Molecular analysis of predation, through polymerase chain reaction amplification of prey remains within the faeces or digestive systems of predators, is a rapidly growing field, impeded by a lack of readily accessible advice on best practice. Here, we review the techniques used to date and provide guidelines accessible to those new to this field or from a different molecular biology background. Optimization begins with field collection, sample preservation, predator dissection and DNA extraction techniques, all designed to ensure good quality, uncontaminated DNA from semidigested samples. The advantages of nuclear vs. mitochondrial DNA as primer targets are reviewed, along with choice of genes and advice on primer design to maximize specificity and detection periods following ingestion of the prey by the predators. Primer and assay optimization are discussed, including cross-amplification tests and calibratory feeding experiments. Once primers have been made, the screening of field samples must guard against (through appropriate controls) cross contamination. Multiplex polymerase chain reactions provide a means of screening for many different species simultaneously. We discuss visualization of amplicons on gels, with and without incorporation of fluorescent primers. In more specialized areas, we examine the utility of temperature and denaturing gradient gel electrophoresis to examine responses of predators to prey diversity, and review the potential of quantitative polymerase chain reaction systems to quantify predation. Alternative routes by which prey DNA might get into the guts of a predator (scavenging, secondary predation) are highlighted. We look ahead to new technologies, including microarrays and pyrosequencing, which might one day be applied to this field.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Polymerase chain reaction (PCR)-based techniques for detecting prey remains in the guts, faeces and regurgitates of predators are being developed to study complex trophic interactions in the field (reviewed in Symondson 2002; Sheppard & Harwood 2005; Sunderland et al. 2005; Gariépy et al. 2007). Research in this area has, however, been hampered by a lack of clear guidance through the many techniques and approaches available. Detecting degraded, semidigested DNA is not always easy. Major areas of difficulty seem to be lack of sensitivity, short post-ingestion detection periods and cross-amplification problems. However, attention to some simple guidance, for example on primer design and assay optimization, can in many instances prevent these problems arising in the first place. At an early stage in the application of a new technology, it is good to see people trying a spectrum of different approaches and we do not wish to inhibit that process; this review is certainly not intended to provide a single blueprint that everyone should follow. We cannot hold up our own papers as examples of how to do it either, because at one time or another we have, between us, made just about every possible mistake. However, we realize that there is a need for guidelines that will assist both molecular ecology novices and those with experience to advance effectively and rapidly in this challenging field.
In addition to the primary description of techniques, and advice on best practice, we have included simple flow diagrams describing a set of basic steps of wide utility that we recommend (Fig. 1) with further details, and alternative strategies, described in the main text.
Vacuum sampling methods (Sunderland et al. 2005) have been used for collecting predators before molecular analysis of their gut contents (e.g. using antibodies, Hagler & Naranjo 1994), but are best avoided wherever possible. Such harsh collecting systems may lead to external contamination of predators by the remains of prey that have broken apart (especially fragile animals such as aphids and whiteflies) or by regurgitated material. Even if the gut is dissected out, rather than the whole predator homogenized, cross-contamination is possible. PCR is such a sensitive technique that the risk of false-positives could be unacceptably high. An even greater danger, possibly, is false-positives from predation occurring within the collection receptacle, immediately after the vacuum is turned off, when high densities of predators and prey are in close proximity. It is our experience that many carabids, spiders and predatory bugs will almost immediately grab the nearest prey. These problems can be mitigated to some extent by using low vacuum pressures and placing samples immediately on ice in the field. If vacuum sampling is used, experiments to check for contamination should be performed.
Pitfall trapping measures activity density (Luff 1975; Sunderland et al. 2005), which can be relevant to predator–prey encounter rates. However, predation events occurring within the trap are again a problem. This may be avoided to some degree by using a mesh insert, allowing smaller animals to fall through out of reach of larger predators (Harper et al. 2005). If the predators are trapped into a preservative (certainly not formaldehyde which is known to inhibit PCR, Gurdebeke & Maelfait 2002) there is a real danger of regurgitates from one predator contaminating others as they drown. A pragmatic compromise is to use raised covers on dry traps (to keep out prey falling from the vegetation above and encourage any that do fall in to crawl out through positive phototaxis), to provide refugia (stones, leaves, etc.) within the traps to cut down on mutual predation (Sunderland et al. 2005) and to collect predators from the traps frequently. The latter is necessary anyway to ensure that they have digested their prey to the minimum extent. A further problem with pitfall traps is that satiated predators are less active than hungry ones (Fournier & Loreau 2001, 2002), potentially causing a bias towards predators with empty guts. Despite these problems, pitfall traps will continue to be used, especially for large, nocturnal, low-density predators such as carabids that bury themselves in the soil during the day and for which no other effective sampling method has been devised.
Ideally, predators required for gut content analysis would be best collected individually, by ground searching within quadrats, using a pooter for small quick-moving predators (e.g. Collembola, Read et al. 2006) and flying insects, or hand searching through soil samples (Juen & Traugott 2007). This is particularly appropriate where whole predators, which might be externally contaminated by vacuum sampling, are homogenized (rather than dissected) before DNA extraction (Harwood et al. 2007). If Malaise traps are used, catching flying predators within the nets by pooter is acceptable, while analysing those in the collection bottle (where cross-contamination is highly likely) is almost certainly not. In practice, combining such collection of high-quality samples for molecular analysis, with more rapid and efficient collection systems (such as vacuum sampling or extraction of soil samples) to obtain population data, would be recommended.
