Experimental analysis of the liquid-feeding mechanism of the butterfly Pieris rapae

Butterflies mainly feed on liquids, such as floral nectar. These insects engage in active suction feeding, and a high pressure gradient is generated in their slender proboscis (Kim et al., 2011b). In previous studies, the nectar-feeding mechanism of a butterfly was investigated to understand its foraging behavior and interaction with plants (e.g. Daniel et al., 1989). In these studies, this liquid-feeding mechanism of butterflies was numerically simulated. In particular, the Hagen–Poiseuille relationship describes the volumetric intake flow rate in terms of morphological parameters (inner radius of the proboscis r; proboscis length l), fluid property (viscosity μ) and driving force (pressure gradient Δp): (May, 1985; Pivnick and McNeil, 1985; Daniel et al., 1989; Kim et al., 2011b). In this equation, the relationship between μ and feeding rate can be determined, and the optimal nectar concentrations for several butterfly species can be predicted (Daniel et al., 1989; Borrell and Krenn, 2006; Kim et al., 2011b).

In liquid feeding, butterflies use specific organs, such as the proboscis and cibarial pump, to extract nectar from flowers; however, these organs have not been investigated in detail in the context of fluid mechanics. Following the Hagen–Poiseuille equation, the flow rate is inversely proportional to the length of the proboscis. A butterfly may compensate for this disadvantage by developing a larger pump (Borrell and Krenn, 2006). The relative sizes of the food canal and the cibarial pump must be in tune to achieve the efficient nectar feeding (Kingsolver and Daniel, 1995). Butterflies should feed efficiently for successful foraging and survival. However, studies have rarely provided sufficient information regarding the strategies used by butterflies to improve liquid feeding with a long proboscis.

The purpose of the present study was to investigate the liquid-feeding strategy of butterflies. The liquid-feeding mechanism of butterflies was visualized in vivo to explain the complete process. Recently, temporal variations in intake flow at the tip of the proboscis were determined using the micro-particle image velocimetry (PIV) velocity-field measurement technique (Lee et al., 2014). The systaltic motion of the cibarial pump in the head was observed in vivo by synchrotron X-ray micro-imaging. The 3D morphological structure of the cibarial pump was also observed by synchrotron X-ray micro-computed tomography (CT). Considering these experimental observations, we developed an ellipsoidal pump model to estimate the intake flow rate in the proboscis. We also measured the flow characteristics of the intake flow using micro-PIV and synchrotron X-ray imaging techniques. Finally, we analyzed the dimensions of the cibarial pump and the proboscis in butterflies and other long-proboscid insects to determine the liquid-feeding strategy.

Fig. 1 depicts typical X-ray images of the temporal variation in volume of the cibarial pump in a liquid-feeding cycle. The volume of the pump chamber varies as a result of the vertical motion of the upper layer of the pump chamber, whereas the bottom layer remains almost fixed (Eastham and Eassa, 1955; Daniel et al., 1989). The feeding process is divided into two stages based on the movement of the upper layer: expansion and contraction (Eberhard and Krenn, 2005; Westneat et al., 2008). The top side of the pump is pulled upward dorsally to fill its chamber with liquid sucked through the proboscis during the expansion stage ranging from 0.125P to 0.750P, where P is the period of cyclic motion (Westneat et al., 2008). The inlet valve connecting the cibarial pump and the proboscis is open in this stage, whereas the outlet

valve connected to the esophagus is closed. The inlet valve is closed and the top part of the organ is pushed downward ventrally to discharge liquid food to the posterior gut through the opened outlet valve in the ejection stage ranging from 0.750P to 1.000P.

The average intake volume flow rate is 68.1±19.0 nl s⁻¹. Table 1 shows the geometric dimensions and the ensemble-averaged liquid-feeding parameters obtained from the four butterfly samples used in this study.

The planar intake flow rates () measured using the micro-PIV technique show a pulsatile phasic variation as a result of the periodic systaltic motion of the cibarial pump (Lee et al., 2014). Fig. 3A shows the temporal variation in the phase-averaged planar intake flow rate at the tip of the proboscis in one feeding cycle. The flow rate and time scale are normalized by the maximum planar flow rate () and the corresponding feeding period (Lee et al., 2014). Flow can be divided into two distinct stages, namely, the intake and ejection stages, which indicate the expansion and contraction of the pump chamber, respectively. In the intake stage, liquid is sucked into the proboscis. In the short ejection stage, the proboscis disgorges liquid outward. The amounts of the reversal flow and the total uptake flow were estimated by integrating the flow rate across the ejection stage and total duration, respectively (Fig. 3A). The amount of the reversal flow was found to be approximately 4% of the total intake flow (Lee et al., 2014).

