Searching for predominant soil bacteria: 16S rDNA cloning versus strain cultivation

Authors

  • Andreas Felske,

    Corresponding author
    1. Laboratory of Microbiology, Department of Biomolecular Sciences, Wageningen Agricultural University, Hesselink van Suchtelenweg 4, 6703 CT Wageningen, Netherlands
    Search for more papers by this author
  • Arthur Wolterink,

    1. Laboratory of Microbiology, Department of Biomolecular Sciences, Wageningen Agricultural University, Hesselink van Suchtelenweg 4, 6703 CT Wageningen, Netherlands
    Search for more papers by this author
  • Robert van Lis,

    1. Laboratory of Microbiology, Department of Biomolecular Sciences, Wageningen Agricultural University, Hesselink van Suchtelenweg 4, 6703 CT Wageningen, Netherlands
    Search for more papers by this author
  • Willem M. de Vos,

    1. Laboratory of Microbiology, Department of Biomolecular Sciences, Wageningen Agricultural University, Hesselink van Suchtelenweg 4, 6703 CT Wageningen, Netherlands
    Search for more papers by this author
  • Antoon D.L. Akkermans

    1. Laboratory of Microbiology, Department of Biomolecular Sciences, Wageningen Agricultural University, Hesselink van Suchtelenweg 4, 6703 CT Wageningen, Netherlands
    Search for more papers by this author

*Corresponding author. Present address: GBF (National Research Institute for Biotechnology, Department of Microbiology, Mascheroder Weg 1, D-38124 Braunschweig, Germany afe@gbf.de

Abstract

The predominant bacteria in Dutch grassland soils, as identified by direct DNA extraction, PCR amplification of 16S rDNA and subsequent cloning and sequencing, were compared to the most abundant culturable bacteria. The 16S rDNAs of the strains from a comprehensive cultivation campaign were compared to some of the predominant cloned sequences by temperature gradient gel electrophoresis (TGGE). Four ribotypes were selected that were found to be abundant in the clone library: two closely related Bacillus-like sequences, a representative from the Verrucomicrobiales cluster and an uncultured member of the Actinobacteria. Using a variety of cultivation approaches a total of 659 pure cultures were isolated. Initially, approximately 8% of all isolates matched any of these ribotypes by same migration speed of their 16S rDNA amplicons on TGGE. However, sequencing analysis of matching isolates indicated that their 16S rDNA sequences were clearly different from the cloned sequences representing the fingerprint bands. Comparing the cultivation approach and the molecular 16S rDNA analysis from the same soil sample, there was no correlation between the collection of cultured strains and the 16S rDNA clone library.

1Introduction

During the last years it has become generally accepted that culture-dependent surveys suffer from the ‘great plate count anomaly’[1]. Most environmental bacterial cells were shown to be refractory to cultivation [2,3]. It has been estimated that less than 1% of all bacterial cells in soil can be cultured in the current types of nutrient media [2]. The culture-independent approach based on direct recovery of bacterial 16S rDNA from soil indicated the predominance of many different uncultured species [4–10]. This has been explained by the presence of a large number of very small and probably inactive cells not able to be recovered on laboratory media [11,12]. It is feasible that the yet uncultured types of bacteria might be grown under laboratory conditions if just the right nutrients are applied. However, the possibility exists that the culturable bacteria represent the viable cells, while cloned 16S rDNA sequences are retrieved from a huge background of inactive cells. If so, this would imply that analysis of cloned 16S rDNA provides information of mainly taxonomic value, with no hint to the main bacterial operators of the biogeochemical processes in soil. A recent molecular study provided information about the metabolically most active members of a bacterial community in soil [13]. The approach used was based on direct ribosome isolation from soil and purification of 16S rRNA [14]. Subsequently, the most active bacteria were detected by temperature gradient gel electrophoresis (TGGE) of the rRNA amplicons obtained by RT-PCR that were compared to those obtained from a cloning approach. It has been demonstrated that 16S rDNA clone libraries fairly well reflected the predominant environmental ribosomes, although appearance of less abundant sequences has been observed [15]. The main goal of the present study was to demonstrate how well the predominant bacteria, as identified by direct DNA/rRNA extraction from soil, were recovered by a cultivation approach. Subsequently, the cultivation of the predominant bacteria would allow further characterization of their physiological properties that might give a hint to their function in the environment. We focused on four predominantly cloned sequences from soil, organisms only identified by their rRNA (ribotypes), to investigate to what extent the results of the 16S rRNA approach could be reproduced by classical cultivation techniques.

