Introduction

The introduction of non-native species to new ecological communities is creating novel and altered predator–prey and parasite–host interactions (Faillace et al. 2017; Garnas et al. 2016; Hobbs et al. 2009; Pearson et al. 2018; Shea and Chesson 2002; Strauss et al. 2012). The species richness of a community can predict the chance that an invasive species will successfully establish. This hypothesis, known as the biotic resistance hypothesis, holds that communities may resist invasions through a combination of factors including predation, competition, parasitism, and disease (Elton 1958; Jeschke et al. 2012; Levine et al. 2004; Maron and Vilà 2001; Sakai et al. 2001; Shea and Chesson 2002). Natural enemies of native or resident species that cross over to use non-native species can influence invasions in complex ways (Dearborn et al. 2016; Faillace et al. 2017; Grabenweger et al. 2010; Strauss et al. 2012) and have the potential to slow down invasions and aid in biological control (Dearborn et al. 2016; Kenis et al. 2008; Maron et al. 2001; Vindstad et al. 2013). These interactions are particularly strong for non-native species with sympatric native congeners and confamilials (Dearborn et al. 2016; Grabenweger et al. 2010; Strauss et al. 2012; Vindstad et al. 2013). Further, biotic resistance may especially affect invasive insects because they are typically r-selected (Sakai et al. 2001) and their population dynamics are closely related to natural enemies, predominantly parasitoids (Hassell 2000; Myers 2018; Waage and Greathead 1985). While it is clear that naturally-occurring biological controls play an important role in pest suppression (Heimpel and Mills 2017), research on the role of parasitism in biological invasions is lagging behind research on biological invasions in general (Poulin 2017) and particularly so for studies of invasive herbivores and their natural enemies (Bürgi et al. 2015).

The European winter moth, Operophtera brumata L. (Lepidoptera: Geometridae), is a well-known defoliator of forest, shade, and fruit trees with multiple invasive populations established in North America (Embree 1965; Myers 2018; Myers and Cory 2013; Roland and Embree 1995), including a recent (circa late 1990s) introduction in the northeastern United States (Elkinton et al. 2010, 2015). Following successful biological control of winter moth in Nova Scotia and British Columbia using the tachinid fly Cyzenis albicans (Fallén) (Diptera: Tachinidae) (Embree 1966; Murdoch et al. 1985; Roland and Embree 1995), similar efforts have been initiated in this most recent invasion (Elkinton et al. 2015). While importation biological control of winter moth in the northeastern United States has shown promising results (Elkinton et al. 2015, 2018), as previously found in Canada (Roland 1988, 1990), the overall success will likely depend on additional mortality from native natural enemies. Parasitoid recruitment from related native species can have significant effects on invasive populations in other insect study systems (e.g. Duan et al. 2013, 2014; Grabenweger et al. 2010; Matosevic and Melika 2013; Schonrogge et al. 1995; Zappala et al. 2012). This is especially true if the parasitoids limit the establishment or facilitate the eradication of a new exotic species (Tobin et al. 2011), if they respond to increases in the invasive population’s densities with density-dependent mortality (Holling 1973), or if the parasitoids parasitize at high rates at the invasion front such that they can slow or stop the spread of the invasive population (Lewis and Kareiva 1993).

Non-native species with native congeners in the introduced range may be less likely to establish an invasive population than species introduced to a range without a native congener; these invasive species face top-down pressure from the natural enemies of their congener (Callaway et al. 2013; Carrillo-Gavilan et al. 2012; Diez et al. 2008; Keane and Crawley 2002; Richardson and Pysek 2006). The native congener of winter moth, Bruce spanworm (Operophtera bruceata Hulst), is a potential source of native parasitoid recruitment to invasive populations of winter moth in North America. Bruce spanworm is present in all regions winter moth has invaded. In addition to having similar life-cycle dynamics, these two congeners use similar hosts, are present at similar times of the year, and can hybridize in the field (Gwiazdowski et al. 2013; Havill et al. 2017). Thus, it is likely that native natural enemies that parasitize Bruce spanworm could use winter moth as a host. Additionally, the life histories of winter moth and Bruce spanworm make their populations particularly vulnerable to pupal mortality by predation and parasitism from native natural enemies. Both winter moth and Bruce spanworm have a long pupation period (6–7 months during the summer, representing the vast majority of its life span) and pupates in the top layer of soil or leaf litter. Together this renders both species particularly vulnerable to pupal mortality by predation, parasitism, and disease.

Parasitoid wasps in the genus Pimpla Fabricius (Hymenoptera: Ichneumonidae) might be an important source of mortality for invasive populations of winter moth in North America. Species of Pimpla are found in all zoogeographic regions, including 27 species in the Nearctic Region (Yu et al. 2012), and typically parasitize Lepidoptera prepupae and pupae (Bennett 2008; Carlson 1979; Gauld 1991; Goulet and Huber 1993). Pimpla species are known to use geometrid pupae concealed in moss or soil (Fitton et al. 1988), suggesting that they may be important natural enemies of Bruce spanworm and winter moth, which fit both of these criteria. Pimpla species are known from the region of this study and have been associated with winter moth populations. Pimpla turionellae L., Pimpla contemplator Müller, and Pimpla flavicoxis Thompson have been recorded as attacking winter moth in its native range in Europe (Fitton et al. 1988; Silvestri 1941; Wylie 1960). P. flavicoxis is currently considered a valid species (Yu et al. 2012) but has also been treated as a junior synonym of Pimpla aquilonia Cresson (Carlson 1979). P. aquilonia is a broadly distributed Holarctic species (covering the region of this study) reported from multiple families of Lepidoptera, including geometrids, but not yet from winter moth (Yu et al. 2012). A small release of P. turionellae to Nova Scotia, Canada in 1955 was conducted to control winter moth; however, recovery collections revealed no evidence of establishment (Graham 1958; Humble 1985). Similarly, P. contemplator was released in Nova Scotia, Canada in 1964 to control winter moth but with no evidence of establishment (Carlson 1979). Pimpla hesperus Townes is native to North America and has been reported as a parasitoid of winter moth and Bruce spanworm in British Columbia, Canada (Humble 1985). Recently, an undetermined species of Pimpla in the United States in Maine was found to be more abundant at sites with high winter moth infestation than moderately infested sites (Morin 2015), and another, or potentially the same, undetermined species of Pimpla has been reported in the northeastern United States as a hyperparasitoid of C. albicans, an introduced biological control agent of winter moth (Broadley et al. 2018). These latter two studies revealed an association of Pimpla species with the invasive population of winter moth in the northeastern U.S., but neither study directly assessed Pimpla wasps as primary parasitoids of winter moth. The assemblage and origin of Pimpla species associated with invasive winter moth populations in the northeastern U.S. is unknown, as is their prevalence, role in causing winter moth mortality, and potential to regulate winter moth densities.

