Volatiles from Aquilaria sinensis damaged by Heortia vitessoides larvae deter the conspecific gravid adults and attract its predator Cantheconidea concinna

The effects of induced plant responses on herbivores are categorised as direct, by reducing herbivore development, or indirect, by affecting the performance of natural enemies. Here, we investigated a tritrophic system, which included the herbivore Heortia vitessoides, its host plant Aquilaria sinensis, and its predator Cantheconidea concinna. Herbivore-damaged A. sinensis plants released significantly greater amounts of volatiles than undamaged and mechanically damaged plants, with an obvious temporal trend. One day after initial herbivore damage, A. sinensis plants released large amounts of volatile compounds. Volatile compounds release gradually decreased over the next 3 d. The composition and relative concentrations of the electroantennographic detection (EAD)-active compounds, emitted after herbivore damage, varied significantly over the 4-d measurement period. In wind tunnel bioassays, mated H. vitessoides females showed a preference for undamaged plants over herbivore and mechanically damaged A. sinensis plants. In Y-tube bioassays, C. concinna preferred odours from herbivore-damaged plants to those from undamaged plants, especially after the early stages of insect attack. Our results indicate that the herbivore-induced compounds produced in response to attack by H. vitessoides larvae on A. sinensis plants could be used by both the herbivores themselves and their natural enemies to locate suitable host plants and prey, respectively.

caterpillars have defoliated large areas of forest in southern China and caused significant economic losses. Apart from the heavy use of pesticides, there is no known effective method for controlling this pest.
Su 13 suggested that the young leaves of A. sinensis were the sole emitters of VOCs attractive to H. vitessoides females seeking oviposition sites. We previously identified and compared VOCs from young and old A. sinensis leaves that potentially attract H. vitessoides. We also tested the behavioural responses of H. vitessoides to synthetic blends of these VOCs in wind tunnel and field tests, and established a relationship between leaf age preference and host plant recognition in H. vitessoides 12 . We found qualitative and quantitative differences between the odour profiles of young and old leaves. Wind tunnel and field tests confirmed that a nine-component mixture based on young leaves (comprised of hexanal, limonene, 2-hexanol, octanal, (Z)-3-hexenyl acetate, (Z)-3-hexen-1-ol, nonanal, decanal, and 2,6,10-trimethyl-dodecane at a ratio of 2:16:9:4:63:100:13:10:5) attracted significantly more moths than the three component mixture based on old leaves (comprised of nonanal, decanal, and 2,6,10-trimethyl-dodecane in a ratio of 11:14:26). The volatile signals from young A. sinensis leaves allowed H. vitessoides females to discriminate suitable larval hosts from the background chemical environment, and guided orientation of flights towards these plants for oviposition.
In a more recent study, we found that female adult oviposition on young A. sinensis leaves was reduced in response to damage caused by H. vitessoides larvae. In other words, female adults preferred to lay eggs on the healthy, intact young leaves. In addition, many natural enemies of H. vitessoides larvae, including Cantheconidea concinna, are found on herbivore-damaged A. sinensis plants 15 . To date many studies have shown that HIPVs can either attract or repel the same or different species of herbivores 16,17 , and even attract their natural enemies 5 . For instance, Tetranycbus evansi adults were more attracted to plants attacked by conspecific larvae than to undamaged plants in olfactometer experiments 18 . Caterpillar-induced nocturnal tobacco plant volatiles were found to repel ovipositing conspecific moths 16 . Kappers et al. 19 suggested that HIPVs play a very important role in plant defences against herbivores, both directly and indirectly, as cues that attract predatory and parasitic natural enemies of herbivores.
