Introduction

It is widely recognized that constitutive defenses limit the scale of the initial herbivore attack, whereas induced defenses determine patterns of subsequent attack (Karban and Myers, 1989; Haukioja, 1990; Jones et al., 1993; Karban and Baldwin, 1997). Induction has been documented in over 100 species of plants (Agrawal and Karban, 1999) and can be morphological, chemical, local, systemic, and highly specific (see Karban and Baldwin, 1997). For the purposes of this review, I define systemic induction as molecular, chemical, or morphological changes that occur in distant undamaged leaves. Recent advances have improved our understanding of plant responses to attack and the effects of these responses to herbivores. We now know that (1) damage results in long-distance transport of signal molecules that elicit changes in distant leaves (Zhang and Baldwin, 1997; Stratmann, 2003); (2) mechanical damage and damage by specific herbivores (and pathogens) generally elicit unique molecular, biochemical, and morphological responses (e.g., Baldwin, 1988; Thomma et al., 1998; Agrawal, 2000; Schittko et al., 2000; Walling, 2000; Voelckel and Baldwin, 2004); (3) induced responses can selectively affect the performance of herbivores and the behavior of natural enemies (Thaler, 1999a,b; Thaler et al., 1999, 2002; Agrawal, 2000; De Moraes et al., 2001; van Poecke et al., 2001); (4) there can be reproductive costs of induction (Baldwin et al., 1990, 1998; Agrawal and Karban, 1999; Heil and Baldwin, 2002; Zavala and Baldwin, 2004; Zavala et al., 2004); and (5) ontogeny sink and strength modify both the strength and pattern ofinduction (e.g., van Dam et al., 2001; Arnold and Schultz, 2002; Arnold etal., 2004).

We have made excellent progress in our understanding of the mechanisms, costs, and consequences of local and systemic induction. Unfortunately, the importance of spatial variation in systemic induction is often ignored but is critical to future progress. In this review, I focus on herbivore-induced induction within plants and argue that (1) vascular architecture, the pathway of signal movement to and away from the damaged leaf, is a key controller of systemic induction patterns; (2) when induction in some species is more widespread than expected, differences in phloem transport, release, and retrieval along the phloem pathway, and volatile production may allow for greater integration ofthe induced response; and (3) experimental designs that fail to account forspatial variation in induction are likely to underestimate or even fail to detect systemic induction. The points made apply to signal transduction following pathogen attack provided that the pathogen does not move systemically itself.

Vascular Architecture

In this section, I first describe how vascular architecture determines patterns of systemic induction and then review inter- and intraspecific factors that control patterns of vascular architecture.

Vascular Architecture and Long-Distance Transport

Botanists have known for years that the transport of photosynthate, signal molecules, and hormones among leaves is controlled by vascular architecture (reviewed by Murray et al., 1982; Watson and Casper, 1984; Watson, 1986; Dickson and Isebrands, 1991; Sachs et al., 1993; Marshall, 1996). There is evidence for both phloem and xylem transport of signal molecules associated with chemical induction (Malone and Alarcon, 1995; Stratmann, 2003).

From a phloem perspective, the anatomy of sieve cells has evolved so that the path of least resistance is longitudinal, through the sieve plates (vanBel et al., 2002), and there is little symplastic transport between mature cells, andthis limits movement among sieve cells (reviewed by Thorpe et al., 2005). The bulk of carbohydrate and signal transport is both longitudinal and toward sink leaves, but modified by vascular connectivity (Watson and Casper, 1984; Dickson and Isebrands, 1991; Preston, 1998; Arnold and Schultz, 2002). Connectivity is based on orthostichy—the phyllotactic arrangement of leaves that describes the distribution of vascular traces. Export from one leaf is greatest to orthostichous leaves because these leaves share primary vascular traces. Export to leaves in adjacent orthostichies is intermediate because these leaves have partial connectivity, and export is absent to leaves in opposite orthostichies because they lack vascular connectivity (Murray et al., 1982; Watson and Casper, 1984). There may be xylem transport of signal molecules as well (Malone and Alarcon, 1995). These authors suggest that upon leaf tissue damage, signal molecules can be drawn into the xylem and transported systemically. Since leaf-to-leaf connectivity is similar for xylem and phloem (see Orians et al., 2000; Zwieniecki etal., 2003; and reviewed by Orians et al., 2005), vascular constraints on signal transport are expected by either mechanism.

