Functional Foveae in an Electrosensory System

Transcription

1 THE JOURNAL OF COMPARATIVE NEUROLOGY 511: (2008) Functional Foveae in an Electrosensory System JOAO BACELO, 1 * JACOB ENGELMANN, 2 MICHAEL HOLLMANN, 2 GERHARD VON DER EMDE, 2 AND KIRSTY GRANT 1 1 Unité de Neurosciences Intégratives et Computationnelles, C.N.R.S., 1 Avenue de la Terrasse, Gif sur Yvette, France 2 Institute of Zoology, Department of Neuroethology, University of Bonn, Bonn, Germany ABSTRACT Several species of Mormyrid weakly electric fish have a mobile chin protuberance that serves as a mobile antenna during prey detection, tracking behaviors, and foraging for food. It has been proposed that it constitutes a fovea of the electrosensory system. The distribution of the three types of receptor organs involved in active imaging of the local surroundings, prey detection, and passive electroreception, and their central projection to the electrosensory lobe (ELL), have been studied in Gnathonemus petersii. Density distributions were compared for different body regions. Primary afferent projections were labeled with biocytin or biotinylated dextrans. This showed that there is considerable central over-representation of the mandibular and nasal regions of the sensory surface involved in electrolocation, at the expense of the other body regions investigated. This over-representation is not a mere effect of the very high density of receptor organs in these areas, but is found to be due to central magnification. This magnification differs between the subclasses of electroreceptors, suggesting a functional segregation in the brain. We conclude that the chin protuberance and the nasal region are the regions of greatest sensitivity for the resistive, capacitive, and low-frequency characteristics of the environment, and are probably most important in prey detection, whereas other regions of the skin with a lesser resolution and sensitivity to phase distortion of the EOD, in particular the trunk, are probably designed for imaging larger, inanimate features of the environment. Our data support the hypothesis that the chin appendage and nasal region are functionally distinct electrosensory foveae. J. Comp. Neurol. 511: , Wiley-Liss, Inc. Indexing terms: electroreception; fovea; Gnathonemus; mormyromast; central magnification; peripheral scaling; mormyrid The natural habitats of many African weakly electric mormyrid fish are turbid tropical rivers (Moller, 1995). Behavioral studies show that mormyrids are active predators of worms and insects, foraging mainly at night (Moller, 1995). Under these conditions, which are unfavorable for visual orientation, mormyrids have developed an electric sense that allows them to detect, analyze, and identify nearby objects or organisms (Lissmann, 1958; von der Emde, 2006). This capability of sensing electric events in the external environment can be divided into three different sensory submodalities active electrolocation, electrocommunication, and passive electrolocation according to the source and type of the stimulus carrier. Mormyrids generate a pulse-type electric field around their body by the discharge of their electric organ, and they monitor modulations of its amplitude, waveform (or phase), and spatial distribution, caused by surrounding objects (Caputi et al., 1998; von der Emde, 1999; Budelli and Caputi, 2000). This is the basis of active electroloca- Grant sponsor: the Portuguese Ministry for Science and Technology; Grant number: doctoral fellowship FCT-SFRH/BD/1424/2000 (to J.B.); Grant sponsor: European Commission; Grant numbers: Marie Curie Postdoctoral Fellowship QLK6-CT (to J.E.) and IST ; Grant sponsor: the French Ministry for Foreign Affairs; Grant number: MAE ECOS U03B01. *Correspondence to: Joao Bacelo, Unité de Neurosciences Intégratives et Computationnelles, C.N.R.S., 1. Avenue de la Terrasse, Gif sur Yvette, France. joaotocs@gmail.com Received 11 December 2007; Revised 5 July 2008; Accepted 31 July 2008 DOI /cne Published online September 19, 2008 in Wiley InterScience (www. interscience.wiley.com) WILEY-LISS, INC.

2 FUNCTIONAL FOVEAE IN AN ELECTROSENSORY SYSTEM 343 tion. The fish can also detect the same sort of electric field generated by nearby conspecifics, allowing the frequency pattern of the electric organ discharge (EOD) to be used for electrocommunication (Moller, 1970, 1976; Hopkins and Bass, 1981; Szabo and Moller, 1984). In addition, they can sense the near-dc low-frequency electric fields typically generated by every living being present in an aquatic environment (Kalmijn, 1974; Szamier and Bennett, 1974; Shieh et al., 1996; Hopkins et al., 1997; von der Emde and Bleckmann, 1998; Bodznick and Montgomery, 2005; Hopkins, 2005). The latter is known as passive electrolocation, whereby the source of energy providing sensory stimulation is in the surrounding environment, rather than generated by the fish itself. The mormyrid electrosensory system possesses three types of electroreceptor organs that have different functions in the transduction of different aspects of the electrosensory world (Hopkins and Bass, 1981; von der Emde and Bleckmann, 1997). These are described in detail below. Behavioral studies show that each of the sensory subsystems can be used independently, as well as in a multimodal manner (von der Emde and Bleckmann, 1998; Moller, 2002; Rojas and Moller, 2002). Early studies described the anatomical differences of the electroreceptor organs (Derbin and Szabo, 1967, 1968; Denizot, 1970, 1971; Szabo and Wersall, 1970; Szabo, 1974) and characterized the sensitivity of the receptor organs (Bennett, 1965; Hopkins and Bass, 1981; Zakon, 1986). The three electrosensory subsystems project to separate somatotopic maps in the electrosensory lateral line lobe (ELL; Bell and Russell, 1978; Bell et al., 1981). Mormyromast system The mormyromast electrosensory system is used particularly in the detection and analysis of the distortions in the self-generated electric field. In adult mormyrids, mormyromast electroreceptors consist of an outer chamber embedded in the outer epidermal layer, and an inner chamber connected to the outer chamber by a narrow canal. Sensory cells known as A-cells are located around the outer chamber, and a second type, B-cells, are enclosed in the inner chamber (Bell and Russell, 1978). The two types of mormyromast sensory cells are differently tuned, further subdividing the system. A-type receptor cells encode primarily the amplitude of the local