Bacterial diversity in maize rhizospheres: conclusions on the use of genetic profiles based on PCR-amplified partial small subunit rRNA genes in ecological studies

Field site and sampling

The field with a total size of 3.5 ha was located in Braunschweig-Völkenrode. The field was homogeneous regarding its previous agricultural history, its height, water regime and other physico-chemical factors. It was divided into three replicated areas. A replicated area consisted of 36 plots, each with an area of 125 m². A single plot on a replicated area represented a different agricultural crop or cultivar (maize, sugar beet, wheat) or treatment (herbicide regime, ploughing). The order of plots was in a randomized block design.

In 2000, five different treatments were sampled in context of this study: the maize cultivar Bosphore (Agrevo, Frankfurt, Germany) with no herbicides or treated with conventional herbicides (Artett and Motivell, active compounds: Terbuthylazin and Nicosulphuran); a transgenic derivative of Bosphore (KX8445, transformed with the pat-gene encoding for a l-phosphinothricin, syn. glufosinate, inactivating transacetylase, Agrevo) treated with conventional herbicides or Liberty (active ingredient: l-glufosinate) and Bosphore grown in nonploughed soil treated with conventional herbicides. Sampling was carried out 7 and 35 days after the herbicide treatments, corresponding to 32 days and 60 days after seeding. The shoot heights of the plants at the first sampling date were approx. 25 cm and at the second sampling 140 cm.

Extraction of bacterial cell consortia and DNA

For each sample, a total of eight plants for the first sampling and six plants for second sampling were selected from random positions within each plot and whole plants were transported to the laboratory. Here, roots were dipped into sterile water to remove large adhering soil particles. Fine roots were cut off and used for bacterial cell extraction. For each sample, 8 g (wet weight) of fine roots were suspended in 20 mL sterile saline solution (0.85% NaCl, wt/vol) for 30 min at 4 °C in an overhead shaker (KH, Guwina-Hoffmann) at 20 r.p.m. After removal of the root material, suspensions were divided into two aliquots and each sample was centrifuged at 4100 g for 30 min at 4 °C. Supernatants were discarded and the pellets were stored at −70 °C.

Frozen bacterial cell pellets of rhizosphere extracts were resuspended in 12 mL sterile lysis buffer (0.05 m Tris-HCl, 0.01 m Na₂EDTA, 0.05 m NaCl; pH 8.0). Cells were lysed by five cycles of freeze-thawing, each cycle consisting of 5 min freezing in liquid nitrogen and 5 min thawing at 65 °C. Between each step, the suspensions were vortexed for 10 s at the highest setting (VF2, IKA Labortechnik). Subsequently, samples were treated with Proteinase K (Roche) for 60 min at 65 °C in a shaking water bath. DNA was extracted by phenol–chloroform (Sambrook et al. 1989). The DNA was purified further with the Wizard DNA Purification Kit (Promega). For this purpose, 10–30 µL of DNA-solution were loaded onto each column and eluted with 40 µL of 75 °C prewarmed 10 mm Tris-HCl, pH 8.0, according to a protocol of the manufacturer. The DNA concentration in the final solution was in the range of 1–10 ng µL⁻¹.

PCR amplification of partial SSU rRNA genes

Amplifications were processed in a final volume of 100 µL with 5 U Platinum-Taq-Polymerase (Gibco Livetech), onefold PCR-buffer, 1.5 mm MgCl₂, 0.5 µm primers, 200 µm of each desoxy-nucleotide (Amersham Pharmacia Biotech) and 2 µL of purified rhizosphere extracted DNA in the thermal cycler Primus 96 (MWG-Biotech). For amplification, we selected primers Com1 and Com2 which hybridize to phylogenetically highly conserved regions of the SSU rRNA gene, corresponding to Escherichia coli position 519–537 and 907–926 (Schwieger & Tebbe 1998; Schmalenberger et al. 2001). The reverse primer was phosphorylated at the 5′ end for single-strand digestion (Schwieger & Tebbe 1998). Primers were synthesized by Gibco Life Technologies. PCR was conducted at 95 °C for 3 min, followed by 35 cycles of 1 min 95 °C, 1 min at 50 °C, 70 s at 72 °C, and a final primer extension for 5 min at 72 °C.

