Comparison of Soil Bacterial Communities Under ... - PubAg - USDA

Microb Ecol (2008) 55:293–310 DOI 10.1007/s00248-007-9276-4

ORIGINAL ARTICLE

Comparison of Soil Bacterial Communities Under Diverse Agricultural Land Management and Crop Production Practices Tiehang Wu & Dan O. Chellemi & Jim H. Graham & Kendall J. Martin & Erin N. Rosskopf

Received: 8 February 2007 / Accepted: 7 May 2007 / Published online: 7 July 2007 # Springer Science + Business Media, LLC 2007

T. Wu : D. O. Chellemi : E. N. Rosskopf US Horticulture Research Laboratory, USDA-ARS, Fort Pierce, FL 34945, USA

effects on how the relative abundance of individual amplicons were distributed (evenness) and not on the total numbers of bacterial 16S rDNA amplicons detected (richness). Similar levels of diversity were detected among all land management programs in soil samples collected after successive years of tomato (Lycopersicon esculentum) cultivation. A different trend was observed after a multivariate examination of the similarities in genetic composition among soil bacterial communities. After 3 years of land management, similarities in genetic composition of soil bacterial communities were observed in plots where disturbance was minimized (bahiagrass and weed fallow). The genetic compositions in plots managed organically were similar to each other and distinct from bacterial communities in other land management programs. After successive years of tomato cultivation and damage from two major hurricanes, only the composition of soil bacterial communities within organically managed plots continued to maintain a high degree of similarity to each other and remain distinct from other bacterial communities. This study reveals the effects of agricultural land management practices on soil bacterial community composition and diversity in a large-scale, long-term replicated study where the effect of soil type on community attributes was removed.

T. Wu : J. H. Graham Citrus Research and Education Center, University of Florida, Lake Alfred, FL 33850, USA

Introduction

Abstract The composition and structure of bacterial communities were examined in soil subjected to a range of diverse agricultural land management and crop production practices. Length heterogeneity polymerase chain reaction (LH-PCR) of bacterial DNA extracted from soil was used to generate amplicon profiles that were analyzed with univariate and multivariate statistical methods. Five land management programs were initiated in July 2000: conventional, organic, continuous removal of vegetation (disk fallow), undisturbed (weed fallow), and bahiagrass pasture (Paspalum notatum var Argentine). Similar levels in the diversity of bacterial 16S rDNA amplicons were detected in soil samples collected from organically and conventionally managed plots 3 and 4 years after initiation of land management programs, whereas significantly lower levels of diversity were observed in samples collected from bahiagrass pasture. Differences in diversity were attributed to

K. J. Martin Department of Biology, William Paterson University, Wayne, NJ 07470, USA T. Wu (*) Division of Cell Biology, Microbiology, and Molecular Biology, University of South Florida, 4202 E. Fowler Avenue SCA110, Tampa, FL 33620, USA e-mail: [email protected]

Bacteria are the most abundant and diverse group of organisms in soil, with estimates of 104–106 distinct genomes per gram of soil [29, 76]. They are critical to many of the biological, chemical, and physical processes that drive terrestrial ecosystems. Plant growth is affected directly through their activities as plant pathogens or plant growth promoters and indirectly via interactions with other soil microorganisms [5, 18, 30, 69]. Several key inorganic

294

nutrients (C, N S, P, Fe, Ni, Ag, etc.) in soil are transformed and cycled through their metabolic activities [17, 40, 47]. Soil structure is affected through their production of organic and inorganic acids, facilitating the weathering of soil minerals and formation of soil aggregates [33, 74]. By virtue of their varied contributions to soil ecosystem function, bacteria are pivotal in the production of food and fiber from agricultural cropping systems. Thus, efforts to develop more sustainable systems that seek to minimize environmental disruption while maintaining plant health are contingent upon defining the impacts of land management and crop production practices on the structure and composition of soil bacterial communities. The activity and diversity of soil bacterial communities are directly influenced by the soil environment [2, 6, 31, 50, 66, 75]. For example, a significant relationship between soil bacterial communities and soil water content was observed in a replicated field trial with winter flooding effects [6]. Bååth [2] differentiated soil bacterial communities of 16 different soils with pH ranging between 4 and 8. Schutter et al. [66] indicated that soil microbial compositions were influenced more by soil type and field properties than by farm management. Anthropogenic activities including agricultural land management practices directly and indirectly affect soil environments and thus may also alter the activity and diversity of soil bacterial communities [4, 6, 11, 20, 21, 49, 67]. The effects of agricultural land management practices on soil microbial communities have been widely studied using culture-based methods that include dilution plating and sole carbon source utilization and biochemical biomarkers. Organic amendments were shown to significantly enhance populations of soil bacteria antagonistic to plant pathogens in the rhizosphere of tomato [20]. Soil bacterial diversity significantly decreased in wheat fields after tillage and was enhanced after crop rotation with red clover green manure or field pea [49]. Winter cover crop residues increased diversity of soil microorganisms including bacteria [66]. The source of plant nutrients (synthetically derived minerals vs decomposed plant material) had a larger effect on microbial communities than land management systems (organic, low-input, or conventional) [7]. Land use history under nine land use types in two coastal valleys in California exhibited different soil bacterial community compositions [67]. The robustness of culture-based and sole carbon source utilization procedures used in the previous community studies is limited by their ability of detecting only 0.1 to 10% of the total bacterial population in soil [1, 77]. Similarly, biochemical procedures based upon phospholipid fatty acid (PLFA) or fatty acid methyl ester (FAME) profiles have several limitations. Other organisms can confound FAME profiles [34], individual species may have numerous fatty acids [7], and interpretations of individual markers depend upon a database derived mostly from

