• Herbivore- and jasmonate-induced volatile organic compounds (VOCs), which mediate indirect defense, must provide reliable information for predators that frequently learn to associate their release with feeding herbivores. Yet little is known about variation of these cues within populations of native plants, on a scale encountered by predators.
• We examined variation in herbivore-elicited VOC emissions and patterns of herbivore-induced jasmonate signaling from accessions of Nicotiana attenuata co-occurring in a native population. VOC emissions elicited by herbivore oral secretions (OS) and by methyl jasmonate (MJ) were characterized using gas chromatography–mass spectrometry (GC-MS), high-resolution two-dimensional gas chromatography–time-of-flight mass spectrometry (GCxGC-ToF-MS) and micro-hydrolysis and micro-hydrogenation reactions.
• Accessions varied in emissions of abundant (trans-α-bergamotene, α-duprezianene, trans-β-ocimene, and cis-3-hexenol) and total detectable VOCs, as well as the accumulation of jasmonates, the jasmonate antagonist salicylic acid (SA), abscisic acid (ABA) and jasmonate signaling-related transcripts after OS elicitation. Yet MJ treatment exacerbated differences in VOC emission, suggesting that much variation in VOC emission is caused by processes downstream of jasmonate signaling.
• Co-occurring N. attenuata plants emit different VOCs following simulated herbivore elicitation as a result in part of differences in jasmonate production and responsiveness, which could reduce the effectiveness of induced indirect defense.
Wild plants use many strategies to survive herbivore attack, including the constitutive or inducible production of toxic and antidigestive compounds (direct defenses); attraction, via volatile compounds or rewards such as extrafloral nectar, of predators and parasitoids of herbivores (indirect defenses); physical barriers to herbivore feeding; and enhanced growth or sugar reallocation, to ensure sufficient energy stores for reproduction regardless of tissue loss to herbivores (Constabel, 1999; Schwachtje & Baldwin, 2008). Indirect defenses mediated by induced volatile signaling can be a particularly efficient strategy, because the investment of fitness-limiting resources to produce volatile compounds is probably less than that required to activate direct defenses (Gershenzon, 1994; Halitschke et al., 2000). The protection afforded by indirect defenses, which can completely remove or disable herbivores at an early stage of attack, can be greater than that resulting from the deployment of direct defenses, which affects herbivore growth and preference but usually does not cause adapted herbivores to leave the plant (Kessler & Baldwin, 2001; Steppuhn et al., 2004).
Plants of the native tobacco Nicotiana attenuata (Solanaceae) germinate after fires from long-lived seed banks to form monocultures in the Great Basin desert, where they are attacked by an unpredictable herbivore community. Feeding by larvae of Manduca sexta and Manduca quinquemaculata (Lepidoptera, Sphingidae), however, frequently accounts for most of the leaf area removed in a given season (Kessler & Baldwin, 2004). Plants employ an effective indirect defense strategy against Manduca via the jasmonate-mediated release of terpenoid and fatty acid-derived volatile organic compounds (VOCs) that begin to attract native predatory insects of the genus Geocoris (Hemiptera, Lygaeidae) by the day following herbivore elicitation (Halitschke et al., 2008; Skibbe et al., 2008).
Monocultures of N. attenuata can be highly diverse: plants within a population are likely to be genetically less similar than plants from different populations (Bahulikar et al., 2004). Nevertheless, phytohormone signaling in response to herbivore attack appears to be consistent in native populations. Wild N. attenuata plants reliably produce a wounding-induced jasmonate burst which is amplified by fatty acid–amino acid conjugates (FACs) in the oral secretions (OS) from M. sexta larvae (Kahl et al., 2000; Halitschke et al., 2001; Schittko et al., 2001; Stork et al., 2009), and jasmonate-mediated responses provide a significant benefit to plants in wild populations (Baldwin, 1998). Yet it is not known how similar the magnitude or quality of these responses is among plants in a population.
For herbivore-induced VOC emission to function as an effective indirect defense, insects must learn to associate bouquets of (variable) scents with their prey, which must indicate the presence or absence of prey feeding on the correct host plant species (Dicke et al., 2003; van Dam & Poppy, 2008). The Geocoris predators native to N. attenuata's habitat might be capable of such adaptive learning, as has been shown for other predators and parasitoids (Takabayashi, 1994; Drukker et al., 2000; de Boer & Dicke, 2006; Tamòet al., 2006). To address mechanisms of predator adaptive learning, we must first understand the variation in the VOC signals that single Geocoris individuals could encounter in N. attenuata populations, and in what ways these signals might be reliable indicators of herbivore presence.
