The transformation of soil to agricultural use causes significant changes in its chemical, physical, and biological features, including changes in the microbial community composition (Kibblewhite et al., 2008). A number of studies have examined the impact of agriculture on microbial community structure. The effect of inorganic and organic fertilizers, tillage, and different crop rotations was analyzed. It was found that nitrogen (N) deposition increases the abundance of certain microbial phyla, such as Actinobacteria, Proteobacteria, Bacteroidetes, and Firmicutes, but it also decreases the abundance of other bacterial phyla, such as Acidobacteria and Verrucomicrobia (Ramirez et al., 2010, 2012; Fierer et al., 2012). An increase in microbial biomass was detected after farmyard manure addition (Kandeler et al., 1999), while previous studies have shown that reducing tillage increased microbial biomass (Buckley and Schmidt, 2001; Plassart et al., 2008). Other studies have focused on the changes of specific microbial taxa involved in nutrient cycling in soil, such as diazotrophs (Meng et al., 2012), nitrifiers (Chu et al., 2008), and denitrifiers (Clark et al., 2012) under specific agricultural management. Therefore, all studies of the agricultural impact on the microbial community are limited either to the phylum level, which represents a large and diverse set of functional microbial groups, or to the specific physiological group. We hypothesize that there are common microbial taxa strongly associated with agricultural land use, and these taxa can be used as indicators to monitor the level of the land disturbance during agricultural management.

One of the abundant groups found in many soils under agricultural use is ammonia-oxidizing archaea (AOA) (He et al., 2007; Shen et al., 2008; Gubry-Rangin et al., 2010; Taketani and Tsai, 2010; Wessén et al., 2010; Pratscher et al., 2011; Xia et al., 2011). In soils such as Amazonian Anthrosol, agricultural plots have a higher archaeal ammonia monooxygenase gene (amoA) copy number than in adjacent soils (Taketani and Tsai, 2010). Wessén et al. (2011) showed that high numbers of archaea are positively correlated with nitrate leaching from both integrated and organic farming systems. Two agricultural soils in Germany, OrthicLuvisol and GleyicCambisol, harbored high abundances of archaeal, but not bacterial amoA genes (Schauss et al., 2009). AOA are important participants in soil nitrification because they are able to perform the first step of nitrification through the ammonia monooxygenase enzyme (AMO) (Könneke et al., 2005; Treusch et al., 2005). The most represented soil AOA was found within was Candidatus Nitrososphaera (Taketani and Tsai, 2010; Xia et al., 2011; Pester et al., 2012).

Agricultural management intensifies all processes related to the N cycle including an increase in nitrification rates (Lu et al., 2011). Increased nitrification can result in nitrate leaching into the surface and ground water, as well as the emission of nitrous oxide to the atmosphere (Kim et al., 2012). Recently, the contribution of AOA to nitrification was shown by assimilation of ¹³C-CO₂ by AOA during soil nitrification (Zhang et al., 2010, 2012; Xia et al., 2011). In addition a high correlation between AOA and nitrification activity was observed (Offre et al., 2009) as well as high abundance of archaeal amoA transcripts in soil (Treusch et al., 2005; Nicol et al., 2008). Hence, AOA may be responsible for all these consequences of intensified nitrification. Knowledge of the factors that may drive the abundance of this group is necessary to prevent agricultural land management from negatively impacting the environment. Previous studies have investigated temperature, soil type, and elevation as drivers of archaeal abundance in soils (Zhang et al., 2009; Taketani and Tsai, 2010). Other studies have examined pH, fertilization, carbon to nitrogen ratio, and tillage as anthropogenic drivers that are a result of land management (Kandeler et al., 1999; Fierer and Jackson, 2006; Enwall et al., 2007; Nicol et al., 2008; Taketani and Tsai, 2010; Bates et al., 2011; Hermansson and Lindgren, 2001). Gubry-Rangin et al. (2011) found different phylogenetic lineages of AOA that were acidophilic, acido-neutrophilic, and alkalinophilic, and these were positively correlated with soil pH levels. Bru et al. (2011) found a positive correlation between AOA abundance and pH. Pereira e Silva et al. (2012) have recently shown that AOA abundance was positively correlated with pH in a temporal study based on eight agricultural soils. However, Nicol et al. (2008) reported results where archaeal amoA gene and transcript abundance decreased with higher pH in acidic soils. The factors responsible for the changes in AOA abundance vary depending on the ecosystem, the geography, and soil type. It remains unclear what factors are mainly responsible for triggering the abundance of archaea in soil.