Aquatic invertebrates present a special problem, particularly marine species that are floating in a planktonic soup of biota. Analysis of the diets of krill, for example, when the water surrounding the predators is teaming with dietary components, is problematic (Martin et al. 2006; Passmore et al. 2006). Special care is needed to avoid contamination of the gut contents during dissection (Passmore et al. 2006) and flaming of dissection tools between samples is always essential. External contamination can also be a problem with soil-dwelling predators. Prey adhering to the outside of very small fragile predators is especially problematic, as such predators may (unlike krill) be impossible to wash. Read et al. (2006) had to microscopically examine a subsample of Collembola to ensure that their nematode prey were not attached to the predator.
Invertebrate predators needed for sequencing must be starved after capture, or body parts such as legs that cannot contain prey DNA, are selected for extraction. The latter can be a problem with spiders, where gut diverticula extend even into the legs (illustrated in Ruppert & Barnes 1994).
Vertebrates can also be sampled by killing them and analysing their gut contents, and this has been used for birds (Scribner & Bowman 1998) and fish (DeWoody et al. 2001; Rosel & Kocher 2002). Although fish are not so problematic (they may be killed for food in any case), we would recommend that most work on vertebrates should now be conducted using noninvasive analysis of faeces, given the proven success of this approach. However, where culling programmes are going ahead in any case, exploitation of the gut contents in order to study the trophic ecology of a predator should be encouraged.
Collection of faecal samples
Most of the literature on faecal analysis is directed towards extraction and analysis of DNA from epithelial cells from the predator (e.g. Garnier et al. 2001; Goossens et al. 2006). This is now a standard method for noninvasive sampling of vertebrate populations and good quality, relatively undegraded DNA can usually be found. However, where very limited quantities of highly degraded prey DNA are sought, the age and condition of faecal samples can be critical. Vertebrates cannot always be induced to defecate on demand, as found by Deagle et al. (2007) with macaroni penguins, Eudyptes chrysolophus, even when subjected to stomach flushing. Reptiles can be more obliging, with many, such as slow worms (Anguis fragilis) and smooth snakes (Coronella austriaca), defecating when gently palpated or as part of a defensive reaction when handled (D.S. Brown, personal communication). Many songbird chicks will produce faecal sacks at the slightest disturbance (Sutherland 2000). Wherever possible, faeces should be collected fresh to minimize further enzymatic action, although freshness is not always easy to determine in the field. Older faecal samples may therefore generate false-negatives for prey that were indeed consumed. Faecal matter in contact with a substrate should be avoided, to minimize contamination, while faeces collected in the sea (Jarman et al. 2004), or potentially in fresh water, may again be contaminated with planktonic organisms (see above).
The temperature at which the predators have been frozen has ranged from –20 °C to –80 °C, and we would recommend the latter where possible to accelerate freezing and halt DNA-destroying enzymatic processes. It is sometimes possible to extract DNA from the predators immediately, without preservation (Cuthbertson et al. 2003), or to store in ethanol and then freeze at –20 °C (Hoogendoorn & Heimpel 2001). Additional techniques that could be applied to the storage of predators come from the literature on the handling and storage of faecal samples for phylogenetic analysis. Faecal samples share many of the same difficulties as gut samples; the DNA is often degraded, the risk of contamination is high and it is likely that PCR inhibitors may be co-extracted. Methods for the storage of faecal samples have included commercial kits such as RNAlater (Ambion) (Nsubuga et al. 2004), two-step storage using ethanol and silica (Roeder et al. 2004), storage in dimethyl sulphoxide (DMSO) salt solution (Seutin et al. 1991; Frantzen et al. 1998) and storage in extraction buffers from commercial kits (Hajkova et al. 2006).
Two factors affect the successful amplification of prey DNA from the guts of invertebrate predators and from faecal material: the amount of target DNA present in the sample and the quantity of that DNA remaining after storage of the predator, dissection (if applicable), then extraction and purification of the DNA. Optimization of each of these steps can enhance the success and consistency of any molecular analysis of predation.
To dissect or not to dissect
Large invertebrate predators such as carabid beetles contain too much tissue for most DNA extraction methods, so the gut may need to be dissected (e.g. Foltan et al. 2005; Harper et al. 2005, 2006; Sheppard et al. 2005), or the amount of predator tissue can be reduced by the removal of legs, elytra and wings (Zaidi et al. 1999), potentially improving sensitivity. It is worth repeating that a major danger at this stage is cross-contamination between predators and this can only be avoided by the use of sterile, DNA-free instruments and clean laboratory practice. Where dissection is impractical (very small predators such as Collembola and mites) or difficult (for example centipedes and some beetle larvae), DNA can be effectively extracted from whole predators (Cuthbertson et al. 2003; de Leon et al. 2004; Read et al. 2006; Juen & Traugott 2007). Predators can be homogenized in extraction buffer using DNA-free pestles or, where large numbers need to be processed, in a ball mill (e.g. Mixer mill MM 301, Retsch).
Several studies have now demonstrated that, where time and resources allow, a combination of visual analysis of prey remains in faeces, and PCR, is optimal for establishing trophic links (Casper et al. 2007a, b). Larger quantities of faeces should therefore be retained for visual analysis than would be needed for DNA extraction alone.
Thanks to medical and veterinary applications, a number of kits are available for the extraction of DNA from faeces such as the QIAamp DNA Stool Mini Kit, QIAGEN (e.g. Deagle et al. 2007) or the Ultra Clean Fecal DNA Isolation Kit, MO BIO Laboratories, Inc. (Casper et al. 2007a). It is advisable to extract several subsamples from each scat to increase prey DNA detection rates (A. Juen & M. Traugott, unpublished data). These kits can prove to be effective at overcoming problems of co-extracted PCR inhibitors from invertebrate guts and were successfully used by Foltan et al. (2005) to analyse decaying and scavenged insect and mollusc remains.