Here, we hypothesise that HIPVs emitted by A. sinensis significantly reduce herbivore oviposition and increase recruitment of their natural enemies, and ask: How do the moth H. vitessoides and its predatory enemy C. concinna respond to HIPV emissions from A. sinensis? We aimed to (1) identify and compare VOCs released by undamaged, mechanically damaged, and herbivore-damaged A. sinensis plants; (2) analyse the antennal and behavioural responses of mated H. vitessoides females to these volatile compounds; and (3) examine whether HIPVs emitted by A. sinensis affect the host-searching behaviour of C. concinna on a host-infested plant. In this study, we sought to elucidate the role of HIPVs emitted from herbivore-damaged plants and improve our understanding of how herbivore insects locate hosts and are located by their predators in a tree ecosystem.
PCA and hierarchical cluster analysis. Principal component analysis (PCA) clearly segregated the overall composition of the headspace volatile blends collected from the six plant treatments (Fig. 4). A scatter plot of the first and second principal components showed that principal component 1 was more discriminating than principal component 2. The two principal component axes accounted for 55.27% of the total variation in VOCs. The first PCA accounted for 37.50% and the second PCA accounted for 17.77%. PCA also segregated the volatile blends into three groups according to the behavioural effects on H. vitessoides females. Group A, comprised of the volatile blends detected in the undamaged and mechanically damaged   (Fig. 4). The number of compounds detected in this unattractive blend (Group B1) was exceptionally high, especially the terpenoids and green leaf volatiles (Table 1).
Hierarchical cluster analysis between-groups linkage was used to analyse the volatiles derived from six treatments, at a distance >5 and <20. They were divided into three clusters (Group A, B1, B2, Fig. 5). System clustering results were consistent with the PCA results.
Wind tunnel bioassays. All six EAD-active blends that were tested stimulated H. vitessoides female upwind flights and approaches to within 5 cm of the source (Fig. 7). Synthetic blends mimicking undamaged plant VOCs (A) had the strongest attraction to females in the wind tunnel; 38.89% of females flew over 120 cm upwind, and 21.11% arrived within 5 cm of the source. The number of female upwind flights elicited by synthetic blend A was significantly higher than flights elicited by the other blends (F = 25.900, DF = 6, P < 0.001). Blend B, containing five compounds identified in the headspace of mechanically damaged plants, was the second most attractive to females. This blend resulted in 27.78% of females flying upwind and 16.67% approaching the source.   Y-tube bioassays. In the dual-choice bioassay, C. concinna, a predator of H. vitessoides larvae, preferred VOCs from herbivore-damaged plants to those from undamaged plants (Fig. 8). The predator was attracted by the odour of 1-day herbivore-damaged plants (X 2 = 111.386, N = 30, P < 0.001), 2-day herbivore-damaged plants (X 2 = 81.820, N = 30, P < 0.001), and 3-day herbivore-damaged plants (X 2 = 50.000, N = 30, P < 0.001). The effect of 4-day herbivore-damaged plants odour was not significant (X 2 = 5.120, N = 30, P = 0.034). Results from the one-way ANOVA showed that olfactory response rates of C. concinna to the odours from A. sinensis plants differed among damage treatments (F = 4.04, P < 0.05). C. concinna adults showed a preference for recently herbivore-damaged plants (1-3 d of herbivore damage) over plants with 4 d of damage.

Discussion
Plants commonly respond to herbivore attacks by releasing HIPVs 6,20 . The production and release of HIPVs can directly and indirectly affect herbivore performance and mediate interactions with other community members. Thus, HIPVs act as signals to herbivores, their natural enemies, and neighbouring plants 6,8 . These components have been well described in agricultural ecological systems, such as maize 17 , rape, and cotton 21,22 . Herein, we studied a tritrophic system which includes the herbivore H. vitessoides, its host tree A. sinensis, and its predator C. concinna. Our studies indicate that previously described tritrophic interactions in agricultural crop systems also apply to a forest ecosystem.
