OUP user menu

Use of the T-RFLP technique to assess spatial and temporal changes in the bacterial community structure within an agricultural soil planted with transgenic and non-transgenic potato plants

Thomas Lukow, Peter F. Dunfield, Werner Liesack
DOI: http://dx.doi.org/10.1111/j.1574-6941.2000.tb00717.x 241-247 First published online: 1 June 2000


The aim of this study was to examine whether the terminal restriction fragment length polymorphism (T-RFLP) analysis represents an appropriate technique for monitoring highly diverse soil bacterial communities, i.e. to assess spatial and/or temporal effects on bacterial community structure. The T-RFLP method, a recently described fingerprinting technique, is based on terminal restriction fragment length polymorphisms between distinct small-subunit rRNA gene sequence types. This technique permits an automated quantification of the fluorescence signal intensities of the individual terminal restriction fragments (T-RFs) in a given community fingerprint pattern. The indigenous bacterial communities of three soil plots located within an agricultural field of 110 m2 were compared. The first site was planted with non-transgenic potato plants, while the other two were planted with transgenic GUS and Barnase/Barstar potato plants, respectively. Once prior to planting and three times after planting, seven parallel samples were taken from each of the three soil plots. The T-RFLP analysis resulted in very complex but highly reproducible community fingerprint patterns. The percentage abundance values of defined T-RFs were calculated for the seven parallel samples of the respective soil plot. A multivariate analysis of variance was used to test T-RFLP data sets for significant differences. The statistical treatments clearly revealed spatial and temporal effects, as well as space×time interaction effects, on the structural composition of the bacterial communities. T-RFs which showed the highest correlations to the discriminant factors were not those T-RFs which showed the largest single variations between the seven-sample means of individual plots. In summary, the T-RFLP technique, although a polymerase chain reaction-based method, proved to be a suitable technique for monitoring highly diverse soil microbial communities for changes over space and/or time.

  • Biomonitoring
  • Transgenic plant
  • Microbial community
  • Terminal restriction fragment length polymorphism analysis
  • Genetic fingerprint
  • Multivariate analysis of variance

1 Introduction

The assessment of spatial as well as temporal changes in the microbial community structure induced by biotic or abiotic factors is one of the major research topics in both applied and basic environmental microbiology [13]. In principle, changes in the composition of microbial communities can be assessed by cultivation-dependent methods, e.g. the community-level catabolic profiling based on the commercially available BiOLOG substrate utilisation system [4,5]. However, cultivation-dependent methods are mostly biased towards the selective enrichment of fast-growing bacteria adapted to high substrate concentrations [6], which often represent only a numerically minor fraction of the total microbial community [7,8]. Therefore, cultivation-independent molecular fingerprinting techniques, based on small-subunit (SSU) rRNA genes (rDNA) [912], have gained in popularity to address questions related to the diversity, structural composition and dynamics of microbial communities. The most common technique is denaturing gradient gel electrophoresis (DGGE) which is based on the electrophoretic separation of partial SSU rDNA fragments of the same length but different base pair composition within a linearly increasing gradient of denaturants. Similar separations can be obtained with temperature gradient gel electrophoresis (TGGE). One of the methodological advantages of both DGGE and TGGE (D/TGGE) is the possibility of directly correlating distinct rDNA bands with phylogenetic information [1316]. However, a major pitfall of the D/TGGE might be that profiles generated for highly complex microbial communities are often characterised by a few dominating bands and a diffuse background caused by a large number of unresolved fragments [8,17]. This hinders the correct qualitative and quantitative interpretation of the community fingerprint patterns obtained. Hence, only a few reports exist on the use of D/TGGE for the analysis of soil and rhizosphere bacterial communities [15,17,18]. One example is provided by Heuer et al. [15], who were able to overcome the limited resolution by polynucleotide probing of D/TGGE patterns.

However, we decided to follow an alternate approach and to evaluate in this study the potential of the terminal restriction fragment length polymorphism (T-RFLP) analysis [19] for biomonitoring of highly diverse soil bacterial communities. The general need for an appropriate technique to monitor microbial communities is stressed by the fact that the European Community plans to enact a regulation that in the future the release of transgenic systems will have to be accompanied by a risk assessment, i.e. by a thorough biomonitoring for possible effects on the indigenous biota.

