Assessment of the Biological Control Potential of Common Carabid Beetle Species for Autumn- and Winter-Active Pests (Gastropoda, Lepidoptera, Diptera: Tipulidae) in Annual Ryegrass in Western Oregon

Simple Summary Many studies have shown that ground beetles feed on different agricultural pests, but little is known about their species communities from US cropping systems. We assessed the biological control potential of the most common carabid beetle species in Oregon annual ryegrass grown for seed by investigating spatial and temporal overlap of the most common species with those of the most damaging autumn- and winter-active pests (slugs, caterpillars and cranefly larvae) and determined the number of field-collected specimens that had fed on the respective pests using molecular gut content analysis. Only the non-native Nebria brevicollis was abundant during pest emergence and tested positive for all three pest groups. While the other common carabid beetle species—Agonum muelleri, Calosoma cancellatum and Poecilus laetulus—were also found to have consumed pests, they were active only during spring and summer, when crop damage by pests is less critical. We also show that disk tilling did not affect any of the four common carabid beetle species and that only N. brevicollis was significantly associated with a vegetated field margin. This study contributes to expanding our knowledge on conservation biological control in a system where chemical pesticides are still the mainstay of control against invertebrate pests. Abstract While carabid beetles have been shown to feed on a variety of crop pests, little is known about their species assemblages in US annual ryegrass crops, where invertebrate pests, particularly slugs, lepidopteran larvae and craneflies, incur major financial costs. This study assesses the biological control potential of carabid beetles for autumn- and winter-active pests in annual ryegrass grown for seed by: (a) investigating the spatial and temporal overlap of carabids with key pests; and (b) molecular gut content analysis using qPCR. Introduced Nebria brevicollis was the only common carabid that was active during pest emergence in autumn, with 18.6% and 8.3% of N. brevicollis collected between September and October testing positive for lepidopteran and cranefly DNA, respectively, but only 1.7% testing positive for slug DNA. While pest DNA was also detected in the guts of the other common carabid species—Agonum muelleri, Calosoma cancellatum and Poecilus laetulus—these were active only during spring and summer, when crop damage by pests is less critical. None of the four carabid species was affected by disk tilling and only N. brevicollis was significantly associated with a vegetated field margin. However, as its impact on native ecosystems is unknown, we do not recommend managing for this species.


Study Sites and Sampling
This study was carried out in ten AR fields grown for seed in Linn County, Oregon, between 23 March 2018 and 20 June 2019. Five of these fields ('experimental sites', as tillage regime changed between the two years) were not tilled in August/September 2017 but tilled in August/September 2018, while the other five fields ('control sites', as tillage regime did not change between the two years) were tilled in both years (Table 1). Tillage at all sites was carried out with a disk ripper at 10-15 cm depth followed by secondary tillage with harrow and roller. AR was seeded in all ten sites in early autumn 2017 using a seed drill and fields were harvested the following June/July before being tilled in August/September and planted again in September 2018. Field sizes ranged from 15 to 81 hectares and soil types were moderately to poorly draining silt loams. More details on each site as well as usage of herbicide, fertilizer and other treatments can be found in Table S1. Two parallel rows comprising five trapping points each were set up in every field. Every trapping point consisted of four pitfall traps surrounding one refuge trap (Figure S1), and one data logger was placed within each row. Rows 1 ('edge traps') and 2 ('field traps') were located at 10 and 70 m, respectively, from the closest field margin ( Figure S1A). These distances were selected as Holland et al. [52] showed that Insects 2020, 11, 722 4 of 22 species associated with the field margin were captured mostly within 60 m from the margin. A distance of 60 m between the traps was deemed sufficient to ensure sample independence, as Thomas et al. [53] demonstrated in a mark-recapture study that even after 30 days, Pterostichus melanarius Illiger rarely traveled further than 55 m from their release site. Within each tillage regime, the closest field margin of three sites consisted of a gravel strip adjacent to a road, which filled with water in the winter, while the closest field margin in the other two sites consisted of taller grasses, shrubs and trees, with the first few meters between the field and the margin usually sprayed with a herbicide. In sites C4, C5 and E4, the vegetation bordered a river, while it bordered a private garden in site E5 ( Figure S2). Traps in site C1 had to be moved by approximately 500 m in September 2018 as the field was split and the initial trap location was planted with meadowfoam Limnanthes alba Benth but they stayed at the same distance to the closest field margin.
Pitfall traps ( Figure S1C) were opened between 18:00 and 20:00 and collected the next morning between 06:00 and 08:00. Trap contents were emptied into labeled resealable plastic bags and placed immediately into a portable freezer containing dry ice to prevent further prey digestion as well as in-trap predation. Upon return to the laboratory, the bags were frozen until identification to species (carabid beetles, following [54]) or at least order (other invertebrates) level, after which samples were put into individual Eppendorf tubes containing 70% ethanol and stored at −20 • C. Refuge traps ( Figure S1B) were deployed immediately after the pitfall traps were collected (to avoid providing a 'slug hotspot' for beetles to feed on right before pitfall trapping began) and were checked for slugs, slug eggs, caterpillars and cranefly larvae one week later between 09:00 and 18:00. After collection, all slugs and caterpillars were weighed and identified to species level. The grey garden slug, Deroceras reticulatum, was classified into four life stages based on weight (small 0-10 mg, intermediate 10-100 mg, large >100-200 mg and very large >200 mg) [55][56][57]. Cranefly larvae and slug eggs were counted in the field.
Sampling took place weekly and alternated between pitfall and refuge traps; the order in which fields were sampled was rotated to avoid sampling-time dependent bias. Not all fields could be sampled on every occasion due to farming operations in the fields, adverse weather conditions or flooding of certain sites. No pitfall sampling was undertaken in February 2019 due to very low overnight temperatures and persistent freezing rain. In total, there were between 23 and 26 refuge trap sampling occasions and between 21 and 25 pitfall trap sampling occasions per field (Table S2).
Three soil samples were collected monthly between 31 October 2018 and 5 May 2019 at 10, 40 and 70 m into the field using a bulb planter (diameter 7 cm, depth 10 cm). They were transferred into resealable bags and immediately put in a cooler chest before being stored at 5 • C in the laboratory. The following day, soil samples were thoroughly mixed within the bags and 25 ± 0.2 g of each sample were weighed out into aluminum foil pans (20 cm × 9.5 cm × 6.3 cm) and put into a drying oven at 105 • C for 24 h, after which they were weighed again to determine the soil moisture content. Plant species and percentage ground cover of vegetation (grouped by vegetation height: small herb layer <0.5 m, large herb layer 0.5-1 m, shrub layer >1-4 m, and tree layer >4 m) were recorded along the field edges adjacent to the traps in June 2019 (Table S3).

