Environmental Variables Shaping the Ecological Niche of Thaumarchaeota in Soil: Direct and Indirect Causal Effects

To find environmental variables (EVs) shaping the ecological niche of the archaeal phylum Thaumarchaeota in terrestrial environments, we determined the abundance of Thaumarchaeota in various soil samples using real-time PCR targeting thaumarchaeotal 16S rRNA gene sequences. We employed our previously developed primer, THAUM-494, which had greater coverage for Thaumarchaeota and lower tolerance to nonthaumarchaeotal taxa than previous Thaumarchaeota-directed primers. The relative abundance estimates (RVs) of Thaumarchaeota (RTHAUM), Archaea (RARCH), and Bacteria (RBACT) were subjected to a series of statistical analyses. Redundancy analysis (RDA) showed a significant (p < 0.05) canonical relationship between RVs and EVs. Negative causal relationships between RTHAUM and nutrient level–related EVs were observed in an RDA biplot. These negative relationships were further confirmed by correlation and regression analyses. Total nitrogen content (TN) appeared to be the EV that affected RTHAUM most strongly, and total carbon content (TC), which reflected the content of organic matter (OM), appeared to be the EV that affected it least. However, in the path analysis, a path model indicated that TN might be a mediator EV that could be controlled directly by the OM. Additionally, another path model implied that water content (WC) might also indirectly affect RTHAUM by controlling ammonium nitrogen (NH4 +-N) level through ammonification. Thus, although most directly affected by NH4 +-N, RTHAUM could be ultimately determined by OM content, suggesting that Thaumarchaeota could prefer low-OM or low-WC conditions, because either of these EVs could subsequently result in low levels of NH4 +-N in soil.


Introduction
The discovery of ammonia-oxidizing archaea (AOA) has changed a century-old paradigm whereby the first step in the nitrification process, chemolithotrophic ammonia oxidation, was thought to be performed exclusively by ammonia-oxidizing bacteria (AOB) belonging to βand γ-Proteobacteria.Archaeal homologs to the bacterial ammonia monooxygenase gene (amoA) were found in environmental genomes, providing initial evidence indicating an archaeal contribution to ammonia oxidation [1,2], and at the same time, a marine AOA strain, Nitrosopumilus maritimus SCM1, was first isolated from a saltwater aquarium by enrichment and filtration techniques [3].AOA were initially classified as mesophilic Crenarchaeota [1,2], but in 2008, the third archaeal phylum, Thaumarchaeota, was proposed, based upon archaeal phylogeny inferred from rRNA and ribosomal protein sequences, to distinguish the mesophilic AOA lineages from hyperthermophilic Crenarchaeota lineages [4].Few years later, this distinction was confirmed by genomic information (e.g., the identification of Thaumarchaeota-specific genes) in Cenarchaeum symbiosum, Nitrosopumilus maritimus, and Nitrososphaera gargensis, representatives of marine and terrestrial AOA lineages [5].To date, ten thaumarchaeotal species (Cenarchaeum symbiosum, Nitrosoarchaeum koreensis, Nitrosoarchaeum limnia, Nitrosocaldus yellowstonii, Nitrosopumilus maritimus, Nitrosopumilus salaria, Nitrososphaera gargensis, Nitrososphaera viennensis, Nitrosotalea devanaterra, and Nitrosotenuis uzonensis) have been isolated from marine and terrestrial environments, and have been well characterized [3,[6][7][8][9][10][11][12][13][14][15][16].However, although extensive effort has been devoted to understanding the biochemical, physiological, and genomic characteristics of Thaumarchaeota, our knowledge of the ecology of Thaumarchaeota is relatively limited.Although findings from ecological studies so far conducted upon Thaumarchaeota have been well discussed in conjunction with its physiology in several review articles [17][18][19], the ecological niche of Thaumarchaeota remains largely unknown.It is important to know the lifestyle of Thaumarchaeota in environments to fully understand this phylum, as well as to provide basal information facilitating further exploration of its as yet unrecognized aspects.
Clues required to elucidate the ecological niche of Thaumarchaeota have been obtained mainly from relative abundance estimates of Thaumarchaeota based on the ratio of the copy number of archaeal amoA to that of bacterial amoA (i.e., the amoA ARCH /amoA BACT ratio) under the assumption that all archaeal amoA-carrying prokaryotes are Thaumarchaeota and vice versa [20][21][22][23][24][25][26][27].However, the conclusions of studies based upon the use of this ratio might have two limitations.First, although it is useful to differentiate the niche of AOA from the niche of AOB, the amoA ARCH /amoA BACT ratio cannot provide substantive information on the Thaumarchaeota-specific niche.The amoA ARCH /amoA BACT ratio reflects only the response of AOA to EVs relative to the analogous response of AOB, and not relative to that of all other nonthaumarchaeotal taxa.Second, the PCR primer pairs currently used to target archaeal amoA may have limited coverage.A high level of sequence variations (identities < 85%) in amoA has been reported [28], and specificity problems of amoA-targeting primers have been pointed out [29,30].In addition to these main limitations, the phylum Thaumarchaeota might contain members that are deficient in ammonia oxidation activity [31].Since almost all methods currently used for cultivating the members of Thaumarchaeota use ammonia as a sole energy source, the members thus cultured could include only ammonia-oxidizing strains.Similarly, as yet undiscovered archaeal groups other than Thaumarchaeota might possess archaeal amoA xenologs.
An alternative approach to determine the Thaumarchaeota-specific ecological niche is to use a quantitative method targeting a universally applicable housekeeping gene such as the 16S rRNA gene, the copy number of which can provide a good abundance estimate regardless of the particular metabolic characteristics of the organisms under study.However, when using this 16S rRNA gene-based approach, it could be difficult to obtain information particularly regarding the response of Thaumarchaeota (AOA) to EVs compared to that of only AOB, unless the abundance data for AOB were collectively obtained from different AOB phylogenetic groups (e.g., β-proteobacterial AOB and γ-proteobacterial AOB).Nonetheless, the 16S rRNA gene-based approach provides very important information especially regarding the response of Thaumarchaeota to EVs compared to that of all other prokaryotic groups, facilitating identification of the Thaumarchaeota-specific niche.However, in our previous study [32], 16S rRNA gene-directed PCR primer pairs frequently used for quantifying Thaumarchaeota, 771F-957R and MCGI391F-MCGC554R (identical to primer pair MGI391-Cren537) [33,34], showed unsatisfactory coverage for Thaumarchaeota.These primer pairs captured sequences belonging to only a single thaumarchaeotal subgroup (either marine group I [MG-I] or soil crenarchaeotic group [SCG]), with insufficient coverage (MCGI391F-MCGC554R, 47.5% of MG-I sequences; 771F-957R, 87.3% of SCG sequences).All previous 16S rRNA gene-based ecological studies upon Thaumarchaeota, except for those studies using a next-generation sequencing approach that employed universal primers, implicitly assumed that MG-I and SCG predominated in marine and terrestrial environments, respectively, and included the use of a primer pair specific to one of these thaumarchaeotal subgroups for estimating the abundance of Thaumarchaeota (or AOA).However, unexpected thaumarchaeotal subgroups (e.g., subgroups MG-I in the terrestrial environment and subgroup SCG in the marine environment) might reside in the environment under study, as observed by Tourna et al. [30] and Beman and Francis [35].Moreover, multiple subgroups or even as yet undiscovered subgroups might possibly reside together.In samples obtained from such environments, the abundance of Thaumarchaeota could be underestimated.In addition, the thaumarchaeotal abundance measured by using the primer pair 771F-957R could be considerably overestimated due to this primer pair's high tolerance (the extent to which it binds to nontarget taxa, ~36.4%) [32].
In this study, to determine EVs shaping the ecological niche of Thaumarchaeota, we estimated its abundance in various soil samples by using real-time PCR that targeted 16S rRNA gene sequences.We employed our newly developed PCR primer, THAUM-494, which has greater coverage (92.9%) for Thaumarchaeota and lower tolerance (0.9%) to nonthaumarchaeotal taxa than other primers targeting this phylum, as we have reported previously [32].The abundance estimate of Thaumarchaeota was compared with that of total prokaryotes, and was subjected to a series of statistical analyses.The EVs affecting the relative abundance of Thaumarchaeota in soil were identified, and their direct and indirect causal effects are described in this paper.

