Microb Ecol DOI 10.1007/s00248-014-0510-6
SOIL MICROBIOLOGY
Distribution and Interaction Patterns of Bacterial Communities in an Ornithogenic Soil of Seymour Island, Antarctica Pabulo Henrique Rampelotto & Anthony Diego Muller Barboza & Antônio Batista Pereira & Eric W. Triplett & Carlos Ernesto G. R. Schaefer & Flávio Anastácio de Oliveira Camargo & Luiz Fernando Wurdig Roesch
Received: 11 August 2014 / Accepted: 8 October 2014 # Springer Science+Business Media New York 2014
Abstract Next-generation, culture-independent sequencing offers an excellent opportunity to examine network interactions among different microbial species. In this study, soil bacterial communities from a penguin rookery site at Seymour Island were analyzed for abundance, structure, diversity, and interaction networks to identify interaction patterns among the various taxa at three soil depths. The analysis revealed the presence of eight phyla distributed in different proportions among the surface layer (0–8 cm), middle layer (20–25 cm), and bottom (35–40 cm). The bottom layer presented the highest values of bacterial richness, diversity, and evenness when compared to surface and middle layers. The network analysis revealed the existence of a unique pattern of interactions in which the soil microbial network formed a clustered topology, rather than a modular structure as is usually found in biological communities. In addition, specific taxa were
Electronic supplementary material The online version of this article (doi:10.1007/s00248-014-0510-6) contains supplementary material, which is available to authorized users. P. H. Rampelotto : A. D. M. Barboza : A. B. Pereira : L. F. W. Roesch (*) Universidade Federal do Pampa, Campus São Gabriel, Av. Antonio Trilha, 1847, 97300-000 São Gabriel, Rio Grande do Sul, Brazil e-mail: [email protected] E. W. Triplett Microbiology and Cell Science Department, Institute of Food and Agricultural Sciences, University of Florida, Gainesville, FL 32611-0700, USA C. E. G. R. Schaefer Departamento de Solos, Universidade Federal de Viçosa, AV PH Rolfs s/n, Viçosa 36570-000, MG, Brazil F. A. de Oliveira Camargo Departamento de Solos, Universidade Federal do Rio Grande do Sul, Av. Bento Gonçalves, 7712, 91540-000 Porto Alegre, Rio Grande do Sul, Brazil
identified as important players in microbial community structure. Furthermore, simulation analyses indicated that the loss of potential keystone groups of microorganisms might alter the patterns of interactions within the microbial community. These findings provide new insights for assessing the consequences of environmental disturbances at the wholecommunity level in Antarctica. Keywords 16S rRNA gene . Co-occurrence . Ion PGM sequencing . Microbiome . Network analysis
Introduction The advent of high-throughput sequencing has allowed for an unprecedented description of microbial communities, opening a new era in microbiology. These advanced new techniques have been used extensively during recent years to improve our understanding of how these communities assemble, evolve, and function [1–3]. Most of the studies to date have focused on the abundance and structure of microbial communities to assess their diversity across different environments [4, 5]. Nevertheless, analytical approaches as network analysis may shed new lights in understanding of microbial ecology and environmental microbiology. Through such network interactions, the ecosystem is capable of performing system-level functions (e.g., nutrient cycling, ecosystem stability) not be achieved by individual species [6, 7]. Detailed information on the microbial taxa is essential to determine whether or not the interaction patterns found are statistically significant [8]. In this context, the large amount of high-quality data generated by the Ion Torrent technology offers an unprecedented opportunity to examine network interactions among different microbial species. Such information is particularly valuable in assessing Antarctic microbial
P. H. Rampelotto et al.
communities, where the basic ecology and evolution of most microbes remains unknown. Furthermore, Antarctica provides key models of ecosystems to study microbial ecology due to the low complexity of its microbial network in comparison to other ecosystems and due to reduced influence of humans, plants, and animals (other than penguins) [9]. Nevertheless, the Antarctic soils are usually poor in organic material and present low levels of weathering and leaching [10], which affects the development of microbial communities. Given the temperature profile of the environment, the vast majority of active microbes are expected to be extremophiles (or at least highly specialized). Because of that, the microbial community and interactions might differ from more “temperate” or “moderate” ecosystems. Ornithogenic soils are of particular interest because they are formed by the constant importation of organic and biological material by the feeding and nesting activity of penguins. Indeed, ornithogenic sites constitute the most important reservoirs of organic carbon in Antarctic terrestrial ecosystems [11]. Recent studies indicate that the bacterial community in penguin guano is not only one of the richest in Antarctica but also extremely diverse phylogenetically, morphologically, and physiologically [12, 13]. The number of microbial diversity studies in different Antarctic environments is still relatively small. Improving our understanding of the interactions among the complex microbial communities in this environment may allow us to predict how these communities react to changes in the environment. This is particularly important given the vulnerability of Antarctica to global climate changes. In this work, soil samples were collected from a rookery site at different depths in order to identify and analyze the bacterial communities at Seymour Island (Antarctic Peninsula) with respect to relative abundance, structure, diversity, and network interactions. These data were then used to identify patterns among the various groups of bacteria present in this ornithogenic site.