In this study we aimed to (1) quantify parasitism by Pimpla wasps on winter moth pupae and C. albicans puparia across a spatial and temporal gradient, (2) test for a density dependent effect of pupal parasitism, and (3) identify the Pimpla species using morphological and molecular characteristics. We discuss our results in relation to their implications for understanding the role and origin of this parasitoid in the control of an introduced lepidopteran pest. These findings demonstrate the importance of evaluating native parasitoids in the establishment, spread, and population dynamics of other invasive species.

Methods

Pupal deployment

To acquire winter moth for deployment as sentinel pupae, we collected winter moth larvae from long-term study sites across eastern Massachusetts each spring from 2014 to 2017 (Elkinton et al. 2015). Larvae were reared in batches of 500 or fewer individuals in ventilated 20 L (5 gallon) buckets with the foliage from the tree species on which they were found. Mortality from viruses, other diseases, and larval parasitism in these collections was minimal (Broadley et al. 2017; Donahue et al. 2018). When the larvae showed signs of pupating (thickening body shape), sifted peat moss was added to the bottom of the buckets for pupation. All winter moth pupae were non-destructively checked under a dissecting microscope (Wild Heerbrugg M5 stereo) for parasitism by C. albicans or other larval parasitoids. The unparasitized winter moth pupae and pupae parasitized by C. albicans were stored at 12 °C in a growth chamber (Percival Scientific) until deployment.

To study pupal parasitism by native parasitoids, winter moth pupae were deployed at sites across eastern Massachusetts, Rhode Island, and Connecticut from 2014 to 2017 (Table 1). The study sites were chosen to coincide with winter moth long-term study sites and to reflect a range of winter moth and C. albicans establishment histories (Elkinton et al. 2014, 2015). The study sites were all in mixed hardwood forests dominated by red oak (Quercus rubra). The pupae were deployed in three to five consecutive rounds from mid-June until the end of October with five deployments (one every 3 weeks) in 2014 and three deployments (one every 6 weeks) in 2015–2017. Each deployment consisted of 100 winter moth pupae attached to small burlap squares with beeswax; these were buried 2 cm below the soil surface haphazardly under the drip line of a red oak tree as had been done for previous studies (Broadley et al. 2018; Whited 2007). The placement depth was chosen to mimic natural pupa depths (East 1974; Embree 1965; Holliday 1977). All pupae in each deployment were the same age; all had pupated at the end of May from larvae collected mid-May. In total 12,420 winter moth pupae were deployed.

Table 1 Percent parasitism by Pimpla for each study site and each year of study. Parasitism values are cumulative pupal-stage parasitism rates. The sample size is included in parentheses
Full size table

To evaluate the effect of native parasitoids on C. albicans puparia (winter moth pupae that were parasitized by C. albicans), in 2014–2016 C. albicans puparia were also deployed at a subset of sites (Table 1). C. albicans eggs are laid on the surface of defoliated leaves, and winter moth can become parasitized by C. albicans if they inadvertently consume a C. albicans egg while feeding. The C. albicans hatches when inside its host, and when the host winter moth pupates in the soil, the fly larva develops and forms a puparium inside the host pupa. Only one puparium can form inside each host (Wylie 1960). The fly puparium is inside the winter moth cocoon and is readily visible by mid- or late June. For this study, a total of 3400 C. albicans puparia were deployed.

Pupal dissections, incubation, and parasitoid collection

After each 3- or 6-week pupal deployment, we retrieved the sentinel pupae and characterized their fate as alive (intact) or dead (consumed by predators, parasitized, or diseased). Without destructive dissection, it is not possible to determine if the retrieved pupae may have a developing endoparasitoid. Thus, the intact pupae were stored in an incubator (Percival Scientific) until the following spring to allow any further parasitoid development. Pupae were stored at 12 °C until the beginning of December, at 9.5 °C until the end of December, and at 4 °C until late March. The pupae were kept in dark with no day/night cycle, and once a month they were sprayed with a sodium propionate solution (5 g sodium propionate/l of water) to prevent mold. Starting in late March, the temperatures were increased 4 °C every 3 days to 22 °C when the pupae were taken out of storage and kept at room temperature. The parasitoids were identified as Pimpla using keys in Townes et al. (1960), Townes (1969) and stored in 95% ethanol at − 20 °C for further molecular or morphological identification.

