Discordant Temporal Turnovers of Sediment Bacterial and Eukaryotic Communities in Response to Dredging: Nonresilience and Functional Changes

ABSTRACT

To study the stability and succession of sediment microbial and macrobenthic communities in response to anthropogenic disturbance, a time-series sampling was conducted before, during, and 1 year after dredging in the Guan River in Changzhou, China, which was performed with cutter suction dredgers from 10 April to 20 May 2014. The microbial communities were analyzed by sequencing bacterial 16S rRNA and eukaryotic 18S rRNA gene amplicons with Illumina MiSeq, and the macrobenthic community was identified using a morphological approach simultaneously. The results indicated that dredging disturbance significantly altered the composition and structures of sediment communities. The succession rates of communities were estimated by comparing the slopes of time-decay relationships. The temporal turnover of microeukaryotes (w = 0.3251, P < 0.001 [where w is a measure of the rate of log(species turnover) across log(time)]) was the highest, followed by that of bacteria (w = 0.2450, P < 0.001), and then macrobenthos (w = 0.1273, P < 0.001). During dredging, the alpha diversities of both bacterial and microeukaryotic communities were more resistant, but their beta diversities were less resistant than that of macrobenthos. After recovery for 1 year, all three sediment communities were not resilient and had reached an alternative state. The alterations in sediment community structure and stability resulted in functional changes in nitrogen and carbon cycling in sediments. Sediment pH, dissolved oxygen, redox potential, and temperature were the most important factors influencing the stability of sediment communities and ecosystem multifunctionality. This study suggests that discordant temporal turnovers and nonresilience of sediment communities under dredging resulted in functional changes, which are important for predicting sediment ecosystem functions under anthropogenic disturbances.

IMPORTANCE Understanding the temporal turnover and stability of biotic communities is crucial for predicting the responses of sediment ecosystems to dredging disturbance. Most studies to date focused on the bacterial or macrobenthic community, only at two discontinuous time points, before and after dredging, and hence, it was difficult to analyze the community succession. This study first compared the stabilities and temporal changes of sediment bacterial, microeukaryotic, and macrobenthic communities at a continuous time course. The results showed that discordant responses of the three communities are mainly related to their different biological inherent attributes, and sensitivities to sediment geochemical variables change with dredging, resulting in changes in sediment ecosystem multifunctionality.

INTRODUCTION

Stability is a crucial factor determining the ability of an ecosystem to maintain functions under changing conditions induced by anthropogenic disturbances (1). Ecosystem stability includes two components: resistance, the property of communities to withstand disturbance; and resilience, the ability of communities return to the original state or reach a new stable state (2). Rapid and significant changes in the structure and functions of freshwater and marine ecosystems caused by human-induced disturbances are of major concern (3). Nutrient cycles and pollutant clearing up in freshwater ecosystems mainly occur in the sediment surface layer through biological metabolic processes (4, 5). The microbial communities can be responsible for 76% to 96% of the total sediment respiration, and the communities are generally known as a dominant driving force in the transformation of complex organic compounds (6, 7). In addition, the majority of macrozoobenthos living at the top 30 cm of sediment also contribute to processes of mineralization through respiration (8, 9). Thus, it is important to study the stability and functional changes of both microcommunities and macrocommunities in response to human-induced disturbance in sediment. Dredging is widely used to improve water resources. Such large-scale anthropogenic disturbance profoundly affects water quality by elevating turbidity, releasing nutrients and toxins from the sediment, and changing the sediment biological system (10–14). For instance, dredging significantly reduces the abundance and diversity of the sediment macrobenthic community, and it takes months to years for a community to reestablish (8, 14–16). Only a few reports are available on the effects of dredging on the bacterial community. A significant increase from 8,200 to 37,000 CFU/ml in the maximum viable bacterial population was observed after a dredging event (17). Among the active members found after dredging, the numbers of Deltaproteobacteria and Spirochaeta decreased, and the members of the classes Gammaproteobacteria and Flavobacteria represented the most dominant growing populations (18). After dredging, most of the extracellular microbial enzymes were significantly depressed (11). However, in these studies, the samples were collected discontinuously at two time points before and after dredging; therefore, it is difficult to analyze the succession and the stability of the bacterial community. Furthermore, little attention has been paid to uncultured microeukaryotes in sediments, despite their crucial roles in the ecosystem (19). It is still unknown whether bacterial, microeukaryotic, and macrobenthic communities exhibit similar responses to dredging, as well as whether they contribute to the resulting ecological functional changes.

Species temporal turnover is defined as the number of species eliminated and replaced per unit of time (20–22). It can be quantified by regression of the changes in community similarity over time, termed the time-decay relationship (TDR) (23–27). Understanding the temporal changes of different biomes in the environment is important in community succession research. A recent meta-analysis of TDRs in aquatic species assemblages across ecosystems indicated that the degree of species turnover in time is jointly driven by several ecological, physical, and geographical factors in aquatic ecosystems, and that the turnover is inconsistent across taxonomic groups (27). Hatosy et al. (20) determined that the temporal turnover of marine bacterial communities is time scale dependent. However, the variations in the temporal turnovers of different sediment communities with dredging disturbance and the underlying mechanisms remain unknown.

To study the stability and functions of sediment communities in response to dredging disturbance, sediment samples were taken continuously over the time course before, during, and 1 year after dredging in this study. The microbial community structures of all 55 samples were analyzed by sequencing bacterial 16S rRNA and eukaryotic 18S rRNA gene amplicons with Illumina MiSeq technology, and the macrobenthic community was identified using a morphological approach simultaneously. This study aimed at addressing three questions. (i) What were the temporal patterns of sediment bacterial, microeukaryotic, and macrobenthic communities in response to dredging disturbance? (ii) Did both sediment microbial and macrobial communities show resistance and resilience to dredging? (iii) What is the linkage between the stability of sediment biotic communities, environmental attributes, and functions of the sediment ecosystem in carbon and nitrogen cycling?

