Isoprene emissions influence herbivore feeding decisions


C. Nicholas Hewitt. Fax: +44 0 1524 593985; e-mail:


Isoprene (C5H8, 2-methyl 1,3-butadiene) is synthesized and emitted by many, but not all, plants. Unlike other related volatile organic compounds (monoterpenes and sesquiterpenes), isoprene has not been shown to mediate plant–herbivore interactions. Here, for the first time, we show, in feeding choice tests using isoprene-emitting transgenic tobacco plants (Nicotiana tabacum cv. Samsun) and non-emitting azygous control plants, that isoprene deters Manduca sexta caterpillars from feeding. This avoidance behaviour was confirmed using an artificial (isoprene-emitting and non-emitting control) diet. Both in vivo and in vitro experiments showed that isoprene can activate feeding avoidance behaviour in this system with a dose–response effect on caterpillar behaviour and an isoprene emission threshold level of <6 nmol m−2 s−1.


Plants have evolved a variety of sophisticated defence mechanisms to counter biotic stresses. One is the synthesis and emission of volatile organic compounds (VOCs) (Dudareva et al. 2006). VOCs help defend plants, either directly, for example, by functioning as insect repellents (De Moraes, Mescher & Tumlinson 2001) and toxins (Vancanneyt et al. 2001), or indirectly, by attracting predators and parasitoids (Schnee et al. 2006). Furthermore, VOCs emitted from infested plants can act as signals between plants by directly inducing defensive responses in neighbouring uninfested plants (Arimura et al. 2000). They can also prepare or prime the plant's defensive machinery for future herbivore attacks (Engelberth et al. 2004).

The large diversity of VOCs produced by plants is mainly derived from three main biochemical pathways: the lipoxygenase (LOX) pathway, the shikimic acid pathway and the terpenoid pathway (Pichersky & Gershenzon 2002). Among these compounds, terpenes, especially monoterpenes and sesquiterpenes, have well-documented roles in mediating interactions between plants and herbivores (Gershenzon & Dudareva 2007).

Isoprene (C5H8; 2-methyl 1,3-butadiene) is produced by the terpenoid pathway in plants. It is derived from the chloroplastic 2-deoxyxylulose 5-phosphate/2-methylerythritol 4-phosphate (MEP) pathway (Lichtenthaler, Rohmer & Schwender 1997; Lichtenthaler 1999), and a considerable amount of energy and newly assimilated carbon may be used in isoprene production (Harley, Guenther & Zimmerman 1996; Sharkey & Yeh 2001). Hence, it is hypothesized that isoprene production must confer a significant benefit to plants. Furthermore, because of its high reactivity and large mass emission rate to the atmosphere (∼500 Tg y−1) (Guenther et al. 2006), isoprene plays an important role in atmospheric chemistry and physics (Fehsenfeld et al. 1992). Given the importance of isoprene in the atmosphere and its high biosynthetic production costs, much work has been carried out to elucidate the biochemical and ecological functions of isoprene in plants. Isoprene has been shown to protect leaf photosynthetic machinery from transient high temperatures (Sharkey & Singsaas 1995; Behnke et al. 2007) and from oxidative stress (Loreto & Velikova 2001). It may also serve as a metabolic overflow mechanism for carbon or photosynthetic energy (Rosenstiel et al. 2004). It has been suggested that isoprene may also act as a signalling compound between plants (Terry et al. 1995).

In contrast to related compounds, especially monoterpenes and sesquiterpenes, isoprene has not been demonstrated to have a clear role in plant–herbivore interactions. It has been shown that forest tent caterpillars feeding on hybrid poplar induced the up-regulation of sesquiterpene synthase gene expression, but the expression of the isoprene synthase gene was unaffected (Arimura, Huber & Bohlmann 2004). Other studies have demonstrated the involvement of isoprene in wound signalling. Defoliation and mechanical stimulation caused a systemic reduction of isoprene emission (Loreto & Sharkey 1993), while application of jasmonic acid (a herbivore-induced signalling molecule) stimulates local and systemic isoprene emissions (Ferrieri et al. 2005).

