Introduction

Assessing the environmental safety of weed biological control agent (BCA hereafter) candidate species is a central aspect of classical biological weed control programs (Hinz et al. 2020). This is typically accomplished through series of no-choice, choice, field cage, and open-field experiments (Schaffner 2001), prior to releasing them into a new environment (Heard 2000; Hinz et al. 2020). No-choice tests are used to assess the fundamental host range, which comprises plant species on which the larvae can complete their development. Choice tests (in cages, field cages, or open-field tests) are used to assess the realized host range, which comprises plant species that a BCA may use under natural conditions (Schaffner 2001). While the centrifugal phylogenetic approach (Wapshere 1974) is instrumental in guiding the selection of test plant species by suggesting that closely related plant species are more likely to serve as suitable alternative hosts for BCA development, studies such as Hinz et al. (2014) and Park et al. (2018) indicate that this method may have limitations in providing detailed insights into the host finding behavior of a candidate BCA. Host finding behavior, which encompasses the behavior exhibited by an insect herbivore during the initial stages of host-plant selection, often requires a more direct observational and experimental approach to capture the nuanced interactions between the herbivore and their potential host plants.

Insect herbivores use olfactory and visual cues from host plants to determine whether to accept or reject a plant during the host-finding stage of host selection (Bernays and Chapman 1994; Balkenius et al. 2009). The cue modalities used during host-finding include volatile organic compounds (VOCs) and foliar and floral colors or structures. The realized host range is typically a subset of the fundamental host range that only includes those species that are visited for potential colonization in the field. Species that are physiologically suitable to support development of a specialist herbivore BCA may be excluded from the realized host range because they lack cues to which the agent is responsive (Schaffner 2001; Schaffner et al. 2018). Understanding a BCA candidate’s host-finding behavior could help explain discrepancies between its fundamental and realized host ranges (Heard 2000; Park et al. 2018) and aid in accurately predicting potential risks of non-target plant attack (Hinz et al. 2014, 2020).

While the host-finding behavior of herbivorous insects has been studied extensively (Knolhoff and Heckel 2014), few of these studies are conducted in the context of biological weed control and most of those involve only one cue modality (either olfactory or visual cues) (Park et al. 2018; but see Andreas et al. 2009; Cossé et al. 2006 [olfactory cues]; Müller and Nentwig 2011; Reddy et al. 2009 [visual cues]). However, both cue modalities may play important roles during host-finding, and should therefore be considered during investigations, ideally simultaneously (Park et al. 2018; Subedi et al. 2023).

The Eurasian invasive herbaceous mustard Isatis tinctoria L. (Brassicaceae), native North American confamilial non-target plant species, and the BCA candidate, the seedpod feeding weevil Ceutorhynchus peyerimhoffi Hustache (Coleoptera: Curculionidae), provide a good study system to investigate the host-finding behavioral responses of a candidate BCA to visual and olfactory plant cues. In host range experiments, C. peyerimhoffi demonstrated a broad fundamental host range, as determined in no-choice tests (Weyl et al. 2017). However, when assessing the realized host range of C. peyerimhoffi in field cage choice tests, heavy attack (potential to kill a plant or reduce reproductive output substantially) was observed for I. tinctoria while other plant species, which were part of the fundamental host range, had only negligible attack (minor damage) in choice field cage tests (Weyl et al. 2017). The data suggest that host selection behavior underlies C. peyerimhoffi’s specialization on I. tinctoria but studies investigating potential mechanisms are lacking. The study system allows us to investigate the integration of multiple sensory modalities that insects use in the wild, which could lead to better understanding and more effective use of weed BCAs.

The aim of this study was to facilitate environmental safety assessments for C. peyerimhoffi through behavioral choice tests using visual and olfactory plant cues of critical native North American non-target plant species. We were interested in examining the host-finding behavior of C. peyerimhoffi with regard to those non-target species that supported the development of the weevil in previous traditional no-choice host range tests but escaped attack in choice field cages tests. We hypothesized that C. peyerimhoffi uses visual and/or olfactory plant cues to discriminate between I. tinctoria and non-target plant species attacked in no-choice larval development tests and that a combination of cues will allow greatest host discrimination. We tested whether C. peyerimhoffi females respond to visual and olfactory cues of confamilial non-target plants in the absence of the Eurasian host, I. tinctoria, cues and whether C. peyerimhoffi females prefer visual and olfactory cues of I. tinctoria over the respective cues of native North American confamilial plant species.