RESULTS

Changes in geochemical variables.

The changes in sediment geochemical variables from the five monitoring sections before, during, and after dredging are shown in Table S1. Sediment pH showed a clear increase during and after dredging, from pH 7.0 to 8.14 at 1 year later. The dissolved oxygen (DO) levels and oxidation-reduction potential (E_h) increased dramatically through dredging and then decreased gradually to 1.43 ± 0.27 mg · liter⁻¹ and −12.80 ± 54.91 mV, respectively. Some fluctuations in nitrate (NO₃⁻-N), ammonium nitrogen (NH₄⁺-N), organic matter (OM), and available phosphorus (P) concentrations were also observed.

Changes of alpha diversity of sediment communities.

Sediment microbial communities and the macrobenthic community were identified at different time stages by high-throughput sequencing and morphotaxonomic inventories, respectively. Overall, the community richness and Shannon-Weiner diversity values are highest for bacteria, for microeukaryotes, and the lowest for macrobenthos in sediment (Fig. S1). Dredging disturbance increased the richness and Shannon diversity of bacterial communities. In contrast, the richness and Shannon diversity of microeukaryotic and macrobenthic communities decreased during dredging, and then a slight increase was observed 1 month after dredging.

The compositions of bacterial and eukaryotic communities changed dramatically at the phylum or class level (Fig. 1). For the bacterial communities, Proteobacteria were dominant through all the samples (Fig. 1a). The relative abundance of Deltaproteobacteria increased significantly, whereas Alphaproteobacteria abundance decreased 1 month after dredging. The abundance of Epsilonproteobacteria increased during dredging but decreased soon after the disturbance. Dredging also altered the abundance of other bacterial groups. Chloroflexi showed a continuous increase since the late period of dredging (day 41). In contrast, Bacteroidetes decreased immediately with dredging, were stable 2 weeks after dredging, and then decreased continuously. For the microeukaryotic community, a clear transition was observed between the community at dredging and 1 month after dredging (Fig. 1b). Alveolata dominated before dredging, with a relative abundance of 24.5% to 62.2% and an average abundance of 46.3%. Then, during dredging, stramenopiles became dominant at 26.8% to 91.9%, with an average abundance of 65.5%. Opisthokonta dominated 1 month after dredging, at 12.5% to 92.8%, with an average abundance of 48.4%.

FIG 1.

(a) Bacterial community composition (derived from Illumina sequencing analysis of 16S rRNA gene amplicons). (b) Microeukaryotic community composition (derived from Illumina sequencing analysis of 18S rRNA gene amplicons) of the 55 sediment samples (S1 to S55).

The top 50 bacterial and eukaryotic operational taxonomic units (OTUs) were visualized through hierarchical cluster analysis (Fig. S2). The samples before day 68 (S1 to S30) were well grouped together, with Acinetobacter, Cloacibacterium, and Sulfuricurvum spp. being dominant. The samples after day 68 (S31 to S55) clustered together, with Moraxellaceae and Thiobacillus spp. being dominant. For microeukaryotic dominant OTUs, the samples during and within 2 weeks after dredging (S6 to S30) grouped together, with Pseudodendromonadales being dominant. The samples before dredging (S1 to S5) were similar to those 1 month after dredging (S31 to SS55), with Annelida and Chytridiomycota being dominant.

Temporal turnovers of sediment communities.

The slopes of bacterial, microeukaryotic, and macrobenthic time-decay relationships (TDRs) were estimated through linear regression of log-transformed community similarity with time (Fig. 2). Overall, significant regression slopes were observed for all three communities. The temporal turnover of microeukaryotes (w = 0.3251, P < 0.0001) was the highest, followed by that of bacteria (w = 0.2450, P < 0.0001), whereas macrobenthos exhibited the lowest succession rate (w = 0.1273, P < 0.0001). In contrast, the bacterial community in the adjacent Beishi River, without dredging, showed no significant TDRs (Fig. S3). The bacterial and eukaryotic TDRs at the phylum/class level were further estimated (Table 1) to obtain detailed insights into the temporal change of different taxonomic groups subject to dredging disturbance. All bacterial phyla showed significant temporal turnover, with considerable variations among different phyla. Nitrospirae (w = 0.3086, P < 0.0001) and Firmicutes (w = 0.2883, P < 0.0001) showed the steepest temporal turnovers, followed by Spirochaetes (w = 0.2621, P < 0.0001) and Proteobacteria (w = 0.2469, P < 0.0001). In contrast, temporal turnovers of Acidobacteria (w = 0.0872, P < 0.0001) and Actinobacteria (w = 0.0607, P < 0.0001) were the lowest. All classes of the phylum Proteobacteria also showed significant temporal turnover, of which classes Epsilonproteobacteria (w = 0.5543, P < 0.0001) and Gammaproteobacteria (w = 0.4191, P < 0.0001) represented the highest sensitivity to the disturbance. Simultaneously, all phyla of the eukaryotic community except Archaeplastida (w = 0.0084, P = 0.662) also showed significant TDRs. The temporal turnover of stramenopiles (w = 0.5798, P < 0.0001) was the highest, followed by that of Alveolata (w = 0.2555, P < 0.0001) and Opisthokonta (w = 0.1945, P < 0.0001). Bicoecea (w = 1.1292, P < 0.0001), Bacillariophyta (w = 0.3205, P < 0.0001), and Basidiomycota (w = 0.2871, P < 0.0001) were the most sensitive.

