A comparison of two Nicotiana attenuata accessions reveals large differences in signaling induced by oral secretions of the specialist herbivore Manduca sexta

Genetic variation within and among populations provides the raw material for evolution. Although many studies describe inter- and intraspecific variation of defensive metabolites, little is known about variation among plant populations within early signaling responses elicited by herbivory, or by herbivore oral secretions (OS) introduced into wounds during feeding. In this study, we compare the OS-elicited early responses as well as the anti-herbivore defensive metabolites in two accessions of the wild tobacco Nicotiana attenuata and show that, compared with an accession collected from Utah (UT, USA), an Arizona (AZ, USA) accession has lower herbivore-elicited activity of the salicylic acid-induced protein kinase (SIPK), an important MAP kinase involved in herbivore resistance. These differences in SIPK activity were associated with substantially different levels of OS-elicited jasmonic acid (JA), JA-isoleucine conjugate (JA-Ile), and ethylene (ET) bursts. Gene expression level polymorphism (ELP) determines phenotypic variation among populations, and we found the two accessions to have significantly different ELPs in the genes involved in early signaling responses to herbivory. In addition, we found differences between the UT and the AZ accessions in the concentrations of several secondary metabolites that contribute to N. attenuata ’s direct and indirect defenses. This study demonstrates significant natural variation in regulatory elements that mediate plant responses to herbivore attack, highlighting the role of ELP in producing a diversity of plant defense phenotypes. early responses to characterize how these genes’ ELP may account for this phenotypic variation. We show that, in addition to the established difference in expression of the TPI gene, many other genes involved in anti-herbivore defense display ELP in these two accessions. Compared with UT, AZ shows lower levels of herbivory-induced salicylic acid-induced protein kinase (SIPK) activity, an important MAPK related to herbivore defense in plants; furthermore, transcripts of SIPK and wound-induced protein kinase ( WIPK ), as well as two other MAPK genes, accumulate at different levels in AZ after wounding or herbivory. Compared with UT, transcripts of genes involved in phytohormone biosynthesis, as well as two transcription factors, have distinct levels of abundance in AZ; more importantly, the herbivore-www Results of the in-gel kinase assay indicate that both AZ and UT plants can perceive FACs from M. sexta OS and, in response, activate SIPK and WIPK. However, a negative regulation system in AZ plants deactivates SIPK more quickly than it does in UT plants. nicotine (A), chlorogenic , rutin were analyzed using HPLC. The emission of cis- α -bergamotene was measured in from treated leaves (+2 position) at 24 to 32 h post-treatment with either W+W or W+OS, with no treatment as control (D) or 150 µg MeJA in lanolin paste, with lanolin as control (E). Absolute emission is not comparable between D and E as these were two separate experiments. *, **,


Introduction
How organisms adapt to their ever-changing environment remains a fundamental and challenging question in biology. Within a species, heritable phenotypic variation among populations, or even individuals within populations, reflects genetic differences -a central requirement for evolutionary change. In recent years, with the completion of several genome sequences, large-scale analyses have revealed substantial intraspecific genetic variation in humans (Morley et al., 2004), yeast (Brem et al., 2002), Drosophila (Nuzhdin et al., 2004), Caenorhabditis elegans (Denver et al., 2004), and Arabidopsis thaliana (Koornneef et al., 2004;Schmid et al., 2003;Nordborg et al., 2005;Schmid et al., 2005;Clark et al., 2007), the model organism of choice for plant genetics.
