Induced defenses are thought to be economical: growth and fitness-limiting resources are only invested into defenses when needed. To date, this putative growth–defense trade-off has not been quantified in a common currency at the level of individual compounds. Here, a quantification method for 15N–labeled proteins enabled a direct comparison of nitrogen (N) allocation to proteins, specifically, ribulose-1,5-bisposphate carboxylase/oxygenase (RuBisCO), as proxy for growth, with that to small N-containing defense metabolites (nicotine and phenolamides), as proxies for defense after herbivory. After repeated simulated herbivory, total N decreased in the shoots of wild-type (WT) Nicotiana attenuata plants, but not in two transgenic lines impaired in jasmonate defense signaling (irLOX3) and phenolamide biosynthesis (irMYB8). N was reallocated among different compounds within elicited rosette leaves: in the WT, a strong decrease in total soluble protein (TSP) and RuBisCO was accompanied by an increase in defense metabolites, irLOX3 showed a similar, albeit attenuated, pattern, whereas irMYB8 rosette leaves were the least responsive to elicitation, with overall higher levels of RuBisCO. Induced defenses were higher in the older compared with the younger rosette leaves, supporting the hypothesis that tissue developmental stage influences defense investments. We propose that MYB8, probably by regulating the production of phenolamides, indirectly mediates protein pool sizes after herbivory. Although the decrease in absolute N invested in TSP and RuBisCO elicited by simulated herbivory was much larger than the N-requirements of nicotine and phenolamide biosynthesis, 15N flux studies revealed that N for phenolamide synthesis originates from recently assimilated N, rather than from RuBisCO turnover.
Plants have evolved two general direct strategies against herbivory: constitutive and inducible defenses. The biosynthesis of these defenses requires fitness-limiting resources that could otherwise be invested into growth and reproduction. Hence, induced plant defenses are thought to be a cost-saving strategy compared with constitutive defenses, as they are only produced when needed, e.g. after herbivory (Karban and Baldwin, 1997), and this cost-saving model plays a central role in most theoretical treatments of induced defenses (for a review of plant defense hypotheses, see Stamp, 2003). Several studies have quantified the costs of induction by measuring photosynthesis rates, plant biomass, size and/or yield associated with an increase in defense metabolites (Bazzaz et al., 1987; Karban and Baldwin, 1997; Zangerl et al., 2002). Although measurements of the impact of anti-herbivore defenses on plant yield are important for understanding their ultimate fitness costs, measurements of plant biomass do not discriminate among the relative investments into compounds that function in growth, storage and defense processes in the tissues analyzed (Chapin et al., 1990). Therefore, the investment into growth is preferably estimated by measuring components of biomass that directly promote the acquisition of resources for growth, such as photosynthetic proteins (Chapin et al., 1990). Additionally, the costs of defense should be measured in the currency of a fitness-limiting resource (Mole, 1994; Baldwin et al., 1998). Nitrogen (N) is often such a fitness-limiting resource, determining the growth and reproduction of plants, and of the herbivores that eat them. N availability also influences N allocation to defense metabolites (Baldwin et al., 1998; Lou and Baldwin, 2004; Simon et al., 2010). Thus, it is an ideal currency to use for the study of growth–defense trade-offs in plant–herbivore interactions.
Ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCO) is the most abundant foliar protein in plants, and is essential for the dark reaction of photosynthesis. RuBisCO constitutes 30–50% of the total soluble protein (TSP) in C3 plants (Ellis, 1979; Makino et al., 1984; Imai et al.,2008), and may function as a potential N storage protein (Millard, 1988); consequently, it represents a major N sink in plants. Its large and small subunits (LSUs and SSUs, respectively) are synthesized from separate precursor pools that have different metabolic origins (Allen et al., 2012). Although the concentration and activity of RuBisCO are not the only factors controlling growth (Stitt and Schulze, 1994), changes in RuBisCO expression influence growth and lead to complex changes in N metabolism (Stitt and Schulze, 1994; Stitt and Krapp, 1999; Matt et al., 2002), making this enzyme a reasonable proxy for growth parameters.
