For the interpretation of T-RFLP data sets, it is necessary to evaluate the reproducibility of all steps of specific PCR-T-RFLP assays, i.e., the PCR, digestion, and electrophoresis. In several investigations, a high reproducibility of T-RFLP analyses targeting SSU rRNA genes of Bacteria
has been shown, e.g., ≈10% absolute peak height variability (36
), 1.5% standard deviation of relative peak area (28
), or 7% standardized peak height variability (13
). Two studies involving PCR-T-RFLP analyses of archaeal SSU rDNA or rRNA found maximum standard deviations of 6% relative peak height (11
) and 6% relative peak area (27
) upon triplicate DNA and RNA extractions, respectively. In the present study, six replicate DNA extractions from three parallel soil slurry incubations and subsequent SSU rDNA-targeted T-RFLP analysis of the archaeal community yielded an average standard deviation of 1% relative amplicon frequency. The highest absolute standard deviation was 33.4% ± 2.3% relative amplicon frequency, and the highest relative standard deviation was 16% (e.g., 5.1% ± 0.8% relative amplicon frequency).
Only a few studies have looked at PCR-related influences on T-RFLP analysis so far. No significant differences in T-RFLP results were reported for a bacterial T-RFLP system between 25 and 35 PCR cycles (36
). To examine the effect of increasing PCR cycle numbers on the archaeal SSU rDNA-targeted T-RFLP system evaluated here, we generated amplicons from identical rice field soil template DNAs with increasing PCR cycle numbers. The relative abundance of T-RFs was stable within standard deviations of maximally 3% between 28 and 45 PCR cycles, and minor peaks did not become more abundant with increasing PCR cycles (data not shown). Therefore, this archaeal T-RFLP system appears to be unaffected by template reannealing. Such a PCR bias would cause less abundant PCR products to become relatively more abundant with increasing PCR cycle numbers (31
Furthermore, we evaluated the possible effects of selective amplification on fingerprinting results by varying the PCR annealing temperature. Different annealing temperatures may alter primer-binding kinetics in a template mixture, especially if primers with degenerate positions are used. The primer pair used for the amplification of archaeal SSU rRNA genes carries one degenerate position, while the mcrA-targeting primers carry nine degenerate nucleotides altogether. Archaeal SSU rDNA relative amplicon frequencies were virtually not affected by annealing temperature variations (Fig. ) within a range from 47.3 to 57°C (within normal standard deviation limits). In contrast, variation of the annealing temperature showed a dramatic effect on mcrA relative amplicon frequencies (Fig. ).
FIG. 1. Effect of various PCR annealing temperatures on the SSU rDNA-targeted (A) and mcrA-targeted (B) T-RFLP analysis of a soil archaeal community. Amplicons were generated from identical soil DNA templates with various PCR annealing temperatures and analyzed (more ...)
This PCR was tested in a range between 45.4 and 55°C (few or no products were obtained below or above these temperatures). Most significantly affected was the 468-bp T-RF, which accounted for almost 30% of the relative amplicon frequency at 45.4°C but was below the detection limit at 55°C. The 405- and 502-bp T-RFs first increased slightly and then decreased again in relative amplicon frequency with increasing annealing temperatures but always remained detectable. Others, especially the 143-, 237-, and 393-bp T-RFs, increased significantly in relative amplicon frequency with annealing temperature. We relate these findings to the highly degenerate mrcA primer set used.
It was shown earlier that different binding energies of primer permutations can lead to PCR selection of GC-rich templates (37
). Similarly, a reduction in T-RFLP-detected diversity upon an increase in the annealing temperature was found within a model bacterial community, although nondegenerate primers were used (9
). In contrast, Osborn and coworkers (36
) observed an increasing number of bacterial T-RFs at a higher annealing temperature but related this “apparent anomaly” to electrophoresis detection problems rather than a PCR artifact. Thus, the influence of the annealing temperature on the differential amplification of certain templates cannot be predicted and should be analyzed for each primer system individually.
Sequence conservation of functional marker genes in comparison to SSU rRNA genes is constrained by the genetic code and its variability, especially in the third codon position. Consequently, highly degenerate primers often have to be used for functional marker genes to cover a wide phylogenetic range. This contradicts the potential of PCR-based fingerprinting to obtain an unskewed picture of community structure, and relative community compositions should be inferred carefully.