Variations in the transcription of particular genes, called expression level polymorphism (ELP), is a hallmark (and perhaps often the cause) of diversified phenotypic traits (Doerge, 2002). ELPs in response to exogenously applied salicylic acid and jasmonic acid (JA) have been shown for many genes among different Arabidopsis accessions (Kliebenstein et al., 2006;Traw et al., 2003;van Leeuwen et al., 2007). In different populations, ELPs in particular genes have been shown to produce variation in pathogen resistance (Grant et al., 1995;Gassmann et al., 1999), flowering time (Johanson et al., 2000;Lempe et al., 2005;Werner et al., 2005), herbivore resistance, and secondary metabolite production (Kliebenstein et al., 2001; Transcription factors play critical roles in regulating gene transcript levels. WRKYs, which are plant-specific transcription factors, regulate various aspects of plant development and stress responses (Eulgem et al., 2000;Ulker and Somssich, 2004). It is known that transcription factors are the main targets of MAPKs (Hazzalin and Mahadevan, 2002;Yang et al., 2003). To investigate whether there is any ELP of WRKY genes between OS-elicited UT and AZ plants, which might result from the ELP of MAPKs, we measured the transcript abundance of three WRKY genes: WRKY3, WRKY6, and SubD48. WRKY6 has been shown to be an important signaling transduction component contributing to N. attenuata's responses to herbivory (Qu et al., unpublished data) and is located downstream of SIPK and WIPK (Wu et al., 2007a). SubD48 is also implicated in herbivore resistance, as W+OS treatment induced higher levels of SubD48 transcript than did W+W (Wu et al., 2007a).
When not treated, both accessions had similar levels of WRKY6 transcript (P = 0.62, unpaired t-test). Transcript abundance of WRKY6 reached maximum levels after only 0.5 h, increasing 450 and 3250 fold after W+W and W+OS treatment, respectively, in UT (Fig. 4A). AZ had remarkably lower WRKY6 transcript levels after both treatments, although both accessions showed similar induction patterns, in that W+OS induced much higher levels of WRKY6 than did W+W. Although WRKY3 did not show greater induction by W+OS, W+W-induced WRKY3 transcript levels decreased faster than those induced by W+OS after reaching maximum levels in both accessions at 0.5 h. No statistical difference in WRKY3 transcript levels was detected between UT and AZ before or after either induction (Fig. 4B). Transcript levels of SubD48 were the same in UT and AZ when plants were not induced, and W+OS induced higher levels of SubD48 transcript than did W+W in both accessions (Fig. 4C). However, SubD48 levels were higher in W+OS-treated AZ than in UT plants.
These data indicate that UT and AZ plants have different expression levels of transcription factors that are involved in herbivore resistance.
The ELP of these trans-regulatory elements may also contribute to further variation in downstream gene transcript levels.
AZ has lower levels of herbivore-induced JA, JA-Ile and ET than does UT JA, JA-Ile, and ET play critical roles in mediating plants' defense against herbivores (Reymond and Farmer, 1998;Halitschke and Baldwin, 2003;Staswick and Tiryaki, 2004;Kang et al., 2006;von Dahl et al., 2007). Evidence that SIPK and WIPK play central roles in mediating plants' JA and ET biosynthesis is increasing (Seo et al., 1995;Liu and Zhang, 2004;Wu et al., 2007a). To investigate whether the different levels of SIPK activity and ELP in many early signaling genes led to variation of phytohormone production in these two N. attenuata accessions, we measured both W+W-and W+OS-induced JA, JA-Ile, and ET levels.
Without treatment, both accessions had similar low levels of JA (~6 ng/g fresh weight (FW), P = 0.28, unpaired t-test). In UT, W+OS treatment induced higher levels of JA than did W+W treatment (Fig. 5A, top), indicating that UT plants recognize herbivory and deploy herbivorespecific defenses. This "JA burst" was much smaller in AZ: W+OS induced only 40% of the JA levels in W+OS-treated UT (Fig. 5A, top). To examine whether the ELP of genes involved in JA biosynthesis contributed to this phenotype, we measured the transcript levels of lipoxygenase 3 (LOX3), which encodes an important enzyme in JA biosynthesis (Fig. 5A, bottom) (Halitschke and Baldwin, 2003;Wasternack, 2007). When not induced, UT and AZ plants had the same levels of LOX3 transcript (P = 0.55, unpaired t-test). Both W+W and W+OS greatly induced LOX3, which declined faster in AZ: at 3 h, W+W-and W+OS-treated AZ had as little as 7% and 0.2%, respectively, of the LOX3 levels found in UT.