HIPVs may discourage oviposition of herbivores on damaged plants and may, therefore, be beneficial in reducing herbivore density 23 . Our behavioural bioassays showed that mated H. vitessoides females preferred VOC blends mimicking healthy, undamaged A. sinensis plants to those containing VOCs emitted by herbivore-damaged plants. This suggests that H. vitessoides females detect and assess VOCs released by A. sinensis to locate suitable oviposition sites and avoid the unsuitable sites. Similarly, in a dual-choice test weevils have been shown to prefer undamaged clover leaves to weevil-damaged leaves 24 . In behavioural bioassays, alate Aphis gossypii preferred the odour from undamaged cotton seedlings to that from A. gossypii-infested plants 25 . Recognition and the ability to  20 . Female insects generally prefer healthy, intact plants as oviposition sites, as these are more likely to provide newly hatched larvae with enough food resources. Thus, this strategy reduces the strength of intraspecific food competition, and increases individual survival and population growth 26 . In a previous study, we showed that mated H. vitessoides females were more attracted to young leaves than to old leaves of A. sinensis plants, and suggest that the former provides suitably tender food for freshly hatched, delicate young larvae 12 .
Many studies have shown that HIPVs may act as an important signal for natural enemies to locate their host/ prey 27,28 . HIPVs are likely to act as important cues for natural enemies to locate damaged plants, and by extension, the herbivores attacking those plants, and, thus, may act as indirect plant defences 23,29 . Females laying eggs on undamaged plants also reduce the risk of parasitoid and predator attacks on larvae 29 . HIPVs are highly detectable and variable, and parasitoids and predators can distinguish these compounds to infer host suitability and even  (Table 2). Females were scored for upwind flights over 120 cm (white columns) and for approaching the source within 5 cm (black columns). Bars with the same colour and different letters were significantly different.  29,30 . Some volatile compounds are continuously emitted, while many others are only released when plants are attacked by herbivores or mechanically damaged 7 . This damage response can play a key role in mediating multitrophic plant-insect interactions 8,31 . These complex volatile compounds resulting from damage tend to be released in greater variety and quantities than those from intact, healthy plants 6,9 . Numerous studies have shown both large quantitative increases and qualitative changes in VOC emissions as a result of mechanical and herbivore damage 20,21 . In particular, obvious differences were found between VOCs from herbivore-damaged A. sinensis plants compared to undamaged or mechanically damaged plants; herbivore damage elicited the release of a greater variety of VOCs, and in far greater quantities than mechanical damage. Herein, using mechanically-damaged plants as one of the treatments was valuable and allowed us to compare the difference between herbivore feeding and mechanical wounding to characterize a set of special HIPVs.

Amount loaded on rubber septum in six treatments b (μg)
In general, the majority of HIPVs belong to green leaf volatiles (GLVs-C 6 aldehydes, alcohols, and their esters), terpenoids, aromatics, and amino acid volatile derivatives 32 . Similarly, these classes detected in our study showed an increased emission pattern in response to herbivore feeding, most of which belonged to the increased groups (IG). However, HIPVs varied considerably over time in response to damage 33 . Different chemical classes have different change rhythms. Some GLVs are produced immediately after initial damage by the larvae of herbivores; the production and emission of these GLVs occurs almost instantaneously during the initial 1-2 hours after herbivore damage. While other chemicals, such as terpenoids, are released several hours after herbivore damage or the following day 21,34 ; these HIPVs are synthetized de novo and emitted later 33 . Some studies have shown that there are different biosynthesis pathways, including autolytic oxidative breakdown of membrane fatty acids or nonmevalonate for different chemical classes.
All of the quantitative and qualitative changes in VOC emissions are always short-lasting. Once the damage ceases, the emission of these VOCs drops rapidly, making this a highly dynamic process 21,22,35 . In our study, the quantity and diversity of several HIPV classes, including GLVs or terpenoids, declined gradually soon after the attacks on the plant stopped. Simultaneously, several other classes, including aldehyde, alcohol, and ketone, almost disappeared after the initial damage, and recovered again during the subsequent 2-4 day sampling dates.