We therefore asked: How do VOC emissions vary among co-occurring accessions of N. attenuata after simulated herbivore attack, and to what extent does variation correspond to differences in herbivore-induced signaling? We collected seeds from native N. attenuata plants in a wild population (accessions 1, 2, 3, and 4, F0) on a spatial scale that might be encountered by an individual predator, and grew them in a controlled environment together with an inbred line (UT, F22) to reveal genetically rather than environmentally controlled variation among individual plants and accessions. We measured the emission of VOCs 24–32 h following treatment of plants with wounding plus M. sexta oral secretions (W+OS) and again 24–32 h after treatment of the same plants with methyl jasmonate (MJ). For each accession, we also determined constitutive (control), wounding (W)-induced, and W+OS-induced production of hormones known to be involved in herbivore-responsive signaling, namely jasmonic acid (JA) and the jasmonoyl–isoleucine conjugate (JA-Ile), salicylic acid (SA) and abscisic acid (ABA), as well as transcript accumulation of genes involved in OS perception, jasmonate biosynthesis and signaling, to better understand the sources of variation in jasmonate and VOC signaling.
Materials and Methods
Plant material and growth conditions
Seeds were collected from four plants growing in a native population of Nicotiana attenuata Torrey ex Watson at a burn site near Santa Clara (UT, USA), in July 2007. Parent plants of accessions 1, 2, and 3 were growing within 1 m of each other near a burned juniper tree (Juniperus spp.; N 37 04.594, W 113 49.994); the parent plant of accession 4 was located c. 65 m downhill (N 37 04.559, W 113 49.979). The well-characterized inbred line ‘UT’ which we used as a control comparison was collected from southwestern UT in 1996 (population U in Glawe et al., 2003, c. 6.5 km west of the 2007 population) and had been self-fertilized in the glasshouse in Jena, Germany for 22 generations. Seed germination and glasshouse growth conditions are described in Krügel et al. (2002).
Analyses of signaling gene transcripts and phytohormones
Plant treatments and sampling were randomized temporally among accessions, and plants were spatially randomized before elicitation and sampling. To determine herbivore-induced transcriptional and phytohormone signaling responses (n = 5), plants were either left untreated (control), or wounded by using a pattern wheel which produced three rows of puncture wounds on either side of the midvein of the first fully expanded leaf (position + 1). Puncture wounds were immediately treated with 20 µl of either distilled water (W) or oral secretions (OS) from M. sexta larvae diluted 1:5 with distilled water (W+OS), referred to as ‘W+OS elicitation’. Water or OS were pipetted onto the wounded leaf and gently dispersed across the surface with a gloved finger. Gloves were changed between treatments. The +1 leaves of treated plants were removed after 1 h and divided lengthwise along each side of the midvein, and each half was separately flash-frozen in liquid nitrogen and stored at −80°C until extraction. Harvesting of +1 leaves in the same manner from control plants was interspersed throughout the treatment and harvesting of W- and W+OS-treated leaves.
For phytohormone analysis, we extracted tissue from the right halves of leaves (as seen from the petiole) in ethyl acetate spiked with 100 ng each of 9,10-dideutero-9,10-dihydro-jasmonic acid (JA-D2), jasmonoyl isoleucine* (JA-Ile-13C6), 3,4,5,6-tetradeutero salicylic acid (SA-D4), and hexadeutero abscisic acid (ABA-D6) as internal standards (ISs), re-suspended extracts in 70% methanol and analyzed phytohormone content by high pressure liquid chromatography tandem mass spectrometry (HPLC-MS/MS) as described in Wang et al. (2007a). Phytohormones were quantified as ng IS g−1 fresh mass (FM).
RNA was extracted from the left halves of leaves with Tri Reagent (Chomczynski, 1993), checked on an agarose gel and quantified on a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA). Synthesis of cDNA from 0.5 µg of RNA per sample and qPCR analyses were conducted as in Wu et al. (2007): TaqMan (Applied Biosystems, Foster, CA, USA) was used for N. attenuata lipoxygenase 3 (Nalox3; Halitschke & Baldwin, 2003) and N. attenuata wound-induced protein kinase (Nawipk; Wu et al., 2007 and SYBR (Invitrogen, Carlsbad, CA, USA) was used for Najaz1, homologous to Nicotiana tabacum JAZ1 in Shoji et al. (2008) (I. Galis, unpublished data; forward primer ACTTACCGCAGTTTTGAACC, reverse primer TGTGCCTTTTCTGGTTCAGA). Expression was quantified relative to N. attenuata actin.
