Supporting Information Wirta et al. 10.1073/pnas.1316990111 SI Text our success levels from field-caught samples may be considered 1. Performance and Validation of the Molecular Methods. Molecular high by comparison. analysis of parasitoid linkages. With PCR-primers designed to se- lectively amplify the DNA of the host order but not the parasitoid, 2. Incidence of Multiple Host and Parasitoid Sequences from the Same we successfully amplified and sequenced a host from 21.9% of 457 Individual. Incidence of multiple host sequences in the same parasitoid parasitoids caught as adults. However, the rate of sequence re- individual. Our MAPL-AP analyses revealed evidence of two different covery was significantly lower for species of braconids than host species in seven individuals of parasitoid wasps (all family ichneumonids and tachinids [Braconidae, n = 34 (2.9%); Ichneumonidae). In a single individual of Aoplus groenlandicus,both Ichneumonidae, n = 346 (24.0%); and Tachinidae, n = 77 Entephria polata and Sympistis nigrita were found; in Buathra labo- (20.8%); χ2 = 8.09, df = 2, P = 0.02). Interestingly, seven rator,bothGynaephora groenlandica and Euxoa adumbrata were ichneumonid specimens yielded sequences from two different found; in two Cryptus arcticus specimens, G. groenlandica was host species, as recovered from two separate body parts or found with Syngrapha parilis in one and S. nigrita in the other; from two different pieces of the abdomen (SI Text, section 2). For in Mesochorus n. sp., both G. groenlandica and S. nigrita were found; eight parasitoid species represented by only a few individuals andintwoPimpla sodalis specimens, G. groenlandica and S. parilis in our sample, no host could be identified through molecular were found in one and E. polata and S. nigrita were found in the analysis of parasitoid linkages–adult parasitoid (MAPL-AP) other. In each case, these different host sequences were recovered (Table S1). from two separate body parts or from two different pieces of the MAPL-host larvae. The selective amplification and sequencing of parasitoid abdomen analyzed as separate samples. For each of these parasitoid DNA from host tissue yielded a parasitoid sequence cases, one of the host sequences likely represents the host that was from 20.9% of 1,195 hosts examined. Of these host larvae, 6.1% fed upon by the parasitoid as a larva, whereas the other sequence were parasitized by an ichneumonid, 12.3% by a braconid, and was acquired from a host that was oviposited upon by the adult 2.4% by a tachinid. The level of parasitism detected varied parasitoids. In three of the seven cases, one host sequence was considerably among host species, ranging from 0% to 44% among unequivocally detected in a tissue sample consisting of the ovipos- species represented by more than one individual. In 2 of 12 host itor only, suggesting acquisition of DNA during oviposition or species included in this study no parasitoids were discovered, while probing a potential host. In the other four cases, oviposition likely due to small sample size (Table S1). emerges as a possible source of the secondary host sequence, al- Recovery of parasitoid sequences from visibly parasitized host larvae. though host feeding by the adult parasitoid cannot be ruled out. When using MAPL-host larvae (HL), the true level of parasit- Incidence of multiple parasitoid sequences in the same host individual. In ism in the sample is unknown (this contrasts with the situation for our MAPL-HL analyses, sequences of two separate parasitoid MAPL-AP, where each parasitoid must have fed on a host as taxa were discovered in 11 host individuals. In most cases (n = 7), a larva). To verify if our MAPL-HL protocol accurately detects the co-occurring taxa were primary parasitoids, whereas in 4 parasitoids when present, we analyzed a subset of host individuals cases, one of the taxa was a secondary parasitoid Mesochorus n. where the presence of a parasitoid larva was detected during sp. In all 11 cases, one of the parasitoids was a member of the sample preparation. Of 30 such samples, 80.0% yielded a para- family Braconidae whereas the other belonged to the Ichneu- sitoid sequence. This suggests that the true level of parasitism is monidae. The co-occurring parasitoids were Cotesia spp. and slightly higher than observed by MAPL-HL in the current study, Mesochorus n. sp. in Boloria chariclea, two cases of Microplitis a discrepancy potentially related to the specificity of the primers lugubris and Mesochorus n. sp. in B. chariclea, Cotesia spp. and used (SI Text, section 6). Nonetheless, the parasitism rate esti- Hyposoter frigidus in Polia richardsoni, M. lugubris and H. frigidus mated by MAPL-HL (20.9%) was still higher than that estimated in another Polia richardsoni, M. lugubris and Mesochorus n. sp. in by rearing (16%; ref. 1). S. nigrita, two cases of M. lugubris and Campoletis horstmanni Recovery of host