Non-destructive environmental safety assessment of threatened and endangered plants in weed biological control

Insects and plants

We received naïve overwintering M. borraginis from a rearing colony at CABI in Switzerland typically in early May of each year of this 4-year study. Because female M. borraginis must feed on flowers of C. officinale to initiate oogenesis, we provided fresh cymes of C. officinale to weevils and kept them in an environmental chamber (E-30B, Percival Scientific, Perry, IA, USA; L18: D6 at 20 °C and 60% relative humidity) at the University of Idaho weed biological control quarantine laboratory. Since T&E plant species under the Endangered Species Act were included on the test plant list for the pre-release risk assessment of M. borraginis, four T&E and one imperiled plant species in Boraginaceae were selected in the United States using royalty-free map software (www.mapsfordesign.com) (Fig. 1). Seeds of Amsinckia grandiflora (Kleeb. Ex A. Gray) Kleeb. Ex Greene (n = 100), Plagiobothrys strictus (Greene) I.M. Johnst. (n = 100), Plagiobothrys hirtus (Greene) I.M. Johnst. (n = 100), and Hackelia venusta (Piper) H. St. John (n = 50) were received from collaborators (see the acknowledgements). They were sown in Sunshine Mix #2 (Sun Gro Horticulture Canada Ltd., Vancouver, Canada) at 4 °C for 12 weeks. Another T&E plant species, Oreocarya crassipes (I. M. Johnst.) Hasenstab & M.G. Simpson, was excluded because it is adapted to an arid environment that both C. officinale and M. borraginis could not tolerate. Rootstocks of Dasynotus daubenmirei I.M. Johnston (n = 10) were collected at Walde Lookout, ID, USA (N 46.23528°, W 115.63528°). Although D. daubenmirei is not a federally listed T&E species, it is the single species in its genus and grows exclusively in one population in north central Idaho (Cohen, 2014). Rootstocks of C. officinale (n = 20) were collected at the Idler’s Rest Nature Preserve, Moscow, ID, USA (N 46.804160°, W 116.948554°). We transplanted them in 11.3 L black plastic pots filled with Sunshine Mix #2 (Sun Gro Horticulture, Agawam, MA, USA) and placed them in an environmentally-controlled greenhouse at the University of Idaho, Moscow, ID, USA in March of the following growing season. Plants were watered as needed. Because many species in Boraginaceae bloom along cymes containing buds, open inflorescences and young fruits, all of which are used by weevils as food and oviposition resource, we used plants bearing cymes for all experiments described below.

For Amsinckia grandiflora, D. daubenmirei and P. hirtus, single-choice oviposition test results were available (M. Schwarzländer, H.L. Hinz, R.L. Winston, 2020, unpublished data), and these species were used as “prove of concept” to study behavioral responses of a biological control agent to olfactory and visual plant cues in lieu of traditional host-specificity tests with potted plants. The assumption was that results of traditional tests would be confirmed and a physiological explanation for results could be provided. For H. venusta and P. strictus no propagules could be obtained. All plant cues for these two plant species were collected in the field and used in behavioral bioassays.

Collecting headspace volatile organic compounds (VOCs) and flowering stems

Polyvinyl acetate (PVA) bags (12 cm × 24 cm; Reynolds, Richmond, VA, USA) were purged of volatile contaminants in a drying oven at 140 °C for 60 min (Park et al., 2018). A flowering stem of the target weed and each test plant species was covered with the PVA bag and sealed with purged cotton balls and a cable tie to minimize potential physical damage to plants. Data were collected as previously described in Park et al. (2018). Each elutant was placed in a screw cap vial and stored at 4 °C until further use. We collected VOCs of C. officinale, H. venusta and P. hirtus at their natural habitats, while VOCs from Amsinckia grandiflora, D. daubenmirei, and P. strictus were collected from plants grown in a greenhouse. In addition, permits were obtained to cut individual flowering stems (5 cm) of the respective three plant individuals per species and placed them into 10 cm transparent aqua-tubes (31-01931-case; Oasis Floral Products, Kent, OH, USA) and stored in a portable cooler to be used as visual cues in bioassays (Fig. 2). Further, to examining the influence of floral size and color on the host recognition by M. borraginis, we measured the size of petals and stamens of each plant species (n = 20, except for P. strictus, n = 10) using a digital caliper.

