Understanding ontogenetic trajectories of indirect defence: ecological and anatomical constraints in the production of extrafloral nectaries

Abstract

Background and Aims

Early ontogenetic stages of myrmecophytic plants are infrequently associated with ants, probably due to constraints on the production of rewards. This study reports for the first time the anatomical and histological limitations constraining the production of extrafloral nectar in young plants, and the implications that the absence of protective ants imposes for plants early during their ontogeny are discussed.

Methods

Juvenile, pre-reproductive and reproductive plants of Turnera velutina were selected in a natural population and their extrafloral nectaries (EFNs) per leaf were quantified. The anatomical and morphological changes in EFNs during plant ontogeny were studied using scanning electron and light microscopy. Extrafloral nectar volume and sugar concentration were determined as well as the number of patrolling ants.

Key Results

Juvenile plants were unable to secrete or contain nectar. Pre-reproductive plants secreted and contained nectar drops, but the highest production was achieved at the reproductive stage when the gland is fully cup-shaped and the secretory epidermis duplicates. No ants were observed in juvenile plants, and reproductive individuals received greater ant patrolling than pre-reproductive individuals. The issue of the mechanism of extrafloral nectar release in T. velutina was solved given that we found an anatomical, transcuticular pore that forms a channel-like structure and allows nectar to flow outward from the gland.

Conclusions

Juvenile stages had no ant protection against herbivores probably due to resource limitation but also due to anatomical constraints. The results are consistent with the growth-differentiation balance hypothesis. As plants age, they increase in size and have larger nutrient-acquiring, photosynthetic and storage capacity, so they are able to invest in defence via specialized organs, such as EFNs. Hence, the more vulnerable juvenile stage should rely on other defensive strategies to reduce the negative impacts of herbivory.

INTRODUCTION

Plants have a limited amount of resources available to perform a set of physiological functions such as establishment, growth, reproduction, storage and resistance. Both resource availability and allocation priorities change during plant development (Stamp, 2003; Weiner, 2004; Boege and Marquis, 2005). In addition, plants face different environmental stresses as they develop, such as herbivore and pathogen attack (Warner and Cushman, 2002; Del Val and Dirzo, 2003; Barrett and Agrawal, 2004; Thomas et al., 2010). Seedlings and juvenile plants may be particularly vulnerable; their priority is often to establish and grow to outcompete neighbours at a time when plants are severely resource-limited due to their small photosynthetic and storage organs (Boege and Marquis, 2005). However, herbivore damage should be proportionally more harmful earlier than later during plant development. Hence, defence strategies are expected to vary during the plants' ontogeny according to the physiological priorities, resource availability and impacts of herbivore damage at each ontogenetic stage (Boege and Marquis, 2005; Barton and Koricheva, 2010). Understanding the processes and mechanisms that constrain plant defence at juvenile stages may contribute to our knowledge of the evolution of ontogenetic trajectories in plant defence.

Plant indirect defensive mechanisms promote the reduction of plant quality to herbivores, by enhancing search efficiency and attack by predators or parasitoids (Boege and Marquis, 2005). Such indirect defences involve mutualistic interactions between plants and predators. For example, in myrmecophytic associations, plants provide food and/or nesting sites for ants, which in turn protect them from attacking herbivores (Bentley, 1977; Del Val and Dirzo, 2003). Although this mutualistic association is an effective way of deterring herbivores (Bentley, 1976; Koptur, 1979; Oliveira, 1997; Dutra et al., 2006), most myrmecophytic plants are not associated with ants during their early ontogeny. But given that plants need particular morphological structures to lodge or feed ant partners (Del Val and Dirzo, 2003) the expression of such traits may be too costly for young plants, given that it is not until later in their development that the protective association begins. However, a few studies have assessed the specific mechanisms by which these ontogenetic changes occur and particularly the constraints that juvenile plants face to produce such indirect defensive traits. The understanding of such constraints could elucidate variation in plant indirect defence found in the field, and the morphological features that constrain natural selection on indirect defences at early ontogenetic stages.