Materials and Methods Site description, Soil Sampling, and Physicochemical Analysis A penguin rookery site (64°17′36.87″ S, 56°41′28.12″ W, 8 m a.s.l.) at Seymour Island, an ice-free island located near the northern tip of the Antarctic Peninsula, was selected because of the constant avian importation of organic/ biological material and lack of vegetative cover. The climate of Seymour is subpolar-semiarid, with an annual mean temperature between −5 and −10 °C. Soil texture is sandy loam and it was classified as Ornithic-Oxyaquic Cryosol (Soil Taxonomy, USA) or Ornithic-Salic Leptosol (WRB/FAO). Triplicate soil samples were collected from a penguin rookery site
at three depths using a clean sterilized stainless steel scoop: surface layer (0–8 cm, actively penguin-colonized soil), middle layer (20–25 cm), and bottom layer (35–40 cm). The sampling strategy and the determination of soil physicochemical properties were described previously [14]. All soil samples (n=9) were placed into 50-ml sterile plastic tubes and stored in ice chests upon collection and transported by the Brazilian Navy at −20 °C to the laboratory (Centro Interdisciplinar de Pesquisas em Biotecnologia–UNIPAMPA, São Gabriel, RS, Brazil) for DNA extraction and physicochemical characterization. Upon arrival (after about 21 days of transportation), the DNA extraction was performed immediately. DNA Extraction, Short Amplicon Libraries Preparation, and Sequencing Microbial DNA was extracted from 1 g of each soil sample using the PowerSoil® DNA Kit (MoBio, USA) according to the manufacturer’s instructions. DNA concentrations and purity were determined using a NanoVue™ spectrophotometer (GE Healthcare, USA), and all DNA samples were stored at −20 °C. Three independent PCR reactions of DNA from each of the soil samples (technical replicates) were performed using the 917F and 1046R primers for the amplification of approximately 130 bp of the V6 region of the 16S rRNA gene. PCR was performed with the High Fidelity PCR Enzyme Mix (Thermo Scientific, USA). The mixtures contained 5 μl of ×10 high-fidelity PCR buffer with 15 mM MgCl2, 0.2 mM of each dNTPs, 100 mM of each primer, 2.5 U of high-fidelity PCR enzyme mix, and approximately 100 ng of DNA template in a final volume of 50 μl. The PCR conditions were 94 °C for 2 min, 25 cycles of 94 °C for 45 s; 56 °C for 30s; and 72 °C for 30s extension; followed by 72 °C for 4 min. Prior to Ion Torrent PGM sequencing, the short amplicon libraries were processed in order to add the barcoded adaptors A and P1 necessary for sequencing. The adaptors were added to the amplicons using the Ion Plus Fragment Library Kit and the Ion Xpress™ Barcode Adapters (Life Technologies, USA). The reactions were performed based on the user bulletin MAN0006846 revision 3.0 available at http:// i o n c o m m u n i t y. l i f e t e c h n o l o g i e s . c o m w i t h m i n o r modifications during the amplicon purification step as follows: the bead suspension with the DNA was incubated with the Agencourt® AMPure® XP Reagent (Beckman Coulter, USA) (×2 sample volume) at room temperature for 10 min, and all washing steps were performed with 500 μl of freshly prepared 80 % ethanol during 30 s. All the other steps for preparing short amplicon libraries (end-repair, barcoded adaptors ligation, and nik-repair) were performed according to the user bulletin mentioned above. The following barcodes TTCCGATAAC, TGAGCGGAAC, CTGACCGAAC, TCCT CGAATC, TAGGTGGTTC, TCTAACGGAC, TTGGAGTG
Bacterial Interaction Patterns in Antarctic Soil
TC, TCTAGAGGTC, and TCTGGATGAC were added to the short amplicons to soil samples 008, 009, and 010 (0–8 cm deep), 011, 012, and 013 (20–22 cm deep), and 014, 015, and 016 (35–40 cm deep), respectively. All barcoded amplicons were quantified by quantitative real-time PCR using Ion Library Quantitation Kit and the Applied Biosystems® 7500 Fast Real-Time PCR System according to the manufacturer’s instructions. The samples were adjusted to 15×106 molecules per microliter and mixed in equal amounts to obtain an equimolar pool of amplicons that was used for template amplification onto Ion Sphere™ Particles (ISPs). The templatepositive ISPs containing clonally amplified 16S rRNA genes were prepared with the Ion OneTouch™ System using the Ion OneTouch™ 200 Template Kit v2 following the user guide Publication Number 4478372 Revision B (available at http:// ioncommunity.lifetechnologies.com). The resulting ISPs were sequenced on an Ion 316™ microchip using the Ion Torrent Personal Genome Machine (Life Technologies, USA) and the Ion PGM™ 200 Sequencing Kit following the workflow from the user guide part number 4474246. After sequencing, the sequence reads were filtered within the PGM software that removed low-quality and polyclonal sequences. All PGMfiltered data were exported as a FastQ file that was used for the subsequent bioinformatics analysis. Raw sequences were submitted to the NCBI Sequence Read Archive under the study number SRP010043, experiment number SRX525897. 