Introduction

Odors of plants can serve as cues to attract natural enemies of the insect herbivores attacking those plants (Vinson, 1976, 1981; Nordlund et al., 1988; Allison and Hare, 2009). These odors may be comprised of hundreds of compounds (Dudareva et al., 2006), many of which are released only by plants after they have been damaged by herbivores (Elzen et al., 1983; Dicke and Sabelis, 1988; Turlings et al., 1991; Tumlinson et al., 1992). One unresolved question in the production of these herbivore-induced plant volatiles (HIPVs) is whether different herbivore species induce consistently different or specific blends, and how specific a blend must be in order to be recognized by the species of natural enemies that effectively suppress the plant's herbivores.

Congenitally fixed responses to herbivore-specific cues were predicted for cases involving extreme specialization by natural enemies on particular herbivore species which themselves are specialized on a particular species of plant, whereas progressively greater levels of learned responses on the part of the natural enemies were predicted for less extreme levels of specialization (Vet and Dicke, 1992; Faeth, 1994). Similarly, specialist natural enemies were predicted to utilize specific cues from their host/prey complex, whereas generalists were predicted to utilize general cues released by all of their different host/prey complexes (Steidle and van Loon, 2003). Although simple predictions may be reasonable for relatively simple trophic chains, especially for plants attacked by a single herbivore species that itself has only one resident species of natural enemy, predictions that concern the responses of plants become more difficult if plants are attacked by a community of generalist and specialist herbivores, themselves attacked by numerous species of natural enemies differing in host specificity.

In some cases, attack by single vs. multiple herbivore species may result in statistically significantly different HIPV blends (Shiojiri et al., 2001; Delphia et al., 2007; De Boer et al., 2008), whereas in other systems, plants infested with multiple herbivore species simply emit a greater quantity of the same volatiles than plants damaged by single herbivore species (Moayeri et al., 2007). In still other systems, damage by one herbivore species may suppress the HIPV response normally induced by other species (Zhang et al., 2009). Therefore, the consequences of infestation of plants by more than one species of herbivore are difficult to predict, and depend upon such factors as the particular combination of plant species, herbivore species, and species of natural enemy (Hare, 2011).

Another controversy in this area focuses upon whether the HIPV blend produced by plants in response to damage by herbivores is anything more than a simple response to mechanical injury. In most studies, damage by herbivores elicits HIPV blends containing a greater range of compounds and in greater concentrations than mechanical damage. Mechanical damage that is most similar to damage by herbivores, both as to the mechanism of damage as well as its duration, elicits HIPV blends most similar to those elicited by natural herbivores (Mithofer et al., 2005; Connor et al., 2007). Plant species may differ in how closely their volatile profiles induced by mechanical damage resemble those induced by herbivore damage, however (Maffei et al., 2007). Even for plant species in which the volatile blends are similar in response either to mechanical or herbivore damage, differences in plant responses can be seen at the molecular level (Bricchi et al., 2010).

Variation in the total quantity of HIPVs as well as the relative proportions of components may vary with plant genotype (Hare, 2007; Schuman et al., 2009), the environmental conditions under which plants are grown (Gouinguene and Turlings, 2002; Kigathi et al., 2009), and the amount of time following damage (Turlings et al., 1998; Kunert et al., 2002). Variation in composition also may be a function of the amount of damage (Gouinguene et al., 2003; Delphia et al., 2007). Such variation raises important questions as to whether any aspect of the HIPV blends of plants induced by different herbivore species may be sufficient to reveal reliably the identities of the herbivores (Turlings et al., 1995; Turlings and Wäckers, 2004; Webster et al., 2010).

When damaged by its primary herbivore, the chrysomelid beetle Lema daturaphila, plants of the undomesticated plant species, Datura wrightii produce an array of at least 17 compounds, most of which are sesquiterpenes (Hare, 2007, 2010). In a laboratory study, total production of volatiles increased 3.9 to 16.2 times among genetic lines of D. wrightii after insect damage. The most abundant compound was (E)-β-caryophyllene, which comprised from 17 to 59% of the volatiles from insect-damaged plants, depending upon genetic line. Other relatively abundant HIPV compounds included (Z)-3-hexenyl acetate, (E)-β-ocimene, (E,E)-4,8,12-trimethyl-1,3,7,11- tridecatetraene (hereafter TMTT), (E)-4,8-dimethyl-1,3,7- nonatriene (hereafter DMNT), and β-selinene, most of which are commonly induced by herbivores in other plant species (Arimura et al., 2005). The array of compounds emitted by young plants induced in the field was similar (Hare, 2010; Hare and Sun, 2011).

Within southern California populations, D. wrightii exhibits two leaf trichome phenotypes, and aspects of the ecology and genetics of the trichome dimorphism have been described elsewhere (van Dam and Hare, 1998; van Dam et al., 1999). The 'velvety' phenotype is densely covered by short, non-glandular trichomes and feels velvety to the touch, whereas the 'sticky' phenotype is less densely covered with glandular trichomes that secrete esters of glucose and aliphatic acids, and these leaves feel sticky to the touch. The trichome phenotype is governed by a single locus and is inherited in a Mendelian fashion; the allele for the sticky phenotype is dominant over that of the velvety phenotype (van Dam et al., 1999).

