Insect herbivores have developed a myriad of strategies to manipulate the defense responses of their host plants. Here we provide evidence that chewing insects differentially alter the oxylipin profiles produced by the two main and competing branches of the plant defensive response pathway, the allene oxide synthase (AOS) and hydroperoxide lyase (HPL) branches, which are responsible for wound-inducible production of jasmonates (JAs), and green leafy volatiles (GLVs) respectively. Specifically, we used three Arabidopsis genotypes that were damaged by mechanical wounding or by insects of various feeding guilds (piercing aphids, generalist chewing caterpillars and specialist chewing caterpillars). We established that emission of GLVs is stimulated by wounding incurred mechanically or by aphids, but release of these volatiles is constitutively impaired by both generalist and specialist chewing insects. Simultaneously, however, these chewing herbivores stimulated JA production, demonstrating targeted insect suppression of the HPL branch of the oxylipin pathway. Use of lines engineered to express HPL constitutively, in conjunction with quantitative RT–PCR-based expression analyses, established a combination of transcriptional and post-transcriptional reprogramming of the HPL pathway genes as the mechanistic basis of insect-mediated suppression of the corresponding metabolites. Feeding studies suggested a potential evolutionary advantage of suppressing GLV production, as caterpillars preferably consumed leaf tissue from plants that had not been primed by these volatile cues.
Throughout their life cycles, plants are attacked by a wide range of organisms, including insect herbivores, with diverse feeding behaviors. Research over the past 40 years has clearly demonstrated a wide range of plant defenses against herbivore attack, but less attention has been given to herbivore adaptations used to overcome these defenses (Karban and Agrawal, 2002; Allmann and Baldwin, 2010). Some of the most potent plant defenses against herbivores are inducible, herbivore-specific and indirect (Baldwin et al., 2006; Heil et al., 2012). For example, insect feeding inevitably inflicts mechanical wounding on plants, with simultaneous release of oral secretions (OS), which are potential carriers of herbivore-specific elicitors into damaged tissues. These elicitors trigger herbivore-specific defense responses in their host plants through large-scale transcriptional, translational and post-translational alteration (Halitschke et al., 2001; Howe and Jander, 2008; Hilker and Meiners, 2010). These large-scale alterations, which are mostly known to induce plant anti-herbivore defense responses, function either directly via mechanisms such as production of amino acid-catabolizing enzymes, anti-digestive proteins or toxic or repelling chemicals (Weber et al., 1999; Chen et al., 2005), or indirectly through production and release of volatile organic compounds (VOC) that attract natural enemies of the herbivores or coordinate other plant defenses (Pare and Tumlinson, 1997, 1999; Kessler and Baldwin, 2001; Engelberth et al., 2004; Kessler et al., 2004; van Poecke and Dicke, 2004; Chehab et al., 2008).
Many inducible defense responses are activated by oxylipins, the oxygenated derivatives of fatty acids (Creelman and Mullet, 1997; Reymond et al., 2000, 2004; Blee, 2002). Allene oxide synthase (AOS) and hydroperoxide lyase (HPL) are the two main competing oxylipin pathway branches producing stress-inducible compounds (Feussner and Wasternack, 2002). The metabolites of the AOS branch are jasmonates [jasmonic acid (JA), methyl jasmonate (MeJA) and their biosynthetic precursor 12–oxophytodienoic acid (12–OPDA)], and the best-characterized metabolites of the HPL branch are the green leafy volatiles (GLVs), which predominantly consist of C6 aldehydes and their corresponding derivatives (Matsui, 2006; Chehab et al., 2008).
Herbivore OS may suppress wound-induced defensive responses of the host plants, analogous to how the saliva from blood-feeding arthropods suppress the defenses of animal hosts (Schittko et al., 2001; Musser, 2005). For example, aphid salivary peroxidases and catechol oxidases inactivate toxic phytochemicals found in host plants (Ribeiro, 1995), and their salivary Ca2+ -binding proteins sabotage the plant's attempts to plug sieve plates in the phloem (Will et al., 2007). Caterpillar OS also suppress plant-induced responses, as exemplified by glucose oxidase in Helicoverpa zea saliva, which reduces nicotine production in Nicotinana tobacum by inhibiting wound signaling (Musser et al., 2002). Thus, these and other reports have established that the biochemical composition of saliva from each insect species has a distinct signature molecule(s) that results in differential host responses (Schittko et al., 2001; Halitschke and Baldwin, 2003). At times, the alteration of plant metabolic profiles by herbivore saliva ultimately benefits the plant, but at other times it benefits the herbivore (Kim et al., 2011).