In this study, isoprene-emitting and non-emitting transgenic tobacco plants (Nicotiana tabacum cv. Samsun) and the tobacco hornworm (Manduca sexta Johan.) were used as a model plant–herbivore system. In order to test the direct effect of isoprene on caterpillar feeding preference, we also used an artificial food medium with and without isoprene in a feeding choice test. Here, we hypothesize that isoprene emission influences the feeding decisions of the tobacco hornworm.


Transgenic tobacco plants

Tobacco (N. tabacum cv. Samsun), the wild type of which does not emit isoprene, was transformed with a Populus alba isoprene synthase so that it emitted isoprene (Vickers et al., unpublished data). Independent, single-locus, homozygous transgenic lines were generated, along with azygous non-isoprene-emitting sibling lines as controls. Homozygous T3 (fourth generation) transgenic plants (labelled as lines 6, 12, 22 and 32), along with their non-emitting azygous controls, were selected for this study. These transgenic plants emit isoprene at rates comparable with those from the poplar species from which the isoprene synthase gene originated, at the same conditions of light and temperature (Vickers et al., unpublished data). They also showed the responses to light, temperature and inhibitor treatments expected from a wild-type isoprene-emitting plant species – that is, isoprene emission rates increased with temperature and light, and were substantially reduced by the application of the chemical inhibitor fosmidomycin. In addition, they exhibited circadian regulation of isoprene emission, as do naturally emitting plants (Wilkinson et al. 2006; Vickers et al., unpublished data). The biochemical, physiological and morphological traits of the homozygous isoprene-emitting plants did not significantly differ from those of their azygous non-emitting controls under non-stress conditions (Vickers et al., unpublished data). Thus, these transgenic lines mimic naturally emitting plants and are excellent tools for examining the ecological roles of isoprene emission.

Plant growth conditions

Transgenic tobacco plants were grown in a glasshouse with an ambient temperature of 28 ± 5 °C day and 16 ± 4 °C night with a supplementary light of 450 ± 100 µmol m−2 s−1 photosynthetically active radiation (PAR) at leaf level for 16 h d−1. One week before the experiments started, the plants were transferred into a controlled environment room with conditions of 28 ± 2 °C day and 20 ± 2 °C night, 450–500 µmol m−2 s−1 PAR, 16 h light, 8 h dark, 70% relative humidity (RH).

Insect growth conditions

Manduca sexta eggs (Department of Biology and Biochemistry, Bath University) were placed in plastic containers in a controlled environment room at 25 °C, 50% RH. Soon after hatching, caterpillars were fed on a wheat germ-based artificial diet until they reached the third instar stage. Newly moulted third-instar caterpillars, which had fed on the diet for not more than 12 h, were used in the experiments. They were starved for 2 h before each experiment. We used the third-instar larvae because their consumption rate is higher than the neonates. This allowed a shorter experimental period to be achieved (within 4 h). A short experimental period is required to allow the experiments to be completed within a single period of daylight so that the leaf biochemistry and physiology, including isoprene emission, would be stable.

In vivo caterpillar preference tests on transgenic plants

A feeding preference index (PI) was used to investigate the preference of M. sexta caterpillars towards isoprene-emitting and non-emitting plants under light and dark conditions. An intact leaf, still attached to the plant, from an isoprene-emitting plant and its azygous non-emitting control of the same line, age and node, were placed side by side in a clip cage (Noble 1958). The cages were made of two clear polystyrene rings (2.5 cm diameter, 2 cm height) connected to a hair clip. The open ends of the rings were sealed by nylon mesh allowing ventilation inside the cage. Care was taken that each leaf shared 50% of the total cage area. An M. sexta caterpillar was placed without specific orientation on the lid of the clip cage above the upper surface of the leaves. In replicate experiments, isoprene-emitting and non-emitting plants were randomly distributed to the left and right sides of the cage. A feeding PI was calculated as PI = [(consumed area of emitting leaf – consumed area of non-emitting leaf) × total area consumed−1] (del Campo & Renwick 1999). The PI values vary between +1 and −1. Positive values indicate a caterpillar feeding preference for isoprene-emitting leaves; negative values indicate a caterpillar feeding preference for non-emitting leaves; zero means no preference. The tests were terminated after 4 h or when the caterpillars had consumed 50% of the total leaf area. There were five independent repeated experiments.

