Microbial community responses to anthropogenically induced environmental change: towards a systems approach


Correspondence: E-mail: Andrew.Bissett@csiro.au


The soil environment is essential to many ecosystem services which are primarily mediated by microbial communities. Soil physical and chemical conditions are altered on local and global scales by anthropogenic activity and which threatens the provision of many soil services. Despite the importance of soil biota for ecosystem function, we have limited ability to predict and manage soil microbial community responses to change. To better understand causal relationships between microbial community structure and ecological function, we argue for a systems approach to prediction and management of microbial response to environmental change. This necessitates moving beyond concepts of resilience, resistance and redundancy that assume single optimum stable states, to ones that better reflect the dynamic and interactive nature of microbial systems. We consider the response of three soil groups (ammonia oxidisers, denitrifiers, symbionts) to anthropogenic perturbation to motivate our discussion. We also present a network re-analysis of a saltmarsh microbial community which illustrates how such approaches can reveal ecologically important connections between functional groups. More generally, we suggest the need for integrative studies which consider how environmental variables moderate interactions between functional groups, how this moderation affects biogeochemical processes and how these feedbacks ultimately drive ecosystem services provided by soil biota.


Soil is essential to life; mediating terrestrial biogeochemical and nutrient cycling, providing a substrate for food production and acting as a filter for the provision of clean groundwater. Soil is complex; containing communities from multiple trophic levels, complex substrate transport mechanisms, extreme temporal and spatial heterogeneity and a myriad of internal and external feedbacks that determine community structure and function. Microbes facilitate many of the processes mediating ecosystem services, including decomposition and mineralisation, inorganic nutrient cycling, disease causation and suppression, and pollutant removal. Any soil disturbance may disrupt microbial activity, alter these processes and impact on ecosystem performance. The complexity and high biological diversity of soils facilitates many feedbacks and potential interactions, most of which are not considered in traditional ‘compartmentalised’ approaches (in which only specific groups are considered in isolation) to investigating soil systems. More holistic consideration of critical biotic and abiotic interactions (e.g. network modelling) will advance understanding of mechanisms governing soil systems and thus improve prediction and management of these systems in the face of perturbation.

Anthropogenic activity is altering soil physical and chemical conditions at many spatio-temporal scales. Soil degradation is caused by erosion, nutrient status change, salinisation, compaction and pollution. It is estimated that > 15% of soils worldwide are degraded (Bridges & Oldeman 1999), but the ramifications of alterations to microbial community structure (composition in terms of organism identity, richness and evenness) and function are poorly understood. Despite the importance of soil to food production, global C cycling and ecosystem health, we still have little insight into how changes to soil ecosystem services are controlled. Are shifts in microbial community composition important or are services only under the control of abiotic factors? If shifts are important do they occur via deterministic or stochastic processes? For example, if community composition is important then we must consider various drivers of microbial community assembly, including adaptation, dispersal, priority effects, selection, legacy effects, dormancy, persistence, co-occurrence and network interactions.

Here, we examine ecological factors involved in the responses of soil microbial communities to anthropogenically mediated biotic and abiotic stressors. Given the difficulty of linking microbial community structure to function, researchers have generally focused on elements of the microbial community responsible for the delivery of specific ecosystem services. This accords with Ridder (2008) who stated that because ecosystem services are not performed by whole ecosystems, but by specific system components, there is no need to consider the whole system. Such approaches serve to simplify microbial systems into compartments that directly impact on a given ecosystem service, and may be more amenable to investigation. However, the complexity of the soil system and the fact that it comprises multiple interacting parts, including direct and indirect feedbacks, suggests that although individual system components may be directly responsible for specific functions, many other components exert significant ‘indirect’ influence over these services. Thus, the compartmentalised approach, in which only specific groups are considered in isolation, can only provide partial understanding of soil processes and may not enable better management of these complex systems.

Kitano (2002) suggested that a systems-level understanding can be achieved by gaining insight into four key biological properties that he termed system structure, system dynamics, the control method and the design method. In relation to developing a soil systems-biology approach, these properties require the following: (1) characterising the network of organisms, gene interactions and biogeochemical pathways in the soil microbial community, (2) understanding temporal and spatial variability in soil microbial systems, (3) understanding how systems structure and variability is modulated by environmental factors, these (2 and 3) could be achieved by examining the effects of environmental gradients, multi-factorial experimental approaches and analysis using metabolic approaches such as metatranscriptomics and metaproteomics, and (4) developing management strategies that modify and control the soil system based on, and informed by, knowledge of the systems structure and dynamics.

To illustrate our thesis, we: (1) introduce concepts of resilience, resistance and redundancy, (2) discuss three components of soil microbial communities that are often considered in a compartmentalised manner in the context of perturbation: ammonia oxidisers (AO), denitrifiers and microbial symbionts, (3) discuss transient states in microbial systems, and (4) discuss systems biology and network analysis approaches to understand soil microbial systems and demonstrate this with an example of such an approach. Together these examples provide an excellent opportunity to compare the responses of different microorganisms to anthropogenic stress and to weigh the merits of compartmentalised vs. system-wide approaches.

With regard to the specific microbial groups we highlight, these were chosen because, (1) they perform important but distinct ecosystem services, (2) their diversity and function are often assessed in relation to disturbance, and (3) they represent differing levels of diversity and dependence on other organisms within related soil processes. Ammonia oxidation and denitrification are both ecologically important. AO and denitrifiers involve differing levels of diversity; AO are phylogenetically limited and thought to have relatively little functional redundancy, while denitrifiers are very diverse and therefore more likely to exhibit high functional redundancy. AO and denitrifiers respond differently to organic matter, which is essential to denitrifiers, but not nitrifiers, and moisture, which reduces oxygen availability and thus increases denitrifier, but constrains nitrifier activity (Stark & Firestone 1995). Thus, distal and proximal regulators of the nitrogen cycle, such as landform or plant type, can have opposing influences on (de)nitrification. These regulators also affect symbiotic organisms. Symbionts (mutualists and pathogens) are broadly reliant on interactions with their hosts and the intimate nature of these associations means they are often considered only in a pair-wise fashion, rather than as components of more complex soil systems. However, these organisms interact with free-living soil biota as well as each other, directly and via host-mediated feedbacks (Bever et al. 2013).

