First report and integrated analysis of two native Trissolcus species utilizing Bagrada hilaris eggs in California

Surveys with sentinel eggs of Bagrada hilaris (Hemiptera: Pentatomidae) in southern California retrieved two parasitoids that were not previously known to be associated with this stink bug, Trissolcus hullensis and T. utahensis (Hymenoptera: Scelionidae). Molecular and morphological analysis of these specimens is used to modify the concept of T. utahensis and assess the factors that contribute to intraspecific variation. We provide an updated couplet to separate T. utahensis from a morphologically similar species, T. cosmopeplae.

Chemical applications have been the main approach used against this pest (Palumbo 2015) and efforts to establish a biological control program were initiated in 2014. Three egg parasitoids of B. hilaris were collected in Pakistan (Mahmood et al. 2015) and brought to the United States to be evaluated as biological control candidates: Trissolcus hyalinipennis Rajmohana & Narendran (Rajmohana 2006), Gryon gonikopalense Sharma (Platygastroidea: Scelionidae) (Sharma 1982), and Ooencyrtus mirus Triapitsyn & Power (Chalcidoidea: Encyrtidae) (Triapitsyn et al. 2020). Laboratory evaluations of these parasitoids are still ongoing, and no release permits have been issued to date (Sforza et al. 2017;Martel et al. 2019;Power et al. 2020a, b). Meanwhile, monthly surveys using B. hilaris sentinel eggs have been conducted in California since the fall of 2017 to identify potential native or introduced parasitoids. Through these surveys, two scelionid species, T. hyalinipennis and Trissolcus basalis (Wollaston) (Ganjisaffar et al. 2018), and a new encyrtid species, Ooencyrtus lucidus Triapitsyn & Ganjisaffar (Triapitsyn et al. 2020), have been identified. This study reports the discovery and identification of two additional scelionid species that parasitized B. hilaris sentinel eggs in southern California: Trissolcus hullensis (Harrington) and Trissolcus utahensis (Ashmead).
Our analysis includes T. utahensis reared from sentinel eggs of Podisus maculiventris (Say) (Hemiptera: Pentatomidae) in British Columbia, Canada, to assess the limits of this species and determine its intraspecific variation using molecular and morphological data. Previous identification of the Canadian specimens brought attention to problems with the couplet that separates T. utahensis and Trissolcus cosmopeplae (Gahan) in the key to Nearctic Trissolcus of Talamas et al. (2015). This key has been used in annual workshops to identify stink bug egg parasitoids throughout North America, during which this couplet has proven to be problematic as well. We here employ molecular phylogenetics to determine which characters are variable, and which are sufficiently stable to be used for identification in these species, and we provide an updated couplet for T. utahensis and T. cosmopeplae.

Survey locations
The Agricultural Operations of the University of California, Riverside, was the main site for the surveys. Fields that had been planted for various research were used for our sentinel egg deployments. A mixed vegetable field available from October 2017 to March 2018, an alfalfa field (Medicago sativa L.) available from October 2017 to January 2019, and a squash field (Cucurbita moschata L., variety black futsu) available from January 2018 to September 2019 were used. The mixed vegetable field was selected particularly because of its two rows of broccoli (Brassica oleracea L., variety Italica), a favorable host for B. hilaris. The alfalfa field hosted several stink bug species during spring and summer, serving as a potential source of stink bug egg parasitoids. The squash field was selected because B. hilaris adults were found on shortpod mustard weeds, Hirschfeldia incana (L.) Lagr.-Foss., within the field. Surveys also were conducted at other locations on the Agricultural Op erations property where mustard weeds were found. In addition, various locations in the urban area with brassicaceous weeds surrounding the University of California, Riverside, were surveyed ( Fig. 1). Previous monthly samplings from January to December 2011 by Reed et al. (2013) had shown that the primary weeds supporting B. hilaris in this area were London rocket (Sisymbrium irio L.), shepherd's purse (Capsella bursa-pastoris (L.) Medik.), and shortpod mustard. Bagrada hilaris also were found occasionally on Russian thistle (Salsola australis R.Br.) and telegraph weed (Heterotheca grandiflora Nutt.) when they were in close proximity to senesced shortpod The black rectangle shows the Agricultural Operations of the University of California, Riverside, where most of our surveys were conducted. Five Trissolcus hullensis were recovered from an alfalfa field (33.96508°N, 117.34084°W), one Trissolcus utahensis was recovered from a squash field with mustard weeds (33.96611°N, 117.34230°W), and eleven T. utahensis were recovered from roadside mustard weeds (33.99105°N, 117.33360°W). mustard. According to Reed et al. (2013), these weeds were available following fall and winter rains through June and the peak B. hilaris abundance was during the spring on these host plants. We used the same locations for our sentinel egg deployments, and surveys were conducted monthly from 21 October 2017 through 27 September 2019.