The herbivore community that attacks D. wrightii in southern California is relatively simple, being comprised of some 5–10 species depending upon season and location, but the herbivore community varies with trichome phenotype (Elle and Hare, 2000; Hare and Elle, 2002). The most abundant herbivore species overall is L. daturaphila, a folivore whose larval and adult stages damage sticky and velvety plants similarly throughout the growing season. Another potentially more damaging insect herbivore, because of its greater rate of food consumption (Bronstein et al., 2006), is the tobacco hornworm, Manduca sexta. This species is more constrained seasonally in its abundance on D. wrightii, but it is more abundant on velvety than sticky plants (Elle and Hare, 2000; Hare and Elle, 2002).

The most abundant herbivore on sticky plants is the mirid bug, Tupiocoris notatus. It utilizes the sugar esters produced by the glandular trichomes of the sticky phenotype of D. wrightii as feeding and oviposition stimulants, but T. notatus can survive and reproduce on velvety plants if given no choice (van Dam and Hare, 1998). Although the effect of T. notatus on HIPV production of D. wrightii has not yet been studied, the omnivorous heteropteran predator, Geocoris pallens, is attracted to plants of Nicotiana attenuata damaged either by T. notatus or Manduca spp. (Kessler and Baldwin, 2004). On the other hand, damage by T. notatus also can suppress some of the jasmonate-mediated responses to other insects in other plant species (Heidel and Baldwin, 2004), which suggests that the HIPV blend induced by T. notatus in combination with L. daturaphila may differ from the HIPV blend induced by L. daturaphila alone.

A number of natural enemies that differ in host range are associated with these herbivore species. In a comprehensive survey of the parasitoids and predators of M. sexta on D. wrightii in Arizona, more than 40 natural enemies were identified, including 34 insect predators and parasitoids from 8 different orders and 14 species of spiders from 4 different families (Mira and Bernays, 2002). Less intensive sampling of the natural enemies associated with D. wrightii in California found many of the same generalist predators, including species of Nabis and Chrysoperla (Gassmann and Hare, 2005), although the most abundant predator in California was G. pallens, a species not reported from the survey in Arizona. Like several other generalist predators, G. pallens exhibited lower rates of predation of L. daturaphila and T. notatus when the herbivores were on sticky plants than when on velvety plants in the greenhouse (Gassmann and Hare, 2005). This predator also inflicted greater levels of mortality of young larvae of L. daturaphila on young D. wrightii plants when plants either were damaged by L. daturaphila or treated with MeJA compared to undamaged or untreated plants. The increase in predation was especially pronounced on the velvety genotype (Hare and Sun, 2011). In field experiments, predation of eggs of L. daturaphila by G. pallens increased with increasing total HIPV production in the early part of the season, when predators were colonizing plants, but predation rates did not vary with the variation in the composition of the HIPV blend produced by different genetic lines of D. wrightii (Hare and Sun, 2011). Common parasitoids of M. sexta in California include the generalists, Hyposoter exiguae and Trichogramma spp. (Oatman et al., 1983). Only two parasitoids of L. daturaphila are observed occasionally in southern California, Conura delumbis (Hymenoptera: Chalcidae) and Myiopharus infernalis (Diptera: Tachinidae). Both of these utilize other coleopteran hosts as well (Arnaud, 1978; Burks, 1979).

Our overall goals were to compare the quantities and compositions of HIPV blends induced by the three predominant herbivores of D. wrightii in southern California, the chewing herbivores L. daturaphila and M. sexta, and the piercing-sucking herbivore, T. notatus, when these insects were feeding on two relatively highly inducible lines of D. wrightii. We asked if the blends differed consistently among inducers to indicate the possible identity of the herbivore causing the damage. The paucity of specialized natural enemies of the herbivores of D. wrightii compared to the abundance of generalist predators also suggested that there may be reduced opportunities for the evolution of herbivore-specific HIPV blends in D. wrightii. We also asked if the quantities and composition of HIPV blends induced by plants infested by both T. notatus and L. daturaphila differed from the blends induced by each species alone. Both the total quantity as well as the composition of blends induced by each herbivore treatment were subjected to multivariate analysis and compared over a period from 1 d prior to induction until 7 d after induction.

Methods and Materials

Two experiments were carried out using similar, though not identical designs. The first compared the HIPVs induced over time by L. daturaphila and M. sexta, whereas the second compared the HIPVs induced by T. notatus and L. daturaphila when each damaged plants alone or in combination.

Plants

Fifth-generation backcross progeny of the highly inducible MVV6 and MVV8 lines (Hare, 2007) were used in both experiments. These lines were established by backcrossing heterozygous sticky progeny to their original velvety pollen parent. Because the genotypes of the progeny of these backcrosses can be only heterozygous or homozygous recessive, then trichome phenotype maps directly to genotype. The expected ratio of adult heterozygous sticky to homozygous recessive velvety sibs within each of the lines is 1:1, and sibs within families are expected to be at least 98.4% similar after five generations of backcrossing. Plants were grown in a greenhouse equipped with high-pressure sodium lamps providing supplemental illumination for 14 hr/d so that mid day light intensities averaged 1250 ± 39 μmol m−2 s−1 PAR illumination (N = 5 readings) at plant height. Greenhouse temperatures ranged between 15° and 35°C. Plants were maintained under these conditions for ca. 2–4 mo until they had produced 6–8 true leaves (first experiment) or 10–12 leaves (second experiment). The first experiment was performed using younger trichome-undifferentiated plants because prior research showed no difference in HIPV blends between sticky and velvety plants when induced by L. daturaphila (Hare, 2007). The effect of trichome genotype was of more interest with regard to the HIPVs induced by T. notatus, so the second experiment utilized older plants for which trichome differentiation had been completed (van Dam et al., 1999).

