•The pitcher-shaped leaves of Nepenthes carnivorous plants have been considered as pitfall traps that essentially rely on slippery surfaces to capture insects. But a recent study of Nepenthes rafflesiana has shown that the viscoelasticity of the digestive fluid inside the pitchers plays a key role.
•Here, we investigated whether Nepenthes species exhibit diverse trapping strategies. We measured the amount of slippery wax on the pitcher walls of 23 taxa and the viscoelasticity of their digestive liquid and compared their retention efficiency on ants and flies.
•The amount of wax was shown to vary greatly between species. Most mountain species exhibited viscoelastic digestive fluids while water-like fluids were predominant in lowland species. Both characteristics contributed to insect trapping but wax was more efficient at trapping ants while viscoelasticity was key in trapping insects and was even more efficient than wax on flies. Trap waxiness and fluid viscoelasticity were inversely related, suggesting the possibility of an investment trade-off for the plants.
•Therefore Nepenthes pitcher plants do not solely employ slippery devices to trap insects but often employ a viscoelastic strategy. The entomofauna specific to the plant’s habitat may exert selective pressures, favouring one trapping strategy at the expense of the other.
Carnivorous plants circumvent the nutrient shortage characterizing the habitats they colonize by deriving key nutrients from arthropods which they attract, trap and digest in specialized leaves (Juniper et al., 1989; Ellison & Gotelli, 2001). Nepenthes (Caryophyllales: Nepenthaceae) is a climbing and carnivorous plant genus characterized by leaves modified as pitcher traps (Fig. 1). It encompasses > 100 species, mainly distributed in southeastern Asia, with the islands of Borneo and Sumatra as hotspots of diversity (Clarke, 1997, 2001; Cheek & Jebb, 2001; McPherson, 2009), and colonize various habitats, including coastal lowlands, cliffs and high-altitude forests, with a high rate of endemism (Clarke, 1997; McPherson, 2009). Nepenthes species show a great diversity of pitcher morphologies, which could reflect differences in trapping strategies and stem from their adaptation to different arthropod fauna found in their habitats. But it is not known whether the trapping strategies of these pitcher plants are actually functionally diverse, and whether this diversity is linked to ecological characteristics of their environment. Nepenthes species are known to vary in their arthropod prey assemblage (Kato et al., 1993; Adam, 1997; Merbach et al., 2002) and even in their N-sequestration strategies (Moran et al., 2010), with some outlying species moving away from a purely carnivorous habit by deriving part of their nitrogen from leaf detritus (Moran et al., 2003), vertebrate faeces (Clarke et al., 2009; Chin et al., 2010), or from the nutritional service of a symbiotic hunter ant (Bonhomme et al., 2011). The trapping strategy of strictly insectivorous species (the vast majority of these pitcher plants) has never been investigated in a comparative study within the genus.
Plants in the Nepenthes genus have long been thought to function as simple pitfall traps relying on slippery surfaces that decrease insect adhesion (Juniper & Burras, 1962; Juniper et al., 1989; Gaume et al., 2002, 2004; Gorb et al., 2005) and wettable surfaces that cause insect aquaplaning (Bohn & Federle, 2004; Bauer et al., 2009). But in 2007, Nepenthes rafflesiana was shown experimentally to use another mechanism. It produces a digestive liquid partly made up of long-chain polymers, the viscoelastic properties of which have a marked effect on insect retention (Gaume & Forterre, 2007; Di Giusto et al., 2008). Even when greatly diluted by water, the digestive liquid in N. rafflesiana has sufficient elastic properties to trap insects (Gaume & Forterre, 2007). This not only means that the digestive liquid might be crucial for the capture success of this tropical pitcher plant that is often subjected to heavy rains, but also that even species with a liquid viscosity similar to that of water may have unsuspected elastic properties that result in high trapping capacities. Therefore, the viscoelastic character of the digestive fluid might have remained cryptic in a number of species and could be far more widespread than expected in the Nepenthes genus. Interestingly, N. rafflesiana var. typica bears pitchers with a waxy zone and mainly traps ants during its juvenile phase; but as the plant ages, the waxy layer is lost (Gaume & Di Giusto, 2009) and the upper pitchers, which are only produced in the adult phase, contain a highly viscoelastic fluid that proves to be very efficient against flying insects (Di Giusto et al., 2008). By contrast, the elongated traps of N. rafflesiana var. elongata keep their waxy layer throughout plant ontogeny and the plant mainly captures ants (Gaume & Di Giusto, 2009).