Nicotiana attenuata is a wild tobacco native to the Great Basin Desert in south-western USA that synchronizes its germination from long-lived seed banks in response to exposure to cues from pyrolized vegetation (Preston and Baldwin, 1999). By timing its germination with the immediate post-fire environment, N. attenuata takes advantage of the abundant, yet ephemeral, pools of inorganic N in burned soil (Lynds and Baldwin, 1998), but is subject to high intraspecific competition for this fitness-limiting resource because of its mass-germination behavior. Furthermore, because it is a pioneer species, N. attenuata is attacked by a diverse herbivore community, including the specialist tobacco hornworm (Manduca sexta). Herbivore attack elicits the jasmonic acid (JA) signaling cascade (Kessler et al., 2004), which activates JA-responsive transcription factors that lead to the biosynthesis of a plethora of induced small metabolites (Figure 1a; Woldemariam et al., 2011), such as the N-intensive alkaloid nicotine, and a variety of phenolamides, which decrease herbivore performance (Baldwin, 1999; Steppuhn et al., 2004; Kaur et al., 2010; Onkokesung et al., 2010). The biosynthesis of nicotine and phenolamides requires the same amino acid precursors (ornithine and arginine for putrescine and spermidine biosynthesis; Kaur et al., 2010; Steppuhn et al., 2004; Takano et al., 2012), but nicotine is produced only in the roots (Hibi et al., 1994), whereas phenolamides are synthesized in the attacked leaf (Kaur et al., 2010).
Nicotine is present constitutively in undamaged N. attenuata tissues, and foliar concentrations increase substantially after herbivory (McCloud and Baldwin, 1997; Baldwin, 1999). The two major phenolamides found in N. attenuata are the N-acylated polyamines caffeoyl-putrescine (CP) and dicaffeoyl-spermidine (DCS), the biosynthesis of which is regulated by the transcription factor NaMYB8 (hereafter MYB8; Figure 1a). Both CP and DCS accumulate constitutively in reproductive tissues, and are strongly induced in leaves by simulated herbivory (Kaur et al., 2010). Herbivory also causes large-scale changes in N. attenuata's transcriptome and proteome, decreasing the levels of photosynthetic genes and proteins, including RuBisCO (Halitschke et al., 2003; Voelckel and Baldwin, 2004b; Giri et al., 2006). Because of the important constraints imposed by N availability upon both its growth and defense, as well as the wealth of understanding of its anti-herbivore defenses and the availability of isogenic transgenic lines impaired in individual classes of defenses, N. attenuata is an ideal model in which to study growth–defense trade-offs in a common N currency.
The induction of defense responses in wild tobacco can be simulated in a standardized and synchronized way by wounding leaves and applying the oral secretions (OS) of M. sexta larvae to the wounds (W + OS, Figure 1). The major elicitors in M. sexta OS are fatty-acid amino acid conjugates (FACs), which are recognized by the plant, triggering defense responses (Schittko et al., 2001; Halitschke et al., 2003; Giri et al., 2006). The FAC composition of OS, and the resulting gene expression and metabolite induction in the plant, differ between specialist and generalist folivores (Voelckel and Baldwin, 2004a; Diezel et al., 2009; Steinbrenner et al., 2011).
Here, we quantified the N investments into different plant parts and among different N pools within a tissue to compare the investments into growth and defense in the same currency after repeated simulated herbivory by W + OS elicitation from a specialist herbivore. Repeated simulated herbivory, in contrast to single W + OS elicitation, more closely mimics natural herbivore feeding, which varies in duration and timing (Van Dam et al., 2001; Skibbe et al., 2008; Stork et al.,2009). A stable isotope labeling technique was used to track N flux among different pools of individual compounds in locally elicited and systemic leaves and seeds. We applied 15N-labeled nitrate to the soil because nitrate is the most common form of N taken up by N. attenuata in nature, after the rapid biological nitrification of the ammonium generated by pyrolysis (Figure 1; Lynds and Baldwin, 1998).
The N flux into the three major N-intensive small metabolites of N. attenuata (nicotine, CP and DCS) was used as a proxy for defense investment that could be directly compared with the N investment into proteins, and in particular the abundant photosynthetic protein, RuBisCO, as a proxy for growth-related investment. These different molecule classes could not be measured in the past with comparable precision and accuracy because of a lack of suitable methods, especially for proteins. Here, we used a high-throughput LC-MSE method for the absolute quantitation of proteins and the incorporation of 15N into peptides (Ullmann-Zeunert et al., 2012), which allows for the quantification of single large proteins with the same accuracy as for the small defense metabolites quantified by UPLC/UV/ToF-MS.
To further disentangle the effects of induced defenses on N allocation after herbivory, we compared two previously described transgenic lines, one deficient in JA signaling, irLOX3 (Allmann et al., 2010), and one deficient in the biosynthesis of phenolamides, irMYB8 (Kaur et al., 2010), with wild-type (WT) plants (Figure 1a). This design allows for a direct comparison of N flux into specific classes of defense compounds with that into growth-related proteins measured in the same N currency, and an evaluation of the hypothesis that RuBisCO is used as an N-storage compound for defense responses.