Collection of Volatiles

Volatiles were collected nondestructively from intact plants in a growth room 14:10 photophase: scotophase, 26°C day and 20°C night temperatures with light intensity of 293 (± 29 SE, n = 5) μmol m−2 s−1 PAR illumination at plant height. Aeration chambers were made from polyester cooking bags (unprinted 45 × 55 cm bags, Terinex, Bedford, England) as described earlier (Hare, 2007).

An aeration chamber was placed over a whole plant, excluding senescing leaves, and secured with a twist-tie around the stem. Intake and exhaust flow rates were regulated with flow meters (Aalborg, Orangeburg, NY, USA). PTFE tubing was used upstream of the adsorbent trap and PVC tubing was used downstream. Compressed air, filtered through activated charcoal, flowed into the bag at 600 ml/min. Periodic collections of volatiles from an empty bag showed that the charcoal filter removed all potentially interfering volatile compounds. Air was exhausted from the bag through an adsorbent trap at the same rate using house vacuum. HIPVs were collected on traps made of glass tubing and filled with Super-Q (25 mg, Alltech, State College, PA, USA). At the termination of the aeration, traps were wrapped in aluminum foil, labeled, and stored at −20°C for extraction later that day.

For the first experiment, 6 plants were assigned to each of two herbivore damage treatments, and 1 plant was assigned to a 'no induction' control treatment within each trial. Control plants were included simply to ensure that plants were not induced by mechanical handling. The greater allocation of replicates to insect-damage treatments reflect the higher priority placed upon comparing the HIPV blends induced by the two different herbivore species than on comparing either to uninduced controls. The whole experiment was repeated five times. One plant in one trial was damaged during the course of the experiment and deleted, leaving a total of 64 experimental plants. For the second experiment, 4 plants, 2 sticky and 2 velvety were assigned to damage by T. notatus alone, L. daturaphila alone, or both, whereas 1 plant of each trichome genotype was assigned to a 'no induction' treatment (control) within each trial of 14 plants. This experiment also was repeated five times, but 3 plants were damaged during handling and so were deleted from the experiment leaving a total of 67 plants.

Induction Treatments

For plants to be induced by L. daturaphila, groups of four fourth (last) instar larvae of L. daturaphila per leaf were placed on each of three leaves, and larvae were allowed to feed for 24 hr. The youngest full-sized leaf, plus the one above and the one below, were chosen, and larvae were confined to their leaf by a small organdy leaf cage. For the plants to be induced by M. sexta, one newly-molted third instar M. sexta per leaf was placed on each of three leaves of similar age, as above, and these larvae also were confined to their leaf with a small leaf cage. These numbers and developmental stages of the two insect species were chosen on the basis of preliminary experiments (not shown) to cause similar levels of damage. In the second experiment, for plants to be induced by T. notatus, plants were infested with 50 nymphs and adults of T. notatus, two pairs of adult L. daturaphila, or both 50 T. notatus individuals and two pairs of adult L. daturaphila, and these insects also were allowed to feed on plants for 24 hr. A single 24-hr-damage-period was chosen to facilitate the calculations of the rate of HIPV emission per unit of remaining plant tissue over time. Although it might have been more realistic to leave insects on the plants until they had ceased feeding, this would have caused an artificial reduction in HIPV production by these plants over time due to the smaller quantities of leaf tissue available to emit HIPVs on each day of the 8-d-observation period.

Preliminary experiments showed little difference in HIPV quantities induced by 25, 50, or 100 T. notatus per plant, and the middle density of L. daturaphila was chosen to minimize leaf tissue losses during the experiment. The treatment with both herbivore species was designed to compare directly the quality and composition of blends induced by both herbivore species with the HIPV blends elicited by the same numbers of each herbivore species alone. In prior experiments, both adults and larvae of L. daturaphila induced the production of similar groups of HIPVs. In total, plants were aerated continuously for 9 d, including the day before induction, the day of induction, and then daily aerations for 7 d following induction, and traps were changed every 24 hr.

At the end of the experiment, all aerated leaves were removed from the plant, and the leaves damaged by L. daturaphila or M. sexta were scanned and digitized to determine the amount of leaf area removed using the Scion Imaging Software for Windows ver. A.4.0.3.2, (Scion Corp., Frederick, MD, USA). Because of the piercing-sucking feeding mode of T. notatus, no attempt was made to quantify damage caused by this insect. After scanning, all aerated leaves were dried (80°C for 48 hr) and weighed to determine total aerated leaf dry mass. Plant growth was slight over the course of the experiment and was ignored in determining total aerated leaf dry mass.

