Olfactory perception and behavioral effects of sex pheromone gland components in Helicoverpa armigera and Helicoverpa assulta

Two sympatric species Helicoverpa armigera and Helicoverpa assulta use (Z)-11-hexadecenal and (Z)-9-hexadecenal as sex pheromone components in reverse ratio. They also share several other pheromone gland components (PGCs). We present a comparative study on the olfactory coding mechanism and behavioral effects of these additional PGCs in pheromone communication of the two species using single sensillum recording, in situ hybridization, calcium imaging, and wind tunnel. We classify antennal sensilla types A, B and C into A, B1, B2, C1, C2 and C3 based on the response profiles, and identify the glomeruli responsible for antagonist detection in both species. The abundance of these sensilla types when compared with the number of OSNs expressing each of six pheromone receptors suggests that HarmOR13 and HassOR13 are expressed in OSNs housed within A type sensilla, HarmOR14b within B and C type sensilla, while HassOR6 and HassOR16 within some of C type sensilla. We find that for H. armigera, (Z)-11-hexadecenol and (Z)-11-hexadecenyl acetate act as behavioral antagonists. For H. assulta, instead, (Z)-11-hexadecenyl acetate acts as an agonist, while (Z)-9-hexadecenol, (Z)-11-hexadecenol and (Z)-9-hexadecenyl acetate are antagonists. The results provide an overall picture of intra- and interspecific olfactory and behavioral responses to all PGCs in two sister species.


Results
Electrophysiological responses of OSNs in antennal sensilla responding to PGCs. Based on the response profiles of their respective OSNs, the sensilla activated by PGCs in the 30-60 annuli of antennal flagella were first classified into A, B, C types consistently with previous reports 10,11 . The latter two types were further categorized into B1 and B2, and C1, C2, and C3, respectively (Figs 1 and 2).  The type A sensilla only responded to Z11-16:Ald (Figs 1A,B and 2A,B). The subtype B1 responded to Z9-14:Ald only, while subtype B2 responded to Z9-14:Ald as well as to an alcohol (Z11-16:OH in H. armigera and Z9-16:OH in H. assulta) (Figs 1C-F and 2C-F).
The relative abundances of A, B1, B2, C1, C2, and C3 sensilla in the male antennae were found to be remarkably different in the two species (84.5%, 1.7%, 3.1%, 4.3%, 6.0%, 0.5% in H. armigera, and 16.7%, 4.4%, 2.2%, 61.4%, 4. 4%, 10.8% in H. assulta) (Fig. 3). In H. armigera, the type A sensilla dominate, being 7.9 times more abundant than those of type C. In H. assulta, instead, type C sensilla are most represented, 4.6 times those of type A. Type B sensilla are a minority in both species (Fig. 3). The dose-response curves of OSNs in the main types and subtypes of sensilla of the two species are shown in Fig. S1.
Topographical expression pattern of PRs. In situ hybridization results show the expression sites of HarmOR13 and HassOR13, HarmOR6 and HassOR6, as well as HarmOR14b and HassOR16 in the 30-60 annuli of male antennal flagella (Fig. 4). The signals were found to be annular, indicating that the riboprobes were hybridized with RNA that is perinuclearly distributed (Fig. 4). OR-positive cells were only detected on the side In the C type sensilla, the total number of spikes from two co-localized neurons was counted. Columns with the same letters are not significantly different at P < 0.05. Paraffin oil was used as control.
bearing sensilla, not on that covered with scale. The distribution of stained cells corresponds to the distribution of trichoid sensilla on the antennal surface. We selected antennae of three males for each species from three independent experimental batches to count the number of OR-positive cells from three kinds of sections. The average numbers of OR-positive cells from various annuli (varying from 3 to 9) of three kinds of sections on each antenna are graphically compared in Fig. 5. In the male antennae of H. armigera, the number of HarmOR13-expressing OSNs is 6.2 and 31.7 times higher than those of HarmOR14b and HarmOR6, respectively (Fig. 5). In the antennae of male H. assulta, the expression levels of HassOR6 and HassOR16 are similar, and about 1.6 times higher than that of HassOR13 (Fig. 5). When comparing the two species, HarmOR13-expressing OSNs in H. armigera are 3.7 times more abundant than those expressing HassOR13 in H. assulta. The ratio is reversed for the numbers of OSNs expressing OR6, 13.9 times higher in H. assulta than in H. armigera. Finally, the number of HassOR16-expressing OSNs in H. assulta is about 2.6 times higher than that of OSNs expressing HarmOR14b in H. armigera. Spatial representation of PGCs in MGC glomeruli. Based on the AL atlas of H. assulta and H. armigera 27,30 , there are three MGC subunits in each species: cumulus, Dm-p, and Dm-a in H. armigera, and cumulus, DM and ventral unit in H. assulta. We identified the MGC subunits according to their positions and monitored their activities after stimulating the antenna with 100 μg of pheromone gland components (Fig. 6).
