Evaluation of denaturing gradient gel electrophoresis in the detection of 16S rDNA sequence variation in rhizobia and methanotrophs


Corresponding author. Tel.: +33 3 80 63 32 98; Fax: +33 3 80 63 32 24; E-mail: vallaeys@dijon.inra.fr


The ability of denaturing gradient gel electrophoresis (DGGE) technique to resolve 16S rDNA products generated from two different collections of bacteria using universal 16S primers was investigated. Alignments of 16S rDNA sequences of known species of rhizobia and methanotrophs were performed in order to determine the genetic variations within a 200 bp product obtained with PCR primers which amplify the 16S rRNA encoding genes from Eubacteria. Theoretical DNA melting curves were obtained with the Melt87 program and found to correlate with the ability to resolve fragments by DGGE. In the case of the rhizobia, the inability of DGGE analysis to resolve the PCR products from closely related species was in accordance with the low polymorphism observed amongst the sequences in the amplified area. In the case of the methanotrophs, the PCR products were surprisingly difficult to resolve given the high degree of sequence polymorphism of the amplified area in some distantly related species. The difference in sequence divergence within the two groups members allowed therefore to scale the resolution ability of the DGGE technique.


Traditional microbiological methods for determining the diversity of microbial communities are biased by the need to culture isolates. There has been great interest recently in molecular techniques which can potentially measure community diversity without the need for culturing. One such technique, denaturing gradient gel electrophoresis (DGGE) can resolve DNA fragments of identical length but with different nucleotide sequences. DGGE can detect up to 95% of all possible single base substitutions amongst sequences of up to 1000 base pairs in length [1]. DGGE has recently been used to estimate the genetic diversity of microbial communities in natural habitats [2–6] and to infer the phylogenetic affiliation of community members [7]. In these studies, a region of suitable length of the 16S rRNA gene is amplified from DNA isolated directly from environmental samples using universal prokaryotic primers. The mixture of PCR products generated from community members is then resolved by DGGE. The complexity of the DGGE ‘fingerprint’ is taken as a measure of the community diversity. However, the elucidation of the diversity of community using universal 16S rDNA primers requires both polymorphism in the PCR product sequences and differential melting behaviour in gradient gels. Preliminary experiments systematically examining collections of known strains are therefore useful for optimizing PCR and DGGE conditions, and for establishing operational limitations of the method, before going to environmental samples. Few studies aimed to scale the resolution of the DGGE technique, and, in particular, its ability to resolve fragments amplified from phylogenetically related groups. Rhizobia and methanotrophs are two microbial groups playing an important role in nitrogen and carbon cycles, respectively. These two functional groups have been widely studied and numbers of 16S rDNA sequences are available from different sequence databases. Rhizobia and methanotrophs are both constituted by phylogenetically related bacteria. The two groups differ by the degree of sequence divergence within the group members allowing consequently to scale the resolution abilities of the DGGE technique.

2Material and methods

2.1Bacterial strains, culture conditions, and sample preparation for PCR

Bacterial strains used in this study and the GenBank accession number of their 16S rDNA sequences are presented in Table 1. Rhizobial cells were grown on TY slopes [8] and then resuspended and washed in distilled water. Cell suspensions were subjected to Proteinase K treatment as described by Lemanceau et al. [9] and then used directly as template DNA in the PCR reaction.

Table 1.  Strains used in this study and GenBank accession number of the 16S rDNA sequence
StrainGenBank accession numberReference
  1. aType strain of the species.

Rhizobium leguminosarum ATCC 10004aD12782[16]
R. tropici IIB CIAT 899aX67233[16]
R. tropici IIA CFN 299X67234[16]
R. galegae ATCC 43677aD12793[16]
Agrobacterium tumefaciens C58 [16]
A. tumefaciens LMG 196X67223[17]
R. giardiniiU86344[18]
Sinorhizobium meliloti ATCC 9930aX67222[16]
S. fredii USDA 205aX67231[16]
S. teranga ORS 1007 [16]
S. teranga ORS 22X68387[19]
R. loti NZP 2213aX67229[16]
R. etli CFN 42aREU28916[16]
R. gallicumU86343[18]
Methanotrophic sp. IMVL20845[20]
Methylobacterium organophilum XXM29028[20]
Methylobacterium extorquens AMIM29027[20]
Methanotrophic sp. ER2L20802[20]
Methylosinus trichosporium OB3bM29024[20]
Pseudomonas putida ATCC 12633 [21]

The methanotrophic strains and their purified DNA were obtained from R.S. Hanson. The DNA solutions were adjusted to a concentration of 0.05 μg per ml as determined by UV absorbance at OD 260 nm.

