1. Nematode assemblages are commonly used as an indicator of ecosystem health; however, conventional approaches to assemblage analyses are restricted by time-consuming processing and declining availability of expertise. Molecular methods offer a rapid and cost-effective alternative.
2. We have designed a molecular profiling system, using directed terminal-restriction fragment length polymorphism (dT-RFLP), to characterise nematode assemblages by relative abundance of feeding guilds.
3. An arable soil was first characterised by cloning and sequencing of small subunit ribosomal DNA, and an enzyme digest selected to discriminate between feeding guilds. This yielded 14 different terminal-restriction fragments (T-RFs) from the sequence set, assigned to five nematode feeding guilds.
4. Robustness of the dT-RFLP methodology was tested. The greatest amount of variation between replicates occurred at the PCR stage, with little variability between replicate digests from the same PCR product or capillary runs.
5. dT-RFLP revealed changes in assemblage composition owing to organic amendments of dairy-cattle slurry and municipal green compost. The proportion of microbial feeding nematodes was higher in compost and slurry plots than in the no amendment control in the first sampling after organic amendment. Plant feeding nematodes composed a significantly greater proportion of the control assemblage during the growing season and post-harvest.
6. The increased throughput of molecular analysis compared with microscopy increases the feasibility of studies involving large-scale sampling and makes nematode assemblage analysis more attractive as an indicator of soil health for environmental assessment.
Owing to population growth, demand for grain is expected to double worldwide by 2050 (Tilman et al. 2002), increasing pressure on soil. With c. 40% of agricultural soils, already suffering degradation (Doran & Zeiss 2000), monitoring and protection of soil health is critical, the alternative being further conversion of natural ecosystems. In Europe, there is an increasing list of legislation relating to the protection of soils as a result of Common Agricultural Policy (CAP) reform (Creamer et al. 2010). Despite this increasing policy requirement for effective monitoring of soils at local-, regional- and national-scales, it remains unclear which properties of soils are most appropriately monitored. This is partly because of the numerous goods and services that soils provide, but also because of their inherent chemical, physical and biological complexity. Given that biota plays such fundamental roles in the majority of soil ecosystem services, biological properties are logical candidates as effective indicators, to complement soil physico-chemical properties (Ritz et al. 2009). The definition of soil health varies but Kibblewhite, Ritz & Swift (2008) describe it as (i) the ability to maintain production while (ii) providing essential ecosystem services and (iii) maintaining biodiversity. All three elements depend on the soil food web that (i) directly affects plant productivity by root herbivory, parasitism and mutualism (ii) effects nutrient cycling (through organic matter decomposition and nutrient mineralisation) and (iii) influences formation and structure of soil.
Nematodes are a key component of the soil food web, occurring at different trophic levels and forming links between plants, bacteria, fungi and other soil fauna (de Ruiter et al. 1993; Traunspurger 2000). Their abundance, diversity and rapid generation times as well as relative ease of extraction from soil (Ritz & Trudgill 1999) make nematodes an ideal indicator group. Inhabiting the film of water surrounding soil particles, the cuticle is in constant contact with the soil environment, and because most species spend their entire life cycle in the soil, nematodes are constantly affected by the surrounding soil conditions (Ritz & Trudgill 1999). Their range of responsiveness to toxins and stresses such as desiccation make them valuable indicators in disturbed systems (Neher 2001). Nematodes are classed in feeding guilds (Yeates et al. 1993) and so function may be inferred according to taxa.
Conventional morphological approaches to assemblage analyses are restricted by the vast diversity within the phylum Nematoda coupled with a lack of skilled taxonomists (Andréet al. 2001). Sample processing is limited to 2–5 per day (assuming 200 nematodes identified to Order), limiting the use of nematodes as indicators in studies on ecological scales. The use of molecular methods for routine assessment of previously described species is increasing (Chen et al. 2010; da Silva et al. 2010).
