Introduction

Extrafloral nectaries (EFNs) are glands that secrete sugars and are located in the vegetative organs of plants (Caspary 1848; Koptur 1992). EFNs attract various insects (Cuautle and Rico-Gray 2003; Oliveira and Freitas 2004), especially ants (Oliveira 1997), which defend the plant against herbivores in return for receiving food, i.e., the nectar (Costa et al. 1992; Katayama and Suzuki 2003). These secretory structures are much common, since around 108 families of flowering plants produce EFNs (Weber and Keeler 2013). Plants bearing EFNs account for 30 % of the species in some environments (Rudgers and Gardener 2004) and are more common in tropical than in temperate communities (Koptur 1992).

In Brazil, there is a very rich species biome regionally called “Cerrado” (Gottsberger and Silberbauer-Gottsberger 2006). “Cerrado” is a neotropical savanna characterized by strong rainfall seasonality, coupled with constantly high diurnal air temperatures, possessing deep and well-drained soils (Furley 1999). The nutrient-poor and acid soils limit plant growth in this region (Haridasan 2000). Furthermore, the presence of fire (Miranda et al. 2004) and herbivory (Silva and Batalha 2011) are common environmental aspects which can determine some adaptive traits in native plants.

In such environmental conditions, more than 25 % of plant species are reported to have EFNs (Oliveira and Leitão 1987; Machado et al. 2008). Experimental studies in “Cerrado” areas showed that ants, which are attracted by extrafloral nectar, defend these plants, demonstrating an anti-herbivory function (Costa et al. 1992; Del-Claro et al. 1996; Oliveira 1997). This interaction can show temporal and spatial variation (Rico-Gray and Oliveira 2007; Marazzi et al. 2013) depending on ontogenical stage (Heil et al. 2000). Also, environmental factors such as the water availability (Delgado et al. 2011a), given the system seasonality (Tilman 1978), presence of fire (Petanidou et al. 1999), abundance and type of ants (Heil et al. 2000), and presence of herbivores (Tilman 1978; Rios et al. 2008) are supposed to affect ant–plant interactions.

EFNs can also show variation in some of their traits, as in their abundance and in the amount of nectar they produce, depending on their ontogenical stage and environmental condition (Heil et al. 2000). For example, EFNs can be functional only on young leaves (Paiva 2009) or during a short period of time (Delgado et al. 2011a). EFNs can also be very plastic due to very distinct environmental conditions and pressures, such as the presence of disturbance factors, as fire (Morais 2007), the insect herbivore pressure (Ness 2003), and the availability of resources for plant growth and reproduction (Mondor et al. 2006).

The abundance of insect herbivore community varies due to presence of fire, which is able to change herbivory pressure, influencing the abundance of insect populations in different ways (Janzen 1973; Diniz et al. 2011; Lepesqueur et al. 2012). The sporadic fire in “Cerrado” areas induces the leaf renovation in a series of shrubs and trees species (Castellani and Stubblebine 1993; Medeiros and Miranda 2005), a fact that attracts a variety of herbivores species (Lepesqueur et al. 2012), given that young leaves possess tender tissues with high nitrogen and low fibers content (Coley 1983). On the other hand, frequent fire, such as fires that occur every 2 years for a decade or with more frequency, reduces the abundance of shrubs and trees, altering the vegetation physiognomy (Dias and Miranda 2010). This leads to change in the insect herbivorous community, decreasing the abundance of herbivores, such as caterpillars (Diniz et al. 2011), due to the direct mortality or forced emigration (Marini Filho 2000).

The abundance of herbivores can also vary according to seasonal aspects, as the most insect damage occurs at the transition of dry season to rainy season (i.e., the end of dry season) because of the greatest amount of new produced leaves in “Cerrado” (Marquis et al. 2001). Besides, the rainfall seasonality in “Cerrado” can also have influence in the community of insects, since the rain benefits the proliferation of fungi which are a natural enemy of insects (Silva et al. 2011). Finally, the vegetation community is also influenced by the higher water availability (Furley 1999). Therefore, the study of different environmental conditions that may be modulating different characteristics of EFNs can assist us to understand changes of interactions mediated by these traits (Marazzi et al. 2013).