The importance of the inclusion of negative controls during DNA extraction cannot be overstated. These controls screen for potential contamination by prey DNA between samples and provide a higher degree of confidence in the assay protocol as a whole.
Nuclear vs. mitochondrial DNA
Both single and multiple-copy DNA regions from both nuclear and mitochondrial genomes have been used for molecular detection of predation (Table 1). After early attempts using nuclear DNA regions, including randomly amplified polymorphic DNA-derived sequence characterized amplified region (SCAR) markers (Agustíet al. 1999), internal transcribed spacer region 1 (ITS-1, Hoogendoorn & Heimpel 2001) and α esterase genes (Zaidi et al. 1999), the majority of recent studies have used mitochondrial DNA (mtDNA) genes as the source of their target-specific primers. There are two main reasons for this. First, hundreds or thousands of copies of the mitochondrial genome may be present within each invertebrate cell (Hoy 1994), greatly increasing sensitivity and hence the probability that prey DNA can be amplified from a predator's gut. Second, there are many published sets of ‘universal’ primers available for the amplification of mtDNA genes (Folmer et al. 1994; Simon et al. 1994, 2006), facilitating the rapid screening of suitable regions from both predator and prey species. From these sequences, prey-specific primers can be designed. Single- or low-copy number nuclear markers, such as most SCARs, are still being used occasionally (de León et al. 2006; Zhang et al. 2007a) but generally show lower sensitivity and are best avoided. The ITS region is probably best avoided because it is subject to intra-individual, as well as intraspecific, variation (Parkin & Butlin 2004). However, the multiple-copy nuclear 18S and 28S ribosomal genes have proved to be useful targets for developing group-specific primers in marine systems (Jarman et al. 2005; Martin et al. 2006; Suzuki et al. 2006; Deagle et al. 2007).
Table 1. Summary of molecular analyses of predation involving terrestrial and marine invertebrate predators and prey, with details of target gene regions and amplicon sizes
The choice of mtDNA gene/region will depend on whether the target is a group of species (e.g. a whole order, family or genus) or a particular species. Protein encoding genes, such as the cytochrome oxidase I and II genes (COI and COII), are less conserved than some other genes and are often appropriate for the design of species-specific primers. The two ribosomal RNA (rRNA) genes within the mitochondria, 12S and 16S, have slower substitution rates and are therefore generally better for the design of group-specific primers (Ballard 2000; Mueller 2006). These ribosomal genes have many indels (insertion/deletion mutations), which can make alignment difficult, but which can provide useful species-specific markers. For example, a 12S primer was developed that was specific to the arionid slugs but, fortuitously, amplified a different sized product for each species tested (Dodd 2004; Harper et al. 2005).
Despite this, the choice of gene region for primer design will in practice depend on the levels of variation found within the target prey group. For instance, Dodd (2004) found that the 12S rRNA gene was more useful than COI when designing primers for species-specific detection of predation on the slugs Deroceras reticulatum and the Arion hortensis agg. Conversely, high levels of intraspecific sequence diversity in the COI gene of lumbricid earthworms foiled attempts to design species-specific primers in this important prey group (Harper et al. 2005, 2006; Admassu et al. 2006), while a lack of suitable primer sites in the 12S gene allowed only for the design of group-specific earthworm primers (Harper et al. 2005). Therefore, before attempting to design primers for any species or group of species, it is advisable to sequence several mtDNA gene regions in order to identify the one most appropriate for the target prey species.
An alternative multicopy region that has been used successfully in predation studies is the nuclear 18S rRNA gene (Jarman et al. 2004). The ribosomal gene cluster, comprising the 5.8S, 18S and 28S rRNA gene, two internal transcribed spacers (ITS) and an external transcribed spacer, is a tandemly repeated region that is found in several hundred copies within nuclear genomes (Beebee & Rowe 2004). The 18S gene has been used extensively in phylogenetic studies and many sequences are available on databases such as GenBank. Using only 18S rRNA gene sequences from public databases, Jarman et al. (2004) were able to design group-specific primers for the amplification of a wide range of potential prey groups found in marine ecosystems. (To avoid any possible confusion, when we design primers for the rRNA genes, we are actually targeting the DNA that codes for these genes, not the RNA produced by the ribosomes in the mitochondria or nucleus).
How many individuals should be sequenced before primer design?
Many recent studies have shown the presence of cryptic species complexes in invertebrates (Hebert et al. 2004; Bickford et al. 2007). This raises questions as to how many individuals per species need to be sequenced and from how many different populations. If primers will be used on numerous field sites, then it would be advisable to sequence several individuals from each of the proposed sites in order to measure the extent of variation within target prey species. Few studies state the number of individuals initially sequenced or the levels of intraspecific variation found. Even within sites, extensive variation can be found, as shown in lumbricid earthworms (Harper et al. 2005, 2006; Admassu et al. 2006; R. A. King, A. L. Tibble, W. O. C. Symondson, unpublished data). This variation can make it difficult to design primers that will amplify all lineages or even haplotypes within a prey species (R. A. King, A. L. Tibble, W. O. C. Symondson, unpublished data). If inadequate preparatory sequencing is performed, there is a danger that substitutions at the primer sites may prevent some haplotypes being detected, leading to false-negatives during predator screening. For this reason and others, it is advisable to do your own sequencing, rather than basing your primer design solely on sequences taken from GenBank or other published sources.
Another potential danger, using mtDNA, is the presence of nuclear copies of mitochondrial genes (NUMTS) in many species (Bensasson et al. 2001). For instance, there are numerous nuclear copies of both COI and COII in Sitobion aphids, many of which are important crop pests (Sunnucks & Hales 1996). Techniques for the detection of NUMTS and methods for avoiding PCR amplification of NUMTS exist and have been reviewed elsewhere (Zhang & Hewitt 1996; Bensasson et al. 2001), and these techniques should be used if their presence is suspected. Designing primers to such NUMTS is best avoided, as they are likely to be single- or low-copy number markers, and this will adversely affect post-ingestion detection times.