The rhythm of these changes in different chemical classes is consistent with the behaviour results in the wind tunnel. The increase in GLVs [(Z)-3-hexenyl acetate, (Z)-3-hexen-1-ol] and terpenoids [(Z)-β-ocimene] and the decrease in aldehydes [octanal, nonanal, and decanal] and alcohols [2-decen-1-ol] during the initial damage correspond to less attraction of H. vitessoides females to the synthetic blends that mimic the initial damaged plant VOCs. Conversely, the decrease of GLVs and terpenoids and the recovery of aldehydes and alcohols during the 2-4 day sampling dates correlated with the recovered attraction of females in the wind tunnel for synthetic blends that mimic the 2-4 day damaged plants. These data imply that HIPVs, including GLVs and terpenoids, are repellent for female H. vitessoides. However, some aldehydes and alcohols are attractive to female H. vitessoides. Most studies have found that adult moths are repelled by host volatiles released by conspecific larval feeding 16,17,36 , although this is by no means universal 37 . Clearly, more behavioural studies are needed to assess the impact of these various volatile components on insect behavior 38 .
Behavioural responses to VOCs from plants do not always mirror electrophysiology results in some insects. Not all GC-EAD active components are attractive in the behaviour assay and may act as a repellent. Wee et al. 24 reported that lemon leaf volatiles elicited electrophysiological responses in weevils, but weevils were repelled by these compounds in behavioural bioassays. In our study, the three GC-EAD components [(Z)-3-hexenyl acetate, (Z)-3-hexen-1-ol, and (Z)-β-ocimene] were repellent in the wind tunnel assay. The amplitude of EAD is also not consistent to its behavioural activity. In our study, mated H. vitessoides females responded consistently and strongly to VOC blends mimicking headspace collections from herbivore-damaged A. sinensis plants in electroantennographic tests, but behavioural responses to these compounds in wind tunnel bioassays were weaker. Some studies have found discrepancies between electrophysiological and behavioural responses to VOCs in herbivores 24,39 . For example, the highest EAD responses from Pandemis heparana moths were obtained with the terpenes, linalool and DMNT, which are often key volatiles in herbivore deterrence 36,40 . The addition of the four compounds that elicited the smallest antennal responses resulted in improved levels of upwind flight of female grape berry moths (Paralobesia viteana) 41 .

Insects. H. vitessoides eggs were provided by Huazhou Green Life Co. Ltd (Guangdong, China) from the A.
sinensis fields in the Chinese Medicinal Material Production Base (CMMPB). Newly hatched larvae were massreared for three instars in glass containers (diameter: 20 cm, height: 30 cm), and then separately transferred to smaller glass containers (diameter: 3 cm, height: 10 cm) with fresh A. sinensis leaves. Adults were provided with a 10% sugar-water solution on water-soaked cotton. All insects were reared in a climate-controlled room (25 ± 2 °C, 70 ± 5% RH, L16:D8).
Late-third to fourth-instar H. vitessoides larvae were used in the tests to induce the HIPVs. Larvae were starved overnight prior to all experiments to encourage active feeding immediately after being placed on plants.
In behavioural assays, mated females were used. To obtain mated females, the newly emerged adult couples were placed in a cage (200 × 200 × 200 cm) with A. sinensis at a 2:1 ratio of male:female to ensure mating. Only females laying eggs were used in the wind tunnel bioassays. None of the females used in tests had previously been exposed to any of the tested odours and each was used only once 42 .
