Introduction

Seed predation is an important process that has a direct influence on plant fitness (Janzen 1978; Crawley 2000) and affects the spatial distribution and abundance of plant populations (Meiners and Stiles 1997; Whitney and Stanton 2004). Because seed predation is often selective on a subset of the species available within a community (Meiners and Stiles 1997; Raju et al. 2009), some characteristics such as seed size (i.e., energy content), nutritional quality, seed phenology, local abundance, and defensive compounds have been evaluated to explain this selectivity (Andresen and Levey 2002; Celis-Diez et al. 2004; Azcátare et al. 2006; Yi and Zhang 2007; Mari et al. 2008). Particularly, the role of chemical compounds against seed predators has been intensively studied (Janzen 1969; Center and Johnson 1974; Rosenthal et al. 1978; Janzen et al. 1986; Rogers et al. 1987; Diaz 1996; Guimarães et al. 2003; Nakagawa and Nakashizuka 2004; Muhlemann et al. 2006; Gallet et al. 2007; Bonal and Muñoz 2008), while other kinds of defensive mechanisms have received less attention (Albrectsen 2000; Albrectsen et al. 2008). Even the role of glandular trichomes as a plant defense has been more often considered as a chemical defense than as a mechanical defense (van Dam and Hare 1998b; Agrawal and Karban 2000; Forkner and Hare 2000; González et al. 2008). Regardless of their mechanism of action, it is well accepted that trichomes may act as herbivore deterrents (Fernandes 1994; Gange 1995; van Dam and Hare 1998a, b; Valverde et al. 2001), but also, in many cases, impair the action of herbivore parasitoids or predators, actually, functioning as a herbivore protection provided by the host plant (Gruenhagen and Perring 1999, 2001; Vinson 1999; Lovinger et al. 2000).

The genus Chamaecrista [Breyne] Moench (Leguminosae) has c. 260 species (Irwin and Barneby 1982), most from the Americas. Chamaecrista seeds are preyed upon in Serra do Cipó (SE Brazil) by endophagous (Coleoptera: Bruchidae) and ectophagous insects that live inside the fruit, but eat the seed from outside (Coleoptera: Curculionidae; Lepidoptera). Bruchids usually have a narrower host range (Center and Johnson 1974; Janzen 1980; Jermy and Szentesi 2003), especially those that oviposit on fruits still attached to the host plant, such as in Chamaecrista. However, seed predation by bruchids can be extremely important because it affects plant population size (Schmale et al. 2002a, b). For example, some studies show that parasitoid wasps reduced bruchid seed predator populations to 20 % (Schmale et al. 2002a, b; Martins 2013). Therefore, mortality rates imposed on bruchids by their natural enemies such as predators and parasitoids may reduce infestation levels and consequently plant damage (Schmale et al. 2003, 2006). In contrast, ectophagous insects usually have a broader geographic distribution, because their interaction with host seeds is not specialized. The larvae of Chamaecrista bruchids are heavily attacked at the study site by parasitoid wasps.

In this study, we determined the influence of some physical and phenological plant traits on seed predation rates of 13 sympatric taxa of Chamaecrista (i.e., 10 species and three varieties of one species), in order to evaluate their potential defensive role and their consequences on parasitism rates suffered by bruchid seed predators. Parasitism rates suffered by bruchids were also measured to evaluate the role of their natural enemies. Plant traits assessed were seed size, fruit pubescence, and seed production during the reproductive stage of all species of Chamaecrista. All plant populations studied are phylogenetically closely related and occur in sympatry. Therefore, all plant species are subjected to similar abiotic conditions. We addressed the following questions: (1) How does seed predation relate to availability (time and abundance) of seeds?; (2) Is seed predation by ectophagous and endophagous insects influenced by fruit pubescence and seed size?; (3) Is parasitism rate of bruchids affected by fruit traits of host plant species?; and finally (4) How is bruchid host range influenced by plant phenology, fruit pubescence, geographic distribution and natural enemies?

