Interactions between hydrology and water chemistry shape bacterioplankton biogeography across boreal freshwater networks

Abstract

Disentangling the mechanisms shaping bacterioplankton communities across freshwater ecosystems requires considering a hydrologic dimension that can influence both dispersal and local sorting, but how the environment and hydrology interact to shape the biogeography of freshwater bacterioplankton over large spatial scales remains unexplored. Using Illumina sequencing of the 16S ribosomal RNA gene, we investigate the large-scale spatial patterns of bacterioplankton across 386 freshwater systems from seven distinct regions in boreal Québec. We show that both hydrology and local water chemistry (mostly pH) interact to shape a sequential structuring of communities from highly diverse assemblages in headwater streams toward larger rivers and lakes dominated by fewer taxa. Increases in water residence time along the hydrologic continuum were accompanied by major losses of bacterial richness and by an increased differentiation of communities driven by local conditions (pH and other related variables). This suggests that hydrology and network position modulate the relative role of environmental sorting and mass effects on community assembly by determining both the time frame for bacterial growth and the composition of the immigrant pool. The apparent low dispersal limitation (that is, the lack of influence of geographic distance on the spatial patterns observed at the taxonomic resolution used) suggests that these boreal bacterioplankton communities derive from a shared bacterial pool that enters the networks through the smallest streams, largely dominated by mass effects, and that is increasingly subjected to local sorting of species during transit along the hydrologic continuum.

Figure 1

Schematic representation of potential large-scale spatial patterns in bacterioplankton taxonomic composition across freshwater networks, where dots represent individual bacterioplankton communities in an ordination space, based on their compositional dissimilarity, and colors indicate the different catchments or geographic regions to which they belong. Owing to the movement of the water in the landscape, the composition of local bacterioplankton communities within a single catchment or region will differ depending on their position along the hydrologic continuum (horizontal axis), as they will be differentially affected by hydrology that regulates the mass effects versus environmental sorting ratio (see Introduction section for further explanation), represented here by the light to dark color gradient. As a result, the compositional variation because of local conditions (vertical axis) will likely increase along the hydrologic continuum because of a gradual intensification of the local sorting of species with increasing WRT. Over broader spatial scales, the increase in environmental differences will likely result in stronger environmentally driven dissimilarities between communities (that is, along the vertical axis), but the overall large-scale spatial patterns may further differ depending on the degree of dispersal of microbes between regions or catchments: For example, in a case of low dispersal between regions (a), a regional structuring of the communities may be detected even in systems with little or no environmental sorting (that is, headwater streams) because the regional bacterial pools are different due to dispersal limitation. In contrast, under a scenario of high dispersal between regions (b), a common bacterial pool will lead to little environmentally driven compositional differences in the headwaters but to a increasing differentiation of communities toward downstream systems (that is, lakes) because of stronger local sorting of species. From this perspective, whereas communities located at the headwaters of the aquatic continuum will mostly reflect the nature of the regional bacterial pools because of strong mass effects, communities located further downstream in the network will be the result of species sorting by the local aquatic conditions. The interpretation of the factors and mechanisms underlying the observed spatial patterns in bacterial community composition will thus depend on the portion of this complex space that is considered (dotted areas, scenarios 1, 2 and 3). For example, studies based on a narrow spatial scale, as depicted in scenario 1, may conclude that hydrology is the main driver of changes in bacterioplankton community composition. In contrast, studies that cover broader environmental gradients but are limited to certain portions of the hydrologic continuum (scenarios 2 and 3) may conclude that either dispersal limitation (2) or local sorting (3) shape local community assembly. We argue that a cross-regional, whole network approach is necessary to disentangle the influence of the mechanisms and factors that are actually influencing the local assembly of bacterioplankton communities in complex freshwater networks.