This casts doubts on the common belief that all Nepenthes species exhibit the same trapping strategy based on the slipperiness of their pitchers. This study explores whether viscoelastic fluids are common among Nepenthes species and whether they are produced in addition to or at the expense of a slippery waxy layer. It also investigates the effects of each of these two retentive devices on the capture of different insect types.
To address these questions, we studied the functional diversity of Nepenthes pitcher plants in a sample of species differing in their geographic origins and habitats. We measured the traits directly involved in the ‘slippery’ and ‘viscoelastic’ strategies, i.e. waxiness (quantity and density of wax coating the inner pitcher walls) and viscoelasticity (relaxation time) of the digestive liquid. We used insect bioassays to compare the retentive ability of different Nepenthes species and measure how efficiently pitcher waxiness and fluid elasticity contribute to the retention of each type of prey. Finally, we investigated whether waxiness and viscoelasticity are correlated.
Materials and Methods
One of the authors (J-J. L.) owns a glasshouse collection of Nepenthes pitcher plants (located in Peyrusse-Massas, Gers, France) which in 1995 was recognized as the National French Conservatory of Carnivorous Plants. The insect bioassays performed in April 2007 employed the following subset of 12 Nepenthes species: Nepenthes ampullaria Jack, N. fusca Danser, N. longifolia Nerz & Wistuba, N. maxima Reinw. ex Nees, N. mirabilis var. echinostoma (Lour.) Druce, N. petiolata Danser, N. rafflesiana var. typica Jack, N. ramispina Ridl., N. spathulata Danser, N. spectabilis Danser, N. tobaica Danser and N. ventricosa Blanco. These species are representative of the ecological and geographical diversities found within the genus (terrestrial/epiphytic climbers; lowland/mountain species; species originating from Borneo, Sumatra, Sulawesi, Philippines and Peninsular Malaysia). Nepenthes species usually produce two types of pitchers: lower terrestrial pitchers, produced during the self-supporting/juvenile stage of the plant, and upper aerial pitchers produced during the climbing/sexually mature stage (Clarke, 1997). Since the glasshouse plants did not all produce upper pitchers, only young individuals with lower pitchers were used. One freshly opened lower pitcher (opening dating from < 1 wk) was selected on three different individuals of approximately the same size (30–60 cm) in each of the 12 species. A total of 36 pitchers were thus used.
In April 2008, a larger sample corresponding to 23 taxa and 21 species was used for comparisons of pitcher waxiness and viscoelasticity within the genus. This sample comprised all the previously cited species, except N. ramispina and N. maxima, which did not produce any lower pitchers at the time of our study, together with 13 other taxa: N. albomarginata T. Lobb ex. Lindl, N. burgbidgeae Hook. f. ex Burb., N. copelandii Merr. ex Macfarl., N. eymae Sh. Kurata, N. glabrata J. R. Turnbull & A. T. Middleton, N. gracilis Korth., N. macrophylla (Marabini) Jebb & Cheek, N. madagascariensis Poir., N. mindanaonensis Sh. Kurata, N. mirabilis var. typica (Lour.) Druce, N. rafflesiana var. elongata Hort., N. tenuis Nerz & Wistuba and N. vogelii Schuit. & de Vogel. Although, the thickness and length of the waxy layer do not to the naked eye seem to vary when the pitcher ages, the fluid viscoelasticities are reported to decrease when the pitchers age, some of them becoming apparently watery 11–14 d after pitcher opening in N. rafflesiana (Bauer et al., 2009). To reduce as far as possible development-induced heterogeneity between pitchers, waxiness and viscoelasticity were measured for each species in the youngest pitcher that had opened in the previous week.
