Two sniffing strategies in palinurid lobsters

Olfaction in a fluid environment can be regarded as a two-stage process in which fluid convection initially carries odor molecules to the region of a sensor and diffusion ultimately transports them to a receptor surface. Organisms have evolved different behavioral mechanisms to facilitate the convective stage of the process. For example, mammals sniff to transport scents over their olfactory surfaces. Snakes vibrate their tongues to move odors over their vomeronasal pouches(Halpern and Kubie, 1980). Male silkworm moths fan their wings to pass pheromones over their antennae(Loudon and Koehl, 2000). Fish flush water through olfactory organs(Døving et al., 1977; Kux et al., 1988). Malacostracan crustaceans, such as crabs, stomatopods and lobsters, flick their antennules to generate convective flows(Goldman and Koehl, 2001; Mead et al., 1999; Moore et al., 1991a; Schmitt and Ache, 1979; Snow, 1973).

Lobsters use their antennules to track the chemical signatures of food and conspecifics (Fig. 1A; Atema, 1995; Atema and Voigt, 1995; Derby and Atema, 1988; Derby et al., 2001; Ratchford and Eggleston, 1998; Reeder and Ache, 1980; Zimmer-Faust et al., 1985). The lateral flagellum of the antennule has rows of specialized chemosensory setae, the aesthetascs (Fig. 1B,C), in addition to other mechano- and chemosensory setae(Cate and Derby, 2001; Grünert and Ache, 1988; Laverack, 1964; Steullet et al., 2002). Lobsters `sniff' by flicking the lateral flagellum through the water(Moore et al., 1991a; Schmitt and Ache, 1979);during the fast closing phase, water flows through the chemosensory aesthetascs, and during the slower opening phase, water is not replaced(Goldman and Koehl, 2001; Koehl et al., 2001). It has been suggested that the generation of such intermittent flows by antennules results in discrete sampling of the surrounding fluid(Moore et al., 1991a; Schmitt and Ache, 1979). Indeed, Koehl et al. (2001)showed empirically that antennule flicking allows spiny lobsters to take discrete samples of the temporal/spatial distribution of fine scale chemical plumes in the environment.

Lobsters locate odor sources by using their antennules to sample an odor plume's structure (Atema, 1996;Moore and Atema, 1988, 1991; Moore et al., 1991b; Zimmer-Faust et al., 1995;reviewed in Weissburg, 2000; Atema, 1995). Within an odor plume, small filaments with high odor concentrations spread from the odor source. When crustaceans flick their antennules, they sample sharp concentration gradients of the plume filaments. By tracking the spatial and temporal distribution of these filaments in the environment, as well as other environmental cues (e.g. current flow; Zimmer-Faust et al., 1995),they can localize the source of an odor.

The amount of fluid that passes through rather than around a lobster's aesthetasc array is determined by the array's leakiness. In an array of cylinders, when the boundary layers around two adjacent cylinders are large relative to the distance between them, some proportion of the incident fluid will be forced around rather than through the array. When the boundary layers are relatively small, more fluid will leak through the array. Thus, the leakiness of an array depends on (1) Re, which determines boundary layer thickness, and (2) the relative size of the gaps between adjacent cylinders, which is expressed as the ratio of the gap to the cylinder diameter(Koehl, 1995, 2000). Quantitatively,leakiness is the volume of fluid that flows through the gap between cylinders in an array divided by the volume of fluid that would flow through the same area had the cylinders been absent (Cheer and Koehl, 1987b).

Across species, changes in velocity and setal spacing can modify the transition point between leaky and non-leaky movements(Koehl, 1995). In a copepod, Centropages furcatus, the setae on the second maxillae are widely spaced and typically function as a rake (capturing particles as water flows through the structure). When maxillae velocity is decreased, the appendage acts like a paddle (most of the particle-laden water flows around rather than through the structure). By contrast, the copepod Temora stylifera,has more closely spaced setae and the maxillae normally function as paddles,but, when velocity is increased, the maxillae act as rakes. These different strategies illuminate the importance of both morphological and behavioral modifications for changing the fluid dynamics of sensory appendages across species (Koehl, 1995). An organism that modifies the speed of its sensory appendage and/or the relative size of the gaps between its sensory hairs can adjust the way it samples chemical information in the environment.