Results and discussion
Anti-herbivore defense elicitation alters the nitrogen content of the shoot
Herbivory is known to change resource allocation within plants (Bazzaz et al., 1987; Frost and Hunter, 2008; Gomez et al., 2010). To estimate the impact of the biosynthesis of N-containing defense metabolites on N accumulation in N. attenuata, we compared the shoot N contents (% dry mass) of the two transgenic lines impaired in defense responses with WT plants after repeated simulated herbivory with W + OS. The isotope ratio mass spectrometry (IRMS) measurements revealed that repeated elicitation reduced the N content of WT shoots (i.e. the total of N per dry mass of shoots; Welch's two-sample t-test, d.f. = 7.24, P = 0.032), but not of the transgenic lines (Figure 2), whereas the N pool sizes were slightly reduced after elicitation for all three genotypes (Figure S1b). The changes in the N pool sizes of elicited irLOX3 and irMYB8 plants were the result of a reduction in shoot dry mass (Figure S1), whereas the elicited WT showed both reduced shoot dry mass and reduced shoot N content, suggesting a possible N reallocation within the plant caused by the biosynthesis of N-containing defense metabolites.
Plants can allocate N to roots to protect this resource from folivores to reduce the nutritional value of the attacked tissues, which, together with increased defenses, can slow herbivore growth and increase their exposure to natural enemies (Trumble et al., 1993). Previous studies with Solanum lycopersicum (tomato) demonstrated that N allocation in the form of amino acids from the shoot to the roots was rapidly induced by methyl-jasmonate (MeJA; Gomez et al., 2010) and M. sexta feeding (Steinbrenner et al., 2011; Gomez et al., 2012). In N. attenuata, OS elicitation has been shown to cause a rapid allocation of carbon from the shoot to the roots, which can later be used for regrowth and reflowering (Schwachtje et al., 2006). The reduced N concentration of WT shoots in our experiment suggests that this species can also allocate N from the shoot to the roots after herbivory. This inference is consistent with the observation that the N contents of WT roots increased after elicitation, as measured in a separate experiment, although the increase was not quite significant (Welch's two-sample t-test, d.f. = 4.71, P = 0.054; inset Figure 2). Alternatively, the increased N content of roots may have resulted from increased N assimilation, but previous 15N labeling experiments in this species have found no evidence for changes in N assimilation rate after herbivory (Baldwin and Ohnmeiss, 1994; Lynds and Baldwin, 1998). Therefore, we conclude that the induced biosynthesis of N-containing metabolites after OS elicitation alters whole-plant N partitioning.
Changes in absolute pool sizes depend on developmental stage
To analyze the influence of anti-herbivore defense induction, especially phenolamide biosynthesis, on within-shoot N allocation, we determined the absolute N pools of different leaf types (hereafter, total N pools) and N allocation to seeds by IRMS. Expressing resource allocation as concentrations reveals proportional allocations within an organ; however, total pools allow for comparisons among organs, as they are a function of both organ size and concentration (Chapin et al., 1990). We analyzed elicited older (oRL) and younger (yRL) rosette leaves to explore the influence of leaf development on N reallocation after elicitation, and the first (unelicited) stem leaf (S1) to examine systemic effects.
Overall, there was no clear effect of genotype or elicitation on the leaf total N pools. Total N pools varied among genotypes only in the S1 leaf (anova, F1,27 = 4.86, P = 0.036), whereas OS elicitation only reduced the total N pool of irLOX3 (two-sample t-test, d.f. = 8, P = 0.006) and WT (Welch's two-sample t-test, d.f. = 8, P = 0.021) in the yRL (anova, F1,28 = 7.40, P = 0.011). The N pool size in oRL was unaffected by genotype and elicitation (Figure 3). As N pool size correlates with biomass at the whole-plant scale (Baldwin and Hamilton, 2000), we evaluated whether the observed changes in total N pools of single leaves could be explained by changes in growth. Although the leaf size of yRL was reduced after elicitation (anova, F1,24 = 12.33, P = 0.002; Figure S2a), it did not correlate with total N pools (ancova, P = 0.187). Similarly, the change in total N pools of S1 leaves was not correlated with changes in leaf size (ancova, P = 0.406).
It is possible that changes in the total N pool of a leaf reflect changes in a major pool within the leaf, such as proteins. Although TSP pool size dramatically decreased in the yRL after elicitation, it did not correlate with the total N pool size in this tissue (ancova, P = 0.122; Figure 3). Thus, we conclude that although both pools are reduced by elicitation, the total N pool of the rosette leaves does not reflect the changes in TSP pool size or leaf size. This result is consistent with the hypothesis that total leaf N content and TSP (RuBisCO content) are controlled by different mechanisms, as has been shown for Oryza sativa (rice; Ishimaru et al., 2001; Makino et al., 2000).