Although the library of available primers targeting DNA of specific prey taxa is steadily growing, new primers usually have to be designed. A simple flow diagram showing the steps involved is shown in Fig. 1a. Primer design is based on alignment of sequences from target prey, nontarget prey and predators. The process can be assisted by inclusion of sequences from GenBank when available, particularly when looking for group-specific primer sites. Alignment and display can be carried out using specific software packages such as bioedit (Hall 1999) or lasergene (DNAStar). In multiple-species systems involving highly generalist predators in biodiverse ecosystems, it is not possible (or necessary) to include sequence information for all potential prey taxa in the alignment. Major alternative prey might be included, especially any species closely related to the target. By aligning target prey and nontarget DNA sequences, primer sites can be identified. For general rules on primer design, see Hawkins (1997) or Apte & Daniel (2003). Primer design software such as primer 3 (Rozen & Skaletsky 2000), primerpremiere (PREMIER Biosoft International) or lasergene (DNAStar) can be used to identify suitable primer sites from multiple-sequence alignments.
Specificity depends critically on the 3′ end of the primer, because extension of the DNA strand only occurs when this end of a primer is fully matched to the template sequence (Hawkins 1997). Therefore, primers that contain several mismatches at the 3′ end are desirable. Wherever possible, primers used for prey DNA detection should have high melting temperatures (Tm > 55 °C) as this allows PCRs to be run with high annealing temperatures (Ta > 60 °C). This reduces the risk of nonspecific amplification and primer annealing to false priming sites, as well as increasing PCR efficiency.
Group-specific primers are sometimes more useful than species-specific ones where, for example, the ecological interaction in question only requires information on predation at this level (e.g. predation on any species of annelid worm, pulmonate mollusc or ‘aphid’). These primers can be used efficiently to screen predators for predation on a ‘group’ followed by re-screening with a set of species-specific primers to species within that group (Fig. 1b). Group-specific primers are also used extensively before a more detailed analysis of gut contents using cloning and sequencing. Such intensive effort is particularly appropriate in vertebrates (Sutherland 2000; Rollo et al. 2002; Jarman et al. 2004; Deagle et al. 2007) and also invertebrates (Blankenship & Yayanos 2005; Deagle et al. 2005b), where the aim is to obtain maximum information on the dietary range of a relatively small number of individuals. Group-specific primers are designed using multiple sequence alignments to identify sites that are conserved within group but unique between groups (e.g. Jarman et al. 2004, 2005). Degeneracy can be tolerated at the 5′ end of the primer, but mismatches at the 3′ end should always be avoided. Apart from commercially available software (see above), amplicon (Jarman 2004) is a software package for designing group-specific PCR primers sets that is freely available on the Web.
Prey DNA detection success in gut and faecal samples has (in most instances) been enhanced by targeting short DNA fragments (< 300 bp) (Agustíet al. 1999; Zaidi et al. 1999; Chen et al. 2000; Hoogendoorn & Heimpel 2001). This is because DNA molecules are broken into smaller fragments during digestion. For that reason, group-specific primers should be designed to be within this size range (e.g. Sutherland 2000; Harper et al. 2005; Jarman et al. 2005). It has been shown, for example, in sea lion faeces that there is a rapid decrease in copy number of prey DNA as fragment size increases (Deagle et al. 2006). However, PCR efficiency is also determined by factors such as the quality of the template DNA extract, PCR reagents, cycle conditions and the efficiency of the primers used. Thus, optimized PCR assays will sometimes allow detection of larger prey DNA fragments, even up to 500 bp, for extended time periods post feeding (Juen & Traugott 2005, 2006, 2007). However, to optimize detectability, shorter fragments < 300 bp should be targeted wherever possible (Table 1).
Assay optimization and evaluation
Optimization to minimize cross-amplification of nontargets
Primers play a key role in determining the sensitivity and specificity of the PCR (He et al. 1994), both critical for accurate analysis of the heterogeneous mixtures of degraded DNA in the guts of predators. Those designed to amplify DNA from a specific prey taxon should have high melting temperatures and be run at the highest annealing temperature possible without compromising amplicon yield. The optimum annealing temperature has to be determined empirically, ideally using a temperature gradient PCR machine. It is advisable to determine this temperature using a low-target DNA concentration, emulating the low quantities of prey DNA that have to be detected in gut samples. Sensitivity and specificity of PCR also depends on other parameters including primer and Mg2+ concentration, polymerase enzymes, cycle number and PCR enhancing agents such as DMSO and bovine serum albumin (BSA, Roux 2003).
The sensitivity of a diagnostic PCR assay to identify prey consumption is best determined by feeding experiments to establish how long after prey ingestion prey DNA can be detected within predator guts (see below). The sensitivity of an assay can be evaluated using serial dilutions of prey DNA in a constant concentration of predator DNA (Chen et al. 2000; Admassu et al. 2006; Traugott et al. 2006). Such tests help to identify, for example, differences in primer sensitivity that can be useful when interpreting detection rates among different prey taxa.