16S rRNA Reads Processing for Downstream Analyses The FastQ file exported from the Ion PGM™ System was processed using Mothur v.1.30.2 [15] with the same pipeline adopted by Lupatini et al. [16]. The multiplexed reads were first filtered for quality and assigned to the starting soil samples. The filtering criteria removed any sequence that (1) contained a homopolymer greater than 8 bases, (2) contained any ambiguous base call, (3) had more than one mismatch to the barcode sequence, and (4) were smaller than 100 bases in length. Also, the sequences were quality screened using a moving window that was 50-base long. Within that window, any read was removed with an average quality score (inferred as Phred score) below 25. Following this first step, the dataset was simplified by obtaining a non-redundant set of sequences that were further aligned against the SILVA reference alignment (http://www.arb-silva.de/). As the barcodes were added after the PCR reaction, sequences in both directions were expected (forward and reverse) than the flip parameter was applied to reverse complement sequences when 50 % of the bases are removed in the alignment, to produce a better alignment and keep all sequences in the same direction. To maximize the number of sequences that overlap over the longest span, any sequence that starts after the position that 85 % of the sequences do, or ends before the position that 85 % of the sequences do, were removed from the alignment.
The alignment was trimmed allowing comparing only those sequences that overlap the same region. Finally, to reduce sequencing noise, a pre-clustering step [17] was applied and the chimeric sequences were checked using the chimera.slayer command. Alpha and Beta Diversity Analysis Community diversity, community evenness, and sequence coverage were estimated using the 97 % similarity cutoff for the definition of an operational taxonomic unit (OTU). The alpha calculators, relating the taxon organization in a single sample, were implemented in Mothur v.1.30.2 [15], and according to the recommendations of Lemos et al. [18], all calculations were performed using a subsample of 7336 sequences (the size of the smallest library). High-quality sequences obtained after data processing were used to generate a distance matrix by calculating the uncorrected pairwise distances between aligned DNA sequences. The distances were used to assign sequences to OTUs at the 97 % similarity level using the cluster command with an average neighbor algorithm. Finally, the output was used for building a table with the OTU abundance for each sample and these abundances were used to calculate the alpha diversity estimators. Beta diversity, relating the taxon organization among different samples, was analyzed by using principal coordinates analysis (PCoA). The calculations were performed within the QIIME pipeline [19]. A matrix using the UniFrac metric (weighted and unweighted) for each pair of environments was calculated. The distances were turned into points in space with the number of dimensions one less than the number of samples. The first three principal dimensions were used to plot a three-dimensional graph that was visualized using KING [20]. To test whether the results were robust to sample size, a sequence-jackknifing technique was applied. The PCoA clusters were regenerated using a subset of 5000 sequences (corresponding to about 70 % of the total number of sequences obtained in the sample with the smaller number of sequences) randomly selected from each soil for 100 replicate trials. In addition, to see which taxa were more prevalent in different areas of the PCoA plot, the ten most abundant class-level taxa were added to the PCoA plots. Interaction Network Analysis The interaction network analysis was performed at 97 % similarity cutoff for operational taxonomic unit (OTU) definition. To test for co-occurrence patterns, taxa that were present in all nine samples were identified and poorly represented OTUs were removed (OTUs with less than five sequences). For network inference, all possible Spearman’s rank correlations were calculated between shared OTUs. Co-occurrence was considered robust when the Spearman’s correlation
P. H. Rampelotto et al.
coefficient was both ≥0.8 and statistically significant (p value ≥0.05). All statistical analyses were carried out using Mothur v.1.30.2 [15]. The network was explored and visualized with the interactive platform gephi [21].
