| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
a Department of Crop and Soil Sciences, Cornell Univ. Ithaca, NY 14853
* Corresponding author (jet25{at}cornell.edu).
 |
ABSTRACT |
|---|
 TOPNOTES ABSTRACT INTRODUCTIONTHE TERMINAL RESTRICTION...APPLICATIONS OF THE METHOD...CONCLUSIONSREFERENCES |
|---|
Abbreviations: AMMI, additive main effects with multiplicative interaction model • bp, base pairs • CA, corresponding analysis • detrended correspondence analysis • ITS, internal transcribed spacer • OTU, operational taxonomic unit • PCA, principal component analysis • PCR, polymerase chain reaction • RA, redundancy analysis • RDPII, Ribosomal Database Project II • RE, restriction endonuclease • rRNA, ribosomal ribonucleic acid • SIP, stable isotope probing • T-RF, terminal restriction fragment • T-RFLP, terminal restriction fragment length polymorphism.
|
INTRODUCTION |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONTHE TERMINAL RESTRICTION...APPLICATIONS OF THE METHOD...CONCLUSIONSREFERENCES |
|---|
In T-RFLP, fluorescent end-labeling of either the 5' or 3' PCR primer allows length separation and sizing of terminal restriction fragments after PCR products have been hydrolyzed with a selected restriction endonuclease (RE), thus producing a molecular "fingerprint" characteristic of the soil community analyzed. Liu et al. (1997) first described the T-RFLP method as it is practiced most commonly today. The potential of the technique for analyzing diverse microbial communities and how the method evolved from earlier work by Cancilla et al. (1992) and Avaniss-Aghajani et al. (1994, 1996) were discussed by Marsh (1999). Variations of the fluorescent end-labeling approach were also published by Bruce (1997) and Clement et al. (1998), but the T-RFLP method as described by Liu et al. (1997) has been most widely adopted for studying soil microbial communities.
The use of T-RFLP for analyzing microbial populations in biofilms (Wuertz et al., 2004), water (Dorigo et al., 2005), and agricultural (Tiedje et al., 1999; Bruce and Hughes, 2000; Arias et al., 2005) and forest soils (Leckie, 2005); for specific groups of organisms, such as fungi (Kennedy and Clipson, 2003; Anderson and Cairney, 2004; Edel-Hermann et al., 2004; Avis et al., 2006), rhizobia (Thies et al., 2001) and other microbial inoculants (van Elsas et al., 2003); or for specific processes, such as monitoring the progress of bioremediation (Mills et al., 2003), have been reviewed recently. Marsh (2005) provided detailed protocols for performing T-RFLP analysis.
The T-RFLP approach has been used to characterize microbial communities in many different environments. Those of most interest to soil scientists include agricultural (Buckley and Schmidt, 2001; Blackwood and Paul, 2003; Devare et al., 2004; Culman et al., 2006), forest (Hackl et al., 2004; Noguez et al., 2005), alpine (Costello and Schmidt, 2006) and polluted soils (Turpeinen et al., 2004; Paul et al., 2006), submerged rice soils (Ramakrishnan et al., 2000; Wu et al., 2006) and sediments (Sjoling et al., 2005; Pereira et al., 2006), soil crusts (Redfield et al., 2002; Yeager et al., 2004), peatlands (Kotsyurbenko et al., 2004; Cadillo-Quiroz et al., 2006), compost (Tiquia, 2005), in and on plant structures (Berg et al., 2005; Reiter and Sessitsch, 2006), and termite (Donovan et al., 2004; Shinzato et al., 2005) and earthworm digestive tracts (Egert et al., 2004).
The growth in the T-RFLP literature has been astounding in recent years, climbing from two refereed journal articles in 1997 to 139 during 2006 (Fig. 1 ). Currently, there is a total of 525 peer-reviewed articles indexed in the Web of Science (accessible through the ISI Web of Knowledge) in which the T-RFLP technique is either used for microbial community analysis or discussed at length. Of these publications, 31% describe the analysis of microbial populations in soils, sediments, the rhizosphere, or the phyllosphere (Fig. 2 ). Clearly, T-RFLP analysis is readily applied to the study of microbial populations in a wide range of environments.
|
|
|
THE TERMINAL RESTRICTION FRAGMENT LENGTH POLYMORPHISM METHOD |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONTHE TERMINAL RESTRICTION...APPLICATIONS OF THE METHOD...CONCLUSIONSREFERENCES |
|---|
Selected gene targets are then amplified from the sample DNA (or cDNA) by PCR (for methodological details see Marsh, 2005). The PCR reactions are prepared using primers where either the forward or the reverse primer is labeled with a fluorophore. Several different fluorophores have been used successfully, including HEX, FAM, and ROX dye chemistries. Fluorophores for primer end-labeling should be chosen based on the type of automated sequencer that will be used for fragment sizing, as recommended by the manufacturer. For any given gene target, PCR reaction conditions will need to be optimized. Suggestions for optimizing reactions are given in Boleda et al. (1996), Ishii and Fukui (2001), and Marsh (2005). Reaction products are typically passed through a PCR clean-up kit, such as those available from Qiagen (Valencia, CA) or Promega Corp. (Madison, WI)..