Sentinel egg card preparation and parasitoid recovery
Bagrada hilaris eggs (≤ 24 hours old) were used for the surveys. Adult mating pairs of greenhouse-grown B. hilaris (Reed et al. 2017) were maintained on organic broccoli florets in plastic containers (15 cm diameter × 6.5 cm height) with 2 screen openings for air circulation in an insectary room at 30 ± 1 °C, 40-50% humidity and 14:10 (L:D) photoperiod. White paper towels were cut in circles to fit the bottom of each container to provide a substrate for oviposition. Approximately 15 mating pairs were placed into each container. Bagrada hilaris eggs were collected daily and glued (Gorilla Super Glue Gel, The Gorilla Glue Co., Ohio, USA) on a 3 × 5 grid of squares on a weatherproof card so that each card contained 15 eggs (Ganjisaffar et al. 2018). For each location, a sentinel card was taped to the wire of a landscape flag, positioning it about 30 cm from the ground. Previous studies showed that eggs placed on cards on the soil surface were eaten by predators (Ganjisaffar et al. 2018). The number of cards used for each survey date varied from 10 to 20, depending on the availability of B. hilaris eggs. Cards were left in the field for 3-5 days (3.7 days on average) to avoid B. hilaris eggs from hatching in the field. According to Reed et al. (2017), B. hilaris eggs hatch after 5 days at temperatures approximating the warmest temperatures on the days of the field surveys. Once collected, the grid was cut and placed in glass vials plugged with cotton. The vials were maintained in the same insectary room that was used for the B. hilaris colony and were examined for parasitism and wasp emergence. Emerged wasps were transferred to vials containing 95% ethanol for identification.

Molecular analysis
Genomic DNA was non-destructively isolated from the entire specimen using the Qiagen DNeasy Blood and Tissue kit (Hilden, Germany) as described in Sabbatini Peverieri et al. (2018). A comprehensive list of all samples extracted with author and year, host, and locality data is given in Table 1 in Supporting Information. The barcode region of the mitochondrial Cytochrome Oxidase Subunit I (COI) was amplified using the universal barcoding primer set LCO1490/HCO2198 (Folmer et al. 1994) (Table 1). The primer set LCO1490puc (Cruaud et al. 2009)/C1-N-2353 (Simon et al. 2006) was used in six samples to amplify a longer region (~850 bp) than the classical barcode region (~710 bp), providing more than enough coverage for the barcode that is not always obtained when using universal primer sets (Table 1). All PCRs were performed as described in Ganjisaffar et al. (2018), except the PCR conditions. The thermocycling conditions were as follows: 1 cycle of denaturation at 94 °C for 3 min, 35 cycles at 94 °C for 30 s, 52 °C (LCO1490/HCO2198) or 50 °C (LCO1490puc/C1-N-2353) for 30 s, 72 °C for 1 min with a final extension step of 10 min at 72 °C. All samples were sequenced bidirectionally using the BigDye chemistry by Genoscreen (Lille, France) or the ABI SeqStudio Platform at FDACS-DPI in Florida (USA). A COI consensus sequence was established for each specimen. All sequences generated from this study are deposited in GenBank, and all residual DNAs are archived at EBCL or FSCA (Table 1). Voucher specimens which have been reexamined following the molecular analysis are presently archived at FSCA (Table 1). All sequences were translated into amino acids to check for stop codons and frame shifts. All sequences obtained were compared with sequences present in GenBank using the Basic Local Alignment Search Tool (http://www. ncbi.nlm.nih.gov/BLASTn). BOLD identification engine (Ratnasingham and Herbert 2007) was similarly datamined for barcodes of Trissolcus species and evaluated for barcode identification success. The 28 sequences obtained in this study were aligned with 28 barcode sequences of Trissolcus retrieved from GenBank. The final alignment of 56 sequences was performed using the default settings of CLUSTAL W (Thompson et al. 1994) as implemented in MEGA X (Kumar et al. 2018) and resulted in 576 characters with 174 parsimony informative sites. The phylogenetic relationships among specimens were reconstructed following a Bayesian analysis as implemented in MrBayes v. 3.2 (Ronquist et al. 2012). Searches were run for 1 million generations, in two independent runs, using default priors and the GTR+I+G substitution model that was selected using the Bayesian information criterion (BIC) in MEGA X. One sequence of Trissolcus thyantae Ashmead (GenBank MN615574.1) was specified as the outgroup based on the results of Talamas et al. (2019). Because a network approach is well adapted to infer intraspecific genealogical relationships, a haplotype network was built for T. utahensis using TCS 1.21 (Clement et al. 2000). As the resulting clades were not connected under 90% statistical parsimony limits, we reran the TCS analysis by fixing connection limits at 50 steps. To estimate the divergence within and between terminal taxa and clades, we calculated the uncorrected p-distance using MEGA X, since the generally used K2P distance (Kimura 1980) could be inappropriate when applied to closely related taxa (Srivathsan and Meier 2012).