Methods

Study site and system

The study was carried out within an area of approximately 28 Km2 of rupestrian grassland vegetation at Serra do Cipó (SE Brazil, 19°12′ and 19°34′S; 43°27′ and 43°38′W), where all the studied taxa naturally co-occur. The area is dominated by grasslands with a few tortuous and sclerophyllous shrubs that grow on nutrient-poor, acid soils. Part of the study area was inside the Serra do Cipó National Park and Reserva Particular Vellozia. The climate of this region is strongly seasonal, with wet summers and dry winters, but a great variation in phenological patterns has been found among the studied taxa (see Madeira and Fernandes 1999).

The 13 studied Chamaecrista taxa belong to three of the six existing taxonomic sections of the genus and vary in terms of geographic distribution from cosmopolite to restricted endemic plant species (Irwin and Barneby 1982). The 13 taxa of Chamaecrista and some of their most important characteristics are listed in Table 1.

Table 1 Chamaecrista taxa studied and their principal characteristics
Full size table

We considered three varieties of C. desvauxii as separate entities, since they show conspicuous differences in phenology, seed packing, seed size, fruit pubescence, microhabitat preference, and plant architecture (Madeira and Fernandes 1999). C. ochnacea is dimorphic in fruit pubescence. We randomly selected 350 plants of all plant species, of which 8.3 % presented simple and glandular trichomes in their fruits and 91.7 % had glabrous fruits. Plants with pubescent and glabrous fruits were considered separately in the analyses.

Sampling procedures

To determine the phenological patterns of all plant species, 20 individuals of each taxon were randomly chosen and labelled. All fruits on each individual plant were counted monthly throughout the year. Because C. choriophylla and C. cathartica are extremely rare, we collected fruits of only 10 individuals of each of these two species.

We calculated a “fruiting dispersion index” (Begon, et al. 1986) that considers richness and abundance of species, using the mean numbers of fruits produced per individual per month, considering a 12-month period. We considered the fruiting period duration as a measure of “window of opportunity” (Hunter 1993) that, collectively with the mean abundance of fruits produced per individual, may be considered as a “diversity index of predation opportunities” (DIPO). In addition, the Simpson diversity index and the equitability index associated with the diversity index were used as “fruiting dispersion index” (FDI). These measures identify the better defended taxa against seed predators with a smaller FDI, implying less diversity of opportunities to seed predators (Madeira and Fernandes 1999).

Fruit pubescence

To determine differences in fruit pubescence, we collected three fruits of 10 individuals of each taxon, and the trichome density per mm2 and the lengths of simple and glandular trichomes were measured. Fruits were collected unripe but totally developed, to allow easy cutting and actual measurements of final trichome density. To obtain density estimations, fruits were transversally half-cut. On each half, four transverse cuts including the two valves were made. The pieces were placed in sodium hypochlorite to clarify the tissue and dissociate the epidermis (Kraus and Arduin 1997). Slides for light microscopy were prepared with four dissociated epidermal pieces and four transversal cuts from each individual plant. We counted the number of trichomes in 1 mm2 on five previously determined standard positions to obtain the mean trichome density of each individual. In addition, on each transverse cut, we measured the first 10 trichomes observed in a fixed focus plane, from a standard start position. A magnification of 100× was used for trichome counts and measurements.

Seed size, seed predation, and parasitism rates

To estimate seed size and seed predation, we collected ripe fruits each month with a limit of 25 individuals per taxa per survey. Sample sizes within and among taxa were quite variable over the year. Collected fruits were kept in the laboratory under controlled environmental conditions (±25 °C, 70 % relative humidity) for insect emergence. Mean individual seed weight was used to estimate seed size. All intact seeds (without signs of predation, fungus, etc.) from all collected fruits were weighted in groups composed of all seeds from each individual plant. Each seed set weight was divided by the number of seeds to obtain individual mean seed weight. The mean seed weight for each taxon was obtained by the mean of the individual plant means.

From all collected ripe fruits, we obtained the number of seeds damaged by herbivores, type of predator (i.e., endophagous or ectophagous), number of predators, and number of parasitized predators. We calculated the seed predation rate per fruit, individual, and taxa. The same procedure was used for parasitism rates. Parasitism rates were assessed as an estimator of the role of natural enemies in the system and to establish its relationship with seed predator host range and plant traits.