Fig. 3B shows the variation in volume of the cibarial pump. The volume variations were ensemble averaged from four butterfly samples (Table 1). The liquid-feeding period measured by micro-PIV is 300 ms and that obtained from the synchrotron X-ray imaging technique is 365 ms. Therefore, the variation in the cibarial pump was synchronized with the corresponding flow rate at the proboscis for analysis. The time period of the intake stage is almost similar to that of the expansion stage in the cibarial pump. In the initial part of the intake stage, the flow rate is remarkably increased and then it maintains the maximum value for a while. This phenomenon indicates the linear increase in the volume of the cibarial pump in the expansion stage (Fig. 3B). Based on these observations, we hypothesized that the proboscis works as a fluid-tight tube without leakage during the liquid-feeding process. In this study, the Reynolds number (Re) based on the liquid-feeding flow inside the proboscis is ~1.40. This low Reynolds number indicates that the viscous effect is greater than the inertia force. Therefore, the curved proboscis and the slight irregularity in the conduit diameter slightly affect the liquid-feeding flow. In addition, the Dean number {D=Re√[r/2(l/π)]}, representing the centrifugal force effect of a curved channel, is ~0.092, where l is the length of the proboscis and r is the inner radius of the proboscis (Table 1). At this low value of the Dean number, the proboscis can be assumed to be a straight channel. Therefore, laminarity in flow is maintained (Kingsolver and Daniel, 1995). The Womersley number (Wo=r√(2π ρ/Tμ), where ρ is density) is one of the most important parameters in pulsatile flow. In the present study, the Womersley number is ~0.129. The unsteady force is negligible and viscous effect is dominant when the Womersley number is <1. Acceleration force is considered important in low Reynolds number flows when the time scale in velocity variation is short. The Strouhal number (St=fr/U, where f is the feeding frequency and U is the average velocity) is used to assess acceleration force; the acceleration force is negligible when the Strouhal number is <1 (Kingsolver and Daniel, 1995). In the present study, the Strouhal number is ~7.55×10⁻³. These non-dimensional parameter results indicate that the liquid-feeding flow inside the food canal is a quasi-steady laminar flow. Therefore, the time-averaged flow inside the food canal is likely in accordance with the parabolic Hagen–Poiseuille relationship, although a short reverse flow is observed in the ejection stage (Lee et al., 2009).

In this study, the liquid-feeding phenomena of butterflies were investigated experimentally using the synchrotron X-ray imaging and micro-PIV velocity field measurement techniques. These advanced imaging techniques have been recognized as effective for studying liquid-feeding insects (Westneat et al., 2008; Kikuchi and Mochizuki, 2011; Kim et al., 2011a; Monaenkova et al., 2012). We analyzed the dynamic motion of the cibarial pump and the corresponding ellipsoidal model. In addition, the feeding flow characteristics were determined to verify the assumption of a simple tube and Hagen–Poiseuille flow. Finally, we analyzed the liquid-feeding strategy from a biomechanical viewpoint.

There have been a few studies on the function of the cibarial pump. Eberhard and Krenn (Eberhard and Krenn, 2005) reported the systaltic operation of the cibarial pump based on functional anatomy. They mentioned that the liquid-feeding cycle is composed of two phases: suction and swallow phases related to expansion and contraction of the cibarial pump, respectively. Westneat et al. (Westneat et al., 2008) observed the cibarial pump of a living butterfly using the synchrotron X-ray imaging technique. The results of these studies agree with those of the present study, although there is not detailed information about the volume and time scale of the previous studies. In the present study, the dynamic behaviors of the cibarial pump were observed consecutively during the liquid-feeding process. The liquid-feeding process consists of two stages: expansion and contraction. In the expansion stage, the cibarial pump is expanded linearly to the maximum volume when the cavity is filled with liquid from the proboscis. In the contraction stage, the pump is contracted rapidly to transport the liquid to the esophagus. The duration of the expansion stage is approximately 2.7 times longer than that of the contraction stage. It seems that the pump muscles are well developed for the contraction (Eastham and Eassa, 1955). In the present study, the test butterfly did not empty its cibarial pump completely after the contraction. The volume of the retained liquid after the contraction (V_min) was 22.4±2.5 nl, ~48% of the maximum pump volume (V_max). If the cibarial pump is completely emptied, discontinuity occurs in the main flow coming from the proboscis toward the esophagus. It seems that butterflies retain some amount of liquid for the next cycle. However, this aspect has not yet been studied in detail in the context of myology.

