Introduction
Natural areas close to agricultural land are valuable sources of natural enemies. Such ‘hinterland’ provides the opportunity for natural enemies to re-invade fields following potentially destructive agricultural activity, such as the use of pesticides or simply from agronomic practices. It has been known for some time that increased landscape complexity, incorporating areas of non-farmed habitat (field margins, hedgerow, woodland, etc.), often results in greater numbers of natural enemies in adjoining farmland (reviewed in Bianchi et al., Reference Bianchi, Booij and Tscharntke2006; Chaplin-Kramer et al., Reference Chaplin-Kramer, O'Rourke, Blitzer and Kremen2011), although beneficial effects in terms of reduced pest damage are rarely reported. This may in part be because such areas can also provide refugia for pests, for example slugs in field margins (e.g., Eggenschwiler et al., Reference Eggenschwiler, Speiser, Bosshard and Jacot2013) or aphids on their non-crop alternative host plants. Beneficial effects of natural areas relate primarily to dispersal of generalist predators into cropped areas (e.g., carabid beetles, Thomas et al., Reference Thomas, Holland, Brown and Holland2002). By contrast, several studies have shown that more specialized parasitoid food webs are often highly compartmentalized, with few parasitoids from natural habitats parasitizing crop pests (Macfadeyn et al., Reference Macfadeyn, Gibson, Symondson and Memmott2011; Derocles et al., Reference Derocles, Le Ralec, Besson, Maret, Walton, Evans and Plantegenest2014).
Assessing diversity and identifying pests and potentially useful natural enemies in natural areas is labour-intensive, complicated by often high levels of heterogeneity among plant (and hence invertebrate) communities and sub-communities within such habitats. Here our aim was to see whether birds caught within large bocks of natural habitat (in this case reedbeds) could be used as natural sampling agents, potentially collecting their prey across sub-habitats within the ecosystem. Direct observation of birds foraging is possible (Tinbergen, Reference Tinbergen1960), but necessarily limited. Cameras at nest boxes can be used to identify insects being brought to chicks (e.g., Blondel et al., Reference Blondel, Dervieux, Maistre and Perret1991). Examination of undigested prey remains in regurgitated pellets, stomach contents or faeces may provide valuable clues to the range of prey taken by birds, but many prey are hard to identify in this way (e.g., many insect eggs and larvae) and the method is strongly biased against soft bodied prey which more rapidly disintegrate within the digestive tract. Stomach flushing can be used to wash prey out of the digestive tract, but is a relatively invasive technique. A disadvantage of these methods is that prey (particularly soft-bodied prey) may already be partially digested, hindering identification. Neck-ligatures can be used to collect ingested food from the buccal cavity of chicks (Rosenberg & Cooper, Reference Rosenberg, Cooper, Morrison, Ralph, Verner and Jehl1990). One advantage of this technique is that food items generally remain intact, facilitating identification, often to species level (Grim, Reference Grim2006). However, neck-ligatures can only be used effectively on nestlings. Such methods are at best invasive or at worst lethal, and are understandably becoming less acceptable.
Molecular biology is revolutionizing the in situ study of trophic relationships (Symondson, Reference Symondson2002; King et al., Reference King, Read, Traugott and Symondson2008; Pompanon et al., Reference Pompanon, Deagle, Symondson, Brown, Jarman and Taberlet2012). It has been clearly demonstrated that prey DNA can still be detected in the guts and faeces of predators after many hours (even days) of digestion (Symondson, Reference Symondson2002; King et al., Reference King, Read, Traugott and Symondson2008). It is now possible to rapidly screen large numbers of predators to identify DNA from the whole range of prey that may be present in their guts or faeces. Prey DNA can be amplified from predator faeces using general primers, followed by Next Generation Sequencing (NGS) (reviewed in Pompanon et al., Reference Pompanon, Deagle, Symondson, Brown, Jarman and Taberlet2012) or, less expensively, by cloning and sequencing, to identify the individual prey taxa present in the sample (Sutherland, Reference Sutherland2000; Jedlicka et al., Reference Jedlicka, Sharma and Almeida2013). This can be done by comparing the sequences with online databases, such as GenBank and the Barcoding of Life Database (BOLD). The rapid expansion of these databases has now made this a viable option for terrestrial invertebrates in relatively well-studied parts of the world such as Europe and North America. Even when exact matches are not possible, because the prey detected is not a species already available in the databases, a nearest match can be obtained, indicating the higher taxonomic level (e.g., family or genus) to which the prey taxa belong.