Statistical Analyses
All statistical analyses were carried out in R 3.6.1 [58]. The soil moisture content of the three samples collected at each time point were compared within each site for each sampling occasion using the Kruskal-Wallis test. Following a non-significant result, the same data were used to compare the soil moisture content between all sites using the Kruskal-Wallis test followed by Dunn's test adjusted for multiple comparisons with 'bh' [59] ( Figure S3, Table S4).
To investigate the effect of field margin structure on the distribution of the most common carabid species, generalized linear mixed-effects models (GLMMs) were created using the lme4 package [60]. Count data of each trapping point per site and negative binomial distributions were used for all species, as the data were overdispersed. They included either the random effect 'site' (categorical with ten levels, see Table 1) only (null model), the random effect plus either of the fixed effects 'field margin' (categorical with two levels: gravel and vegetation, Table 1) or 'soil moisture' (numerical as 'mean  water loss', Table S4) or all three effects (full model). The latter effect was included to detect possibly false associations with field margin type, as all four fields with vegetation margin were also those with the lowest soil moisture. The models were compared using parametric bootstrapping (N = 10,000) with the package pbkrtest [61]. Adjusted R 2 values were calculated using rsq [62].
To investigate the influence of disk tilling on the activity density of the most common carabid species, a Before/After and Control/Impact (BACI) analysis was carried out using a two-way ANOVA with the factors 'tillage regime' (categorical with two levels: control and experimental) and 'year' (categorical with two levels: 'year 1' (23 March-13 June 2018) and 'year 2' (26 March-20 June 2019)) with the square root-transformed total number of specimens per site.