Morphology
Terminology follows that of Talamas et al. (2017). Following non-destructive DNA extraction, six specimens used in the molecular analysis were photographed (1 T. hullensis and 7 T. utahensis) to document morphology of specimens from different haplogroups and those reared from different hosts. Images were produced with a Macropod imaging system. Image stacks were rendered with Helicon Focus and further processed in Adobe Photoshop CS6.

Field surveys
Our survey period of March 6-9, 2018, in the alfalfa field, yielded a sentinel card with 6 parasitized eggs from which 5 specimens of T. hullensis and 1 specimen of T. basalis wasps emerged. A month later during our April 6-9, 2018, survey in the roadside mustard weeds (33.99105N, 117.33360W), two sentinel cards were parasitized; one of them had 5, and the other one had 7 parasitized eggs, which yielded 11 T. utahensis. One T. utahensis wasp was recovered from a sentinel card that was deployed in the squash field with mustard weeds on October 12-16, 2018.

Molecular analysis
The phylogenetic analysis based on the COI barcode data revealed a relatively wellresolved and supported topology identifying six terminal taxa (Fig. 2). The deepest node corresponds to the split between T. utahensis and the other species, including T. hullensis. The T. hullensis clade comprised two specimens reared from B. hilaris eggs and one specimen reared from frozen sentinel eggs of Halyomorpha halys (Stål) (Hemiptera: Pentatomidae) in Napa County, California. The T. utahensis cluster contains four main clades. The TCS analysis yielded four haplotype sub-networks, which were not connected under the 90% statistical parsimony limits, corresponding to these four clades of which clade 2 and clade 3 are each represented by a single haplotype, and clade 1 and clade 4 by three haplotypes each (Fig. 3). Clades 1, 3 and 4 contained specimens from British Columbia that were reared from P. maculiventris eggs. Clade 1 contained specimens from Canada and southern California, reared from the eggs of P. maculiventris and B. hilaris, respectively, indicating that they are conspecific. Interestingly, clade 2 included only Californian specimens reared from B. hilaris. Genetic distance between all clades ranged from 2.8% (clade 1/clade 2) to 9.9% (clade 1/clade 4) but remained lower than the range (from 10.6% to 14.3%) of interspecific distances obtained with the five other species in our analysis (Table 2).

Trissolcus hullensis (Harrington)
Remarks. The identification of Trissolcus hullensis is straightforward using the characters presented in Johnson (1985) and repeated in Talamas et al. (2015). Characters of the posterior mesosoma are particularly useful for determining this species: In T. hullensis the propodeum and metanotum are directly adjacent between the metapostnotum and metascutellum (Fig. 4), whereas in other species of Nearctic Trissolcus, the metapostnotum extends medially toward the metascutellum, and separates the propodeum from the metanotum (Fig. 5).