We had previously reported that Z11-16:Ald and Z9-16:Ald evoked responses in the cumulus and Dm-p in H. armigera respectively, while in H. assulta they elicited activities in the ventral unit and cumulus respectively 8 .
Here we use the evoked area of these two components as landmarks within the MGC to determine the relative position of activities elicited by the other components. We found that Z11-16: OH and Z11-16: Ac both activated Dm-a in H. armigera, while Z9-16:OH and Z11-16:OH both evoked responses in DM in H. assulta (Fig. 6).
Behavioral effects of PGC blends on male attraction in wind tunnel. Based on the relative amounts of PGC in pheromone gland extracts of the two species, as previously studied in our laboratory 3 , ternary and quaternary blends were prepared by adding one or two components to the mixtures of Z9-16:Ald and Z11-16:Ald  (Table 2), and the behavioral responses of males of the two species to each blend were studied in wind tunnel. (Fig. 7 and Table 2).
In H. armigera, addition of Z11-16:OH or Z11-16:Ac to the two-component blend inhibited a series of attractive steps during male pheromone-directed behavior, while Z9-16:OH and Z9-16:Ac had no significant effect (Fig. 7A,C).
In H. assulta, addition of Z11-16:Ac alone or together with Z9-16:Ac to the binary pheromone blend significantly increased males' landing and displaying the copulation behavior (Fig. 7D). However, addition of Z9-16:OH or Z9-16:Ac significantly reduced the flight, upwind and close behaviors to the binary blend (Fig. 7B,D). Adding Z11-16:OH only reduced the landing behavior of H. assulta males (Fig. 7B).

Discussion
Sex pheromone communication in moths allows males to find conspecific females over a long distance and also serves as a primary barrier of premating isolation among moth species 31,32 . Besides the principal pheromone components Z11-16:Ald and Z9-16:Ald, other chemicals in pheromone glands also play crucial roles in H. armigera and H. assulta. We comparatively study behavioral effects of PGCs and the olfactory coding mechanisms for PGCs in these two species, using state-of-the-art molecular, physiological and behavioral techniques to shed new light onto chemosensory bases of reproductive isolation between these species. Peripheral coding relative to these components in male is more complicated than previously suspected. Of the three types of pheromone sensitive sensilla (A, B and C) previously reported 10,11,13 , type B can be further classified into two subtypes and type C into three subtypes in both species. The abundance of OSNs expressing each of the 6 PRs in male antennae represents a guide to assign each PR to the OSNs in different types of sensilla. We found that Z11-16:OH and Z11-16:Ac act as antagonists for H. armigera in pheromone communication, while in H. assulta Z11-16:Ac is an agonist, and Z11-16:OH as well as Z9-16:Ac are antagonists. We also proved that Z9-16:OH acts as an antagonist for H. assulta 1 .