Molecular Gut Content Analysis
A total of 681 guts were dissected from the four most commonly found carabid species, considering both overall abundance and occurrence at eight or more sites (Table S5). Dissection of the crop and proventriculus was carried out in a sterile Petri dish using forceps that were sterilized with 70% ethanol and heat (using either an open flame or the ErgoSteri™ VT Glass Bead Sterilizer (PhytoTech Labs, Lenexa, KS, USA) 110 V at 250 • C) between each beetle. The guts were individually collected in 1.5 mL Eppendorf tubes and stored at −20 • C until DNA extraction.
The beetle guts were homogenized using a Kimble Kontes Pellet Pestle (DKW Life Sciences, Milville, NJ, USA) or bead beating and left until complete lysis in a water bath with frequent vortexing (usually approximately 5 h). Total DNA was extracted from 534 beetle gut samples ( Table 2) in batches of 24 (23 samples and one blank extraction) using the DNeasy Blood and Tissue Kit (QIAGEN, Hilden, Germany), according to the manufacturer's instructions. Guts that were damaged during dissection or that contained little or no food were not used for further analysis. The gut extracts were screened using qPCR to identify feeding incidences on the commonly encountered pest groups in the fields: slugs, craneflies and caterpillars. Fewer gut samples were screened for craneflies compared to the other pests as larvae were only observed in the fields between December and May. Due to the variety of slug, caterpillar and cranefly species in the fields and our focus on pest groups rather than individual pest species, class-(Gastropoda), order-(Lepidoptera) or family (Tipulidae) specific primers were selected (GastNLSf1/Sr1 [63], Tip-gen-S267/A268 and Lep-gen-S274/A275 (both [64])) after tests confirmed the amplification of target and non-amplification of non-target species which were commonly found during this study (Table S6). While the GastNLS primer set would also amplify snail DNA, we did not see this as a problem as these gastropods can also cause significant economical damage to crops [65].
qPCR reactions (20 µL) contained 1X PowerUp SYBR Green Master Mix (Thermo Fisher Scientific, Waltham, MA, USA), 0.25 µM of each primer, 1 µL of template DNA and nuclease-free water. Reactions were run on an Applied Biosystems QuantStudio 3 or StepOnePlus Real-Time PCR System (Thermo Fisher Scientific, Waltham, MA, USA) using the following conditions for amplification: 95 • C for 10 min, followed by 40 cycles of 95 • C for 15 s and 60 • C for 1 min. To ensure that only the target product was amplified, Insects 2020, 11, 722 6 of 22 a melting curve analysis was conducted after amplifications by heating samples to 95 • C for 15 s followed by 60 • C for 1 min and steadily increasing the temperature at 0.3 • C per min to 95 • C. It was confirmed that the PCR products dissociated between 87.5 and 88.1 • C (Gastropoda), 75.8 and 77.8 • C (Tipulidae) and 76.9 and 78.2 • C (Lepidoptera), respectively. Each 96-well plate contained 92 samples, two positive controls of target species DNA and two negative controls (non-target species and water).
A total of 1840 slugs belonging to five species were collected: the most abundant was non-native D. reticulatum (86.3%) followed by the Holarctic Deroceras laeve M

Temporal and Spatial Overlap of Carabids and Annual Ryegrass Pests
The most abundant carabid species had highest activity densities in spring and summer, beginning from April (A. muelleri and N. brevicollis) or May (P. laetulus and C. cancellatum). Harvest operations started during the final week of June and halted sampling until the end of July or start of August (depending on the field), at which time only C. cancellatum was still captured in significant numbers ( Figure 1A-D). Nebria brevicollis had a second activity phase in October and November with single specimens found throughout the winter as well. Additionally, its surface-active larvae were trapped between November and May, with a peak in larval activity in March and April ( Figure 1A). Tenerals (newly emerged adults) were found between late March and early June, with a peak in late April/early May. ller (6.6%), of which both, native and introduced populations are found within the US, and non-native Arion circumscriptus Johnston (5.8%) ( Table S7). A total of 1026 cranefly larvae (Tipula spp.) and 675 caterpillars (Lepidoptera) were also recorded during this study. The most abundant caterpillar was N. pronuba (89%) followed by Cnephasia longana Haworth (1.7%) and Mythimna unipuncta Haworth (1.6%).

Temporal and Spatial Overlap of Carabids and Annual Ryegrass Pests
The most abundant carabid species had highest activity densities in spring and summer, beginning from April (A. muelleri and N. brevicollis) or May (P. laetulus and C. cancellatum). Harvest operations started during the final week of June and halted sampling until the end of July or start of August (depending on the field), at which time only C. cancellatum was still captured in significant numbers ( Figure 1A-D). Nebria brevicollis had a second activity phase in October and November with single specimens found throughout the winter as well. Additionally, its surface-active larvae were trapped between November and May, with a peak in larval activity in March and April ( Figure 1A). Tenerals (newly emerged adults) were found between late March and early June, with a peak in late April/early May.
Slug activity was highest in May and relatively consistent for the rest of the year except during October 2018, when no slug activity was recorded ( Figure 1E; due to summer aestivation no sampling for slugs took place between July and September). Eggs were found between April and May 2018 and between March and June 2019. While all life stages were found on nearly every sampling occasion, neonates were most abundant between April and May 2018 and again in June 2019, and juveniles were most abundant from May until December 2018 and a little less so from January to March 2019. Main activity for caterpillars was between November and March, when 83% of the total number were caught ( Figure 1F). Cranefly larvae were observed only between December and May with 97% found between December and March ( Figure 1G). Insects 2020, 11, x 7 of 22 Slug activity was highest in May and relatively consistent for the rest of the year except during October 2018, when no slug activity was recorded ( Figure 1E; due to summer aestivation no sampling for slugs took place between July and September). Eggs were found between April and May 2018 and between March and June 2019. While all life stages were found on nearly every sampling occasion, neonates were most abundant between April and May 2018 and again in June 2019, and juveniles were most abundant from May until December 2018 and a little less so from January to March 2019. Main activity for caterpillars was between November and March, when 83% of the total number were caught ( Figure 1F). Cranefly larvae were observed only between December and May with 97% found between December and March ( Figure 1G).
Nebria brevicollis was found in all ten sites but was most commonly found in three of the four sites with a vegetation field margin and tended to be on the drier end of the soil moisture spectrum ( Figure 2A). The other three common carabid species were found more commonly in the slightly wetter sites with gravelly field margins ( Figure 2B-D). Calosoma cancellatum was completely absent from sites C3 and C5, while P. laetulus was absent from site E5. The largest number of slugs was found at sites E4 and E5 ( Figure 2E), while caterpillars were most abundant at site C3 ( Figure 2F) and cranefly larvae in site E3 ( Figure 2G). Nebria brevicollis was found in all ten sites but was most commonly found in three of the four sites with a vegetation field margin and tended to be on the drier end of the soil moisture spectrum (Figure 2A). The other three common carabid species were found more commonly in the slightly wetter sites with gravelly field margins ( Figure 2B-D). Calosoma cancellatum was completely absent from sites C3 and C5, while P. laetulus was absent from site E5. The largest number of slugs was found at sites E4 and E5 ( Figure 2E), while caterpillars were most abundant at site C3 ( Figure 2F) and cranefly larvae in site E3 ( Figure 2G).  Table S4). Sites to the left of the broken line are those with a gravel margin, sites to the right have a vegetation margin.  Table S4). Sites to the left of the broken line are those with a gravel margin, sites to the right have a vegetation margin.