EOD, whereas the B-type receptor cells encode both the amplitude and the waveform distortions of the local EOD (von der Emde and Bleckmann, 1992a,b). Amplitude changes occur due to the resistive properties of nearby objects interacting with the electric field, whereas waveform alterations are due to the capacitive properties of nearby objects. In nature, resistive electric properties predominate in inanimate objects, whereas living organisms typically have more complex impedances. It has been suggested that the dissociation of these parameters by the electrosensory system is exploited by the animal as an additional cue for DLZ ELL MZ VLZ Abbreviations dorsolateral zone electrosensory lateral line lobe medial zone ventrolateral zone prey detection and identification (von der Emde, 1993; Budelli and Caputi, 2000). Knollenorgan system The knollenorgan system constitutes the second component of the active electric sense. Knollenorgans can be recognized by the distinct epithelial cell layer that covers the sensory organ, which contains a variable number of sensory cells. They are tuned to the higher frequencies of the fish s species-specific EOD (Hopkins and Bass, 1981). In contrast to the mormyromast system, Knollenorgans are used in electro-communication, involving schooling behavior, courtship, and individual identification (Moller, 1976; Hopkins and Bass, 1981; Moller et al., 1989; Friedman and Hopkins, 1998). Ampullary system Ampullary organs consist of a long, slender epidermal channel that ends in a terminal chamber enclosing a variable number of sensory cells. Low-frequency, near-dc, electric fields originating from the interaction between the electrochemical gradient of cell membranes and muscle activity are detected by the ampullary electrosensory system. Such fields are typically generated by living organisms in the water (Kalmijn, 1974), and a large number of species of different phylogenetic origin are able to sense and use this information (Bullock et al., 1983; Alves- Gomes, 2001). The frequency and DC components of these fields can constitute an electric signature typical of different swimming organisms (Kalmijn, 1974; Peters and Bretschneider, 1972; Taylor et al., 1992; Kalmijn et al., 2002). Thus, this system seems to be used functionally to detect and identify living prey, or other organisms in the nearby aquatic environment. Following on from the studies of Maler et al. (1973a,b), the somatotopic organization of mormyromast primary afferents to the electrosensory lobe was shown, although only for the trunk, by Bell and Russell (1978). Bell et al. (1989) later showed that primary afferents from the mormyromast A-afferents terminate in the medial zone (MZ) of the ELL, whereas the B-afferents terminate in the dorsolateral zone (DLZ; Fig. 1A). Primary afferents innervating knollenorgan receptor organs project to the nucleus of the electrosensory lobe (nell; Enger et al., 1976; Bell et al., 1981). The primary afferents innervating the ampullary system convey their information somatotopically to the ventrolateral zone (VLZ) of the ELL (Bell and Russell, 1978). The aim of the present study has been to examine the central representation of the electrosensory surface in greater detail, with respect to the distributions of the different types of electroreceptors, and in particular to search for evidence of anatomical foveae of the electrosensory system. Electric foveae Early studies showed that mormyromast electroreceptors are particularly abundant on the elongated chin appendage characteristic of Gnathonemus petersii (Harder, 1968; Quinet, 1971; von der Emde and Schwarz, 2002), which is so long that it has earned this species the popular name elephant-nose fish, but in fact some sort of similar protuberance or swelling is present in almost all mormyrid species. The appendage is used in a finger-like manner to dig between stones and mud, or to explore the

3 344 J. BACELO ET AL. surface of the substrate or objects (von der Emde, 2006). A high density of mormyromasts is also present in the nasal region above the mouth around the nares (Harder, 1968). The lowest density was observed on the trunk region of the body. Knollenorgans were found to have a patchy distribution, which is symmetrical on the two sides of the fish s body (Harder, 1968): an individual knollenorgan on one side of the body has a corresponding knollenorgan on the other side. Of all electroreceptor organs, the knollenorgans occur with the lowest density. The distribution of ampullary electroreceptors has been studied previously, but data are inconsistent: in some studies the chin appendage was not analyzed (Harder, 1968), whereas later studies did not find ampullary receptor organs in the chin appendage (Quinet, 1971). Although incomplete, these studies showed that ampullary organs have their highest densities in the nasal region, and that density decreases caudally (Harder, 1968; Quinet, 1971). The idea that the chin appendage or the head regions of weakly electric fish might contain electrosensory foveae was first suggested by Caputi and Budelli (1995), based on predictions from modeling the electric image showing that, in this region that is rich in electroreceptors, body geometry would produce a tip effect that would induce maximal current densities in the perioral region. Preliminary studies of the electroreceptor organ densities, the observation of foraging behavior, and measures of the physical properties of the self-generated electric field in both mormyrid and gymnotiform species all support this interpretation (Caputi and Budelli, 1995, 2006; Castello et al., 1998; Castello et al., 2000; Caputi et al., 2002; von der Emde and Schwarz, 2002; Caputi, 2004; Migliaro et al., 2005). Caputi was also the first to speculate that the chin appendage and the nasal region might indeed have different electrosensory functions: a short-range food classification/detection fovea, and a long-range object detection and guidance system, respectively (see von der Emde and Schwarz, 2002; von der Emde, 2006). Based on its high receptor organ density and movement patterns during foraging, the mobile chin appendage in G. petersii resembles the star of the star-nosed mole (Catania, 1995; Catania and Kaas, 1995) or a human finger (Marshall et al., 1937, 1941). In fact, early investigators attributed a purely tactile function to the chin appendage (Stendell, 1914a,b), but this was at a time when electroreception was still unknown. In comparison with