Terminal-restriction fragment length polymorphism (T-RFLP) is a semi-quantitative PCR-based fingerprinting technique, most frequently used for bacterial community profiling (Liu et al. 1997) but has also been applied to, for example, archaeal (Chin, Lukow & Conrad 1999), ectomychorrizal (Zhou & Hogetsu 2002) and arbuscular mycorrhizal (Uibopuu et al. 2009) communities. Use of T-RFLP for non-fungal eukaryotic organisms has been more limited but includes ciliates (Marsh et al. 1998), marine picoeukaryotes (Díez et al. 2001), protists (Countway et al. 2005), mites (Gibb et al. 2008) and nematodes (Donn et al. 2008; Edel-Hermann et al. 2008). The main advantage of T-RFLP over similar fingerprinting methods is the ability to compare data across electrophoretic runs, as the inclusion of a size standard with every sample allows accurate sizing of fragments (Marsh 1999; Nunan et al. 2005). T-RFLP is frequently used as a community fingerprinting method (with no inference of the identity of the analysed taxa) or coupled with in silico digest of a sequence library to infer identity of T-RFs. Provided the sequences within the assemblage are known, a more informative way of using T-RFLP is to select restriction endonucleases, which cut DNA in positions that discriminate between taxa of interest, termed here as directed T-RFLP (dT-RFLP).
We first characterised an arable nematode assemblage by sequencing the small subunit ribosomal gene (SSU rDNA) and based the dT-RFLP design on these sequences. While it is unlikely that molecular profiles will fully match morphological analyses because of differences in target gene copy number between nematode species (Donn et al. 2011), the ability of molecular fingerprinting to reflect changes in nematode assemblages is tested by comparing assemblages under different organic amendments (cattle slurry and green compost). Organic amendments act as a plant nutrient supply stimulating the microbial community (Bünemann, Schwenke & Van Zwieten 2006) and have previously been shown to increase numbers of bacterial feeding nematodes (Griffiths 1994; Neher & Olson 1999; Forge, Bittman & Kowalenko 2005). Our hypotheses are (i) both slurry and compost amendment will result in an increased proportion of nematode Orders including microbial feeding nematodes and (ii) amendments should improve soil fertility resulting in greater plant biomass, with an associated increased proportion of plant feeding nematodes.
We further hypothesise that dT-RFLP provides the accuracy to provide information about the composition of nematode assemblages suitable for the application of soil health indices. To this end, the robustness of the dT-RFLP strategy is tested, and the main sources of variation are assessed.
Materials and methods
Sequencing from Nematode Assemblages
To characterise the nematode assemblage from an arable soil, sequencing of small subunit ribosomal DNA was performed. Nematodes were extracted from 200 g soil sampled from six plots of barley (Hordeum vulgare L.) at the James Hutton Institute (56′46 N, 03′04 W). These comprised two replicate plots of three treatments – deep plough, no till and compaction, which were part of a larger field trial (described by Sun et al. 2011). It was expected that sequencing of clones generated from PCR products originating from those three different treatments would reveal the greatest range of nematode types present at the field site, from which the dT-RFLP could be designed. Nematodes were sampled in October 2004 as described by Griffiths et al. (2006).
DNA was extracted by bead beating with purification through a Purelink PCR purification column (Invitrogen, Paisley, UK) as described by Donn et al. (2008). Near full-length SSU rDNA was amplified by PCR using primers Nem_SSU_F74 and SSU_R81 (Table 1). PCR was performed in a 25-μL final volume including 1× buffer (60 mM Tris–SO4, 18 mM NH2SO4), 2 mM MgSO4, 0·2 mM dNTPs (Promega, Southampton, UK), 0·5 μM each primer (unless indicated otherwise, all primers were synthesised by VH Bio, Gateshead, UK) and 1 unit High Fidelity Platinum Taq DNA polymerase (Invitrogen). Cycling was performed on a MJ DNA engine PTC-200 PCR machine (MJ Research, Reno, NV, USA) as follows: 94 °C, 2 min then 35 cycles of 94 °C, 30 s; 55 °C, 30 s; 68 °C, 2 min and a final extension step of 68 °C for 10 min. Cloning and sequencing were performed as described by Griffiths et al. (2006). Briefly, PCR product was ligated into pGEM-T Easy vector (Promega) and transformed into electrocompetent DH5-αEscherichia coli with selection performed on LBAIX plates. Plasmid was recovered using a MultiScreenHTS Vacuum Manifold (Millipore, Bedford, MA, USA) with MultiScreen Clearing and Multiscreen FB plates. One hundred and twenty-eight clones from each of the three treatments were sequenced with primer Nem_SSU_F74, giving c. 600 bp of single-strand sequence from the 5′ end of the gene.