Measurement of insect retention ability
Experiments comparing the trapping ability of the 12 Nepenthes species were carried out in April 2007 under homogeneous temperature and hygrometry conditions (26–27°C, 80–90%). Retention rates in the 12 species were compared for the ant Lasius niger L. (Hymenoptera, Formicidae, Formicinae) and the fly Calliphora vomitoria L. (Diptera, Calliphoridae). According to the prey spectra of Nepenthes species published so far (reviewed by Juniper et al., 1989; Ellison & Gotelli, 2009), ants and flies are the most commonly trapped insects. Although the species used in our bioassays do not occur, at least for the ants, in the natural prey fauna of the Nepenthes species tested, they represent insects typically trapped by Nepenthes as regards their size and shape and the insect order to which they belong. Our aim here was simply to compare the ability of different Nepenthes species to retain flying insects vs crawling insects. A colony of L. niger found near the glasshouses provided us with worker ants. They were transferred to a plastic tube and placed on the pitcher rim, the so-called peristome (Fig. 1a). About 500 laboratory-bred C. vomitoria larvae were kept at 30°C for 7 d until adults emerged. The adults were then collected and confined to a glass jar connected to a cylindrical mesh, the aperture of which was closed by a piece of string. Once some of the flies emerged, the string was removed and the few emerging flies were collected in an inverted beaker connected to the mesh. The beaker and mesh were then disconnected from the jar, which was rapidly closed with a temporary cap, then ‘slipped’ onto a pitcher, enclosing pitcher and flies together (Fig. 1b) by attaching the end of the mesh around the tendril sustaining the pitcher. These experimental set-ups allowed ants and flies direct access to the pitcher. Each trial consisted of an insect’s fall into the pitcher and the insect’s fate was recorded as the binary outcome, retained/not retained within the pitcher. We observed each insect until it died or escaped from the pitcher. An insect was considered as retained if it did not successfully escape from the pitcher. We considered three different pitchers for each of the 12 Nepenthes species, and performed 10 tests per pitcher, each time with a different ant, and 10 tests, each time with a different fly, gathering a total of 720 binary responses and 72 retention rates.
Quantitative measurement of the characters involved in trapping
All the (intra- + epicuticular) wax in the pitchers was extracted using warm chloroform as described by Riedel et al. (2003) and weighed to within 1 μg on a Sartorius MC5 balance (Gottingen, Switzerland –Fig. 1c). Because the plants are fragile, only one pitcher per species was used for wax extraction and measurement. The total weight of wax was used as an explanatory variable for retention rates: it takes account not only of the thickness but also of the length of the waxy zone, and both parameters are probably important characteristics of the slippery trap. Each species’ relative investment in ‘wax’ and ‘liquid viscoelasticity’ was determined by measuring the density of wax, estimated as the weight of wax per cm² of pitcher wall. This measurement of wax density is relevant because the viscoelasticity measurement also reflects density (i.e. that of polymers) in the fluid. In any case, the two different measurements of wax quantity plotted against fluid viscoelasticity showed similar statistical trends.
