Spatial Changes in the Bacterial Community Structure along a Vertical Oxygen Gradient in Flooded Paddy Soil Cores

Abstract

Molecular ecology techniques were applied to assess changes in the bacterial community structure along a vertical oxygen gradient in flooded paddy soil cores. Microsensor measurements showed that oxygen was depleted from 140 μM at the floodwater/soil interface to nondetectable amounts at a depth of approximately 2.0 mm and below. Bacterial 16S rRNA gene (rDNA)-based community fingerprint patterns were obtained from 200-μm-thick soil slices of both the oxic and anoxic zones by using the T-RFLP (terminal restriction fragment length polymorphism) technique. The fingerprints revealed a tremendous shift in the community patterns in correlation to the oxygen depletion measured with depth. 16S rDNA clone sequences recovered from the oxic or anoxic zone directly corresponded to those terminal restriction fragments which were highly characteristic of the respective zone. Comparative sequence analysis of these clones identified members of the α and β subclasses of Proteobacteria as the abundant populations in the oxic zone. In contrast, members of clostridial cluster I were determined to be the predominant bacterial group in the oxygen-depleted soil. The extraction of total RNA followed by reverse transcription-PCR of the bacterial 16S rRNA and T-RFLP analysis resulted for both oxic and anoxic zones of flooded soil cores in community fingerprint patterns similar to those obtained by the rDNA-based analysis. This finding suggests that the microbial groups detected on the rDNA level are the metabolically active populations within the oxic and anoxic soil slices examined.

Sediments and wetland soils are mostly characterized by an oxic surface layer and a redox stratification of the oxygen-depleted zone (56). In principle, the stratification follows the thermodynamic theory and thus creates distinct niches in which the different redox processes often react at separate localities (27, 48). Despite their different dimensions, the same principles, i.e., spatial stratification of the electron donors and acceptors, were reported for stratified water columns (34), for biofilms and microbial mats (10, 26, 35, 39), for freshwater and marine sediments (21, 24, 27, 48, 51), and for wetland soils (9, 15, 36).

Microsensors enable the measurement of redox stratifications including their diurnal and seasonal fluctuations and thereby the activity of distinct physiological groups in such gradient systems. The most important factor which determines the development of naturally occurring redox gradient systems is the presence or absence of oxygen. The steepness of the oxygen gradient depends on the amount of bioavailable organic matter in the surface layer and, as a consequence, on the biological oxygen demand (37, 40). This relationship becomes most obvious in eutrophic habitats, such as biofilms, where oxygen is depleted within 1 mm beneath the surface (26, 35, 39). A major factor which may influence the steepness of the oxygen gradient is the production of reduced compounds, such as ammonia, ferrous iron, sulfide, and methane, within the anoxic zone. At the oxic/anoxic transition zone, the reduced compounds may be reoxidized and therefore influence the depth of the oxygen penetration (27, 51).

Research on the abundance and spatial distribution of defined microbial groups within such gradient systems has focused mainly on sulfate-reducing bacteria because of their biogeochemical importance in marine sediments. This body of work includes cultivation studies (4, 23, 46), but these may have led to an underestimation of the naturally occurring abundance of sulfate reducers. The latter assumption, as concluded in one of these studies (23) from the numbers of viable cells determined in relation to the respiration rates measured in situ, is consistent with the more general view that cultivation techniques are inadequate to describe complex microbial communities (2, 29). Thus, cultivation-independent rRNA-based techniques, partly in combination with microsensor measurements (11, 34, 35, 43), were applied to achieve a more objective view of the spatial and temporal distribution of sulfate reducers in redox gradient systems. These studies included fluorescence in situ hybridization (34, 35), quantitative slot blot hybridization (11, 43), and denaturing gradient gel electrophoresis (DGGE) (45, 52, 53). The spatial distribution of methanogens in relation to chemical gradients was investigated in a freshwater sediment from Lake Rotsee, Switzerland (57).

