Comparative Analysis of Prokaryotic Communities Associated with Organic and Conventional Farming Systems

One of the most important challenges in agriculture is to determine the effectiveness and environmental impact of certain farming practices. The aim of present study was to determine and compare the taxonomic composition of the microbiomes established in soil following long-term exposure (14 years) to a conventional and organic farming systems (CFS and OFS accordingly). Soil from unclared forest next to the fields was used as a control. The analysis was based on RT-PCR and pyrosequencing of 16S rRNA genes of bacteria and archaea. The number of bacteria was significantly lower in CFS than in OFS and woodland. The highest amount of archaea was detected in woodland, whereas the amounts in CFS and OFS were lower and similar. The most common phyla in the soil microbial communities analyzed were Proteobacteria (57.9%), Acidobacteria (16.1%), Actinobacteria (7.9%), Verrucomicrobia (2.0%), Bacteroidetes (2.7%) and Firmicutes (4.8%). Woodland soil differed from croplands in the taxonomic composition of microbial phyla. Croplands were enriched with Proteobacteria (mainly the genus Pseudomonas), while Acidobacteria were detected almost exclusively in woodland soil. The most pronounced differences between the CFS and OFS microbiomes were found within the genus Pseudomonas, which significantly (p<0,05) increased its number in CFS soil compared to OFS. Other differences in microbiomes of cropping systems concerned minor taxa. A higher relative abundance of bacteria belonging to the families Oxalobacteriaceae, Koribacteriaceae, Nakamurellaceae and genera Ralstonia, Paenibacillus and Pedobacter was found in CFS as compared with OFS. On the other hand, microbiomes of OFS were enriched with proteobacteria of the family Comamonadaceae (genera Hylemonella) and Hyphomicrobiaceae, actinobacteria from the family Micrococcaceae, and bacteria of the genera Geobacter, Methylotenera, Rhizobium (mainly Rhizobium leguminosarum) and Clostridium. Thus, the fields under OFS and CFS did not differ greatly for the composition of the microbiome. These results, which were also confirmed by cluster analysis, indicated that microbial communities in the field soil do not necessarily differ largely between conventional and organic farming systems.

confirmed by cluster analysis, indicated that microbial communities in the field soil do not necessarily differ largely between conventional and organic farming systems.
Ilia Korvigo, Elizaveta Pershina, Evgeny Andronov): http://grant.rscf.ru/forms?rid=2Wv-G10JrV8-3i8XBU2cLb00. Discussion chapter was prepared together with the research scientists participating in Russian Science Foundation (RSF) grant 14-26-00079: "Bio-physical and chemical diagnosis of soil organic matter quality for the development of scientific and theoretical basis of Agrobiotechnology" (Author: Ekaterina Ivanova, http://grant.rscf.ru/forms? rid=1IgCW10JqtDX3i8WD72cLb00.) The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
horizons [21,22,23] and on the different distances from the plant roots [24,25]. As it was shown by van Diepeningen and co-workers the composition of soil microbial community oscillated, depending on the distance remaining from the root. These wavelike patterns were detected both for oligotrophic and copiotrophic bacteria both in OFS and CFS soils but were significantly stronger in conventional croplands [24].
One of the main advantages of employing the pyrosequencing techniques in biodiversity studies is the improvement of knowledge about the impact of agriculture on unculturable microorganisms. The most pronounced effect on soil microbiome revealed in the DOK experiment was the impact of organic fertilizers to the abundance of Acidobacteria in soil. In vitro cultivation methods for this bacterial phylum are lacking for most of its members, except in rare attempts to define the role of these bacteria in the agricultural systems managed with organic fertilizers amendments [26]. Among acidobacteria, genera Cand. Solibacter and Cand. Koribacter have been found exclusively associate with CFS, whereas Chloracidobacteria and the RB25 group have been found associated with OFS in previous studies [10].
The aim of this study was to compare long-term impacts of OFS and CFS on microbial diversity in soil. In the experimental station Karila (Mikkeli, Finland) where OFS and CFS have been carried out in adjacent fields for 14 years. The aim was to compare the taxonomic structure of microbiomes in OFS and CFS and to identify microbes specifically inhabiting these ecosystems.