In this study, we analyze the effect of ontogenical leaf stage, presence of fire, and seasonal water availability on EFNs’ traits. As EFNs are considered facultative defenses of plants (Heil and McKey 2003), being dependent on herbivory pressure (Mondor and Addicot 2003; Morais 2007), it is expected that plants have more EFNs or show functional EFNs when the probability of herbivory is high. Thus, we believe that plants which grow in areas with frequent fires may have lower quantity of EFNs than plants growing in controlled areas (free from fire). Furthermore, if EFNs are modulated by herbivory pressure, we expect to find more EFNs in leaves that flushed during the dry season than in leaves flushed during the rainy season, since those leaves have higher probability to be damaged by herbivorous insects. On the other hand, if EFNs are not influenced by herbivory damage, we suppose that there is no difference between plants grown in area with frequent fire to control areas. In this way, only the difference in water available may influence extrafloral nectary production, finding more EFNs in leaves flushed during the rainy season since there is more resource available to plants develop and grow.

In this study, we described structural and ontogenical aspects of EFNs in leaves of Maprounea brasiliensis A. St.-Hill (Euphorbiaceae), a perennial, evergreen, and small shrub 0.5–2 m tall, commonly found in Cerrado (Esser 1999), and analyzed some of EFNs traits under different environmental conditions. First, we asked how the EFNs are formed and morphologically arranged in leaves of M. brasiliensis. Second, we tested if EFNs traits are modulated by environmental conditions and disturbances, such as fire and rainfall seasonality. For these reasons, anatomical and ultrastructure analyses as well as field experiments were conducted.

Materials and methods

Study area and plant sampling

This study was conducted at Reserva Ecológica do IBGE (RECOR) (15°56′41″S and 47°53′07″W) in Distrito Federal, Brazil. The dominant physiognomy of Brazilian savanna is the cerrado sensu stricto, a regionally type of savanna vegetation consisting of a dominant herbaceous stratum with grasses and a discontinuous woody stratum (Eiten 1972). The climate has a distinct dry season from May to September and a rainy season from October to early May. The soils are deep, well drained, acidic, and dystrophic (Franco et al. 2005). As an object of study, we have chosen Maprounea brasiliensis A. St.-Hill (Euphorbiaceae), a small and evergreen shrub that possesses EFNs in their leaves (Senna 1984). In this study, we sampled leaves from distinct developmental stages, including young leaves and senescent leaves. Specific methodologies were described below according to the type of study. A voucher specimen is deposited in the Herbarium of Reserva Ecológica do IBGE: 41883.

EFN morphology and nectar composition

To describe the morphology of EFNs and how they are formed, we performed anatomical and ultrastructural analysis. We also quantified some extrafloral nectary traits such as the quantity, location, and size. We also made chemical analysis to obtain the carbohydrate composition of nectar. Ants found on EFNs from different M. brasiliensis specimens were collected, fixed, and identified.

The number of EFNs on the base and on the apex of the leaf was determined in leaves (one leaf per individual) collected from 30 distinct plants with similar size distributed in a natural cerrado environment in June of 2009. These EFNs were also classified into small (0.1–1.0 mm diameter) and large (1.1–2.5 mm diameter), with the diameter being measured with a digital caliper. We also determined the leaf area of those leaves in order to know if the leaf area may influence the EFNs’ traits. The leaf area was obtained by linear regression whose line equation [foliar area = (8.35 × leaf width) − 16.76; R 2 = 0.806] was determined by the area of those 30 leaves drawn on squared paper.