Defined template mixtures.
The essential question in T-RFLP fingerprinting, how the relative abundance of amplicons (and thus of T-RFs) reflects template ratios, cannot be answered by the analysis of environmental DNA extracts of unknown composition. Therefore, we used artificial DNA template mixtures of defined gene content to evaluate the quantitative potential of archaeal T-RFLP analysis. Genomic DNA from four pure cultures of methanogens (M. bryantii
, M. concilii
, M. hungatei
, and M. jannaschii
) was used, because templates in environmental DNA extracts are also genomic and the flanking regions of target genes have been shown to influence amplification efficiency (17
). Moreover, genomic templates allow comparison of results obtained via T-RFLP analyses of identical mixtures targeting different marker genes, such as SSU ribosomal DNA (rDNA) and mcrA
Prior to the mixing experiments, the pure-culture DNAs were analyzed by both SSU rDNA- and mcrA-targeted PCR-T-RFLP. This control experiment was conducted to verify the sole appearance of T-RFs in silico predicted from sequence data. Results are summarized in Table , and the measured T-RFs of pure cultures were in good agreement with the predicted T-RFs.
Measured and predicted in silico T-RFs of SSU rDNA and mcrA/mrtA amplicons of pure cultures of methanogensa
For the defined mixtures, the SSU rRNA gene content of freshly diluted working solutions (≈2 ng μl−1
, Pico Green measurement) of the genomic DNAs was determined by real-time PCR (18
). The measurement was standardized with a 10-fold dilution series of M. jannaschii
genomic DNA (1.665 Mbp, two rrn
]), and was found to be reliable between at least 107
copies. The average standard deviation of measurements was below 2% of means. For each of the four methanogens, a series with 25%, 40%, 57.1%, and 78.6% SSU rDNA template ratio was prepared, while the remaining three species equally accounted for the residual percentage (25%, 20%, 14.3%, and 7.1% each, respectively). Three entire sets of defined template mixtures were prepared in completely independent mixing procedures (starting from freshly diluted DNA working solutions) to account for pipetting errors and to obtain statistically significant results. Immediately after preparation, the defined genomic DNA mixtures were analyzed by both SSU rDNA- and mcrA
-targeted T-RFLP analyses.
As an example for the T-RFLP analysis of defined template mixtures, electropherograms from a series with increasing M. bryantii SSU rDNA template ratios are shown in Fig. . An adequate representation of the defined template ratios was visually apparent in the electropherograms. Precise amplicon ratios were determined by inferring both the peak height and peak area of T-RFs. Summarized PCR-T-RFLP results of the three independent sets of defined template mixtures are shown in Fig. . Increasing template ratios were precisely reflected in the relative abundance of T-RFs for the four methanogens tested and in average in absolute correlation when peak height analysis was used (Fig. ). Peak area integration, on the other hand, resulted in a significant overrepresentation of M. hungatei and M. jannaschii relative amplicon frequencies (Fig. ). Both were characterized by long SSU rDNA T-RFs (395 and 789 bp, respectively), which migrated slowly during electrophoresis, resulting in broad, “fuzzy” T-RF peaks.
FIG. 2. Electropherograms of T-RFLP analysis of defined template mixtures of M. bryantii (MB), M. concilii (MX), M. hungatei (MP), and M. jannaschii (MC) pure-culture DNA. Shown is a series with increasing M. bryantii template ratios from 25 to 78.6% of SSU rRNA (more ...)
FIG. 3. Relationship between SSU rDNA template ratios and relative amplicon frequencies as determined by T-RFLP analysis of defined template mixtures. Relative amplicon frequencies (RAF) were determined by fluorescence peak height analysis (A) or peak area integration (more ...)
We assume that the automated area integration of T-RFs by the Genescan analysis software overestimates the area representation of such fuzzy peaks. This would explain the constant overrepresentation of M. jannaschii
and, to a lesser extent, also of M. hungatei
template ratios when the relative peak areas of T-RFs were compared. Inconsistencies in the Genescan software peak-calling routine that prevented the use of peak area data have been observed before (9
). Consequently, for the evaluation of relative signal intensities in archaeal T-RFLP analysis, peak height comparison should be used.