Host-finding bioassays with double-stacked y-tube device (D-SYD)

A double-stacked y-tube device (D-SYD; diameter: 2 cm, length of the arms: 12 cm) was used to assess quantitatively the host selection behavior of M. borraginis based on olfactory (VOCs) and visual cues (a flowering stem) either individually or simultaneously (Park et al., 2019). Olfactory cues were offered to the insect through a pull pump in a closed system. Because olfactory and visual cues were separated by glass, there was no possibility that visual cues emitting VOCs could have contaminated the bioassays. The D-SYD was installed in a darkened room with a full spectrum light bulb (ES5M827FS; 27 Watts, Home Depot, Atlanta, GA, USA; 350 nm to 850 nm wavelength) diffused through a white polyethylene dome (40 cm × 30 cm × 20 cm) placed 20 cm above the D-SYD. The device was rinsed with 70% ethanol following each bioassay and rotated 180° after every five replicates to eliminate left or right-handed bias (Park et al., 2018). Twenty female weevils were used in each of the four experimental set ups per plant species (total of 80 female weevils for each plant species). Each female was only used once, and each individual weevil was treated as one replicate. For P. strictus, only ten females were used in each bioassay due to the limited number of M. borraginis available. In each replicate, the female was placed at the release point in the bottom Y-tube in each bioassay and observed for up to 5 min. A weevil was considered to have made a choice if it passed a decision line that was 3 cm into one of the arms within 5 min (Tooker, Crumrin & Hanks, 2005).

A total of four bioassays were conducted. Before conducting abovementioned experiments, neither visual nor olfactory cues were placed in the D-SYD and 30 female weevils tested, to confirm no bias exists (experiment 1). The next two experiments examined the role of a single cue of a T&E plant species and C. officinale. For the visual cue (experiment 2), 5 cm of a flowering stem was placed in the D-SYD; to test the effect of olfactory cues (experiment 3), 1 μl of headspace VOCs, which was eluted directly from the identical plants as in experiment 2, was placed in the D-SYD. The other two experiments evaluated the combined effect of visual and olfactory cues; the only difference was the presence (experiment 4) or absence (experiment 5) of C. officinale in addition to a plant species in the D-SYD.

GC-MS and GC-EAD/FID analysis

Gas chromatography-mass spectrometry (GC-MS) was performed to check the identity of electrophysiologically active chemical compounds and quantifying their concentration in floral scents. An Agilent 7890A (Agilent Technologies Inc., Santa Clara, CA, USA), with an HP-5MS column (30 m × 250 μm × 0.25 μm; Agilent Technologies Inc., Santa Clara, CA, USA), was coupled with a Hewlett Packard 5973 mass selective detector (Agilent Technologies Inc., Santa Clara, CA, USA). A total of 10 ng of nonyl acetate (W278807, Sigma Aldrich) was injected with each elutant (1 μl) as an internal standard in the splitless mode, injector temperature 250 °C. Column specification and the temperature program were collected as previously described in Park et al. (2019). To confirm the chirality of an enantiomeric compound, α-copaene, the floral blends were analyzed using an Agilent J&W Cyclodex-B column (Agilent Technologies Inc., Santa Clara, CA, USA). Retention time of the α-copaene was compared with the compound’s retention times in two essential oils (Young Living Essential Oils, Lehi, UT): Angelica in which (−) predominates and Copaiba in which (+) predominates.