the representation of the 11th ray of the star-nosed mole, it might be expected that the high density of mormyromasts on the chin appendage and its role during foraging would be reflected in the central representation (Catania, 1995; Catania and Kaas, 1995, 1997). The aim of this present work was therefore to study quantitatively the distribution of all types of electroreceptors and their representations in the ELL of G. petersii. Biocytin and biotinylated dextran amines were used to trace the central projections of electrosensory primary afferents from different regions of the body. The surface area of different sensory regions, together with the number of receptor organs present in those regions, was correlated with the extent of the central projections. This made it possible to estimate two magnification factors for each of the body regions studied: one characterizes the relation between the peripheral sensory surface area and the space occupied by the corresponding primary afferent terminal field projections in the ELL; the second relates the proportion of peripheral receptor organs on each body region to the size of their corresponding central representation. MATERIALS AND METHODS The data described are based on a total of 33 specimens of Gnathonemus petersii, ranging in length from 7 to 17 cm. Fish were obtained from registered dealers (Aquarélite, Aufargis, France and Aquarien-Gläser, Frankfurt, Germany) and housed in registered facilities conforming to French, European, and international standards of animal care. All experimental procedures were carried out in accordance with the recommendations of the Guide for the Care and Use of Laboratory Animals, Institute of Laboratory Animal Resources (NIH), the European Directive 86/ 609/EEC concerning the protection of animals used for experimental and scientific purposes, and the Treaty of Amsterdam Protocol on Animal Welfare (1997). The innervation of the electroreceptive epithelium has been studied by several authors (Stendell, 1914a,b; Cordier, 1937; Derbin and Szabo, 1968; Harder, 1968; Szabo and Wersall, 1970; Quinet, 1971) and is summarized for the mormyromast system in Figure 1A. Based on the pattern of ramification of the electrosensory eighth nerve, we have defined the following areas: the chin appendage, the nasal region, and the head and the trunk, as illustrated in Figure 3A. Spatial distribution of electroreceptor organs To study the spatial distribution of the electroreceptor organs in the different regions of the body, three fish, 8 10 cm long, were deeply anesthetized with MS222 (3- aminobenzoic acid ethyl ester methanesulfonate salt; Sigma, St. Louis, MO; concentration 0.2 g/liter) and fixed by perfusion with 4% glutaraldehyde in 0.1 M phosphate buffer, ph 7.2. The complete electroreceptive skin was removed from each of the four regions designated above. The individual pieces were sandwiched between a glass slide and a coverslip and then dehydrated in an ascending ethanol series, cleared in xylene, and mounted in DePeX (Gurr-BDH, Poole, UK) without additional staining. The sections were inspected under the microscope (Orthoplan, Leitz, Wetzlar, Germany), and the following parameters were used to classify sensory organs: ampullary organs were discriminated by their small size and a lack of specialized surrounding epidermal structures (Fig. 2F); knollenorgans were transparent, which made it possible to see the four or five receptor cells inside the organ (Fig. 2E); and mormyromasts could be distinguished from the other organs by their central epidermal pore surrounded by paler supporting tissue (Fig. 2F). The total number of electroreceptor organs, and the number of each of the organ subtypes, were counted in the skin from each of the designated regions, and the surface area of each of the pieces of skin was measured, by using a CCD camera system (SPOT Insight 2MP Firewire Color Mosaic, Diagnostic Instruments, Sterling Heights, MI) equipped with digital image analysis software (SPOT imaging, Diagnostic Instruments). The size of the individual mormyromast receptor organs found in the different body regions was compared in one fish, by measuring 100 or-

4 FUNCTIONAL FOVEAE IN AN ELECTROSENSORY SYSTEM 345 Fig. 1. A: Schematic drawing of the structural organization of the mormyromast receptor organ (left) and of primary afferent projection to the ELL of G. petersii. Note that the different type A and type B afferents terminate in separate regions of the electrosensory lobe (ELL) in the medial zone (MZ) and the dorsolateral zone (DLZ), respectively. Primary afferents from ampullary organs (not shown) terminate in the ventrolateral zone (VLZ). This schematic is modified after Szabo and Wersäll (1970). B: Schematic of two successive 60- thick transverse sections of the ELL showing the three zones of ELL with the landmarks used to analyze the size of terminal fields. The granular layer and the overlying ganglionic layer are shown by the solid gray and black lines, respectively. The length of this line was determined in each section and each zone separately, by using Spot Imaging software. The labeled terminal fields are shown in black, and their length was measured in each zone. For each of the three zones, the overall surface area of the granular layer was calculated as the area of the corresponding trapezoid indicated in light gray shading. The area of the labeled terminal fields in the granular layer was calculated similarly, as shown by the dark gray trapezoids. gans randomly sampled from each region. To do this, using the SPOT imaging software a circle was drawn around the pore and the surrounding supporting tissue of each organ, and the surface area was calculated from the calibrated digital images. The total number of mormyromast electroreceptor organs in the chin appendage was counted in a further 14 fish ranging in size from 7 to 17 cm. To do this, the chin appendage was cut longitudinally into four quadrants: dorsal, ventral, right lateral, and left lateral. In order to be able to normalize measurements from fish of different lengths, the image of each quadrant was divided into 10 sections equal in length. The receptor organs were then counted in each section. Innervation of mormyromasts and A-cell density For the comparison of the innervation of mormyromasts three fish (S.L.: 9 11 cm) were used. These fish were fixed by immersion for at least 1 week in 2% glutaraldehyde and 2% paraformaldehyde in phosphate buffer (0.1 M, ph 7.2). Following