In order to collect nectar, branches of M. brasiliensis were bagged with plastic for 96 h to exclude ant’s visits. The nectar was obtained from seven plants, using four EFNs at the base and the apex of leaf per plant, in June of 2009. Nectar was collected only from young leaves, because the EFNs of senescent leaves did not produce more nectar as observed at the field. The nectar was collected with a 0.5 ml syringe, diluted in 500 μl of distillated water, and stored in Eppendorff tubes in liquid nitrogen in the field. In the lab, the samples were kept at −80 °C until analyzed. The composition of sugars was determined by high-performance anion-exchange chromatography with integrated pulsed amperometric detection (HPAEC-IPAD model ICS3000) on a CarboPac™ PA10 (2 × 250 mm) column (Dionex Corp, Sunnyvale, CA, USA) using an isocratic eluent with 200 mM NaOH in water for 30 min. Sugars (glucose, fructose, and sucrose) and total nectar sugar concentrations were calculated with Chromeleon Software (Thermo Scientific, Chromatography Data System) from a standard curve of glucose, fructose, and sucrose in the following concentrations: 0.625, 1.25, 2.5, 5.0, and 10.0 μM.

The study of the EFN morphology was done with plants that were growing in a natural cerrado environment. For light microscopy, leaves from different developmental stages, i.e., leaves of first to third node and adult and senescent leaves, were fixed in FAA (5 % acetic acid, 5 % formaldehyde, 90 % ethanol 70 %), stored in 70 % ethanol (Johansen 1940), dehydrated in alcohol, and infiltrated with historesin (Methyl Methacrylate Embedding Kit). The transversal, longitudinal, and paradermic sections were cut with a microtome (5 μm thick), mounted on slides, and stained with Toluidine Blue (O’Brien et al. 1964). For histochemical tests, hand cut sections were treated with the acid phloroglucin for the detection of lignin and Lugol for starch (Johansen 1940). For scanning electron microscopy studies, completely expanded leaves with EFNs were fixed in 2.5 % glutaraldehyde in 0.05 M phosphate buffer at pH 7.2 for 24 h, dehydrated in a graded acetone series (30–100 %), dried in critical point CO2, covered with a layer of gold (Bozzola and Russel 1992), and observed on a Jeol® 840A scanning electron microscope at an accelerating voltage of 5.0 kV. For transmission electron microscopy studies, fragments of recently expanded leaves with EFNs were fixed in Karnovsky solution (Karnovsky 1965) for 24 h, post-fixed in 1 % osmium tetroxide in 0.1 M phosphate buffer pH 7.2 and processed using standard methods (Roland 1978). Ultrathin sections were post-stained with uranyl acetate and lead citrate and observed on a Jeol® 1011 transmission electron microscope at 80 kV.

Variation in EFNs traits given frequent fire and water seasonality

In order to verify the existence of variation in EFN production and EFN size due to the effect of common “Cerrado” environmental variables, we performed two experiments to test the effect of (1) fire frequency and (2) seasonality. The experiments were located in a continuous area with cerrado sensu strict. Therefore, there was no variation among sites.

To assess the role of fire in EFNs traits, we selected 15 adult individuals with similar size in an area where fire was absent for more than 35 years and 15 adult individuals surviving in an area with frequent fire whose area was 10,000 m2, which suffered experimental burn from 1991 to 2008 during the middle of drought (at the beginning of August) each every 2 years. The distance of experimental to control area was more than 400 m (Dias and Miranda 2010). In each case, we collected three completely expanded leaves from each selected individual in May 2009 (the end of the rainy season). In this experiment, we counted the number of EFNs and measured their larger diameter. These traits were chosen because the EFNs’ number and size can be used as an indicator of potential food source to animals which deter herbivorous in return of nectar.

To assess the effect of rainfall seasonality in EFNs traits, we marked 20 healthy adult individuals with similar size that were growing in a cerrado area without any fire management. In the beginning of August 2009, in each of those plants, we marked two sun exposed branches above the site where the leaves were located. We counted the number of leaves in each branch. Month by month, we went to the field in order to monitor the development of new leaves. In December 2009, the new leaves above the mark were completely expanded. Then, the leaves located below the marked branch were considered as leaves produced during the previous rainy season and the leaves above the marked branch were considered as leaves produced during the dry season. In each case, we collected completely expanded leaves from each selected branch in each individual. In this experiment, we counted the number of EFNs and measured their larger diameter.