Soil sampling and DNA extraction
Twenty-seven soil samples were collected from sites with and without vegetation (grassland, 25.9%; forest, 40.7%; arid soil, 25.9%) in Korea.Soil samples belonged to andisol (11.1%), entisol (51.9%), inceptisol (14.8%), and ultisol (18.5%), and the majority (70.4%) of them were sandy loam.The soil samples were taken below 10 cm from the soil surface, after surface litter (e.g., plant debris) was removed.All sampling sites were located in private land.We obtained landowners' permission to collect soil samples from their property land prior to sampling.
After sieving (sieve size = 2 mm) on site, the soil samples were immediately stored at 4°C for transportation to the laboratory, and were subjected to DNA extraction upon arrival.Community DNAs were directly extracted from the soil samples by using a PowerSoil DNA isolation kit (MoBio Laboratories, Carlsbad, CA, USA) according to the manufacturer's protocol.To avoid subsample bias, DNA obtained from four subsamples per soil sample was pooled to prepare each sample's PCR template.

Physicochemical analysis
Soil temperature was measured at ca. 10 cm depth by using a bimetallic thermometer.Water content (WC) was determined from the percent weight loss after oven-drying at 105°C for 18 h.Soil pH was determined as pH 1:5 after mixing with distilled water (soil-to-water ratio, 1:5 [w/v]).Total carbon (TC) content was measured by a combustion method [36] by using a carbon analyzer (TOC-L, Shimadzu, Kyoto, Japan) equipped with a combustion module (SSM5000A, Shimadzu).Total nitrogen (TN) content was determined by means of Kjeldahl digestion [37].Total phosphorus (TP) and total sulfur (TS) contents were determined by using an inductively coupled plasma-atomic emission spectrometer (ICP-AES 7510, Shimadzu), after extraction of samples using a hydrogen chloride and nitric acid solution [38][39][40][41].Ammonium nitrogen (NH 4 + -N) and nitrate nitrogen (NO 3 − -N) were respectively determined by means of Kjeldahl digestion [36] and ion chromatography (ICS 5000, Thermo Fisher Scientific, Sunnyvale, CA, USA) [42,43].

