The Spatial Factor, Rather than Elevated CO₂, Controls the Soil Bacterial Community in a Temperate Forest Ecosystem

Abstract

The global atmospheric carbon dioxide (CO₂) concentration is expected to increase continuously over the next century. However, little is known about the responses of soil bacterial communities to elevated CO₂ in terrestrial ecosystems. This study aimed to partition the relative influences of CO₂, nitrogen (N), and the spatial factor (different sampling plots) on soil bacterial communities at the free-air CO₂ enrichment research site in Duke Forest, North Carolina, by two independent techniques: an entirely sequencing-based approach and denaturing gradient gel electrophoresis. Multivariate regression tree analysis demonstrated that the spatial factor could explain more than 70% of the variation in soil bacterial diversity and 20% of the variation in community structure, while CO₂ or N treatment explains less than 3% of the variation. For the effects of soil environmental heterogeneity, the diversity estimates were distinguished mainly by the total soil N and C/N ratio. Bacterial diversity estimates were positively correlated with total soil N and negatively correlated with C/N ratio. There was no correlation between the overall bacterial community structures and the soil properties investigated. This study contributes to the information about the effects of elevated CO₂ and soil fertility on soil bacterial communities and the environmental factors shaping the distribution patterns of bacterial community diversity and structure in temperate forest soils.

The global atmospheric carbon dioxide (CO₂) concentration has increased from a preindustrial value of about 280 μmol mol⁻¹ to 379 μmol mol⁻¹ in 2005 and is predicted to increase continuously over this century (38). The increase of atmospheric CO₂ concentration may result in a series of profound changes in terrestrial ecosystems (2, 33, 36). Some studies reported enhanced plant productivity under elevated CO₂ (37, 53), which implied a potential negative feedback where the CO₂ increase could be partially abated by enhanced carbon (C) storage in terrestrial ecosystems. However, because plant production in many temperate forest ecosystems is nitrogen (N) limited, the sustainability of CO₂-enhanced plant productivity may depend largely on the N availability and balance of nutrients in soil (16, 48, 60). Consequently, although most previous studies on the effects of elevated CO₂ on biological systems have been single-factor experiments, there is considerable need to perform such work in a multifactorial framework that considers the concurrent effects of nutrient amendment (2, 68).

Soil bacteria are vital to ecosystem function and health through their roles in residue decomposition and nutrient cycling and their associations with other organisms (13, 27), which make them a key player in maintaining the productivity and sustainability of terrestrial ecosystems (4, 7, 55). Despite the potential significance of soil bacteria in maintaining the sustainability of higher plant productivity under elevated CO₂ condition, few large-scale field studies have examined the effects of elevated CO₂ on bacterial communities in temperate forest soils over a range of N availability using clone library techniques with high taxonomic resolution. Most previous studies on microbe-related responses to elevated CO₂ have focused on fungi (42, 64) or on gross microbial parameters such as microbial biomass, soil respiration, and soil enzyme activity (5, 15, 23, 41, 43, 46, 76, 77). Because of the much higher concentrations of CO₂ in soils than in the atmosphere, it is likely that elevated CO₂ has little direct effect on soil bacterial communities (11, 39). It is most probable that the elevated CO₂ affects soil bacterial communities indirectly through enhanced soil organic C input (40), changed chemistry of leaf litters and root exudates (57, 71, 72), and other soil properties (58, 59). Enhanced C inputs may result in a surplus of C relative to N in soil, especially where N is limiting, and thus benefit fungi over bacteria, leading to a decrease of soil bacterial abundance and diversity (34). Changes in leaf litter and root exudates may also affect soil bacterial communities (73, 79, 80), partially due to the differences of bacterial taxa in their physiological abilities to utilize different compounds. However, soil chemical properties depend on the quality and amount of organic C inputs caused by contemporary disturbances such as elevated CO₂, as well as the soil conditions at the initiation of the study, reflecting microsite variation due to topography, hydrology and historical land use-related activities (19, 64, 72, 73). These factors introduce high heterogeneity to field sites, making it more difficult to observe significant effects of experimental elevation of CO₂ on soil bacterial communities. Accommodating this reality, it is necessary to explicitly consider the effects of spatial factors when evaluating the effect of elevated CO₂ on these communities.