A high level of specificity is essential if pest species, and natural enemies known to have utility as biocontrol agents, are to be distinguished from close relatives. Molecular approaches have an advantage over morphological methods in that they allow detection and identification of very small, soft-bodied or morphologically cryptic prey taxa. Molecular diagnostics have been widely used to analyse the diets of both terrestrial and aquatic invertebrate and vertebrate predators, but rarely birds. The bird dietary work has been mainly on seabirds, including Glaucous Gulls (Larus hyperboreus) (Scribner & Bowman, Reference Scribner and Bowman1998), Macaroni Penguins (Eudyptes chrysolophus) (Deagle et al., Reference Deagle, Gales, Evans, Jarman, Robinson, Trebilco and Hindell2007), Little Penguins (Eudyptula minor) (Deagle et al., Reference Deagle, Chiaradia, McInnes and Jarman2010), European Storm Petrels (Hydrobates pelagicus) (Medeiros, Reference Medeiros2011) and Cory's Shearwaters (Alonso et al., Reference Alonso, Granadeiro, Waap, Xavier, Symondson, Ramos and Catry2014). Among terrestrial birds, molecular analyses have been used to identify invertebrate DNA in the faeces of tits (Parus spp.) (Sutherland, Reference Sutherland2000), Western Bluebirds (Sialia mexicana) (Jedlicka et al., Reference Jedlicka, Sharma and Almeida2013) and captive Carrion Crows (Corvus corone) (Oehm et al., Reference Oehm, Juen, Nagiller, Neuhauser and Traugott2011), and to identify prey species from bones in owl (Strigiformes) pellets (Poulakakis et al., Reference Poulakakis, Lymberakis, Paragamian and Mylonas2005; Kim et al., Reference Kim, Kim, Lee, Won and Lee2009). However, we know of no studies of prey DNA from song bird faeces other than that of Sutherland (Reference Sutherland2000) working on Blue Tits (Parus caeruleus) and Great Tits (P. major), and the work of Jedlicka et al. (Reference Jedlicka, Sharma and Almeida2013) (to order level only) on Western Bluebirds. None of these studies was designed to screen for the presence of pest and natural enemy species in natural habitats.
This proof of concept study was designed in part to establish how effective identification of prey from DNA in faecal samples might be, given the current coverage of invertebrate barcodes on BOLD and GenBank which are becoming increasingly comprehensive for invertebrate species in well-studied regions. The better the coverage the more practical it becomes to identify pest and natural enemy taxa to species level. To ensure that prey taxa were representative of the habitat in which the birds lived, and that sampling by the birds was not limited by species-specific prey preferences, we took faecal samples from representatives of three bird species. Reed Warblers (Acrocephalus scirpaceus) and Sedge Warblers (Acrocephalus schoenobaenus) are closely related trans-Saharan migrant songbirds, which differ markedly in their foraging ecology. Reed Warblers are thought to be dietary generalists and Sedge Warblers to be more specialist foragers (Bibby & Green, Reference Bibby and Green1981). Previous studies have investigated the diet of both of these Acrocephalus species using stomach flushing of migrating birds (Chernetsov & Manukyan, Reference Chernetsov and Manukyan2000), neck ligatures to retrieve food delivered to nestlings (Davies & Green, Reference Davies and Green1976; Grim, Reference Grim2006) or hard part analysis of nestling faeces (Bibby & Thomas, Reference Bibby and Thomas1985). As two of the most abundant bird species in Phragmites stands in NW Europe during summer and autumn, they are major consumers of arthropods in the reedbed ecosystem. Cetti's Warbler (Cettia cetti) is unusual among warblers in that it is resident in UK reedbeds throughout the year. The diet of Cetti's Warbler is known mainly from observations of foraging behaviour and analysis of hard-part remains in faeces (Cramp et al., Reference Cramp, Simmons and Perrins2004 and references therein). Previous studies have shown that all three of these warbler species consume a wide range of invertebrates including Araneae, Diptera, Trichoptera, Homoptera and Hymenoptera species. However, because of the difficulty of distinguishing morphologically similar and often semi-digested insect taxa, the majority of such studies have identified prey to family level only.
Here we used a sensitive and practical DNA barcoding-based approach, applied to reedbed warblers, that combines amplification of prey DNA from faeces with cloning and sequencing, allowing the simultaneous detection of predation on multiple invertebrate species. Although this was a proof of technique study, and small numbers of birds were analysed (n = 11 faecal samples from 11 different individuals at 3 reedbeds), we wished to determine whether the combined diets of the three species would provide us with useful information on the prey species present within the reedbeds, particularly the presence of any pests of agricultural crops and useful natural enemies.