Our research aimed at an assessment of the impact of oxygen depletion on the bacterial community structure at floodwater/soil interfaces. Flooded paddy soil cores under defined experimental conditions were used as a model system in this study. The combined use of molecular ecology techniques and oxygen microsensors clearly showed that changes in the bacterial community structure directly corresponded to the oxygen depletion measured within the upper 2.0-mm soil layer. The comparative sequence analysis of bacterial 16S rRNA genes (rDNA) and reverse transcription-PCR (RT-PCR) of total 16S rRNA followed by terminal restriction fragment length polymorphism (T-RFLP) analysis (30) permitted the characterization of the abundant and metabolically active populations in both the oxic and anoxic zones of flooded soil cores.

MATERIALS AND METHODS

Flooded soil cores.

Soil was taken from drained paddy fields of the Istituto Sperimentale della Risicoltura (Vercelli, Italy) in the spring of 1997. The soil was air dried and stored as dry lumps at room temperature. The soil characteristics were described previously (22). The soil was sieved immediately prior to its use, which resulted in a homogeneous fraction of soil particles with a diameter of less than 2 mm. The soil was mixed with deionized water in a ratio of 2:1 (wt/vol) and filled into 60-ml reaction vessels (Sarstedt, Nümbrecht, Germany). In total, seven unplanted soil cores were incubated with a 1-cm floodwater layer at 30°C for 7 days in the dark. After incubation, six soil cores were immediately frozen at −80°C and used for the molecular analysis. The upper 1.0 cm of each core was cut into 200-μm-thick slices in a precooled microtome (Microm, Walldorf, Germany). The slices were transferred into 1.5-ml reaction tubes and stored at −80°C until use (see below). The remaining soil core was used to determine the oxygen depth profile which had developed after 7 days of incubation.

Microsensor measurements.

Oxygen profiles were measured with Clark-type microelectrodes equipped with an additional guard cathode. The microelectrodes, manufactured according to the description by Revsbech (38), were obtained from Mas Com (Bremen, Germany). The tip diameter was between 50 and 100 μm. Measurements were carried out as described by Frenzel et al. (15).

Extraction of total DNA.

Total community DNA was extracted from all individual slices of the upper 3.0 mm of three flooded soil cores and from additional slices taken at different depths within the anoxic zone. The procedure applied to extract total DNA was a modified version of a previously reported protocol (18). For each DNA extraction, a single soil slice (approximately 250 mg [wet weight] per soil slice) was mixed with 500 μl of a sodium phosphate buffer (0.10 M, pH 8.0) and 125 μl of a 10% (wt/vol) solution of sodium dodecyl sulfate. After incubation for 10 min at 60°C, 0.5 g of glass beads (0.17- to 0.18-mm diameter) was added, and the suspension was shaken for 1 min at maximum speed in a bead beater (Dismembrator-S; B. Braun Biotech, Melsungen, Germany). Soil particles, glass beads, and cell debris were pelleted at 13,000 × g for 10 min at 4°C, and the supernatant was extracted three times with chloroform-isoamyl alcohol (24:1 [vol/vol]). The DNA in the aqueous phase was precipitated with 2.5 volumes of ethanol. For further purification, the DNA was treated with cesium chloride as described previously (47). Finally, the DNA was resuspended in 50 μl of TE buffer (10 mM Tris-HCl, 1 mM EDTA [pH 8.0]).

Extraction of total RNA.