2.2PCR conditions

16S rDNA was amplified from the collection of rhizobial strains and methanotrophs using primers (p2 and p3) described by Muyzer et al. [2]. Primers positions correspond to position 341 and 534 in the 16S rDNA of E. coli, respectively. PCR reagents and conditions were those described by Muyzer et al. [2]. Ten μl of template DNA solution were used for the PCRs. Amplification was done with a Hybaid Omnigene thermocycler. The PCR amplified fragments were first visualized on a 4% (wt/vol) NuSieve agarose gel (FMC bioproducts) in TAE-buffer (20 mM Tris-acetate, 10 mM sodium acetate, 0.5 mM Na2 EDTA, pH 7.4). Ten μl of the PCR solution containing the amplified DNA fragments were mixed with 5 μl of loading solution and applied onto the gel [10]. Electrophoresis was performed for 2 h.

2.3DGGE conditions

DGGE was performed using the Protean II system (Bio-Rad) as previously described by Muyzer et al. [2]. The following conditions were used for the DGGE analysis. The fragments obtained from the rhizobial strains were separated using an 8% (wt/vol) acrylamide gel (acrylamide−N/N′-methylenebisacrylamide, 37/1) prepared in 0.5×TAE buffer with a denaturing gradient ranging from 25% to 55% (100% denaturant corresponds to 7 M urea and 40% (vol/vol) formamide deionized with AG501-X8 (Bio-Rad) resin). The gradient gel was cast using the Bio-Rad 385 gradient former and the 101 U peristaltic pump (Watson-Marlow). Polymerization was enhanced with TEMED (0.05% vol/vol) and ammonium persulfate (100 μl of a 10% (wt/vol) solution in 20 ml of gel mix). Electrophoresis for separation of PCR fragments was performed at 200 V and 60°C. The electrophoresis was run for 4 h for separation of products amplified from rhizobia and 5 h for products amplified from methanotrophs. After electrophoresis, the gels were incubated for 20 min in a 1 mg per liter ethidium bromide solution, rinsed for 10 min in distilled water and photographed under UV illumination with Polaroid Type 665 positive/negative films.


Alignments of the partial 16S rDNA sequences from the collection strains were obtained with the CLUSTAL V subroutine of the BISANCE program [11] using the default parameters.The mobility of DNA molecules of the sequences of interest in a denaturing gradient gel were predicted using the Melt87 computer program [12]. However, this approach was inapplicable to sequence data presenting undetermined nucleotides. Theoretical DNA melting curves obtained with the program and theoretical running time predicted by the program (TRAVEL subroutine) were then used to optimize DGGE running conditions.


3.1Sequence alignments and PCR amplification

Sequences of 16S rDNA fragments located between the two PCR primers p2 and p3 were available for all strains of rhizobia and methanotrophs studied except for Sinorhizobium teranga ORS 1007 and Agrobacterium tumefaciens C58. However, the corresponding sequence of another strain of the same species was known (Table 1). The number of nucleotide differences between the 200 bp long 16S rDNA fragments was determined from the multiple alignments of the partial 16S rDNA sequences of rhizobial strains (Table 2) and of methanotrophic strains (Table 3). Within the rhizobial strain collection, the number of differences ranged between 0 and 8 nucleotides (Table 2). Three sets of strains (set 1: R. leguminosarum and the R. tropici IIA and IIB; set 2: R. etli strain CFN42 and R. gallicum R602; set 3: the 3 Sinorhizobium species) presented no differences between their partial 16S rDNA sequences.