Functional classification of the sensilla and their sensory neurons. The types of sensilla identified
in H. armigera and H. assulta males seem similar to, but not as stereotypical as those described in several other Heliothine species before 10,11 . The A type sensilla always contain an OSN tuned to Z11-16:Ald, while the B and C type sensilla can be further classified into subtypes, B1, B2, C1, C2, and C3 (Figs 1 and 2). Such flexibility in the same types of sensilla implies that different groups of ORs could be expressed in the subtypes of B or C type sensilla. Sensilla of types B and C could be grouped together as both are mainly tuned to Z9-14:Ald, but may also respond to other chemicals. The presence of subtypes could be a phenomenon of chemosensory adaptation in speciation. The versatility of sensilla responding to a range of stimuli could increase the capability of males to detect other semiochemicals present in the environment, and may provide better adaptation to changes in pheromone components 33 . Such a divergence of chemical channels between the two sympatric species may be caused by either reinforcement or communication interference, but how changes in these communication systems evolve remains elusive 34-38 . Comparison of PR expression level and abundance of different types of sensilla. The correlation analysis between the population sizes of the different types of sensilla and the numbers of OSNs expressing the associated ORs can help assigning different ORs to OSNs and types of sensilla in the olfactory system. HarmOR13 and HassOR13 when expressed ectopically in oocytes of Xenopus or in insect cells 24   males of H. armigera are 3.7 times more abundant than in H. assulta, in agreement with the ratio of type A sensilla between the two species statistically (5.1, Fig. 8 14 . In the present study, we find that the HassOR6-positive cells as well as the HassOR16-positive cells significantly outnumber the OR13-positive cells in antennae of male H. assulta, but their population ratio of only 1.6 : 1, does not match the population ratio (4.6 : 1) between sensilla of types C and A in male H. assulta (Fig. 8). It is thus reasonable to suggest that HassOR6 and HassOR16 are expressed in OSNs of only some of the type C sensilla, and HassOR6 is not the PR tuned to the major pheromone component Z9-16:Ald in H. assulta.
HarmOR14b and HassOR16 have been found to be tuned to Z9-14:Ald 18 . We hypothesized that the number of HarmOR14b-and HassOR16-positive cells would be mirrored by the abundance of sensilla containing OSNs responsive to Z9-14:Ald, i.e. the sum of type B and C sensilla. The in situ hybridization results showed that the expression ratio of HarmOR13 to HarmOR14b in H. armigera is 6.2:1, in agreement with the ratio of type A to type B + C sensilla (5.5:1, Fig. 8). However, the ratio of HassOR13 to HassOR16 is 1:1.6 in H. assulta, much higher than the ratio of the related sensilla, 1:5.0 (Fig. 8). Thus, the expression level of HarmOR14b matches the abundance of OSNs tuned to Z9-14:Ald, i.e. a sum of type B and type C sensilla in H. armigera, but a similar model cannot be applied to HassOR16 in H. assulta.
To reliably characterize the different types of sensilla, the function and expression pattern of PRs they house need to be determined. Using specific riboprobes for each OR, in situ hybridization provides strong evidence to accurately define the topographical distribution of OSNs and sensillum types of interest in the antennae, especially for non-model animals like Helicoverpa species. Response profiles and distribution of OSNs in sensilla of different types (subtypes) provide the basis for further functional studies of ORs. It will be worthwhile to revisit the ligand profile of the various PRs, especially for OR6 and OR16 in this system.
Behavioral valence of pheromone gland components. In our wind tunnel experiment, Z11-16:Ac was shown to act as an agonist for H. assulta males and as antagonist for H. armigera when added to the binary sex pheromone blends (Fig. 7D,C). Whether Z11-16:Ac works as a third sex pheromone component in H. assulta still requires further field tests. Our additional electroantennographic (EAG) experiments demonstrated that Z11-16:Ac is detected by the H. assulta antenna (Fig. S2). In ensuing SSR experiments we also found OSNs in several C1-type sensilla weakly responding to higher doses of Z11-16:Ac (Fig. S3 ). However, calcium imaging could not visualize representative areas of Z11-16:Ac activation in the MGC. Z11-16:Ac might not activate a single olfactory pathway, but may be involved in detecting the sex pheromones. In further studies we will add this compound to each sex pheromone component and their blends to see if it affects responses of the related OSNs and spatial representative patterns in ALs to sex pheromones. Z11-16:OH and Z9-16:OH are common PGCs in pheromone gland extracts of Heliothine species 3,39-42 . Z11-16:OH is a minor sex pheromone component of Heliothis subflexa 43 but an antagonist of many Heliothine species 2,44,45 . Our wind tunnel data indicate that Z11-16:OH and Z9-16:OH act as antagonists for males of H. assulta, in agreement with field data reported by Cork et al. 1 . Z11-16:OH is also proven to be an antagonist to male H. armigera. We find the OSNs in type C1 sensilla are tuned to Z9-16:OH and Z11-16:OH in H. assulta, which is different from the previous electrophysiological recordings done by Berg and Mustaparta 9 , in which no OSN responding to Z11-16:OH was found. Considering that the insect population in the previous study was from Korea, the difference might be related to this fact.
