Shifts in the bacterial community composition along deep soil profiles in monospecific and mixed stands of Eucalyptus grandis and Acacia mangium

Our knowledge of the rhizosphere bacterial communities in deep soils and the role of Eucalyptus and Acacia on the structure of these communities remains very limited. In this study, we targeted the bacterial community along a depth profile (0 to 800 cm) and compared community structure in monospecific or mixed plantations of Acacia mangium and Eucalyptus grandis. We applied quantitative PCR (qPCR) and sequence the V6 region of the 16S rRNA gene to characterize composition of bacterial communities. We identified a decrease in bacterial abundance with soil depth, and differences in community patterns between monospecific and mixed cultivations. Sequence analysis indicated a prevalent effect of soil depth on bacterial communities in the mixed plant cultivation system, and a remarkable differentiation of bacterial communities in areas solely cultivated with Eucalyptus. The groups most influenced by soil depth were Proteobacteria and Acidobacteria (more frequent in samples between 0 and 300 cm). The predominant bacterial groups differentially displayed in the monospecific stands of Eucalyptus were Firmicutes and Proteobacteria. Our results suggest that the addition of an N2-fixing tree in a monospecific cultivation system modulates bacterial community composition even at a great depth. We conclude that co-cultivation systems may represent a key strategy to improve soil resources and to establish more sustainable cultivation of Eucalyptus in Brazil.


Introduction
Eucalyptus is one of the most common genera in commercial forest plantations worldwide [1]. In Brazil there are 5.10 million hectares of planted Eucalyptus, and accordingly the species plays an important role in the country's economy [2]. Most of the Eucalyptus plantations in Brazil are monocultures [2], which may cause several problems in soils such as an imbalanced PLOS ONE | https://doi.org/10.1371/journal.pone.0180371 July 7, 2017 1 / 15 a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 (treatment E(A+E)). Mixed stands had a 1:1 ratio of A. mangium and E. grandis. Spacing between trees was set at 3 x 3 m and plot areas at 30 x 30 m. Within each plot, we focused on areas of 18 x 18 m, in order to avoid bordering effects. We opened trenches and discarded edges in each of those plots [6]. Trenches measured 0.6 x 1.65 m and were 8m deep. We discarded approximately 25 cm of the external soil border from each trench to avoid cross-contamination between soil layers. Samples were collected in October 2013, when the trees were 4 years old and approximately 12 m high. Five soil layers (0-100, 100-300, 300-500, 500-700, and 700-800 cm) were sampled from three trenches, one from each plot. From each soil layer, we collected and homogenized four subsamples to create composite samples [26]. The resulting 60 samples thus represented four treatments [100A, 100E, 50A-50E: A(A+E) and E(A+E)], three replicates, and five soil depths. From each sample, a 400-g soil portion (after sieving and air drying) was stored in plastic bags, kept at 4˚C, and subsequently subjected to soil chemical and physical analyses, following the methodology proposed by Raij et al. [27] and by the Brazilian Agricultural Research Corporation [28]. Samples were placed in centrifuge tubes in the field and then kept at -80˚C for DNA extraction and molecular analysis at the "Luiz de Queiroz" College of Agriculture.

Quantitative real-time PCR (qPCR)
The abundance of copies of the 16S rRNA gene (g soil -1 ) was determined using quantitative PCR (qPCR). The qPCR mixture (25 μL) contained 0.3 μL of each primer (0.4 mM) 341F (5'-CCTACGGGAGGCAGCAG -3') and 534R (5'-ATTACCGCGGCTGCTTGG -3'), 12.5 μL of the fluorescent marker SYBR1 Green PCR Master Mix 2x (Applied Biosystems1) and 1 μL of DNA template (approx. 50 μg mL -1 ), generating fragments of 193 base pairs [29]. Each reaction was duplicated, and we also included positive and negative control samples (free of DNA) to monitor for potential contamination. Samples were loaded into the StepOne™ Real-Time PCR System (Applied Biosystems 1 ), and reactions were subjected to denaturation at 95˚C for 3 min; 35 cycles of 94˚C for 30 s, 55˚C for 30 s, and 72˚C for 30 s; and a final extension of 72˚C for 45 s [30]. Amplification specificity was confirmed by melting curve analyses obtained from serial dilutions (10 −2 to 10 −8 gene copies μL -1 ). Consistency in Ct values and consequently in quantification values validated our approach. Amplification reaction efficiency was 0.98, with R 2 values higher than 0.98 for all calibration curves.

