Repeated introduction of micropollutants enhances microbial succession despite stable degradation patterns

The increasing-volume release of micropollutants into natural surface waters has raised great concern due to their environmental accumulation. Persisting micropollutants can impact multiple generations of organisms, but their microbially-mediated degradation and their influence on community assembly remain understudied. Here, freshwater microbes were treated with several common micropollutants, alone or in combination, and then transferred every 5 days to fresh medium containing the same micropollutants to mimic the repeated exposure of microbes. Metabarcoding of 16S rRNA gene makers was chosen to study the succession of bacterial assemblages following micropollutant exposure. The removal rates of micropollutants were then measured to assess degradation capacity of the associated communities. The degradation of micropollutants did not accelerate over time but altered the microbial community composition. Community assembly was dominated by stochastic processes during early exposure, via random community changes and emergence of seedbanks, and deterministic processes later in the exposure, via advanced community succession. Early exposure stages were characterized by the presence of sensitive microorganisms such as Actinobacteria and Planctomycetes, which were then replaced by more tolerant bacteria such as Bacteroidetes and Gammaproteobacteria. Our findings have important implication for ecological feedback between microbe-micropollutants under anthropogenic climate change scenarios.


Supplemental text Part1) Additional experiments to test micropollutant degradation
Page 3 Part2) Calculation of the weighted mean rRNA gene copy number Page 4 Part3) Analysis of line regression model Page 5 Supplemental Tables  Table S1  Page 6  Table S2 Page 7 Table S3 Pages 8-14 Table S4 Pages 15-18 Figure S1 Page 19 Figure S2 Page 20 Figure S3 Page 21 Figure S4 Page 22 Figure S5 Page 23 Figure S6 Page 24 Figure S7 Page 25 Figure S8 Page 26 Figure S9 Page 27 3 Supplemental methods

Part 1) Additional experiments to test micropollutant degradation
An additional experiment was conducted to test the removal rates of micropollutants in the absence of microorganisms. Reservoir water was filtered through 0.22-μm filters and then amended with a mixture consisting of 1 μg BPA/L, 0.1 μg BPS/L, 1 μg TCS/L, and 0.1 μg TCC/L as a blank control. A second additional experiment was conducted to assess whether the ability of prokaryotic assemblages to degrade BPA and TCS was enhanced by the evolving community succession following micropollutant addition. Thus, 40 ml of the initial inoculum and 40 ml of the bacterial assemblage at the end of the experiment (after incubation of the B7 microcosms) were each amended with 160 ml of BPA, 160 ml of TCS, or a mixture thereof, with each micropollutant present at a final concentration of 1 mg/L.
The removal rates of BPA and TCS in these treatments after 7 days were measured and the rates in the initial inocula vs. those in the B7 microcosms were compared. The incubation conditions were the same for each treatment and the experiment was conducted in triplicate.
The first experiment showed that the removal proportion of the four micropollutants in the prokaryote-free controls was stable during the 7-day incubation (Supplemental Fig. S2), and therefore the chemical stability and negligible transformation of the micropollutants in the absence of microorganisms. The second experiment showed that the concentration of BPA declined later in the initial community without pre-exposure than in the community with a 35-day history of exposure (Supplemental Fig. S3). In the case of TCS, the concentration did not change over time in the initial community, while in the pre-exposed community it decreased sharply at day 4 of the incubation. These results demonstrated that pre-exposure to micropollutants impacted the ability of the bacterial community to transform micropollutants during a later exposure and that this ability differed between communities exposed to BPA and TCS. 4 Part 2) Calculation of the weighted mean rRNA gene copy number The rrnDB database catalogues the 16S rRNA gene copy number of an organism with its NCBI or RDP taxonomy [1]. The 16S rRNA gene copy number of an OTU is estimated to be the number of its assigned genus. If the 16S rRNA gene copy number of the corresponding genus is not provided in the rrnDB database, then number of the higher phylogenetic level (i.e., family/order/class/phylum) is used. The abundance-weighed average gene copy number Nrrn of each community is calculated using the relative abundance P of each OTU in the community, the taxonomic assignment of each OTU, and the estimated 16S rRNA gene copy number n of each OTU based on rrnDB database v5.5 [2]: where ni is the estimated 16S rRNA gene copy number of OTU i, Pi is the relative abundance of OTU i, and x is the total number of OTU in the microbial community.

Part 3) Analysis of line regression model
A linear regression model was used to evaluate the rate of community turnover over time: where D is the Bray-Curtis dissimilarity between the micropollutant treated vs. untreated communities, T is the incubation time, and the slope v is the turnover rate. The time-decay relationship (TDR) of the PTC of each treatment group and control was evaluated using a log-transformed power law model [3,4]: where S is the pairwise Bray-Curtis similarity in community composition across time interval T, and the slope w is the temporal turnover rate of the community within each treatment group or control.        Figure S6. Non-metric multidimensional scaling (NMDS) based on Bray-Curtis dissimilarity matric illustrating the beta-diversity (between-sample differences in community composition) of the initial inoculum, the controls, and the treatments. The analysis was done using the exact sequence variants generated from DADA2 pipeline (Callahan et al., 2017). Raw sequences were processed using the DADA2 pipeline according to the DADA2 tutorial (v1.14) in R. The primers of sequences were trimmed using Cutadapt, and the resulting sequences were quality filtered with customized modifications as follows: truncLen=c(220, 190), maxEE=2, truncQ=2, maxN=0, rm.phix=TRUE, trimLeft=c(0,0). Subsequently, denoising, merging and chimera removal were completed according to the DADA2 pipeline tutorial. All sequences were aligned and assigned taxonomically using the SILVA v.132 reference database. The M 2 value and P-value were obtained from Procrustes statistic: M 2 value represents sum of squared deviations between sample pairs, and the lower value mean a better fit between matrices.

Reference:
Callahan BJ, McMurdie PJ, and Holmes SP. (2017). Exact sequence variants should replace operational taxonomic units in marker-gene data analysis. ISME J, 11:2639-2643.  to ecological categories found in the treatments but not shared in the controls when analyzed with the same filtering criteria as done for ecological grouping. The percentage in each sector of the Venn diagram presents the proportion of OTUs that are unique to either the controls or treatments, or that are shared between the two. A large fraction of the bacterial OTUs was unique to each of three ecological categories.
This suggests that the identification of ecological strategies was not biased towards the selection of bacterial taxa whose high abundance in each defined phase was due to simple enrichment in the controls over time.