Cross-amplification tests on nontarget organisms
It is important to empirically test whether a PCR assay, designed to detect a specific prey, also amplifies DNA from the predators themselves or nontarget organisms that might be consumed by the predators in the system under study. Some compromise is inevitable here, given that field sites will normally contain thousands of species (including microorganisms). Species chosen for cross-reactivity tests should include those known from the literature to be consumed by the predator plus representative species from other major taxonomic groups. The number of species chosen depends to some extent upon the complexity of the system under study, but should be as large as is practicable. For example, Juen & Traugott (2007) and Harwood et al. (2007) tested their primers against 93 and 84 nontarget invertebrate taxa, respectively. To avoid the danger of false-negatives, the DNA from these nontarget organisms should also be tested with general invertebrate primers, to ensure that they contain amplifiable DNA. If prey-specific primers do cross-amplify, but the amplicons from nontarget species are clearly of a different size to that of the target prey, this may not matter, as long as they are not dominating the PCR and potentially reducing detectability of any target DNA in the same gut sample. Within multiplex PCRs, where amplicons of several sizes are generated, these nonspecific fragments may, however, interfere with prey fragment assignment and thus extra bands should be avoided whenever possible.
Screening of predators
The importance of controls
As in all PCR work, but especially in diagnostic applications, it is essential to include positive and negative controls, which indicate whether the reaction was successful and whether DNA contamination occurred. As positive controls, DNA from the organisms targeted by the primers should be used. It is advisable to minimize the amount of target DNA in the positive controls to ensure robust amplification and to reduce the chances that this DNA might itself serve as a source of contamination (Neumaier et al. 1998). The best positive control might be DNA from a predator fed with the target prey or a mixture of predator and prey DNA.
Negative PCR controls are critical and serve to check for sample-to-sample contamination and contamination of PCR reagents. The negative controls should include all reagents except the DNA, which is substituted by PCR water. It is strongly recommended that several negative controls be used for each PCR (n > 5) as low levels of DNA contamination can result in random amplification patterns (i.e. some controls may show a band while others do not). In addition, DNA from the predator may be used as a negative control, ensuring that nothing endogenous to the predators is being amplified.
False-negatives can arise when prey DNA cannot be amplified because of failures during DNA extraction, the presence of PCR inhibitors or simply errors during PCR. To avoid this type of error, all samples should ideally be tested in parallel with an internal control in the form of a primer that will amplify DNA from any invertebrate, including the predator. This has been carried out, for example, by Zaidi et al. (1999) with primers targeting the actin gene; only samples that were actin-positive were included in the results. De León et al. (2006) used a 28S primer in a similar way. Multiplexing allows primers targeting the predator and prey to be amplified simultaneously (Juen & Traugott 2006, 2007).
The high sensitivity of PCR is a double-edged sword as it also facilitates the amplification of minute quantities of contaminating DNA. PCR to detect prey remains has to follow the same stringent contamination control strategies as used by those working with ancient DNA. Contamination-free conditions can be best ensured by physically separating pre- (DNA extraction, PCR preparation) and post-PCR (PCR execution, visualization of PCR products) activities and the workflow should always be from pre-PCR to post-PCR areas. Separate sets of pipettes should be used with filter tips. For further information on setting up a contamination-free environment in a PCR laboratory, see Neumaier et al. (1998).
Overcoming PCR inhibition
PCR inhibition can lead to false-negative results. Inhibitors may originate from the invertebrate or faecal sample itself or from the environment from which it is taken (Rossen et al. 1992). Modern DNA extraction and purification kits have greatly reduced this problem which can be further reduced using PCR facilitators such as BSA, Betaine, or TritonX-100. Juen & Traugott (2006) found that addition of BSA greatly improved amplification of DNA from soil-dwelling predators. Inhibitor-tolerant thermostable polymerases can also be used to overcome PCR inhibition (Rådström et al. 2004). Simply diluting the DNA extract is an easy, and in many cases successful, approach to overcoming PCR inhibition (Muelhardt 2000), but cannot be applied when using PCR to detect prey remains, simply because concentrations of undigested prey DNA will usually be low and any further dilution will jeopardise detection.
Singleplex and multiplex PCR
If predators or faecal material need to be screened for many prey taxa, a number of separate singleplex PCRs may be conducted, one PCR for each prey type (Agustíet al. 2003a). If the predator needs to be tested for many targets, this becomes a lengthy, costly and tedious process, and effectively precludes the analysis of the hundreds of predators that might be required for a meaningful ecological study.
Multiplex PCR offers a more rapid approach. The use of fluorescent primers, to improve sensitivity, to allow separation of fragments only 1 bp apart and to allow multiloading of gel channels, was first exploited in predation studies by Dodd (2004) and reported later in Dodd et al. (2005). This was further developed by Harper et al. (2005), who showed that it is possible to simultaneously amplify mitochondrial DNA fragments of up to 12 different prey from the gut content of invertebrate predators within a single PCR. Fluorescently labelled primers were used to separate PCR products by size on a sequencer and record the results on electropherograms. Multiplexing (without fluorescent labelling) has also been used to simultaneously test for PCR inhibition and check for false-negative results (Juen & Traugott 2006, 2007). By including predator-specific primer pairs, it is possible to identify the producer of faecal material or the identity of the predator (Juen & Traugott 2006). By comparing prey DNA detection rates in laboratory-fed beetle larvae assayed with either singleplex (targeting prey DNA only) or multiplex PCR (targeting both prey and predator DNA), Juen & Traugott (2006) found that prey DNA detection success was not significantly different between the two PCR methods.
Harper et al. (2005) were greatly assisted in their work by the availability of multiplex PCR kits (QIAGEN). These kits are highly tolerant to differences between primers in, for example, annealing temperatures. Indeed, the multiplex PCR mastermix provided in these kits can often be used to get ‘difficult’ primers to work in singleplex, overcoming problems of inhibition.
