Application of the microbial community coalescence concept to riverine networks

Flows of water, soil, litter, and anthropogenic materials in and around rivers lead to the mixing of their resident microbial communities and subsequently to a resultant community distinct from its precursors. Consideration of these events through a new conceptual lens, namely, community coalescence, could provide a means of integrating physical, environmental, and ecological mechanisms to predict microbial community assembly patterns better in these habitats. Here, we review field studies of microbial communities in riverine habitats where environmental mixing regularly occurs, interpret some of these studies within the community coalescence framework and posit novel hypotheses and insights that may be gained in riverine microbial ecology through the application of this concept. Particularly in the face of a changing climate and rivers under increasing anthropogenic pressures, knowledge about the factors governing microbial community assembly is essential to forecast and/or respond to changes in ecosystem function. Additionally, there is the potential for microbial ecology studies in rivers to become a driver of theory development: riverine systems are ideal for coalescence studies because regular and predictable environmental mixing occurs. Data appropriate for testing community coalescence theory could be collected with minimal alteration to existing study designs.


I. INTRODUCTION
Lotic ecosystems hold some of the richest biodiversity on the globe (Tockner & Stanford, 2002;Dudgeon et al., 2006;Vörösmarty et al., 2010). River systems are particularly unique as hierarchical dendritic networks of ecological corridors, influenced by directional connectivity that links processes across scales longitudinally, latitudinally, and vertically (Ward, 1989;McCluney et al., 2014). Local aquatic communities are influenced by the interaction between local habitat conditions and factors at the regional scale (Heino, 2011), such as interactions with adjacent riparian ecosystems and material and energy transport within dendritic networks from upstream to downstream. These processes are driven by the temporally dynamic nature of river discharge (Poff et al., 1997), maintaining spatio-temporally heterogeneous microbial community structures.
Microorganisms play fundamental ecological roles in rivers: they are largely responsible for decomposition, drive biogeochemical cycles, interact intimately with both macroscopic organisms and the abiotic environment, degrade pollutants and serve as the base of brown food webs. Through their growth and activities microorganisms exert an influence on all the assessment criteria laid out by the Water Framework Directive that determine the ecological status of a river, namely biological, hydromorphological (e.g. Mardhiah et al., 2014), physicochemical and chemical characteristics. Despite their essential functions, microbial groups comprise only a small percentage of lotic ecological studies since publication of the paradigm-shifting River Continuum Concept (Vannote et al., 1980;Tornwall et al., 2015).
Microbial communities occupy many habitats in riverine ecosystems. In the water column, microorganisms can be free-living bacterioplankton or associated with particulate matter or flocs that provide a heterogeneous habitat and a carbon source, and are sites of high activity (Crump, Baross, & Simenstad, 1998). Organisms inhabiting the water column will be passively moved by flows of water, downstream or into the hyporheic zone for example, where they will encounter other microbial communities. Microbial communities also inhabit benthic and hyporheic sediments, typically forming biofilms (Besemer, 2015;Battin et al., 2016).
Metacommunity theory has been applied to explain lotic community assembly from a multi-scale perspective (Tornwall et al., 2015). This framework is concerned with the dispersal of organisms from a regional pool of diversity to local communities (Leibold et al., 2004). Traditionally, it has been applied to fairly distinct local communities connected by the movement of organisms, seeds or propagules. It has the advantage of integrating regional-and local-scale dynamics, but there are limits and challenges in applying this framework to describe the complexity of biotic patterns in riverine systems (Brown et al., 2011;Liu et al., 2013). Some theoretical advances have been made to expand the application of metacommunity theory to dendritic networks (Brown et al., 2011;Altermatt, 2013). For example, dendritic structures drive higher spatial heterogeneity in community composition than lattice-cell structures (e.g. terrestrial and lentic systems) (Carrara et al., 2012). While these advances improve predictions about community assembly in dendritic networks, they are still notoriously unpredictable (Altermatt, 2013). The current theoretical framework may be critically lacking to adequately incorporate the movement and mixing of abiotic factors, which often move over similar spatial and temporal scales as microbial communities, and is restricted to the mechanisms occurring within the watercourse while neglecting interactions with adjacent environments (e.g. environmental mixing across ecosystem boundaries). Rillig et al. (2015) proposed a novel ecological framework for environmental microbial assemblages involving a multi-scale approach appropriate for the complexity of microbial community assembly in lotic systems. The concept, community coalescence, is an emerging paradigm unique to microorganisms that considers the wholesale mixing of multiple communities and their surrounding environments, which can occur on a regular or intermittent basis and within an abruptly short period of time (even within a few seconds). Associated biotic and abiotic fluxes (e.g. flowing water carrying a microbial community, carbon resources, and nutrients) are considered simultaneously as interactive predictors of the observed local microbial community structure.