Statistical analysis

For all data, normality was tested by Kolmogorov–Smirnov test and the homogeneity of variances by Levene test. Data were transformed when non-normality was found. Student t test was undertaken in order to compare the number of EFNs at the base and at the apex of the leaves of plants in a cerrado area, the number of EFNs per leaf (EFN nº) and the average of the largest diameter of EFN per leaf (DIAM) among plants of M. brasiliensis in the experimental area of frequent fire and in a control area. Student t-test for dependent samples was done in order to compare the number of EFNs per leaf (EFN nº) and the average of the largest diameter of EFN per leaf (DIAM) from leaves flushed during the dry period and the rainy season. Mann–Whitney was carried out to compare the quantity of large and small EFNs at the base and at the apex of the leaves of plants in a cerrado area. Furthermore, the correlation (Spearman correlation) between the number of EFNs and leaf area was determined. Statistical analyses were carried out using Statistica 7.0, and significance was taken at P < 0.05.

Results

EFN morphology and nectar composition

EFNs are conspicuous, flattened, and either circular or elliptical. They appeared on the abaxial leaf surface (Figs. 1, 2), occurring at the base and at the apex of the leaf. In terms of number, they were more common at the base of the leaf (2.854 ± 1.068) instead of the apex (1.417 ± 0.907) (t = 4.105; P < 0.01). There are more large EFNs (2.063 ± 1.237) than small EFNs (0.792 ± 0.942) at the base of the leaf (Z = 2.933; P < 0.01). On the other hand, there are more small EFNs (1.396 ± 0.913) at the apex of the leaf (Z = −4.917; P < 0.01). There is no correlation between number of EFNs and leaf area (Spearman = −0.054; P > 0.05).

Figs. 1–8
figure 1

EFN of Maprounea brasiliensis. (1) EFNs without nectar secretion at the base of the leaf. (2) EFN with a nectar droplet (head of arrow). (3) Scanning electron microscope image of an EFN. (4) An EFN with one layer of secretory cells. (5) Detail of the first anticlinal division (arrow). (6) A completely expanded leaf with two mature EFNs. (7) The subglandular tissue. (8) Cell walls of the subglandular tissue impregnated with lignin as shown by a positive (red) stain with floroglucinol. PC parenchyma cells, Ph phloem, SC secretory cells, ST subglandular tissue, V vascularization. Scale bars (3) 10 μm. (4) 100 μm. (5) 40 μm. (6) 20 μm. (7) 10 μm. (8) 5 μm

Full size image

EFNs begin secretory activity in young leaves and lose functionality in senescent leaves. The nectar volume decreases as the leaf matures, making it impossible to collect nectar drops on senescent leaves. Ants, such as Cephalotes sp., visit only EFNs of young leaves, patrolling branches with immature leaves.

The nectar carbohydrate composition remained essentially the same between EFNs at the base and at the apex of the leaf. The nectar is composed by hexoses (fructose and glucose) and sucrose, with the average concentration of glucose being 24.310 ± 33.366, fructose 23.160 ± 33.577, and sucrose 12.750 ± 21.937 μM.

Stomata and trichomes are not observed in the EFN secretory surface (Fig. 3). The secretory tissue originates from the epidermis, and the first cells are elongated anticlinally with a conspicuous nucleus (Figs. 4, 5). These cells divide anticlinally at first and then periclinally (Fig. 5) to produce a cell group toward the abaxial mesophyll. Below the initial cells of the EFN, we observed several layers of parenchyma cells with primary walls (Fig. 5). These parenchyma cells, which divided into various planes, are localized between the secretory cells and the vascular tissue (Fig. 5), which is denser and more ramified in the EFN region than in the other parts of the leaf (Figs. 6, 7). After many anticlinal and periclinal divisions, the secretory tissue is completely formed (Fig. 6). During these divisions, the parenchyma cells, which are below the secretory cells, start expanding and impregnating lignin into their walls (Fig. 8). At the end of EFN ontogenesis, the secretory cells which have no starch are surrounded by three or more layers of cells with lignified walls.