Materials and methods

Study system

Isatis tinctoria (dyer’s woad) is a Eurasian winter annual or biennial mustard that is invasive in the western USA (Gaskin et al. 2018). It is a declared noxious weed in 11 western US states (USDA NRCS 2023). The seedpod weevil C. peyerimhoffi is under consideration as a BCA for the control of I. tinctoria in the USA (Weyl et al. 2022). Females C. peyerimhoffi emerge during spring and feed on I. tinctoria inflorescence structures (Fig. S1). From late March to August, females typically lay eggs singly in developing I. tinctoria seedpods and the larvae feed on the seeds in the pod as they mature (Hinz et al. 2016). C. peyerimhoffi attack can reduce the seed output of I. tinctoria plants by 40–70% (Hinz et al. 2016).

Three North American non-target species confamilial to I. tinctoria were selected to study the pre-alightment host selection behavior of C. peyerimhoffi: Braya alpina Sternb. & Hoppe, Boechera Hoffmannii (Munz) Al-Shehbaz, and Caulanthus heterophyllus (Nutt.) Payson (see Table S1 for plant species details). These plant species were selected because they supported C. peyerimhoffi in no-choice developmental tests (Weyl et al. 2022). In addition, B. hoffmannii is restricted to a few locales in San Bernardino and San Diego Counties, California, USA, with the largest populations occurring on Santa Rosa and Santa Cruz Islands in Santa Barbara County, California (FNA 2023). The plant is a US federally listed threatened and endangered species in the USA (ECOS 2023).

All plant species were grown from seeds received from CABI Switzerland. Plants were grown in 3 l black plastic pots (diameter: 25 cm, height: 17 cm; T-pot Three, Stuewe and Sons, Inc, Tangent, OR, USA) using the following soil mix: one part sand and three parts Sunshine Mix No. 4 (Sun Gro Horticulture Canada Ltd). All plants were watered as needed and maintained at 26 °C day and 15 °C night, and a L:D 16:8 photoperiod. Since the flowering phenostage of the plant species was required to conduct experiments with C. peyerimhoffi, 12-week-old plants, except C. heterophyllus, were vernalized in a cold room for 10–12 weeks at a L:D 12:12 photoperiod, and a constant temperature of 4 °C. As C. heterophyllus is an annual plant, it was only germinated once the other test plants were moved into the cooling room.

Adult C. peyerimhoffi used in this study were originally collected in Italy and laboratory reared at CABI Switzerland since 2008, with a single generation per year. We received adult C. peyerimhoffi from CABI Switzerland, in March 2019 (n = 300), and maintained in a biological control quarantine laboratory at the University of Idaho, Moscow, ID, USA. The presence of ventral abdominal depressions was used to differentiate male weevils from females (Jordan et al. 1993). Twenty adults each, 15 females and five males, were maintained in transparent plastic cylinders (diameter: 11 cm, height: 15 cm, Semadeni AG, Ostermundigen, Switzerland) covered with a mesh lid in an environmental chamber (I-35 VL, Percival Mfg. Co., Boone, Iowa, USA), with fresh I. tinctoria flowers at 17 °C during the day and 8 °C at night, 60–70% RH, and a L:D 16:8 photoperiod. Since these weevils were allowed to feed on I. tinctoria flowers, they were classified as ‘experienced’ during choice tests. In April 2020, we received another shipment of C. peyerimhoffi larvae from CABI Switzerland. Larvae were kept at 5 °C in a rearing chamber (Percival Scientific Incubator, Model C-30, Percival Scientific, Inc., Perry, Iowa, USA) for emergence. A total of 67 adult C. peyerimhoffi emerged from these larvae. These weevils were separated into males and females as described above. Since these weevils had no feeding experience, they were classified ‘naïve’ and used immediately for choice tests.

For most experiments, each weevil was used only once to eliminate the possibility of learning or habituation affecting the outcomes. This was feasible due to the large number of weevils initially available. However, for some experiments conducted later, such as experiment 7, which compared the combined cues of test species versus I. tinctoria, limitations in the number of unused weevils available forced us to re-use weevils that had previously been assayed. In these instances, we selected 13 unused and seven previously used weevils for each plant species to be tested in the experiment. The ratio of unused to previously used weevils were chosen to standardize the experiment among plant species within the constraints of our weevil population. To minimize confounding factors such as learning or habituation to I. tinctoria cues, we avoided weevils previously exposed to these cues. Furthermore, when re-using weevils, their selection was randomized to ensure no prior knowledge of their involvement in earlier tests, thereby safeguarding against inadvertent bias in the experimental design.