Impact of Field Margin and Disk Tilling on Carabid Species Distributions
The model containing the fixed effect 'field margin' was a significantly better fit than the model containing only the random effect 'site' for explaining the distribution of N. brevicollis, P. laetulus and A. muelleri and a better fit than those with 'soil moisture' or 'field margin' + 'soil moisture'. Neither of the models containing any of the fixed effects was a significantly better fit than the model with only the random effect for C. cancellatum (Table 3). Association with the fixed effect was low in all cases and including 'site' produced a worse model fit than if it was omitted for N. brevicollis and P. laetulus. Table 3. Results of the generalized linear mixed model analysis showing the best scoring model for each carabid species. Shown are the X 2 and p-value of the comparison with the model containing only the random effect 'site' following parametric bootstrapping (N = 10,000) as well as the adjusted R 2 values of the fixed effect (FE), random effect (RE) and full model (F). When comparing the spring and summer data from 2018 and 2019, N. brevicollis numbers decreased sharply in two of the experimental sites ( Figure 3), but no significant effects of 'tillage regime', 'year' or the interaction of both factors were found in the BACI analysis for this or any of the other three carabid species (Table 4).

Impact of Field Margin and Disk Tilling on Carabid Species Distributions
The model containing the fixed effect 'field margin' was a significantly better fit than the model containing only the random effect 'site' for explaining the distribution of N. brevicollis, P. laetulus and A. muelleri and a better fit than those with 'soil moisture' or 'field margin' + 'soil moisture'. Neither of the models containing any of the fixed effects was a significantly better fit than the model with only the random effect for C. cancellatum (Table 3). Association with the fixed effect was low in all cases and including 'site' produced a worse model fit than if it was omitted for N. brevicollis and P. laetulus. When comparing the spring and summer data from 2018 and 2019, N. brevicollis numbers decreased sharply in two of the experimental sites ( Figure 3), but no significant effects of 'tillage regime', 'year' or the interaction of both factors were found in the BACI analysis for this or any of the other three carabid species (Table 4).

Identification of Pest DNA in Carabid Gut Contents by qPCR
Gastropod-specific DNA was amplified from three of the commonly found carabid species: N. brevicollis (26 out of 280 adult beetles (9.3%), and one out of 36 larvae (2.8%)), P. laetulus (four out of 99 beetles (4.04%)) and A. muelleri (one out of 51 beetles (1.96%)). All but one of these were from specimens collected between April and June; one N. brevicollis adult tested positive in October, when no slugs were found in the fields ( Figure 4A). The majority of these beetles were collected from the two sites which had the largest number of slugs, 13 from site E4 and nine from site E5.
Screening for other pests in AR revealed that lepidopteran DNA was detected in 21 out of 280 (7.5%) N. brevicollis, one out of 44 (2.3%) A. muelleri and four out of 68 (5.9%) C. cancellatum. Fourteen out of 26 positive Lepidoptera samples were from N. brevicollis collected between September and December 2018, seven from specimens collected in April and May and five from carabids collected in June ( Figure 4B). Only the three positive N. brevicollis that were trapped in November and December coincided with a period when caterpillars were found in large numbers. All C. cancellatum testing positive for Lepidoptera were collected in June, at a time when only a few caterpillars were detected using refuge and pitfall traps. Most positive testing carabids (nine) were collected in site C1, the site with the second largest total number of caterpillars ( Figure 2F).
Tipulidae-specific DNA was detected from 14 out of 228 (6.1%) N. brevicollis, three out of 45 (6.7%) A. muelleri and three out of 61 (4.9%) P. laetulus. All but three of the 20 carabids that tested positive for Tipulidae were collected in April and early May of both years, at times when only a few or no cranefly larvae were observed in the fields, the other three were from N. brevicollis adults collected between October and December, two at times when no cranefly larvae were found ( Figure 4C). The majority of all carabids testing positive for cranefly DNA (five) were collected from site E3, again the site where cranefly larvae were most abundant.