Materials and methods
Sample collection and DNA extraction
Adult birds were sampled from May to August 2009 by mist-netting at three reedbed sites along the South Wales coast (table 1; Cosmeston Lakes Country Park 51°24′54″N, 3°11′4″W; Cardiff Bay Wetlands 51°27′32″N, 3°10′11″W and Magor Marsh 51°34′35″N, 2°49′48″W) during regular monitoring as part of the British Trust for Ornithology's ringing scheme (Robinson et al., Reference Robinson, Julliard and Saracco2009). Faeces were collected opportunistically from warblers during the bird-ringing sessions, immediately preserved in 100% ethanol on collection and stored at room temperature. Feather samples were collected (dropped by some birds as adventitious moult during handling) from the three warbler species to provide samples of their DNA and stored in dry Eppendorf tubes at room temperature. A reference collection of invertebrates for barcoding, representing species and higher taxonomic groups likely to form part of the prey spectrum of the three warbler species, was collected using sticky-trap sampling (yellow, double sided, effective area 100 cm2, Oecos, Hertfordshire, UK), from the three field sites (table 2) and stored in a −80°C freezer. Sequences from additional invertebrate species were obtained from GenBank (table 2).
Details of the individual Cetti's, Reed and Sedge Warblers from which faecal samples were collected and analysed.
Reference invertebrate species used in the phylogenetic analysis of cloned invertebrate sequences. Accession numbers beginning ‘HE’ were new sequences for reedbed insects that we uploaded, the remainder were downloaded from GenBank.
1 Identified as Cleigastra apicalis based on comparison with sequences in the BOLD (100% identity).
2 Identified as Holocentropus picicornis based on comparison with sequences in the BOLD (100% identity).
DNA was extracted from reference invertebrate material and faecal samples using DNeasy Tissue Kits and QIAamp DNA Stool Mini Kit (Qiagen, Manchester, UK), respectively, following the manufacturer's instructions. DNA was extracted from feather samples using InstaGene Matrix (BioRad, Hemel Hempstead, UK). The base of the feather was cleaned by removing the barbs towards the end of the feather and approximately 5 mm was then cut off from the calamus (Horvath et al., Reference Horvath, Martinez-Cruz, Negro, Kalmar and Godoy2005). 50 μl of distilled water was then added to each sample and centrifuged at 13,000 rpm for 3 min. 20 μl of InstaGene Matrix was added, the samples were incubated at 50°C for 30 min and then at 100°C for 8 min as described by Medeiros et al. (Reference Medeiros, King, Symondson, Cadiou, Zonfrillo, Bolton, Morton, Howell, Clinton, Felgueiras and Thomas2012).
PCR amplification, cloning and sequencing of mitochondrial DNA
Universal primers LCO1490 and HCO2198 (Folmer et al., Reference Folmer, Black, Hoeh, Lutz and Vrijenhoek1994) were used to amplify a 710 base pair (bp) region of the mitochondrial cytochrome oxidase subunit I (COI) gene from all reference invertebrate samples. All warbler faecal DNA samples were initially screened with the universal primers BirdF1 and BirdR2 (Hebert et al., Reference Hebert, Stoeckle, Zemlak and Francis2004) that amplify a 751 bp fragment of the avian COI gene. This was to check for successful extraction of amplifiable DNA. The general invertebrate primers LCO1490 and HCO1777 (Brown et al., Reference Brown, Jarman and Symondson2012) were used to amplify a 282 bp region of the COI gene from invertebrate remains in the warbler faecal DNA. Primers that amplify short fragments such as this are needed to ensure PCR success from semi-digested samples (Symondson, Reference Symondson2002; King et al., Reference King, Read, Traugott and Symondson2008). PCR reactions were carried out in a total volume of 25 μl consisting of 1 × PCR buffer (Invitrogen, Paisley, UK), 4 mM MgCl2 (Invitrogen), 0.05 mM dNTP mix (Invitrogen), 2.5 μg BSA (bovine serum albumin, New England Biolabs, Hitchin, UK), 0.1 μM of each forward and reverse primer, 0.625 U Taq polymerase (Invitrogen) and 2.5 μl DNA. After an initial denaturing step at 94°C for 2 min 30 s, amplification proceeded for 35 cycles at 94°C for 30 s, 47°C (LCO1490–HCO2198), 55°C (BirdF1–BirdR2) or 44°C (LCO1490–HCO1777) for 30 s, 72°C for 45 s and a final extension at 72°C for 10 min.