Collection of floral headspace volatiles for olfactory cues and flowering stems for visual cues

We used a portable volatile collection system (PVCS) for dynamic headspace VOC collections, which consists of a push pump to push the air into the volatile collection bag and a pull pump to pull out the equal amount of air, and flowmeters to regulate the amount of airflow (Park et al. 2019). Based on previous studies (e.g., Oluwafemi et al. 2011; Park et al. 2018, 2019), the concentration and airflow rates were chosen to balance the need for a realistic representation of natural VOC levels with practical considerations of the experimental design. For each collection, purified air (300 ml min−1) was pushed using a Rena Air 400 pump (RENA, Chalfont, PA, USA) into a sealed sterilized polyvinyl bag (14 × 24 cm; Reynolds, Richmond, VA, USA) enclosing a flowering stem of a test plant at one end. The enclosed air, along with emitted floral headspace VOCs from the enclosed plastic bag, was pulled out at the same rate at the other end, passing through a VOC trap (40 mg Porapak Q, 80–100 mesh; Southern Scientific Inc. Micanopy, FL, USA) to absorb the emitted headspace VOCs (Park et al. 2019). Volatile collections were conducted for 3 h between (10:00 to 13:00) on sunny days for a group of six plants for a plant species with one empty bag as a control. Trapped VOCs were eluted with 200 µl of dichloromethane into screw cap vials (Supelco, Bellefonte, PA, USA) and stored at − 80 °C until further use. Flowering stems of individual test plants were cut and kept in 10 cm transparent aqua tubes (Syndicate Sales Inc., Kokomo, IN, USA) to use as a visual cue in visual choice tests.

Olfactometer choice tests

A modified glass Y-tube device (4 cm Y-stem, 12 cm arms, 2 cm internal diameter) was used to assess the weevil’s behavioral responses to host and non-host visual and olfactory cues (see Fig. S2 for set-up). This set-up is adapted from Park et al. (2019). Purified air carrying eluted VOCs from a group of six plants is pushed into Y-tube arms from the end of each arm, and an equal amount of air is pulled out from the stem end. The Y-tube olfactometer was placed on a cylindrical plastic ring (internal diameter: 16 cm, external diameter: 20 cm; height: 5 cm; Fig. S2) with a transparent plastic top (Plastic Wrap, WinCo Foods, Boise, ID, USA). The cylindrical ring consisted of two openings (diameter: 3 cm) on the side that aligned with the arms of the Y-tube resting on it and introduced visual cues into the choice test (Fig. S2). After each replicate, the Y-tube was cleaned with 70% ethanol to remove residual olfactory cues and rotated 180º to avoid any positional effects (Park et al. 2019). In a sterilized 2 mm2 filter paper, eluted VOCs (1 µl) were applied using a 10 µl manual syringe (Agilent Technologies, Sydney, Australia). Our olfactometer experiments proceeded under the assumption, based on Park et al. (2019), that insect responses to collected VOCs are comparable to responses to VOCs from whole-potted plants, without direct testing of this variable. Filter paper with eluted VOCs was then placed into a Tygon tube (R-3603, Saint-Gobain Corporation, Valley Forge, PA, USA) connected to the Y-tube arm (Fig. S2). The Rena ® 400 pumps were used to push the purified air with eluted VOCs into the Y-tube’s arm and pull it out through the stem end. Flowmeters (MR3000, Key Instruments, Hatfield, PA, USA) regulated the amount of airflow at 300 ml min−1 into each Y-tube arm. Both olfactory cues (VOCs) and visual cues (flowering stems) collected from six plants, each of the target and non-targets, were replaced every five replicates during the choice test. The Y-tube olfactometer arena was illuminated using a LED light (Jansjö ® LED lamp, Inter Ikea System B. V., Delft, The Netherlands) by placing it directly above Y-tube (Fig. S2). The testing arena was enclosed in a double-layered rectangular box (180 × 90 × 60 cm), with the inner layer being a white cloth to minimize visual distraction.