Total RNA was extracted from soil slices taken from the floodwater/soil interface (oxic zone) and from a depth of approximately 1 cm (anoxic zone). To obtain sufficient amounts of total RNA for further analysis, the 200-μm slices of three flooded soil cores were pooled. Each of the composite samples (≈1 g of soil [wet weight]) was placed into a 2.0-ml reaction tube containing 1 g of glass beads (0.17- to 0.18-mm diameter) and 700 μl of precooled TPM buffer (50 mM Tris-HCl [pH 7.5], 1.7% [wt/vol] polyvinylpyrrolidone, 10 mM MgCl₂ [12]). The suspension was shaken for 60 s at maximum speed in a bead beater (Dismembrator-S; B. Braun Biotech). Glass beads, soil particles, and cell debris were pelleted by centrifugation for 10 min at 4°C, and the supernatant was transferred to a new reaction tube. The pellet was suspended in 700 μl of a phenol-based lysis buffer (50 mM Tris-HCl [pH 7.5], 10 mM EDTA, 1% [wt/vol] sodium dodecyl sulfate, 6% [vol/vol] water-saturated phenol), followed by a second round of bead beating (see above). After centrifugation at 13,000 × g, the supernatants of the two bead beating treatments were pooled and were extracted with water-saturated phenol, phenol-chloroform-isoamyl alcohol (25:24:1 [vol/vol/vol]), and finally chloroform-isoamyl alcohol (24:1 [vol/vol]). The total nucleic acids were precipitated from the aqueous phase with 3 volumes of ethanol and, after being dried, were resuspended in 100 μl of diethyl pyrocarbonate (DEPC)-pretreated water. If necessary, the nucleic acids were further purified by a Sephadex G-75 column filtration as described by Moran et al. (32). For the removal of coextracted DNA, 1 volume of TMC buffer (10 mM Tris-HCl [pH 7.5], 5 mM MgCl₂, 0.1 mM CsCl₂ [12]) and 5 U of RNase-free DNase (Promega, Madison, Wis.) were added. Incubation was at 37°C for 30 min, and the reaction was stopped by extraction with 1 volume of chloroform. The total RNA was precipitated from the aqueous phase as described above. Finally, the rRNA was resuspended in 100 μl of DEPC-pretreated water. The integrity of the 16S and 23S rRNA fragments was checked by electrophoresis on a 1.2% agarose gel and comparison to an rRNA standard from Escherichia coli (Roche Diagnostics, Mannheim, Germany). The gel was stained with ethidium bromide.

PCR of bacterial 16S rRNA genes.

PCR was carried out using the oligonucleotide primers 27f and 1492r (28), which amplify 16S rRNA genes of a wide range of members of the domain Bacteria from positions 28 through 1491 (E. coli numbering [5]). For T-RFLP analysis, the 5′ primer was labeled with the dye carboxyfluorescein. The reaction mixture contained 1 μl of template DNA, 10 μl of 10× reaction buffer (PCR buffer II; PE Applied Biosystems, Foster City, Calif.), 1.5 mM MgCl₂, 200 μM (each) deoxynucleoside triphosphate (dNTP) (U.S. Biochemical, Cleveland, Ohio), 0.3 μM (each) primer (MWG-Biotech, Ebersberg, Germany), and 2.5 U of Taq DNA polymerase (AmpliTaq; PE Applied Biosystems). The thermal PCR profile was as follows: initial denaturation for 2 min at 94°C; 32 cycles (total DNA extracted from soil slices) or 25 cycles (cloned 16S rDNA for subsequent T-RFLP analysis) consisting of denaturation at 94°C for 45 s, primer annealing at 48°C for 60 s, and elongation at 72°C for 120 s. The final elongation step was extended to 12 min. Amplification was performed in a total volume of 100 μl in 0.2-ml reaction tubes and a DNA thermal cycler (model 2400; PE Applied Biosystems). Aliquots of the 16S rRNA gene amplicons (10 μl) were checked by electrophoresis on a 1% agarose gel.

RT-PCR of bacterial 16S rRNA.

Ribosomal copy DNA (rcDNA) synthesis was performed using the Moloney murine leukemia virus reverse transcriptase, RNase H minus (Promega). An aliquot (1 μl) of the specified extract of total RNA was mixed with 60 pmol of primer 907r (28), and the mixture was filled with sterile, RNase-free water (Sigma, St. Louis, Mo.) to a volume of 15 μl. To denature the secondary structure of 16S rRNA, the template-primer mixture was incubated at 70°C for 5 min and then immediately stored on ice. For rcDNA synthesis, the following agents were added: 1× reaction buffer (Promega), 2.5 mM (each) dNTP (U.S. Biochemical), 40 U of RNase inhibitor (Promega), and 200 U of Moloney murine leukemia virus reverse transcriptase. The reaction was performed in a total volume of 25 μl at 42°C for 1 h and stopped by incubation at 70°C for 5 min. Aliquots (1 μl) of the rcDNA solution were used for subsequent PCR amplification with primers 27f (labeled for T-RFLP analysis with carboxyfluorescein) and 907r. The composition of the reaction mixtures was the same as described for the amplification of 16S rRNA genes (see above). The thermal PCR profile was as follows: initial denaturation at 94°C for 2 min; 28 cycles consisting of denaturation at 94°C for 45 s, primer annealing at 52°C for 60 s, and elongation at 72°C for 60 s. The final elongation step was extended to 12 min. Aliquots (10 μl) of the 16S rcDNA amplicons were checked by electrophoresis on a 1% agarose gel.