fixation the skin was bleached for 2 days by adding 4 10% H 2 O 2 to the fixative. After sufficient clearing of the skin, the fish were transferred via an increasing ethanol series to 70% ethanol containing 0.5% Sudan black (Sudan Black B, Sigma S2380, St. Quentin Fallavier, France) and were left in this solution overnight. The superficial layer of the specialized epidermis covering the electroreceptive regions was removed, and the skin was gently dissected and counterstained in 5% toluidine blue solution. The counterstaining was differentiated in 70% ethanol, and the skin was dehydrated and mounted in DePeX (Gurr-BDH). Nerve fibers innervating single mormyromasts were counted and, when possible, the number of A-cells per organ was determined. Surgical preparation and labeling of primary afferents The different skin regions distinguished for analysis were defined following von der Emde and Schwarz (2002), who previously described the putative foveal regions. To label the electrosensory nerves, fish (n 13) were anesthetized by immersion in an aerated MS222 solution (0.2 g/liter; Sigma A-5040: 3-aminobenzoic acid ethyl ester methansulfonate salt) until nociceptive reflexes were absent. Then fish were transferred to an operation tank where they were artificially respirated with aerated water containing 0.06 g MS222/l (flow rate 40 ml/min). Small skin incisions were made to expose the chosen branches of the nerves (Fig. 3A). These branches were: 1) the most rostral ventral branch of the anterior lateral line nerve (ALLN) together with the mandibular branch of the trigeminal nerve, just below the mouth, in order to label the chin appendage (n 7); 2) the maxillary branch of the ALLN and all small nerve branchlets that innervate the nasal region above the mouth and around the nares (n 3); and 3) the dorsal branch of the posterior lateral line nerve (PLLN), in order to label the dorsal trunk (n 3). The nerve branches were gently dissected and cut. One of the following markers: biotinylated dextran amine MW 3000 (BDA; Vector, Burlingame, CA), or biocytin (Sigma-Aldrich), was applied directly to the cut nerve, either as crystals or dissolved in DMSO (dimethlysulfoxide: Sigma-Aldrich). The incisions were closed with sterile sutures (Ethicon Endo-Surgery Vicryl Mersutures S14 with 6/0 Sabreloc needles, Issy-Les-Moulineaux, France), and the artificial respiration with anesthetic was replaced with fresh aerated water. Following surgery, fish were returned to individual aquaria. After 1 2 days (for biocytin labeling) or 4 7 days (for BDA labeling), fish were re-anesthetized deeply with MS222 (0.2 g/liter) and perfused via the heart with teleost Ringer s solution (Wolf,

5 346 J. BACELO ET AL. Fig. 2. A F: Photographs of the epidermal layer of the skin in different regions of the body. Figures A-D appear at the same magnification and show approximately 1.3 mm 2 of skin. A: Dorsal trunk. B: Head. C: Nasal region. D: Chin appendage. In this material stained with toluidine blue, the outer chambers of the mormyromasts are visible as red spots. The white spots visible in addition in D are the pores of ampullary organs. E: Knollenorgan from the nasal region, at higher magnification (unstained). F: Ampullary organ and mormyromast from chin appendage stained with toluidine blue, at higher magnification. Scale bar 100 m in A F. 1963), followed by fixative containing 2% glutaraldehyde and 2% formaldehyde buffered with 0.1 M sodium phosphate, ph 7.4. Following perfusion, brains were removed and stored at 4 C overnight in the fixative solution and cut the next day on a vibrating microtome (Leica VT100M, Rueil-Malmaison, France) in 60- m-thick sections in the coronal plane. Endogenous peroxidase activity in the sections was blocked with ethanol (Metz et al., 1989), and labeling was developed by using the avidin-biotin complex (ABC) technique (Vector), by using a diaminobenzidine chromogen (DAB, Sigma-Aldrich). Sections were counterstained with Richardson s stain (Azure II and methylene blue) or Neutral Red and mounted in DePeX (Gurr-BDH). Morphometric measurements Electrosensory primary afferents terminate in the granular layer of the ELL. The labeled areas of the central projection to the medial, dorsolateral, and ventrolateral zones of the ELL were estimated from serial sections. In each brain section, the total width of each zone was measured by using the SPOT imaging software to draw a curve that followed the boundary between the granular cell layer and the plexiform layer (see Meek et al., 1999). If the section contained labeled afferent fibers, the width of the terminal zone in the granular cell layer was calculated in a similar manner (see black shading in Fig. 1B).

6 FUNCTIONAL FOVEAE IN AN ELECTROSENSORY SYSTEM 347 Fig. 3. A: Schematic drawing of the head and rostral part of the trunk of G. petersii showing the branches of the anterior lateral line nerve (ALLn) innervating electroreceptors of different skin regions (after Harder, 1968). Note that the anterior lateral line nerve (black) and the trigeminal nerve (ntrig, gray) run in close proximity, and both innervate the chin appendix. For tracing studies, different nerves were cut and labeled at the positions indicated by black bars. The areas referred to as the chin appendage and the nasal region are shown in dark gray. The area defined as the rest of the head is white. The trunk region is defined as all that region posterior to an imaginary contour passing through the ventral edge of the operculum and the first branch point of the dorsal ramus of the PLLn (dotted line). B: Photographs of mormyromast receptor organs following bleaching of skin and Sudan black labeling of nerve fibers. In the upper photograph three nerve fibers innervate six A-cells. In the lower photograph there are four nerve fibers and eight A-cells. C: Number of A-cells found per mormyromast in the different regions of the body; bars show standard deviation. The significance of differences was assessed by using the Student-Newman-Keuls test (df [124]; receptor organs contained significantly more cells in the trunk region, indicated by an asterisk (P 0.001). D: Surface area of mormyromast organs in different regions of the body determined by measuring the specialized epidermal area surrounding each receptor organ (see shaded area in Fig. 1A). Bars show standard deviation, also calculated as above; receptor organ size is significantly different in the body regions studied. Scale bar 50 m inb.