T. Wu et al.

information on isolated pure cultures of bacteria [79]. Direct comparisons PLFA and sole carbon source utilization procedures have produced conflicting results [6, 38, 80]. Recently developed polymerase chain reaction (PCR) based approaches for ribosomal DNA (rDNA) amplification have been widely used to study soil bacterial diversity and community structure [44]. These techniques include denaturing gradient gel electrophoresis (PCR-DGGE), terminal restriction fragment length polymorphism (T-RFLP), and length heterogeneity PCR (LH-PCR). Using PCR-DGGE profiles and phenotypical testing (Biolog), organic soil from pastureland (53% organic matter) exhibited a higher soil bacterial diversity than a sandy soil (10% organic matter) previously used for conventional vegetable production [57]. T-RFLP was used to distinguish bacterial diversity and community structures in soil [31, 46, 48, 75]. In an agricultural soil planted with transgenic and nontransgenic potato plants, spatial and temporal changes in bacterial community structure were observed using different T-RFLP profiles [48]. Similarly, field soil amended with copper greatly decreased bacterial diversity as detected by T-RFLP profiles [75]. Combining T-RFLP with DGGE and Biolog methods, Girvan et al. [31] compared soil bacterial communities under a number of geographically distant agricultural sites comprising two distinct soil types. Their results indicated that soil type was the key factor determining bacterial community composition in these arable soils [31]. LH-PCR is based upon natural variation in the length of bacterial ribosomal DNA and provided a better description of soil bacterial communities than T-RFLP in a recent study [54]. Omission of the restriction step decreases processing time, making it more applicable for large-scale microbial community studies [54, 64, 70]. Ritchie et al. [64] used LH-PCR and FAME to examine bacterial community composition in different soil types and crop management practices. Similar community patterns were revealed by the two methods. Soil bacterial diversity and community structures of a vineyard amended with compost were studied using automated ribosomal intergenic spacer analysis (A-RISA), a method of LH-PCR using a different region of rDNA and PLFA [65]. Although the size and structure of soil bacterial communities were altered with the addition of compost, the changes disappeared rapidly in the low-level compost treatment but persisted for months in the high-level treatment [65]. Although PCR-based molecular techniques provide great advantages for understanding the overall community structures of soil bacteria, few studies have applied the techniques to field studies replicated over a single soil taxonomic unit large enough to accommodate commercial crop production practices. This aspect is critical, as soil type is the primary determinant of soil bacterial community composition and diversity in studies involving both culture [7, 66] and nonculture-based molecular methods [31, 78].

Comparison of Soil Bacterial Communities

The combined application of univariate and nonparametric multivariate statistical methods to analyze microbial community data may further an understanding of their attributes under different agricultural land management regimes. Traditionally, univariate analysis of diversity indices and principal component analysis (PCA) of community composition have been employed to study community attributes [28, 37, 49, 54, 55, 57]. A single index, such as the Shannon–Weiner index of diversity, collapse or condense species counts from a sample into a single coefficient and may fail to discern significant attributes related to the composition of microbial communities [22]. PCA has been widely used in microbial community studies to graphically display differences in microbial community structure [28, 37, 49]. However, PCA requires biological populations to have a linear response to ecological variation [28] and be normally or log normally distributed [23] to make strong conclusions from the statistical analysis [71, 72]. Data obtained from culture-independent methods, including PCR-DGGE, TRFLP, or LH-PCR, are not normally distributed, even after log transformation [23]. Alternative multivariate methods such as Kohonen self-organizing maps, hierarchical clustering, and nonmetric multidimensional scaling (MDS) have been used to address this type of data [23, 42]. Hierarchical clustering and nonmetric MDS analysis are based upon a matrix of pairwise similarity coefficients, such as Bray– Curtis similarity coefficient. Methods based upon the ranked similarities among species are not subjected to the same set of assumptions associated with Euclidian geometry-based methods, i.e., a normal distribution of sample means. Thus, they may provide a more robust conclusion related to community structures [15]. These methods have been more commonly applied to analysis of marine communities [36, 43, 58, 61], but there are limited examples of the application of these methods to soil microbial community analysis [62, 65]. The objective of this study is to identify the long-term impacts of agricultural land management and crop production practices on the structure and composition of soil bacterial communities in a humid, subtropical region. Experiments were conducted in large, replicated, 0.16 ha field plots arranged in a randomized complete block design and were continued over a 5-year period. The influence of soil type on soil microbial communities was mitigated by employing the different land management practices across a single soil taxonomic unit. Plots were large enough to simulate commercial crop production practices. A culture-independent microbial community fingerprinting method (LH-PCR) was combined with univariate and multivariate analysis to statistically discriminate soil bacterial diversity and community structures under five diverse land management treatments before and after the initiation of tomato crop production practices. This study focused on land management and crop

295

production practices that provide chemical and nonchemical alternatives to soil fumigation with methyl bromide for the management of soilborne pests of fresh market tomato. It is part of a larger, multidisciplinary effort to develop economically feasible fresh market vegetable production systems that minimize environmental disruption.