In vitro caterpillar preference tests on artificial diet

In order to eliminate any other confounding variables, which might exist in our transgenic tobacco plant system, an in vitro experiment was used to confirm whether isoprene has a direct effect on caterpillar feeding decisions. A wheat germ-based artificial diet was cooled to 37 °C before adding varying amount of liquid isoprene (99%; Aldrich, Steinheim, Germany) to give four isoprene emission rates of 5, 28, 55 and 57 nmol m−2 s−1 (mean emission rates from the diet surface over a 19 h experimental period). These emission levels were comparable with those of the transgenic tobacco used in the in vivo experiment. The diet was then poured into plastic cube moulds (8 cm3) and left for 15 min at room temperature.

The isoprene-emitting and non-emitting diet was cut into halves. One half of each type of diet was offered to a caterpillar. The other half of the cubes with and without isoprene was used as control in order to account for weight loss due to desiccation. The isoprene-emitting and non-emitting diet were placed 2 mm apart. An M. sexta caterpillar was put on a filter paper platform in front of the diet. The head of the caterpillar was placed centrally between the isoprene-emitting and non-emitting diet and about 3 mm away from each. The caterpillar was left to feed for 19 h. A feeding PI was calculated as PI = [(weight consumed on emitting diet − weight consumed on non-emitting diet) × total weight consumed−1]. There were three independent repeated experiments.

Isoprene emission measurement methods

Isoprene emissions were measured from the sixth-node leaves of 6-week-old transgenic tobacco plants of all selected transgenic lines. A leaf apex was clamped with a light- and temperature-controlled leaf cuvette connected to a CIRAS-1 portable photosynthesis system (PP System, Hitchin, UK). The conditions inside the cuvette were set to 28 °C and 500 µmol m−2 s−1 PAR. All measurements were made at least 20 min after the point at which a steady rate of photosynthesis had been reached.

A light-to-dark transition experiment was also performed in order to investigate whether, in the dark, isoprene production and emission from transgenic tobacco plants was similar to that of plants that naturally emit isoprene. The information obtained from this experiment was used to support the results obtained from in vivo caterpillar preference tests in the dark. For the light-to-dark transition experiment, leaves were maintained in the cuvette at the conditions described previously for 3 h, before the light was turned off. Isoprene emission measurement continued for 1 h in the dark. Part of the outflow of air passing from the leaf cuvette to the CIRAS-I was diverted to the analytical instrument for determination of isoprene concentrations and hence emission rate.

For the artificial diet, the food cubes were released from the mould and were placed in a plastic box (1200 cm3). Air was pumped through the box at a rate of 400 mL min−1. The outlet of the box was connected to the analytical instrument for determination of isoprene concentration and emission rate over a period of 19 h.

The concentration of isoprene in air was measured using a proton transfer reaction mass spectrometer (Ionicon GmBH, Innsbruck, Austria) through detection of mass 69, using a total dwell period (duty cycle) of 11 s (Hewitt, Hayward & Tani 2003; Hayward et al. 2004). Confirmation of the identity of mass 69 as isoprene was obtained using gas chromatography–mass spectrometry (Possell, Hewitt & Beerling 2005).

Statistical analysis

Statistical analyses were performed by using SPSS and Sigma Plot for Windows (Version 10.0) (SPSS Inc., Chicago, IL, USA).The replicates of the experiments were not statistically heterogeneous and therefore pooled prior to the statistical analysis. Analysis of feeding preference was by one-sample t-test using no preference (0) as the null hypothesis (two-tailed). Non-linear regression analysis (one-tailed) was used to examine the relationship between feeding PI and isoprene emission rate. The regression lines from both the in vivo and in vitro experiments were combined within statistical criteria (univariate test).