In this study, 12 soil samples associated with 10-year CO₂ treatment and 1-year N treatment were collected from six plots at the Duke Forest free-air CO₂ enrichment (FACE) experiment site in a temperate forest ecosystem. Duke FACE is a large-scale field experiment utilizing a technology allowing intact portions of forest to be exposed to current and elevated levels of CO₂ in an otherwise unaltered environment (30). From these samples, a total of 809 bacterial clones were sequenced and analyzed, in combination with denaturing gradient gel electrophoresis (DGGE) analysis, to examine the variations in overall soil bacterial diversity and structure associated with different plots (spatial factor) as well as CO₂ and N treatments.

This study represents the first attempt to quantify the relative importance of the spatial factor, elevated CO₂, and N amendment in influencing the soil bacterial community diversity and structure of an intact forest ecosystem through a combination of an entirely sequencing-based approach, DGGE, and advanced statistical analysis. We tested the following hypotheses. (i) CO₂ and N treatments will dominate the makeup of soil bacterial communities as reflected in soil bacterial community diversity and structure; otherwise, the makeup of soil bacterial communities will be mainly dominated by the spatial factors. (ii) Soil chemical heterogeneity, affected either by long-term historical processes or contemporary environmental disturbances, will determine the distribution patterns of the soil bacterial community.

MATERIALS AND METHODS

Site and sampling.

A total of 12 soil samples associated with CO₂ and N treatments were collected from six plots at the Duke FACE experiment site in the Blackwood Division of Duke Forest, North Carolina (35°58′41″N, 79°05′39″W). The site has a temperate climate with average annual precipitation of 1,145 mm and an average annual temperature of 15.5°C. The FACE plots were established in a loblolly pine (Pinus taeda L.) stand planted in 1983. In August 1996, six plots began operation; three were fumigated and three remained under ambient CO₂, forming together six plots in a randomized block design (n = 3). The plots are 30 m in diameter, with three under the ambient CO₂ concentration (i.e., 382 μmol mol⁻¹ in 2000) and three treated with CO₂ to a target of 200 μmol mol⁻¹ above the ambient concentration. In 2005, each of the plots (and additional plots not used in this study) was split in half, with one part fertilized at a rate of 11.2 g N m⁻² year⁻¹ in the form of ammonium nitrate to examine the effect of interaction of elevated CO₂ and N nutrition (60). Thus, the treatments (three replicates of each treatment) were designated as follows: AC-AN for ambient CO₂ plus ambient nutrient, EC-AN for elevated CO₂ plus ambient nutrient, AC-EN for ambient CO₂ plus elevated nutrient, and EC-EN for elevated CO₂ plus elevated nutrient.

The soil type at the experiment site was a low-fertility acidic Hapludalf (Ultic Alfisol in the Enon series) with clay loam texture (61, 64). Soils were collected in 2006. To minimize intraplot variability derived from the inherent heterogeneity of soil, five soil cores (0 to 10 cm) were randomly collected using a 5-cm-diameter hammer corer from each part (fertilized and unfertilized) of three elevated-CO₂ plots and three ambient-CO₂ plots and then well mixed to form one composite sample. This sampling strategy allowed each soil sample to represent one specific CO₂ treatment (ambient or elevated), N treatment (ambient or elevated), and spatial factor (different plot location). Each sample was placed in a sterile plastic bag, sealed, and transported to the laboratory on ice. All samples were passed through a 2.0-mm sieve and stored at −80°C for DNA analysis. The spatial information for and chemical properties of the soils used in this study are shown in Table S1 in the supplemental material.

Soil DNA extraction, PCR, DGGE, cloning, and sequencing analyses.

Soil DNA was extracted using the Powersoil DNA isolation kit (Mo Bio, Carlsbad, CA) according to the manufacturer's instructions. The 16S rRNA gene was amplified by PCR based on a protocol described previously, using the extracted DNA as the template and universal bacterial primers 954F and 1369R (20, 75). Amplicons (about 200 ng) were analyzed by DGGE, and the cloning experiments were performed following a protocol described previously (18, 19).

