Oil adsorption ability of three-dimensional epicuticular wax coverages in plants

Primary aerial surfaces of terrestrial plants are very often covered with three-dimensional epicuticular waxes. Such wax coverages play an important role in insect-plant interactions. Wax blooms have been experimentally shown in numerous previous studies to be impeding locomotion and reducing attachment of insects. Among the mechanisms responsible for these effects, a possible adsorption of insect adhesive fluid by highly porous wax coverage has been proposed (adsorption hypothesis). Recently, a great decrease in insect attachment force on artificial adsorbing materials was revealed in a few studies. However, adsorption ability of plant wax blooms was still not tested. Using a cryo scanning electron microscopy approach and high-speed video recordings of fluid drops behavior, followed by numerical analysis of experimental data, we show here that the three-dimensional epicuticular wax coverage in the waxy zone of Nepenthes alata pitcher adsorbs oil: we detected changes in the base, height, and volume of the oil drops. The wax layer thickness, differing in samples with untreated two-layered wax coverage and treated one-layered wax, did not significantly affect the drop behavior. These results provide strong evidence that three-dimensional plant wax coverages due to their adsorption capability are in general anti-adhesive for insects, which rely on wet adhesion.

Scientific RepoRts | 7:45483 | DOI: 10.1038/srep45483 We have previously proposed four hypotheses explaining the mechanisms of insect adhesion reduction on plant surfaces covered with three-dimensional epicuticular waxes 17 . (1) Wax projections create micro-roughness, which greatly decreases the real contact area between the plant surface and insect attachment organs called adhesive pads (roughness hypothesis). (2) Wax projections are easily detachable structures contaminating insect pads (contamination hypothesis). (3) Highly porous wax coverage may absorb the fluid secretion from the insect pad surface (fluid absorption hypothesis). (4) Insect pad secretion may dissolve epicuticular plant waxes (wax dissolving hypothesis). This would result in the appearance of a thick layer of fluid, making the plant surface slippery.
Recently, only the first two hypotheses have been experimentally tested. The effect of surface roughness on insect attachment has been revealed in a number of studies showing the worst insect attachment on rough substrates with a nominal asperity size of 0.3 and 1.0 μ m e.g. refs 18-21 corresponding to those in most wax-covered plant surfaces. Also the contamination of insect pads by the plant waxes has been detected for several insect and plant species [22][23][24][25][26][27] . It has been found that plants differ in their contaminating effects on insect pads: the contaminating ability depended on the micro-morphology of the wax projections and was related to both the largest dimension and the aspect ratio of the projections 27 .
As for the fluid adsorption hypothesis, only a very few relevant studies demonstrating the effect of the fluid reduction on the insect attachment to artificial surfaces have been performed. The essential reduction or even the loss of the attachment ability on smooth substrates after treating the pads of the bug Rhodnius prolixus with lipid solvent or after walking of the aphid Aphis fabae on silica gel for a certain time were recorded 28,29 . A significant attachment force reduction caused by a decrease of the pad secretion was measured in the stick insect Carausius morosus on smooth polyimide substrates that selectively adsorbed the watery component of the secretion 30 . Also rough nano-porous substrates highly reduced the attachment forces of the beetle Coccinella septempunctata if compared to smooth solid substrates 31 . Since the comparison of the changes in contact angles of water and oil between solid and nano-porous surfaces indicated a strong adsorption ability of the nano-porous samples for both polar and non-polar fluids, it was concluded that due to their high porosity, the nano-porous samples can adsorb the fluid from insect pads, thereby reducing insect attachment.
However, information on the adsorption ability of three-dimensional epicuticular waxes in plants is lacking in literature. The aim of the present study was to test whether the wax coverage of the waxy zone in the pitcher of the carnivorous plant Nepenthes alata is able to adsorb fluids. The wax bloom of N. alata was used here as a model plant system because it has been comprehensively studied previously in regards to its micro-morphology 26,[32][33][34] as well as physicochemical 35 and material properties 26,36 . The main functions of this pitcher surface are prey trapping and retention fulfilled through the reduction of insect attachment e.g. refs 25,26,37 and 38. Additionally, the two-layered structure of the wax coverage ( Fig. 1) made it possible to vary the porous layer thickness.
We performed experiments with the N. alata pitcher surfaces bearing either intact, two-layered wax coverage (UL) or only one, lower wax layer (LL) after the upper wax projections were mechanically removed. As in nature, the upper wax projections can be dislodged from the waxy zone by insects trying to escape from the trap, we used both UL and LL pitcher samples in this study. The artificial smooth solid sample (SS) served as a reference surface. We measured adsorption rates of two fluids -polar double-distilled water and non-polar mineral oil. Water was used to simulate the watery phase of the bi-phasic (water-oil emulsion) adhesive pad secretion reported for locusts, stick insect, ants, and flies e.g. refs 18,39-45. The oil was employed as a rough approximation of the oily pad secretion found previously in beetles e.g. refs 46-50.