Fluid viscoelasticity was first estimated by observing the presence or absence of a filament when the fluid was stretched between two fingers. This qualitative measurement of viscoelasticity was performed on the fluid of each of the three pitchers in the 12 Nepenthes species studied in April 2007. Appropriate equipment for rheological studies was acquired in April 2008 and fluid from the 23 taxa was subject to a quantitative measurement of its viscoelasticity. An estimation of pitcher fluid viscoelasticity was obtained by measuring elastic relaxation time, defined as the time required for a filament of fluid to break, as described by Gaume & Forterre (2007). The fluid was subject to vertical strain by rapidly lifting a thin rod (diameter D0 = 3.0 mm) 3 cm vertically from a 40 μl sample of liquid, thus creating an elongated liquid filament. The subsequent thinning speed and time to filament rupture were recorded at a high spatial and temporal resolution (31 pixels mm−1 and up to 3500 frames s−1) using a Phantom Miro IV high-speed camera (Vision Research, Wayne, NJ, USA) and a Nikkor 60 mm macro lens. The recordings were then analysed using an Image J – R script (Abramoff et al., 2004; R Development Core Team, 2009) that we developed for this purpose (Fig. 1d). Each fluid was tested in triplicate and mean relaxation time calculated as an estimation of its viscoelasticity. The relaxation time of distilled water was measured six times as the nonviscoelastic reference fluid. All measurements were performed under homogeneous temperature conditions (25–26°C).
Logistic regressions were used to explain the variability in insect retention success. Backward procedures were adopted for model selection, starting with removal of the nonsignificant, highest-order interactions.
Logistic models were used to address the following questions: does retention success vary between Nepenthes species, types of insect (ant/fly) and pitchers within a given species; and could this variability be explained by the trapping features of the species, for example, the waxiness of pitcher walls and the viscoelasticity of pitcher fluid? We first performed a mixed logistic model on observed retention successes, which set ‘species’, ‘insect type’ and ‘species’ × ‘insect type’ interaction as fixed explanatory factors and ‘pitcher’ as a random factor nested within species. This mixed model was run with SAS v. 9.2 software package using the GLIMMIX procedure (SAS Institute, Cary, NC, USA), but all the further statistical analyses were carried out using R software (R Development Core Team, 2009). As the pitcher effect was not significant, we then pooled retention success data for the three pitchers of each species. In order to test whether species retention rates (number of tests per species = 30) could be explained by their trapping features, we thus performed two other logistic regression models (one for each type of insect) on retention rates with ‘presence/absence of a viscoelastic liquid’ and ‘quantity of wax in the trap’ as explanatory variables. Corrections for overdispersion were applied when necessary using the quasibinomial error distribution implemented in R.
Relaxation times for the 23 digestive fluids measured in April 2008 were compared using Student’s t-tests with the capillary pinch-off time for water, that is, the shortest breaking time for a filament. It should be noted that elastic relaxation times shorter than the capillary pinch-off time for water cannot be measured using this capillary break-up method (Rodd et al., 2005). A fluid was conservatively qualified as viscoelastic if its relaxation time was significantly longer than the capillary pinch-off time for water, with P <0.01. To determine whether there was a correlation between quantity of wax and viscoelasticity of the digestive liquids, we selected the species shown to have a viscoelastic fluid and tested whether the quantity of wax fitted a linear then a hyperbolic function of liquid viscoelasticity. Goodness of fit of these two regressions was then compared using Akaike’s information criterion in R.
Fisher’s exact test was performed on a contingency table (viscoelastic/nonviscoelastic liquid vs mountain/lowland species) to statistically compare the frequencies of species associated with viscoelastic fluid in mountain and lowland species. The altitude range of the studied species was obtained from different bibliographic sources (Clarke, 1997, 2001; Cheek & Jebb, 2001; McPherson, 2009) and the rough threshold of 1000 m was applied to distinguish between lowland and mountain species. This threshold is commonly used in tropical areas (Richards, 1996) to distinguish between species, including Nepenthes spp. (Clarke, 1997), of low and high altitudes. In our Nepenthes sample, species that were considered as lowland species generally occur from sea level up to 1000 m at the most (apart from N. longifolia which is not found below 300 m and could be considered as an intermediate species), while species that were considered as montane species occur from 700 up to 2900 m. The average altitudinal range of our lowland species is 265 ± 236 m, while the average altitudinal range of our montane species is 1550 ± 399 m. Measurements on each Nepenthes species are not truly independent because species share evolutionary history, and comparative methods that take into account the phylogenetic relatedness of the studied species should be used to test the relationship between traits (Revell, 2010). But available phylogenies for the genus Nepenthes (Meimberg & Heubl, 2006) are only poorly resolved, precluding applications of these tests. However, species used here are scattered throughout the currently available phylogenetic trees of the genus, and therefore statistical biases as a result of close relatedness among species should be minimized.