Monthly and annual parasitism rate estimates

To calculate monthly and annual mortality from parasitism, our estimate of parasitism included pupae that had developing wasps (larvae and pharate adults) and pupae with wasp emergence holes (as shown in Fig. S1). The proportion of pupae parasitized by Pimpla wasps for each deployment was calculated by dividing the total number of parasitized pupae by the total number of intact pupae retrieved from the field (i.e., number of pupae retrieved excluding the number of pupae that were lost due to predation). This method of calculating percent parasitism incorporates the fact that parasitism rates can be obscured by predation rates because predation typically occurs on the pupae whether or not they were parasitized. This method aims to estimate the true underlying mortality rate of each source in the system (Buonaccorsi and Elkinton 1990; Elkinton et al. 1996; Royama 1981; Van Driesche 1983). To test for any relationship between the rates of Pimpla wasp parasitism and predation, we regressed the proportion parasitized by the proportion lost to predation and no trend was detected (Fig. S2); this further suggests that predators do not discriminate between parasitized and unparasitized pupae.

To calculate the standardized mortality from parasitism for each deployment over 31 days (S31), we used the following equation:

$${\text{S}}_{31} = [({\text{S}}_{\text{p}} )^{{1/{\text{n}}}} ]^{31}$$

where Sp is the pupal survivorship from parasitism as a proportion and n is the true number of days the pupae were deployed (which ranged from 19 to 45 days with a mean of 31 days). To calculate the cumulative pupal stage parasitism (Pc), we used the following pair of equations:

$${\text{S}}_{\text{c}} = \, \left( {{\text{S}}_{{{\text{p}}1}} \times {\text{ S}}_{{{\text{p}}2}} \times {\text{ S}}_{{{\text{p}}3}} \times {\text{ S}}_{{{\text{p}}4}} \times {\text{ S}}_{{{\text{p}}5}} } \right)$$

and

$${\text{P}}_{\text{c}} = 1 \, {-}{\text{ S}}_{\text{c}}$$

where Sc is the cumulative pupal stage survivorship from parasitism and Sp1 to Sp5 are the proportion survival from parasitism for successive deployments for a particular site and year. No value was included for deployments four and five when only three deployments were used to span the pupal life stage.

Host selection, development, and sex ratio

In 2016, we deployed an additional set of 2000 winter moth pupae to study parasitism depth and host searching behavior of Pimpla wasps. The pupae were spread one layer thick across the base of wire mesh cages (mesh size: 6.4 mm sides with a 17 mm mesh lid) to keep predators out. Two cages each were deployed in two of the study sites, Wellesley and Hanson, Massachusetts (Table 1) for three consecutive rounds for 35 days each (deployment 1: 7 July–8 August; deployment 2: 8 August–9 September; deployment 3: 9 September–11 October). Half of the pupae were deployed with a thin layer of leaves over top, and the other half had 2 cm of soil then leaves over top. After deployment, the retrieved pupae were sifted from the soil and stored at room temperature with natural light cycling to allow wasps to complete their development and emerge. Every 3 days, the pupae were sprayed with the water-sodium propionate solution, and wasp emergence was recorded. We tested for an effect of soil and leaf coverage on the emergence counts using a generalized linear model (GLM) with a negative binomial fit.

To study wasp development time, sex ratio, and host choice, Pimpla wasps that emerged from field-deployed pupae in 2016 were reared in the laboratory in cages (BugDorm 4F4545 Insect Rearing Cage). Dead wasps were replaced, and newly emerged wasps were added, but we always kept 30 wasps per cage with an equal number of males and females. At night, the cages were stored in full dark at 12 °C; during the day, they were kept at 23 °C and exposed to ambient light. The wasps were sprayed with water twice a day and given honey water. To study development time, the wasps were given access to 100–200 unparasitized winter moth pupae, which were replaced every 5–6 days for a total of 12 rounds between 15 August and 17 October. Oviposition was monitored during the first hour of exposure to new pupae. When a female oviposited into a pupa, we moved the pupa singly to a tube (15 ml Falcon centrifuge tube with a ventilation hole). The exposed pupae were stored at room temperature in ambient light and monitored for subsequent wasp emergence. Of these 120 pupae, 42 wasps emerged. Wasp development time was calculated. The sex of wasps that emerged from the field-deployed pupae and lab colony was noted and sex-ratio was evaluated by exposure treatment (field or lab pupae), by date of exposure to parasitism, and by season of emergence (fall or spring) using binomial GLM. To assess host choice, a subset of 30 wasps was given both winter moth pupae and C. albicans puparia in a choice test. We ran three trials in September, each with an equal number of winter moth pupae and C. albicans puparia ranging from 100 to 200 pupae for each pupa type. We also compared the monthly parasitism rates from pupae and puparia deployed in the main study plots that received C. albicans puparia (Table 1, sites A–F, 2014–2016). We analyzed both monthly and cumulative pupal stage parasitism by pupal type (winter moth pupae or parasitized by C. albicans), year, site, and deployment using a GLM with a quasibinomial fit.

Parasitism seasonality and year-to-year variation

To assess seasonality of parasitism rates, we used a logistic regression to analyze the monthly rate of parasitism on winter moth pupae weighted by the total pupae analyzed against the main effect of deployment date and included year and site. We used data from the main study sites (Table 1, sites A–H). Similarly, to assess seasonality of parasitism, we used a negative binomial GLM to evaluate counts of adult wasps that emerged from the additional pupae deployed in 2016 (“Pimpla host selection, development time, and sex ratio” section) by date of exposure to parasitism. To assess pupal parasitism across years, parasitism rates from the winter moth deployed across all years and sites (Table 1, sites A–H) were compared using a logistic model using both our estimation of monthly and annual parasitism rates weighted by the total pupae analyzed. For our analysis of monthly parasitism, we included year, site, and deployment as predictors. For our analysis of cumulative pupal stage parasitism, we weighted the logistic regression by the average number of pupae analyzed across the deployments and used year and site as predictors.