layer and at low abundance in the middle layer. Approximately 16.7±5.5 % of the sequences remained unclassified. However, the unclassified reads were more abundant in layer 3 compared to the surface and middle layers. Diversity and Structure of Bacterial Communities
Results Soil Physicochemical Properties Significant differences in soil physiochemical properties were observed in the surface and middle layers when compared to the bottom layer (Table 1). The surface layer and the middle layer presented higher levels of P, K, Na, and Mg compared to the bottom layer, but lower concentrations of Ca. These layers also presented high H+Al in the surface/sub-surface. Furthermore, the neutral pH observed indicates that the organic materials deposited by penguins are poorly decomposed. Recent deposits of guano usually have neutral to alkaline pH, becoming progressively more acidic with the microbial transformation of these materials, which release strong acids in the soil as nitric and sulfuric acids. Composition and Distribution of Bacterial Communities Eight bacterial phyla were observed in these samples (Table 2). The dominant phyla were Actinobacteria, Proteobacteria, Firmicutes, and Bacteriodetes. The four phyla in relatively low abundance were Verrucomicrobia, TM7, Fusobacteria, and Aquificae. Significant differences in the distribution of these phyla among the three layers were observed. Members of the phylum Actinobacteria were found in high concentration at the surface layer when compared to the middle and bottom layers. Actinomycetales was the most abundant order within this phylum. In contrast, the phylum Proteobacteria, the second most abundant, was in low abundance at the surface layer but in high abundance in the middle layer. Gammaproteobacteria was the dominant group within this phylum. Members of the phylum Firmicutes were present at a high abundance in the surface layer and at a low abundance in the bottom layer. The phylum Bacteroidetes, specially the order Flavobacteriales, was in high abundance at the surface
To analyze how well each sample was representative of the bacterial community in the environment, sequence coverage was calculated (Table 3). Even in the sample with the lowest number of sequences (i.e., 7336 sequences), it was possible to achieve more than 93 % coverage using grouping criteria of 3 % dissimilarity. The diversity indices indicate that the samples presented different degrees of richness, diversity, and evenness (Table 3). The highest values of richness, diversity, and evenness were in the bottom layer compared to the surface and middle layers. In terms of bacterial community structure, the unweighted and weighted UniFrac analyses showed that the soil bacterial communities from the three layers were quite different from each other (Fig. 1a, b). Relatively low variation (52 %) was explained by the first three axes with jackknifed unweighted PCoA. On the other hand, the first three axis of the weighted jackknifed PCoA accounted for 93.7 % of the variation, indicating that the overall differences between the clusters were more related to the abundance of specific OTUs than to their presence or absence. The orders Lactobacillales (Firmicutes), Desulfuromonadales (Proteobacteria), and Fusobacteriales (Fusobacteria) were the main groups responsible for the specificity of the surface layer. Solirubrobacterales (Actinobacteria), Xanthomonadales, and Enterobacteriales (Proteobacteria) were the main groups responsible for the specificity of the bottom layer. For both unweighted and weighted PCoA, the middle and bottom layers were more associated to each other than the surface layer. Shared OTUs were detected within the three soil profiles. Assuming that the three soil layers are different (see Table 1), the shared OTUs were divided into the following three general categories based on how widespread or restricted each OTU was: restricted (i.e., bacterial taxa ∼10 times more abundant in one of the three layers), mid-restricted (i.e., bacterial taxa ∼3– 4 times more abundant in one of the three layers), and widespread (bacterial taxa broadly distributed in the three layers).
Table 1 Soil physicochemical properties of a penguin rookery site at Seymour Island, Antarctica Depth (cm)
pH
P mg/dm3
K
Na
Mg2+ Ca2+ cmolc/dm3
0–8 20–22 35–40
7.01 7.34 6.04
1927.7 1984.6 34.2
2150.0 1501.0 877.0
2571.0 1418.5 241.0
0.42 1.94 6.00
Surface layer Middle layer Bottom layer a
Soil organic carbon
5.06 6.75 0.58
Al3+
H+Al
SOCa g/kg
0.0 0.0 0.0
9.8 4.7 2.6
9.78 4.56 0.78
Bacterial Interaction Patterns in Antarctic Soil Table 2 Relative abundance of phyla in three different depths of a penguin rookery site at Seymour Island, Antarctica Taxonomy
Total
SD
Surface layer 0–8 cm
SD
Middle layer 20–22 cm
SD
Bottom layer 35–40 cm
SD
38.5 7.6 23.7 17.8 0 0 0 0 12.4
0.4 1.6 5.8 4.2 0 0 0 0 1.0
25.2 39.8 14.6 6.3 0.1 0.1 0.1 0.1 13.9
2.2 1.6 3.8 2.5 0 0 0 0 0.8
24.9 28.5 10.6 11.0 1 0.1 0 0.1 23.8
0.9 1.6 1.5 1.5 0.1 0 0 0 2.6
Percent of total Actinobacteria Proteobacteria Firmicutes Bacteroidetes Verrucomicrobia TM7 Fusobacteria Aquificae Other
29.5 25.3 16.3 11.7 0.3
SOIL MICROBIOLOGY
Distribution and Interaction Patterns of Bacterial Communities in an Ornithogenic Soil of Seymour Island, Antarctica Pabulo Henrique Rampelotto & Anthony Diego Muller Barboza & Antônio Batista Pereira & Eric W. Triplett & Carlos Ernesto G. R. Schaefer & Flávio Anastácio de Oliveira Camargo & Luiz Fernando Wurdig Roesch
Received: 11 August 2014 / Accepted: 8 October 2014 # Springer Science+Business Media New York 2014