OTU definition, bacterial identification, and diversity indices calculation.

The DNA sequences obtained were manually proofed using Sequencer 4.7 (Gene Codes, Ann Arbor, MI). Possible chimeric sequences were checked using the Bellerophon chimera detection program (http://foo.maths.uq.edu.au/∼huber/bellerophon.pl) (35). Valid sequences were clustered into operational taxonomic units (OTUs) based on cutoff evolutionary distances of 0.03 using the program DOTUR (67). The sequences were identified with the taxonomic affiliation using the program Classifier in the Ribosomal Database Project II (http://rdp.cme.msu.edu/) (49). Nonparametric richness estimators (Chao1 and ACE) were used to estimate the total bacterial species richness at the experiment field site.

For the DGGE data set, each detected band in the DGGE gel was defined as an OTU according to its relative front value, and the relative pixel intensity for each band was defined as the relative abundance (66). DGGE is a commonly used DNA-based community fingerprinting method for detecting the differences between microbial communities (66). Although DGGE patterns provide a relatively lower taxonomic resolution than species identification and members of different species may share the same DGGE band (20), they provide a consistent measure of community composition and diversity (3).

Genotypic richness (J) and the Shannon index (H′), the most widely used diversity estimates, were selected to compare the variation of soil bacterial diversity among the soil samples (19). Because the numbers of successfully sequenced clones differed among soil samples, we increased the reliability of the comparison by estimating rarefied diversity measures (genotypic richness and Shannon index) using DOTUR and rarefying clones from each soil sample to the number in the plot with the smallest number of sequenced clones.

Comparison of bacterial genotypic diversity and community structure.

Analysis of variance (ANOVA) for split-plot design followed by Student's t means comparison was performed to compare mean diversity estimates among the treatments. Principal-coordinate analysis (PCoA), an unconstrained ordination technique, was conducted to examine the Bray-Curtis distances between bacterial communities. Nonparametric multivariate ANOVA (NPMANOVA), analysis of similarities (ANOSIM), and the multiresponse permutation procedure (MRPP) were conducted to further confirm the results of PCoA and test for significant differences in overall community structure among the treatments. These nonparametric multivariate tests are well suited for a complex community data set, as they do not require the data set to meet statistical assumptions such as normality, equal variances, and independence (29). ANOVA, PCoA, NPMANOVA, ANOSIM, and MRPP analyses were conducted using the “R” statistical software (http://www.r-project.org/) (69).

Partitioning the relative importance of categorical factors and quantitative variables.

Multivariate regression tree (MRT) analysis was conducted to partition the relative effects of the spatial factor, elevated CO₂, and N amendment on soil chemical properties and thus on soil bacterial community diversity and structure. For MRT analysis, dissimilarity among the response variables is defined as the total sum of squares of the responding variable values, and the least-sum-of-squares criterion is used to repeatedly split data into two groups based on one of the classification variables. Each split tends to minimize the dissimilarity within the two resulting groups and maximize the dissimilarity between the two resulting groups based on a single environmental classification by comparing all the potential splits in the data (9). Multivariate regression tree analysis was conducted by using the package “mvpart” of the “R” statistical programming environment (http://cran.r-project.org/web/packages/mvpart/) (9).

Identification of the key environmental variables determining the distribution patterns of soil bacterial diversity and structure.

Linear regressions were conducted to explore the potential distribution pattern of soil bacterial diversity along gradients of soil chemical properties. The Mantel test was conducted to explore the environmental and spatial heterogeneities of the soil bacterial community by examining the Pearson correlation between the environmental distance (Euclidean distance) of the soil chemical properties, the spatial distance of the sampling plots, and the ecological distance (Bray-Curtis distance) of the bacterial communities with 1,000 permutations. Linear regressions and the Mantel test were conducted using the “R” statistical software (69).

Nucleotide sequence accession numbers.

The 16S rRNA gene sequences have been deposited in the GenBank database and assigned accession numbers FJ432787 to FJ432986, FJ433055 to FJ433253, FJ433323 to FJ433524, and FJ433593 to FJ433800.

