Drivers of microbial food-web structure along productivity gradients

Abstract

Ratios between viruses, heterotrophic prokaryotes and chlorophyll a are key indicators of microbial food structure and both virus-prokaryote and prokaryote-chlorophyll ratios have been proposed to decrease with system productivity. However, the mechanisms underlying these responses are still insufficiently resolved and their consistency across aquatic ecosystem types requires critical evaluation. We assessed microbial community ratios in highly productive African soda-lakes and used our data from naturally hypereutrophic systems which are largely underrepresented in literature, to complement earlier across-system meta-analyses. In contrast to marine and freshwater systems, prokaryote-chlorophyll ratios in African soda-lakes did not decrease along productivity gradients. High-resolution time series from two soda-lakes indicated that this lack of response could be driven by a weakened top-down control of heterotrophic prokaryotes. Our analysis of virus-prokaryote relationships, revealed a reduction of virus-prokaryote ratios by high suspended particle concentrations in soda-lakes. This effect, likely driven by the adsorption of free-living viruses, was also found in three out of four additionally analysed marine datasets. However, the decrease of virus-prokaryote ratios previously reported in highly productive marine systems, was neither detectable in soda-lakes nor freshwaters. Hence, our study demonstrates that system-specific analyses can reveal the diversity of mechanisms that structure microbial food-webs and shape their response to productivity increases.

Keywords: chlorophyll a; eutrophication; heterotrophic bacteria; lysis of viruses; microbial loop; viral shunt.

Figure 1.

(a,b) Trophic biomass pyramids of plankton communities and (c,d) coefficient of variation (standard deviation/mean) of different plankton groups in (a,c) Lake Nakuru and (b,d) Lake Bogoria during the 16 months sampling period (n = 67). Plankton is categorized into primary producers (first level), heterotrophic prokaryotes (second level), picoplanktivores (HNF and small ciliates, which feed on heterotrophic prokaryotes and small algae; third level) and non-picoplanktivorous zooplankton (rotifers and ciliates with a body size of over 60 µm; fourth level).

Figure 2.

Structural equation model (SEM) describing the effects of environmental factors and biological interactions on the density of heterotrophic prokaryotes in two African soda-lakes. Beige boxes represent abiotic variables, the green box chlorophyll a concentration, blue boxes zooplankton and orange boxes heterotrophic prokaryotes and virus-like particles (VLP). Positive and negative interactions are indicated as black and red arrows, respectively. Grey lines represent tested interactions that did not improve model fit. Numbers represent standardized path coefficients and indicate how many standard deviations the response variable changes when the predictor changes by one standard deviation. If two values are stated, lake identity had a significant impact and coefficients represents values for L. Bogoria and L. Nakuru, respectively. ^aOnly marginally significant. ^bModel II regressions of trophic interactions led to slightly higher path coefficients (e.g. 0.40 instead of 0.26 for the relationship between virus and bacteria densities).

Figure 3.

Across soda-lake comparisons (a) between virus-like particles and heterotrophic prokaryotes, and (b) between heterotrophic prokaryotes and chlorophyll a. Dotted lines indicate regression lines, and the shaded area their confidence bands. Whiskers represent the standard deviations of the two soda-lakes, which have been weekly sampled for a 1.5 year period. Slopes of log–log regressions were 1.12 (CI: 0.96–1.28) for (a) and 0.80 (CI: 0.50–1.11) for (b).

Figure 4.

Relationships (a) between the concentration of viruses and heterotrophic prokaryotes, and (b) between prokaryotes and chlorophyll a, in across-system meta-analyses. Each data point represents average values of one ecosystem or distinct habitat. Solid lines represent regressions between log–log-transformed data (logarithmic relationship), dashed lines represent linear relationships. ‘Others’ refers to rare ecosystem types such as e.g. ice or non-soda salt lakes. (c,d) Dependency of the regression slopes on ecosystem type and latitude. Error bars denote 95% confidence intervals. O (others), S (soda-lakes), M (marine systems), F (freshwater habitats), Ar (arctic systems), Te (temperate systems), St (sub-tropical system), Tr (tropical systems).

Figure 5.

Relationships between the virus–prokaryote ratios and turbidity as proxy for suspended particle concentrations in marine systems. (a) Atlantic, latitudinal transect. (b) North Sea. (c) Baltic Sea. (d–e) Northwest Passage. Turbidity was a significant predictor of virus–prokaryote ratios in (a–c) and of virus particle abundances in (e). In all three significant models explaining virus–prokaryotes ratios, variability decreased significantly at higher turbidities (accounted for in GLS-model structure) and ratios showed a logarithmic decrease with turbidity.

Figure 6.

Relationship between the depth of the euphotic zone (dotted red line), the mixing depth (whole water body in polymictic shallow lakes), and the phytoplankton primary production per lake volume. Green arrows indicate decreasing lake levels and red arrows changes in the euphotic zone due to decreases in chlorophyll a.