Data analysis

To determine the effects of plant traits on seed predation rates by endophagous and ectophagous herbivores, we used stepwise multiple linear regressions (F to enter = 2.0). In these analyses, taxa means were used as sample units, since it would be impossible to obtain all desired data on the same individual plants used to measure seed predation rates. Analyses were performed on individual taxa to determine the predation levels by bruchids and ectophagous herbivores, and in each case, data were transformed into square-root arcsin to obtain a normal distribution (Zar 1984).

In the plant taxa that suffered seed predation by bruchids, we applied a stepwise multiple regression between parasitism rates of bruchids and plant traits. Seed size was not included in the analysis because it did not affect parasitoid attack rate. Parasitism data were also normalized through square-root arcsin transformation. C. choriophylla was not included in any statistical analyses due to the small number of individuals with collected ripe fruits (n = 6), which did not allow a normal distribution of data.

As a further test of the hypothesis that fruit pubescence would diminish seed predation, we performed an intraspecific analysis to compare seed predation in plants with glabrous fruits and plants with pubescent fruits of C. ochnacea. Parasitism rates on bruchids attacking glabrous and pubescent fruits were also compared, including only plants attacked by bruchids.

Seed predation rates on C. cipoana and C. semaphora were also contrasted separately because fruit production of both species occur simultaneously in the dry season (Madeira and Fernandes 1999) and because they were predated by the same bruchid species (Sennius bruneus), offering an opportunity to test the preference of bruchids. Therefore, parasitism rates of S. bruneus were also compared to verify the parasitoid preferences on bruchids that feed on different plant species.

Trichome density and simple and glandular trichome length of C. cipoana and C. semaphora were also contrasted to determine the potential relationship between seed predation and parasitism rates with fruit pubescence. All pairwise comparisons were performed by t tests.

Results

Phenology

Phenologically, aseasonal species had greater fruiting dispersal indexes (FDI), especially C. dentata that produced fruits throughout the year. Most taxa showed a low FDI (Fig. 1a), indicating a restricted fruiting period. Among different sections, Absus showed higher variation in FDI while Xerocalyx was the most homogeneous section, with low fruit dispersion indexes for all taxa (Fig. 1a).

Fig. 1
figure 1

a Fruiting dispersion index for 13 Chamaecrista taxa on Serra do Cipó (Brazil) (see text for details). b Seed weight (mean ± standard error) for 13 Chamaecrista taxa on Serra do Cipó (Brazil) in 1996. gray box = Section Absus; open box = Section Chamaecrista; right to left striped box Section Xerocalyx. Cca Chamaecrista cathartica, Cci C. cipoana, Cde C. dentata, Coc C. ochnacea, Cse C. semaphora, Cch C. Choriophylla, Cmu C. mucronata, Cro C. rotundifolia, Cve C. venulosa, Cla C. desvauxii latistipula, Cma C. d. malacophylla, Cmo C. d. mollissima, Cra C. ramosa

Full size image

Fruit pubescence

Only section Absus had species with glandular trichomes on fruits, some of them in combination with simple trichomes. Most taxa from sections Chamaecrista and Xerocalyx had simple trichomes, and one taxon from each section had glabrous fruits (Table 2).

Table 2 Fruit trichome density per mm2 (simple + glandular) and simple and glandular trichome length (mm) on 13 Chamaecrista taxa
Full size table

Seed size

Mean seed weight varied between 2.61 ± 0.06 mg (C. rotundifolia) and 25.23 ± 0.244 mg (C. dentata). Species with round seeds from section Absus had higher seed size than other sections, while species of the Section Chamaecrista with flattened seed showed a higher seed size. Seeds of all taxa from section Xerocalyx were among the smallest and tended to be triangular shaped (Fig. 1b).

Insect herbivores

Adult bruchids were classified into four new species of Sennius (Silva et al. 2003). All parasitoids attacking bruchids were hymenopterans. We could not rear any adult lepidopteran, as they died easily when fruits were removed from plants, and quickly became dried and unrecognizable. Based on larval morphology, we could identify at least two morphospecies of lepidopterans.

