Bacterial communities in human-impacted rivers and streams are exposed to multiple
anthropogenic contaminants, which can eventually lead to biodiversity loss and function.
The Wonderfonteinspruit catchment area is impacted by operational and abandoned gold
mines, farms, and formal and informal settlements. In this study, we used 16S rRNA gene
high-throughput sequencing to characterize bacterial communities in the lower Wonderfonteinspruit
and their response to various contaminant sources. The results showed that composition
and structure of bacterial communities differed significantly (P<0.05) between less
(downstream) and more (upstream) polluted sites. The taxonomic and functional gene dissimilarities
significantly correlated with each other, while downstream sites had more distinct
functional genes. The relative abundance of Proteobacteria, Bacteroidetes and Actinobacteria
was higher at upstream sites, while Acidobacteria, Cyanobacteria, Firmicutes and Verrucomicrobia
were prominent at downstream sites. In addition, upstream sites were rich in
genera pathogenic and/or potentially pathogenic to humans. Multivariate and correlation
analyses suggest that bacterial diversity was significantly (P<0.05) impacted by pH and
heavy metals (cobalt, arsenic, chromium, nickel and uranium). A significant fraction (~14%)
of the compositional variation was explained by a combination of anthropogenic inputs, of
which mining (~6%) was the main contributor to bacterial community variation. Network
analysis indicated that bacterial communities had non-random inter- and intra-phyla associations
and that the main taxa showed both positive and negative linkages to environmental
parameters. Our results suggest that species sorting, due to environmental parameters,
was the main process that structured bacterial communities. Furthermore, upstream sites had higher relative abundances of genes involved in xenobiotic degradation, suggesting
stronger removal of polycyclic aromatic hydrocarbons and other organic compounds. This study provides insights into the influences of anthropogenic land use on bacterial community
structure and functions in the lower Wonderfonteinspruit.
S1 Fig. Rarefaction curves for all sequences at all sampling locations estimating the number
of bacterial OTUs at the 97% identity level.
S2 Fig. Stacked column bar graph representing the predicted metabolic attributes between
upstream and downstream sites against the KEGG database implemented in PICRUSt at
(A) tier level-1 and (B) tier level-2. The mean Nearest Sequenced Taxon Index (NSTI) value
for all samples was 0.188 ± 0.000125 s.d.
S3 Fig. Heat map and cluster analysis of bacterial core 97% identity OTUs between the different
sampling locations. Samples were grouped using hierarchical clustering (complete
linkage) based on the Bray–Curtis distance matrix calculated from the relative abundances (in
percent) of the OTUs. The colour code goes from blue (not detected) to yellow (low abundance)
to orange (medium abundance) to red (high abundance) on a logarithmic scale to
improve visualization between low and medium abundance.
S4 Fig. Network analysis of bacterial core OTUs (i.e. OTUs with relative abundance
>0.2%) and environmental parameters. Co-occurrence patterns were based on significant
(P<0.05) Spearman correlations with ρ ±0.6 showing the entire network structured according
to site (upstream, downstream and/or both). Each node (circle) in the network represents
a unique OTU and the size is proportional to node degree. Each edge (connection) represents
a strong and significant correlation (P<0.05), while the colour relates to the type of interaction:
positive (grey solid lines) or negative (red dashed lines). Environmental parameters are presented
by purple rectangles.
S1 Table. Summary of the sampling area, anthropogenic activities performed in the lower
WFS and associated contaminants.
S2 Table. Recommended Target Water Quality Range (TWQR) for the lower Wonderfonteinspruit.
The authors kindly thank IBIS/Universite´ Laval Plate-forme d’Analyses Ge´nomiques for 454-
pyrosequencing and support, especially Dr. Brian Boyle, as well as Marie-E`ve Beaulieu (manager
of DPK’s laboratory). We also acknowledge the Unit for Environmental Sciences and
Management (UESM), and David Hamman for assistance with sampling.