RESULTS

Overall bacterial community composition and diversity at the experimental site.

In total, 809 valid sequences were obtained by sequencing 838 clones. These sequences yielded 331 OTUs at a Jukes-Cantor evolutionary distance of 0.03, which is a much higher number than those detected by DGGE (41 OTUs). The Chao1 method and the ACE method yielded estimates of 678 (95% confidence interval, 560 to 855) and 712 (95% confidence interval, 585 to 836) total species, respectively, at a Jukes-Cantor evolutionary distance of 0.03. Therefore, approximately 46% of the bacterial taxa at the experimental site were sampled. The rank-abundance plot demonstrated that, consistent with other bacterial community studies, many rare and a few common taxa were encountered, with 211 OTUs (64%) detected only once and 10 OTUs (3%) detected more than 10 times.

The program Classifier in the Ribosomal Database Project II was used to identify the taxonomic affiliations of the 809 sequences (331 OTUs) at hierarchical taxonomic levels (see Table S2 in the supplemental material). At the phylum level, Proteobacteria (50.7%) was the dominant taxonomic group, followed by Acidobacteria (26.2%), Actinobacteria (10.5%), Bacteroidetes (5.3%), Gemmatimonadetes (0.5%), and Firmicutes (0.2%). Alphaproteobacteria (39.7%), Acidobacteria (26.2%), and Actinobacteria (10.5%) were the most abundant groups at the class level; Acidobacteriales (26.2%), Rhizobiales (21.9%), and Actinomycetales (7.3%) were the most abundant groups at the order level; Acidobacteriaceae (26.2%) and Bradyrhizobiaceae (11.2%) were the most abundant groups at the family level; and Gp1 (22.6%) and Bradyrhizobium (8.9%) were the most abundant groups at the genus level.

Effects of elevated CO₂ and N amendment on soil bacterial communities.

Although lower values for mean rarefied richness and the Shannon index of soil bacterial communities under elevated CO₂ were observed (Table 1), the global test of ANOVA and the following Student's t means comparison showed that neither CO₂ nor N affected the mean rarefied richness and Shannon index significantly. Similar to the results of the comparison of soil bacterial diversity estimates, the results of PCoA suggested that neither CO₂ nor N affected the overall bacterial community structure, because for none of the treatments were the associated samples clustered in the two-dimensional PCoA graph (Fig. 1). The significance test of community difference using NPMANOVA, ANOSIM, and MRPP further confirmed the results of PCoA that there is no significant difference among the treatments (Table 2).

TABLE 1.

Genotypic diversity properties of soil bacteria associated with 10 years of CO₂ treatment and 1 year of N treatment

Treatmentb	Genotypic richnessa determined by:		Shannon indexa determined by:
	Sequencing	DGGE	Sequencing	DGGE
AC-AN	46.33 ± 1.76	31.33 ± 0.67	3.68 ± 0.09	3.22 ± 0.03
EC-AN	44.33 ± 4.70	28.00 ± 1.53	3.55 ± 0.21	3.12 ± 0.04
AC-EN	46.00 ± 0.58	30.67 ± 0.33	3.60 ± 0.01	3.20 ± 0.02
EC-EN	42.00 ± 2.89	30.33 ± 1.76	3.50 ± 0.09	3.19 ± 0.06

^aValues are means ± standard errors; none of the values within the same column were different at a P value of <0.05.

^bAC, ambient CO₂; AN, ambient nutrient; EC, elevated CO₂ (200 μmol mol⁻¹ above ambient CO₂); EN, elevated N nutrient (fertilized at a rate of 11.2 g N m⁻² year⁻¹).

FIG. 1.

Principal-coordinate analysis (PCoA) of the soil bacterial community structures associated with 10-year CO₂ treatment and 1-year N treatment using the sequencing-based data set (a) and the DGGE-based data set (b). Each point in the PCoA graph represents a sample associated with a treatment. A near distance of two points in the graph indicates a small ecological distance between the communities. AC, ambient CO₂; AN, ambient nutrient; EC, elevated CO₂ (200 μmol mol⁻¹ above ambient CO₂); EN, elevated nutrient (fertilized at a rate of 11.2 g N m⁻² year⁻¹); DGGE, denaturing gradient gel electrophoresis.

