Introduction

Plants release a variety of volatile organic compounds that play multiple roles in interactions with other plants and animals (Dudareva et al., 2004; Knudsen et al., 2006; Pichersky et al., 2006). Insect herbivores exploit these volatiles to locate their host plants at a distance, for feeding, mating, and egg-laying (Linn et al., 2003; Bengtsson et al., 2006; Tasin et al., 2006; Pinero and Dorn, 2009; Cha et al., 2008; Schmidt-Busser et al., 2009; Sole et al., 2010). In addition to olfaction, vision and contact chemoreception (taste) could play a role at a short distance (Renwick and Chew, 1994; Calatayud et al., 2008; Pontes et al., 2010). Knowledge of the chemicals and mechanisms that mediate host plant location by insect herbivores is essential for our understanding of plant—insect relationships, and also will lead to the development of novel tools for insect management (Szendrei and Rodriguez-Saona, 2010; Witzgall et al., 2010).

Plant volatiles are products of diverse metabolic pathways, but many are derived from the isoprenoid or terpenoid pathways (Sacchettini and Poulter, 1997; Degenhardt et al., 2009). Solanaceous plants, like tomato, produce a suite of terpenes that likely serve as defense agents against herbivores (Kennedy, 2003; Bleeker et al., 2009; Kang et al., 2010). The constitutive headspace of undamaged plants varies with genotype, phenological stage, and environmental conditions. It is conceivable that insects use the volatile signals that correlate with this variation to distinguish the most suitable hosts (Bengtsson et al., 2001; Vallat and Dorn, 2005; Karlsson et al., 2009).

The tomato leafminer Tuta absoluta (Meyrick) (Lepidoptera, Gelechiidae) feeds on several solanaceous species, and preferentially on tomato Solanum lycopersicum (Siqueira et al., 2000). It is a serious threat to commercial tomato production in South and Central America. Since its introduction, it has become an economically important insect in Europe (Torres et al., 2001; Desneux et al., 2010). It is anticipated that T. absoluta will even spread from Central America into Mexico and the USA. Control of T. absoluta relies on multiple insecticide sprays, but resistance to these chemicals is growing (Siqueira et al. 2000; Lietti et al., 2005). Synthetic sex pheromones are used to monitor population levels, but also for control by mass trapping of males in greenhouses (Salas, 2004; Witzgall et al., 2010).

Tomato plants are infested at any developmental stage, with females ovipositing preferentially on leaves. After hatching, larvae feed on leaf parenchyma tissue, on tender portions of the stems (especially axillary buds), and in developing and mature fruit, causing bud drop, fruit malformation, fruit rot, and a drastic reduction in leaf area (Torres et al., 2001). No tomato cultivars are entirely resistant to T. absoluta, but not all cultivars are equally susceptible (Bogorni et al., 2003; Oliveira et al., 2009). In addition, the wild tomato species S. habrochaites (syn. Lycopersicon hirsutum) is more resistant to T. absoluta and several other tomato herbivores than S. lycopersicum (Leite et al., 2001; Kennedy, 2003).

The aim of this study was to determine the role of tomato volatiles in host finding and oviposition behavior of tomato leafminer T. absoluta. We identified the volatiles released from leaves of three different cultivars of tomato S. lycopersicum and investigated the question whether tomato leafminer uses these volatiles for the discrimination of suitable larval hosts.

Methods and Materials

Insect Rearing

A laboratory colony of tomato leafminer T. absoluta originated from pupae and larvae collected in a commercial tomato plantation located in Viçosa, Minas Gerais (Brazil). The colony was maintained at 24 ± 1°C, 70 ± 10% r.h., and a 12:12 h L:D photoperiod. Adults were fed with sugar water and provided with tomato (cv. Aromata) leaves for egg-laying. Leaf stems were cut and placed through a hole in the transparent plastic lid of a 500-ml recipient containing water and a nutrient solution. Larvae were provided with fresh leaves every 2nd d, and rolls of corrugated paper were added for pupation. Pupae were sexed and placed in plastic cages for eclosion (100 × 40 × 40 cm).