Trissolcus utahensis (Ashmead)
Remarks. The concept of Trissolcus utahensis was most recently treated in a revisionary context by Johnson (1985). This treatment separated T. utahensis from T. cosmopeplae Figure 3. TCS COI haplotype network for the four clades of T. utahensis by fixing connection limits at 50 steps. Each haplotype is represented by a colored circle. Lines represent one mutational step between haplotypes, and dark circles represent unsampled haplotypes inferred from the data. Interrupted lines were used when haplotypes were separated by a long branch of more than 7 mutation steps.
based on the length of the anteroventral extension of the metapleuron, the absence of a genal carina, the shape of the gena in lateral view, and if notauli could be distinguished from the surface sculpture of the posterior mesoscutum. These characters were used again in the key to Nearctic Trissolcus by Talamas et al. (2015), with the addition of the form of the mesoscutal humeral sulcus, which was treated as variable within T. cosmopeplae. Talamas et al. (2015) also treated the anteroventral extension of the metapleuron as variable within T. cosmopeplae and emphasized the shape of the gena in lateral view to separate these species. This modification to the key sought to reconcile variability in the shape of the gena with other, seemingly variable characters. The shape of the gena has proven to be one of the more difficult characters to interpret because there is not a discrete boundary between "narrow" and "bulging". Because the variation in these characters does not correspond to clades in our phylogeny, we treat them as intraspecifically variable and the T. utahensis clade as a single species (Fig. 2). Based on the morphological analysis provided below we propose the following replacement for couplet 14 in Talamas et al. (2015):

T. utahensis
Clade 4 11.9 ± 1.2 12.9 ± 1.3 12.6 ± 1.3 14.7 ± 1.5 11.9 ± 1.3 9.97 ± 1.1 8.65 ± 1.1 9.61 ± 1.2 0.4 ± 0.1 antennal scrobe. Fig. 8 illustrates a specimen that emerged from a B. hilaris egg. As was found in T. basalis, specimens that developed in B. hilaris eggs have reduced sculpture relative to those that developed in other hosts (Ganjisaffar et al. 2018). The specimens in Figs 9, 11 were both reared from the eggs of P. maculiventris in British Columbia, and have identical COI barcode sequences, yet the size of smooth area on the dorsal frons differs between them. The specimen in Fig. 11 is the largest (1.35 mm) among these, and the specimen in Fig. 8 is the smallest (0.93 mm). The specimens in Figs 9, 10 have the smooth area on the frons about equal in size and these specimens are also approximately equal in length (1.11 and 1.07 mm, respectively). These two specimens were retrieved in different haplogroups (clades 1 and 4), and we thus postulate that sculptural differences on the frons are size dependent. Variation on the gena. The shape of the gena varies between and within the four clades of T. utahensis. In clades 1 and 3, the specimens have a rather narrow gena, and   in clades 2 and 4 the gena is moderately to distinctly bulging in lateral view. Figs 12-15, 24 and 27 illustrate this variation. The degree to which the gena is bulging in lateral view does not appear to be host or size related. Specimens reared from B. hilaris eggs are the smallest and have the gena moderately (Fig. 12) to distinctly (Fig. 24) bulging. The specimens with the most distinctly bulging gena (clade 4,Figs 15,19) and the narrowest gena (clade 3, Figs 27, 28) were both reared from eggs of P. maculiventris and the specimens are larger than those reared from B. hilaris eggs.
Specimens with a bulging gena tend not to have the genal carina indicated, whereas specimens with a narrow gena often have it clearly expressed, but this is not an exact correlation. The specimen in Figs 12, 16 has a moderately bulging gena and the genal carina is distinctly present.
Microsculpture on the poster gena is less developed in specimens reared from B. hilaris eggs (Fig. 12), and this area is noticeably smoother than in specimens reared from P. maculiventris eggs (Figs 13-15). This phenomenon is consistent with a general pattern of reduced sculpture in smaller specimens.
Occipital carina. Tortorici et al. (2019) presented a new character to distinguish T. semistriatus (Nees von Esenbeck) from closely related species: the form of the occipital carina in dorsal view. In most species of Trissolcus the occipital carina is evenly convex, but in a few species, including T. semistriatus and T. utahensis, the occipital carina forms a distinct angle and the vertex of this angle may have a short carina directed toward the median ocellus (Figs 5,22). In some specimens that emerged from B. hilaris eggs, the vertex of the angle formed by the occipital carina is less sharp, perhaps due to its diminutive size (Fig. 25). The occipital carina is not visible in the available images of T. cosmopeplae, and it is obscured by glue in the holotype specimen of T. utahensis. This character deserves further attention in Nearctic Trissolcus although we are not presently able to determine if it can separate T. cosmopeplae and T. utahensis.
Notaulus. Specimens of T. utahensis reared from both B. hilaris and P. maculiventris have the notaulus indicated by short, shallow grooves present at the posterior margin of the mesoscutum (Figs 5,(20)(21)(22). These are visible in the holotype of T. utahensis (see fig. 100 in Talamas et al. (2015)), and they have the same form in the holotype of T. cosmopeplae (Fig. 7). Based on re-examination of images of the types and the specimens at hand, we conclude that this character does not separate these species.
Mesoscutal humeral sulcus. The form of the mesoscutal humeral sulcus was used by Tortorici et al. (2019) to separate very similar Palearctic Trissolcus species. This sulcus is clearly indicated by cells in the holotype of T. cosmopeplae (Figs 6, 7) and it is present as a smooth furrow in all specimens of T. utahensis that we have examined (Figs 5,(20)(21), including the holotype.
Anteroventral extension of the metapleuron. The length of this structure, reaching to the mesocoxa in T. cosmopeplae (Fig. 6) and very short in T. utahensis (Figs 12-19 was the first character listed in the couplet that separates these species in Johnson (1985). Our analysis of specimens in this study lends weight to the reliability of this character.
Color. Most species in Trissolcus have a black metasoma. The most notable exception is a Palearctic species, T. rufiventris Mayr, in which T2-T7 vary from bright yellow to dark brown. The specimens reared from B. hilaris eggs have T2-T7 notably lighter in color than the head and mesosoma (Figs 20,25), and the body overall is lighter in color than in specimens reared from P. maculiventris eggs (compare Fig. 8 to Figs 9-11, Fig. 12 to Figs 13-15).
The most obvious color variation in T. utahensis is in the legs. Specimens that parasitized B. hilaris eggs have legs that are pale brown to orange distal to the coxae (clades 1, 2; Figs 12, 24). Brightly colored legs were found in T. utahensis that emerged from P. maculiventris eggs (clade 3, Fig. 29), although the clear majority of specimens reared from P. maculiventris eggs in British Columbia had dark brown legs. Interestingly, FSCA 00033239, with pale brown legs, in clade 1 is sister to FSCA 00000320, with dark brown legs. The antennae vary in color in accordance with the legs, although the variation is less pronounced and ranges from medium to dark brown. These data do not indicate that T. utahensis exhibits a direct correlation of appendage color with host, size, or lineage.