T-RFLP analysis.

Both 16S rRNA genes and 16S rcDNA were amplified by PCR as described above. After purification with Qiaquick spin columns (Qiagen, Hilden, Germany), approximately 100 ng of the amplicons was digested with 10 U of the restriction endonuclease MspI (Promega). The digestions were carried out in a total volume of 10 μl for 3 h at 37°C. Aliquots (2.5 μl) of the digested amplicons were mixed with 2.0 μl of formamide and 0.5 μl of an internal lane standard (GeneScan-1000 ROX; PE Applied Biosystems). The mixtures were denatured at 100°C for 3 min and then chilled on ice. Electrophoresis on a polyacrylamide gel was performed using an automated DNA sequencer (model 373; PE Applied Biosystems) for 6 h at the following settings: 2,500 V, 40 mA, and 27 W (24-cm gel length). After electrophoresis, the sizes of the 5′-terminal restriction fragments (T-RFs) and the intensities (=peak areas) of their fluorescence emission signals were automatically calculated by the GeneScan Analysis software, version 2.1 (PE Applied Biosystems). The accuracy of size calling between replicates was ±1 bp. This permitted the comparison of T-RFLP community fingerprint patterns obtained from different soil slices for similarities and dissimilarities, i.e., permitted the unambiguous decision as to whether T-RFs had to be considered identical or not. The proportional abundance of individual T-RFs within a given T-RFLP pattern was determined as the peak area of the respective T-RF divided by the total peak area of all T-RFs detected within a fragment size range between 50 and 928 bp and was expressed as a fraction based on 1.0.

Cloning and sequencing.

The bacterial 16S rRNA gene pools recovered from both the oxic and anoxic zones of the same flooded soil core (oxic and anoxic soil core clones [oxSCC and anoxSCC, respectively]) were cloned using a TOPO TA cloning kit (Invitrogen Corp., San Diego, Calif.) in accordance with the manufacturer's instructions. The preparation of plasmid DNA of randomly selected clones followed by PCR-mediated amplification of cloned 16S rDNA inserts and their nonradioactive sequencing were carried out as described previously (20).

Phylogenetic analysis.

Comparative sequence analysis of cloned 16S rDNA, i.e., data processing and reconstruction of trees, was done by use of the ARB program package (developed by O. Strunk and W. Ludwig; available online at http://www.biol.chemie.tu-muenchen.de/pub/ARB/). The 16S rDNA clones, each at least 800 nucleotides in length, were added to a database of about 7,000 complete or partial bacterial 16S rRNA sequences (31, 41). This database was part of the ARB program package. Phylogenetic placement was done in more detail by comparing the 16S rDNA clone sequences to reference sequences of the α and β subclasses of the class Proteobacteria (α- and β-Proteobacteria) or to reference sequences of the clostridial cluster I of Collins et al. (7). The tree topologies were evaluated by distance matrix analyses. To avoid possible treeing artifacts caused by nucleotide sequence positions that are subject to multiple mutational changes and/or cannot be aligned unambiguously, we used a 50% invariance criterion for the inclusion of individual nucleotide sequence positions in the treeing analyses (14, 16). Evolutionary distance values between pairs of microorganisms were calculated by using the Jukes-Cantor equation (25) and only those positions present in both sequences of the various sequence pairs. The trees were constructed by using the neighbor-joining algorithm (44). Overall 16S rDNA similarities were determined by using the appropriate tool of the ARB program package.

Nucleotide sequence accession numbers.

The environmental 16S rDNA clone sequences recovered in this study from the oxic and anoxic zones (clones oxSCC-1 to oxSCC-40 and anoxSCC-7 to anoxSCC-44, respectively) of a flooded paddy soil core have been deposited in the EMBL, GenBank, and DDBJ nucleotide sequence databases under accession no. AJ387860 through AJ387886.

