The plants used for the experiments with all parasitoids, except for tests with Diaeretiella rapae M’Intosh (Hymenoptera: Braconidae), came from a wild accession of B. rapa; seeds were collected in 2012 and 2013 near Maarssen, Netherlands. Plants used for experiments with D. rapae were Arabidopsis thaliana (L.) (wild-type, Columbia-0) instead of B. rapa because the aphid and parasitoid colony had been maintained on that plant at the University of Zurich. All plants were grown in controlled growth chambers under 16/8 L: D light regime at constant 25°C, light intensity 240–260 μmol. They were grown in cylindrical plastic pots (4 × 10 cm), with fertilized commercial soil (Ricoter Aussaaterde, Aarberg, Switzerland) and were watered every other day without supplemental nutrients. Plants used for experiments were approximately 3 weeks old; at that age, B. rapa plants had three to five fully expanded leaves and A. thaliana had 10–15 rosette leaves.

A total of five parasitoids species, all naturally occurring in Europe and representative of a range of ecological characteristics regarding their host range, host life stage targeted, and known use of plant volatiles were used in the study. Specifically, we used the following species: C. glomerata, Hyposoter ebeninus (Gravenhorst) (Hymenoptera: Ichneumonidae), Pteromalus puparum L. (Hymenoptera: Pteromalidae), Microplitis mediator (Haliday) (Hymenoptera: Braconidae), and D. rapae (Hymenoptera: Braconidae) (Table 1 and Figure 2). Three of these species (C. glomerata, H. ebeninus, and M. mediator) parasitize larvae of caterpillars, one (D. rapae) parasitizes aphids, and the fifth one (P. puparum) parasitizes lepidopteran pupae. The use of HIPVs for host location is well documented for parasitoids of caterpillars and aphids in general (Godfray, 1994), and particularly well for C. glomerata (Geervliet et al., 1996; Bukovinszky et al., 2012; de Rijk et al., 2016; Desurmont et al., 2016a, b) and D. rapae (Reed et al., 1995; Girling et al., 2006; Blande et al., 2007) but remains poorly known for pupal parasitoids (Wajnberg et al., 2008). In regard to their degree of dietary specialization, two of the species (C. glomerata and H. ebeninus) are considered to be oligophagous parasitoids that can only attack a few host species within the family Pieridae (Lepidoptera) (Harvey et al., 2010b), whereas the three other species are generalist species able to parasitize hosts belonging to different insect families (Peck, 1963; Arthur and Mason, 1986; Pike et al., 1999). Parasitoids used for the study originated from laboratory colonies maintained on P. brassicae L. (Lepidoptera: Pieridae) larvae (parasitoid species C. glomerata and H. ebeninus), P. brassicae pupae (parasitoid species P. puparum), and Brevicoryne brassicae (Hemiptera: Aphididae) (parasitoid species D. rapae). P. brassicae caterpillars originated from a laboratory rearing maintained in our laboratory on B. rapa for oviposition and Brassica oleracea (Brassicales: Brassicaceae) for larval development. The rearing was initiated with field-collected individuals from various locations in Switzerland. B. brassicae aphids came from a laboratory rearing maintained at the University of Zurich.

Cotesia glomerata originated from field-collected individuals from the Wageningen area (Netherlands) later supplemented with individuals collected in the Neuchâtel and Zurich areas (Switzerland). P. puparum and H. ebeninus originated from individuals collected in the field in the Neuchâtel area. M. mediator originated from individuals collected in the Frick area (Switzerland). Finally, D. rapae originated from individuals collected in the Zurich area. Parasitoids were maintained under similar conditions before being used for experimental purposes: they were collected as cocoons or adults after emergence from their hosts and were placed in cages (30 cm × 30 cm × 30 cm, model: bugdorm DP1000, MegaView Science Co., Taichung, Taiwan) without any host or plant material. They were provided water and droplets of honey and were kept in an incubator at 25°C and a 12:12 (l:d) photoperiod for 48 h to allow mating. Then the cages were transferred to an incubator at 13°C and a 12:12 (l:d) photoperiod. In the case of C. glomerata, most males were removed from the cages before the transfer at 13°C. Mated naïve females (i.e., females that had never been exposed to plant odors) that were 2–4 weeks old were used for all experiments. This rearing protocol was used with success in previous studies investigating the behavior of C. glomerata (Desurmont et al., 2016a, b).