Predation and parasitism rates

Mean seed predation rates per taxon varied from zero (C. d. malacophylla, C. choriophylla, and C. cathartica) to 45.72 % ± 3.05 (C. ochnacea) (Table 3). We found that higher seed predation rates were caused by bruchids associated with plants of a narrow distribution range. Bruchids showed higher incidence on locally distributed hosts. For example, Sennius maculates was associated only with C. ochnacea seeds and represented the highest predation rate, Sennius kingsolveri was associated with C. dentata (18.07 % ± 1.5) and Sennius bruneus preyed on seeds of C. cipoana (15.5 % ± 2.6) (Table 3). In contrast, we found that predators with more broad geographic distributions destroyed smaller proportions of their host seeds. Particularly, Sennius niger preyed on seeds of four taxa, but the greatest damage was caused on C. desvauxii mollissima (8.5 % ± 3.2). Larvae of all the bruchid species were capable of preying on more than one seed of different hosts, indicating that Sennius species may be considered as an endophagous guild. Within the ectophagous guild, we found that the highest seed predation rate was imposed by curculionid larvae on C. venulosa (8.6 % ± 1.9).

Table 3 Number of individual plants with fruits collected (Plant indiv.), number of individual plants with any seed predation (predated indiv.), total seed predation rates (mean ± 1 SE) on each Chamaecrista taxon, and percentage of predation due to each predator kind and guild
Full size table

The species from section Absus were exclusively preyed upon by Sennius. It was also the section that suffered more seed damage. In contrast, lower rates of seed predation by Sennius niger and ectophagous predators were found in taxa from section Xerocalyx. Within section Chamaecrista, C. mucronata was preyed upon by S. kingsolveri, C. venulosa by S. niger and ectophagous predators, and C. rotundifolia only by ectophagous predators (Table 3).

Parasitism rates on Sennius varied between 0.9 % ± 0.9 (Sennius bruneus on C. cipoana) and 84.2 % ± 5.2 (S. niger on C. ramosa) (Table 4). In a comparison between Sennius species host range and parasitism rates, no relationship was found between the attack from natural enemies and bruchid host ranges (Table 4). However, we found a lower rate of parasitism in the two more specialized bruchid species (Sennius maculatus and Sennius bruneus).

Table 4 Number of plants with seed preyed upon by bruchids, number of plants which had any parasitism on bruchids preying upon their seeds (individuals with any parasitism), percentage of parasitized individuals (mean ± 1 SE) on each host plant (% P), and specialization level of bruchid species, represented by the number of host plant taxa used (in brackets after taxon name) among the 13 Chamaecrista taxa and by the contribution in number of seeds of each host taxon to total seed consumption (% C) of each bruchid species
Full size table

The stepwise multiple linear regression model that best explained variation of bruchid seed predation among the different taxa of Chamaecrista included seed weight and trichome density of fruits. We found that seed weight was the only important variable influencing bruchid seed predation on Chamaecrista taxa (Table 5). Similarly, seed weight was the only variable left in the multiple model that affected ectophagous predation on the Chamaecrsita taxa, although the relationship was negative (Table 5). The regression model that best explained parasitism rates on Sennius included only glandular trichome length, revealing a significant negative correlation between parasitism on Sennius and glandular trichome length (Table 5).

Table 5 Results of multiple linear regression analyses between plant traits and seed predation
Full size table

The analysis of glabrous and pubescent fruit-bearing plants of C. ochnacea revealed no significant differences in seed predation rates (Table 6) and parasitism rates on Sennius maculatus (Table 6). However, we found that predation and parasitism rates were greater among plants bearing glabrous fruit, supporting our hypothesis. Although fruit pubescence was significantly greater in C. cipoana than in C. semaphora (Tables 2, 6), seed predation rate by Sennius bruneus was significantly greater on C. cipoana than in C. semaphora (Table 6). Parasitism rate on Sennius bruneus, in contrast, was greater on C. semaphora than on C. cipoana seeds (Table 6).