Materials and Methods Experimental Design In July 2000, a field experiment was initiated in Fort Pierce, Florida on land previously subjected to 10 consecutive years of conventional tomato production including soil fumigation with methyl bromide/chloropicrin. The soil type was a Pineda fine sand (loamy, siliceous, hyperthermic, Arenic Glossaqualfs). Five land management treatments were arranged in a randomized complete block design with six replications. The size of each replicate plot was 0.16 ha. Treatments were (1) ‘conventional’—annual tomato production after conventional guidelines that included soil fumigation with a 62:35 formulation of 1,3-dichloropropene (1,3-D) and chloropicrin (PIC) and application of the herbicides napropamide and trifluralin; (2) ‘disk fallow’—soil maintained free of vegetation through periodic cultivation; (3) ‘weed fallow’—soil left undisturbed and weeds permitted to revegetate naturally; (4) ‘organic production’—annual cover crops of sunn hemp (Crotalaria juncea L.) and Japanese millet (Echinochloa crusgalli var. frumentacea (L.) Beauv.) combined with annual applications of broiler litter and urban plant debris (UPD); and (5) ‘bahiagrass pasture’—an improved stand of a perennial pasture grass (Paspalum notatum Flügge var. Argentine) was established and maintained. During years 1–3, only plots in the conventional land management program were cultivated to tomato. Tomato was cultivated in the south half of each replicate plot in year 4 (fall 2003) and in both plot halves in year 5 (fall 2004; Table 1). Organic production standards were used to cultivate tomato in the organic land management treatment [56]. Applications of 22,000 kg ha−1 broiler litter and 67,000 kg ha−1 UPD in the organic production system corresponded to 671 kg N ha−1, 798 kg P2O5 ha−1, and 595 kg K2O ha−1 in 2003; and 609 kg N ha−1, 698 kg P2O5 ha−1 and 517 kg K2O ha−1 in 2004. By contrast, applications of synthetic mineral fertilizer used to produce the tomato crops in the conventional, disk, and weed fallow management programs corresponded to a fertility rate of 485 kg N ha−1, 56 kg P2O5 ha−1, and 749 kg K2O ha−1 in 2003; and 525 kg N ha−1, 92 kg P2O5 ha−1 and 1012 kg K2O ha−1 in 2004. Conventional fresh market tomato production practices were applied to the weed and disk fallow land management treatments except that soil fumigation was omitted. An alternative low-input production

296

T. Wu et al.

Table 1 Chronological sequence for collection of soil samples Sample date

Relation to the stage of the experimental design

July 2003 Sep. 2003

Beginning of year 4 Before transplanting tomato (year 4) End of tomato crop (year 4) Beginning of year 5 Before transplanting tomato (year 5) End of tomato crop (year 5)

Dec. 2003 July 2004 Sep. 2004 Jan. 2005

Location in plots and additional notes

South South

Plots subjected to 3 years of different land management treatments

South South South

Plots subjected to 3 years of land treatments and 1 year of tomato production

South

North North

Plots subjected to 4 years of land management treatments

North

Land management treatments initiated July 2000

Soil samples were collected at the beginning of year four (July 2003) and five (July 2004; Table 1). In each replicate plot, soil cores from 14 evenly spaced quadrats were removed using a 2.5 cm wide×15 cm deep probe and bulked before analysis. Soil samples were also collected before transplanting tomato in year 4 (Sep. 2003) and year 5 (Sep. 2004) and again when tomatoes were harvested in year 4 (Dec. 2003) and year 5 (Jan. 2005). Samples at tomato harvest (Dec. 2003 and Jan 2005) were collected by bulking soil extracted from the root zone of four tomato plants from each plot. No effort was made to distinguish the rhizhosphere and no-rhizosphere soil, as tomato roots were well established in the soil systems of all tomato plots. DNA was extracted from the collected soil samples using a commercially available soil DNA extraction kit (UltraClean™ Soil DNA Isolation Kit, Mo Bio Laboratories, Solana Beach, CA, USA). Extracted DNA samples were preserved at −20°C until use.

1× buffer, 3 mM MgCl2, 0.5 mg ml−1 bovine serum albumen, 0.2 mM diethylnitrophenyl thiophosphate (dNTP), 0.05 units μl−1 Taq DNA polymerase (Roche, Cat. No. 12 032 953 001, Indianapolis, IN, USA), and 0.2 μM of each primer for PCR reaction. The reactions were performed with an initial denaturation step at 95°C (10 min), followed by 25 cycles of denaturation at 95°C for 30 s, annealing at 55°C for 1 min and extension at 72°C for 1 min, plus a final elongation step at 72°C for 7 min, in a reaction volume of 50 μl using a DNA Engine® Thermal Cycler (Model PTC-200, MJ Research, Waltham, MA). The primer pair 27F (5′AGAGTTTGATCMTGGCTCAG-3′), labeled with 6-FAM, and 355R (5′-GCTGCCTCCCGTAGGAGT-3′) were used to amplify a partial fragment of the 16S rDNA gene [54, 70]. Labeled 5′ end of 27F with 6-FAM was used for the discrimination of the different 16S rDNA amplicon lengths. PCR products were purified with Millipore ultrafilters (Montage PCR) and run on 2% synergel/0.7% agarose gels to quantify the DNA loading amount. About 3 fmole of PCR products were mixed with 7.5 μl deionized formamide and 0.5 μl MapMarker™ 1000 (BioVentures, Murfreesboro, TN). The mixture was held 95°C for 5 min followed by immediate chilling on ice. The sample was loaded on Applied Biosystems 3730xl DNA Analyzer (Applied Biosystems, Foster City, CA) discriminating 16S rDNA amplicon lengths. Labeled amplicons were separated by capillary electrophoresis and detected by laser-induced fluorescence with the 3730xl DNA Analyzer after the spectral calibration with Matrix Standard DS-30 (Applied Biosystems).