Abstract Next-generation, culture-independent sequencing offers an excellent opportunity to examine network interactions among different microbial species. In this study, soil bacterial communities from a penguin rookery site at Seymour Island were analyzed for abundance, structure, diversity, and interaction networks to identify interaction patterns among the various taxa at three soil depths. The analysis revealed the presence of eight phyla distributed in different proportions among the surface layer (0–8 cm), middle layer (20–25 cm), and bottom (35–40 cm). The bottom layer presented the highest values of bacterial richness, diversity, and evenness when compared to surface and middle layers. The network analysis revealed the existence of a unique pattern of interactions in which the soil microbial network formed a clustered topology, rather than a modular structure as is usually found in biological communities. In addition, specific taxa were
Electronic supplementary material The online version of this article (doi:10.1007/s00248-014-0510-6) contains supplementary material, which is available to authorized users. P. H. Rampelotto : A. D. M. Barboza : A. B. Pereira : L. F. W. Roesch (*) Universidade Federal do Pampa, Campus São Gabriel, Av. Antonio Trilha, 1847, 97300-000 São Gabriel, Rio Grande do Sul, Brazil e-mail: [email protected] E. W. Triplett Microbiology and Cell Science Department, Institute of Food and Agricultural Sciences, University of Florida, Gainesville, FL 32611-0700, USA C. E. G. R. Schaefer Departamento de Solos, Universidade Federal de Viçosa, AV PH Rolfs s/n, Viçosa 36570-000, MG, Brazil F. A. de Oliveira Camargo Departamento de Solos, Universidade Federal do Rio Grande do Sul, Av. Bento Gonçalves, 7712, 91540-000 Porto Alegre, Rio Grande do Sul, Brazil
identified as important players in microbial community structure. Furthermore, simulation analyses indicated that the loss of potential keystone groups of microorganisms might alter the patterns of interactions within the microbial community. These findings provide new insights for assessing the consequences of environmental disturbances at the wholecommunity level in Antarctica. Keywords 16S rRNA gene . Co-occurrence . Ion PGM sequencing . Microbiome . Network analysis
Introduction The advent of high-throughput sequencing has allowed for an unprecedented description of microbial communities, opening a new era in microbiology. These advanced new techniques have been used extensively during recent years to improve our understanding of how these communities assemble, evolve, and function [1–3]. Most of the studies to date have focused on the abundance and structure of microbial communities to assess their diversity across different environments [4, 5]. Nevertheless, analytical approaches as network analysis may shed new lights in understanding of microbial ecology and environmental microbiology. Through such network interactions, the ecosystem is capable of performing system-level functions (e.g., nutrient cycling, ecosystem stability) not be achieved by individual species [6, 7]. Detailed information on the microbial taxa is essential to determine whether or not the interaction patterns found are statistically significant [8]. In this context, the large amount of high-quality data generated by the Ion Torrent technology offers an unprecedented opportunity to examine network interactions among different microbial species. Such information is particularly valuable in assessing Antarctic microbial
P. H. Rampelotto et al.
communities, where the basic ecology and evolution of most microbes remains unknown. Furthermore, Antarctica provides key models of ecosystems to study microbial ecology due to the low complexity of its microbial network in comparison to other ecosystems and due to reduced influence of humans, plants, and animals (other than penguins) [9]. Nevertheless, the Antarctic soils are usually poor in organic material and present low levels of weathering and leaching [10], which affects the development of microbial communities. Given the temperature profile of the environment, the vast majority of active microbes are expected to be extremophiles (or at least highly specialized). Because of that, the microbial community and interactions might differ from more “temperate” or “moderate” ecosystems. Ornithogenic soils are of particular interest because they are formed by the constant importation of organic and biological material by the feeding and nesting activity of penguins. Indeed, ornithogenic sites constitute the most important reservoirs of organic carbon in Antarctic terrestrial ecosystems [11]. Recent studies indicate that the bacterial community in penguin guano is not only one of the richest in Antarctica but also extremely diverse phylogenetically, morphologically, and physiologically [12, 13]. The number of microbial diversity studies in different Antarctic environments is still relatively small. Improving our understanding of the interactions among the complex microbial communities in this environment may allow us to predict how these communities react to changes in the environment. This is particularly important given the vulnerability of Antarctica to global climate changes. In this work, soil samples were collected from a rookery site at different depths in order to identify and analyze the bacterial communities at Seymour Island (Antarctic Peninsula) with respect to relative abundance, structure, diversity, and network interactions. These data were then used to identify patterns among the various groups of bacteria present in this ornithogenic site.