Volatiles Analysis

Volatiles were eluted from the traps with 150 μl of CH2Cl2 containing 4 ng/μl of 1-bromoheptane (Sigma-Aldrich) as an internal standard into autosampler vials with 250 μl glass inserts, and vials were sealed with crimp caps and PTFE-lined rubber septa. Samples were analyzed by gas-liquid chromatography largely as described in Hare (2007) except that a shorter column (DB-5, 30 m, 0.25 mm ID, 0.25 μm film thickness, J & W Scientific, Folsom, CA, USA) and an autoinjector (Hewlett-Packard 7673) were used. To confirm identifications, selected samples were analyzed with a Hewlett-Packard 6890 gas chromatograph coupled to a Hewlett-Packard 5973 mass selective detector, and the mass spectra and retention times of compounds were matched with those of authentic standards as described in Hare (2007). Peaks were quantified in units of ng g−1 leaf (dry wt.) hr−1 using Agilent ChemStation® software based on comparison of the peak height of each HIPV component with that of the internal standard.

Statistical Analyses

We performed an exploratory principal components analysis (PCA) using PROC PRINCOMP of SAS (SAS Institute, 2008). The covariance matrix of log10 –transformed quantities of the individual HIPVS from induced plants was used to calculate a smaller number of uncorrelated components that were linear, independent combinations of the original variables. The factor loadings show how much each of the original variables contributes to each new component, and factor scores for the new components can be calculated for each original observation. This allows one to examine the composition of the blend as a whole instead of as an ad hoc combination of individual, potentially correlated variables. A repeated measures analysis of covariance on these factor scores (PCA-ANCOVA) was used in the experiment comparing the HIPV blends induced by M. sexta with L. daturaphila to detect significant variation in the composition of the blends induced by the different herbivore species. The analysis was performed using PROC MIXED of SAS that included: 1) insect species, aeration day, and their interaction as fixed effects; 2) block as a random effect; 3) percent leaf area removed as a covariate; and 4) the individual plant as the repeated subject effect over the 9 d of aeration. The results from PCA then guided analyses of individual compounds to explore changes in the composition of the HIPV blend with herbivore species or over time. Where significant interactions were obtained, then they were decomposed by using the SLICE option of PROC MIXED. Similar repeated-measures analyses were performed on the data from the experiment by comparing HIPV blends from plants damaged by L. daturaphila, T. notatus, or both except 1) there was no need to include defoliation level as a covariate, and 2) the effect of trichome genotype and all interactions with trichome genotype were included as fixed effects.

To test for nonadditive effects on the total HIPV production of both L. daturaphila and T. notatus feeding simultaneously on D. wrightii, we calculated the expected HIPV production of plants damaged by each herbivore species alone for each sampling day and block. These expected values were calculated after pooling plants from both trichome genotypes because HIPV production did not differ between trichome genotypes (see Results). The observed HIPV production of plants damaged by both herbivore species simultaneously was compared to these expected values, and the deviation from the expected value was calculated by subtraction. The deviations were not amenable to parametric statistical analysis because they did not meet the assumption of normally-distributed errors and homogeneity of variances, and no common transformation rectified these problems. We, therefore, utilized a non-parametric method of analysis. The frequencies of plants producing greater or lesser quantities of HIPVs than predicted was analyzed by a replicated goodness-of fit G-test (Sokal and Rohlf, 1995) replicated over each sampling day. Under the null hypothesis of additive effects of the two herbivore species on HIPV production, the frequencies of plants producing greater quantities of HIPVs than predicted would not differ significantly from the frequencies of plants producing less HIPVs than predicted; synergism would be consistent with a significantly greater number of plants producing higher quantities of HIPVs than predicted, and antagonism would be consistent with a significantly greater number of plants producing lower quantities of HIPVs than predicted. Data for the pretreatment aeration were not included in this analysis.

Results

Lema and Manduca

A preliminary analysis of total HIPV production showed that HIPV production was low and did not differ among plants assigned to the various experimental treatments on the day prior to induction (F 2, 57 = 0.74, P = 0.48, Fig. 1a). HIPV production of control plants did not differ among days (F 8, 535 = 0.81, P = 0.59), whereas that of the herbivore-damaged plants did differ over time (F 8, 535 = 13.49, P < 0.001 for plants induced by L. daturaphila and F 8, 535 = 12.38, P < 0.001 for plants induced by M. sexta). Highest average total HIPV production of 279 ± 70 ng HIPVs g−1 hr−1 occurred in the plants damaged by L. daturaphila 1 d after induction (Fig. 1a). Because the control plants remained uninduced through the experiment, they were deleted to simplify subsequent analyses of potential differences in the HIPV quantities and blends induced by the two different herbivore species. Mean leaf area removed by L. daturaphila was 10.1 ± 1.0 cm2, or 20.0 ± 2.5% of the available leaf area. Mean leaf area removed by M. sexta was less than expected based upon the preliminary experiment (5.9 ± 1.0 cm2 or 10.1 ± 1.3% of available leaf area), but the distributions of leaf area removed by the two insect species overlapped, with a maximum leaf area removed of 46% for L. daturaphila and 31.5% for M. sexta. In general, the group of HIPVs obtained was similar to those reported in a previous laboratory study (Hare, 2007). Abundant HIPVs included (E)-β-ocimene, (E)-β-caryophyllene, (E,E)-α-farnesene, (Z)-3-hexenyl acetate, and DMNT. Other components included limonene, linalool, an isomer of bergamotene, α-humulene, β-selinene, nerolidol, TMTT, and methyl salicylate (MeSA).