Three long-term experiments that study crop production, nutrient cycling, and environmental impact of agriculture were included in this work. The first experiment, the Broadbalk Rothamsted Experiment, was designed more than 170 years ago to test the effects of various combinations of inorganic fertilizers and farmyard manure on the yield of the wheat. The second study site, the Kellogg Biological Station long-term ecological research project, was initiated to examine the basic ecological relationships in field-crop ecosystems typical of the Midwestern U. S. Early in the last century, 280,000 ha of primary rich with organic matter histosols in South Florida were drained to create the Everglades Agricultural Area, which was the third experiment for our study. About 25% of US winter vegetables and sugarcane are cultured in the EAA. Despite the many years of study at the Broadbalk Rothamsted Experiment, the Everglades Agricultural Area, and the Kellogg Biological Station, there are few studies describing how agricultural management practices affect microbial taxa at these long-term sites (Castro et al., 2005; Ogilvie et al., 2008; Ramirez et al., 2010; Delmont et al., 2011; Clark et al., 2012; Fierer et al., 2012). None of these show the detailed differences in taxonomic groups that occur with agricultural land use and none of them show this across multiple sites. Also, none of these studies provide evidence for biomarkers of land use or show how microbial taxa change with land use succession. The goals of this study were to examine changes in archaeal and bacterial community composition in response to land-use with a particular emphasis on ammonia oxidizers. Microbial communities in the soils of three long-term agricultural sites were examined using 16S rRNA barcoded Illumina sequencing. Sites were also chosen so that the effect of succession on microbial taxa could be examined. Soil community composition was also compared to several soil properties to identify the drivers of microbial diversity and abundance.

No previous work shows the dramatic effects of land use management on soil microbial diversity on adjacent plots using long-term field experiments. In addition, the observation that these effects are reversible, as seen through the study of the Kellogg successional plots is also novel. In addition, no previous work identifies specific markers of land use change as has been done here with the discovery of Ca. Nitrososphaera as an abundant organism in agricultural soils and Bradyrhizobium as an abundant organism in non-agricultural soils and that these land use markers are reproducible across a suite of diverse sites and highly statistically significant. And no previous work has shown that ammonia is the driver of archaeal relative abundance in soils. These major findings are discussed below in the context of previous work.

This work was inspired by the results of Roesch et al. (2007) who found a high relative abundance of archaea in three agricultural soils (4–12%) and a vanishingly low number in a boreal forest soil (0.01%). This result led to several questions. Why were the archaea so low in the boreal compared to the others sites? Was it the colder climate in the boreal forest site from northern Ontario, Canada compared to the other sites? Was it the likely lower nutrient status of the forest site compared to the agricultural sites? Just prior to the publication of Roesch et al. (2007), the discovery of ammonia-oxidizing archaea in soil was made (Leininger et al., 2006). It was reasonable to assume that nitrogen fertilization of agricultural sites might contribute to the higher relative abundance of archaea in agricultural soils. To test this notion, three long-term experimental sites were chosen that met the following three important criteria. First, the sites had to be long-term sites with the availability of nutrient and other environmental data. Second, the sites had to have adjacent plots that were cultivated and uncultivated. Third, collectively the sites had to differ significantly in mean annual temperature to be able to test a climate effect on soil archaea. Using these criteria, three sites were chosen: the Broadbalk experiment at Rothamsted Research in the UK, the Kellogg Biological Station in Michigan USA, and the Everglades Agricultural Area in South Florida, USA. The Broadbalk and Kellogg sites had the added advantage of having experimental plots with varying amounts of N fertilizer applied annually.

In addition to the archaea, bacterial taxa were also examined for their changes with land use. Only about 24% of all reads could be classified to known genera. Of the more than 700 known genera found at each site, an average of 20 genera at each site were of reasonably high abundance that also differed in relative abundance between agricultural and non-agricultural plots. Of those, only two differed in relative abundance by land use at all three sites. One of these, Bradyrhizobium, is involved in ammonia production and is best known for its role as a nitrogen-fixing symbiont on legume roots. The other, Ca. Nitrososphaera, is best known for its role in ammonia oxidation. The two genera, Bradyrhizobium and Ca. Nitrososphaera, are negatively and positively correlated with ammonia levels, respectively. This also makes sense biochemically. Addition of fixed N inhibits nitrogenase and nitrogen fixation in the lab and field (Sekhon et al., 1987; Peoples et al., 1995; Halbleib and Ludden, 2000). In addition, as the product of nitrogenase, ammonia is a product inhibitor of nitrogenase and also blocks transcription of the nifregulon (Halbleib and Ludden, 2000). As a result, higher levels of N would reduce the role for nitrogen-fixing organisms in non-legume agriculture at each of the three sites studied here. The observation that the proportion of Ca. Nitrososphaera to Bradyrhizobium increases slightly, but significantly statistically, over time at Broadbalk supports the notion that Ca. Nitrososphaera's relative numbers compared to Bradyrhizobium increase with increasing ammonia concentrations in soil. This may occur because Bradyrhizobium is providing more ammonia to soil, which encourages a population increase among the ammonia oxidizing archaea in a non-agricultural soil. In an agricultural soil where N fertilizer is applied, the number of free-living bradyrhiozbia declines since fixed N inhibits N₂ fixation. So the numbers of ammonia oxidizing archaea increase in an agricultural soil as a result of added N and a higher soil pH. Both factors result in the increased availability of ammonia.