When highly conserved DNA (or cDNA) sequences are amplified, the resulting PCR products will be of a similar size. To separate them and obtain a fingerprint, the PCR products are hydrolyzed by one or several REs in separate reactions. Amplified DNA from different organisms that contain different restriction sites will yield terminally labeled fragments of different sizes (Fig. 3 ). These terminal restriction fragments (T-RFs) are then sized on an automated DNA sequencer using either gel-based or capillary methods, thus yielding a fingerprint that is characteristic of the community from which the DNA was extracted (Fig. 4 ). To reliably compare T-RFLP fingerprints between different samples, it is imperative that concentrations of (i) template DNA before PCR amplification, (ii) PCR-amplified DNA before restriction digestion, and (iii) hydrolyzed PCR products before sizing the terminally labeled fragments are standardized between the samples to be compared.
|
|
For sizing terminal fragments using a capillary sequencer, the hydrolyzed DNA must be desalted by either precipitating it in ethanol or purifying it using a Performa DTR Edge Plate (Edge BioSystems, Gaithersburg, MD). The Performa DTR Edge Plate allows 96 samples to be desalted simultaneously and can also be fitted into a speed vacuum where the samples can be evaporated completely. Just before sizing, samples are reconstituted in loading buffer (deionized formamide) along with the manufacturer's recommended concentration of a size standard, such as MM1000 (Bioventures Inc., Murfreesboro, TN) or Liz 500 (Applied Biosystems, Foster City, CA). The molecular size standard added must be labeled with a fluorophore that is different from that used to label the amplified sample DNA. In this way, samples in each well are sized precisely against a standard marker, thus allowing, in theory, terminal fragment sizes to be reported to within ±0.5 to 1.0 base pairs (bp).
For sizing on a polyacrylamide gel, sample restriction digests are used directly. A standardized concentration of each restriction digest is mixed with an appropriate molecular size standard, such as TAMRA 2500 (Applied Biosystems) and deionized formamide. The mixtures are then denatured by heating before loading into the gel. For the ABI 373A (Applied Biosytems) or similar systems, a 36-cm, 6% denaturing polyacrylamide gel is used for T-RF sizing (Marsh, 2005).
Output from automated sequencers is in the form of an electropherogram, with peaks representing fluorescently labeled T-RFs detected over time in relation to the size standard. The duration and intensity of the fluorescent signal from T-RFs is reflected in the area and height of each peak detected, respectively (Fig. 4). Software specific to each sequencing unit collects data from each run. GeneScan software is used in conjunction with the ABI 373A and data output must be imported into the Genotyper software package to set peak height detection thresholds, align peaks, and set the width for assigning a peak to a size bin. The ABI 3730 capillary sequencer operates a completely different software package, GeneMapper v3.5 (Applied Biosystems), which performs the functions of both GeneScan and Genotyper. Either data collection program provides researchers with several algorithms for sizing sample fragments by comparing their mobility with that of the size standard. The Local Southern Method is used most commonly for determining sample fragment lengths.
Once data are processed and fragment lengths assigned, the data set is typically imported into a spreadsheet program, such as Microsoft Excel (Microsoft Corp., Redmond, WA). In the spreadsheet, sample identifiers can be added and presence/absence (1, 0) matrices developed. Other manipulations, such as matrix inversions, can also be performed. The T-RFs for each sample run should be closely examined and the entire run evaluated for the average number of T-RFs detected per sample and the number of T-RFs contained in the various size classes. Fingerprints of complex communities from environmental samples with overall low numbers of terminal fragments are probably compromised and should be rerun. Individual lanes should also be evaluated for consistency and completeness. Details for performing these steps to prepare T-RFLP for statistical analysis are given in Blackwood et al. (2003a, 2003b), Kim and Marsh (2004), and Abdo et al. (2006). Presence/absence matrices can be exported to a variety of multivariate statistical analysis software programs as desired, such as Canoco (Microcomputer Power, Ithaca, NY), MATMODEL (Microcomputer Power), or Bionumerics (Applied Maths, Sint-Martens-Latem, Belgium). In addition to multivariate analyses, the on-line analysis page in the RDPII, Release 8 (Cole et al., 2003) allows similarity matrices between samples in a data set to be developed and provides the opportunity to compare fragment lengths of interest with database sequences subjected to an in silico digest using the same RE used for sample analysis. This latter analysis should be approached with caution, however, since T-RF lengths reported may be as much as 7 bp different from the true length of a given T-RF (see below).