Collection of plant VOCs
Plant treatments and sampling were randomized temporally among accessions, and plants were spatially randomized before elicitation and sampling. Eight rosette-stage plants per accession were W+OS-elicited on the second fully expanded (+2) leaf and placed c. 0.3 m apart on a single table in the glasshouse. Leaves near the source–sink transition leaf respond the most to treatment (van Dam et al., 2001), but VOC emission may be greatest after treatment of position +2 (Halitschke et al., 2000). From 24 to 32 h after treatment, treated leaves were enclosed in two shallow 50-ml plastic cups (Huhtamaki, Espoo, Finland) secured with miniature claw-style hair clips, and headspace VOCs were collected on 20 mg of SuperQ (Alltech, Deerfield, IL, USA) by drawing ambient air through these clip cages as described in Wu et al. (2008) and Halitschke et al. (2000) for 8 h. Background VOCs present in ambient air were collected using empty trapping containers. SuperQ-packed filters were either immediately spiked with 400 ng of tetraline as an IS and eluted with 250 µl of dichloromethane into a GC vial containing a glass insert, or stored at −20°C until elution.
The same eight plants per accession were allowed to recover for 1 wk, at which point plants were elongated but not flowering. To determine the average maximum jasmonate-elicited VOC response of each accession, plants were treated with 150 µg of MJ (Sigma-Aldrich, St. Louis, MO, USA; 95%, mixture of isomers) dissolved in 20 µl of lanolin paste, which was gently spread across the base of the upper leaf surface (position +2, now the oldest stem leaf) using a blunt spatula. We again collected VOCs from 24 to 32 h from the treated leaf.
To obtain an indication of developmental differences in VOC emission, an additional group of UT plants was treated at the rosette stage with W+OS, MJ, or no treatment (control), n = 6/treatment, and VOCs were collected from 24 to 32 h after treatment from the treated leaf (+2). The plants were allowed to rest for a week, after which time elongated W+OS-treated and control plants were elicited with MJ, and VOCs were collected after 24 to 32 h from the treated leaves (+2).
Analysis of VOC emissions
For GC-ion trap MS analysis, analytes in plant total VOC samples were separated on a nonpolar FactorFour VF-5ms column (30 m × 0.25 mm i.d. × 0.25 mm; Varian Inc., Lake Forest, CA, USA) in a Varian CP-3800 GC (Varian Inc., Palo Alto, CA, USA) equipped with a CP-8400 auto-injector operated in splitless mode, and analyzed on a coupled Varian Saturn 4000 ion trap mass spectrometer: GC, helium carrier gas, 1 µl of 30-s splitless injection at 250°C, initial temperature 40°C for 5 min, increased at 5°C min−1 to 185°C, and at 30°C min−1 to 300°C, 10-min hold; MS, transfer line at 230°C, trap temperature 180°C, scan range from 40 to 399 m/z at 1 spectra s−1. The most abundant analytes across all samples (trans-α-bergamotene, α-duprezianene, cis-3-hexenol, and trans-β-ocimene) were identified for further analysis. Individual VOCs were quantified relative to the peak area of the tetraline IS in each sample as either the relative amount of VOC in ng IS per g FM of the trapped leaf (Fig. 1) or as percentage IS (untargeted analysis), which did not change significant relationships among samples.
For the GCxGC-ToF analysis, the same VOC samples were run on an Agilent 6890N gas chromatograph equipped with an Agilent 7683 auto-injector (Agilent Technologies, Santa Clara, CA, USA) coupled with a LECO Pegasus III time-of-flight mass spectrometer with a 4D thermal modulator upgrade (LECO, St. Joseph, MI, USA). Injected samples were separated first on a nonpolar column (C1 RTX-5MS, 20 m × 250 µm i.d. × 0.5 µm; Restek, Bellefonte, PA, USA) and every 6 s (modulation time) transferred to a midpolar column (DB-17, 0.890 m × 100 µm i.d. × 0.1 µm; Agilent Technologies) for the second separation. Chromatography and analysis conditions as well as deconvolution, alignment, and integration of VOC analyte peaks are described in Gaquerel et al. (2009). During peak table alignment using the comparison feature imbedded in the ChromaToF software (LECO), mass spectra alignment was accepted at a similarity threshold of 500/1000.
A C9-C24 n-alkane series was used to determine the Kovats retention index (RI; Kovats, 1965). Linear retention times (RTs) on the GCxGC-ToF were obtained for RI calculation from the sum of the RT values of each compound on C1 and C2, as previously reported (Kusano et al., 2007). All identifications were based on mass spectra and RTs compared with those of authentic reference compounds. When identification was not possible, we tried to determine the class of VOC by typical fragmentation patterns combined with an RT similar to those of standard compounds of the same class; otherwise the compound was listed as unknown.