Despite progress in understanding the biochemistry and genetics of plant–insect interactions, our knowledge of the underlying mechanisms is incomplete. Here we have performed metabolic analyses of oxylipins and expression profiling of the corresponding pathway genes in mechanically wounded and insect-damaged Arabidopsis genotypes at various intervals after feeding initiation. These data indicate that both specialist and generalist chewing insects selectively suppress GLV production via transcriptional and post-transcriptional reprogramming of the HPL pathway genes, a strategy that counteracts the plant's ability to produce defensive metabolites and reduce nutritional quality for these herbivores. We suggest that this represents an herbivore strategy to manipulate intra- and inter-plant signaling in order to counteract plant inducible defenses.
Chewing insects suppress emission of wound-inducible HPL-derived volatiles
We previously characterized GLVs emitted from mechanically wounded or aphid-infested Arabidopsis plants. In these studies, we used the Col–0 accession, with a natural loss-of-function mutation in the hpl gene (Duan et al., 2005; Chehab et al., 2008) as the control, and engineered Col–0 lines that over-express a rice HPL gene under the control of the CaMV 35S promoter (OsHPL3 OE). We identified cis–3–hexenyl acetate as the predominant volatile released from mechanically wounded and aphid-damaged over-expressing plants (OsHPL3 OE), but not from damaged Col–0 (Chehab et al., 2008). In the present study, we also used Ws, an Arabidopsis ecotype with a functional HPL gene expressed from the native promoter, and further confirmed cis–3–hexenyl acetate and its precursor cis–3–hexenol as the predominant volatiles released from mechanically wounded and aphid-infested plants (Figure 1a,b and Figure S1).
Next we examined the levels and profiles of GLVs emitted by unwounded, mechanically wounded (by forceps) and caterpillar-damaged Ws plants (Figure 1c–e). Specifically, we compared Ws plants damaged at various time points post feeding initiation by both a specialist caterpillar (Pieres rapae) and a generalist caterpillar (Spodoptera exigua). These data show that while mechanical wounding induced elevated emissions of cis–3–hexenyl acetate and cis–3–hexenol from Ws, release of these GLVs in plants damaged by both the generalist and the specialist insects remained at the basal unwounded levels at all intervals examined after feeding initiation (Figure 1d,e and Figure S2).
Next we questioned whether the wound-induced production of HPL-derived metabolites (cis–3–hexenol) observed in plants damaged by forceps differed from that in plants wounded by removal of tissue using scissors (mimicking the caterpillar feeding pattern). As a control, we also analyzed the levels of JA, another oxylipin metabolite that is produced by the pathway parallel to and competing with the HPL pathway (Figure 1a and Figure S3). These data clearly indicate that both methods of damage delivery lead to comparable induction in the levels of these oxylipins.
Herbivore oral secretions suppress emission of GLVs
To determine whether insect-induced suppression of GLV emission is caused by herbivore OS, we either applied caterpillar regurgitant, which contains some OS, or water or boiled regurgitant as a control, to mechanically wounded leaves. In agreement with a previous finding (Ellinger et al., 2010), treatment of plants with regurgitant from P. rapae or S. exigua produced almost identical results; hence subsequent experiments were performed with the regurgitant from P. rapae. Plants wounded and treated with regurgitant emitted approximately 50% less hexenol and approximately 70% less hexenyl acetate than wounded and water- or boiled regurgitant-treated controls (Figure 2a,b). Next, we determined whether herbivore feeding suppresses GLV emissions even in conjunction with subsequent mechanical damage. Specifically, we compared GLV emissions from mechanically wounded leaves that had been previously fed on by P. rapae, or have been mechanically wounded only, or had experienced no previous damage. Analyses of GLVs at 90 min after wounding established that plants first damaged by insects and subsequently mechanically wounded emitted approximately 60–70% less hexenol and hexenyl acetate than those damaged only mechanically (Figure 2a,b). Collectively, these results established that mechanical wounding after insect damage did not restore insect-inflicted suppression of GLV emission.
