Niche Separation Increases With Genetic Distance Among Bloom-Forming Cyanobacteria

Abstract

Bacterial communities are composed of distinct groups of potentially interacting lineages, each thought to occupy a distinct ecological niche. It remains unclear, however, how quickly niche preference evolves and whether more closely related lineages are more likely to share ecological niches. We addressed these questions by following the dynamics of two bloom-forming cyanobacterial genera over an 8-year time-course in Lake Champlain, Canada, using 16S amplicon sequencing and measurements of several environmental parameters. The two genera, Microcystis (M) and Dolichospermum (D), are frequently observed simultaneously during bloom events and thus have partially overlapping niches. However, the extent of their niche overlap is debated, and it is also unclear to what extent niche partitioning occurs among strains within each genus. To identify strains within each genus, we applied minimum entropy decomposition (MED) to 16S rRNA gene sequences. We confirmed that at a genus level, M and D have different preferences for nitrogen and phosphorus concentrations. Within each genus, we also identified strains differentially associated with temperature, precipitation, and concentrations of nutrients and toxins. In general, niche similarity between strains (as measured by co-occurrence over time) declined with genetic distance. This pattern is consistent with habitat filtering - in which closely related taxa are ecologically similar, and therefore tend to co-occur under similar environmental conditions. In contrast with this general pattern, similarity in certain niche dimensions (notably particulate nitrogen and phosphorus) did not decline linearly with genetic distance, and instead showed a complex polynomial relationship. This observation suggests the importance of processes other than habitat filtering - such as competition between closely related taxa, or convergent trait evolution in distantly related taxa - in shaping particular traits in microbial communities.

Keywords: Dolichospermum; Microcystis; competition; cyanobacteria; ecological niche; habitat filtering; niche partitioning.

FIGURE 1

Temporal dynamics of the two dominant cyanobacterial genera over an 8-year time-course in littoral (A) and pelagic (B) sampling sites. Relative abundance of Microcystis is shown in solid red, Dolichospermum in dashed green and the other members of the bacterial community in dashed light violet. The time scale (x-axis) is in units of years.

FIGURE 2

Niche partitioning at the sub-genus level. Best-fit polynomial models of the response of Dolichospermum (A) and Microcystis (B) nodes to abiotic factors. The relative abundance of each MED node (strain) was computed relative to the total number of reads within each genus using the centered-log ratio (clr) transform. Significant relationships are shown by solid lines and colored confidence intervals. In most cases, the degree-1 polynomial (linear model) provided the best-fit (see Supplementary Table S1 for details). For Dolichospermum, only the dominant nodes (observed in at least 70% of samples) are shown.

FIGURE 3

Co-occurrence of strains declines with their pairwise genetic distance. Relationship between co-occurrence (significant SparCC correlation, P < 0.05) and genetic distance (p-distance) between Dolichospermum (A) and Microcystis (B) nodes. Blue diamonds represents the mean SparCC correlation for each distance. Boxplots show the median (horizontal line), the 25th and 75th percentile (enclosed in box) and 95% confidence intervals (whiskers). The discreteness observed in the x-axis is due to the discrete number of substitutions in the 16S rRNA gene sequence (e.g., exactly 1, 2, 3, …nucleotide differences between pairs).

FIGURE 4

The relationship between genetic distance and the co-response (niche similarity) for Dolichospermum. LVMs were used to identify correlations between the responses of MED nodes to each measured environmental parameter (separate panels). Positive correlations of fitted responses indicate similar niches; negative correlations indicate different niche preferences. Genetic distances were computed using the p-distance. Separate model fits were tested with the Akaike Information Criterion (AIC) for the relationship between each niche dimension and genetic distance. See Supplementary Table S4 for details of model fits. Significant model fits are shown with thick solid lines; non-significant fits are shown with dashed lines.