Western flower thrips avoid sweet pepper plants previously infested by an omnivorous predator

* Corresponding author. Tel: 0623112757

E-mail address: daanrgvw@gmail.com (D.R.G. van Wieringen)

Western flower thrips avoid sweet pepper plants previously infested by an

omnivorous predator

Daan van Wieringen, 10658513

ARTICLE INFO

Keywords: Induced plant defence,

host plant selection, biological

control, Frankliniella occidentalis, Macrolophus pygmaeus, Capsicum annuum

ABSTRACT

Plant defences are an important factor in the selection of host plants by herbivores. Plant damage causes the induction of plant defences, lowering the performance of the herbivores through direct defence mechanisms and defending the plant indirectly by secretion of plant volatiles. Recent research has shown that omnivorous predators, commonly used in biological control, also induce plant defences and subsequently affect host plant selection of herbivores and parasitoids. I investigated the effects of previous infestation of a sweet pepper plant by an omnivorous predator on the host plant selection of western flower thrips with short and long range cues. I show that the omnivore Macrolophus pygmaeus induces plant defence in sweet pepper plants, making it less attractive for the greenhouse pest Frankliniella occidentalis. Thus, omnivores affect the behaviour of pests not only through entomophagy, but also through induction of plant defences.

1.

Introduction

There are many ways in which a plant can defend itself against herbivores: constitutive defences and induced defences, which can both be direct and indirect (Turlings et al., 1990; Dicke et al, 1988). Constitutive defences are present continuously and include traits such as thorns, toxins and compounds that slow down digestion of plant material by herbivores. Other constitutive defences combat herbivores with structures that can harbour and protect natural enemies and by providing food to the natural enemies (Dicke & Sabelis, 1988). Induced plant defences are activated by herbivore damage and herbivore-associated compounds including saliva and regurgitant (Delphia et al., 2007), but also by oviposition by the herbivores. Induced direct defences include toxins and proteinase inhibitors, which slow down the digestion and overall performance of the herbivore (Green & Ryan, 1972; Haukioia, 1980). Indirect induced plant defences consist of the release of volatile chemical signals that are attractive to natural enemies and parasitoids of the herbivores (Turlings et al., 1990, Dicke and Sabelis 1988).

Many factors affect host plant choice by herbivores, such as plant quality (Delphia et al., 2007) and the predation risk associated with different host plants (Nomikou et al., 2003). Volatile signals are important factors for assessing host plant quality and the host plant selection of herbivores and predators (Bell and Cardé, 1984). These

volatile signals can attract as well as repel organisms. Spider mites for instance, are slightly attracted to volatiles of plants with conspecifics and strongly repelled by volatiles of plants with thrips, which are competitors and predators of spider mites (Pallini et al., 1997). In the case of bark beetles, volatiles produced by groups of heterospecifics are used to avoid increased competition on trees (Byers, 1993). Herbivores can use signals associated with their predators’ presence to avoid plants with them, as shown with spider mites and predatory mites (Pallini et al., 1999). They can also use cues of artificially damaged conspecifics to choose their host plant (Janssen et al. 1997). Thrips can also use cues from infested plants to avoid intraspecific competition (Agrawal and Klein, 2000). Research by Delphia et al. (2007) shows that thrips prefer untreated host plants over plants with defences induced by their conspecifics. Host plants with mechanical damage were also found to be less attractive in a choice experiment with thrips and tobacco plants (Delphia et al. 2007).

Research has shown that plant defences can also be induced by omnivores (Puysseleyr et al. 2011, Pappas et al., 2015), which could subsequently affect host plant selection by herbivores and parasitoids (Pérez-Hedo et al., 2015). A large part of the arthropod predators of pests are omnivores, making it important to understand their effects on plants and pests for improving biological control. The use of the omnivore Macrolophus pygmaeus is considered safe to combat herbivores such as thrips, because damage inflicted by the predator on the plant is minimal. Only when prey availability is low and the density

Bachelor Thesis Biological Sciences, UvA

of the predator is very high will the predator cause economic damage (Castañé et al., 2011). Research by Pappas et al. (2015) has shown that the zoophytophagous predator M. pygmaeus can induce direct plant defence responses in tomato plants, which negatively affects herbivore performance, lowering fecundity and plant damage. A study by Zhang (unpublished) has shown that M. pygmaeus can also induce plant defences in sweet pepper plants, which reduces the performance of western flower thrips. This means that omnivorous predators can affect herbivorous prey directly through predation and also indirectly via plant-mediated effects. It has been shown that this effect can last for at least 2 weeks (Pappas et al. 2015). The combination of combatting pests directly as well as indirectly through the plant, could make M. pygmaeus more suitable for biological control of pests than previously thought (Pappas et al., 2015).