7 348 J. BACELO ET AL. This was done for each section. Given the known thickness of the sections (60 m), the extent of the labeled zone in each section and the total extent of the particular zone in each section, the size of the labeled projection area through the whole ELL was then calculated as a trapezoid (see black-shaded area connecting labeled zones in Fig. 1B) from the serial measurements. Thus the size of the labeled afferent projection could be expressed as a proportion of the total area of the zone in question. The primary aim of the present experiments was to study the central representation of the putative foveal regions, and the full innervation of the trunk and head regions was not measured directly. For the trunk, we labeled only the dorsal branch of the PLLn, and the area representing the ventral trunk was therefore calculated by subtracting the area representing the dorsal trunk from the whole, making the assumption that the terminal fields of the electrosensory projection from the dorsal and the ventral part of the trunk both extend to the same rostrocaudal point in the ELL (Bell and Russell, 1978). Due to the wide branching of the nerves in the head region, it was difficult to be sure that all were labeled in any given experiment. Thus the size of the region representing the head was calculated by subtracting the sum of all the other labeled areas (representing the chin appendage, the nasal region, and the dorsal and ventral trunk region) from the total measured area in each zone. For each area of the skin that was innervated by a given labeled nerve branch, we calculated the relative size of this skin region with respect to the whole surface of the electroreceptive skin, as follows: relative skin area area of investigated region/total area of electroreceptive skin. In the same way we calculated the relative size of the projections from each skin area to each zone of the ELL: relative projection area labeled area/total area of ELL zone. To estimate the representation in the ELL relative to the skin area, we then calculated the total magnification factor: Mtotal relative projection area/relative skin area. An Mtotal factor of 1 means that the electrosensory surface is represented proportionally according to its relative skin area. Mtotal factors 1 indicate overrepresentation of that skin region compared with its relative size, and Mtotal factors between 0 and 1 indicate a representation in the brain than would be less than expected based on the relative area of that region. Because the density of electroreceptors is different in each of the investigated regions, we needed to scale this magnification factor with respect to the number of electroreceptors present. The relative number of electroreceptors was calculated as: relative number of electroreceptors number of electroreceptors in investigated area/total number of electroreceptors over the whole body. By using this density we calculated a central magnification factor: Mcentral relative projection area/relative number of electroreceptors. Mcentral gives an estimate of the central magnification. Thus Mcentral 1 indicates that more area within the brain is used to process the input than would be expected based on the density of the receptor organs. For statistical comparisons, data were transformed with an arcsin function to avoid deviations from normality due to the use of percentages: 0.5 arcsin x X 1 arcsin x 1 X 1 where x labeled area and X total area. Photographs were taken with a Canon EOS 300 digital camera (Canon Deutschland, Krefeld, Germany), or a Spot Insight CCD camera (Diagnostic Instruments), attached to the microscope. When appropriate, contrast and brightness were optimized by using Corel PhotoPaint software from the Corel Graphics suite (Version 12) (COREL, Ottawa, Canada). RESULTS Distribution of the different receptor organs The distribution and density of the three types of electroreceptors was mapped by using flat-mounts of the epidermal layer of the electroreceptive skin. As described by Harder (1968), the three types of receptor organs can be identified by their different coloration following staining with toluidine blue and by their morphology. The knollenorgans have no opening to the skin surface, they are the biggest organs that can be found, they stain red with toluidine blue, and very large cells are visible within each receptor organ (Fig. 2E). The number of sensory cells visible within each knollenorgan is variable. Within each receptor organ, the sensory cells are enclosed in a cavity within a large capsule that is filled by epithelial cells. Previous studies have shown that each sensory cell receives a branch of a single afferent nerve fiber (Derbin and Szabo, 1968). When the epidermal layer is removed, the whole intact capsule of the knollenorgan usually comes with it. Ampullary organs are distinguished by their opening to the surface of the epithelium; they are small and stain bluish/greenish with toluidine blue (Fig. 2D,F). These receptor organs contain multiple sensory cells located at the base of a short canal. Mormyromast electroreceptor organs are intermediate in size; they have no opening to the surface but are covered by a well-defined, dome-shaped supporting structure (Fig. 2F). In contrast to the knollenorgans, they lack the very large, visible sensory cells and they stain bluish/violet with toluidine blue. The mormyromast organ consists of

8 FUNCTIONAL FOVEAE IN AN ELECTROSENSORY SYSTEM 349 Fig. 4. A: Mean number of mormyromasts (circles) and mean sensory skin surface area (triangles) (calculated from mean receptor organ surface area) in the different body regions of an 8-cm fish. B D: Densities of different types of electroreceptor organs in the different body regions. B: Mormyromasts. C: Ampullary organs. D: Knollenorgans. Differences in densities were tested with the Student-Newman- Keuls post hoc test: **, P 0.001; *, P Note that only knollenorgans were evenly distributed over the body (one-way ANOVA, F 3,8 1.43; P 0.3). an intraepidermal cavity that is filled with polysaccharides and is covered by epithelial cells (Szabo and Wersall, 1970; Denizot, 1971). A second, inner chamber is located below the first cavity, to which it is connected via a narrow canal (Fig. 1A). Mormyromast electroreceptors. Extending previous data that focused on the head and trunk (Harder, 1968; Quinet, 1971; Bell et al., 1981; von der Emde, 2006), the present study has examined receptor organ distribution over the whole electroreceptive surface. Quantitative measurements show that the number and density of mormyromast electroreceptor organs on the chin appendage are much greater than in any other body region (Fig. 4A,B; one-way ANOVA, F 3,8 1289; P 0.001, Student- Newman-Keuls test, df (8), P 0.001). At the chin appendage the mormyromast receptor organs occur with a density more than 20 times higher than on the trunk (Figs. 3, 4B). From the chin appendage to the head there is a gradual decrease in mormyromast receptor organ density (Student-Newman-Keuls test, P 0.001), and from the caudal region of the head to the trunk the density does not change ((Figs. 4B, 5A; Student-Newman-Keuls test, P 0.4). The number of mormyromasts is unevenly distributed along the length of the chin appendage (Fig. 5A, two-way ANOVA, F 9, ; P 0.001), with the dorsal quadrant having the highest density (F 3, ; P 0.007; Student-Newman-Keuls test; df [480]; P 0.02; data were transformed to natural logarithms to ensure homogeneity of variances before testing). Ampullary organs. In general, the ampullary organ density follows the spatial profile of the mormyromasts closely (Fig. 4B,C; one-way ANOVA, F 3,8 246; P 0.001, Student-Newman-Keuls test, df [8], head: P 0.001; nasal region: P 0.02; head and trunk: P 0.76). At the chin appendage the ampullary receptor organs occur with their highest density, up to 50 times higher than on the trunk