Fig. 1
figure 1

Mean (± s.e.) total HIPVs by day and herbivory treatment for plants damaged for 24 hr between Day 0 and Day 1. Plot a: plants damaged by 12 fourth-instar Lema daturaphila or three third-instar Manduca sexta larvae; plot b: plants damaged by four Lema daturaphila adults, 50 adults and nymphs of Tupiocoris notatus, or the same numbers of both insect species simultaneously

Full size image

Results of the PCA analysis on HIPVs showed that the first component (PC 1) accounted for 63% of the total variation and PC 2 accounted for an additional 12%. PC 3 accounted for 5.5% and the remaining PCs accounted for progressively smaller percentages of the total variance and will not be considered further. All original variables had positive loadings on PC1, thus indicating that this component is largely a "general production" factor much like a "general size" factor in morphological studies (Fig. 2a). The highest loadings were for DMNT, TMTT, (E,E)-α-farnesene, (E)-β-ocimene, (Z)-3-hexenyl acetate, and (E)-β-caryophyllene (Fig. 2a). PC 2 is an index of the composition of the blend and contrasted the abundance of TMTT with DMNT, and to a lesser extent, TMTT with (E,E)-α-farnesene, (E)-β-ocimene, (Z)-3-hexenyl acetate, and (E)-β-caryophyllene. Most of the rest of the components had factor loadings near zero on PC 2 (Fig. 2b). PC 3 independently contrasts plants producing relatively high levels of (E)-β-caryophyllene with plants producing relatively high levels of MeSA, (Z)-3-hexenyl acetate, and DMNT (Fig. 2b).

Fig. 2
figure 2

Factor loadings for individual compounds for plants damaged either by larvae of Lema daturaphila or Manduca sexta for plants damaged for 24 hr between Day 0 and Day 1 as in Fig. 1a . Plot a: PC1 and PC 2; plot b: PC1 and PC 3. Chemical abbreviations: Cary: (E)-β-caryophyllene; DMNT: (E)-4,8-dimethyl-1,3,7-nonatriene; Far: (E,E)-α-farnesene; Hex Ac: (Z)-3-hexenyl acetate; Oci: (E)-β-ocimene; Lol: linalool; M: methyl salicylate; TMTT: (E)-4-8,12-trimethyl-1,3,7,11-tridecatetraene. Overlapping near the origin: Hum: α-humulene; J: jasmone; Lim: limonene; Ner: nerolidol Hex-al: trans-2 hexenal; Hex-ol: cis-3-hexen-1-ol; Hex-1-ol: trans 2-hexen-1-ol; apin: α-pinene B: α-bergamotene; G: geranyl acetone S: β-selinene

Full size image

Scores for PC 1 increased with increasing defoliation (test of covariate, F 1, 506 = 7.87, P = 0.005). There were no differences overall in the factor scores for PC 1 between the two induction treatments (F 1, 506 = 0.00, P = 0.97) after correcting for differences in defoliation, but there was significant variation in factor scores over time (F 8, 506 = 19.14). Scores for PC 1 also differed significantly due to the day by treatment interaction (F 8, 506 = 3.82, P = 0.002, Fig. 3a). Decomposing this interaction showed that PC 1 scores differed between induction treatments starting on the day after damage and continued to differ 6 d after insects were removed (all P ≤ 0.005). During this time period, plants induced by M. sexta had significantly lower scores for PC 1 than plants induced by L. daturaphila, even after correcting for the differences in defoliation; per unit leaf area removed, M. sexta induced a lower production of HIPVs than did L. daturaphila.

Fig. 3
figure 3

Mean (± standard error) factor scores for principal components 1–3 for plants damaged either by larvae of Lema daturaphila or Manduca sexta for 24 hr between Day 0 and Day 1 as in Fig. 1a

Full size image

ANCOVA on the scores for PC 2 showed significant variation among days (F 8, 506 = 9.95, P < 0.001, Fig. 3b) and was due to the day by treatment interaction (F 8, 506 = 3.36, P < 0.001) but not among induction treatments overall (F 1, 506 = 1.34, P = 0.25). Scores for PC 2 were independent of the level of defoliation (test of covariate, F 1, 506 = 1.33, P = 0.25). In general, plants produced blends relatively richer in HIPVs that loaded positively on PC 2 during the day of damage. Afterward, they produced blends relatively richer in HIPVs that loaded negatively on PC 2, principally TMTT, during the remainder of the experiment (Fig. 3b). Although the day by treatment interaction was statistically significant overall, on no day did the scores for PC 2 differ significantly among induction treatments. The closest difference to statistical significance occurred 3 d after induction (P = 0.074), in which the blend from plants induced by L. daturaphila was marginally richer in compounds with positive scores for PC 2 compared to plants induced by M. sexta.

ANCOVA on the factor scores for PC 3 also showed statistically significant variation over days (F 8, 506 = 9.79, P < 0.001, Fig. 3c), and the day by treatment interaction (F 8, 506 = 4.33, P < 0.001). There was no overall variation in PC scores between herbivore species (F 1, 506 = 0.00, P = 0.99) but scores for PC 3 declined significantly with percent defoliation (test of covariate, F 1, 506 = 7.23, P = 0.007). Apparently higher levels of damage were associated with higher proportions of (E)-β-caryophyllene and lower concentrations of most other compounds. Decomposing the day by treatment interaction showed that treatments differed only on the day of induction and the day following (P = 0.003 and 0.03, respectively, Fig. 3c).