Parasitism with winter moth spread and density

To compare Pimpla wasp parasitism on winter moth pupae between sites infested by winter moth (the first eight sites listed in Table 1, to the east of the light dotted line in Fig. 1) compared to sites on the edge of the infestation or outside the infestation area (the last six sites listed in Table 1, to the west of the light dotted line in Fig. 1), we used a logistic regression to analyze the monthly rate of parasitism weighted by the total pupae analyzed with infestation status, year, and site as predictors. We also analyzed the cumulative rate of parasitism weighted by the average number of pupae. Only years 2015 and 2016 were used for these comparisons as they included both heavily infested sites and sites outside the heavily infested area. The designation of a site as winter moth-infested was determined from previous studies of winter moth spread (Elkinton et al. 2014, 2015, 2018).

Fig. 1
figure 1

Average (2014–2017) percent parasitism by Pimpla on winter moth pupae across the pupal deployment sites. The letters for each study site correspond to Table 1 with the six main study sites (A–F) indicated by the gray boxes. The area to the right of the dashed lines approximates the winter moth infestation area for 2007 and 2014 (Elkinton et al. 2014, 2015)

Full size image

To estimate winter moth pupal density at each long-term study plot (Table 1 sites A–H), 16 plastic buckets (16 cm width × 28 cm length × 10 cm height) filled 3 cm deep with sifted peat moss and rainwater drainage holes were placed under each study tree in late May before pre-pupal winter moth caterpillars began to spin down from the tree canopies. Each bucket was placed at a randomly selected distance between the tree stem and the edge of the tree canopy along one of eight evenly spaced directions. To test for density-dependence, we analyzed Pimpla wasp parasitism by winter moth pupae density using a logistic regression of monthly parasitism rates weighted by the total pupae analyzed regressed against the corresponding density of winter moth pupae (log-transformed) with site, year, and deployment as predictors. We also analyzed the cumulative pupal stage parasitism against the log-transformed pupal density with year and site effects.

Statistical analyses

All analyses were performed in R 3.4.4 (RCoreTeam 2013) using RStudio, version 1.1.442 (RstudioTeam 2015). For each analysis, the full model was run initially (including site, year, and deployment effects, etc.), the model was evaluated for evidence skew in the residuals or outliers, and any insignificant predictors were dropped sequentially until the best fit model was selected using AIC comparisons. We checked for overdispersion, and when evidence of overdispersion was noted, we applied a quasibinomial or quasipoisson distribution. Quasibinomial and quasipoisson analyses do not generate AIC values; thus, to select the best fit model, we compared the residual deviance of the fit model to that of the null model. A pseudo-R2 was calculated by comparing the residual deviance of the fit model against the null model (deviance null model − deviance fit model/deviance null model). All graphical data were displayed using ggplot2 (Wickham 2009).

Morphological comparative analyses

Following initial identification of our specimens as belonging to the genus Pimpla, further morphological and molecular identification was conducted using a subset of 302 samples (289 from winter moth and 13 from C. albicans puparia). This subset included wasps reared from all collection sites, seasons, and years, and included males and females. The specimens were initially sorted into putative species based on morphology; species identification was attempted using keys and diagnostic information in Townes (1940) and Townes et al. (1960). Specimens were also compared with authoritatively determined specimens of Pimpla turionellae L., Pimpla contemplator Müller, Pimpla hesperus Townes, Pimpla aquilonia Cresson, and Pimpla disparis Viereck in the Smithsonian Institution National Museum of Natural History, Washington, DC (USNM). All but the last two species have been recorded as attacking winter moth (Silvestri 1941; Wylie 1960; Humble 1985). Pimpla flavicoxis has been reported from winter moth in the British Isles and was previously treated as a junior synonym of P. aquilonia. However, there are no specimens at the USNM identified as P. flavicoxis. P. disparis was introduced into Canada and the U.S. to control other lepidopteran pests. It has been reported as a parasitoid of lepidopterans in multiple families, including geometrids (Yu et al. 2012). We also compared our specimens to the lectotype for Pimpla aequalis Provancher from the Université Laval, Quebec City, Quebec, Canada (ULQC). Vouchers for parasitoid species in this research are deposited in the University of Massachusetts Insect Collection, Amherst, MA (UMEC).

Molecular comparative analyses

DNA extraction, amplification, and sequencing

A subset of the Pimpla wasps that emerged or were dissected from winter moth or C. albicans pupae were selected for molecular analyses. When possible, three adult samples and one larval sample were selected for each location and study year; otherwise, up to four wasps of any life stage were selected for a total of 77 field-collected individuals and 20 laboratory-reared wasps (Table S1). DNA was extracted using the QIAGEN DNeasy Blood and Tissue Kits following the company protocol with the following modifications: for larvae, individuals were destructively sampled by grinding with a mortar and pestle; for adults, DNA was extracted from a single leg removed from the specimen; for both life stages, DNA was eluted twice in 100 μl Buffer AE instead of once with 200 μl. All DNA extractions were stored at − 20 °C for subsequent analysis.

A portion of the mitochondrial locus cytochrome c oxidase subunit I (COI) was amplified using standard PCR techniques. For a subset of individuals collected across years, sites, life stages, and life history (Table S1) fragments from three additional nuclear gene regions were amplified: the carbomoylphosphate synthase domain (Cadherin, rudimentary, CAD), elongation factor 1-α (EF1-α), and the D2 and D3 expansion segments of the large subunit ribosomal RNA gene (28S). The primers and temperature profiles used are outlined in Table S2. For each locus, a master mix was prepared using the following ratios of reagents per sample: 17.3 μl nuclease free water, 0.5 μl dNTPs, 5 μl 5 × GoTaq Buffer (Promega), 0.2 μl GoTaq (Promega), 0.5 μl of both the forward and reverse primer (10 μM each), and 1 μl of eluted DNA. PCR reactions were run on a BioRad T100 thermocycler, and the resulting PCR products were visualized on a 1.5% agarose gel stained with SYBERsafe (Invitrogen, Carlsbad, CA) to verify amplification. Samples that produced bands of the expected fragment size for each locus were then cleaned prior to sequencing using Exonuclease 1 (ThermoScientific) and Thermolabile Recombinant Shrimp Alkaline Phosphatase (New England BioLabs). The resulting products were submitted to The Yale University DNA Analysis Facility on Science Hill for Sanger sequencing in both sense and anti-sense orientations.