Our investigation started with a thorough evaluation of the accuracy and reproducibility of T-RFLP analysis, as well as an assessment of the range of bacterial diversity detectable by direct analysis of total community DNA in relation to that fraction detectable by community-level catabolic BiOLOG profiling. As a second step, bacterial communities of three distinct soil plots located within the same agricultural field and planted with different plant variants, i.e. (I) transgenic Barnase/Barstar (tBB), (II) transgenic GUS (tGUS) and (III) non-transgenic (WT=wild-type) potato plants, were biomonitored. This was done to determine, in the format of a preliminary examination, whether T-RFLP analysis represents an appropriate molecular tool for biomonitoring possible spatial and/or temporal effects of transgenic systems on microbial community structure in soils. To obtain high statistical confidence for possible effects we analysed seven parallel samples for each of the three plots at a given time-point.

2 Materials and methods

2.1 Field trials

Field trials were conducted at the Max-Planck-Institut für Züchtungsforschung (Cologne, Germany) in 1998. The three soil plots under investigation were planted with different potato plant variants as follows: (I) tBB, (II) tGUS and (III) WT (Solanum tuberosum L. cv. ‘Bintje’). The total field trial covered an area of about 110 m2. The dimension of the field was 13.30×8.25 m. The dimension of each plot was 0.65×2.30 m. The three soil plots were separated by between 2.5 and 6 m. Seven plants of the same variant were grown in one plot. For each of the three plots seven soil cores were taken with a drill from a depth of 20 cm on 5 May (two days before planting), 24 June (elongation growth stage), 11 August (flowering stage) and 14 September 1998 (senescent plant stage). The soil cores taken from one plot and time-point were sampled in close proximity to the plants in this plot. Detectable root material was removed. Remaining root material did not have any effect on the T-RFLP-based results because a 413-bp terminal restriction fragment (T-RF) indicative of plastid SSU rDNA of potato plants was never observed in any of the community fingerprint patterns compared, while this T-RF represented one of the major fragments in community fingerprint patterns obtained from root-associated bacterial populations (data not shown). The samples were stored at −20°C until used.

2.2 Transgenic plants

The transgenic Barnase/Barstar potato plants carried two modified gene constructs. One gene construct consisted of a bacterial ribonuclease from Bacillus amyloliquefaciens, termed Barnase, coupled with the gst1 promoter [20]. Selective induction of the promoter after infection with various types of pathogenic, e.g. Phytophthora infestans, or symbiotic organisms should lead to a suicide of the infected cells and thus prevent spreading of the pathogen [21]. To minimise the detrimental effects of a potential background activity of the barnase gene in non-infected tissue, the gene encoding Barstar coupled with the constitutively expressed viral CaMV 35S promoter was also inserted into the genome of the transgenic plants [20]. The barstar gene inhibits Barnase synthesis and is also derived from B. amyloliquefaciens. Instead of the barnase/barstar gene construct the transgenic GUS potato plants carried a gus gene (uidA gene) coding for β-glucuronidase [21].

2.3 DNA extraction

Total community DNA was extracted from 1-g aliquots (fresh weight) of the soil samples by a procedure which combined elements of the protocols described by Ogram et al. [22] and Smalla et al. [23]. One ml of sodium phosphate buffer (0.1 M, pH 8.0) and 240 μl of a 10% SDS solution (w/v) were added to each soil sample. The samples were mixed thoroughly and incubated at 65°C for 10 min. Then the suspensions were mixed with 1 g of glass beads (∅=0.17–0.18 mm) and shaken at 2000 rpm (Dismembrator, B. Braun, Melsungen, Germany) for 1 min to lyse the bacterial cells. The lysate of each sample was clarified by centrifugation (20 800×g for 10 min at 4°C). The supernatants were recovered, transferred to new 2.0-ml reaction tubes and extracted three times with equal volumes of chloroform:isoamyl alcohol (24:1). After ethanol precipitation, the DNA pellets were resuspended in 200 μl of TE buffer (10 mM Tris, 1 mM EDTA, pH 8.0). For further purification, 200 mg of CsCl was added, mixed well and the suspensions were incubated at room temperature for 3 h. The CsCl was pelleted by centrifugation (20 800×g for 5 min at 20°C). The supernatants were transferred to new 2.0-ml reaction tubes and an isopropanol precipitation of the DNA was performed. The DNA pellets were resuspended in 100 μl of TE buffer and stored at −20°C. This procedure resulted in the extraction of high molecular mass DNA of approximately 20-kb size fragments which showed only a low degree of fragmentation [24].