Target Genes for Polymerase Chain Reaction
While the small-subunit ribosomal RNA (rRNA) genes are by far the more popular markers used for T-RFLP analysis, virtually any gene for which sequence information is available and that is sufficiently polymorphic (i.e., contains restriction sites that will yield terminal fragments of varying size) can be targeted. For 16S rRNA genes, a number of primer sequences have been published that are complementary to highly conserved sequences of the Bacteria or the Archaea or are conserved among specific subgroups within these domains, such as the alpha- or beta-Proteobacteria (Table 1). While the 16S rRNA genes are the more commonly amplified DNA targets used for bacterial and archael community comparisons, the 18S rRNA genes are less frequently used for characterizing eukaryotic communities, such as fungi. The 18S rRNA genes lack the sequence variation across the major fungal taxa needed for differentiating clearly between them when using the T-RFLP approach, thus the use of this gene for community comparisons has been limited. Instead, the internal transcribed spacer (ITS) region (Martin and Rygiewicz, 2005) or the nrDNA ITS region (Edwards and Turco, 2005) have been targeted more commonly for comparisons of fungal community composition between samples.
|
Several online tools are available for designing primers for use in T-RFLP experiments. These tools can be found at the National Center for Biotechnology Information (http://www.ncbi.nlm.nih.gov/, verified 18 Jan. 2007), probeBase (http://www.microbial-ecology.de/probebase/, verified 18 Jan. 2007), and the RDPII (http://rdp.cme.msu.edu/, verified 18 Jan. 2007) websites. Von Wintzingerode et al. (1997) discussed the more important points to consider in primer design, including matching the melting temperature for the forward and reverse primers and avoiding secondary structure, among other considerations.
The ability to target a wide range of functional genes or phylogenetic groups lends power to the T-RFLP approach and makes it an appealing precursor to other, more demanding protocols such as stable isotope probing (SIP) or standard cloning and sequencing.
Data Analysis: Separating Signals from the Noise
In their analysis of a simple bioreactor community, Liu et al. (1997) targeted the 16S rRNA gene for T-RFLP profiling. At the same time, they analyzed the potential for the T-RFLP technique to characterize more complex communities. With the RDPII (Maidak et al., 1999) well underway, they were able to perform in silico digests of database sequences and analyze the potential number of terminal fragments that might result from the use of a variety of REs. They found that HhaI, MspI, and RsaI yielded the best results, that is, a greater number of potential T-RFs than the other REs tested (Marsh et al., 2000). During the first several years of experimenting with the method, one exciting possibility was the potential to link the data generated by use of the T-RFLP technique to the rRNA gene sequence database and therefore provide both a rapid community analysis method and a means to potentially identify influential members of microbial communities in environmental samples (Marsh, 1999). To this end, investigators in several laboratories developed programs that could be used to compare fragment length data between community profiles and to existing sequence databases to make tentative phylogenetic placements (Marsh et al., 2000; Kent et al., 2003; Grant and Ogilvie, 2004; Matsumoto et al., 2005; Smith et al., 2005; Nakano et al., 2006). By 2003, however, it was becoming clear that variations in the base composition, particularly the purine content of terminal fragments (Kaplan and Kitts, 2003), and differences in the mobility of fragments labeled with different fluorophores may lead to sizing inaccuracies of up to 7 bp (Marsh, 2005). In light of this, trying to make phylogenetic placements for specific T-RFs from complex communities based on their length alone should be approached with caution. Some studies have shown the potential for a strong tentative identification of specific T-RFs by using T-RFLP in conjunction with group-specific primers and site-specific sequence databases. This approach has been successfully applied in studies by Rich et al. (2003), who worked with denitrifying bacterial communities in meadow and forest soils, and by Cadillo-Quiroz et al. (2006), who worked with methanogen communities in acidic peat bogs. In the latter study, amplicons derived from specific, methanogenic archael clones were digested and analyzed on a capillary sequencer for T-RF sizes that were then compared with T-RF sizes found in T-RFLP profiles from the bog samples. Since the diversity of methanogens in the bog was significantly less than the whole-community diversity, specific T-RFs could be assigned to the broad taxa present at the site with confidence (see Fig. 4C and 4D). In the majority of studies, T-RFLP data are analyzed statistically and phylogenetic placements are not attempted.