Ontogenetic patterns of indirect defence and their consequences for plant fitness are poorly understood (see Trager and Bruna, 2006; Wooley et al., 2007; Rios et al., 2008). In general, plants tend to increase the number of extrafloral nectaries (EFNs), domatia and the production of rewards for their mutualists as they age. For example, as Acacia drepanolobium plants age, the proportion of leaves bearing nectaries and ant patrolling increases (Young et al., 1997). A similar pattern is seen in cholla cacti plants (Opuntia imbricatai) in which EFN density peaks at the onset of reproduction (Miller, 2007). The same ontogenetic pattern has been reported for the total amount of food bodies in Macaranga triloba (Heil et al., 1997). These ontogenetic trajectories have been mostly explained in terms of resource limitation for young plants to produce such traits (Del Val and Dirzo, 2003; Izzo and Vasconcelos, 2005; Miller, 2007), or due to the architectural constraints associated with a plant's ability to attract a whole colony of ants (Fonseca, 1993, 1999; Boege, 2005). However, to date, no study has assessed the anatomical and histological constraints promoting the observed ontogenetic patterns in indirect defences, particularly in EFNs.

As plants grow, their ability to produce tissues changes; this influences the morphology, anatomy and physiology of different plant organs. For instance, as woody plants age, their flexible stems harden by the accumulation of lignin fibres (Hatfield and Vermerris, 2001; Bowes and Mauseth, 2008) and the stem epidermis is replaced by bark (Mauseth, 2003; Lendzian, 2006). Secondary growth by itself is an example of an ontogenetically induced anatomical change (Blatrix et al., 2012). In addition, the stem's photosynthetic function is lost at the onset of secondary growth (Mauseth, 2003). Given that cork is impermeable to gaseous diffusion (Bowes and Mauseth, 2008) it also imposes new ontogenetic challenges to gas exchange. In the stems of young plants gas exchange occurs through stomata (Lendzian, 2006), an epidermic pore which consists of two guard cells that regulate its opening (Mauseth, 2003). But as woody plants age, they develop cork openings called lenticels, functionally analogous to stomata (Lendzian, 2006) through which gas exchange is possible in lignified stems (Topa and McLeod, 1986).

For the particular case of EFNs, their development has been widely explored at particular ontogenetic stages but studies have focused primarily at reproductive plants (Kobayashi et al., 2008; Andrade de Aguiar-Dias et al., 2011; Delgado et al., 2011). Several aspects of EFN emergence, differentiation, senescence, changes in cytoplasm and nuclei density, chemical nature, the axis of the mitotic divisions (Dave and Patel, 1975; Durkee, 1982; Thadeo et al., 2008), and even the genes that promote nectar development (Lee et al., 2005) have been reported, but there are no assessments of the anatomy and histology of fully developed EFNs at different plant ontogenetic stages (Escalante-Pérez and Heil, 2012). Such information could help to understand the ontogenetic trajectories in this indirect defence. Because many ecological studies have focused exclusively on the differences in resource allocation to several plant functions and structures (Schwinning and Weiner, 1998), we still understand little about the physiological, cellular and molecular mechanisms influencing their expression throughout development (Escalante-Pérez and Heil, 2012). The aim of this study was to describe the underlying structural and physiological mechanisms of the ontogenetic trajectories of EFNs, and their consequences for the interaction between plants and their associated ant defenders.

MATERIALS AND METHODS

Study site and system

The study was conducted in a population of Turnera velutina (Passifloraceae) growing in stabilized sand dunes at field station Centro de Investigaciones Costeras La Mancha (19 °35′N, 96 °22′W, <100 m above sea level), on the Gulf of Mexico, Veracruz, Mexico. The climate is warm and subhumid with rains in the summer (June–September) (Travieso-Bello and Campos, 2006). Fieldwork was conducted from June 2009 to September 2010.