Materials and Methods Site description, Soil Sampling, and Physicochemical Analysis A penguin rookery site (64°17′36.87″ S, 56°41′28.12″ W, 8 m a.s.l.) at Seymour Island, an ice-free island located near the northern tip of the Antarctic Peninsula, was selected because of the constant avian importation of organic/ biological material and lack of vegetative cover. The climate of Seymour is subpolar-semiarid, with an annual mean temperature between −5 and −10 °C. Soil texture is sandy loam and it was classified as Ornithic-Oxyaquic Cryosol (Soil Taxonomy, USA) or Ornithic-Salic Leptosol (WRB/FAO). Triplicate soil samples were collected from a penguin rookery site
at three depths using a clean sterilized stainless steel scoop: surface layer (0–8 cm, actively penguin-colonized soil), middle layer (20–25 cm), and bottom layer (35–40 cm). The sampling strategy and the determination of soil physicochemical properties were described previously [14]. All soil samples (n=9) were placed into 50-ml sterile plastic tubes and stored in ice chests upon collection and transported by the Brazilian Navy at −20 °C to the laboratory (Centro Interdisciplinar de Pesquisas em Biotecnologia–UNIPAMPA, São Gabriel, RS, Brazil) for DNA extraction and physicochemical characterization. Upon arrival (after about 21 days of transportation), the DNA extraction was performed immediately. DNA Extraction, Short Amplicon Libraries Preparation, and Sequencing Microbial DNA was extracted from 1 g of each soil sample using the PowerSoil® DNA Kit (MoBio, USA) according to the manufacturer’s instructions. DNA concentrations and purity were determined using a NanoVue™ spectrophotometer (GE Healthcare, USA), and all DNA samples were stored at −20 °C. Three independent PCR reactions of DNA from each of the soil samples (technical replicates) were performed using the 917F and 1046R primers for the amplification of approximately 130 bp of the V6 region of the 16S rRNA gene. PCR was performed with the High Fidelity PCR Enzyme Mix (Thermo Scientific, USA). The mixtures contained 5 μl of ×10 high-fidelity PCR buffer with 15 mM MgCl2, 0.2 mM of each dNTPs, 100 mM of each primer, 2.5 U of high-fidelity PCR enzyme mix, and approximately 100 ng of DNA template in a final volume of 50 μl. The PCR conditions were 94 °C for 2 min, 25 cycles of 94 °C for 45 s; 56 °C for 30s; and 72 °C for 30s extension; followed by 72 °C for 4 min. Prior to Ion Torrent PGM sequencing, the short amplicon libraries were processed in order to add the barcoded adaptors A and P1 necessary for sequencing. The adaptors were added to the amplicons using the Ion Plus Fragment Library Kit and the Ion Xpress™ Barcode Adapters (Life Technologies, USA). The reactions were performed based on the user bulletin MAN0006846 revision 3.0 available at http:// i o n c o m m u n i t y. l i f e t e c h n o l o g i e s . c o m w i t h m i n o r modifications during the amplicon purification step as follows: the bead suspension with the DNA was incubated with the Agencourt® AMPure® XP Reagent (Beckman Coulter, USA) (×2 sample volume) at room temperature for 10 min, and all washing steps were performed with 500 μl of freshly prepared 80 % ethanol during 30 s. All the other steps for preparing short amplicon libraries (end-repair, barcoded adaptors ligation, and nik-repair) were performed according to the user bulletin mentioned above. The following barcodes TTCCGATAAC, TGAGCGGAAC, CTGACCGAAC, TCCT CGAATC, TAGGTGGTTC, TCTAACGGAC, TTGGAGTG
Bacterial Interaction Patterns in Antarctic Soil
TC, TCTAGAGGTC, and TCTGGATGAC were added to the short amplicons to soil samples 008, 009, and 010 (0–8 cm deep), 011, 012, and 013 (20–22 cm deep), and 014, 015, and 016 (35–40 cm deep), respectively. All barcoded amplicons were quantified by quantitative real-time PCR using Ion Library Quantitation Kit and the Applied Biosystems® 7500 Fast Real-Time PCR System according to the manufacturer’s instructions. The samples were adjusted to 15×106 molecules per microliter and mixed in equal amounts to obtain an equimolar pool of amplicons that was used for template amplification onto Ion Sphere™ Particles (ISPs). The templatepositive ISPs containing clonally amplified 16S rRNA genes were prepared with the Ion OneTouch™ System using the Ion OneTouch™ 200 Template Kit v2 following the user guide Publication Number 4478372 Revision B (available at http:// ioncommunity.lifetechnologies.com). The resulting ISPs were sequenced on an Ion 316™ microchip using the Ion Torrent Personal Genome Machine (Life Technologies, USA) and the Ion PGM™ 200 Sequencing Kit following the workflow from the user guide part number 4474246. After sequencing, the sequence reads were filtered within the PGM software that removed low-quality and polyclonal sequences. All PGMfiltered data were exported as a FastQ file that was used for the subsequent bioinformatics analysis. Raw sequences were submitted to the NCBI Sequence Read Archive under the study number SRP010043, experiment number SRX525897. 