Global Biogeography and Quantitative Seasonal Dynamics of Gemmatimonadetes in Soil

Abstract

Bacteria belonging to phylum Gemmatimonadetes comprise approximately 2% of soil bacterial communities. However, little is known of their ecology due to a lack of cultured representation. Here we present evidence from biogeographical analyses and seasonal quantification of Gemmatimonadetes in soils, which suggests an adaptation to low soil moisture.

TEXT

Bacteria belonging to phylum Gemmatimonadetes are frequently detected in environmental 16S rRNA gene libraries and have been identified as one of the top nine phyla found in soils, comprising ca. 2% of soil bacterial communities (23). More recent high-throughput sequencing approaches have confirmed these estimates: Gemmatimonadetes relative abundances in large libraries (>500 sequences) from soils range from 0.2% to 6.5%, with a mean of 2.2% (Fig. 1).

Fig. 1.

Relative abundances of Gemmatimonadetes (as a proportion of total bacterial community) in soils and sediments reported in published 16S rRNA gene libraries. Studies are grouped by library size as follows: >5,000 sequences (pyrosequencing libraries) (black squares), 500 to 2000 (small pyrosequencing libraries and large clone libraries) (gray squares), and <500 sequences (clone libraries) (white squares). Studies referenced are as follows: Antarctic dry soils (6, 14), alpine hyper arid soils (9), hot desert (8, 39, 47), alpine tundra (37), Antarctic peninsula/coastal (6, 45), semiarid soils (1, 26, 32), prairies/grasslands (12, 41), pasture (22), crop agriculture (7, 22, 29, 41, 50), temperate forest (17, 22, 29, 30, 41), tropical forest (13), moist acidic tundra (5), and freshwater sediments (27, 48).

Despite their frequency and persistent abundance in soils, only one representative from this phylum has been isolated and characterized, Gemmatimonas aurantiaca strain T-27, a polyphosphate-accumulating isolate from wastewater (51). The isolation of four other Gemmatimonadetes strains has been reported (10, 25); however, none of these strains have been characterized to date. The highest proportions of Gemmatimonadetes were found in arid soils (Fig. 1, left), suggesting an adaption to low-moisture environments. A limitation of these studies is that they are restricted to a singular time or treatment.

To gain further insight into Gemmatimonadetes ecology, we tested the hypothesis that Gemmatimonadetes are more prevalent in drier soils. A 2-fold approach was employed: first, a biogeographic analysis of Gemmatimonadetes sequences deposited in public databases was performed to determine patterns of environmental distribution. Second, seasonal Gemmatimonadetes abundances were quantified in five land management systems at a long-term ecological research (LTER) site in Michigan to determine relationships to edaphic factors.

Biogeography of Gemmatimonadetes.

Gemmatimonadetes 16S rRNA gene sequences (>1,200 bp in length) deposited in the RDP were used to analyze biogeographical patterns. A neighbor-joining phylogenetic tree of 456 sequences was constructed in MEGA v4 (44) with Fibrobacter succinogenes (accession no. AJ496032) as an outgroup (Fig. 2). NCBI GenBank metadata were retrieved to classify soils by dominant vegetation and land use. The resulting tree (Fig. 2) confirmed a cosmopolitan distribution of the phylum, which was not unexpected as it has been observed for other prokaryotic phyla (40). The sequences were grouped into three clades, here arbitrarily designated G1, G2, and G3. The greatest number of Gemmatimonadetes phylotypes was from soils, including grassland/prairie/pasture soil (26.4% of sequences), agricultural soil (13.1%), forest soil (11.1%), and contaminated soil (20.6%), confirming its place as one of the nine dominant soil phyla (23). Gemmatimonadetes phylotypes have also been recovered from sediments (4, 19, 38), and other nonsoil locations (16, 20, 24, 33, 35, 46, 49). Their presence in environments with a wide range of nutrient concentrations (e.g., eutrophic lake sediments to alpine soils) and redox states (anoxic sediments or inner soil aggregates to airborne dust) suggests versatile metabolisms which have contributed to their cosmopolitan success.