FIG 2.

Time-decay curves for sediment microbial and macrobenthic communities. The slopes of all lines are significantly less than zero and significantly different by pairwise comparison.

TABLE 1.

Temporal turnover of bacterial communities and microeukaryotic communities among different phylogenetic groups

Phylogenetic group	w	P	SDa
Bacterial communities
Acidobacteria	0.0872	<0.0001	0.0169
Actinobacteria	0.0607	<0.0001	0.0134
Bacteroidetes	0.2132	<0.0001	0.0145
Chlorobi	0.1241	<0.0001	0.0237
Chloroflexi	0.1181	<0.0001	0.0095
Cyanobacteria	0.1778	<0.0001	0.0306
Firmicutes	0.2883	<0.0001	0.0177
Nitrospirae	0.3086	<0.0001	0.0417
Planctomycetes	0.1383	<0.0001	0.0200
Proteobacteria	0.2469	<0.0001	0.0115
Alphaproteobacteria	0.3276	<0.0001	0.0147
Betaproteobacteria	0.1551	<0.0001	0.0083
Gammaproteobacteria	0.4191	<0.0001	0.0209
Deltaproteobacteria	0.2793	<0.0001	0.0198
Epsilonproteobacteria	0.5543	<0.0001	0.0438
Spirochaetes	0.2621	<0.0001	0.0338
Verrucomicrobia	0.1419	<0.0001	0.0217
Eukaryotic communities
Alveolata	0.2555	<0.0001	0.0283
Amoebozoa	0.1319	<0.0001	0.0309
Archaeplastida	0.0084	0.662	0.0189
Excavata	0.1284	<0.0001	0.0275
Opisthokonta	0.1945	<0.0001	0.0225
Ascomycota	0.1883	<0.0001	0.0228
Basidiomycota	0.2871	<0.0001	0.0234
Chytridiomycota	0.0193	0.235	0.0165
Cryptomycota	0.0687	<0.0001	0.0183
Nematoda	0.1659	<0.0001	0.0413
Rhizaria	0.1026	<0.0001	0.0271
Stramenopiles	0.5798	<0.0001	0.0386
Bacillariophyta	0.3205	<0.0001	0.0331
Bicoecea	1.1292	<0.0001	0.0556
Chrysophyceae-Synurophyceae	0.0684	0.005	0.0263
Labyrinthulea	0.1294	<0.0001	0.0322

^aObtained by bootstrapping (999 times).

The overall distribution patterns of sediment communities were visualized on the first two coordinates of the nonmetric multidimensional scaling (NMDS) ordination based on the Bray-Curtis distance (Fig. S4). The difference among the samples was tested through analysis of similarities (ANOSIM) (Table 2). Dredging significantly altered the bacterial and microeukaryotic community compositions (P < 0.0001), and a minimal effect was posed on macrobenthos (P = 0.054). A clear transition was observed 1 month after dredging. The bacterial and microeukaryotic communities were quite different from their original state (P < 0.0001). The dissimilarity continuously increased for the bacterial community, as quantitatively estimated by statistic R of ANOSIM (Table 2). In contrast, the dissimilarity for the microeukaryotic community was the highest shortly after dredging and decreased later.

TABLE 2.

Dissimilarity test of the community structures in response to dredging via ANOSIM analysis^a

Time around dredging by community	R or P value by time around dredging
	Before	Dredging	1–2 wk	1–1.5 mo	2–3 mo	1 yr
Bacteria
Before		0.0001	0.0004	0.0006	0.0004	0.0106
Dredging	0.762		0.0767	0.0001	0.0001	0.0001
1–2 wk	0.691	0.107		0.0001	0.0002	0.0002
1–1.5 mo	0.981	0.945	0.831		0.0002	0.0005
2–3 mo	0.981	0.986	0.901	0.413		0.0033
1 yr	0.952	0.987	0.949	0.667	0.442
Microeukaryotes
Before		0.0002	0.0004	0.0006	0.0006	0.0191
Dredging	0.822		0.8464	0.0001	0.0001	0.0001
1–2 wk	0.932	0.062		0.0001	0.0002	0.0002
1–1.5 mo	0.945	0.973	0.986		0.0002	0.0066
2–3 mo	0.552	0.902	0.932	0.542		0.0073
1 yr	0.700	0.956	0.979	0.394	0.413
Macrobenthos
Before		0.054	0.0057	0.004	0.0012	0.7222
Dredging	0.237		0.291	0.0736	0.2378	0.2852
1–2 wk	0.430	0.023		0.6395	0.6091	0.0392
1–1.5 mo	0.531	0.104	0.033		0.084	0.0091
2–3 mo	0.674	0.034	0.025	0.090		0.0047
1 yr	0.088	0.053	0.254	0.456	0.532

^aThe ANOSIM statistic R is based on the difference of mean ranks between groups and within groups. The values in italics are the R values, and those without italics are the P values. Significant P values (≤0.05) are indicated in boldface.

Resistance and resilience of sediment communities.

The resistance (RS) and resilience (RL) of the alpha diversity (richness) of the sediment communities were further estimated (Fig. S5). During dredging, the three communities showed different tolerance to the disturbance. The alpha diversity of both bacterial and microeukaryotic communities was more resistant (RS, >0.5) than that of the macrobenthic community (RS, ≈0.25) (Fig. S5a). Bacterial and eukaryotic phyla also represented different sensitivities to dredging. The beta diversity (community similarity) of the bacterial phyla Betaproteobacteria, Bacteroidetes, Cyanobacteria, and Gammaproteobacteria showed higher resistance (RS, 0.5 to 0.8), whereas Epsilonproteobacteria, Verrucomicrobia, Chloroflexi, and Acidobacteria showed the lowest resistance (RS, <0) (Fig. S5b). The eukaryotic phyla Alveolata, Archaeplastida, and Rhizaria were most resistant to the disturbance (RS = 0.3 to 0.8) (Fig. S5c).