LH-PCR

Cloning and Sequencing

Amplicon length heterogeneity PCR (LH-PCR) was used to amplify the template bacteria 16S rDNA following procedures outlined by Mills et al. [54]. PCR reaction mixtures included

Soil samples collected from Sep. 2003 under five different above land managements were used to generate sequences of different amplicons in LH-PCR. PCR products of bacterial

system, including strip-tillage for tomato cultivation, was adhered to in the bahiagrass pasture treatment [12]. The fertility rates in the bahiagrass pasture treatment were 559 kg N ha−1, 56 kg P2O5 ha−1, and 883 kg K2O ha−1 in 2003; and 552 kg N ha−1, 50 kg P2O5 ha−1, and 1026 kg K2O ha−1 in 2004. After land was prepared for the cultivation of tomato in year 5, two major hurricanes (Francis and Jeanne) struck the experimental farm site on September 5 and 25, 2004, causing significant damage to the farm and surrounding communities. Soil Sample Collection and DNA Extraction


16S rDNA using above correspondent nonfluorescent primers were purified using the QIAquick® PCR Purification Kit (Qiagen, Valencia, CA). About 300 ng purified PCR products were cloned using the Topo TA cloning kit (Invitrogen, Carlsbad, CA) as per the manufacture’s instructions and transferred into E. coli Top 10 cells (Invitrogen, Carlsbad, CA) by heat shock (42°C for 30 s). After incubating in Luria–Bertani kanamycin broth at 37°C for 1 h and plating on Luria–Bertani agar plate containing kanamycin (50 μg ml−1), only transformed cells grew on the plates. Colonies were selected with sterile tooth pickers and grown in 96-well plates in Luria–Bertani kanamycin broth on an orbital shaker incubator at 215 rpm for 24 h. Cell lyses and DNA extraction of cloned products were performed on Bio Robot 9600 (Qiagen) to obtain DNA sequence templates. Sequencing reactions were performed using BigDye Terminator v 3.1 cycle sequencing kit as described by the manufacturer, using the primers of M13 forward and reserve, respectively. Sequencing was performed on ABI Prism® 3730xl Genetic Analyzer, and sequence data were analyzed using Sequencher version 3.0 (Gene Codes, Ann Arbor, MI). The sequences obtained from clones with 97% sequence similarity were assigned into the same sequences and were subjected to basic local alignment search tool (BLAST) searches on NCBI to obtain the closest matches to bacterial sequences in the database. All of the nucleotide sequences generated were submitted to NCBI GenBank and assigned accession numbers from EF043537 to EF043566. The obtained clones were also used as template for LH-PCR to confirm the size of the fragments by LH-PCR detections. Collation of Data The resulting electropherogram profiles were analyzed using ABI Prism® GeneMapper™ v. 3.0 software. Electropherogram peaks represented amplicons of different lengths, and the areas under the peaks represented the relative proportion of each amplicon. The relative abundance of each DNA amplicon was determined by calculating the ratio between the peak area at a specific amplicon length and the total peak area of all DNA amplicon lengths in the sample. To remove the error introduced by the collection and analysis software, only the amplicons with a relative abundance greater than 0.01 were included in additional analyses. Peaks separated by 1 base pair (bp) or less were lumped together and considered as one amplicon length for all statistical analyses. Representation of Soil Bacterial Communities Graphical descriptions of soil bacterial communities subjected to the different land management and tomato production treatments were performed using several nonmetric multivariate procedures outlined by Clarke [13].

297

Pairwise Bray–Curtis similarity matrices for samples were constructed using relative abundance values for each DNA amplicon length [9]. A square root transformation was applied to the data before construction of the similarity matrices. Hierarchical agglomerative clustering with groupaverage linking was used to classify the resulting triangular matrix of similarities between pairs of samples into groups [16]. Representation of communities was accomplished using dendrograms to link samples into hierarchical groups. Inter-relationships between samples were mapped in ordination plots using nonmetric MDS [45], where the position of each sample is determined by its distance from all other points in the analysis and the “stress” represents a measure of the goodness of fit of the plot. A stress value less than 0.2 indicates a useful two-dimensional picture [13, 61]. Discrimination of Land Management and Tomato Production Treatments Univariate indices of diversity were used to characterize community structure under the different land management and tomato production practices. Genetic richness (S) was expressed as the total number of amplicons identified. Diversity was calculated using the Shannon–Weiner (Weaver) P index as follows: diversity (H′): H 0 ¼ ðpi Þðloge pi Þ; where pi is the proportion of an individual peak area relative to the sum of all peak areas detected in a sample. Evenness, or the equitability of the observed amplicon lengths, was calculated by Pielou’s evenness index as follow: J′=H′/Log (S). Stastica™ (Stat Soft, Tulsa, OK) was used for univariate analyses of diversity and richness indices. Data collected in 2003 was analyzed as a randomized complete block design. Data collected in 2004 and 2005 were analyzed as a split plot design with land management practice as the main plot and location within the plot (north vs south) as the split plot. Fisher’s protected LSD was used to separate the means. Statistical discrimination of land management and tomato production practices using nonparametric multivariate methods was performed with the analysis of similarities (ANOSIM) procedure [14]. Using the ranked order of similarities in the original Bray–Curtis matrices, a test statistic (R) was computed as: R¼

rtrt rrep 1=2M

Where rtrt is defined as the average of rank similarities arising from all pairs of replicates between treatments, rrep is the average of all rank similarities among replicates within treatments and M=n(n−1)/2 where n is the total number of samples under consideration. R=1 when replicate samples within a treatment are more similar to each other than any samples across treatments, i.e., treatments are dissimilar.

298

R=0 when similarities among samples across treatments is equivalent to similarities among samples within a treatment, i.e., no treatment effects. To test the significance of the experimental R, the same statistic was recomputed 462 times using random permutations of the sample labels and the number of simulated R’s that exceeded the experimental R was recorded. The 16S rDNA amplicons primarily responsible for distinguishing treatments was determined using the similarity percentages (SIMPER) procedure outlined by Clarke and Warwick [15]. All nonparametric multivariate analyses procedures, including calculation of Bray–Curtis similarity matrices, cluster analysis, MDS, SIMPER analysis, and ANOSIM, were conducted using the Primer-E statistical software (Primer-E, Plymouth Marine Laboratory, UK).