Visualizing of PCR products
Gel systems and fluorescent primers
When the aim is to measure predation on a single target pest species, agarose gel electrophoresis of prey-specific PCR amplicons will usually be adequate. However, as the number of target prey species increases, so will the time and effort needed to screen each predator for multiple prey items. Given adequate size differences between the amplicons for different prey targets, agarose gels can still be used to detect multiple prey within a single predator following multiplex PCR (Juen & Traugott 2007). However, when differences between amplicon sizes are smaller, polyacrylamide gels can give a higher resolution. Other high-resolution gel systems, including eGene (http://www.egeneinc.com/) and Elchrom (http://www.elchrom.com), are also available. For analysis of many targets within highly generalist predators, multiplex PCR utilizing fluorescently labelled primers and separation on highly sensitive DNA sequencer-based detection systems allows the simultaneous detection of multiple prey species from a single predator (Harper et al. 2005).
DGGE and TGGE
Recent studies have utilized ‘universal’ and group-specific primers for detection of prey DNA on temperature or denaturing gradient gels (TGGE and DGGE) (Deagle et al. 2005a, b; Harper et al. 2006; Martin et al. 2006). These gels allow amplicons of the same length to be separated on the basis of differences in their sequence and may be useful when trying to distinguish closely related species for which specific primers could not be developed. This approach is potentially excellent for examining predator responses to prey diversity, overall or within prey groups (Harper et al. 2006). Control samples must be run on the same gel, in parallel, in order to identify the prey species consumed (Harper et al. 2006). PCR artefacts, as encountered by Deagle et al. (2005b), may complicate the interpretation of gels. TGGE and DGGE can be difficult techniques to master and can give confusing results where, for example, haplotype diversity within target species may produce very different band mobilities on the gels (Harper et al. 2006). Although others may have had more success, in our own experience it is difficult to get consistently good gels using these techniques for gut analyses and, in view of the other problems outlined, this is not an approach we could recommend.
It is not usually possible to quantify the amount of prey DNA amplified from the gut or faeces of a predator using conventional PCR. Deagle et al. (2005a, 2007) used both clone libraries and quantitative PCR (qPCR) to estimate proportions of the different fish species in captive sea lion faeces. Both studies showed that the prey proportions estimated from faecal DNA samples corresponded reasonably well with the known dietary proportions. Zhang et al. (2007b) went further, quantifying the number of copies of their DNA target that equated to a Bemisia tabaci egg, nymph or adult. They then screened a range of predators from the field and estimated the numbers of nymph equivalents present in the guts of each predator from the DNA copy number. They also showed that qPCR, using TaqMan, improved sensitivity compared with conventional PCR. Two recent studies, both using qPCR, have attempted to quantify ingestion rates of algal-feeding zooplankton species. Troedsson et al. (2007) demonstrated that, in the appendicularian Oikopleura dioica, algae of varying sizes were trapped and ingested at different rates, while Nejstgaard et al. (in press) quantified feeding rates in three calanid copepod species in both laboratory and seminatural conditions.
Before quantitative methods can routinely be used for molecular detection of predation, several barriers have to be overcome. Given that DNA copy number will often be very different between species and inevitable differences in primer efficiency, calibration of such quantitative methods will require considerable effort, especially when the aim is to detect the remains of many prey targets simultaneously in highly polyphagous predators. For instance, Nejstgaard et al. (in press) found differences in 18S rDNA copy number between different aged cultures of Emiliania huxleyi which has major implications for the choice of standards for qPCR. There are also questions as to the applicability of quantitative approaches outside of the controlled settings under which the above studies have been conducted. It is not clear how quantitative approaches should be calibrated when applied in field situations and further experiments are also needed to investigate the effects of meals of different ages within the same predator on qPCR signal. At best, it can provide a semiquantitative measure of the biomass of undigested prey at the time of analysis (a valuable quantitative measure in its own right), but not the biomass consumed and certainly not the number of prey killed and/or eaten. High-specification qPCR machines can detect many different fluorescent labels simultaneously, offering the prospect of multiplex qPCRs in future predation work. At its simplest, qPCR can be recommended for quantifying predation on a single target prey species using a single primer, but only in well-calibrated systems, when levels of predation are likely to be low. Such a system could be used in a similar way to enzyme-linked immunosorbent assays using monoclonal antibodies, to obtain data on the relative quantities of prey consumed over time and between locations or field treatments (e.g. Symondson et al. 1999).
Calibratory feeding experiments to determine DNA survival during digestion
Before a new PCR primer can be used to assess predation in the field, its ability to detect prey DNA in the gut of a predator has to be assessed (Fig. 1a). For invertebrates, this is achieved by carrying out a feeding trial, in which predators are fed the target prey and then killed at various time points post-ingestion of the prey. After DNA extraction, the predators are screened for prey DNA, and the number testing positive at each time point can be expressed as a percentage or proportion. As the ability to detect DNA in the guts of invertebrate predators can be affected by temperature (Hoogendoorn & Heimpel 2001), feeding trials should be conducted in a controlled environment emulating conditions (day/night cycles, temperature) found at the field site (e.g. Greenstone et al. 2007). Predator activity levels may also influence digestion rates; high activity has been shown to increase the production of haematin (a by-product of digestion) in tsetse flies (Loder et al. 1998). Therefore, if feeding trials are conducted indoors, predators should be maintained in a light : dark cycle that is similar to field conditions. Many invertebrate predators are nocturnal, so it is advisable to begin the feeding trials (by feeding the predators) during the period when predation is likely to be occurring in the field.