The resulting sequences were visualized, and the forward and reverse sequences aligned and edited using Geneious R8.1.8 and R9 (Biomatters Ltd.). The ends of the aligned sequences were trimmed by hand to remove primer sequences and so that all sequences had a high-quality score (> 90% HQ nucleotide reads). The presence of heterozygous sites was determined by Geneious and encoded using IUPAC-IUB ambiguity codes. All ambiguous regions were subsequently inspected by eye. For our COI fragment sequences, we looked for evidence of nuclear mitochondrial DNAs (NUMTs) or pseudogenes by examining for the presence of stop codons based on translation with Invertebrate Mitochondrial DNA genetic code.

Phylogenetic analysis

If any of our sequences were identical, we only included a single representative haplotype. The number of samples and sample identification for each haplotype is outlined in Table S3. To compare our Pimpla COI sequences to previously published Pimpla COI sequences, we searched the National Center for Biotechnology Information (NCBI) GenBank database and the University of Guelph Centre for Biodiversity Genomics’s Barcode of Life Data Systems (BOLD). We initially downloaded triplicate sequences from each Pimpla species available across the two repositories, but when the triplicates were identical to each other or nearly identical (> 99% identical), we then retained one representative sequence for each Pimpla species available. From the BOLD sequences, we prioritized sequences acquired from Pimpla samples in the hymenopteran collection of the Canadian Natural Collection of Insects, Arachnids and Nematodes (Agriculture and Agri-Food Canada) accessed by A. Bennett (accession numbers start with ‘BOLD HYCNG’). These sequences included 57 Pimpla specimens representing 13 of the 19 extant described species in the Nearctic Region (Yu et al. 2012). When no representative sequence was available for a particular Pimpla species by the Canadian National Collection, we searched GenBank for a representative sample, followed by any other sequences available in BOLD. Sequences identified as P. aequalis and as Pimpla sp. that were the closest matches in GenBank to sequences of our each of our two Pimpla clades were included; these were associated with publications (Carpenter and Wheeler 1999; Hebert et al. 2016) and from the International Barcode of Life and NCBI GenBank.

JModelTest was used to select the best substitution model for nucleotide evolution, as implemented in the CIPRES Science Gateway (Miller et al. 2010). We performed neighbor-joining, maximum likelihood, and Bayesian reconstructions using the GTR substitution model. Neighbor-joining analyses were run in Geneious using 1000 bootstrap replications and a majority rule (50%) consensus threshold. Maximum likelihood analyses were run using PhyML (Guindon et al. 2010) with 100 bootstrap replications. Bayesian analyses were run using MrBayes 3.2.6 (Huelsenbeck and Ronquist 2001) with a MCMC chain length of 1,000,000 and a burn in length of 10%. The resulting gene trees were then visualized using FigTree Version 1.4.2 (Rambaut 2014).

To determine whether specimens identified as P. aequalis might be members of a cryptic species complex, we used a multilocus genealogical concordance approach (Andersen et al. 2010; Dettman et al. 2006; Groeneveld et al. 2009; Starrett and Hedin 2007) to estimate the number of species present in our dataset. This method considers lineage sorting in multiple, independent loci and has become a common approach for species delineation. For these analyses, we created separate alignments for each gene fragment including each specimen from which all target loci were successfully amplified. In addition, we included publicly available sequences from Labena grallator (Say) (Hymenoptera: Ichneumonidae) as the outgroup for each alignment. Individual gene trees were estimated for each locus, and the congruence of the topologies of the reconstructed gene trees were then visualized by inferring a majority-rule consensus tree using PAUP (Swofford 2003).

Results

Parasitoid collection

Of the 6580 retrieved pupae and puparia that escaped predation (5009 winter moth pupae and 1571 C. albicans puparia, Table 1) over the study period (2014–2017), 342 were parasitized by Pimpla wasps (305 winter moth pupae and 37 C. albicans puparia). Besides the Pimpla wasps, we recovered only two other species: two winter moth pupae (Wellesley, MA and Pawckatuck, CT; 24 June–5 August 2015) were parasitized by a species of Cratichneumon, and four C. albicans puparia (Kingston, RI; 5 August–18 September 2015) were parasitized by a brood of diapriid wasps. Of the Pimpla wasps recovered from field-deployed sentinel pupae and puparia, 46% were adults and the rest larvae.

Parasitism seasonality, overwintering, and year-to-year variation

Monthly parasitism on winter moth pupae varied from 0 to 52% (Fig. S3). Deployment was marginally significant with parasitism rates from early August to mid-September slighter higher than those of either late June to early August or mid-September to late October (df = 100, pseudoR2 = 0.34, p = 0.06). Pupae exposed to Pimpla wasps after the first week of August had wasps that did not emerge until the following spring (the overwintering generation), and most of the overwintering wasps (91%) were from winter moth pupae that were exposed to wasps after the first week of September. Cumulative pupal stage parasitism (one minus the product of the survivorship of each pupal deployment for the duration of the pupal life stage, see “Monthly and annual parasitism rate estimates” section) ranged from 0 to 92% (Table 1). Pimpla wasps were recovered from all years and sites, though some sites had a year without recoveries. There was a significant effect of year but not site or deployment date when analyzing monthly parasitism rates (df = 57, pseudoR2 = 0.40, p = 0.034), and site was significant when analyzing the cumulative pupal stage parasitism rates (Fig. S4; df = 13, pseudoR2 = 0.72, p = 0.009). Parasitism rates were highest in 2015.