Plant Material

For behavioral experiments and volatile analysis, three tomato Solanum lycopersicum Mill. (Solanaceae) cultivars were used: cv. Santa Clara and cv. Carmen, which are known to be highly susceptible, and cv. Aromata, which is less susceptible (Svalöf-Weibull Trädgård AB, Hammenhög, Sweden) (Bogorni et al., 2003; Oliveira et al., 2009; Lima et al. unpublished data). In addition, the wild tomato species S. habrochaites, which is resistant to T. absoluta, was used for wind tunnel and oviposition bioassays. Tomato seeds were pre-cultivated in rock wool cubes in the greenhouse for 4 week (r.h.: 75%, light intensity: 200 μmolm−2 s - 1). Plants were transplanted to a hydroponic system and fertilized daily with a nutrient solution. Standard values (in mM) for the composition of the nutrient solution were: 6.0 Ca(NO3)2, 8.0 KNO3, 5.0 MgSO4, 1.0 NH4NO3, 2.3 KH2PO4 and for micronutrients: 0.4 Fe–EDTA, 0.01 MnSO4, 0.025 H3BO3, 0.005 ZnSO4, 0.00075 CuCl2,2 H2O, 0.0005 Na2MoO4,and pH 5.8 ± 0.1.

Flight Behavior Bioassay

Female attraction to plants was tested in a wind tunnel (flight section 63 × 63 × 200 cm; Tasin et al., 2006). Filtered air (24 cylinders with active charcoal, 14.5 × 32.5 cm, Camfil, Trosa, Sweden) was blown into the tunnel by a centrifugal fan (Fischbach GmbH, Neunkirchen, Germany) at 25 cm/s. Outcoming air was aspirated by another fan and cleaned by two additional sets of charcoal filters. The flight section was lit diffusely from the side at 10 lux, and the room was kept at 23 ± 2°C and 40–60% r. h. Cut tomato plants (ca. 35 cm high) were placed into vials with water in the center of the upwind end of the flight tunnel. Two h before the end of the photophase, females were placed individually into cylindrical plastic containers. They were kept until testing in the wind tunnel room. Moth behavior was scored for upwind oriented flight in the center of the wind tunnel (over at least 60 cm from the downwind end) and for approaching tomato plants. By using this method, we tested (1) the attraction of 2-d-old mated (N = 190) and virgin (N = 150) females to S. lycopersicum cv. Aromata, and (2) the attraction of mated female (N = 150) to S. lycopersicum cv. Santa Clara and S. habrochaites.

Oviposition Bioassays

Experiments were done with 2-d-old mated females during 24 hr. Females were individually kept in a glass tube (15.0 × 3.0 cm) covered on both sides with a plastic film. To keep the females in one side of the tube, it was divided in half with a nylon grid. The female and the tomato leaf were put on the same (N = 48) or different sides of the tube (N = 48). This experiment was done with the three tomato S. lycopersicum cultivars. After 24 h, the eggs laid on the glass wall and the nylon grid were counted.

In choice tests, mated females were kept in a plastic container (12 × 10 cm) either with one leaf of cv. Aromata, cv. Carmen and cv. Santa Clara (N = 81 females), or with one leaf of Santa Clara and wild tomato S. habrochaites (N = 37 females), and the leaves were scored for eggs after 24 h.