Mature EFNs are embedded in the leaf mesophyll and isolated from the vascular tissue by several layers of cells, which show thick walls impregnated with lignin (Figs. 6–8). Although vascular tissue is denser in the leaf portion where the EFN is located, neither the phloem nor the xylem reaches it (Fig. 7). The secretory tissue exhibits radially elongated cells with dense cytoplasm and a conspicuous nucleus. The nectar appears to be accumulated inside subcuticular spaces (Fig. 3) between the secretory cells and the cuticle.

The ultrastructure of EFNs was investigated in recently expanded leaves. Microscopy shows that EFN cells have a cytoplasm with a large amount of mitochondria (Figs. 11, 12, 18, 19, 21), smooth endoplasmic reticulum (Figs. 16, 19), Golgi apparatus (Fig. 12), and free ribosomes. Some secretory cells contain vacuoles with secretion or many multivesicular bodies (Fig. 12), and the large nuclei contain uncondensed chromatin (Fig. 9). The connection between the secretory cells was maintained by plasmodesmata (Figs. 10, 11). Furthermore, granulocrine secretion might be inferred due to the vesicles observed near the plasmallema (Fig. 13). Below the EFN, living cells with thick walls were identified (Figs. 14–16, 19–21), which contained granular cytoplasm with many free ribosomes (Figs. 16–20), rough endoplasmic reticulum (Figs. 16, 18, 19, 21), large vacuoles (Figs. 17, 20), mitochondria with dense stroma and developed cristae (Figs. 18, 19), and irregular nuclei (Figs. 14–17). These cells, which were between the secretory cells and the phloem, formed the subglandular tissue (Fig. 14). They had symplastic connections between them, such as plasmodesmata (Figs. 16, 18), which were sometimes localized inside primary pit-fields (Fig. 19). We did not find plasmodesmata between the thick-walled cells, which form the barrier layer, and the secretory cells (Figs. 14, 15). Some sieve-tube elements (Fig. 21) were observed distributed among the cells of the subglandular tissue, which were more concentrated in layers near the vascular bundle.

Figs. 9–13
figure 2

Transmission electron microscope images of EFN from recently expanded Maprounea brasiliensis leaves. (9) Elongated cells that form the EFN. (10) Detail of plasmodesmata and vesicles. (11) Detail of two secretory cells. (12) Secretory cell. (13) The formation of vesicles at the apex of the secretory cell. GA Golgi apparatus, M mitochondria, N nucleus, n nucleolus, Pl plasmodesmata, V vacuole, ve vesicles, W cell wall. Scale bars (9) 5 μm. (10) 0.2 μm. (11–13) 0.5 μm

Full size image
Figs. 14–21
figure 3

Transmission electron microscope images of cells of the subglandular tissue and the phloem in Maprounea brasiliensis leaves. (14) Detail of the interface region between the secretory cells and the cells of the subglandular tissue. (15) Detail of secretory cell and subglandular tissue (16) Detail of the plasmodesmata and endoplasmic reticulum. (17) Cells with vacuoles, nuclei, and nucleoli. (18) Cells with mitochondria, plasmodesmata, and endoplasmic reticulum. (19) Detail of a punctuation. (20) General view of the cells that constitute the subglandular tissue. (21) Sieve-tube element. CW cell with a thick wall, ER endoplasmatic reticulum, M mitochondria, N nucleus, n nucleolus, Pl plasmodesmata, P punctuation, R ribosomes, SC secretory cell, ST subglandular tissue, STE Sieve-tube element, V vacuole. Scale bars (14) 5 μm. (15) 2.5 μm. (16) 0.2 μm. (17, 20–21) 2 μm. (18–19) 1 μm

Full size image

Variation in EFNs traits given frequent fire and water seasonality

We did not verify an effect of frequent fire on EFN production. The average number of EFNs per leaf found in the control area was 3.02, which was similar to the value (3.55 EFNs per leaf) found at the frequent fire area (t = −0.772; P = 0.446). The average of the largest EFN diameter found in the control area was 4.82 mm and in areas with frequent fire was 5.17 mm, which was not significantly distinct according to Student t test (t = −0.358; P = 0.723).