One of the major pests in greenhouses and the cause of considerable damage to many crops is the western flower thrips Frankliniella occidentalis. Insecticides are not very effective to control Western flower thrips because of their secluded behaviour and resistance to most insecticides (Jensen, 2000), which makes biological control essential. The omnivorous predator M. pygmaeus is commonly used to control various pests on different crops in greenhouses and is effective for the biological control of thrips (Blaeser et al., 2004).

By infesting sweet pepper plants (Capsicum annuum) with M. pygmaeus the induced direct and indirect defences will be activated, lowering the performance of subsequent herbivores, which was shown with spider mites and thrips (Zhang, unpublished). The compounds which may be produced, such as proteinase inhibitors and other secondary plant compounds and volatiles, will probably make the sweet pepper a less attractive host plant to thrips compared to uninfested plants. Research by Zhang (unpublished) has shown that previous infestation by M. pygmaeus can affect host plant choice of another common greenhouse pest, the two spotted-spider mites Tetranychus urticae, even without M. pygmaeus being present on the plant.

Thus, we investigated how previous infestation of sweet pepper plants by M. pygmaeus affected the host plant preference of the western flower thrips. Two experiments were performed to answer this question. The first experiment focused on the role of long range cues, such as volatiles, but also visual cues, in the host plant selection by thrips. This was done offering adult thrips a choice between clean plants and plants treated with M. pygmaeus. A second experiment was performed where thrips were given more time to choose among plants and could therefore visit several plants to assess host plant quality. Thus, they could use both long-range and short-range cues to assess host plant quality. The two experiments were compared to determine whether thrips also need short range cues to properly assess host plant quality. Reproduction of thrips on the plants was estimated in the last experiment by counting numbers of larvae 5-6 days after the last counting. This was used to assess whether the host plant preference of thrips was also reflected in different oviposition rates.

2.

Material & Methods

The herbivore Macrolophus pygmaeus was obtained from a laboratory culture at the University of Amsterdam. It was reared on C. annuum at

(25±1o

C, 60-70 % RH, 16:8 L: D) in mesh cages (45 x 45 x 45 cm) and were fed with twice a week with Ephestia kuehniella eggs. Western flower thrips were originally obtained from cucumber plants in a commercial greenhouse near Pijnacker, The Netherlands, in February

2006. The thrips were reared on C. annuum at (25±1o

C, 60-70 % RH, 16:8 L: D) in mesh cages (45 x 45 x 45 cm). Their diet was supplemented by Typha latifolia L. pollen twice a week. Plants were watered twice weekly and replaced with new plants whenever necessary.

All sweet pepper plants were grown from seed (Spider F1, Enza) until four weeks old in a climate chamber containing only clean plants. All plants were watered twice a week and grown under the same

conditions (25±1o

C, 60-70 % RH, 16:8 L: D) on Jongkind potsoil in round plastic pots (12 cm high and 15 cm in diameter). The plants were infested by placing them individually in a mesh cage (45 x 45 x 45 cm) with five female and five male M. pygmaeus inside for four days. The predators remaining at the end of this period were counted and removed from the treated plants. The plants were subsequently transferred to a greenhouse compartment the day before the experiments. For the choice experiment, two treated and two clean plants were placed inside a mesh cage in a greenhouse compartment (84 x 86.5 x 203 cm and 84 x 103 x 203 cm). Plants were placed in a square with an equal distance of 35.5 cm to the release point in the centre and the positions of treated and clean plants were alternated (Figure 1). The plants were oriented such that every third leaf pointed to the release point in the centre. The order of treated and untreated plants was alternated with each replicate. A group of 100 adult thrips were collected from the culture with an aspirator and starved for 90 minutes in four 1000 µL Eppendorf pipette tips with 25 thrips in each tip, which were closed with Parafilm. The tips were placed in the centre of the setup and the thrips were released by removing the Parafilm.