The method of using odor blends created by mixing pure synthetic compounds in behavioral bioassays is ideal to work with consistent blends whose concentrations can be easily changed and has been used with success in previous studies (Schiestl et al., 2014; Desurmont et al., 2015). The synthetic floral odor mix used for parasitoid behavioral bioassays was based on the six most abundant compounds produced by flowers of this B. rapa population found in headspace collections (Schiestl et al., 2014; Desurmont et al., 2015). The concentration of each compound in the solution was adjusted so that two rubber septa (GR-2, 5 mm Supelco, Bellefonte, PA, United States) soaked in the floral mix solution would emit the emission rate of each compound comparable to one inflorescence of B. rapa (ca. 30 flowers, hereafter named 1 inflorescence equivalent, IE), as described in a previous study (Schiestl et al., 2014). The six compounds composing the floral mix were: phenylacetaldehyde (≥90% purity, Sigma-Aldrich, Buchs, Switzerland) 3 μL/mL, nonanal (Givaudan, Dübendorf, Switherland) 9 μL/mL, decanal (Givaudan) 4 μL/mL, acetophenone (Givaudan) 24.5 μL/mL, p-anisaldehyde (≥99% purity, Sigma-Aldrich) 27 μL/mL, and α-farnesene (mixture of isomers, Sigma-Aldrich) 492 μL/mL, diluted in dichloromethane (HPLC grade, Sigma-Aldrich). Rubber septa (GR-2, 5 mm Supelco, Bellefonte, PA, United States) were left to soak in the floral mix solution for a duration of 1 h, then were allowed to dry for 2 h before being used for olfactometer or wind tunnel bioassays with parasitoids.

The preferences of parasitoids for different odors were measured in a 4-arm olfactometer setting (D’alessandro and Turlings, 2005). In this setting, wasps were given the choice between four odor sources (= treatments) contained in separated glass bottles. Individual air flows were connected to each odor source and all converged to a central glass piece, where the wasps were released. After 30 min, wasps were recollected and the treatment they chose was recorded. For all parasitoids, except for H. ebeninus, one olfactometer test (= 1 replicate) consisted of five consecutive releases of five wasps (25 wasps tested per replicate) with the same set of plants. For H. ebeninus, which is significantly larger than the other species tested, parasitoids were released in groups of three instead of five, and one replicate consisted of five consecutive releases of three wasps with the same set of plants. Parasitoids were given the choice among two olfactometer arms carrying only purified air (= blanks), one arm with the odor of a vegetative plant infested by their host herbivore (herbivore treatment), and one arm with the odor of a vegetative plant infested by their host with the addition of synthetic floral odors (herbivore + floral odors treatment). To add the floral odor, two rubber septa spiked with a blend of volatiles (see Materials and Methods section: preparation of synthetic floral odor mixes) were placed above each plant from the herbivore + floral odors treatment prior to the tests. Two control septa that had been treated only with solvent were added to each plant from the herbivore treatment. Plants were changed and the olfactometer glassware was cleaned between replicates, and a total of six replicates was done for each parasitoid species tested (totaling 6 × 25 or 6 × 15 wasps per species), except for C. glomerata (nine replicates; 9 × 25 wasps). The modalities of plant infestation by herbivores prior to the tests (timing of infestation, number of herbivores used) is summarized in Table 1.

Olfactomer tests conducted to investigate dose-dependent effects of floral odors on C. glomerata attraction used the same 4-arm olfactometer setup. For these tests, parasitoids has to choose among two odor sources with purified air (= blanks), one vegetative plant infested by their host (herbivore treatment), and one vegetative plant infested by their host with the addition of synthetic floral odors (herbivore + floral odors treatment). Five concentrations of floral odors were tested: 0.25, 0.5, 1, 2, and 4 IE (see section “Materials and Methods”: preparation of synthetic floral odor mixes). One replicate consisted in five consecutive releases of five wasps with the same set of plants (25 wasps tested per replicate), and a number of five to nine replicates per concentration were conducted (0.25 IE, N = 7; 0.5 IE, N = 8, 1 IE, N = 9; 2 IE, N = 5, 4 IE, N = 5).