Turnera velutina is a Mexican endemic shrub with a ramified, reddish, woody stem and pubescent, linear or ovate leaves with serrate borders, growing in an alternate phyllotaxis (Arbo, 2005). The population growing at our study site was recently reclassified as T. velutina having previously been reported as T. ulmifolia (Arbo, 2005). Sexual maturity is reached in about 4–7 months and plants can live for several years. Although it flowers year round, reproductive output peaks during the summer. T. velutina has floral and foliar EFNs at the base of the leaf blade or at the union between the petiole and the lamina, and the EFNs have a paired, lateral and opposite position and a well-defined cup cavity (Arbo, 2005). The main extrafloral nectar consumers are wasps, ants and bees. Besides defending the leaves of T. velutina, ants also disperse the seeds and eat the elaiosome, allowing them to germinate. The main foliar herbivores are Euptoieta hegesia (Lepidoptera: Nymphalidae) larvae which are most active during the summer (Cuaultle and Rico-Gray, 2003). Preliminary field observations suggest that juvenile plants are damaged more by herbivores (N. Villamil, unpubl. res.). To understand the ontogenetic trajectories in EFN production, three ontogenetic stages were defined as follows: juvenile (J) (plants measuring 15 cm or less, without reproductive structures), pre-reproductive (Pr) (plants measuring more than 15 cm without reproductive structures) and reproductive (R) (plants measuring more than 15 cm with reproductive structures or scars left by flowers and fruits).

EFN abundance

Juvenile, pre-reproductive and reproductive plants were haphazardly selected in the field (n = 55 plants per stage). The abundance of EFNs was assessed in each plant during six bimonthly censuses. Plants that outgrew their ontogenetic stage or died were not included in further censuses. The number of EFNs per leaf was counted in 15 mature leaves of each plant. To ensure that leaves were representative of the entire plant canopy, EFN abundance was quantified in every x mature leaves, where x is the total number of leaves/15. If a plant had fewer than 15 leaves, all were surveyed. Because data could not be normalized, the influence of ontogeny on the average number of EFNs per census per plant was assessed using a Kruskal–Wallis test, and Mann–Whitney U tests for a priori contrasts.

Structural and morphological studies

Samples for light and scanning electron microscopy were collected in June 2009. Five plants per stage were randomly chosen, and from each, five mature leaves were fixed in FAA (formaldehyde, 95 % ethanol, glacial acetic acid, distilled water 1 : 5 : 0·5 : 3·5, v/v). Due to the severe defoliation that five leaves represented for juvenile plants, only two leaves per plant were collected for this stage. When processed, the tissue was rinsed in water for 30 min to remove the fixator excess that the storage time might have caused, and then dehydrated in a graded ethanol series.

After the graded ethanol series, samples used for structural studies with light microscopy followed a gradual substitution of ethanol by xylene before infiltration and embedding in Leica Paraplast embedding medium at 52 °C for 12 h. Longitudinal sections (6–8 µm) were cut with a rotary microtome and stained with safranin and fast green. Once stained, sections were preserved with Enthellan resin and polymerized for 48 h at 56 °C. The plates were examined with a photomicroscope (Provis AX70; Olympus, Tokyo, Japan) using light and phase contrast microscopy. Detailed studies of nectary morphology were performed using scanning electron microscopy (SEM). Samples were isolated after the graded ethanol dehydration series, dried to critical point with liquid CO₂, mounted on aluminium stubs with two-sided tape and sputter-coated with a gold–palladium alloy. The samples were observed and photographed with an electron microscope (JSM-5310 LV; JEOL, Tokyo, Japan). Due to its resolution power, light microscopy and SEM techniques are suitable to assess changes in EFN tissue structure, rather than transmission electron microscopy that focuses on intracellular, ultrastructural features.

Nectar collection

The influence of ontogeny on extrafloral nectar secretion was assessed in September 2010 in 79 plants (19 Pr and 60 R). Juvenile individuals were not considered in this survey because most plants had either outgrown this stage or died by the time nectar was assessed. All plants were double-bagged the night before with a wire-tied organza bag inside a waxy paper bag sealed with masking tape. The organza bag prevented ants from reaching the EFNs, while the waxy paper retained humidity, thus avoiding changes in nectar volume caused by sun dehydration or dew hydration. Finally, a strip of masking tape with Tanglefoot (Forestry Suppliers Inc., Jackson, MS, USA) was placed around the plant's stem below the bags, as an additional ant barrier. Extrafloral secretion was assessed at the peak of nectar secretion, which occurred between 1100 and 1300 h. Nectar was collected from each isolated leaf using 1-μL microcapillary pipettes. Nectar volume (μL) was calculated by measuring the height of the nectar column with a digital vernier. The content was then released onto the prismatic surface of a hand-held refractometer (0–50 °Brix) to determine its sugar concentration. Differences in the average volume (μL) and sugar concentration (°Brix) between ontogenetic stages were examined using a Mann–Whitney U test.