2.1. First choice experiment

In the first choice experiment, the thrips were released and counted by removing them from the plants with an aspirator at intervals of one hour. Samples were taken one hour after the initial release and then with intervals of one hour between the end of the previous sampling and the start of the new sampling. This short interval between sampling ensured that most thrips were caught on the host plant of their first choice (Joost & Riley, 2008). Six samples were collected on the day of the release and one final sample was taken 24 hours after the initial release. The experiment was replicated nine times.

2.2. Final choice experiment

In the final choice experiment, thrips were counted six times without removing them from the plants. Twenty-four hours after the release, the thrips were again counted and removed with an aspirator. Thus, the thrips could visit several plants and also use short-range cues to assess plant quality. This experiment was replicated six times.

2.3. Reproduction

The six replicates of the final choice experiment were also used to estimate the reproduction of the thrips on the plants, by counting the number of nymphs after five days. After the thrips had been removed from the plants, these were placed in mesh cages (45 x 45 x 90 cm) on top of tables in a greenhouse compartment with pairs of clean plants and treated plants in the same cage. The pots were placed on trays (22.5 x 17.5 cm), which were placed inside larger trays (30 x 23 cm) filled with water, to prevent the nymphs from escaping. The nymphs were counted by removing them with an aspirator.

The cumulative number of thrips collected during the experiments were log transformed and the effects of treatment and time were analysed using a linear mixed effects model. Differences between time points were analysed using simultaneous tests for general linear hypotheses from the multcomp package (Crawley, 2013). The first and final choice experiment were compared with a generalized linear model with a quasibinomial error distribution. Reproduction was compared using a one-sided t-test. The ratio of larvae on clean and treated plants was compared to the ratio of the cumulative number of thrips on clean and treated plants of the final choice experiment, using a binomial test for proportions (Crawley, 2013). All data analysis was performed using R software (version 3.3.0).

Figure 1: Experimental setup. Two replicates were performed at the same time in two mesh cages, shown as rectangles with a solid blue line, inside the greenhouse compartment. The clean plants (A1, A2) and treated plants (B1, B2) were placed inside the center of the mesh cages, in a square with equal distances between the plants, represented by an intermittent line. The position of clean and infested plants was alternated between replicates. Thrips were released from the blue dot in the center of the square.

3.

Results

The plants were treated with 5 male and 5 female M. pygmaeus, of which 51% survived the four days of treatment, ranging from 1 to 8 individuals. The large majority of M. pygmaeus were present on the plants when they were removed.

The large majority of thrips dispersed from the release point, on average only c. 2.1 thrips were found dead (ranging from 0 to 9 individuals) at the release site. About 46.2% of the released thrips were recaptured. Some thrips roamed the cage, while others escaped from

the mesh cages during the experiment and were recaptured on a trap plant, which was placed in the greenhouse compartment after the experiment was finished.

3.1. First choice experiment

The first choice experiment was repeated nine times. Out of the 900 thrips which were released, 439 were recaptured (48.8%) from the plants. With 255 thrips recaptured from the clean plants (58.1%) and 184 on treated plants (41.95%) the cumulative number of thrips that chose for the clean plants was significantly higher (Figure 2, LME:

Chi2

= 27.7, df = 1, P < 0.0001). The cumulative number of thrips

increased significantly over time (Chi2

= 136.2, df = 6, P <0.0001). There was no interaction between treatment and time (Chi = 0.72, df = 6, P = 0.99). Tests per recapture time showed that significantly more thrips were found on the clean plants from the third sample and onwards (Figure 2).

Figure 2. The average cumulative number of thrips and standard error per plant as a function of sample number of the first choice experiment. In this experiment, thrips were removed during counting at each sample. Analyses using simultaneous tests for general linear hypotheses showed a significant preference (marked with an asterisk) from the third sample onwards (SE = 0.14, df = 230, t-value = 2.1, P = 0.037).