Experiments were conducted during the summer of 2019 and 2020 at the weevils’ peak activity (i.e., feeding and oviposition). Since the females were kept together with males in a cylinder, they were assumed to have mated and ready for oviposition. All C. peyerimhoffi females were starved for 24 h prior to experiments to increase the responsiveness to cues (Defagó et al. 2016). Single C. peyerimhoffi females were introduced in the Y-tube at the weevil-release point (i.e., at the end of the Y-tube stem; Fig. S2) using forceps by briefly disconnecting the outlet hose at the stem end. A camera (Contour Roam 2, Contour Ins., and Seattle, Washington, USA) was used to record the weevil behavior. For each choice test, a replicate lasted 10 min. If the C. peyerimhoffi female did not cross a decision line (3 cm from the release point into the Y-tube arms; Fig. S2), the female was considered unresponsive, and was discarded from analyses. Approximately 88% of the weevils exhibited a response in the Y-tube olfactometer experiments, with this response rate being consistent across various experiments. For each choice test, the total time spent, i.e., time spent by weevils in each arm of the Y-tube, initial (first arm) choice, time taken to reach the decision line in the Y-tube arm (response time), and weevils’ position in the Y-tube at the end of the 10 min recording period or final choice was recorded for each replicate. Total time spent was used to measure the strength of a female’s preference for cues offered. Initial choice (choosing the first arm) and response time variables were used to measure the weevil’s ability and agility to discriminate between the offered cues. Final choice was noted to explain the weevil’s ultimate preference for the offered cues. There were three possible outcomes: indifference, repellence, and attraction. For this study, we defined indifference (a response where the weevils did not exhibit a statistically significant preference for or against the test stimuli compared to the control treatments) as occurring when a plant’s VOCs and/or visual cues were not more or less preferred by C. peyerimhoffi than control cues in choice test (purified air and/or absence of a visual cue; Martini et al. 2015). Attraction occurred if plant VOCs and/or visual cues were significantly greater over the control treatment and repellence occurred if C. peyerimhoffi preferred control treatments over test plant cues (Vet et al. 1983). For each trial, the sample size ranged between 10 and 25 individual females. All choice tests were conducted in the biological control quarantine laboratory between 9:00 and 14:00 h, aligning with the diurnal activity patterns of the weevils as observed in the field (Weyl, unpublished data). The tests were carried out at a controlled temperature of 21 °C and 50% RH. At the beginning of experimentation, a blank test with purified air was conducted to test for orientation bias between the Y-tube arms.

Olfactometer choice experiments were conducted with experienced females in the summer of 2019 and naïve weevils in the summer of 2020. The first three experiments focused on naïve weevils and were conducted with all test plant species, except for B. hoffmannii, which was excluded due to the unavailability of its flowering phenostage. Experiment 1 compared olfactory plant cues (eluted VOCs) to purified air. Experiment 2 assessed visual plant cues (flowering stems) against empty arms, and experiment 3 evaluated weevil responses to a combined (olfactory and visual) plant cues. These experiments were replicated in experiments 4, 5, and 6 with experienced weevils. Experiment 7, conducted only with experienced weevils, explored their preference for combined plant cues of non-target species versus those of I. tinctoria, presenting these cues in a modified Y-tube device to understand the weevils’ relative preference in the presence of I. tinctoria cues. All experiments employed a Y-tube set-up, with visual cues presented under one arm and olfactory cues pushed through one arm, as detailed above (Fig. S3).

For the experiments with experienced weevils, the trial sequence was structured as follows: (1) Olfactory cues versus control (experiment 4), with one day dedicated to each of the four plant species under study for a total of four days; (2) visual cues versus control (experiment 5), conducted over the subsequent four days, again with one plant species tested each day; (3) combined cues versus control (experiment 6), again conducted during the subsequent four days; and (4) Test plant species combined cues versus I. tinctoria combined cues (experiment 7), conducted during three days since only three plant species were tested. For the experiment with naïve weevils, we replicated the sequence used for experienced weevils, completing this over a span of three days.

Statistical analysis

Statistical analyses were conducted using R version 4.3.1 (R Core Team 2024), with a focus on the key variables: time spent, initial choice, final choice, and response time. To standardize the weevil time spent response, we measured the proportion of time spent in each arm of the Y-tube, excluding the time spent in the weevil-release area.

Using the ‘cor’ function from the ‘stats’ package, pairwise correlations were calculated to assess relationships among the variables. Data were grouped by experimental parameters, including plant species (I. tinctoria, B. alpina, B. hoffmannii, C. heterophyllus), type of experiment, and weevil status (naïve or experienced). This analysis pinpointed a moderate and consistent correlation between initial and final choice variables (Table S2). Consequently, we focus primarily on initial choice, relegating final choice details to supplementary materials (Fig. S4, Table S3). Additionally, response time, which displayed negligible correlations with other variables but lacked significant treatment differences across most experiments, is reported in supplementary materials (Table S4).