After the disturbance, the alpha diversity of bacterial, microeukaryotic, and macrobenthic communities was nonresilient 1 year later (Fig. S5a), indicating an alternative state of community. The change in the bacterial community was the greatest (RL, −0.61), followed by that in the microeukaryotic community (RL, −0.33), and then the macrobenthic community (RL, −0.14). Taxon groups exhibited differential resilience. The majority of the bacterial phyla were nonresilient (RL, <0), except for Epsilonproteobacteria (RL, 0.61) and Firmicutes (RL, 0.09) (Fig. S5b). Similarly, the majority of eukaryotic taxon groups were nonresilient. Only the phyla Excavata (RL, 0.42), Amoebozoa (RL, 0.12), Basidiomycota (RL, 0.08), Cryptomycota (RL, 0.06), and Chytridiomycota (RL, 0.003) were partially recovered (Fig. S5c).

Functional changes in nitrification, denitrification, and respiration.

Sediment nitrification potential, denitrification potential, and respiration activities were determined through cultivation methods to investigate the impact of dredging on the ecological functional processes. Sediment nitrification potential increased during dredging, from 38 NO₃⁻ g/(kg · day) to an average of 45 NO₃⁻ g/(kg · day). Sediment nitrification potential decreased after dredging, with some fluctuations, and reached approximately 20 NO₃⁻ g/(kg · day) 1 year later (Fig. 3a). In contrast, dredging significantly decreased sediment denitrification potential, from 107 NO₃⁻ g/(kg · day) to the lowest level of 78 NO₃⁻ g/(kg · day). Sediment denitrification potential recovered gradually to the original state, approximately 109 NO₃⁻ g/(kg · day).

FIG 3.

(a) Nitrification potential and denitrification potential in sediment (mean ± standard deviation, n = 5). The different letters above the bars indicate significant differences (P < 0.05). (b) Respiration rates in sediment based on the mineralized carbon.

Sediment carbon mineralization activity was highest before dredging, with a carbon mineralization rate of 11.27 g/(kg · day) (Fig. 3b and Table 3). Dredging dramatically decreased carbon mineralization activity, from 4.70 to 6.28 g/(kg · day). The carbon mineralization rate within 2 months after dredging (from days 49 to 68) was still as low as that during dredging, which ranged from 4.85 to 7.32 g/(kg · day). Then, the carbon mineralization rate gradually increased to 10.94 g/(kg · day) 1 year later.

TABLE 3.

Exponential regression of sediment carbon mineralization^a

Time (day) with regard to dredging	C0k (g/kg of sediment/day)b	C0 (g/kg of sediment)	C1 (g/kg of sediment)	k (liters/day)	r2	P
Before (0)	11.27 A	375.65	2.89	0.030	0.999	<0.001
During (15)	4.70 B	187.90	1.27	0.025	0.998	<0.001
During (30)	6.28 B	149.46	1.64	0.042	0.998	<0.001
During (41)	6.06 B	147.69	2.40	0.041	0.997	<0.001
After (49)	4.85 B	131.00	6.00	0.037	0.998	<0.001
After (58)	5.11 B	222.22	0.69	0.023	0.997	<0.001
After (68)	5.94 B	129.16	0.99	0.046	0.998	<0.001
After (86)	7.32 B	170.15	1.73	0.043	0.996	<0.001
After (106)	6.73 B	172.66	1.46	0.039	0.998	<0.001
After (136)	9.83 AB	178.76	2.89	0.055	0.998	<0.001
After (365)	10.94 AB	195.31	5.90	0.056	0.998	<0.001

^aRespiration activity was indicated by the initial potential rate of carbon mineralization (C₀k).

^bThe difference in respiration activity, indicated by different letters, is significant at a P value of <0.05.

Linkage between geochemical variables and sediment community structure and functions.

Canonical correspondence analysis (CCA) was performed to link the relationship between geochemical attributes (i.e., pH, OM, NH₄⁺, NO₃⁻, P, SO₄²⁻, DO, Eh, and temperature) and sediment community structure (Fig. 4). Similar to the NMDS ordination results, the samples were clearly separated before, during, and after dredging, particularly for bacterial and microeukaryotic communities. Of all the environmental attributes measured, pH, DO, E_h, and temperature were the most significant factors influencing bacterial community structure (P = 0.005), E_h, DO, NH₄⁺, and pH were strongly correlated with microeukaryotic community structure (P = 0.005), and pH and temperature were closely related to the patterns of the macrobenthic community (P = 0.01).

FIG 4.

CCA of bacterial communities (a), microeukaryotic communities (b), and macrobenthic communities (c) and geochemical variables that were significantly related to community variations: pH, organic matter (OM), nitrate (NO₃⁻), ammonium nitrogen (NH₄⁺), available phosphorus (P), sulfate radical (SO₄²⁻), dissolved oxygen (DO), redox potential (E_h), and temperature (Temp).