C. concinna nymphs, the primary predator of H. vitessoides larvae, were collected from the same fields as the original H. vitessoides eggs and reared in smaller glass containers (diameter: 3 cm, height: 10 cm) under the same conditions as their hosts. Late-third to fourth-instar H. vitessoides larvae were provided as food to C. concinna nymphs and adults. C. concinna adults used in Y-tube trials were 1-2 d old. All adults were starved overnight prior to trials and none had been exposed to any host plant or prey odour. For mechanical-damage treatment 2, the plants were immediately placed inside the volatile collection system (see below) after mechanical damage. Collections were performed for 8 h. For the herbivory treatments 3-6, each entire plant was individually wrapped in gauze mesh to prevent the larvae from escaping during the experiment. After 8 h of feeding, the gauze mesh and larvae were removed from the infested plant. Volatile collection experiments began after the removal of the larvae. Collections were conducted for 8 h every day for a successive 4 d period, corresponding to the four treatments, 3-6. Each collection was made at the same time (20:00-04:00) each day, corresponding to the oviposition peak period of H. vitessoides. The plants were weighed immediately after collection. Each treatment was repeated six times, with different plants, cut damage, and test larvae.

Plant volatile collections.
We used a headspace collection system to collect headspace volatiles from plants. Living test plants were placed in a clean roasting bag. The bag was sealed around the plant stem with a self-sealing strip about 20 cm above soil-height 34 . Humidified, charcoal-filtered air was pulled through the bag with a pump (Beijing Institute of Labour Instruments, China) at 300 ml·min −1 and passed over an adsorbent cartridge. The adsorbent cartridge was a 0.5 × 10 cm glass column containing 50 mg of adsorbent (80/100 mesh, Supelco, Bellefonte, PA, USA). The Porapak Q (50 mg, 80-100 mesh, Supelco, Bellefonte, PA, USA) was held between plugs of glass wool. Each sample was aerated for 8 h. Volatiles were eluted from the adsorbent cartridge with 500 μl redistilled hexane at room temperature. An internal standard of 0.5 μg of benzaldehyde (99%, Fluka Production) was added to the extract for chemical quantification 43 . The final extracts were reduced to 50 μl using a slow stream of nitrogen and then subjected to gas chromatography mass spectrometry (GC-MS) and gas chromatography-electroantennographic detection (GC-EAD). If not used immediately, extracts were stored in glass vials at −18 °C until use.

GC-MS.
Headspace extracts were analysed with an Agilent Technologies 6890 N gas chromatograph linked to a 5973 mass spectrometer (Palo Alto, CA, USA) with a polar DB-Wax or non-polar DB-5 fused silica column (both 30 m × 0.25 mm × 0.25 μm; J&W Scientific, Folsom, CA, USA). The column oven temperature was held at 50 °C for 1 min, raised to 120 °C at 3 °C·min −1 , and then increased to 240 °C at 10 °C·min −1 for 10 min. Helium (1.0 ml·min −1 ) was used as the carrier gas. Splitless injection (2 μl) was used with an injector temperature of Wind tunnel bioassays. Assessments of the attractiveness of synthetic chemical blends to mated H. vitessoides females were carried out in a plexiglas wind tunnel (flight section: 200 × 60 × 60 cm). Incoming air was filtered through activated charcoal and was blown by a horizontal fan at 0.3 m·s −1 at the point of release of the moths. The upwind and downwind ends of the tunnel were covered with gauze to prevent escape of moths 41,46 . All bioassays were performed from 20:00 to 04:00, which is the oviposition peak period of H. vitessoides 12 . During the tests, temperature and relative humidity of the wind tunnel were kept at 25 ± 2 °C and 75 ± 5%, respectively.
On the basis of the results of the GC-EAD analyses, VOCs from the six different treatments of A. sinensis eliciting antennal responses in the female of H. vitessoides were formulated in blends for the wind tunnel tests. Six blends of synthetic compounds were prepared in the ratios of GC-EAD-active VOCs as emitted by the corresponding treatments ( Table 2). Chemicals were diluted with redistilled hexane (Sigma-Aldrich, St. Louis, MO, USA). For each blend, the formulations contained 0.5 mg of the most abundant compound and the others compounds were added in the same proportion as in the natural volatile mixture. Preliminary tests confirmed that these concentrations were adequate to elicit moth responses in the wind tunnel. The blends were released into the wind tunnel by means of a green rubber septum. A septum treated with hexane only and no scent lure served as a control 12,43,47 .