TABLE 2.

Significance test of dissimilarity of soil bacterial communities associated with 10 years of CO₂ treatment and 1 year of N treatment^a

Data set	NPMANOVA		ANOSIM		MRPP
	F	P	R	P	δ	P
Sequencing	0.815	0.970	−0.33	0.988	0.788	0.969
DGGE	0.735	0.779	−0.139	0.836	0.192	0.821

^aNPMANOVA, nonparametric multivariate ANOVA; ANOSIM, analysis of similarities; MRPP, multiresponse permutation procedure.

Effects of the spatial factor on soil bacterial communities.

If the bacterial diversity or community structure shows substantial spatial heterogeneity, the values of diversity estimates or community similarity between pairs of nearby plots should be more similar than those between distant plots. The results of the Mantel test of the Pearson correlation demonstrated that, generally, there was a significant positive correlation between the spatial distance of the sampling plots and the ecological distance of the diversity estimates or the bacterial genotypic community among the different plots, although the detailed results derived from the two different molecular approaches are not always similar. For example, there was a significant positive correlation between the spatial distance and the Bray-Curtis distances of bacterial genotypic richness (sequencing data set, r = 0.47 and P = 0.002; DGGE data set, r = 0.29 and P = 0.020), the sequencing-based Shannon index (r = 0.38 and P = 0.008), and the DGGE-based community distance (r = 0.32 and P = 0.010).

Relative importance of categorical factors and quantitative variables.

Multivariate regression tree (MRT) analysis was conducted for partitioning the relative effects of the spatial factor, elevated CO₂, and N amendment on soil chemical properties and, in turn, soil bacterial community diversity and structure. The results of MRT analyses show that the variation in soil chemical properties explained by the spatial factor (50.6%) was much higher than that explained by CO₂ treatment (14.7%) (Fig. 2). Of the effects of categorical variables on the soil bacterial community, more than 70% of the variation in soil bacterial diversity and 20% of the variation in community structure were explained by the spatial factor, while less than 3% of the variation was explained by CO₂ or N treatment (Fig. 2). Of the effects of quantitative variables, the variation in soil bacterial diversity explained by total N or the C/N ratio was much higher than that explained by other variables, while there was no dominant variable explaining the variation in the overall soil bacterial community (Fig. 2).

FIG. 2.

Summary of multivariate regression tree (MRT) analysis to illustrate the relative contributions of the categorical or quantitative environmental variables in influencing soil bacterial diversity (a and c) and structure (b and d) derived from the sequencing-based data set (a and b) and the DGGE-based data set (c and d). Solid lines represent the response variables which can be explained by the environmental variables through multivariate regression tree analysis, and line width is proportional to the percentage of the explanation, while dashed lines represent the contributions of the environmental variables for which the variation of the response variables is negligible. DGGE, denaturing gradient gel electrophoresis.

Distribution patterns of soil bacterial communities.

The linear regressions between the bacterial diversity properties and soil chemical properties show that there was a significant negative linear relationship between the C/N ratio and either sequencing-based or DGGE-based genotypic richness or Shannon index, as well as a significant positive linear relationship between total N and sequencing-based genotypic richness (Fig. 3). It seems that there were no significant correlations between diversity estimates and the soil properties pH, total organic C, δ¹³C, and δ¹⁵N. The results of the Mantel test of the Pearson correlation show that there was no significant correlation between the ecological distances of bacterial communities and the environmental distance of the soil properties investigated, including soil pH, total organic C, C/N ratio, total N, δ¹³C, and δ¹⁵N.

FIG. 3.