PCR products amplified from warbler faeces using the LCO1490 and HCO1777 primers were purified using QIAquick PCR Purification Kits (Qiagen) and cloned using the TOPO TA Cloning® kit (Invitrogen) following the manufacturer's instructions. Cloned inserts were amplified using M13 primers. PCR reactions were carried out in total volumes of 15 μl consisting of 1 × PCR buffer (Invitrogen), 1.5 mM MgCl2 (Invitrogen), 0.015 mM dNTP mix (Invitrogen), 1.5 μg BSA, 0.1 μM each forward and reverse primer (10 μM), 0.625 U Taq DNA polymerase (Invitrogen) and 1 μl of overnight-grown colony suspension. After an initial denaturing step at 94°C for 2 min 30 s, amplification proceeded for 35 cycles at 94°C for 30 s, 55°C for 30 s, 72°C for 45 s and a final extension at 72°C for 10 min.
Full length reference invertebrate and cloned faecal DNA COI PCR products (10 μl) were purified using 0.25 U each of Exonuclease I (New England Biolabs) and Shrimp Alkaline Phosphotase (USB) incubated at 37°C for 45 min and 80°C for 15 min. Sequencing was carried out on a 3130 × l Genetic Analyzer (Applied Biosystems, Warrington, UK) using Big Dye (v1.1) sequencing chemistry. PCR products from reference invertebrate samples were sequenced in both forward and reverse orientations using the original PCR primers. Cloned products were sequenced directly using the M13R primer. GenBank accession numbers are given in tables 2 and 3.
Occurrence of 45 distinct cloned COI sequences in each of 11 individual reedbed warblers, of 3 species (Cetti's, Reed and Sedge Warblers).
1 An additional six sequences were derived from bird mtDNA (98.68% similarity to C. cetti sequences in BOLD).
Data analysis
Sequences were aligned using SEQUENCHER version 3.1.2 (Gene Codes, Brighton, UK). Phylogenetic relationships between cloned and reference sequences were determined by neighbour-joining (NJ) analysis using pairwise Kimura-2-parameter distances (K2P, Kimura, Reference Kimura1980) between sequences in MEGA v4 (Tamura et al., Reference Tamura, Dudle, Nei and Kumar2007). The Barcode of Life Data Systems Identification Engine (BOLD-IDS, http://www.boldsystems.org/) was used to provide a taxonomic assignment for the sequences cloned from warbler faeces. Sequences were initially queried against the Species Level Barcode Records database (barcodes with a species-level identification). If a match was not found, then the sequence was queried against the All Barcode Records database. This search returns a list of the nearest matches to the query sequence and includes barcodes that do not have species-level identification. A sequence having similarity of >98% to either reference invertebrates or BOLD-IDS sequences was considered as being derived from that taxon (Clare et al., Reference Clare, Barber, Sweeney, Hebert and Fenton2011).
Results
Initial testing of the LCO1490–HCO1777 primer pair against reference invertebrate DNA and DNA from all three warbler species confirmed that the primers were specific to invertebrates, with no visible amplification (when screened on 0.8% agarose gels) seen with bird DNA.
A total of 152 clones were sequenced from the 11 warbler faecal samples (range 10–16 clones per faecal sample, table 3). Of these, only six clones from a single Cetti's Warbler were identified as being bird mtDNA (98.68% identity to C. cetti sequences in BOLD-IDS). The remaining 146 clones revealed a total of 45 distinct mtDNA sequences. The number of distinct sequences cloned per individual bird ranged from 2 to 10 (table 3). Only three sequences were identified in multiple warbler species. Sequence RBW 03 (Braconidae, Hymenoptera) was found in both Cetti's and Sedge Warblers, RBW 13 (Strictochironomus niger, Chironomidae) in Reed and Sedge Warblers and RBW 14 (94.1% similarity to Culex nigripalpus) in Cetti's and Reed Warblers (table 3, Online Appendix 1).
The NJ analysis based on a 232 bp COI fragment for cloned and reference DNA sequences revealed five distinct groups of sequences: general Diptera (predominantly cyclorrhaphid flies), chironomid midges, Araneae, Lepidoptera and a mixed group containing Trichoptera, Hymenoptera and Hemiptera sequences (fig. 1). There were marked differences between the three bird species in terms of the number of sequences from each of these prey groups that were detected in their faeces samples. Cetti's Warbler prey sequences were derived only from the Diptera and mixed groups, while Reed and Sedge Warblers consumed a more varied diet with prey from four and five of the main groupings, respectively. For Sedge Warblers, 67.8% of sequences were from chironomid midge species, whereas for Reed Warblers (27.7%) and Cetti's Warblers (0%) chironomids appear to be a much less frequent component of the overall diet (fig. 1).
