Frequency of disturbance alters diversity, function, and underlying assembly mechanisms of complex bacterial communities

Abstract

Disturbance is known to affect the ecosystem structure, but predicting its outcomes remains elusive. Similarly, community diversity is believed to relate to ecosystem functions, yet the underlying mechanisms are poorly understood. Here, we tested the effect of disturbance on the structure, assembly, and ecosystem function of complex microbial communities within an engineered system. We carried out a microcosm experiment where activated sludge bioreactors operated in daily cycles were subjected to eight different frequency levels of augmentation with a toxic pollutant, from never (undisturbed) to every day (press-disturbed), for 35 days. Microbial communities were assessed by combining distance-based methods, general linear multivariate models, α-diversity indices, and null model analyses on metagenomics and 16S rRNA gene amplicon data. A stronger temporal decrease in α-diversity at the extreme, undisturbed and press-disturbed, ends of the disturbance range led to a hump-backed pattern, with the highest diversity found at intermediate levels of disturbance. Undisturbed and press-disturbed levels displayed the highest community and functional similarity across replicates, suggesting deterministic processes were dominating. The opposite was observed amongst intermediately disturbed levels, indicating stronger stochastic assembly mechanisms. Trade-offs were observed in the ecosystem function between organic carbon removal and both nitrification and biomass productivity, as well as between diversity and these functions. Hence, not every ecosystem function was favoured by higher community diversity. Our results show that the assessment of changes in diversity, along with the underlying stochastic-deterministic assembly processes, is essential to understanding the impact of disturbance in complex microbial communities.

Fig. 1

Microbial community dynamics across disturbance frequencies and time, as assessed by 16S rRNA gene terminal restriction fragment length polymorphism (T-RFLP) fingerprinting. a Canonical analysis of principal coordinates (CAP, constrained ordination) plot, with disturbance levels as differentiation criteria, shows cluster differentiation for L0 (CAP1 axis) and L7 (CAP2 axis) from intermediately disturbed levels (L1–6). Disturbance levels: L0 [light-green triangles], L1 [blue upside-down triangles], L2 [light-blue open squares], L3 [open red rhombuses], L4 [purple circles], L5 [black crosses], L6 [green x-symbols], and L7 [blue stars]. b Non-metric multidimensional scaling (NMDS, unconstrained ordination) shows temporal dispersion effect. Days: 14 [open triangles], 21 [light-grey upside-down triangles], 28 [dark-grey squares], and 35 [black rhombuses]

Fig. 2

Process performance indicators across disturbance levels. Effects include temporal changes and trade-offs in community function. a, c Percentage of organic carbon as chemical oxygen demand (COD, black circles) and 3-CA (open purple rhombuses) removal for all levels (negative values represent accumulation). c Biomass as volatile suspended solids (VSS, open green squares). b, d Concentration of ammonium (black rhombuses), nitrite (open blue triangles), and nitrate (open red circles) as nitrogen for all levels. Data are from days 7 (a, b) and 35 (c, d) of the study (for all time points sampled, see Supplementary Figure 2). Mean ± s.d. (n = 3) are shown. Undisturbed L0 replicates had consistent organic carbon removal and complete nitrification, whereas press-disturbed L7 never showed nitrification and had the lowest final biomass. Intermediate levels L1–6 displayed changing functionality with higher s.d. values that increased over time

Fig. 3

Community dissimilarity assessed by principal coordinates analysis (PCO) plots for all disturbance levels on T-RFLP datasets on days a 14 and b 35 of the study. Ovals with dashed lines represent 80% similarity calculated by group average clustering. Disturbance levels: L0 [light-green triangles], L1 [blue upside-down triangles], L2 [light-blue open squares], L3 [open red rhombuses], L4 [purple circles], L5 [black crosses], L6 [green x-symbols], and L7 [blue stars]. c Procrustes analysis on PCO at day 35 comparing metagenomics (circles) and T-RFLP (triangles) datasets. Lines unite data points from the same reactor (n = 24). Same colour palette as for disturbance levels. Tests comparing both methods were statistically significant (Supplementary Table 2). Intermediate treatments’ (L1–6) within-treatment dissimilarity increased with time. L0 and L7 clusters consistently displayed higher similarity after 14 days

Fig. 4

α-Diversity patterns. a Temporal dynamics of Hill number ²D for abundant OTUs, calculated from T-RFLP data across disturbance levels. b Hill number ²D calculated from T-RFLP (black dashed bars) and metagenomics (grey solid bars) data at days 0 (seed) and 35 (disturbance levels L0–L7). c Hill numbers ⁰D (black solid bars) and ¹D (blue solid bars) from metagenomics data on days 0 (seed) and 35 (L0–L7). Values represent mean ± s.d. (n = 3). Characters above bars indicate Games–Howell post-hoc grouping

Fig. 5

Influence of stochastic assembly mechanisms in bacterial communities as assessed by a stochastic intensity and b standard effect size (SES). Both metrics were calculated through null model analysis on the metagenomics genus-level dataset at days 0 (seed) and 35 (disturbance levels L0–L7). Each calculation involved all replicates of each treatment (n_seed = 2, n_L0–L7 = 3) evaluated over 10,000 null model iterations. SES values closer to zero represent communities less deviant from the null expectation, implying stronger stochastic assembly processes. Overall, stochasticity was stronger for intermediate disturbance levels L2–L5 and also increased with respect to the sludge inoculum

Fig. 6

Intermediate stochasticity hypothesis (ISH) for community assembly under varying disturbances. Conceptual representation of the classic relationship between α-diversity and disturbance, including the effect of underlying stochastic and deterministic processes driving bacterial community assembly. When intermediate disturbance regimes result in less predictable environments, specialized traits would be less advantageous to taxa, and the stochastic equalization of competitive advantages would lead to higher α-diversity. On the contrary, extreme ends of the range where conditions are recurrent would select for adapted organisms whose dominance would result in a lower α-diversity