Sequencing of the bacterial gene 16S rRNA and data analysis
To analyse the structure and composition of the bacterial community, the V6 region of the 16S rRNA gene was amplified with primers A-967F and B-1046R [31]. Primers were amended with adapters for the Ion Torrent Personal Genome Machine System sequencing (PGM) (Life Technologies, USA). Primer A-967F received an additional 5-bp tag to identify sequences from each sample. PCR reactions were performed following an initial denaturation step at 94˚C for 3 min; 30 cycles at 94˚C for 30 s, 57˚C for 45 s, and 72˚C for 1 min; and a final extension of 72˚C for 2 min [31]. PCR products were purified with the ChargeSwitch1 PCR Clean-Up kit (Life Technologies) and subsequently sequenced by the Ion Torrent Personal Genome Machine System (PGM) available at the Microbiology Laboratory of the National Centre for Environmental Research, Assessment and Impact Evaluation-EMBRAPA (Jaguariúna, São Paulo, Brazil).
Sequence analysis was conducted using QIIME (Quantitative Insights into Microbial Ecology) software [32]. Sequences were separated by sample according to their 5-bp tags. Adapter and primer sequences were removed, and low-quality sequences (Q<20) were discarded [33]. Obtained data were rarefied at 14,404 sequences per sample to avoid bias caused by the different numbers of individuals sampled from each soil core. The resulting sequences were grouped into operational taxonomic units (OTUs) based on 97% similarity using the UCLUST method [34]. Representative sequences of each OTU were subjected to taxonomic analysis through the PYNAST method against the Greengenes database [32]. Sequence datasets were uploaded to the MG-RAST (Metagenomics Analysis Server) databank (accession number 265089).

Statistical analyses
Data were examined for homogeneity of variances and normal distribution. Furthermore, the dataset was subjected to ANOVA and mean comparisons by Tukey's test (p<0.05). Bacterial community similarity was represented by Principal Coordinate Analysis (PCoA), using the Bray-Curtis dissimilarity metric distance after Hellinger's standardization implemented with the software Primer-E 6.0 [35,36]. To test whether bacterial composition differed between pure Eucalyptus and co-cultivation with Acacia, we applied the PERMANOVA statistical test. [37,38]. The bacterial phyla and classes most contributing to differences among treatments were determined by the SIMPER test (Similarity percentages species contributions) using the Chord algorithm from the PAST software [38].

Soil properties
In general, soil pH values slightly increased with depth but remained at a value of 5.5 in most of the assessed cultivations and layers. When comparing the first soil layer (0-100 cm) with all the other layers, total N was significantly higher in treatments E(A+E) (1649.7 mg kg -1 ) and A (A E) (1641.9 mg kg -1 ). Treatment 100E presented the lowest total N content (1334.7 mg kg -1 ) (S1 Table). The mean values of available P (in mg kg -1 ), K + , Ca 2+ , Mg 2+ , and Al 3+ (in mmol c kg -1 ) in the 0-100 cm layer were respectively 2.3, 0.4, 2.0, 2.1, and 7.5, showing no significant difference among the treatments (S1 Table). Most of the abovementioned nutrients were below the detection limit in the layers below 100 cm, making it impossible to determine the exact content in the sample. Even in surface layers (0-100 cm) the soil showed a low percentage of the sum of bases and consequently limited values of base saturation (%V) below 50%. Organic matter (OM) values decreased with depth from 12.25 to less than 4.0 g kg -1 (average values of the treatments) from layers 0-100 to 700-800 cm. However, treatment 100E showed the highest OM values in all analysed soil layers (S1 Table). In particle size analysis, soil samples from all layers were classified as medium-sandy except for samples from the A. mangium cultivation classified as sandy between 100 and 300 cm (S1 Table).

Abundance of the 16S rRNA gene along soil depth and among treatments
There was a significant reduction in the abundance of the bacterial 16S rRNA gene with increasing soil depth, with log copies (g soil -1 ) ranging from 10.1 to 10.4 at 0-100 cm and 7.37 to 6.38 at 700-800 cm (p<0.05). We found no significant changes in the abundance of the 16S rRNA gene across treatments in samples from the surface layers (0-100 cm). Nevertheless, we found higher amounts of the target gene from 0 to 300 cm for treatments including Acacia, either in monospecific cultivation (100A, log copies g soil -1 = 8.73) or in mixed cultivation (A (A+E), log copies g soil -1 = 8.08) (p<0.05). Considering the layers of soil below 300 cm and down to 800 cm, the monocultures (100E and 100A) had a higher abundance of the 16S rRNA gene than other treatments (p<0.05). However, treatment 100E stood out from the others, showing the highest mean of log copies of the 16S rRNA genes (g soil -1 ): 100E = 7.58, whereas 100A = 7.02, E(A+E) = 6.20, and A(A+E) = 6.31 (Fig 1).