NJ tree for the mitochondrial COI sequence data for 45 distinct sequences cloned from reedbed warbler faeces and 30 reference invertebrate species. Reference species abbreviations are as given in table 2. Star, circle and square symbols represent a sequence cloned from Cetti's, Reed and Sedge Warbler faeces, respectively (table 3). Panels show the occurrence of each group of sequences cloned from each warbler species as a percentage of the total sequences for that species (CW, Cetti's Warbler; RW, Reed Warbler; SW, Sedge Warbler).
A total of 39 distinct sequences could be identified to either species or genus level (table 4). Thirty-five sequences could be identified to species level, while four were identified to genus only. Twenty-two sequences were identified as Diptera species, with the majority of these being derived from chironomid midge species. The remaining sequences were identified as being derived from Hymenoptera, Trichoptera, Lepidoptera, Araneae and Collembola species (table 4). For some sequences, e.g., RBW 01, RBW 03 and RBW 20, BOLD-IDS returned multiple possible species-level identifications based on the 98% sequence similarity level. Some cloned sequences could not be identified to species or genus level, but could be tentatively assigned to taxonomic grouping based on the NJ analysis. For instance, sequences RBW 25 and 44 were found in the chironomid midge group (fig. 1) and had 97.0 and 97.6% identity to Chironomidae specimens in BOLD-IDS (Online Appendix 1). Online Appendix 1 provides a complete summary of the BOLD-IDS results for all cloned sequences.
Sequences cloned from Cetti's, Reed and Sedge Warbler faeces identified to genus and species level by comparison (greater than 98% similarity) to either reference sequences (table 2) or sequences in the BOLD. Sequences underlined indicate pests or potentially useful generalist predators.
1 This species has not been recorded from the British Isles.
2 British status unclear (Hopkin, Reference Hopkin2007).
Analysis of the species detected in the faecal samples revealed potential pest species. These included the noctuid moth Lacanobia oleracea (the Tomato Moth) (table 4, RBW 32, eaten by Cetti's Warbler); the mirid bug Phytocoris ulmi (RBW 40, eaten by Sedge Warbler); and tenthredinid sawflies (RBW 41, eaten by Sedge Warbler). Also present were potentially valuable natural enemies, that may help control pests in agriculture and horticulture, including the syrphids Melanostoma scalare/mellinum (RBW 01, eaten by Cetti's warbler); and braconid wasps, Aleiodes sp. (RBW 03, eaten by Sedge Warbler).
Discussion
Our study showed clearly that barcode databases (plus supplementary sequencing of DNA from some of the prey species in the target habitat) can now provide genus- and in most cases species-level identification of invertebrates in the diets of songbirds, even in less-studied habitats such as reedbeds. Overall, 39 out of 45 distinct species were identified to species or genus. Such a high success rate does not apply in many other parts of the world where biota are far less well barcoded, particularly in the tropics. For example Brown et al. (Reference Brown, Burger, Cole, Vencatasamy, Clare, Montazam and Symondson2014), in their study of invertebrates eaten by the Asian Musk Shrew (Suncus murinus) and the Telfair's Skink (Leiolopisma telfairii) on a Mauritian island were not able to identify any of the 119 molecular operational taxonomic units detected using NGS of faecal DNA. However, identifying the insect diets of, for example, bats in well-barcoded regions such as Europe and North America, using molecular diagnostics, is now far more successful (e.g., Clare et al., Reference Clare, Symondson and Fenton2014; Krüger et al., Reference Krüger, Clare, Greif, Siemers, Symondson and Sommer2014), with a high proportion of sequences matched to species.