Studies of chemoreception in aquatic crustaceans have focused on a broad array of taxa, including spiny lobsters (Palinuridae: Derby and Atema, 1988; Grünert and Ache, 1988; Reeder and Ache, 1980), clawed lobsters (Nephropidae: Atema,1995; Atema and Voigt,1995), crayfish (Cambaridae: Dunham et al., 1997; Moore and Grills, 1999; Oh and Dunham, 1991) and crabs(Portunidae: Weissburg and Zimmer-Faust, 1993, 1994). Most studies of lobster chemoreception have focused on Panulirus argus (Palinuridae) and Homarus americanus (Nephropidae), with few studies comparing species within these families (Cate and Derby,2002), although some studies have compared these model species to other crustaceans (e.g. Ghiradella et al.,1968; Laverack,1985).

In the present study, we compare Panulirus argus (Caribbean spiny lobster) to Palinurus elephas (common spiny lobster) to determine how major structural differences in chemoreceptor organs influence the capture of odor molecules from the environment and to assess the relevance of the P. argus model to other palinurid lobsters. (Note: these species are in different genera with similar names.) We describe the morphology and kinematics of the antennular flagella in P. elephas through the use of scanning electron microscopy, high-speed videography and a new method of digital video analysis. We compare Re and leakiness parameters to those previously described in P. argus and propose a functional interpretation for the two dramatically different strategies used by these species in acquiring chemical information. We measure antennule structures across all major genera in the Palinuridae and estimate the amount of water sampled by different antennule sizes. To our knowledge, this is the first time that the antennular kinematics of a palinurid species other than P. argus has been studied. Furthermore, this comparative study offers a first step towards understanding the behavioral function of the remarkable antennule variation across the palinurid family.

Palinurus elephas (Decapoda: Palinuridae) specimens were collected by Cleggan Lobster Fisheries Ltd (Galway, Ireland) and shipped to Duke University. They were housed in filtered, recirculating, synthetic seawater at 10-11°C in 2271 glass aquaria.

The microscopic anatomy of the flagella was examined using stereo light microscopy and scanning electron microscopy (SEM). We removed the antennules from freshly euthanized lobsters. The tissue was immediately placed in 2%glutaraldehyde in phosphate-buffered saline and fixed for 1.5 h [see Dykstra(1992)]. The tissue was rinsed in distilled water and dehydrated in an ethanol series. The specimens were stored in 100% ethanol until critical point drying and then were sputter coated (60:40 gold:paladium mix, Anatech Hummer V, Anatech Ltd, Springfield,VA, USA) and observed at up to 2500× magnification with a scanning electron microscope (Philips 501 SEM, Oregon, USA). The lateral and medial flagella were cut into four and two smaller sections, respectively, in order to mount on stubs in the scanning electron microscope. The guard hairs were removed in some samples in order to view the aesthetascs.

Aesthetasc diameter and the gap distance between rows of aesthetascs were measured from digital images captured from the SEM (Scion Image, v.4.0.2,Maryland, USA). Aesthetasc diameter and gap distance between each aesthetasc were measured at the base of the aesthetascs near their insertion on the flagellum. Average measurement error was 2.3% of the mean. We sampled ten aesthetascs and ten gaps from four individuals. Digital SEM images of a calibration disk (Ted Pella, Inc., Redding, CA, USA) were used for calibrating images of antennules.

We measured lengths of the peduncle segments and flagella of 17 lobster species using preserved specimens housed at the National Museum of Natural History, Smithsonian Institution, Washington, DC, USA. Each major genus of the Palinuridae was represented, as well as members of the Nephropidae(Homarus americanus) and the Scyllaridae (Parribacus antarcticus,Scyllarus arctus). Body size was approximated as carapace length (base of the rostrum to the posterior margin of the carapace). Two specimens per species were measured when possible.

A high-speed video system (HR1000, Redlake Motionscope Systems, San Diego,CA, USA) recorded images of the P. elephas antennules as they flicked. The system's tripod-mounted camera recorded images of the two flagella on each antennule. Images of antennule flicking in which the lateral flagellum was perpendicular to the camera's lens were collected at 500 frames s^-1 and stored on SVHS videotape. Later, video fields were digitized and stored as sequences of individual bitmap images (TIFF format).