Similar results were obtained for measurements of JA-Ile levels. W+OS treatment induced higher levels of JA-Ile than did W+W treatment in both accessions (Fig. 5B, top). 1 h after either treatment, UT showed about 70% higher levels of JA-Ile than did AZ. JAR4 and JAR6 have been identified as the enzymes that conjugate JA and Ile to form JA-Ile (Wang et al., 2007); we measured the transcript levels of both genes (Fig. 5B, middle and bottom). AZ had basal transcript levels of both JAR4 and JAR6 that were almost 10-fold lower than those in UT (P = 0.038 and 0.005, respectively, unpaired t-test). However, 1 to1.5 h after induction, JAR4 and JAR6 transcript levels were similar in both accessions. The difference between JA-Ile levels in UT and AZ may have resulted from the different levels of both JA a substrate for the conjugation reaction and JAR activity, since the basal levels of JAR protein might also be considerably lower in AZ.
W+W treatment did not induce a detectable amount of ET in UT or AZ (data not shown). After W+OS treatment, ~25 nL ET was produced for every gram of UT leaf tissue (Fig. 5C, top); AZ produced only one-third as much. We measured the transcript profiles of two genes involved in ET biosynthesis, ACC oxidase 1 (ACO1) and ACC synthase 3a (ACS3a) in both accessions (Wang et al., 2002;von Dahl et al., 2007). Unexpectedly, levels of ACO1 transcript in untreated AZ plants were about 6-fold higher than in untreated UT plants (Fig. 5C, middle). In UT, W+OS-induced very similar ACS3a transcript levels (P = 0.48, unpaired t-test), and W+W treatment induced similar ACS3a levels in both accessions (P > 0.064, unpaired t-test); however, after W+OS treatment, ACS3a levels were several-fold higher in AZ than in UT (Fig. 5C, bottom).
These data show that UT and AZ have distinct herbivore-induced JA, JA-Ile, and ET responses. The ELP of phytohormone biosynthetic genes may contribute to the phenotypic variation between these accessions, but the differential regulation of ET production between the two accessions is clearly more complex.

UT and AZ produce different levels of anti-herbivory secondary metabolites
Herbivore-induced kinase activation, and changes in phytohormones and the transcriptome, alter the accumulation of secondary metabolites to raise plant defenses against herbivore attackers. In N. attenuata, these herbivore-induced defenses include nicotine and TPI (Steppuhn et al., 2004;Zavala et al., 2004b). Because AZ plants have lost their TPI defense mechanism (Wu et al., 2007b), we analyzed the levels of nicotine as well as two phenolic compounds, chlorogenic acid and rutin, which are putative anti-herbivore defenses (Isman and Duffey, 1982;Bi et al., 1997), and may constitute part of N. attenuata's direct defense.
Four days after plants were treated with W+W or W+OS, when levels of nicotine were at their highest (Baldwin et al., 1998), samples were harvested and analyzed by HPLC (Fig. 6A). Without treatment, AZ had ~50% more nicotine than did UT. Nicotine concentrations increased in both UT and AZ after treatment; nevertheless, AZ still contained 35% and 20% more nicotine compared to UT plants identically treated with W+OS and W+W, respectively.
Compared with UT, more chlorogenic acid and rutin were detected in AZ (Fig. 6B, C). Under normal conditions, AZ contained 1.3-fold more chlorogenic acid than did UT. After W+W and W+OS treatments, AZ still contained higher levels of chlorogenic acid. Rutin levels were not changed after either W+W or W+OS treatment; similarly to chlorogenic acid, concentrations of rutin in AZ were about 50% higher than in UT.
In addition to direct defenses, N. attenuata deploys indirect defenses to attract predators of herbivores. One of the most important indirect defense compounds is cis-α-bergamotene (CAB), an herbivory-inducible VOC (Kessler and Baldwin, 2001). We found that W+OS treatment enhanced CAB emission in UT ~7 fold; however, neither W+W nor W+OS induced higher levels of CAB emission in AZ (Fig. 6D). Since CAB biosynthesis is regulated by JA (Halitschke et al., 2000), we examined whether the decreased JA levels in AZ after W+OS treatment were associated with diminished CAB levels. We applied 150 µ g of methyl jasmonate (MeJA) dissolved in lanolin paste or pure lanolin paste (lanolin) as a control to both accessions and measured CAB emission. In both accessions, CAB emission was highly induced, but MeJA treatment did not recover the difference in CAB production between AZ and UT (Fig. 6E). This suggests that during evolution, AZ may have lost or reconfigured cis-or transelements involved in the regulation of CAB biosynthesis.