Chewing insects selectively and constitutively suppress GLV production
To determine whether the endogenous levels of HPL pathway products are differentially altered in mechanically wounded and insect-damaged plants, we analyzed the levels of the first HPL pathway product, cis–3–hexenal, and its derivative products, cis–3–hexenol and cis–3–hexenyl acetate in leaves. These analyses showed basal levels of cis–3–hexenal in mechanically damaged plants (Figure S4A), in agreement with previous reports using the No–0 accession of Arabidopsis (Matsui et al., 2000). However, in contrast to a previous study that showed enhanced levels of cis–3–hexenal in P. rapae-infested No–0 plants (Shiojiri et al., 2006), we did not detect statistically significant differences in the levels of this metabolite in undamaged versus insect-damaged plants. This discrepancy is potentially due to differences in that Arabidopsis accessions employed, as Ws contains considerably higher basal levels of this metabolite than No–0. This suggests that the high basal levels present in Ws plants may not have permitted detection of potentially small alterations caused by mechanical or insect-inflicted damage. Interestingly, however, the endogenous level of cis–3–hexenol, the downstream product of cis–3–hexenal, was elevated in mechanically wounded plants but remained at the basal control levels in insect-damaged plants (Figure S4B).
Next, we analyzed levels of cis–3–hexenol and JA at various intervals after P. rapae feeding initiation on Ws plants. These data clearly show similar levels of cis–3–hexenol, a product of the HPL branch of the oxylipin pathway, in control and insect-damaged plants examined at various intervals after insect feeding initiation (Figure 3a). In contrast, the levels of JA, produced by the competing AOS branch of the oxylipin pathway (Figure 1a), increased steadily with increasing damage inflicted by prolonged insect feeding (Figure 3b). This latter finding is in agreement with numerous earlier studies demonstrating enhanced expression levels of JA biosynthetic pathway genes and production of the corresponding metabolites after herbivore feeding initiation (Reymond et al., 2000, 2004; De Vos et al., 2005; Ralph et al., 2006; Howe and Jander, 2008). Collectively, these data clearly support a selective and constitutive suppression of HPL pathway-derived metabolites by chewing insects.
Insect-mediated transcriptional and post-transcriptional reprogramming of HPL pathway genes
Next, we explored the molecular basis of altered oxylipin levels. We first compared the relative expression levels of corresponding pathway genes in control, mechanically damaged and P. rapae-damaged plants 72 h after feeding initiation using quantitative RT-PCR (Figure 4a). We specifically focused on plants damaged 72 h after insect feeding initiation, as these plants accumulated the highest JA levels within the experimental time frame. Specifically, we assayed the expression of genes encoding two phospholipases [DONGLE (DGL) and phospholipase A (PLA)–Ιγ1 ], five lipoxygenases (LOX1–4 and LOX6), allene oxide synthase (AOS), 12–OPDA reductase 3 (OPR3), hydroperoxide lyase (HPL) and acetyl CoA:cis–3–hexenol acetyltransferase (CHAT). The phospholipase DGL is required for production of basal and wound-induced levels of JA, and PLA–Ιγ1 phospholipase activity contributes to formation of the JA precursor 12–OPDA within the first hour after wounding (Hyun et al., 2008; Ellinger et al., 2010), by releasing unesterified fatty acids. The free fatty acids are then oxygenated by LOXs and converted to their corresponding hydroperoxides, as substrates for the several competing oxylipin branch pathway enzymes. AOS is the major control point in the JA biosynthetic pathway, and OPR3 is the enzyme that converts OPDA to 3–oxo-pentenyl-cyclopentane-1–octanoic acid, which undergoes three rounds of β–oxidation in the peroxisomes to yield JA (Schaller et al., 2004). Examination of HPL pathway genes was necessarily restricted to the two that are known (Figure 1a), namely HPL, which encodes the enzyme catalyzing the cleavage of fatty acid hydroperoxides into aldehydes, and CHAT, which encodes the enzyme that catalyzes conversion of cis–3–hexenol to cis–3–hexenyl acetate (Matsui, 2006; D'Auria et al., 2007).