Discussion
The three pest groups that were monitored for this study were selected as they are active between autumn and spring and are known to cause serious damage to the emerging AR crop [47,49].

Discussion
The three pest groups that were monitored for this study were selected as they are active between autumn and spring and are known to cause serious damage to the emerging AR crop [47,49]. Thus, the identification of natural enemies that are active during the same time period is an important consideration for biological control. Of the common carabid species in AR, only N. brevicollis was found in large numbers during the early parts of this critical period, when it emerges from summer diapause to mate and lay eggs [66]. Abundance of this nocturnal carabid, which overwinters mainly in its larval stage [67], decreased in November but activity of N. brevicollis briefly overlapped with that of slugs and N. pronuba in late October/early November ( Figure 1A,E,F). Nebria brevicollis larvae were found between December and April, with the highest activity between March and April followed by the emergence of the first tenerals around the same time, starting the spring activity phase of this species. Other common carabids in the fields also started to become active in late March/early April (A. muelleri, Figure 1D) or late April/early May (P. laetulus, Figure 2C), while C. cancellatum emerged in May ( Figure 1B). Even if these three carabid species are not active during the time critical for crop damage in AR, slugs and their eggs and, to a lesser degree, caterpillars and cranefly larvae were still present in the fields and predation on these pests might still occur. This can lessen pest pressure the following autumn either indirectly by reducing the number of caterpillars and cranefly larvae metamorphosing into egg laying adults or directly by feeding on slug eggs and juvenile slugs that would otherwise have fed on emerging AR later in the year. The only distinct spatial overlap of carabid and pest abundances was between N. brevicollis and slugs, which were most abundant in the sites with vegetation margins.
Unlike with quadrats or soil sampling, the refuge traps we used in our study do not provide data on the actual density but rather the activity of the monitored organisms. This is dependent on conditions such as temperature, humidity or soil moisture (e.g., [68,69]), which can have different effects on activity depending on vegetation cover in the habitat [70,71]. While refuge traps were also reported to overestimate the abundance of surface-active species such as D. reticulatum [72] and underestimate juveniles [73] in comparison to soil sampling, Clements and Murray [74] found that 'wadding square traps', which are most similar to the refuge traps used in this study, caught more juveniles than saucer traps. The primary advantage of refuge traps over quadrats or soil samples is the comparably low sampling effort, which is crucial for large-scale experiments. Hence, they are frequently used to monitor slugs (e.g., [75][76][77]), but to our knowledge have not been used previously for cranefly larvae and caterpillars. As we were interested mainly in the activity periods of the pests and the differences in relative pest numbers between the sites, refuge traps were the ideal choice for this study.
Nebria brevicollis was the only common carabid in this study whose distribution was significantly associated with a vegetation margin. This species has traditionally been considered a woodland species associated with litter of deciduous trees and high soil moisture [67]. However, more recent publications report it to be rather eurytopic and abundant in a range of different habitats including agricultural grasslands and even urban areas [78,79]. While it occurred in all sites during this study, it was most abundant in those with a vegetation margin. It is likely that the vegetation field margins provide suitable habitat and refuge during harvest and post-harvest operations, which, in AR, also correspond with the timing of summer diapause and subsequent mating in N. brevicollis. However, its high mobility both as adult and larva [80] make it an effective colonizer of nearby fields that might offer suboptimal conditions for breeding or summer diapause. While the model containing the effect 'field margin' was also a significantly better fit than the null model for P. laetulus and A. muelleri, their activity densities in the different fields indicate an association with a gravel rather than a vegetation margin ( Figure 2C,D). Agonum muelleri is described by some authors as a species that prefers open and dry habitats [81], while other authors classify this species as 'hygrophilous' [82] and list damp grasslands or even open woodlands as habitats [79]. No other studies have investigated the habitat preferences of P. laetulus to date and further work is needed to determine the ecological requirements of this species. 'Field margin' or 'soil moisture' did not explain the site preference of C. cancellatum, which is a species that is associated with open and dry country [54]. However, other soil characteristics such as grain size, pH and organic matter content (not measured in this study) could be important for this species as it was completely absent from site C3 (Figure 2B), the only site with slightly gravelly soil. In general, none of the associations with 'field margin' were strong (R 2 < 0.21; Table 3) for any of the common carabid species, indicating that other factors not measured in this study are likely of greater importance.
Tilling can influence carabid beetles either directly by killing adults, larvae or pupae or indirectly, by altering biotic and abiotic habitat characteristics making the environment more or less suitable for different species [27,28]. None of the common carabid species in this study showed a significant response to disk tilling in August following the BACI analysis (Table 4), which is comparable with the findings of other studies. Holland and Reynolds [83] determined that tilling in October or February had no effect on the abundance of emerging N. brevicollis in fields of wheat stubble and undersown grass, and other studies even indicate that tilling might have a favorable impact on the species (summarized in [84]). Hatten et al. [85] state that, while C. cancellatum shows no consistent response to tillage, trap catches of this species were higher in conventional as opposed to no-till fields of peas, wheat and winter wheat. Agonum muelleri activity has been shown to be reduced by moldboard ploughing in July in a previous timothy/red clover field [86], but in Gareau et al. [87] it was significantly more abundant in full October tillage compared to reduced October tillage in cereal rye. While no studies have been carried out involving P. laetulus, related species P. scitulus was more abundant in conventional tillage systems compared to no till in spring wheat [88] and peas, wheat and winter wheat [85].