Functional diversity in the trapping success of Nepenthes
The species studied differed significantly in terms of their retention success on the insects employed in our bioassays (67 ± 21% of insects retained, significant species effect on retention success in the mixed logistic regression model; Table 1, Fig. 2) and they were on average more efficient on ants than on flies (significant insect effect, percentage of ants retained = 72 ± 18%, flies = 62 ± 27%, SD given in the text; Table 1). However, some species were more efficient on ants while others retained flies more easily (significant insect × species interaction; Table 1). There was no significant difference in the retention success of pitchers of the same species (no random effect of pitcher: pitcher variance estimate = 1.73 × 10−19, tests of covariance based on the residual pseudo-likelihood: P ∼ 1).
Table 1. Results of the mixed logistic model testing for the fixed effects ‘Nepenthes species’ and ‘insect type’ and the random effect ‘pitcher within species’ on trapping success
Variability in insect retention
Response variable: retention success
P (> F)
The effect of pitcher was not significant (variance = 1.73 × 10−19, P ∼ 1). ndf, numerator degrees of freedom; ddf, denominator degrees of freedom.
Species × Insect
The efficiency of pitcher wax and viscoelastic fluid is insect-dependent
The total weight of wax averaged 1.85 mg and ranged from 0.28 mg in N. ampullaria, which has no visible epicuticular wax in its pitchers, to 3.40 mg in N. maxima, which bears a thick layer of epicuticular wax in its pitchers. Fluid viscoelasticity also varied between species. The pitchers of six species (N. ampullaria, N. longifolia, N. mirabilis, N. ramispina, N. spectabilis and N. ventricosa) were shown to contain a nonviscoelastic, apparently water-like fluid, whereas the pitchers of six other species (N. fusca, N. maxima, N. petiolata, N. rafflesiana, N. spathulata and N. tobaica) contained a fluid shown to be viscoelastic by the creation of a filament when stretched between two fingers. The highest retention rates were observed in species with viscoelastic fluids, and the lowest were observed in species with water-like fluids (Fig. 2).
Differences in wax quantity and fluid viscoelasticity explained most of the variations observed in the trapping ability of the species which also differed with regard to insect type. For ants, retention rates increased significantly with wax quantity and fluid viscoelasticity (Table 2a, Fig. 3a). For flies, retention rates did not significantly depend on the amount of wax but on fluid rheometry as retention rates were far higher when the fluid was viscoelastic (Table 2b, Fig. 3b). Our observations of insect behaviour corroborated these results. Ants that fell into the liquid close to the pitcher wall were not extensively wetted and were often observed to reach and climb up the pitcher wall but they then slipped frequently when reaching the waxy zone. By contrast, flies that were not extensively wetted by the liquid were sometimes observed to climb up the pitcher wall and then successfully take off from there, occasionally without even touching the waxy layer. If wetted by the viscoelastic liquid, the insects had little chance of escaping. The more they struggled in the fluid, the greater it resisted their movements and the insects rapidly became exhausted and drowned.