We further fitted structural equation modelings (SEMs) to investigate the direct and indirect effects of pH, DO nutrients (NH₄⁺-N, NO₃⁻-N, and P), temperature, and sediment community alpha diversity (Shannon index) and beta diversity (community similarity) on ecosystem multifunctionality (EMF) (Fig. 5). Overall, the SEMs demonstrated that EMF was directly and indirectly influenced by pH, DO, nutrients, temperature, and alpha diversity and beta diversity of sediment bacterial, microeukaryotic, and macrobenthic communities. pH controlled EMF both directly and indirectly by strongly affecting nutrients and the alpha diversity and beta diversity of sediment bacterial, microeukaryotic, and macrobenthic communities (Fig. 5a to c), and it was the strongest predictor of such attributes for bacteria (Fig. 5a). DO was a significant parameter influencing EMF directly and indirectly through pH, alpha diversity, and beta diversity. The nutrients were found to be the strongest direct driver of EMF. The alpha diversity of both bacterial (standardized regression weights, r = 0.161, P < 0.001) and microeukaryotic (r = 0.081, P < 0.001) communities positively impacted EMF, but the alpha diversity of the macrobenthic community showed a negative effect (r = −0.119, P < 0.001). The beta diversity of both bacterial (P = 0.057) and microeukaryotic (P = 0.774) communities exhibited no significant impact on EMF, while the macrobenthic community beta diversity showed a significant positive effect (r = 0.078, P = 0.001). Temperature showed the strongest effect on macrobenthic community alpha diversity and beta diversity, with r = −0.458 (P < 0.001) and r = 0.249 (P < 0.001), respectively. Temperature also significantly affected bacterial alpha diversity, with r = 0.355 (P < 0.001).

FIG 5.

Structural equation models of environmental properties, alpha diversity, and beta diversity as predictors of ecosystem multifunctionality (EMF). (a) Bacteria. (b) Eukaryotes. (c) Benthos. Solid red arrows represent positive paths (P < 0.05), solid black arrows represent negative paths (P < 0.05), and dotted gray arrows represent a nonsignificant path (P > 0.05). The path coefficients are indicated as standardized effect sizes of the relationship. Temp, temperature; DO, dissolved oxygen; EMF, ecosystem multifunctionality.

DISCUSSION

Predicting how the functions and composition of biotic communities respond to disturbance is a central challenge in environmental biology research. Dredging significantly influences sediment communities in freshwater ecosystems (8, 11, 13). In this study, the diversity, structure, stability, and functions of sediment bacterial, microeukaryotic, and macrobenthic communities in response to dredging were investigated simultaneously. An increase in the richness and Shannon diversity of the bacterial community and a decrease in both microeukaryotic and macrobenthic communities occurred. This phenomenon can be explained by an abundance that is higher and a wider distribution of bacteria living in the sediment compared to eukaryotic organisms. The majority of the macrofaunal community lives at the top 30 cm of surface sediment, and thus the reduction in macrobenthic species abundance and diversity is directly related to the removal of substrates by dredging (9, 15, 28). In our study, the composition and structure of sediment communities were also significantly altered by dredging disturbance, particularly for bacteria and microeukaryotes. A clear transition was observed 1 month after dredging, and no recovery was observed 1 year later. The difference can be ascribed to the alterations in sediment geochemical attributes. Oxygen status is the most important factor that strongly regulates cell metabolism in sediments (19). Dredging disturbance causes a direct increase in sediment DO and E_h (Table S1), resulting in the stimulation of aerobic populations and repression of anaerobic groups. For example, a significant negative correlation was observed between the bacterial phylum Deltaproteobacteria and DO and E_h (r = −0.55, P < 0.001) (Table S2), of which the strictly anaerobic families Desulfobulbaceae, Geobacteraceae, and Syntrophaceae were dominant. In the eukaryotic community, DO was positively correlated with Bicoecea (r = 0.589, P < 0.001) but negatively correlated with Bacillariophyta (r = −0.376, P = 0.005). The results further proved that oxygen was a key driver of protist community structure (19), which creates an apparent phylogenetic dichotomy between oxic and anoxic assemblages (29–31). In addition, pH and seasonal temperature variation were important factors influencing sediment community structure (32). CCA results further indicated that the main controlling factors were DO, pH, E_h, and the seasonal variations in temperature in this study (Fig. 4). Furthermore, variations in temperature, DO, pH, and nutrients were confirmed to directly, and indirectly through sediment community biodiversity, influence EMF (Fig. 5). Another explanation for the obvious changes in community structures before and after dredging can be caused by succession, given that the upper 10- to 20-cm sediment layers were completely removed by dredging. Therefore, the community structure after dredging may reflect a new niche recolonized by the immigration and recruitment (33) of the biological community from the water column and the surrounding environment.

The temporal turnover of TDRs is a reliable indicator of the succession dynamics of biological communities (26). In this study, microbial communities showed a higher succession rate than the macrofaunal community in response to dredging disturbance. A quantitative analysis of temporal turnover in aquatic species assemblages indicated that the degree of temporal turnover was related to organism characteristics, given that larger organisms with high mobility showed slower temporal turnover than small organisms (27). In contrast to large benthos, microbes exhibit extremely high cell densities, small body size, and high generation time. The rate of the life cycle is directly influenced by the body size through metabolic constraints (34). In addition, small organisms probably show fast fluctuations in population dynamics because they present large species pools from which local sites can be rapidly colonized, thereby exhibiting fast turnover (27, 35). Microbial communities are able to adapt to changing environmental conditions at very short time scales for their unique metabolic and genetic capabilities (36). The high succession rate of sediment microbes may be relatively important to maintain ecosystem functions under disturbance.

The turnover is not uniform across taxonomic groups. Different subsets of the microbial community showed varied temporal turnover in response to dredging disturbance. The possible explanation for the considerable variations of temporal turnovers can be deterministic fitness differences between taxonomic groups (37). The ecological succession largely attributes to more or less niche-based deterministic development of community structure after perturbations (38). Disturbances resulted in changes in various environmental factors. The abiotic factors caused different selection in determining microbial taxon assemblage structure and diversity. Changing the environmental conditions after disturbance may also alter the interspecies interactions (e.g., competition, predation, mutualisms, and tradeoffs) (39–41), thereby resulting in considerable variations in microbial taxon assemblage.