For these behavioral tests, a 200 cm × 80 cm × 80 cm wind tunnel was used. A speed-controller fan (D340/E1, FDR32, Neunkirchen, Germany) pushed air through the tunnel, and the air velocity of 0.35 m s^–1. Four charcoal filters (145.457 mm, carbon thickness 16 mm, Camfil Farr, Reinfeld, Germany) cleaned the incoming air. The tests were conducted during daytime at ambient room temperature (ca. 22°C). Naïve parasitoid females were tested individually in the wind tunnel. They were introduced in the wind tunnel in open plastic tubes placed on a stand at the downwind side of the tunnel. For the first experiment (effects of real B. rapa flowers on the behavior of C. glomerata), a 3-week old vegetative B. rapa plant with 3–5 fully expanded leaves in a cylindrical plastic pot (4 cm × 10 cm) infested with 15 first instar P. brassicae 24 h prior to the test was placed at the upwind end of the wind tunnel. Either a blooming B. rapa plant with one inflorescence (ca 30 flowers) or a yellow cardboard-made odorless decoy inflorescence of the same dimensions was placed 5 cm behind the infested B. rapa plant. Once an individual parasitoid flew away from the open tube to track an odor plume, the following variables were recorded: the total flight duration until landing, the outcome of parasitoid foraging after landing (flew away or attacked a P. brassicae caterpillar) and, for parasitoids that did attack a caterpillar, the duration between landing on the plant and attacking a caterpillar. The parasitoid was then removed from the wind tunnel. For each herbivore-infested B. rapa plant, five parasitoids were tested with a blooming B. rapa plant and five were tested with an odorless decoy, then the plants were replaced. The experiment was repeated five times (N = 5, 10 parasitoids tested per replicate). For the second experiment (effects of different concentrations of floral odors on the behavior of C. glomerata), two 3-week old vegetative B. rapa plants infested with 15 first instar P. brassicae were placed at the upwind end of the wind tunnel. The two plants were placed side by side and were separated by a distance of 20 cm. Two rubber septa soaked in floral odors prior to the experiment were placed along a vertical metal rod placed on a support behind one of the two B. rapa plants, and two septa that had been soaked only in solvent were placed along a vertical metal rod behind the second B. rapa plant. Once an individual parasitoid flew away from the open tube to track an odor plume, the plant it landed on was recorded and the parasitoid was removed. To avoid positional biases (i.e., parasitoid flying preferentially to one side of the wind tunnel), the position (left/right) of the rods or the position of the B. rapa plants was switched in the wind tunnel every five parasitoids tested so that all possible positional combinations were covered every 20 parasitoids tested. The B. rapa plants and the septa were changed once 20 parasitoids were tested (= 1 replicate), and the experiment was replicated five times per concentration of floral odors. Four concentrations of floral odors were tested (0.5, 1, 2, and 4 IE) for a total of 100 individual parasitoids tested per concentration. For all wind tunnel tests, parasitoids that did not fly after 5 min or flew directly to the ceiling or the side of the tunnel without tracking an odor plume were discarded and not counted in the total of parasitoids tested.