Soil sampling
Field experiments were carried out under permission of the Natural Resources Institute Finland (formerly MTT AgriFood Research Finland). The field studies did not involve endangered or protected species.
Sampling was done at once from CFS and OFS fields in the experimental station Karila (Mikkeli, Finland) during the season of active plant growth in July 2011. The fields had been cleared from pine-spruce forest in the beginning of 20th century and therefore soil samples were collected from the pine-spruce forest next to the fields included for comparison ( Table 1). The soil type was a coarser fine sand in both sampling fields. According to US soil taxonomy soil was sandy Aquic Haplocryod. Soil samples were taken from the top soil layer (10 cm) using soil drill (Ø 1 cm).
At each sampling site three circles (Ø 1 m) were marked and 10 soil subsamples were taken from inside each circle and combined. Hence, three samples (replicates) were obtained for analysis from each type of soil (woodland, CFS and OFS). The distance between the sampling sites was 45 m in average. All samples were immediately transported to the laboratory and Table 1. Summary of the cultivation history of the fields sampled in the experimental station "Karila" (Mikkeli, Finland). For details, see S1 Table. Year The cultivation history of the fields in Karila is presented in Table 1. Details of the cultivation practices during the last three growing seasons prior to sampling are provided in S1 Table. At the time of sample collection timothy grass (Phleum pratense) and meadow fescue (Festuca pratensis) were grown as a mixture in both sampled fields (OFS and CFS). Besides analysis of the microbiome, the soil samples from OFS, CFS and woodland were subjected to agrochemical analyses ( Table 2).

DNA extraction
DNA was extracted from 0.2 g of soil using PowerSoil DNA Isolation Kit (Mobio Laboratories, Solana Beach, CA, USA), which included a bead-beating step, according to the manufacturer's specifications. Homogenization of the samples was performed using FatsPrep (MP Biomedicals, Santa Ana, CA, USA). The purity and quantity of DNA were tested by electrophoresis in 0.5× TAE buffer on 1% agarose. DNA concentrations were measured at 260 nm using SPECTROStar Nano (BMG LABTECH, Ortenberg, Germany). The average DNA yield was 2-5 μg DNA with the concentration of 10-50 ng/μl.

Quantitative PCR analyses
Relative abundances of bacterial and fungal small subunit rRNA gene copies were analyzed by quantitative PCR (qPCR) (reaction volume 25 μl) using iQ™ SYBR Green Supermix (BIO RAD, Hercules, USA) and 10 ng of sample DNA. For bacteria, the forward primer Eub338 and reverse primer Eub518 were used [27]. The forward primer arc915 and the reverse primer arc1059r were used for archaea [28]. To estimate bacterial and archaeal small-subunit rRNA gene abundances, standard curves were generated using a 10-fold serial dilution of a plasmid containing a full-length copy of 16S rRNA gene belonging either to the Escherichia coli or FG-07 strain of Halobacterium salinarum (courtesy of G. Jurgens, University of Helsinki).
All qPCR reactions were run in triplicate. The reaction was carried out in iCycler (BIO RAD, Hercules, USA) using the following: 94°C for 15 min, followed by 40 cycles of 94°C for 30 s, 50°C for 30 s and 72°C for 30 s. Melting curve analyses were done to verify that the amplified products were of the expected size. Fungal and bacterial gene copy numbers were estimated using a regression equation for each assay relating the cycle threshold (Ct) value to the known number of copies in the standards. Statistical analysis of the qPCR data was carried out using one-way ANOVA in STATISTICA10 Enterprise (www.statsoft.com). Statistical significance was tested by Fischer's least significant difference (LSD) and Bonferroni adjusted p-values.

Bar-coded pyrosequencing of bacterial and archaeal communities
The purified DNA templates were amplified with universal multiplex primers F515 5'-GTGCCAGCMGCCGCGGTAA-3' and R806 5'-GGACTACVSGGGTATCTAAT-3' [29] targeting the variable region V4 of bacterial and archaeal 16S rRNA genes. Each multiplex primer contained the adapter, 4-bp key (TCAG), 10-bp barcode and primer sequences. The expected length of the amplification product was 400 bp. Purification, pooling and pyrosequencing of the amplicons were performed with reagents according to manufacturer's instructions (Roche, Branford, USA). Pyrosequencing was carried out using GS Junior system (Roche).