Considering the resistance and resilience of sediment community in alpha diversity and beta diversity, functions were further estimated to reveal the ecosystem stability. During disturbance, the alpha diversity of both microbial communities was more resistant, but the beta diversity was less resistant than that of macrobenthos. The reduction in the richness and diversity index values of benthic communities because of dredging was generally reported in previous studies (8, 14, 15). In contrast, microorganisms usually feature high population density, and more individuals with versatile physiologies contributed to the high resistance of bacterial and microeukaryotic communities in alpha diversity (42). Both the alpha diversity and beta diversity of sediment bacterial, microeukaryotic, and macrobenthic communities were nonresilient 1 year after the disturbance. Qian et al. (13) also reported that the dredged cores were sensitive to disturbance, and recovery of the microbial community was not observed after 15 months. Growth rates, population sizes, and dispersal abilities play important roles in community recovery after disturbance (43). Disturbance can kill or inactivate local resident taxa, and empty niches can be created. Dispersal can facilitate the dissemination of disturbance-tolerant taxa among localities, and these empty niches can be filled, thereby altering the fundamental membership of a community and reducing the overall community resilience (44). The new sediment community may form interactions with overlying water and be laid down in the period postdredging. However, the current study could not determine it. In future work, the sedimentary age of the material, or the first cores by depth, should be examined and then see if the new community exhumed by dredging is sourced from deeper in the initial sediments, or if they form from interactions with overlying water. The resistance and resilience of microbial taxonomic groups were differently affected by dredging. This phenomenon may be explained well by their individual ability to accommodate and exploit environmental changes. Mixotrophy, or the ability to use various energy and nutrient resources in taxonomic groups, provided different support for individual flexibility in fluctuating environments (45, 46). Therefore, physiological plasticity and stress tolerance were likely to contribute to different taxonomic resistance and resilience responses.

Resistance and resilience responses to disturbances may vary between functional microbial groups, depending on their level of functional redundancy (47). A high-order species performing the same function within a functional community may act as a buffer against the influence of biodiversity loss on functioning (48). The change in community composition is not strictly accompanied by a change in function provided by the communities after the disturbance (47, 49, 50). In the present study, although the mineralization of organic matter was considerably decreased during dredging, it nearly recovered to the original state. A similar trend was observed for the microbe-mediated denitrification process. In contrast, nitrification function was only half of the original function. Denitrifiers were less affected than nitrite oxidizers after disturbance (51). Given that denitrifiers are more diverse than nitrite oxidizers, functional redundancy was hypothetically high for denitrifiers that reduced the impact of their decreased diversity (47).

In conclusion, bacterial, microeukaryotic, and macrobenthic communities were discordantly affected by dredging. The temporal turnover of microeukaryotic community is the highest, followed by that of bacteria, and that of macrobenthos community is the lowest. This variation is related to different sensitivities to sediment geochemical variable changes and biological inherent attributes, such as different generation times, body sizes, and their respective nutrient demands and capability in limited resource competition. Given the varied stress sensitivities and tolerances, different taxonomic groups exhibited various responses and succession rates to dredging. All communities were nonresilient, and another alternative biotic state was reached. The ecological processes of carbon and nitrogen cycling in sediments altered differently. This study showed different temporal turnovers and nonresilience of sediment communities and functional changes in the sediment ecosystem in response to dredging disturbance, which provides new insights into the impacts of dredging on the fundamental temporal changes, the stability of sediment communities, and the prediction of sediment ecosystem functions under anthropogenic disturbance. However, in addition to bacterial, microeukaryotic, and macrobenthic communities, further studies of the archaeal community are needed in future work to make more comprehensive explanations for predicting the responses of sediment ecosystems to dredging disturbance.

MATERIALS AND METHODS

Site description and sample collections.

The 6-km-long, 17.5- to 35.8-m-wide Guan River connected to the JingHang Canals is one of the main urban rivers in Changzhou, Jiangsu Province, China (119°93′E and 31°78′N) (Fig. S6), which plays an important role in city flood control and ecological water diversion. Before the dredging, the accumulation of sludge in the river amounted to 130,000 m³ for long-term sedimentation and no navigation, seriously affecting flood flow and the water environment. In order to improve the water conservancy to meet flood in the flood season, a dredging project was performed with cutter suction dredgers from 10 April to 20 May 2014 and took 41 days. During dredging, although not all sampling sections were dredged at the same time, they were all disturbed as stirring in a fluidic system. The dredged part was 2,602 m long and about 2.5 m deep, and approximately 91,920 m³ of sediment was dredged. Five monitoring sections (Fig. S6, points A to E) distributed evenly along the river were selected as the biological replicates along the Guan River from upstream to downstream. The top 20-cm surface sediment was sampled using grab samplers. The sampling was performed once before dredging (day 0) on 9 April 2014, three times during dredging (days 15, 30, and 41), and seven times after dredging (days 49, 58, 68, 86, 106, 136, and 365). A total of 55 surface sediment samples were collected. At the same time, we used the same method to collect sediment samples in the adjacent Beishi River (Fig. S6, points a to e) without dredging and sequenced bacterial 16S rRNA gene amplicons as a baseline. The Beishi River is connected to the upstream section of the Guan River (Fig. S6). They are part of the Grand Canal net and exhibited similar environmental attributes. Each sample was divided into three subsamples: a subset of fresh sample was used for the sediment macrobenthos analysis, the second subset was stored at 4°C to measure the sediment geochemical attributes within 3 days after the samples were collected, and the third subset was stored at −80°C for bacterial and microeukaryotic community analyses. For macrobenthos analysis, each fresh sample was washed in a sieve with a mesh size of 0.5 mm, and the retained fraction was preserved in 4% neutral formalin solution. In the laboratory, the material from each sieved sample was carefully sorted, and the macrofauna was identified using a binocular microscope (×16 magnification) (52). Sediment geochemical attributes were measured as follows. NO₃⁻-N and NH₄⁺-N were measured using the Kjeldahl method (53). P was extracted with sodium bicarbonate and determined using the molybdenum blue method (54). Sulfate radical (SO₄²⁻) was measured through an indirect titration method using ethylenedinitrilotetraacetic acid disodium salt. OM content was determined through the dichromate oxidation method (55). The pH, DO, and E_h were recorded by a calibrated Myron L Ultrameter (Myron L Company) in situ. Fresh sediment samples were cultured aerobically in 250-ml flasks for 2 weeks at 28°C with the addition of 200 ppm NH₄⁺-N; concentrations of NO₃⁻-N were measured before and after the incubation to determine the sediment nitrification potential (56). The sediment denitrification potential was estimated through anaerobic incubation at 28°C for 3 days in sealed 250-ml flasks filled with N₂; the consumption of NO₃⁻ was calculated with an initial addition of 50 ppm NO₃⁻-N (57). Sediment C mineralization activity was estimated by the potential of organic carbon mineralization with aerobic incubation at 25°C, using an alkali absorption method during the incubation days 1, 3, 6, 10, 15, 21, and 28 (58). The exponential regression model was used to describe soil carbon mineralization as follows (58):