16S rRNA Reads Processing for Downstream Analyses The FastQ file exported from the Ion PGM™ System was processed using Mothur v.1.30.2 [15] with the same pipeline adopted by Lupatini et al. [16]. The multiplexed reads were first filtered for quality and assigned to the starting soil samples. The filtering criteria removed any sequence that (1) contained a homopolymer greater than 8 bases, (2) contained any ambiguous base call, (3) had more than one mismatch to the barcode sequence, and (4) were smaller than 100 bases in length. Also, the sequences were quality screened using a moving window that was 50-base long. Within that window, any read was removed with an average quality score (inferred as Phred score) below 25. Following this first step, the dataset was simplified by obtaining a non-redundant set of sequences that were further aligned against the SILVA reference alignment (http://www.arb-silva.de/). As the barcodes were added after the PCR reaction, sequences in both directions were expected (forward and reverse) than the flip parameter was applied to reverse complement sequences when 50 % of the bases are removed in the alignment, to produce a better alignment and keep all sequences in the same direction. To maximize the number of sequences that overlap over the longest span, any sequence that starts after the position that 85 % of the sequences do, or ends before the position that 85 % of the sequences do, were removed from the alignment.
The alignment was trimmed allowing comparing only those sequences that overlap the same region. Finally, to reduce sequencing noise, a pre-clustering step [17] was applied and the chimeric sequences were checked using the chimera.slayer command. Alpha and Beta Diversity Analysis Community diversity, community evenness, and sequence coverage were estimated using the 97 % similarity cutoff for the definition of an operational taxonomic unit (OTU). The alpha calculators, relating the taxon organization in a single sample, were implemented in Mothur v.1.30.2 [15], and according to the recommendations of Lemos et al. [18], all calculations were performed using a subsample of 7336 sequences (the size of the smallest library). High-quality sequences obtained after data processing were used to generate a distance matrix by calculating the uncorrected pairwise distances between aligned DNA sequences. The distances were used to assign sequences to OTUs at the 97 % similarity level using the cluster command with an average neighbor algorithm. Finally, the output was used for building a table with the OTU abundance for each sample and these abundances were used to calculate the alpha diversity estimators. Beta diversity, relating the taxon organization among different samples, was analyzed by using principal coordinates analysis (PCoA). The calculations were performed within the QIIME pipeline [19]. A matrix using the UniFrac metric (weighted and unweighted) for each pair of environments was calculated. The distances were turned into points in space with the number of dimensions one less than the number of samples. The first three principal dimensions were used to plot a three-dimensional graph that was visualized using KING [20]. To test whether the results were robust to sample size, a sequence-jackknifing technique was applied. The PCoA clusters were regenerated using a subset of 5000 sequences (corresponding to about 70 % of the total number of sequences obtained in the sample with the smaller number of sequences) randomly selected from each soil for 100 replicate trials. In addition, to see which taxa were more prevalent in different areas of the PCoA plot, the ten most abundant class-level taxa were added to the PCoA plots. Interaction Network Analysis The interaction network analysis was performed at 97 % similarity cutoff for operational taxonomic unit (OTU) definition. To test for co-occurrence patterns, taxa that were present in all nine samples were identified and poorly represented OTUs were removed (OTUs with less than five sequences). For network inference, all possible Spearman’s rank correlations were calculated between shared OTUs. Co-occurrence was considered robust when the Spearman’s correlation
P. H. Rampelotto et al.
coefficient was both ≥0.8 and statistically significant (p value ≥0.05). All statistical analyses were carried out using Mothur v.1.30.2 [15]. The network was explored and visualized with the interactive platform gephi [21].