Thus community coalescence asks the same question as metacommunity theory (how do local communities assemble given fluxes of community members across the landscape?) while taking a metaecosystem perspective [what environmental flows and subsidies (sensu Polis, Anderson, & Holt, 1997) are occurring in this system?] to ultimately understand how, where, and when entire habitat patches, and their associated microbial communities will collide and what resulting environment and community might arise (Fig. 1). Thus it is in line with theoretical work (Logue et al., 2011) that has called for an integration of meta-ecosystem dynamics into metacommunity studies. However, it is noteworthy that immediate wholesale mixing of communities rarely happens to larger organisms and that community coalescence can occur both within a single ecosystem and across ecosystems, distinguishing community coalescence from metacommunity and meta-ecosystem theory, respectively.
Since the publication of meta-ecosystem framework in 2003 (Loreau, Mouquet, & Holt, 2003), it has been applied to riverine networks -particularly to explain carbon dynamics (Battin et al., 2009;Larned et al., 2010). When flows of organisms are included in meta-ecosystem studies, they are often broadly classed as 'organic matter' or 'nutrient sources' rather than described with the phylogenetic and functional precision necessary to understand compositional and trait-space changes following their dispersal -rather this is the territory of metacommunity theory. While meta-ecosystem theory is concerned with dynamic resource flows across the landscape, metacommunity studies often consider abiotic conditions to be static in a local habitat and consider dynamic biotic flows across the landscape. By contrast, the community coalescence concept is concerned with the alteration of both biotic and abiotic components as Fig. 1. A generalized visual scheme for community interactions in space. C i represents different communities (each with a different species composition) and E i represents environmental constituents, consisting of various resources or other abiotic factors. Metacommunity theory (A) envisages that exchanges are a subset of C i (i.e. a few species move by dispersal rather than entire communities), and it does not consider movement of environmental factors (E). On the other hand, during community coalescence entire patches collide and all biotic (C i ) and environmental (E i ) components mix. Depending on the types of environmental pieces that are mixing, communities and environmental constituents may initially contact one another through a 'touching' type interface (B) or in a wholesale mixing scenario (C). This mixing is driven by environmental flows, illustrated by multicoloured arrows in the landscape (blue: hydrological flows; green: biological flows; red: anthropogenic flows; brown: precipitation-driven flows). a result of environmental mixing and in turn, the resultant changes in the microbial community.
The community coalescence concept has been theoretically developed using a top-down (communities as integrated units) modelling approach (Tikhonov, 2016). It has also been experimentally applied in restoration ecology: deterministic processes were found to drive coalescence outcomes in a soil microbial community transplant study (Calderón et al., 2016). An investigation of microbial dynamics in anaerobic digesters found that mixing multiple methanogenic communities increased methane production efficiency, and that the most efficient initial community dominated following community coalescence (Sierocinski et al., 2017). In rivers, consideration of mixing of entire microbial communities has been lacking as an explanation for community assembly, even though interchange between several environmental bodies, including surface water, groundwater, soil, leaves, wastewater plant effluent, and saltwater -and the microbes inhabiting them -occur on a regular basis. Thus, riverine systems are apt for theoretical and experimental application of the community coalescence concept.
Herein we present a holistic view of microbial community assembly in rivers and floodplain habitats through the lens of the community coalescence concept and discuss hydrological, chemical and physical characteristics that drive or could influence the outcome of the coalescence event. First, we discuss community coalescence events along a longitudinal gradient from headwaters to estuaries. The second part takes a cross-sectional view of the river and considers environmental mixing resulting from lateral and vertical flows (e.g. floodplain to/from river and groundwater-surface water interactions). Lotic hydrologists and ecologists have previously utilized similar hydrological frameworks for theoretical work (Vannote et al., 1980;Boulton et al., 1998) and study designs (Savio et al., 2015).
Community coalescence in riverine systems may include a broad range of organisms, including but not limited to prokaryotes, viruses, fungi, diatoms, algae and protists (e.g. Livingston et al., 2013) because of their size in relation to the environmental components that they inhabit. The scope of this review has been limited to only bacteria, archaea and fungi. We were interested in studies investigating microbial community dynamics in zones of mixing in rivers or streams. To target these studies, we conducted a systematic review. First, we made an a priori list of potential mixing zones based on river ecology and hydrology (i.e. headwaters, litter fall, etc.) and searched for field studies investigating microbial community structures in these compartments or during these events (e.g. flooding, litter fall). Search strings used were: bacteria OR fungi OR archaea OR microbial AND river OR stream AND community AND [compartment/event] in Google Scholar; the literature search concluded in December 2016. We excluded review papers, theoretical works and laboratory/mesocosm-based experiments. With this subset of field-based studies, we further applied criteria to determine if the minimum requirements to study community coalescence were met (i.e. reporting the community composition of both initial communities as well as the resultant community) and if so, whether evidence was found supporting the occurrence of community coalescence. Results of the 79 studies (see online supporting information, Table S1) included in this review are presented in Figs 2 and 3. It is noteworthy that few fungal studies met the criteria to make inferences about community coalescence (Fig. 2), and that a sixth of the 79 studies met these criteria and provided evidence for community coalescence (Fig. 3).

II. LONGITUDINAL COMMUNITY COALESCENCE
Microbial community coalescence occurs along the length of the river, across the terrestrial-aquatic interface from upslope soils to headwaters and downstream to estuaries  . + indicates those studies meeting the criteria to study community coalescence and − indicates that criteria were not met. The minimum criteria to make inferences about community coalescence dynamics were that the study reported the community composition of all of the initial communities and the resultant community following mixing.