Gas chromatography-electroantennographic detection and flame ionization detection (GC-EAD/FID) was performed to detect electrophysiologically active VOCs from C. officinale. If all female weevils (n = 6) responded to a VOC in the blend of C. officinale, the specific compound was considered as the electrophysiologically active VOC. An Agilent-6890N (Agilent Technologies, Santa Clara, CA) was equipped with an identical column specification and using the same temperature program as described in Park et al. (2019). A 1:1 column splitter delivered effluent to the FID detector and to the EAD. Effluent was delivered via a GC effluent conditioner (Syntech, Hilversum, Netherlands) into humidified air flowing at 10 ml/sec directed through a glass tube to the antenna of a female weevil. Depolarization of the antenna was recorded and plotted with the FID signal using Syntech GC-EAD software.

To prepare antennae for recording, female M. borraginis were decapitated using a scalpel under a microscope. The decapitated head was placed on a ground probe with Spectra 360 electrode gel (Parker Laboratories, Fairfield, NJ, USA). The distal tip of the antenna was punctured with a minute insect pin (1208SA; Bioquip, Czech Republic), which was placed in contact with the recording probe (Ockenfels Syntech GmbH, Buchenbach, Germany). The undamaged antenna from the head was positioned to receive the entrained effluent from the GC column. The performance of the system was checked before each recording using an antenna simulator (Ockenfels Syntech GmbH, Buchenbach, Germany).

Statistical analysis

We used a generalized linear model with an expected null ratio of 50:50 and a binominal distribution to analyze the number of weevils in the either side of testing device in each assay assuming a completely random design. We compared the least square means of responses to test whether behavioral responses of M. borraginis differed among control (experiment 1), single modalities (experiments 2 & 3), and combined cues (experiments 4 & 5) that compared the effect of absence/presence of C. officinale on weevil response in the two cue bioassays. We used a generalized linear model with a Poisson distribution to analyze the decision time (i.e., the time delay from the initial movement of a female to passing the decision line) in each assay. Least squares mean differences were calculated to compare the decision time of females between arms of the D-SYD in each behavioral assay. Data from tests on all plants were pooled for analysis because there was no evidence of effects of individual plants on the weevil responses (Park et al., 2018). We performed two principal component analyses (PCA) based on the entire volatile profile of plants and the presence of electrophysiologically active VOCs identified from C. officinale. We compared the size of stamens and corollas among plant species with a one-way ANOVA. All analyses were carried out using SAS 9.4 (SAS Institute, Cary, NC, USA).

Experiment 1: Empty (control) arms of D-SYD

The number of female weevils selecting either side of the empty D-SYD did not differ in all bioassays: (A. grandiflora: Z = 0, P = 1; D. daubenmirei: Z = 0.45, P = 0.65; H. venusta: Z = 0, P = 1; P. hirtus: Z = 0.45, P = 0.65; P. strictus: Z = 0.63, P = 0.52; Figs. 3A–3E, top bars in each panel). However, the searching time of the weevils in either side of the D-SYD differed for some experiments (Figs. 3F–3J, first bar from top). Time spent by female weevils on one side of the D-SYD was longer for A. grandiflora (Z = 3.69, P < 0.001; Fig. 3F) and D. daubenmirei (Z = 4.52, P < 0.0001; Fig. 3G), but not different for H. venusta (Z = 0.74, P = 0.45), P. hirtus (Z = −0.19, P = 0.85), and P. strictus (Z = 0.43, P = 0.66).

Experiment 2: Visual cues

Female weevils preferred flowering stems of C. officinale over all plant species in this study (A. grandiflora: Z = 2.87, P < 0.01; D. daubenmirei: Z = 2.77, P < 0.01; H. venusta: Z = 2.48, P < 0.05; P. hirtus: Z = 2.95, P < 0.01; P. strictus: Z = 0, P = 1; Figs. 3A–3E, second bars from top). The decision time of female weevils did not differ for P. hirtus (Z = 0.08, P = 0.93), but was longer for D. daubenmirei (Z = 2.00, P = 0.04) and H. venusta (Z = 2.24, P = 0.02) when compared with C. officinale (Figs. 3F–3J, second bars from top). Only one female weevil chose A. grandiflora and P. strictus over C. officinale, which did not allow for statistical inferences of the decision time.

