Reconstructing the Microbial Diversity and Function of Pre-Agricultural Tallgrass Prairie Soils in the United States

+ See all authors and affiliations

Science  01 Nov 2013:
Vol. 342, Issue 6158, pp. 621-624
DOI: 10.1126/science.1243768

Prairie Redux

Tallgrass prairie is extinct across much of its former range in the midwestern United States, but relicts preserved in cemeteries and nature reserves allow functional comparison of former grassland soils with modern agricultural soils. Fierer et al. (p. 621; see the Perspective by Scholes and Scholes) took matched soil samples from sites representing the gamut of climate conditions and modeled the combination of genomic analysis and environmental data to resurrect the historical prairie soil communities, identifying the nutrient-scavenging Verrucomicrobia as keystone bacteria in functioning prairie.


Native tallgrass prairie once dominated much of the midwestern United States, but this biome and the soil microbial diversity that once sustained this highly productive system have been almost completely eradicated by decades of agricultural practices. We reconstructed the soil microbial diversity that once existed in this biome by analyzing relict prairie soils and found that the biogeographical patterns were largely driven by changes in the relative abundance of Verrucomicrobia, a poorly studied bacterial phylum that appears to dominate many prairie soils. Shotgun metagenomic data suggested that these spatial patterns were associated with strong shifts in carbon dynamics. We show that metagenomic approaches can be used to reconstruct below-ground biogeochemical and diversity gradients in endangered ecosystems; such information could be used to improve restoration efforts, given that even small changes in below-ground microbial diversity can have important impacts on ecosystem processes.

After the European settlement of the midwestern United States in the mid-19th century, the tallgrass prairie ecosystem was profoundly altered by the removal of key animal taxa (including bison), fire suppression, and the plowing under of native grasses. Together these land-use changes contributed to the most substantial decline of any major ecosystem in North America (1, 2). This ecosystem, which once covered nearly 10% of the contiguous United States (>65 million ha), has been reduced to a small fraction of its historical extent (3). Cultivation and row crop agriculture, now practiced across most of the tallgrass prairie biome, not only replaced species-rich plant communities with monoculture croplands, but also drastically altered the physicochemical and biological characteristics of prairie soils. Except for a few prairie relicts that have never been tilled, the soils currently found throughout the region bear little resemblance to their pre-agricultural state (47). We confirmed the effect of cultivation on soil microbial communities by directly comparing bacterial communities in cultivated soils with paired uncultivated soils collected from throughout the native tallgrass prairie range (table S1), and found that the cultivated soils harbored bacterial communities that were significantly distinct in composition from those found in the corresponding native prairie soils (fig. S1).

Owing to the historical and biological importance of the native tallgrass prairie, there have been various attempts to predict the historical distributions of plants and animals across this ecosystem [e.g., (8)]. However, comparable reconstructions of below-ground microbial diversity have, to our knowledge, never been attempted, hindering our understanding of how soil microbes may have once influenced plant production, nutrient retention, and soil carbon dynamics in this ecosystem. By coupling metagenomic sequence data, which capture the phylogenetic and functional diversity of existing soil microbial communities (9, 10) found in tallgrass prairie remnants, to spatially explicit models (11), which predict the structure of soil microbial communities across environmental gradients, we can reconstruct these communities across the historical extent of the tallgrass prairie ecosystem.

We collected surface soils (top 10 cm) from 31 remnant native prairie sites (found primarily in cemeteries or nature preserves) that were carefully selected to span the range of climate conditions found throughout the tallgrass prairie ecosystem (fig. S2 and table S1). To the best of our knowledge, none of the sampled sites were ever tilled and all were dominated by native tallgrass prairie plant species (e.g., Andropogon, Panicum, and Sorghastrum) when soil samples were collected at the height of the growing season. To characterize the bacterial and archaeal communities, we pyrosequenced a polymerase chain reaction (PCR)–amplified region of the 16S rRNA gene (12) and compared relative diversity levels across the samples by standardizing sequencing depth to 940 reads per sample (13). Microbial α diversity (measured using the Shannon index, H′) was not uniformly distributed across the tallgrass prairie (Fig. 1A and table S2). However, none of the measured edaphic variables were significantly correlated with taxonomic diversity (P > 0.1 in all cases). Instead, taxonomic diversity was most strongly correlated with precipitation levels (table S3), just as moisture availability is often a strong predictor of plant and animal diversity at comparable spatial scales (14).

Fig. 1 Relationships between taxonomic and functional diversity (A) and community similarity patterns (B) across the 31 sites, with x axes showing taxonomic comparisons and y axes showing comparisons based on functional genes.