Figure 2

(a) Distribution of sampling sites across the seven sampled regions in Northern Québec (Canada). (b) Principal component analysis (PCA) of the sites based on the measured environmental and geographic parameters. Different regions are indicated by different colors and symbols indicate rivers (triangle) or lakes (circles). The two first axes explain 44.5% of the variance. Chl a, chlorophyll a; Cond, conductivity; DIC, dissolved inorganic carbon; %C2, percentage of fluorescent component C2 (humic-like DOM); %C5, percentage of fluorescent component C5 (freshly produced labile DOM); Elevation, mean catchment elevation; Mean T, mean annual temperature; NPP, net primary productivity; Precipitation, mean annual precipitation; Runoff, mean annual runoff; %Forest/%Shrubland/%Wetland, %forest/shrubland/wetland covered area in the catchment; Temp, water temperature; TP;TN, total phosphorus and nitrogen.

Figure 3

NMDS ordination of bacterioplankton communities based on the Bray–Curtis dissimilarity of community composition (stress =0.13). Shape indicate rivers (triangles) or lakes (circles), and sites are colored according to geographical region (a) and water pH (b), which was the variable that best fitted the ordination space (R²=0.61, see Table 1). The arrows indicate the direction at which the environmental and hydrological vectors fit the best (using envfit function) onto the NMDS ordination space. The size of the arrow is proportional to the strength of the correlation of each variable. AWC, area of water in the catchment; CA, catchment area; Cond, conductivity; DIC, dissolved inorganic carbon; %C2, percentage of fluorescent component C2; %C5, percentage of fluorescent component C5 (freshly produced labile DOM); T, water temperature.

Figure 4

Relationship between geographic (a), environmental (b) and hydrological (c) distances and pairwise bacterial community dissimilarity (Bray–Curtis). pH and WRT (log10-WRT) were used as proxies of environmental and hydrological distances, respectively. Mantel correlations (R) an the probabilities are provided for each case.

Figure 5

Changes in OTU richness (a), Pielou's evenness (b) and taxonomic composition (at the phylum level, c) for sites grouped according to their position on the NMDS1 axis (G1 to G7, n=174, 74, 37, 18, 26, 39, 18), which mostly represents an hydrologic continuum from the smallest headwater streams (G7) to larger rivers and lakes (G1, for details see Results section). Dots are means and error bars represent the standard error of the values for the sites within each NMDS1 group. Colors indicate different phyla (and different classes within the Phylum Proteobacteria, indicated by the dashed line) and the heights of the bars represent the percentage of sequences associated to each taxonomic rank relative to the total number of sequences within each group of sites.

Figure 6

(a) Variation in the R coefficient of the Mantel correlations between the taxonomic and the environmental (pH) dissimilarity matrices for sites grouped according to their position on the NMDS1 axis (G1 to G7). Note that G1 includes most lakes and the largest rivers, and G7 (G1 to G7, n=174, 74, 37, 18, 26, 39, 18) contains mostly small headwater streams (for details see Results section). All the correlations were significant (P<0.01). (b) Changes in bacterial community composition along the NMDS1 axis as a function of WRT. The NMDS1 scores of the sites were binned into 16 equal groups (n=20) based on ranked WRT. The dots are the average NMDS1 scores within each bin and represent the mean position of the sites within a given bin on the NMDS1 axis. Error bars are the standard errors for either WRT and scores within each bin. Note that increases in WRT result in gradual changes in taxonomic composition along the NMDS1 axis until a WRT of 10 days is reached, above which increases in WRT do not translate into further changes in taxonomic composition along the NMDS1 axis. (c) Percentages of variation in the NMDS ordination space explained by the environment and hydrology, for sites with WRTs above and below 10 days (WRT>10 and WRT<10, respectively). Non-explained and shared variation (environment+hydrology) are also shown. The ‘environment' category included the variables DIC, dissolved inorganic carbon; %C2, percentage of fluorescent component C2 (humic-like DOM); %C5, percentage of fluorescent component C5 (freshly produced labile DOM) and conductivity. T, water temperature; CA, catchment area; AWC, area of water in the catchment; WRT, and d-excess were included within the ‘hydrology' category.