Table 2. Results of logistic regression testing for wax quantity and fluid viscoelasticity effects on trapping success for ants (a) and flies (b)
Response variable: retention rates
P (> χ²)
(a) Ant retention success dependent on viscoelasticity and wax quantity, Cox–Snell pseudo r² = 0.53
Relative investments in wax and liquid viscoelasticity
Wax quantities measured in April 2008 were consistent with those obtained in 2007. Wax densities ranged from 0.022 mg cm−2 (N. ampullaria, wax quantity = 0.229 mg) to 0.608 mg cm−2 (N. macrophylla, wax quantity = 3.180 mg), averaging 0.124 mg cm−2 (median = 0.087 mg cm−2, SD = 0.135). Thirteen of the 23 species were viscoelastic, that is, their relaxation time was significantly longer than that of water, the nonviscoelastic reference fluid (t-test results in Supporting Information, Table S1). N. longifolia and N. spectabilis appeared to be slightly viscoelastic but were not classified as viscoelastic in 2007, probably because they create viscoelastic filaments that are not visible by direct observation. Among the 13 mountain species, 12 were found to be viscoelastic whereas only three lowland taxa (two species, N. longifolia, N. rafflesiana var. typica and N. rafflesiana var. elongata) out of the 10 lowland taxa (nine species) in our sample were found to be viscoelastic. These proportions differed significantly (Fisher’s exact test performed on the 23 taxa, P = 0.006; on the 21 species, P =0.003), demonstrating that mountain species more often exhibit viscoelastic fluids. By way of comparison, the values obtained for water ranged from 0.024 to 0.028 s (n =6). None of the species with very waxy pitchers was found also to exhibit a very viscoelastic fluid, and vice versa. Moreover, the quantity of wax produced in the pitchers of viscoelastic species were seen to better fit a hyperbolic function (Akaike information criterion (AIC) = −29.72) than a linear function (AIC = −11.00) of the fluid viscoelasticity. The more wax a species produces, the less viscoelastic is its fluid (wax density (mg cm−2) = 0.013 + 0.014/relaxation time (s), F1,12 = 14.19, P = 0.001, R² = 0.75; Fig. 4).
Our comparative study of the trapping systems of Nepenthes pitcher plants generated three important results. First, and contrary to common belief (but see Gaume & Di Giusto, 2009; Bonhomme et al., 2011), the different species show functional diversity in their retentive devices and do not rely solely on the slipperiness of their trap to capture insects. The results of this study show that N. rafflesiana is not the only species to possess a viscoelastic fluid. This character may be widespread within the genus as found in two-thirds of the species in our study. Secondly, wax and viscoelatic fluids do not target the same type of prey: wax appears to be efficient only for ants, whereas viscoelasticity proved to be a powerful trapping device for both insect types and is more often found in mountain than in lowland species. Thirdly, our cross-species comparison suggests that investments in the ‘waxy’ trait and in ‘viscoelastic fluid’ could be made at the expense of the other. At the least, species that are very viscoelastic do produce very little wax. Altogether this suggests that there are two different trapping strategies in these pitcher plants, a ‘waxy’ strategy and a ‘viscoelastic’ strategy.
A widely shared viscoelastic trap in Nepenthes pitcher plants and the question of its origin
The fluid contained in the pitchers of most of the Nepenthes species studied was viscoelastic, and this might therefore represent the rule in the genus rather than the exception. Contrary to wax, whose efficiency as a trapping device has been shown to be quantity-dependent, even low viscoelasticity appears to contribute strongly to a plant’s trapping ability. Therefore, it is surprising to note that the fluids of some Nepenthes species are highly viscoelastic. It is possible that this viscoelasticity increases with plant age (Gaume & Di Giusto, 2009). But, for all species, the plants used in our study were carefully chosen to be of similar age. Another explanation could be that, as the plant species differ in their habitats, some are more subject than others to rainfall and humidity (e.g. those of altitudinal mossy forests or those that have a reduced lid that can protect the pitcher) and to subsequent fluid dilution by water. Greater production of the polymers that are assumed to cause the viscoelasticity of pitcher fluid might have been selected in some species, helping them to cope with the problem of daily dilution. This hypothesis is corroborated by the results of Gaume & Forterre (2007), who showed that the elastic fluid of N. rafflesiana (which here appears to be among the most viscoelastic species), when diluted in 95% of water, was still viscoelastic enough to capture all the insects dropped into the pitchers.