Resilience, Resistance and Redundancy

It is widely thought that microbial communities are highly functionally redundant and therefore largely resistant (able to withstand disturbance) and resilient (able to recover from disturbance) to perturbation (Allison & Martiny 2008). This view stems from the assumptions that microbes collectively have high biomass and growth potential, are universally dispersed, have low extinction potential and a high incidence of horizontal gene transfer (Allison & Martiny 2008). Although universal occurrence of microbes is a contentious issue (Hughes-Martiny et al. 2006), if true, it implies that community composition is of little importance to ecosystem function. Support for these ideas is evidenced by the relative success of ‘black-box’ models that assume biogeochemical cycling is not limited by microbial abundance or diversity and employ single mathematical functions for microbially driven processes across a range of environmental conditions (Schimel et al. 2001). While this area of ecosystem modelling is rapidly developing and there is debate as to the utility of generalised functions, these functions do approximate broad-scale changes over medium term time-scales relatively well. It must be noted, however, that black-box models are better at predicting some pathways than others and are often not able to compartmentalise specific pathways, but concentrate instead on more general functions, for example, gross nitrogen mineralisation (Friedlingstein et al. 2006; Treseder et al. 2012).

The success of black-box models could be due to three factors. First, microbial communities may be resistant, resilient and functionally redundant. These terms are discussed in Box 1. Second, function and diversity may be completely decoupled in microbial communities. Or, lastly, even if function and diversity are coupled, diversity is constant across temporal and spatial scales (Allison & Martiny 2008). These concepts are different from what is observed for higher levels of ecological organisation, where individual species losses can be explicitly linked to functional changes (Hooper et al. 2012). A recent review of resilience and resistance in soil microbial communities (Griffiths & Philippot 2012) suggested that many factors influence soil community stability to disturbance, and that response to disturbance is therefore difficult to predict. Griffiths & Philippot (2012) utilised the engineering definition of resilience (Box 1), which is the one most commonly employed in microbial ecology since it is the easiest to measure. Despite predominance of the engineering resilience concept in the microbial literature, we argue that this concept does not adequately account for the possible responses of soil communities to perturbation because it assumes a single optimum stable state, departure from which is classified as negative. Such a focus on ‘maintenance of the status-quo’, rather than on the mechanisms and processes driving change, has the potential to restrict further advances in prediction and management of disturbance responses in soil microbial systems.

Variation in Resistance and Resilience among Microbial Functional Groups

Microbial ecologists often compartmentalise ecosystems to investigate disturbance responses. AO and denitrifiers are frequently used to investigate disturbance responses because of their defined functions and varying phylogenetic diversity. Hydrocarbons and metals, two common anthropogenic pollutants, have been investigated extensively for their effects on ammonia oxidising bacteria (AOB). Acquired specific resistance (e.g. via mobile genetic elements) is an important component of system recovery from specific effects of these pollutants, while it likely plays no part in community resilience to their indirect effects (e.g. oxygen depletion). For example, hydrocarbons can compete with ammonia for the active site of ammonia monooxygenase (AMO; (Deni & Penninckx 1999), n-alkynes irreversibly inhibit AMO and aromatic hydrocarbons deplete cellular reductants that are needed to oxidise ammonia (Keener & Arp 1994). Thus, these toxicants have specific modes of action that act uniquely on AOB. These modes of action are reflected in the high sensitivity of AO to hydrocarbons; typically, AO are the most hydrocarbon sensitive organisms present in ecosystems (vanBeelen & Doelman 1997). In addition, hydrocarbons have a non-specific toxic effect as non-polar narcotics that partition into lipid-rich microbial membranes AO are sensitive to three key events that occur after disturbances such as those from hydrocarbon spills (Aislabie et al. 2004): (1) toxicity, both specific and non-specific, (2) oxygen limitation caused by hydrocarbon-degrading organisms rapidly consuming oxygen, (3) and competition for essential nutrients, and in the case of AOs, ammonium. Hydrocarbons also have direct abiotic effects on soil microbial communities, typically increasing soil temperature and reducing water retention by increasing soil hydrophobicity (Aislabie et al. 2004). While no study has demonstrated it specifically, data suggest that indirect effects of oxygen limitation and competition, and their cascading abiotic effects are likely the dominant mechanism of hydrocarbon impact on AO. For example, AO exposed to hydrocarbons (e.g. petroleum products) can recover by increasing their affinity for ammonia (Deni & Penninckx 1999). In extreme systems such as the Antarctic, AO do not recover from hydrocarbon impact (Schafer et al. 2007). In temperate systems, it is likely that the AO community adapts to these indirect effects by altering enzyme specificity and this adaptive ability is linked to levels of soil organic matter or soil microbial biomass (Dawson et al. 2007). This adaptation to indirect pollutant effects is important in system response to perturbation and is influenced by a myriad of factors including system energy status, other biota (e.g. earthworm and plant presence; Dawson et al. 2007) and duration/frequency of pollutant load. In contrast, in extreme environments, the AO community do not adapt to these indirect pollutant effects, possibly because Antarctic soils are so energy poor (Schafer et al. 2007). Thus, it is probable that ecosystem energy status, soil chemical environment and interactions between seemingly unrelated organisms are key components of AO community resilience.