Fig. 2.

Neighbor-joining phylogenetic tree of 456 Gemmatimonadetes 16S rRNA gene sequences deposited in RDP (>1,200 bp in length). Fibrobacter succinogenes was used as an outgroup. Sequences are listed by GenBank accession number and colored according to the sample type as follows: arid soils and deserts (tan), forest soil (dark green), prairie and grassland soil (light green), agricultural crop soil (yellow), alpine and tundra soil (brown), rock surfaces (black), soil contaminated with organics and hydrocarbons (pink), soil contaminated with radioactive waste (red), wastewater and activated sludge (light blue), freshwater sediments and saturated soils (medium blue), and marine sediments (navy blue). The three major clade divisions are labeled G1, G2, and G3.

The phylogenetic tree shown in Fig. 2 was statistically analyzed in UniFrac (31) to determine possible biogeographic patterns. A principal components analysis followed by cluster analysis with jackknife resampling (100 permutations) revealed two significant (P < 0.01) environmental clusters (Fig. 3). Lineage-specific analysis using the G test corrected for multiple comparisons (31) was used to determine if certain clades were significantly enriched in particular environments. G1, G2, and G3 have environmental patterns that are significantly nonrandom (P < 0.001). G1 has an overrepresentation of sequences from grassland and prairie soils; G2 has an overrepresentation of agricultural soils and organically contaminated soils. At a finer phylogenetic resolution (family and genus levels), most small clades are randomly distributed and include members of the Gemmatimonadetes from a variety of environments and locations, suggesting a generalist ecological strategy and adaptation to a variety of environments.

Fig. 3.

Principal components analysis of sequences of the phylum Gemmatimonadetes shown in Fig. 2 (n = 456) by environment type. Soil types are as follows: crop agriculture (SoilAg); alpine (SoilAlpine); arid and desert (SoilArid); grassland, prairie, and pasture (SoilGrass); and forest (SoilForest). Also included are soils contaminated with organics (SoilContamOrg) and radioactive wastes (SoilContamRadio), freshwater sediments and saturated soils (SedimentsFW), marine sediments (SedimentsMar), rock surfaces (Rock), and wastewater and activated sludge (WWAS). Significant clusters are circled.

Seasonal quantification of Gemmatimonadetes in KBS soils.

A quantitative PCR assay targeting Gemmatimonadetes was used to quantify this phylum in soils at the Kellogg Biological Station (KBS) LTER in Michigan. Samples were collected over the 2008 season from replicate plots under five different land treatments. Plots included two agriculture types: (i) conventional till and chemical input with a corn-soybean-wheat rotation (replicate plots T1R1, T1R4, and T1R5) and (ii) organically managed plots with a corn-soybean-wheat rotation and a vetch winter cover crop (T4R2, T4R3, T4R5). Other plots included early succession fields maintained by annual burning (T7R1, T7R3, T7R5) and annual mowing (T8R1, T8R2, T8R3) and a mid-succession forest plot ca. 50 years postagricultural abandonment (SF2). For each date, composite soil samples (15 cores) were collected from the upper 15 cm of each plot and sieved through 2-mm mesh. DNA was extracted using a PowerSoil DNA extraction kit (MoBio) according to the manufacturer's protocol and quantified using a NanoDrop spectrophotometer.