Ant patrolling

The number of ants patrolling each plant was considered as an estimator of the level of biotic defence. In June 2010, 87 plants (18 J, 23 Pr and 46 R) were monitored during three consecutive days, every 2 h during a 12-h period (0700–1900 h), counting for 1 min the number of ants in each plant. The total number of ants was divided by the total number of leaves per plant to assess the influence of ontogeny upon ant attraction independently of plant size. Differences in ant patrolling between ontogenetic stages were examined using a Mann–Whitney U test.

RESULTS

EFN abundance

The number of EFNs per leaf ranged from zero to five, in general showing a paired distribution, although occasionally they were present in uneven numbers. Plant ontogeny had a significant effect on EFN frequency, increasing up to 40 % as plants developed (χ² = 51·02, P < 0·0001). Reproductive plants consistently had the most EFNs per leaf, followed by the pre-reproductive and then juvenile plants (Fig. 1).

Fig. 1.

Ontogenetic variation in leaf extrafloral nectary density in Turnera velutina. Different letters indicate significant differences among ontogenetic stages within each census (P < 0·05).

EFN structure and morphology

Although morphologically paired, the EFNs in T. velutina have an asymmetric petiole insertion (Fig. 2A). These glands, once mature, have a cup shape conferred by the elevated hems that contain nectar drops. At the centre of the cup is an elevated plateau. Across all stages, their anatomy consists of an external cuticle, a layer of secretory epidermis, a tight parenchyma that forms most of the gland's body, vascular elements, and a loose parenchyma that forms a shaft and might contain calcium oxalate crystals and tanniferous cells (Fig. 2B). The main physiological contributions of these anatomical features are: water loss prevention (cuticle), nectar exudation (secretory epidermis), structural support (parenchyma), herbivore deterrence (tannins and calcium oxalate crystals), and water and photosynthate supply (vascular elements). In spite of these functional elements, gland morphology changed considerably between ontogenetic stages (Fig. 2C–H). Juvenile plant EFNs were not yet cup-shaped, bearing descendent hems and a convex secretory surface, although an elevated plateau was observable in the centre of the secretory surface. Gland diameter ranged from 370 to 580 µm, and the plateau measured approx. 108 µm (Fig. 2C). The cuticle was thick. The underlying secretory epidermis was unistratified, formed by palisade cells that were slightly elongated and with a densely stained cytoplasm. The gland's body consists of a parenchyma with isodiametric cells with a lightly stained cytoplasm, abundant calcium oxalate crystals and just a few tanniferous cells. Vascular elements were found near the base of the gland (Fig. 2D). EFNs in pre-reproductive plants did not have a fully developed cup shape, but the secretory surface was concave and the hems were flat (Fig. 2E). Cuticle thickness was uneven throughout the entire surface. The elevated plateau in the centre of the secretory surface was fully developed, exhibiting a crater in the middle where the cuticle thins considerably (Fig. 2F). Gland size ranged from 400 to 740 µm, and the apical plateau measured approx. 100 µm (Fig. 2E). The underlying secretory epidermis was unistratified, formed by slightly elongated, palisade cells. The gland body was formed by a tight parenchyma and contained a vast array of calcium oxalate crystals. Beneath, a loose parenchyma with lightly stained cytoplasm cells formed a bottlenecked shaft where the vascular elements were found. Tanniferous cells were distributed along the petiole tissue in a ringed pattern (Fig. 2F). In reproductive plants, EFNs had a fully developed cup shape with a concave secretory epidermis and prominent ascendant hems (Fig. 2G). In this stage there are usually several nectaries on each side of the petiole that differ in size, being at least one main nectary considerably larger, next to the secondary, smaller glands. The elevated plateau in the centre of the secretory surface was fully developed, exhibiting a crater in the middle, where the cuticle thins. Gland size ranged from 700 to 1000 µm, and the apical plateau measured approx. 100 µm (Fig. 2G). Cuticle thickness was uneven throughout the entire surface. The underlying secretory epidermis was bistratified, formed by slightly elongated, palisade cells with a densely stained cytoplasm. The gland's body and shaft parenchyma had the same anatomical features as those in the pre-reproductive stage. However, in the reproductive stage vascular elements extended up through the parenchyma and very close to the secretory epidermis. A vast array of calcium oxalate crystals was found in the central portion of the parenchyma, while tanniferous cells were mainly distinguished along the external borders of the gland (Fig. 2H).