RESULTS

After incubation in the dark for 7 days at 30°C, gas bubbles became visible within the flooded soil cores, indicating a high level of gas production due to microbial activity. The surface layer at the floodwater/soil interface was red colored, which may point to the presence of oxidized iron compounds. The oxygen depth profile was measured with a microelectrode in triplicate. Two measurements indicated that oxygen was depleted from 140 μM at the floodwater/soil interface to nondetectable amounts at a depth of 1.6 mm and below, while one measurement detected oxygen down to a depth of 2.2 mm (Fig. 1). Hence, the depth of oxygen penetration into the flooded soil cores was considered to be approximately 2.0 mm. This corresponds well to oxygen profiles determined in flooded, unplanted rice paddy soils in previous studies, regardless of the period of incubation after which the oxygen measurements had been performed, e.g., 1 day (15), 14 days (17), or 7 weeks (36) after flooding.

FIG. 1.

Oxygen depth profile as determined for one flooded, unplanted paddy soil core after 7 days of incubation at 30°C in the dark. Microelectrode measurements were conducted in triplicate.

Three flooded soil cores were examined on the DNA level by separate extraction of total community DNA from individual slices. For each of the three cores, T-RFLP community fingerprint patterns were obtained from all 200-μm slices of the upper 3 mm and from a few additional slices taken at a depth between 3 and 7 mm. The community fingerprint patterns obtained from the same depth but different soil cores were highly similar. The fingerprints showed a tremendous shift in the community patterns with depth, which directly corresponded to the oxygen depletion measured within the upper 2-mm surface layer (Fig. 2). Distinct sets of T-RFs were identified for the oxic and anoxic zone. The community fingerprint patterns obtained from the oxic soil slices were characterized by T-RFs with sizes of 90, 141, 148, 160, 436, 454, 486/489 (the 486- and 489-bp T-RFs could not be clearly resolved by T-RFLP analysis), and 496 bp (Fig. 2A-I). These T-RFs decreased rapidly within the upper 2.0-mm soil layer (Fig. 2B) and were almost undetectable at soil depths where oxygen was depleted (Fig. 2A-III and B). Toward fully anoxic conditions, the high abundance of T-RFs with sizes of 270, 510, and 519 bp became obvious. The latter three T-RFs contributed a proportional abundance of 0.61 to the total fluorescence signal intensity of all T-RFs in the community fingerprint pattern obtained from a depth of 6.0 to 6.2 mm (Fig. 2B). T-RFs indicative of either the oxic or anoxic soil slices were detected in the community fingerprint patterns obtained from the oxic/anoxic transition zone, as exemplified by the T-RFLP pattern obtained from a depth of 1.0 to 1.2 mm (Fig. 2A-II).

FIG. 2.

Changes in the bacterial community structure along the depth profile of a flooded paddy soil core as determined by T-RFLP analysis. (A) Representative 16S rDNA-based community fingerprint patterns obtained from the depths as indicated in panel B (2A-I, -II, and -III) or from the air-dried paddy soil fraction used for the preparation of the flooded soil cores (2A-IV). The numbers indicate sizes of T-RFs which clearly changed in proportional abundance with soil depth. (B) Changes in the proportional abundance of predominant T-RFs in relation to soil depth and the oxygen profile (idealized curve; see also Fig. 1).

All community fingerprint patterns obtained from the flooded soil cores after 7 days of incubation clearly differed from the T-RFLP pattern of that soil which had been used as the starting material for preparation of the flooded soil cores (Fig. 2A-IV). Obviously, the abundant T-RFs detected in the oxic soil slices, e.g., T-RFs with sizes of 148, 436, and 486/489 bp, were detected only with low fluorescence signal intensities or not at all in the nonincubated soil (Fig. 2A-I versus 2A-IV). Some of the T-RFs characteristic of the anoxic zone were present in the bacterial diversity pattern obtained from the nonincubated soil. However, their proportional abundances were clearly different, with a strong increase of the 270- and 510-bp T-RFs and a decrease of the 143- and 152-bp T-RFs in the community fingerprint patterns obtained from the anoxic zone (Fig. 2A-III versus 2A-IV).