Table 1. Small subunit rDNA primers used in sequencing and/or dT-RFLP. For dT-RFLP, the flurophore VIC was added to Nem_18S_R. Letters denoting nucleotides follow the standard IUPAC codes
*Position on Caenorhabditis elegans sequence X03680.
Sequences were aligned using POA (Lee, Grasso & Sharlow 2002) and phylogenetic analysis performed in Topali (Milne et al. 2004) by constructing a neighbour-joining tree (F84 and gamma rates of substitution), as described by Griffiths et al. (2006), with clones grouped into sequence types at 0·03 substitutions per base. Close to full-length double-strand sequence was obtained for a minimum of three clones from each of the sequence types or as many clones as were available by further sequencing using primers SSU_F_22, SSU_F23, SSU_R_09, SSU_R_23 and SSU_R_81 (Table 1), with reaction conditions described by Griffiths et al. (2006). Chimeric sequences were identified by creating three phylogenetic trees from the front, middle and end subsections of the near full-length sequences. Any sequences moving between groupings across the trees were identified as chimeras and removed from further analyses. The near full-length sequences were aligned with 273 database SSU nematode sequences and the alignment trimmed to 163–1670 (corresponding to Caenorhabditis elegans sequence X03680). A neighbour-joining tree (F84 and gamma rates of substitution) was drawn with bootstrapping over 1000 iterations. Sequence types were assigned to nematode Order, or Family where possible, based on grouping with database sequences at >80% bootstrap support.
Design of dT-RFLP Strategy
Sequence data collected previously were used to design an enzyme digest to separate sequences into terminal-restriction fragments (T-RFs) relating to nematode Order. Design of the dT-RFLP strategy was based on a sequence segment of c. 900 bp close to the 5′ end of the gene, bounded by general nematode primers Nem_SSU_F74 and Nem_18S_R (Table 1). The 5′ end of the gene was selected as it contains the most variability (Floyd et al. 2002) and been used in nematode sequencing studies (Floyd et al. 2002; Griffiths et al. 2006). Restriction enzymes were selected using the freeware application DRAT (Roberts et al. 2011) (http://www.hutton.ac.uk/drat). Sequence types from the three treatments were trimmed to include the Nem_SSU_F74 and Nem_18S_R primer sites. Initial input into DRAT set the maximum number of enzymes as 1 with successive cycles increasing the ‘maxenz’ parameter by 1 until a solution was found. Minimum distance between peaks was set to 5, as this allows peaks in the resulting dT-RFLP profiles to be discriminated without confusion between neighbouring peaks. Each DRAT run was set to search for a solution with either the 5′ or 3′ primer labelled. While DRAT can search for a solution with both primers labelled, a single label was preferred for ease of data processing removing the calibration requirement for dye efficiencies and avoiding any problems of interference between dye channels.
Choosing an Enzyme – Comparing Efficiencies of Enzyme Digests
Several enzyme combinations were identified, which separated the sequence types by Order with a 3′ fluorescent PCR primer. Solutions included six combinations of either BtsCI or FokI in combination with one of MlyI, PleI or HinfI. BtsCI and FokI are neoschizomers, sharing the same recognition sequence but cleaving at different positions. The same is true in MlyI and PleI. HinfI includes a degenerate base and therefore cleaves at additional sites to MlyI and PleI. Digests were performed on six clones of known sequence with each of the six enzyme combinations to determine which enzyme combination cuts most efficiently. Success of an enzyme digest was determined by the percentage of the total peak area lying in the predicted T-RF size. Mean percentage peak area in the predicted peak was calculated over all clone digests for each enzyme combination. Fluorescence not in the predicted peak was classed as uncut, PCR artefacts (also present in uncut profiles) or ‘other’ peaks.