Real-time PCR
Abundance estimates of Thaumarchaeota were determined by using a real-time PCR (RTi-PCR) assay employing the Thaumarchaeota-specific primer THAUM-494, which we developed previously, and demonstrated to be more inclusive and specific than other primers used for quantifying Thaumarchaeota, in terms of coverage for Thaumarchaeota and tolerance to nonthaumarchaeotal taxa [32].THAUM-494 was paired with an archaeal universal primer 917R [44].The primer pair (THAUM-494-917R) used in this study had a higher coverage (89.3%) for Thaumarchaeota and lower tolerance (0.9%) to nonthaumarchaeotal taxa than previously used primer pairs (coverage, 21.9-33.7%;tolerance, ~36.4%) (S1 Table ); thus our primer pair was expected to provide a considerably better estimate of the abundance of Thaumarchaeota.Primer pairs ARC806-ARC915 [45,46] and Eub338-BAC515 [47,48] were used for quantifying Bacteria and Archaea, respectively (S2 Table lists coverage values of the universal primers used).SYBR Premix Ex Taq reagent (Takara, Shiga, Japan) was used for the RTi-PCR.SYBR Green I and ROX were used as reporter and passive reference dyes, respectively.Reactions were carried out in MicroAmp optical eight-tube strips (Applied Biosystems, Foster City, CA, USA) by using a ABI Prism 7300 sequence detection system (Applied Biosystems).Each reaction contained 2 μl of template DNA, 25 μl of SYBR Premix Ex Taq, 1 μl of ROX dye, 1 μl (20 pmol) of each primer, and sterile water to bring the total reaction volume to 50 μl.Thermal cycling parameters were as follows: initial denaturation (95°C) for 5 min, followed by 35 cycles of denaturation (95°C) for 30 s, primer annealing (primer pairs for Thaumarchaeota, 55°C; Archaea, 60°C; Bacteria, 55°C) for 30 s, and final extension (72°C) for 30 s. Fluorescence signals were measured during the extension step.RTi-PCR amplifications were performed in triplicate.The collected fluorescence signals were analyzed by using ABI Sequence Detection Software version 1.4 (Applied Biosystems).The fluorescence signal intensity of the reporter dye was normalized by using the signal intensity of the passive reference dye to correct for fluctuations in the fluorescence signal due to the changes in the concentration and volume of the reaction mixture.The threshold cycle (C T ) was taken to be the PCR cycle at which a significant increase in the normalized fluorescence signal was first detected.The threshold level was determined automatically by the detection system, using the default settings.
Recombinant plasmid DNAs were used to construct standard curves for determining the copy numbers of 16S rRNA gene sequences of Thaumarchaeota, Archaea, and Bacteria (S1 Fig) .Purified DNAs from environmental clones obtained in our preliminary study, which contained 16S rRNA gene sequences of Thaumarchaeota, Archaea, and Bacteria (GenBank accession numbers KF275705, KF276604, and GQ143752, respectively) were used as template DNAs for constructing the standard curves.RTi-PCR amplifications used in constructing the standard curves were performed in quadruplicate using a range of template DNA concentrations (ca. 10 1 -10 9 copies of 16S rRNA gene sequences per reaction).The slopes of the standard curves ranged from −3.29 to −3.20 (R 2 = 0.994-0.999),indicating PCR efficiency near 100% (PCR efficiency = e −ln(10)/slope − 1).

Statistical Analysis
Based on the copy numbers of 16S rRNA genes of Thaumarchaeota (N THAUM ), Archaea (N ARCH ), and Bacteria (N BACT ) determined in RTi-PCR assays, relative abundance estimates of Thaumarchaeota (R THAUM ), Archaea (R ARCH ), Bacteria (R BACT ) were calculated as R THAUM = N THAUM /N PROK , R ARCH = N ARCH /N PROK , and R BACT = N BACT /N PROK , respectively, where N PROK = N ARCH + N BACT .Multivariate statistical analyses were performed to establish relationships between the relative abundance estimates (RVs) and environmental variables (EVs).Log-transformed data (except pH data) were subjected to RDA [49] using RVs as response variables and EVs as explanatory variables, in which the ordination of RVs is constrained in a way such that the resulting ordination vectors are linear combinations of the EVs [50].For statistical ordination, we preferred RDA employing a linear model rather than canonical correspondence analysis employing a unimodal model [51] because the maximum length of the gradient determined by means of detrended correspondence analysis [52] was less than two standard deviations [53].The statistical significance of the overall RDA results was assessed (α = 0.05) by a permutation test (of 499 iterations) on the null hypothesis that the RVs and EVs are not linearly related.In a resulting RDA biplot, the lengthening factor of 1.5 was applied to EV vectors to improve the legibility of the biplot.Pearson correlation coefficients (r) between the RVs and EVs were calculated to quantify the symmetric relationship between them, and the dependency (asymmetric relationship) between them was individually evaluated by means of simple linear repression analyses using the ordinary least squares (OLS) method.Multiple linear regression (MLR) analysis using the OLS method was also performed where appropriate.
To determine the direct and indirect effects of EVs upon R THAUM , path analysis [54, 55] was performed.In a path model, path coefficients (PCs) were calculated using OLS-MLR to assess the contributions of EVs to R THAUM by means of both direct and indirect paths.The significance of the path model was evaluated by using the goodness of fit chi-square (χ 2 GoF ) as well as the root mean square error of approximation (RMSEA).In each significant path model, covariations among variables were decomposed to distinguish noncausal covariation from total covariation.All statistical data analyses were performed using the software packages PAST (Paleontological Statistics, Natural History Museum, University of Oslo, Oslo, Norway), Onyx (University of Virginia and Max Planck Institute for Human Development, http://onyx.brandmaier.de),IBM SPSS-AMOS (IBM Software, Armonk, NY, USA), and XLSTAT (Addinsoft, Paris, France).All software packages were used according to the instructions in their manufacturers' manuals.