The TSP pools differed between the lines in all three leaf types (anova; oRL, F1,27 = 8.70, P = 0.007; yRL, F1,27 = 12.95, P = 0.001; S1, F1,27 = 44.77, P = 3.5*10−07; Figure 3). The TSP pools of irLOX3 and irMYB8 in the yRL were reduced by about 50% after OS elicitation, whereas WT TSP pools were reduced by 91%. Both transgenic lines had constitutively larger TSP pools in the S1 leaf than the WT, and although S1 TSP decreased after elicitation in irLOX3 plants, TSP pools of both transgenic lines were still 2.5–3.0 times larger than those of WT after elicitation (Figure 3), suggesting that the biosynthesis of N-containing metabolites affects protein pool sizes, and that inducible defenses also have constitutive costs.
The recently developed method for the absolute quantification of single proteins allowed us to quantitatively compare the investment in defense metabolites with that in growth-related compounds, specifically, the photosynthetic protein RuBisCO, with similar accuracy (Ullmann-Zeunert et al., 2012). Being the most abundant soluble protein in plants, the total level of RuBisCO (sum of LSU and SSU) reflected the TSP pattern in the different leaves, independent of genotypes (ancova; oRL, P < 0.0001; yRL, P < 0.0001; S1, P = 0.42; Figures 3 and S3a). Overall, the data revealed a decrease in pool sizes of both RuBisCO subunits after OS elicitation (Figure S3a), which coincided with an increase in N-containing defense metabolites (Figure S3b), but the effects differed among lines, and for most traits measured irLOX3 showed an intermediate phenotype between WT and irMYB8. The two transgenic lines with either reduced (irLOX3) or undetectable levels of CP and DCS (irMYB8; Figure S3b) showed a smaller decrease of RuBisCO LSU and SSU than the WT in the elicited yRL (47–59% in irMYB8/irLOX3 compared with 92–95% in WT; Figure S3a). RuBisCO LSU and SSU levels were unaltered after elicitation in the systemic S1 leaf of irMYB8, but strongly declined in WT and irLOX3. The nicotine pool sizes showed similar induction patterns for all lines, except in the yRL, where the OS-elicited nicotine levels were higher in the WT than in the transgenic lines. These data suggest that the growth–defense trade-offs at the leaf scale are probably influenced by the capacity to biosynthesize and accumulate phenolamides, and that this also affects growth investments in the systemic S1 leaf.
As all transgenic lines used in this study accumulated similar levels of nicotine, it is unclear whether the biosynthesis of this alkaloid might affect N allocation to proteins (Figure S3b). To answer this question rigorously, experiments with transgenic lines completely with no flux of N into nicotine biosynthesis are needed. In the nicotine-silenced transgenic lines we have produced in our laboratory by silencing putrescine N-methyl transferase, nicotine biosynthesis is silenced, but the elicited flux of N into other alkaloids (anatabine) is not (Steppuhn et al., 2004).
Similarly, the induction of proteinase inhibitors could have additional influence on N allocation; however, preliminary experiments with virus-induced empty vector and MYB8-silenced plants showed a similar trypsin proteinase activity in both plants after elicitation (H. Kaur, personal communication), indicating that the synthesis of proteinase inhibitors does not seem to play a key role in the reallocation of N from primary to secondary metabolism.
A comparison of the two locally elicited leaves revealed differences in their defense and growth pool sizes: whereas the oRL accumulated the largest defense metabolite pools, with only slight reductions in TSP and both RuBisCO subunits after elicitation, the yRL had the strongest reductions in protein pools, with less pronounced increases in N-containing defense metabolite levels than the oRL (Figures 3 and S3). The optimal defense theory predicts that the allocation of defense metabolites is directly proportional to the fitness value of different plant parts (McKey, 1974, 1979; Rhoades, 1979), and many studies have demonstrated that younger leaves of N. attenuata, presumed to have a higher fitness value than older leaves, contain higher defense metabolite levels (Zavala et al., 2004a; Kaur et al., 2010; Onkokesung et al., 2012). These results appear to contradict our findings, because the oRL contained higher metabolite levels than the yRL; however, the previous studies compared concentrations of metabolites in elicited rosette leaves at different stages of plant development, whereas here we analyzed metabolite pool sizes of two elicited rosette leaves, of different maturity, harvested simultaneously from the same plant. As the plants were just beginning stalk elongation at the time of OS elicitation, both oRL and yRL are likely to be important tissues for later plant growth and reproduction. Thus the larger defense metabolite pools of the mature oRL – which was a source leaf at the time of the first elicitation – may result from its larger nutrient pools, which are probably important for regrowth capacity. Meanwhile, the smaller pools of TSP and RuBisCO in elicited yRL – which was in the transition stage from sink to source during the first W + OS treatment – may reflect a lower N allocation to proteins in developing leaves, which could enhance their defense status by reducing the food quality for herbivores. This is in agreement with the model from Orians et al. (2011), assuming that the mature source leaf allocates resources not only to defense and growth, but also to storage, thus making it relatively more valuable for the whole plant, and therefore better protected. Regardless of their ultimate explanations, these data demonstrate that growth–defense trade-offs are dependent on leaf development.