Genetic profiles by single strand conformation polymorphism (SSCP)

PCR-products were purified with the Qiaquick PCR purification kit (Qiagen) as recommended by the manufacturer. Half of the purified PCR product was digested with 10 U lambda-exonuclease (Amersham Pharmacia Biotech) at 37 °C for 45 min according to the single-strand community approach described previously (Schwieger & Tebbe 1998). Proteins were removed by phenol–chloroform extraction, DNA was precipitated, collected and dried (Schwieger & Tebbe 1998). Prior to SSCP, the samples were resuspended in 8 µL of 10 mm Tris-HCl, pH 8.0, and 8 µL of denaturing loading buffer (95% v/v formamide, 10 mm NaOH, 0.025% wt/vol bromophenol blue and 0.025% wt/vol xylene cyanole). Samples were incubated for 2 min at 95 °C and cooled immediately on ice.

For electrophoresis, we diluted a twofold stock solution to 0.6-fold (MDE, FMC Bioproducts) and prepared a gel of 21 cm length in 1 X TBE (Sambrook et al. 1989). A Macrophor sequencing apparatus (Amersham Pharmacia Biotech) was used for SSCP. Gels were run at 20 °C at 8 mA and 400 V for 16 h. DNA in the gels was visualized by silver staining (Bassam et al. 1991).

Extraction, reamplification and sequencing of DNA from silver-stained SSCP-profiles

Selected bands of community profiles were cut out with a sterile razor blade and the single-stranded DNA of these bands was eluted for 3 h at 37 °C and 500 r.p.m. in 50 µL ‘crush and soak’ solution (Sambrook et al. 1989). The eluted DNA was precipitated with ethanol, centrifuged and dried as previously described. The precipitated DNA was finally resuspended in 10 µL of 10 mm Tris-HCl, pH 8.0. The sequences were recovered by PCR with the Com-primers as described above in a 50-µL scale and with a total of 1.25 U of Platinum Taq polymerase (Gibco Life Technologies). PCR products were purified as previously described. Half the products were digested to obtain single stranded DNA (5 U of lambda-exonuclease, Amersham Pharmacia Biotech) to check the correct position of reamplified single bands on SSCP gels. The reamplified DNA molecules recovered from bands of community profiles were used for cloning and sequencing (Peters et al. 2000; Schmalenberger et al. 2001).

Southern blotting of SSCP gels and gene probe hybridizations

In SSCP analyses that were conducted for Southern blot analysis, both thermostatic plate and glass plate were treated with Repel Silane (Amersham Pharmacia Biotech) to allow separation of the acrylamide gel from the glass plates. For transfer of DNA, SSCP-gels were blotted onto positively charged nylon membranes (Hybond N +, Amersham Pharmacia Biotech or Biodyne plus Nylon membranes, respectively) in an electric field for 5 h with 10 mA/cm², using the Nova Blot Unit (Amersham Pharmacia Biotech) and 0.5 × TBE (Sambrook et al. 1989). Membranes were dried at room temperature and DNA was cross-linked to the membranes with 1.5 J/cm² in UV light at 266 nm (FluoLink).

Specific probes (Table 1) of 27–28 base length matching inside the SSU rRNA gene position 538–906 (corresponding to the numbering of E. coli) were designed with the ARB software environment (www.arb-home.de) and the oligo 6 software (MBI, Cascade). Probes were ordered at Gibco Life Technologies and labelled with the alkaline phosphatase direct labelling kit (Amerham Pharmacia Biotech) according to a protocol provided by the manufacturer.

