Genetic Engineering of Terpenoid Metabolism Attracts Bodyguards to Arabidopsis

See allHide authors and affiliations

Science  23 Sep 2005:
Vol. 309, Issue 5743, pp. 2070-2072
DOI: 10.1126/science.1116232


Herbivore-damaged plants release complex mixtures of volatiles that attract natural enemies of the herbivore. To study the relevance of individual components of these mixtures for predator attraction, we manipulated herbivory-induced volatiles through genetic engineering. Metabolic engineering of terpenoids, which dominate the composition of many induced plant volatile bouquets, holds particular promise. By switching the subcellular localization of the introduced sesquiterpene synthase to the mitochondria, we obtained transgenic Arabidopsis thaliana plants emitting two new isoprenoids. These altered plants attracted carnivorous predatory mites (Phytoseiulus persimilis) that aid the plants' defense mechanisms.

The integration of ecology and molecular biology has yielded important progress in understanding complex interactions between organisms and the underlying mechanisms (1). In recent years, Arabidopsis thaliana L. was shown to be an excellent model plant for investigating ecological interactions such as induced indirect defense. In these tritrophic interactions, plants defend themselves against feeding by herbivorous arthropods by producing volatiles that attract the natural enemies of the herbivores (24). The activities of these natural enemies benefit the plant's fitness, and this defense therefore is evolutionarily advantageous (57). Hence, the term “bodyguard” has been used to describe the function of carnivorous arthropods such as predatory mites (5). For example, feeding by caterpillars of the crucifer pest Pieris rapae resulted in the emission of volatiles that attracted the parasitoid wasp Cotesia rubecula (8, 9). The parasitization of P. rapae caterpillars by C. rubecula resulted in an increase in plant fitness in terms of seed production and thus benefited the plant (7). The individual components of herbivore-induced volatile blends originate from various chemical classes, but isoprenoids dominate the composition of many of these blends (1013) and are known to attract carnivorous arthropods (214). One of the herbivore-induced volatiles in Arabidopsis is the C16-homoterpene 4,8,12-trimethyl-1,3(E),7(E),11-tridecatetraene [(E,E)-TMTT)] (8). A related compound, the C11-homoterpene 4,8-dimethyl-1,3(E),7-nonatriene [(E)-DMNT], has been detected in the headspace of many plant species after herbivory (12, 14, 15), but not in the (induced or noninduced) volatile mixture of Arabidopsis (8, 16).

Our first step in studying the ecological relevance of individual compounds of the complex herbivory-induced volatile mixture was to generate transgenic plants that constitutively emit these chemicals. We chose to engineer the sesquiterpene (3S)-(E)-nerolidol, a component of the herbivore-induced volatile blend of, for example, maize (17) and tomato (18) and the first dedicated intermediate en route to (E)-DMNT (15, 17) (Fig. 1A). Earlier attempts to produce substantial amounts of sesquiterpenes in transgenic plants failed, most probably because of a lack of sufficient precursors (1921). In these experiments, sesquiterpene synthases were targeted either to the cytosol (19, 20)—which is the expected location of farnesyl diphosphate (FPP), the precursor for sesquiterpenes—or to the plastids (21). Here, we targeted FaNES1, a strawberry linalool/nerolidol synthase, specifically to the mitochondria. We reasoned that because the mitochondria are the site of ubiquinone biosynthesis and Arabidopsis possesses an FPP synthase isoform with a mitochondrial targeting signal (22, 23), FPP should be available in this cell compartment.

Fig. 1.

Generation of transgenic Arabidopsis plants emitting (3S)-(E)-nerolidol and its derivative 4,8-dimethyl-1,3(E),7-nonatriene [(E)-DMNT]. (A) Schematic representation of biosynthetic pathway involved in the formation of (E)-DMNT. (B) CoxIV-FaNES1 construct scheme. (C) CoxIV-FaNES1 mRNA accumulation (upper lane) determined by reverse transcription polymerase chain reaction with actin as control (lower lane). From left to right: wild type (wT); five individual primary transformants. (D) Arabidopsis plants as used for behavior experiments: left, wild type; right, transgenic plant (both 4 weeks after sowing).