Plant Volatile Analysis

Freshly cut plants of the three cultivars of S. lycopersicum were confined separately in a 2-L glass jar that was closed with a ground-glass fitting. The cut end of the branch was held in a 10-ml vial of water. A charcoal-filtered airstream was pulled over the plant material from the bottom to the top of the jar, and over a 35-mg Super Q trap (80/100 mesh; Alltech, Deerfield, IL, USA) which was held between plugs of glass-wool in a 4 × 40 mm glass tube. Before use, traps were rinsed sequentially with 3 ml methanol, ether, and redistilled hexane, after 15 min treatments in ultrasonic baths in ether and hexane, respectively. The air flow was 150 ml/min, exchanging the headspace in the jar 4.5 times/h. Collections were done for 24 h, at 20–22°C and 10–30 lux. The charcoal filter for incoming air and the Super Q trap for outcoming air were connected with glass fittings to the jar. All glassware was heated to 350°C for 10 h before use. After volatile collections, traps were extracted immediately with 0.5 ml hexane (redistilled; LabScan, Malmö, Sweden). Sample volumes were reduced to 50–60 μl, at ambient temperature in Francke-vials with an elongated tip (5 cm × 2 mm i.d.). Samples were stored in sealed glass capillary tubes at −18°C (Bengtsson et al., 2001).

Odor samples were analyzed on a gas-chromatograph coupled with a mass-spectrometer (GC-MS). Volatile compounds in trap eluents were injected in aliquots of 2 μl into a Hewlett- Packard 5970 B MS instrument, with electron impact ionization at 70 eV, and interfaced to an HP 5890 GC (Hewlett- Packard, Palo Alto, CA, USA). The GC was equipped with a fused silica capillary column (30 m × 0.25 mm) coated with DB-wax (df = 0.25 μm; J&W Scientific, Folsom, CA, USA) using helium as carrier gas (35 cm/s). The temperature program was from 35°C (hold 5 min) at 8°C/min to 230°C (hold 5 min). The Kovats retention index (KI) of each compound was calculated. Tentative identification of compounds was based on comparison of matching of mass spectra with the NIST 98 MS and Wiley databases and a reference library built by measuring commercially available standards, and comparison of KIs on similar columns. Identification of some components was confirmed by comparison with mass spectra and GC retention data with those of standards.

Statistical Analysis

All statistical analysis was done using R (version 2.8.0; R Development Core Team, www.R-project.org) following the method of Crawley (2005).

The effect of either mating status or test plant (S. lycopersicum and S. habrochaites), on the proportion of females, grouped per day, that were attracted to and that landed on the plant was tested with a Generalized Linear Modelling (GLM) under quasibinomial distribution.

Similarly, the effect of contact with the host plant and the effect of the tomato cultivar (Carmen, Santa Clara, Aromata) on the total number of eggs laid in the non-choice bioassays were tested using a GLM under quasipoisson distribution. In the choice experiments, the effect of S. lycopersicum cultivars or tomato species on the proportion of eggs laid on leaves was tested using a GLM with quasibinomial distribution. In all analyses, a full model was fitted, from which terms were deleted in a stepwise fashion to obtain the minimal adequate model (MAM). Significance (P < 0.05) was assessed by testing the change in deviance after the removal of a term from the model. Differences among treatment levels were examined by contrast analysis. Candidate similarity levels were amalgamated to compose a new model, which was then compared with the previous un-amalgamated model. The new, simplified model was accepted if it did not differ from the previous, more complex model. Amalgamation proceeded until all treatment levels were checked. Models were checked by residual analyses, and eventually corrected for overdispersion as well as for the correctness of the assumed distribution (Crawley, 2005).

Headspace composition was compared among samples by multivariate analyses using R with Vegan and labdsv packages (Oksanen et al., 2008; Roberts, 2010). Non-metric multidimensional scaling (NMDS) was used to detect similarities among samples (using amounts relative to total peak area). Data were square-root transformed and standardized before calculating Bray–Curtis similarities (Bray and Curtis, 1957). In oviposition bioassays, T. absoluta prefers cv. Santa Clara and cv. Carmen over cv. Aromata with no difference between Santa Clara and Carmen. Thus, based on T. absoluta oviposition preference, variation in odor composition among tomato cultivars and between cv. Aromata vs. cvs. Santa Clara and Carmen (pooled), was tested for significance using permutational multivariate analysis of variance (permanova) with the Bray–Curtis dissimilarity measurement and 10000 permutations (McArdle and Anderson, 2001).