2Materials and methods

2.1Soil sampling

The investigated site was located in the Drentse A agricultural research area in the Netherlands (06°41′E, 53°03′N), representing a 1.5-km stretch of grassland along the Anlooër Diepje brook. The different cultivation histories of the Drentse A plots were considered by sampling six plots representing different years of last fertilization for agricultural hay production. Details of the soil properties have been published [18]. In total 360 soil cores (<10 cm depth) of approximately 50 g were taken with a drill (0–10 cm depth) and transferred into sterile sample bags. Pooled samples were prepared by sieving and mixing single samples (5 g input each) from each plot. A general sample was prepared for isolation experiments by mixing together the pooled samples from the single plots.

2.2Enrichment and cultivation of soil bacteria

Soil samples (1 g) were suspended in sterile 0.85% NaCl solution and diluted in 10-fold steps. Agar media were inoculated with 100 μl of these suspensions, corresponding to 10−6–10−9 g soil per plate. All media were inoculated in two different ways. One set of agar media was inoculated with original soil suspensions while another, identical set of plates was inoculated with pasteurized soil suspensions (heated for 15 min at 80°C) to select for Bacillus endospores. A total of 35 different types of media were used at two different pH values (Table 1). Most of the media were mineral media [20] with different components as sole carbon source. Also rich nutrient media, like DSM medium 1, 6, 16 and 78 [19], were used in the original composition and in 10-fold dilutions. Since the predominant environmental ribotype DA001 was closely related to Bacillus benzoevorans (97.3% sequence similarity), we also applied selective media for B. benzoevorans consisting of a mineral medium containing 0.2% benzoate, its derivatives, or other phenylated compounds as sole carbon source [20,21]. Some media were prepared in soil extract [22] instead of water (Table 1). All media contained 15 g l−1 agar and 50 mg l−1 cycloheximide to prevent fungal growth and had a pH of 7. A parallel version of all media was prepared with a pH of 4.0, corresponding to the natural soil pH. All cultures were incubated at 20°C and sampled after 4 days (89 isolates), 2 weeks (232 isolates), 4 weeks (242 isolates) and 3 months (96 isolates), giving a total of 659 separate microbial isolates (Table 1). The colonies appearing on the media inoculated with 10−8 and 10−9 g soil represented the most abundant culturable isolates and were picked entirely (a total of 234). The other isolates were selected for different colony morphologies from the plates inoculated with 10−6 and 10−7 g soil.

Table 1.  The number of strains isolated from the different media of pH 7 or 4 (media inoculated with either untreated or pasteurized soil suspensions)
Medium+carbon source (0.2% w/v or v/v)Soil suspension not pasteurizedSoil suspension pasteurized
 pH 7pH 4pH 7pH 4
  1. aMM=mineral medium.

  2. bMedia selective for B. benzoevorans.

  3. c–=no growth observed.

MMa+nothing6322
MM without yeast extract111
MM+acetate3214142
MM+acetate+soil extract3222
Acetate+soil extract1111
MM+benzoateb2c
MM+casein3919152
MM+casein+soil extract2121
Casein+soil extract4231
MM+chitin3431
MM+corn steep liquor201181
MM+gelatin5221
MM+glucose14664
MM+glycerol3100
MM+humic acids221
MM+malt extract14230
MM+meat extract19372
MM+methyl benzoateb2
MM+m-hydroxybenzoateb1
MM+polyvinylpyrrolidone12831
MM+starch2112102
MM+tannin2
MM+tryptose238143
MM+tryptose+soil extract1111
Tryptose+soil extract3233
MM+soil extract4232
Soil extract2101
DSM 119991
DSM 1 (1/10)3515163
DSM 64311
DSM 6 (1/10)6421
DSM 165211
DSM 16 (1/10)7252
DSM 782231
DSM 78 (1/10)12270