Insect feeding led to induction of measurable levels of LOX2, LOX3, LOX4, AOS and OPR3, but expression levels of the other genes remained at or below the basal levels (Figure 4a). In contrast, mechanical wounding induced expression of all genes examined except for LOX1, whose expression is known to be reduced in wounded plants (Reymond et al., 2000; De Vos et al., 2005), and LOX6 (Figure 4b).
The unexpected lack of induction of DGL and PLA–Ιγ1 expression in insect-damaged plants led us to further examine expression of these genes at 90 min after feeding initiation, as they are known to be induced only during early stages of JA production (Hyun et al., 2008; Ellinger et al., 2010). We therefore performed the quantitative RT-PCR on control leaves (unwounded), mechanically wounded leaves, and leaves damaged by 90 min of P. rapae feeding (Figure S5A). Wounding of plants was carefully performed using forceps in order to wound leaves at levels similar to damage found 90 min after initiation of insect feeding. The expression analysis indicated that, in contrast to wounding, insect feeding did not induce expression of DGL and PLA–Ιγ1 (Figure S5B), suggesting that other yet to be identified lipases are induced in response to insect attack.
The lack of induction of HPL and CHAT expression in response to insect feeding, but their up-regulation in response to mechanical damage, provides clear evidence for herbivore-mediated transcriptional reprogramming of the pathway genes as the molecular basis responsible for constitutive suppression of the respective metabolites.
To determine whether insect-inflicted suppression of DGL, PLA–Ιγ1, HPL and CHAT is caused by herbivore OS, we applied either caterpillar (P. rapae or S. exigua) regurgitant, or water as a control, to mechanically wounded leaves, and examined the transcript levels of these four genes (Figure 5). Application of P. rapae or S. exigua regurgitant resulted in very similar data, and therefore only the data from S. exigua regurgitant application are included here. These results clearly show that expression of all the genes except that of the HPL is suppressed in the presence of regurgitant. Among the three genes responding to regurgitant application, the CHAT expression level is most notably reduced as compared to the mock treatment (Figure 5). The lack of suppression of HPL expression in these plants as opposed to that obtained after insect feeding assays (Figure 4) may be a function of OS concentration in the exogenously applied regurgitant. These data are in agreement with recent findings demonstrating that insect OS from two unrelated insects, a specialist (Pieris brassicae) and a generalist (Spodoptera littoralis) suppress the expression of several wound-inducible genes in Arabidopsis, suggesting that this suppressive activity is a general property of insect OS (Consales et al., 2012).
Next we investigated the involvement of potential mechanisms in the suppression of GLV production, other than transcriptional regulation. We compared the emission of GLVs (specifically cis–3–hexenol and cis–3–hexenyl acetate) between Col–0 and OsHPL3 OE lines, which have been engineered to over-express the rice HPL3 gene under the control of the constitutive CaMV 35S promoter in the Col–0 background, in response to wounding and insect damage inflicted by both specialist (P. rapae) and generalist (S. exigua) caterpillars (Figure 6a,b and Figure S6). These results collectively indicate that mechanical wounding enhanced GLV emission, whereas both specialist and generalist caterpillars impaired production of these volatiles at all time points after feeding initiation. The insect-mediated suppression of GLV production in plants expressing HPL under the control of the constitutive promoter suggests additional regulatory mechanisms beyond HPL transcriptional control.
In summary, evidence from comparison of transgenic lines constitutively expressing HPL, in conjunction with quantitative RT-PCR-based expression analyses, suggest that both insect-mediated transcriptional changes as well as post-transcriptional reprogramming of HPL pathway genes are responsible for the suppressed production of the corresponding metabolites.