Many previous studies have demonstrated that inducible defenses are costly, often leading to a decrease in reproductive performance (Heil and Baldwin, 2002): e.g. growth–defense trade-offs at the leaf scale affect the N allocation to capsules in Nicotiana sylvestris (Ohnmeiss and Baldwin, 2000). However, here, neither the time of flowering and seed ripening, nor the number of mature capsules, the mass of the first mature seed capsule, nor the total N content of the first seed capsule were significantly different from controls after repeated simulated herbivory (Figure S4). This lack of observed fitness effects could result from species-specific differences or differences in the experimental design. In our experiment, OS elicitation may have been too early to affect seed set (the first capsules were harvested on average 18 days after the last elicitation), or the W + OS treatment was too weak to elicit changes in allocation to seeds, compared with the relatively stronger MeJA elicitation used in other experiments (Voelckel et al., 2001). In nature, wild tobacco faces strong intraspecific competition because of its mass-germination behavior, and strong alterations in N allocation to reproductive units in glasshouse-cultivated tobacco were only found when MeJA-elicited plants competed with control plants for the same limited resources (Van Dam and Baldwin, 2001). Thus, the costs and benefits of N allocation for a plant after herbivore attack may only become obvious if neighboring plants competing for the same limited resources are present. Additional experiments with plants grown in competition and exposed to simulated and natural herbivory are necessary to further explore the impact of growth–defense trade-offs within the leaf on plant fitness.
MYB8 indirectly affects nitrogen investment into proteins
The pool sizes of proteins and defense metabolites of the two transgenic lines suggest an influence of N-containing metabolite biosynthesis on the observed growth–defense trade-offs, but did not allow for a direct comparison of the levels of N demanded for metabolite biosynthesis, and the decreased N partitioned into TSP and RuBisCO after herbivory. By calculating the N investment into growth and defense per mg of fresh tissue mass after elicitation, we were able to further explore the role of phenolamide biosynthesis on N reallocation. We combined this approach with 15N pulse labeling to follow the investment of a defined N pool into both plant functions.
For all lines, and in locally treated leaves, elicitation decreased the N investment into rest TSP and RuBisCO per mg fresh mass, compared with controls. Particularly in yRLs, the decrease in N investment into TSP (rest TSP and RuBisCO) was much more pronounced in the WT (89%) compared with 28% in irMYB8 and 47% in irLOX3 (Figure 4a). IrLOX3 plants, for all parameters measured here, showed similar but less pronounced N-allocation patterns after elicitation as WT. These patterns are consistent with the correlation analysis of all measured N pools (Figure 4a, heat maps). Correlating all genotype/treatment groups with each other revealed that OS-elicited WT plants did not correlate with the other genotype/treatment groups in all three leaf types. Only OS-elicited irLOX3 oRL and S1 leaves showed a weak correlation with WT-OS. In contrast, irMYB8-OS did not correlate with any other genotype by treatment group.
Interestingly, the observed N-investment pattern is congruent with previous results on the patterns of MYB8 transcript accumulation in N. attenuata asLOX3 plants (which are comparable with irLOX3; Allmann et al., 2010; Halitschke et al., 2004). After elicitation, asLOX3 leaves have four times lower MYB8 transcript levels, whereas irMYB8 have 10 times lower levels than WT leaves (Kaur et al., 2010; Onkokesung et al., 2012). Furthermore, MYB8 functions downstream of JA signaling, and OS-elicited JA levels are not altered in irMYB8 plants (Kaur et al., 2010), whereas they are significantly reduced in LOX3-silenced lines (to about one-third of that in the WT, but roughly six to seven times higher than in untreated controls; Allmann et al., 2010). Thus, the observed phenotypes of the two transgenic lines are consistent with their respective MYB8 transcript levels, but not their JA levels; the MYB8 expression after elicitation in the three lines used in this study is inversely proportional to the N investments into soluble proteins. Based on these results we conclude that the observed changes in N allocation after simulated herbivory only indirectly depend on JA signaling, and are probably caused by differences in MYB8 expression or the MYB8-regulated synthesis of phenolamides. MYB8 could regulate defense induction by playing a role in N assimilation and allocation. In other plants and algae, members of the R2R3-MYB transcription factor family, to which NaMYB8 belongs, have been shown to be crucial for increases in the abundance of transcripts of N assimilation genes (Miyake et al., 2003; Imamura et al., 2009). To further elucidate the putative role of MYB8 in N reallocation, more detailed expression and enzyme activity studies targeting N metabolism at later time points after herbivory are necessary.