Physicochemical and microbiological variables
In total, 27 soil samples were collected from various terrestrial sites; Table 1 summarizes their physicochemical properties.In our samples, variations of the EVs studied were measured as relative standard deviations (RSDs), and ranged from 10.5% to 196.1% (average 102.4%).Soil pH varied the least among samples, and the largest RSD was observed for TN.In every soil sample analyzed, the total inorganic carbon content was below the detection limit; thus we considered the total organic carbon content (TOC) to be equal to the TC; the TOC in turn was taken to reflect the OM content.The correlation coefficient between TC and OM was 0.814 (p < 0.05) in our preliminary study.
Our results for the relative abundance of Thaumarchaeota, R THAUM , suggested that, in terms of abundance per se, Thaumarchaeota might not be a major constituent of soil prokaryotes in our soil samples, although it was a numerically dominant member of soil Archaea.Our results are consistent with those of Ochsenreiter et al. [34] and Lehtovirta et al. [56], who reported that Thaumarchaeota represented up to 5% of the total prokaryotes in many soils.However, although such abundance figures may seem small, evaluations of the contribution of Thaumarchaeota to soil ecosystems should include the consideration that all the currently recognized members of Thaumarchaeota (at least all cultured or enriched Thaumarchaeota) are considered to be ammonia oxidizers.Because of the low energy yields (ΔG 0 0 = −235 kJ/mol)  produced in the oxidation of ammonia [57] relative to those of the complete oxidation of OM by organotrophs, ammonia oxidizers have to respire much more than organotrophs, resulting in a considerably larger contribution by Thaumarchaeota to geochemical cycling.Thus, in terms of their ecological importance, conversion factors might be applied to the number of Thaumarchaeota that are orders of magnitude higher than those for organotrophs.It should also be noted that Thaumarchaeota is the only prokaryotic phylum all of whose known members participate in the ammonia oxidation process, unlike Proteobacteria, only a few of whose members at the genus level are ammonia oxidizers (e.g., Nitrosomonas, Nitrosospira, Nitrosolobus, and Nitrosococcus).