Discussion
In this study, we examined the early herbivore-induced responses in two N. attenuata accessions collected from Utah and Arizona, USA. We measured 1) the activity of two MAPKs, SIPK and WIPK; 2) the levels of the herbivore-induced phytohormones JA, JA-Ile, and ET; 3) differential transcript accumulation of several genes encoding MAPKs, transcription factors, and phytohormone biosynthesis enzymes, which we took to be indicative of ELPs in these genes, and furthermore, 4) the associated variation in concentrations of direct and indirect defense-related secondary metabolites.

Natural variation in herbivore-induced signaling events
MAPK signaling has been shown to play a critical role in mediating plant resistance to herbivores (Kandoth et al., 2007;Wu et al., 2007a).
FACs from M. sexta OS are introduced into plant cells through wounds generated by herbivore feeding. These FACs bind to putative receptors in N. attenuata, enabling plants to perceive herbivory. This perception results in the activation of MAPKs, most importantly SIPK and WIPK (Wu et al., 2007a). We examined kinase activity and kinase gene transcript levels in both UT and AZ. Activity of SIPK and WIPK reached the same maximum levels in both accessions, indicating that both are equally able to perceive herbivory and initiate herbivory-induced signal cascades. Nevertheless, the speed with which SIPK activity levels decrease in AZ following simulated herbivory (W+OS) suggests that the negative regulation of SIPK is stronger in AZ than in UT. A recent study in Arabidopsis has shown that a phosphatase, AP2C1, might be responsible for this negative regulation (Schweighofer et al., 2007).
Both SIPK and WIPK modulate herbivory-induced JA biosynthesis (Wu et al., 2007a). AZ plants accumulate lower levels of JA after simulated herbivory, consistent with rapidly decreasing levels of herbivore-induced SIPK activity. Compared with JA biosynthesis, ET biosynthesis is simpler: only 2 enzymes, ACO and ACS, are responsible for converting S-adenosylmethionine to ET (Wang et al., 2002). SIPK directly phosphorylates certain (but not all) ACSs, dramatically increasing their stability (Liu and Zhang, 2004). Longer-lasting herbivore-induced SIPK activity levels may partly account for the greater amount of ET produced in UT plants than in AZ plants, and could also explain why AZ produces less ET despite accumulating significantly higher levels of ACO and ACS transcripts than UT. However, the large differences in herbivore-induced JA and ET levels between UT and AZ plants are unlikely to have resulted solely from differences in SIPK activity. Variation of many genes involved in JA and ET biosynthesis probably also contributed to differences in the production of these two phytohormones in UT and AZ plants.
In addition to SIPK and WIPK, we found ELP in two other MAPK genes: Naf4 and MPK4. Naf4 shows high sequence homology to SIPK and thus may have a similar function (Ren et al., 2006). MPK4 is known to be involved in pathogen resistance and wound-induced responses, but has as yet no known function in herbivore resistance (Petersen et al., 2000;Gomi et al., 2005). Since kinases are upstream regulatory elements that influence the transcription of a wide rage of genes, variation in their constitutive and/or induced levels is likely to produce large sets of ELPs.
Most MAPK targets are transcription factors (Hazzalin and Mahadevan, 2002). Two WRKY transcription factors, WRKY6 and SubD48, displayed ELP between UT and AZ; however, there was no ELP in a third WRKY (WRKY3). Transcript levels of WRKY6 have been shown to be positively regulated by both SIPK and WIPK (Wu et al., 2007a). Thus, we can assume that the different herbivory-induced SIPK activity in UT and AZ may have resulted in the observed ELP of WRKY6. MAPKs, JA, and ET are known to regulate expression of a wide array of genes. Therefore, many other transcription factors downstream of herbivory-induced MAPKs, JA and ET, are likely to also show ELP between UT and AZ.
The large and sometimes unpredictable variation between these two N. attenuata accessions in their herbivore-induced early responses is indicative not only the complexity of the signaling networks, but also of the cross-talk and feedback regulation that characterize the relationships among their components, which include receptors, kinases, phytohormones, and transcription factors, with each group of molecules potentially participating in multiple layers of regulation. Studying the molecular mechanisms that underlie the phenotypic changes in these early signaling events will eventually provide us with a detailed understanding of how plants fine-tune their signaling systems, as well as how evolutionary forces have shaped genetic variation among natural populations.