There were few pests of agriculture or horticulture amongst the reedbed invertebrates eaten by the warblers, which should be welcome news to growers adjoining such habitats. This assumes that the warblers were mainly hunting within the reedbeds, where they were normally found, but we cannot exclude the possibility that they sometimes strayed onto farmland. The Tomato Moth, L. oleracea, is a highly polyphagous species that attacks Tomatoes, Solanum lycopersicum, both outdoors and in glasshouses (e.g., Jarrett & Burges, Reference Jarrett and Burges2011). It is probable that this moth is exploiting alternative host plants within the reedbeds but could well move onto tomato crops near by, or onto ornamentals. The mirid bug P. ulmi is common in the UK and feeds on the fruit and flowers of many plants as well as small insects. It has been found on commercial fruit and berry crops in Lithuania (Rakauskas, Reference Rakauskas1984) but does not appear to be a major pest in the UK. Tenthredinid sawflies, most likely Nematus viridissimus, attack trees and other plants (Nyman et al., Reference Nyman, Vikberg, Smith and Boevé2010) and may be pests of ornamentals. By far the most speciose family detected was the Chironomidae, with 20 species identified (around half of the taxonomic units identified to species or genus using sequencing). These non-biting flies play a major role in the reedbed ecosystem, sustaining not just birds but also bats (Krüger et al., Reference Krüger, Clare, Greif, Siemers, Symondson and Sommer2014) and many invertebrate predators (e.g., Gratton et al., Reference Gratton, Donaldson and vander Zanden2008). However, their large numbers can sometimes create a nuisance, especially in urban areas, including at one of our study sites at Cardiff Bay where efforts are being made to control them (Vaughan et al., Reference Vaughan, Newberry, Hall, Liggett and Ormerod2008). In affected regions chironomids are implicated in the spread of cholera (e.g., Halpern et al., Reference Halpern, Raats, Lavion and Mittler2006).
Potentially valuable natural enemies included aphidophagous syrphids, M. scalare/mellinum, known to attack aphid colonies within crops (Rotheray, Reference Rotheray1983; Dziock, Reference Dziock2005). Although recorded attacking the bean aphid, Aphis fabae, these syrphids, like many others, will attack the larvae of a range of species of aphid found in colonies, helping potentially to reduce aphid numbers in several crops. There were also braconid wasps, Aleiodes sp., which generally attack moth larvae, although whether the species found attack pest species is unknown. Generalist spiders were found, including Linyphiidae which are major predators of aphids and other pests in crops (e.g., Harwood et al., Reference Harwood, Sunderland and Symondson2004). However, this is probably misleading as the species identified, Gnathonarium dentatum, is primarily associated with aquatic environments and waste land, clearly showing the value of species-level identification.
Extraction, amplification and identification of prey DNA from predator faeces depends on the survival of amplifiable DNA through the predator's digestive tract. There is increasing evidence that large DNA fragments are digested relatively quickly within predator guts and so even with advanced PCR methodologies, small DNA fragments (<300 bp) are usually targeted to ensure successful amplification of prey DNA (e.g., Greenstone et al., Reference Greenstone, Payton, Weber and Simmons2014). In the present study, a fragment of 282 bp of invertebrate DNA was successfully amplified from all 11 warbler faecal samples tested. The LCO1490–HCO1777 primer pair used failed initially to amplify bird DNA on a gel, although bird DNA was found in just 6 out of 152 clones. In other respects the primers were specific to invertebrates across multiple orders. The primers provided a good compromise between shortness (to survive digestion) and a length sufficient to provide a reasonable level of taxonomic resolution. The target fragment was within the standard COI barcoding region, maximizing returns from database searches. Although we recommend this primer combination for further work, using either cloning and sequencing (as here) or NGS (Pompanon et al., Reference Pompanon, Deagle, Symondson, Brown, Jarman and Taberlet2012), all primers should be tested against the predator at an early stage. We know, for example, that this primer pair co-amplifies a legless lizard, the Slow Worm Anguis fragilis (Brown et al., Reference Brown, Jarman and Symondson2012), and a seabird, the European Storm Petrel, Hydrobates pelagicus (unpublished data).
Although sample sizes were low, we were able to get useful information on the diets of the three warbler species and their ecological roles in and around the reedbed ecosystem. Overall the Cetti's Warblers appeared to have a restricted diet compared with the more polyphagous Reed Warbler, while the Sedge Warbler diet was dominated by chironomids (67.8% of cloned sequences). Choosing all three of the warbler species to sample the habitat for invertebrates therefore probably considerably increased the breadth of invertebrate species that we were able to detect. These warblers appeared to have been foraging not only on prey originating from the reeds themselves, but also from adjoining or interspersed areas of scrub and alder, Alnus, and willow, Salix, carr. For example, one of the Cetti's Warblers was shown to have consumed the hoverfly M. scalare, a species often found in woodland and scrub (Stubbs & Falk, Reference Stubbs and Falk2002). A sequence from the sawfly genus Nematus was cloned from Sedge Warbler 7, trapped at Cardiff Bay Wetlands. The typical host plants for Nematus sawflies are trees from the Alnus, Betula and Salix genera (Nyman et al., Reference Nyman, Vikberg, Smith and Boevé2010). Thus, dietary analyses such as this can reveal where birds are most likely to have been foraging as well as what they have eaten.