Whichever pathway of transport, xylem or phloem, within-plant variation in systemic induction is commonplace (Davis et al., 1991; Shulaev et al., 1995; Rhodes et al., 1999; Orians et al., 2000; Schittko and Baldwin, 2003). In cottonwood (Populus trichocarpa × Populus deltoides, Davis et al., 1991), tobacco (Nicotiana sp., Shulaev et al., 1995; Schittko and Baldwin, 2003), and tomato (Lycopersicon esculentum, Rhodes et al., 1999), leaves with direct vascular connections to the damaged leaf show greater defensive chemical induction than leaves without vascular connections. Systemic induction in tomato not only exhibits leaf-to-leaf variation, but within-leaf variation as well (Orians et al., 2000). Orians et al. (2000) found that proteinase inhibitor induction in tomato was greatest in leaflets that have direct vascular connections and lower in leaflets without direct connections. That tomato leaflets without direct connections exhibited induction suggest that signal transduction is not completely isolated to specific sectors (see Beyond Vascular Architecture for discussion of possible mechanisms).

Such chemical changes are ecologically relevant. Jones et al. (1993) demonstrated that differential induction alters leaf beetle performance on cottonwood (Populus deltoides). More recently, Viswanathan and Thaler (2004) found induced resistance in orthostichous leaves, but induced susceptibility in non-orthostichous leaves of Solanum dulcamara when damage was intense. This suggests that induction changes source–sink relationships (sensu Arnold and Schultz, 2002) and may result in greater allocation of resources toward resistance traits in the damaged sector and lowered allocation of resources toward resistance traits in other sectors.

Inter- and Intraspecific Variation in Phyllotaxy

Differences in vascular architecture among plants can alter patterns of systemic induction. Most angiosperms have a 2/5 phyllotaxy (Moore et al., 1998). A 2/5 phyllotaxy reflects the fact that every fifth leaf is vertically aligned after two spirals around the plant. In this arrangement, carbon exported from leaf 1 accumulates most in leaves 4 and 6, to a lesser extent in leaf 3, and minimally in leaves 2 and 5 (Dickson, 1991). However, some species have a 1/2 or 3/8 phyllotaxy (Moore etal., 1998). Moreover, since phyllotaxy may change through development—the phyllotactic arrangement of leaves in Populus changes from 2/5 to 3/8 to 5/13 with plant development (Larson, 1977)—it is essential to map the phyllotaxy with dyes (Orians et al., 2000) or isotopes (Larson, 1977) to determine the operative vascular connectivity.

Beyond Vascular Architecture

Vascular constraints on systemic induction are common; however, there is increasing evidence that induction in some species is less constrained (i.e., more integrated) (Shea and Watson, 1989; Mutikainen et al., 1996). Shea and Watson (1989) found that localized damage to fireweed, Chamaenerion angustifolium, despite having a 3/8 phyllotaxy, had little effect on fruit production within that orthostichy. They also report that 14C flow was less restricted. More recently, Orians et al. (2000) and Kiefer and Slusarenko (2003) showed that leaf tissues lacking direct vascular connections are induced in tomato and Arabidopsis, respectively. What might lead to induction in other sectors? I suggest that two factors may facilitate greater integration: (1) symplastic and apoplastic pathways in long-distance phloem transport and (2) volatile release at the site of damage and subsequent induction within distant plant tissues.

Throughout development, there is extensive apoplastic exchange—release and retrieval—of phloem constituents along the phloem pathway (Minchin and Thorpe, 1984,1987; van Bel et al., 2002; van Bel, 2003; Thorpe et al., 2005). Such release and retrieval requires an extensive array of specialized transfer cells and membrane-bound transporters (Pate and Dieter Jeschke, 1995; Fisher, 2000). Most importantly, from the perspective of integration, species differ both in type and in density of transporters (van Bel, 2003). Clearly, transport of damage-induced signal molecules could lead to more widespread induction. Further work is necessary to examine how interspecific variation in apoplastic exchange alters the degree of symplastic isolation and subsequent spatial patterning of systemic induction.