Multivariate causal relationship projected onto a reduced space
Multivariate causal relationships between the EVs and the RVs determined from our soil samples were inferred by means of a gradient analysis using RDA.Because the maximum length of gradient determined by using detrended correspondence analysis was less than 2 standard deviations, RDA was used in lieu of canonical correspondence analysis, which is more appropriate for data sets with maximum gradient lengths greater than 3-4 standard deviations and assumes an unimodal model for response variables.To avoid the multicollinearity problem (association between NH 4 + -N and TN) as well as to simplify our initial inference, the EVs of inorganic nitrogen contents (NH 4 + -N and NO 3 − -N) were not included in the RDA.
The test of significance on the overall RDA results performed using 499 permutations showed that the canonical relationship between the RVs and EVs was highly significant (p = 0.018), indicating that the relative abundance of prokaryotic taxa estimated in this study and the environmental variables measured in this study were linearly related.The two canonical axes (CA-I and CA-II) shown in the RDA biplot (Fig 1) explained almost all (>99.9%) the variance in the RV data set, and the first axis (CA-I) explained the great majority (91.5%) of the variance.The relative abundance of the phylum Thaumarchaeota, R THAUM , contributed the most (81.6%) to the first canonical axis, and the relative abundances of Archaea and Bacteria (R ARCH and R BACT ) respectively showed a moderate contribution (18.4%) and the smallest contribution (<0.001%) to the first canonical axis.The biplot score of R BACT on the first canonical axis was very small (|−0.010|)compared to those of R THAUM (|+0.957|) and R ARCH (| +0.455|).Among the EVs, WC, TC, TN, and TS showed biplot scores greater than 0.9 on the first canonical axis, whereas temperature, pH, and TP showed biplot scores less than 0.3.In addition, when the causal effects of temperature, pH, and TP were controlled in partial RDA, the linear relationship between EVs and RVs was significant (p < 0.05), indicating that these EVs have marginal or no effect on RVs.The angles (directions) between RVs and EVs in the RDA biplot suggested negative causal relationships of R THAUM with nutrition level-related environmental variables (TC, TN, and TS) and with WC.R ARCH was also shown to be negatively related to these four EVs (TC, TN, TS, and WC), but the causal relationships between R ARCH and these EVs appeared to be less strong than the relationships between the R THAUM and these EVs.
We considered that raw values of thaumarchaeotal abundance estimates (N THAUM , thaumarchaeotal 16S rRNA gene copy number) could be a function of EV effects specific to Thaumarchaeota (E-EV THAUM ) as well as of EV effects affecting all prokaryotes (E-EV PROK ) (N THAUM = f[E-EV THAUM , E-EV PROK ], where EV THAUM E-EV PROK ).If E-EV PROK is considerably larger than E-EV THAUM , N THAUM could depend largely upon E-EV PROK , which could potentially lead us to misinterpret the results.In such cases, EVs substantially affecting all prokaryotes might be erroneously identified as EVs that affect Thaumarchaeota specifically.Therefore, N THAUM values normalized by the total 16S rRNA gene copy number (N PROK ), R THAUM = (N THAUM /N PROK ), could provide better abundance estimates for an inference on E-EV THAUM (R THAUM f(E-EV THAUM ).Correspondingly, when archaeal and bacterial amoA copy numbers (N ARCH-amoA and N BACT-amoA ) are compared, the raw values of archaeal amoA copy numbers (N ARCH-amoA ) represent the effects of EVs upon both ammonia oxidizers generally (E-EV PROK-amoA ) and archaeal ammonia oxidizers specifically (E-EV ARCH-amoA ).Hence, only the relative proportion, R ARCH-amoA = N ARCH-amoA /(N ARCH-amoA + N BACT-amoA ), could reflect the EV effects specifically upon archaeal ammonia oxidizers compared to those upon total ammonia oxidizers (R ARCH-amoA f[E-EV ARCH-amoA ]).As expected based upon the above considerations, the EV effects upon Bacteria (E-EV BACT ) could not be distinguished from the EV effects upon total prokaryotes (E-EV PROK ) in this study, because Bacteria represented a great majority of the prokaryotes in our soil samples and the sample variation in N BACT could dominate the sample variation in N PROK (if N BACT !N PROK [R BACT 1], then E-EV BACT E-EV PROK ).Also, the sample deviation coefficient (RSD) of R BACT was about 1/ 50 times those of R ARCH and R THAUM , indicating that EV effects upon many bacterial phyla or a number of subphylum-level taxa (E-EV i ) might be averaged or mutually cancelled out (∑[E-EV i ] 0), which could explain the low RDA biplot score of R BACT .Similarly, because almost 50% of archaeal members in our soil samples were estimated to be Thaumarchaeota, N THAUM became a major term in R ARCH .Hence, the responses of soil Archaea to EVs tended to resemble those of Thaumarchaeota in this study.

Individual causal effect of environmental variables
The RDA inference of the potential causal relationships between the EVs and the RVs was further examined by Pearson's correlation (symmetric analysis for reciprocal [bidirectional] relationships) and regression analyses (asymmetric analysis for causal [unidirectional] relationships).R THAUM showed significant (p < 0.05) negative correlations to TC (r = −0.563),TN (r = −0.708),and TS (r = −0.612),and to WC (r = −0.599)(Table 2).Correlations between R THAUM and EVs of inorganic nitrogen (NH 4 + -N and NO 3 − -N) were also significant (p < 0.05) and negative (−0.675 and −0.569, respectively).R ARCH showed correlation results similar to those of R THAUM , but its correlation with NO 3