A critical examination of the random effects in the Generalized Linear Mixed Model (GLMM; ‘glmmTMB’ package; Brooks et al. 2017) for time spent and initial choice indicated an inconsequential variance attributed to the ‘replicate’ or ‘repeated use of weevils.’ This lack of significant variability led to the exclusion of random effects, leading to the use of Generalized Linear Models (GLMs; ‘lme4’ package; Bates et al. 2015). A Generalized Linear Model (GLM) with a Gaussian error distributions was utilized to analyze time spent, followed by a likelihood ratio test (ANOVA; 'car' package; Fox and Weisberg 2019). For initial choice, we conducted a GLM with Binomial error distributions and logit link function for different plant species in each experiment, where Wald tests evaluated the significance of individual predictors, and the Z-value was determined as the ratio of the estimate to its SE. These models included ‘treatment’ as the explanatory variable, discerning between cues from plant species versus control (purified air/empty arm), or I. tinctoria versus non-target species cues. The outcomes from these GLMs are presented graphically in the results section, with time spent data depicted as proportions and initial choice data as percentages. To determine their relative influence of plant cues on time spent, behavioral responses of female C. peyerimhoffi were analyzed using a GLM with Gamma error distributions and inverse link function, contrasting olfactory, visual, and combined cues. Post-hoc comparisons of cue modalities were executed using estimated marginal means, facilitating a detailed assessment of individual versus combined cue effects. For all tests, P-values < 0.05 were considered significant.

Results

Experiment 1: olfactory cues versus purified air with naïve weevils

For time spent, naïve C. peyerimhoffi females spent more time in the arm with I. tinctoria volatiles than those containing purified air (χ2(1) = 37.815, P < 0001; Fig. 1a). In contrast, females spent more time in arms with purified air compared to B. alpina olfactory cues (χ2(1) = 19.606, P < 0.0001; Fig. 1a), and there was no significant difference for the time spent in arms filled with purified air or C. heterophyllus volatiles (χ2(1) = 0.076, P = 0.782; Fig. 1a). Weevils did not prefer any tested species for initial choice (I. tinctoria: Z = 1.754, P = 0.080, B. alpina: Z = 1.829, P = 0.067, and C. heterophyllus: Z = 0.301, P = 0.763; Fig. 1b).

Fig. 1
figure 1

Proportion of time spent (mean ± SE) and percent initial choice (mean ± SE) by naïve Ceutorhynchus peyerimhoffi in a modified Y-tube choice tests with olfactory cues (time spent: a, initial choice: b), visual cues (time spent: c, initial choice: d), combined cues (time spent: e, initial choice: f) of native North American confamilial plant species and Isatis tinctoria against control (purified air and/or empty arm). Olfactory cues were derived from eluted floral volatiles, with flowering stems providing visual cues. Generalized Linear Model of individual experiments (*P < 0.05, ***P < 0.001; n.s. not significant (P > 0.05)). Significance (*, ***) are for the null hypothesis of 50% time spent in or initial choice for the treatment (or control) arm. Numbers next to each right bar on time spent indicate the sample size for that experiment (N Responders, N.R. Non responders (discarded from analysis))

Full size image

Experiment 2: visual test plant cues (flowering stems) versus empty arms with naïve females

Response of naïve C. peyerimhoffi females differed for time spent between visual cues of I. tinctoria and the empty arm (χ2(1) = 4.182, P = 0.0409; Fig. 1c). For all other test plant species, time spent by naïve C. peyerimhoffi females were not significantly different from the empty arm (B. alpina: χ2(1) = 0.129, P = 0.719, and C. heterophyllus: χ2(1) = 1.548, P = 0.213; Fig. 1c). Weevils were indifferent to flowering stems of all tested species for initial choice (I. tinctoria (Z = 1.648, P = 0.099), B. alpina (Z = 0, P = 1), and C. heterophyllus (Z = 0, P = 1); Fig. 1d).