3.2. Final choice experiment

For the final choice experiment, six replicates were performed. Out of the 600 thrips that were released, 239 were recaptured (41.9%). Of the 239, 166 were recaptured from the clean plants (69.5%) and 73 from the treated plants (30.5%). The thrips showed a clear preference for the

clean plants over the treated plants (Figure 3, LME: Chi2 _{= 99.1, df =}

1, P < 0.0001). The number of thrips increased over time (Chi2

= 52.7, df = 4, P < 0.0001). No interaction between treatment and time was

found (Chi2 _{= 2.05, df = 10, P = 0.92). The host plant preference for}

the clean plants was significantly different from the first sample and onwards (Figure 3). No significant difference was found between the proportion of thrips choosing the clean host plant when comparing the first and final choice experiments (Figure 4, quasibinomial GLM: df = 1,28, F = 1.5, P = 0.23). There also seemed to be more feeding scars on the clean plants than on the treated plants, although this was not quantified. A2 A1 B1 B2 B2 B1 A1 A2 Door NW 1 2 3 4 5 6 0 24 -4 -2 0 2 4 6 8 10 12 14 16 18

Time since release

cum ul at iv e #t hr ips r ec ap tur ed Contr Macro * * * * *

Figure 3: The average number of thrips and standard error per plant as a function of sample number of the final choice experiment. In this experiment thrips were removed only at the end. The preference was significant from the first sample onwards (SE = 0.1623, t-value = 3.5, P = 0.00055), indicated by an asterisk.

3.3. Reproduction

Reproduction was only determined in the final choice experiment. The preference for the clean host plant was also found in oviposition choice rates, with a significantly higher number of larvae which hatched and survived on the clean plants (Figure 5, One sided t-test: t = 1.9, df = 15.8, p-value = 0.038). The ratio of larvae found on the treated and clean plants (170/85 = 2) was very similar to ratio of thrips in the final

choice experiment (166/73 = 2.27). A binomial proportions test

showed no significant difference between the two (SE = 0.19, z value = -0.66, p-value = 0.51).

Figure 4: The proportion of thrips, comparing the first and final choice experiments. The grey bars represent the first choice experiment and the final choice experiment is represented by the black bars. There was no significant between the proportion of thrips choosing the higher quality host plant when comparing the first and final choice experiments (df = 28, F = 1.5, P = 0.23).

Figure 5: The average number of thrips larvae sampled and standard error. Significantly more larvae were found on the clean plants than the plants previously infested with M. pygmaeus (t = 1.9, df = 15.8, P = 0.038). The ratio of clean/treated was very similar in the oviposition experiment (170/85 = 2) and the final choice experiment (166/73 = 2.28) and no significant difference was found between the two ratios (SE = 0.19, z value = -0.66, P = 0.51).

4.

Discussion

Research has shown that plant defences are induced by infestation of tomato plants by the omnivorous M. pygmaeus and that this reduces the plant quality for subsequently arriving herbivores (Pappas et al., 2015). Experiments by Zhang (unpublished) demonstrate that infestation by M. pygmaeus is also detrimental to the performance of western flower thrips on sweet pepper plants. Research by Delphia et al. (2007) has shown that host plant selection by thrips is affected by the induction of plant defences by conspecifics, but not by spider mites (Pallini et al., 1999). The induction of plant defences by another omnivore, N. tenuis also influenced the host plant selection of herbivore B. tabaci (Rérez-Hedo et al., 2015).

The results of this research confirm the expectation that the induction of defences by the omnivore M. pygmaeus also affects host plant selection of a subsequently arriving herbivore. Western flower thrips can use long range cues, such as volatiles, to avoid the treated plants, which are of lower quality due to defences induced by infestation of M. pygmaeus. Fewer thrips were collected from the treated plants, which was also reflected in lower reproduction on these plants. The lower reproduction rates could be explained by oviposition rates as well as lower hatching rates (Pappas et al., 2015) on the treated plants. While the thrips did show a significant preference for the clean plants, there was still a large portion of the thrips which choose for the treated plants. This was especially the case in the first choice experiment, where 42% choose for host plants previously infested with M. pygmaeus. There are several possible causes for the fact that such a large portion still choose for the plants with induced defences. For instance, plants were in close proximity, which might have complicated the use of volatiles. It could be interesting for future research to compare long and short distance cues and how this affects their ability to choose the higher quality host plant.