−
-N was insignificant (p = 0.308).On the other hand, R BACT showed significant (p < 0.05) positive correlations only with TN (r = +0.420)and NH 4 + -N (r = +0.474).The relationship between R BACT and TN was not apparent in the RDA biplot, but was significant according to the symmetric analysis that did not assume an underlying causal relationship between these two variables.Moreover, R BACT showed significant negative correlations with R THAUM (r = −0.414,p < 0.05) and R ARCH (r = −0.944,p < 0.05), while R THAUM and R ARCH had a significant positive correlation (r = +0.459,p < 0.05).In asymmetric analyses using simple linear regression, results similar to those of the correlation analyses were obtained.All the EVs that showed strong correlations with RVs also resulted in significant (p < 0.05, ANOVA) regression coefficients (slope, β 1 ) (Table 3 and S2-S4 Figs).Each of these EVs explained >30% (R 2 = 0.317-0.501) of the variation in R THAUM .WC appeared to have the most prominent effect (the largest slope, β 1 ) on R THAUM in nonstandardized scale, while TN was the strongest determinant of R THAUM in the standardized (zscore-transformed) scale.Significant (p < 0.05, ANOVA) dependencies of R ARCH on TC, TN, WC, and NH 4 + -N were also observed, but the explanatory powers of these EVs were relatively low (<30%, R 2 = 0.170-0.302).Correlation results showed that R BACT was affected only by TN and one of its inorganic forms, NH 4 + -N.
Although pH was previously suggested as an important factor that could shape the niche of Thaumarchaeota (or AOA) [21,24,[58][59][60][61][62][63], no evidence suggesting a role of pH in controlling R THAUM was observed in this study.It was considered that the insignificant relationship observed between pH and R THAUM could arise from the relatively small sample variation (as reflected by RSD) of pH in our soil samples compared to those of other EVs that showed prominent effects on R THAUM .The lack of a significant relationship between temperature and R THAUM might also be attributed to our limited sampling; our soil samples were collected from temperate sites in Korea.On the other hand, despite the fact that the sample variation of TP was comparable to those of the prominent EVs (TC, TN, TS, and WC), TP appeared not to affect R THAUM .Although the causal effects of pH, temperature, and TP on R THAUM appeared insignificant in this study, the roles of these EVs in shaping the niche of soil Thaumarchaeota should be further examined by comprehensive surveys using wider ranges of sample EVs than those used herein.
To the best of our knowledge, all the physiologically described members of Thaumarchaeota were cultured (or enriched) in growth media containing bicarbonate (hydrogen carbonate, HCO 3 − ) as the sole source of carbon [3,[6][7][8][9][10][11][12][13][14][15][16], suggesting an autotrophic lifestyle of Thaumarchaeota.Evidence of autotrophy by Thaumarchaeota include the genomic components of a modified 3-hydroxypropionate/4-hydroxybutyrate (3HP/4HB) pathway [6,8,12,16,[64][65][66] and a reductive (or reverse) tricarboxylic acid (rTCA) cycle [66][67][68] found in genomes of the cultured (or enriched) members of Thaumarchaeota.Along with this genomic evidence, experimental results obtained using stable isotope probing confirmed the autotrophic growth and activity of Thaumarchaeota [27,69].However, heterotrophy (mixotrophy, more likely) has been also suggested for thaumarchaeotal physiology.In addition to genes encoding the 3HP/4HB pathway, oxidative TCA cycle genes were retrieved in genomes of C. symbiosum [64,67] and N. maritimus [66].Moreover, many physiological studies suggested that Thaumarchaeota has the potential to uptake OM for growth [16,57,[70][71][72][73][74].However, regardless of whether Thaumarchaeota is autotrophic or mixotrophic, cultures of Thaumarchaeota have been shown to be susceptible to organic carbon content in culture media, and our result regarding the effect of TC on R THAUM is consistent with these findings.Namely, addition of organic compounds, even in very low concentrations, inhibited the growth of N. maritimus [3], and slowed the ammonia oxidation of N. yellowstonii [7].Although Tourna et al. [16] reported that growth of N. viennensis was enhanced by the addition of a small amount of pyruvate to the culture medium, such growth-stimulating effects were observed only at very low concentrations (~0.05 mM).Suppressive effects of organic carbon on the abundance of Thaumarchaeota were also observed in a study by Wessen et al. [75].They reported a significant (p < 0.05) negative correlation (r = −0.60) between Thaumarchaeota (AOA) abundance and organic carbon content in soil samples; our results are consistent with this finding.Di et al. [76] also reported that the populations of Thaumarchaeota (AOA) were more numerous in soils with lower organic carbon content.In addition to such growth-suppressing effects, organic carbon content has been shown to be negatively related to Thaumarchaeota (AOA) species richness [69].Similar to the effect of TC on Thaumarchaeota, TN as well as the two inorganic forms of nitrogen (NH 4 + -N, and NO 3 − -N) negatively affected R THAUM in this study.While Hofferle et al. [22] and Stopnisek et al. [77] reported that ammonium concentration had a marginal or nonexistent effect on thaumarchaeotal (AOA) abundance in soil, a strong negative relationship between thaumarchaeotal growth and ammonium concentration in soil has been observed in many environmental studies [20,69,78], indicating that Thaumarchaeota prefer low ammonia concentration for growth and ammonia oxidation activity.Many studies on cultured members of Thaumarchaeota also showed that thaumarchaeotal growth rate decreased with increasing ammonium concentration.Growth of N. gargensis and of N. maritimus was inhibited at very low ammonium concentrations (2 mM and 3 mM, respectively) [8,79].Studies by Tourna et al. [16], Jung et al. [9], and Morley et al. [12] demonstrated growth inhibition of Thaumarchaeota (N.viennensis, N. koreensis, and N. devanaterra) at ammonium concentrations that were slightly higher (>20-50 mM), but still much lower than those that inhibited AOB growth.Some AOB were reported to be able to grow at >200 mM NH 4 + [80][81][82].Moreover, Tourna et al. [16] considered that the accumulation of toxic metabolic intermediates including nitrate (NO 2 − ) might inhibit the growth of AOA (N.viennensis) at high (>3.5 mM) ammonium concentrations.On the other hand, unlike those of ammonium, the effects of nitrate and TN upon thaumarchaeotal growth in culture media or upon estimates of its abundance in the environment have not yet been well studied.This could possibly be due to the fact that most physiological and ecological studies concerning the effects of nitrogen upon Thaumarchaeota have been focused mainly on ammonium (or ammonia), the substrate of ammonia oxidation.In our study, although we did not determine the ammonia oxidation rate, the NO 3 − /NH 4 + ratio, which might partly explain the rate of nitrification, correlated positively to R THAUM (r = 0.456, p = 0.017) and negatively to organic carbon content (r = −0.512,p = 0.006).