The large decreases in N. brevicollis numbers in sites E4 and E5 ( Figure 3A) might be due to other effects: soil cultivation in this study took place at a time when N. brevicollis is in summer diapause, which is likely taking place in the field margin-Kromp [27] refers to an unpublished manuscript, that mentions N. brevicollis moving to the permanent boundary strip and woodland edge of the field during soil cultivation and seedbed preparation. If tilling had a direct mortality effect, it should have impacted the number of specimens found in autumn and subsequently also the number of larvae. However, activity density of the autumn 2018 population of N. brevicollis was high in site E4 and the number of larvae trapped the following spring was similar to those of the previous year. It is rather more likely that the larvae, all of which were caught in the edge traps in site E4, were pupating in or close to the field margins and died when the adjacent river burst its banks in early April 2019. Only one larva was caught at site E5 in spring 2018, followed by many adults. The autumn captures for N. brevicollis at this site were low, only two specimens were trapped in late October followed again by only a few larvae caught in winter 2018/spring 2019. This indicates that the species primarily inhabits the field margin at this site and only sporadically ventures into the field. This is supported by the finding that, unlike in site E4 where N. brevicollis was also found in large numbers in the field traps, nearly 90% of the captures at site E5 were from the edge traps. Why significantly fewer adults were captured at this site in the second year is unclear, but conditions might have been more suitable in the adjacent field margin and garden than within the field.
The majority of feeding incidences on gastropods as established by qPCR was accounted for by N. brevicollis, which is plausible given the temporal and spatial overlap between this beetle species and slugs. However, only one specimen tested positive for gastropod DNA during pest emergence in the autumn when slugs would pose the greatest problem. The remainder of the screened beetles that tested positive for Gastropoda were collected in spring/summer, possibly due to the higher abundance of smaller slugs at this time which are much easier to consume than larger adults [3] or due to a lack of more palatable prey. Feeding incidences on slugs were also highest in the two fields where these gastropods were most abundant and in the year in which they were present in especially high numbers, indicating opportunistic feeding events including scavenging rather than targeted predation. The overall feeding incidence was just under 10%, compared with 37% of N. brevicollis testing positive for slug DNA from a study in the United Kingdom [89], but according to the farmers of our study fields, pest pressure by slugs was much lower in the two years of the study compared to previous years, likely owing to the unusually warm and dry autumn weather in 2017 and 2018. Thus, slug numbers in the fields and associated feeding incidences by carabids might generally be higher than those recorded during the current study. Lower feeding rates for gastropods were observed for P. laetulus and A. muelleri, which is likely due to these beetles having lower abundances in the sites with high slug numbers (E4 and E5), and thus fewer encounters.
Nebria brevicollis also tested positive for Tipulidae and Lepidoptera DNA, and the beetles were observed feeding on live mature N. pronuba caterpillars as well as Tipula spp. larvae in the laboratory. Mair and Port [4] reported that N. brevicollis readily fed on Diptera larvae and Seric Jelaska et al. [90] showed that 20% of specimens tested positive when screened for Lepidoptera DNA with PCR, compared to 7.4% in this study. However, when looking solely at the presence of positive feeding events in September and October when early season predation would be at its most beneficial as the predator to pest ratio is largest, it was found that 18.6% of the 59 screened N. brevicollis, tested positive for Lepidoptera. This indicates that caterpillars might be an important food source during the second activity phase of this carabid species. While no species-specific primers were used, N. pronuba was the only abundant caterpillar in the AR fields during the autumn and winter season. This study provides the first feeding record of carabid beetles on this relatively new introduced pest species. Both caterpillars and cranefly larvae were consumed during pest emergence in the autumn, even when they were not found under refuge traps. This indicates that consumption of these two prey groups is probably not just the result of opportunistic feeding, as simple chance encounters with the prey would be less likely.
Lepidoptera were also consumed by C. cancellatum, which is described as 'purely predaceous' [91] and one pair of beetles in captivity was reported to have killed between five and twelve Noctua clandestine larvae a day [92]. In addition, Gidaspow [93] stated that Calosoma spp. could kill approximately 300 caterpillars per season with even their newly hatched larvae being able to overcome caterpillars of considerable size. It is hence surprising that only a few individuals (5.9%) tested positive for Lepidoptera DNA. However, only a few caterpillars were detected in our sites during peak C. cancellatum activity. The same applies for P. laetulus-while this latter species as well as A. muelleri showed some potential for contributing to pest control in AR, this study is the first to investigate their predation on these pest groups; still more studies are needed to elucidate their feeding preferences. A limitation of molecular gut content analysis is that it cannot distinguish between predation, scavenging [94,95] or secondary feeding [96,97]. Additionally, not every positive qPCR result necessarily indicates that the prey was killed. While the latter issue is difficult to resolve, as even an exact quantification of the amount of prey being consumed cannot prove the death of that prey, the ability and/or likelihood of a generalist predator actually feeding on live rather than dead prey can be determined through feeding experiments. We carried out two different laboratory trials (at 21 and 16 • C) using 20 adult N. brevicollis that were starved for seven days, which showed the species neither consuming eggs nor live juvenile D. reticulatum (~0.05 g; Appendix A) while dead slugs were readily eaten on other occasions, even when beetles were not starved (Inga Reich, personal observation). Other laboratory studies on the predation of slugs by N. brevicollis also showed very limited feeding on D. reticulatum, with most being restricted to very small, injured or dead individuals or eggs [3,4,98,99]. However, Ayre [2] found that up to 80% of one-day-old slugs presented to N. brevicollis at an ambient temperature of 8 • C were killed but mortality rate of slugs dropped at higher temperatures. Nebria brevicollis larvae were also found to feed on slugs, with ambient temperature being less important [2] but only one of the 36 larvae tested in the latter study were positive. It is hence questionable whether the positive results herein show actual predation events, even though they mainly occurred at a time where many neonate slugs were present. On the other hand, N. brevicollis was observed to feed on both, live mature cranefly larvae and final stage N. pronuba caterpillars in the laboratory, demonstrating that the positive qPCR results for these pests might constitute actual predation events rather than scavenging or secondary predation. Another limitation of molecular gut content analysis is that digestion will break down the prey DNA in the predator gut as time progresses. Using qPCR technique, slug DNA could still be detected in 42 out of 47 N. brevicollis even after 24 h after feeding on~0.001 g tissue (Appendix B).
The next step in investigating the impact of commonly occurring carabids in AR fields should be to quantify the effect of predation by manipulating pest and predator densities in field cage experiments [100][101][102]. This is essential as Firlej et al. [13] found that despite a high incidence of feeding on soybean aphid Aphis glycines by P. melanarius (up to 33.7% of field-collected carabids screened positive for aphid DNA), a field cage experiment showed very little impact of carabid feeding on A. glycines population growth at moderate infestation levels.