sequences from wild-caught versus reared parasitoid in S. nigrita, M. lugubris and H. frigidus in another S. nigrita, individuals. To verify that MAPL-AP correctly identifies the host M. lugubris and C. horstmanni in S. parilis,andM. lugubris and use of adult parasitoids, we compared a set of 422 wild-caught H. frigidus in S. parilis (the first listed parasitoid is the braconid individuals to a set of 35 parasitoids reared from known hosts. and the second is the ichneumonid). Within these sets, a host sequence was successfully recovered from The incidence of multiple parasitoids in the same hosts (mul- 22.5% of the wild-caught individuals, compared with 17% of the tiparasitism) might have been underestimated by the methods that reared individuals. Hence, the recovery rate was similar among we used. As we used Sanger sequencing (3), and sequenced only materials (Fisher’s exact test, P = 0.50). one PCR product, we might have failed to detect co-occurrence Overall, our success in detecting the host sequence from adult of two confamilial parasitoids in the same host individual. In parasitoids was comparable to that of Rougerie et al. (2), who future studies, additional sequencing of PCR products by next applied a method similar to MAPL-AP to reared parasitoid generation techniques may improve quantification of cases of specimens. Targeting adults of 3 parasitoid species in 2 families, multiparasitism. Rougerie et al. (2) detected a host in 24% of individuals, whereas our detection rate of 20.9% applies to a wider range of 3. Estimations of Link Richness and Sample Coverage. Methods. No parasitoid species (20 species from 2 wasp families and 1 fly finite sample will reveal all species or all interactions in a food web. family). Ultimately, the success rate in discovering remains of To examine how well our samples and methods describe the set of host DNA will be constrained by the limited amounts of DNA trophic links in the Lepidoptera–parasitoid community at Zack- remaining in the gut of the parasitoid after metamorphosis, as well enberg, we used methods originally developed for the estimation as by the degradation of this DNA during both larval development of species richness to estimate the total number of different tro- and adult life span before collection (discussed by ref. 2). As phic links in our target community. For this, we used two non- Rougerie et al. (2) sampled their parasitoids directly after emer- parametric estimators: Chao1 (4, 5) and the abundance-based gence (thus minimizing the degradation of DNA after emergence), coverage estimator (ACE) (6, 7). Chao1 provides an estimate of Wirta et al. www.pnas.org/cgi/content/short/1316990111 1of6 yet-undetected links based on the frequency of singletons provide a molecular toolset of prime utility in exploring the (trophic links revealed only once) and doubletons (trophic links structure of host–parasitoid food webs. revealed only twice) in the sample. ACE employs information on link types encountered at slightly higher frequencies. 4. Methodological Provisos. Using host larvae as the only in- Sample size variation not only influences the detection of in- formation source to describe trophic interactions may lead to dividual links by a particular method, but it also affects the links involving idiobiont parasitoids going undetected (see Results number of links jointly detected by multiple methods. Consider and Fig. 1). On the other hand, the use of adult parasitoids as the the case where a set of interactions is sampled by two methods, sole source of information might cause similar problems. As each with a small sample size. Under this scenario, one method DNA degrades in the gut of the parasitoid, individuals caught long will reveal a subset of all interactions, whereas the other method after their emergence may no longer carry host DNA of sufficient will detect another subset, with limited overlap between methods. quality to allow their identification. Finally, a sample of adult par- Although our sample sizes were reasonably large, they varied asitoids alone would not reveal parasitism levels suffered by the among methods (n = 457 for MAPL-AP, n = 1195 for MAPL- host, which is indeed an important dimension of interaction struc- HL, and n = 1420 for rearing). Naturally, more species will be ture. Thus, an adequate description of the interaction web can only detected by any methodological approach if sample size is in- be built on a combination of multiple sources of information. creased. To examine the complementarity of different methods Interpretation of trophic associations. Although the detection of para- while adjusting for differences in sample size, we applied the sitoid DNA in a living host larva (MAPL-HL) offers proof of an Chao1 estimator modified to derive the true number of species interspecific interaction, it is not sufficient proof of a feeding as- sociation that will lead to host death.