Fig. 2.

Topography, morphology and structure of extrafloral nectaries in Turnera velutina among ontogenetic stages. (A) Leaf underside showing the petiole and extrafloral nectaries in a reproductive plant. (B) Structural scheme of an extrafloral nectary in longitudinal section. (C) Scanning electron micrograph of the extrafloral nectary's external morphology from a juvenile plant (200×). (D) Median, longitudinal section of an extrafloral nectary from a juvenile plant (phase contrast). (E) Scanning electron micrograph of the extrafloral nectary's external morphology from a pre-reproductive plant (100×). (F) Close-up at the apical portion of the elevated plateau in a median, longitudinal section from a pre-reproductive plant (phase contrast). (G) Scanning electron micrograph of the extrafloral nectary's external morphology from a reproductive plant (100×). (H) Median, longitudinal section of the extrafloral nectary from a reproductive plant (light microscopy). Abbreviations: ct, cuticle; me, elevated plateau; es, secretory epidermis; pq, tight parenchyma; pql, loose parenchyma; t, tanniferous cells; d, calcium oxalate crystals; hv, vascular elements; tr, simple unicellular trichome; trg, glandular trichome; ne, accumulated extrafloral nectar.

Nectar release mechanism: pore

Extrafloral nectar is produced by the secretory epidermis and accumulates between the epidermis and the cuticle (Fig. 3A, B). As the volume of nectar increases, cuticle sections rise separating from the epidermis but never become detached or burst (Fig. 3C, D). Finally, the nectar flows out through a conspicuous anatomical opening: a transcuticular pore found in the middle of the plateau crater. In this region the cuticle becomes considerably thinner and invaginates in the centre, forming a channel that connects the gland's interior with the external environment (Fig. 3E, F). Once outside the gland, the liquid is contained by the prominent, ascendant hems of the nectar and the cup reservoir fills with macroscopic nectar drops (Fig. 2A).

Fig. 3.

Secretion route and mechanism of extrafloral nectar in a reproductive plant of Turnera velutina. (A–D) Progressive accumulation of produced nectar under the raising cuticle (A, C: light microscopy; B, D: phase contrast). (E) Median, longitudinal section of the apical portion in a nectar-secreting gland, showing the transcuticular pore (phase contrast). (F) High magnification at the transcuticular pore through which nectar flows out. Abbreviations: ct, cuticle; es, secretory epidermis; pq, tight parenchyma; t, tanniferous cells; d, calcium oxalate crystals; tr, simple unicellular trichome; ne, accumulated extrafloral nectar; rs, secretion route; po, transcuticular pore.

Extrafloral nectar secretion and ant patrolling

Plant ontogeny significantly influenced both nectar secretion and ant patrolling. Reproductive plants secreted almost ten times more nectar than pre-reproductive plants (U₁ = 25·55, P < 0·0001, Fig. 4A). In terms of quality, reproductive plants also exceeded pre-reproductive plants, secreting nectar with ten times more sugar (U₁ = 21·54, P < 0·0001, Fig. 4B). Although juvenile plants were not considered in this survey, our anatomical evidence suggests an inability to secrete extrafloral nectar. Consequently, no ant patrolling was observed at this stage. Forty per cent more ants per leaf were observed in reproductive than in pre-reproductive plants (0·05 ± 0·025 and 0·07 ± 0·026 ants per leaf, respectively; Fig. 5).

Fig. 4.

Differences among ontogenetic stages in (A) the volume and (B) sugar concentration of extrafloral nectar secreted in Turnera velutina. Different letters indicate significant differences among ontogenetic stages (P < 0·05).

Fig. 5.

Differences among ontogenetic stages of Turnera velutina in ant patrolling frequency. Different letters indicate significant differences among ontogenetic stages (P < 0·05).