Negative linear relationships between C/N ratio and sequencing-based genotypic richness (a) and Shannon index (b), positive linear relationships between total N and rarefied genotypic richness (c) and Shannon index (d), and negative linear relationships between C/N ratio and DGGE-based genotypic richness (e) and Shannon index (f). Each point on the graph represents a sample associated with CO₂ and N treatments. AC, ambient CO₂; AN, ambient nutrient; EC, elevated CO₂ (200 μmol mol⁻¹ above ambient CO₂); EN, elevated nutrient (fertilized at a rate of 11.2 g N m⁻² year⁻¹); OTU, operational taxonomic unit.

DISCUSSION

Differences in soil bacterial community and structure.

Our first hypothesis was that elevated CO₂ and increased N availability would strongly affect the community diversity and composition of soil bacteria by generating more and different soil organic C inputs, thus changing the resources of soil bacteria. Our results do not support this hypothesis and showed that there was no significant effect of elevated CO₂ or N amendment on the overall bacterial community structure and diversity. This is consistent with results from previous studies showing no effect of elevated CO₂ on overall bacterial community composition, richness, and total abundance (1, 8, 22, 44, 78), although some studies reported significant changes in specific bacterial groups such as deltaproteobacteria (8), heterotrophic decomposers (29, 44), rhizosphere colonizers (12), and ammonia-oxidizing bacteria (32). Our results are further supported by the finding at the same site that overall ectomycorrhizal fungal richness and diversity were not affected by elevated CO₂ (64). Consistent with our observation of no effect of N amendment on overall bacterial community properties, previous studies also demonstrated relatively weak effects of N-containing inorganic fertilizers on bacterial communities in temperate agricultural soil after 17 years of continuous treatment (20) and no effect of inorganic N fertilizer on the Nitrobacter community (17) and the nitrate-reducing community (10).

There are a number of plausible explanations as to why the overall soil bacterial community structure and diversity did not significantly change following the treatments with elevated CO₂ and N. First, elevated CO₂ can affect soil bacterial communities only indirectly through changes on resource quality and availability. However, soil chemical properties are not merely a reflection of recently applied treatments but mostly reflect the initial conditions of the soil determined by a series of historical, long-term processes (19, 64, 72). Furthermore, it has been reported that slight topographical relief at the experimental site in Duke Forest generated a “treatment-similar” effect on plant productivity and C input to the soil (51, 53, 62), thus further disturbing the effects of CO₂ and N treatments. In addition, recent weather events contributed to increased spatial heterogeneity through uneven impact across the site (54). The resulting high heterogeneity of soil chemical properties makes it difficult to isolate the signal of the effect of elevated CO₂ or N amendment on soil bacterial diversity and community structure.

Consisting with previous examination of soil C sequestration in Duke Forest after 9 years of atmospheric CO₂ enrichment studies (45), we did not observe an obvious link between surface soil C and N pools and the imposed treatments. However, we did observe distinct distribution patterns of soil bacterial diversity along the gradients of C/N ratio and total N. These results support our second hypothesis that resource availability plays an important role in shaping the distribution patterns of the soil bacterial community. The variation of resource availability at the experimental site in Duke Forest was derived mainly from the initial conditions of the soil.

Relative importance of CO₂, N, and the spatial factor.

We quantitatively disentangled the relative importance of space, CO₂, and N in influencing soil chemical properties and thus soil bacterial diversity and community structure in a multifactor hierarchical framework. The simultaneous consideration of the relative importance of multiple factors such as contemporary environmental disturbances (CO₂ and N treatments) and historical and topographical processes (spatial factor) in influencing soil bacterial communities could form a meaningful framework for understanding the effect of elevated CO₂ in a hierarchical context and thus for identifying the dominant factor and environmental variables shaping the distribution patterns of soil bacterial communities (19, 50).

The partitioning analyses demonstrated that the spatial factor is a major driver of the variations in soil bacterial community diversity and structure. The strong effects of the spatial factor on the soil bacterial community are further supported in that the structure of the bacterial diversity was more similar in pairs of nearby plots than in distant plots, reflected in significant positive correlation between the spatial distance of the sampling plots and the ecological distance of the diversity estimates or community structure. Although to date no study has explicitly considered the effects of spatial factors when evaluating the effect of elevated CO₂ on soil bacterial communities, several previous studies have emphasized the dominant role of spatial and historical factors in shaping the distribution patterns of microbial communities (19, 50, 70, 74). Taxon-area relationships have also been repeatedly reported for microbial communities, providing further evidence for microbial biogeography (3, 21, 31). The observations in this study indicate a strong effect of historical and spatial processes (e.g., dispersal history and long-term integrated environmental heterogeneity) on soil bacterial diversity (19, 50, 70). It is likely that location is an index of the possibility that past divergence and diversification of microbial assemblages, whether due to neutral genetic drift or to adaptation to past environments, are inherited through genetic isolation because of spatial separation (6, 65).

