The conceptual framework for understanding primary succession was initially developed from the study of plant systems, with the earliest work dating to 1685 (Clements 1916), but also grew out of studies of pedogenesis (soil development, e.g. Jenny 1980) and more recently from generalized community assembly theory (Vellend 2010; Nemergut et al. 2013). Plants, microbes and soils share obvious connections, but it was not until the rise of high-throughput sequencing technologies that phylogenetic community profiles of multiple microbial communities could be rapidly inventoried. We now have the opportunity to test ecological hypotheses developed from the study of macroscopic plants and animals, as well as emerging hypotheses specific to microorganisms (Nemergut et al. 2013). Studies of microbial succession are also becoming more important because global warming is causing unprecedented rates of glacial retreat especially in high-elevation and high-latitude environments where the rate of plant colonization is quite slow compared with that of microbes (Fig. 1).
Brown & Jumpponen (2014) studied bacterial and fungal community succession and assembly along a chronosequence of soils created by the receding Lyman Glacier in the northern Cascade Range. By comparing community assembly of bacteria and the fungi in these new soils, they address several specific questions, including: Do bacterial and fungal communities follow similar assembly trajectories along the chronosequence? Does the presence of plants alter the assembly trajectories for bacteria and fungi? The study of Brown & Jumpponen (2014) can also be viewed in light of the debate about whether historical or deterministic factors are more important in controlling community assembly during succession. This debate originated from the deterministic view of Clements (e.g. 1916, 1936) and the more historical view of Gleason (e.g. 1926) and has recently been addressed in studies of microbial systems (e.g. Peay et al. 2012; Ferrenberg et al. 2013). Here, we discuss the results of Brown & Jumpponen (2014) from the perspective of this debate and with respect to the growing body of knowledge about microbial dispersal, biogeography and biogeochemistry.
Perhaps the most interesting finding of Brown & Jumpponen (2014) was that a greater fraction (19%) of fungal OTUs displays nonrandom patterns of occurrence along the chronosequence compared with bacterial OTUs (9.5%). However, when examined as a whole, bacterial communities were more affected by both distance along the chronosequence and vegetation cover than were fungal communities. In addition, bacterial communities converged along the chronosequence, whereas fungal community assembly appeared to be more stochastic (less deterministic) and showed no evidence of convergence towards one community type (Brown & Jumpponen 2014).
Although the data of Brown & Jumpponen (2014) are by no means conclusive, they do hint at different trajectories of succession for bacteria and fungi at this site. If bacterial communities converge on one community type during succession, and fungi do not, this may indicate that bacterial communities assemble in a more deterministic fashion than do fungi. One likely explanation for this pattern is that fungi may be more dispersal limited than bacteria and therefore more prone to historical (stochastic) effects at this site (Fig. 2). Some recent studies indicate that early-successional Betaproteobacteria (e.g. Polaromonas spp.) are not dispersal limited at the global scale (Darcy et al. 2011), whereas larger microbes like algae and zoosporic fungi (both common in periglacial environments) show more divergent biogeographic patterns (and therefore perhaps dispersal limitation) at global and regional scales (De Wever et al. 2009; Schmidt et al. 2011; Naff et al. 2013). Recent modelling studies also support a difference in dispersal capabilities between smaller (e.g. bacteria) and larger (e.g. fungi) microbes. Wilkinson et al. (2012) showed that there is a very low probability that microbes greater than 20 μm in diameter can undergo passive dispersal between continents and that the successful dispersal of small microbes is due to their greater abundance and their longer residence times in the atmosphere compared with larger microbes. Thus, both empirical and theoretical studies point to a more consistent ‘propagule rain’ (sensu Brown & Jumpponen 2014) of bacteria at early-successional sites (Fig. 2).
A more consistent propagule rain for bacteria could reduce or eliminate priority effects for bacterial communities resulting in more deterministic community assembly across the landscape compared with fungi. Other research has shown that priority effects can lead to greater divergence in fungal community assembly (Peay et al. 2012) and are more likely to occur when the rate of propagule input is low (Chase 2003). For example, priority effects determined the ultimate fungal community structure of two ectomycorrhizal species on pine roots (Kennedy & Bruns 2005). Likewise, complex wood decomposing fungal communities can be driven to significantly different successional outcomes by the order of addition of community members (Fukami et al. 2010).