Mated C. glomerata females were anesthetized using CO₂ and their head was excised below the prothoracic segment with micro scissors. The segment was inserted into the recording electrode (AgCl electrodes bathed in 0.1 M KCl solution contained in borosilicate glass capillaries, OD = 1.5 mm, ID = 1.05 mm, Hilgenberg, Malsfeld, Germany). Subsequently, the distal annulum of both antennae was cut off and the ends were inserted into the indifferent electrode. The head was held in a humidified, charcoal-filtered air stream (90–100% RH) delivered at 1 m/s via a metal tube (ID 6.9 mm) whose outlet was about 1 cm from the preparation. EAG responses were recorded on a computer using Syntech hardware and software (Syntech, Kirchzarten, Germany). The responsiveness of antennae was tested at the start with an air puff from a 5-mL stimulus syringe containing 1 μg of (Z)-3-hexenol on a filter paper strip introduced into the metal tube through a teflon tape-covered hole (Taneja and Guerin, 1997). If the EAG response to this 1 s air puff (1 mL/s) from the stimulus syringe was less than 6 mV, another head was mounted. Headspace collections were obtained by placing a 3-week old vegetative B. rapa plant with 3–5 fully expanded leaves that had been infested with 15 1st instar P. brassicae caterpillars 24 h prior in a volatile collection setup (Ton et al., 2007) for 6 h. In this setup, plant volatiles were collected using trapping filters containing 25 mg of 80–100 mesh SuperQ absorbent and extracted from the filters at the end of the collection with 150 μL of dichloromethane. Solvent extracts were obtained by rinsing leaves of vegetative B. rapa plants that had been infested with 15 1st instar P. brassicae caterpillars for 24 h directly with dichloromethane. Solutions containing synthetic plant volatiles (0.1 μg/μL) and headspace collections were placed in an aliquot of 10 μL on a filter paper strip that was placed in the stimulus syringes after evaporation of the solvent, and solvent extracts were applied to rubber septa similar to the ones used for behavioral bioassays and introduced into the stimulus syringe after being allowed to dry for 30 min.

In order to test how mixtures of B. rapa floral volatiles affect the EAG responses of C. glomerata females to leaf volatiles, three synthetic flower compounds separately or in mixture (flower mix) were presented to C. glomerata females in combination with three synthetic leaf compounds separately or in mixture (leaf mix) or in combination with headspace collections and solvent extractions of herbivore infested B. rapa leaves. The synthetic leaf chemicals used were three common leaf volatiles produced by herbivore-infested B. rapa plants (Danner et al., 2017; Desurmont et al., 2018): benzyl nitrile, (E)-2-hexenal, and (Z)-3-hexenyl acetate, and the synthetic flower chemicals used were three common compounds produced by B. rapa flowers: p-anisaldehyde, (E,E)-α-farnesene, and phenylacetaldehyde (Schiestl et al., 2014). The following stimuli were tested: solvent control (dichloromethane, N = 27), benzyl nitrile (benznit, N = 9), (E)-2-hexenal (e2hex, N = 9), (Z)-3-hexenyl acetate (z3ac, N = 10), mixture of benzyl nitrile, (E)-2-hexenal and (Z)-3-hexenyl acetate (leafmix, N = 10), p-anisaldehyde (anis, N = 9), mixture of leafmix and anis (leafmixanis, N = 10), α-farnesene (farn, N = 9), mixture of leafmix and farn (N = 10), phenylacetaldehyde (paa, N = 10), mixture of leafmix and paa (leafmixpaa, N = 9), mixture of α-farnesene, p-anisaldehyde, and phenylacetaldehyde (flower mix, N = 10), a mixture of leafmix and flower mix (N = 10), headspace collection of B. rapa leaves (headspace, N = 9), a mixture of headspace and flower mix (N = 10), solvent extract of B. rapa leaves (extract, N = 9), a mixture of flower mix and extract (extractflower, N = 9). Odor stimuli were tested in a random order. For each odor stimulus 9–10 C. glomerata females were used.

Olfactometer bioassays: preferences of C. glomerata females were analyzed for each test using a generalized linear model (GLM) with a Poisson distribution fitted by maximum quasi-likelihood estimation (Turlings et al., 2004), with the number of wasps counted in the different branches of the olfactometer as the dependent variable. Means were then compared using a non-parametric multiple comparison Wilcoxon test (α = 0.01, JMP12). The number of wasps tested in each olfactometer test varied depending on the species tested (15 wasps per replicate for H. ebeninus, 25 per replicate for all other parasitoids), and the olfactometer data were analyzed for each parasitoid species separately. Results are presented as percentage attractiveness in the figures illustrating olfactometer tests for easier comprehension: percentage attractiveness of a treatment was calculated as the number of wasps that chose a treatment divided by the total number of wasps that made a choice during the test × 100. Wasps that did not make a choice during the olfactometer tests were not included in the analyses but were included to calculate parasitoid participation (i.e., number of parasitoids making a choice divided by the number of parasitoids tested × 100) after each test. The attractiveness values for the two empty odor sources (Blank1 and Blank2) were averaged in Figures 2, 4 to facilitate the visual representation of the data.