Modification of the water drops.
With the only exception of the contact angles (94.50 ± 1.09° on SS, 140.76 ± 3.67° on UL, 151.60 ± 2.84° on LL) (Fig. 2a), the behavior of the water drops was very similar on all tested surfaces. We observed no changes in the drop shape during the experimental time of 30 s (see Supplementary Movie S1). The drop base usually remained the same. The drop height either did not change or was reduced by up to 1.5% of the initial value on both UL and LL or by 0.5-5.0% of the initial value on SS. The loss of the drop volume accounted for 0.5-2% of the initial value. The volume V decreased gradually and showed similar linear dependence on time t for all tested surfaces: V = − 0.001 t (n = 7) on SS, V = − 0.0013 t (n = 6) on UL and V = − 0.0008 t (n = 6) on LL (one way ANOVA: F 2,18 = 3.118, P = 0.072) (Fig. 2b).

Modification of the oil drops.
In the case of the oil, which readily wetted all test surfaces (Fig. 3, upper row), the drops behaved differently during 30 s of the experiment depending on the surface (Fig. 3). On SS ( Fig. 4 and Supplementary Movie S1), the spreading was accompanied by the continuous growth of the drop base (by 21-38% of the initial value) and continuous decrease in the drop height (by 25-40% of the initial value) (Fig. 3 (SS) and 4a,b). These processes occurred much faster at small times and were slower at large times. Some reduction in the drop volume (by 5-8.5% of the initial value) happened at three different regimes/stages: the initial stage 1 at small times (up to 3-6 s), some intermediate stage 2 and the stage 3 at large times (Fig. 4c). The relative loss of the volume V/V 0 showed the exponential character ln(V/V 0 ) = γ t with γ = − 0.0020 ± 0.0025 during the "quick" stage 1, whereas it could be described well by the power-law function ln(V/V 0 ) = α ln(t) with α = − 0.0121 ± 0.0103 during the "slow" stage 3 (Fig. 4d,e).
Cryo SEM investigations of the oil drops applied onto the pitcher wax coverage revealed the high affinity between oil and the wax material: fluid readily wetted and completely penetrated the wax coverage down to the cuticle ( Fig. 1b-g).
The oil drops showed extremely quick spreading over the UL surface during the first several seconds; then after a certain time without observable changes, they sunk in either over the same drop base or with its shrinkage (Fig. 3 (UL) and Supplementary Movie S3). The intensive spread of the drop base (by 30-220% of the initial value) occurred extremely fast during the first several seconds, then the base either almost did not change or went down (by 33-57% of the highest value) (Fig. 5a). The drop height decreased by 62-88% of the initial value exceedingly Scientific RepoRts | 7:45483 | DOI: 10.1038/srep45483 quickly at the very beginning and much slower at larger times (Fig. 5b). The great reduction in the drop volume, which accounted by 40-88% of the initial value, happened through three stages: the "quick" stage 1 at small (first 2-5 s, up to10 s) times (exponential function ln(V/V 0 ) = γ t, γ = − 0.0783 ± 0.0513), some intermediate stage 2 and the "slow" stage 3 at large times (power-law function ln(V/V 0 ) = α ln(t) with α = − 0.4525 ± 0.1993) (Fig. 5c-e).
On LL, the oil drops spread very quickly over the surface during the first several up to 10 s, then did not change in shape for a short time and afterwards started to shrink/disappear ( Fig. 3 (LL) and Supplementary Movie S4). The drop base increased extremely fast during the first few (up to maximum 10) seconds by 41-160% of the initial value, then decreased either slightly (only by 10% of the maximal value) or greatly (by 40-75% of the maximal   value) (Fig. 6a). The drop height decreased by 73-95% of the initial value: first (from several to up to 10 s) faster, then slower (Fig. 6b). The great decrease in the drop volume (by 32-94% of the initial value) occurred during the "quick" stage at small (0-10 s) times (exponential function ln(V/V 0 ) = γ t with γ = − 0.0644 ± 0.0262) and the "slow" stage at larger times (power-law function ln(V/V 0 ) = α ln(t) with α = − 0.5100 ± 0.1790) (Fig. 6c-e). The intermediate stage 2 was either not present or very short.
The comparisons of γ -values in the exponential functions (for the "quick" stage at small times) and α -values in the power-law functions (for the "slow" stage at large times) (Fig. 7) showed significant differences between surfaces (γ : H 2,22 = 12.904, P = 0.002; α : H 2,22 = 10.857, P = 0.002; both Kruskal-Wallis one way ANOVA on ranks). Both values were considerably lower on SS compared to both UL and LL and did not differ between the latter two surface samples (Table 1).

