Herbivore feeding preference corroborates optimal defense theory for specialized metabolites within plants

Spatial and temporal variation of metabolically costly defenses within a plant is rationalized by the optimal defense theory, which predicts that defenses have evolved and are allocated based on the fitness value and vulnerability of individual plant parts to simultaneously optimize growth and defense (1–3). Much research in plant defense theory is focused on its ecological–evolutionary facets—studying whether organisms evolve defenses proportional to their risk of being attacked and inversely proportional to the cost of defense—which addresses one important aspect of the optimal defense theory (4, 5). A significant number of physiological investigations into optimal defense within individual plants reported correlation between defense distribution and either herbivore feeding preference, risk of attack, or plant fitness (4, 6). Moving beyond correlational tests of predictions of the optimal defense theory for within-plant defense distribution, plants genetically modified to eliminate biosynthesis or activation of defensive metabolites were exploited to indicate a causal link between defensive metabolites and herbivore feeding preference (7–9). Although these findings reveal the benefit of defense for the plant, they are based on comparing mutant plants without such metabolites with wild-type (WT) plants in which the defense is not uniformly distributed. Because of the lack of genetic tools to manipulate chemical defense distribution without eliminating them, empirical studies addressing the effect of optimal defense distribution within a plant on animal feeding preference and behavior are still missing. Here, we asked how the allocation of chemical defense compounds affects herbivore feeding preference using glucosinolates in Arabidopsis thaliana as a model.

Glucosinolates are amino acid–derived thioglucosides that yield herbivore-deterrent catabolites upon tissue damage–triggered enzymatic hydrolysis by myrosinases (10). However, many herbivores circumvent glucosinolate-based defenses via diverse behavioral and physiological counteradaptations (11). High selective pressure by coevolving predators conserved the production of glucosinolates despite large energy costs (12, 13). A. thaliana meets the predictions of the optimal defense theory: Young leaves are more valuable than old ones and display much higher glucosinolate concentration (14, 15). Similar ontogenetic patterns were reported for many other species of the Brassicales order, indicating existence of a universal glucosinolate distribution pattern (16–20). To examine the glucosinolate distribution in rosettes of the reference plant A. thaliana, we grouped leaves of 5-wk-old plants into three cohorts according to their age (Fig. 1A). Liquid chromatography–mass spectrometry (LC-MS) revealed that the total glucosinolate concentration is highest in young leaves and decreases gradually with leaf age (Fig. 1B). To investigate whether the ontogenetic pattern of glucosinolate concentrations across the rosette correlate with transcript levels of key enzymes involved in glucosinolate biosynthesis, we measured the expression of CYP79B2, CYP83A1, MAM1, MAM3, SUR1, and UGT74C1 by qRT-PCR. Gene expression was lowest in young leaves and gradually increased with leaf age. Although tissue-specific biosynthetic rates cannot be rigorously determined by expression analysis of biosynthetic genes, these findings suggest that glucosinolates are primarily produced in old leaves and subsequently transported to young ones (Fig. 1C). To test this hypothesis, we analyzed the tissue-specific glucosinolate distribution in knockout mutants of genes encoding the well-characterized H⁺/glucosinolate symporters NPF2.10/GTR1, NPF2.11/GTR2, and NPF2.9/GTR3 (21, 22). In gtr1 gtr2 double knockout mutants, the gradient of total glucosinolates across leaf groups was abolished, demonstrating that GTR1/GTR2-mediated transport establishes the ontogenetic glucosinolate distribution in the rosette (Fig. 1B). Compared to WT, total glucosinolate levels were significantly reduced in young leaves and increased in old and mature ones, indicating that GTR1 and GTR2 are responsible for the export of glucosinolates from old leaves. Intriguingly, indolic glucosinolates— particularly 1-methoxy-indol-3-ylmethyl glucosinolate—retained a gradual distribution in gtr1 gtr2 (SI Appendix, Fig. S1). Analysis of the tissue-specific glucosinolate distribution in gtr1 gtr2 gtr3 triple knockout mutants—GTR3 has high affinity toward indolic glucosinolates—showed that high indolic glucosinolate levels in young leaves are still maintained in the absence of all three glucosinolate transporters, despite significant accumulation in old and mature leaves leading to their uniform distribution among leaf cohorts (SI Appendix, Fig. S1). To test whether absence of a glucosinolate supply from old and mature leaves affects tissue-specific expression of biosynthetic genes, we analyzed the spatial gene expression of key biosynthetic enzymes in gtr1 gtr2 and found significant genotype × tissue interactions for CYP79B2, CYP83A1, MAM3, and SUR1 (Fig. 1C). The transcript levels of these genes were reduced in old leaves, indicating that GTR1/GTR2-dependent overaccumulation of glucosinolates in old leaves may lead to feedback inhibition of glucosinolate biosynthesis gene expression (Fig. 1 B and C). This supports the idea that the altered glucosinolate distribution in gtr1 gtr2 is caused by perturbation of transport that cannot be entirely compensated for by inducing the expression of biosynthetic genes in young leaves.