Bioinformatics of the pyrosequencing-derived dataset
The raw sequences were processed using QIIME ver. 1.8.0 [30]. To reduce sequencing errors, the multiplexed reads were first filtered for quality and grouped according to barcode sequences. Sequences were omitted from the analysis if they were less than 200 bp, had a quality score less than 25, contained uncorrectable barcodes, primers, ambiguous characters or a homopolymer length equal or greater than 8 bp. All non-bacterial ribosomal sequences and chimeras were also removed from the database. In total, 17 311 sequences were obtained with an average of 1923 sequences per library. The dataset was subjected to the normalization procedure resulting in 1100 sequences per sample. The minimum, median and maximum lengths of sequences were 200, 355 and 313 bp, respectively. Similar sequences were clustered into operational taxonomic units (OTUs) with a minimum identity of 97% using de novo and closed reference algorithms. A representative set of sequences was chosen by selecting the most abundant sequence from each OTU. Representative sequences from each OTU were subjected to RDP naïve Bayesian rRNA Classifier [31] with a confidence level of 80% and aligned using PyNast [32] and Greengenes database [33]. Aligned sequences were used to build a distance matrix with a distance threshold of 0.1 and phylogenetic tree necessary for downstream analysis. Sequence data were archived in SRA database with accession SUB473223.
To compare microbial communities the alpha and beta diversity analyses were performed. To estimate alpha diversity, the indices for richness (observed species, ChaoI) and evenness (PD_whole tree, Shannon evenness, Simpson index) were calculated. The t-test was performed to verify the observed differences. For beta diversity the weighted Unifrac metrics [34] was used to calculate the amount of dissimilarity (distance) between the compared bacterial communities. The results were presented in PCoA analysis using QIIME ver. 1.8.0 [30]. All estimates were measured for the normalized data (normalization was carried out up to the smallest number of sequences present in the sample).
The multiple matrix regression based on Mantel permutations [35] implemented in the phytools R package (http://www.phytools.org) was conducted to reveal the relationships between community composition and different agrochemical properties of soil. To reduce factor space dimensionality (by removing redundant variables) we performed multiple pairwise tests for Spearman rank-order correlation. Significant dependency observed between pH and P allowed us to remove the latter from our feature set.
The abundances of OTUs were compared between samples by calculating the median relative change values for all groups of triplicates. A positive median indicated an increase in abundance, whereas a negative median was taken as evidence for decline of abundance. A basic permutation test was used to infer significance, whereas a jackknife-like resampling approach was applied to test the stability of median estimates.

Land use effects on edaphic soil properties
The agrochemical properties of cropland soils managed according to the two different farming systems were rather similar, but differed from the woodland soil despite of the similar soil type ( Table 2). The woodland soil had the highest content of organic matter and C/N index, whereas the lowest C/N value was observed in CFS. Soil pH was lowest in the woodland. As for the main biogenic elements, woodland soil was rich in sulfur and manganese, while the croplands were higher in magnesium, calcium and phosphorus ( Table 2).

Relative quantities of bacteria and archaea estimated by qPCR
The amounts of the bacterial and archaeal biomass estimated by qPCR were expressed as the copy number of rRNA operons per gram of soil and used for comparing the relative abundances of microorganisms in the soil samples. The copy number of ribosomal operons in the genomes of microorganisms varies and is, in average, 4.09 for bacteria and 1.76 for archaea according to the rrnDB database [36]. The experimental data on the average copy numbers of E. coli and H. salinarum rRNA operons in soil samples were used to calculate the abundance of bacterial and archaeal communities, respectively. The average number of bacteria in soil was 8.37·10 8 for CFS, 1.56·10 9 for OFS and 2.19·10 9 for woodland (Fig 1). Archaea were about three folds of magnitude less abundant and their average numbers were 8.15·10 5 for CFS, 2.41·10 6 for OFS, and 1.37·10 7 for woodland (Fig 1). These results showed that the population densities of bacteria and archaea were lowest in CFS and highest in woodland (p<0.05). This tendency was particularly noticeable for archaea, whose numbers in the woodland were 2 orders of magnitude higher than in the croplands. The number of bacteria in OFS was significantly higher than in CFS (p < 0.05), whereas the total counts of archaea did not vary between CFS and OFS (Fig 1).