In vivo caterpillar preference tests on transgenic plants

Fourth-generation (T3) homozygous plants showed a range of isoprene emission rates: 2.1 (±0.2), 10.5 (±0.9), 10.9 (±2.9) and 17.7 (±2.2) nmol m−2 s−1 at the environmental conditions of 28 °C and 500 µ mol m−2 s−1 PAR (Fig. 1a). As expected (Monson et al. 1991; Hayward et al. 2002), isoprene emission rates rapidly fell when the plants were placed in the dark, with a characteristic 'post-illumination burst' visible (Fig. 1b). They reached zero within 20 min, confirming that the internal pool of isoprene in the transgenic tobacco plants is very small, and that isoprene production does not occur in the dark (Tingey et al. 1979; Hayward et al. 2002).

Figure 1.

Isoprene emission measurements. (a) Mean isoprene emission rates (n = 5) from wild-type tobacco (Nicotiana tabaccum cv. Samsun) and transgenic tobacco plants (lines 22, 6, 12 and 32) containing the isoprene synthase gene from poplar (Populus alba L.). (b) Light-to-dark transition of isoprene emission of transgenic tobacco plants (n = 4). Emission rates were measured at 28 °C and 500 µmol m−2 s−1 of photosynthetically active radiation (PAR). Whisker bars indicate standard errors.

In the light, the caterpillars showed a significant feeding preference for the non-isoprene-emitting azygous control plants rather than the corresponding isoprene-emitting plants, for three of the four transgenic lines (Fig. 2: one sample t-test, P < 0.05, n ≥ 30 per line). The caterpillars showed no feeding preference for plants of line 22, the line with the lowest isoprene emission rate, compared with its azygous controls (Fig. 2: one sample t-test, P > 0.05, n = 30). Thus, the caterpillars could discriminate between emitting and non-emitting leaves for plants emitting isoprene at ∼10 nmol m−2 s−1 or greater, but not for those emitting ∼2 nmol m−2 s−1. The inability to discriminate non-isoprene-emitting leaves from emitting leaves for line 22 implies that there is a threshold level of the caterpillar response to isoprene-emitting transgenic plants. Although a feeding deterrance effect of isoprene-emitting transgenic plants was observed in lines 6, 12 and 32, a direct effect of isoprene on caterpillar feeding preference could not be concluded based on this information alone, because different lines might differ in other ways than only isoprene emission. Therefore, further studies were designed to clarify the direct effect of isoprene on caterpillar choice.

Figure 2.

In vivo caterpillar feeding choice test on transgenic tobacco plants in the light compared with azygous controls. Feeding preference index of third-instar Manduca sexta caterpillars on emitting and non-emitting transgenic tobacco plants (lines 6, 12, 22 and 32; n ≥ 30 per line). P indicates probability values. Asterisks indicate significant differences, P < 0.05.

In the dark, when isoprene production and emission does not occur (Fig. 1b), caterpillars lost their ability to differentiate between leaves from transgenic and azygous control plants (Fig. 3: one sample t-test, P > 0.05, n ≥ 30 per plant line). Because the experimental period was relatively short (≤4 h), the significant results found for plants of lines 6, 12 and 32 in the light (Fig. 2) were unlikely to be due to differences in stable aspects of leaf biochemistry and physiology (such as leaf thickness and availability of stable secondary metabolites), but rather imply that caterpillars use a light-dependent cue to distinguish between emitting and non-emitting leaves. This cue is most likely isoprene.

Figure 3.

In vivo third-instar Manduca sexta caterpillar feeding choice test on transgenic tobacco plants in the dark (isoprene emission rates from all lines = zero, n = 30) compared with azygous controls. P indicates probability values. There was no significant statistical difference in preference index from zero for all lines (P > 0.05). Whisker bars indicate standard errors.

In vitro caterpillar preference tests on artificial diet

The results showed that caterpillars were deterred by the isoprene-emitting artificial diet, at the three higher emission rates (one sample t-test, P < 0.05, n = 45 per diet type), but not at the lowest (5 nmol m−2 s−1: one sample t-test, P > 0.05, n = 45) (Fig. 4). As for the feeding tests using tobacco, this also indicates that there is an isoprene emission threshold level in caterpillar feeding preference. We therefore conclude that isoprene has a direct negative impact on the feeding decisions of the tobacco hornworm.