Discussion
The reports of this study indicate that at least four species of Trissolcus (T. basalis, T. hyalinipennis, T. hullensis, and T. utahensis) are actively parasitizing B. hilaris eggs in southern California (Ganjisaffar et al. 2018). Three other scelionids have been reported parasitizing B. hilaris eggs in Mexico: Telenomus podisi (Ashmead), Gryon myrmecophilum (Ashmead), and Idris elba Talamas (Felipe-Victoriano et al. 2019, Lomeli-Flores et al. 2019). This suggests that B. hilaris eggs may be broadly suitable for Nearctic parasitoids. We conducted surveys in the same locations as Reed et al. (2013) did monthly samplings of B. hilaris in 2011, and the fact that we did not find any B. hilaris in those areas during our two-year survey, suggests that populations of B. hilaris have steadily declined in recent years, and it is likely that parasitoids are part of the cause.
Trissolcus utahensis exhibits a striking degree of variation in COI among specimens from only a few localities. An expansion of this analysis with additional samples is certain to expand our understanding of genetic diversity in this species. It is worth noting that specimens FSCA 00094713-00094715 emerged from the same egg mass and were retrieved in different clades (1 and 3), indicating that interbreeding opportunities exist between these maternally definable populations. Future work should include nuclear genes to provide a broader view of population dynamics and genetic diversity in this species.
The systematics of Trissolcus is undergoing perpetual improvement as it continues to receive attention for the species that attack the eggs of economically important stink bugs. This study demonstrates how a seemingly routine activity of rearing and identifying specimens can require a multifaceted research endeavor to reach a satisfactory answer while also providing new lines of inquiry. For example, our molecular analysis retrieved T. colemani in two clades, one comprised of specimens from East Asia and one from Europe and the Middle East. This study also demonstrates how ongoing parasitoid surveys continue to be productive by providing fresh specimens with host association data. These specimens and data are instrumental for an integrated approach to systematics in which morphological, molecular, and behavioral data are combined to provide robust and holistic species concepts.