Identification of α-duprezianane by high-resolution GC-MS of crude and hydrogenated plant volatile extract
GC-MS experiments were performed using an HP 8570 series gas chromatograph (Agilent) connected to a MasSpec sector-field mass spectrometer (Micromass, Manchester, UK). A nonpolar DB-5 MS column (30 m long, 0.25 mm i.d., phase thickness 0.25 mm; Agilent J & W Scientific, Santa Clara, CA, USA) was used for separations. The injector was operated in splitless mode at 220°C. The detector temperature was set to 200°C; standard 70-eV spectra were recorded at 1 scan s−1. The temperature of the GC oven was programmed as: 60°C, hold for 2 min; 10°C min−1 to 320°C and hold at 320°C for 10 min. Helium was used as a carrier gas at constant flow of 0.7 ml min−1. Data were analyzed using the opus software (San Francisco, CA, USA) and Wiley version 6 (Wiley, Hoboken, NJ, USA) and NIST libraries (National Institute of Standards and Technology, Gaithersburg, MD, USA) were used for spectra data searches. High-resolution MS electron ionization (EI) data were obtained using a MasSpec 2 instrument (Micromass) in positive ion mode using 70-eV ionization energy (see Supporting Information Fig. S1). A perfluorokerosine mixture was used as an IS.
Dichloromethane from a plant volatile trapping extract was evaporated and ethylacetate (0.5 ml) was added. A platinum on carbon (Pt/C, 10%) catalyst was added (c. 2 mg), and the reaction vial was evacuated and purged with hydrogen three times. Micro-scale hydrogenation (Marques et al., 2004) was performed under vigorous stirring at normal pressure for 1 h. The black suspension was filtered over cotton wool in a Pasteur pipette and the clear solution was analyzed on GC-MS as described at the beginning of this section.
Spectra and RT in the crude plant volatile extract were compared with those in a sample of α-duprezianane kindly provided by Dr A. F. Barrero run sequentially on both a nonpolar and a polar GC column, and by co-injection.
Levene tests for homogeneity of variance were performed on untransformed data. To meet requirements of normality and homogeneity of variance, and to include 0 values, 1 was added to the raw areas of all VOC analytes from 1D and 2D analyses, and, after normalization to the IS, data were log2-transformed; transcript and phytohormone levels were also –log2-transformed (transcripts) or log2-transformed. Differentially emitted VOCs were tested by one-way ANOVAs (P < 0.05) across all accessions, and those that showed significant differences were then tested by MANOVA (Table S7), to determine whether the significance remained when the individual VOC concentrations were considered to be a multivariate data set; because of the number of detected VOCs it was not possible to test all by MANOVA. Scheffe's post-hoc tests were conducted for the four most abundant VOCs. Transcripts and phytohormones were analyzed by two-way ANOVAs across treatments (control, W and W+OS) and accessions followed by Scheffe's post-hoc tests. Pearson's product-moment correlation analyses were performed using R (R Development Core Team, 2005) on untransformed data. When samples were excluded from analysis, this was always because of problems arising during sampling or processing (very poor signal quality for all compounds measured or from IS(s); loss of sample during extraction). ANOVAs were calculated using TIGR MeV 3.1 software (TM4 Microarray Software Suite, Boston, MA, USA, http://www.tm4.org/mev.html; time of flight (ToF) measurements) or StatView (version 5.0; SAS Institute, Cary, NC, USA) and MANOVAs and the correlation table were calculated in StatView.
Identification of α-duprezianane
A previously unidentified peak at 11.45 min showed a typical bi- or tricyclic sequiterpene mass spectrum (abundant M+• at 204 (C15H24), and 161, 119, 93 and 69 fragments) with a prominent m/z 148: presumably a neutral loss of butene (C4H8) (Fig. S1). The number of double bonds was determined by hydrogenation over Pt/C (10%), providing evidence for one unsaturation (peak at 11.50 min; m/z 206). Notably, a more intense m/z 206 peak at a later RT (12.08 min) was detected showing two dominant neutral losses of C2H4 at m/z 178 and C4H8 at m/z 150. The double bond in the original C15H24 alkene was not reduced, as the C4H8 loss was still apparent, but rather one of the rings was hydrogenolyzed. The unidentified sesquiterpene was tentatively identified as a tricyclic alkene. Based on structures available in the literature (Barrero et al., 1996, 2000) this compound was determined to be α-duprezianane. This assignment was fully supported by co-injections with an authentic standard from Dr A. F. Barrero.
Emission of VOCs differs significantly among accessions when plants are elicited with M. sexta OS
The most abundant VOCs detected in emissions from accessions after W+OS treatment were two sesquiterpenes, the monoterpene trans-β-ocimene, and the fatty acid derivative cis-3-hexenol. The sesquiterpenes were subsequently identified as trans-α-bergamotene (rather than the cis isomer previously reported in N. attenuata) and α-duprezianene. The corrected identification of the α-bergamotene isomer does not influence previous work in this system, because the isomer of α-bergamotene purified for use in former experiments was also determined to be trans- rather than cis-α-bergamotene (N. Heinzel, unpublished data).