To identify the bacterial groups which corresponded to the predominant T-RFs, 16S rDNA clone libraries were generated for both the oxic and anoxic zones. Randomly selected clones were examined by T-RFLP analysis in comparison to the corresponding community fingerprint pattern. In total, 24 16S rDNA clones obtained from the oxic zone and 19 recovered from the anoxic zone were partially sequenced; i.e., at least 800 nucleotide sequence positions were determined for each of the 16S rDNA clones.

The majority of the oxSCC sequences belonged to the α- and β-Proteobacteria (6 and 12 clones, respectively [Fig. 3 and 4]). The four α-proteobacterial clones oxSCC-5, oxSCC-25, oxSCC-29, and oxSCC-36 formed a tight cluster and grouped together with oxSCC-22 within the phylogenetic radiation of the genus Sphingomonas (Fig. 3). The overall 16S rDNA sequence similarity values to Sphingomonas spp. ranged between 96.4 and 99.7%. oxSCC-1 showed a moderate relationship to Beijerinckia indica (overall 16S rDNA sequence similarity of 93.3%). Six clone sequences (oxSCC-13 through oxSCC-40, Fig. 4) formed a tight cluster within the β-Proteobacteria. Members of this cluster plus clones oxSCC-3, oxSCC-6, and oxSCC-26 exhibited, with overall 16S rDNA similarity values not higher than 94%, only moderate relationships to cultured β-Proteobacteria. Three other oxSCC sequences could be closely related to members of the β-Proteobacteria (overall 16S rDNA similarity values are given in parentheses): oxSCC-4 to Aquaspirillum delicatum (97.5%), oxSCC-12 to Burkholderia cepacia (99.5%), and oxSCC-14 to Ralstonia pickettii (99.6%) (Fig. 4).

FIG. 3.

Phylogenetic dendrogram showing oxSCC sequences in relation to representative members of the α-Proteobacteria. The dendrogram was inferred by distance matrix analysis based on 742 nucleotide sequence positions. The root was determined by using the 16S rDNA sequence of E. coli as the outgroup reference. The T-RFs indicated for the oxSCC sequences correspond to the community fingerprint patterns shown in Fig. 2 and 6. Scale bar = 10% estimated difference in nucleotide sequence positions.

FIG. 4.

Phylogenetic dendrogram showing oxSCC sequences in relation to representative members of the β-Proteobacteria. The dendrogram was inferred by distance matrix analysis based on 791 nucleotide sequence positions. The root was determined by using the 16S rDNA sequence of E. coli as the outgroup reference. The T-RFs indicated for the oxSCC clone sequences correspond to the community fingerprint patterns shown in Fig. 2 and 6. Scale bar = 10% estimated difference in nucleotide sequence positions.

Most of the oxSCC sequences placed into the α- and β-Proteobacteria could also be assigned to one of the abundant T-RFs of the oxic zone (compare Fig. 2 with Fig. 3 and 4). Except for one 16S rDNA clone, the α-proteobacterial Sphingomonas-like sequences corresponded to the 148-bp T-RF. The majority of the β-proteobacterial clone sequences shared the 486/489-bp T-RF, which was the most abundant T-RF in the community fingerprint patterns obtained directly from the oxic soil slices. Similarly, the abundant T-RFs with sizes of 141, 436, and 454 bp could be assigned to members of the β-Proteobacteria. In addition, the 90- and 160-bp T-RFs could also be linked to distinct phylotypes; i.e., oxSCC-15 (90-bp T-RF) belonged to the Cytophaga/Flavobacterium/Bacteroides phylum, and oxSCC-16 (160-bp T-RF) grouped within the δ-Proteobacteria. The remaining four oxSCC sequences could not be assigned to any of the abundant T-RFs. These 16S rDNA clones were affiliated with various bacterial lineages but not with the α- and β-Proteobacteria (data not shown in the format of a tree).