9 350 J. BACELO ET AL. Fig. 5. A: Mean number of mormyromasts along the chin appendage. (Data are normalized for 20 chin appendages from fish of different lengths.) Each chin appendage was divided into 10 segments of equal length. These segments were further separated into four quadrants: dorsal (black), ventral (white), left (dark gray), and right (light gray). Note that the mean number of mormyromasts per segment was significantly higher toward the tip of the chin appendage (see statistics in text) and in the dorsal quadrant. B: Cumulative count of ampullary organs (circles), mormyromasts (stars), and knollenorgans (squares) through the four body regions investigated. The gray surface represents the cumulative distribution of the sensory surface. Note that for ampullary and mormyromast receptor organ distributions, the thick dotted vertical lines, indicating the point in front of which 50% of these receptor organs are situated, is shifted rostrally compared with the 50% level of the sensory surface (fine dotted line). (Figs. 2D, 4C). This contrasts with previous studies that failed to identify this kind of electroreceptor on the chin appendage (Quinet, 1971). Knollenorgans. The density of knollenorgans is the lowest of all receptor organs ( mm 2 ; Fig. 4D), and no clear regional differences were found (one-way ANOVA, F 3,8 1.43; P 0.3). However, their distribution is very stereotyped and particular clusters occur in specific regions, symmetrical on either side of the fish s body and easily identifiable from one fish to another. Knollenorgans are not present on the chin appendage itself. A single knollenorgan is present at the base of the chin appendage on either side of the mouth and then, moving caudally, these receptors occur in particular around the nares, around the eye, along the margin of the operculum, and also in a row along the trunk at the margin of the specialized electrosensory epithelium. In summary, both ampullary organs and mormyromasts have their highest densities at the chin appendage. Their densities then decrease steeply along the chin appendage and the body, reducing fourfold from the chin to the nasal region in the case of mormyromasts, and sixfold in the case of ampullary organs. In both cases, when the figures for the chin appendage and the nasal region are grouped together, it can be seen that almost 50% of all receptor organs are located on the most rostral region of the head, as illustrated in the cumulative distribution graph in Figure 5B. Thus, although the two regions that have been putatively termed electric fovea constitute less than 5% of the sensory surface, they bear about half the total number of ampullary and mormyromast organs. The irregular, clustered distribution of knollenorgans suggests that they may be less concerned with spatial resolution at the sensory surface and possibly more involved in marking the timing of electrical events. Innervation and morphology Although the density of the mormyromast distribution decreased toward the trunk, the size of individual mormyromast organs increased from rostral to caudal (Fig. 3D). Corresponding to their larger size, the mean number of A-cells/mormyromast was highest in receptor organs on the trunk (mean 6.4 1; one-way ANOVA, F 3, ; P 0.001, Student-Newman-Keuls test, df [124], P 0.001). Thus the caudal body has bigger mormyromasts that contain more sensory A-cells (Fig. 3C,D) compared with those in the more rostral regions. Although some preparations allowed visualization of B-cells in the inner chamber (see upper image in Fig. 3B), this was not usually the case, and the sample size was too low to test for any trend. However, as was shown by Bell et al. (1989), the number of A- and B-cells in a mormyromast is usually equal. Almost all mormyromasts were innervated by three afferent nerve fibers, regardless of their size and position on the body. This finding agrees with observations made on the head region only: other authors have shown that one of the three fibers innervates all B-cells whereas the remaining two fibers innervate the A-cells (Cordier et al., 1937; Szabo and Wersall, 1970; Bell et al., 1989) (see schematic in Fig 1A). In only a few mormyromasts did we find four afferents fibers, and when this was the case, these receptor organs always contained more than 7 A-cells. For knollenorgans and ampullary organs we never found more than one afferent fiber innervating the receptor organ. Central representation of the electrosensory surface Figure 6 shows examples of the labeled primary afferent terminal fields in the granular layer of the three zones of the ELL cortex and in the nucleus of the ELL. Primary

10 FUNCTIONAL FOVEAE IN AN ELECTROSENSORY SYSTEM 351 Fig. 6. Examples of primary afferent terminal fields in the granular layer of ELL, labeled with biocytin. A: Schematic diagram showing zones of ELL. B: DLZ and VLZ. C: MZ. D: The nucleus of the ELL. Primary afferent terminals were found in the deep and superficial granular cell layers but did not extend into the molecular layer. Scale bar 100 m inb;50 m in C,D. Fig. 7. Comparison of the extent of labeling in the different zones of ELL following labeling of nerves to different body regions (chin appendix, nasal region, and trunk; pooled data from 10 fish 8 cm long.) MZ, black; DLZ, light gray; VLZ, dark gray. Significant differences (*) stood out for the representation of the chin appendage in DLZ and for the representation of the trunk in MZ following labeling of the dorsal ramus of the PLLn. afferent terminals were found in the intermediate layer and the deep and superficial granular cell layers but did not extend into the molecular layer. This is in contrast to the earlier findings using nerve degeneration techniques (Maler et al., 1973a,b) but in agreement with later studies of Bell and Russell (1978). The distribution of labeling also showed a somatotopic representation of the electrosensory periphery in the three zones of the ELL cortex, as previously proposed