2.4 Polymerase chain reaction (PCR) amplification

The bacterial SSU rDNA was PCR-amplified from total DNA using oligonucleotide primers which target a wide range of members of the domain Bacteria (27F [25], 1378R [26]). For T-RFLP analysis the 5′ primer (27F) was labelled with the dye 5-carboxyfluorescein. The reaction cocktail contained 1 μl of template DNA, 10 μl of 10×reaction buffer (PCR buffer II, PE Applied Biosystems, Weiterstadt, Germany), 1.5 mM of MgCl2, 200 μM of each deoxynucleoside triphosphate (AGS, Heidelberg, Germany), 0.3 μM of each primer 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 and 28 (total DNA extracted from BiOLOG plates) or 30 cycles (total DNA extracted from soil samples) consisting of denaturation at 94°C for 45 s, primer annealing at 48°C for 60 s and elongation at 72°C for 2 min. The final elongation step was extended to 8 min. Amplification was performed with 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 SSU rDNA amplicons (10 μl) were checked by electrophoresis on a 1% agarose gel and stained with ethidium bromide.

2.5 T-RFLP analysis

The SSU rDNA amplicons were purified using Qiaquick spin columns (Qiagen, Hilden, Germany) according to the instructions of the manufacturer. Aliquots of the purified amplicons (100 ng) were digested with the restriction endonuclease MspI (Promega, Mannheim, Germany). The reaction cocktails contained 2 to 8 μl of the purified PCR products, 1 μl of the appropriate incubation buffer supplied by the manufacturer (Promega) and 1 μl of the endonuclease (10 U). The digestions were performed with a total volume of 10 μl at 37°C for 3 h. Aliquots of the digests (2.5 μl) were mixed with 2.0 μl of formamide and 0.5 μl of an internal lane standard (GeneScan-1000 ROX, PE Applied Biosystems). The samples were denatured at 94°C for 2 min and then immediately placed on ice until loading onto the gel. Electrophoresis was carried out on a 12-cm 6% (w/v) polyacrylamide gel containing 8.3 M urea and 1×Tris–borate–EDTA buffer (89 mM Tris, 89 mM borate and 2 mM EDTA). The restricted SSU rDNA fragments were size-separated in relation to the internal lane standard in the GeneScan mode of an automated ABI DNA sequencer (model 373, PE Applied Biosystems) for 6 h using the following settings: 2500 V, 40 mA and 27 W. Only the fluorescently-labelled 5′-terminal restriction fragments were detected and further analysed. The length and the fluorescence signal intensity of each T-RF in a given community fingerprint pattern were automatically calculated by the ‘GeneScan Analysis Software’ (Version 2.1, PE Applied Biosystem).

2.6 Statistics

The percentage abundance (Ap) of each T-RF was calculated as Embedded Image in which ni represents the peak area of one distinct T-RF and N is the sum of all peak areas in a given T-RFLP pattern. In total, 84 different soil samples (seven samples×three soil plots×four time-points) were comparatively analysed by using the T-RFLP fingerprinting technique. Ap values were determined for all T-RFs detected in a size range between 50 and 700 bp for a given T-RFLP pattern. Some T-RFs were detected only in single T-RFLP patterns. To exclude these T-RFs from statistical analysis, only T-RFs were considered which had Ap≥1% in at least one of the four sampling time-points in all seven samples. This resulted in the consideration of in total 40 different T-RFs. Analyses of variance (ANOVA) were performed on each of the 40 peaks, with plot and time as grouping factors. Agreement with parametric assumptions (normality and homogeneity of variance) was examined using residual scatter plots, probability plots and skewness and kurtosis values. A total of 29 peaks were accepted as parametric (many after transformation by a logarithm or square root) and then analysed in a single multivariate analysis of variance (MANOVA). The remainder were analysed with nonparametric Kruskal–Wallis tests. Interpretation of the data followed two paths: (i) examination of global effects for the 29 variables in the MANOVA and (ii) examination of univariate analyses (ANOVA F-tests or Kruskal–Wallis tests) for all 40 variables. In the latter case Bonferroni adjustments were used to compensate for the total number of T-RFs and maintain the experiment-wide Type I error at P<0.05. Multiple comparisons are Bonferroni contrasts. All analyses were performed using SYSTAT 9 (SPSS, Chicago, IL, USA).