Treatment of the sweet pepper plants could also partly explain the large amount of thrips choosing the lower quality host plant. After four

clean macro 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Treatment C u mu la ti v e p ro p o rt io n o f th ri p s final first Contr Macro 0 2 4 6 8 10 12 14 16

Treatment

Time since release

N u mb er o f th ri p s co u n te d Contr Macro clean 0.54 0.56 0.58 0.6 0.62 0.64 0.66 0.68 Treatment Cu m u lativ e p ro p o rti o n o f th rip s final first * * * * * * *

days of treatment only half of the M. pygmaeus survived, which might have led to a lower level of induced plant defences, making it harder for thrips to properly asses host plant quality. In future research, more M. pygmaeus could be used to ensure that there is a high level of induced defences in all treated plants. The level of induced plant defences and the production of cues was probably not influenced by the presence of thrips and plant damage caused by them. Research by Kant et al. (2004) on the timing of spider mite induced defences has shown that although activity in JA systems increased on the first day of infestation already, the emission of volatiles started on day four. This suggests that herbivores need more time to induce the production of volatiles, making it unlikely to have influenced the results of this research.

The thrips might have also used other cues which could have influenced their host plant choice, such as colours outside the human spectrum which might have changed due to infestation by M. pygmaeus. Thus, it is interesting to investigate thrips host plant choice when visual and volatile cues are separated. This could be achieved with a y-olfactometer experiment, in which the thrips are unable to use visual cues, or a choice experiment in which the thrips make the choice based only on visual cues with plants downwind.

Previous research by Zhang (unpublished) suggests that spider mites are better able to assess host plant quality when allowed more time to make a choice and to sample the plants. The results of this research also hint at a stronger preference in the final choice experiment, where thrips could use short range cues as well as long range cues and were had more time for host plant selection. The ratio of thrips which choose for the clean plants in the final choice experiment was indeed higher (69.5% in the final choice experiment compared to 58.1% in the first choice experiment). However, the preference for the untreated host plant in this research was not significantly different between the first and final choice experiments. It is not known how much time thrips need to accurately assess host plant quality. Perhaps the preference would be significantly different between first and final choice experiments when allowed more time to choose, however 24 hours was considered to be sufficient. Another factor could have been the difference in total recapture rates between first and final choice experiments. Fewer thrips were recaptured in the final choice experiment (41.9%) than in the first choice experiment (48.8%). Most thrips probably escaped during sampling, when the cage was opened. Because the thrips in the final choice experiment were not removed until the end, they therefore had more time and more opportunities to escape from the mesh cage. The average recapture rate of thrips was less than half (46.2%), similar to the low recapture rates found by Pallini et al. (1999). It could be that a large proportion of thrips tried to escape the mesh cage altogether, due to the presence of plants damaged by M. pygmaeus. It would be interesting to first determine the recapture rate of thrips in an experiment using only clean plants. Since the results and the difference in ratios do hint at a difference in the ability to assess host plant quality between the two experiments, it might be valuable to perform more replicates in future research.

To conclude, infestation of the sweet pepper plant induces the creation of cues which significantly affect host plant selection of thrips. For future research it would be interesting to quantify the cues created and ascertain which type of cues influence thrips host plant selection. Distinguishing these cues and understanding their composition and effect on behaviour could allow for manipulation of pest host plant selection as part of integrated pest management programs (Kaplan,

2012). That omnivores induce cues which influence host plant selection of the pest, as well as decrease performance of subsequent infestation by thrips and combating thrips directly, should be taken into account when using omnivores in biological control (Zhang, unpublished; Pappas et al., 2015; Blaeser et al., 2004). Since omnivores such as the M. pygmaeus do not cause economic damage unless in very high densities (Castañé et al., 2011) it could be beneficial to have omnivores permanently present to protect crops against pests directly and indirectly through induced defences.

5.

Acknowledgements

I thank Arne Jansen and Nina Zhang for giving me the opportunity to do this research, their guidance, help during experiments and processing the results, comments and making this research an interesting as well as enjoyable project. I thank Joke Andringa for helping to take care of the colonies. Lastly I thank everyone in the research group for their critical questions.

R

Agrawal, A. A., & Klein, C. N. (2000). What omnivores eat: direct effects of induced plant resistance on herbivores and indirect consequences for diet selection by omnivores. Journal of Animal Ecology, 69(3), 525-535.

Blaeser, P., Sengonca, C., & Zegula, T. (2004). The potential use of different predatory bug species in the biological control of Frankliniella occidentalis (Pergande) (Thysanoptera: Thripidae). Journal of Pest Science, 77(4), 211-219.

Bell, W. J., & Cardé, R. T. (2013). Chemical ecology of insects. Springer. Byers, J. A. (1993). Avoidance of competition by spruce bark beetles, Ips typographus and Pityogenes chalcographus. Experientia, 49(3), 272-275.