In a previous study on the modeling of the cibarial pump, Daniel et al. (Daniel et al., 1989) developed a cylindrical model based on anatomical observation. Their model assumed that the cylindrical pump has constant radius and the expansion state is caused by a change in height, somewhat similar to the model established in the present study. However, our elliptical model was based on the dynamic motions of live butterflies observed using a synchrotron X-ray imaging system.

It is difficult to estimate precisely the mechanism of the inlet and outlet valves of the pump from the X-ray images in this study. Therefore, we studied the opening and closing operation of the valves based on the boundary images of the cibarial pump (Fig. 1). The inlet valve located in front of the pump did not perfectly seal the discharged flow during the contraction phase, because a slight reversal flow occurred (Fig. 3A). However, the effect of imperfect sealing seems to be nearly negligible, because the reversal flow in the ejection stage was only 4% of the total intake flow (Lee et al., 2014). Therefore, the intake flow in the cibarial pumps is relatively well regulated by the valves located at the inlet and outlet of the pump system.

The flow around the tip of the proboscis and the volume change of the cibarial pump are similar in terms of time scale (Fig. 3). This implies that the expansion and contraction of the cibarial pump induce the intake and ejection stages, respectively. The intake flow exhibits large standard deviations at the beginning of the intake stage (Fig. 3A). This results from the dramatically increased intake flow rate in the initial stage. The initial rapid expansion seems to induce large fluctuations in the intake flow rate, although the standard deviations in the expansion stage are small (Fig. 3B). Likewise, the standard deviations of the volume in the contraction stage are relatively large because the volume decreased rapidly. However, these large standard deviations do not affect on the reversal flow in the ejection stage because it is regulated by inlet valve.

The liquid-feeding flow inside the proboscis has been considered as a quasi-steady state, laminar flow, because the Reynolds, Womersley and Strouhal numbers are estimated to be very small (Kingsolver and Daniel, 1995; Lee et al., 2009; Lee et al., 2014). In this study, we obtained those governing parameters under in vivo condition.

The liquid-feeding strategy of butterflies was investigated based on the Hagen–Poiseuille relationship in the present study. We derived the relationship between the proboscis and the cibarial pump in terms of energy consumption. Minimum energy consumption may occur when the volume of the cibarial pump is proportional to the cube of the radius of the proboscis, r³~V_cp. The liquid-feeding organs of the test butterfly, P. rapae, seem to bear this cube relationship. We checked other insects having a cibarial pump with a long movable proboscis (Fig. 4). These insects are considered to have the same liquid-feeding mechanism as that of the butterfly in the viewpoint of fluid mechanics (Kim and Bush, 2012). These insects also follow the cube relationship. The present study thus provides insights into the liquid-feeding systems of butterflies and other long-proboscid insects. In addition, the experimental methodology employed in this study will be helpful for investigating the biomechanical aspects of liquid-feeding insects.

In this study, the cabbage white butterfly [Pieris rapae (Linnaeus 1758)] was selected as a test subject for observing liquid-feeding phenomena. It is the most common flower-visiting butterfly in the Republic of Korea. It has a well-developed proboscis and a cibarial pump for nectar feeding (Eastham and Eassa, 1955). Adult butterflies and pupae were bought from a merchant in Goyang, Republic of Korea. These butterflies were reared at 27°C with 80% relative humidity under 16 h:8 h light:dark condition. The butterflies were fed with artificial nectar using a cotton rod soaked in 10% sucrose. In this experiment, the butterflies were evaluated at 4 days post-emergence (Lee et al., 2009; Lee et al., 2014; Kim et al., 2011a; Kim et al., 2013).

The proboscis is typically coiled under a butterfly's head in a resting state. Therefore, the coiled proboscis was manually straightened using a slender needle to imitate the posture of liquid feeding. The straightened proboscis was then fixed using sticky tape. However, internal flow is hardly observed because the proboscis is opaque. As such, the uptake flow around the tip of the proboscis was measured using the micro-PIV technique. The straightened proboscis was then immersed in a test liquid to induce water intake (Lee et al., 2014).

The wings of live butterflies were laterally fixed on a two-axis stage to visualize the systaltic motion of the pump organs using the synchrotron X-ray imaging technique (Fig. 6A).