Bacterial community structures and composition
Structure and composition of the general bacterial community also exhibited clear and significant differences across soil layers and treatments (Fig 2A, S1 Fig).
A principal coordinate analysis revealed a first separation (PCO 1 = 63.7%) corresponding to the influence of soil depth, separating shallow soil samples (0-100 cm) from the others (below 100 cm). Accordingly, the PERMANOVA test showed significant separation (R = 3.22; p = 0.0003) (Fig 2A). The major bacterial groups responding to differences in soil depth were the phyla Acidobacteria and Verrucomicrobia prevalent in the shallow layers of 0-100 cm (S1 Fig, S3 Table) (p<0.05). They were followed by an increase in the frequency of phyla Proteobacteria, Firmicutes, and Bacteroidetes at a greater depth. The relative abundance of Proteobacteria, for instance, was twice as high in the 500-700 cm and 700-800 cm layers as in the 0-100 cm layer. Phylum Firmicutes showed almost four times higher abundance in deep layers than in the surface layer. The Acidobacteria community was significantly less prevalent at greater depths, at 38% in the 0-100 cm layer and 2.9% in the deepest layer (700-800 cm).
The secondary separation (PCO 2 = 16.1%) shows a sample cluster originating from treatments 100A, A(A+E), and E(A+E) that is distinct from samples 100E based on the PERMA-NOVA results (R = 1.97; p = 0.0001) (Fig 2A). We found that the surface (0-100 and 100-300 cm) samples of treatment 100E had a clearly different bacterial community pattern compared with those from areas 100A, A(A+E), and E(A+E) (Fig 2A-2C). Moreover, from top to bottom communities became more dispersed and less separated from each other. Thus, similar patterns were seen between areas at depths below 300 cm (Fig 2D-2F). Furthermore, the A(A+E) and E(A+E) samples shared many bacterial groups with the A. mangium monoculture (Fig  2A-2D).
These major groups occurred in distinct ordinations across treatments. Proteobacteria was the most abundant phylum in samples from treatments A(A+E) and E(A+E) or in the monoculture of Acacia mangium (100A), while in the Eucalypt monoculture (100E) the prevalent phylum was Firmicutes (S1 Fig, S2 Table) (p<0.05). The SIMPER test discriminated the main groups that contributed to the shifts in bacterial communities between the 100E treatment and the 100A, A(A+E), and E(A+E) treatments (Table 1).
In this context, phylum Proteobacteria (26.4%) contributed the most to the distinction between the above-mentioned treatments. Within the phylum Proteobacteria, Betaproteobacteria (16.6%) was the most representative class, expressing the highest percentage of dissimilarity among the tested patterns, followed by Alphaproteobacteria (4.3%), Gammaproteobacteria (3.6%), and Deltaproteobacteria (1.7%) (S2 Fig, Table 1, S4 and S5 Tables). A large fraction of the variation between the treatments can be explained by the presence of the phyla Firmicutes (19.6%), Acidobacteria (19.3%), and Bacteroidetes (12.6%) in the 100E treatment, which were represented by the classes Bacilli and Flavobacteria, while other phyla and classes, as well as unclassified sequences, totalled 11.3% of the group variations between the treatments.