Unlike hydrocarbons, metals do not have specific modes of toxic action on AO metabolism, but rather more general modes, for example, generation of reactive oxygen species and disruption of enzyme-metallic catalytic centres. Despite this, AO are very sensitive to and rapidly inhibited by metals, and maintenance of AO-mediated ecosystem function under metal stress occurs via changes in AO community composition. In temperate ecosystems, AO recover from metal pollutants within a year (Mertens et al. 2007), as they adapt to metal loads (Mertens et al. 2006). Further, recovery of nitrification is linked to community energy status: the more ammonia provided, the more rapid the recovery (Ruyters et al. 2010) and adaptation of AOB – rather than archaea – to the pollutant (Mertens et al. 2009). Mertens et al. (2009) found that changes in soil abiotic parameters due to processes such as metal ageing did not account for AO recovery, but recovery was linked to changes in community structure, which were accelerated when additional energy was provided. In simple terms, hydrocarbons cause changes in AMO substrate specificity which allows the AO community to function, whilst metals drive changes in AO community structure which also allow the AO community to function. Like adaptation to hydrocarbon pollution, these changes occur at different temporal scales and affect other ecosystem components, making a network approach suitable for characterising the systemic effects of perturbation.

Unlike AO, denitrifiers are fairly insensitive to many toxicants (Schafer et al. 2007). In fact, their numbers may increase in contaminated sites because of their role in hydrocarbon remediation (Powell et al. 2006). Further, the direct modes of hydrocarbon toxicity that impact AO do not occur for denitrifiers. Instead, there is clear differential sensitivity among denitrifiers in their response to hydrocarbon pollution. Metabolically, the enzymatic pathway containing NOR (nitrous oxide reductase, commonly assessed through the diversity and prevalence of nosZ) is more sensitive to hydrocarbons and metals than the pathway using nitrite reductase (NIR) (Siciliano et al. 2000; Holtan-Hartwig et al. 2002). However, for reasons that are unclear, the nirS genotype is more sensitive to hydrocarbons than the nirK genotype (Guo et al. 2011) and nirK is more sensitive to trinitrotoluene than nirS (Siciliano et al. 2000). There are no similar genotype divisions within nosZ containing denitrifier communities (not all denitrifiers contain nosZ). However, because of the sensitivity of NOR to pollution generally, investigators have explored how NOR recovers from hydrocarbon and metal toxicants and found that the denitrifier community can recover NOR function (De Brouwere et al. 2007). This NOR restoration is not reflected in changes in nosZ structure (Ruyters et al. 2010); instead the fitness of the community has been degraded (Holtan-Hartwig et al. 2002). Because these studies involved the use of metals, investigators postulated that a mobile genetic element was conferring metal resistance, the cost of which was reduced growth rates (Holtan-Hartwig et al. 2002). The effects of abiotic stressors on denitrifiers are much smaller than on AO. Unlike AO, denitrifiers are able to adapt to many abiotic factors (Cavigelli & Robertson 2000). Niche breadth and adaptive ability, therefore, may contribute to resilience in that many potential stressors could be circumvented by shifting within a broader niche space. For example, the facultative anaerobic ability of denitrifiers and their ability to use many organic carbon sources imply that they do not suffer from the niche limitations observed for autotrophic AO. Thus, denitrifiers appear largely resilient to many pollutants in terms of gene function, but soil denitrifiers are also influenced indirectly by the success of other system components. For example, nitrification and dentrification are often tightly coupled. Therefore, failure of nitrification may negatively impact on denitrifcation, despite denitrifiers being both resistant and resilient to both metal and hydrocarbon pollution.

Box 1. Resistance, resilience and redundancy, definitions and conceptual overview

Ecosystem stability is determined by resistance, resilience and redundancy, making these concepts central to any discussion of perturbation. Resistance and resilience may refer to community composition (species diversity, relative or absolute abundance) or function (e.g. biogeochemical process rates, disease occurrence) and may result from growth, physiological or genetic adaptation. Resistance is the degree to which a system changes following disturbance (Pimm 1984) and is sometimes considered a component of resilience (Walker et al. 2004). Resilience has been promoted widely in ecology since its introduction by (Holling 1973) and several definitions have been used (Gunderson 2000). Resilience was originally defined as the persistence of relationships within a system and suggested as a measure of that system's ability to endure changes (Holling 1973). Resilience was later used by (Pimm 1984) to describe the speed with which a system returns to its original state after perturbation, a definition widely used today (Griffiths & Philippot 2012). (Gunderson 2000) further differentiated ‘ecological resilience’ as the amount of disturbance a system can absorb without moving to a new ‘stable state’, while ‘engineering resilience’ is the rate at which a system returns to its original state after disturbance. These two definitions differ primarily in that the latter assumes a single ideal stable state, and that the system is in that state prior to disturbance, while the former does not. While the concept of engineering resilience is convenient in terms of measuring change, it is not necessarily the most appropriate when applied to dynamic systems such as soils. Because this perspective tends to neglect the dynamic and interactive nature of complex systems, it is also more likely to encourage monitoring of single parameters or specific organisms, thus offering little insight into mechanisms governing whole system functioning.