The individual compounds that contributed to the differences in PC 2 and PC 3 over time were revealed by comparing the temporal pattern of HIPV emission for the six compounds with the greatest absolute value for factor loadings on PC 2 and PC 3. For (Z)-3-hexenyl acetate, (E)-β-ocimene DMNT, (E)-β-caryophyllene, and (E,E)-α-farnesene, peak concentrations were reached 1 d after herbivores were removed, and HIPV concentration declined regularly over the next 6 d (Fig. 4). By contrast, concentrations of TMTT persisted at near peak concentration for 3 d after insects were removed before declining. The persistence of TMTT during the middle of the observation period, while other HIPVs declined, accounted for the change, with the composition of the blend becoming relatively richer in components loading negatively on PC 2 (i.e., TMTT) over time. The temporal pattern of variation in PC 3 is best reflected in the temporal pattern of variation in (Z)-3-hexenyl acetate (Fig. 4).

Fig. 4
figure 4

Production of (Z)-3-hexenyl acetate, DMNT, (E)-β-ocimene, (E)-β caryophyllene, (E,E)-α-farnesene, and TMTT for plants damaged by Lema daturaphila or Manduca sexta for 24 hr between Day 0 and Day 1 as in Fig. 1a. Mean (± standard errors) in ng g−1 (dry wt.) leaf hr−1 are shown

Full size image

Lema, Tupiocoris, and Both

A preliminary analysis on total HIPV production showed, once again, that HIPV production did not differ among plants to be assigned to the various experiments prior to induction (F 3, 58 = 0.51, P = 0.67, Fig. 1b). HIPV production of control plants again did not differ among days (F 8, 545 = 0.70, P = 0.69) but that for the other three treatments did differ among days (all P ≤ 0.02). Highest average HIPV production of 198 ± 72 ng HIPVs g−1 hr−1 occurred in the plants damaged by both L. daturaphila and T. notatus, 1 d after all insects were removed (Fig. 1b). Because the control plants also remained uninduced through the course of this experiment, they were removed from subsequent analyses to simplify the comparison of HIPVs produced by the different herbivore treatments. Mean leaf area removed by L. daturaphila was 5.06 ± 0.9 cm2, or 5.80 ± 0.66% of available leaf tissue and did not differ between plants attacked by L. daturaphila or by L. daturaphila and T. notatus. In this experiment, the variation in total HIPV production was independent of the variation in herbivore damage (F 1, 330 = 0.96, P = 0.33), probably because of the more consistent level of feeding compared to the first experiment, and there was no need to use defoliation as a covariate in subsequent analyses.

Results of the PCA analysis on HIPVs from insect-damaged plants showed that the first component (PC 1) accounted for 49% of the total variance, and PC 2 accounted for an additional 14.5%. PC 3 accounted for 10% of the total variance; the remaining principal components accounted for less than 10% each of the remaining variance and will not be considered in detail. All original variables except nerolidol had positive loadings on PC 1. The most abundant compounds were generally similar to those in the experiment comparing damage by L. daturaphila and M. sexta (Fig. 5a). In this experiment, PC 2 largely contrasted the abundance of (E)-β-caryophyllene with most of the rest of the HIPVs (Fig. 5a), and PC 3 independently contrasts plants producing high levels of TMTT with plants producing higher DMNT and (Z)-3-hexenyl acetate (Fig. 5b).

Fig. 5
figure 5

Factor loadings for individual compounds for plants damaged by Lema daturaphila, Tupiocoris notatus, or both insect species simultaneously as in Fig. 1b. Plot a: PC1 and PC 2; plot b: PC1 and PC 3. Chemical abbreviations are as in Fig. 2

Full size image

An analysis of variance on the factor scores for PC 1 showed significant differences among induction treatments overall (F 2, 445 = 11.06, P < 0.001), over time (F 8, 445 = 16.58, P < 0.001), and scores for PC 1 also differed significantly due to the day by treatment interaction (F 16, 445 = 1.85, P < 0.001, Fig. 6a). Scores for PC 1 did not differ significantly due to trichome genotype or any interactions involving trichome genotype (all P ≥ 0.16). Decomposing the day by treatment interaction showed that PC 1 scores did not differ before insect damage (F 2, 445 = 0.81, P = 0.45) but did differ among induction treatments starting the day of damage and continued to differ through the end of the experiment (all P ≤ 0.05). On all days, plants damaged by both Lema and Tupiocoris emitted more HIPVs than did plants damaged by either alone, and for all but the last 4 d, plants damaged by Lema emitted greater quantities of HIPVs than plants damaged by Tupiocoris (Fig. 6a).

Fig. 6
figure 6

Mean (± standard error) factor scores for principal components 1–3 for plants damaged either by Lema daturaphila, Tupiocoris notatus, or both insect species simultaneously as in Fig. 1b, for 24 hr between Day 0 and Day 1

Full size image

For the plants damaged by both herbivores, the nonparametric analysis showed that the increased emission of plants damaged did not differ significantly from the null hypothesis of additive effects of damage by each herbivore species individually. Over the 160 observations in the full experiment (4 plants per block infested with both L. daturaphila and T. notatus, 5 blocks, and 8 d of observations), in 86 cases, plants produced lower total HIPV quantities than expected and in 74 cases they produced higher HIPV quantities than expected. These frequencies did not differ from the expected distribution of 80:80 (G 1 = 0.901, P = 0.34). There also was no heterogeneity in the frequency distributions on different days (G7 = 4.359, P = 0.74).