Cork et al. reported that addition of Z9-16:Ac and Z11-16:Ac to the main sex pheromone components decreased the trapped number of male H. assulta in China 1 . In our study, Z9-16:Ac is also found to be an antagonist. However, when both Z11-16:Ac and Z9-16:Ac were mixed with sex pheromone blends, the agonistic effect of Z11-16:Ac was so strong that it overshadowed the antagonistic effect of Z9-16:Ac (Fig. 7D). Our previous study showed that Z9-14:Ald acted as an agonist when minor amounts were added to the binary pheromone blends of H. armigera, but as an antagonist at higher levels. For H. assulta, Z9-14:Ald acted as an antagonist especially when presented with higher concentration 6,7 . These results indicate that these PGCs, Z9-16:OH, Z11-16:OH, Z9-16:Ac, Z11-16:Ac, and Z9-14:Ald also play roles in species isolation of related species of Heliothine moth.
In this study, we elucidate the behavioral valence of long overlooked additional sex pheromone gland components in the two species. Combining the results of the in situ hybridization, SSR, and calcium imaging in this study and from Wu et al. 7 , we can provide an overall picture of intra-and interspecific behavioral and olfactory  Fig. 8. We identify different types and subtypes of sensilla containing OSNs responsive to PGCs and evaluate their abundance. We also confirm that HarmOR13 and HassOR13 are located in the type A sensilla and are involved in detecting Z11-16:Ald, a pheromone component of the main binary mixture in both species. Furthermore, based on the agreement between the ratios of sensillum numbers and OR abundance, we suggest that OR6 is just expressed in OSNs of some C type sensilla, while the PR tuned to the major pheromone component Z9-16:Ald in H. assulta need to be further identified.

Materials and Methods
Insects. H. armigera and H. assulta were originally collected as larvae in tobacco fields in Zhengzhou, Henan province of China, and successive generations were maintained in the laboratory under a 16 L: 8D photoperiod cycle at 26 ± 1 °C and 55-65% relative humidity. The larvae were reared on two kinds of artificial diets whose main component is wheat germ 46 . Pupae were sexed and males and females were put into separate cages for eclosion. After emergence, moths were fed with 10% honey in water. 2-4 days old virgin males were used in wind tunnel and in situ hybridization experiments. 2-6 days old virgin males were used for calcium imaging and electrophysiological studies. In SSR experiments we used tungsten electrodes to record sensilla in the front of the antennae, and glass electrodes for the trichoid sensilla on the lateral sides of the antennae. We selected 140 sensilla from 23 antennae in H. armigera and 120 sensilla from 20 antennae in H. assulta, which were responsive to PGCs and located at 30-60 annuli of the antennae. In both cases, half of them were probed with glass electrodes and half with tungsten electrode. Three replications were performed for each species, and thus in total 420 sensilla in H. armigera and 360 sensilla in H. assutla were recorded. 100 μg (10 μL of 10 μg/μL solutions) of the seven compounds were used to determine sensilla types, while 100 ng, 1 μg, 10 μg, 100 μg, 1 mg were used for dose-response curves. Paraffin oil was adopted as control stimulus.
Single sensillum recording (SSR). The insect was placed inside a 1 mL disposable Eppendorf pipette tip with the narrow end cut to allow the head and the antenna to protrude. The head and antenna were immobilized with dental wax under a stereomicroscope. When recording with the tungsten electrode, the reference electrode was inserted into a compound eye, and the sharpened recording electrode was inserted into the base of a single sensillum in the front area of the antenna. When using the glass electrode for recording, sensilla on the lateral side of the antenna had their tips cut off using custom sharpened forceps, and inserted into glass capillaries filled with receptor lymph saline 47 . An Ag-AgCl electrode was placed in the glass micropipette to record action potentials of the receptor neurons. The recorded signals were then amplified through a IDAC interface amplifier (IDAC-4, Syntech, Germany). The software Autospike, version 3.4 (Syntech, Germany), was used to store and analyze data.
A continuous stream of purified and humidified air was directed on the antenna (12.5 mL/s) from the outlet of a steel tube (i.d. 6 mm, length 15 cm), positioned 2 cm from the antenna. Test odors were injected into the air stream using a stimulus flow controller (CS-55, Syntech, Germany), which generated 200 ms air pulses through the odor cartridge at a flow rate of 10 mL/s, and a compensating air flow was provided to keep a constant current. The odor cartridge was made with 10 μL stimulus loaded on a filter paper strip (0.7 cm × 2.5 cm) in a Pasteur pipette (15 cm long).