Table 6 Results of pairwise comparisons by t test
Full size table

Discussion

This study is the first to document tri-trophic interactions analyzing the effects of several plant traits on seed predation and parasitism rates on thirteen taxa of the widespread Chamaecrista genus in Brazilian rupestrian grasslands. In general, phenologically aseasonal Chamaecrista species had greater fruiting dispersal indexes (FDI). However, most of the taxa had lower FDI indicating a restricted fruiting period. In seasonal systems, massive fruit flushing at the beginning of the dry season is a general phenological pattern, whereas in riverine forests, plants are evergreen and produce new fruits throughout a more extended period during the wet season and a short period in the dry season (Frankie et al. 1974; Opler et al. 1980, Bullock and Solís-Magallanes 1990; van Schaik et al. 1993). Therefore, high availability of fruits and seeds to predators occurs during a short period of time in seasonal systems. Phenological patterns have wide ecological implications for the behavior of the animals that feed on plants. Fruit abundance is usually an important parameter associated with the behavior of frugivores and seed predators (Koenig 1997; Renton 2001). However, in our study, plant reproductive phenology had no significant influence on seed predation rates by both bruchids and ectophagous herbivores. It also did not affect parasitism rates on Sennius. This result may reflect the limited range among the observed phenological behaviors, with no host taxon concentrating its fruit production in an especially narrow period of time (van Schaik et al. 1993). The season when a taxon produces its seeds seemed to affect, in some cases, the identity of the predators preying upon it, but not the predation rates.

Fruit pubescence is poorly studied as a defensive trait, but should be more effective against generalist predators (Levin 1973; Johnson 1975; Woodman and Fernandes 1991; van Dam and Hare 1998a). Despite the lack of a correlation across genetic families, variation in latex and trichomes was negatively correlated with herbivore damage for A. syriaca (Agrawal 2005). Other studies indicated that trichomes on the fruits of Brassica hirta deter feeding of the flea beetle Phyllotreta cruciferae (Lamb 1980; Handley et al. 2005), or herbivores that oviposit inside the host’s tissue (Ågren and Schemke 1994; Valverde et al. 2001). Nevertheless, in this study, fruit pubescence did not show a significant correlation with seed predation rates by bruchids or by ectophagous herbivores. It is not surprising that highly specialized seed predators such as bruchids were unaffected by fruit pubescence (Gannon and Bach 1996). Reinforcing this result, there were no significant differences on seed predation rates suffered by C. ochnacea individuals with glabrous and pubescent fruits. However, the interspecific comparison between seed predation by Sennius bruneus on C. cipoana and C. semaphora revealed that the species with greater trichome density and size (C. cipoana) had significantly greater losses due to seed predation.

We found a relationship between seed size and seed predation for both predator types. Bruchid seed predation was positively related with seed size. Components of seed morphology such as size and shape can be important for the preference of endophagous seed predators (Szentesi and Jermy 2003). Some studies indicated that large seeds represent sites of higher nutritional quality and lower content of chemical defense to seed predators (Janzen 1969; Brewer 2001; Gómez 2004). In addition, preference for larger seeds is a widespread trend in the family. It must be observed, however, that the studied bruchid species belong, more specifically, to the “quasi-endophagous” guild, which means that a larva is capable of using the first attacked seed as well as the neighboring seeds to complete its development. This capacity reduces the constraint represented by seed size to real endophagous species, whose entire development take place inside a single seed (Center and Johnson 1973). In spite of this consideration, excluding locally very rare species (C. cathartica and C. choriophylla) that probably escaped seed predation due to their rarity (Chung and Waller 1986), the only two taxa not predated at all by bruchids were those with the smallest mean seed sizes (C. rotundifolia and C. desvauxii malacophylla) while those with larger seeds (C. ochnacea, C. dentata, and C. cipoana) suffered greater predation.

