, marine microalgae communities disturbed by a simulated heat wave were sensitive but not resilient because of a shift in community dominance toward a temperature-tolerant species. This species also became prevalent in other patches connected by dispersal, even though the conditions there were less suitable for it (Eggers et al., 2012).

Furthermore, niches opened by disturbance events are subject to stochastic colonization events by dispersed microorganisms. “Priority effects” refers to the impact that successful early colonizers may have on community re-assembly after disturbance, which can affect the likelihood of colonization by subsequently dispersed microorganisms (e.g., Shulman et al., 1983). Early post-disturbance colonizers that adapt rapidly to local conditions may persist long-term, out-competing native community members and impeding community resilience (Urban and De Meester, 2009).

Together these studies and others demonstrate that microbial dispersal, in interaction with biological attributes of local resident populations (plasticity, dormancy, or evolution) and disturbance characteristics, can have important implications for the resilience of microbial communities.

In general, diversity is thought to influence how communities respond to disturbance. Aspects of alpha diversity, such as species richness and evenness, have been shown to enhance the functional resilience of communities of larger organisms (Allison, 2004; Downing and Leibold, 2010; Van Ruijven and Berendse, 2010), whereas evidence about the impact of richness and evenness on microbial community resilience is mixed (e.g., Griffiths et al., 2000a; Wertz et al., 2007; Wittebolle et al., 2009; van Elsas et al., 2012).

One challenge lies in clarifying the functional and compositional responses to disturbances of diverse communities. The underlying mechanisms behind a positive relationship between taxon diversity and resilience may be related to a buffering effect, called the Insurance Hypothesis (Yachi and Loreau, 1999). More genetically diverse communities are more likely to contain taxa with complementary response traits (e.g., Tilman et al., 1997, 2006; Lavorel and Garnier, 2002; Elmqvist et al., 2003) and the ability for rapid compensatory growth after a disturbance (e.g., Flöder et al., 2010), which may promote resilience. Also, rare microbial taxa (potentially below the limit of detection, and therefore not counted in estimates of community diversity) may quickly respond to altered environmental conditions and become abundant. This is exemplified by the case of a rare Vibrio species that was below detection in the majority of the time points over a 6-year study in the Western English Channel, but then bloomed to become a prevalent member of the community at one time point (Caporaso et al., 2011). Co-occurrence networks were used to find that the Vibrio bloom was correlated to a bloom of a diatom species (Gilbert et al., 2012). Though models are constantly improved for predicting conditions for microbial blooms (Larsen et al., 2012), niche spaces for many environmental microorganisms remain uncharacterized, which further veils the relationship between compositional and functional diversity and microbial community stability.

Here we discuss a few examples (of many in the literature) of the impact of diversity on microbial community stability. In an early experiment to tease apart the relations between diversity and community functions, Griffiths et al. (2000b) examined the impact of soil microbial diversity on functional stability using a range of intensities of soil fumigation followed by a disturbance, either a heat shock (pulse disturbance) or the addition of the heavy metal copper (II) sulfate (press disturbance). The results indicated that microbial production (rate of generation of biomass, here measured as thymidine incorporation) was not affected, but specific functions, such as nitrification, decreased when diversity was lower. The lower-diversity communities were less functionally resistant (measured as grass residue decomposition rate) to the press disturbance and unable to recover, but were sometimes more resistant to the pulse disturbance than the higher-diversity communities. The control community, which was not fumigated and had the highest diversity, was often the most resilient to both pulse and press disturbances.

Building on the Griffiths et al. (2000b) results, the relationship between species richness and stability was investigated for denitrifiers and nitrite oxidizers (Wertz et al., 2007), two specialized functional groups of soil microorganisms important for nitrogen cycling. In this work, microbial abundance (and, as an extension, richness) was altered by inoculating soil microcosms with different dilutions of microbial cells from non-sterile soil. After a period of incubation, the microcosms were subjected to a pulse heating disturbance. Denaturing gel gradient electrophoresis fingerprints of the denitrifiers and the nitrite oxidizers were coupled with measurements of the two processes, and each responded differently to heating. Though both processes were sensitive, denitrification was resilient after 3 months, while nitrite oxidation did not completely recover. The response of richness to heating was variable across dilutions, and richness did not recover to pre-disturbance levels after 3 months. These results suggest that a community may recover in function even if diversity remains altered after disturbance, and that the initial richness may not necessarily impact functional recovery.