Results Representation of Soil Bacterial Communities The July 2003 samples were collected after the completion of 3 years under the different land management treatment (Table 1). With regard to their respective treatments, organic and weed fallow plots shared >80% similarity (Fig. 1a). Five of the six plots in the bahiagrass treatment also shared >80% similarity. The composition of 16S rDNA amplicons in the five bahiagrass plots was closely associated with amplicons in the weed fallow plots, whereas the organic plots remained uniquely grouped (Fig. 1b). The Sep. 2003 sample was collected after tomato crop production practices were initiated but before transplanting tomato seedlings. In general, associations among communities appeared to diverge with less than half the plots having a similarity level >80% (Fig. 1c). Except for one conventional plot, bacterial communities within each treatment shared a higher degree of similarity with each other than with communities in other treatments (Fig. 1c). Nonmetric MDS indicated several distinct separations, with only communities in the organic treatment maintaining a close association among all six replicate plots (Fig. 1d). The Dec. 2003 samples were collected upon completion of the tomato crop. Except for communities in the organically managed plots, disk fallow plot 3, and conventional plot 5, bacterial communities shared a higher degree of similarity with each other (Fig. 1e). Communities in the organic plots remained more closely associated with each other and distinct from communities in the other land management programs (Fig. 1f). The July 2004 north-side samples were collected after completion of 4 years under their respective land management programs (Table 1). Similarity groupings of the bacterial communities were not as readily pronounced as groupings from the July 2003 samples. Only four of six organic plots

T. Wu et al.

shared >80% similarity (Fig. 2a). In all other treatments, no more than two plots shared >80% similarity. Mapping of the ranked similarities revealed a high degree of association among plots within the conventional and disk fallow treatments, respectively (Fig. 2b). The Sep. 2004 north-side sample was collected after tomato crop production practices were initiated but before transplanting tomato seedlings. As observed in the Sep. 2003 sample, associations among communities appeared to diverge except in the organic and bahiagrass plots (Fig. 2c, d). Upon completion of the crop in the January 2005 sample, a higher degree of association was observed among communities within organically managed plots. Communities in five of the six bahiagrass plots also exhibited a higher degree of association with each other (Fig. 2e, f). The July 2004 south-side samples were collected after completion of 3 years under the different land management treatments and 1 year of tomato production (Table 1). Only bacterial communities subjected to organic land management practices maintained a high degree of association with each other (Fig. 3a) and were readily distinguishable from communities subjected to other land management practices (Fig. 3b). The Sep. 2004 south-side sample continued the trend in divergence of community associations based upon the composition of 16S rDNA amplicons (Fig. 3c, d). Similarity rankings among communities in the organic plots revealed two distinct groups. Completion of 2 years tomato production after 3 years of land management (Jan 2005 southside sample) revealed a divergence in community composition in all land management programs except organic, where communities remained uniquely clustered based upon their Bray–Curtis similarity coefficients (Fig. 3e, f). Discrimination of Land Management and Tomato Production Treatments The number of 16S rDNA amplicons amplified with primer pair 27F-355R, an indicator of genetic richness, ranged from 15.0 in the weed fallow December 2003 sample to 20.5 in the September 2003 organic sample. The number of amplicons detected was not significantly impacted by the five agricultural land management programs (July 2003 and July 2004 north samples, Table 2). After the initiation of tomato production practices but before transplanting tomato, amplicon numbers were significantly lower in the weed and disk fallow treatment when compared to the organic and conventional treatments in 2003. In 2004, only the organic system maintained a higher number of amplicon lengths before transplanting tomato. At the end of tomato production in Jan. 2005, an interaction between treatment and location was detected. Plots on the north side of the disk and weed fallow treatments (subjected to 4 years of treatment) had significantly fewer amplicons detected than plots in the south


299 Organic 5 Organic 4 Organic 3 Organic 2 Organic 6 Organic 1 Conventional 6 Conventional 4 Conventional 5 Conventional 1 Conventional 2 Conventional 3 Disk Fallow 1 Disk Fallow 4 Disk Fallow 3 Disk Fallow 2 Disk Fallow 5 Disk Fallow 6 Bahia 5 Bahia 6 Bahia 1 Bahia 2 Bahia 4 Bahia 3 Weed Fallow 6 Weed Fallow 2 Weed Fallow 5 Weed Fallow 3 Weed Fallow 1 Weed Fallow 4

a

60

70

80 Bray-Curtis Similarity (%)

90

60

70


90

60

70


90

2D Stress: 0.16

d Prior to transplanting tomato (year 4)

100

Organic 6 Organic 3 Organic 5 Organic 4 Organic 1 Organic 2 Weed fallow 1 Bahia 2 Weed fallow 6 Bahia 4 Weed fallow 5 Weed fallow 3 Bahia 5 Bahia 1 Weed fallow 4 Weed fallow 2 Bahia 6 Bahia 3 Disk fallow 5 Disk fallow 2 Conventional 3 Conventional 2 Conventional 4 Conventional 1 Disk fallow 6 Disk fallow 1 Disk fallow 4 Conventional 6 Disk fallow 3 Conventional 5

e

End of year 3

100

Weed fallow 4 Weed fallow 1 Weed fallow 2 Weed fallow 3 Weed fallow 6 Weed fallow 5 Disk fallow 4 Disk fallow 1 Disk fallow 6 Disk fallow 2 Disk fallow 5 Disk fallow 3 Conventional 6 Conventional 4 Conventional 2 Conventional 5 Conventional 3 Conventional 1 Organic 4 Organic 3 Organic 6 Organic 2 Organic 5 Organic 1 Bahia 6 Bahia 3 Bahia 1 Bahia 5 Bahia 2 Bahia 4

c

2D Stress: 0.18

b

2D Stress: 0.12

f

End of tomato crop (year 4)