We designed an automated method that does not rely on traditional motion analysis techniques to track flagellum movement. Video images could not resolve natural markings on the flagella and we were unable to use physical markers because they either damaged the delicate antennules or were removed by the animal. To circumvent these problems, we quantified the motion of the antennule's lateral flagellum by tracking the intersection between the edge of the flagellum and a line fixed in the image plane(Fig. 2A). The line intersected the distal end of the aesthetasc-bearing region of the flagellum and ran tangent to the arc traced by a point on the flagellum when it moved. For the small angular excursions subtended by the flagella during a flick(approximately 5°), the maximum deviation between the tangent line and the circular path of a point on the flagellum was approximately 0.1% of the diameter and thus was neglected.

Image processing consisted of: (1) filtering, (2) sampling pixel intensity along the tangent line and (3) edge detection. First, an adaptive Wiener algorithm acted as a low-pass filter to reduce noise in the image [local means and standard deviations were calculated for each pixel using an area set at 5 pixel × 5 pixel window (Matlab v. 6, The Mathworks Inc., Natick, MA,USA)]. Second, a tangent line was drawn on the image for each flick sequence using a software graphical interface. Image-processing software sampled the intensity of each pixel along the line for each image in a sequence. Third,the edge was detected along the tangent line at locations where the intensity of the pixels changed from light (background) to dark (flagellum). We tracked the position of the intensity transition over the course of each flick(Fig. 2B).

We determined the position of the transition from background to flagellum by comparing the intensity difference between the end pixels of a 3-pixel wide window passed along the tangent line (Fig. 2). The position of the transition depended on a threshold value of intensity difference. Within a range of threshold values, our method detected the same edge but located it at slightly different positions. Outside of the threshold-value range, the method either detected an edge in the background (resulting from noise in the pixel intensity of the background) or did not detect an edge at all. Rather than choose a single threshold value arbitrarily, we determined a range of threshold values that properly detected all of the edges in a flick sequence and used the mean edge position determined over that range of threshold values. The standard deviation of edge position measurements over the range of threshold values never exceeded 3% of the mean.

To track the position of the flagellum over time we first filtered the raw position data to remove high-frequency noise and then calculated the speed. The filtering algorithm consisted of four steps: (1) filtering the data with a second-order low-pass Butterworth filter, (2) reversing, with respect to time,the output, (3) filtering with the same filter again and (4) reversing, with respect to time, the output of the second filter(Winter, 1990). Re-filtering of the reversed data introduced a phase shift equal and opposite to that introduced by the first filter, and thus the algorithm resulted in filtered data in phase with the raw data. Residual analysis revealed that a cutoff frequency of 100 Hz balanced the amount of noise passed through the filter with the amount of signal distortion introduced by the filter(Winter, 1990). We applied a numeric differentiation technique to calculate the speed of the flagellum. The speed at a time point was estimated to be the distance moved by the flagellum divided by the time increment between two video frames.

We converted kinematic measurements from pixel units to SI units by measuring the pixel dimensions of a known antennule structure in the image using Scion Image (the actual size previously was measured on the live specimens). Imprecision, expressed as the standard deviation of repeated measures of the same structure, was typically 5% of the mean within an image and 3% of the mean between different images within a time sequence.

In assigning the variables used to calculate Re (equation 1), we recognized that to an observer fixed in still fluid during a flick, the flagellum's speed equals the freestream flow speed in the flagellum's reference frame. Thus, we followed convention and chose U as either the peak or mean speed of the lateral flagellum during each phase of a flick(Goldman and Koehl, 2001; Mead et al., 1999). Because a flicking flagellum rotates about a fixed point, the freestream flow, and thus Re, will vary linearly along the flagellum's length. We report speeds and Re values encountered at the distal end of the aesthetasc-bearing region of the antennule. Aesthetasc diameter is used as the length scale(L) to be consistent with both empirical studies of flow through arrays of cylinders and studies of other biological sensors that bear hair-like sensilla (Goldman and Koehl,2001; Hansen and Tiselius,1992; Koehl, 1993, 1996, 2000; Mead et al., 1999). The kinematic viscosity (ν) of seawater (35‰) at 10°C is 1.36×10^-6 m² s^-1(Sverdrup et al., 1942; Vogel, 1994).