Discussion
Our experiments with the water drops on the artificial smooth solid sample (SS) and two N. alata pitcher samples with either one (LL) or two wax layers (UL) demonstrated very similar behavior of the fluid, despite of some differences in the surface wettability between the samples. Permanence of the drop shape and base as well as very little, linear decrease in the drop height and volume during the experimental time indicate that neither spreading nor adsorption of water took place on the tested samples. Little changes in the drop height and volume resulted most likely from the evaporation of the fluid. Hence, non-significant difference in k-values (V = kt) between different samples means that similar processes, apparently evaporation of water, occur at similar rates on the tested surfaces. We have found that performance of the oil drops differed essentially from that of the water drops. On all tested surfaces, distinct changes in shape of the oil drops as well as growth of the drop base together with the reduction of the drop height were clearly seen, especially at small times, and losses of the drop volume were revealed. We also detected certain dissimilarities in the drop behavior if compare SS with both UL and LL.
The continuous, but relatively moderate spreading of the oil drop over SS, which was relatively rapid at the beginning and slow at high times, resulted in the expanding drop base and the reducing drop height. Changes in both drop parameters were essentially lower than those in UL and LL. The obtained decrease in the oil drop volume on SS appeared to be higher than the expected one caused by the evaporation of oil. This happened possibly due to errors of experimental and/or calculation methods used in the study. However, taking into account great differences in the volume loss values between SS compared to both UL and LL, these errors can be neglected. As both γ -value (ln(V/V 0 ) = γ t) and α -value (ln(V/V 0 ) = α ln(t)) in SS were nearly 35 times lower than those in UL and LL, we can conclude that SS did not adsorb oil, whereas both UL and LL could do this.
For the oil drops on both UL and LL, we suggest the following scenario. During the initial stage, which was usually shorter (just several seconds) in UL and longer (up to 10 s) in LL, the liquid simultaneously quickly penetrated into the porous media located under the drop and spread over the surface. This stage contributed to the most decrease of the drop volume observed in the experiments, especially on UL. At large times (starting from ca. 10 th s after the drop deposition on the surface), mainly imbibition of the liquid to lateral directions into the porous media, which resulted in the shrinkage of the drop base (more pronounced on LL compared to UL) and the expanding of the wetted region inside the porous layer, took place. During the intermediate stage, which was noticeable in UL, the pores under the drop base became saturated and the adsorption of the fluid towards the lateral directions into the porous media started. At that time, the drop base did not expand or even slightly   Our previous study on the insect attachment to nano-porous artificial samples, showing the pronounced ability to oil adsorption, revealed extremely poor attachment of ladybird beetles Coccinella septempunctata to these substrates due to possible absorption of the secretion fluid from insect adhesive pads by the porous media 31 . In the case of epicuticular plant waxes, due to the high affinity between the wax material and oily insect pad secretion, the fluid disappears rapidly from the contact. That is why all kinds of forces contributing to the wet adhesion (capillary interactions and viscous adhesion) will be reduced or even completely eliminated. This will cause a great reduction in the attachment of insect to such substrates. Since the amount of the fluid secreted by insects is extremely small (average amount in the footprint is 0.986 μ m 3 in the beetle and 0.019 μ m 3 in the fly 52 ) and much smaller than the adsorption capability of the waxy zone in Nepenthes pitcher shown in the present paper, the fluid will be immediately and completely absorbed from the insect pad by the plant wax. The results of the present study provide strong evidence that three-dimensional epicuticular wax coverages in plants are in general anti-adhesive for insects, which rely on wet adhesion.

Methods
Samples. Three types of surface samples were used in the study: (1) the N. alata pitcher sample with one wax layer of the three-dimensional epicuticular wax coverage (LL), (2) the pitcher sample with the two-layered wax (UL), and (3) the artificial smooth solid sample (SS) served as a reference surface.
Pitchers of the carnivorous plant species N. alata Blanco (Nepenthaceae), endemic to the Philippines 53 , were obtained from plants grown in the greenhouse at the Botanical Garden of the Kiel University (Kiel, Germany). The waxy zone of the freshly harvested pitchers was used to get both UL and LL. The untreated pitcher surface bearing both wax layers 26,32,33,36 was used as UL (Fig. 1a). The LL was obtained after mechanical removal of the upper wax layer by treating the waxy zone with a two-component polyvinylsiloxane (Coltène Whaledent Dentalvertriebs GmbH, Konstanz, Germany). For this purpose, the fluid polyvinylsiloxane was applied to the pitcher surface and then peeled off after 5 min of polymerization. As SS, we used the epoxy resin sample obtained from a smooth clean glass slide by applying a two-step molding method according to Gorb 54 .