Experiment 3: combined olfactory and visual test plant cues versus control with naïve females

Naïve weevils preferred combined olfactory and visual cues of I. tinctoria over purified air and empty arms (χ2(1) = 20.715, P < 0.0001; Fig. 1e), but there was no difference in time spent of females to combined cues of B. alpina (χ2(1) = 0.067, P = 0.795; Fig. 1e) and C. heterophyllus (χ2(1) = 0.077, P = 0.780; Fig. 1e), and control treatments, respectively. Naïve weevils chose I. tinctoria combined cues over the control arm (Z = 2.195, P = 0.002; Fig. 1f). No significant preference for combined cues of B. alpina (Z = 0.575, P = 0.566; Fig. 1f) and C. heterophyllus (Z = − 0.824, P = 0.41; Fig. 1f) was found for initial choice.

Responses of naïve female weevils to combined olfactory and visual cues of tested species was not different from their respective individual olfactory and visual cues for time spent (P > 0.05; Table S5). In a test of the relative strength of cues, no difference in naive C. peyerimhoffi response between olfactory cues, visual cues, and combined olfactory-visual cues was detected (P > 0.05; Table S6).

Experiment 4: olfactory test plant cues versus purified air with experienced females

Experienced C. peyerimhoffi females spent more time in the arm with I. tinctoria volatiles than with purified air (χ2(1) = 27.197, P < 0.0001; Fig. 2a). There was no preference for time spent between olfactory cues of non-target plant species and purified air (B. alpina: χ2(1) = 0.069, P = 0.793, B. hoffmannii: χ2(1) = 0.038, P = 0.845, and C. heterophyllus: χ2(1) = 1.250, P = 0.263; Fig. 2a). When VOCs from plants were presented in the Y-tube against purified air, experienced female C. peyerimhoffi preferred I. tinctoria for initial choice (Z = 2.545, P = 0.011; Fig. 2b) and were indifferent to all other test species (B. alpina: Z = 0, P = 1, B. hoffmannii: Z = 0.446, P = 0.655, and C. heterophyllus: Z = 0.724, P = 0.469; Fig. 2b).

Fig. 2
figure 2

Proportion of time spent (mean ± SE) and percent initial choice (mean ± SE) by experienced Ceutorhynchus peyerimhoffi in a modified Y-tube choice tests with olfactory cues (time spent: a, initial choice: b), visual cues (time spent: c, initial choice: d), combined cues (time spent: e, initial choice: f) of test plant species against control (purified air and/or empty arm, white bars), and proportion of time spent (mean ± SE) and percent initial choice (mean ± SE) by experienced C. peyerimhoffi in a modified Y-tube choice tests with combined cues of native North American confamilial plant species against Isatis tinctoria cues (time spent: g, initial choice: h). Olfactory cues were derived from eluted floral volatiles, with flowering stems providing visual cues. Generalized Linear Model of individual experiments (*P < 0.05, **P < 0.01, ***P < 0.001; n.s.: not significant (P > 0.05)). Significance (*, **, ***) are for the null hypothesis of 50% time spent in or initial choice for the treatment (or control) arm. Numbers next to each right bar on time spent indicate the sample size for that experiment (N Responders, N.R. Non responders (discarded from analysis))

Full size image

Experiment 5: visual plant cues versus an empty arm with experienced females

Experienced female C. peyerimhoffi spent more time in the arm with I. tinctoria flowering sprigs compared to the empty control arm (χ2(1) = 30.867, P < 0.0001; Fig. 2c). There was no difference in time spent between arms with C. heterophyllus visual cues and empty control arms (χ2(1) = 0.412, P = 0.520; Fig. 2c). In contrast, weevils spent more time in the empty control arms than those with visual cues of B. alpina (χ2(1) = 6.741, P = 0.009; Fig. 2c) and B. hoffmannii (χ2(1) = 5.247, P = 0.022; Fig. 2c), respectively. Experienced females preferred I. tinctoria flowering stems over the empty arm for initial choice (Z = 2.832, P = 0.005; Fig. 2d) and were indifferent to flowering stems of other test species (B. alpina: Z = 1.381, P = 0.167, B. hoffmannii: Z = 1.758, P = 0.079, and C. heterophyllus: Z = − 0.446, P = 0.655; Fig. 2d).