Wind tunnel bioassays: Data from the bioassays with real B. rapa flowers were analyzed using a student t-test (α = 0.05) testing the hypothesis that the parasitoids spend more time flying toward a plant and foraging on a plant in presence of floral odors than in their absence. Replicate was used as a blocking factor. The proportion of parasitoids that attacked a caterpillar after landing on a plant in presence and in absence of floral odors was compared using a Chi-square test (α = 0.05). Data from the bioassays comparing the effects of different concentrations of synthetic floral odors on the attractiveness of infested B. rapa plants were analyzed with a paired t-test (α = 0.05) comparing the number of parasitoids that landed on the plant with floral odors and the plant without floral odors for each positional combination of every replicate.

Electrophysiology: All multiple comparisons were analyzed in R (R Core Team, 2013). Analysis of variance (α = 0.05) was used to test whether the electroantenogram responses were significantly different between the different odor sources. When the ANOVA was significant (P < 0.05) multiple comparisons (R-package: Multcomp) were made using Tukey-contrasts with Bonferroni correction.

There was a 43.5% overall decrease in the attractiveness of host-infested plants to parasitoids due to the addition of floral odors in 4-arm olfactometer tests (Figure 2). Specifically, floral odors caused a 44.9% decrease in plant attractiveness to D. rapae, 39.1% to C. glomerata, 73.7% to H. ebeninus, 42.2% to M. mediator, and 17.8% (non-significant decrease) to P. puparum. The pattern of parasitoid preferences was the same for four parasitoid species: parasitoids preferred the “herbivore” treatment more than the “herbivore + floral odors” treatment, and the two blanks were the least attractive treatments (D. rapae: df = 8, χ² = 65.5, P < 0.0001; C. glomerata: df = 11, χ² = 119.5, P < 0.0001; H. ebeninus: df = 8, χ² = 80.9, P < 0.0001; M. mediator: df = 8, χ² = 64.0, P < 0.0001). For the fifth parasitoid species, P. puparum the host and host + floral odors treatments were comparably attractive, and the two blanks were less attractive (df = 8, χ² = 27.8, P < 0.001) (Figure 2). The percentage participation (percentage of parasitoids choosing one of the olfactometer arms) of the parasitoids was as follows: 48% for D. rapae, 77% for M. mediator, 85% for C. glomerata, 91% for H. ebeninus, and 96% for P. puparum.

In the wind tunnel no-choice bioassays, C. glomerata females took 66% more time flying to the caterpillar-infested plant when in presence of a real inflorescence than in presence of an odorless decoy (df = 44, T = 1.73, P = 0.04) (seconds, mean + SE). After landing, the proportion of wasps attacking a caterpillar and the proportion of wasps leaving the plant without attacking a caterpillar were comparable in presence of a real inflorescence and in presence of a decoy (df = 1, χ² = 2.2, P = 0.14). Finally, for the subset of observations where parasitoids attacked a caterpillar after landing on a plant, time spend foraging on the plant between landing and attacking a caterpillar was comparable in presence of a real inflorescence and in presence of a decoy (df = 45, T = 1.2, P = 0.88) (Figure 3).