The emission of the four most abundant VOCs (Fig. 1, Table S7) differed significantly among accessions following W+OS treatment; however, trans-β-ocimene emission did not differ significantly among the accessions producing this VOC (1, 2, 3, and 4, but not UT) (one-way ANOVAs; final n for each accession: 1, n = 8; 2, n = 8; 3, n = 4; 4, n = 8; UT, n = 7; trans-α-bergamotene: F4,30 = 9.17, P < 0.0001; trans-β-ocimene: F4,30 = 34.3, P < 0.0001; cis-3-hexenol: F4,30 = 2.91, P = 0.038; α-duprezianene: F4,30 = 2.76, P = 0.046). A Pearson's correlation matrix showed that emission levels were unrelated except for trans-β-ocimene and cis-3-hexenol which were marginally correlated (Table S8). On average, accessions from neighboring plants (1, 2, and 3) were no more similar to each other in their emission of these VOCs than they were to an accession collected from a different location in the same population (4) (Fig. 1). Individual plants within each field-collected accession did not differ more from sibling plants in their VOC emissions than did plants of the F22 UT line: Levene's test for homogeneity of variance revealed nonsignificant differences in the variance of VOC emission within each accession after W+OS treatment, with the exception of trans-α-bergamotene, which was attributable to the fact that only one plant of accession 1 emitted trans-α-bergamotene (trans-α-bergamotene: F4,30 = 2.89, P = 0.039; trans-β-ocimene: F4,30 = 2.37, P = 0.075; cis-3-hexenol: F4,30 = 1.00, P = 0.422; α-duprezianene: F4,30 = 1.38, P = 0.27). When accession 1 was removed from the analysis, the P-value was no longer significant (F3,23 = 1.84, P = 0.168).
We extended our investigation to the large-scale profiling of VOC blends emitted by each accession. Because plant VOCs are highly structurally diverse (Holopainen, 2004) and emitted in a large range of concentrations, we analyzed the same volatile extracts by two-dimensional gas chromatography coupled to time-of-flight mass spectrometry (GCxGC-ToF-MS) to fully resolve these complex mixtures. Seventeen additional volatiles and trans-α-bergamotene, trans-β-ocimene and cis-3-hexenol were detected as statistically different among accessions (ANOVAs for each compound; Fig. 2, Tables S1, S7).
Accumulation of phytohormones and transcripts of signaling genes amongst accessions
We tested the hypothesis that differences in VOC emissions from W+OS-treated plants correlate with differences in W+OS-induced signaling. Control, W-, and OS-induced concentrations of phytohormones known to mediate plant–herbivore interactions (JA, JA-Ile, SA and ABA) differed significantly among accessions (P ≤ 0.05 in two-way ANOVAs with accession and treatment as factors; Table S2). Maximum concentrations of JA and JA-Ile occur at approx. 1 h after wounding and are amplified by OS treatment of the puncture wounds in the inbred UT line and in native accessions of N. attenuata (Wang et al., 2007a; Stork et al., 2009). By contrast, concentrations of SA increase within 1 h of wounding and are not further elevated by treatment with M. sexta OS (C. Diezel, C.C. von Dahl, R. Halitschke and I.T. Baldwin, unpublished data). Thus we found no significant effect of W+OS treatment on SA concentrations at 1 h. For all three phytohormones, the greatest differences among accessions were seen after W+OS elicitation (Table S2, Fig. 3), suggesting different abilities to transduce OS perception into phytohormone production among accessions (Fig. 4d). W+OS-induced phytohormone values were similar among plants within each accession (Levene's test of variance; JA: F4,20 = 0.988, P = 0.437; JA-Ile: F4,20 = 0.987, P = 0.437; SA: F4,20 = 0.870, P = 0.500; ABA: F4,20 = 0.548, P = 0.702).
JA and SA concentrations in the same leaves were strongly and negatively correlated across all accessions after W+OS treatment, whereas JA-Ile and ABA concentrations in the same leaves were positively correlated after W+OS treatment (Fig. 3e,f; Pearson's product-moment correlations; JA vs SA: r = −0.401, P = 0.047; JA-Ile vs ABA: r = 0.497, P = 0.012). Interestingly, only JA-Ile and ABA showed significantly different patterns of induction by W and W+OS treatments across accessions (Table S2). There was no correlation between JA and JA-Ile, JA-Ile and SA, or SA and ABA concentrations.