Parasitism with winter moth spread and density

Pupal density of the study sites had a significant effect on the monthly (df = 62, pseudoR2 = 0.25, p = 0.039) and cumulative pupal stage parasitism (df = 19, pseudoR2 = 0.39, p = 0.036), with a significant effect of year in both models (Fig. 2). No significant difference was found in percent Pimpla wasp parasitism on winter moth pupae that were deployed in sites on the edge of the current winter moth infestation area compared to the heavy infestation area (Fig. 1). However, there was a trend toward higher parasitism rates at the edge of the winter moth infestation (Fig. S5). Pupae deployed in sites at the edge of the winter moth infestation had a mean cumulative pupal stage parasitism rate of 0.53 ± 0.08 (mean ± SE), while the cumulative parasitism was 0.32 ± 0.03 in the infested sites.

Fig. 2
figure 2

Logistic relationship between monthly (left) and cumulative pupal stage (right) Pimpla parasitism on pupae by winter moth pupal density across sites for the six main study sites. Each point indicates each site for each year, the solid lines show the fit model, and the dashed lines show confidence intervals

Full size image

Wasp development time and sex ratio

It took 21.2 ± 0.6 SE days from the date of parasitism to adult wasp emergence (n = 39) for pupae parasitized in the laboratory. No wasp emergence was noted after 9 October until spring. Thus, if Pimpla wasps parasitize between 1 June and 1 October (122 days) and if wasps take 21 days from oviposition to emergence, then we can expect up to 5 generations per season and a final 6th overwintering generation. This estimate is from individuals held at a constant 23 °C temperature; however, in the field average temperatures are slightly cooler (~ 19 °C, Table S4), so development may be slower under field conditions. Across studies (field deployed pupae and laboratory rearing), there were 305 Pimpla females and 238 Pimpla males, which suggests a 1:1 female to male ratio.

Pimpla host selection

From the three deployments of pupae placed in Wellesley, MA and Hanson, MA in 2016, there was a significant effect of soil treatment (df = 8, pseudoR2 = 0.57, p = 0.0043) and deployment (p = 0.0018). More wasps emerged from the pupae that were covered only by a layer of leaves than from those buried under soil and leaves. In the host choice study, we did not observe any wasps attempting to oviposit in the C. albicans puparia, and no wasps emerged from these trials. However, from the field studies, we found that C. albicans puparia can be parasitized by Pimpla spp. but at a significantly lower rate than winter moth; this was true for the model that included monthly parasitism rates (df = 83, pseudoR2 = 0.39, p = 0.00012) and the model with cumulative pupal stage parasitism rates (df = 22, pseudoR2 = 0.64, p = 0.0018). Mean cumulative parasitism by Pimpla on C. albicans puparia was 0.15 as compared to 0.27 for winter moth pupae.

Molecular and morphological comparative analyses

Based on morphological features, a subset of Pimpla wasps were identified by Dr. David Wahl (Utah State University) as P. aequalis. Subsequently, additional samples were identified by the second author (RRK) as P. aequalis using the key presented in Townes (1940) and Townes et al. (1960) and comparison with the lectotype for P. aequalis (a female, Fig. S7). The identification was confirmed by another ichneumonid systematist (B. Santos, Smithsonian Institution NMNH) through examination of a subset of the identified specimens. However, results from analysis of molecular data (see below) suggest that the Pimpla wasps in this study actually comprise two species hereafter referred to as Pimpla sp. 1 and Pimpla sp. 2. Images of a female of Pimpla sp. 1 and a male of Pimpla sp. 2 are presented in Fig. 5a–f, respectively (note: we did not rear females of Pimpla sp. 2). While there appear to be subtle morphological differences between Pimpla sp. 1 and Pimpla sp. 2, particularly shape of the mesosoma, whether they are reliable for differentiating these species is equivocal because we have only one adult male of Pimpla sp. 2. All other specimens of Pimpla sp. 2 were larvae from parasitized hosts.

We acquired high quality COI sequences for 74 individuals representing all sites and years (50 adults and 24 larvae; Table S1). Based on reconstruction of the phylogeny using the COI gene fragment, our samples separated into two distinct clades that exhibited 9.7–10.1% sequence divergence (Fig. 3). All nucleotide differences between the two clades represented third-codon substitutions and thus likely represent genetic differences accumulated due to genetic drift and not selection. Both Pimpla clades included wasps acting as primary parasitoids and hyperparasitoids (49 primary and 26 hyperparasitoids; Table S1). Additionally, we obtained high quality sequences for fragments of CAD, EF1-α, and 28S from 26 Pimpla wasps (19 Pimpla sp. 1 individuals and 7 Pimpla sp. 2 individuals as from the COI analyses). For CAD, sequences from the two Pimpla clades were 1.5–3.4% different from each other, with six base-pair differences fixed between the two Pimpla clades (Fig. 4, Table S7). Similarly, for EF1-α the two clades were between 0.7 and 1.6% different with 3 base-pairs consistently different between clades. See Tables S6–S8 for distance matrices for each alignment. For 28S, there were no differences between individuals from the two COI clades, with all but two individuals having identical sequences; the two individuals differed from the other by a single base-pair substitution. Because 28S was invariant, it was left out of the multilocus analyses. The trees constructed from each of these three loci (Fig. 4) and the majority rule consensus tree (Fig. S6) all supported the presence in our samples of two cryptic Pimpla species within what we considered P. aequalis based on morphology (Fig. 5).