Larger seeds are costly, but usually have higher germination rates (Fenner 1985; Azcátare et al. 2006). On the other hand, their higher energetic content exerts a greater attractiveness to seed predators, justifying a heavier investment on chemical defense. Therefore, well-defended seeds tend to be preyed upon by specialists (Center and Johnson 1974; Janzen 1980; Mari et al. 2008). The higher production costs of these seeds leads to a selective pressure for harder seed coats that maximize seed germination chances through the possibility of a long dormancy period (Janzen 1980). Taxa with less well-defended seeds should tend to be preyed upon by generalists and to produce smaller seeds in greater number (Janzen 1980). This is in accord with our results; the overall trend was that species from section Absus, which have the largest seeds and smallest number of seeds per fruit, were preyed upon by the more specialized predators. In contrast, species from section Chamaecrista were more variable in seed size and number of seeds per fruit, and only one of them was preyed upon by a relatively specialized predator (S. kingsolveri on C. mucronata). All taxa from section Xerocalyx had large numbers of small seeds per fruit and were preyed upon only by generalists. It is unlikely that bruchids could have competitively excluded ectophagous seed predators on species from section Absus, because predation rates suffered by these species were not strong enough to impair ectophage occurrence. The complete absence of ectophagous seed predation on section Absus may have two not mutually exclusive explanations. Absus seeds may have a very hard seed coat and/or be well defended chemically. A hard seed coat does not represent a barrier to these bruchids, since they enter the seeds when they are still immature, but it may exclude ectophagous insects that eat seeds from the outside, when they are already mature. Chemically well-defended seeds may pose a problem to generalist seed predators like ectophagous insects, but can be tolerated, or even preferred (Rosenthal et al. 1978; Agrawal et al. 1999), by specialists like Sennius species that prey on seeds from section Absus.

Sennius niger, the more generalist bruchid species, preyed predominantly on seeds of three geographically widespread host taxa (C. desvauxii latistipula, C. d. mollissima, and C. ramosa, see Table 1) and on seeds of one intermediately distributed species (C. venulosa, distributed along Espinhaço Mountain Chain, see Table 1). This fact strongly reinforces the conclusion that S. niger is a more generalist species, since widespread host plants allow their predators to come in contact with a wider range of new potential hosts through time, as predator distribution also tends to be widespread (Johnson and Siemens 1995). The other three Sennius species preyed on seeds of more narrowly distributed species, all from section Absus, excepting C. mucronata (from section Chamaecrista, intermediately distributed), preyed upon by S. kingsolveri, but in much lower rates than its preferred host species, the more restricted C. dentata.

The negative correlation between glandular trichome size and parasitism rate on Sennius is striking. Natural enemies may be an important factor determining herbivore host range (Bernays and Graham 1988; Helms et al. 2004; van Veen et al. 2006). Parasitoids seem to exert an effective population regulation on some of the studied bruchid species. Many studies have shown that host plant traits may act as herbivore protections by impairing their enemies’ actions (Gruenhagen and Perring 1999, 2001; Vinson, 1999; Lovinger et al. 2000). Hence, in a tri-trophic interaction, seed predation pressure by bruchids may be determining a selective advantage on fruit glandular trichome size reduction, through the consequent facilitation of parasitoid action. Micro-hymenopteran difficulty with large trichomes and their secretions may be related to oviposition behavior and/or body size. Additional support for this conclusion is given by the significant difference found in parasitism rates on Sennius bruneus attacking C. cipoana and C. semaphora. C. cipoana had 15.55 % of their seeds destroyed by Sennius bruneus, with 0.9 % of parasitism. C. semaphora had only 4.69 % of its seeds preyed upon, but on this species, Sennius bruneus suffered a 8.3 % parasitism rate. C. cipoana had significantly larger glandular trichomes and higher trichome densities. Hence, female parasitoids found a better resource on the scarcer bruchids on C. semaphora and bruchids on C. cipoana seem to be protected by their own host plants. Moreover, Fernandes and Bicalho (1995) did not find significant correlation between trichome density and herbivory on Chamaecrista dentata leaves, steams, and fruits, but fruit trichome density was the smallest among the three organs. This result suggests that the three organs may have been subjected to different selective pressures (Valverde, et al. 2001). As only fruits deal with seed predators, they may be the selective pressure that caused a reduction in trichome density and size on C. dentata fruits, favouring parasitoid action. We found no effects of trichomes on the incidence of parasitoids and parasitism rates of C ochnacea between pubescent and glabrous fruits.