Data obtained from this small number of warbler samples demonstrated that the primers and protocols work well. Once the molecular taxonomy of the main prey consumed has been described, the PCR-based approach can be developed in several ways to address different ecological questions. For example, species-specific primers can be designed, targeting the most frequent invertebrate species found within the three warbler species (i.e., prey species making a significant contribution to their energetics). Large numbers of individual faecal samples can then be rapidly screened for these target taxa, to compare diet between warbler species, age classes, seasons and years. In an applied context they could be screened for particular pest or natural enemy species to obtain more quantitative information (i.e., proportion of birds testing positive). It is now also possible to screen for many targets simultaneously using multiplex PCRs with fluorescent labels, as demonstrated in our previous analyses of the gut contents of generalist invertebrate predators (Harper et al., Reference Harper, King, Dodd, Harwood, Glen, Bruford and Symondson2005; Bell et al., Reference Bell, King, Bohan and Symondson2010; King et al., Reference King, Vaughan, Bell, Bohan and Symondson2010a , Reference King, Moreno-Ripoll, Agustí, Shayler, Bell, Bohan and Symondson b ). Prey choice can also be analysed by comparing the relative abundance of different prey in the field with the relative frequencies with which those prey are detected (i.e., predator gut or faecal samples test positive) (Agustí et al., Reference Agustí, Shayler, Harwood, Vaughan, Sunderland and Symondson2003; King et al., Reference King, Vaughan, Bell, Bohan and Symondson2010a ; Davey et al., Reference Davey, Vaughan, King, Bell, Bohan, Bruford, Holland and Symondson2013).
Although the molecular approach has great power to identify prey DNA from predator faeces, it also has limitations. For example, many of the insects identified in warbler diet in the present study could have been consumed as eggs, larvae or adult prey, or any combination of these stages. It is also not usually possible to distinguish primary predation from secondary predation (i.e., where the focal predator has consumed a second predator or parasite, that itself has consumed a third taxon of prey) using molecular-based methods (Sheppard et al., Reference Sheppard, Bell, Sunderland, Fenlon, Skirvin and Symondson2005; King et al., Reference King, Read, Traugott and Symondson2008). Similarly, parasites and parasitoids may inadvertently be eaten with their hosts. In our study, the finding of sequences from parasitic braconid hymenopteran species (Aleiodes spp., RBW03, table 3) in warbler faeces may indicate such a scenario. This species was cloned from both Cetti's and Sedge Warbler faeces and is a parasitoid of noctuid moths. A sequence from the noctuid L. oleracea was cloned from Sedge Warbler 2, a bird from which Aleiodes sequences were also obtained (table 3). Alternatively, the birds could be feeding upon adult Aleoides. A second sequence, RBW31, showed 99.57% similarity to the ichneumonid wasp Acrodactyla sp., a genus whose larvae are ectoparasites of Tetragnatha spiders and demonstrates the possibility of ‘unintentional’ predation of the ichneumonid by a bird foraging on spiders. Traugott et al. (Reference Traugott, Bell, Raso, Sint and Symondson2012) were able to conclude, in a molecular analysis of an invertebrate food web, that carabid beetles were mainly consuming parasitoids within their aphid hosts because DNA of both parasitoids and hosts appeared simultaneously in their guts. However, in the same study, spiders were eating both parasitized aphids plus adult parasitoids, which they caught directly in their webs, because parasitoid DNA was frequently found in their guts in the absence of host DNA.
Despite success at identifying many taxa, species-level identification of prey DNA was not always possible. For nine of the prey DNA sequences identified to species level in the present study, BOLD-IDS returned multiple likely species-level identifications (table 4). DNA barcoding has routinely found that while the majority of species can be reliably separated on the basis of their COI sequences, a small proportion cannot. For instance, Park et al. (Reference Park, Foottit, Maw and Hebert2011) found 15 cases of hemipteran species having identical or very similar barcodes while Kerr et al. (Reference Kerr, Lijtmaer, Barreira, Hebert and Tubaro2009) found that six species of Sporophila shared DNA barcodes. The sharing of barcodes between species tends to involve species that are morphologically very similar or where hybridization and mitochondrial introgression is suspected. Additionally, the failure of the short COI region utilized here to discriminate between some prey species could be due to this region lacking the discriminating base positions that would separate these particular taxa. Roe & Sperling (Reference Roe and Sperling2007) found that the location of regions of maximum divergence between species within the COI gene was highly variable among taxa and that maximizing sequence length would increase the probability of sampling regions with the highest phylogenetic resolution. Clearly correct identification, to species, is essential when the aim is to obtain warning of the presence of source areas for recognized pest species.