2.3Screening of isolates for matching sequences

Total DNA was extracted from Drentse A isolates by taking up single clone colonies with sterile toothpicks and transferring these into 1.5-ml micro-centrifuge tubes containing 50 μl TE buffer. The tubes were heated for 15 min at 95°C to lyse the cells and then chilled on ice. Amplification of 16S rDNA sequences was performed with a GeneAmp PCR System 2400 thermocycler (Perkin-Elmer-Cetus, Norwalk, CT, USA), using 35 cycles of 94°C for 10 s, 54°C for 20 s and 68°C for 40 s. The PCR reactions (10 μl) contained 10 mM Tris-HCl (pH 8.3), 50 mM KCl, 3 mM MgCl2, 0.05% detergent W-1 (Life Technologies, Paisley, UK), 50 μM each of dATP, dCTP, dGTP, and dTTP, 30 pmol each of primers GC968f and 1401r [15], 0.5 units of Taq DNA polymerase (Life Technologies), and 1 μl cell lysate. Amplification products were confirmed by visualization on 1.4% agarose gels.

The Diagen TGGE system (Diagen, Düsseldorf, Germany) was used for sequence-specific separation of PCR products [23]. Electrophoresis took place in polyacrylamide gels (0.8 mm thick, 6% w/v acrylamide, 0.1% w/v bis-acrylamide, 8 M urea, 20% v/v formamide, 2% v/v glycerol) with 1×TA buffer (40 mM Tris-acetate, pH 8.0) at a fixed current of 9 mA (approximately 120 V) for 16 h. The manufacturer's gel casting setup produces 27 slots of approximately 7.5 μl sample capacity. For the preliminary TGGE screen a slot former (64-slot blunt end comb) from the Li-Cor 4000L Sequencer (Li-Cor, Lincoln, NE, USA) was applied, providing 70 slots of approximately 1 μl sample capacity each. A temperature gradient was built up in the direction of electrophoresis from 37°C to 46°C. After electrophoresis, the gels were silver-stained [24]. Subsequently, the gels were inspected for matches between the clone signals (DA001, DA011, DA079 and DA101) and the bands of the isolates from soil. Visual matches were confirmed by a second TGGE analysis (conventional 27 slots), where PCR products from the matching isolate and the according clones were loaded in the same slot. The isolates that yielded amplicons covering those from a clone, resulting in a single band, were selected for partial sequencing.

2.4Sequencing of PCR products from isolates

Amplification of 16S rDNA sequences was performed with a thermocycler (as above) using 30 cycles of 94°C for 10 s, 46°C for 20 s and 68°C for 100 s. The PCR reactions (2×100 μl) contained 10 mM Tris-HCl (pH 8.3), 50 mM KCl, 3 mM MgCl2, 150 μM each of dATP, dCTP, dGTP, and dTTP, 100 pmol each of primers 7 and 1512 [15], 2.5 units of Taq DNA polymerase (Life Technologies), and 1 μl cell lysate (see above). PCR products were purified and concentrated (from 200 to 50 μl) with glass fiber spin columns following the manufacturer's instructions (High Pure PCR Product Purification Kit, Boehringer-Mannheim, Germany). Purified DNA was eluted from the columns with 50 μl deionized water. The sequencing was done by using a Sequenase (T7) sequencing kit (Amersham, Slough, England). Each 4-μl reaction (A, C, G and T) contained 2.5 μl template, 0.5 μl labeled primer (Infra-Red Dye 41, MWG-Biotech, Ebersberg, Germany) and 1 μl reaction mix (A, C, G or T, Amersham). Inserts were read with primer seq968 (primer GC968f without GC-clamp), sequencing the last approximately 500 bases of the sequence. The reaction was performed in a thermocycler (as above) with 35 cycles at 94°C for 5 s, 56°C for 10 s and 68°C for 10 s. After addition of 3 μl loading dye (Amersham) the reactions were run on a Li-Cor Sequencer 4000L.