Chewing insects favor plants that are not primed by GLVs
GLVs are known to attract natural enemies of herbivores in many systems, including Arabidopsis (Baldwin et al., 2002; Chehab et al., 2008; Peng et al., 2011; Wei and Kang, 2011). More recently, GLVs have been shown to play a role in defensive coordination within and between plants (Arimura et al., 2001; Giron-Calva et al., 2012). Thus interruption of GLV signaling may be beneficial to herbivore fitness. In order to test whether GLV inter-plant signaling may have a negative impact on herbivores (and thus be a target for suppression), we performed no-choice feeding trials on plants that had or had not been exposed to GLVs from a neighboring damaged plant. Specifically, we arranged groups of plant such that two damaged or undamaged OsHPL3 OE plants (emitters), which produce large quantities of GLVs upon wounding (Figure 6a,b), were placed on either side of one Col–0 assay plant, without physically touching (Figure 7a). After 2 days, emitter plants were removed, and the assay plants were challenged with one neonate P. rapae caterpillar. Caterpillars fed more on plants that had not been primed by GLVs emitted by wounded plants (Figure 7b), suggesting that plant–plant communication triggered by these volatiles altered the insect feeding behavior.
Plant signals co-ordinate defenses both within and between plants, and these signals may be a target for herbivore adaptations to overcome plant defenses (Kim et al., 2011). Among the plant-produced signaling molecules are JA, an indispensable component of the defense pathway that promotes resistance to a wide spectrum of insects (Howe et al., 1996; McConn et al., 1997; Baldwin, 1998; Howe and Jander, 2008), and GLVs, which are known to be involved in intra- and inter-plant defense signaling cascades as well as tritrophic interactions (Bate and Rothstein, 1998; Pare et al., 1998; Stotz et al., 1999; Arimura et al., 2001; Engelberth et al., 2004; Chehab et al., 2008).
In this study, we found that feeding by both generalist and specialist caterpillars suppressed the production of various GLV metabolites but increased the production of JA (Figures 1 and 3). Aphids, on the other hand, elicited a GLV response that was similar to that elicited in response to mechanical wounding. In agreement with previous findings (Reymond et al., 2000; De Vos et al., 2005), we found that different herbivores elicited unique signaling responses from the plant, and that these responses often differed from those elicited in response to mechanical wounding alone. Specifically, caterpillars constitutively reduced the production of cis–3–hexenol and cis–3–hexenyl acetate, metabolites downstream of cis–3–hexenal, the first and most abundant product of the HPL pathway. At first glance, these results are in apparent disagreement with a previous report, which found significant increases in GLVs in response to P. rapae infestation (Shiojiri et al., 2006). However, their study only assayed levels of cis–3–hexenal and neither of the downstream products that show the greatest suppression in our study. Similar to our results, feeding by Colorado potato beetle suppressed cis–3–hexenol production in potatoes (Gosset et al., 2009), suggesting that GLV suppression may be a strategy employed by taxonomically diverse herbivores.
We further showed that insect-derived elicitors are responsible for the various plant responses to different herbivores or damage treatments, and that the suppression of GLVs by caterpillars lasts for a relatively long time. Addition of caterpillar regurgitant, which is thought to contain a small amount of salivary enzymes and elicitors (Peiffer and Felton, 2009), to mechanically damaged leaves reduced GLV emission (Figure 2). Caterpillar feeding continued to suppress GLV emission even when leaves experienced additional mechanical damage 3 h after caterpillar feeding (Figure 2), suggesting that the suppression of GLVs is long-lasting.
The change in signaling dynamics due to different types of herbivores or mechanical damage is probably caused by transcriptional modifications of critical genes in the HPL pathway. Specifically, we show that the transcriptional levels of the HPL pathway-related genes HPL and CHAT are not altered in caterpillar-damaged plants, but these same genes are up-regulated by mechanical damage (Figure 4). In contrast, general oxylipin pathway genes, including those of the JA branch pathway, such as LOX2, LOX3, LOX4, OPR3 and AOS, were up-regulated by both mechanical damage and by caterpillar feeding (Figure 4). This resulted in a burst of JA in caterpillar-infested or mechanically damaged plants, consistent with numerous studies of JA signaling in plants (De Vos et al., 2005; Howe and Jander, 2008). Using transformed Arabidopsis lines that over-express HPL, we also show that HPL-derived metabolites are suppressed in the absence of transcriptional regulation (Figure 6). This suggests the existence of additional post-transcriptional mechanisms or other downstream processes that inhibit emission of GLVs.