DISCUSSION

We have shown that the abundance of leaf EFNs, gland histological complexity, the volume and quality of the secreted nectar, and the number of ants attracted to plants increased as T. velutina plants age. This ontogenetic pattern could be associated with a reduced ability of young plants to acquire and fix carbon, but also with their inability to produce structurally complex and productive EFNs. Mature plants are able to produce more and better quality nectar rewards and are likely to benefit from greater defensive activity of their ant mutualists. In addition, we have described detailed changes in EFN structure and histology throughout plant ontogeny, elucidating the mechanism of extrafloral nectar release in T. velutina (Elias et al., 1975; González, 1996; González and Ocantos, 2006) and partially explaining why younger ontogenetic stages are unable to produce extrafloral nectar.

EFNs in T. velutina have been described as derived and specialized glands based on the secreted carbohydrates, their vascularization pattern (Elias et al., 1975), elevated cup-shaped morphology and the apical plateau (González, 1996; Díaz-Castelazo et al., 2005). Based on the histological complexity of these glands and their secretion, EFNs might be costly structures for plants (Southwick, 1984; Mondor et al., 2006). Besides their contribution as indirect plant defences, EFNs in T. velutina could also have direct effects on herbivore deterrence given their contents of calcium oxalate crystals and tanniferous cells. In particular, tannins have a negative influence on the amount of herbivore damage (Feeny, 1970) and have been considered as direct plant defences (Hagerman and Butler, 1991) due to the negative effects their oxidative properties have upon herbivore metabolism (Salminen and Karonen, 2011). Although the precise function of crystal formation is unclear, much attention has focused on its role as a noxious anti-foraging mechanism (Finley, 1999). Many plants possess calcium oxalate crystals with a well-described protective function (Franceschi and Horner, 1980; Ward et al., 1997; Hanley et al., 2007). The barbed and grooved edges of raphides and druses, two types of calcium oxalate crystals, are particularly irritating to mouth and throat tissues when eaten (Prychid and Rudall, 1999) with crystal microstructure and size both contributing to irritation (Sakai et al., 1984).

EFN production and accumulation mechanisms in T. velutina have been described by various authors (Elias et al., 1975; González, 1996; González and Ocantos, 2006). Pre-nectar sugars and water are transported from the phloem by diffusion. Once mixed, the nectar moves from the intercellular spaces to the surface by pressure-driven mass flow, generated by the active transport of sugars from the cytoplasm of secretory cells (Vassilyev, 2010). However, there is some debate regarding the release mechanism and cuticle structure in T. velutina. On the one hand, Elias et al. (1975) stated that the release mechanism consisted of a cuticle burst promoted by the pressure of the liquid accumulated between the secretory epidermis and the cuticle (see also Nepi, 2007). Nevertheless, although a cuticle regeneration process after each secretion is plausible, there is no histological evidence to support this assertion (González, 1996; González and Ocantos, 2006). The alternative hypothesis proposed secretion as a process of nectar leak through the thinnest portion of the cuticle, which coincides with a pore-like structure (González, 1996; González and Ocantos, 2006). The main caveat, however, was the definition of a pore as a parenchyma protuberance without any anatomical opening. This hypothesis requires that secretion should occur through the cuticle, but this would entail a very thin, waxy layer to become permeable or hydrophyllic, allowing the aqueous nectar solution through (González, 1996; González and Ocantos, 2006). The cross-sections presented here reveal a transcuticular pore: an anatomical opening consisting of a channel-like structure made by the invagination of the cuticle at the thinnest portion (Fig. 2E, F). This pore connects the inner gland tissue with the external environment. Hence, the nectar accumulated beneath the cuticle flows outwards driven by mass flow pressure (Vassilyev, 2010), through the pore, and is contained by the cup-shaped nectary.