Feeding trials are usually conducted using field-caught predators, which reflect the ranges of ages, sizes and physiological states that will later be encountered when screening predators for predation. Previous feeding trial studies have starved the predators from 48 h to 14 days before the start of the feeding trial (de León et al. 2006; Harper et al. 2006). This is advisable to ensure predators have empty guts at the start of the feeding trial and equal hunger levels. It also increases the chances that most of the predators will consume the prey when presented with it, facilitating the experiment greatly. However, this may not be analogous to conditions in the field, as many predators will have the remains of previous meals in their guts before feeding. The consumption of alternative prey after feeding on target prey has been shown to increase prey DNA detection times in laboratory feeding trials (Dodd 2004). So far, no studies have investigated the influence of the consumption of alternative prey before feeding on target prey, but it is likely this will have a similar effect. In some predators, such as spiders, the quantity of food eaten (or the degree of hunger) can affect metabolic rates (Anderson 1974), probably explaining the relatively long post-ingestion detection periods found for spiders not fed alternative prey in feeding trials (Sheppard et al. 2005). Similarly, in a study using C-14 labelled food, starvation after feeding by the collembolan Tomocerus flavescens caused an increase in gut retention times (Wolters 1985).
Most previous studies have involved feeding predators in Petri dishes or vials, with damp filter paper or cotton wool as a substrate (Agustíet al. 1999; Chen et al. 2000; Greenstone et al. 2007). This approach avoids contamination of the predator with extraneous material, and ensures that the predator can find the prey quickly and that the experimenter can observe and confirm predation taking place. Once feeding has occurred, the predator should be moved to a more natural environment and substrate with suitable refugia, otherwise stress might affect digestion rates. Food intake during the feeding period is either limited to a set number or weight of prey, or feeding is ad libitum. Limiting food intake is preferable, reducing a potential source of error in the results and probably reflecting better the limited prey availability predators may often experience in the field (Lovei et al. 1985). For example, the mean foregut biomass of the carabid beetle Pterostichus melanarius caught in the field over 5 years was ~7.5 mg (Symondson et al. 2002a), whereas when allowed to feed ad libitum on slugs in the laboratory for 2.5 h, the mean foregut biomass was much greater (females 22.4 mg, males 16.7 mg) (Symondson et al. 1999). Meal size has been shown to influence the detection of prey DNA in the gut of predators, where larger meals caused an increase in prey detection times (de León et al. 2006), although other studies have observed no correlation (Zaidi et al. 1999; Hoogendoorn & Heimpel 2001; Juen & Traugott 2005).
The number and range of the postfeeding sampling times depends upon the type of predator being studied and must be planned from the start. Maximum detection times have ranged from a few hours to 5 days postfeeding (Chen et al. 2000; Sheppard et al. 2005). Greenstone et al. (2007) found that the detection times for prey DNA in the guts of predators that process their prey in very different ways can vary considerably. Colorado beetles, Leptinotarsa decemlineata, were fed to a soldier bug, Podisus maculiventris, a fluid feeder that pre-orally partially digests its prey, and a ladybird, Coleomegilla maculata, a predator that consumes macerated prey. The mean detection period was seven times longer in the former than in the latter. Prey DNA in the spider Tenuiphantes tenuis fed on the aphids Sitobion avenae was still being detected 120 h after ingestion (Sheppard et al. 2005). The more time points that can be included in a feeding trial, the more accurately the decay of prey DNA can be modelled. The number of replicates at each time point in previous feeding trials has ranged from 1 (Cuthbertson et al. 2003) to over 30 (Juen & Traugott 2006), but we would recommend no less than 8–10 to allow for the many sources of error associated with predator physiological state and quantities of prey consumed.
Calibratory feeding experiments have also been applied to vertebrates, although time periods cannot be very precise as one has little if any control over when (or where) a vertebrate will defecate. With vertebrates, the aim is usually to determine whether the DNA survives complete transit through the gut. Deagle et al. (2005a) examined the effects of feeding different quantities and ratios of prey to sea lions and used clone numbers (and later real-time PCR, Deagle & Tollit 2007) to quantify prey detectability. The main parallel with invertebrate work is to determine how long, postfeeding, prey DNA can still be found in fresh faeces (Deagle et al. 2005a; Casper et al. 2007b), and this is recommended where practicable, before application in the field.
One method for comparison of feeding trial data, for each predator–prey and primer combination, is to calculate a median detection time or molecular half-life (Greenstone & Hunt 1993). This is defined as the time after feeding when 50% of the predators test positive for the target prey DNA. For this to be calculated, the feeding trial data need to be fitted with a regression model that describes the decline in positives over time. The most frequent method of analysis has been to fit a linear regression to the number of positives against time since feeding (Agustíet al. 2003a; Harper et al. 2005, 2006; Sheppard et al. 2005; Read et al. 2006), although Probit models (Chen et al. 2000; Ma et al. 2005; Greenstone et al. 2007) and logisitic regressions (Foltan et al. 2005) have also been used. Additionally, 95% confidence limits can be calculated for the observed amplification success at specific time points after feeding (Juen & Traugott 2007) and differences in prey DNA detection compared by G-test (Dytham 2003). A comparative weighting can then be placed on prey positives obtained from field-collected predators, so that predation by predators with long detection times is not overestimated in comparison with predators with shorter detection times.
Scavenging and secondary predation
The fact that a predator tests positive for a target prey species does not necessarily mean that it killed that prey animal nor even that it intended to eat it. Scavenging has been shown to be a potential cause of false-positives for predation by carabids feeding on slugs (Deroceras reticulatum) and aphids (Sitobion avenae) (Foltan et al. 2005), and chafer larvae (Melolontha melolontha) (Juen & Traugott 2005). Foltan et al. (2005) found that DNA could still be amplified from 50% of dead slugs and aphids, placed on field soil, after 175 h and 134 h, respectively. They went on to show that these dead and decaying slugs were readily consumed by carabids after more than a week and that the DNA from scavenged slugs could be amplified from the guts of the beetles. Juen & Traugott (2005) obtained similar results with chafer grubs. Both studies showed that acceptance of dead prey was negatively correlated with carrion age. DNA-based detection of prey cannot currently distinguish between scavenging and predation, and to date, nobody has managed to effectively monitor the availability of dead prey in the field. In some studies, it may not matter whether the prey are dead or alive when eaten, if the aim is simply to identify sources of predator nutrition. However, in most cases, the main interest is in the population dynamics of the predator–prey interaction and for that we really do need to know who is killing whom and in what numbers.