ANOVA on the scores for PC 2, one index of the composition of the HIPV blend, showed no significant variation over time, treatment, or the time by treatment interaction (all P ≥ 0.14, Fig. 6b). The scores for PC 3 differed over time (F 8, 445 = 8.54, P < 0.001) and due to the day by treatment interaction (F 16, 445 = 1.94, P = 0.016). Scores for PC 3 did not differ overall due to treatment (P = 0.19), nor due to trichome genotype (P = 0.15), nor due to any interaction including trichome genotype (all P ≥ 0.15, Fig. 6c). Decomposing the day by treatment interaction showed that scores for PC 3 differed significantly only on the last day of the experiment (F 2, 445 = 4.21, P = 0.016). The largest difference was between plants damaged by L. daturaphila and plants damaged by both herbivores.

Individual Components

The differences in PC 2 and PC 3 over time were revealed by comparing the temporal pattern of HIPV emission for the same 6 compounds as in Experiment # 1, but the temporal dynamics differed somewhat between the experiments. For (Z)-3-hexenyl acetate, (E)-β-ocimene, and DMNT, the highest concentrations were seen on the day after herbivores had been removed, as was the case in the first experiment. Also similar to the first experiment, peak concentrations of TMTT were somewhat delayed. In contrast to the first experiment, peak concentrations of (E)-β-caryophyllene and (E,E)-α-farnesene were not reached on the 1st d after insects were removed, but one to 2 d later, depending upon treatment (Fig. 7).

Fig. 7
figure 7

Production of (Z)-3-hexenyl acetate, DMNT, (E)-β-ocimene, (E)-β caryophyllene, (E,E)-α-farnesene, and TMTT for plants damaged by Lema daturaphila, Tupiocoris notatus, or both insect species simultaneously as in Fig. 1b, for 24 hr between Day 0 and Day 1. Mean (± standard errors) in ng g−1 (dry wt.) leaf hr−1 are shown

Full size image

In summary, induction of plants by both L. daturaphila and T. notatus caused a greater emission of HIPVs than induction of plants by either herbivore species alone, but the increased emissions did not differ significantly from that predicted as the sum of the effects of each herbivore species' damage individually. There was no evidence that the differences in modes of feeding by T. notatus and L. daturaphila resulted in blends of different composition. The greatest source of variation overall was that due to the number of days since damage. The blends never differed consistently due to the differences in trichome genotype.

Discussion

The composition of HIPV blends from D. wrightii induced by feeding damage from M. sexta and L. daturaphila, or L. daturaphila and T. notatus, either alone or in combination, varied substantially with the number of days since damage, but there was no evidence that the HIPV blends characteristically and consistently differed between species of herbivore. Most of the induced monoterpenes [e.g., (E)-β-ocimene) and sesquiterpenes (e.g., (E,E)-α-farnesene, (E)-β-caryophyllene] reached peak emission rates within 1–3 d after the termination of plant damage, then declined over the next 4–6 d. In contrast, concentrations of the 16-carbon homosesquiterpene, TMTT, rose more slowly and persisted longer than the major mono- and sesquiterpenes. As a result, the HIPV blends became richer in TMTT in all treatments as the time since damage increased.

Temporal variation in HIPV production has been noted before in other systems. Commonly, C6 green leaf volatiles are emitted from damaged leaves almost immediately after induction, and the emission of other compounds begins some time later. The amount of time between the emission of green leaf volatiles and other compounds varies among plant species. For example, Turlings et al. (1998) found significant temporal variation in the production of HIPVs in maize over a period of only 12 hr. In cotton, however, inducible terpene production peaked about a day after insect damage was terminated (Loughrin et al., 1994). In Brussels sprouts, 3 d of damage by 100 first- or second-instar caterpillars of Pieris brassicae was necessary to induce sufficient HIPV production to attract its parasitoid Cotesia glomerata, and this response persisted for a day after insect herbivory ceased (Mattiacci et al., 2001). Similarly in Brassica oleraceae, plants subjected to 24 hr of herbivory or treated with JA remained attractive to parasitoids for 3 d after insects were removed and 5 d after JA treatment (Bruinsma et al., 2009).

Not only may total quantities vary temporally, but the production of individual components may follow independent time courses as well. In maize, DMNT was one of the first compounds produced, and the sesquiterpenes (E)-β-caryophyllene and (E)-β-farnesene were produced relatively later (Turlings et al., 1998). In tomato, TMTT and several terpenoids were induced within 3 hr of damage by M. sexta, but MeSA was not emitted until the 2nd d after damage (Farag and Pare, 2002). Different proportions of classes of volatile compounds also are emitted over a period from 1–37 hr in Nicotiana attenuata, with hexenyl esters predominating in the first 13 hr after damage and terpenoids in later time periods (Gaquerel et al., 2009).

Interestingly, different genotypes of plant species may exhibit different temporal profiles of HIPV emission; in one cultivar of maize, indole was emitted as early as DMNT, but in another, its peak emission was delayed to about the same time as the sesquiterpenes (Turlings et al., 1998). Thus, there are numerous precedents from other systems for the temporal variation seen in D. wrightii, not only in the total quantities of HIPVs, but also in the composition of the HIPV blends elicited by a single inducer. Such results suggest that the synthesis of different classes of compounds is regulated on different time scales following damage within species, but temporal pattern of regulation also can vary among species or genotypes within species.