Fig. 3.
Percentage of studies meeting criteria to assess community coalescence and the degree of evidence provided or not provided by the studies reviewed (N = 79; see Table S1). CC, community coalescence.
( Fig. 4). Coalescence events are conceptually organized from upstream to downstream in this section. This perspective harkens to the meta-ecosystem perspective of the River Continuum Concept (Vannote et al., 1980), which postulates that there is a relationship between carbon processing and biological communities along a longitudinal continuum. In the case of runoff and litter entering low-order streams, microbial communities are concomitantly transported to the freshwater environment. From highly diverse headwaters, these communities shift in composition and become less diverse as they travel downstream (Besemer et al., 2013;Savio et al., 2015).
These shifts may occur in part because of increasingly recalcitrant carbon resources and a changing chemical environment further downstream, but also involve the ability of the component taxa to compete in a freshwater environment. Hydrological events (e.g. large storms, snowmelt) may transport large quantities of particulate organic matter (POM) into riverine systems in a short time period, which is quickly shunted downstream due to increased discharge (Raymond, Saiers, & Sobczak, 2015). Microbial communities are transported along with POM -the concurrent increase in velocity during a pulse-shunt event may allow otherwise uncompetitive

Fig. 4.
A longitudinal view of riverine community coalescence events; from headwaters (A) to estuary (E). The collision of two communities and their inhabiting environments are illustrated by two circles in each box (green and yellow). The brown area represents the mixture of source communities and environmental components, and arrows indicate the direction of community exchange (cf . Fig. 1). GW, groundwater; WW, wastewater; WWTP, wastewater treatment plant. microorganisms protected by particulate matter to survive for quite some distance along the longitudinal gradient. Such a longitudinal perspective also offers a means by which these communities might be studied: because the coalescence events discussed in this section occur in compartments along a longitudinal gradient in a moving water body, to study the resultant microbial community, it is necessary to examine samples upstream and downstream of the site of coalescence and consider the driving force of hydrology.

(1) Community coalescence in headwaters
Microbial community coalescence can occur when microorganisms of terrestrial origin enter rivers and streams via runoff and mix with aquatic microorganisms (Fig. 5). In a permeable system the majority of runoff will first be filtered through soil (Sophocleous, 2002), where many microorganisms will be lost via sorption to soil particles and some gained via detachment (Buchan & Flury, 2004). By contrast, in less-permeable systems (e.g. clay-dominated geology), given the same rainfall event, a proportionally greater volume of water is likely to flow over the surrounding floodplain soil, picking up soil particles and associated microorganisms before reaching a receiving stream or river. Microbial size (Gannon, Manilal, & Alexander, 1991) and cell surface characteristics, as well as soil properties (Huysman & Verstraete, 1993) and the presence of preferential flow pathways in the soil such as macropores (Martins et al., 2013) could affect the size and composition of the transported community. In addition to geology, magnitude and duration of rainfall event, and time of year are also important; greater volumes of runoff will enter streams following high-intensity and/or long-duration storm events especially when antecedent soil moisture deficit is low (Heppell et al., 2002).
Coalescence of terrestrial with freshwater communities may be particularly important in headwaters and low-order streams due to the relatively high proportion of soil-derived water entering the stream. Several recent studies in tundra (Crump, Amaral-Zettler, & Kling, 2012), boreal (Ruiz-González, Niño-García, & del Giorgio, 2015) and temperate (Besemer et al., 2013;Read et al., 2015) ecosystems as well as along the entire length of the Danube River (Savio et al., 2015) have observed considerable abundances of terrestrial microorganisms in headwater microbial communities. Some studies also assessed microbial diversity in soils upslope of the headwaters and found that the majority of taxa in headwaters and downstream sites were first detected in soil or soil water Ruiz-González et al., 2015), providing evidence for the coalescence of terrestrial and aquatic communities.
However, other source communities that coalesce in headwaters, such as those associated with groundwater-fed springs, were not assessed in these studies. Savio et al. (2015) indirectly accounted for the contribution of groundwater-derived microorganisms (via comparison of river-derived microorganisms to a database of groundwater-associated microorganisms) and Read et al. (2015) mentions the contribution briefly. Neither studied groundwater communities prior to coalescence in headwaters directly, and no studies considered litter as an additional source of headwater microbial diversity.
A majority of microbial community members found in higher order streams in these longitudinal studies were also detected in headwaters and upslope soils. Thus, community coalescence in headwaters is a defining event in riverine microbial diversity downstream. Further downstream some new taxa are introduced Ruiz-González et al., 2015), but what largely shapes the community are shifts in abundance of taxa detected in headwaters -those microorganisms that are introduced and survive in the highly diverse headwater environment are likely going to be the major downstream players. Because coalescence in headwaters is such an important event for the microbial biodiversity of the entire river, it is imperative to investigate potential controls on the process of microbial community assembly here. Framing these investigations within the community coalescence concept may lead to new and insightful research questions, for example: how is the headwater community shaped by runoff-to-groundwater mixing ratios? How do anthropogenic activities affect source communities and their propagation downstream?