Differences in the occurrence and abundance of compounds in the headspace of the cv. Aromata vs. the cvs. Santa Clara/Carmen were determined according to the method of Dufrêne and Legendre (1997) by calculating ‘indicator compound’ (e.g., Proffit et al., 2009). For each compound, an observed indicator value (IV) was calculated. The deviation of the observed IV from a random distribution of IVs was tested using a Monte Carlo test with 10000 randomizations. Indicator compounds have an observed IV significantly different from the random IV, and designate the most characteristic compounds in headspace samples of a group, either because of their occurrence in different samples and/or because of their relative abundance.

Results

Wind Tunnel Experiments

Mated tomato leafminer T. absoluta females flew upwind over 180 cm towards tomato plants (45% of the females tested) and landed (42%) on these plants. In contrast, unmated females showed no attraction response to tomato (N = 17; for flight, F 1,15  = 192.63, P < 0.001; for landing, F 1,15  = 122.09, P < 0.001; Fig. 1).

Fig. 1
figure 1

Proportion of mated and virgin females of tomato leafminer Tuta absoluta flying upwind and landing on tomato leaves (Solanum lycopersicum cv. Aromata) in the wind tunnel (mean ± SE/day). For both behavioral responses, differences between mated (N = 190) and virgin females (N = 150) are significant (N = 17 days, GLM, P < 0.05)

Full size image

Oviposition Experiments with Tomato Cultivars

In non-choice tests, the cultivar had a significant effect on oviposition (df = 2, χ 2 = 132.71, P < 0.001; Fig. 2a and b). Tomato leafminer females laid more eggs in response to cvs. Santa Clara and Carmen as compared cv. Aromata (df = 1, χ 2 = 110.39, P < 0.001 and df = 1, χ 2 = 95.75, P < 0.001), both with and without leaf contact: the interaction between leaf contact and cultivar was not significant (df = 2, χ 2 = 3.38, P = 0.82).

Fig. 2
figure 2

Oviposition rate (mean ± SE) of mated tomato leafminer Tuta absoluta females on three tomato cultivars Aromata, Santa Clara, and Carmen, in (a) a non-choice experiment without leaf contact (N = 48), (b) a non-choice experiment with leaf contact (N = 48), and (c) a choice experiment with leaf contact (N = 81). For each experiment, treatments represented by filled and empty bars are significantly different (GLM with contrast, P < 0.05)

Full size image

A choice test (Fig. 2c) confirmed the effect of the cultivar on oviposition (F 1,78 32.08, P < 0.001). Here again, the females preferred Santa Clara and Carmen over Aromata (F 1,78  = 22.96, P < 0.001 and F 1,78  = 36.11, P < 0.001, respectively). In both non-choice and choice tests, there was no difference in the preference between Santa Clara and Carmen (df = 1, χ 2 = 0.56, P = 0.79, and F 1,78  = 1.53, P = 0.22, respectively). Females laid significantly more eggs when in contact with tomato leaves of S. lycopersicum cv. Aromata, Carmen and Santa Clara, compared to when they were exposed only to leaf headspace (df = 1, χ 2 = 766.55, P < 0.001; Fig. 2).

Comparison Between Cultivated and Wild Tomato

Both in the wind tunnel as well as in oviposition bioassays, T. absoluta preferred cultivated tomato S. lycopersicum over wild tomato S. habrochaites (F 1,20  = 109.30, P < 0.001; F 1,72  = 46.42, P < 0.001, respectively). The proportion of insects preferring S. lycopersicum was similar in both experiments (Fig. 3).

Fig. 3
figure 3

Upwind attraction (N = 22 days, mean ± SE/day) and oviposition choice tests (N = 37, mean ± SE) of mated tomato leafminer Tuta absoluta with cultivated Solanum lycopersicum cv. Santa Clara and wild tomato Solanum habrochaites. The effect of the tomato species was significant in both experiments (GLM, P < 0.05)

Full size image

Chemical Analysis of Tomato Headspace

We found 52 volatile compounds in the headspace collection of cv. Aromata, Santa Clara, and Carmen (Table 1). Terpenoid compounds largely dominated the tomato leaf headspace. Beta-phellandrene was the most abundant compound by far and represented more than half of the total volatile emission. In addition, limonene, 2-carene, as well as (E)-β-caryophyllene were major compounds in tomato headspace of the three cultivars.