FokI and MlyI performed poorly in initial clone digests (data not shown) and were eliminated. The two most efficient enzyme combinations, BtsCI/HinfI and BtsCI/PleI, were each further tested on 55 clones, representing the 12 predicted T-RF sizes. PCR was performed using primers Nem_SSU_F74 and VIC-Nem_18S_R (Applied Biosystems, Foster City, CA, USA) (Table 1) as above but with the extension step at 68 °C reduced to 45 s, with 1 μL of a 1 in 10 dilution of plasmid template. PCR products were digested with each enzyme combination in 20 μL final volume including 6 μL PCR product, two units of each enzyme, 0·7× buffer and 100 μg mL−1 BSA both supplied with the enzymes (New England Biolabs, Ipswich, MA, USA). All enzymes were supplied with the same buffer with concentration of buffer added adjusted to account for the presence of PCR buffer. Two units of PleI or HinfI was added to the digest mix, and reactions were incubated at 37 °C for 1 h followed by 65 °C for 20 min. Two units BtsCI was then added to each and the reactions incubated at 50 °C for a further 1 h. Digest products were diluted 1 in 10 in sterile distilled water, 1 μL mixed with 9 μL Hi-Di formamide and 0·05 μL ROX-labelled MapMarker 1000 (BioVentures, Murfreesboro, TN, USA). Fragments were analysed on an ABI 3730 capillary sequencer and data analysed in Genemapper (Applied Biosystems) with a baseline threshold of 50 fluorescence units. Uncut PCR product was also run for one clone from each of the T-RF groupings. Any peaks, smaller than the uncut PCR product, found in the uncut profiles were assumed to be PCR artefacts and subtracted from the digested samples. These typically represented 4% of the total fluorescence.
Digest efficiency was assessed by calculating the percentage of the total peak area in the dT-RFLP profile of the expected peak size (predicted by DRAT). Percentage abundance of fluorescence in the predicted peak was logit-transformed, to achieve normality, and analysed by two-way anova with fragment size and enzyme (PleI or Hinfl) as factors in GenStat Version 13 (VSN International, Hemel Hempstead, UK).
Testing dT-RFLP on Clones – Sizing
To compare experimental data with in silico predicted T-RF sizes, Genemapper size-calling settings were compared. dT-RFLP data from clone digests using enzymes BtsCI and PleI were analysed using each of the five algorithms available in Genemapper: second order, third order, cubic spline, local southern and global southern. Experimentally derived T-RF sizes for each setting were compared with T-RF sizes predicted by DRAT.
Testing dT-RFLP on Environmental Samples
The main possible sources of variation arising from dT-RFLP methodology are (i) PCR, (ii) digest and (iii) sequencer capillary run. The repeatability of the BtsCI/PleI digest was tested on a nematode assemblage extracted from an arable field site. PCR, digest and sequencer run were replicated. Three PCRs were performed from a single DNA sample; for each single PCR product, the enzyme digest was replicated three times, and the product of these digests was separately analysed in three capillaries during the fragment analysis. For each peak in the resulting profiles, mean fluorescence was calculated across each set of replicates, and the coefficient of variation was calculated.
Nematode Assemblage Composition Under Amendment
Plots were amended with dairy-cattle slurry, municipal green compost or no amendment (control), prior to being sown with spring barley. Further details of the experiment are described by Griffiths et al. (2010). Six composite samples from each treatment were taken in April (after amendment, pre-sowing), June (in-crop) and September (post-harvest). Composite soil sample collection, nematode isolation, DNA extraction and dT-RFLP were performed as described earlier.
T-RFs of nematodes belonging to different feeding guilds within Orders were separated by at least two base pairs. Thus, it was possible to analyse relative abundance of T-RFs belonging to feeding guilds. Each of the sequence types was assigned to feeding guilds according to grouping with database sequences and with reference to Yeates et al. (1993) although we recognise the existence of intra-generic trophic variability (Okada, Harada & Kadota 2005). Percentage composition of feeding guilds was calculated by combining fluorescence from peaks belonging to the same guild and dividing by the total fluorescence for the profile after removal of artefactual peaks. Mean percentage of each feeding guild was compared by repeated measures analysis of variance with treatment as the main factor. Data for fungal feeders and carnivores were first logit-transformed to achieve homogeneity of variance.
Sequencing from Nematode Assemblages
In total, 516 single-strand sequences were obtained from nematode assemblages extracted from the three sampled treatments (deep plough, no till and compaction). The majority of these were sequences between 500 and 800 bases in length close to the 5′ end of the SSU gene. From these, 32 sequence types were identified at 0·03 substitutions per base. Near full-length double-strand SSU sequence was obtained for representative clones of each of those sequence types to ensure high-quality sequence; all sequences generated have been submitted to GenBank under accession numbers JN049656–JN049687. Sequence types were assigned to Order or Family based on phylogenetic analyses of the near full-length sequences. All obtained sequences grouped with named database sequences with high (>80%) bootstrap support (Fig. 1).