(2) Terrestrial/freshwater coalescence via litter fall Microorganisms of terrestrial origin can enter lotic systems through litter fall. In northern, temperature systems, the majority of litter fall enters rivers during autumnal senescence Biological Reviews 93 (2018)   and in headwaters and low-order streams (Abelho, 2001). On trees, leaf surface (phyllosphere) communities are dominated by bacteria, and also include archaea, filamentous fungi and yeasts (Lindow & Brandl, 2003). Studies of cultivated plants and tree species have observed Actinobacteria, Firmicutes, Bacteroides and Proteobacteria to be the most dominant bacterial phyla, and bacterial community structure has been shown to correlate with plant host, geographical location and climate (Lindow & Brandl, 2003;Bulgarelli et al., 2013;Meaden, Metcalf, & Koskella, 2016). Following senescence, litter from riparian trees can enter streams and rivers, where coalescence of terrestrial microbial phyllosphere communities and freshwater microbial communities occurs.
While leaves on trees in terrestrial environments are numerically dominated by bacterial taxa, fungi are the most abundant group following stream incubation (Das, Royer, & Leff, 2007;Manerkar, Seena, & Bärlocher, 2008). The best-studied group of leaf-colonizing fungi in running freshwaters is aquatic hyphomycetes; this group of fungi is largely responsible for microbial decomposition of leaf litter in streams (Krauss et al., 2011). Aquatic hyphomycete growth is typically highest at 15-25 • C (Krauss et al., 2011) and reduced diversity has been reported from studies artificially raising stream  and microcosm (Fernandes et al., 2012) temperatures. Stream pH also significantly affects aquatic hyphomycete communities: both species richness and activity are reduced in acidified streams (Krauss et al., 2011;Clivot et al., 2014). A small meta-analysis of studies comparing pairs of control and human-impacted streams reported that eutrophication and mining consistently reduced aquatic hyphomycete richness, while changes in riparian vegetation (e.g. forestry or plant invasion) did not (Lecerf & Chauvet, 2008).
Because of the importance of allochthonous carbon dynamics in many river systems, litter decomposition rates and decomposer communities inhabiting leaf litter are well studied. Investigations of litter microbial communities over time report that fungal diversity and evenness is highest in the first few days following submersion, while one or two fungal decomposer species typically dominate after a few weeks of incubation (Nikolcheva, Cockshutt, & Bärlocher, 2003). One explanation offered to explain this phenomenon is that terrestrial taxa are present during the first days of litter incubation and decline in abundance resulting from competition with aquatic hyphomycetes (Nikolcheva et al., 2003).
It does not appear that coalescence of phyllosphere and freshwater microbial communities has been studied directly: in all studies reviewed here, litter microbial communities were only assessed after days to weeks of in-stream incubation. However, the incorporation of the phyllosphere community into the catchment microbiome was recently suggested (Battin et al., 2016). Further investigating the mechanisms behind community assembly on leaf litter is warranted: when and why do terrestrial taxa persist on leaf litter in streams and what is their ecological function? How do anthropogenic activities affect phyllosphere/freshwater coalescence events, and what are the functional consequences?

(3) Serial discontinuities: anthropogenic/freshwater coalescence events
Community coalescence occurs not only as a result of natural events but also at artificial discontinuous points along the river, such as when dam release and wastewater effluent are discharged into running waters. Both wastewater and reservoirs formed by dams have microbial communities and abiotic conditions that differ from running waters. These point-source discrete influences are considered within the context of the Serial Discontinuity Concept (Ward & Stanford, 1983, 2001, which predicts a decrease in the biodiversity of regulated rivers. Artificial reservoirs frequently associated with man-made dams create a lentic habitat that fragments the river continuity, with inherently different chemical and thermal conditions from the lotic continuum (Friedl & Wüest, 2002). Consequently new bacterial and archaeal lineages, not detected upstream, can appear in the reservoir (Dumestre et al., 2002). Upon dam release, the microbial community in the reservoir water will mix with the lotic community downstream. Physical properties of the dam, such as size and operational rules, may govern whether a unique microbial community is formed in the reservoir (and therefore whether a downstream coalescence event is likely to occur) (Mašín et al., 2003) by determining hydraulic retention time and the associated biophysical alterations (Poff & Hart, 2002).
Another man-made structure that may influence riverine microbial communities is a wastewater treatment plant. The contribution of wastewater effluent to the volume of stream water may range widely and in some cases contributes to 90% or more of instream flow (Brooks, Riley, & Taylor, 2006). The proportion of effluent to stream discharge affects the mixing ratio of the transported (effluent) and receiving (stream) microbial communities, and therefore could influence the outcome of the coalescence event.
Wastewater effluent contains nutrients, such as nitrogen and phosphorous, as well as antimicrobials and pharmaceuticals, that can all impact microbial community composition (Rubin & Leff, 2007;Barra Caracciolo, Topp, & Grenni, 2015) and affect the success of various taxa following mixing.