100

Figure 1 Dendrograms (a, c, e) and MDS ordination (b, d, f) of 16S rDNA by primer 27F-355R from 2003 samples. Samples collected in July (a, b), Sep. (c, d) and Dec. (e, f). Vertical line in a, c, and e and

circle in b, d, and f represent a similarity level of 71%. Symbols in MDS ordinations are as follow: triangle bahiagrass; inverted triangle conventional; square disk fallow; diamond organic; circle weed fallow

300

T. Wu et al. Organic N 4 Organic N 3 Organic N 2 Organic N 1 Organic N 5 Organic N 6 Weed fallow N 3 Conventional N 5 Conventional N 2 Conventional N 6 Conventional N 3 Conventional N 4 Conventional N 1 Weed fallow N 5 Bahia N 4 Weed fallow N 4 Weed fallow N 2 Weed fallow N 6 Weed fallow N 1 Disk fallow N 6 Disk fallow N 5 Disk fallow N 3 Disk fallow N 1 Disk fallow N 4 Disk fallow N 2 Bahia N 3 Bahia N 2 Bahia N 5 Bahia N 1 Bahia N 6

a

50

60

70

80

90

2D Stress: 0.18

b End of year 4 North Side

100

Bray-Curtis Similarity (%)

Conventional N 6 Disk Fallow N 4 Conventional N 5 Conventional N 1 Conventional N 3 Conventional N 4 Weed Fallow N 4 Weed Fallow N 2 Weed Fallow N 6 Weed Fallow N 3 Weed Fallow N 1 Weed Fallow N 5 Disk Fallow N 3 Conventional N 2 Disk Fallow N 6 Disk Fallow N 2 Disk Fallow N 1 Disk Fallow N 5 Organic N 6 Organic N 2 Organic N 3 Organic N 5 Organic N 4 Organic N 1 Bahia N 3 Bahia N 2 Bahia N 6 Bahia N 4 Bahia N 5 Bahia N 1

c

50

60

70 80 Bray-Curtis Similarity (%)

90

50

60


90

Prior to transplanting tomato (year 5) North Side

f

End of tomato crop (year 5) North Side

100

Disk Fallow 6 N Disk Fallow 4 N Disk Fallow 5 N Disk Fallow 2 N Conventional 2 N Disk Fallow 1 N Conventional 4 N Conventional 6 N Conventional 5 N Weed Fallow 6 N Weed Fallow 3 N Weed Fallow 5 N Weed Fallow 2 N Bahia 4 N Weed Fallow 1 N Disk Fallow 3 N Conventional 3 N Conventional 1 N Weed Fallow 4 N Bahia 6 N Bahia 5 N Bahia 2 N Bahia 1 N Bahia 3 N Organic 5 N Organic 3 N Organic 6 N Organic 2 N Organic 4 N Organic 1 N

e

2D Stress: 0.14

d

2D Stress: 0.12

100

Figure 2 Dendrograms (a, c, e) and MDS ordination (b, d, f) of 16S rDNA by primer 27F-355R from 2004/2005 samples collected on the north side of plots. Samples collected in July 2004 (a, b), Sep. 2004 (c, d) and Jan. 2005 (e, f). Vertical line in a, c, and e and circle

in b, d, and f represent a similarity level of 71%. Symbols in MDS ordinations are as follow: triangle bahiagrass; inverted triangle conventional; square disk fallow; diamond organic; circle weed fallow


301 Disk fallow S 5 Disk fallow S 3 Conventional S 2 Disk fallow S 4 Conventional S 3 Conventional S 1 Disk fallow S 1 Conventional S 6 Conventional S 4 Disk fallow S 2 Conventional S 5 Weed fallow S 6 Weed fallow S 3 Weed fallow S 2 Weed fallow S 1 Weed fallow S 5 Weed fallow S 4 Disk fallow S 6 Bahia S 4 Bahia S 1 Bahia S 5 Bahia S 2 Bahia S 6 Bahia S 3 Organic S 3 Organic S 1 Organic S 2 Organic S 6 Organic S 5 Organic S 4

a

50

60


90

50

60


90

50

60


90

d

Prior to transplanting tomato (year 5) South Side

2D Stress: 0.16

100

Organic 4 S Organic 3 S Organic 2 S Organic 1 S Organic 6 S Organic 5 S Weed Fallow 4 S Weed Fallow 2 S Weed Fallow 1 S Disk Fallow 3 S Disk Fallow 6 S Weed Fallow 6 S Weed Fallow 3 S Weed Fallow 5 S Disk Fallow 2 S Disk Fallow 5 S Conventional 2 S Disk Fallow 4 S Disk Fallow 1 S Conventional 4 S Conventional 3 S Conventional 5 S Conventional 1 S Bahia 5 S Bahia 2 S Bahia 4 S Bahia 1 S Conventional 6 S Bahia 6 S Bahia 3 S

e

End of year 4 South Side

100

Bahia S 6 Bahia S 3 Bahia S 2 Bahia S 1 Bahia S 4 Bahia S 5 Weed Fallow S 6 Weed Fallow S 3 Weed Fallow S 5 Weed Fallow S 4 Weed Fallow S 2 Disk Fallow S 3 Disk Fallow S 4 Disk Fallow S 2 Disk Fallow S 1 Disk Fallow S 5 Conventional S 4 Conventional S 2 Conventional S 3 Conventional S 6 Conventional S 5 Conventional S 1 Weed Fallow S 1 Disk Fallow S 6 Organic S 5 Organic S 1 Organic S 6 Organic S 3 Organic S 2 Organic S 4

c

2D Stress: 0.21

b

2D Stress: 0.17

f End of tomato crop (year 5) South Side

100

Figure 3 Dendrograms (a, c, e) and MDS ordination (b, d, f) of 16S rDNA by primer 27F-355R from 2004/2005 samples collected on the south side of plots. Samples collected in July 2004 (a, b), Sep. 2004 (c, d) and Jan. 2005 (e, f). Vertical line in a, c, and e and circle

in b, d, and f represent a similarity level of 71%. Symbols in MDS ordinations are as follow: triangle bahiagrass; inverted triangle conventional; square disk fallow; diamond organic; circle weed fallow