We tested whether the following kinematics parameters varied between the closing and opening phases of each flick: Re, maximum and average flagellum speed, duration and distance moved (excursion). For duration measurements, 26-39 flicks from each of four individuals were measured. For all other kinematic measurements, 11-39 flicks from each of four individuals were measured. Data were tested for normality using a Shapiro—Wilk test(Zar, 1999). As the data did not conform to a normal distribution, differences between flick phases across individuals were tested using a nonparametric Kruskal—Wallis test(Zar, 1999). Significance level was set at P=0.01 to account for multiple comparisons(Sokal and Rohlf, 1981).

On the lateral flagellum of Palinurus elephas, the ventro-laterally projecting aesthetascs form paired rows on each annulus of the distal half of the flagellum (Fig. 3). The tips of the aesthetascs form a diffuse cluster, which lacks the zigzag pattern described in Panulirus argus(Fig. 1C; Goldman and Koehl, 2001). The number of aesthetascs per row decreases as the flagellum tapers to a thin,flexible point. Dimensions of the aesthetascs and their spacing on the flagellum are listed in Table 1. Asymmetric setae are found on each annulus immediately adjacent to the medial edge of the aesthetasc rows(Fig. 3C). Dwarfing these setae are guard setae that flank each pair of aesthetasc rows and form a canopy over the aesthetascs (Fig. 3A). Guard hairs on the lateral side of the aesthetasc tuft are plumose. Outside the guard hairs are pairs of companion setae(Cate and Derby, 2001; Fig. 3A-C). The companion setae found on the lateral side of the aesthetasc rows are simple(Fig. 3A); however, the companion setae on the medial side are plumose(Fig. 3B) [similar to the pattern found in the Red rock lobster, Jasus edwarsii(Cate and Derby, 2002)].

An array of plumose setae extends along the length of the medial flagellum(Fig. 3E,F). These setae extend ventro-laterally and remain ventral to the lateral flagellum during a flick(Fig. 4). An array of long,simple setae extends along the dorsal surface of the flagellum(Fig. 4). The long axis of the simple setae extends dorso-ventrally, whereas the plumose setae run perpendicular to the lateral flagellum's flick (Figs 3, 4).

Panulirus species have considerably longer flagella per body length than any other measured species, with the medial flagellum extending to up to three times the carapace length(Table 2; Fig. 5). The medial flagellum is longer than the lateral flagellum in all the species. The total peduncle length is relatively constant across taxa, although Justitia longimanus (West Indian furrow lobster) has a relatively high total peduncle length. Excepting Parribacus spp. and Scyllarusspp., the first segment of the peduncle is longer than the subsequent segments(Fig. 5; Table 2). The second and third segments of the peduncle are similar to each other in length. Lateral flagellum length varies dramatically across the lobsters(Fig. 6A). By contrast, most variation in total peduncle length appears to be explained by body size across taxa (Fig. 6B; Table 2).

The flick of the lateral flagellum of a P. elephas antennule is a short reciprocal motion during which the flagellum executes a ventral excursion (closing phase) followed by an immediate return to or near the initiation point (opening phase) (Fig. 4). The medial flagellum is not actively moved during a flick but it does flex slightly in passive response to the impulsive motion of the lateral flagellum. Flicks were often executed in rapid succession with interspersed periods of quiescence (Fig. 7).

We found significant differences between flick phases in all measured parameters except excursion distance (Table 3). Individuals differed significantly (P≪0.01, N=4) in all parameters except flick duration, so flick phase comparisons were calculated within each individual. Both the peak and the mean speed of the initial closing phase were approximately twice those of the opening phase of the flick (Fig. 7; Table 3). Furthermore, as the distance traveled by the flagellum was nearly equal in both phases of the flick, the duration of the flick phases also differed by a factor of approximately two (Fig. 7; Table 3). The Re of an aesthetasc during both phases of a flick is in the order of one. During the closing phase, the aesthetasc Re is approximately twice the aesthetasc Re during the opening phase(Table 3).

Palinurus elephas and Panulirus argus have strikingly different antennule structures and kinematics yet, in spite of these differences, both species sample the fluid environment in a fundamentally similar way. We begin by discussing our findings in P. elephas and compare them with published studies of P. argus. Second, we make a broader comparison across palinurid genera and consider how the two flicking strategies in these species could influence the acquisition of spatial and temporal information contained in odor plumes.