In our 4-arm olfactometer tests, the effect of floral odors on plant attractiveness was dose-dependent: it was negligible at concentrations of floral odors inferior to one IE, and became increasingly significant at higher floral odors concentrations (39.1% reduction in attractiveness at 1 IE, 60.3% at 2 IE, and 86.4% at 4 IE). At the two highest concentrations (2 and 4 IE), the attractiveness of the herbivore + floral odors treatment was as low as the blanks (Figure 4). Specifically, the patterns of parasitoid preferences were as follows: the herbivore and herbivore + floral odors treatments were the most attractive and the blanks were least attractive at concentration 0.25 IE (df = 9, χ² = 80.2, P < 0.0001); the herbivore treatment was the most attractive, the herbivore + floral odors treatment was not significantly different from the other treatments, and the blanks were the least attractive treatment at concentration 0.5 IE (df = 10, χ² = 41.4, P < 0.0001); the herbivore treatment was the most attractive, the herbivore + floral odors treatment was intermediate, and the blanks were the least attractive at concentration 1 IE (df = 11, χ² = 119.5; P < 0.0001); the herbivore treatment was the most attractive and the herbivore + floral odors and the blanks were comparably less attractive at concentration 2 IE (df = 7, χ² = 55.87, P < 0.0001) and 4 IE (df = 7, χ² = 123.3, P < 0.0001) (Figure 4).

In a wind tunnel setting, the effect of floral odors on plant attractiveness to parasitoids was also dose-dependent. Indeed, plants with and without floral odors were similarly attractive to parasitoids when the concentration of floral odors was inferior or equal to 1 IE (df = 19, T = 0.29, P = 0.39 for concentration 0.5 IE; df = 19, T = 0.48, P = 0.68 for concentration 1 IE), and plants without floral odors were more attractive than plant with floral odors when the concentration of floral odors was superior to 1E (df = 19, T = 2.26, P = 0.02 for concentration 2 IE; df = 19, T = 2.44, P = 0.01 for concentration 4 IE) (% attraction, mean ± SE), with a reduction in attractiveness due to the addition of floral odors of 46% for concentration 2 IE and 53% for concentration 4 IE (Figure 5).

The three B. rapa leaf volatiles, benzyl nitrile, (E)-2-hexenal, (Z)-3-hexenylacetate and a mixture of the three compounds (leafmix, a proximate for a leaf volatile blend) elicited significantly higher EAG responses in C. glomerata females than dichloromethane alone (F₄,₆₀ = 36.5, P < 0.001, Figure 6A). Pairwise comparisons between the four odor stimuli indicated that the leafmix did not elicit higher EAG responses, except compared to (Z)-3-hexenylacetate (Tukey post hoc, P < 0.05). In order to test if floral volatiles of B. rapa affect the EAG responses of C. glomerata females to leaf volatiles, p-anisaldehyde, α-farnesene and phenylacetaldehyde were presented in combination with the leafmix. While p-anisaldehyde on its own did not elicit significantly higher EAG responses compared to the control the compound significantly increased the EAG response of C. glomerata females to the leafmix (F₃,₅₂ = 68.3, P < 0.001, Tukey post hoc, P < 0.05, Figure 6B). The same pattern was observed with α-farnesene, with the compound not eliciting significantly different EAG responses singly compared to the control but increasing the EAG response to the leafmix (F₃,₅₂ = 65.9, P < 0.001, Tukey post hoc, P < 0.05, Figure 6C). Phenylacetaldehyde elicited significantly higher EAG responses when presented singly compared to the solvent control and increased the EAG response of C. glomerata females to the leafmix as well (F₃,₅₂ = 94.4, P < 0.001, Tukey post hoc, P < 0.05, Figure 6D).

The flower mix elicited significantly higher EAG responses when presented singly compared to the solvent control and increased the EAG response of C. glomerata females to the leafmix as well (F₃,₅₃ = 76.6, P < 0.001, Tukey post hoc, P < 0.05, Figure 6E). While headspace collections from herbivore infested B. rapa leaves elicited significantly higher EAG responses compared to the control, the combination with the flower mix elicited similar EAG responses as the floral mix alone (F₃,₅₂ = 58, P < 0.001, Tukey post hoc, P < 0.05, Figure 6F). The same pattern was observed with solvent extracts of herbivore-infested B. rapa leaves, with the blend eliciting significantly higher EAG responses compared to the control but not eliciting different EAG responses when added to the floral mix compared to the floral mix alone (F₃,₅₁ = 7.32, P < 0.001, Tukey post hoc, P < 0.05, Figure 6G).