Defined SSU rDNA template mixtures for PCR and subsequent amplicon ratio analysis have been prepared earlier, but the results were inconsistent. Mostly pairwise mixtures of either genomic or amplicon standard DNAs have been prepared, and post-PCR amplicon ratios were either in good concordance with the prepared template ratios (14
) or indicated significantly skewed product ratios, depending on the templates or primers used (17
). Furthermore, single-species standards have been seeded to natural microbial communities in different ratios, and similar 1:1 relationships between spiked cell numbers (10
) or seeded DNA quantities (12
) and post-PCR T-RF ratios have been reported. However, in these analyses, it was not tested whether this correlation remained constant, if the seeding had been conducted with different strains, varying in rrn
copy number, either separately or simultaneously.
In addition to SSU rDNA as a marker gene, we also analyzed the reliability of mcrA
-targeted T-RFLP fingerprinting of methanogenic communities. This was of special interest because earlier investigations showed that the mcrA
genes of members of the Methanobacteriaceae
were of increased relative abundance in T-RFLP fingerprints compared to SSU rDNA-derived relative amplicon frequencies, while members of the Methanomicrobiaceae
were underrepresented (25
We compared the measured SSU rRNA gene ratios of defined template mixtures to the corresponding mcrA gene ratios. Correlation was low, and pronounced representational differences for the different species analyzed became apparent (Fig. ). Although increasing ratios were also reflected, the mcrA genes of M. hungatei were strongly underrepresented, while the combined mcrA and mrtA genes of M. bryantii and M. jannaschii were overrepresented. The mcrA genes of M. concilii were of a more consistent representation in both fingerprinting methods. Differences in the respective operon copy numbers may have contributed to the observed distortion.
Relationship between SSU rDNA and mcrA/mrtA gene relative amplicon frequencies (RAF) as determined by T-RFLP analysis of defined template mixtures. Lines indicate estimated fits.
Members of the Methanococcales
harbor two MCR isoenzymes, encoded by the mcr
). The α-subunit genes of both operons are detected by T-RFLP analysis (25
). On the other hand, M. concilii
and M. hungatei
are known to contain only one mcr
operon; a second isoenzyme has not been detected (42
). Given that all four methanogens contain the same rrn
operon copy number, this could explain the systematic overrepresentation of M. bryantii
and M. jannaschii
, but not the dramatic underrepresentation of M. hungatei
. Again, a possible explanation for these findings may be the highly degenerate primer set used for the mcrA
fingerprinting analyses. The reliability of post-PCR amplicon ratios has earlier been shown to depend on the primer pair used (45
), and degenerate primers should be avoided (37
). Unfortunately, the complete mcrA
sequences of M. concilii
, M. hungatei
, and M. bryantii
are not yet available, and therefore the precise primer binding sites cannot be determined. This would be necessary to test the model of PCR selection of mcrA/mrtA
templates with GC-rich primer target sites.
In this study, we showed that the SSU rDNA-targeted PCR system used is a precise and reproducible tool for the determination of archaeal template ratios as analyzed by T-RFLP. A broad range of different SSU rRNA gene ratios for a four-membered community was absolutely reflected in relative amplicon frequencies, as shown by fluorescence peak height comparison, which excludes PCR bias such as PCR selection or template reannealing. For natural communities, variations in neither PCR annealing temperature nor cycle number affected relative amplicon frequencies, which supports the quantitative potential of SSU rDNA-targeted T-RFLP fingerprinting for analysis of archaeal communities. In mcrA-targeted T-RFLP analysis, increasing template ratios were also detected, but representation varied significantly for the four different methanogens tested, probably due to preferential amplification of certain templates caused by the highly degenerate mcrA primer set.
Possible bias involved in PCR should be evaluated carefully for each primer system used if quantitative inferences of relative amplicon frequencies are intended. In fact, our study has shown that T-RFLP analysis can be used as a powerful tool to detect PCR bias of individual PCR assays by the precise quantification of the PCR product pool. But, as noted earlier (30
), because of potential PCR bias and largely unknown rrn
copy numbers in microbial genomes (15
), T-RFLP profiles do not reflect cell counts of natural microbial communities.