Natural variation of secondary metabolites
Plants' reconfiguration of their transcriptomes and proteomes following herbivore attack results in the deployment of the last layer of the defense system, which comprises various anti-herbivore secondary metabolites (Baldwin, 2001). We found that AZ produces greater amounts of both nicotine and phenolic compounds (chlorogenic acid and rutin) than does UT. Nicotine is a potent chemical defense that has a strong negative effect on M. sexta and on many other herbivores and which has been previously shown to be a critical component of JA-mediated direct defense in AZ as well as in UT (Steppuhn et al., 2004). Nicotine biosynthesis is both upregulated by JA and downregulated by OS-induced ET (Baldwin et al., 1994;von Dahl et al., 2007;Voelckel et al., 2001). Thus, higher nicotine levels in AZ could indicate that decreased negative regulation by ET plays a greater role than decreased positive regulation by JA in nicotine accumulation after herbivory. Alternately, they could reflect a greater availability of nitrogen for nicotine biosynthesis due to the lack of nitrogen-intensive TPI production in AZ. However, given that AZ produces constitutively higher amounts of nicotine, whereas ET and JA can only influence induced nicotine production; and given that such high constitutive nicotine production was not seen in a line of the UT accession which was silenced for TPI (Zavala et al., 2004a), we assume that changes in certain trans-or cis-element(s) in the regulation of nicotine biosynthesis have enabled AZ to produce more nicotine than UT.
The low level of CAB production in AZ illustrates the complex changes in its regulatory system: that AZ is able to produce CAB, but releases less than 10% as much as UT after OS-elicitation indicates that the biosynthesis machinery is intact, but the regulatory system is constrained. That the decrease in production cannot be recovered with MeJA application shows that the bottleneck in regulation is downstream of the JA. Little is known about how CAB and other VOCs in N. attenuata, which may also act as indirect defense compounds, are synthesized and regulated. The synthesis of phenolic compounds in N. attenuata and their possible influence on M. sexta performance also remains to be characterized; some studies have revealed potential anti-herbivore functions of phenolic compounds (Isman and Duffey, 1982;Bi et al., 1997).
Further biochemical and genetic studies are needed to elucidate the ELPs in biosynthetic genes which may underlie the differential regulation of these compounds.

Ecological and evolutionary significance
Understanding the mechanisms responsible for the phenotypic variation in herbivore defense systems between these two N. attenuata accessions provides insights into their evolution. The TPI gene has been identified as a null allele in AZ (Zavala et al., 2004b;Wu et al., 2007b). In this study, we show that despite (or perhaps due to?) the loss of its TPI defense mechanism, AZ produces greater amounts of nicotine and phenolics.
The higher concentrations of these compounds in AZ may be a result of natural selection following the event of TPI mutation, enabling AZ to compensate for the loss of TPI. How the TPI null allele was fixed in AZ is unclear, since this mutation could have considerably lowered the ability of AZ to survive under herbivory pressure. Using TPI-silenced transgenic plants, Zavala et al. (2004a) demonstrated that TPI production is costly under normal conditions, because plants having lower TPI levels grow faster and have higher fitness due to the re-allocation of nitrogen; however, under selection pressure from herbivory, TPI-containing plants realize greater fitness due to the effectiveness of this defense (Zavala et al., 2004a, b).
One hypothesis is that after the TPI mutation event in AZ, there was very low selection pressure from herbivores; therefore, this null TPI allele was quickly fixed, since it provided plants with higher fitness. An alternate scenario which could lead to the fixation of the null TPI allele would be genetic drift and small original population size. In either case, over time, (re-imposed) selection pressure from herbivory could have led to higher concentrations of other anti-herbivore compounds, i.e., nicotine and phenolics. In nature, indirect defense compounds (e.g. CAB) are also highly effective in protecting plants from herbivory (Kessler and Baldwin, 2001). It is unclear why AZ has lost most of its CAB production. Steppuhn et al.