Membranes were placed into glass tubes in a mini hybridization oven (Oncor Appligene) with 20 mL of the hybridization solution according to the manufacturer (RPN3680, Amersham Pharmacia Biotech). Depending on the gene probe and annealing conditions, hybridization temperatures were optimized empirically in the range of 45–50 °C (except probe PROC907 which hybridized at 35 °C). The probe concentration was set to 10 ng/mL and the incubation time was 16 h. Washing procedures were conducted according to the manufacturer with the supplied urea containing buffer 1 and the Tris-base containing buffer 2 (RPN3680, Amersham Pharmacia Biotech). Buffer 1 (50 mL per tube) was applied for 10 min at 10 °C below hybridization temperature. The washing procedure was repeated once. Subsequently the double concentrated buffer 2 was used twice for 10 min and 15 °C (for PROC907: 5 °C) below hybridization temperature to reduce the content of probes which hybridized nonspecifically with the nylon membranes. Finally, buffer 2 was applied twice for 10 min at 30 °C (for PROC907: 25 °C). Afterwards, membranes were incubated with 30 µL cm⁻² of the CDP-Star detection reagent for 5 min (Amersham Pharmacia Biotech). Membranes were placed into plastic bags and chemoluminescence was detected with Hyperfilm ECL (Amersham Pharmacia Biotech) incubated for 20 min to 18 h, depending on the expected strength of the hybridization signal, in a film cassette.

DNA sequence analysis

DNA was sequenced to obtain the base compositions in both directions and consensus sequences were generated as described elsewhere (Schmalenberger et al. 2001). For further analysis the consensus sequences were loaded into the ARB Database (http://www.arb-home.de). After alignment of the partial SSU rRNA genes, sequences were integrated with the parsimony iteractive function into an existing phylogenetic tree consisting of more than 10 000 complete SSU rRNA genes. Distance matrices were calculated with a 50% filter for each phylogenetic group.

Nucleotide sequence accession nos

The partial SSU rRNA gene sequences were deposited in GenBank under the following accession nos: AJ431194, AJ431195, AJ431276 to AJ431327 and AJ 437404 to AJ437477.

Comparison of genetic profiles at the level of pattern similarity

A total of 30 SSCP patterns were generated from the bacterial cell consortia extracted from rhizospheres of maize plants of different age (see Materials and methods). Each of the SSCP patterns consisted of approx. 50 distinguishable bands and all patterns from one sampling date were highly similar (Fig. 1). Some unique bands existed in single profiles but these bands did not occur in all of three replicate samples or in samples from the same field block. Thus, an altered composition of the bacterial community in response to the genetic modification, to herbicides, to soil tillage or to heterogeneities in the field could not be detected. In addition, no significant differences were found when profiles obtained from the two different sampling dates were compared with each other by digital image analysis (data not shown).

In a previous study with maize plants grown at the same field site only one year earlier, in 1999, small but significant differences in the composition of the bacterial communities in rhizospheres as a response to plant age (32 days and 70 days after seeding) could be detected (Schmalenberger & Tebbe 2002). As the same set of primers and plants of nearly the same age and from the same field were used in both studies, it is likely that plant-age dependent differences actually exist, but that these differences are close to the threshold of detection with PCR-SSCP, the chosen primer pair, and digital image analysis.

Identification of sequences contributing to the genetic profiles and indications for the size of the rRNA gene pool

In order to characterize the richness of partial rRNA gene sequences contributing to the patterns, we selected bands from different profiles at different positions for further analysis. The positions of the bands that were cloned and sequenced are shown in Fig. 1. We included bands which were consistently observed in the profiles (numbers on the right side of the panel) and those which occurred inconsistently only in single profiles. PCR-reamplification of some eluted bands, i.e. no. 28 and no. 29 from the first sampling (Fig. 1A) and no. 2 and no. 8 from the second sampling (Fig. 1B) resulted in one or two additional bands, as detected by a second SSCP-analysis (no Figure). All these bands were identified by DNA-sequencing. In total, we cloned and sequenced 54 DNA fragments from the SSCP-profiles shown in Fig. 1.