Based on our data we cannot differentiate whether MYB8 itself or the synthesis of phenolamides, in particular CP and DCS, mediate the changes in N investment into growth and defense. Silencing MYB8 also silences genes further downstream of the transcription factor, and in addition to CP and DCS, the synthesis of at least 29 different coumaroyl-, caffeoyl- and feruloyl-containing metabolites (Onkokesung et al., 2012). It is difficult to pinpoint the effects of single compounds in the complex biosynthetic network of a leaf, but applying phenolamides in different concentrations to control and elicited leaves of irMYB8 plants, and evaluating their effects on protein (RuBisCO) levels, or using plants silenced in genes affecting phenolamide biosynthesis downstream of MYB8, can help to evaluate if either MYB8 alone or MYB8 indirectly through phenolamide biosynthesis mediates the changes in N investment into proteins.
A comparison of the total N investment with the 15N investment per mg fresh mass revealed a similar pattern, with increased 15N in defense compounds and decreased 15N in both RuBisCO subunits after elicitation. One major difference was that WT and irLOX3 plants allocated proportionally more 15N than total N into CP and DCS, and less into nicotine, after elicitation, whereas the 15N investment into the RuBisCO subunits was proportionally similar to the total N investment in both control and elicited leaves (Figure 4b; for a clearer comparison of N and 15N investment, see Figure S5). Larger investments of recently assimilated 15N into CP and DCS, compared with nicotine, makes ecological sense, because the OS used was from M. sexta larvae, a tobacco specialist, which is nicotine-tolerant but negatively affected by phenolamides (Kaur et al., 2010).
A comparison of the decrease in total N investment into RuBisCO and TSP after OS elicitation with the N requirements of nicotine and phenolamide biosynthesis (Figures 4a and S6, showing a time-course analysis) suggested that RuBisCO metabolism could be a source of reallocated N to defense metabolite biosynthesis. Based on concentrations in the yRL, about 54% of N from RuBisCO or 13% of N from TSP could have been invested into phenolamides and nicotine (Figure S5). This comparison does not take into account the N requirements of biosynthetic enzymes or other N-containing inducible defense compounds, such as proteinase inhibitors (Zavala et al., 2004b). Hence, the N demands for defense metabolite biosynthesis are likely to be underestimated; however, considering the dramatic decline in TSP it is likely that more N is released from the turnover of primary metabolism than N invested into defense metabolites.
Nitrogen invested into phenolamides does not originate from RuBisCO after herbivory
To further elucidate the N flux into defense metabolites and to investigate whether RuBisCO N is used as a source of N for CP and DCS biosynthesis after OS elicitation, the 15N-incorporation (atomic percentage, At%) into N-containing metabolites and RuBisCO was determined in a time-course experiment (for details, see Figure 1b). This approach allows us to follow the N flux of a known quantity of 15N, independently of within-leaf N pool sizes. The experiment was carried out with the yRL, because this leaf showed the greatest differences in N investment after elicitation (Figure 4a,b). It is important to note that during the experimental period, the 15N-incorporation of the whole leaf was constant in all three lines, independent of elicitation (Figure S7), indicating that N is mainly redistributed within the leaves and that there is no increased net N influx into the leaf after elicitation.
As the 15N-incorporation into RuBisCO LSUs and SSUs was similar, we only report on the incorporation into LSUs. Incorporation into RuBisCO increased at a constant rate until it reached a maximum of about 8 At% between 4 and 7 days after the first OS elicitation in all three lines, independent of elicitation (Figure 5). In contrast, 15N was rapidly incorporated into CP and DCS in OS-elicited leaves until these compounds attained a maximum of about 10–12 At%, 4 days after the first elicitation in WT and irLOX3 plants. Had RuBisCO degradation provided the precursors for PA biosynthesis, it should have a similar or higher 15N-incorporation as phenolamides, because the precursor pools will have similar or higher labeled isotope incorporation rates as their derived compounds. The large differences in 15N-incorporation between CP and DCS and LSU make it unlikely that N derived from RuBisCO was used for CP and DCS biosynthesis. This result challenges the common conception that N released from products of primary metabolism (proteins) is a direct source for the production of defense metabolites (Herms and Mattson, 1992; Schwachtje et al., 2006). In contrast, the data indicate that recently assimilated N is channeled into defense metabolite synthesis (Figure 4b). We hypothesize that N released from TSP turnover is mainly reinvested into other compounds, enabling the plant to react in different ways upon attack. Thus, plants may reduce the nutritive value of the tissue by reducing the level of TSP, and at the same time investing N not only in defense metabolites, but also in other N-containing compounds that are less digestible for the herbivore or more easily reallocated.