It appears that signals are most likely to reach young developing tissues in other sectors. There is little symplastic exchange among mature sieve elements, but early in tissue development there can be extensive symplastic transport due to high sieve element and plasmodesmal density (van Bel, 2003; Thorpe et al., 2005). This allows for greater movement between adjacent sieve elements in young developing tissues, and modification of microchannel size of the plasmodesmata can facilitate even greater exchange of carbohydrates, viruses, and perhaps signal molecules (Lucas, 1997; Ishiwatari et al., 1998; Ruiz-Medrano et al., 1999). Although many details remain unresolved, recent evidence suggests that modification of microchannel size is a highly regulated and active process (Ruiz-Medrano et al., 2004). Thus, symplastic transport, especially in combination with apoplastic exchange, could result in widespread induction in young developing tissues. Whether plants have evolved mechanisms to facilitate movement of signal molecules from one sector to another requires further study.

I suggest that volatile release from damaged tissues may also result in more widespread induction. In fact, volatiles may be the only mechanism to induce leaves on distant branches where vascular connections are lacking. It is well known that volatiles released from damaged tissues and can cause molecular and chemical changes in adjacent plants, and alter subsequent resistance (Farmer, 2001; Kessler and Baldwin, 2001). The implications of this work are great because they suggest that plants may eavesdrop on their neighbors and adjust their allocation to defenses accordingly. However, it appears that eavesdropping requires that plants be in close proximity (within 15 cm) (Karban et al., 1997). Air turbulence and movement serves to dramatically reduce the concentration of volatiles arriving at distant plants, and as a consequence, there is considerable debate concerning the importance of interplant communication under field conditions (Lerdau, 2002; Dicke et al., 2003). Given that volatile signals become increasingly dilute with distance, volatiles may be more important for intraplant communication. Such tissues are in close proximity and likely to have the appropriate receptors to respond but may be on different branches with no vascular connections (Watson, 1986). Although intraplant volatile communication has been hypothesized to occur (Farmer, 2001), no studies have examined the importance of intraplant volatile communication as a mechanism to bypass vascular constraints in systemic induction.

Experimental Approaches

Despite the potential for more widespread induction than would be predicted by vascular architecture alone, the presence of vascular constraints iscommon. Therefore, failure to account for vascular constraints in systemic induction can: (1) underestimate the strength of induction; (2) lead to erroneous conclusions about specificity of induction; and (3) hide important within-plant heterogeneity in leaf quality. In this section, I review the approaches researchers have taken in their experiments and highlight which of the approaches are likely to underestimate the strength of induction and lead to inappropriate conclusions. It is not mygoal to provide an exhaustive literature review, but to use a few case studies tohighlight how knowledge of vascular architecture might improve our understanding of systemic induction and ultimately lead to more carefully designed experiments.

To date there have been three approaches to studying induction and each differs in the importance of vascular architecture to the results (Table 1). (1) Some studies have released herbivores on young plants and then, later in development, measured induction of specific traits or induced resistance (hereafter termed “immunization”; Karban, 1986; Agrawal et al., 1999). (2) Other studies have isolated herbivores on specific leaves (or manually damaged the leaf) and then measured induction in undamaged leaves shortly thereafter (hereafter termed “localized damage”; Davis et al., 1991; Jones et al., 1993; Stout et al., 1996a,b). (3) More recently, researchers have sprayed plants with powerful elicitors such as jasmonic acid and salicylic acid (Cipollini and Redman, 1999; Thaler, 1999a,b; Thaler et al., 1999; Moran and Thompson, 2001; Redman et al., 2001; Cipollini, 2002). Most of the elicitor studies applied inducing agents to the entire plant and then measured whole-plant induction (or if applied to a specific leaf, induction was only measured in that same leaf; Cipollini and Redman, 1999). Because most experiments involving elicitors have effectively measured local induction, vascular architecture should have no effect on the outcome of experiments and will not be discussed further.