W- and OS-elicited accumulation of signaling gene transcripts was analyzed in the same leaves used for phytohormone analysis. We tested for differences among accessions in their induction of gene transcripts involved in OS perception (Nawipk), JA biosynthesis (Nalox3), and JA perception (Najaz1) (P ≤ 0.05 in two-way ANOVAs with accession and treatment as factors; Fig. 4, Table S3). Transcripts of these genes are known to be up-regulated within 1 h of W or W+OS treatment (Skibbe et al., 2008; Meldau et al., 2009; I. Galis, unpublished data). Transcripts were significantly elevated in response to wounding and amplified by OS in all accessions, but overall transcript accumulation of Nalox3 and Najaz1 differed significantly, and accessions differed in their treatment response for Nawipk (Table S3). Levene's tests of variance revealed that individuals from field-collected seed did not differ from each other more than did individuals of the inbred line UT (Nawipk: F4,20 = 0.502, P = 0.735; Nalox3: F4,20 = 0.805, P = 0.537; Najaz1: F4,20 = 0.348, P = 0.842).
Elicited concentrations of JA and JA-Ile were highly correlated with transcript accumulation as predicted by the current model of herbivore-responsive JA signaling in N. attenuata (Fig. 4a,b,d): all Pearson's correlation coefficients were ≥ 0.728 and highly significant (P < 0.0001). Thus the model applies to genetically variable native plants as well as to the inbred line UT in which it has been elucidated (see for example Wu et al., 2007; Wang et al., 2007a,b; Paschold et al., 2008). There was also a significant but weak correlation between transcript accumulation and ABA concentration after W+OS treatment across accessions (Pearson's product-moment correlations; ABA vs Nalox3: r = 0.264, P = 0.025; ABA vs Nawipk: r = 0.241, P = 0.042; ABA vs Najaz1: r = 0.339, P = 0.004) However, in W+OS-elicited samples, only differences in JA, not differences in JA-Ile or ABA, were correlated to differences in transcript accumulation (Fig. 4c).
Correlation between W+OS-induced endogenous jasmonate signaling and VOC emissions
Only average emissions of trans-α-bergamotene, an unidentified sesquiterpene, and methyl salicylate showed strong correlations across accessions to average JA and SA concentrations after W+OS treatment: trans-α-bergamotene and the unidentified sesquiterpene were positively correlated to JA and negatively correlated to SA, whereas methyl salicylate showed the opposite pattern (Fig. 5). This indicates that variability in W+OS-elicited JA production and signaling accounted for few of the differences in VOC release.
W-induced JA concentrations showed a similar correlation to these three VOCs (Pearson's product-moment correlation; JA vs trans-α-bergamotene: r = 0.910, P = 0.032; JA vs unidentified sesquiterpene: r = 0.952, P = 0.013; JA vs methyl salicylate: r = −0.924, P = 0.025). Control concentrations of JA were too low to allow detection of any relationship to volatile emission after W+OS treatment. Although there was no significant induction of SA by W or W+OS, control concentrations of SA were correlated only to methyl salicylate emission (r = 0.986, P = 0.002), perhaps reflecting an effect of induction on SA–JA crosstalk. Average JA-Ile and ABA concentrations in W+OS-treated plants were not correlated to any of the VOCs which were differentially emitted among accessions after W+OS treatment.
Total detectable VOC emission after MJ treatment
To determine whether differences in OS-elicited volatile emission are explained by differences in endogenous jasmonate signaling, we conducted a separate experiment to measure MJ-elicited VOC emissions. MJ treatment infuses plants with jasmonates and thereby short-circuits the transduction of OS into endogenous JA production. Because we did not yet know how great the variance among individuals within an accession might be, we used the same plants but allowed them to recover for 1 wk from the previous OS elicitation, at which time plants were elongated but not yet flowering. In a previous experiment using the same elicitation and trapping methods, we found that, for the UT line, 1 wk is sufficient to eliminate any effect of OS elicitation on MJ-elicited VOC emissions (Table S4), and the VOCs that are emitted by elongated plants after MJ treatment often differ quantitatively, but generally not qualitatively, from those emitted by rosette-stage plants (Table S5). A total of 43 compounds differed among MJ-elicited accessions, but only 20 differed after W+OS elicitation (Tables S1, S6, S7, Figs 2, 6), showing that MJ treatment did not produce comparable emissions among accessions but rather may have increased differences. A Pearson's correlation matrix of the four most abundant VOCs showed a significant positive correlation between trans-β-ocimene and trans-α-bergamotene, and a significant negative correlation between trans-β-ocimene and α-duprezianene emissions (Table S8).
We tested the hypothesis that co-occurring accessions (1–4) of the wild tobacco N. attenuata would emit similar VOC profiles after herbivore elicitation, and if not, that differences could be related to differences in endogenous jasmonate signaling. VOC blends emitted after simulated herbivory (W+OS) or jasmonate supplementation (MJ) were characterized in two separate experiments on the same plants, using both targeted (most abundant components) and untargeted approaches (Figs 1, 2), for a period approximating the time at which predators begin to respond to indirect defense signals (24–32 h after treatment; see for example Skibbe et al., 2008). Accessions displayed striking differences in VOC emissions and elicited signaling traits (Figs 1–4), but phenotypes were stable within accessions (Levene's tests of variance for individual traits).