Topographical expression pattern of PRs. To visualize the antennal expression sites of genes encoding
HarmOR13 and HassOR13, HarmOR6 and HassOR6, and HarmOR14b, and HassOR16, we used in situ hybridization. We prepared slices of the 30-60 annuli of antennal flagella from the proximal end, making a vertical deep section, a horizontal deep section and a horizontal superficial section. Three replications were performed for each species.
In situ hybridization. The protocol for in situ hybridization was adapted from Krieger et al. 48 . Briefly, fresh antennae were embedded in JUNG tissue freezing medium (Leica Microsystems, Germany). Sections of 12 μm were prepared with a Leica CM 1950 microtome at − 22 °C, and, mounted on Superfrost Plus Glass Slides (Electron Microscope Science, Hatfield, USA). Slides were dried in air for 10 min, followed by fixing with  Calcium imaging. The calcium imaging method was adopted from Galizia and Vetter 49 . Preparation of moths and optical recording were performed as described previously 7,8 . Briefly, virgin male moths were restrained in plastic tubes and fixed with dental wax. After dissecting and exposing the brain, we used a calcium-sensitive dye, CaGR-2-AM (Molecular Probes, Eugene, OR, USA) to stain the antennal lobes for one hour, and then thoroughly rinsed with Ringer solution. Imaging data were collected using a Till-Photonics imaging system (Till Photonics, Germany). The antennal lobe was illuminated at 475 nm. Stimulation started at frame 12 and lasted 500 ms in the recording sequence of 40 frames. For the false color images, the relative calcium change of each frame was calculated as relative changes in fluorescence (ΔF/F) by MATLAB software. 100 μg (10 μL of 10 μg/ μL solutions) of six compounds (Z11-16:Ald, Z9-16:Ald, Z9-16:OH, Z11-16:OH, Z9-16:Ac, and Z11-16:Ac) was used as stimuli. Stimuli were applied in random order. Paraffin oil was used as control.
Behavioral effects of PGCs. A wind tunnel was used to investigate the role of PGCs in the attraction behavior of H. armigera and H. assulta. A series of ternary and quaternary blends (Table 2) prepared with one or two more components added to the binary mixture (Z9-16:Ald and Z11-16:Ald) based on the PGC ratios 3 were used as stimuli. Because Z9-16:Ac was found to be present at a quite high concentration in H. assulta female pheromone gland extracts (1.8 times that of the basic pheromone blend ) 3 , we also used it at a lower dose (15% of the basic pheromone blend) according to concentrations identified in the related species H. subflexa 42 . The binary mixture was adopted as the control, and hexane was used as solvent. In each experiment, 10-20 male moths were tested and at least three replicates were performed. The order of treatments was randomized.
Wind tunnel experiments. The males were tested during scotophase (4-6 hr) in a Plexiglas wind tunnel  Data analyses. In SSR, spike frequencies (spikes/s) were calculated by counting the number of spikes during the first 200 ms of the response. For both non-responding and poorly responding OSNs, spikes were counted during the same 200 ms post-stimulus interval. Data of SSR were analyzed by the one-way ANOVA for analysis of variance, and the least significant difference (LSD) test was used for means multiple comparisons. To compare the number of olfactory receptor neurons on each antenna, neurons were counted from 3-9 flagellomeres. The number of olfactory receptor neurons between species, and between sections was compared by two-way ANOVA based on the Student-Newman-Keuls Tests. The ratios of different types of sensilla and OSNs expressing associated ORs were compared using Χ 2 2 × 2 test of independence with Yates' continuity correction. In calcium imaging, data were acquired by the software Till-vision (Till photonics) and further analyzed by software ImageJ (NIH, USA) and custom-made programs in MATLAB (The Math Works, Inc). One-way ANOVA and LSD were used to compare the response intensity of different glomeruli in the MGC of tested compounds. In wind tunnel experiment, percentages of males in performing sequential behaviors were subjected to Χ 2 2 × 2 test of independence with Yates' continuity correction. The level of significance was set as P < 0.05.