2.7 BiOLOG microtitre plates

Soil (1-g aliquots, fresh weight) from each of the seven sampling sites of the same soil plot were pooled and suspended in 70 ml of a sterile 0.85% NaCl solution (w/v). The suspension was thoroughly mixed and an aliquot was diluted 1:10 twice. The terminal dilution step was used to inoculate each of the 96 wells of one BiOLOG GN microtitre plate (Merlin, Bornheim-Hersel, Germany) with 150-μl aliquots. The plates were incubated at 25°C until wells became visibly violet (up to 7 days). For T-RFLP analysis of the bacterial consortia grown in the BiOLOG plates, the inocula of all wells of the same microtitre plate were pooled and the cells were pelleted by centrifugation (4200×g for 5 min at 4°C). The cell pellet was resuspended in 1 ml of sodium phosphate buffer. The suspension was used as starting material for the extraction of total DNA as described above. This treatment was carried out for each of the three different soil plots at all four sampling time-points.

3 Results and discussion

A prerequisite for a meaningful comparative analysis of T-RFLP patterns generated from different environmental samples is high accuracy in the determination of both size and signal intensity of individual T-RFs. The accuracy of the automated size determination has been demonstrated [19,27,28]. In our study, the length of defined T-RFs varied by not more than ±1 bp within a size range between 50 and 700 bp. T-RFLP analysis of a 10-fold dilution series of SSU rDNA amplicons of Escherichia coli showed that the signal intensity of an individual T-RF linearly correlates to the amount of fluorescently-labelled DNA, which means that the signal intensity of an individual peak directly corresponds to the percentage abundance of the respective T-RF within a SSU rDNA amplicon. The repeated examination of three replicates based on the same rhizosphere soil sample, but each time with separate extraction of total community DNA, PCR-mediated amplification of SSU rDNA and T-RFLP analysis showed that the percentage abundances of distinct T-RFs with peak areas ≥1% of the total peak area were highly reproducible. From these results it can be concluded that the T-RFLP technique generates community fingerprint patterns in a format which provides a means for the estimation of dissimilarities (or similarities) between bacterial diversity profiles obtained by this technique.

We assessed the range of diversity detectable by T-RFLP analysis in cultivation-independent approaches in relation to the diversity detectable by community-level catabolic BiOLOG profiling. All T-RFLP patterns generated from the pooled cell fractions of BiOLOG microtitre plates showed one dominant T-RF at 489 bp and never more than three different T-RFs (Fig. 1). Based on SSU rDNA sequences deposited in public databases, the 489-bp T-RF is mainly indicative of β- and γ-Proteobacteria. This correlates well with results obtained by Smalla et al. [6] which showed that TGGE community patterns generated from individual BiOLOG GN wells corresponded mainly to members of the γ-Proteobacteria. In contrast, each of the community fingerprint patterns obtained from the 84 soil samples showed a richness between 44 and 53 different T-RFs. This finding clearly documents that BiOLOG profiling only focuses on a minor fraction of the total bacterial community, while the cultivation-independent analysis covers a broad range of phylogenetic diversity which may also include noncultivable but predominant bacterial populations.