Nematode clone libraries were dominated by Mononchid and Dorylaimid sequences (Fig. 1). The most abundant sequence type closely matched Anatonchus tridentatus and the dominant Dorylaimid sequence grouped with Aporcelaimellus obtusicaudatus. Sequences were obtained for 32 nematode types present at the field site, which are likely to represent the most abundant types at this arable site. This provided the starting point for the design of an enzyme digest strategy to profile the nematode assemblage.
Design of a Directed T-RFLP Strategy
No solution was possible with a single enzyme digest, but with the maximum enzyme number increased to two, several possible solutions were found with a single 3′ label. DRAT reported that Bse118I/BshFI separated Orders as required with the greatest T-RF size fidelity within Orders. This combination was eliminated because the in silico digest left some sequences uncut (data not shown). Although this does discriminate them from other groups in the design, these sequences could not be distinguished from sequences uncut as a result of inefficient digestion. DRAT returned six further enzyme combinations that separated all the Orders, and after initial testing (see Materials and methods), this was reduced to two possible combinations: BtsCI/HinfI and BtsCI/PleI.
Choosing an Enzyme – Comparing the Efficiencies of Enzyme Digests
Mean percentage fluorescence within the expected T-RFs was higher with the BtsCI/PleI double digest than with BtsCI/HinfI (anovaP < 0·001, Fig. 2); therefore, PleI was chosen as the second enzyme.
Percentage fluorescence in the expected peak varied with cut site (anovaP < 0·001). For Cephalobid sequences with a T-RF of 771 bases, only 79% of the total fluorescence was in the peak of the expected size for Plel (Fig. 2). Approximately 3% of the total peak area lay within the uncut peak, while the remainder was distributed between three peaks between 100 and 150 bp of unknown origin (Fig. 3).
Predicted T-RF size for each of the sequence types identified when digested with BtsCI and PleI is shown in Table 2.
Table 2. Predicted T-RF size and terminal cutting enzyme for each sequence type in Fig. 1 based on a BtsCI/PleI double digest. Type numbers correspond to Fig. 1. Predicted T-RF size is from in silico digest using DRAT and observed from T-RFLP of three clones per sequence group tested, or a single clone where the clone was a singleton in the library. Observed sizes >1 bp different from the predicted size are shown in bold
Of the five sizing algorithms in Genemapper, cubic spline and local southern methods matched the predicted fragment sizes most closely and gave very similar results (data not shown). Observed size deviated from the expected size by >1 bp for four of the T-RFs though this was not correlated to increasing fragment length and was not consistent with being an effect related to the enzyme making the terminal cut (Table 2). Where sizing differed from the expected size, it was consistent across clones tested with the same predicted T-RF.
Testing dT-RFLP on Environmental Samples
An example of a dT-RFLP profile from an environmental nematode assemblage sample is shown in Fig. 4. Peaks belonging to distinct taxa can clearly be discriminated, and levels of background fluorescence were low. The variability at each of the three stages tested: PCR, restriction enzyme digestion and capillary run is compared in Table 3. The greatest variability is between PCR replicates with least variability between capillary runs of the same digest product.