A variety of multivariate statistical approaches have been used to analyze T-RFLP data and derive meaning from community fingerprint patterns in relation to experimental treatments or naturally occurring environmental gradients (e.g., Dollhopf et al., 2001; Blackwood et al., 2003a, 2003b; Rees et al., 2004; Hartmann et al., 2005; Wang et al., 2004; Abdo et al., 2006; Culman et al., 2006). The T-RFLP data analysis usually involves one of two statistical approaches: ordination or classification. Ordination involves extracting a small number of dominant patterns (axes) from complex relationships and the output is commonly represented as a two-dimensional scatterplot. Classification group samples based on similarity and common output from these analyses are tree-like dendrograms, reflecting how the data cluster in relation to one another. Both approaches provide ways to summarize T-RFLP data for exploratory purposes rather than to test specific hypotheses (e.g., that there is a significant difference between two microbial communities). Final results of community comparisons are often presented as cluster diagrams with branch lengths marked and bootstrap values given for each branch for classification approaches or as plots in which compared communities are aligned along two dimensions in a scatterplot representing the key results of multidimensional analyses for ordination methods.
Ordination approaches, or multivariate statistical approaches, have been most frequently used for analyzing T-RFLP data. The use of principal component analysis (PCA, Dollhopf et al., 2001), redundancy analysis (RA, Blackwood et al., 2003a, 2003b), additive main effects and multiplicative interaction (AMMI) model (Culman et al., 2006), nonmetric multidimensional scaling or multidimensional scaling (NMS, Rees et al., 2004; Osborne et al., 2006), correspondence analysis (CA or reciprocal averaging), and detrended correspondence analysis (DCA) have all been used. Although the analysis of T-RFLP data has developed considerably during the last decade, there remains a lack of consensus about which statistical analyses offer the best means for finding trends in these data. The more common ordination methods used in the literature for analyzing T-RFLP data have been evaluated by Culman et al. (unpublished data, 2006), who compared (i) PCA, (ii) CA, (iii) DCA, (iv) NMS using the Sørensen (Bray-Curtis), Jaccard, and Euclidean distance measures, and (v) the AMMI model.
The AMMI model, also known as "doubly centered PCA," has only recently been applied to T-RFLP data analysis (Culman et al., 2006). The AMMI model uses ANOVA to first partition the variation into main effects (T-RFs and treatments) and interactions, and then applies PCA to the interactions to create interaction principal components (Gauch, 1992). Hence, instead of examining the overall variability of the data, the AMMI analysis focuses on the differential responses of T-RFs to treatments, a more fruitful approach for analyzing T-RFLP data where a treatment structure can be identified. Distance-based RA (Legendre and Anderson, 1999) is a useful extension of RA that has some of the advantages of AMMI and also allows for conducting ANOVA-like tests.
Culman et al. (unpublished data, 2006) analyzed seven different data sets obtained from a wide variety of environments and found that AMMI and DCA were the more discriminatory approaches and recommended them as the two "methods of choice" for T-RFLP community analysis. The AMMI analysis yields an ANOVA table that specifies the amount of variation in each axis, and thus provides valuable insight to guide researchers in data interpretation. The DCA method yielded the most discriminatory analysis overall and performed well even with uncommonly heterogeneous data. The NMS analysis with Sørensen and Jaccard distance measures was recommended only for use with very heterogeneous data. The CA or NMS methods with the Euclidean distance measure performed very poorly and had no advantages over the other methods for data analysis.
Research reports vary considerably in the type of data used for statistical analysis. Culman et al. (unpublished data, 2006) found that the data input type: binary (presence/absence), peak heights, peak areas, relativized peak heights, or relativized peak areas generally did not affect the results of the ordinations overall. Binary data were less prone to variable results than peak height or area, making them a more robust measure than relativized peak height or relativized peak area. Relativized peak height generally had the greatest interaction signal to interaction noise ratio of all three data types and thus they recommended the use of binary data and relativized peak height for analysis, rather than relativized peak area. Investigators should carefully consider experimental aims and the strengths and weaknesses of the different statistical models before deciding on the best model for use with their system.