Before the bioassays, the synthetic lures were loaded in individual rubber septa respectively. Each septum loaded with one of the test samples was placed in the centre of the upwind end of the tunnel (30 cm from upwind end), affixed to a holder, and used only once per day. In order to reduce the experimental error between different blends or same blends with different replications, all odour blends were deployed based on the same criteria. After each treatment, the flight section of the wind tunnel was washed with hexane, and then dried with an electric hair drier (HP 8200; Philips, Zhuhai, China) 43,48 .
One hour before the trial, all mated females were transferred to the wind tunnel room and allowed to acclimate to the conditions. Test females were introduced into the downwind end of the wind tunnel one by one. Three groups were run for each blend tested. The number of females tested in each group ranged from 25-30 depending on availability of mated females. Females were placed in a cylindrical gauze cage (diameter: 10 cm, height: 15 cm). The cylinder was closed with a solid lid on one side and placed in a holder at a height of 30 cm in the centre of the downwind end of the wind tunnel. At the beginning of the bioassays, the lid was removed, allowing the moths to leave the cage. Moth behaviour was scored as follows: (1) for >120 cm upwind oriented flight in the centre of the wind tunnel and (2) for coming within 5 cm of the odour source. The behaviour of each batch was observed for 20 min. Each female was used only once. Y-tube olfactometry. We tested the attraction of predatory C. concinna to damaged and undamaged plant tissue in a glass Y-tube olfactometer. Undamaged A. sinensis plants served as a control, and damaged plants were the same damage treatments described in 'Plant materials and treatments' . The olfactometer consisted of two glass chambers (diameter: 10 cm) that were each connected with one of the two 20-cm-long arms of the olfactometer, and joined with a 20-cm-long common arm. The odour sources (potted plant tissue) were placed inside clean roasting bags, which were then connected to the extremities of each arm: one arm served as a control (undamaged plants) and the other held the test material (H. vitessoides damaged plants). Fine-meshed nylon gauze was inserted at the ends of the two arms of the Y-tube to prevent insects from reaching the plant tissue. Moistened, activated-charcoal-filtered air with the odour source was pumped into each arm at a flow rate of 300 ml·min −1 . All bioassays were conducted during the photophase, which is the feeding period of C. concinna. Temperature and relative humidity in the Y-tube olfactometer were maintained at 25 ± 2 °C and 75 ± 5%, respectively.
One hour before the start of the bioassays, groups of insects were transferred to the Y-tube room and allowed to acclimatize in the observation room inside glass vials (15 × 3 cm). For the observations, insects were placed individually at the beginning of the common arm and observed for 10 min. Behaviour was recorded as choosing the test odour or the control if the insects entered the respective chamber. If the insects remained in the common arm of the Y-tube it was recorded that no choice had been made 49 . For each bioassay, 30 replicates were performed. Each insect was used only once in the bioassays. After each test, the Y-tube was washed with distilled water, acetone, and alcohol (v/v 90%), and then dried with an electric hair drier (HP 8200; Philips, Zhuhai, China). Data Analysis. Mean volatile concentrations in the headspace samples from different treatments, mean numbers of H. vitessoides females responding to each VOC blend in the wind tunnel, and percentage of C. concinna adults making each choice in Y-tube were each compared using one-way analysis of variance (ANOVA). Significant differences in the means were assessed using Tukey's multiple range test (α = 0.01). Chi-square test was applied to analyse results from the Y-tube behavioural tests. To reduce the complexity of the multivariate VOC data, principal component analysis (PCA) was performed. PCA was applied to yield a 2D display of the multivariable data set and to graphically determine whether clustering of the six damage treatments (undamaged plants, mechanically damaged plants, and H. vitessoides damaged plants) occurred based on their overall VOC profiles 50 . All statistical analyses were performed using SPSS, version 16.0 (SPSS Inc. Chicago, IL, USA).