A microcosm experiment was performed to specifically examine the role of initial community evenness (equitability of taxa abundances) on functional stability of denitrifiers subjected to a press increase in salinity and temperature (Wittebolle et al., 2009). The authors found that high initial evenness was important for functional stability of microbial communities. However, the communities were not observed after the disturbance ceased to assess resilience of function or evenness. This work demonstrated that aspects of diversity other than species richness can play a role for community functional stability. Given that many environmental microbial communities are characteristically uneven because they contain a large number of rare taxa (sometimes referred to as the “rare biosphere,” e.g., Casamayor et al., 2001; Sogin et al., 2006), the implications of small differences in evenness among generally uneven communities may be of interest for further investigation.

A very recent experiment demonstrated the impact of diversity on stability by creating synthetic combinations of soil isolates, rather than confronting the complexity and unknown organisms of the natural soil (van Elsas et al., 2012). With constructed mixtures containing various numbers of random representatives of a collection of cultured isolates, the authors demonstrated a highly significant correlation between species richness of the community and its resistance to invasion by an E. coli strain (van Elsas et al., 2012).

Together, these studies and others reveal the complex relationships between microbial community diversity, function, and stability. Multiple aspects of diversity (richness, evenness) can affect microbial functional resistance and resilience, and general and specific community functions may have different overall responses to pulse disturbance. Importantly, these studies suggest that there may not be a “one-size-fits-all” response of microbial diversity and function to disturbance. For diversity-stability relations, more work must be done to understand system-specific trends before it will be possible to determine which patterns, if any, are general across microbial systems.

Community turnover is the replacement and substitution of community members along an environmental gradient or with time (Wilson and Shmida, 1984). Turnover is directed by the growth of populations, and partially determines how quickly a community recovers from a pulse disturbance (Figure 4). High-throughput fingerprinting and sequencing tools that enable experiments involving longer time series demonstrate that many microbial communities have clear trajectories. For example, successional patterns have been observed in tree leaf bacterial communities over the growing season (Redford and Fierer, 2009; Redford et al., 2010), in the human gut after antibiotic treatment (Antonopoulos et al., 2009; Dethlefsen and Relman, 2011), and in the infant gut during the first two and a half years of development (Koenig et al., 2011). Also, seasonal trajectories are common in freshwater, marine, and sediment systems (Fuhrman et al., 2006; Christian and Lind, 2007; Shade et al., 2007; Nelson, 2008; Andersson et al., 2009; Crump et al., 2009; Gilbert et al., 2009, 2010). Together these studies provide insight into the temporal scale on which community turnover and disturbance responses may be anticipated in similar systems, but turnover is unknown for many other habitats. Further investigation is needed to describe, quantify, and compare turnover within and across habitats.

Species-time relationships, which quantify the accumulation of new species in a community with time, provide a method to calculate community turnover (White et al., 2006). There is preliminary evidence that species-time relationships of community turnover may be common among microbial communities from very different environments. For example, microbial communities in streams (Portillo et al., 2012), on leaf surfaces (Redford et al., 2010), across a set of newly deglaciated soils (Nemergut et al., 2007), and in bioreactors (Van Der Gast et al., 2008), have exhibited species-time relationships with turnover rates comparable to those of larger organisms. If most microbial communities display a characteristic turnover rate, as observed for communities of larger organisms (White et al., 2006), variation around the expected range of community turnover rates may be a reasonable starting point for comparing microbial community responses to disturbance.