Fig. 3
figure 3

Phylogentic inference of our Pimpla samples (bolded) and representative sequences. Representative sequences were downloaded from NCBI GenBank and BOLD. The tree was constructed using a Bayesian analysis with a 604 bp region of the COI locus. Where our sequences were identical, the groups were collapsed as outlined in Table S2. Pimpla aequalis sp. 1 are noted with red and sp. 2 with blue. Branch lengths are drawn proportional to the rate of change observed. The number to the left each node represents the bootstrap support value for the branch (Bayesian over the Maximum Likelihood)

Full size image
Fig. 4
figure 4

Phylogentic inferences using a COI, b CAD, and c EF1-α gene regions for only our P. aequalis sp. 1 (red) and sp. 2 (blue) samples. Branch lengths are drawn proportional to the rate of change observed. The number to the left each node represents the bootstrap support value for the branch (Bayesian over the Maximum Likelihood)

Full size image
Fig. 5
figure 5

The two Pimpla species reared from winter moth pupae in this study. a–cPimpla sp. 1, female a Lateral habitus; b lateral of head and mesosoma, c Dorsal of head and mesosoma, d–fPimpla sp. 2, male, d Lateral habitus, e lateral of head and mesosoma, f Dorsal of head and mesosoma. Scale bars = 1.00 mm

Full size image

Based on these comparisons, we appear to have two distinct species of Pimpla that fit P. aequalis sensu Townes (1940) and Townes et al. (1960), and these species are not any Palearctic species available to us known to attack winter moth in Europe. It is likely that either Pimpla sp. 1 or Pimpla sp. 2 is P. aequalis. The lectotype of P. aequalis is more similar morphologically to specimens of Pimpla sp. 1 than Pimpla sp. 2; however, Pimpla sp. 2 is known from a male only, and the P. aequalis lectotype is a female. Determining the identities of these clades, and discerning the identities of Holarctic species of Pimpla in general, would require extensive analysis of morphological and molecular data for Pimpla species in the Nearctic and Palearctic regions, including primary type specimens. The sequences generated from our specimens did not match any published sequences available for other Pimpla species, which includes 13 of the 19 extant described species in the Nearctic Region (Yu et al. 2012) and an additional six species of Pimpla from outside the Nearctic Region. The only molecular matches were for other unknown Pimpla species collected from the northeastern United States and southeastern Canada. Thus, Pimpla sp. 1 and Pimpla sp. 2 are presumably native to North America.

Discussion

Biotic resistance is the process by which native natural enemies spill over from native species to attack an invasive species and reduce the success of that invader (Diez et al. 2008; Elton 1958; Levine et al. 2004; Sakai et al. 2001; Shea and Chesson 2002). We found evidence of biotic resistance to invasive winter moth populations in the northeastern United States; in this most recent invasion, winter moth populations are sustaining heavy, density dependent parasitism by two cryptic species of Pimpla. One of the Pimpla wasps is likely P. aequalis, while the other appears to be a related unknown species. While the association of these Pimpla wasps with winter moth is recent, their impact is notable; estimates of parasitism across our study sites and years were on average 6% and were found to be as high as 52% on winter moth pupae in infested areas. Pimpla wasps were found across all of our study plots. Where winter moth has invaded, Pimpla wasp parasitism responded to winter moth pupal density in a positive density dependent manner and thus has the potential to be regulatory.

The ability of a native predator or parasitoid to respond functionally or numerically to a primary host population, while also using an alternative host when the primary host is at low densities, may result in particularly effective suppression of populations of non-native invader by a native species (Nechols et al. 1992; Shea and Chesson 2002). In this way, the native natural enemy has the ability to build up using the novel host but can also use other host species when this new host is less abundant. As a result, the natural enemy is maintained in the community and can aggregate or respond numerically when the invasive alien population outbreaks or spreads. This pattern of host use is common for a number of generalist predators, parasitoids, and pathogens (deRivera et al. 2005; Hassell and Rogers 1972; Holling 1973; Murdoch 1969; Oaten and Murdoch 1975; Schenk and Bacher 2002; Strauss et al. 2012). Parasitism by Pimpla wasps in this study exhibited these characteristics. While Pimpla wasps may build up using winter moth as a host species and respond to high densities with higher rates of parasitism, it also appears to be able to use other host species when winter moth populations are not at high densities.

We detected a positive density dependent relationship of parasitism to winter moth density, suggesting that Pimpla wasps may have a regulatory effect on the winter moth population densities. Further, since the Pimpla wasps are multivoltine, the wasp population may be able to build up in outbreak populations in a numerical response, which could control an outbreaking population (Holling 1973). However, we did not detect increased parasitism rates over either season or years and were not able to test for a numerical response to densities. In the absence of a documented numerical response, it may be that the density dependent response we found may arise from a Type III functional response driven by Pimpla wasp host switching behavior (Holling 1959; Murdoch 1969). While pimplines typically have a preferred host species (niche specialization), they are facultative generalists and can use a broad host range spanning multiple lepidopteran families (Bennett 2008; Fitton et al. 1988; Krombein et al. 1979). Parasitism of winter moth by Pimpla wasps is likely the result of natural enemy spillover from the native congener Bruce spanworm or other related lepidopteran species in the area. We found that Pimpla wasps exhibited particularly high parasitism 95 km beyond the nearest high density winter moth infestation (i.e., Framingham, MA) and 22 km from the nearest capture of winter moths in pheromone traps (i.e., Orange, MA). The finding that Pimpla wasps were present across the study area, including areas beyond the range that winter moth has spread, supports the hypothesis that the Pimpla species are native. Pimpla wasp population densities respond to, but do not depend on, the presence of winter moth. In fact, we found particular high parasitism rates at the leading edge of the winter moth infestation. This is likely because in these interior sites Pimpla wasps, as native generalist parasitoids, are maintained at high densities by a robust Lepidoptera community. In this way, Pimpla wasps may provide a biotic resistance barrier to the spread of winter moth, which was found by Elkinton et al. (2014) to be slowing. Pimpla wasps may help control winter moth spread, but this warrants further study.