$$C_t=C_0(1-e^{-kt})+C_1$$

where t is the time from start of incubation, C_t is the mineralized carbon at time t, C₀ is the maximum potentially mineralized carbon, and C₁ is the easily mineralized carbon.

Illumina sequencing analysis of bacterial 16S rRNA and eukaryotic small subunit rRNA gene amplicons.

Microbial genomic DNA was extracted from 2 g of well-mixed sediment for each sample by combining freeze-grinding and sodium dodecyl sulfate for cell lysis, and then purified by agarose gel electrophoresis, followed by phenol-chloroform-butanol extraction, as previously described (59). The primers 515F (5′-GTGCCAGCMGCCGCGGTAA-3′) and 909R (5′-CCCCGYCAATTCMTTTRAGT-3′) targeting the bacterial V4–V5 region in the 16S rRNA genes (60), and TAReuk454FWD1 (5′-CCAGCASCYGCGGTAATTCC-3′) and TAReukREV3 (5′-ACTTTCGTTCTTGATYRA-3′), targeting the eukaryotic V4 region in the 18S rRNA genes (61), were selected. The forward and reverse primers were tagged with adapter, pad, and linker sequences. Each barcode sequence (12-mer for bacteria and 80-mer for eukaryotes) was added to the reverse primer for pooling of multiple samples in one run of MiSeq sequencing. PCR amplification was performed in triplicate with a GeneAmp PCR-system 9700 (Applied Biosystems, Foster City, CA, USA) in a total volume of 25 μl, which contained 2.5 μl of 10× PCR buffer II and 0.5 units of Herculase II DNA polymerase high fidelity (Agilent, USA), 0.4 μM each primer, and 10 ng of template DNA. PCR was performed to target the 16S rRNA gene using cycling conditions of initial denaturation at 94°C for 3 min, followed by 30 cycles at 94°C for 40 s, 56°C for 60 s, and 72°C for 60 s, with a final extension at 72°C for 10 min. The PCR products from the three replicates were combined and purified with an Agencourt AMPure XP kit (Beckman Coulter, Brea, CA, USA). The PCR products were examined by electrophoresis with 1% agarose gel and quantified by PicoGreen with FLUOstar Optima (BMG Labtech, Jena, Germany). The PCR products were pooled from different samples in equal amounts, purified using a Qiagen gel extraction kit (Qiagen Sciences, Germantown, MD, USA), in accordance with the manufacturer's instruction, and requantified by PicoGreen. The amplification steps for the eukaryotic 18S rRNA were similar to those of the 16S rRNA, except for the changes in the PCR conditions, in which the initial denaturation was at 95°C for 2 min, followed by 25 cycles at 95°C for 20 s, 55°C for 20 s, and 72°C for 45 s, and termination with extension at 72°C for 3 min.

The sequencing samples were prepared using the TruSeq DNA kit in accordance with the manufacturer's instructions. The purified library was diluted, denatured, rediluted, and mixed with PhiX (equal to 30% of final DNA amount), as described in the Illumina library preparation protocols. Afterward, the purified library was applied to an Illumina MiSeq system for sequencing with the reagent kit version 2, 2 × 250 bp, as described in the manufacturer's manual (Illumina, San Diego, CA, USA).

After assigning each sequence to its sample according to its barcode, 6,139,308 reads for bacteria and 2,899,731 reads for eukaryotes from both ends were obtained for all 55 samples. The sequence data were processed using the QIIME pipeline version 1.7.0 (http://qiime.org/). All sequence reads were trimmed and assigned to each sample on the basis of their barcodes. The sequences with high quality (length >200 bp, without ambiguous base N, and average base quality score of >30) were used for downstream analysis. Operational taxonomic units (OTUs) were clustered using the recently introduced program UPARSE (62) at a 97% similarity level (63). Final OTUs were generated using the clustering results, and taxonomic annotations were assigned to each OTU's representative sequence by RDP 16S Classifier (64) and the SILVA database (https://www.arb-silva.de/) for bacterial and eukaryotic classifiers, respectively. Singletons were removed for downstream analyses.