To quantify Gemmatimonadetes seasonal abundances, a quantitative PCR (qPCR) assay targeting their 16S rRNA genes was developed. A primer to target phylum Gemmatimonadetes had been previously designed in silico (34); however, a search of this primer against the RDP database revealed matches to other phyla, especially Actinobacteria and Firmicutes. For our study, primer G1G3-673F (5′-GAATGCGTAGAGATCC) was designed in Primrose (2) to target clades G1 and G3. In silico, primer G1G3-673F targeted 55.6% of all full-length sequences classified as Gemmatimonadetes in RDP (2,312 sequences). The primer had a perfect match to 95.6%, 0%, and 100% of sequences in clades G1, G2, and G3, respectively. G1G3-673F matched only 11 of 7,446 type strains from other phyla and was thus considered to be Gemmatimonadetes specific. Primer specificity was confirmed by melt curve analysis after each run and by sequencing PCR products (reactions described below) from three randomly selected samples: PCR products (ca. 243 bp) were cloned into the pGEM-T Easy vector (Promega) and sequenced on an ABI 3730 DNA analyzer (Applied Biosystems) at the Molecular Biology Resource Facility at the University of Tennessee. Of the 27 sequences obtained, 70% were classified as Gemmatimonadetes with RDP Classifier (>80% similarity) and represented a diversity comparable to that of clades G1 and G3 (>78% sequence identity). Quantitation standards for qPCR were created by cloning the 16S rRNA gene from strain KBS708, a Gemmatimonadetes isolate from KBS soil (J. M. DeBruyn et al., unpublished data). The 16S rRNA gene was PCR amplified using universal bacterial primers 8F and 1392R and cloned and sequenced as described above (GenBank accession no. HM154525). Plasmids were diluted to create an 8-point standard curve (10¹ to 10⁸ copies per reaction). PCRs were done in triplicate and consisted of 12.5 μl SYBR Premix II Taq (Takara), 400 nM G1G3-637F primer, 400 nM universal 16S rRNA primer 907R, and 2.5 ng template DNA extracted from soil. Dilutions of template DNA (1:10 and 1:100) were used to determine a 93% PCR efficiency with a lower detection limit of 100 gene copies per reaction. Template-free negative controls were run in parallel. Reactions were performed on a Bio-Rad CFX96 thermocycler as follows: 95°C for 30 s, followed by 40 cycles of 95°C for 5 s, 53°C for 20 s, and 72°C for 20 s. Gemmatimonadetes abundances were expressed as a proportion of total 16S rRNA genes, quantified using universal bacterial primers 1055F and 1392R as described previously (18).

Gemmatimonadetes 16S rRNA gene copies were detected in all KBS plots, indicating that they are a persistent member of these communities (Fig. 4). Mean relative abundances of G1 and G3 ranged from 0.09% to 5.31% of total 16S rRNA gene copies. These percentages are in agreement with results from other soil studies (Fig. 1). A nonparametric analysis of variance (ANOVA) (Kruskal-Wallis; F) revealed no differences by land treatment. However, there were significant differences by date (F = 19.32; P < 0.0001); relative abundances in October and November were significantly different from those of the other dates (post hoc multiple comparison Z test). The only significant difference between the five land treatments was observed in September, when the relative abundance of Gemmatimonadetes was higher in agriculture and field plots (T1, T4, T7, T8) than in forest plots (SF2) (Mann-Whitney U test; P < 0.001). The relative increase in Gemmatimonadetes abundance in September is largely driven by a decrease in total bacterial abundance. There were no significant changes in other parameters, and field logs do not indicate activity around this time, so it is unclear what may have driven this change. Other studies have indicated that forest and agriculture soil communities have seasonal dynamics (22) and that agricultural (crop and pasture) soils are enriched with Gemmatimonadetes compared to forest soils (22, 29).

Fig. 4.

Relative abundance of Gemmatimonadetes bacteria (clades G1 and G3) in five soil types at the Kellogg Biological Station over time. Quantities are expressed as a percentage of the total bacterial 16S rRNA genes. Land treatments include conventional agriculture (T1), organic agriculture (T4), early succession meadows (T7 and T8), and mid-succession forest (SF2) (see text for descriptions). Each point represents the mean and standard deviation of 3 replicate land plots.