Functional redundancy is often seen as important in maintaining microbial community stability and may be evident as a lack of change in rates of ecosystem processes despite change in community structure after return of pre-disturbance conditions (Mertens et al. 2009). Potentially, both small and large changes in structure could lead to large or small changes in function (Fig. 1). Functional redundancy may result in functional stability, despite resilience and resistance not being features of community structure. For example, when ecosystem components are easily substitutable, community structure is of little consequence. Thus, testing microbial responses to disturbance, and the effect of these responses on function, is difficult because microbial function depends on a myriad of abiotic and biotic factors. A commonly employed definition of functional redundancy stipulates that the process be performed similarly under identical conditions by different organisms (Allison & Martiny 2008). A disturbance, by its very nature, implies that conditions have changed. Functional redundancy should, perhaps, refer to the ability of a group of organisms to perform the same processes at the same rate across a range of conditions. Recent work suggests that the high functional redundancy of microbial communities results from niches being filled stochastically, so that community composition is less relevant than functional composition. Microbial communities should perhaps be assessed not on species composition, but on the presence of core functional pathways (Burke et al. 2011) and interactions. Thus, if function is not phylogenetically conserved, assessment of response to change based on phylogenetic structure may lead to the erroneous conclusion that a fundamental shift has occurred, rather than merely a refiltering of the community such that core functions are being performed by different organisms. This suggests that a systems approach is needed, integrating community structure, process rates, temporal fluctuations, interactions and feedbacks.

Figure 1.

Potential shifts in community structure and function after disturbance. Left-hand panels represent microbial communities in community structure space, right-hand panels represent the same communities mapped onto functional space. Black stars represent initial communities and functions, which could undergo shifts to any of the coloured dots post-disturbance. (a) Large community shifts, small functional changes (e.g. high functional redundancy with many organisms performing the same task); (b) large community shifts, large functional changes (high coupling between function and identity, thus low functional redundancy and large niche changes); (c) small community shifts, large functional change (e.g. when mobile genetic elements confer functional changes, but community phylogeny does not change). Key questions regarding microbial response to perturbation include: What is the relative importance of biotic and abiotic factors in determining function? Do phylogenetically similar communities perform more similar functions? What aspects of structure (phylogenetic vs. functional) are most useful to map for prediction and management of ecological systems?

Figure 2.

Network analysis of saltmarsh sediment microbial communities. (a) Overview of a saltmarsh microbial network. Nodes (triangle, square, etc.) represent bacterial OTUs defined by 0.03% clustering of ribosomal rRNA V6 regions or environmental factors (green triangles) measured by (Bowen et al. 2011). Lines connecting nodes (edges) represent strong (R > 0.5) and significant (< 0.05) positive (blue) or negative (dashed black) co-occurrence relationships. Node size is proportional to an OTU's relative abundance. Node shape denotes phylum level classification (e.g. circle = Proteobacteria), node colour denotes class level classification (e.g. red = Alphaproteobacteria) and node name is defined by genus. Edge values represent between-node relationship strengths (R). The circular layout is ordered anticlockwise from the network bottom based on betweeness-centrality calculations for each node, starting with WS3 (pink diamond) as the node with the highest score. (b) Level of betweeness-centrality (y-axis) of each node as a function of number of neighbours (x-axis). The node with the highest betweeness-centrality = candidate division WS3. (c) Sub-network highlighting central position of WS3 (pink diamond) in the network topology. Secondary circles in Fig. 1a represent groups of highly connected nodes – two examples are expanded as sub-networks in (d) and (e).

To summarise, resistance and resilience vary among microbial functional groups due to a range of factors: (1) differences in specific function (e.g. nitrifier enzyme specificity under hydrocarbon pollution), (2) ability to acquire mobile genetic elements (e.g. denitrifer response to metals); the genetic structure and plasticity of the group (e.g. nitrifier response to metals), (3) sensitivity to environmental factors (e.g. energy influences on nitrifier resistance and adaptation), (4) and the effects of these factors on other organisms that a given group interacts with (e.g. in undisturbed ecosystems, denitrifier and nitrifier activity is linked to interactions with other organisms [e.g. see Siciliano et al. 2009; ]. Resistance and resilience therefore depend intimately on a complex network of interactions. This dependence on interactions is explicitly the case for symbiotic interactions involving multiple trophic levels. For example, changes in soil salinity not only directly affect rhizobial community composition and abundance, but may also mediate shifts in host dependence on the symbiosis (Thrall et al. 2008). Overall, despite some knowledge of the responses of specific functional groups to perturbation it is difficult to predict or understand whole ecosystem stability without taking a systems perspective.

Temporal Stability of Microbial Communities

As indicated above, previous work on ecosystem stability has frequently adopted an engineering concept of resilience (Box 1) which assumes a single stable state. This concept seems unlikely to apply to soil systems, particularly given their heterogeneity and the continued natural disturbance they endure (rainfall, drought, point source nutrient inputs from animals, bioturbation, etc.). Conceptual frameworks developed in plant ecology that assume natural communities constantly shift between alternative transient states that vary in community structure or function (Fukami & Nakajima 2011) may be appropriate for soil microbial systems. Transitions between different transient states occur because of variable immigration, selection, drift, dispersal and mutation histories and other stochastic processes. Different states may therefore exist despite communities being assembled under similar environmental conditions and receiving the same set of species multiple times.

