Lake microbial communities are resilient after a whole-ecosystem disturbance

Abstract

Disturbances act as powerful structuring forces on ecosystems. To ask whether environmental microbial communities have capacity to recover after a large disturbance event, we conducted a whole-ecosystem manipulation, during which we imposed an intense disturbance on freshwater microbial communities by artificially mixing a temperate lake during peak summer thermal stratification. We employed environmental sensors and water chemistry analyses to evaluate the physical and chemical responses of the lake, and bar-coded 16S ribosomal RNA gene pyrosequencing and automated ribosomal intergenic spacer analysis (ARISA) to assess the bacterial community responses. The artificial mixing increased mean lake temperature from 14 to 20 °C for seven weeks after mixing ended, and exposed the microorganisms to very different environmental conditions, including increased hypolimnion oxygen and increased epilimnion carbon dioxide concentrations. Though overall ecosystem conditions remained altered (with hypolimnion temperatures elevated from 6 to 20 °C), bacterial communities returned to their pre-manipulation state as some environmental conditions, such as oxygen concentration, recovered. Recovery to pre-disturbance community composition and diversity was observed within 7 (epilimnion) and 11 (hypolimnion) days after mixing. Our results suggest that some microbial communities have capacity to recover after a major disturbance.

Figure 1

Temperature and dissolved oxygen profiles of North Sparkling Bog during the year before mixing (2007; a, c) and the year of the mixing manipulation (2008; b, d). Profiles were observed with a hand-held dissolved oxygen and temperature sensor. Mixing began on 02 July 2008 (day −8) and ended on 10 July 2008 (day 0, Mix) when homogenous temperature was observed. Black ticks above panels indicate sampling points.

Figure 2

Changes in measured water chemistry before and after lake mixing. (a) Microbial oxidation–reduction chemistry measured at 4 m depth (hypolimnion) over the mixing experiment. Pre-mixing, Mix and post-mixing time points are on the x axis. (b) Dissolved gas concentrations over the mixing experiment. Daily average carbon dioxide concentration in the surface waters of North Sparkling Bog during the ice-free period of 2008, measured by a carbon dioxide sensor deployed on the lake. (c) Methane (open symbols) and carbon dioxide (closed symbols) profiles for day −9 (pre-manipulation), for (d) Mix (when complete mixing was achieved), and for (e) day 20 (post-mixing). c–e were measured using standard analyses of depth-discrete, manually-collected samples.

Figure 3

Bacterial response to lake mixing, assessed with bar-coded pyrosequencing of the 16S rrn gene V1–V2 region and analyzed for changes in epilimnion (0 m, open symbols) and hypolimnion (4 m, closed symbols) community (a) richness, (b) Pielou's evenness and (c) PD. Error bars on the rarefied values are too small to visualize on the chart; overlap in s.ds. between epilimnion and hypolimnion values occurred on Mix for PD and on day 3 for richness.

Figure 4

Log-normal models fit to ranked taxa abundance to assess changes in the rarity and dominance of members in the (a) epilimnion (0 m) and (b) hypolimnion (4 m). Red arrows below the x axis show the shift in community structure after mixing, and black arrows show recovery. Each line color is a community structure at a different sampling time. In the epilimnion, communities experienced a second, greater shift in structure on day 3, shown by the two red arrows.

Figure 5

Changes in bacterial community composition assessed with pyrosequencing and visualized by non-metric multidimensional scaling analysis (NMDS) of weighted UniFrac similarities in the (a) epilimnion (0 m), (b) hypolimnion (4 m), (c) epilimnion (red lines) and hypolimnion (blue lines). The communities change along trajectories shown by the dashed lines. Orange circles identify the pre-mix time point (day −9), green circles identify the day complete mixing was achieved (Mix), and yellow circles identify the recovery time point (day 20). (d) Differences in weighted UniFrac distance between the epilimnion (0 m) and hypolimnion (4 m) through the experiment. Note that epilimnion and hypolimnion are operationally defined on Mix because strata do not exist when water column temperature is homogeneous.

Figure 6

Phylum-level (and Proteobacterial class level) changes in epilimnion (0 m) and hypolimnion (4 m) composition through the lake mixing experiment. Relative abundances of phylotyped reads are shown, identified to their closest match in the Ribosomal Database Project. Asterisks indicate that no members associated with the phylum were detected. In the x axis, orange circles identify the pre-mix time point (day −9), green circles identify the day complete mixing was achieved (Mix) and yellow circles identify the recovery time point (day 20). See Supplementary Figure 3 for a different perspective of phylum-level dynamics, normalized by total occurrences within each phylum to visualize maximum changes.

Figure 7

Comparison of composition across integrated epilimnion and integrated hypolimnion communities fingerprinted with ARISA, collected 2 days before the mixing manipulation began (day −10) and on the day that complete water column mixing was achieved (Mix). (a) ARISA OTUs detected in and shared across the epilimnion and hypolimnion on day −10 and Mix. Numbers in parentheses show the proportion of detected OTUs belonging to the group. (b) For the epilimnion and hypolimnion Mix communities, the proportion of OTUs that originated in the epilimnion on day −10, OTUs that originated in the hypolimnion on day −10, OTUs that were associated with both layers on day −10, and OTUs that were uniquely observed on the day of Mix (labeled as ‘unknown origin'). (c) Heatmap and cluster analysis of community composition of epilimnion and hypolimnion day −10 and epilimnion and hypolimnion Mix. OTUs are in columns (ARISA OTU IDs are provided along the x axis) and rows are separate communities. OTUs are clustered based on similar abundance and occurrence patterns across the four samples.