We verified an effect of water seasonality on EFN traits when we compared leaves flushed during the dry season to leaves flushed during the rainy season. Leaves flushed during the rainy period showed more EFNs per leaf (Fig. 22) (t = 2,611; P < 0.05) and higher values in the largest diameter of these structures (Fig. 23) (t = 3,397; P < 0.05) than leaves flushed in the dry season.

Figs. 22–23
figure 4

EFN and leaf traits of Maprounea brasiliensis produced at different times of the year: leaves flushed during the rainy season and leaves flushed during the dry season. (22) The number of EFNs per leaf. (23) The average of the largest diameter of EFN per leaf. Different letters indicate statistically significant differences, with P < 0.05. Bars represent the standard deviation

Full size image

Discussion

In this study, we showed how the EFN is formed in leaves of M. brasiliensis and described its anatomical and morphological arrangement, as well as, we provided its nectar sugar composition. We showed that EFN traits are modulated by rainfall seasonality, but not by frequent fire.

EFNs of M. brasiliensis are predominant at the leaf abaxial surface, near the leaf base; however, some of them are observed on the leaf apex. The presence of EFNs at leaf base and apex may suggest an adaptive value in terms of defense herbivory, as this arrangement suppose that ants need to patrol an entire leaf in order to obtain a certain amount of nectar (Delgado et al. 2011a; b; Morellato and Oliveira 1994). Nevertheless, the bigger size of nectaries at the leaf base favors the ants’ presence at this region. This phenomenon might be considered advantageous for the plant because from the adaptive perspective, it is better to keep the base of the leaf healthier than its apex because the basal portion is responsible for fixing the leaf to the stem.

The carbohydrate composition of extrafloral nectar revealed the presence of hexoses (fructose and glucose) and sucrose. Glucose, fructose, and sucrose are commonly found in the nectar of various plant species (Paiva 2009). Additionally, it is known that the chemical composition of nectar affects the identity of the attending ants (Baker and Baker 1982; Heil et al. 2005) since sucrose is highly attractive to many different ant species, including the nonsymbiotic ones (Heil et al. 2005). For example, in M. brasiliensis, we collected individuals of Cephalotes, which are known to have a varied diet (de Andrade and Baroni Urbani 1999). Although Cephalotes are not a mutualistic species, their presence on the plant can inhibit potential herbivory attack since many herbivores are visually oriented (Del-Claro et al. 1996; Oliveira 1997).

EFNs from M. brasiliensis consist of layers of small secretory columnar cells in palisade arrangements (Elias and Gelband 1976; McDade and Turner 1997). The pre-nectar is derived from phloem sap located below some of the parenchyma cells found between the secretory cells and the vascular tissue, since neither chloroplasts nor starch was observed in the secretory cells or in their surrounding parenchymatic cells. Therefore, as the pre-nectar must be transported through the parenchyma cells until the secretory cells, this parenchyma could be functioning as a subglandular tissue (Wergin et al. 1975). In completely expanded leaves, the walls of the subglandular tissue become totally impregnated with lignin. Because of this, we believe that the process of pre-nectar transport may cease once the subglandular tissue makes an apoplastic and symplastic barrier. The apoplastic barrier may appear due to deposition of lignin, suberin, or cutin on the walls of cells under the secretory tissue (Lüttge 1971; McDade and Turner 1997; Paiva and Machado 2006; present work). On the other hand, the symplastic barrier was evidenced by the absence of plasmodesmata between secretory cells and cells of the subglandular tissue in the species studied. Therefore, the nectar must be synthesized from the sugars stored in vacuoles of secretory cells until its total consumption when the pre-nectar flow ceases.