Targeted analysis of the most abundant VOCs emitted by accessions after W+OS elicitation (Fig. 1, Table S7) revealed substantial variation, which motivated us to investigate the total volatile blends trapped from the leaves of each accession by GCxGC-ToF analysis. This sensitive detection technique has been shown to yield high-quality mass spectra obtained from the enhanced chromatographic resolution (Shellie et al., 2001) which accounts for its rapid adoption in nontargeted studies (Shellie et al., 2005; Kusano et al., 2007; Tu et al., 2007; Gaquerel et al., 2009). Untargeted analyses revealed differential emission among accessions for 20 VOCs after W+OS treatment (Figs 2, 6, Tables S1, S7) and for 43 after subsequent MJ treatment of the same plants (Fig. 6, Tables S6, S7). Because VOC emissions were found to be qualitatively if not quantitatively similar across the time frame of the two experiments, a comprehensive qualitative analysis of VOCs emitted by N. attenuata was possible (Tables S5 and S6), which revealed that differences among accessions were increased rather than obviated by MJ treatment, in terms of both the number of VOCs and the magnitude of some differences (Fig. 6).
Endogenous differences in concentrations of jasmonates (JA and JA-Ile) correlated well with differences in induced transcript accumulation among accessions (Fig. 4a,b). This pattern is consistent with the model of jasmonate-mediated, W+OS-elicited defense responses deduced from extensive work with transformed lines of N. attenuata silenced in various genes in these signaling pathways (Fig. 4d). Correlations among phytohormones demonstrated cross-talk between JA and SA signaling in all accessions, as has been shown by Spoel et al. (2003), Traw et al. (2003) and van Leeuwen et al. (2007) to occur variably in accessions of Arabidopsis thaliana. Interestingly, a positive correlation between JA-Ile and ABA was also discovered (Fig. 3c,d). ABA is thought to be involved in both synergistic and antagonistic modulation of JA signaling in defense, but its role, especially in herbivore defense, is unclear (Asselbergh et al., 2008; Bodenhausen & Reymond, 2007). The induction of JA-Ile and ABA relative to control concentrations, but not of JA or SA, varied significantly among accessions after W and W+OS treatment (Table S2). Thus, patterns of herbivore-elicited accumulation may be more conserved for JA and SA than for JA-Ile and ABA among individual N. attenuata plants. This is interesting given that only differences in JA concentrations corresponded to different transcript levels of signaling genes in W+OS-treated leaves (Fig. 4c). Also, only differences in JA and SA accumulation were correlated with differences in VOC emission after W+OS treatment; this was true for only three of 20 VOCs (Figs 2, 5). Because elicitation of the same plants with MJ resulted in more rather than fewer differences (Fig. 6), these combined results suggest that most of the significant qualitative differences in VOC profiles after W+OS treatment are explained by differences independent from, or downstream of, OS-elicited jasmonate production.
Perhaps the most interesting result from the correlation of phytohormones with other responses is that endogenous peak JA concentrations are more indicative of transcript regulation and jasmonate-mediated VOC production than are peak JA-Ile concentrations, although JA-Ile and not JA has been shown to play a direct role in jasmonate signal perception (Chini et al., 2007; Thines et al., 2007; Katsir et al., 2008). Van Poecke & Dicke (2003) found that indirect defense ability was not affected in JA-Ile-deficient A. thaliana plants. Wang et al. (2007b) showed that many jasmonate-mediated responses in N. attenuata are dependent on jasmonates other than JA-Ile; JA-Ile also seems not to play a role in the indirect defense of N. attenuata (E. Gaquerel, N. Heinzel, M. Schuman and S. Meldau, unpublished data). Yet N. attenuata transformants with significantly impaired jasmonate biosynthesis and signaling showed reduced indirect defense ability in the same environment from which these accessions originated (Halitschke et al., 2008; Skibbe et al., 2008). Thus, although most differences in volatile profiles do not correlate well to differences in jasmonate signaling, it is likely that those that do are important for indirect defense, and the mechanism underlying this relationship is still poorly understood. Comparisons of two N. attenuata inbred lines originating from different accessions, the UT line and a line from Arizona (AZ), revealed that AZ is not impaired in indirect defense when compared with UT in the UT native habitat, despite having c. 50% less JA and producing no trans-α-bergamotene (Glawe et al., 2003; Steppuhn et al., 2008; Wu et al., 2008), similar to accession 1. Many VOCs may be able to serve as effective indirect defense signals (Dicke et al., 2003), and there may be species-specific threshold levels of JA which are important for their elicitation (Stork et al., 2009).