by the study of Bell and Russell (1978). As described previously for the posterior lateral line nerve, primary afferents coming from the ventral nerve branches of the body project ipsilaterally to the lateral part of the MZ and to the dorsal parts of the DLZ and VLZ. However, the projections from the chin appendage, in spite of its purely ventral innervation by the infraorbital branch of the anterior lateral line nerve (Fig. 3A), occupy the whole rostral ELL (see Fig. 11), thus resulting in a somatotopic representation of the chin appendage aligned (in register) with the rest of the body, and not merely ventrally. The granular layers of the three zones have different surface areas (MZ mm 2 ; DLZ mm 2 ; VLZ mm 2 ), and the present results are in agreement with the calculations of Bell et al. (1989), who noted that the surface area of the ganglionic cell layer of the MZ is roughly 50% bigger than that of the DLZ and 2.7 times bigger than that of the VLZ. The areas occupied by labeled terminations for the four skin regions in the medial, dorsolateral, and ventrolateral zone of the ELL were quantified as described in Materials and Methods (Figs. 7, 8). The afferents innervating the chin appendage give rise to a very large projection to ELL, where their terminal fields occupy roughly 40% of the total ELL surface. The major part of the projection from the chin appendage goes to the DLZ ( mm 2 ). Projections to the MZ ( mm 2 ) and the VLZ zones ( mm 2 ) are each approximately 50 60% of the size of the projection to the DLZ (Fig. 8; one-way ANOVA, F 2, ; P 0.001; Student-Newman-Keuls test; df [18]; P 0.001). In contrast, the projections from the nasal region are equally distributed over the three zones of the ELL (oneway ANOVA, F 2,6 2.04; P 0.2) and comprise about mm 2 in each zone. The afferents from the trunk occupy a very large portion of the MZ ( mm 2 ), and constitute the main source of input to this zone. In contrast, much smaller termination areas of trunk afferents are found in the DLZ ( mm 2 ) and the VLZ ( mm 2 ; one-way ANOVA, F 2,6 56.1; P 0.001; Student-Newman-Keuls test; df [6]; P 0.001). Thus the MZ receives its major input (40%) from the trunk (one-way ANOVA, F 2, ; P 0.001; Student- Newman-Keuls test; df [10]; P 0.001), followed by the head (31%) and the chin appendage (22%) (Fig. 8A). This is in contrast to both the DLZ and the VLZ, where the trunk is relatively much less represented (Fig. 8B,C). These zones receive half of their input from the chin appendage (50%; one-way ANOVA, F 2,10 664; P 0.001; Student-Newman-Keuls test; df [10]; P 0.001; and 47%; one-way ANOVA, F 2, ; P 0.001; Student- Newman-Keuls test; df [10]; P 0.001) followed by the head, the trunk, and the nasal region. The representation of the fish s body in the DLZ is even more polarized or skewed than would be expected based on the peripheral distribution of the mormyromasts. This is illustrated in the cumulative distribution count of primary afferent terminal field area shown in Figure 8D. The 50% point of the cumulative distribution for the DLZ (Fig. 8D, gray trace) is located further rostrally than the 50%

11 352 J. BACELO ET AL. Fig. 8. A D: Areas of primary afferent terminal fields in the ELL labeled following local injections, expressed as a percent of total area: (A) labeling of the MZ, (B) DLZ, and (C) VLZ. D: Cumulative distribution of sensory representation in the VLZ (gray), DLZ (open circles), and MZ (black), in the rostrocaudal axis. The gray surface represents the cumulative distribution of the area of the sensory surface. Vertical dotted lines represent the point at which sensory representation fills half of each zone. Note that all are shifted rostrally compared with the 50% level of the sensory surface (see dotted lines). This gives a measure of the relative magnifications apparent in the central representation of the different body regions. Letters above measurements indicate the results of the Student-Newman-Keuls post hoc tests. Different letters within one graph indicate that there were significant differences (P 0.001) between populations, whereas where letters are identical, there was no significant difference between measures (P 0.05). Because the head-projection representations were calculated by subtraction from 100%, as the area missing following summation of results for all other injection sites, no statistics could be performed for this region. point in the cumulative count of peripheral mormyromast receptor organs (Fig. 5B, stars), indicating that most neurons in the DLZ receive input from rostral areas. A similar polarization of ampullary organ input is seen in the VLZ (Fig. 8D, dotted trace; Fig. 5B, circles), where the size of the rostral central representation was relatively much greater than the polarization of the peripheral distribution. In contrast, for the MZ, the relative polarization of the central (Fig. 8D, black trace) and the peripheral distributions (Fig. 5B, stars) is similar. In the following section we present the results on the magnification factors for the three zones of the ELL, considering the termination of mormyromast afferents first, followed by the termination of ampullary organs. For

12 FUNCTIONAL FOVEAE IN AN ELECTROSENSORY SYSTEM 353 Fig. 9. Magnification factors of the mormyromast pathway compared in the medial zone (MZ: black) and dorsolateral zone (DLZ: gray). A: Magnification factor relative to the electroreceptive surface area. This shows that the chin appendix is over-represented in both the dorsolateral and medial zones of the ELL, whereas the trunk is under-represented. B: Magnification factor relative to the proportion of electroreceptors. Note that the results for the medial and dorsolateral zones differ: in the MZ there is no central magnification, whereas in the DLZ the chin appendage is over-represented at cost to representation of the trunk. Horizontal dashed lines mark the level corresponding to a magnification faction of 1. Statistics based on the t-test: **, P 0.01; *, P 0.05; all others P 0.2. mormyromasts, the total magnification factor, relative to the skin area, is shown in Figure 9A. For both the medial and the dorsolateral zones, the chin appendage is strongly over-represented by a total magnification factor of 14 (t0.05[2],8 14.9; P 0.001) and 32 (t0.05[2],8 156; P 0.001), respectively. Likewise, the nasal region, even though it represents only 2.1% of the sensory surface, is over-represented in both zones, i.e., MZ 3.4 (t0.05[2],4 3.5; P 0.02) and DLZ 4.6 (t0.05[2],4 7.3; P 0.002)). The trunk, comprising 55% of the total sensory surface, is strongly under-represented in both zones, i.e., MZ 0.72 ; (t0.05[2],4 7.9; P 0.001) and DLZ 0.34 (t0.05[2],4 34.7; P 0.001)). Scaling the projections to the different zones according to the area of skin gives an estimate of the central representation of the electrosensory periphery. However, because the densities of the receptor organs are uneven, the total