2.5Sequence analyses

Phylogenetic analysis of the sequences was performed by alignment of the partial isolate sequences to the according clone sequence and also to the EMBL database of 16S rRNA sequences. The software programs used were BestFit and FASTA from the GCG software package [25]. The computer-aided simulation of melting behavior and migration of the amplicons during TGGE was performed with the software Poland V1.0 [26,27]. This simulation was done with n=72°C without considering ionic strength of the buffer system or denaturing agents. The denaturing agents urea and formamide in the TGGE gel push the melting curves to colder temperatures of approximately n=40°C. The analyzed sequences can be found in the EMBL database by their accession numbers AJ132748–AJ132749 and AJ236889–AJ236893.

3Results and discussion

3.1Isolation of microorganisms from Drentse A soil

The uncultured bacteria in Drentse A soil were represented by the Bacillus-like ribotype DA001, which was previously detected in soil by whole-cell in situ hybridization as apparently active, vegetative rods [16]. Furthermore, we selected another Bacillus-like ribotype DA011, ribotype DA079, which was closely related to some uncultured Actinobacteria from German peat [15], and sequence DA101 that fits into the Verrucomicrobiales cluster [17]. Since the optimal growth conditions for the four bacteria containing the four ribotypes were unknown, various types of media with different carbon sources were selected. Since the predominant ribotype DA001 was closely related to B. benzoevorans, we applied three types of media containing the carbon sources benzoate or its derivatives m-hydroxybenzoate or methyl benzoate that have been used to grow B. benzoevorans. However, no fast-growing colonies or B. benzoevorans-like cell types as described by Pichinoty and Asselineau [20] were found. Colonies that grew on all the other media consisted of cell types of typical Bacillus appearance and endospore formation. Many agar plates were dominated by very rapidly growing B. cereus-like colonies showing a mycoides phenotype. Ultimately 659 isolates were obtained that were subsequently screened by analyzing their rDNA amplicons via TGGE.

3.2TGGE screen of the bacterial strains isolated from soil

In order to compare the ribotypes DA001, DA011, DA079 and DA101 with the cultured isolates, we applied the same screening method to both approaches. The cloned 16S rDNA inserts of the directly extracted soil DNA were amplified with the same primer pair used to analyze the isolates. Previous results have demonstrated that 16S rDNA cloning was a good approach to reveal the predominant ribotypes [15]. Comparing the cloned insert bands with TGGE fingerprints from directly extracted soil rRNA, matches were found with the most intense bands in the soil fingerprint [13]. The clone library contained 128 different 16S rDNA sequences and additionally 37 sequences identical to another. Of all 165 insert-containing clones, 42 clones matched the 15 most intense TGGE fingerprint bands (Fig. 1). This correlation indicated that both approaches detected the same predominant soil sequences. All remaining clones could not be affiliated to any bands [15]. They might have originated from maybe thousands of rare bacterial species to be expected in soil [28]. The four selected ribotypes DA001, DA011, DA079, DA101 accounted for 17 of 165 cloned 16S rRNA sequences. They also represented some of the strongest bands in the TGGE fingerprints from soil ribosomes [13]. Therefore, they appeared to be major contributors to the ribosome fraction in soil, and therefore should be considerably active.

Figure 1.

Sequence redundancy in the 16S rDNA clone library. The most abundant clones correlated with the most intense bands in TGGE fingerprints (black bars). Others matched some weaker bands (striped bars) or did not match at all (white bars). The stars indicate the four ribotypes of this study.

Amplicons obtained by PCR of the 16S rDNA of the 659 isolated strains were screened by TGGE for possible matches with either of the investigated ribotypes. Next to the isolates, markers were loaded on the TGGE gel consisting of equal amounts of PCR products generated from the cloned 16S rDNA of DA001, DA011, DA079 and DA101 (Fig. 2, lane M). Within the first TGGE screen we found that 54 of the 659 isolate signals might match one of the clone signals. If the PCR amplicons are all running next to each other, the visual resolution of differences is not much better than 1 mm. Therefore, we added a second TGGE screen for verification of the approximately matching isolates. Here, the PCR products from the clone and from one possibly matching isolate were loaded in the same slot (Fig. 2). In this way very small migration differences could also be detected since this would result in two bands. Fourteen of the 54 tested strains showed no identified migration difference and 10 were selected for subsequent partial sequence analysis. The other four, matching DA079 and DA101, were identified as Bacillus by endospore formation and consequently excluded from further analysis. The 14 matching strains preferentially appeared at neutral pH on mineral medium with acetate as carbon source (Table 2).