Conclusions
Based on temporal overlap with all three common pest groups and on the results of the molecular gut content analysis in combination with feeding observations, N. brevicollis appears to be the most promising carabid for conservation biological control of cranefly larvae and caterpillars in the surveyed AR fields. Due to its autumn activity phase, this species could exert early season predation, which is critically important as it can delay pest population increase up to a point, after which the crop is less susceptible to damage. A potential problem is that N. brevicollis activity abundance decreases sharply in mid-November ( Figure 1A) when N. pronuba and cranefly larvae are increasing in numbers ( Figure 1F,G) and none of the larvae of N. brevicollis, which are active throughout the winter and early spring, were found to have consumed caterpillars or cranefly larvae.
While feeding incidences were slightly lower and the spatial and temporal overlap with any of the pest taxa was not as distinct, the other common spring-active species P. laetulus and A. muelleri and, to a lesser degree, the summer-active C. cancellatum still have some potential to reduce the populations of the pests at the times critical for crop damage. While none of the summer-active larvae of these three species were captured during this study, and neither A. muelleri or P. laetulus larvae have been reported to feed on any of the studied pest groups, those of C. cancellatum have been found to feed on caterpillars [92] and might contribute to conservation biological control of lepidopterans.
Disk tilling appeared to have no effect on the most common carabid species found in AR in this study. This was probably because these sites were under a no-till regime for only one year and it likely takes longer for carabid assemblages to reflect such reduced disturbance. Of the common carabid species, only the distribution of N. brevicollis was significantly associated with a vegetation field margin. The latter habitat is likely used by the species as a refuge during summer diapause and then as a source from where fields are re-colonized by either the larvae in the winter or the emerging tenerals the following spring. However, we do not recommend managing for this species, which is a relatively new introduction to Oregon because its impact on the native invertebrate fauna is not yet known [78]. The ecology and implications for the native carabid fauna of N. brevicollis in AR in Oregon will be addressed in a subsequent paper.
All raw data from this study that are not found in the Supplementary Materials have been uploaded to a public repository and can be accessed at https://doi.org/10.7267/p2677251h.