Table 2.  Number of differences between partial 16S rRNA sequences of rhizobia
 Number of different nucleotides
1. R. leguminosarum0           
2. R. tropici IIB00          
3. R. tropici IIA000         
4. R. galegae3330        
5. A. tumefaciens C5855560       
6. R. giardinii111250      
7. S. meliloti2225330     
8. S. fredii22253300    
9. S. teranga ORS 1007222533000   
10. R. loti5558663330  
11. R. etli CFN4211146233360 
12. R. gallicum000351222510
Table 3.  Number of differences between partial 16S rRNA sequences of metanotrohs
 Number of different nucleotides
  1. aThe number of differences includes deletion/insertion events.

1. IMV0    
2. OB3b5a0   
3. AMI57a59a0  
4. ER21114a60a  
5. XX1920a69a240

The methanotrophs presented much more sequence polymorphism in the region of the 16S rRNA gene under study. Numbers of differences between these sequences are presented for this group in Table 3 and ranged from 5 up to 69 nucleotides between some strains. However, insertion/deletion events were partially responsible for some high sequence divergences observed in this 16S rRNA gene region which in some cases resulted in a length polymorphism between the sequences. Moreover, sequence polymorphism was also observed in the area of the primers leading to low PCR yield for some of the methanotrophic strains. Methylobacterium extorquens AMI gave an extremely low yield of amplification, presumably because of five base differences between the priming site and the primer sequence (and one difference at the 3′ end of both priming sites).

3.2Theoretical melting curves

The theoretical melting curves, predicted by the Melt87 program [12] were identical for the strains with identical rDNA partial sequences. Sequence divergence of at least one base pair resulted in theoretical melting curves presenting slightly different shapes. A theoretical range of temperatures and migration time for which denaturation of two double stranded DNA molecules would occur separately could be defined from such analysis. As denaturation temperature is equivalent to a specified concentration of denaturant in a gradient gel [12], such results allow theoretically the optimization of the resolution of the corresponding PCR products. However, the theoretical analysis of the melting behaviour of the partial 16S rDNA sequences of interest suggested that the number of differences which could be detected was limited: increasing number of sequence differences as observed in the rhizobial group led to obtaining more distinct theoretical denaturation curves. However, when a high number of differences was observed between two sequences, shapes of the corresponding denaturation curves could be really similar as shown in Fig. 1. In the case of the methanotrophs, another difficulty was introduced by DNA fragments presenting at the same time different sequences and different lengths, as both parameters have an influence on the migration of the molecules in the gradient gel.

Figure 1.

Theoretical melting curves obtained using the Melt87 program (Lerman and Silverstein, 1987) for prediction and optimization of the denaturant gradient electrophoresis for two rhizobia (A) R. fredii, (B) R. galegae and one methanotroph (C) strain IMV.

3.3DGGE analysis of PCR amplified rDNA fragments

The results of the DGGE analysis of the PCR amplified 16S rDNA fragments are presented in Fig. 2. As expected, PCR products from the rhizobial strains which did not have any sequence difference in the 16S rDNA region amplified migrated to the same point in the denaturing gradient gel. On the other hand, strains which had at least one or more nucleotide differences in the amplified area were resolved in the denaturing gradient gel with the exception of R. etli CFN42 and R. gallicum R602, for which the G/C substitution between these two molecules was not detected by DGGE analysis (Fig. 2A, lanes 12 and 13). In any case, the DGGE results were in accordance with the results predicted by the theoretical melting curves and sequence data analysis. In addition, a correlation was observed between the number of differences present between two sequences and the difference in distances of migration of the corresponding PCR products in the denaturing gradient gel.

Figure 2.

Results of DGGE analysis for A: rhizobia (1: R. leguminosarum; 2: R. tropici IIB; 3: R. tropici IIA; 4: R. galegae; 5: A. tumefaciens; 6: R. giardinii H152; 7: R. giardinii H152; 8: R. meliloti; 9: R. fredii; 10: S. teranga; 11: R. loti; 12: R. etli; 13: R. gallicum R602), B: methanotrophs and control strains (1: Methylotrophic sp. IMV; 2: R. meliloti; 3: R. fredii; 4: Methylobacterium organophilum XX; 5: Methylobacterium extorquens AM1; 6: Methylotrophic sp. ER2; 7: Pseudomonas putida; 8: Methylosinus trichosporium OB3B).