Potential Pitfalls and Sources of Bias in the Method
While T-RFLP has many strengths, particularly the high sample throughput needed for broad-scale community comparisons, it has many potential pitfalls as well. Some of the pitfalls are those encountered in all methods where DNA or RNA is extracted from environmental samples and analyzed subsequently by PCR (von Wintzingerode et al., 1997), such as variation in cell lysis efficiency between different organisms during nucleic acid extraction, variation in nucleic acid recovery between different soil types, and coextraction of contaminants such as humic acids that may interfere with downstream PCR. Extraction methods that include a bead-beating step tend to shear DNA. Amplification of fragmented nucleic acids can generate PCR artifacts, such as chimeras (Liesack et al., 1991; Wang and Wang, 1996; von Wintzingerode et al., 1997).
Pitfalls arising from PCR amplification include inhibition of the reaction by coextracted contaminants, differential amplification (Lueders and Friedrich, 2003), formation of PCR artifacts (Suzuki and Giovannoni, 1996; Osborne et al., 2005), and biases associated with 16S rRNA sequence variations due to rrn operon heterogeneity (Farrelly et al., 1995; Klappenbach et al., 2001). All of these have the potential to influence subsequent measures of the extant microbial diversity (Liesack et al., 1991; von Wintzingerode et al., 1997). Coextraction of contaminants is a common problem, particularly for high-organic-matter soils and composts. The use of BSA in the PCR can help reduce the interference by coextracted contaminants. Additional purification steps, such as commercially available columns or diluting DNA extracts may also resolve problems with PCR amplification failure. Loss of DNA from organisms present in low numbers, however, will introduce yet another bias into the analysis. For PCR products to represent the existing diversity of nucleic acid templates, the following underlying assumptions must be met: all nucleic acid molecules are equally accessible to hybridize to the primers; primer–template hybrids form with equal efficiencies; the extension efficiency of DNA polymerase is the same for all templates; and substrate exhaustion limits the extension of all templates equally (von Wintzingerode et al., 1997). These assumptions are very rarely met, which is why many researchers do not attempt to derive standard ecological diversity indices from T-RFLP or other data obtained from PCR-based experiments.
The choice of primer for amplifying universal or taxon-specific 16S rDNA from complex microbial communities strongly influences the recovery of target sequences. Careful primer design and selection may help alleviate some PCR biases. Von Wintzingerode et al. (1997) strongly recommended the use of computer algorithms, such as CHECK_PROBE in the RDPII, PROBE_MATCH in the ARB software package (Ludwig et al., 2004), or the program described by Brunk et al. (1996) to refine primer choices and to check their relative efficiencies for amplifying different target sequences.
Polymerase chain reaction artifacts such as single-stranded products, template-primer hybrids, unextended PCR products produced during that last few cycles of the PCR, and the formation of chimeras (Wang and Wang, 1996) are also potential pitfalls of PCR-based analyses (Suzuki and Giovannoni, 1996; von Wintzingerode et al., 1997; Osborne et al., 2005). Products of PCR that have not been copied completely and chimeras often result from the use of too many cycles in the PCR. Reducing the number of cycles is helpful for reducing these PCR artifacts, but may also lead to failure to detect community members present in low numbers in the community. Egert and Friedrich (2003) reported the presence of single-stranded amplicons post-PCR that led subsequently to the presence of pseudoterminal fragments in T-RFLP electropherograms. They observed that single-stranded products were not cut by REs and thus were detected during T-RFLP product sizing as additional fragments larger than the expected fragment size. The use of mung bean exonuclease as a post-PCR cleanup step eliminated pseudoterminal fragments, but also reduced the peak height of T-RF products because single-stranded products were hydrolyzed and thus not detected in the electropherogram (Egert and Friedrich, 2003). Egert and Friedrich (2005) went on to discover that the use of postamplification incubation with Klenow fragment aided in reconstituting single- to double-stranded PCR products and thus restored the amplitude of the detected T-RFs. Despite these findings, few studies have reported the regular use of either mung bean exonuclease or Klenow fragment for eliminating the potential bias associated with single-stranded PCR products.
Inaccuracies during sizing of T-RFs due to differences in purine content of the amplicons have also been reported (Kaplan and Kitts, 2003), as was discussed above. Despite running the size marker in every lane of the sizing gel, which is meant to increase accuracy in sizing, errors in sizing calls are still common. Marsh (2005) reported that fragments labeled with different fluorophores may have altered mobility in DNA sizing gels or capillaries. Thus, when sample T-RFs are compared with the size marker, each labeled with a different fluorophore, base pair lengths are inaccurately calculated. This may lead to miscalls in T-RF length of up to 7 bp (Marsh, 2005).