Points are colored according to latitude (red, <35°N; yellow, 35° to 40°N; green, 40° to 45°N; blue, >45°N). The Shannon index (H′) and the Bray-Curtis index of community similarity were used as measures of α diversity and β diversity, respectively. The general β diversity patterns are shown here using the principal coordinates score for the first axis; Mantel tests conducted directly from the distance matrices confirm the significance of the correlation shown in (B) (Mantel r = 0.95, P < 0.001).

To complement the taxonomic analyses and to characterize microbial functional diversity, we used shotgun metagenomics to determine the diversity of known protein-coding genes in each community (10, 12). All samples were compared at an equivalent survey depth of 1.8 million randomly selected annotated reads per sample. Although these annotations are based only on previously described genes and much of the functional diversity contained within soil [or other environments where the majority of taxa have not been well characterized (15)] is therefore missed, we did find that the functional diversity that could be annotated within the microbial communities correlated well with taxonomic diversity (Fig. 1A). There has been considerable debate in the field of ecology as to how the taxonomic diversity of communities relates to the observed functional, or trait-level, diversity (16); we found a strong positive correlation between the functional and taxonomic diversity of soil microbial communities (Fig. 1A). This pattern is also commonly observed in plant and animal communities (17, 18). Therefore, the abundant soil microbial taxa do not appear to exhibit a high degree of functional redundancy (or equivalence), as is often assumed (19); reductions in taxonomic diversity were associated with decreases in the breadth of functional traits contained within these soil communities.

To extend our results beyond the 31 soils directly assayed, we constructed spatially explicit models of the taxonomic and functional diversity of soil bacterial communities throughout the historical extent of the native tallgrass prairie using a species distribution modeling approach (20, 21). These models were based solely on climatic variables (tables S3 and S4), not soil characteristics, because we do not know what soil characteristics would have been found across the biome >150 years ago when the native tallgrass prairie ecosystem was still intact (current soil maps are not likely to reflect historical soil properties that have been profoundly altered by decades of cultivation). The best-fit models of Shannon diversity were able to predict >50% of the variance in bacterial taxonomic and functional diversity across the sampled sites (table S3), a predictive power similar to comparable models of regional plant or animal diversity (20), with soils from the middle latitudes having less taxonomic and functional diversity than those in the northernmost and southernmost portions of the range (Fig. 2). This pattern contrasts with the biogeographical patterns observed in many plant and animal communities where diversity often peaks in the middle of biomes, which suggests that the “mid-domain effect” (22), or related explanations for such diversity gradients, may not be broadly applicable across the tree of life.

Fig. 2 Predicted diversity patterns (calculated using the Shannon index, H′) of bacterial taxa (A) and functional genes (B).

Inset plots show the cross-validation results, comparing observed and predicted diversity values for the 31 sampled locations.

The taxonomic composition of the microbial communities and their functional attributes varied considerably across the 31 sampled tallgrass prairie sites. Communities that were taxonomically similar were also similar with respect to their functional characteristics as these two distinct measures were well correlated (i.e., communities that shared many taxa in common also shared many functional attributes) (Fig. 1B). Model predictions of the patterns in community similarity show that the soils in the central portion of the tallgrass prairie range harbored communities that were taxonomically (Fig. 3A) and functionally (Fig. 3B) distinct from those found on the edges of the range. The relative abundances of many taxa changed across the sampled soils (table S2), but the α diversity patterns (Fig. 2) and the patterns in community similarity (Fig. 3) were both closely tied to changes in the relative abundance of Verrucomicrobia (Fig. 3C and table S2), the dominant bacterial phylum across the collected soils (fig. S2 and table S2). Although often underestimated (23), Verrucomicrobia represented >50% of the bacterial 16S rRNA sequences in the prairie soils from the mid-latitudes but only <15% of the sequences on the edges of the range (Fig. 3C and table S2). These patterns in verrucomicrobial abundances were nearly identical whether abundances were determined by the PCR-based 16S rRNA gene analyses or by analyzing the 16S rRNA genes recovered from the shotgun metagenomic data (fig. S3); this finding provides independent evidence for the high relative abundances of Verrucomicrobia in many of these soils.

Fig. 3 Maps showing bacterial community types based on their taxonomic composition (A) or functional gene composition (B).