Discussion
The bacterial community inhabiting the surface and subsurface of forest soils plays crucial roles in biogeochemical cycles [39,40]. Although the physical-chemical soil attributes did not present significant differences among treatments (with the exception of the total mineral N), bacterial community structure presented a clear differentiation in the abundance and composition of the rRNA gene along soil layers in response to the established treatments. We did not find any published analysis of the abundance of the 16S rRNA gene in soil layers below a 300-cm depth and thus our inferences regarding deeper soil layers (down to 800 cm) rely exclusively on our own insights.
Previous studies in other plant systems have shown clear changes in the microbial community abundance and composition in soil layers of medium depth. Some authors have stated that the low carbon quantity available in deeper horizons is the prevalent modulating factor for a reduced bacterial community in the subsurface layers [16,26]. The decrease in the bacterial community seems to be related to deeper soil layers, which may act as ecological filters, reducing the overall biomass [13]. Thus, the subsurface environment might select for microbes that can survive under reduced energy sources, limiting the development of highly complex bacterial communities; alternatively, the subsurface soil life must have distinct substrate utilization, leading to the niche colonization of a more specialized and slow-growing bacterial community structure [17,19,23,41]. The shift in the bacterial community along soil layers is not exclusively related to the quantity of bacterial gene copies.
mangium and E. grandis with sampling at the Acacia base; 100E: E. grandis in a monospecific system; and E(A+E): mixed plantation of A. mangium and E. grandis with sampling at the Eucalyptus tree base. (A) refers to a total PCoA including all soil layers (0 to 800 cm) and all treatments; (B) refers to a soil layer depth between 0 and 100 cm, whereas (C) evaluates between 100 and 300 cm, (D) between 300 and 500 cm, (E) between500 and 700 cm and (F) from700 to 800 cm. We found a shift in the relative abundance of the phyla Acidobacteria and Proteobacteria. While Acidobacteria predominated in the surface layer of the treatments, subsurface layers (below 300 cm) were dominated by an exponential increase in the relative abundance of the phylum Proteobacteria, making it the most abundant phylum in the subsurface. In an investigation of the bacterial structure of rice crops, we had previously found increased numbers of the phylum Proteobacteria in soil layers down to 300 cm in depth [20]. This result suggests that the members of the phylum Proteobacteria can survive in a variety of soil types, including those that present higher pH values [42], those that are acidic and those that have a low availability of nutrients [43]. The greatest decrease in abundance of Acidobacteria with increased depth is probably due to the aerophilic and acidophilic characteristics of the microorganisms that compose this phylum. We found a small increase in soil pH values with increased soil depth (S1 Table), which may be a factor discouraging the presence of Acidobacteria [16,44]. The abundance of Firmicutes also showed high peaks in deep soil layers, mainly between 300 and 500 cm, consistently with the findings reported by Li et al. [20] for the subsoil layer. This phylum includes the class Bacilli, which is resistant to pH variations in the soil [45]. The prevalence of the phyla Proteobacteria, Acidobacteria, Actinobacteria, and Bacteroidetes was repeatedly reported, and we previously attributed this dominance to the acidic nature of forest soils [40,46,47]. Actinobacteria, in turn, are K-strategists, slow growing and have a marked ability to grow in oligotrophic environments, in addition to being drought-tolerant [48]. These considerations, and the fact that the phylum Acidobacteria is still unexplored due to its difficulty to be cultivated in laboratory conditions, indicate the importance of more detailed studies of these groups and their community structures and roles in forest ecosystems [40,49,50].
With regard to bacterial community structure, we found great differences even between the surface soil samples (0-100 cm) of the Eucalyptus stand and the three other treatments (Fig  2A). This finding suggests a more potent "rhizospheric effect" caused by Acacia, probably related to its particular supply of carbon exuding from the roots. Another possible reason for this influence of A. mangium is its great acidification of the rhizosphere due to an intense absorption of cations and a constant liberation of H+ ions from the roots, which may contribute to the modification of the bacterial community structures in this environment [4]. Moreover, Acacia is a leguminous N 2 -fixing tree, and the resulting increased N availability may be the most important factor modulating the soil microbiome. Furthermore, N fixation is an important strategy for replenishing N in the soil and for the development of Eucalyptus [11] and considered the most important stimulus for improving soil C and N concentrations, stimulating the development of the nitrifying community, and optimizing intercrop growth [4,12]. N fixation by A. mangium is due to an association between this plant and the N-fixing bacterium Rhizobium of the phylum Proteobacteria [4,8]. However, our results indicate a much greater enrichment by Betaproteobacteria than by Alphaproteobacteria. Thus, we suggest that A. mangium may be another tropical leguminous tree modulated by Betaproteobacteria, alongside Ensifer or Burkholderia among others [51].
Generally, bacterial groups at the soil surface (0-100 cm) were the most compacted, possibly as a result of greater daily and seasonal variation. However, in soil layers below 300 cm we found a smaller allocation of OM, a limiting condition for many bacterial groups [52]. The lower variation of the chemicals attributes in deep soil suggests that there is little microbial genetic variability in deeper horizons [19]. Such conditions can cause genetic and metabolic processes to occur more randomly, leading to the formation of scattered and more dissimilar clusters. Thus, we suggest that the possible dominance effect of A. mangium on the bacterial community is limited to the surface layers of the soil.
Analysing results from the cultivation of Eucalyptus and leguminous tree intercrops, some authors have detected higher productivity in intercropped stands, a feature correlated with nitrogen-fixing ability, biomass production, nutrient cycling, and other benefits [53]. Corroborating these statements, we can add that intercropping plays a crucial role in shifts of the bacterial community structure and composition in comparison to monoculture. The diversification of the stand may help the rhizosphere to acquire new bacterial groups [4]. Changing environmental conditions are linked to changes in profiles of soil microbial communities in response to different management systems [4,53,54] and to changes imposed by different plant species [20]. The diversity of plant species [55][56][57] may cause a variability of C compounds to be released into rhizosphere ecosystems [58], rooting zone depletion [59], litter deposition, diffusion of allelopathic compounds [60,61], and even effects on soil pH, humidity, and nutrient levels [62,63]. Therefore, in light of our data we must assume that the co-cultivation of E. grandis with A. mangium is a successful methodology promoting the diversification and increase in functionality of the rhizosphere microbiome.
Proteobacteria may play a key role in structuring the bacterial community in soil, mainly due to the greater abundance of this phylum in the treatments containing leguminous trees (S2 and S4 Tables). N is one of the most important mineral nutrients, establishing healthy plant growth and rhizosphere bacteria modulation. This nutrient seems to be highly influenced by the dynamics resulting from insertion of a new plant species, resulting in modified bacterial community structure in intercropped plantations [4]. Many authors have demonstrated enhanced N availability in intercropped systems [5,7,8,12], which is key to achieving soil microbial management, acting as a powerful microbial tool and thereby contributing to a better understanding of how to obtain ecological intensification in forests.
Our data suggest that it is possible to "manage" the soil bacterial community by using the adopted tree cultivation system, thereby increasing inherent gains in plantations of interest. This is relevant since microbial groups of interest can be introduced into soils as a result of associations between forest trees. This can give rise to a system of ecological intensification where both economic gain (genus Eucalyptus) and ecological interest (genus Acacia) are found in the same agroecosystem, with improved productivity and sustainability [4]. Our study pioneers the evaluation of bacterial communities in soil layers down to a depth of 800 cm in mixed plantations of Eucalyptus and Acacia, presents initial insights into the rhizosphere bacterial community structure, and will be complemented by our ongoing research.