α-biodiversity of the soil microbial communities
The biodiversity within each individual sample was estimated using richness (number of observed species, Chao1) and evenness (Shannon evenness, Simpson) indices (Table 3).
Woodland samples had the highest percent of coverage (the number of OTUs to chao1 ratio expressed as a percentage) per library (82.7% in average). The coverage values for OFS and CFS samples were 65.1% and 65.8%, respectively. The observed species richness and Simpson index of dominance were not significantly different between the samples (Table 3).
At the phylum level, only minor differences were found between the bacterial communities of CFS and OFS, whereas the differences between woodland soil and croplands were more apparent (Fig 2). Proteobacteria were among the most abundant phyla in croplands, whereas Acidobacteria dominated in the woodland soil (Fig 2). In general, different microbial taxa in woodland soil were more evenly represented, including Firmicutes, Actinobacteria, Nitrospira, Gemmatimonadetes and Chloroflexi.
The microbiomes of croplands and woodland soils differed markedly in the composition of Proteobacteria (Fig 2). Woodland soil was dominated by Alphaproteobacteria and Gammaproteobacteria, while in croplands Betaproteobacteria and Gammaproteobacteria were substituting Alphaproteobacteria. Bacteria from the families Pseudomonadaceae and Enterobacteriaceae accounted for more than 80% of the gammaproteobacteria. The family Pseudomonadaceae was almost exclusively represented by the genus Pseudomonas (Fig 3). The abundance of bacteria of this genus varied between the croplands (16.0% in CFS and 13.2% in OFS). In the woodland soil the number of pseudomonads was only 5.3%. On the other hand, the proteobacterial family Sinobacteriaceae counted for more than 7.2% in the woodland soil, as compared with only 1.7% in OFS and CFS. Similarly, betaproteobacteria of the genus Burkholderia and alphaproteobacteria of the family Bradyrhizobiaceae and the genus Rhodoplanes were substantially more common in the woodland soil than in cropland soils (Fig 4). Among the most notable differences in the microbial taxonomic composition between woodland and croplands was the much higher prevalence of the phylum Acidobacteria in woodland, in particular, bacteria of the   Figs 3 and 4), the order Ellin6513 and Solibacterales (Fig 4).
The main bacterial genera found in soil microbiomes are shown in Fig 3. The most pronounced differences between the CFS and OFS microbiomes were within the genus Pseudomonas, which were significantly (p<0,05) more abundant in CFS soil, as compared with OFS. Other statistically significant (p<0,05) differences in the taxonomic composition between CFS and OFS microbiomes were minor and found among the bacterial genera with frequencies rarely exceeding 1% of all taxa. Compared with OFS, CFS had higher relative abundances of the actinobacteria belonging to the family Nakamurellaceae, acidobacteria of the family  Koribacteraceae, proteobacteria of the groups SC-I-84, Ellin6067, Oxalobacteriaceae (particularly the genus Janthinobacterium) and Ralstonia and bacteria, belonging to the genera Paenibacillus and Pedobacter. Microbial community in OFS was enriched with proteobacteria of the families Comamonadaceae (in particular bacteria of the genera Hylemonella), Sinobacteraceae, Geobacteraceae (Geobacter sp.) and Hyphomicrobiaceae, actinobacteria from the family Micrococcaceae and bacteria from the genera Methylotenera and Clostridium (Fig 4). In OFS samples, the relative densities of the alphaproteobacterial population were found to be increased, which (according to the overall genera composition presented on Fig 3) can be entirely associated with the genus Rhizobium (mainly Rhizobium leguminosarum).

