Community recovery dynamics in yellow perch microbiome after gradual and constant metallic perturbations

Abstract

Background: The eco-evolutionary processes ruling post-disturbance microbial assembly remain poorly studied, particularly in host-microbiome systems. The community recovery depends not only on the type, duration, intensity, and gradient of disturbance, but also on the initial community structure, phylogenetic composition, legacy, and habitat (soil, water, host). In this study, yellow perch (Perca flavescens) juveniles were exposed over 90 days to constant and gradual sublethal doses of cadmium chloride. Afterward, the exposure of aquaria tank system to cadmium was ceased for 60 days. The skin, gut and water tank microbiomes in control and treatment groups, were characterized before, during and after the cadmium exposure using 16s rDNA libraries and high throughput sequencing technology (Illumina, Miseq).

Results: Our data exhibited long-term bioaccumulation of cadmium salts in the liver even after two months since ceasing the exposure. The gradient of cadmium disturbance had differential effects on the perch microbiota recovery, including increases in evenness, taxonomic composition shifts, as well as functional and phylogenetic divergence. The perch microbiome reached an alternative stable state in the skin and nearly complete recovery trajectories in the gut communities. The recovery of skin communities showed a significant proliferation of opportunistic fish pathogens (i.e., Flavobacterium). Our findings provide evidence that neutral processes were a much more significant contributor to microbial community turnover in control treatments than in those treated with cadmium, suggesting the role of selective processes in driving community recovery.

Conclusions: The short-term metallic disturbance of fish development has important long-term implications for host health. The recovery of microbial communities after metallic exposure depends on the magnitude of exposure (constant, gradual), and the nature of the ecological niche (water, skin, and gut). The skin and gut microbiota of fish exposed to constant concentrations of cadmium (CC) were closer to the control negative than those exposed to the gradual concentrations (CV). Overall, our results show that the microbial assembly during the community recovery were both orchestrated by neutral and deterministic processes. Video Abtract.

Keywords: Community assembly; Disturbance; Evolutionary forces; Fish microbiome; Metagenomics; Neutrality; Pathogens; Recovery; Stress gradient.

Fig. 1

Schematic illustration of the perch microbiome recovery experiment

Fig. 2

Predicted alpha-diversity plots by linear mixed model. Alpha-diversity in water and host-microbial communities over time and among treatments is predicted using the linear mixed model. The richness/evenness ratio were considered as response variables, the fixed effects were defined by time and cadmium concentration (in water and liver), and tanks were taken as random effects. Over time, the predicted alpha diversity in host microbial communities (skin, gut) highlights stable trends of the Control group compared to the treatments. However, all groups of the water microbial communities decrease overtime. Constant cadmium regime (CC) is in orange, variable cadmium regime (CV) is in yellow, and control (control) is in green

Fig. 3

Taxonomic composition dynamics of host communities. Stacked barplots show the most abundant taxa (> 0.5%) overtime in the gut, skin and water microbiomes. The genera that significantly changed among treatments and control at T5 are summarized in Additional file 8: Table S2

Fig. 4

Heatmaps of differential abundance among host and water communities. This figure from left to right includes 9 heatmaps of the significant taxonomic fingerprints at the genus level between gut, the skin and the water at times T0 (first column), T3 (second column), and T5 (third column) in the control (first row), the CV (second row), and the CC (third row) groups. The hierarchical clustering of the relative abundance of phyla which significantly changed over time was performed using Ward’s method and Bray–Curtis dissimilarity distance. Vegan package and pheatmap () function in R were used

Fig. 5

Recovery dynamics of the networks of host communities. The networks organization is based on nodes betweenness centrality among treatments and Control. Unstructured patterns in the networks were observed at T0. Node size represents sample richness. The strength of correlation (Spearman correlation from 0.3 to 1) between two nodes is inversely proportional to the size of the edge. This network was built using R and Cytoscape software. Constant Cadmium samples (CC) are in orange, variable cadmium samples (CV) are in yellow, and control (Control) samples are in green

Fig. 6

Recovery dynamics of the network of water communities. The networks organization at every resilience time TR1, TR2, TR3, TR4, and WT5 is based on nodes betweenness centrality among treatments and control. The network modules easliy distinguishable between groups since T0. The size of nodes (sample richness) at the beginning of TR1 and at the end of the time TR4 showed shifts in the community richness. The strength of correlation (Corr. Spearman from 0.5 to 1) between two nodes is inversely proportional to the size of the edge. This network was built using R and Cytoscape software. constant cadmium samples (CC) are in orange, variable cadmium samples (CV) are in yellow, and control (control) samples are in green

Fig. 7

Centrality plots of host microbiome networks. This figure summarizes relationships of betweenness centrality versus eigenvalue centrality of host microbiome networks among treatments and at each time point. The results show evidence of shift in centrality medians between the control regime (which is higher) and the cadmium selection regimes. The plots of eigenvalue centrality versus betweenness centrality clearly reveal that centrality shift at time T3 for skin and gut microbiome

Fig. 8

Function diversity dynamics in host and water microbiome. Boxplots of functions profiles were predicted from the matrices of taxa count using the software Tax4Fun. The statistical significance (p value < 0.05) found using ANOVA followed by FDR (false discovery rate) test are represented with asterisks points (0.001: “***,” 0.01: “**,” 0.05: “*”)

Fig. 9

Percentage of neutral OTUs over time and treatment. Using the non-linear least squares model (NLS), the percentage of OTUs that fit the neutral model within a confidence interval of 95% showed variable trends between communities across time and treatments. A goodness of fit R²> 0.5 was considered as the significant threshold of neutrality fit. The cadmium treatment invoked stochasticity in the water communities, while in gut and skin communities, the percentage of neutral OTUs remained higher in the control compared to treatments.

Fig. 10

Demographic variation of metacommunity neutrality across water and host microbiome. This figure summarizes the scatterplots of neutral model fitting the whole metacommunity ( gut skin and water) at times T0 (first column), T3 (second column) and T5 (third column) in the control (first row), the CV (second row), and the CC (third row) groups. Neutral OTUs are shown in black, non-neutral are depicted in grey, while the red is Mycoplasma sp. OTUs. We see no Mycoplasma sp. OTUs that fit the neutral model in the whole metacommunity