The sequences could be attributed to several phylogenetic groups (Table 2). In accordance with our own previous studies (Schmalenberger et al. 2001; Schmalenberger & Tebbe 2002) and with another study on bacterial communities from maize rhizospheres (Chelius & Triplett 2001) we found that most sequences were related to Proteobacteria from different subgroups and to members of the Cytophaga–Flavobacterium–Bacteroides (CFB) group. The bacterial species Paenibacillus polymyxa (Seldin et al. 1998; von der Weid et al. 2000) and Burkholderia cepacia (Balandreau et al. 2001) can be found as common inhabitants of maize rhizospheres. However, we could not detect rRNA sequences indicating their presence. Variability in microbial community compositions can be explained by soil factors (Miethling et al. 2000) or cultivars (Dalmastri et al. 1999), but also by methodological biases, e.g. caused by the selection of the primers that hybridize in the PCR to evolutionarily conserved regions (universal primers) of the SSU rRNA genes (Schmalenberger et al. 2001), or by varying lysis efficiencies for bacteria with different cell walls (Kuske et al. 1998).

Distance matrices were calculated in order to compare the similarity of sequences isolated from our SSCP profiles. In this analysis, we included 30 additional sequences obtained from the same maize cultivars grown on the same field, only 1 year earlier (Schmalenberger & Tebbe 2002), and we placed the sequences into operational taxonomical units (OTUs). In another study, a threshold value of 1% sequence divergence was chosen for the definition of same operational taxonomical unit (OTU) and was found to be useful when SSU rRNA genes were sequenced directly from environmental DNA (Ogino et al. 2001). In our study, we increased the OTU threshold value to 1.5% because our approach incorporated the reamplification of SSCP-gel eluted bands as an additional PCR-step that might have been an additional source of sequence error.

Surprisingly, only five of 40 OTUs were detected at both sampling dates in one year and only two OTUs were found in samples from two different years. The latter OTUs included a member of the α-subgroup of Proteobacteria (AJ308248) and one was from the CFB group (AJ308288). Since all SSCP-patterns from maize rhizosphere were similar, we concluded that there was either an unintended coincidence that different sequences would make up similar patterns, or that there were more sequences contributing to one band than suggested by the sequencing of only a single clone.

How many rDNA-sequences hide behind one band?

To determine the number of possible sequences that make up one single band in an SSCP profile we selected two dominant bands, band nos 32 and 34 from the first sampling date (Fig. 1A), referred to as nos 1–32, 1–34 and one faint band (2–1) (Fig. 1B) from the second sampling date, for further analysis. The band positions are indicated in Fig. 1. The preliminary identifications of these bands are shown in Table 2. Instead of one clone, we sequenced from band 1–32 a total of 26 clones, from 1–34 a total of 36 and from band 2–1 12 clones. We found that in a total of 74 sequences 64 were different. These different sequences could be placed into 50 OTUs. From the faint band (2–1) two OTUs were recovered with one dominant OTU consisting of 11 sequences of which five were identical. The dominant band nos 1–34 consisted of sequences which could be placed into 22 OTUs. Three of these OTUs consisted of four to eight sequences but the others only of one sequence. The diversity of sequences was even higher for band 1–32: only one OTU was represented by three sequences and each of the remaining OTUs consisted only of one sequence.

Sequences from the clone libraries of the bands nos 1–32, 1–34 and 2–1 were computed with the parsimony interactive function into a tree consisting of 62 complete SSU rRNA genes shown in Fig. 2. Sequences obtained from the two strong bands were scattered over a large number of phylogenetic groups, including Proteobacteria of the α-, β-, γ- and δ-subgroups, the CFB-group, Firmicutes with a high G + C content and others. In contrast, sequences recovered from the faint band (no. 2–1) were related closely to each other and could be placed into a deep-rooting branch of the β/γ-subgroup of the Proteobacteria. The closest relatives of these sequences were mainly uncultured bacteria from agricultural soil of a potato field in Germany (AJ252661 and AJ252637; T. Lukow, pers. comm.).