P. elephas sniffs by flusing more water through the aesthetasc array during the relatively rapid closing phase of each flick than during the slower opening phase. By flicking its lateral flagella such that the aesthetasc tuft operates at Re values of approximately one, the leakiness of the aesthetasc array is sensitive to small changes in Re. If the aesthetasc array operated in an Re range that was an order of magnitude higher or lower, small Re changes would not affect leakiness: at Re values of approximately one, small differences in the speed of flick phases cause relatively dramatic differences in the leakiness of the aesthetasc array. If we apply the leakiness estimates generated by Hansen and Tiselius(1992; pp. 826-827) to the present study, we find that the leakiness of P. elephas aesthetasc arrays averages approximately 0.40 (0.70 peak) during the closing phase and 0.25 (0.30 peak) during the opening phase of flicks. Hence, P. elephas sniffs by varying the flagellum's speed between the two flick phases, such that the closing phase is 60-130% more leaky than the opening phase.

Considering that mammals, snakes and other crustaceans discretely sample odors, it comes as no particular surprise that P. elephas also does so. It is surprising, however, that none of the factors that determine Re — flow speed, aesthetasc diameter and kinematic viscosity— are the same in P. elephas and P. argus, yet the mean Re of the closing phase is the same in each species and the mean Re of the opening phase is remarkably similar(Table 4). Both species operate their antennules in the Re range where leakiness is sensitive to the speed of the flick. If the flagella of P. elephas were merely geometrically scaled up to be the same length as those of P. argusand the angular velocity held constant, the relative spacing of the aesthetascs would remain constant but the Re range would cause both phases of the flick to be very leaky and the animal could not take discrete odor samples. On the other hand, if the flagella of P. argus were simply scaled down to the same length as P. elephas, neither flick phase would allow much fluid to enter the aesthetasc array and diffusion would move molecules from the fluid to the surface of the sensilla [although see Trapido-Rosenthal et al.(1987) for a discussion of enzyme-mediated chemoreception at sensillar surfaces]. Thus, in spite of both morphological and kinematic differences, P. elephas and P. argus discretely sample the fluid by flushing it through their aesthetasc arrays during the fast, leaky closing phase of a flick and retaining that sample of fluid during the slower, less leaky opening phase(Koehl, 2000; Goldman and Koehl, 2001; Koehl et al., 2001).

Even though P. elephas and P. argus use the same mechanism to discretely sample odors, the spatial and temporal characteristics of the water samples are distinctly different. The duration of each flick phase is nearly an order of magnitude lower in P. elephas than in P. argus (Table 4),hence the two species have vastly different flicking frequencies. P. argus flick at a frequency between 0.4 Hz and 1.5 Hz and increase flicking rates up to 3.5 Hz in the presence of food scent(Gleeson et al., 1993; Goldman and Koehl, 2001). We found that P. elephas flick at approximately 20 Hz; in one case this rate was sustained for over 0.8 s. In addition, the lateral flagella of P. argus are nearly seven times longer and the distal end of each flagellum travels four to five times further during a flick than do the lateral flagella of P. elephas (Tables 3, 4; Figs 5, 6).

Differences in the setae on the medial and lateral flagella also distinguish P. elephas and P. argus. Non-aesthetasc setae are important in acquiring mechanical and chemical information, complementary to that acquired with the aesthetascs(Steullet et al., 2001). On the lateral flagellum, P. argus has simple companion setae, whereas P. elephas and other palinurids have plumose companion setae (pc, Fig. 3; Cate and Derby, 2001, 2002). The function of these companion setae has yet to be determined, but the absence/loss of setules in Panulirus species suggests a change in mechanical sensitivity through modification of flow around the setae. On the medial flagellum, the plumose setae of P. elephas appear as though they would influence flow into the aesthetasc array during the closing phase of the flick(Fig. 4), but, at these Re values, motion near surfaces does not alter the leakiness of cylinder arrays (Loudon et al.,1994). P. elephas lacks the peculiar zigzag orientation of aesthetasc tips found in P. argus, which has been proposed to channel fluid between neighboring aesthetascs and thereby facilitate diffusion across the laminar region surrounding aesthetascs(Fig. 1C; Gleeson et al., 1993; Goldman and Koehl, 2001). In P. argus, fixation of the antennules can disrupt the zigzag arrangement of the aesthetasc tips; thus, the clustering of aesthetasc tips towards the center of rows in P. elephas may be an artifact of the glutaraldehyde and alcohol tissue preservation applied to the flagella.