Experiment 6: combined olfactory and visual cues versus control treatments with experienced females

Experienced females preferred olfactory and visual cues of I. tinctoria (χ2(1) = 32.576, P < 0.0001; Fig. 2e) and those of C. heterophyllus (χ2(1) = 6.694, P = 0.009; Fig. 2e) over purified air and empty arms for the time spent. There was no difference in the time spent for combined cues of B. alpina (χ2(1) = 3.222, P = 0.072; Fig. 2e) and B. hoffmannii (χ2(1) = 0.343, P = 0.557; Fig. 2e) against the control. For initial choice, females preferred I. tinctoria (Z = 2.622, P = 0.009; Fig. 2f) over an empty arm and were indifferent to B. alpina (Z = 0.724, P = 0.469; Fig. 2f), B. hoffmannii (Z = 0.989, P = 0.323; Fig. 2f), and C. heterophyllus (Z = 1.462, P = 0.144; Fig. 2f).

The response of the weevils to combined olfactory and visual cues did not differ from the individual olfactory and visual cues for I. tinctoria for time spent (P > 0.05; Table S5). In a test of the relative strength of cues, no difference in experienced C. peyerimhoffi response between olfactory cues, visual cues, and combined olfactory-visual cues was detected for time spent (P > 0.05; Table S6).

Experiment 7: combined olfactory and visual cues of non-target species versus combined cues of I. tinctoria

In choice tests for olfactory and visual plant cues of confamilial non-target plant species against those of I. tinctoria, experienced female weevils preferred I. tinctoria over all non-target species tested based on time spent in arms (B. alpina: χ2(1) = 7.079, P = 0.007, B. hoffmannii: χ2(1) = 4.473, P = 0.034, and C. heterophyllus: χ2(1) = 12.747, P = 0.004; Fig. 2g). For initial choice, weevils preferred I. tinctoria over B. alpina (Z = 2.349, P = 0.019; Fig. 2h) and C. heterophyllus (Z = -2.421, P = 0. 015; Fig. 2h), but not over B. hoffmannii (Z = 1.386, P = 0.166; Fig. 2h).

Discussion

In this study, we investigated the host-finding behavior of the biological control agent (BCA) candidate C. peyerimhoffi in relation to its Eurasian field host I. tinctoria and three native North American confamilial non-target plant species (B. alpina, B. hoffmannii, and C. heterophyllus). We conducted choice tests to assess the weevil’s responses to both visual and olfactory plant cues in a modified Y-tube olfactometer, measuring four variables (time spent, initial choice, final choice, and response time). Our results indicate that C. peyerimhoffi females prefer I. tinctoria plant cues while exhibiting no preference for either visual or olfactory cues, or their combination, in almost all tests with the non-target species. Most responses to non-targets during choice tests indicated indifference or repellence, except experienced C. peyerimhoffi females that were attracted to the combined olfactory and visual cues of C. heterophyllus. Our results suggest that the host range of C. peyerimhoffi is limited by its host-finding behavior, similar to other biological control agents (Chu et al. 2019; Kafle 2016). Our data are consistent with the concept that the realized host range of a BCA candidate species is typically narrower than its fundamental host range (Schaffner 2001).

The data indicate that C. peyerimhoffi females are attracted to both olfactory and visual cues of I. tinctoria, regardless of prior feeding experience. Both olfactory and visual cues alone were sufficient to elicit attraction in both naïve and experienced females, suggesting that C. peyerimhoffi utilizes both olfactory and visual cues in host finding and discrimination. Additionally, in choice tests between I. tinctoria and non-target species with combined cues, experienced C. peyerimhoffi females strongly preferred their Eurasian field host I. tinctoria over native North American species tested, suggesting that the weevil can discriminate between I. tinctoria and all three tested native North American plant species. This finding could explain the observed differences in the C. peyerimhoffi attack between I. tinctoria and each test plant species in choice field cage tests (see Weyl et al. 2017). Our findings support prior results of traditional host range investigations of C. peyerimhoffi and suggest that visual and olfactory cues are not only crucial in C. peyerimhoffi’s host finding and acceptance but that they also contribute to its narrow realized host range.

Indifferent and repellant responses of naïve and experienced C. peyerimhoffi females to various plant cues of B. alpina and B. hoffmannii suggest that the weevil is unable to identify these plant species as potential hosts in the field during pre-alightment stages of host finding. Olfactory cues are an essential component in host finding (Bernays and Chapman 1994) as they can provide reliable information regarding host-plant quality (host suitability, nutritional quality; Tasin et al. 2011). Similarly, Reeves (2011) suggests that host and non-host visual cues are perceptible and reliably used by herbivores. Therefore, the observed indifference or repellent response of C. peyerimhoffi to olfactory and visual cues of B. alpina and B. hoffmannii suggest these species are non-hosts (Tasin et al. 2011) that go undetected or even repel the weevil in the field (Byers et al. 2004; Deletre et al. 2016). Specialist herbivores with narrow host ranges, such as C. peyerimhoffi, may use olfactory (Tasin et al. 2011) and/or visual cues (Stenberg and Ericson 2007) to efficiently distinguish suitable hosts from non-hosts to maintain their host fidelity.