Castañé, C., Arnó, J., Gabarra, R., & Alomar, O. (2011). Plant damage to vegetable crops by zoophytophagous mirid predators. Biological control, 59(1), 22-29.

Crawley, M. J. (2013). The R book second edition. John Wiley & Sons Ltd., Chichester, UK.

De Puysseleyr, V., Höfte, M., & De Clercq, P. (2011). Ovipositing Orius laevigatus increase tomato resistance against Frankliniella occidentalis feeding by inducing the wound response. Arthropod-Plant Interactions, 5(1), 71-80.

Delphia, C. M., Mescher, M. C., & De Moraes, C. M. (2007). Induction of plant volatiles by herbivores with different feeding habits and the effects of induced defences on host-plant selection by thrips. Journal of chemical ecology, 33(5), 997-1012.

Dicke, M., & Sabelis, M. W. (1988). How plants obtain predatory mites as bodyguards. Netherlands journal of zoology, 38(2), 148-165. Grostal, P., & Dicke, M. (1999). Direct and indirect cues of predation risk

influence behavior and reproduction of prey: a case for acarine interactions. Behavioral Ecology, 10(4), 422-427.

Kaplan, I. (2012). Attracting carnivorous arthropods with plant volatiles: the future of biocontrol or playing with fire? Biological

control, 60(2), 77-89.

mites than clean plants under greenhouse conditions. Entomologia Experimentalis et Applicata, 90(2), 191-198.

Jensen, S. E. (2000). Insecticide resistance in the western flower thrips, Frankliniella occidentalis. Integrated Pest Management

Reviews, 5(2), 131-146.

Joost, P. H., & Riley, D. G. (2008). Tomato plant and leaf age effects on the probing and settling behavior of Frankliniella fusca and Frankliniella occidentalis (Thysanoptera: Thripidae). Environmental entomology, 37(1), 213-223.

Kant, M. R., Jonckheere, W., Knegt, B., Lemos, F., Liu, J., Schimmel, B. C. J., & Egas, M. (2015). Mechanisms and ecological consequences of plant defence induction and suppression in herbivore

communities. Annals of Botany, 115(7), 1015-1051.

Lucas‐Barbosa, D., Loon, J. J., Gols, R., Beek, T. A., & Dicke, M. (2013). Reproductive escape: annual plant responds to butterfly eggs by accelerating seed production. Functional Ecology, 27(1), 245-254. Nomikou M., Janssen A., & Sabelis M. W., et al. (2003). Herbivore host

plant selection: whitefly learns to avoid host plants that harbour predators of her offspring. Oecologia, 136(3), 484–488. Pallini, A., Janssen, A., & Sabelis, M. W. (1997). Odour-mediated

responses of phytophagous mites to conspecific and heterospecific competitors. Oecologia, 110 (2), 179-185.

Pallini, A., Janssen, A., & Sabelis, M. W. (1999). Spider mites avoid plants with predators. Experimental & Applied Acarology, 23(10),

803-815.

Pappas ML., Steppuhn A., Geuss D., Topalidou N., Zografou A., Sabelis M.W., et al. (2015) Beyond Predation: The zoophytophagous zredator Macrolophus pygmaeus induces tomato resistance against spider mites. PLoS ONE 10(5): e0127251.

doi:10.1371/journal.pone.0127251

Pashalidou, F. G., Lucas-Barbosa, D., van Loon, J. J., Dicke, M., & Fatouros, N. E. (2013). Phenotypic plasticity of plant response to herbivore eggs: effects on resistance to caterpillars and plant development. Ecology, 94(3), 702-713.

Pérez-Hedo, M., Urbaneja-Bernat, P., Jaques, J. A., Flors, V., & Urbaneja, A. (2015). Defensive plant responses induced by Nesidiocoris tenuis (Hemiptera: Miridae) on tomato plants. Journal of Pest

Science, 88(3), 543-554.

Sarmento, R. A., Lemos, F., Bleeker, P. M., Schuurink, R. C., Pallini, A., Oliveira, M. G. A., Lima, E. R., Kant, M., Sabelis, M. W., Janssen, A. (2011). An herbivore that manipulates plant defence. Ecology letters, 14(3), 229-236.

Turlings, T.C.J., Tumlinson, J.H., Lewis, W.J., 1990. Exploitation of herbivore-induced plant odours by host seeking parasitic wasps. Science 250, 1251–1253.