The T-RFLP analysis method, like other PCR-fingerprinting approaches, is limited by a lack of resolving power and, thus, extant diversity is frequently underestimated. The characteristic attribute of each community member measured is the length of the T-RFs of hydrolyzed PCR amplicons. Each T-RF resolved during sizing is termed an operational taxonomic unit, which reflects our understanding that, in complex communities, there is rarely a one-to-one correspondence between a T-RF length and a specific member of the microbial community. The sites at which REs cut double-stranded DNA are not necessarily unique to a particular taxonomic group. Thus, many sequence types can share the same T-RF length. This is termed OTU overlap or OTU homoplasy, and the result is the masking of community members that share the same T-RF, which leads to underestimates of community diversity. Engebretson and Moyer (2003) tested 18 restriction endonucleases for the number of unique T-RFs produced and their ability to resolve unique sequence variants in microbial populations of varying complexity. Of the REs tested, BstUI, DdeI, Sau96I, and MspI had the highest frequency of resolving single populations in their model communities. All REs used in their study detected 
70% of the OTUs at community richness values >50 OTUs per modeled community, leading these resesarchers to conclude that T-RFLP might best be applied to less complex communities, such as biofilms, than to soils where diversity is expected to be high (Engebretson and Moyer, 2003). Rosch and Bothe (2005) examined the ability of 13 REs to resolve individual community members based on 16S rRNA genes, and genes for N2 fixation (nifH) and denitrification (nosZ). They developed a computer program, TReFID, to aid in their data analysis. The use of multiple RE digests to characterize communities increased the ability to discriminate among them (Rosch and Bothe, 2005). A similar program, TRF-CUT, that runs in conjunction with ARB, was developed by Ricke et al. (2005). The TRF-CUT program may also be used to make in silico predictions of T-RFs of aligned small-subunit rRNA genes or functional gene sequences. Since phylogenetic assignment when using these programs is based on T-RF length, it will be affected by length miscalling, hence, results should be interpreted with caution.
A key limitation of the T-RFLP technique is that it not possible to obtain sequence information that can be directly associated with specific OTUs, as is possible with denaturing gradient gel electrophoresis (DGGE, see Nakatsu, 2007). In the process of running samples through the sequencing gels or capillaries during T-RF sizing, the samples are lost. Two recent innovations may help to overcome this limitation. One such approach is called paramagnetic bead-enabled T-RFLP (Blackwood and Buyer, 2007). The procedure is in many ways the same as a standard T-RFLP. Community DNA is extracted and the target gene is amplified by PCR. Instead of a fluorescently labeled primer, however, this method uses a biotinylated primer that contains a unique restriction site. Once sample DNA is amplified, it is cut with a different RE, which yields fragments of various lengths. The sample is then incubated with streptavidin-coated paramagnetic beads, which bind all the biotinylated terminal fragments. The beads are washed and treated with the RE that releases the terminal fragments from the paramagnetic beads. These restriction fragments are then run on a standard agarose gel to separate them based on their length. Use of an agarose gel allows DNA from fragments of interest of be recovered and subsequently cloned and sequenced. In this approach, there is some loss of band resolution through the use of an agarose gel and visual, rather than automated, detection of resulting bands. Another approach for recovering sequence information from T-RFLP analysis involves adaptor ligation, fragment size selection, and reamplification with adaptor site-specific PCR to obtain a desired T-RF fraction. Cloning of the size-selected T-RF fraction allowed clones specific to the T-RF to be isolated efficiently (Widmer et al., 2006). These adaptations to the T-RFLP technique allow investigators to have the high throughput needed for broad surveys, while also enjoying the possibility of recovering sequence information from T-RFs of interest.
Avis et al. (2006) recently reviewed the biases that may arise that are unique to analyzing fungal communities. For community composition analysis, most investigators are concerned with characterizing active members of the community. Avis et al. (2006) found that spores present in analyzed samples could contribute significantly to the pool of DNA extracted and thus community members present primarily in a resting stage may be disproportionately represented in the analysis. They also found that use of some REs led to multiple fragments (i.e., "extra peaks" or "pseudo-terminal fragments") being detected and that this was caused by restriction digestion inefficiency in combination with intracollection rDNA ITS variation (Avis et al., 2006). The ITS region amplified by PCR contains both variable and highly conserved regions. In fact, the gene that codes for the 5.8S rRNA subunit, which is highly conserved, is located in this region. In fungi, genes encoding ribosomal RNA are numerous and distributed in tandem arrays along the same chromosome or among different chromosomes (Rooney and Ward, 2005). While the ITS within these gene copies is assumed to have a similar sequence, sequence variations in ITS regions within the same fungal species have been reported (Pawlowska and Taylor, 2004; Rooney and Ward, 2005). This may result in more than one T-RF being resolved from analysis of a single species, thus overestimating sample richness. On the other hand, since the fungal rRNA ITS regions contain both conserved and variable sequences, if selected REs are cut within the highly conserved 5.8S sequences, different fungal genotypes may not be resolved from each other, thus underestimating sample richness. In their analysis, Avis et al. (2006) found that the use of REs that cut in conserved vs. variable sequence regions did indeed lead to different community diversity analysis results for the samples tested. They suggested that researchers include DNA from identified fungi from studied sites to improve T-RFLP data interpretations.