It is becoming clear that the history of microbial community assembly can play a large role in community function and response to external stimuli (Elgersma et al. 2011; Keiser et al. 2011). The source of a given stressor and frequency with which it is found in a system is a critical determinant of system response when considering legacy effects and microbial response to disturbance. For example, where a pollutant occurs frequently in time and space, many organisms will likely be able to process it, even if it requires specialised adaptation. This occurs in ecosystems routinely exposed to toluene, for example, where the TOL plasmid confers the ability to degrade toluene. In some cases, shifts in community structure and function are secondary consequences of pollutant breakdown. Decomposition of diesel oil, for example, is mediated by many organisms and leads to reduced oxygen, nitrogen and phosphorus concentrations, shifting communities to favour anaerobes (Aislabie et al. 2004). The presence of microbes capable of degrading common pollutants will, therefore, add to a system's resilience and resistance to the pollutant. However, in the case of pollutants-containing xenobiotics (e.g. pesticides) specialised decomposition pathways are necessary and resilience is linked to metabolic pathway mobility. Many of these degradation pathways are readily transmissible (e.g. 2,4-D, alkane, nitrotoluene degradation) and contained in specific catabolic plasmids. Plasmid mobility allows rapid assimilation and transmission of plasmids (metabolic functions), and changes to function without altering community phylogenetic composition. Bacterial adaptation to xenobiotics appears to be mediated in part by horizontal transfer of mobile genetic elements into a single host that mediates the process and can pass this catabolic pathway on to other microbes once assembled (Top & Springael 2003). Typically, upon stressor removal, plasmid prevalence decreases and function returns to its original state. Specific catabolic plasmids are common among bacteria, but not known to be important in fungi. Instead, fungal responses can involve general C metabolism pathways, such as lignin degradation (Christian et al. 2005).

Soil Systems Biology

In the last decade, systems biology has emerged as a holistic scientific discipline targeted at elucidating emergent properties from large datasets describing highly complex interactions. It aims to understand how complex biological and ecological processes arise from interactions occurring across different scales of biological organisation (e.g. from DNA/RNA/protein through to cells, organisms and up to community structure and function) and environmental features (e.g. from nutrients and soil moisture up to local and global climatic conditions). Systems biology is inherently interdisciplinary (Karsenti 2012) and requires a wide range of approaches. These include graphical network models that integrate high information content data streams and identify putatively interacting variables, multi-factorial experiments to examine the causality and directionality of interactions, and theoretical and simulation models to explore population and community dynamical responses to these interactions.

The methods used to identify associations between variables in large data sets may vary depending on the questions being asked, and ecosystems generally display important traits that need to be considered. For example, complex systems can display nonlinear behaviours with time lags and thresholds (tipping points). Many components may display only weak to moderate coupling (McCann et al. 1998), and there may be built-in flexibility and redundancy (Wright et al. 2012). Further, many species share similar abiotic environments which can lead to correlations and apparent synchrony among non-interacting species (Sugihara et al. 2012).

Network modelling approaches have only rarely been applied to microbial communities [e.g. Chaffron et al. 2010; Barberan et al. 2011; ] or to investigate perturbation (Zhou et al. 2011), but there is considerable scope for network approaches to clarify the roles of microbes and link them to emergent ecological properties of complex systems.

Network analysis has identified extensive phylogenetic and functional trait associations among soil bacteria generally (Barberan et al. 2011) and in response to disturbance (Zhou et al. 2011) and specific antagonistic effects (Prasad et al. 2011). Similar to other systems, work describing microbial network associations suggests they follow a small world, scale-free model which may be visualised as a network with a few nodes displaying high connectedness (‘hubs’) and many nodes with only one or a few connections, such that the network follows a power law distribution (Janssen et al. 2006; Wright et al. 2012). In such networks, high abundance organisms are not necessarily highly interconnected and keystone species are those with high connectivity relative to their abundance or those which provide critical links between nodes nested in otherwise disparate local networks. Thus, resilience is directly related to the degree of disturbance required to ‘fracture’ the service network into disconnected subnets, and functionality may be maintained if a disturbance removes only low connectivity nodes. Resilience is thus reliant on network topology, including network diameter as well as various metrics of network connectivity, for instance, high network density can facilitate rapid recolonisation following disturbance (Janssen et al. 2006). Scale-free networks are generally robust to the random removal of links (Janssen et al. 2006); however, targeted removal of hubs or keystone species will result in rapid fragmentation. Hence, the presence of keystone species may actually decrease resilience (Levin 2001). Interestingly, food-web networks, which have been used for several decades in ecology, and which are based on observations of system trophic interactions and energy flow, may be among the few naturally occurring networks that do not follow the small world, scale-free network topology (Dunne et al. 2002). However, such networks generally have fewer nodes than the types of microbial association network generated using molecular data which we discuss here.

As is the case with other complex biological networks, such as those describing cellular or neurological processes, an understanding of soil association networks requires experimental verification of the strength, direction and reliability of large numbers of potential positive and negative biotic and abiotic interactions (the sum of the soil interactome), a goal which, while not currently feasible (Koch 2012), is one that the field is moving towards. Under laboratory conditions it has been observed that species in mixed communities adapted to perturbation more readily than when individual community components were challenged in isolation (Lawrence et al. 2012). Hence, identification of groups of components that behave as a single functional module will reduce the complexity of associations needed for evaluation, enhancing prospects of gaining a mechanistic understanding of system processes (Koch 2012). Modular redundancy may be identified using network diagrams (Wright et al. 2012), and redundant modules may facilitate a ‘species change-over’ in community composition.

Below, we first illustrate how graphical network analysis can provide additional insights beyond traditional approaches using a specific example related to pollution of salt marshes. We then discuss abiotic and biotic factors that can interactively influence the eco-evolutionary dynamics of plant–soil symbiont associations, suggesting the need for moving beyond pair-wise investigation of such interactions (Bever 2003; Thrall et al. 2007; van der Putten et al. 2007).

Network re-analysis of bacterial saltmarsh communities

Bowen et al. (2011) examined microbial community response to experimental nutrient loading in saltmarsh sediments where widespread ecological responses to pollution had previously been documented. These responses included macro-faunal community shifts and altered ecosystem functions, including increases in%C,%N, bacterial production and alterations in sediment redox chemistry (Bowen et al. 2011 and references therein). However, Bowen et al. (2011) found that microbial community composition was remarkably resistant to increased nutrient loading despite functional changes, suggesting decoupling of microbial community structure and biogeochemical processes.