Another source of error can be secondary predation (Harwood et al. 2001). Secondary predation errors are a consequence of a predator consuming a second predator, shortly after the latter has consumed the target prey. In a DNA-based study, secondary predation was found to be a significant potential source of error (Sheppard et al. 2005). In the worst case a secondary predator (the carabid Pterostichus melanarius) tested positive for aphid 4 h after eating the primary predator (the spider Tenuiphantes tenuis), that itself had been digesting its aphid prey (S. avenae) for 4 h. It is likely that such sensitivity to secondary predation was helped by the long post-ingestion detection times recorded for the spiders, with a median detection period of ~60 h, twice as long as that recorded for the beetles. Thus, the importance of secondary predation as a potential source of error may depend critically upon the digestion rates of the predators involved.
As far as we know, the problems of scavenging and secondary predation have not as yet been addressed in vertebrate systems, but are probably an equally serious source of error.
The use of gel-based PCR detection systems to monitor predation on single or small numbers of prey species will continue to expand. However, the real prize will be to exploit invertebrate genomics to develop mass-target detection systems, allowing us to identify the complete prey range even of highly generalist species in biodiverse communities. This is important, because rates of predation on a target species (such as an agricultural pest) are strongly affected by prey choice and the relative densities of alternative prey species (Symondson et al. 2002b). How far we go down this route is currently limited mainly by cost rather than technology. To date, the nearest approach has been through the use of multiplexing with fluorescent-labelled primers (Harper et al. 2005), which managed to detect and identify up to 12 targets simultaneously. However, microarray technology is now well established, with the potential to identify tens of thousands of targets on a single surface-based chip. There are many formats available and these are used extensively to detect viruses and bacteria in medical diagnostics (reviewed in Striebel et al. 2003) as well as the identification of fungi and other plant pathogens (Lievens et al. 2005; Szemes et al. 2005). Bead-based microarrays are also available, in which DNA is hybridized to fluorescent-labelled reporters on the beads which are read individually by laser (Armstrong et al. 2000). At present, up to 100 distinct bead populations can be generated, more than enough for most predation work. Where necessary microarrays can be constructed that are capable of detecting species-specific single nucleotide polymorphisms (SNP, Pastinen et al. 2000; Erdogan et al. 2001), that will separate closely related taxa where suitable primer sites cannot be found. Such an approach could allow predation to be monitored not just on different species but even on individual haplotypes. Such a tool would permit comparison between genotypes existing in the field and those found in the guts of predators, allowing direct study of genetic selection through predation.
Another limitation on the use of such technology is the need to sequence and find primer sites for all the target species. In time, these will become available, but this process could be accelerated by use of mass pyrosequencing, which is capable of detecting many thousands of sequences simultaneously from mixed samples (Margulies et al. 2005). This approach is already being used to mass sequence marine bacteria (Goldberg et al. 2006). At present, it is prohibitively expensive, costing thousand of pounds per run, but may one day become cheap enough to use for direct screening of predator gut and faecal samples.
However, the main task ahead is to start applying molecular analysis techniques to study complex trophic interactions in the field. The potential is enormous but at present we have barely scratched the surface. A great deal of valuable background work in going on, developing primers, characterizing their use in the laboratory, developing new quantitative methods, factoring in, for example, the effects of temperature, amplicon size, primer efficiency, meal size, predator activity and sex, but all this now needs to be applied to address fundamental ecological questions. The few PCR-based field studies to date have done little more than record semiquantitatively a few trophic links. There are notable exceptions. For example, the study by Kasper et al. (2004) showed, through field data, the degree of prey overlap and resource partitioning between a native and alien wasp. Resource partitioning, between closely related birds feeding on Lepidoptera, was also addressed by Sutherland (2000). The study by Agustíet al. (2003a) compared the proportions of different prey (Collembola) in the field with the ratios found in predator (spider) guts and was thus able to demonstrate clear prey choice. DeWoody et al. (2001) were able, with the aid of microsatellites, to study filial cannibalism in fish. We have some way to go yet but potentially molecular analysis of the diets of generalist predators, vertebrate or invertebrate, should allow us to construct quantitative foodwebs, similar to those currently published for host–parasitoid interactions (e.g. Henneman & Memmott 2001). Major areas with future potential include molecular tracking of the diets of individuals from faeces (individual profiling of DNA from the predator (Reed et al. 1997) combined with use of prey-specific markers), analysis of herbivory, and evolutionary processes such as selection among prey genotypes by predators (Taberlet & Fumagalli 1996). Molecular analysis of predation can also make major contributions in the more practical fields of biological control and conservation, and is clearly beginning to do so. Given all this potential, the current almost exponential rise in the numbers of papers published in this field is likely to continue.
This review has emerged from the research group of Dr Bill Symondson at Cardiff University, a group that specialises in the development and application of new techniques for studying predator–prey interactions in the field. The authors include Dr Michael Traugott (now developing a similar research group at Innsbruck), Dr Andrew King (senior postdoc developing multiplex approaches to analysis of the spatial dynamics of predation by carabid beetles), Dr Daniel Read (working on micro-arthropod predators and soil foodwebs) and Dr Bill Symondson (Reader in Invertebrate Molecular Ecology). All authors contributed equally.