The incorporation of 15N into nicotine only increased slightly after elicitation, and reached a maximum of around 2 At% in all three lines (Figure 5), although roots had a labeling of about 8 At%, similar to leaves (Figure S6). These findings differ from previous results showing the rapid incorporation of recently assimilated 15N into nicotine after elicitation, but those results were obtained from plants that were starved of N for 24 h before application of the 15N pulse, and 15N was applied at the same time as MeJA to the roots (Baldwin et al., 1994, 1998; Lynds and Baldwin, 1998). Elicitation of roots and shoots is known to differentially affect the accumulation of defense metabolites (van Dam and Oomen, 2008). Furthermore, MeJA is a stronger elicitor than OS elicitation (Voelckel et al., 2001), and N-starved plants are known to transport N preferentially to the strongest sink (Ohtake et al., 2001). These differences in experimental design probably led to different source–sink relationships within the plant, resulting in different patterns of 15N investments.
Nicotine is a constitutively synthesized pool in the roots of N. attenuata that is transported to the shoot, but not metabolized, and contains 5–8% of the total N in the plant (Baldwin and Hamilton, 2000). It is possible that the newly synthesized nicotine might be diluted by the large pool of previously synthesized unlabeled nicotine, resulting in a low 15N-incorporation. Alternatively, it may be derived from previously synthesized (and therefore unlabeled) precursors.
In summary, the 15N-incoporation illustrates the flux of a defined 15N pulse, independent of pool size, and indicates that N invested into CP and DCS is unlikely to be derived from RuBisCO, but is allocated directly to defense processes after assimilation instead of growth processes.
In this study, we quantified simulated herbivory-induced growth–defense trade-offs in a unified currency by measuring the investments of the limited resource of N into RuBisCO as proxy for growth and into small defense-related compounds (nicotine and phenolamides). In N. attenuata, OS elicitation reconfigures N allocation on multiple scales. Figure 6 summarizes the relative changes in the different N pool sizes after repeated simulated herbivory in the yRL. At the whole-plant scale, OS elicitation induced a weak N reallocation from the shoot to the root, thus presumably giving attacked plants a higher tolerance against herbivores by reducing the chances that valuable resources are removed by herbivores, and by increasing regrowth capacity after attack. At the within-leaf scale, changes between different N pools are much more dramatic. Taking the N level of RuBisCO after elicitation as a reference (x), RuBisCO-N declined 21x, and TSP declined from 73x to 9x, whereas N investment into defense metabolites increased, but to a far lesser extent (an increase from 5.5x to 11.5x for nicotine, and of 5x for CP and DCS).
The transcription factor NaMYB8, possibly by regulating the production of metabolically dynamic phenolamides, CP and DCS, indirectly mediates the reconfiguration of N allocation after elicitation. The comparison of two elicited rosette leaves indicated that the extent of reconfiguration and the total concentration of defense metabolites produced depends on the developmental stage of the leaf, and on source–sink relationships.
Flux studies with 15N strongly indicated that the N for PA biosynthesis comes from recently assimilated N rather than RuBisCO turnover. These results suggest that the drastic reallocation of resources and the shut-down of investments into growth within the leaf are not primarily driven by the direct costs for defense metabolite biosynthesis, but rather that N release from primary metabolism may enable the plants to react to attack in multiple ways. It remains to be elucidated if and how these allocation costs are translated into ecological costs. This question can only be answered if plants are grown in competition under different levels of herbivore attack.
Future experiments will seek to validate these results under more natural settings by comparing the results shown here for simulated herbivory, using OS of a specialist folivore, with damage by the natural herbivore community. An additional focus will be on tracing N investments into further metabolites and non-soluble proteins, and following the flux of N at the whole-plant level in more detail.
Plant germination and growth conditions
Seeds of the 31st generation of an inbred WT line of N. attenuata Torr. ex. Watts (Solanaceae) and two stably transformed lines, irMYB8 with reduced expression of the transcription factor NaMYB8 (A-08-810, Kaur et al., 2010), and irLOX3 silenced in lipoxygenase 3 (NaLOX3, A-03-562-2, Allmann et al., 2010), were sterilized and germinated according to Kruegel et al. (2002) and cultivated in 1-L pots. For details on cultivation and fertilization, see the Appendix S1. The transgenic lines were homozygous, near-isogenic to the WT and representative of several independent transformation events.