Conclusions
We verified the influence of the deep-soil and forest-tree composition on soil bacterial communities. In the co-cultivation of E. grandis and A. mangium, the legume exerted a greater influence on the soil bacterial community. Moreover, the soil bacterial community tends to be modulated by soil depth, and as a result it is possible to verify the formation of specific groups in each studied soil layer as a mechanism of plant recruitment. However, in deeper soil horizons, we could not detect typical and distinct rhizosphere communities. The major responsive bacterial groups were Proteobacteria and Acidobacteria (more frequent in samples between 0-100 and 100-300 cm). The major bacterial groups differentially displayed in the monospecific cultivation of Eucalyptus were Firmicutes and Proteobacteria. The use of mixed forests of Eucalyptus with leguminous trees emerges as a promising alternative for sustainability, and should be recognized as a highly important strategy for the conservation of soil biodiversity.
Supporting information S1 Table. Chemical and physical properties of soil at various depths and in the different plantation systems. Table. Average abundance (n = 3) of bacterial phyla across treatments. 100A (A. mangium in a monospecific plantation system); A(A+E) (mixed plantation of A. mangium and E. grandis, with sampling at the Acacia base; 100E (E. grandis in a monospecific plantation system); and E(A+E) (plantation of A. mangium and E. grandis, with sampling at the Eucalyptus base). "Others" represents unclassified sequences. (DOCX) S3 Table. Average abundance (n = 3) of bacterial phyla in the soil layer. 100A (A. mangium in a monospecific plantation system); A(A+E) (mixed plantation of A. mangium and E. grandis, with sampling at the Acacia base; 100E (E. grandis in a monospecific plantation system); and E(A+E) (plantation of A. mangium and E. grandis, with sampling at the Eucalyptus base). "Others" represents unclassified sequences. (DOCX) S4 Table. Average abundance (n = 3) of bacterial classes across treatments. 100A (A. mangium in a monospecific plantation system); A(A+E) (mixed plantation of A. mangium and E. grandis, with sampling at the Acacia base; 100E (E. grandis in a monospecific plantation system); and E(A+E) (plantation of A. mangium and E. grandis, with sampling at the Eucalyptus base). "Others" represents unclassified sequences. (DOCX) S5 Table. Average abundance (n = 3) of bacterial classes in the soil layer. 100A (A. mangium in a monospecific plantation system); A(A+E) (mixed plantation of A. mangium and E. grandis, with sampling at the Acacia base; 100E (E. grandis in a monospecific plantation system); and E(A+E) (plantation of A. mangium and E. grandis, with sampling at the Eucalyptus base). "Others" represents unclassified sequences.