Supplementary Materials:
The following are available online at http://www.mdpi.com/2075-4450/11/11/722/s1, Figure S1: Details of the study setup and sampling equipment. Figure S2: Photograph of each field margin. Figure S3: Mean weight loss in grams of 25 g soil after 24 h at 105 • C for each site ± standard deviation on eight occasions between 31 October 2018 and 20 May 2019. Table S1: Field features and treatments used at each site between autumn 2017 and autumn summer 2019. Table S2: Sampling dates for pitfall and refuge traps at each site. Table S3: Plant species recorded in the field margins closest to the first row of traps in sites with a vegetation field margin (C4, C5, E4, E5) and general percentage of ground cover offered by each layer. Table S4: Mean weight loss in grams and standard deviation of all soil samples taken at each site and mean difference in weight loss in grams between sites.  Table S6: Species that were commonly found in the fields during this study and that were tested with the primer pairs GastNLSf1/GastNLSr1, Tip-gen-S267/Tip-gen-A268 and Lep-gen-S274/Lep-gen-A275.

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

Appendix A. Feeding Trials
Nebria brevicollis were collected in the field two weeks prior to the experiment. They were kept in an incubator at 16 • C and a 12/12 light/dark cycle and maintained on frozen Noctua pronuba caterpillars. After being starved for seven days, they were put in a plastic container (height 12 cm, diameter 9 cm) lined with damp tissue paper and were presented with a juvenile D. reticulatum slug (~0.05 g) and/or two eggs from the laboratory colonies. The containers were left at 21 • C (trial 1) or 16 • C (trial 2) and at a 12/12 light/dark cycle for four days. The dead slug (Table A2) showed no signs of injury, and the chorion of the disturbed eggs was not damaged or ruptured (Tables A1 and A2).

Appendix B. DNA Detection
Nebria brevicollis were collected in the field several weeks prior to the experiment. They were kept in an incubator at 16 • C and a 12/12 light/dark cycle and maintained on frozen Noctua pronuba caterpillars. A total of 50 adult N. brevicollis were provided with 0.01 g of slug tissue following a three day starvation period. Beetles were allowed to feed for 30 min and were killed by freezing at −20 • C in groups of five after 0, 4, 8, and 12 h. As three beetles did not consume slug tissue within the 30 min, only three (kept at 21 • C) and four (kept at 5 • C) were killed after 24 h. Even after 24 h post-feeding on slug tissue, DNA could still be detected in 6 out of 7 N. brevicollis. However, DNA detection was not determined for 5 out of 40 beetles in the other time intervals. Two generalized linear models (GLMs) with a binomial distribution were created in R, one for each temperature. The resulting regression equations were Logit (probability of DNA detection at 5 • C) = 2.228 − 0.029 * time and Logit (probability of DNA detection at 21 • C) = 2.314 − 0.045 * time and median DNA detection times were calculated as 76.5 h at 5 • C and 51.5 h at 21 • C. Predicted detection probabilities for longer time periods post-feeding were calculated and plotted ( Figure A1A,B), but the range of values within the 95% confidence interval is large. Table A1. Feeding trial 1 with 20 Nebria brevicollis kept at 21 • C and a 12/12 light/dark cycle and presented with two Deroceras reticulatum eggs each. The feeding trial was not checked on 2 June 2019. Eggs disturbed Eggs disturbed 14 15 16 17 Carabid dead trial ended 18 19 20 Table A2. Feeding trial 2 with 20 Nebria brevicollis kept at 16 • C and a 12/12 light/dark cycle and presented with two Deroceras reticulatum eggs and a juvenile D. reticulatum (~0.05 g) each. The feeding trial was not checked on 12 June 2019. Eggs disturbed 8 9 10 1 slug hatched 11

Appendix B-DNA Detection
Nebria brevicollis were collected in the field several weeks prior to the experiment. They were kept in an incubator at 16 °C and a 12/12 light/dark cycle and maintained on frozen Noctua pronuba caterpillars. A total of 50 adult N. brevicollis were provided with 0.01 g of slug tissue following a three day starvation period. Beetles were allowed to feed for 30 min and were killed by freezing at −20 °C in groups of five after 0, 4, 8, and 12 h. As three beetles did not consume slug tissue within the 30 min, only three (kept at 21 °C) and four (kept at 5 °C) were killed after 24 h. Even after 24 h post-feeding on slug tissue, DNA could still be detected in 6 out of 7 N. brevicollis. However, DNA detection was not determined for 5 out of 40 beetles in the other time intervals. Two generalized linear models (GLMs) with a binomial distribution were created in R, one for each temperature. The resulting regression equations were Logit (probability of DNA detection at 5° C) = 2.228 − 0.029 * time and Logit (probability of DNA detection at 21 °C) = 2.314 − 0.045 * time and median DNA detection times were calculated as 76.5 h at 5 °C and 51.5 h at 21 °C. Predicted detection probabilities for longer time periods post-feeding were calculated and plotted ( Figure A1A,B), but the range of values within the 95% confidence interval is large.