Pre-experiment to determine elicitation time points
Seven days after transfer to 1-L single pots, rosette-stage plants were pulse-labeled with 5.1 mg 15N in 50 ml of a 0.694 g l−1 solution of K15NO3 (modified from Van Dam and Baldwin, 2001), and the oldest sink leaf (hereafter yRL) and the youngest source leaf (hereafter oRL; Pluskota et al., 2007) were labeled for later sampling. The leaves and roots were harvested at 0, 4 and 12 h, and at 4, 7 and 10 days after the 15N pulse. Roots were washed to remove excess soil and all samples were dried for 48 h at 60°C. Between 3 and 10 days after the pulse, the leaves and roots had a constant 15N concentration (Figure 1b), indicating that an equilibrium had been reached. This time period was chosen for further experiments (Figure 1b, indicated by the grey arrows), as a stable 15N-incorporation facilitates the analysis of proportional allocation to single compounds. The N pulse did not have any obvious effects on plant growth.
Three days after the 15N pulse the oldest sink, youngest source and transition leaf at the time point of labeling were wounded with a pattern wheel and treated with M. sexta OS (10 μl per leaf per day, 1:5 diluted) on three consecutive days (Ullmann-Zeunert et al., 2012). Unelicited plants were used as controls. For the whole-shoot N analysis, the aboveground biomass of control and elicited plants was harvested 4 days after the first elicitation, and dried as above.
For the N-partitioning analysis, both the locally elicited yRL and oRL were harvested 4 days after the first elicitation and flash-frozen in liquid N2. After stalk elongation, the first stem leaf (S1) was harvested when it reached the source–sink transition stage. For all three leaves, only the right leaf blade was harvested to standardize sampling and minimize changes in source–sink relationships arising from repeated sampling. The harvest time points of the S1 leaf differed depending on plant development. The first mature seed capsules were harvested at the day of opening: seeds were counted, weighed and analyzed for N content. For the kinetic analysis, plants received a 15N pulse and were elicited as described above, and the locally elicited yRL was harvested at 0, 1, 3, 4 and 7 days after the first elicitation (Figure 1b). The sample size for all analyses was five.
Protein extraction and quantification
The TSP and RuBisCO LSUs and SSUs were extracted and quantified by Bradford assay and LC-MSE, respectively, as described by Ullmann-Zeunert et al. (2012). The 15N-incorporation of RuBisCO was determined with the Excel spreadsheet ProSIPQuant (Taubert et al., 2011).
Metabolite extraction and quantification
Small metabolites were extracted as in Gaquerel et al. (2010) and analyzed by UPLC/UV/ToF-MS, using a Dionex RSLC system with a diode array detector (Dionex, http://www.dionex.com) and a Micro-ToF Mass Spectrometer (Bruker, http://www.bruker.com). Further details on instrument parameters and quantification are described in Appendix S1. Average mass spectra were extracted for 15N-incorporations using the Excel spreadsheet ProSIPQuant (Taubert et al., 2011), modified for small metabolites based on compound sum formulae.
Isotope ratio mass spectrometry analysis (IRMS)
The IRMS sample preparation, analysis and following calculations of total N content (% dry mass) and 15N-incorporation were carried out as described in Meldau et al. (2012).
The r environment was used for statistical analysis (Team, 2009). For anova and ancova analyses, if the assumption of homoscedasticity of variances was violated or the residuals did not follow a normal distribution, response variables were transformed prior to the analyses using Box–Cox transformation (see Appendix S2). The Box–Cox lambda was estimated using Venables’ and Ripley's mass library for r. All anova models were simplified to the minimum adequate model using Aikaike's information criterion (Ronchetti, 1985). For the correlation analysis (Figure 4a, heat maps) the data were imported into the environment and vectors containing the following variables were generated: N-rest protein μg mg−1, N-RuBisCO LSU μg mg−1, N-RuBisCO SSU μg mg−1, N-nicotine μg mg−1, N-CP μg mg−1 and N-DCS μg mg−1. These vectors were pairwise correlated, calculating Kendall's τ coefficient (Kendall, 1938). In contrast to Pearson's correlation coefficient, Kendall's τ is more robust and not sensitive to the data distribution.
The authors thank Franziska Hufsky for bioinformatics help with RuBisCO quantification and Dr Matthias Schöttner for technical support with metabolite measurements. This research was supported by the Max Planck Society, M.A.S. was supported by a grant of the International Max Planck Research School and I.T.B. was supported by an advanced ERC grant, ClockworkGreen (293926).