Experiment 3: Olfactory cues

Female weevils preferred VOCs of C. officinale to all T&E plant species (A. grandiflora: Z = 2.48, P < 0.05; D. daubenmirei: Z = −2.95, P < 0.01; H. venusta: Z = 2.77, P < 0.01; P. hirtus: Z = 2.87, P < 0.01; P. strictus: Z = 0, P = 1; Figs. 3A–3E, third bars from top). Compared to C. officinale, the decision time of females did not differ for H. venusta (Z = 1.33, P = 0.18), was longer for A. grandiflora (Z = 2.86, P < 0.01) and was shorter for D. daubenmirei (Z = 3.90, P < 0.0001) (Figs. 3F–3H, third bars from top). The decision time for P. hirtus and P. strictus could not be statistically analyzed because only one weevil chose them over C. officinale (Figs. 3I–3J, third bars from top).

Experiment 4: Combined visual and olfactory cues

When combining both cues, M. borraginis female strongly preferred C. officinale over all T&E plant species (A. grandiflora: Z = 0, P = 1; D. daubenmirei: Z = 0, P = 1; H. venusta: Z = 0, P = 1; P. hirtus: Z = 0, P = 1; P. strictus: Z = 2.08, P = 0.03; Figs. 3A–3E, fourth bars from top). Since no weevils chose any three T&E and one imperiled plant species, except one M. borraginis that chose P. strictus, we were unable to compare decision time between C. officinale and plant species (Figs. 3F–3J, fourth bars from top). Behavioral responses of M. borraginis did not differ when comparing response to a single plant cue with those to combined plant cues for three T&E plant species: A. grandiflora (χ² = 3.28, P = 0.07), P. hirtus (χ² = 2.34, P = 0.12), and P. strictus (χ² = 2.27, P = 0.13) (Figs. 3A–3E, brackets to right of bars). The behavioral responses of M. borraginis were stronger for combined cues than a single cue for D. daubenmirei (χ² = 4.18, P = 0.04) and H. venusta (χ² = 6.06, P < 0.05) (Figs. 3A–3E, brackets to the right).

Experiment 5: Combined cues vs. control

Female weevils strongly preferred the empty arm of the D-SYD to the combined visual and olfactory cues of all four T&E plant species and one imperiled plant species (A. grandiflora: Z = −2.95, P < 0.01; D. daubenmirei: Z = 0, P = 1; H. venusta: Z = 2.77, P < 0.01; P. hirtus: Z = 0, P = 1; P. strictus: Z = 2.08, P = 0.03; Figs. 3A–3E, bottom bars). The strength of this preference for combined cues was similar when the opposing arm of the D-SYD presented combined cues from C. officinale (Figs. 3A–3E; this comparison could not be made for P. hirtus, and D. daubenmirei because 100% of females chose C. officinale). The decision time of females did not differ for A. grandiflora (Z = −1.02, P = 0.30) and H. venusta (Z = −1.65, P = 0.09) (Figs. 3F–3J, bottom bars).

Volatile profiles of plant species

Sixty-one volatile compounds were identified in the floral scents of the six plant species included in this study. Of these, ten volatile compounds detected in the floral scent of C. officinale were electrophysiologically active based on EAD responses of M. borraginis females. Among them, two sesquiterpenes, (−)-α-copaene and (E)-β-farnesene, were unique to C. officinale when compared to the five T&E plant species. A PCA for VOC similarity among plant species explained 75.28% of variation (PC1: 41.57%, PC2: 33.72%) based on the ten electrophysiologically active VOCs found in C. officinale (Fig. 4A). When based on all 61 VOCs identified among the six plant species, the PCA explained 45.24% of the variation of the VOC similarity (PC1: 23.22%, PC2: 22.02%, Fig. 4B).