Similarity-decay (Nekola and White, 1999) is a common method for understanding changes in microbial community structure over space (e.g., Horner-Devine et al., 2004; Jones et al., 2012), but has also been applied over time in communities of larger organisms (Korhonen et al., 2010). For temporal similarity-decay, all pairs of community resemblance (a.k.a. similarity or distance) are regressed against time (or space) between community observations. The slope of this regression is analogous to a turnover rate, but has additional utility from species-time relationships because the resemblance metric can be chosen to include properties of community composition (e.g., Sørensen), structure (e.g., Bray–Curtis), and phylogenetic representation (e.g., UniFrac distance, based on the underlying calculation of phylogenetic diversity, Faith, 1992). Similarity-decay has been applied to understand temporal dynamics of microbial communities (e.g., Wittebolle et al., 2008; Bürgmann et al., 2011). As one example, similarity-decay was used to compare resilience of lake microbial communities across treatments in an experiment designed to separate the environmental drivers of oxygen and nutrients from the physical process of water column mixing, an important seasonal disturbance to temperate lake bacterial communities (Shade et al., 2011). In this study, similarity-decay relationships were used in part to quantify the relative robustness of the communities to these environmental disturbances, such that the hypolimnion community was found to be the most sensitive to oxygen addition, but also the most resilient. Though similarity-decay clearly is useful as a baseline descriptor of temporal community turnover, this study demonstrated the additional utility of similarity-decay for comparing microbial community resilience when challenged with different disturbances.

Interactions between community members, including competition and mutualism (Little et al., 2008), are also important in determining community response to disturbances. For example, a mesocosm experiment with protist communities demonstrated that competitive interactions and disturbance characteristics together determined the community disturbance response, where competition between species became increasingly important in driving extinction as the intensity of disturbance was amplified (Violle et al., 2010). This trend was observed despite the fact that the species were chosen for the experiment because they collectively represented a wide range of competitive ability and disturbance tolerance. Thus, the community response was more complex than the sum of the individual species’ traits, and instead hinged on interspecific interactions.

As time series analyses of microbial communities become increasingly available, we will be better able to quantify complex interactions among community members. For instance, recently developed statistical tools enable investigation of microbial interactions by studying co-occurrences of microbial taxa through time. Networks can be built from co-occurrences, with nodes representing taxa or operational taxonomic units (OTUs) and connecting edges representing correlation over time (e.g., Ruan et al., 2006). These networks assess effects of disturbance on community dynamics (Montoya et al., 2006). Positive co-occurrence may indicate common preferred environmental conditions or cooperative activities (facilitation and/or syntrophy). Similarly, negative correlations may represent the outcome of competition (i.e., displacement) or negative interactions, such as allelopathy or predation (Fuhrman, 2009). Co-occurrence networks are gaining popularity in microbial ecology (Faust and Raes, 2012), and recently have been applied to observational studies of marine and soil systems (Steele et al., 2011; Barberan et al., 2012; Eiler et al., 2012; Gilbert et al., 2012). Network analysis applied to controlled experiments can provide insight into the associations between community members that are lost, gained, or maintained after a disturbance (Shade et al., 2010). Networks can also be used to discover “core” community members shared across communities from similar habitats (Shade and Handelsman, 2012), serving as a method to identify taxa shared across localities within a larger metacommunity.

Theoretical and empirical work on food webs indicate that the robustness of such networks is affected by several attributes, such as the strength of links among species (Montoya et al., 2006). In particular, architectural properties of trophic networks affect the relationship between network complexity and stability (May, 1974; Pimm, 1984; McCann et al., 1998; Rozdilsky and Stone, 2001). For example, in mutualistic networks, compositional diversity and connectivity lead to higher resilience, whereas these properties destabilize trophic networks (Thébault and Fontaine, 2010). Microbial communities are probably affected by both mutualistic and trophic interactions, and thus may provide a unique system for exploring the relationship between network attributes and community stability. Though we hypothesize that microbial interactions and the emergent network properties of a microbial community likely contribute to both resistance and resilience to pulse and press disturbances (Figure 4), the exact mechanisms underlying this stability often are unknown, and may be challenging to unravel for many species-rich microbial communities that maintain both mutualistic and trophic interactions among their members.