The statistical methods used by Bowen et al. (2011) examined changes in relative abundance of each operational taxonomic unit (OTU) individually across paired experimental treatment and control samples, or over four time points in temporal experiments. An alternative approach is to examine associations between OTU's and environmental parameters across all samples. Using a subset (approximately 36% of the total community composition) of the same data analysed by Bowen et al. (2011), we identified co-occurrence relationships of bacteria and environmental variables across all marsh sediment samples (Supp. Methods) and generated an association network (Fig. 2a). Here, network nodes are defined as bacteria (OTU based on 16S rRNA gene sequence identity) or environmental factors, and significant co-occurrence relationships define network edges, where OTUs were determined to co-exist with a ‘neighbour’ more frequently than expected by chance across the 24 samples, as defined by the local similarity analysis algorithm (Ruan et al. 2006). Co-occurrence associations could be positive or negative and linear or nonlinear.

The topology of the resultant network informs questions regarding ecosystem function. The network topology conforms to the scale-free model (Barabási & Oltvai 2004; Chaffron et al. 2010) (Fig. 2). Most of the environmental parameters (green triangles) are correlated with each other, but are not correlated with members of the bacterial community (Fig. 2a). Hence, changes in environmental factors would not be expected to influence microbial community structure – the bacterial community appears resistant to environmental changes, as observed by Bowen et al. (2011). We can make some other general observations from the network, providing avenues for further hypothesis testing. For example, (1) levels of N and P fertilisation were only associated with actual measured environmental levels of C and N, not with any microbial node and (2) and rates of N2 fixation were associated with some environmental variables (C:N, N), but were also positively associated with the Chromatiales (known N2 fixers) implicating them as important nitrogen fixers in this system.

Clearly, some of these results could have been obtained from other statistical approaches routinely used in ecology such as permanova and indicator species analysis. However, these approaches do not explicitly identify the biotic linkages between species that we explore here, being more generally used to identify environmental drivers or shifts in community composition. Network analysis does not result in a large reduction of the dimensionality of data as occurs in distance-based statistical approaches, nor does it require a priori knowledge or designation of sample status. Networks allow observations and exploration of the data underlying more traditional methods, and are inherently useful for developing hypotheses and research directions.

The overall degree of connectivity (average number of connections per node) may prove informative about the complexity of biogeochemical transformations or general ecosystem stability. Nodes may have a large number of neighbours but low betweeness if they associate only with their immediate neighbours or module and not with other modules (see below). Hence, betweeness-centrality is an important indicator of which organisms act as intermediaries between groups of other organisms (nodes with high betweeness have large influence on ‘information’ flow through the network). They may have fewer connections, but mediate more associations. Analysis of the network highlighted one node in particular as displaying high betweeness-centrality (Fig. 2b). Although this node, representing candidate bacterial division WS3 (pink diamond), does not have the most neighbours, its position within the network suggests it provides an important link between the connections of many other organisms (Fig. 2c). Interestingly, this node is not one of the most abundant in the network. Given this, it appears possible that WS3, about which little is known, acts as a keystone species in this environment, and highlights the ability of the network approach to reveal the potential importance of numerically non-abundant or cryptic groups.

Modules are highly interconnected network regions with fewer node connections outside the module than inside. Although the entire network in Fig. 2a highlights co-occurrence and not necessarily interactions, modular associations are likely to infer actual interactions (Zhou et al. 2011). Modules could originate from many sources, (resource partitioning, ecological niche overlap, convergent evolution, phylogenetic relatedness) and could be important for system stability (Olsen et al. 2007). We highlight two modules identified from the saltmarsh data set.

The first (Fig. 2d) consists of two sub-modules displaying strong negative co-occurrence between them. These modules represent organisms that appear in distinct subsets of marsh samples, highlighting patchiness in community composition and potentially of functionality. Such patchiness may reflect functional differences between marshes, or it may be a result of modular redundancy, wherein different groups of co-occurring organism maintain a common functionality. The positive relationships between 11 Alphaproteobacteria taxa (red circles) may signify that functionally similar, closely related taxa do not compete with each other in this environment, and potentially act synergistically. This has also been found in the human gut microbiome – where closely related taxa a positively correlated, whilst functionally similar, but distantly related taxa appear to compete and are negatively correlated (Faust et al. 2012). The second module (Fig. 2e) is in fact an extension from members of the first module. It links two groups of organisms involved in both the nitrogen and sulphur cycles. Photosynthetic Rhodospirillaceae and nitrifying Nitrosococcus, are linked with the denitryfying Sinobacteria and several nodes representing the sulphate reducing Desulfobulbacaea.

In network theory, overlap in modular structure, such as that due to species that belong to multiple modules, is a critical feature allowing effective ‘information transfer’ between processes occurring independently. Whether modularity increases resilience in environmental networks has long been a subject of debate, mostly concerning food-web networks, and remains ‘theoretically unclear and empirically controversial’ (Ruiz-Moreno et al. 2006). For example, it has been argued that modularity may act to retain the impact of disturbances within ‘compartments’ (Kokkoris et al. 2002), and also that modularity is favoured as an adaptation to short rather than long-term disturbances (Ruiz-Moreno et al. 2006). In microbial systems, simplifying the vast complexity of interactions through the identification of modular components relating to functionally coherent units is an important step towards predictive models, and can uncover useful biomarkers for effective monitoring. However, even for defined functions, such as nitrification, linking composition to function is exceptionally difficult. Many organisms have multiple modes of metabolism, hence their presence may not be indicative of the same functional trait across multiple samples, and the taxonomic co-occurrence patterns would potentially be misleading. For example, AO can consume ammonia by reducing oxygen (classic nitrification) or by reducing nitrite (nitrifier-denitrification). Further, some AO can opportunistically oxidise methane or live heterotrophically, oxidising suitable organic carbon substrates (heterotrophic nitrification) (Arp & Stein 2003). Thus, it has been suggested that rather than AO communities being regarded as filling a soil ‘niche’ dictated by available nutrients and energy, ammonia oxidation could be thought of as a metabolic ‘niche’ in which how an organism utilise oxidised or reduced N is the critical concept. Rather than considering taxonomic networks as the principal factor determining community response, one can envision a nitrogen gene network and how it links to reduced nitrogen species metabolism that determines community response to disturbance as discussed for general functional gene assemblies (Burke et al. 2011).