Two recent studies provide evidence for lotic/wastewater effluent coalescence. In one case, seasonal differences in the outcome of lotic/wastewater effluent coalescence were observed: higher abundances of wastewater-derived taxa were observed in downstream sites in autumn and winter (García-Armisen et al., 2014). In another study, a lineage of ammonia-oxidizing bacteria (AOB) introduced in wastewater effluent represented 40% of all AOB further downstream in the estuary (Cebron et al., 2004). This group of bacteria is responsible for a rate-limiting step in the nitrogen cycle, so this result indicates the potential for coalescence events to affect organisms that drive ecosystem functions. Shifts in benthic microbial communities downstream of wastewater treatment plants have also been observed (Wakelin, Colloff, & Kookana, 2008;Drury, Rosi-Marshall, & Kelly, 2013), but the wastewater effluent community was not assessed in these cases and community differences were attributed solely to physicochemical mechanisms. Thus whether established benthic communities are as strongly influenced as planktonic communities by wastewater effluent communities is an open question.
Studying wastewater effluent/lotic community coalescence could provide insight about how lotic communities with altered functional and biological properties arise. From a public health perspective, the presence of antimicrobial compounds in wastewater may result in the proliferation of organisms containing antibiotic-resistance genes originating from the wastewater effluent community (Kristiansson et al., 2011). The coalescence of wastewater communities and lotic communities could result in the transfer of resistance genes to native lotic organisms. In this context it is important to consider wastewater treatment processes, particularly the use of chlorination or other tertiary treatments, as these will preferentially select resistant taxa and decrease microbial abundance in wastewater effluent. Microorganisms have been shown to survive even chlorination treatment and regrow in environmental waters (Bolster, Bromley, & Jones, 2005).

(4) Estuaries: mixing of freshwater and marine microbial communities
Located at the end of the longitudinal gradient from headwaters downstream are estuaries, where freshwater from rivers and marine waters meet. Mixing of freshwater and marine water creates a salinity gradient exerting a strong influence on microbial community composition (Lozupone & Knight, 2007). Across studies, salinity has been linked to community composition and/or abundance of bacterial (Bouvier & del Giorgio, 2002;Fortunato et al., 2012), archaeal (Crump & Baross, 2000;Webster et al., 2015) and ammonia-oxidizing (Mosier & Francis, 2008) taxa. This type of coalescence is continually occurring, dominantly driven by river inflow and the tide. River inflow relative to tidal intrusion also affects the temporal variability of the salinity gradient (Wolanski, 2007), thus the mixing ratios of the microbial communities and the chemical environment in which they meet are linked to hydrology. Microbial community structure in estuaries is frequently attributed to the mixing of riverine and coastal microbial communities, i.e. community coalescence.
Many studies in estuaries meet the minimum criteria to assess the consequences of community coalescence; namely microbial diversity is frequently assessed in both upstream freshwaters and downstream marine waters as well as in the zone of mixing, i.e. the estuary. Reports suggest that estuarine microbial communities are comprised of members of the upstream and downstream communities (Crump, Armbrust, & Baross, 1999;Crump & Baross, 2000;Crump et al., 2004;Fortunato et al., 2012). Compared to proximal sampling locations, both high and low diversity in the zone of mixing has been reported in studies employing next-generation sequencing (Campbell & Kirchman, 2012;Fortunato et al., 2012). Particle-associated and free-living estuarine communities may behave quite differently in community coalescence situations. For example, free-living bacterial communities were found to be more similar to proximal river and marine (source) samples than particle-associated communities, which the authors attributed to the longer residence time of particles than water in the estuary (Crump et al., 1999). While many studies assess bacterial and archaeal communities in estuaries, to our knowledge, no studies exist in which fungal community composition was assessed in hydrologically connected freshwater and marine waters and the zone of their mixing.  1C). Mixing events resulting in community coalescence are shown with black arrows and text. Chemistry likely to impact the outcome of the community coalescence event is shown in white (Tockner et al., 1999). Hydrological flows driving the events are in blue and physical controls on the hydrology in green. DOC, dissolved organic matter; POM, particulate organic matter.

III. LATERAL AND VERTICAL COMMUNITY COALESCENCE
Processes driven by the longitudinal movement of water downstream were considered in Section II. Now our perspective will shift and consider flows in the remaining two dimensions of the river (Fig. 6). Lateral movement of water onto floodplains and vertical exchange of surface water and groundwater communities in the hyporheic zone will be discussed.
(1) Overbank flooding: a soil/freshwater community coalescence event Freshwater and terrestrial microbial communities mix during overbank flooding. River-derived particulate matter and floodplain-derived soil particles and litter are exchanged, thus community coalescence occurs in both habitats. Floodplain inundations resulting from heavy precipitation are more likely to occur in impermeable or minimally permeable (e.g. clay-rich) systems where precipitation cannot infiltrate the soil before reaching the recipient stream, and river levels rise rapidly in response to rainfall (Sophocleous, 2002). In more permeable systems, rising groundwater and/or growth of autochthonous vegetation (in highly vegetated systems) can displace water onto banks, where aquatic and soil communities likely interact for longer periods of time than in flashier impermeable systems. The flood pulse concept predicts that flood duration and predictability (timing) correlate with organismal adaptations to exploit the aquatic/terrestrial transition zone created by flood pulses (Junk, Bayley, & Sparks, 1989). It is likely that community coalescence outcomes would differ in short and irregularly flooded systems compared to regularly flooded streams with longer flood duration.