Another potential drawback of the molecular approach to analysing insectivorous bird diets is whether the proportions of prey determined from cloned sequence number are representative of the proportions of actual prey individuals or biomass consumed. This is not so important where the primary aim is to gather presence/absence data but can be a problem when constructing quantitative food webs. Deagle et al. (Reference Deagle, Chiaradia, McInnes and Jarman2010) investigated this in a molecular-based assessment of the diet of captive Little Penguins, Eudyptula minor, where the proportion of one prey species represented in faecal DNA samples was significantly lower than expected from mass of prey consumed. There are several possible explanations for this. There may be differences in the digestibility of different prey items or variation in the efficiency with which the general primers amplify across prey species. Copy number of mtDNA in the cells of different species can also vary widely. Furthermore, in the case of the prey species found to have been eaten by the three warbler species, there are marked differences in prey size that could potentially bias estimates of the number of individual prey consumed, based on faecal DNA proportions. These factors are unlikely to be a problem where prey species are closely related and similar morphologically, but the greater the taxonomic distance the less valid quantitative comparisons become. Where dietary components are highly diverse, as here, this approach to determining the diet of generalist predators may only provide semi-quantitative estimates of prey ratios in the diet.
Despite these caveats regarding the interpretation of prey DNA sequences and sequence-abundance results, differences in prey DNA ratios in samples obtained from directly comparable groups of predators (e.g., species, sexes or age-classes of warblers, or temporal changes within individuals) can be considered to be an index of real differences in diet composition. Molecular methods provide a highly effective, specific and objective method for the identification of dietary components, which greatly increases the scope for accurate identification of even very small and soft-bodied insects. The present study confirms the practicality of applying the molecular approach to identifying the insect and spider prey of songbirds, and paves the way for future studies to apply this powerful and relatively non-invasive new technique to a range of applied ecological and conservation contexts. This includes the use of birds as powerful tools that allow us to search in any ecosystem for invertebrates of applied importance, including both natural enemies and crop pests. It is an approach that could equally well be used in a conservation context, to search for the presence of rare species, or to monitor invasive invertebrate species.
We were able, for this project, to take advantage of on-going mist netting operations in the reedbeds and obtained faecal samples from birds that would have been caught anyway for ringing (banding). Mist netting is widely used by conservation bodies and others in the UK, to gain information on changes in bird populations and migration patters. It is usually harmless to the birds and helps provide the data needed to improve their conservation. The cost of the consumables for the molecular work came to approximately £80 per bird (at 16 clones per bird). The alternative would have been to use a combination of Malaise traps, sticky traps and vacuum sampling in all the sub-habitats across the the reedbeds. This would have been a major undertaking, not costly in consumables (if you have the equipment) but certainly in man-hours. Even if collected by entomologists, many of the taxa, particularly difficult groups such as chironomids and parasitic hymenoptera, may need to be identified to species by the few experts that still exist, who usually charge for their services. Museums, for example, can charge up to £80 per species for invertebrate identification (Michael R. Wilson personel communication). Even then identity may have to be confirmed using PCR. Here we are proposing to let the birds do the sampling work and the DNA do the identification. There are pros and cons to both approaches but what we have done here is to demonstrate that an alternative strategy exists which we believe could be valuable.
Supplementary Material
The supplementary material for this article can be found at http://www.journals.cambridge.org/BER
Acknowledgements
The warbler faeces samples were collected by various licensed bird-ringers, primarily James Vafidis, Richard Facey, Ian Vaughan and Dave Beveridge. We are grateful to Gwent Wildlife Trust for permissions to ring birds at Magor Marsh, to Cardiff Harbour Authority for permissions at Cardiff Bay Wetlands, and to Vale of Glamorgan Council for permissions at Cosmeston Lakes Country Park. We thank John Deeming of the National Museum, Cardiff for identifying specimens of intact Diptera. We are grateful to Dr David Brown for kindly providing us with information, pre-publication, on the new general invertebrate primer he had designed (HCO1777). James Vafidis, Richard Facey and anonymous reviewers provided comments on earlier drafts of the manuscript.