Comparison of 16S rDNA-based T-RFLP community fingerprint patterns obtained from the same rhizosphere soil sample of non-transgenic potato plants after cultivation on a BiOLOG GN microtitre plate (A) and by a cultivation-independent approach (B). Based on MANOVA the T-RFs with sizes of 83, 206, 291, 404 and 436 bp allowed the best separation between the seven-sample means of the T-RFLP patterns, while the 489-bp T-RF showed the single largest variances. Based on sequence information obtained from SSU rDNA databases, these T-RFs correspond to the following major phylogenetic groups: 83 and 206=Cytophaga/Bacteroides/Flavobacterium phylum, 291=Holophaga/Acidobacterium phylum, 404=α-Proteobacteria, 436=γ-Proteobacteria and 489=β- and γ-Proteobacteria.

For the MANOVA, 29 T-RFs were considered (Table 1). The MANOVA test statistics for each factor, i.e. plot, time and plot×time, all had P<0.001, indicating that highly significant differences in the T-RFLP community patterns were detectable over space and time. Based on the comparison of univariate tests for each of the 40 T-RFs, seven T-RFs varied significantly among plots, while 19 T-RFs varied significantly over time. This clearly indicated that, in our study, time effects seemed to be of more relevance for changes in the bacterial community structure than spatial effects. Most of these time effects appeared to be attributable to the community fingerprint patterns generated from the soil samples taken at time-point four, i.e. on 14 September 1998. The overall T-RFLP community pattern of time-point four clearly differed from those generated for the time-points one, two and three. In contrast, the overall T-RFLP community fingerprint patterns of the first three time-points showed only a low number of significant differences. The analysis of how many T-RFs varied with time in each of the three plots indicated that most of these differences were related to the soil plot planted with tGUS, i.e. 14 of the 40 T-RFs varied significantly. In contrast, the plots planted with tBB and WT potato plants showed only 4 and 0 significant differences over time, respectively.

View this table:

Number of T-RFs which differ significantly (P<0.05 after a Bonferroni adjustment for the total number of comparisons) over various treatment combinations

ComparisonnNumber of T-RFs which differ
Overall plot effects (space)407 (6)a
Overall time effects4019 (12)
Between plotsb:
tBB versus WT294
tBB versus tGUS290
WT versus tGUS293
Between time-pointsc:
1 versus 2293
1 versus 3290
1 versus 4298
2 versus 3291
2 versus 4296
3 versus 4298
Time within plots (i.e. the number of peaks which vary with time in each plot)
tBB404 (4)
WT400 (0)
tGUS4014 (10)
  • aThe numbers given in brackets were calculated based on those T-RFs which were parametric (n=29).

  • bPlots planted with different varieties of potato plants, i.e. transgenic Barnase/Bastar (tBB), transgenic GUS (tGUS) and non-transgenic (WT) plants.

  • cT-RFLP community fingerprint patterns generated for soil samples taken in 1998 on 5 May (1), 24 June (2), 11 August (3) and 14 September (4).

Because the individual T-RFs have to be treated as proportions rather than absolute values, a major change in one T-RF will by definition affect the relative proportion of the other T-RFs. However, three significant discriminant factors for time effects and two for plot effects were found in the MANOVA (Chi-squared, P<0.05). This means that the data varied in several dimensions, i.e. that all T-RFs did not vary together in a single correlated fashion. Based on these results, it can be concluded that both time and plot effects were attributable to changes in at least two distinct T-RF-defined bacterial populations, if not more. In this context, it should be mentioned that the Ap values of individual T-RFs were in the range between 0 and 14.4%±1.5 (mean±standard deviation) and that, of the T-RFs included in the MANOVA, the largest single variances between the seven-sample means of individual plots were detected for the 143- and 489-bp T-RFs. Ap values for these two T-RFs ranged from 3.1%±0.6 to 7.6%±1.8 and 10.0%±2.1 to 14.4%±1.5, respectively. However, these two T-RFs were not those which best correlated with the discriminant factors, i.e. which allowed the best separation of the various T-RFLP community patterns. Rather, the canonical loadings calculated in the MANOVA identified the 206- and 291-bp T-RFs as those which showed the highest correlations to the most significant discriminant factors for plot effects and also plot×time effects. These plot effects were for both T-RFs mainly due to significantly different abundances in tGUS and tBB plots compared to the WT plot, as shown for the 291-bp T-RF as an example (Fig. 2, see also Table 1). This is especially evident for time-point three. The plot×time effects for the 291-bp T-RF were based on significant changes of this T-RF in the tGUS and tBB plots, while its percentage abundance did not change in the WT plot. Other T-RFs which had high correlations to discriminant factors were those at 404 and 436 bp (time effects) and 83 bp (plot×time effects). The time effects determined for the 404- and 436-bp T-RFs corresponded to a large change of their percentage abundances at time-point four, which is shown for the 404-bp T-RF as an example (Fig. 2, see also Table 1). Each of these peaks which were most correlated to the discriminant factors had highly significant F-statistics (P<0.001). Thus, the T-RFs with sizes of 83, 206, 291, 404 and 436 bp and the bacterial groups which correspond to these T-RFs, can be considered ‘indicators’ of community changes over space and/or time (Fig. 1). Only one of these indicators, i.e. the 404-bp T-RF, was also detected by the BiOLOG-dependent cultivation approach. This highlights the importance of cultivation-independent methods for monitoring complex microbial communities for changes over space and/or time.