Locations similar in hue indicate those communities that are more similar in composition. (C and D) The patterns shown in (A) and (B) appear to be largely driven by variation in verrucomicrobial abundances (% of reads) (C). Shown in (D) are functional gene categories strongly correlated with verrucomicrobial abundances, only including those gene categories represented by >0.001% of the shotgun metagenomic reads with r values greater than 0.8 or less than –0.8.

Verrucomicrobia are clearly dominant in the tallgrass prairie soils, but their ecology remains poorly understood because members of this group are difficult to culture and study in the laboratory (23, 24). The Verrucomicrobia identified from the prairie soils were not diverse; only five phylotypes accounted for >75% of the verrucomicrobial sequences from this data set (fig. S4). All of the more abundant verrucomicrobial taxa were classified as belonging to the Spartobacteria class, but the taxa were not closely related to previously cultivated verrucomicrobial isolates (fig. S4), making it difficult to determine the ecological attributes of these taxa. Our understanding of soil microbial communities in prairie soils will clearly benefit from efforts to directly determine the phenotypes of the Verrucomicrobia that dominated the majority of the native tallgrass prairie soils.

Although the ecological attributes of these verrucomicrobial taxa cannot be directly assessed with these data, we used a niche-modeling approach in combination with the shotgun metagenomic data to gain some insight into the ecology of the Verrucomicrobia and to try to explain the distribution patterns shown in Fig. 3C. We found that spatial variability in the abundance of Verrucomicrobia could be predicted from climatic conditions (table S3) and that this group was most abundant in soils exposed to intermediate temperature and precipitation conditions (Fig. 3C and fig. S2). However, this correlation with climatic conditions may represent only a distal control on their distribution patterns. The shotgun metagenomic data lend support to the hypothesis that Verrucomicrobia are relatively slow-growing taxa that thrive under conditions of limited nutrient availability (25, 26). Specifically, verrucomicrobial abundances were positively correlated with a variety of genes associated with carbohydrate metabolism but were negatively correlated with genes associated with nitrogen metabolism and cell division (Fig. 3D). Verrucomicrobia may thus represent a large component of below-ground communities in regions where changes in the quantity or quality of plant organic matter inputs constrain the growth of more copiotrophic taxa. This hypothesis is congruent with results indicating consistent declines in the relative abundances of Verrucomicrobia when soils from across North America were amended with nutrients (27). Likewise, this hypothesis is consistent with recent genomic information obtained from Spartobacteria aquaticum, an aquatic Verrucomicrobia that is within the same class as the dominant soil Verrucomicrobia observed here, that appears to specialize on the degradation of more recalcitrant carbon compounds (28).

Our reconstructions of microbial diversity and functional capabilities across the tallgrass prairie ecosystem could be used to guide and monitor the hundreds of prairie restoration efforts currently underway throughout the midwestern United States (29). Maps of the soil microbial communities that once existed in this ecosystem may provide targets to help improve the long-term success of prairie restoration efforts, as restoration efforts are often more successful when they also try to restore below-ground communities (30). Such information may be particularly important if the goal is to restore key ecosystem functions, such as soil carbon sequestration, that are strongly controlled by the below-ground communities. Likewise, deviation in soil microbial communities from the predicted pre-agricultural state could be used to quantify the extent of degradation experienced by soils throughout the native prairie range. More generally, this work demonstrates that we can use recent advances in high-throughput microbial community characterization to reconstruct the biogeographical patterns in the diversity and functional capabilities of microbes across a nearly extinct ecosystem. This approach could be extended more broadly to quantify how historical changes in environmental conditions may have altered the diversity and function of below-ground communities in other systems or to determine how human-induced climate change may alter ecosystem properties in the future.

Supplementary Materials

Materials and Methods

Figs. S1 to S4

Tables S1 to S4

References (3150)

References and Notes

  1. See supplementary materials on Science Online.
  2. Acknowledgments: We thank K. McLauchlan and three anonymous reviewers for their critical feedback on earlier versions of the manuscript; R. Jackson for his help with soil collection and analyses; and J. Henley for her help with the laboratory analyses. Supported by NSF grants DEB-0953331 (N.F.) and DMS-1069303 (K.S.P.), the Howard Hughes Medical Institute (R.K.), Gordon and Betty Moore Foundation grant 3300 (K.S.P.), U.S. Department of Energy contract DE-AC02-06CH11357 (J.A.G.), and USDA National Research Initiative 2005-35101-15335/17371 (R.L.M.). All amplicon data have been deposited in the European Nucleotide Archive under accession number ERP003610; the accession number for the shotgun metagenomic data is ERP003954. Data have also been made available through the Dryad data depository.
View Abstract


Navigate This Article