Although most of the variations in bacterial diversity are related to predefined categorical factors, around 20% of the variation in the bacterial diversity and more than half of the variation in the community structure remained unexplained. This indicates that factors that we did not study are responsible for the diversity variation. These factors may include environmental variability in space and time not being measured at the appropriate scale, sampling effects, and neutral ecological drift (65).

Potential distribution patterns of soil bacterial diversity.

Bacterial diversity estimates were negatively correlated with C/N ratio and positively correlated with total N. Uncovering relationships between the distribution patterns of soil bacterial diversity and quantitative environmental variables may lead to improved prediction of the potential variations in soil bacterial diversity along environmental gradients. Several previous studies have linked soil and climate properties to soil bacterial diversity, showing that bacterial diversity correlated positively with soil pH (3.5 to 7.5) on a regional or continental scale (14, 24). We show that the C/N ratio and total N were potentially important for predicting the variations in soil bacterial diversity, at least at the local scale, where soil pH varied relatively little (4.1 to 5.2) and was uncorrelated to bacterial diversity. Because at this site N acts as a resource that limits plant growth (16, 52, 60, 63), it is not surprising that a positive correlation between total N and soil bacterial diversity emerged. The competitive stress for limited N may eliminate more bacteria from the communities than the increase of bacteria which are adapted to a lower-N environment. Furthermore, considering the fact that the optimal C/N ratio for most bacteria is about 20, it is also not surprising that the soil bacterial diversity is higher at a relatively lower C/N ratio (about 20) than that at a higher C/N ratio (about 30), as reflected in the observed negative correlation between C/N ratio and soil bacterial diversity. The C/N ratio was shown to act as a major factor for predicting the effects of organic wastes on soil bacterial communities in temperate agricultural soils (18).

The δ¹³C and δ¹⁵N in soils were selected as reflective of the quality of soil organic matter. Generally, lower values of δ¹³C and δ¹⁵N represent labile material of relatively recent origin and higher values represent more humified and amorphous organic compounds (28). Here, δ¹³C or δ¹⁵N performed poorly in predicting the variations in soil bacterial diversity, possibly because of treatment-induced changes that are divorced from the quality of the litter.

The variation of bacterial diversity and community structure along environmental gradients may reflect changes in biogeochemical cycling. It is plausible that bacterial communities themselves brought about the observed differences in soil chemical properties, because they directly mediate a wide range of ecosystem processes such as residue decomposition and nutrient cycling (13, 25, 27). Although specific bacterial groups have been linked to biogeochemical cycles in terrestrial ecosystems (26, 47, 81), elemental cycling may be affected by previously unknown organisms (24). Therefore, identifying the environmental variables shaping the distribution pattern of uncultured soil bacterial communities is a significant first step in delineating functional guilds (24, 56).

In conclusion, by combining two independent molecular techniques, i.e., an entirely sequence-based approach and DGGE, and advanced statistical analysis of data from samples taken in an intact temperate forest ecosystem, this study quantitatively partitioned the relative importance of the spatial factor, elevated CO₂, and nutrient amendment in influencing soil chemical properties and thus bacterial diversity and community structure. Our results revealed that the spatial factor was the primary determinant of soil bacterial diversity and community structure. The soil C/N ratio and total N, affected by either the spatial factor or anthropogenic disturbances, were key determinants of soil bacterial diversity. These findings have advanced our understanding of the effects of elevated CO₂ on soil bacterial communities and the effects of the spatial factor and environmental variables in shaping the distribution patterns of bacterial community diversity and structure in temperate forest soils.