Table 3. DNA was extracted from a nematode assemblage and amplified in triplicate using primers NEM_18S_F74 and VIC-Nem_SSU_R. Products were digested with BtsCI and PleI and T-RFs analysed by capillary electrophoresis. PCR 1 was digested in triplicate, and the electrophoresis run replicated in triplicate for digest A. Results are shown as absolute peak area fluorescence values. The mean and standard deviations were calculated across the three replicates. Standard deviation is also shown as a percentage of the mean. Observed fragment lengths are listed
Digest replicate (PCR 1)
Capillary run replicate (PCR 1, digest A)
5580 ± 568
5179 ± 57
5397 ± 142
17 906 ± 635
17 450 ± 895
18 562 ± 478
9600 ± 241
9233 ± 121
9450 ± 211
4006 ± 201
4002 ± 221
4137 ± 231
5355 ± 932
6056 ± 591
6417 ± 89
3125 ± 394
3089 ± 113
3292 ± 173
15 779 ± 2134
17 046 ± 729
17 920 ± 444
18 880 ± 1612
20 041 ± 261
20 538 ± 451
8120 ± 630
7667 ± 312
8070 ± 74
6887 ± 268
6589 ± 185
6671 ± 169
95 237 ± 4791
96 350 ± 3402
100 453 ± 2165
Nematode Assemblage Composition Under Organic Amendment
In April, immediately post-amendment, the proportion of bacterial feeding nematodes in compost and slurry amended plots was higher than the control. In compost plots, bacterial feeders continued to comprise a greater proportion of the population throughout the year; however, by June, the proportion of bacterial feeders in slurry plots was not significantly different from the control (Fig. 5) (anova main effects amendment P = 0·011, time P < 0·001). Fungal feeders composed a significantly smaller proportion of the assemblage in the compost treatment throughout (anova main effect amendment P = 0·01). In April, fungal feeding nematodes accounted for a greater proportion of the assemblage in slurry plots compared with the control though by June there was no difference. The proportion of plant feeding nematodes was lower under compost compared with the control throughout the year (anova main effects amendment P = 0·007, time P = 0·002) (Fig. 5). Carnivorous and omnivorous nematodes formed a smaller proportion of the compost assemblage in April, though by June, proportion was not significantly different to the control (carnivores: anova time × treatment interaction P = 0·003; omnivores: anova time × treatment interaction P = 0·014). In September, carnivores composed a significantly greater proportion of the assemblage in slurry plots compared with the control.
Many of the nematode sequence types found here corresponded to those found previously at this site (Griffiths et al. 2006). New sequence types, not previously found, grouped with Panagrolaimus, Aphelenchoides, Telotylenchidae, Hoplolaimoidea, Criconematina, Sectonema, Pungentus and Longidoridae. That more Tylenchida were revealed by sequencing is unsurprising, as a newly designed more inclusive primer set was utilised here (Donn et al. 2011). Sequence types found by Griffiths et al. (2006), but not seen here were all present previously as a single clone suggesting that they were rare types – these included sequences grouping with Eumonhystera, Alaimus and Tripyla. As observed by Griffiths et al. (2006), clone libraries were dominated by large biovolume nematode sequences (Mononchids and Dorylaimids). The rarity of the smaller species sequences (e.g. the Tylenchida that appear mostly as singletons, Fig. 1) is potentially explained by the dominance of large biovolume groups by Griffiths et al. (2006) and Donn et al. (2011).
To design a dT-RFLP strategy to discriminate between nematode taxa, several solutions were presented by DRAT. These enzyme combinations were tested empirically to select the pair that cuts most efficiently. BtsCI in combination with any of the set two enzymes performed consistently better than FokI. Of the set two enzymes, MlyI performed poorly in comparison with HinfI or PleI. One possible reason for this could be the reaction conditions of the digest and highlights the need to test enzyme combinations prior to application. Here, the difference may have been driven by enzyme sensitivity to reaction conditions as for convenience digestion was performed in PCR buffer, supplemented with the recommended enzyme buffer. It is possible to purify the PCR product to remove the PCR buffer and then to carry out the digest under the recommended conditions; however, this adds a time-consuming step to the process that is avoided when a suitable enzyme performs well in DNA polymerase buffer. The enzymes tested and finally selected for the digest (BtsCI/PleI) did not cut 100% efficiently when tested on clones. The sequence of the amplified product appeared to affect the efficiency with which the enzymes performed. While the BtsCI/PleI combination was c. 95% efficient at cutting Mononchid types with a T-RF of 327 and Tylenchid types 19 and 31 (T-RFs 593 and 587), only 80% of the Rhabditid T-RFs of 132 and Cephalobid type 8 (T-RF 771) peak area was observed as predicted. This may be due to site preference of the enzymes. The efficiency of digestion at recognition sites can in some cases be affected by the surrounding sequence or position of the recognition site in the molecule (New England Biolabs Technical Reference 2009).