Some mesocosm studies have been conducted to investigate the impact of flooding on community composition (Drenovsky et al., 2004;Wilson et al., 2011). Deionized water lacking a natural microbial community was used to simulate floodwater and thus no coalescence event took place in these studies. However, in both cases, shifts in soil microbial community composition were observed, indicating that the chemical and physical impacts of flooding are influential and likely to affect the community coalescence outcome. One manipulative field study investigated the effects of flooding and observed a decrease in microbial biomass and bacterial-to-fungal ratio in the stagnant flooding treatment (Unger, Kennedy, & Muzika, 2009) -conversely, a field study across short-, intermediate-and long-term flooded soils found no fungal marker in the long-term flooded site (Rinklebe & Langer, 2006). While hydrologic connectivity has been reported to influence floodplain microbial activity and microbial community structure (Freimann et al., 2015), further investigation is necessary to resolve the influence of abiotic factors on floodplain microbial communities and to clarify the extent to which microorganisms are exchanged between floodplain and stream.
Studying this exchange by applying the community coalescence framework in floodplains might provide insight into dispersal or survival tactics by particular groups, i.e. some microorganisms may be able to take advantage of floods temporarily moving them out of their 'typical' habitat (e.g. Chauvet et al., 2016). For example, terrestrial sites may serve as a reservoir for taxa whose preferred stream habitat is always moving them further downstream. In one study, aquatic hyphomycete species were identified on leaves submerged within and proximal to a Canadian stream (Sridhar & Bärlocher, 1993). The highest number of species was observed on litter from within the stream, followed by litter in occasionally flooded sites, and the fewest in never-inundated sites. Conversely, typically terrestrial taxa may take advantage of running waters for dispersal, as has been proposed for plants (Gurnell et al., 2007). For example, mycorrhizal fungal spores have been observed in freshly deposited flood sediments (Harner et al., 2011).
(2) Hyporheic zone: freshwater/groundwater community coalescence The hyporheic zone is the interface of surface water and groundwater in riverine systems, and thus a site of vertical exchange between surface water and groundwater microbial communities. At the catchment scale, rivers can be categorized as 'gaining' or 'losing' depending on the net surface water volume gained from or lost to groundwater, and most river systems are dependent to some extent on base flow (upwelling groundwater) (Boulton & Hancock, 2006). On the reach scale, however, the riverbed has both upwelling and downwelling zones, for example in riffle-pool sequences (Hendricks, 1993), termed hyporheic exchange flows. Such flows may shape distinct microbial communities within meters of one another: distinct hyporheic bacterial communities inhabited upwelling and downwelling zones in a gravel bar (Lowell et al., 2009).
Organisms tolerant of low-oxygen conditions will likely be more competitive in zones of upwelling of oxygen-depleted, dissolved organic matter (DOM)-poor and sometimes nutrient-rich groundwater, whereas aerobes and decomposers would have an advantage in zones of intrusion of DOM-rich, oxygenated surface water (Hendricks, 1993;Boulton et al., 1998). A variety of studies have investigated controls on hyporheic bacterial community structure and report that nutrients (Lowell et al., 2009;Febria et al., 2012), carbon quality and availability (Lowell et al., 2009;Febria et al., 2012;Wagner et al., 2014) and temperature (Febria et al., 2012) are influential. Physical filtering of microbial communities controlled by sediment characteristics and bed permeability could affect the density and composition of the transported community in downwelling zones. For example, sediment grain size and conidia morphology have been found to limit fungal dispersal into hyporheic sediment .
To our knowledge, only one study has directly assessed the mixing of surface water and groundwater microbial communities and the resultant community composition in the hyporheic zone (Stegen et al., 2016). The authors attributed compositional changes along the groundwater-to-surface water gradient to both organismal dispersal via hyporheic flows and to changes in the resource environment (e.g. organic carbon concentrations). This study is a good model for further community coalescence studies in the hyporheic zone: samples were collected across different river-stage conditions, organic carbon was well characterized, indices of community compositional change were calculated and mechanistic hypotheses to explain their observations linked ecology, chemistry and hydrology (i.e. flows of resources and communities).

IV. WHY AND HOW TO STUDY COMMUNITY COALESCENCE
Application of the microbial community coalescence concept in riverine systems may help target relevant sampling locations and explain the microbial community assembly by considering the influence of the intermingling of microbial communities from multiple sources within this dynamic environment. In addition to providing valuable ecological information, riverine systems are ideal habitats for testing community coalescence and further theory development.