|
APPLICATIONS OF THE METHOD IN SOIL MICROBIAL ECOLOGY STUDIES |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONTHE TERMINAL RESTRICTION...APPLICATIONS OF THE METHOD...CONCLUSIONSREFERENCES |
|---|
Microbial Surveys: Analyses of Community Composition
The T-RFLP analysis method has been used to characterize microbial communities in relation to naturally occurring environmental gradients both extensively and intensively. Extensive surveys have explored microbial biogeography, that is large-scale spatial patterns of microbial diversity, in marine systems (Baldwin et al., 2005; Polymenakou et al., 2005), tropical deciduous forests (Noguez et al., 2005), and soil (Fierer and Jackson, 2006). Fierer and Jackson (2006) derived indices of bacterial community diversity and richness from T-RFLP fingerprints of 98 soils collected from North and South America. These indices were not correlated with variables that normally drive patterns of diversity for macroorganisms, such as temperature and latitude. Of the variables examined, only soil pH was strongly and consistently correlated with the diversity and richness of the soil bacterial communities. Diversity was highest in neutral soils and consistently lower in acidic soils, indicating the strong influence of this edaphic factor on bacterial community composition.
Intensive surveys of microbial population distribution and community composition have been conducted in a range of environments including: forest (Hackl et al., 2004; Burke et al., 2006a), serpentine (Mengoni et al., 2004), and grassland soils (Brodie et al., 2002; Kuske et al., 2002; Mummey and Stahl, 2003; Nunan et al., 2005; Kennedy et al., 2005); rhizosphere soils from anaerobic (Scheid and Stubner, 2001; Weber et al., 2001; Culman et al., 2006; Lu et al., 2006; Wu et al., 2006) and aerobic (Dunbar et al., 2000; Sakai et al., 2001; Kerkhof et al., 2000; Tesar et al., 2002; Matsuka et al., 2003; Idris et al., 2004; Culman et al., 2006) systems; soil crusts (Redfield et al., 2002; Yeager et al., 2004); along land use gradients (Mintie et al., 2003; Rich et al., 2003; Gomez et al., 2004); in different soil types (Girvan et al., 2003; Singh et al., 2006; Ulrich and Becker, 2006); in different soil aggregate size classes (Mummey and Stahl, 2004; Vaisanen et al., 2005; Blackwood et al., 2006); in compost (Tiquia, 2005); and in biofilms (Lunsdorf et al., 2001). Bacterial population surveys have also included the guts of fungal and soil-feeding termites (Donovan et al., 2004) and earthworms (Egert et al., 2004) and populations of endophytic bacteria (Sessitsch et al., 2002b; Conn and Franco, 2004a; Miyamoto et al., 2004; Sessitsch et al., 2004; Berg et al., 2005). Intensive surveys of general fungi (Dickie et al., 2002; Lord et al., 2002; Zhou and Hogetsu, 2002; Brodie et al., 2003; Klamer and Hedlund, 2004; Allmer et al., 2006; Burke et al., 2006b; Genney et al., 2006) and arbuscular mycorrhizae (Tonin et al., 2001; Vandenkoornhuyse et al., 2003; Johnson et al., 2004; Mummey et al., 2005) have also been conducted using T-RFLP analysis. Surveys such as these document changes in microbial community composition in relation to naturally occurring gradients and attempt to link findings with edaphic and climatic factors to better understand which factors are most influential in driving changes in communities across time and space. These surveys also serve to establish baseline conditions against which changes in soil microbial communities induced by changes in management or the environment can be measured. Recently, T-RFLP has been used in forensic studies (Horswell et al., 2002), where analysis of soil evidence could potentially be used to place items, such as shoes, at a particular scene by comparative community analysis.