layer and at low abundance in the middle layer. Approximately 16.7±5.5 % of the sequences remained unclassified. However, the unclassified reads were more abundant in layer 3 compared to the surface and middle layers. Diversity and Structure of Bacterial Communities
Results Soil Physicochemical Properties Significant differences in soil physiochemical properties were observed in the surface and middle layers when compared to the bottom layer (Table 1). The surface layer and the middle layer presented higher levels of P, K, Na, and Mg compared to the bottom layer, but lower concentrations of Ca. These layers also presented high H+Al in the surface/sub-surface. Furthermore, the neutral pH observed indicates that the organic materials deposited by penguins are poorly decomposed. Recent deposits of guano usually have neutral to alkaline pH, becoming progressively more acidic with the microbial transformation of these materials, which release strong acids in the soil as nitric and sulfuric acids. Composition and Distribution of Bacterial Communities Eight bacterial phyla were observed in these samples (Table 2). The dominant phyla were Actinobacteria, Proteobacteria, Firmicutes, and Bacteriodetes. The four phyla in relatively low abundance were Verrucomicrobia, TM7, Fusobacteria, and Aquificae. Significant differences in the distribution of these phyla among the three layers were observed. Members of the phylum Actinobacteria were found in high concentration at the surface layer when compared to the middle and bottom layers. Actinomycetales was the most abundant order within this phylum. In contrast, the phylum Proteobacteria, the second most abundant, was in low abundance at the surface layer but in high abundance in the middle layer. Gammaproteobacteria was the dominant group within this phylum. Members of the phylum Firmicutes were present at a high abundance in the surface layer and at a low abundance in the bottom layer. The phylum Bacteroidetes, specially the order Flavobacteriales, was in high abundance at the surface
To analyze how well each sample was representative of the bacterial community in the environment, sequence coverage was calculated (Table 3). Even in the sample with the lowest number of sequences (i.e., 7336 sequences), it was possible to achieve more than 93 % coverage using grouping criteria of 3 % dissimilarity. The diversity indices indicate that the samples presented different degrees of richness, diversity, and evenness (Table 3). The highest values of richness, diversity, and evenness were in the bottom layer compared to the surface and middle layers. In terms of bacterial community structure, the unweighted and weighted UniFrac analyses showed that the soil bacterial communities from the three layers were quite different from each other (Fig. 1a, b). Relatively low variation (52 %) was explained by the first three axes with jackknifed unweighted PCoA. On the other hand, the first three axis of the weighted jackknifed PCoA accounted for 93.7 % of the variation, indicating that the overall differences between the clusters were more related to the abundance of specific OTUs than to their presence or absence. The orders Lactobacillales (Firmicutes), Desulfuromonadales (Proteobacteria), and Fusobacteriales (Fusobacteria) were the main groups responsible for the specificity of the surface layer. Solirubrobacterales (Actinobacteria), Xanthomonadales, and Enterobacteriales (Proteobacteria) were the main groups responsible for the specificity of the bottom layer. For both unweighted and weighted PCoA, the middle and bottom layers were more associated to each other than the surface layer. Shared OTUs were detected within the three soil profiles. Assuming that the three soil layers are different (see Table 1), the shared OTUs were divided into the following three general categories based on how widespread or restricted each OTU was: restricted (i.e., bacterial taxa ∼10 times more abundant in one of the three layers), mid-restricted (i.e., bacterial taxa ∼3– 4 times more abundant in one of the three layers), and widespread (bacterial taxa broadly distributed in the three layers).
Table 1 Soil physicochemical properties of a penguin rookery site at Seymour Island, Antarctica Depth (cm)
pH
P mg/dm3
K
Na
Mg2+ Ca2+ cmolc/dm3
0–8 20–22 35–40
7.01 7.34 6.04
1927.7 1984.6 34.2
2150.0 1501.0 877.0
2571.0 1418.5 241.0
0.42 1.94 6.00
Surface layer Middle layer Bottom layer a
Soil organic carbon
5.06 6.75 0.58
Al3+
H+Al
SOCa g/kg
0.0 0.0 0.0
9.8 4.7 2.6
9.78 4.56 0.78
Bacterial Interaction Patterns in Antarctic Soil Table 2 Relative abundance of phyla in three different depths of a penguin rookery site at Seymour Island, Antarctica Taxonomy
Total
SD
Surface layer 0–8 cm
SD
Middle layer 20–22 cm
SD
Bottom layer 35–40 cm
SD
38.5 7.6 23.7 17.8 0 0 0 0 12.4
0.4 1.6 5.8 4.2 0 0 0 0 1.0
25.2 39.8 14.6 6.3 0.1 0.1 0.1 0.1 13.9
2.2 1.6 3.8 2.5 0 0 0 0 0.8
24.9 28.5 10.6 11.0 1 0.1 0 0.1 23.8
0.9 1.6 1.5 1.5 0.1 0 0 0 2.6
Percent of total Actinobacteria Proteobacteria Firmicutes Bacteroidetes Verrucomicrobia TM7 Fusobacteria Aquificae Other
29.5 25.3 16.3 11.7 0.3