Nine of the anoxSCC sequences were related to members of the clostridial cluster I of Collins et al. (7) (Fig. 5), with eight of them forming a coherent cluster. This cluster was phylogenetically intertwined with environmental 16S rDNA sequences (BSV clones) and strain RPec1, which had been retrieved from the same type of anoxic paddy soil in previous studies (6, 20). The intracluster 16S rDNA similarity values ranged between 92 and 98%, while similarities to the next closely related cultured outgroup organism (Clostridium fallax [Fig. 5]) were between 91 and 93%. The anoxSCC-BSV cluster could be subdivided into two distinct groups characterized by T-RFs highly indicative of the anoxic zone, i.e., T-RFs with sizes of 270 and 510 bp, respectively (compare Fig. 2 with Fig. 5). Eight further anoxSCC sequences were related to various sublineages of the low-G+C gram-positive bacteria. However, none of these clones could be assigned to one of the abundant T-RFs indicative of the anoxic soil slices. This is also true for the remaining two anoxSCC sequences, which grouped within the Cytophaga/Flavobacterium/Bacteroides phylum and the division Verrucomicrobia (19), respectively (data not shown in the format of a tree).

FIG. 5.

Phylogenetic dendrogram showing anoxSCC sequences in relation to representative members of the clostridial cluster I of Collins et al. (7), as well as to Oxobacter pfennigii, Caloramator indicus, and Caloramator fervidus. The dendrogram was inferred by distance matrix analysis based on 766 nucleotide sequence positions. The root was determined by using the 16S rDNA sequence of Bacillus subtilis as the outgroup reference. The T-RFs indicated for anoxSCC and BSV sequences correspond to the community fingerprint patterns shown in Fig. 2 and 6. Scale bar = 10% estimated difference in nucleotide sequence positions.

The 16S rRNA-based community fingerprint pattern obtained from the oxic zone corresponded well to those obtained on the DNA level (compare Fig. 2A-I with Fig. 6A); i.e., T-RFs with sizes of 148, 436, 454, and 486/489 bp were detected with a high proportional abundance in both approaches. A major difference between the DNA and RNA approaches was the detection of a 137-bp T-RF in the oxic zone only by the 16S rRNA-based analysis. Comparative analysis of 16S rRNA-derived clone sequences showed that the 137-bp T-RF corresponded like the 486/489-bp T-RF to members of the β-Proteobacteria (A. Henkel and W. Liesack, unpublished data). The 16S rRNA-based community fingerprint pattern obtained from the anoxic zone was also very similar to those obtained on the DNA level; i.e., T-RFs with sizes of 270, 510, and 519 bp were most abundant.

FIG. 6.

16S rRNA-based community fingerprint patterns obtained from the oxic (A) or anoxic (B) zone of flooded paddy soil cores. The community fingerprint patterns are based on the extraction of total RNA from the different soil layers followed by RT-PCR of the bacterial 16S rRNA fraction and T-RFLP analysis. The numbers indicate sizes of major T-RFs which are indicative of the oxic and anoxic zones, respectively (see also Fig. 2).

DISCUSSION

Flooded paddy soil cores were chosen as a model system to assess the impact of oxygen depletion on the bacterial community structure. The active respiration processes of the microbial community at the floodwater/soil interface in conjunction with the reduced diffusivity of oxygen in water (10,000 times slower than in air) leads to a rapid depletion of oxygen with soil depth. Apart from the content of degradable organic matter in the surface layer (37, 40), the production of methane and its diffusion toward the oxic surface layer are known to cause an increase of the biological oxygen demand at the oxic/anoxic interfaces of such environments in which sulfate reduction is not a predominant process, e.g., flooded paddy soils, natural wetland soils, and freshwater sediments (17, 42, 51).

The community fingerprint patterns of all soil slices examined were clearly different from that of the air-dried paddy soil which had been used for preparation of the flooded soil cores. This finding suggests that the bacterial populations detected after 7 days of incubation were actively growing. This was further indicated by the results of the rRNA approach (see below). The gradual changes in the community fingerprint patterns directly corresponded to the depletion of oxygen with depth. Although not a final proof, this close correlation provides strong evidence that the presence or absence of oxygen was the major factor which determined the changes in the bacterial community structure. This conclusion was further supported by the assignment of T-RFs indicative of the oxic or anoxic zones to phylogenetic groups which fitted well into the ecological context of such a gradient system (see below). The latter point, and also the gradual way in which the community fingerprint patterns changed from fully oxic to fully anoxic conditions, suggests that the molecular retrieval was not strongly affected by PCR-based artifacts like preferential amplification of unique 16S rDNA sequence types (29, 50).