302

T. Wu et al.

Table 2 Genetic richness (S), Shannon–Weiner diversity (H′), and Pielou’s evenness (J′) of bacterial 16S rDNA amplicons by primers 27F-355R in soil subjected to various land management treatments Date

Genetic richness July 2003 Sep. 2003 Dec. 2003 July 2004

Location

(S) South South South North South Mean Sep. 2004 North South Mean Jan. 2005 North South Mean Shannon–Weiner diversity (H′) July 2003 South Sep. 2003 South Dec. 2003 South July 2004 North South Mean Sep. 2004 North South Mean Jan. 2005 North South Mean Pielou’s evenness index (J′) July 2003 South Sep. 2003 South Dec. 2003 South July 2004 North South Mean Sep. 2004 North South Mean Jan. 2005 North South Mean

Land management treatments Bahiagrass

Conventional

Disk fallow

Organic

Weed fallow

17.5±0.7A 19.2±1.1AB 15.2±0.3A 16.2±0.9 18.3±0.8 17.3±0.6A 16.7±0.8 17.2±0.9 16.9±0.6B 17.3±0.6a 16.7±0.8abc 17.0±0.5

18.0±0.7A 20.3±0.5A 15.8±0.6A 18.5±0.4 17.3±0.8 17.9±0.5A 18.2±0.7 18.3±0.6 18.3±0.4AB 15.7±0.3bcd 16.2±0.5abcd 15.9±0.3

16.2±0.9A 17.5±0.9BC 16.5±1.0A 17.3±0.8 19.0±0.5 18.2±0.5A 17.3±0.7 17.2±1.2 17.3±0.7B 15.2±0.8d 16.7±0.9abc 15.9±0.6

17.8±0.5A 20.5±0.8A 17.2±0.4A 18.0±0.6 18.5±1.0 18.3±0.6A 19.5±0.6 18.7±0.7 19.1±0.5A 16.8±0.6ab 15.8±0.7bcd 16.3±0.5

18.7±0.6A 16.8±0.6C 15.0±0.4A 16.3±0.8 17.5±0.3 16.9±0.4A 16.3±0.6 17.0±0.5 16.7±0.4B 15.3±1.0cd 16.8±0.4ab 16.1±0.5

2.24±0.03C 2.41±0.07B 2.31±0.04B 2.18±0.03 2.31±0.09 2.24±0.05B 1.99±0.09f 2.12±0.09e 2.05±0.07 2.33±0.05 2.38±0.03 2.36±0.03A

2.59±0.04A 2.64±0.03A 2.41±0.09AB 2.54±0.02 2.41±0.09 2.48±0.05A 2.56±0.06b 2.58±0.05b 2.57±0.04 2.40±0.06 2.38±0.09 2.38±0.05A

2.39±0.07B 2.38±0.07B 2.42±0.06AB 2.40±0.05 2.46±0.07 2.43±0.04A 2.49±0.04bc 2.42±0.06cd 2.46±0.04 2.36±0.06 2.51±0.07 2.43±0.05A

2.59±0.02A 2.71±0.04A 2.53±0.04A 2.52±0.03 2.53±0.05 2.53±0.03A 2.71±0.04a 2.59±0.04ab 2.65±0.03 2.39±0.04 2.39±0.06 2.39±0.04A

2.55±0.05A 2.34±0.04B 2.29±0.05B 2.25±0.08 2.28±0.05 2.27±0.05B 2.12±0.04e 2.34±0.05d 2.23±0.05 2.29±0.08 2.48±0.04 2.39±0.05A

0.79±0.01C 0.82±0.01B 0.85±0.01A 0.79±0.01 0.79±0.02 0.79±0.01C 0.71±0.02f 0.75±0.03e 0.73±0.02 0.82±0.02 0.85±0.01 0.84±0.01A

0.89±0.01A 0.88±0.01A 0.87±0.03A 0.87±0.01 0.85±0.03 0.86±0.01AB 0.88±0.01abc 0.89±0.01ab 0.88±0.01 0.87±0.02 0.85±0.03 0.86±0.02A

0.86±0.01B 0.83±0.01B 0.87±0.01A 0.84±0.01 0.83±0.02 0.84±0.01B 0.87±0.01bc 0.86±0.01cd 0.87±0.01 0.87±0.01 0.89±0.01 0.88±0.01A

0.90±0.01A 0.89±0.01A 0.89±0.01A 0.87±0.01 0.87±0.00 0.87±0.01A 0.91±0.01a 0.88±0.01abc 0.90±0.01 0.85±0.01 0.87±0.02 0.86±0.01A

0.87±0.01B 0.83±0.01B 0.85±0.01A 0.81±0.02 0.80±0.02 0.80±0.01C 0.75±0.01e 0.83±0.02d 0.79±0.01 0.84±0.01 0.88±0.01 0.86±0.01A