Figure 2.

Verification screen of isolates on TGGE. The ‘M’ indicates the marker lanes. Of 54 checked isolates six are presented on this gel. The marker consists of a mixture of four PCR products from DA001, DA079, DA101 and DA011, following the order from top to bottom. The other lanes contain PCR products of the isolates or a mixture of PCR products from an isolate and a cloned sequence.

Table 2.  Sequence analysis results of 10 isolates that showed exactly the same migration distance as the according clones
Isolated strainSequence similarity to cloneSequence similarity to next related species
  1. aIdentical sequences.

IDA113 no93.7% to DA00199.0% to Bacillus cereus
IDA216 ac98.2% to DA00196.4% to Bacillus megaterium
IDA234 csl76.4% to DA07996.7% to Bacillus megateriuma
IDA465 ac+p87.5% to DA10199.0% to Bacillus cereus
IDA533 dsm 185.2% to DA01196.3% to Arthrobacter nicotinovorans
IDA600 ac76.4% to DA07996.7% to Bacillus megateriuma
IDA624 ac95.7% to DA00194.8% to Bacillus cohnii
IDA627 ac94.8% to DA00194.8% to Bacillus benzoevorans
IDA629 ac97.6% to DA00194.6% to Bacillus benzoevorans
IDA647 dsm 1/1093.8% to DA00195.4% to Bacillus pseudomegaterium
ac, grown on mineral medium with acetic acid; csl, grown on mineral medium with corn steep liquor; dsm 1, grown on medium DSM 1; dsm 1/10, grown on 10-fold diluted medium DSM 1; no, grown on mineral medium without carbon source; +p, pasteurized soil suspension.

3.3Sequence analysis and phylogenetic assignment of isolates

Although matching on TGGE, the partial 16S rRNA sequences of the isolated bacterial strains were not identical to the according sequences of the cloned amplicons. All isolates showing equal migration distances on TGGE with the B. benzoevorans relatives DA001 could be identified as Bacillus strains by sequence analysis and microscopic detection of endospores. Here the best fit was found with isolate IDA216 showing 98.2% sequence similarity to clone DA001 and equal migration distances on TGGE (Table 2). Therefore, there is no indication that we succeeded in isolating the predominant bacteria. The sequence differences between clone DA001 and isolate IDA216 are unlikely to be due to PCR errors or reading errors during sequence analysis, because the base differences are all located in the highly variable regions of the 16S rRNA, which is the most likely location of base exchanges. Furthermore, the average sequence reading error for the applied protocol has been estimated to be less than 0.5% of all nucleotides [13]. In contrast, the isolates with rDNA amplicons showing equal migration distances following TGGE as that of DA011, DA079 and DA101, had approximately 90% (or less) sequence similarity to the respective cloned sequences. While DA079 and DA101 are located in the Actinobacteria and the Verrucomicrobiales cluster, the matching isolates were found to belong to Bacillus and Arthrobacter spp. These genera were already known to appear frequently during former cultivation approaches with Drentse A soil, and a culture collection of 120 different Arthrobacter strains from Drentse A grassland soils was previously set up (de Vrijer, unpublished data). However, the 16S rDNA cloning approach could not detect any 16S rDNA sequences of Arthrobacter. The conventional cultivation approaches apparently had a biased preference for Arthrobacter and also Bacillus strains from the Drentse A soil samples. Due to the strong appearance of Bacillus-strains one might expect that a strain collection of similar size to Arthrobacter is also achievable for the genus Bacillus.