Table 1 Headspace of tomato foliage of three cultivars Aromata, Santa Clara, and Carmen (Solanum lycopersicum). Abundance of volatile compounds (mean ± SD), relative to the total amount or released volatiles
Full size table

Although most volatiles were released by all three cultivars (Table 1), leaf headspace composition differed among them (Permanova F 2,19  = 5.95, P < 0.001; Fig. 4), headspace differed between Aromata and Santa Clara/Carmen (Permanova F 1,20  = 2.23, P = 0.001). According to a Monte Carlo test, 26 of the 52 compounds were good indicators of the leaf headspace of either Aromata or Santa Clara/Carmen (Table 1), most of which are minor terpenoid compounds, with a mean leaf headspace proportion of < 1%. Aromata did not produce extra compounds compared to the two other cultivars.

Fig. 4
figure 4

Non-metric multidimensional scaling (NMDS) plot visualizing the samples of three tomato cultivars, Aromata, Santa Clara, and Carmen based on the similarity (Bray-Curtis distance) of their leaf headspace composition, rotated by principal component, so that the variance of points is maximized on the first dimension (stress = 0.15). Samples are grouped (dashed lines) according to tomato leafminer Tuta absoluta oviposition preference. The centroids of cv. Aromata vs. the preferred cvs. Carmen/Santa Clara are the focal points of the solid lines

Full size image

In contrast, Carmen and Santa Clara released 6 compounds that were absent from the headspace of Aromata: carvacrol, α-copaene, (Z)-limonene oxide, an unidentified monoterpene (unknown 4), an unidentified sesquiterpene (unknown 11), and one unindentified compound (unknown 18). In addition, 9 other compounds that were characteristic for Santa Clara/Carmen were present in smaller amounts in Aromata leaf headspace. On the other hand, 10 compounds were characteristic for Aromata and more abundant than in Santa Clara and Carmen headspace (Table 1).

Discussion

Combined headspace analysis and behavioral assays show that tomato leaf volatiles are essential cues for host finding and oviposition in tomato leafminer T. absoluta (Figs. 1, 2 and 3). The females detected small variations in volatile signatures of the different tomato cultivars (Table 1, Fig. 4), and the ensueing behavioral response (Figs. 2 and 3) matched the suitability of tomato cultivars and species as larval hosts (Leite et al., 1999, 2001; Oliveira et al., 2009).

Long Range Attraction Cues

Mated T. absoluta females responded to tomato leaves by upwind oriented flight, followed by landing (Fig. 1). Host volatiles have been shown to elicit attraction of several other moths (Tasin et al., 2006; Arab et al., 2007; Masante-Roca et al., 2007; Knudsen et al., 2008; Pinero and Dorn, 2009; Sole et al., 2010).

Only mated females were attracted to tomato leaves (Fig. 1). In most herbivorous insects, host-searching behavior is based not only on the perception of suitable olfactory information, but also on the insect’s internal state. Physiological changes such as ovarian maturation, egg production, and mating status can influence the response of insect females to host cues (Thompson and Pellmyr, 1991; Yan et al., 1999; Masante-Roca et al., 2007). For example, mated grapevine moth Lobesia botrana or Egyptian cotton leafworm Spodoptera littoralis females are attracted to their larval host plants, not unmated females (Masante-Roca et al., 2007; Saveer et al. unpublished data).

Oviposition Cues

Tomato leaf volatiles alone elicited an oviposition response in T. absoluta (Fig. 2a). Moreover, ovipositing females discriminated among different tomato cultivars and between cultivated and wild tomato on the basis of volatile cues (Figs. 2 and 3). However, leaf contact significantly increased the number of eggs laid (Fig. 2a and b), and leaf surface morphology and chemistry is accordingly of importance for oviposition. Tuta absoluta females lay eggs uniformly on the under- and upper-side of leaves that are covered with trichomes that provide chemical and mechanical stimuli (Torres et al., 2001).