TDR, community resistance and resilience, and other statistical analyses.

We used linear regression to examine the relationship between the temporal distance among samples and the similarity in microbial and benthic composition. Bray-Curtis distance was used as a taxon-based metric of differences in community composition. An Arrhenius (log-log) plot was used to model the species-time relationship in the form ln(S_s) = constant − w ln(T), where S_s is the pairwise similarity in community composition and T (day) is the time interval of sampling the two compared samples. w and constant were the regression coefficient and intercept of simple linear regression of ln(S_s) with ln(T), respectively, and w is a measure of the rate of log(species turnover) across log(time).The significance comparison of w values among different estimations was also achieved by bootstrapping (999 times), followed by a pairwise t test.

The resistance (RS) and resilience (RL) of sediment communities during dredging (days 15, 30, and 41) and after dredging (days 49, 58, 68, 86, 106, 136, and 365) were calculated by comparing the alpha diversity (richness) and beta diversity (community similarity) between samples predisturbance and postdisturbance. The indices of resistance and resilience of community richness measured are calculated as follows (1):

$$\text{RS}=1-\frac{2\left|D_0\right|}{(C_0+\left|D_0\right|)}$$

$$\text{RL}=\frac{2\left|D_E\right|}{\left(\left|D_E\right|+\left|D_x\right|\right)}-1$$

where D₀ is the difference in community richness between the control (C₀) (in this study, C₀ was calculated at day 0) and the disturbed sediment on days 15, 30, and 41 during the disturbance, D_E is the difference in community richness between C₀ and the sediment at the end of the disturbance (day 41), and D_x is the difference in community richness between the control (C_x) and the sediment samples at seven time points after dredging (days 49, 58, 68, 86, 106, 136, and 365). RS is bounded by −1 and +1, with a value of +1 showing that the disturbance exerted no effect (maximal resistance) and low values demonstrating strong effects (less resistance). RL is also bounded by −1 and +1. A value of +1 at the time of measurement indicates full recovery (maximal resilience), whereas low values indicate a low rate of recovery. An index value of 0 indicates that the disturbed sediment has not recovered at all since the disturbance ended. A negative value may occur when the disturbance initially reduces the response variable being measured. However, this value also increases substrate availability, which subsequently causes a large increase in the response variable.

The resistance and resilience of community beta diversity were calculated using dissimilarity test statistics. The dissimilarity test of the community structures in response to dredging was performed with ANOSIM (65). The ANOSIM statistic R is based on the difference of mean ranks between groups and within groups and calculated as follows:

$$R=\frac{r_b-r_w}{\frac{1}{4}\left[n(n-1)\right]}$$

where r_b is the mean rank of between group dissimilarities, r_w is the mean rank of within group dissimilarities, and n is the total number. According to the ranks of the dissimilarities, statistic R ranges from −1 to +1. Ecological communities rarely achieve an R value of <0. Thus, an R value of ≈0 indicates no difference among groups, and an R value of >0 shows that groups differ in community composition. In this paper, we use 1 − R to indicate the resistance and resilience of beta diversity of the sediment communities during and after dredging, respectively.

Hierarchical cluster analysis of bacterial and eukaryotic genera was performed using a pairwise average-linkage clustering algorithm (66). The microbial and macrobenthic community distribution patterns were determined by NMDS (67). CCA was used to identify the effects of sediment geochemical variables on the microbial and macrobenthic community composition (68). Univariate patterns of association were tested using Pearson correlation coefficients (69). The significance of differences in nitrification potential, denitrification potential, and sediment C mineralization activity was tested using Duncan's multiple-range test at a P value of <0.05 after one-way analysis of variance (ANOVA). All the above-mentioned analyses were performed in the R program (version 3.0.2; http://www.r-project.org/) with the vegan and ecodist packages.

EMF, and direct and indirect effects of environmental parameters and biodiversity on EMF.

The EMF index was calculated from sediment C mineralization activity, nitrification, and denitrification potential, with an averaging approach (70). The Z-scores for the measured variables were averaged to obtain an EMF for each sample.

To evaluate direct and indirect effects of environmental parameters and biodiversity on EMF, we constructed SEMs including pH, dissolved oxygen (DO), temperature (temp), nutrients, α diversity (Shannon index), β diversity (community similarity), and EMF. E_h was not included for its redundancy with DO, and OM was not included for weak correlations in the CCA results. The nutrients were obtained as the first principal component (PC1) of principal-component analysis (PCA) with NH₄⁺-N, NO₃⁻-N and P. Before modeling, we examined the distributions of all our variables and tested their normality. We used multiple criteria to test the overall goodness of fit for the SEMs. We used the χ² test (the model has a good fit when χ² is low and the probability value is high, traditionally P > 0.05, χ²/df of <2), the root mean square error of approximation (RMSEA; the model has a good fit when RMSEA is near 0 and the probability is high, traditionally P > 0.05), and adjusted goodness-of-fit index (AGFI; the model has a good fit when AGFI is over 0.9) (71). Because some of the variables introduced did not demonstrate a normal distribution, the probability of a path coefficient to differ from zero was tested by using bootstrap tests (72). Bootstrapping is preferred to the classical maximum likelihood estimation. Data are randomly sampled with replacement to arrive at estimates of SEMs that are empirically associated with the distribution of the data in the sample. With a reasonable model fit, we interpret the path coefficients and their associated P values to describe the strength of the relationships between two variables. The statistical analyses were performed using SPSS Amos (version 23; IBM).

Accession number(s).

The paired-end sequence data were deposited in the NCBI Sequence Read Archive under BioProject accession no. SRP082463 (http://www.ncbi.nlm.nih.gov/bioproject/PRJNA339333).