Table 1 Typical Treatments Imposed in Induction Studies, the Types of Responses Measured, and the Importance of Vascular Architecture to Patterns of Systemic Induction
Full size table

Immunization studies that give herbivores free access to plants early in development and then measure systemic induction later in development appear robust. For example, Karban (1986, 1987) showed that when young cotton, Gossypium hirsutum, seedlings are attacked by mites, the plants often show increased resistance to subsequent mite attack. Agrawal et al. (1999) obtained similar results with cucumber, Cucumus sativus. In these studies, cotyledons/leaves were damaged and the whole plant was assayed for subsequent resistance. Since the initial damage was systemic, we would not expect induction to vary among leaves (Table 1).

Problems may arise when initial damaged is localized to specific leaves and induced responses are measured in other parts of the plant. Induction could be underestimated if damage is concentrated on specific leaves, but induced responses are examined at the whole-plant level or might be severely underestimated in studies that only assay specific leaves for induced responses when the previous damage was concentrated on other leaves (Table 1). As illustration, Agrawal (2000) placed a caterpillar of one of four herbivore species (Plutella xylostella, Spodoptera exigua, Pieris rapae, or Trichoplusia ni) on one leaf and then measured how these herbivores performed on the remainder of the damaged plant. He found that previous damage by Plutella (a specialist) and Spodoptera (a generalist) resulted in increased resistance to all species. In contrast, previous damage by Pieris only induced resistance to Spodoptera and Pieris, and damage by Trichoplusia (a generalist) failed to induce resistance to the other herbivores. Although these results clearly indicated specificity of induction, some of the differences in herbivore responses may reflect differences in feeding behavior. Perhaps, Plutella—the species that appeared to be least affected by previous feeding—tended to avoid feeding on leaves with direct vascular connections to the damaged leaf and, thus, avoided the most heavily induced leaves. The later explanation is intriguing and deserves further examination.

More recently, Traw and Dawson (2002) released herbivores on four-leafed Brassica nigra and then quantified differences in foliar trichome density and sinigrin concentration on specific leaves between damaged and undamaged plants. They found evidence for trichome induction, but the effect was often insignificant and varied among plants. Perhaps differences in initial damage, which they report, but did not control for in their analysis, limited their ability to detect differences. Because they did not quantify the pattern of initial damage, they could have detected (1) no induction—if damaged was concentrated on a single leaf that did not have direct connections to the sampled leaf; (2) moderate induction—if damage was dispersed; or (3) high induction—if damage was concentrated on the leaf with direct vascular connections to the sampled leaf.

An extreme example of uneven damage exists in studies in which damage is applied to one leaf and induction is measured in other leaves. Since some leaves lack direct vascular connections and others have only partial connections, induction might be underestimated. For example, Stout et al. (1996a) used different leaflets from a leaf adjacent to the damaged leaf to examine how different enzymes are induced following damage. Since some of the leaflets had direct connections whereas others did not, it is possible that differences in induction reflect the experimental design, and not differences in the enzymes. If true, failure to detect induction could be an artifact of the experimental design. It would be better to measure induction in leaflets or leaves known to share direct vascular connections. Alternatively, researchers could distribute damage across several consecutive leaves to make sure all sectors were damaged (as done by Constabel et al., 2000).

Conclusion

Whether studying molecular mechanisms of systemic induction, cross-talk between different signal transduction pathways, specificity of induction, costs and benefits of systemic induction, or the effects of induced changes on herbivores and their natural enemies, a priori knowledge of vascular constraints on systemic induction will help researchers make robust conclusions. Care should be taken to control for (or to examine) the effects of vascular architecture on induction. Controlling for vascular architecture necessitates that initial damage is dispersed across all sectors of the plant. Examining its effects requires mapping the vascular connections and quantifying the amount of damage to specific leaves. Since within-plant heterogeneity in leaf quality may affect herbivore behavior or performance (Denno and McClure, 1983) and may even influence foraging patterns of natural enemies, we can even use vascular architecture to create and study the effects of heterogeneity on herbivores (reviewed by Orians and Jones, 2001). As we learn more about the mechanisms of within-plant communication, both by long-distance signal transport of dissolved substances within the vascular system and by volatiles, we will gain an even greater understanding of the importance of vascular architecture in systemic induction across species.