Tomato leafminer females lay their eggs directly on leaves (Torres et al., 2001). Accordingly, there is no spatial separation between the oviposition site and plant part that releases the chemical signal that elicits host-finding. In contrast, two closely related species from potato, Tecia solanivora and Phthorimaea operculella, oviposit in the soil close to the plant, rather than on leaves or stems (Fenemore, 1988; Horgan et al., 2007; Karlsson et al., 2009).

Variation of Volatile Profile Among Tomato Cultivars

The volatile profile of tomato leaves of the cultivars Aromata, Santa Clara, and Carmen was dominated by monoterpenes, in particular β-phellandrene (major compound), limonene, and 2-carene, and the sesquiterpene (E)-β-caryophyllene (Table 1), which is in accordance with headspace analysis of other tomato cultivars (Buttery et al., 1987; Zhang et al., 2008). Nonetheless, overall headspace composition of Aromata differed significantly from Santa Clara and Carmen (Fig. 4), due to differences in blend proportions of minor compounds and due to the absence of several compounds, mostly terpenes, in Aromata (Table 1). Our oviposition bioassays shows that T. absoluta females were able to detect this inter-cultivar variation in leaf headspace, since they laid more eggs in response to headspace from Santa Clara and Carmen than Aromata (Fig. 2).

Tomato Leaf Headspace as an Indicator of Host Suitability

Herbivorous moths are attracted to host plant volatiles at specific blend ratios in laboratory assays (Fraser et al., 2003; Tasin et al., 2006). However, more recent studies in Oriental fruit moth Cydia molesta and grapevine moth Lobesia botrana show that females are attracted to plant volatile blends that show considerable variation with respect to proportions and composition (Tasin et al., 2007, 2010; Najar-Rodriguez et al. 2010). A response to a broad spectrum of plant volatile blends could represent an adaptation to differences in volatile signatures from genotypes and to variation during phenological changes and would accordingly facilitate host finding even in specialist herbivores.

This is contrasted by the differential response of tomato leafminer to comparatively small variations in tomato headspace components and proportions (Table 1, Fig. 2). A tentative explanation for discrimination of minor blend variations is that the females are less tolerant and respond strongly to those compounds that are direct cues of host plant suitability. Trichome-borne terpenoids, which account for much of the difference between the cultivars Aromata, Santa Clara, and Carmen (Table 1) have been suggested to influence host plant selection and fitness in tomato insects (Bleeker et al., 2009, 2011; Kang et al., 2010).

The idea that volatile cues provide information of host suitability at a distance is further substantiated by attraction and egg-laying choice tests with wild tomato S. habrochaites, where females clearly preferred cultivated over wild tomato (Fig. 3). Leaf headspace of wild tomato S. habrochaites was dominated by β-ocimene, β-myrcene, and undecan-2-one (Smith et al., 1996; Bleeker et al., 2009). Glandular trichomes of wild tomato also contain many sesquiterpenes, including caryophyllene, curcumene, elemene, humulene, and zingiberene, that play a role in tomato resistance against beet armyworm and whitefly (Eigenbrode et al., 1994; Smith et al., 1996; van der Hoeven et al., 2000; Bleeker et al., 2009, 2011). Resistance of wild tomato to tomato leafminer has been attributed to tridecan-2-one, undecan-2-one, and zingiberene, which have not been found in the three cultivars studied here (Leite et al., 1999; Azevedo et al., 2003).

This initial study of the attraction and oviposition response of tomato leafminer T. absoluta in response to its primary host tomato will support the identification of the odor template that encodes host recognition, and the development of new methods to control this important tomato pest worldwide. Tomato volatiles can be used directly for trapping mated females, and the differential behavioral response to tomato cultivars indicates the potential of plant breeding for improved resistance. While a complete description of the bioactive compounds is a considerable undertaking, the wind tunnel bioassay is already available for efficient screening of plant genotypes.