β-diversity analysis
The Unifrac distance matrices were tested to determine if land use had a significant effect on the bacterial and archaeal communities. According to the regression analysis (Table 4), unweighted unifrac distances could be quite precisely predicted based on the data on Ca, Mg, pH and total N in the soil (model R squared~0.95, p-level < 0.0005).
The community composition analysis as well as the regression results were summarized in PCoA analysis. The croplands formed a separate group clearly separated from the woodland soil (Fig 5). CFS and OFS microbiomes were closer to each other, but there was a trend suggesting a closer relationship of OFS than CFS with woodland soils in the composition of microbiome, as supported by the high percent of explained variation (76,51%) on the corresponding axis.

Discussion
The highest C/N value and Shannon evenness parameter were observed in the woodland soil as compared with cropland soils, suggesting a more stable microbial community in the woodland than cropland soil. Cropland microbiomes seemed less stable and represented transitional type of the microbial communities, including a few dominant groups [16]. This conclusion is supported by the parameters of α-diversity, which were generally higher in the woodland. The overall community composition, e.g., the presence of physiologically diverse microbiomes in woodland differed from croplands containing mainly copiotrophic bacteria. Thus, regardless of the OFS or CFS practiced, microbiomes in cropland systems were less stable and likely dependent on the external flow of organic matter and macronutrients. We didn't find any noticeable differences in α-diversity parameters between OFS and CFS samples, despite of the different farming practices. This result contradicts with the data obtained in a few previous studies reporting on increased richness and decreased evenness parameters of soil microbiomes in OFS [10,11,12].
The results of quantitative PCR revealed a higher total bacterial and archaeal abundance in woodland and OFS soil compared to relatively low amount of microorganisms in CFS soil. These findings may correlate with the changes in the pore-space structure of the soil. Cropland soils usually have big pores with high connectivity, favoring the growth of fungi, whereas woodlands tend to have small isolated pores, creating favorable conditions for development of bacteria [37]. The increased number of bacteria in OFS soil may be also be taken as indication of the inputs of organic fertilizers, which carry not only the various types of organic compounds, but also the indigenous bacteria of manure that may remain in soil for a certain period of time.
Comparison of woodland and cropland soil microbiomes revealed significant differences in the composition of soil bacterial taxa at the phylum level. Woodland microbiome was dominated by acidobacteria including bacteria belonging to the groups Koribacteraceae, Ellin6513 and Solibacteraceae, considered as oligotrophs in some reports [38]. Increasing numbers of acidobacteria in woodland are likely associated with low pH, as proposed in recent studies [16,19]. Other groups of oligotrophic bacteria were also abundant in woodlands including hemo-and phototrophic bacteria from the groups Rhodoplanes, Nitrospira, Rhodospirillaceae and Bradyrhizobiaceae [2]. They are known to fix atmospheric nitrogen, which is an advantage in woodland soil due to the smaller content of available forms of nitrogen, as compared with cropland soils [39]. The presence of oligotrophic bacteria, especially those contributing to autotrophic groups, may indicate high levels of functional diversity leading to diversification of ecological niches. Among other oligotrophic soil bacteria the significant increase in the amount of the genus Burkholderia was detected. The representatives of this genus are known to degrade recalcitrant organic matter in soil.
In contrast, the cropland soils were dominated by copiotrophic bacteria from the phyla Proteobacteria, including the orders Pseudomonadales and Enterobacteriales. These bacteria are typical for agroecosystems due to regular mixing of soil and the introduction of nutrients, as well as specific nutritional substrates, such as manure, silage and xenobiotics, whose biodegradability is well described for the Pseudomonas group [40]. Predominance of the bacteria belonging to Pseudomonas in croplands has been described in several papers using classical microbiological approaches, as well as the new-generation sequencing methods [16,41]. Pseudomonads might be used as the primary bioindicators of the ecological status of the soil due to several reasons: a) they are clearly responding to the changes in the edaphic characteristics of the soil, b) they have high population sizes, and c) they can be detected using in vitro cultivation methods and modern molecular methods [16,41,42].
Among the most pronounced differences in the composition of soil microbiomes of croplands soils was the significant increase in the relative amounts of pseudomonads in CFS soils as compared with OFS. Thus, it seems possible to use pseudomonad's diversity data to distinguish not only woodland and cropland soils but CFS and OFS soils as well. These results are supported by other studies, where it was shown that fluorescent pseudomonads are suppressed in OFS soils [43,44].