Since relaxation times of < 100 ms are impossible to detect without the use of a high-speed camera, viscoelastic fluids have probably gone unnoticed in many species and may be far more common than suspected. This raises the question of whether the viscoelastic fluids in Nepenthes species have a common origin. Interestingly, the glue secreted by the leaves of Drosera, another carnivorous genus, is composed of acid polysaccharides (Gowda et al., 1982) and these have been demonstrated to be viscoelastic (Erni et al., 2008). This is also probably the case for the glue secreted by Drosophyllum (V. Bonhomme, unpublished). Both of these genera are the closest relatives of Nepenthes in the phylogeny of the Caryophyllales (Heubl et al., 2006) and their traps are known to function like flypaper (Juniper et al., 1989). We can therefore advance the hypothesis that the viscoelastic polysaccharide fluids in Nepenthes and in these other carnivorous genera have a common and thus plesiomorphic origin, with the glue of the other carnivorous genera simply containing a far higher concentration of polysaccharides than the fluid in Nepenthes.
Relative investment in different trapping devices
The appearance of botanical carnivory and the evolution of specialized traps are subject to powerful cost–benefit constraints (Givnish et al., 1984; Ellison & Gotelli, 2001; Pavlovičet al., 2007). We can therefore assume that it is costly for carnivorous plants to produce modified leaves with lower photosynthetic capacities (Pavlovičet al., 2007, 2009), and that the biosynthesis of trapping features is subject to selective pressure and can be maintained throughout evolution only if the cost of these features is exceeded by the benefits they provide in terms of insect-derived nutrients. Development of the waxy zone, mainly composed of aliphatic compounds dominated by very long-chain aldehydes (e.g. triacontanal or dotriacontanal containing 30 or 32 carbon atoms, respectively Riedel et al., 2003, 2007), is metabolically costly for the plant. The molecules responsible for digestive fluid viscoelasticity are assumed to be long-chain polysaccharides (Gaume & Forterre, 2007) that must also be costly to synthesize. This could provide part of the explanation as to why none of the species we tested possesses both very viscoelastic fluids and very waxy pitchers. The inverse relationship that seems to link these two quantitative traits in our cross-species comparisons might illustrate the existence in the plant of an investment trade-off. This will need to be tested at the genus scale with an analysis that takes into account the phylogenetic relationships of the measured species. A test of this hypothesis will also necessitate studies of several populations of a given species showing some variations in these traits.
Interestingly, a few plant species in our study showed both nonviscoelastic fluids and only slightly waxy pitchers, or pitchers that contained no epicuticular wax at all, such as N. ampullaria and N. ventricosa. These plants are outliers in the Nepenthes genus. Perhaps the pitchers do not have a strictly carnivorous diet; this is the case for N. ampullaria, which obtains part of its nitrogen from leaf debris (Moran et al., 2003). An additional explanation is that they utilize other trapping strategies. It is possible that N. ampullaria relies uniquely on its peristome, which forms a steep slop, to trap its prey. Several other features, such as water-dependent structures facilitating insect-aquaplaning (Bohn & Federle, 2004; Bauer et al., 2008) or specific pitcher morphology, might favour both insect capture and retention. But as attested by the high coefficients of determination of the models testing for the effect of wax and viscoelasticity, these other features only played a minor role in insect retention.