Percentage abundance for T-RFs with sizes of 291 and 404 bp in the three plots and four sampling times. Data are least-squared means from the MANOVA. For the 404-bp T-RF an exponential transform has been reversed. The 291-bp T-RF shows significant plot effects and plot×time effects. The 404-bp T-RF shows a significant time effect.

Taken together, the results of this study showed that the T-RFLP fingerprinting technique enables the detection of both spatial heterogeneities and changes over time in the structural composition of highly diverse bacterial communities. However, the analysis of replicate samples seems to be a crucial point for the generation of meaningful data sets for monitoring microbial communities. This is indicated by the disparity between the T-RFs which showed the largest single variances between the seven-sample means and those T-RFs which were most correlated to the discriminant factors. Fingerprinting techniques based on PCR amplification of SSU rDNA are more amenable to replication than conventional SSU rDNA clone libraries. This approach is, therefore, the current ‘state-of-the-art’ tool for rapidly investigating microbial diversity when a large number of samples has to be processed, such as for the assessment of the possible impact of the release of transgenic systems on the composition of indigenous microbial communities. This is true despite their inability to accurately assess community structure [29].

The decision whether the T-RFLP technique or DGGE is preferable will depend on the research goals and the complexity of the microbial community to be biomonitored. In studies on marine bacterioplankton communities the use of both techniques resulted in fragment patterns of similar complexity for all samples examined [30]. The D/TGGE permits the direct correlation of the banding patterns with phylogenetic information either by probing or by recovery of distinct SSU rDNA bands for further analysis. However, comparative assignment and determination of the relative abundance of individual fragments between D/TGGE community fingerprint patterns is possible only after a rather complex procedure, including staining, generation of a digitised photo using a CCD camera and image analysis [31]. In contrast, the T-RFLP technique enables a more objective comparison of community fingerprint patterns because of the automated quantification of both size and relative abundance of individual T-RFs [32]. In addition, this technique is easier to handle and more sensitive [30]. The D/TGGE might especially be favourable for the analysis of microbial communities of low or moderate complexity, while the T-RFLP technique seems to represent the superior tool for the biomonitoring of highly diverse communities. Nonetheless, for future tasks in environmental biomonitoring there exists a strong need for the development of novel fingerprinting techniques with an improved qualitative and quantitative resolution of microbial community structure. One of the future directions might be microarray-based techniques which enable a multiple phylogenetic and/or functional probing [33].


We thank Sonja Fleissner for excellent technical assistance. This work was supported by a grant from the Bundesministerium für Bildung, Wissenschaft, Forschung und Technologie (contract 0311199).


  1. [1]
  2. [2]
  3. [3]
  4. [4]
  5. [5]
  6. [6]
  7. [7]
  8. [8]
  9. [9]
  10. [10]
  11. [11]
  12. [12]
  13. [13]
  14. [14]
  15. [15]
  16. [16]
  17. [17]
  18. [18]
  19. [19]
  20. [20]
  21. [21]
  22. [22]
  23. [23]
  24. [24]
  25. [25]
  26. [26]
  27. [27]
  28. [28]
  29. [29]
  30. [30]
  31. [31]
  32. [32]
  33. [33]
View Abstract