We have discussed interactions within microbial communities and how these link to shifts in abiotic factors, yet these communities also exist within higher levels of ecological organisation and thus, interactions with other organisms such as plants can have profound effects on microbial community response to disturbance. Below we focus on plant–soil symbiont interactions and examine impacts on these associations.

Plant–Symbiont Interactions

As discussed above, soil communities can be strongly impacted by anthropogenically driven biotic (invasive weeds, agricultural crops, shifts in plant community) and abiotic (pollution, physical impacts on soil structure) disturbances. Microbial symbionts are no different in this respect, with the exception that in addition to direct ecological and evolutionary impacts of perturbation, shifts in co-evolutionary interactions are of potentially greater significance. For example, a recent review (Kiers et al. 2010) supports the idea that human impacts at all spatial scales are increasingly affecting mutualisms. These include ecological impacts on the abundance and diversity of soil mutualists, as well as evolutionary shifts along the mutualism-parasitism continuum. Given the importance of plant–soil feedbacks as major drivers of ecological function (Bever et al. 2012), it is of concern that we still have little ability to predict how complex interactions between soil mutualists, pathogens and host communities could be altered by perturbation, particularly with regard to evolutionary dynamics or ecological function (Thrall et al. 2011; de Vries et al. 2012). Below, we discuss some factors that may drive shifts in host–symbiont interactions, and use these to further illustrate the need for the development of theoretical and experimental studies that explicitly focus on ecological and evolutionary feedbacks involving plants, soil biota and abiotic factors.

Biotic stressors

In natural systems, plant biodiversity loss via native vegetation clearing or shifts in plant community structure due to weed invasion could alter the relative or absolute abundance of soil mutualists and pathogens (Ceulemans et al. 2011). Experimental studies of above-ground plant-pathogen associations provide empirical support for the prediction that more specific and more aggressive plant pathogens will be favoured in less complex host communities (Mitchell et al. 2003). Recent work also suggests that more complex plant communities likely promote higher abundances of soil bacteria which protect against plant infection by pathogens (Latz et al. 2011). However, the question of how soil symbionts respond generally to shifts in host community structure remains largely unexplored (Thrall et al. 2007).

Crops and invasive weeds represent examples of specific biotic stressors that can alter soil function (Rout & Callaway 2012), and invasion is strongly influenced by anthropogenic activities. While relatively little is known about the eco-evolutionary impacts of invasive species on soil symbionts, there is increasing evidence that plant–soil feedbacks may differ between native and exotic species in invaded habitats (Engelkes et al. 2008; Callaway et al. 2011). Invaders can also degrade mutualisms, and there is evidence for reduced host dependence on mutualists for plant invaders (van der Putten et al. 2007; Seifert et al. 2009). For example, invasive species have been shown to produce allelopathic compounds that can reduce the abundance, diversity and effectiveness of mycorrhizal fungi (Hale et al. 2011; Koch et al. 2011). Ultimately, this may lead to further loss of plant diversity through demographic impacts on mycorrhizal-dependent plants (Hale et al. 2011), with potential consequences for ecosystem function. Invasive plants can also alter soil communities to either suppress (Zhang et al. 2009) or favour the accumulation of generalist pathogens which further impact on native species (Mangla & Callaway 2008), or could co-invade with introduced symbionts (Porter et al. 2011). Plant introductions, including new crop varieties and weeds, could also provide opportunities for the evolution and emergence of soil pathogens (e.g. fusarium wilt in cotton, Wang et al. 2010; soil dieback in invasive weeds, Aghighi et al. 2012). Such situations are likely to occur with increasing frequency, given trends in land-use change and global connectivity (Thrall et al. 2011). In some cases, invasive symbionts themselves may constitute a stressor with potential for both positive and negative consequences where they have been introduced (Schwartz et al. 2006; Rout & Callaway 2012).

Abiotic stressors

In both agricultural and natural systems, increased nutrient loads can select for reduced mutualist effectiveness or even parasitism, as demonstrated for both nitrogen-fixing rhizobial bacteria and mycorrhizal fungi (Kiers et al. 2010; Lau et al. 2012). Consistent with this, modern crop varieties may have inadvertently been selected to be less dependent on soil mutualists, and their introduction can impact symbiont diversity (Xing et al. 2012). Interestingly, analysis of a long-term (50+ years) field study which included high and low fertiliser regimes, found that fungal strains from low fertility treatments grew better, but were less likely to pass benefits on to their hosts (Antunes et al. 2012); consistent with other reports, this study also showed that high fertiliser levels favour more parasitic strains. With regard to soil-borne diseases, there has been considerable theoretical work examining the impacts of environmental productivity on the evolutionary dynamics of pathogen populations ((Thrall et al. 2007) and references therein; (Poisot et al. 2013). One prediction is that increased nutrient inputs might select for more aggressive pathogens, and while the ecological effects of such enrichment are likely to be complex, there is empirical evidence that disease emergence and pathogen impacts increase in such situations (Johnson et al. 2010).