Other fluctuations in the certain bacterial groups inhabiting two types of croplands concerned the bacterial genera, whose relative amounts rarely exceeded 1% of the total bacterial counts. Under the conditions of CFS the proportion of bacteria capable of biodegradation of various xenobiotics increased, including the genera Ralstonia, Pseudomonas, Paenibacillus and Pedobacter. Another significant finding was the statistically supported increase in the relative abundance of the bacteria from the family Koribacteriaceae in CFS compared to OFS, which agrees with results of Hartmann et al. [10]. Thus, we can speculate on the valuable ecological properties of these bacteria in cropland system supplied with mineral fertilizers.
OFS microbiome was dominated by several groups of bacteria, whose appearance in the soil may be caused by the imputes of the spectrum of organic compounds. Particularly this may be concerned as a reason for the increase rates of the bacteria, belonging to the genera Methylotenera and Clostridium. Methylotrophic bacteria from the genera Methylotenera are capable for the utilizing of the methane and its derivatives that accumulate in soil as a result of decomposition of the introduced organic matter [45,46]. Members of the genus Clostridium are known to be one of the main anaerobic decomposers of the soil organic matter [47]. The increase in the proportion of clostridia in OFS indicates appearance of anaerobic zones in the soil, which generally form within the soil aggregates [48]. The increase in the proportion of nitrogen-fixing bacteria (Rhizobium sp.) and some bacterial genera of the PGPR-group (e.g. the bacteria from the family Hyphomicrobiaceae) in OFS soil agrees with previous studies [49,50] and can be explained by the increasing demand for growth factors and mineral elements exhibited by the plant in the absence of mineral fertilizers [51].
The soil microbial communities in cropland soils differed greatly from the microbiome in the soil of the woodland, which was once (ca. 100 years earlier) declared from forest to arable fields. In contrast, the differences between the CFS and OFS microbiomes were much less pronounced, affecting the community composition mainly at the genus level. These conclusions are largely supported by the PCoA analysis, in which OFS and CFS microbiomes are clearly separated from woodland group, whereas the differences between the two cropland systems are minor. Regression analysis showed that the differences observed in microbial composition could be explained by soil chemical properties, among which the soil pH seemed to be the most significant parameter at the phylum level. The amount of biological macroelements, such as Ca and Mg, could be treated in turn as predictors of genus-level variance in the composition of cropland microbiomes. Lower amounts of these mineral nutrients in OFS soil may be one of the reason for the appearance of plant growth promoting bacteria in the corresponding microbiomes, which also agrees with previously reported data [10,13,16].
It is worth mentioning that the data presented in this study are likely dependent on the applied methods of soil sampling, DNA sequencing and bioinformatics analysis. It is well known that soil is highly heterogeneous and sampling may not capture its entire variability. Hence, the results may not be able to explain all the spatial variance in microbiome of the studied habitat. Furthermore, in this study we didn't investigate the seasonal dynamics of microbial communities, which is likely to exist. Other limitations of the molecular studies of soil microbiomes, such as DNA-extraction and PCR biases [52] also must be considered when interpreting the results of the study.
Keeping the aforementioned limitations in mind, we may summarize the main conclusions as follows. Comparison of the soil microbiomes in two cropping systems (CFS and OFS) with the microbiome in soil of woodland revealed major differences in the agrochemical parameters of soil and the taxonomic composition of microbiomes. Higher C/N ratios and α-diversity parameters as well as the presence of many oligotrophic bacteria in woodland indicate active participation of microorganisms in the deposition of the organic matter increasing its availability for further biodegradation. Croplands represent systems depending partially on the influx of organic or mineral fertilizers, which seems to lead to the predominance of bacteria capable of biodegrading xenobiotics in CFS soils and degrading various organic compounds in OFS soils. Additionally due to the relatively low concentrations of the available mineral macronutrients, OFS soils seem to be dominated by plant growth promoting bacteria. Generally, the use of OFS or CFS had only minor influence on microbial biodiversity in the fields of this study, affecting primarily the genus-level composition of microbiomes. The results provide valuable new information, indicating that carefully managed conventional and organic farming systems may maintain similarly diverse microbial communities, which creates prospects for further research in this area.
Supporting Information S1