Another potential reason that bacteria and fungi have different early-successional trajectories is that bacteria exhibit a broader range of physiologies than do fungi and thus are more likely to be successful colonists of the oligotrophic, plant-free soils near the glacier terminus. Early-successional bacteria can be photoautotrophs, heterotrophs or chemoautotrophs and many can fix atmospheric nitrogen (Nemergut et al. 2007; Schmidt et al. 2008b; Duc et al. 2009), whereas fungi are all heterotrophs and none can fix nitrogen. Thus, fungi are more dependent than bacteria on fixed sources of carbon and nitrogen and may not have as many available niches before there is significant organic matter build-up during succession. Indeed, it may be that many of the fungi present in recently deglaciated sites are actually dormant (due to lack of organic matter) as originally suggested by Jumpponen (2003), in which case their distribution across the landscape would probably be more stochastic due to the dispersal constraints discussed above. Alternatively, wind-blown particles tend to accumulate in protected pockets, for example next to rocks (Swan 1992), which would cause a very nonuniform (seemingly stochastic) accumulation of organic matter (and therefore fungi) across the early-successional landscape. Brown & Jumpponen (2014) also point out that many of the fungal OTUs at their sites are related to fungi known to be associated with insects. This observation supports the hypothesis of Hodkinson et al. (2002) and Swan (1992) that wind-blown arthropods can be important sources of organic matter in recently deglaciated environments.
As highlighted by Brown & Jumpponen (2014) and other authors, one of the big mysteries surrounding the early stages of microbial succession is determining the source of carbon and energy for the early colonists. In addition to allochthonous (mostly wind-blown) carbon (C) inputs discussed above, the two other major sources of C to early-successional soils are ancient C (Welker et al. 2002; Bardgett et al. 2007; Sattin et al. 2009) and C fixed autochthonously by photo- and chemoautotrophs (Nemergut et al. 2007; Schmidt et al. 2008b). The relative contribution of each of these C sources to the overall pool of C in early-successional soils probably has a major effect on the structure of the resulting microbial communities (Fierer et al. 2010). For example, sites with relatively high levels of ancient C, or high inputs of wind-blown C, might be dominated by heterotrophs early in succession. By contrast, sites with low levels of wind-blown and ancient C would probably be dominated by autotrophs early in microbial succession. As sunlight is abundant at the soil surface before plants colonize periglacial soils, it is logical to assume that photoautotrophs will be important players in most newly deglaciated soils especially in areas with low levels of soil carbon. However, other factors may stunt the development of microbes early in succession, such as limitations of essential nutrients (Yoshitake et al. 2007; Göransson et al. 2011; Schmidt et al. 2012). Given the importance of C and other elements for the development of microbial communities, more information is needed on the geochemistry and aeolian inputs of nutrients to all of the early-successional sites presently being studied by microbial ecologists (Mladenov et al. 2012).
Finally, the relationship between plant and soil microbial communities in early-successional sites should be highly correlated (Blaalid et al. 2012; Knelman et al. 2012) as the nutrient cycling cascades catalysed by microorganisms in plant-free soils are likely to strongly influence the success of plant colonization and because plants add substantial inputs of carbon to the soils through both litter and root exudates. Most importantly, plants form symbiotic relationships with many soil bacteria and fungi and should therefore skew microbial community assembly towards symbiotic heterotrophs. Therefore, it was somewhat surprising that Brown & Jumpponen (2014) found that the presence of plants played a very minor role in the distribution of fungal OTUs relative to bacterial OTUs at the community level. This difference could be explained by dispersal limitations of fungi relative to bacteria as discussed above. Previous studies have shown that inoculum levels for the most common type of plant symbionts, arbuscular mycorrhizal fungi, are often below detection limits in early-successional soils – favouring weedy, nonmycorrhizal plants early in succession (Miller 1979; Schmidt et al. 2008a). However, it is notable that, when examined at the community level, the presence of specific plant taxa had a significant influence on fungal, as well as bacterial composition (Brown & Jumpponen 2014).
It would be interesting to re-examine the data of Brown & Jumpponen (2014) using a nested, multivariate approach to partition the relative influence of distance along the chronosequence vs. plant cover (presence or absence of vegetation as well as plant species type) on bacterial and fungal community composition as well as the relationships between these two groups. Likewise, a deeper exploration of soil physiochemical parameters and/or the use of null deviation analyses could help identify the cause of the difference in patterns of bacterial distribution along the chronosequence. Brown & Jumpponen (2014) posit that bacterial assembly processes are more stochastic very early in succession, but not later in succession, a pattern that has been documented in other systems (e.g. Ferrenberg et al. 2013). However, it is also possible that these seemingly stochastic patterns are actually due to heterogeneity in environmental filters (e.g. soil organic matter as discussed above) and that plant invasion serves to homogenize the landscape, making assembly appear to be more deterministic in older soils.
Overall, the study of Brown & Jumpponen (2014) and other recent studies (e.g. Zumsteg et al. 2012) illuminate the importance of studying more than just one component of the microbial community and point the way to future work needed to understand the patterns they report. Especially needed are studies of the relative dispersal abilities of microbial groups and studies of nutrient levels, inputs and limitations in early-successional systems. New, high-throughput sequencing technologies have allowed us to finally describe the spatial and temporal patterns of microbial communities, but mechanistic studies that isolate the relative importance of the individual drivers of microbial community assembly lag far behind.