The above hypothesis depends on nitrogen fixation by free-living Bradyrhizobium in soil since the samples collected in this work were from soil and did not contain roots. There are several examples of free-living nitrogen fixation by Bradyrhizobium (formerly referred to as slow-growing Rhizobium). This was first shown by Pagan et al. (1975) who demonstrated free-living nitrogen fixation by strain 32H1 as well as other strains. This work was followed shortly by experiments that optimized the conditions for nitrogenase activity by free-living bradyrhizobia including O₂ requirements as well as the cell morphology changes that occur under nitrogen fixation conditions (Gibson et al., 1976; Keister and Evans, 1976; van Brussel et al., 1979). These observations were soon expanded to still more strains of Bradyrhizobium (Subba-Rao, 1977; Skotnicki et al., 1979). Nevertheless, attempts to obtain free-living nitrogen fixation in many bradyrhizobia have failed (Pagan et al., 1975; Skotnicki et al., 1979). However, this may be caused by the lack of a specific nutrient in medium. For example, it was recently shown that symbiotic rhizobia lack nifV, a gene essential for the production of homocitrate, a necessary component of the FeMo cofactor present in dinitrogenase and that the host plant provides homocitrate to the nodule bacteria to compensate for the lack of nifV (Hakoyama et al., 2009).

Nevertheless, free-living bradyrhizobia have been shown to fix ¹⁵N₂ in soil using stable isotope probing. ¹⁵N label was found in bradyrhizobial 16S rRNA sequences after feeding ¹⁵N₂ to soil mesocosms (Buckley et al., 2007). In addition the genome of a free-living Bradyrhizobium strains isolated from a rice paddy was recently sequenced and found to contain the same complement of nitrogen fixation genes found in the genome of a nitrogen-fixing symbiont of Bradyrhizobium. However, this paddy soil Bradyrhizobium strain did not nodulate any legume tested, lacked a symbiosis island of genes often found in N₂-fixing legume symbionts, and did not possess any of the nodulation genes. All of these results taken together with the results presented here are expected to encourage an examination of free-living bradyrhizobia in uncultivated soils to determine their ability to provide fixed N to unmanaged ecosystems.

There is also a sound biological basis for Ca. Nitrososphaera to be relatively more abundant in agricultural soils than non-agricultural soils. As ammonia, not ammonium, is the substrate for ammonia monooxygenase (Suzuki et al., 1974; Arp et al., 2002), it is not surprising that ammonia levels, not ammonia plus amonium levels, correlate well with the relative abundance of Ca. Nitrososphaera, particularly at higher pH levels. However, discovering this required that ammonia levels be calculated from the total ammonia plus ammonium levels and soil pH as current methods of measuring ammonia/ammonium levels in soil do not distinguish between ionized ammonium and non-ionized ammonia.

The results here show that agricultural management causes significant changes in soil, which leads to an increase in AOA abundance. Ca. Nitrososphaera, the most abundant soil AOA, was present in a greater abundance at all three sites in response to agriculture. Of all factors examined, pH mediated NH₃ accumulation was the primary driver of the AOA community in soil.

In addition, this work shows the effect of agriculture on the relative abundance of other organisms involved in the nitrogen cycle. At each site, the relative abundance of Bradyrhizobium, a nitrogen-fixing symbiont, was strongly negatively correlated with agricultural land use, pH, and NH₃ levels. The reciprocal responses of Bradyrhizobium and Ca. Nitrososphaera appear to be excellent biological markers for land use. For further validation of these microorganisms as biological markers, these results should encourage the testing of these genera as markers for land use at other sites.

All sequences have been deposited in the GenBank database with Accession No. PRJNA191521, RJNA191098, and PRJNA191523.