magnification factor will also depend on the peripheral density. To compensate for this, the magnification factors were rescaled by the relative number of electroreceptors in each zone investigated. The central magnification factor then shows the relative neuronal space devoted to the processing of each electroreceptor. The results are shown in Figure 9B. For the DLZ, the central magnification factors follow a rostrocaudal profile comparable to that of the total magnification factor. This means that the chin appendage has a representation two times greater than what would be expected based on the receptor organ distribution (t0.05[2],8 62.2; P 0.001), the nasal region is represented in proportion to the peripheral receptor organ density (1.17, t0.05[2],4 0.99; P 0.4), and the trunk is under-represented (central magnification factor 0.5, t0.05[2],4 3.00; P 0.04). In contrast to the DLZ, no significant difference in the central representation scaled by the relative receptor organ numbers was found in the MZ (P 0.2). Taken together, these two factors show that in both the MZ and the DLZ, representation of the nasal region and the chin appendage (the two foveae) is significantly magnified. The regions differ in that in the case of the MZ the high degree of central representation is proportional to the polarized distribution of mormyromasts at the periphery, whereas in the DLZ an additional central magnification occurs, by which an even greater proportion of the zone is devoted to representing the already highly polarized chin appendage and nasal region foveae. This further strengthens the representation of the chin appendage in the DLZ, thereby enhancing the representation of the capacitancesensitive pathway for this structure. The results for the ampullary system (Fig 10) are comparable to those of the mormyromast DLZ pathway. The chin appendage is strongly over-represented centrally, i.e., 29 (t0.05[2],6 18.5; P 0.001) ), followed by the nasal region, i.e., 5 (t0.05[2],4 92.5; P 0.001)). This reflects the rostrally polarized distribution of the peripheral receptor organ population, but in addition, rescaling the projection values by the relative number of ampullary organs (Fig. 10B) again demonstrates a significant central magnification. Ampullary organs in the chin appendage are centrally over-represented by a factor of 1.27, and representation of the nasal region occupies roughly 1.26 times as much space per organ as expected (t0.05[2],4 2.85; P 0.05). Again, this over-representation leads to a strong underrepresentation of the trunk by a factor of 0.5 (t0.05[2],4 2.9; P 0.04)).

13 354 J. BACELO ET AL. Fig. 10. Magnification factors of the ampullary pathway. A: Magnification factor of the central representation relative to the total electroreceptive surface. B: Magnification factor of the central representation relative to the proportion of electroreceptors in each body region. Note that the chin appendage and the nasal region are overrepresented in both measures at the expense of the trunk. Statistical significance was tested with the t-test: **, P 0.01; *, P DISCUSSION Based on receptor organ densities, the results show that the chin appendage, and probably also the nasal region, can be considered as foveae of both the mormyromast and ampullary systems. The labeling of primary afferent projections has shown that in the pathways of the DLZ and VLZ of the ELL, a further magnification of the polarized distribution of electroreceptive input takes place, which exceeds a prediction based on the principle of peripheral scaling. To distinguish this centrally occurring magnification, we have used the term central magnification factor. The strong central magnification found for the primary afferent projection from mormyromast B cells of the chin appendage suggests that this structure may serve a different purpose in electrolocation compared with the rest of the body. Behavioral observations suggest that it is used primarily during foraging behavior and because living prey are characterized by complex impedances, unlike inanimate objects, which tend to be purely resistive, this suggests that the enhancement of the representation of the phase-sensitive pathway of the chin appendage provides an advantage in locating and discriminating prey items against a more resistive background. The resistance-sensitive pathway of the mormyromast A-cells may have greater importance for general orientation and navigational tasks because the central representation remains proportional to the peripheral density distribution and obeys the principle of peripheral scaling without enhanced central magnification. The similar rostral polarization of the ampullary system gives further support to the interpretation of functional segregation between the chin and the rest of the body. These receptor organs are particularly sensitive to lowfrequency (near-dc) electric fields of a biological nature (Szamier and Bennett, 1974) and have been shown to be used in prey localization in other species of electric fish that rely on external electric cues only (passive electrolocation; (for review, see Wilkens and Hofmann, 2005). What constitutes a fovea? Studies of electroreceptor densities and modeling of electrosensory images in the gymnotiform fish Gymnotus omari or Gymnotus carapo have identified a fovea with a very high receptor organ density on the snout below the mouth and a parafovea with a slightly lesser density above the mouth (Carr et al., 1982; Migliaro et al., 2005; Caputi and Budelli, 2006). In the mormyrid G. petersii, the present description, showing a very high electroreceptor organ density on the chin appendage and a slightly lesser, but still high-density distribution on the nasal region of the snout, gives a similar picture. However, observations of foraging behavior and physical measurements of the electrosensory image in G. petersii (von der Emde and Schwarz, 2002), suggest that the chin appendage and the nasal region could possibly be considered as two separate electric foveae with different functions: a short-range food classification/detection fovea, and a long-range object detection and guidance system, respectively (von der Emde and Schwarz, 2002; von der Emde, 2006). Two or more functionally different foveae have been identified in a number of other sensory systems, for example, in the visual system of many birds of prey and pigeons (Friedman, 1975; Galifret, 1968), or the dual fovea of the foureyed fish Anableps (Oliveira et al., 2006). Before addressing this hypothesis of a functional separation, we must consider what constitutes a fovea and which of these requirements are met according to our data. The Latin term fovea means little pit and was used historically by anatomists to describe the small depression in the retina of the human eye where receptor organ density is greatest and visual acuity is highest. In most sensory systems, including the retina, the spatial distri-