Experiments
Cryo scanning electron microscopy. Pieces of ca. 5 × 5 mm were cut out of the central part in the waxy zone of the N. alata pitcher using a razor blade. Two types of the pitcher samples were examined: (1) with the intact epicuticular wax coverage and (2) after applying a small droplet of mineral oil (light oil (neat) BioReagent M8410, Sigma-Aldrich Chemie GmbH, Munich, Germany; density 0.84 g/cm 3 at 25 °C, viscosity 20.5 mm 2 /s at 40 °C, Brookfield viscosity 30.0 cps at 25 °C) onto the wax coverage. This oil was selected, because its viscosity corresponds to that previously estimated experimentally for the footprints of beetles 55 . The total mount specimens with the inner pitcher surface facing up and fractured samples were examined in a scanning electron microscope Hitachi S-4800 (Hitachi High-Technologies Corp., Tokyo, Japan) equipped with a Gatan ALTO 2500 cryo preparation system (Gatan Inc., Abingdon, UK). Sample preparation and mounting are described in detail by Gorb and Gorb 32 and Benz et al. 33 .  35 . We applied 2 μ l drops and circle or ellipse fitting for evaluation of the apparent contact angles. On each surface, the contact angles of 10 drops were measured. Data are given as mean ± SD.

Adsorption experiments.
In the adsorption experiments, we used the sessile drops (ca. 2 μ l) of two fluids -polar double-distilled water and non-polar mineral oil (light oil (neat) BioReagent M8410, Sigma-Aldrich Chemie GmbH, Munich, Germany). The behavior of the drops during the first 30 s after the drop deposition on the test surfaces was video recorded (10 frames per second) using the contact angle measuring device mentioned above. On each sample type, 6 -9 experiments were carried out with each fluid. By applying the numerical analysis of the experimental results, described below, we obtained time-dependencies for the drop base, height and volume.
Numerical analysis of experimental data. Numerical analysis of the experimental results was organized as follows. We copied the sequence of graphic images from an experimentally recorded movie to MatLab software and transformed them into black and white maps. Numerically, these black and white maps corresponded to the matrices, which contained only 0 and 1 values. This allowed one to determine formally borders of the visualized liquid drops. As during our numerous experiments on contact angles measurements, both previous 32,35 and performed in this study, we observed mostly symmetric drops, we treated here every liquid drop as cylindrically symmetric.
In a further numerical procedure, the instant height of the drop h was divided into a large number of as small as possible layers dh«1. In fact, the number of the layers was defined by the number of pixels in the recorded HD-movie. Each layer in an instant drop image had a diameter, which was determined by the borders of its projection on the vertical plane. It can be formally done in MatLab, where a black and white image corresponded to a matrix containing 0 and 1 elements only.
Using the obtained array of the diameters for all layers, one can easily calculate the areas of the circles corresponding to each layer. In particular, the lowest (first) layer of this array defined instant contact area (drop base). One could also multiply the area of every layer by the height dh to calculate its volume and thus collecting the volumes of all the layers up to the total volume of the drop, performing a numerical summation on all of them.
The above procedure could be easily repeated for every frame of the movie. As a result, we obtained time-dependant values of the drop volume, base and height. This procedure is illustrated in the numerically generated Supplementary Movie S5. Mutual correspondence between the real image of the process, a black and white map of the vertical drop projection (with its borders marked by the colored points), and time-dependant volume, the base and height of the drop were directly seen from a comparison between different subplots of the movie frames.
In particular, it could be seen directly from the frames that different stages of the process had different rates and different (functional) time dependence. Because of this, it was also useful to plot all time-dependant values in the log-log scale. It allowed us to separate the time intervals, where the corresponding values behaved as a linear, exponential or scaling (power-law) function of time.
Determination of different functional time-dependencies in log-linear and log-log plots (recorded for the same process, shown in the Supplementary Movie S5) is illustrated in the Supplementary Movie S6. Straight lines touching the experimental curve in the first and second subplots of this movie correspond to the exponential and scaling hypothetic functional dependencies in different time intervals, respectively.