Indirect but ultimate causal effect of OM
Considering that most soil nitrogen is in organic form [83], the effect of TN on R THAUM could overlap with the effect of TC on R THAUM .Because no samples in this study were collected from sites to which fertilizer (artificial sources of nitrogen and carbon) had been applied, both the TN and TC in our samples were thought to have originated from indigenous OM such as plant debris and other types of indigenous biomass, under the assumption that carbon fixation and nitrogen fixation are the only routes of carbon and nitrogen input to our soil samples.Thus, the level of OM could determine the levels of TC and TN, and the levels of TC and TN could subsequently be interrelated.In fact, TC and TN showed a significant positive correlation (r = 0.653, p < 0.05) in this study.Bearing in mind these points, we questioned if the observed EV effects of TC or TN on R THAUM were indirect causal effects.Even though only TN actually affects Thaumarchaeota and TC actually does not, a superficial relationship between TC and R THAUM likely appeared because TC correlated with TN.Hence, we applied path analysis [54,55] to determine the direct and indirect causal effects of TC and TN upon R THAUM , as well as to test their causal ordering.
In the path analysis, we found a significant (χ 2 GoF < 0.001, RMSEA < 0.001) path model that explained the causal effects of TC and TN upon R THAUM (Fig 2).In this path model, only TN directly affected R THAUM (PC = −0.593,p = 0.004).The effect of TC upon R THAUM was indirect (PC = −0.176,p = 0.356), and TN was a mediator EV.This result suggested that the relationship between TC and R THAUM was chiefly an indirect effect mediated by TN.In addition, when bivariate covariations among the variables were decomposed (Table 4), only 31.3% of the total covariation between TC and R THAUM was explained by direct causation, whereas 83.8% of the total covariation between TN and R THAUM was explained by direct causation.Note that a finding of an indirect effect does not mean that there is no effect, but does mean that the effect is mediated by another variable.Additionally, we supposed that the TC level measured in this study (recall that TC TOC, as total inorganic carbon was below the detection limit for all samples) could be used as a proxy for the level of OM in our soil samples.Since carbon content is generally about 10 times that of nitrogen content in natural OM (i.e., its C/N ratio is ca.10), the observed TC effect was considered to dominate the expected OM effect.Besides, the TC level strongly corresponded to the OM level in our preliminary study.Thus, in terms of causal ordering, we concluded that OM might be an indirect but ultimate EV controlling the abundance of Thaumarchaeota, and that this causal effect of OM could be mediated by TN.Although insignificant and weak, the direct causal effect of OM upon R THAUM might correspond to the growth-suppressive effect of organic carbons (mostly carbohydrates) observed in the culture-based studies mentioned above.On the other hand, in the path analysis that was applied to TS under the assumption that TS also originates from OM, no significant path models were found.This result suggested that our assumption regarding TS could be false, or that there might be latent mediator EVs that were not measured in this study (e.g., SO 4 − and HS − ).A recent study by Park et al. [65] reported that AOA were selectively enriched over AOB in the presence of sulfur-oxidizing bacteria.Although they concluded that the selective enrichment of AOA was likely due to oxygen depletion caused by the rapid growth of sulfur-oxidizing bacteria, an oxidized sulfur species might enhance the growth of AOA, or, inversely, a reduced sulfur species might inhibit the growth of AOA.Except for the nutrient-related EVs, only WC showed a significant correlation with R THAUM (r = −0.599,p < 0.05) and was a significant EV explaining the variance of R THAUM (R 2 = 0.359, p < 0.05), implying that the members of Thaumarchaeota in soil could prefer low-WC conditions or that they could be less vulnerable to desiccation stress than their competitors.Microorganisms exposed to low-WC environments must possess mechanisms to avoid water loss by osmosis, and a well-known strategy to maintain turgor is to accumulate intracellular compatible solutes (osmolytes).All xerophiles (including some halophiles) produce and accumulate low-molecular-mass organic compounds that have osmotic potential (organic osmolytes) [84,85].In Thaumarchaeota, genes coding for the biosynthesis of ectoine, an organic osmolyte, were observed in the genome of N. maritimus [66].However, no experimental evidence has been reported regarding xerophilic (or xerotolerant) characteristics of Thaumarchaeota.Even though not being xerophiles, soil microorganisms are generally considered to be adapted to cope with water stress because water levels in soil fluctuate.Hence, we first hypothesized that the terrestrial members of Thaumarchaeota might have better mechanisms to respond to desiccation stress than their competitors do.However, this hypothesis was inconsistent with our conclusions regarding TC and TN because WC showed positive correlations with TC and TN, both of which negatively affect R THAUM .To avoid this contradiction, an alternative hypothesis on the effect of WC on R THAUM was formulated by using path analysis.
In the path analyses employing WC as a fixed explanatory variable and the other EVs as inand-out variables, we found a significant (χ 2 GoF < 0.001, RMSEA < 0.001) path suggesting an effect of WC, mediated by ammonium, upon R THAUM (Fig 2).Unlike the SLR model, in which WC appeared to be one of the significant determinants of R THAUM , the direct effect of WC on R THAUM in this path was almost zero (PC = 0.024) and insignificant (p = 0.939).WC directly affected ammonium concentration (PC = 0.879, p < 0.001) rather than R THAUM , and the ammonium concentration subsequently affected R THAUM .Only 4.0% of the total covariation between WC and R THAUM was explained by direct causation, whereas 96.9% of the total covariation between ammonium and R THAUM was explained by direct causation.We hypothesized based on the path analysis results that WC negatively affected R THAUM by means of the mineralization of organic nitrogen (ammonification).Considering that WC positively regulates the rate of nitrogen mineralization (a major route of inorganic nitrogen input in soil) [86][87][88][89], low WC could result in increased R THAUM by decreasing the ammonium level, because Thaumarchaeota might prefer low-ammonium conditions.This hypothesis was consistent with our suggestion that, among the nutrient-related EVs, nitrogen content might be the most direct factor controlling Thaumarchaeota in soil.To present a possible scenario describing all the prominent relationships between soil EVs and Thaumarchaeota identified in this study, we developed a combined path model that can be examined in future studies (Fig 2).Although this model was not significant (χ 2 GoF = 60.9,RMSEA = 0.377), we tried to incorporate all the direct and indirect causal effects of EVs on R THAUM in this single model.There might be latent variables (e.g., inorganic sulfur/phosphorus) that were not included in this study, or missing links between variables.