Interestingly, we found that GLV production may itself be used as a signal by plants to coordinate their defensive response to herbivores. We showed that plants that were exposed to GLVs from a damaged neighbor experienced less herbivore damage than plants without such exposure (Figure 7), although such treatment had no effect on caterpillar weight gain (P =0. 97, data not shown). We do not know why caterpillars consumed less foliage but were able to compensate in terms of weight gained over the short term. These results suggest another benefit of herbivore-induced GLV production for plants and another potential cost of GLVs to herbivores. Additionally, numerous previous studies have found strong evidence that GLV production attracts natural enemies of herbivores, thus serving as an even broader ecological signal. In each of these cases, herbivores may benefit from removal of the plant signal. We found strong evidence that some herbivores at least are able to diminish plant signaling. The co-evolution of defense and counter-defense in herbivore–plant systems is probably not limited to ‘defenses’ per se, but also includes the development and coercion of signaling processes.
Plant growth and treatment conditions
Arabidopsis wild-type accessions Wassilewskija (Ws) and Columbia–0 (Col–0), and transgenic Col–0 plants over-expressing the rice HPL3 gene under the control of the CaMV 35S promoter (OsHPL3 OE) were grown as previously described (Chehab et al., 2008).
To mechanically damage plants, leaves were wounded either using a forceps, or by tissue removal using scissors, to produce damage at levels similar to that inflicted by chewing insects 90 min after feeding initiation. The forceps was used to pinch each damaged leaf perpendicular to the main vein.
To establish caterpillar damage, we placed 4th-instar larvae on rosette leaves of 3-week-old plants for experiments lasting from 90 min to 3 h, and used 2nd-instar larvae for longer experiments. Volatiles from undamaged controls, mechanically wounded and caterpillar-damaged plants were collected for 3 h. Analyses of endogenous metabolite levels and expression profiling of their respective genes were performed on undamaged and damaged tissue collected at the indicated intervals after feeding initiation, and at 90 min after mechanical wounding. Collected tissues were immediately frozen in liquid nitrogen and stored at −80°C prior to analyses. Damage by M. persicae and metabolic analyses of aphid-infested and control plants were performed as previously described (Chehab et al., 2008).
Regurgitant preparation and application
Regurgitant was collected from 4th- to 5th-instar P. rapae fed on Arabidopsis, and was diluted 1:1 with water. Subsequently, 10 μl of water (control), boiled regurgitate or diluted regurgitant was applied to the wounded area.
Analyses of HPL-derived metabolites
Volatiles and endogenous HPL-derived metabolites were extracted and analyzed by GC–MS as described previously (Chehab et al., 2008), except that the analyses were performed in scan acquisition mode, and target compounds were quantified using extracted ion chromatograms: m/z 69 for cis–3–hexenal, m/z 67 for cis–3–hexenol and cis–3–hexenyl acetate, and m/z 59 for the internal standard 3–octanol.
Analysis of JA was performed by GC–MS as previously described (Savchenko et al., 2010).
Expression analyses were performed by quantitative RT-PCR, and At4g34270 and At4g26410 were used as internal controls for transcript normalization as previously described (Walley et al., 2008), using primers listed in Table S1.
Statistical analysis was performed as previously described (Chehab et al., 2008).
Feeding assays were performed on independent groups consisting of one Col–0 assay plant placed between two OsHPL3 OE plants (emitters). Plants were not allowed to physically touch one another: emitter plants were 3 cm away from the assay plants and groups of three plants were separated from other groups by more than 1 m to reduce volatile contamination. Two leaves of half of the GLV emitting plants neighboring the Col–0 plants were wounded using the forceps, and the other Col–0 assay plants were flanked by unwounded GLV emitters as controls. Two days after wounding, one neonate P. rapae caterpillar was placed on each assay plant (n =84), and the emitters were removed. Caterpillars fed for 4 days before leaves were removed, and the area consumed was measured. Each caterpillar was weighed at the beginning and the end of the feeding assay. Statistical analysis was performed using anova.
This project was supported by US National Science Foundation grant IOS-1036491 (K.D.).