Soil fertility could also mediate more complex interactions involving symbionts and host plants. For example, Sigüenza et al. (2006) showed that mycorrhizal inocula from soils high in nitrogen reduced the growth of a native shrub but not an invasive grass. Such negative feedbacks could have accelerating effects whereby not only is the invasion process facilitated, but further changes in both soil and native plant communities might follow. Similarly, nutrient run-off might increase diversity losses in fragmented ecosystems, which are often already stressed, particularly in situations where native plants have evolved in low fertility soils and may be relatively sensitive to high N and P that may contaminate fragmented non-agricultural areas. In such cases, major effects on soil biota might occur indirectly through reduced diversity of native hosts and increased capacity for weed invasion, ultimately impacting on soil community structure and function. Unfortunately, detailed studies of such interactions across the agro-ecological interface are scarce; there are still almost no data on how environmental degradation via agricultural run-off might alter the abundance of soil mutualists and pathogens.

In agricultural contexts, agro-chemicals and other soil amendments can be important (Bünemann et al. 2006). Thus, while the application of herbicides, fungicides and pesticides has benefits with regard to reducing the impact of weeds, pests and soil-borne diseases, rhizobia and mycorrhizae generally react negatively to agro-chemicals, which can inhibit nitrogen fixation, and mycorrhizal root association (Hussain et al. 2009). Both biological and chemical fumigants can reduce the abundance of soil-borne pathogens, but differ with regard to their impacts on other aspects of soil community structure and function (Omirou et al. 2011). Interestingly in this context, there is also good evidence that soil mutualists can increase plant performance (or at least provide a buffer against adverse effects) in areas of chemical or heavy metal pollution (Glassman & Casper 2012).

Physical disturbances could also impact soil symbionts at various spatio-temporal scales. For example, local increases in soil salinity can inhibit rhizobial nitrogen fixation (Andrés et al. 2012) and fire can reduce both mycorrhizal and rhizobial colonisation of host plants, in some cases favouring exotic weeds over native species (Carvalho et al. 2010). In agro-ecosystems, there is considerable interest in understanding how tillage affects soil systems in the context of soil health and agricultural sustainability. For example, it has been suggested that conservation farming might promote mycorrhizal associations by reducing disruption of hyphal networks, and that symbiont diversity may be higher under conservation farming (Ceja-Navarro et al. 2010). However, temporal components also play a role, with cropping cycles perhaps having greater effects on soil communities than specific levels of physical disturbance (Ng et al. 2012). In general, it is thought that mycorrhizal associations are impacted by, in decreasing order of importance, disruption of hyphae via tilling, chemical-fallow and non-host crops being planted in previous years (Jansa et al. 2006). In the case of soil pathogens, population dynamical models suggest that the impact of different cropping cycles on disease prevalence and pathogen persistence will at least partly depend on pathogen life-history (Thrall et al. 1997).

Overall, while there is evidence that anthropogenic disturbance can have significant impacts on soil symbiont community structure and function there are still considerable gaps in our understanding of how abiotic and biotic factors influence eco-evolutionary feedbacks between plants, pathogens and mutualists. These gaps may only be bridged by considering systems approaches. Even more critically, we still have little ability to predict how shifts in these interactions might in turn influence ecological functions (e.g. nitrogen fixation, disease suppression, water cycle, nutrient cycling/partitioning).

Future directions for research on symbiont response to perturbation include assessing the extent to which the ability to tolerate novel environments correlates with symbiotic properties, and characterising how impacts of perturbation on symbiont community dynamics depend on life histories. How important is nutrient enrichment likely to be in natural systems and would shifts in other trophic interactions involving soil mutualists or above-ground herbivores occur? What are the causal factors associated with development of disease suppressive soils or emergence of new pathogens? Are the eco-evolutionary effects of chemical pollutants primarily mediated by direct effects on soil mutualists or indirectly through changes in plant dependencies?

Conclusions and Future Directions

Our understanding of links between changes in the soil microbial community and the remainder of the soil system are improving, but results are often confounding and specific links remain unclear. For example, large shifts in plant communities are not always reflected by changes in microbial community composition and function (Lamb et al. 2011). Investigators have discovered that some specific microbial community structures are very sensitive to anthropogenic impacts but, how these sensitivities may influence the wider soil system remains elusive.

Progress in system wide understanding is limited by (1) insufficient mechanistic understanding of the ecological, evolutionary and functional consequences of direct and indirect feedbacks between perturbation and soil microbial communities, (2) lack of ability to integrate knowledge of individual ecosystem components across whole ecosystems in ways that enable prediction of eco-evolutionary changes caused by anthropogenically driven impacts, and (3) use of inappropriate definitions and concepts of resilience, largely driven by a compartmentalised approach that has favoured using engineering definitions to investigate change and recovery. Soil community responses depend on many factors, including biotic and abiotic interactions, the level of redundancy of a given soil community component, spatio-temporal factors and the type of perturbation. Mapping soil community structure to function thus remains a central challenge in microbial ecology and will only be possible by considering the myriad interactions and feedbacks that occur in complex systems and including robust assessment of temporal and spatial aspects of community composition and process rates.

Developing a broader, more integrative conceptual framework that encompasses both biotic and abiotic factors will greatly enhance our ability to understand, predict and manage critical aspects of soil microbial communities and the processes they mediate in managed and natural ecosystems. This framework should identify coherent, potentially interchangeable functional modules which relate to one or more biogeochemical pathways and inform how particular environmental variables moderate interactions between modules, how this moderation affects biogeochemical processes and how this feeds back to biotic and abiotic components and ultimately affects ecosystem services. Development of this framework requires moving beyond the engineering definition of resilience to one that encompasses transitional states and the temporal dynamics of soil microbial communities.


AB developed the conceptual framework for the article, and all authors contributed equally to writing the manuscript.