Assessing the Effects of Soil Management and Environmental Change on Soil Biota
Many studies have focused on how microbial community composition is affected by changes in soil or crop management. Heightened awareness of how intensive management may damage soil structure and thereby compromise soil microbial population function(s) has stimulated a greater interest in monitoring these populations using molecular analysis tools. The T-RFLP method has been used to assess changes in bacterial and fungal community composition in relation to inorganic fertilizer use in forest (Edwards et al., 2004; Mohanty et al., 2006), rice (Mohanty et al., 2006), arctic (Deslippe et al., 2005), grassland (Kennedy et al., 2004), and arable (Sessitsch et al., 2001; Tiquia et al., 2002; Tom-Petersen et al., 2003; Wolsing and Prieme, 2004) soils. The effects of the use of microbial inoculants (Conn and Franco, 2004b), organic amendments (Jordan et al., 2002; Tiquia et al., 2002; Nelson and Mele, 2006; Perez-Piqueres et al., 2006; Wang et al., 2006), and tillage (Buckley and Schmidt, 2001) have also been explored.
The T-RFLP method has been used frequently to examine how changes in crop management, such as the introduction of transgenic crops (Lukow et al., 2000; Bankhead et al., 2004; Blackwood and Buyer, 2004; Devare et al., 2004; Hsieh and Pan, 2006; Rasche et al., 2006a, 2006b) and changes in farming systems (Culman et al., 2006; Hartmann et al., 2006) affect soil microbial communities. Results from studies with transgenic crops have so far not detected any adverse changes in soil microbial populations as a result of their use. Instead, these studies invariably show significant changes in community composition in relation to seasonal and annual sampling, probably reflecting changes in rainfall and temperature across growing seasons and between years.
Lasting changes in the composition of soil bacterial populations due to soil solarization (Culman et al., 2006), herbicide (Moran et al., 2006) and pesticide (Rousseaux et al., 2003) use, soil wetting and redrying (Fierer et al., 2003; Pesaro et al., 2004), disease promulgation (McSpadden Gardener and Weller, 2001; Reiter et al., 2002), warming trends (Deslippe et al., 2005), elevated CO2 (Klamer et al., 2002; Janus et al., 2005), fire (Tobin-Janzen et al., 2005; Yeager et al., 2005), and soil pollutants (Sanchez et al., 2004; Jung et al., 2005) were readily detected by T-RFLP analysis, thus making it a highly useful tool for soil monitoring.
Linking Community Structure and Function
The information gained on the composition of microbial communities can be greatly enriched by combining T-RFLP with morphological, biochemical, or other molecular analyses. Combined approaches help us to understand how population changes may affect soil microbial function. Combining T-RFLP with SIP (Radajewski et al., 2000; Manefield et al., 2002) is a promising new avenue of exploration (Lueders et al., 2004). This approach has been used to examine the community dynamics of methanogens associated with rice roots (Lu et al., 2005). By incubating rice roots in sand with H2–13CO2 vs. N2–13CO2 in either a phosphate or carbonate buffer, extracting nucleic acids, and analyzing CsCl2 gradient fractions by T-RFLP analysis and 16S rRNA gene sequencing, these investigators were able to establish that Methanosarcinaceae and Rice Cluster I (RCI) predominated when N2 was in the headspace and carbonate buffer was used, and that RCI members were more active than the Methanosarcinacea. The Methanobacteriaceae were more active in phosphate buffer with H2 in the headspace. In a similar study, agricultural soil was incubated with 15N2. Soil DNA was extracted and centrifuged in a CsCl2 gradient to identify the dominant free-living diazotrophs in this soil. Fractions recovered from the CsCl2 gradient were analyzed by 16S rRNA targeted T-RFLP, which was used to distinguish which T-RFs were associated with the higher density fractions. This information was then used to select fractions for follow-up cloning and sequencing to identify active members of the free-living diazotrophic community (D.H. Buckley, personal communication, 2006).
|
CONCLUSIONS |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONTHE TERMINAL RESTRICTION...APPLICATIONS OF THE METHOD...CONCLUSIONSREFERENCES |
|---|
|
ACKNOWLEDGMENTS |
|---|
My special thanks go to Terence Marsh at Michigan State University, who introduced me to the T-RFLP method and worked with me over the years to train others to use this technique for soil microbial community analysis. The author also acknowledges R. Kantety, Alabama A&M University, for his help in creating the T-RFLP flow chart in Figure 3.
|
NOTES |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONTHE TERMINAL RESTRICTION...APPLICATIONS OF THE METHOD...CONCLUSIONSREFERENCES |
|---|
Received for publication September 9, 2007.
|
REFERENCES |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONTHE TERMINAL RESTRICTION...APPLICATIONS OF THE METHOD...CONCLUSIONSREFERENCES |
|---|
from microbial community DNA extracts.