If Pimpla wasps act as hyperparasitoids (a parasitoid of a parasitoid) of any native or introduced parasitoids (biological control agents), then they have the potential to reduce population control of the invasive species by inflicting more mortality on the biological control agent than on the invasive species itself. However, if a parasitoid inflicts more mortality on the invasive species than on any introduced biological control agent, then it aids in controlling the pest population (Brodeur 2000; Nechols et al. 1992; Sullivan 1987). While we found that Pimpla wasps can hyperparasitize, as is typical of many pimpline wasps (Bennett 2008; Fitton et al. 1988) and was found previously (Broadley et al. 2018), they appear to do so only facultatively. From our laboratory host range study, when given a choice of winter moth pupae or winter moth pupae parasitized by C. albicans, the wasps only parasitized winter moth pupae. From our field study, both species of Pimpla wasps can act as hyperparasitoids, but they do so at rates significantly lower than their rate of primary parasitism. Hyperparasitism on C. albicans at these sites was primarily caused by wasps from the genus Phygadeuon rather than Pimpla (Broadley et al. 2018). Together, this further demonstrates that Pimpla wasps contribute to the population control of winter moth.

We detected two species of Pimpla parasitizing winter moth pupae; one is likely P. aequalis, the other is unknown and is possibly an undescribed species. For our estimates of percent parasitism, we did not distinguish between the two species of Pimpla, but from the molecular work, one clade is better represented than the other; the majority (88%) of the randomized samples we tested belonged to Pimpla sp. 1. Both species were found across study site, season, and year, and both species acted as both primary and hyperparasitoids. While both species were initially considered P. aequalis based on morphology, we were not able to acquire DNA from the lectotype to discern if one of our two species is P. aequalis. Pimpla aequalis is known only from North America, and we consider both Pimpla species we recovered from winter moth as native to North America. However, some species of Pimpla are known to have a Holarctic distribution (Yu et al. 2012). The overall geographic distribution of the Pimpla species reported here cannot be determined until their identities have been discerned; this would require more extensive taxonomic research on Pimpla species in the Nearctic and Palearctic regions, such as a revision that includes both morphological and molecular data.

Winter moth has been extensively studied in its native range in Europe (e.g. Klemola et al. 2008; Myers and Cory 2013; Tenow et al. 2013; Varley et al. 1973; Vindstad et al. 2011, 2013), as well as in the prior accidental introductions to North America (e.g. Embree 1965; Roland 1990; Roland and Embree 1995); however, this is the first report of Pimpla species as important parasitoids of winter moth. To our knowledge none of the studies of winter moth pupal mortality in Nova Scotia recorded parasitism by Pimpla wasps, although native parasitoids were noted (Embree 1965; Graham 1958; Macphee et al. 1988; Pearsall and Walde 1994; Roland 1990). In British Columbia, Coccygomimus (= Pimpla) hesperus was recorded from Operophtera spp. pupae (Humble 1985) but was not recorded in later studies (Roland 1990; Roland and Embree 1995). This suggests that the wasps were accidently overlooked in the prior studies, that Pimpla wasps now show more host switching to winter moth than when winter moth was first introduced to North America in the 1930s, or that the region of this study has more abundant Pimpla wasp populations with a wider host range than was found in Nova Scotia or British Columbia. Surveys were conducted in winter moth infested sites in coastal Maine, and species of Pimpla, likely P. aequalis and other morphologically similar congeners, were found there (Morin 2015); however, the study did not include sentinel winter moth and thus showed co-occurrence but not parasitism.

We were surprised that Pimpla sp. 1 detected in this study seemed to be the only major species that parasitized winter moth pupae. The lack of additional parasitoids and slow recruitment of Pimpla wasps may help explain why winter moth has been such a high density pest in its introduced region. We looked for parasitism by P. contemplator and P. turionellae, which are known parasitoids of winter moth in Europe (Wylie 1960) and were introduced to southeastern Canada in an attempt to control winter moth, but there is no evidence that they established (Carlson 1979; Graham 1958). Further, P. disparis, a parasitoid introduced in this region to control gypsy moth, was also a potential candidate since P. disparis has a broad host range, attacking lepidopterans of at least 14 families (Schaefer et al. 1989). However, these Pimpla species were not detected. Besides the two species of Pimpla reported here, we only had a few cases of parasitism by diapriid and Cratichneumon wasps. Cratichneumon culex (Muell.) has been recorded as an important parasitoid of winter moth pupae in Europe (East 1974; Hassell 1969; Varley et al. 1973; Wylie 1960), and an undescribed Cratichneumon species was reared from winter moth in British Columbia (Humble 1985). As far as we know, C. culex has not been introduced to North America (Embree 1966; Graham 1958).

Conclusions

Our findings suggest an important role of Pimpla wasps in the population dynamics of winter moth, an invasive forest pest. Overall, this study shows that biotic resistance from a native parasitoid is affecting the dynamics of the winter moth invasion. We urge future research on parasitism by native species in the study of biological invasions, as knowledge of biotic resistance from native natural enemies on invasive populations is essential to our overall understanding of invasions. We also encourage insect biological control practitioners to consider not only the effect of an introduced biocontrol agent but also the effect of native parasitoids and the potential interactions between introduced biocontrol agents and native natural enemies.