Next, we leveraged the uniform distribution of glucosinolates in the rosette of gtr1 gtr2 mutants and used glucosinolate-based chemical defense as a model to explore the effect of defense distribution on herbivore feeding preference. We performed a bioassay using larvae of Spodoptera littoralis, a generalist insect herbivore that infests plants from at least 40 families, including the families of the Brassicales order and many economically important species (23). S. littoralis showed a strong preference (100%, n = 90) for feeding on old and mature leaves of the WT, as S. littoralis larvae caused severe feeding damage on old and mature leaves while avoiding the young ones (Fig. 2 A and B and SI Appendix, Fig. S2). This feeding preference inversely correlates with leaf glucosinolate concentrations and is consistent with the predictions of the optimal defense theory for defense distribution within the plant. Strikingly, we found that the feeding preference is inverted (98.5%, n = 67) on gtr1 gtr2 (Fig. 2 A and B). Thus, the spatial pattern of glucosinolate distribution governs the feeding preference of S. littoralis. This conclusion is further supported by the observation of a preference for young leaves (100%, n = 56) on glucosinolate-deficient myb28 myb29 cyp79b2 cyp79b3 quadruple mutants (Fig. 2B and SI Appendix, Fig. S3). As GTR1 and GTR2 expression is not affected in the quadruple mutant (24), the finding demonstrates that the feeding preference specifically depends on glucosinolates, thus eliminating a potential role of GTR1/GTR2-mediated distribution of other metabolites (e.g., nutrients or hormones). Overaccumulation of glucosinolates in old leaves of gtr1 gtr2 cannot explain the feeding preference, as larvae also avoided feeding on old leaves of glucosinolate-deficient plants, probably due to the relative nutritional value and digestibility (Figs. 1B and 2B). The feeding preference for young leaves on biosynthetic and transport-deficient mutants supports the idea that high glucosinolate concentrations in young leaves are crucial to their protection. Consistent with previous studies showing that glucosinolates delay larval development of S. littoralis because of energy-consumptive detoxification (25), larvae gained significantly more weight on glucosinolate-deficient plants compared to WT (Fig. 2C). By contrast, S. littoralis feeding on glucosinolate transporter mutants did not affect the weight of larvae (Fig. 2C), possibly because the ingestion of the residual glucosinolates in young leaves of gtr1 gtr2 is sufficient to delay growth compared to the glucosinolate-free young leaves of the quadruple mutant. Hence, herbivore performance in terms of larval weight gain is not affected by shifting herbivore feeding to young leaves. We set out to assess plant fitness by quantifying fecundity, even though it already seems reasonable to infer that fitness cost will be incurred for plants that have lost most of their young leaves. However, when we exposed plants to larvae long term, gtr1 gtr2, gtr1 gtr2 gtr3, and myb28 myb29 cyp79b2 cyp79b3 mutants were not surviving to fruiting. We observed an overall survival rate of 0% (n = 11) for gtr1 gtr2, gtr1 gtr2 gtr3, and myb28 myb29 cyp79b2 cyp79b3 mutant plants, while WT plants survived to bolting (100%, n = 11) in two experimental replicates. In conclusion, A. thaliana profits from transporter-mediated, nonuniform glucosinolate distribution in the rosette, which prevents herbivore attacks against the most valuable and vulnerable tissues, thus providing physiological evidence for the central hypothesis of the optimal defense theory.