We found increasing EFN abundance per leaf, nectar production and ant attraction during T. velutina development in accordance with observations in other systems (Young et al., 1997; Veena et al., 1989; Falcao et al., 2003; Trager and Bruna, 2006; Miller, 2007). For the first time we report increasing EFN structural complexity and histological changes throughout plant development. EFN structure, morphology, physiology and ecology changed markedly as plants developed. In juvenile plants, EFNs are immature and histologically unable to produce nectar due to their incipient secretory epidermis. Secretion is also impossible because the cuticle is continuous and thick, even above the apical plateau where the pore is located in further stages. These features prevent juveniles from releasing nectar and, in addition, from a morphological point of view these EFNs are unable to hold a nectar drop given that the hems are descendent and thus the cup shape is not formed – hence young plants are anatomically unable to continuously offer EFN to their mutualists. This probably explains why no ants were observed to patrol plants at the juvenile stage. In contrast, EFNs from pre-reproductive plants presented a secretory epidermis fully differentiated and productive with a thinned cuticle over the apical plateau, allowing nectar release. However, at this stage the hems were still flat and large nectar drops cannot be successfully contained. Finally, at the reproductive stage plants presented fully developed and mature EFNs. A duplicated secretory epidermis could explain the patterns in nectar production found in the field, which was ten times greater in volume and sugar concentration in reproductive plants than in earlier ontogenetic stages. The elevated hems forming the cup shape probably allow concentration of the nectar until harvested by ants, in contrast to the flat hems of pre-reproductive plant that cannot contain large nectar drops. To sustain the double-layered secretory epidermis and the abundant nectar production at this stage, vascular elements extend up through the gland's parenchyma and are visible until below the epidermis. Overall, all these structural elements promote the secretion and concentration of larger nectar drops in reproductive plants, probably explaining the observed patterns in ant patrolling in the field, which was 1·4-fold greater at this stage than in pre-reproductive plants and absent from juvenile plants. What remains to be investigated are the internal or external cues to which plants respond to start producing fully functional EFNs and determine the precise ontogenetic stage in which this occurs. The lack of biotic defences early in T. velutina plants suggests that plants at this stage should rely on other defensive mechanisms to deal with herbivores, as apparently they receive more damage than older stages in the field (N. Villamil, pers. observ.) We suggest that the anatomical transition from non-functional to fully functional EFNs we have described throughout the ontogeny of T. velutina could correspond to the transition in which plants should be switching from one defensive strategy to another, as has been found for other systems (Boege et al., 2007). This, however, warrants further investigation.

The increase in EFN abundance as well as the morpho-histological changes and nectar production throughout ontogeny follow the pattern predicted by the growth-differentiation balance hypothesis (Herms and Mattson, 1992). This hypothesis assumes that after fulfilling the growth priorities, the surplus of carbon fixed through photosynthesis should be allocated to tissue differentiation, maturation and specialization (Stamp, 2003). As plants age, they increase in size and therefore have larger nutrient-acquiring and photosynthate-producing organs, such as roots and leaves (Boege and Marquis, 2005). Accordingly, a greater investment in defence via specialized tissues is expected in plant stages with an excess of fixed carbon, which coincides with the greater production of structurally more complex EFNs in reproductive plants than in earlier stages of T. velutina. If the priority of young plants is to grow until they start accumulating excess fixed carbon, tolerance is a more likely defensive strategy. However, this remains to be investigated.

In the last decade, it has become clear that defences change throughout plant ontogeny with patterns depending on plant life history, defence type and herbivore identity (Boege and Marquis, 2005; Elger et al., 2009; Barton and Koricheva, 2010). In the particular case of EFNs there are too few studies to elucidate the general ontogenetic trajectories of this indirect defence (Barton and Koricheva, 2010), but most authors report a positive relationship between EFN production and plant development (Veena et al., 1989; Young et al., 1997; Falcao et al., 2003; Miller, 2007). Understanding the physiological, structural and histological constraints of the production of both direct and indirect defences throughout ontogeny may contribute to our general knowledge of the evolution of plant defence in the context of multitrophic interactions, but also can help to detect changes in the agents and targets of natural selections as plants develop. In the particular case of seedlings or young plants, for example, even when mutualistic ants could act as selective agents favouring the production of rewards in the presence of herbivores, anatomical and/or resource-related constraints seem to restrict such possibility. The detection of these constraints, in turn, represents an excellent opportunity to investigate the evolutionary alternatives plants have evolved to deal with the negative impacts of herbivores early during their development, given the challenge that establishment, growth and defence represent.