Although the initiation of HIPV production in D. wrightii began on the day of damage and peaked 1–2 d later, the production of induced volatiles persisted for nearly a week. Such long relaxation times seen here and in other systems (Bruinsma et al., 2009) may complicate the efficient use of HIPVs as indicators of herbivores on plants because the HIPVs may be providing false information about the presence of actively-feeding herbivores (Puente et al., 2008a,b). One possible solution to this problem would be if the blend changed sufficiently over time that natural enemies could discriminate old damage from current damage. The solitary specialist parasitoid, Cotesia vestalis, for example, can make such discrimination on the basis of plant breakdown products that are emitted during herbivory but not after. The discrimination is imprecise because wasps still are attracted to plants with old damage when offered plants with old damage and plants with no damage (Kugimiya et al., 2010). The context-dependent nature of the responses of this parasitoid suggests that if there were natural selection on plants for the production of honest cues, then selection should favor the rapid cessation of HIPV emission when herbivores are no longer on those plants, rather than the continued emission of a blend of a different composition.

These experiments with D. wrightii in the laboratory, like most other laboratory experiments, greatly simplify how plants are attacked in nature. Under natural conditions, individual D. wrightii plants occur in patches that persist for several years (Hare and Elle, 2004). Plants quickly become colonized by overwintering adults of L. daturaphila, T. notatus, and other herbivores in the early spring, then suffer chronic herbivory throughout the growing season (Elle and Hare, 2000; Hare and Elle, 2002; Hare, 2010). Commonly, plants are infested with multiple life stages of L. daturaphila and other herbivorous insects throughout the season. It, therefore, is possible that damage inflicted by a single herbivore species for only a single day, as in this laboratory study, exaggerates the potential change in HIPVs over time compared to plants that are attacked under more natural conditions in the field.

Velvety plants infested with T. notatus were included in our experiments even though such infestations rarely occur in the field. We chose to include both trichome genotypes in order to avoid confounding herbivore species with trichome genotype when comparing the HIPV blends of sticky plants infested with T. notatus and L. daturaphila with velvety plants infested only with L. daturaphila¸ as is usually the case in the field. Despite the ability of feeding by T. notatus to suppress some jasmonate-mediated responses in other plant species (Heidel and Baldwin, 2004), we found no persistent differences in the HIPV production and composition of either plant genotype when infested with the combination of T. notatus and L. daturaphila compared to plants infested by L. daturaphila alone. One of our initial motivations to study the HIPV production of D. wrightii was to determine if predators like G. pallens might be able to recognize and avoid sticky plants, on which they are less effective predators (Gassmann and Hare, 2005). This might be possible if 1) the two trichome genotypes differed intrinsically in HIPV production, or 2) if the HIPV production of sticky plants was altered from that of velvety plants by the presence of T. notatus on sticky plants. Taken together with previous research (Hare, 2007), it appears that the HIPV blends of D. wrightii do not vary among trichome genotypes either intrinsically or when the trichome genotypes are attacked by L. daturaphila or L. daturaphila in combination with T. notatus. Moreover, sticky and velvety plants in the field do not differ in the composition of their HIPV blends even though the sticky plants were attacked by T. notatus in combination with L. daturaphila, whereas only the velvety plants were attacked by the latter species (Hare, 2010).

In systems in which HIPV blends are observed to vary with the inducing herbivore species, it may be a consequence of additional elicitors being associated with particular herbivores (Bruinsma et al., 2009; Peiffer and Felton, 2009). The evidence here suggests that there may be little difference in the response of D. wrightii to elicitors from different herbivore species, for there were no consistent, statistically different blends induced by the different herbivore species or combinations of species in the two experiments. Even if there were differences not detectable by the instrumentation, the biological significance of such differences may be slight, especially with regard to generalist predators like G. pallens that utilize all three herbivore species locally as prey. This conclusion is strengthened by the results of field experiments in other systems that show that predation rates of G. pallens can be increased when plant volatiles are augmented simply with a single HIPV component (James, 2003; Halitschke et al., 2008), suggesting that the precise composition of HIPV blends may be of less importance for G. pallens than for other natural enemies.

In summary, our original goals were to determine if three herbivores of D. wrightii produced HIPV blends that might differ consistently to possibly indicate the identity of the herbivore causing the damage. This was not the case. We also asked if simultaneous attack of plants by T. notatus and L. daturaphila might induce HIPV blends that differed qualitatively from those elicited by each insect alone and this also was not the case. The blends did not differ significantly and consistently in composition, rather, simultaneous damage by T. notatus and L. daturaphila simply elevated in an additive manner the HIPV production of plants with both herbivores over that induced by each species independently. Because the HIPV blends did not differ significantly among herbivore species or between trichome types for any combination of herbivores tested, this suggests that HIPV production by D. wrightii is more likely a general than specific indicator of damage by herbivores (Kessler and Baldwin, 2001). The absence of abundant specialist natural enemies of M. sexta, L. daturaphila, or T. notatus that might differentially respond to herbivore-specific HIPV blends may limit the potential for evolution of such herbivore-specific HIPV blends in D. wrightii.