Linking niche size and phylogenetic signals to predict future soil microbial relative abundances

Abstract

Bacteria provide ecosystem services (e.g., biogeochemical cycling) that regulate climate, purify water, and produce food and other commodities, yet their distribution and likely responses to change or intervention are difficult to predict. Using bacterial 16S rRNA gene surveys of 1,381 soil samples from the Biomes of Australian Soil Environment (BASE) dataset, we were able to model relative abundances of soil bacterial taxonomic groups and describe bacterial niche space and optima. Hold out sample validated hypothetical causal networks (structural equation models; SEM) were able to predict the relative abundances of bacterial taxa from environmental data and elucidate soil bacterial niche space. By using explanatory SEM properties as indicators of microbial traits, we successfully predicted soil bacterial response, and in turn potential ecosystem service response, to near-term expected changes in the Australian climate. The methods developed enable prediction of continental-scale changes in bacterial relative abundances, and demonstrate their utility in predicting changes in bacterial function and thereby ecosystem services. These capabilities will be strengthened in the future with growing genome-level data.

Keywords: environmental change; microbial traits; phylogenetic signal; soil; soil bacteria; structural equation model.

Figure 1

Structural equation model of Proteobacteria relative abundance across Australia. A SEM with nine endogenous and one exogenous variable was constructed and found congruent with the observations from 1,381 soil samples from across Australia. (A) The SEM model for the phylum Proteobacteria, see Supplementary Figures S1–S56 for SEMS of other phyla. Solid arrows represent significant relationships (p < 0.05, dashed lines are non-significant) and the thickness of the arrow indicates the strength of the relationship, and the color indicates the origin of the path (blue = climate, green = vegetation, and brown = soil). Standardized path coefficients are shown next to each path. Standardized path coefficients can be interpreted as follows: if, for example, temperature increases by one standard deviation from the mean, then Proteobacterial relative abundance would increase by 0.25 standard deviations from its own mean. (B) Evaluation of SEM predictive ability, using a SEM calibrated from 1,000 samples to predict Proteobacterial relative abundance across 381 not in the calibration data set, (C) Spatial dispersal of Australian Proteobacterial relative abundance in 2016.

Figure 2

Ternary diagram of relative influence of soil, vegetation, and climate on bacterial relative abundance at the phyla level. Path coefficients were calculated between each phyla and endogenous variables: soil properties (brown; OC, conductivity, and pH), vegetation characteristics (green; C3 macrothermal plants, C3 mesothermal plants, and C4 megathermal plants), and climate (blue; maximum annual temperature, humidity, and precipitation). Relative influence of soil, vegetation, and climate is indicated by the thickness and contrast of the triangles along each axis (darker, thicker = greater influence; range from 0 to 100). Upper left: full ternary diagram with thick, dark triangle highlighting the region enlarged in the lower right. Point positions represent absolute values of standardized direct-path SEM coefficients scaled to sum to 100. For example, the Acidobacteria soil path coefficients were: OC = 0.051, EC = −0.358, pH = −0.165, and the sum of all 9 was 3.689. Therefore: ∑[(0.051/3.689)*100 + (0.358/3.689)*100 + (0.165/3.689)*100] = 15.6. Doing the same for vegetation and climate yields 70.5 and 13.9, respectively, the sum of all three equates to 100—15.6 + 70.5 + 13.9—which equates to the ternary coordinates for Acidobacteria. Point sizes represent median relative abundance. Only phyla with a median relative abundance greater or equal to 1 across 1,381 samples are included here. Point colors correspond to the 10 groups determined through hierarchical clustering (see Figure 3).

Figure 3

Clustering of soil, vegetation and climate drivers of bacterial relative abundances and links to soil niche space of bacterial genera. Standardized path coefficients of bacterial links to temperature, humidity, precipitation, C3 mesothermal plants, C3 macrothermal plants, C4 megathermal plants, OC, conductivity, and pH were associated based on Euclidean distance and then hierarchically clustered into 10 groups (represented by colored boxes) based on Ward’s minimum variance method. Extended Huisman-Olff-Fresco models were used to calculate the niche space ranges (solid lines), niche optimum (closed circles) where niche optimum was present for soil parameters. A continuous dashed line indicates a phylum for which a niche optimum was not detected.

Figure 4

Microbial functional traits related to soil salinity and pH are more commonly shared among related taxa. (A) Standardized path coefficients (bars) between salinity and pH of the SEM model presented in Figure 1 applied to genus level relative abundances that displayed significant local areas of phylogenetic association are displayed in color, gray indicates no significant phylogenetic associations. Branch tips with no associated bar were present in less than 200 samples and were therefore excluded. (B) Local areas of phylogenetic conservation or dispersion significantly different from zero for all genera present in more than 200 samples (p < 0.05). Points indicate local Moran’s I for each genus for salinity (inner ring) or pH (outer-ring). Points on the ring = 0, points outside that vary in value (>0–1) and are indicated approximately by the scale indicators for salinity and pH. Bars, points, and de novo trees are color-coded according to phylum.

Figure 5

Changes in key microbial community composition and ecological function across Australia from 2016 to 2030. (A,B) ln-fold changes in Cyanobacteria Synechococcophycideae and Firmicutes Bacilli relative abundance. Change in n-dimensional space of PCoAs Bray–Curtis dissimilarity matrices of current and future (C) microbial composition and (D) methane metabolism. Color indicates the magnitude of change between time periods: (A,B) pink/violet = increased and yellow/orange = decreased future relative abundances; (C,D) pink/violet indicates relatively large changes and yellow/orange relatively small changes in community composition/methane metabolism. Standardized path coefficients from SEMs of soil factors, vegetation, and climate on bacterial relative abundance are presented for Cyanobacteria Synechococcophycideae and Firmicutes Bacilli. Soil influences (brown): OC, soil OC; EC, electrical conductivity. Vegetation influences (green): C3A, C3 macrothermal plants, C3E, C3 mesothermal plants; C4M, C4 megathermal plants. Climatic influences (blue): Tmp, temperature; Hmd, humidity; and Prc, precipitation.