(1) Why? Structure/function, altered ecosystems, and theory development Attributing variation in community assembly patterns to particular factors is often difficult in the field due to a multitude of concurrently acting environmental factors. The amount of variance between observed communities explained by measured environmental variables is often quite low (i.e. how well the metacommunity paradigm of 'species sorting' explains observed community structure): a recent analysis found the mean variance explained across >300 ecological studies to be 26% and to be particularly low for decomposers (Soininen, 2014). Including mixing-related parameters (e.g. direction and volume of groundwater flow, time since last flood) may improve the amount of variance explained. This may prove easily accomplished as often these hydrological and climactic variables are already monitored. Assessment of other source communities that may mix with stream communities could also improve predictions about how microbial communities will change in time and space along the river continuum. This is important, because ecosystem function is linked with community composition. For example, in estuaries bacterial community composition has been linked to functions including CO 2 production, CH 4 flux, nitrification and enzyme activities (Reed & Martiny, 2013). The functional capacity of a community is dependent upon the traits of the taxa comprising that community, i.e. the trait space. Following community coalescence, the functional trait space of the resultant community may expand or contract, and the maintenance of functions carried out by specialized groups of organisms will depend on the success of those groups following mixing.
The community coalescence framework makes explicit the potential for environmental mixing to serve as an additional driver of microbial community structure, and one that may be particularly important in places where mixing frequently occurs, like in rivers. As a result, application of the microbial community coalescence concept leads to new types of hypotheses and study designs that may improve our ability to make predictions about microbial community structure in the field. At the catchment scale (e.g how microbial communities are assembling and changing from headwaters downstream to estuaries), zones or time periods where/when environmental mixing is occurring may correlate with points in space and/or time where large shifts in microbial communities occur as a result of community coalescence. These points in space/time will be specific to a particular river system, but are often easily observable. Sampling before/after these points may resolve some of the shifts in community structure along the river continuum. For example, a hypothesis that could be posited is that large shifts in microbial community structure along the river continuum will occur following environmental mixing events.
At the reach scale, applying this framework might improve the amount of variance explained between different observed local communities or identify easily measurable environmental parameters that correlate with local community structure. For example, a comparison of application of the species sorting paradigm with community coalescence in the hyporheic zone: in both cases, local communities inhabiting hyporheic sediments would be sampled and local environmental variables of interest (e.g. O 2 concentrations, nutrient status) would be measured. In the species sorting paradigm, the 'metacommunity' would be subjectively defined by the researcher, and may simply include the other local hyporheic communities. Often, variance partitioning is applied to determine the environmental variables predictive of the local observed communities. By contrast, the community coalescence framework explicitly considers the forces of environmental mixing that drive passive dispersal of new microorganisms to these local communities. So, departing from metacommunity theory, these 'source' communities (i.e. surface water, groundwater) would also be sampled, and the driver of the environmental mixing, i.e. hyporheic flows, would also be quantified. Some additional hypotheses that arise could be, for example: (i) the direction and volume of hyporheic flows are predictive of local microbial community structure in hyporheic sediments because they drive passive dispersal of new microorganisms to these sediments and control fluxes in local physiocochemical conditions; (ii) surface water assemblages with characteristics X and Y are competitive in hyporheic sediments and therefore are likely to make up a significant proportion of the hyporheic community in downwelling zones.
Community coalescence may also provide insight into altered ecosystems, a critical topic in the context of global change and increasing anthropogenic pressure on riverine systems. For example, terrestrial fungal taxa were found to make a large contribution to fungal diversity on leaf litter in an acidic, polluted stream compared to an unpolluted stream where typical freshwater taxa dominated (Clivot et al., 2014). This illustrates an unintended consequence of community coalescence: typically the terrestrial taxa present on leaf litter do not persist long on leaves in the aquatic environment; however, in the presence of mining pollution they persisted. Because aquatic fungi are largely responsible for in-stream litter decomposition (Krauss et al., 2011), this could have functional consequences.
In addition to providing a framework for studying microbial communities in both natural and altered riverine systems, the community coalescence concept also provides the opportunity for riverine microbial ecology studies to drive theory development. Riverine systems are uniquely suited for community coalescence studies because of the regular and often predictable mixing that occurs (as compared to mixing that would be present within terrestrial systems). Many studies examine the influence of riverine mixing events on various biological and chemical parameters; therefore, suitable study designs are already employed (e.g. litter bag studies, sampling in mixing zones) or require only slight alteration.
(2) How? Source/resultant communities and methodologies Perhaps more so in rivers than in other habitats (e.g. soils) microbial ecologists are already (and necessarily) considering environmental mixing as it relates to community assembly. However, inhabited pieces of the environment are often considered in a strictly abiotic sense. For example, studies investigating aquatic microbial communities on leaf litter often take the perspective of the leaf as a substrate for decomposition by the aquatic community and do not consider the microbial community associated with that leaf prior to its entry in the stream, or the leaf as a source of microbial cells for the river. In longitudinal studies of microbial communities, hydrology is related to the temporal dynamics of community shifts from headwaters to estuary within the watercourse, but groundwater and upwelling hyporheic samples are often overlooked as additional sources of microbial communities themselves.