To obtain further evidence for our conclusion that the bacterial populations detected on the DNA level were the metabolically active groups, an rRNA-based approach was applied. In general, metabolically active cells have a higher ribosome content than those which are in a stationary growth phase or even in a dormant stage (3, 33). Thus, the rRNA content of a defined microbial group in a given environmental sample is a function of both metabolic activity and population size. Several studies focused on the identification of the metabolically active portion of microbial communities via the extraction of total RNA (1, 13, 49, 52, 55). In some of these studies, DGGE or temperature gradient gel electrophoresis (D/TGGE) was applied to obtain from the same environment community banding profiles on the DNA and RNA levels (13, 52). Comparison of D/TGGE banding patterns always revealed clear differences between the DNA-based and RNA-based analyses. This finding was interpreted as the difference between the genetic potential (DNA level) and the active portion (RNA level) of a microbial community (13, 52). However, a differently biased retrieval of the molecular data could not be excluded. In contrast to these D/TGGE-based studies, the T-RFLP patterns obtained in the course of this study from both total DNA and RNA were, except for the rRNA-based detection of a 137-bp T-RF in the oxic zone and a 151-bp T-RF in the anoxic zone, highly similar. This obvious difference between our T-RFLP-based investigation and previous D/TGGE-based studies may reflect a different in situ situation within the environments examined, but it could also be due to the different levels of molecular resolution of the two fingerprinting techniques used. Individual bands of D/TGGE-based fingerprint patterns are considered to correspond to one or only a few distinct phylotypes, while T-RFs mostly represent clearly more than one defined sequence type. These sequence types can be part of a monophyletic group but may also scatter over a broad phylogenetic range. Nonetheless, the recovery of the same predominant T-RFs on the DNA and RNA levels from flooded paddy soil cores suggests that the same bacterial groups were detected in the two separate approaches.

The bacterial groups identified and their spatial distribution within the oxic/anoxic interface fit well with our understanding about such a gradient system, including the predominance of bacteria of mainly the α- and β-Proteobacteria in the oxic zone. The majority of the α-proteobacterial clone sequences were affiliated with Sphingomonas spp. All cultured members of Sphingomonas are obligate aerobes which can utilize a rather broad range of carbon sources for growth. The presence of the 148-bp T-RF as one of the major fragments especially in the rRNA-based community fingerprint pattern suggests a high metabolic activity of Sphingomonas spp. in the oxic zone (compare Fig. 2 and 6), which might be a reflection of their phenotypic capabilities.

A similar argumentation should also be true for the phenotypes which correspond to those oxSCC sequences affiliated with the β-Proteobacteria. However, phenotypic characteristics and thus a possible ecological role can be deduced only when pure cultures become available or at least the environmental sequence data can be related very closely to already described species (29).

The high proportional abundance of T-RFs which correspond to members of clostridial cluster I agrees well with the results of previous studies on the bacterial diversity in anoxic rice paddy soil (20). This is documented by the finding that eight anoxSCC sequences belonging to the clostridial cluster I are closely intertwined with BSV clones (20) and strain RPec1 (6). The main carbon sources in this methanogenic soil are rice plant residues and rice straw. Because polysaccharides, such as cellulose, hemicellulose, pectin, and xylan, are major components of rice straw (54), their hydrolysis is the primary mineralization step in the anaerobic degradation of organic matter in flooded paddy soils (8). Thus, the anoxSCC and BSV sequences can be considered molecular markers for a predominant polysaccharide-degrading population. This view is supported by the capability of strain RPec1 to utilize, in addition to monosaccharides, pectin and xylan as growth substrates (6).

Taken together, the results of this study demonstrated for a given time point spatial changes in the bacterial community structure within a few millimeters, which correlated to the depletion of oxygen. Community fingerprint patterns obtained from the same depth but different soil cores were highly similar, even between the DNA and RNA approaches. The identification of members of the α- and β-Proteobacteria and clostridia as predominant inhabitants of the oxic and anoxic zones, respectively, may point to general principles on how the presence or absence of oxygen determines bacterial community structure. The latter will have to be examined in more detail by analysis of both the spatial and temporal development of defined populations within oxic/anoxic interfaces.