Osborn, Moore & Timmis (2000) recommend empirical determination of the most suitable algorithm for each new system tested. Here, cubic spline and local southern methods most closely matched the expected T-RF sizes. Four peak sizes differed from the predicted fragment size by more than one base pair: Aphelenchida groups 10, 15, 20, 21 at 228 bp, Dorylaimida 6 at 603 bp, Dorylaimida 16 at 605 bp and Cephalobida 8 at 771 bp (Table 2). Pandey, Ganesan & Jain (2007) cite differences in dye chemistry as a possible root of inaccurate sizing as any difference in molecular weight of the dyes could alter the migration rate of fragments. Similarly, with ROX-labelled ladder and FAM-labelled T-RFs, Kaplan & Kitts (2003) found that as T-RF size increased, T-RF length was increasingly underestimated. ROX label has a higher molecular weight than FAM and therefore migrated more slowly although sequencer controls should mitigate this issue. This was not the case here where we observed sizing inaccuracies occurring in both long terminal fragments (Cephalobida type 8, 771 bp) and relatively short terminal fragments (Aphelenchida type 10, 228 bp, Table 2) while other fragments of 132 (Rhabditida type 17) and 639 (Cephalobida type 12) were sized accurately. Kaplan & Kitts (2003) also found that a difference of 1% purine content of T-RFs could affect sizing. In this study, purine content of T-RFs varied by up to 5%; however, no relationship was found between this and the discrepancy between observed and expected T-RF sizes (data not shown). Other possible explanations for the disparity between observed and expected T-RF lengths include residual polymerase, terminal transferase (Clark 1988; Hartmann, Enkerli & Widmer 2007) or exonuclease (Gury et al. 2008) activity at the 3′ end of overhanging restriction fragments. The difference in observed and expected fragment sizes was not consistent across recognition sequences (Table 2); however, enzyme activity may be dependent on surrounding sequence composition (Brownstein, Carpten & Smith 1996; Magnuson et al. 1996) and therefore may not affect all T-RFs in the same way.
The PCR stage was the source of greatest variability in the process (Table 3), followed by restriction enzyme digestion and capillary run. Little variation was seen in capillary run with the coefficient of variation ranging from 0·9 to 5·6%. This is lower than that observed by Osborn, Moore & Timmis (2000) when acrylamide gels were loaded manually and depended on line-based detection, showing that the use of automated capillary sequencers has improved the reproducibility of T-RFLP, though this will vary with the capillary sequencer model used. Variability between digests was likely caused by pipetting error. The major source of variability was PCR, with the coefficient of variation up to 17%. In Table 3, absolute fluorescence was compared between runs; however, in T-RFLP analysis, relative abundance of T-RFs is routinely calculated, which will reduce the variability caused by different total fluorescence between profiles. Remaining variability caused by PCR bias may be reduced by performing several PCRs and combining the products before digest (Blackwood et al. 2003), though some studies have found this to make little difference to the end result (Osborn, Moore & Timmis 2000).
While we set out to design a dT-RFLP to profile taxonomic groupings of nematodes, the resulting profiles were easily analysed as feeding guilds, meaning function could be inferred. dT-RFLP revealed changes in nematode assemblage composition under compost and slurry amendments. The greatest differences were seen in April, just after amendments were applied. Higher proportions of bacterial feeding nematodes were seen in the compost and slurry amended plots compared with the control, and fungal feeders were also higher in the slurry plots consistent with greater microbial activity following nutrient inputs. While the high proportion of bacterial feeders persisted in compost plots, by June, the proportion of microbial feeding nematodes in slurry plots was not different to the control. By September, plant feeders made up a greater proportion of the assemblage in the controls despite greater yields from amended plots. This may be related to the trend showing an increased proportion of carnivores and omnivores following amendment, which would indicate greater development of the food web and a more structured biological system (Ferris, Bongers & de Goede 2001).
In this study, resolution of the nematode assemblage is low, being classified only as far as feeding guild. Finer resolution towards species level may be achieved by selecting more variable genes (e.g. internal transcribed spacer regions or large subunit rDNA) as PCR targets. All steps in this method may be carried out in 96- or 384-well plate format, meaning the number of samples to be processed is limited primarily by the extraction of nematodes from soil. A rapid analysis method opens up the use of nematodes as indicators in sample-intensive studies such as spatial and temporal variation of assemblages just not only in an agricultural setting but also in natural, disturbed, polluted soils and other habitats aquatic or terrestrial. While we have used dT-RFLP to study whole nematode assemblages, the methodology is easily adapted and specific primers could be used to restrict the analysis to groups of interest or expand and be applied to whole soil eukaryotic faunal assemblages or food webs.
This work was supported by the Scottish Government through the Rural and Environmental Research and Analysis Directorate and a BBSRC Quota studentship (BBS/S/K/2004/11271). We recognise statistical advice given by Biomathematics and Statistics Scotland.