Gemmatimonadetes abundances were compared to edaphic parameters. Percent soil moisture (θ) was determined gravimetrically after oven drying. Soil pH was determined by combining 10 g soil with 10 ml deionized H₂O. Nitrogen and carbon content were determined by using dry combustion on a Carlo-Erba C/N analyzer, and phosphorus was measured by the molybdate-blue colorimetric procedure after Mehlich 3 extraction (43). Plant diversity data (Shannon H index), determined by noncrop biomass and richness at harvest, were obtained from the KBS LTER website (http://lter.kbs.msu.edu/) and are used with permission. Pearson's correlation analysis revealed that total bacterial 16S rRNA gene copies were significantly correlated with several edaphic parameters (Table 1). In contrast, the only parameter that significantly correlated with Gemmatimonadetes abundances was moisture, which ranged from 8.7% to 61.7% during this study. The highest relative Gemmatimonadetes abundances were observed during periods with the lowest soil moisture (Table 1; Fig. 5 A). (March samples were excluded from the analysis because much of the moisture was ice.) Gemmatimonadetes proportions were best predicted by the following linear regression model: log (%G1G3) = −1.957 − 1.363 × θ (r² = 0.113; F = 9.538; P = 0.0028), where θ is percent soil moisture and %G1G3 is the percentage of Gemmatimonadetes in the bacterial community. Multivariate regression was attempted; however, the fit of the univariate models could not be improved by addition of the other parameters measured in this study (data not shown). Many microbial activities are known to increase with soil moisture (15, 21, 42); however, microbial community structure responses to moisture are less clear. Some studies have found shifts in soil microbial community composition under different moisture regimens (11, 15), while others have reported no relationship with soil moisture changes (3, 29). In this study, Gemmatimonadetes relative abundances were inversely correlated to moisture. There were also many Gemmatimonadetes phylotypes in libraries from semiarid and arid soils and deserts (1, 6, 8, 9, 26, 32) and higher relative abundances in these environments (Fig. 1 and 2). In combination, this evidence suggests either an adaptation to drier soils or, alternatively, that Gemmatimonadetes may be outcompeted when moisture is available. In microbial communities associated with soil microaggregates, a much higher percentage of Gemmatimonadetes (10 to 32% of the community) was found in the inner microaggregates than in the whole aggregates (36). This may be indicative of an adaptation to low-moisture (or low-oxygen) conditions typical of inner aggregates. Desiccation tolerance may have contributed to high dispersal rates, leading to the observed cosmopolitan biogeography of this phylum.

Table 1.

Correlation coefficients between log transformed gene quantities and soil parameters

Parametera	Pearson coefficient (r) forb:
	Total bacterial 16S rRNA gene copies	%G1G3
%G1G3	−0.676
Moisture	0.409	−0.336
pH	−0.239	0.181
N	0.307	−0.020
C	0.295	−0.038
C/N	0.049	−0.144
PO4	−0.015	0.032
PlantH	0.357	−0.082

^aGemmatimonadetes 16S rRNA gene quantities (%G1G3) are normalized to total 16S quantities. C/N, carbon-to-nitrogen ratio; PlantH, Shannon diversity of noncrop plants.

^bThe significance of each correlation is indicated by italics (P < 0.05) and bold (P < 0.01) (n = 77).

Fig. 5.

Gemmatimonadetes 16S rRNA gene copies (expressed as a percentage of total bacterial 16S rRNA gene copies) related to percent soil moisture (A) and soil pH (B). Land treatments include conventional agriculture (T1), organic agriculture (T4), early succession meadows (T7 and T8), and mid-succession forest (SF2) (see text for descriptions).

Other studies have reported higher relative abundances of Gemmatimonadetes in soils near neutral pH than in acidic soils (28, 32, 48). We observed a slight (but not significant) relationship to soil pH, which ranged from 3.7 to 6.3 over the course of the study (Fig. 5B). The highest relative abundances were observed near neutral pH, while low abundances were observed across all pHs, suggesting that after moisture, pH may act as a secondary constraint on Gemmatimonadetes in these soils.

Gemmatimonadetes bacteria have a cosmopolitan distribution in terrestrial systems, and their consistent abundance implicate them as persistent and important members of soil communities. Spatial and temporal abundances of Gemmatimonadetes bacteria across five land treatments at KBS has demonstrated that their relative abundances were not related to land management but were inversely correlated to soil moisture, suggesting an adaptation to drier soils.