(in review) have shown that AZ plants emit much higher levels of some monoterpenes than UT plants, both constitutively and inducibly. Whether in its natural habitat, AZ depends more on direct defense rather than indirect defense against herbivory, or whether the predators in this habitat respond to other volatile signals remains to be studied. native environment via as-yet unidentified defense or tolerance mechanisms. Coevolution has long been considered to have shaped the diversity of both plants and herbivores (Ehrlich and Raven, 1964;Mauricio, 2001;Zangerl and Berenbaum, 2005). Further ecological studies could provide the necessary data to connect diversity in herbivore-induced plant responses to differences in native herbivore population structures and their dynamics, and to determine whether different selection forces imposed by diverse insect populations could have caused polymorphisms in defense compound production between native plant populations.
Recent studies of Arabidopsis populations have revealed considerable variation on both the genomic and the transcriptomic level (Kliebenstein et al., 2006;Clark et al., 2007;Kim et al., 2007). It is very unlikely that only one or a few genetic changes differentiate AZ and UT.
We predict that thousands of genes show polymorphism in sequence and/or expression between these two accessions. This unknown pool of polymorphisms greatly restricts our ability to explain changes in many phenotypic traits between the UT and AZ accessions, especially since most of these traits are polygenic. Nevertheless, ever-increasing theoretical and practical advances have provided us with the ability to unravel the complex genetic basis that is responsible for the variations we have detected (Kliebenstein et al., 2001;Kliebenstein et al., 2002;Wentzell et al., 2007). Using biochemical, genetic, and population genetic tools, we expect to see variations in many loci that contribute to the phenotypic variation between UT and AZ; it will not be surprising if these loci turn out to be mostly regulatory elements: transcription factor binding sites in promoters (cis-elements), genes coding for transcription factors (trans-elements), or even genes involved in modifying transcription factors, such as their binding activity, stability, and localization.

Plant material and experimental conditions
Seed collections of Nicotiana attenuata Utah (UT) and Arizona (AZ) accessions are described in Glawe et al. (2003). Nearly homozygous UT and AZ accessions were obtained by inbreeding for 17 and 7 generations in the glasshouse, respectively. In each generation, seeds were collected from one randomly selected plant for each accession. No artificial selection was conducted.

Quantitative real time-PCR (qRT-PCR) analyses
Total RNA was extracted from ~100 mg leaf tissue using the Trizol reagent (Invitrogen, http://www.invitrogen.com). The qRT-PCR procedure is described in detail in Wu et al. (2007a). All RNA samples were reverse transcribed using oligo(dT) and the Superscript II enzyme (Invitrogen). qRT-PCR was performed on an ABI PRISM 7700 sequence detection system (Applied Biosystems, http://www.appliedbiosystems.com), following conditions recommended by the manufacturer. In all analyses, a N. attenuata actin gene was used to normalize the concentrations of cDNA samples. The primers and probes used for the Taqman-based analysis, as well as the primers used for SYBR Green-based analyses are listed in Tables S1 & S2, respectively. The headspace was flushed into a photoacoustic laser spectrometer with hydrocarbon-free clean air, and the ET concentration was quantified by comparing ET peak areas with peak areas generated by a standard ET gas.
quality plastic containers secured with miniature claw-style hair clips. Ambient air flowed into the cage primarily through a trimmed P1000 pipette tip inserted into the bottom container, and was pulled out through a self-packed glass tube containing glass wool and 20 mg of SuperQ (Alltech, http://www.discoverysciences.com), secured in a second trimmed P1000 pipette tip inserted into the top container. Airflow was powered by a manifold vacuum pump as described in Halitschke et al. (2000). Background volatiles present in ambient air were also collected using empty trapping containers; CAB was not present in quantifiable levels in background controls. Trapping was stopped after 8 h and traps were either immediately eluted by spiking each trap with 400 ng tetraline as an internal standard (IS) and pushing through 250 µ L dichloromethane into a GC vial, or stored at -20°C until elution. Flow rates through each trap were measured and included as correlates in statistical analyses.
Eluents were separated on a DB-5 column (Agilent, http://www.agilent.com) in a Varian CP-3800 GC coupled with a Varian Saturn 4000 ion trap MS in EI mode (Varian, http://www.varianinc.com). Relative emission of CAB was expressed in ng tetraline/g FW/h. Data were logtransformed prior to analysis to meet requirements of normality and homogeneity of variance; 1 was added to all values prior to transformation to accommodate 0 values.

Statistical analysis
Statistics were done using the StatView software (SAS Institute, http://www.sas.com).