For trans-α-bergamotene, which has been shown to function as an indirect defense in nature (Kessler & Baldwin, 2001; Halitschke et al., 2008), differences in emission after W+OS treatment were significantly and positively correlated with average endogenous JA concentrations in each accession 1 h after W+OS treatment, and negatively correlated with endogenous SA concentrations after W+OS treatment (Fig. 4). The same was true of only two other components found in total detectable leaf VOC profiles: an unidentified sesquiterpene which is thought to be co-regulated with trans-α-bergamotene (Gaquerel et al., 2009), and methyl salicylate, for which emission levels corresponded well to the evident cross-talk between JA and SA in phytohormone analyses of individual leaves (Fig. 3). Supplementing plants with MJ to eliminate differences among accessions in OS-elicited jasmonate production did reduce differences in the emission of trans-α-bergamotene, which we may directly compare between the separate W+OS and MJ elicitations, because it is emitted at comparable levels across the plant growth stages used in these experiments (Table S5). Thus jasmonate deficiency may account for reduced trans-α-bergamotene production in some, but not all accessions: accession 1 emitted little or no trans-α-bergamotene after either treatment, and is probably mutated in the biosynthesis in this sesquiterpene rather than its signaling.
Because we conducted our experiments in a controlled environment, we assume that the differences we detected have a genetic basis. Delphia et al. (2009) found that variation in VOC emissions in co-occurring genotypes of horse nettle (Solanum carolinense) was heritable, and that inbreeding reduced variety, which is interesting given that we found several VOCs present in wild accessions that were not detected in UT (Fig. 2). Accessions of N. attenuata are probably polymorphic for one or more signaling genes as well as one or more genes directly controlling volatile biosynthesis (accession 1 produces almost no trans-α-bergamotene, and UT does not produce the monoterpene trans-β-ocimene: Fig. 1, Table S6). Wu et al. (2008) and Steppuhn et al. (2008) found many of the same differences between the UT and AZ accessions, and Wu et al. (2008) showed that these differences could be accounted for in part by expression level polymorphisms (ELPs) in known signaling genes such as Nawipk (see Fig. 4); ELPs may be particularly important in explaining intraspecific variation (Caroll, 2008). Polymorphism in a single biosynthetic gene can also have a significant impact on defense signaling: a study by Pajerowska-Mukhtar et al. (2008) showed that different alleles of the Solanum tuberosum gene for allene oxide synthase (AOS, an enzyme necessary for jasmonate biosynthesis; Fig. 4d) resulted in different levels of resistance to pathogens when expressed in A. thaliana AOS null transformants. However, mechanisms underlying variation may be multiple and complex. Kliebenstein et al. (2002) found that, for most glucosinolates produced by A. thaliana, differences among accessions are influenced by multiple quantitative trait loci (QTLs), which also determine the plasticity of glucosinolate production in response to MJ and SA induction. Furthermore, epigenetic mechanisms are emerging as important players in the polymorphism of herbivore-induced responses (Pandey & Baldwin, 2007).
Little is known about the mechanisms of functional genetic diversity in N. attenuata or its role in the fitness of individual plants. Polymorphism in the signaling response to herbivore attack may be widespread within as well as between populations, as is thought to be the case for neutral genetic variation in this species as a result of its long and variable seed dormancy and recent spread (Bahulikar et al., 2004). We speculate that herbivores of N. attenuata and their predators must cope with a great variety of signaling and defense responses, even on a spatial scale relevant for a single insect. Perhaps the compounds that are most important for this interaction are limited to those that do not vary after elicitation, although this seems unlikely given the importance and variability of trans-α-bergamotene (Figs 1, 2, Tables S1, S6) and of other volatiles, including linalool, which are emitted by only a few genotypes of N. attenuata (Halitschke et al., 2000; Kessler & Baldwin, 2001; I. T. Baldwin, pers. comm.). Even if this is the case, the question remains: why should total VOC emission profiles vary so dramatically when plants are elicited? Is this variation a result of competing evolutionary pressures to be more or less apparent, or to employ a more effective indirect defense than a neighboring plant – which in populations of N. attenuata is unlikely to be a sibling plant? Is it just a by-product of variation in biosynthetic pathways?
Results of studies like this are a prerequisite to understanding how insects learn to respond to varying volatile signals in a polymorphic plant population, and how these responses may affect plant evolutionary fitness.
We thank Danny Kessler for collecting seeds from the field, Brigham Young University for the use of their field station, the Lytle Ranch Preserve, Dr Stephan von Reuss for providing standards of cis- and trans-α-bergamotene, Dr A. F. Barrero for providing α-duprezianene, Firmenich (Switzerland) for providing a standard mixture of cis- and trans-β-ocimene, Emily Wheeler for editorial assistance, and the Max Planck Society for financial support.