The CoxIV (cytochrome oxidase subunit IV) sequence, a bona fide mitochondrial targeting signal (24), was used to localize FaNES1 to the mitochondria (25). Transgenic Arabidopsis plants harboring the CoxIV-FaNES1 construct (Fig. 1B) were generated, and FaNES1 expression was detected in leaves of primary transformants (Fig. 1C). In earlier work we showed that in protoplasts, CoxIV when fused to green fluorescent protein (GFP) efficiently targeted GFP to the mitochondria (26). The headspace of rosette leaves from 4-week-old plants was analyzed using solid-phase microextraction (SPME) as described (21), and 9 of 12 primary transformants emitted (3S)-(E)-nerolidol (Fig. 2, D and E, peak b). The levels of (3S)-(E)-nerolidol emitted were 20 to 30 times those from plants with plastid-targeted FaNES1 (21). (3S)-(E)-Nerolidol could not be detected in rosette leaves of wild-type plants (Fig. 2A). Because the heterologous protein was targeted to the mitochondria, these results demonstrate that the sesquiterpene precursor, FPP, is indeed available in this organelle. Interestingly, we also detected (E)-DMNT in the headspace of five of the nine (3S)-(E)-nerolidol–producing plants (Fig. 2E, peak a). (E)-DMNT has not been detected before in Arabidopsis foliage after herbivory (8) nor in the flower headspace (16). In our own headspace analyses of rosette leaves derived from wild-type Arabidopsis plants, we also did not detect (E)-DMNT (Fig. 2A).

Fig. 2.

Gas chromatography–mass spectrometry profile of volatiles emitted by Arabidopsis leaves. (A) Undamaged wild type; (B) wild type, infested by spider mites for 10 days; (C) wild type, infested by P. rapae for 24 hours; (D) undamaged CoxIV-FaNES1, plant 2.2; (E) undamaged CoxIV-FaNES1, plant 7.3; (F) wild type, detached leaves placed with petiole in (E)-nerolidol (10 μg/ml) for 4 hours before headspace sampling; (G) CoxIV-FaNES1, plant 2.2 sprayed with 5 μM jasmonic acid 24 hours before headspace sampling. Identified compounds: (a) (E)-DMNT; (b) (3S)-(E)-nerolidol; (c) (E)-TMTT. The y axis shows peak area of mass/charge ratio 69 + 93.

Our results show that neither spider mites (Tetranychus urticae) nor caterpillars (P. rapae) induce (3S)-(E)-nerolidol or (E)-DMNT formation in wild-type Arabidopsis (Fig. 2, B and C), although both herbivores are known to induce (E)-DMNT in a number of other plant species (2, 12, 14, 15). Boland and co-workers have shown that leaves and flowers of several plant species are capable of converting (3S)-(E)-nerolidol into (E)-DMNT constitutively (27), and others have shown that the rate-limiting step in the herbivory-induced release of (E)-DMNT is the formation of (3S)-(E)-nerolidol (15, 17). Our results show that the specific introduction of a linalool/nerolidol synthase into the mitochondria of Arabidopsis results in the formation of substantial amounts of (3S)-(E)-nerolidol and that Arabidopsis apparently possesses the enzymes that convert (3S)-(E)-nerolidol into (E)-DMNT. This was further confirmed by feeding of (E)-nerolidol to leaves of wild-type Arabidopsis plants, which resulted in the formation of (E)-DMNT (Fig. 2F). We suggest that the enzymes responsible for the conversion of geranyl-linalool into (E,E)-TMTT (C20 → C16), which is emitted by Arabidopsis after herbivory (8), are also capable of the C15 → C11 conversion from (3S)-(E)-nerolidol into (E)-DMNT. Jasmonic acid is a known mediator of herbivory-induced signaling (28, 29) and has also been shown to induce (E,E)-TMTT formation in Arabidopsis (30). When individual transgenic lines that only emitted (3S)-(E)-nerolidol were sprayed with jasmonic acid, (E)-DMNT was subsequently detected in the headspace (Fig. 2G). Therefore, it is likely that the enzymes en route to (E,E)-TMTT are up-regulated by jasmonic acid, as was previously shown for (E)-DMNT formation in maize (17).