The ultrastructure of the EFNs elucidated the final phase of nectar secretion in the completely expanded leaves. We found mitochondria, segments of endoplasmic reticulum, Golgi apparatus, ribosomes, and conspicuous nuclei with uncondensed chromatin which are reported to appear in active secretory cells (Delgado et al. 2011a; Fahn 1988). ATP requirements for active transport of nectar exudation (Vassilyev 2010) and the synthesis of the nectar itself (Delgado et al. 2011a) can explain the higher number of mitochondria observed in the secretory cells. The presence of uncondensed chromatin and free ribosomes may prove that proteins are synthesized inside the nectary cells (Alberts et al. 2002). The pre-nectar transport among the secretory cells is made by a symplastic mechanism, such as plasmodesmata (Fahn 1979a, b) by diffusion (Vassilyev 2010). Furthermore, the presence of endoplasmatic reticulum, Golgi apparatus, and vesicles near the plasmallema may be considered as a feature of granulocrine secretion (Fahn 1979b; Durkee 1982; O’Brien et al. 1996).

In M. brasiliensis, EFNs are functional from young to recently expanded leaves. It was observed that the loss of functionality in the EFNs is correlated to ant behavior since these insects were found patrolling only young or recently expanded leaves with active EFNs (personal communication). Other studies also showed that the functionality of EFN can be restricted to young leaves (Paiva 2009) or during a short time of the leaf life according to Delgado et al. (2011a). The presence of ants suggests a mechanism of biotic defense in M. brasiliensis because, in return for the ants’ protection, plants offer them food rewards such as extrafloral nectar (Janzen 1966; Heil et al. 2005). Moreover, a concentration of ants on young leaves is advantageous for the plant because both unexpanded and recently expanded leaves suffer from more herbivore attack than mature leaves (Heil and McKey 2003) and these are leaves with greater life expectancy (photosynthesis) for the plant.

If EFNs of M. brasiliensis were damage-dependent, we expected to find low production of EFNs in plants subjected to frequent fires in comparison with control plants because the frequence fire decreases the abundance of herbivores (Diniz et al. 2011). Besides, as some plants have induced defense only when they are in contact with the herbivorous saliva (Ness 2003), the production of EFN would not be stimulated in plants with low herbivore pressure. However, we did not identified difference in the EFN production when compared plants submitted to frequent fire and control plants. Our result demonstrate that the herbivory damage is unnecessary to raise the EFN production in M. brasiliensis or the herbivory damage is not enough to trigger a defense response. Probably, the availability of additional resource (in this case, water) is also important to this species to raise EFN production. As pointed out by Mondor et al. (2006), the production of conspicuous nectaries can be simultaneously damage and resource dependent.

On the other hand, our results demonstrated that EFN traits are modulated by environmental conditions, such as rainfall seasonality, because leaves flushed during the rainy period showed more EFNs per leaf and the largest diameter of these structures than leaves flushed during the dry season. However, we did not corroborate the hypothesis that production of EFNs might be higher in the leaves flushed during the dry season than in the rainy season since the insect herbivory rate is normally higher at the end of the dry season (Marquis et al. 2001). Our result can be explained because mature leaves are able to detect and transmit changes from the external environment to new leaves of the same plant and induce an appropriate developmental response in new leaves (Lake et al. 2001). Thus, in this perspective, the leaves flushed during dry season and that suffered higher insect damage may transmit this sign to leaves that were expanding during wet season. Although, the experiment of frequent fire showed that herbivory pressure is not enough to raise EFN production. Then, another explanation to the raise of EFN production on leaves flushed during the rainy season is the high availability of water that stimulates the plant to develop. Therefore, M. brasiliensis may be evolutionary programed to produce more EFN when the availability of water is higher. In that case, further experiments with different levels of water addition are necessary to be done in order to measure the EFN production to analyze the real influence of water on EFN production.

In this study, we demonstrated that the structure and function of EFNs of M. brasiliensis are determined by the developmental stage of the leaf, which results in the termination of phloem supply to the EFN cells. The ecological consequence due to anatomical and ultrastructural changes is the end of plant–ant mutualistic interaction via EFN. Furthermore, we demonstrated that the production of EFNs in M. brasiliensis is not responsive to fire, which showed that herbivory damage is not necessary to raise the EFN production. The production of nectaries is responsive to the effect of season, being higher in leaves that flushed during the rainy season, the period of the highest availability of water. Therefore, the defenses in M. brasiliensis vary throughout the year and at different stages of leaf development, being resource dependent and probably evolutionary programed to be higher during the rainy season.