Mean

17.3±0.3A 18.1±0.3A 17.6±0.4A 17.7±0.4A 16.07±0.32 16.43±0.29

2.38±0.03A 2.40±0.04A 2.37±0.06 2.41±0.04 2.36±0.03B 2.42±0.03A

0.84±0.01A 0.83±0.01A 0.82±0.02 0.84±0.01 0.85±0.01A 0.87±0.01A

Means ± SE are reported. Means for main effects (land management treatment or location) followed by different capital letters are significantly different at p=0.05. Means for the interaction between land management treatments and location followed by different small letters are significant at p=0.05.

side (subjected to 3 years followed by 1 year of tomato production). Plots on the north side of the disk and weed fallow treatment also had fewer amplicons detected than plots in the bahiagrass and organic treatment. The diversity of bacterial communities, as indicated by the Shannon–Weiner index ranged from 1.99 in the bahiagrass September 2004 north sample to 2.71 in the organic September 2004 north sample (Table 2). After 3 years (July

2003), significantly lower diversity was detected in the bahiagrass and disk fallow treatments when compared to the conventional, organic, and weed fallow treatments. Diversity in the organic treatment remained higher than in the bahiagrass and disk fallow treatment after the production of tomato (Dec. 2003 sample). After 4 years, diversity was lower in the bahiagrass and weed fallow treatments when compared to the conventional, organic, and disk fallow


303

Table 3 Analysis of similarity (ANOSIM) on 16S rDNA amplicons by the global and pairwise test Pairwise R Statistic Pairwise tests

July 03

Sep. 03

Dec. 03

July 04 N

Sep. 04 N

Jan. 05 N

July 04 S

Sep. 04 S

Jan. 05 S

Bahiagrass, conventional Bahiagrass, disk fallow Bahiagrass, organic Bahiagrass, weed fallow Conventional, disk fallow Conventional, organic Conventional, weed fallow Disk fallow, organic Disk fallow, weed fallow Organic, weed fallow Global R P (%)

0.96 (1)a 0.92 (1) 1 (1) 0.79 (1) 0.75 (1) 0.88 (1) 0.80 (1) 0.90 (1) 0.99 (1) 1 (1) 0.88 0.1

0.97 (1) 0.94 (1) 1 (1) 0.79 (1) 0.65 (1) 0.83 (1) 0.81 (1) 0.94 (1) 0.74 (1) 0.96 (1) 0.82 0.1

0.76 0.64 0.89 0.33 0.34 0.94 0.71 0.99 0.53 0.80 0.63 0.1

0.75 0.82 0.80 0.54 0.83 0.63 0.82 0.95 0.92 0.75 0.74 0.1

0.99 (1) 0.99 (1) 1 (1) 0.99 (1) 0.38 (2) 0.72 (1) 0.58 (1) 0.69 (1) 0.60 (1) 0.85 (1) 0.78 0.1

0.80 0.78 0.99 0.38 0.09 0.99 0.44 0.96 0.36 0.75 0.62 0.1

0.67 0.69 0.94 0.48 0.38 0.71 0.26 0.74 0.40 0.83 0.56 0.1

0.91 0.70 0.99 0.82 0.48 0.85 0.68 0.79 0.41 0.80 0.80 0.1

0.51 0.54 0.80 0.55 0.52 0.91 0.68 0.85 0.33 0.82 0.60 0.1

(1) (1) (1) (6) (1) (1) (1) (1) (1) (1)

(1) (1) (1) (2) (1) (1) (1) (1) (1) (1)

(1) (2) (1) (7) (100) (1) (4) (1) (9) (1)

(1) (1) (1) (1) (4) (1) (4) (1) (1) (1)

(1) (1) (1) (1) (1) (1) (1) (1) (1) (1)

(2) (3) (1) (1) (1) (1) (1) (1) (2) (1)

a

Numbers in parenthesis represent the number of R values from simulated permutations that exceed the experimental R (462 total permutations tested).

treatments. An interaction between land management treatment and location was observed after tomato crop production practices were initiated in Sep. 2004. In the north side, where the land management practices had not been interrupted for four consecutive years, diversity was significantly higher in the organic treatment when compared to the other treatments. Diversity also increased in the south side of the bahiagrass and weed fallow plots when compared to the north side. At the end of tomato production of Jan. 2005, no difference was observed in diversity for different land management treatments; however, significantly lower diversity existed in the north side than in south side (Table 2). Agricultural land management practices also significantly affected the evenness of the distribution of amplicon lengths (Table 2). Pielou’s evenness index ranged from 0.71 in the bahiagrass September 2004 north sample to 0.91 in the organic September 2004 north sample. After 3 years under the different land management treatments (July 2003), the distribution of amplicon lengths was more uniform in the organic and conventional treatments. Bacterial communities in the bahiagrass treatment had the most uneven distribution of 16S rDNA amplicon fragments. Upon completion of a tomato crop, the evenness index was similar in all five treatments (Dec. 2003). Similar trends occurred in the 2004 and 2005 samples except that an interaction between land management treatment and location was observed in samples collected after initiation of tomato crop production practices but before the planting of tomato. The global ANOSIM test indicated significant differences among soil bacterial communities at all sample dates. The highest global R value (0.88) was obtained in the July 2003 sample date, indicating that bacterial communities within replicated plots for each treatment were more similar to each

other than communities from other treatments. R values remained high after initiation of tomato production in the Sep. 2003 and Sep. 2004 samples but declined at the end of the crop cycle, indicating that the land management effects on community composition was negated after the cultivation of tomato. The global R value was also lower in the July 2004 south sample, which indicated that the effect carried over to the next season. Pairwise comparison between individual treatments indicated a high degree of similarity among communities in some treatments. For example, comparisons between the conventional and disk fallow treatment revealed an R value