Both first- and second-generation transgenic plants displayed some growth retardation of the basal rosette (Fig. 1D), but their flowering stems appeared at approximately the same time as in nontransgenic control plants. Mitochondria use FPP for production of ubiquinone and heme A, but apparently the introduction of a sesquiterpene synthase into these organelles does not divert so much of the available FPP that this leads to growth inhibition. This is of importance when considering the impact of our metabolic engineering strategy on plant performance.

Transgenic plants emitting the two new signaling compounds were used to examine the effect on bodyguard attraction. Undamaged transgenic plants (4 weeks old) producing (E)-DMNT and (3S)-(E)-nerolidol in a ratio of approximately 2:1 were tested in a closed-system Y-tube olfactometer (25) against undamaged wild-type Arabidopsis plants of the same age. The predatory mites (Phytoseiulus persimilis) highly significantly preferred the volatiles emitted by CoxIV-FaNES1 plants to those of wild-type plants (binomial test, P < 0.001; Fig. 3A). An infestation with spider mites (T. urticae) that did not result in emission of (3S)-(E)-nerolidol and (E)-DMNT did not make wild-type Arabidopsis attractive to predatory mites, the natural enemies of the spider mites (Fig. 3A).

Fig. 3.

Responses of P. persimilis predatory mites to volatiles released by Arabidopsis. (A) Response tested in a Y-tube olfactometer. (B) Response tested under semi-natural conditions (22). Bars represent the overall percentages of predatory mites choosing either of the odor sources; numbers in bars are the total numbers of predators choosing that odor source. Error bars represent SE (n = 10 independent tests). Choices between odor sources were analyzed with a two-sided binominal test on numbers (ns, P > 0.05; **P < 0.01; ***P < 0.001). §Data from (27).

Because CoxIV-FaNES1 plants emitted both (E)-DMNT and (3S)-(E)-nerolidol, we assessed which of the two volatiles attracts the predators. (E)-DMNT was previously shown to attract P. persimilis (2, 31) (Fig. 3A). However, CoxIV-FaNES1 plants that only emitted (3S)-(E)-nerolidol and no (E)-DMNT were also attractive to P. persimilis (Fig. 3A). We then tested the attraction of P. persimilis to racemic (E)-nerolidol and found that the predators were significantly attracted. Although nerolidol is often reported as a component in the volatile blend induced by herbivory, to our knowledge, attraction of P. persimilis or any other carnivorous arthropod to (3S)-(E)-nerolidol has not been reported previously. Thus, the introduction of a mitochondrially targeted FaNES1 into Arabidopsis resulted in the emission of two terpenoids that both attract the predatory mite P. persimilis. These signaling molecules, (E)-DMNT and (3S)-(E)-nerolidol, are known to be induced by P. persimilis' prey in several plant species (15, 17, 18), but not in wild-type Arabidopsis (Fig. 2B).

Attraction of predators to CoxIV-FaNES1 plants was also tested, using plants in soil under more natural conditions, in an octagon setup (Fig. 3B). In this open setup, the odor spreads through diffusion rather than by directing the odor of enclosed plants through a closed container with an air stream. In 10 independent experiments, we found that the majority of the predatory mites made their first visit to the CoxIV-FaNES1 plants, which demonstrates a clear preference (P < 0.001) for the undamaged transgenic plants that emit (E)-DMNT and (3S)-(E)-nerolidol (Fig. 3B).

We have shown that genetic engineering of Arabidopsis, resulting in plants that emit one or two novel volatiles, provides a novel tool to investigate the role of signaling compounds in mediating tritrophic interactions. This is especially true for compounds that are not commercially available and not easy to synthesize in enantiomer-pure form, such as sesquiterpenoids [e.g., (3S)-(E)-nerolidol] and homoterpenes [e.g., (E)-DMNT]. The levels of the sesquiterpene alcohol (3S)-(E)-nerolidol as well as the homoterpene (E)-DMNT that were emitted by the transgenic plants are the highest reported so far, indicating that FPP is readily available in the mitochondria for metabolic engineering. Emission of these signaling chemicals from engineered plants demonstrated that these volatiles influence bodyguard behavior in vivo. Our results show that the transgenic approach holds considerable promise for improving crop protection through a transgenic approach (e.g., by exploiting herbivore-inducible promoters coupled to genes responsible for biosynthesis of signaling compounds), so that crop plants can be generated that more effectively recruit biological control agents after infestation with arthropod pests.

Supporting Online Material

Materials and Methods


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article