The high diversity of sequences obtained from the dominant bands was so surprising that we included an additional control experiment with cloned sequences from band 1–34. The vectors with the inserted sequences were used as a template in a PCR reaction with Com-primers. After a single strand digest of the reverse DNA strand, the PCR products were analysed by SSCP with the original community profile as a reference. In fact, all 19 clones checked with this procedure were located in the position of band 1–34 in the community profile (no Figure). This confirmed that several unrelated sequences were found within the same band.

Recent references indicate that also bands in DGGE or TGGE (temperature gradient gel electrophoresis)-community profiles may contain more than one sequence (Gomes et al. 2001; Sekiguchi et al. 2001; Smalla et al. 2001). It is probable that that this phenomenon is much more common than acknowledged at present in the literature. In fact, both methods SSCP and D/TGGE amplify rRNA genes from community-DNA and detect the diversity of PCR products on gels. However, for cloning procedures with D/TGGE, staining is normally performed with ethidium bromide or SYBR green. In particular, ethidium bromide staining is less sensitive than silver staining (Muyzer et al. 1998) and, thus, more PCR-product needs to be loaded onto a gel for pattern generation. The probability that a visible band is composed of different sequences may therefore even be higher with D/TGGE than reported here for SSCP.

Gene probes test the presence of rDNA sequences in different SSCP-profiles for comparative community analyses

Up to this point only bands isolated from single profiles were analysed. In order to characterize the abundance of specific sequences in apparently similar profiles, we decided to develop gene probes for Southern blots with community SSCP-profiles (Table 1). Three probes were selected which were specific for different members of the CFB-group. The probe SPOC621 was derived from the sequence found in band 1–32 (see Table 2), probe RUNE655 was derived from clones of band 1–34, i.e. −115, −219, −306, and −319 (see Fig. 2) and probe CYTJ625 was derived from band 2–8b (Table 2). The probe CYTJ625 also matched to a sequence (AJ308288) that was found in the preceding growing season, in 1999. In addition a probe was included that was designed to be highly specific for a sequence which we had found only in 1999 but not in 2000 (AJ308290) (Table 1). This probe, BREV816 was related to members of the α-Proteobacteria. Finally, a universal gene probe (PROC907, corresponds to primer Com2) was used as a positive control to confirm that at the end of the sequential hybridization procedures blotted DNA from the SSCP profiles was still present on the membranes (Fig. 3E).

Probe SPOC621 hybridized with the band from which it had been isolated and it occurred at the same positions in all SSCP patterns that were analysed, including rhizosphere samples from different treatments, from both cultivars and from two growing seasons (Fig. 3A). In addition the probe hybridized weakly with DNA above the positive band, equally for all community patterns that were analysed. The probe also hybridized with Flavobacterium johnsoniae, a member of the CFB group which was a component of our standard mixture flanking the community profiles, even though sequence comparisons suggested five mismatching positions. The positive hybridization signal can be explained by the fact that the amounts of standard DNA were several fold higher for each band than for the community SSCP profiles, amplifying the signal for a weak hybridization. In dot blot analyses with same amounts of reamplified band nos 1–32 and standard DNA, no nonspecific hybridization signals were detected (data not shown).

With probe RUNE655 four additional sequences which would hybridize without any mismatch were found in the database (ARB, see Materials and methods). Gene probes hybridized to the band position from which the sequence was isolated and, in addition, to a band in its vicinity (Fig. 3B). The latter band was at the position of band no. 18 (Fig. 1B), designated as nos 2–18, which had also been identified as a partial sequence belonging to the CFB-group (Table 2). The similarity between the gene probe and that sequence was 100%. As expected from the previous sequence information, probe CYTJ625 hybridized to profiles obtained from both growing seasons (Fig. 3C). A strong hybridization was also observed with F. johnsonae in the standard DNA, which is in accordance to the predicted specificity of that probe. Samples from the previous growing season showed an additional weakly hybridizing band. This pointed to a possible difference between profiles of both years.

Probe BREV816, which was selected to be specific for samples from the 1999 growing season, hybridized with the same intensity to samples from maize rhizospheres collected in 2000. Thus, despite the large number of sequences that were recovered in the context of this study that particular sequence was obviously missed (Fig. 3D).