In addition to defense allocation within the plant, the optimal defense theory also predicts that defense is increased upon attack. To test how the tissue-specific distribution of glucosinolates is affected by the plant–herbivore interaction, we measured their concentration in young WT leaves upon S. littoralis feeding. As young leaves of the mutants were completely consumed during bioassays, we further used mechanical wounding to mimic generalist herbivores. Indolic glucosinolate levels were significantly (approximately threefold) induced by both treatments, indicating that herbivore-associated molecular patterns are negligible for this response (Fig. 3). These increases also occurred in gtr1 gtr2 and gtr1 gtr2 gtr3, suggesting tissue-specific induction of de novo indolic glucosinolate biosynthesis upon wounding (Fig. 3B and SI Appendix, Fig. S5), which is consistent with the reported up-regulation of genes involved in indolic glucosinolate biosynthesis upon S. littoralis feeding and wounding (26). Furthermore, S. littoralis exhibited a feeding preference for young leaves on aliphatic glucosinolate-deficient myb28 myb29, and not indolic glucosinolate-deficient cyp79b2 cyp79b3 and myb34 myb51 myb122 mutants, demonstrating that aliphatic, but not indolic glucosinolates, are critical for establishing the feeding preference (Fig. 2B and SI Appendix, Fig. S3). Indeed, the levels of long-chain aliphatic glucosinolates in young leaves increased significantly upon both feeding and wounding (SI Appendix, Figs. S4 and S5). Moreover, specifically long-chain methylsulfinylalkyl glucosinolates accumulate, indicating that feeding-inflicted wounding triggers S-oxygenation on the side chain in young leaves (Fig. 3 and SI Appendix, Figs. S4 and S5). S-oxygenation in response to wounding was attenuated in gtr1 gtr2 and, interestingly, the mutant overaccumulated the corresponding precursors (i.e., methylthioalkyl glucosinolates) in unwounded mature leaves (Fig. 3B and SI Appendix, Fig. S5). Yet the accumulation and modification of long-chain aliphatic glucosinolates is not required for the defense of young leaves, as S. littoralis showed a feeding preference for old leaves on long-chain aliphatic glucosinolate-deficient mam3 mutants (Fig. 2B and SI Appendix, Fig. S3). Hence, basal levels of total aliphatic glucosinolates are critical for optimal defense of young tissues against generalist herbivores such as S. littoralis.

Based on our results, we propose a model that predicts a function for mature tissues in serving as relays that orchestrate glucosinolate biosynthesis and transport to protect tissues with higher fitness values from attack (SI Appendix, Fig. S6). Briefly, glucosinolates are translocated from root to shoot via the xylem, which is controlled by a GTR1/GTR2/GTR3-mediated retention mechanism that is inactivated upon feeding-induced wounding. This is supported by our finding that mature leaves of untreated gtr1 gtr2 gtr3 mutants display elevated long-chain methylthioalkyl glucosinolate concentrations, which could not be further induced by wounding (SI Appendix, Fig. S5). In mature leaves, glucosinolates are GTR1/GTR2-dependently loaded into the phloem for transport to young leaves, which is supported by the expression of GTR1 and GTR2 in the vasculature and hydathodes (SI Appendix, Fig. S7). Young, immature plant tissues are well known to be strong sinks for carbon and nitrogen that are supplied by mature tissues. Because of metabolic constraints, biosynthesis of specialized metabolites—such as glucosinolates—in mature leaves may thus be more adaptive for plants than producing them in young leaves, which is consistent with the subhypothesis of the optimal defense theory predicting a trade-off between defense and other plant functions such as growth and reproduction. Higher defensive investment in young tissue is not limited to glucosinolates in Brassicales. Direct defenses, such as alkaloids, phenolic compounds, iridoid glycosides, cyanogenic glycosides, and furanocoumarins, were also shown to be higher in young leaves than old ones (8). Similar within-plant distributions were also found for indirect defenses (e.g., volatile organic compounds) and defensive structures (e.g., trichomes) (8). Most of the nonuniformly distributed direct and indirect defenses are subject to transport, as they can be detected in phloem and/or xylem exudates (27). For example, pyrrolizidine alkaloids act as a defense against generalist herbivores and are transported from old to young leaves of Cynoglossum officinale (28). Benzylisoquinoline alkaloids (e.g., morphine in Papaver somniferum) are (at least partly) synthesized in sieve elements of the phloem and stored in laticifers (29). In tobacco, nicotine is synthesized in the root and transported to the leaves via the xylem in response to herbivore feeding (30, 31). In recent years, numerous transporters with affinities toward defensive metabolites have been identified (32). Based on the current knowledge, it is conceivable that transporters are important for the distribution of other types of chemical defenses in other species (33). As the relative value and vulnerability of plant tissues change during development, defenses need to be tightly regulated to reduce metabolic costs. Metabolic costs for defense transport are largely unknown, but it seems likely that total costs are reduced if defenses are (re)allocated to the target tissues. To sum up, using glucosinolate-based defense of A. thaliana against the generalist herbivore S. littoralis, we were able to provide experimental evidence for the physiological aspects of the optimal defense theory and highlight the importance of coordinating biosynthesis and transport processes of defensive metabolites to obtain the most favorable distribution within a plant for optimizing growth and defense.