On the contrary, the PCR products obtained from the methanotrophs were difficult to resolve and required a narrow denaturation gradient which was surprising considering the number of differences revealed by the sequence data analysis. This was particularly true for the PCR product from the Methylobacterium extorquens AMI strain which, although differing in length and base composition from the PCR product of Methylobacterium organophilum XX strain, presented similar migration behaviour in the DGGE gel (Fig. 2B, lanes 4 and 5). There was no correlation between the number of sequence differences and the distance of migration in the DGGE gel prior to denaturation.


This work clearly demonstrates the limits of the DGGE analysis in the measurement of the global diversity within a complex microbial population when using universal primers for amplification of small fragments of the 16S rRNA gene. The number of fragments which could be visualized on a DGGE gel may in some cases underestimate the actual diversity of a microbial community and should be considered as a lower limit of an estimation of the total number of bacterial species present. In this work, we showed that within the rhizobial group, some related strains of the genus Rhizobium such as R. leguminosarum and R. tropici, but also the Sinorhizobia (S. meliloti, S. teranga, S. fredii) could not be separated. This was also true when too many differences in base composition were present between two sequences, as shown for the methanotrophs. In addition, the migration in a denaturing gel was difficult to predict considering the number of undetermined nucleotides in 16S rDNA sequences of the methanotrophs used in this study and migration conditions could not be systematically optimized.

The amplification of parts of the 16S rRNA gene using universal prokaryotic primers consequently underestimates the biodiversity within some bacterial groups. The amplification of larger fragments of the 16S rRNA gene could avoid such results. Other conserved areas have been published at both extremities of the 16S rRNA gene [13]. However, the DGGE technique can not resolve DNA fragments of over about 1000 bp in size [1]. It therefore appears tempting to design primers for each specific group of interest in an appropriate area of the 16S rRNA gene, allowing an optimal discrimination of the group members. In addition, DGGE patterns obtained by loading 16S rRNA genes amplified directly from total soil DNA may reflect the relative abundance of the community members. Such an approach will therefore fail to reflect the diversity of a particular group, particularly if it represents a low proportion of the soil microflora. Recently, Yanagi and Yamasato [14], described 16S rDNA regions presenting a high sequence differentiation amongst the rhizobial species which could be used to discriminate rhizobial strains by designing primers flanking this highly variable area. We therefore tried to define two short sequences flanking this variable area of the 16S rDNA gene, distant by approximately 200 bp and conserved amongst all the rhizobial species. A set of primers (one specific to the rhizobial group and one universal primer carrying the GC clamp) was successfully designed at the position corresponding to positions 811 and 1031 according to the numbering of the 16S rDNA sequence of E. coli, respectively (unpublished results). However, the wobbles which had to be introduced to obtain PCR products from all the rhizobial strains resulted in multiple bands obtained from each strain in the DGGE gel (results not shown).

Relative abundance of the microbial species among soil community members is only one of the bias to be expected while using universal primers. In this work, we also showed that yields of amplification could be dramatically different between different strains even within a phylogenetically related microbial group. In the case of methanotrophs, differences of one to ten in the PCR yield could be correlated to the number of dissimilarities in the priming site sequences with the previously described primer set [2]. Other authors demonstrated that the genome size and the number of copies of the 16S rRNA genes were also affecting the yield of amplification of partial 16S rDNA fragments leading to different quantities of PCR products from different species in a mixed bacterial community [15]. Preferential amplification phenomena during the PCR step may therefore interfere with the precision of the DGGE results, when this technique is used to measure the bacterial diversity within a complex community.

DGGE appears a powerful tool for the study of the diversity of microbial communities. However, when applied to the study of the diversity of phylogenetic groups of bacteria, caution should be taken while elaborating the DGGE primer set. Using of universal primers should be avoided. However, in most cases, the choice for conserved and specific sequences within a particular phylogenetic group is limited. The conditions for separation of the PCR products should then be carefully optimized. The software to predict melting behaviour would be really helpful in this regard.

In conclusion, our results suggest that when working with environmental communities of interest, caution should be taken when preparing for and interpreting DGGE analyses. In any case, such studies should be preceded by careful and systematic preliminary optimization using standard strains.


We thank R.S. Hanson from the University of Minesota (USA) for providing us methanotrophic DNA and strains, Richard Hamelin and Laurent Puig from Institut Curie (Paris, France) for their help with the Melt87 program.