A test butterfly was decapitated, and the head was attached to the sample holder using instant glue to obtain the 3D morphological structures of the pump organs by synchrotron X-ray CT. After the test sample was mounted, the sample holder was placed in liquid nitrogen for a few seconds to freeze the head. The sample holder was then placed on a rotating stage, and cold nitrogen gas was supplied continuously using a cryojet to avoid dehydration after the sample was exposed to an X-ray beam (Kim et al., 2012).

A micro-PIV system was used to determine the characteristics of the temporal variations in intake flow during the liquid-feeding process. The micro-PIV system comprises a vertical epi-fluorescence microscope (Eclipse 50i, Nikon, Tokyo, Japan), a continuous Nd:YAG laser (λ=532 nm, SLOC, Shanghai, China) as a light source, a high-speed CMOS camera (Photron ultima APX, Fujimi, Tokyo, Japan), and a personal computer for data processing and control (Lee et al., 2009; Lee et al., 2014; Kim et al., 2011a; Kim et al., 2013). Fluorescent particles (Molecular Probes, Eugene, OR, USA) with a mean diameter of ~1.0 μm were mixed with DI water (~2.5×10⁹ beads ml⁻¹) (Lee et al., 2014). These particles emit fluorescence at a peak wavelength of 554 nm after absorbing green light at a peak wavelength of 534 nm. The fluorescence images of the tracer particles were captured using a high-speed camera after passing an optical long-pass filter (λ=550 nm). Clear fluorescent-particle images were obtained from the thin measurement plane by adjusting the depth-of-focus of the microscope (Meinhart et al., 1999; Lee et al., 2009). A 10× objective lens with a numerical aperture of 0.25 was attached to the microscope to magnify the flow images. These flow images were recorded consecutively at a frame rate of 500 frames s⁻¹. A fast Fourier transform-based cross-correlation PIV algorithm was used to obtain instantaneous velocity field information from two consecutive flow images (Lee et al., 2009; Lee et al., 2014). Typical velocity vector fields obtained at the intake and ejection stages are depicted in Fig. 5. The intake flow rate () was evaluated by calculating the planar flow rate (mm² s⁻¹), which was obtained by integrating the velocity data along the borderline 100 μm away from the tip of the proboscis (Lee et al., 2014). The intake flow rates of five feeding cycle at the same phase angle were ensemble-averaged (Fig. 3A).

The cibarial pump of a butterfly was visualized in vivo using the synchrotron X-ray micro-imaging system of the Pohang Accelerator Laboratory (PAL; Pohang, Republic of Korea). The 6C biomedical imaging (BMI) beamline provides monochromatic X-rays with energy at 10–60 keV (Kim et al., 2011a). The transmitting X-rays were converted to visible light by use of a scintillation crystal of CdWO₄, which was then reflected 90 deg by a gold-coated mirror and registered by a CCD camera (MegaPlusII ES2001, Redlake, Tallahassee, FL, USA) at a frame rate of 24 frames s⁻¹. The test butterflies ingested a diluted iopamidol solution (600 mg ml⁻¹; Fig. 6A). Iopamidol solution was used as a contrast agent in this study because the boundary of the pump chamber may not be clearly detected; this result could be attributed to the complicated anatomical structure of the cibarial pump chamber, which is covered with layers of muscles (Kim et al., 2011a). Fig. 1 shows typical X-ray images showing the volume variation of the cibarial pump in the corresponding feeding cycle (0–1P). The light intensity in the cibarial pump filled with the contrast medium was lower than that in the surrounding regions. Therefore, the boundary of the pump chamber is defined at the area where the intensity was remarkably changed (Fig. 1). ImageJ software was used to determine the edges in the captured X-ray images.

X-ray CT was applied to obtain the 3D morphological structure of the pump system of a butterfly. Fig. 6B shows a schematic of the synchrotron X-ray micro-CT system established at PAL. The test butterfly was decapitated, and the head was fixed on a rotating stage. The sample attached on the stage was rotated from 0 to 180 deg at intervals of 0.5 deg. A total of 361 X-ray images were obtained using a CCD camera (VM-11M5, Vieworks, Gyeonggi-do, Republic of Korea). The 3D structure of the sample was numerically reconstructed from the projection slice images using the Octopus (in CT, Ghent, Belgium) image calculation program. Amira image analysis software (Visualization Sciences Group, Burlington, MA, USA) was used to analyze the 3D reconstructed images according to a previously described method (Kim et al., 2012).

Synchrotron X-ray imaging experiments were conducted at the 6C BMI beamline of the Pohang Accelerator Laboratory (PAL), Pohang, Repulic of Korea.