Both antennular flagellum length and total peduncle length vary across the Palinuridae (Table 2; Figs 5, 6). Peduncle length is relatively short in the nephropid lobsters, however most variability across the palinurids can be explained by body size (Figs 5, 6). Antennular flagellum length, by contrast, is highly variable across the palinurids.

Flicking antennules collect both spatial and temporal information about a plume's structure (reviewed in Crimaldi et al., 2002; Weissburg,2000). Models by Crimaldi et al.(2002) show that flicking (1)increases the number of concentration peaks (plume filaments) sampled per unit time, (2) increases the probability of sampling a high concentration plume filament and (3) permits two-dimensional sampling of the spatial structure of the plume. The fast response and integration times of both Homarus americanus (Nephropidae) and Panulirus argus (Palinuridae)receptor cells suggest that information about odor concentration and filament encounter can be acquired during and across flicks(Fadool et al., 1993; Gomez and Atema, 1996a,b;gomez et al., 1994a,b, 1999; Moore and Atema, 1988; Schmiedel-Jakob et al.,1989).

P. argus, with its long antennules and slow flick rates, and P. elephas, with its short antennules and fast flick rates, use two different flicking strategies to sample the environment. Higher flick rates permit P. elephas to acquire shorter temporal samples of the environment than P. argus and, up to the temporal limits of the sensory receptors, can resolve shorter temporal fluctuations of filament concentrations as they move relative to the animal. Shorter antennules permit P. elephas to sample smaller areas of the plume than P. argus (Fig. 8) and can potentially resolve smaller spatial scale variations in plume filaments. P. argus samples a larger area of the plume per flick, which increases the probability of encountering an odor filament, but at the expense of small scale spatial resolution of the plume. By contrast, P. elephas samples small areas of water over shorter time scales and thus has higher spatial resolution of the plume compared with P. argus. However, if we take into account both (1) the probability of encountering an odor filament in the sampled area of a flick (high in P. argus and low in P. elephas) and (2) the probability of odor encounter based on the rate of flicking (low in P. argus and high in P. elephas), it remains to be determined whether these two species actually differ in the probability of encountering odor filaments.

Clearly, many factors influence evolutionary diversification of antennules in spiny lobsters. Spiny lobsters inhabit a wide range of habitats, from shallow reefs to deep-sea mud flats(George and Main, 1967; Holthuis, 1991; Kanciruk, 1980; Lozano-Alvarez and Biornes-Fourzan,2001; Sharp et al.,1997), and exhibit a range of social behaviors across species and in the course of development (Childress and Herrnkind, 1996; Phillips et al., 1980). Odors play a central role in social behavior, and detection is limited by concentration(Childress and Herrnkind,2001; Ratchford and Eggleston,1998). While little is known of social and food searching behavior in P. elephas (Hunter,1999), other spiny lobsters have been shown to rely on odor to locate food (Zimmer-Faust and Case,1983; Zimmer-Faust et al.,1985), with Panulirus species foraging exceptionally long distances compared with other palinurids(Butler et al., 1999; MacDiarmid et al., 1991). The interplay between fluid dynamics, size, habitat and behavior is complex but critical to informing our understanding of evolutionary changes in antennule morphology. Perhaps the most fundamental conclusion to be drawn from the present study is that, with their extraordinarily long flagella, Panulirus is the unusual genus in palinurid lobsters. While P. argus continues to be an informative model system, the next step is to examine the evolutionary history and current function of these two palinurid flicking strategies.

D. Schmitt and the Duke University Animal Locomotion laboratory (NSF No. SBR-9904-401) generously loaned their high-speed video system. We thank M. Koehl and K. Mead for valuable discussions. J. Birch, N. Danos, M. McHenry and D. Simon provided assistance with data analysis and implementation. We thank L. Eibest and the Duke University Scanning Electron Microscopy laboratory. This manuscript benefited from the comments of two anonymous reviewers. Funding was provided to S. N. P. by the Miller Institute for Basic Research in Science and the National Science Foundation DIG No. 9972597.