To study community coalescence, minimally the source (starting) and resultant (final) communities must be characterized. Methodologies differentiating living/active from dead organisms are critical (e.g. Campbell & Kirchman, 2012), as community changes can occur rapidly in response to changing biotic and environmental conditions following mixing, but imprints of dead organisms very likely persist as environmental DNA (Carini et al., 2016). Additionally, key environmental variables that characterize mixing events should be quantified. Many of these variables are already measured in lotic microbial ecology studies, e.g. mixing ratios, nutrient contents, etc.; however, the composition of source communities is often overlooked. Collecting this additional data would allow testing of hypotheses that explicitly consider how conditions created by environmental mixing govern the composition of the resultant community. Runoff and groundwater mixing ratios, for example, may be the best predictive parameters for headwater microbial diversity because these ratios control the abundance of the source communities as well as the chemistry of the headwaters. This could be tested statistically by calculating community dissimilarity measures among the initial and resultant communities and then determining if the measures are correlated with the relative proportion of mixing and/or the values of abiotic factors of the associated communities. If several headwaters with differing mixing ratios are tested, the importance of this ratio for headwater community structure could be resolved using ordination techniques.
For those compartments where little or no evidence has been reported supporting community coalescence dynamics, field investigations are an appropriate first step. Field studies will elucidate the degree to which community coalescence dynamics may or may not help to explain differences in community variation. Mixing is likely to play a larger role in community assembly in certain compartments than others and some mixing-related predictor variables may be more easily measured or ascertained than others -e.g. wastewater treatment plant effluent volumes versus estimating the amount of litter entering a stream. It is important to ensure that native microbial communities are not altered during manipulative studies in the field -i.e. leaf litter should not be extensively handled or autoclaved before litter bag experiments intending to investigate community coalescence dynamics.
In compartments where evidence for coalescence has been observed in the field (e.g. estuaries, wastewater treatment plant effluent), experimental microcosm studies may be appropriate for testing mechanistic hypotheses. Mixing ratios could easily be manipulated (e.g. to simulate various wastewater effluent volumes) and different source communities (e.g. wastewater subjected to different treatment regimes) could be tested. Mechanisms that could drive changes in ecosystem function following the community coalescence event (e.g. horizontal gene transfer or shifts in functional gene abundance) could be investigated without the confounding effects of the other environmental variables present in the field. Additionally, mechanistic hypotheses about community assembly could be tested, such as the influence of priority effects, mixing ratios (Rillig et al., 2015), and number of communities (Sierocinski et al., 2017) on the resultant community.
Some established ecological theories could provide mechanisms for the outcome of microbial community coalescence events in different lotic habitats. Priority effects explain how an established biofilm might inhibit immigration by members of a transported community by occupying advantageous patches and utilizing carbon and nutrients. Particle-associated microbial communities might be described by nestedness: resulting from proximity and utilization of common resources, the inhabitants of a floc or a leaf are more likely to interact with one another than with other inhabitants of the water column. Oligotrophs may be likely to persist following coalescence in the water column of higher-order rivers, whereas copiotrophs may prevail after freshwater coalescence with nutrient-rich wastewater effluent.
Ecological mechanisms explain part of the story of community assembly following environmental mixing but to get the whole picture we must ask not only how the communities are interacting with one another, but also how their changing environment is acting upon them. What are the resultant chemical characteristics of the new environment created by mixing (e.g. shifts in O 2 , N, or labile carbon concentrations) and does it confer a competitive advantage to a particular source community or subset of organisms (i.e. species sorting)? Has regularly occurring mixing 'primed' one or both source communities for persistence following coalescence?

V. CONCLUSIONS
(1) The recently proposed, microbial-specific community coalescence concept provides a holistic framework to examine microbial community assembly in rivers, departing from the classical metacommunity framework. The ecology of all source microbial communities and the physicochemical properties of the environmental bodies that are mixing are considered in concert, discussed here in the context of the physical and hydrological controls constraining the occurrence of communities and environmental factors in space and time.

Fig. 7.
Conceptual diagram illustrating putative connections between chemistry, hydrology, and temporality in community coalescence events along temporal (x) and spatial (y) axes.
(2) Studies of wastewater treatment plant outflow, as well as research in headwaters, estuaries, and the hyporheic zone provide preliminary evidence for community coalescence. The influence of environmental mixing on flood-influenced habitats and phyllosphere communities has not yet been investigated.
(3) It seems very likely that the hydrological and chemical environment as well as spatial and temporal controls influence the ecological outcome of microbial community coalescence events (Fig. 7), but the process as a whole and the relative contributions of the components in explaining composition and diversity requires more integrated studies. Riverine systems are ideal for the experimental and theoretical development of the community coalescence concept.
(4) This framework may aid in identifying spatiotemporally relevant sampling points where potential community shifts occur. Further, defining the environmental mixing events of interest for the community under study dictates the appropriate spatial scale and timing of sample collection and narrows the scope of physical and chemical parameters of interest.

VI. ACKNOWLEDGMENTS
The authors would like to thank Janis Antonovics and two anonymous reviewers for providing valuable commentary on this manuscript. This work was carried out within the SMART (Science for the Management of Rivers and their Tidal Systems) Joint Doctorate funded with the support of the Erasmus Mundus programme of the European Union.