The role of prey in the evolution of different trapping strategies
Whether wax and viscoelastic fluids are produced at each other’s expense, as these devices are necessarily costly, the question arises as to the selective factors that have favoured the evolution of these traits. Pitcher wax causes insects to slide and is thus implicated in both capture and retention (Juniper & Burras, 1962; Gaume et al., 2002), while the viscoelastic fluid acts on retention (Gaume & Forterre, 2007). The results obtained in our study show that the efficiency of these strategies is prey-dependent. Wax is more efficient on ants than on flies, whereas viscoelasticity is very efficient on both insect types and definitely more efficient than wax on flies. Winged insects are able to take off from the pitcher wall without even touching the waxy surface, and even if they do enter into contact with it, the wax acts only on their attachment systems (Gaume et al., 2004; Gorb et al., 2005) not on their flying system. By contrast, crawling insects have no other option than to cope with the wax that contaminates their pads and causes them to lose adhesion. Moreover, since winged insects have a higher surface : volume ratio than crawling insects, they offer a larger surface area for the viscoelastic fluid to exert its retentive force and this may explain why they are more often retained in Nepenthes liquids (Gaume et al., 2002; Gaume & Forterre, 2007).
Hence wax and viscoelastic fluid clearly do not have the same function and do not target the same types of insect. This suggests that they might represent adaptation to different prey spectra and that local differences in entomofauna might exert different selective pressure on the development of wax and/or viscoelasticity. For any given pitcher waxiness, a ‘viscoelastic strategy’ is needed to trap flies with the same efficiency as ants. This means that habitats dominated by ants, such as the lowland forests of Borneo (Gunsalam, 1999; Davidson et al., 2003), may favour the development of a waxy ‘slippery’ strategy. On the other hand, habitats dominated by flying insects may favour the development of a ‘sticky’, viscoelasticity-based strategy. Such habitats are generally found at higher altitudes where ants are few in number but flying insects are relatively more abundant (Collins, 1980). This can also temporarily be the case for lowland, open, and regularly flooded habitats such as those inhabited by N. rafflesiana var. typica (Gaume & Di Giusto, 2009), which is associated with a flower scent cue that more specifically targets flying insects (Di Giusto et al., 2010).
We therefore advance the hypothesis that the scarcity of ants in tropical mountains (Borneo (Collins, 1980; Clarke et al., 2009), the Philippines (Samson et al., 1997)) and the relative abundance of flying insects (Collins, 1980) provide part of the explanation for the widespread viscoelastic strategy among mountain Nepenthes species. A comparative study (Adam, 1997) corroborates this hypothesis by showing that mountain species tend to trap a larger prey spectrum, including more dipterans and coleopterans, than lowland species, which were recorded to trap mostly ants. Furthermore, at least seven species in mountain mossy forests, and known to possess a highly viscous fluid, are reported to be specialized in the capture of flying insects: N. inermis has been reported to be (under the name of N. bongso) specialized in trapping midges (Kato, 1993); N. aristolochioides is specialized in trapping midges; N. dubia, N. jamban, N. eymae and N. talangensis are specialized in trapping small dipterans (McPherson 2009); and N. jacquelinae has been observed to trap mainly larger flying prey (Clarke, 2001). Interestingly, the pitchers of such species do not contain a waxy zone and are all funnel-shaped.
Few comparative studies have been conducted on the prey spectra of Nepenthes species (but see Kato et al., 1993; Adam, 1997). To test our hypotheses on the evolution of Nepenthes trapping devices, we need to carry out studies comparing the prey spectra of Nepenthes species with the entomofauna found in their habitats, and relating this to their insect-trapping devices. This would help in understanding the ecological mechanisms underlying the evolution and diversification of these pitcher plants.
We wish to thank B. Buatois, M. Guéroult and R. Leclerc for their helpful assistance in the laboratory and glasshouses, G. Le Moguédec for statistical advice, and S. Martinez, J-M. Felio and P. Cervetti (IUSTI) for technical assistance in designing the portable device (capillary break-up rheometer) used to measure fluid viscoelasticity. A high-speed video camera was purchased thanks to a CIRAD grant ‘Structuring scientific equipment’ delivered to AMAP in 2008. This work was also funded through a PEPS CNRS/ST2I contract obtained in 2007 by Y.F. and L.G. Three anonymous reviewers are acknowledged for their review of the manuscript and M. Jones (Transcriptum) for his review of the manuscript’s English.