In choice tests, both naïve and experienced C. peyerimhoffi females showed no attraction to individual olfactory, visual or combined cues of C. heterophyllus for both initial choice and final choice parameters. However, experienced females preferred combined cues over the control based on the amount of time spent in the Y-tube arm, indicating that prior feeding experience on I. tinctoria allows C. peyerimhoffi to identify C. heterophyllus as a potential host. The reason for this incongruent pattern of preference is unclear. Both naïve and experienced C. peyerimhoffi tended to prefer olfactory cues of C. heterophyllus over the control based on time spent in the olfactometer arm, suggesting that the floral headspace volatile profile of C. heterophyllus includes at least one potentially attractive VOC. Previous research on Mogulones crucifer Pallas (Coleoptera: Curculionidae) has shown that its attraction to C. officinale can be triggered by a single bioactive VOC (methyl isovalerate) (Kafle 2016). However, the lack of attraction based on both initial and final choice suggests that C. heterophyllus cues may present a broken or incomplete signal to which the weevils respond inconsistently. Furthermore, we found that C. peyerimhoffi displayed a greater preference for combined visual and olfactory cues of its host plant I. tinctoria over those of C. heterophyllus. This suggests that the floral headspace volatile profile of I. tinctoria contains attractive VOCs not present in the profile of C. heterophyllus. Weevils may show increased sensitivity and fidelity to complete cues, leading to more consistent responses (Li et al. 2017), underscoring the importance of complete and reliable information during decision-making.

In our experiments we used yellow-flowered C. heterophyllus, but it has been reported to have different inflorescence morphotypes with varying sepal and petal colors (Brassibase 2023). Because our study tested one morphotype of C. heterophyllus, namely the one most resembling I. tinctoria, our results may not reflect responses by the weevil to the morphotype diversity of the species. This limitation could explain the observed attraction of C. peyerimhoffi to the combined visual and olfactory cues of C. heterophyllus in our choice tests (Lyu et al. 2021). Additionally, metabolic pathways for floral color and volatiles in inflorescences may be shared, resulting in the correlation among these traits. Selection of specific floral volatile compounds may also be due to the direct selection of floral color (Zvi et al. 2008). Therefore, the differing inflorescence morphotypes of C. heterophyllus may not only differ in floral coloring but also volatile profiles. For example, Ascrizzi and Flamini (2020) found that the floral volatile profile of Iris lutescens Lam. (Iridaceae) differed between two inflorescence morphotypes growing in serpentine soil. Different morphotypes of C. heterophyllus may produce different floral reflectance and VOC profiles, potentially leading to different behavioral responses from C. peyerimhoffi than observed in our study.

Ideally, experiments should be conducted using naïve insects to exclude potentially confounding factors like female age or gravidity that could affect test results (Thompson 1988). Here, that was not possible because the availability and logistics of obtaining weevils prevented the use of naïve individuals.

Our experiments show that C. peyerimhoffi utilizes both visual and olfactory cues to distinguish its host plant, I. tinctoria, from three native North American non-target confamilial species. This utilization of host-finding cues may help maintain the weevil’s host fidelity. Also, these results highlight the importance of including both cue modalities in behavioral experiments used for environmental safety assessments. This information can assist in differentiating the realized and fundamental host range of C. peyerimhoffi as assessed in traditional host range investigations. Our findings also suggest that C. peyerimhoffi is unlikely to cause any non-target damage to B. alpina or the endangered B. hoffmannii post-release. While the results of our experiments on pre-alightment host-finding are encouraging, it should be noted that post-alightment cues can also contribute to host selection by weed biological control agents (Das et al. 2019) limiting their realized host range. Studies like ours, as well as those conducted by Park et al. (2018), Fung et al. (2021), and Subedi et al. (2023), can provide insight into the mechanisms underlying host selection among biological control agent candidates and complement traditional methods of testing host specificity. Further research linking host selection behavior and traditional field testing of BCAs can aid in designing effective pre-release field studies and improve the accuracy of conclusions drawn from these studies.