Concluding Remarks
The relative abundance patterns of Thaumarchaeota in this study demonstrated the oligotrophic lifestyle of Thaumarchaeota in soil.The oligotrophy of Thaumarchaeota might be attributed to their enzymes and transporters with high affinity, which provide them a competitive advantage under nutrient-limiting conditions.In fact, it has been suggested that the reason why Thaumarchaeota favor low ammonium concentrations is their exceptionally low half-saturation constant (K m = 0.13-0.61μM) of ammonia monooxygenase [9,65,79], which was 1/40 to 1/10,000 times that of the K m values of AOB (K m = 30-1,300 μM) [65,90,91].Another hypothesis regarding energy metabolism was also previously formulated based on thaumarchaeotal genomic sequences to explain the ammonium oligotrophy of Thaumarchaeota.Genomes of Thaumarchaeota (C.symbiosum and N. maritimus) [64,66] were reported to lack genes encoding the hydroxylamine-ubiquinone redox module, which has been considered indispensable in electron recycling during the initial oxidation of ammonium, generating proton motive force and reducing power.To explain this unusual genomic feature, Walker et [66] hypothesized that Thaumarchaeota might not require the electron recycling module because archaeal ammonia monooxygenase produces nitroxyl hydride (HNO) instead of hydroxylamine (NH 2 OH), suggesting that this simplified ammonia oxidation process could provide an ecological advantage to Thaumarchaeota under ammonium-limited conditions.Along with its high affinity to substrate and its efficient energy metabolism, Thaumarchaeota might have other characteristics common in oligotrophs (e.g., high vulnerability to toxic metabolites rapidly accumulated under nutrient-rich conditions and to energy depletion caused when excessive transportable nonmetabolic substances suddenly become available) [92], a topic that remains to be studied.Although published thaumarchaeotal genome sequences have shown only one copy of the 16S rRNA gene in each thaumarchaeotal genome [6,10,11,13,14,64,66,93], many other prokaryotes have been reported to possess multiple copies of the 16S rRNA gene in each cell [94].Errors related to such variations in the 16S rRNA gene copy number per genome among prokaryotic taxa could affect our results on the relative abundance of Thaumarchaeota, possibly causing underestimation of the actual abundance of Thaumarchaeota in soil.Also, environmental variables measured in this study do not represent the full spectrum of environmental variables that Thaumarchaeota interact with.Hence, our conclusion might be only a snapshot of the ecology of Thaumarchaeota.Comprehensive ecological studies overcoming our limitations, and future research on the physiology and molecular biology of cultured/enriched cells, will allow us to more fully understand the ecological niche of the intriguing phylum Thaumarchaeota.

Fig 1 .
Fig 1. RDA biplot representing the relative abundance of prokaryotic taxa and environmental variables.Solid-line arrows and dashed-line arrows represent the biplot scores of the relative abundances of prokaryotic taxa and of the environmental variables, respectively.Values in parentheses indicate the percentages of the total variation that are explained by each canonical axis.doi:10.1371/journal.pone.0133763.g001

Fig 2 .
Fig 2. Path diagram of the effects of environmental variables upon the relative abundance of Thaumarchaeota in soil.Path model A shows the causal effects of TC and TN upon R THAUM , and path model B shows the causal effects of WC and NH 4 + -N on R THAUM .Path model C combines A and B in a single concatenated model; the link between models A and B is indicated by a grey arrow.Causal ordering is represented by arrows.Solid lines and dashed lines respectively indicate direct and indirect causal effects in path models A and B, and the thickness of each solid line represents its PC.In path model C, all environmental variables are included that showed significant effects on R THAUM in simple linear regression analysis.doi:10.1371/journal.pone.0133763.g002

Table 1 .
Summary of physicochemical and microbiological properties of soils used in this study.

Table 2 .
Correlations between environmental variables and relative abundances of Thaumarchaeota, Archaea, and Bacteria.Lower left half, Pearson correlation coefficients (r); upper right half, p values.

Table 3 .
Results of regressions between environmental variables and relative abundances of Thaumarchaeota, Archaea, and Bacteria.

Table 4 .
Direct and indirect causal effects of environmental variables upon the relative abundance of Thaumarchaeota in hypothesized path models.Significant (p < 0.05, t-statistic calculated from MLR using the OLS method) direct effects are displayed in bold.doi:10.1371/journal.pone.0133763.t004 b