3.4Simulation of the 16S rDNA melting process during TGGE

The apparently equal migration distance of completely different sequences on TGGE demanded further investigation, since this might be a major source of bias when the TGGE approach is used to screen 16S rDNA sequences retrieved by cloning or isolating bacterial strains. Some of the isolates giving identical TGGE signals to the four cloned sequences were quite similar to the corresponding clone (IDA216 to DA001) while others were completely different (IDA600 to DA079, Table 2). The theoretical migration of the amplicons was investigated by computer simulations. Following the computation algorithm of Poland [27], the melting behavior could be predicted by the nucleotide sequence of the amplicons generated with the primers GC968f and 1401r, which were used for TGGE. Here, the three completely different sequences from the Bacillus DA001, the Actinobacteria relative DA079 and the Verrucomicrobiales relative DA101 showed a quite similar melting behavior (Fig. 3a) and were consequently running quite close to each other (Fig. 2). The amplicon of DA001 melted first and stopped approximately 1.5 mm earlier than the amplicon of DA079, which stopped about 2 mm earlier than DA101 (Fig. 2). These sequences were clearly separated on TGGE. It is more difficult to distinguish sequences closely related to each other. The melting behavior of DA001 and the three closest isolate sequences IDA216 (98.2% sequence similarity to DA001), IDA624 (95.7%) and IDA629 (97.6%) was determined (Fig. 3b). These sequences showed almost identical melting curves and could not be separated on TGGE (Fig. 2). However, another closely related sequence, DA011 (96.8% sequence similarity to DA001), showed a completely different melting behavior and migrated much further into the gel (Fig. 2). Indeed, closely related sequences might migrate to the same position [29,30]. However, the distance between two bands on a gel is not proportional to the sequence difference as demonstrated by amplicon DA011 (Fig. 3b).

Figure 3.

Computer simulation of amplicon migration speeds in TGGE. A relative immobility of 0 is free mobility, while a relative immobility of 1 is no mobility at all. a: The three completely different sequences DA001, DA079 and DA101 show a quite similar melting behavior. b: The highly similar sequences DA001, IDA216, IDA624 and IDA629 show an almost identical melting behavior while the also closely related sequence DA011 is melting at much higher temperatures. c: The sequence DA079 and its TGGE matches show an identical melting behavior. DA101 and IDA465 show different, crossing melting curves, but on TGGE they end on the same position. The calculated melting curves are sequence-specific, the temperature constant ‘n’ is determined by electrophoresis-specific parameters.

The isolates with rDNA amplicons that matched those of clones DA079 and DA101 upon TGGE differed considerably with respect to their 16S rRNA sequences. The melting curve of the Bacillus sequence IDA234/IDA600 was almost perfectly identical to DA079, although the sequence similarity was only 76.4% (Table 2). Remarkably, sequences DA101 and IDA465 generated amplicons that both showed the same migration speed but a quite different melting behavior (Fig. 3c). In the beginning, the IDA465 amplicon appeared to slow down earlier than DA101, but due to a subsequent more dramatic pausing of DA101, IDA465 might close up again or even overtake. With increasing temperature, the amplicon DA101 showed a slightly faster migration, and finally both amplicons apparently ended at the same position (Fig. 2). This peculiar difference in melting behavior was proven by premature termination of the TGGE, at the time both sequences were indeed still separated (data not shown). In summary, slightly different amplicons might be separated with a resolution down to one base pair difference [31] or not at all (Fig. 3b). Completely different amplicons are likely to show different migration patterns following TGGE, but by accident they might end at exactly the same position (Fig. 3c).

3.5Conclusions

TGGE-supported screening of isolated strains was found to be a convenient and efficient way to process large numbers of colonies. This study clearly demonstrated that very different sequences from the same source samples might show the same migration speed during TGGE. Therefore, the assignment of isolates to matching bands of according environmental fingerprints without an additional check is not acceptable. This additional confirmation can be given by sequence comparison of the rDNA amplicons or by V6 probe hybridization, which does not require previous sequencing of the 16S rDNA fragments [32]. Cloning of environmental 16S rDNA yielded phylogenetic information that was highly correlated to the TGGE analysis of environmental 16S rRNA. The most abundant clones were identical in sequence to the most intense bands in environmental TGGE fingerprints. The predominant bacteria in Drentse A grassland soils remained uncultured. This might be due to hitherto unknown nutrition and growth requirements, low growth rates of the predominant soil bacteria, or the inhibition by other microorganisms during growth on agar.

Acknowledgements

This work was supported by a grant from the European Communities EC project ‘High Resolution Automated Microbial Identification’ (EC-HRAMI Project BIO2-CT94-3098). A.F. is supported by the EC Project BIO4-98-0168. We would like to thank the Dutch State Forestry Commission, who allowed us access to the nature reserve.

Ancillary