Figure 4.

In vitro third-instar Manduca sexta caterpillar feeding choice tests on artificial diets emitting isoprene (5, 28, 55 and 57 nmol m−2 s−1, diets 1–4, respectively) compared with unadulterated non-emitting artificial diet. Feeding preference index of third-instar M. sexta caterpillars on emitting and non-emitting diet (n = 45 per emission level). P indicates probability values. Asterisks indicate significant difference, P < 0.05.


The experiments described in this paper using a novel plant–insect interaction model indicate that isoprene can be detected by the caterpillars, and that it deters them from feeding. When results from the in vivo and in vitro experiments are combined, a clear picture of dose–response effects and threshold level is obtained (R2 = 0.99, P < 0.0001, Fig. 5), with an estimated threshold isoprene emission rate of between 0 and 6 nmol m−2 s−1 (Fig. 5). To our knowledge, the avoidance behaviour of a herbivore towards isoprene-emitting plants has not been reported previously. Given that the trait of isoprene emission is found throughout the plant kingdom (from mosses to trees: Harley, Monson & Lerdau 1999), it is not surprising that herbivores, which share co-evolutionary history with plants (Ehrlich & Raven 1964; Farrell 1998), can sense isoprene. One other insect detritivore (Sinella coeca feeding on isoprene-spiked artificial food) has been shown to be capable of responding to isoprene (Michelozzi et al. 1997), suggesting that the capacity to detect this compound could be widespread in insects.

Figure 5.

Regression between preference index and isoprene emission rate from the combined results of experiments using in vivo and in vitro systems (Figs 2 and 4). The regression lines from these two systems were not significantly different (univariate test, P = 0.374). The regression of the best fit curve for all data points was significant at P < 0.0001, R2 = 0.99. Replication was n ≥ 30 on the vertical axis, and n ≥ 3 on the horizontal axis. The shaded area represents the estimated standard errors.

Not all plants emit isoprene (Harley et al. 1999), and it has been suggested from recent molecular phylogenetic work that, at least for flowering plants, the isoprene synthase gene evolved several times independently (Sharkey et al. 2005). The uneven distribution of isoprene emission across plant species holds true even within plant families (Harley et al. 1999). The isoprene emission rates of many plants under natural growth conditions are in the range of 10–200 nmol m−2 s−1 (Guenther et al. 1995; Lerdau & Keller 1997; Street, Hewitt & Mennicken 1997; Owen, Boissard & Hewitt 2001), which is above the threshold rate identified for M. sexta larvae here (approximately 6 nmol m−2 s−1, Fig 5). Thus, variation in isoprene emission may be one factor that mediates herbivore food choice in nature. Clearly, the role of isoprene needs to be confirmed in a wider range of plant–herbivore systems, including not only choice by larvae, as here, but also choice of plant by egg-laying females.

In addition to the previously defined ecological functions of isoprene emission from plants, that is, protecting leaf metabolic machinery from abiotic stresses including temperature (Sharkey & Singsaas 1995; Behnke et al. 2007) and oxidants (Loreto & Velikova 2001), our data show isoprene can play a direct role in plant–herbivore interactions in our model system. Whether or not this feeding deterrence effect occurs in natural systems is not yet clear. If it does, isoprene has multiple physiological and ecological functions in plants, and the trait of isoprene synthesis may be much more important in conferring competitive advantage on plants than previously recognized.


We thank E. Ackerley for advice on statistical analysis; F. Wäckers, K. Wilson and S. Cotter for valuable discussions; and P. Nott for technical assistance. This work was supported by a Royal Society Dorothy Hodgkin Postgraduate Award, the Biotechnology and Biological Sciences Research Council (BBS/B/12172), the Marie Curie Research Training Network ‘ISONET’, and the European Science Foundation 'VOCBAS' programme.